Arginine‐Rich Cell‐Penetrating Peptides Induce Lipid Rearrangements for Their Active Translocation across Laterally Heterogeneous Membranes

Abstract Arginine‐rich cell‐penetrating peptides (CPPs) have emerged as valuable tools for the intracellular delivery of bioactive molecules, but their membrane perturbation during cell penetration is not fully understood. Here, nona‐arginine (R9)‐mediated membrane reorganization that facilitates the translocation of peptides across laterally heterogeneous membranes is directly visualized. The electrostatic binding of cationic R9 to anionic phosphatidylserine (PS)‐enriched domains on a freestanding lipid bilayer induces lateral lipid rearrangements; in particular, in real‐time it is observed that R9 fluidizes PS‐rich liquid‐ordered (Lo) domains into liquid‐disordered (Ld) domains, resulting in the membrane permeabilization. The experiments with giant unilamellar vesicles (GUVs) confirm the preferential translocation of R9 through Ld domains without pore formation, even when Lo domains are more negatively charged. Indeed, whenever R9 comes into contact with negatively charged Lo domains, it dissolves the Lo domains first, promoting translocation across phase‐separated membranes. Collectively, the findings imply that arginine‐rich CPPs modulate lateral membrane heterogeneity, including membrane fluidization, as one of the fundamental processes for their effective cell penetration across densely packed lipid bilayers.


Introduction
The transport of active molecules (e.g., ions, metabolites, and drugs) across cell membranes has been significantly challenging due to their limited diffusion through phospholipid bilayers.To address this, cellpenetrating peptides (CPPs) have attracted great attention because of their ability to cross cell membranes with cargoes. [1,2]In particular, arginine-rich peptides, including HIV-1 TAT peptide and polyarginines (R n ), are widely used for imaging and therapeutic applications. [3]6][7][8] The direct translocation of arginine-rich CPPs has been suggested to be related to their strong interactions with lipid headgroups.Specifically, the positively charged arginine prefers to bind to negatively charged lipids (e.g., phosphatidylserine (PS) and phosphatidylglycerol (PG)), and the guanidinium group of arginine even forms hydrogen bonds with phosphate, carboxylate or sulfate groups of lipids, causing structural alterations of peptide-bound membranes. [9,10]ultiple mechanisms have been proposed to explain how CPPs traverse cell membranes.One mechanism is that the binding of CPPs induces stable pores in lipid membranes for subsequent cell penetration of molecules.The strong interactions between arginine-rich CPPs and lipids are well known to significantly reduce the thermodynamic cost of membrane pore formation; both HIV-1 TAT [11][12][13] and R n [14][15][16] have proven to induce pores in membranes, allowing the passage of small dyes or ions.The other mechanism suggests that when transient pores naturally occur with thermal fluctuations, CPPs utilize these pores in translocation. [15,17]Based on the classical theory of pore formation, [18] the probability of both stable and transient pore formation is mainly determined by two factors; the lateral tension of the membrane, which promotes pore expansion, and the line tension of the pore edge, which promotes pore closure.21][22] Another fundamental feature of cellular membranes is their lateral heterogeneity. [23,24]Several studies have shown that the alterations in lateral membrane organization have a significant role in the functionality of diverse membrane-active peptides.27][28] Phase separation induced by a few CPPs has also been reported, raising the possibility of peptide translocation via phase boundary defects [29][30][31] or modulated membrane curvatures. [32]While it is evident that the translocation ability of the CPPs could not be fully understood without membrane heterogeneity, the relationship between CPP translocation and dynamic membrane rearrangements in heterogeneous membranes remains poorly understood.Instead, penetration efficiencies have typically been evaluated by externally modulating overall membrane properties by changing membrane compositions. [7,12,15,33,34]n this work, we demonstrate the dissolution of ordered lipid domains in heterogeneous membranes upon the binding of nona-arginine (R 9 ), a representative arginine-rich CPP, [35,36] that can facilitate peptide translocation across the membrane.Using planar freestanding lipid membranes, [37] we successfully visualized the lateral rearrangements induced by R 9 in real-time.In particular, we found that the lipid rearrangements could lead the liquid-ordered (L o ) phase to fluidize into the liquid-disordered (L d ) phase, thereby lowering the phase transition temperature of the lipid bilayers.Penetration analysis of R 9 into phase-separated giant unilamellar vesicles (GUVs) further confirmed that the dissolving of the L o domains accelerates the direct translocation of R 9 .These findings suggest that peptide-induced membrane remodeling may accelerate membrane translocation of CPPs by altering the fluidity of tightly packed lipid membranes.

Experimental Setup for Real-Time Monitoring of Lateral Membrane Reorganization
To visualize the dynamic membrane reorganization induced by nona-arginine (R 9 ), we used a freestanding planar lipid membrane array in a transmission electron microscopy (TEM) grid (Figure 1a; Figure S1, Supporting Information). [37,38]We prepared simplified lipid mixtures composed of phosphatidylcholine (PC), phosphatidylserine (PS), and cholesterol, mimicking mammalian plasma membranes.The preparation of phase-separated membranes began with the mixing of unsaturated phospholipid (DOPC) and saturated phospholipid (DPPC), both of which are neutral in charge, and the resulting DOPC/DPPC/cholesterol membrane underwent phase separation into a bright L d phase and a dark L o phase at 24 °C with 0.5 mol% of Texas Red 1,2dihexadecanoyl-sn-glycero-3-phosphoethanolamine (TR-DHPE) (Figure 1b).We note that the L d phase is enriched in unsaturated phospholipids and fluorescent TR-DHPE, and the L o phase is enriched in saturated phospholipids and cholesterol.Either the L d or L o phase can be more negatively charged to strongly interact with the positively charged R 9 ; by introducing PS, an anionic phospholipid, in a saturated or unsaturated form (i.e., DPPS or DOPS), we could determine which phase preferentially interacted with the cationic peptide.For example, upon the inclusion of DPPS, the fluorescence of FITC-labeled R 9 (FITC-R 9 ) was predominately found in the L o phase, where anionic saturated phospholipids were more abundant (Figure 1b).Even with GUVs, we successfully visualized the phase-preferential R 9 binding (Figure S2, Supporting Information), and the R 9 preferred binding to the L d or L o phase with 10 mol% DOPS or DPPS, respectively, whereas the binding did not occur on the neutral membrane of DOPC/DPPC/cholesterol.

R 9 -Induced Lipid Rearrangements on Phase-Separated Membranes
On heterogeneous membranes, the binding of cationic R 9 induced lateral rearrangements of anionic lipids (Figure 2).The phase behaviors of freestanding lipid bilayers were monitored in real-time by introducing 50 μm R 9 to the membranes while varying the membrane compositions.On neutral DOPC/DPPC/cholesterol membranes, the addition of positively charged R 9 had no effect due to the absence of negatively charged lipids (Figure 2a).However, when cationic DPPS was included in the heterogeneous membranes, two significant changes in phase behavior occurred (Figure 2b; Movie S1, Supporting Information).First, immediately upon the addition of R 9 , small L o domains were formed within the pre-existing L d domains as a result of R 9 binding (Figure S3, Supporting Information).It is known that electrostatic interactions between anionic lipids and cationic peptides can induce lateral segregation of anionic lipids, [28,30,39,40] and we also confirmed that initially homogeneous DOPC/DPPS membranes underwent phase separation due to the DPPS clustering upon the addition of R 9 (Figure S3, Supporting Information), making FITC-R 9 highly concentrated on the surface of newly formed L o domains (Figure S4, Supporting Information).Second, over several tens of minutes, the total area of L o domains decreased dramatically as a result of domain dissolution, that is, phase mixing.On membranes containing 10 mol% DPPS, for example, L o domains decreased to 49.5 ± 31.4% (n = 4) of the initial area after 30 min (Figure 2d).Similarly, with 4 μm of HIV-1 TAT, another representative arginine-rich CPP, the L o domains of membranes containing 16 mol% DMPG were reduced to 53.7 ± 7.1% (n = 10) of the initial area after 1 h (Figure S5, Supporting Information).We note that there were no alterations in the phase behavior with membranes containing DOPS (Figure S6, Supporting Information).It was anticipated that these membranes would undergo lateral rearrangements in a similar manner; however, the inability of unsaturated lipid DOPS to form dark ordered phase domains made their observation technically challenging through domain visualization based on dye partitioning. [41,42]espite the early temporal clustering of DPPS-rich ordered domains within the L d phase, DPPS-rich L o domains, including newly formed ones, eventually dissolved over time.The observed phase mixing suggests that the binding of cationic R 9 to anionic L o domains would ultimately decrease the miscibility temperature (T m ) of lipid phase separation.To analyze the temperature-dependent phase separation behavior of laterally heterogeneous membranes, the temperature of an imaging chamber was increased at a rate of ≈0.4 °C min −1 , and the number of phase-separated membranes was counted every 2 °C.By ) is introduced, it electrostatically binds to the surface of lipid membranes, specifically interacting with anionic phospholipids.The heterogeneous membrane is formed using a lipid oil mixture of DOPC/DPPC/DPPS + TR-DHPE (60/30/10 + 0.5 mol%), and cholesterol is added via MCD.The prepared FITC-R 9 solution was injected at a concentration of 3 μm into the aqueous solution above the lipid membranes.The images were obtained ≈1 min after the FITC-R 9 addition.
linearly fitting the collected data, the apparent T m was determined as the temperature at which 50% of the total membranes undergo phase separation (Figure 2e).For example, the T m of bare DOPC/DPPC/DPPS/cholesterol membranes was measured to be 30.1 °C, but 30 min incubation with 50 μm R 9 lowered the T m to 28.2 °C.This decrease in T m indicates that the cationic R 9 fluidizes the anionic lipid membranes, leading to the observed L o domain dissolution by phase mixing.

Electrostatic Interactions of R 9 with Heterogeneous Membranes in Phase Mixing
As the dissolution of L o domains is likely due to the electrostatic interactions between the cationic R 9 and the anionic DPPS in the DPPS-containing heterogeneous membranes, we next investigated the effect of DPPS and R 9 concentration, respectively.We first varied the molar content of DPPS in the phase-separated membranes, and the total area of L o domains was measured over time (Figure 2b-d).While 10 mol% DPPS resulted in a decrease of ≈50.5% in the area of the L o domains (Figure 2b), 5 mol% DPPS did not cause such a significant change (Figure 2c).It should be noted that we could not investigate the effect of higher mol% DPPS; as the percentage of DPPS increased, the resulting lipid bilayer became increasingly unstable, impeding the formation of freestanding planar lipid membranes in a TEM grid.However, the effect of R 9 concentrations on the extent of L o domain dissolution was well evaluated when the DPPS content was up to 10 mol%.When the R 9 concentration was gradually changed from 5 to 50 μm for 180 min, the decrease in the total area of L o domains became more significant at higher R 9 concentrations (Figure 2f).Although we only observed the membranes for 30 min at each concentration, phase mixing occurred slowly and continuously even at low R 9 concentrations, and were able to identify the trend of R 9 concentration-dependent dissolution of the L o domains.Note that in our experiment system, the lipid-to-peptide ratio is expected to be ≈1:10 at 50 μm R 9 , potentially leading to dramatic phase mixing at this high peptide concentration.
The real-time observations indicated that the rate of phase mixing would be accelerated with higher levels of anionic lipids and cationic peptides.When there were stronger interactions between the anionic lipids and the cationic peptides, the T m of phase-separated membranes would decrease further, leading to more domain dissolution by phase mixing.Indeed, the dissolu-tion of L o domains on heterogeneous membranes became faster with increasing concentrations of anionic DPPS and cationic R 9 (Figure 2d-f); even at low concentrations of DPPS and R 9 , the membrane reorganization would occur, although with a limited extent of phase mixing.Thus, we concluded that the presence of R 9 would induce membrane reorganization and subsequent fluidization, allowing us to speculate that arginine-rich CPPmediated lipid rearrangements would be directly related to its active translocation across laterally heterogeneous membranes.

Investigating Penetration of R 9 into Homogeneous and Heterogeneous GUVs
The energy-independent cell penetration of arginine-rich CPPs is generally explained by the aforementioned two mechanisms, with the critical difference being whether or not the peptide is directly involved in pore formation at plasma membranes.Using PS-containing heterogeneous membranes, we explored the possibility of R 9 -induced pore formation and its membrane penetration.To evaluate the pore-forming ability of R 9 , we encapsulated calcein, a water-soluble fluorescent dye (≈6.5 Å in radius), into DOPC/DPPC/DPPS/cholesterol or DOPC/DOPS/DPPC/cholesterol GUVs, and observed the dye leakage from the interior to the exterior of the GUVs in real-time (Figure 3).Because the GUVs were unstable at high R 9 concentrations (>50 μm), our experimental setup was limited to 10 μm R 9 , making our leakage measurements reliable and reproducible.It should be noted that this concentration of peptide was still ca-pable of inducing a phase transition from L o to L d phase.Over 30 min, the addition of R 9 did not cause notable fluorescence changes inside the GUVs (Figure 3a), indicating that there were no large pores for the calcein to pass through.Moreover, there was no big difference in dye leakage between DPPS-and DOPScontaining membranes, confirming that phase-preferential R 9 binding would be irrelevant to pore formation as well (Figure 3b).
Without the formation of large pores, arginine-rich CPPs can be internalized into PS-containing GUVs, but at different rates for different phases of lipid membranes (Figure 4).To determine which phase is more helpful for the penetration of R 9 , we prepared anionic L d or L o phase GUVs, which can be represented by DOPC/DOPS/cholesterol or DPPC/DPPS/cholesterol, respectively.To visualize and quantify the penetration of R 9 , we used FITC as a fluorophore conjugated to R 9 .We note that due to its anionic nature under the experimental pH, the physicochemical properties of the peptide can be altered, potentially affecting its ability to penetrate the membrane. [43,44]Into the L d phase GUVs, the penetration of FITC-R 9 was extremely rapid even at 5 μm FITC-R 9 , so that the fraction of R 9 -internalized GUVs reached 100% in just 10 min (Figure 4a, empty red diamond); note that we extrapolated the FITC-R 9 concentrations inside GUVs from their fluorescence intensities using confocal fluorescence microscopy images, such that GUVs with >0.2 μm FITC-R 9 were regarded as the R 9 -internalized GUVs (Figure S7, Supporting Information).On the other hand, the entry of FITC-R 9 into the L o phase GUVs occurred less frequently than that into the L d phase GUVs (empty black diamond).In particular, when comparing the quantities of FITC-R 9 between the two different GUVs, the L d phase GUVs composed of unsaturated lipids (Figure 4b) exhibited a higher accumulation of FITC-R 9 in their lumens compared to the L o phase GUVs composed of saturated lipids (Figure 4c), which is in good agreement with previous studies showing a rapid penetration of arginine-rich CPPs through the membranes composed of unsaturated lipids. [15,16]Higher concentrations of FITC-R 9 accelerated its accumulation in both the L d and L o phase GUVs (Figure S8a,b, Supporting Information), and the absence of anionic lipids caused neither calcein leakage nor FITC-R 9 penetration even at high R 9 concentrations (≈50 μm) (Figure S9, Supporting Information).
As we have previously demonstrated in planar membranes, the presence of arginine-rich CPPs in phase-separated GUVs can cause a further phase transition from L o to L d phase, facilitating their translocation across more fluidized membranes.Similar to the single-phase GUVs, PS-containing phase-separated GUVs allowed consistent penetration of FITC-R 9 , and the presence of anionic L d domains was more effective for membrane penetration than that of anionic L o domains; at 5 μm FITC-R 9 , 97.1% DOPScontaining GUVs were determined as R 9 -internalized GUVs within 10 min (Figure 4a, solid red square), and the DPPScontaining GUVs displayed 74.8% during the same period (solid black square).However, compared to the DPPS-containing homogeneous GUVs (empty black diamond) with the same content of PS and cholesterol (10 and 30 mol%, respectively), the DPPS-containing heterogeneous GUVs could accelerate the penetration of R 9 , attributed to the dissolution of L o domains into L d ones upon the phase-preferential R 9 binding.Lipid packing defects at the domain boundary of phase-separated membranes may also have a supplementary effect on enhancing R 9 internalization. [45]Even the time-dependent R 9 accumulations in the DPPS-containing heterogeneous GUVs (Figure 4e) were much larger than those in the DPPS-containing homogeneous GUVs at 5 μm FITC-R 9 (Figure 4c).Similar trends were observed at higher R 9 concentrations (Figure S8b-d, Supporting Information).Although the time-dependent R 9 accumula-tions were smaller in the DOPS-containing heterogeneous GUVs (Figure 4d; Figure S8c, Supporting Information) than those in the DOPS-containing homogeneous GUVs (Figure 4c; Figure S8a, Supporting Information), these would be due to the different area fraction of L d domains relevant to the total amount of unsaturated lipids.

Dynamics of R 9 Binding and Its Penetration into Phase-Separated GUVs
To further understand the arginine-rich CPP-mediated lipid rearrangements and its active translocation across laterally heterogeneous membranes, we visualized the R 9 binding and its penetration in real-time in the coexistence of L d and L o phase domains (Figure 5).Using FITC-R 9 (green) and TR-DHPE (red), we readily assess peptide binding at L d domains with strong red fluorescence and L o domains with weak red fluorescence; the phase-preferential R 9 binding and subsequent R 9 accumulation were evaluated with the intensity of FITC fluorescence at the rim of each phase and in the lumen of GUVs composed of DOPC/DPPC/DPPS/cholesterol (Figure 5a).In measuring the rim intensities, a straight line was drawn through three points: the L o phase region, the GUV center, and the L d phase region (Figure S10, Supporting Information).The first and second highest peaks in the TR-DHPE intensity profile corresponded to the L d and L o phase regions, respectively.The FITC intensity at each peak represented the concentration of FITC-R 9 at each phase domain, and we obtained the average values from 5 distinct lines.
As soon as the DPPS-containing heterogeneous GUVs were exposed to FITC-R 9 , their outer surfaces were fully covered with the bound FITC-R 9 , with a larger amount bound to the L o domain than at the L d domain (point (1) in Figure 5).For the following 200 s (from point (1) to ( 2)), the normalized intensity of L o phase rim (solid black square) was maintained at ≈0.5, while that of L d phase rim (solid red square) slowly increased, inferring that the FITC-R 9 would not pass through the L o domain but through the L d domain.However, at point (2), the intensity of L o phase rim increased with the time delay; presumably, the FITC-R 9 that entered the GUV through the L d domain would be bound to the inner surface of the L o domain rather than being free in the lumen, as there was no significant increase of accumulated FITC-R 9 (empty green square).Nevertheless, a certain fraction of the peptide did not adhere to the inner membrane surface and instead remained within the lumen.At point (3), the intensity of L d phase rim reached a plateau with no further increase, but that of GUV lumen began to increase rapidly, indicating continuous penetration of FITC-R 9 and its subsequent accumulation in the GUV.As the concentration of FITC-R 9 inside the GUV became sufficiently high, at point (4) the intensity of L o phase rim also reached a plateau, suggesting that even the inner surface of the L o domain would be completely occupied with the internalized FITC-R 9 .However, the intensity of the GUV lumen consistently increased with FITC-R 9 accumulation over time (from point (4) to ( 5)).
The preferential penetration of R 9 through the L d domain suggests that when L d and L o domains coexist in phase-separated membranes, translocation of arginine-rich CPPs would favor the thinner and less rigid domains; in other words, the ordered domains have a higher concentration threshold for peptide penetration.Based on the previously reported phase diagrams, [46][47][48] we estimated that with our membrane composition (10 mol% DPPS), the L d and L o phase domains would contain ≈7-13 mol% DPPS, respectively.Because of lower charge density, the L d domain was less covered with the FITC-R 9 than the L o domain, but the initial increase in the intensity of L d phase rim validated that the membrane flexibility would be more crucial for the pen-  2), the intensity of L o phase rim began to increase as the FITC-R 9 in the lumen bound to the inner surface of the L o domain.At (3), the intensity of L d phase rim reached a plateau, but that of the lumen began to increase rapidly.At (4), the intensity of L o phase rim also reached a plateau, but that of the lumen consistently increased from (4) to (5).The images in the last row are intensity mappings of the GUV lumen (white dashed squares in merged channels).Membrane composition: DOPC/DPPC/DPPS/cholesterol/biotin-cap-PE + TR-DHPE (33/25/10/30/2 + 0.2 mol%).Imaging was performed at 20 °C.Scale bar, 5 μm.
etration of FITC-R 9 .As the penetration of R 9 across the less anionic L d domains prevails over that across the more anionic L o domains, it is evident that the R 9 -induced domain dissolution from a more ordered state to a less ordered state would actively lead to their easier translocation across laterally heterogeneous membranes.This implies that the dissolution of the L o phase into the

Conclusion
In exploring how arginine-rich CPPs traverse laterally heterogeneous cell membranes, we directly visualized in real-time the dynamic reorganization of membranes with coexisting L d and L o phases mediated by R 9 binding and its penetration.By analyzing the fluorescence intensities of FITC-R 9 at the surface and in the lumen of phase-separated GUVs, we observed that R 9 translocated through the disordered domains rather than the ordered domains.This result is consistent with previous findings that the penetration of R 9 was readily observed when the membrane was mechanically weak and deformable, but became rare when it was not. [15,45,49]However, we also demonstrated that R 9 eventually dissolved the L o domains containing anionic lipids, suggesting that the arginine-rich CPPs actively alter the spatial and temporal distribution of membrane components.Presumably, in the heterogeneous GUVs, additional R 9 would lead to lateral expansion of ordered domains to form energetically unfavorable voids, and hydrocarbon chains of neighboring lipids would rearrange into a more disordered conformation to remove the free volumes in the bilayer. [50,51]Owing to the R 9 -induced membrane reorganization , the heterogeneous membranes would become fluidized, thereby facilitating subsequent translocation of R 9 across otherwise highly ordered, impermeable membranes (Figure 6).
The finding of this study, demonstrating that the dissolution of ordered membrane domains promotes the penetration of CPPs, is in accordance with established explanations underlying mem-brane permeability.Membrane thickness is an important factor in determining permeability across the membrane; the disordering of membrane lipids typically leads to a decrease in membrane thickness, [52] subsequently lowering the energy barrier for peptide penetration. [49,53]In particular, the transition of the membrane from L o to L d phase dramatically improves the permeability of small molecules through the membrane, [54,55] which is consistent with our results with R 9 .Furthermore, an increase in the lipid bilayer fluidity makes the bilayer more susceptible to further membrane reconstructions, such as alterations in membrane curvature [16,32,56] and redistribution of membrane components, [13,29,32,57] which have been proposed as a potential mechanism for CPP translocation.
In a cell, negatively charged lipids are not uniformly distributed throughout the plasma membrane.For instance, PS is predominantly located at the inner leaflet of the eukaryotic plasma membrane. [58]This asymmetry is disrupted in cancer cells, as well as other pathological and apoptotic cells. [59,60]Consequently, PS has emerged as a promising target for anticancer therapeutics, [61] including arginine-rich CPPs. [62,63]Additionally, a considerable fraction of PS is found to co-segregate with cholesterol to form distinct microdomains called lipid rafts, [64][65][66][67] and the lipid rafts are generally accepted to exist as L o domains. [68,69]Accordingly, it is typically regarded that penetration of CPPs would be significantly challenging through the highly ordered lipid rafts; previous studies have reported that increasing the structural order of membranes, particularly with increasing cholesterol content, suppresses the translocation of CPPs. [15,33,34]][72] In this regard, our current findings expand previous works by demonstrating that CPPs themselves can locally loosen the packing of pathological cell membranes through lateral membrane reorganization prior to their active translocation.We believe that this complements existing model membrane studies that have assessed the translocation of CPPs only with overall membrane properties, [73] by elucidating how locally concentrated CPPs reconstitute heterogeneous lipid membranes, thereby providing insight into highly effective drug delivery.Peptide Synthesis: All peptides were synthesized with the solid-phase peptide synthesis method.Rink amide MBHA resins (100 mg, 45 μmoL) were swollen with DMF (2 mL) in a 5 mL fritted syringe overnight.The Fmoc protecting group was removed by treatment with 20% piperidine in DMF (2 × 10 min).On the resin, arginine residues were added by a repeated standard route.In detail, for amino-functionalized resin conjugation, Fmoc-Arg(pbf)─OH (5 equiv.)were treated with 1 m HOBT (5 equiv.), 1 m HBTU (5 equiv.), and 1 m DIC (10 equiv.) in DMF.After shaking for 2 h, the reaction was finished and mixture was drained out and the resin was washed with DMF (7×).Unless noted, this washing step was repeated at each reaction step.This process was repeated until desired sequences of nona-arginine (R 9 ) were obtained.For cleavage, the beads were treated with 1 mL of a cleavage cocktail (95% TFA, 2.5% TIS, and 2.5% distilled water) for 2 h.To synthesize FITC-labeled R 9 (FITC-R 9 ), the N-terminus of R 9 was coupled with Abu linker by standard Fmoc chemistry.After deprotection, the amine group of the Abu was coupled with FITC (10 equiv.)using the peptide coupling chemistry.For cleavage, the beads were treated with 1 mL of a cleavage cocktail (95% TFA, 2.5% TIS, and 2.5% distilled water) for 2 h.

Experimental Section
After reactions were completed, the crude products were purified by high performance liquid chromatography (HPLC), and their purity and identity were analyzed by liquid chromatography-mass spectrometry (LC-MS) and matrix-assisted laser desorption ionization-time of flight mass spectrometry (MALDI-TOF MS).Analytical HPLC and LC-MS characterization were performed on an Agilent system with a C18 reversed-phase HPLC column (Agilent, 3.5 μm, 4.6 mm × 150 mm).A gradient elution of 10%-100% B in 7 min (keep 100% B till 13 min) was used at a flow rate of 0.7 mL mi −1 n (solvent A: distilled water, 0.01% TFA; solvent B: acetonitrile, 0.01% TFA).With a C18 reversed-phase column (Agilent, 5 μm, 21.2 mm × 150 mm) changing solvent composition with a linear gradient of 100% solvent A in 5 min followed by 100% solvent B in 65 min.MALDI-TOF MS was performed on 4700 Proteomics Analyzer (Applied Biosystems) and Autoflex speed LRF (Bruker) using -cyano-4-hydroxycinnamic acid and 2,5-dihydroxybenzoic acid as matrices.Small aliquots of the obtained R 9 and FITC-R 9 were flash-frozen using liquid nitrogen and stored at −20 °C.Peptides for long-term storage were freeze-dried.
Preparation of Planar Lipid Membrane Imaging: The imaging chamber was assembled with a bottom coverslip and side slide glasses.PLL-PEG was then used to passivate the glass coverslips.PLL-PEG was prepared by conjugation of amine-reactive PEG with PLL at a molar ratio of 1:5.The reaction was performed in 50 mm sodium tetraborate pH 8.5 solution.The mixture was continuously stirred overnight at room temperature, buffer exchanged into 10 mm HEPES, 100 mm NaCl, pH 7.4 using a D-Tube Dialyzer (MWCO 6-8 kDa, Novagen), and stored at 4 °C.The 200 μL of 0.5% PLL-PEG solution was added to each imaging chamber.After 20 min of incubation, the chamber was rinsed with 10 mm HEPES, 100 NaCl, and pH 7.4 buffer.
Freestanding Planar Lipid Membranes: Freestanding planar lipid membranes were prepared using an ultra-stable freestanding planar lipid membrane array, as described in previous studies. [37,38]First, the lipids dissolved in chloroform were mixed in a glass vial in the desired molar ratio and dried under a gentle N 2 stream.The dried lipids were dissolved in a mixture of hexadecane and silicone oil (1:1 v/v) to obtain a lipid oil solution with a total lipid concentration of 5 mm.Each oil was filtered through a 0.2 μm syringe filter (Whatman) to remove impurities before use.The lipid oil solution was sonicated for 1 h and used for experiments within 5 h.
A total of 1 mL of the aqueous buffer was injected into the PLL-PEG passivated imaging chamber, and 2 μL of lipid oil solution was gently dropped and spread on the air-aqueous buffer interface.≈3 min after dropping the lipid oil solution, a hexagonal TEM grid (G200HEX, G150HEX, G100HEX, Gilder Grids) treated with 1-dodecanethiol for hydrophobic coating was gently placed on the air-oil interface using tweezers.For the next 2 min, the hydrophobic surface of the TEM grid was wetted with the lipid oil solution, and the holes of the grid were filled with thin oil films.The grid was submerged into the aqueous buffer using a syringe needle to position it on the PLL-PEG-coated glass surface.As the oil drained, the thickness of the oil film decreased sufficiently for zipping of the two lipid monolayers, which resulted in the formation of the planar freestanding lipid bilayers.The lipid membrane array was incubated at 45 °C for at least 30 min and cooled to 24 °C.≈1 h after cooling, a cholesterol-methyl--cyclodextrin solution (4 mm, 20:1 mol%) was injected into the aqueous buffer at a concentration of 60 mm to increase the cholesterol concentration in the lipid membranes to ≈30 mol%.After ≈1 h of incubation, the prepared R 9 or FITC-R 9 solution was injected into the aqueous solution to the desired concentration.Experiments were done at 24 °C.The lipid membranes were directly visualized using an inverted fluorescence microscopy (IX73, Olympus) with an iXon EMCCD camera (X-6880, Andor Solis).Images were analyzed using the open-source program, Fiji. [74]UV Preparation: The lipids dissolved in chloroform were mixed to a desired ratio at a total concentration of 1 mg mL −1 .For all GUV experiments, 2 mol% of biotin-cap-PE was included for vesicle tethering on the glass surface.GUVs were prepared based on the electroformation method. [75]Briefly, the lipid mixture in chloroform was spread on indiumtin-oxide coated glass slides (30 Ω, Omniscience) and dried under vacuum for at least 2 h.Electroporation was performed at 55 °C in 200 mOsm L −1 glucose solution or 200 mOsm L −1 glucose solution with 10 μm calcein.The voltage was increased from 50 to 1400 mVpp for the first 30 min at 10 Hz, and then held at 1400 mVpp for 120 min.For the last 30 min, voltage and frequency were adjusted to 2200 mVpp and 5 Hz, respectively.
Coating of Glass Surface for GUV Tethering: GUVs were tethered on a glass coverslip using the strong interaction between biotin and streptavidin.The imaging chamber was assembled with a coverslip and 2 mm thick silicon gaskets with 3 mm diameter holes.Prior to chamber assembly, gaskets, and coverslips were cleaned with 2 vol% Micro-90 concentrated cleaning solution (International Products Corporation), rinsed thoroughly with water, and dried under a nitrogen stream.10 μL of 1 mg mL −1 biotin-BSA solution was added to the chamber and incubated for at least 30 min.Excess biotin-BSA was washed with 10 mm HEPES, 100 mm NaCl, and pH 7.4 buffer by gentle pipetting.Next, 0.2 mg mL −1 streptavidin was added to the chamber, incubated for at least 30 min, and then rinsed with the same buffer.
Peptide Internalization Assay and Fluorescence Leakage Assay with GUVs: GUVs were observed using a confocal laser scanning microscope (LSM880, Carl Zeiss) in the KAIST Analysis Center for Research Advancement (KARA).To investigate R 9 -induced pore formation of GUVs, calceinencapsulated GUVs were tethered on a glass coverslip and R 9 solution was added to the desired concentration.To investigate the internalization of fluorescence-labeled peptides into GUVs, calcein-free GUVs were tethered on a glass coverslip and FITC-R 9 solution was added to the desired concentration.GUVs were exposed to the laser only when imaging to minimize the photobleaching of the fluorescence probes.Imaging was done at 20 °C and images were analyzed using the open source program, Fiji. [74]tatistical Analysis: The normalized area and normalized calcein fluorescence intensities inside the GUVs values were obtained by normalizing to the initial values before peptide addition.The FITC fluorescence intensities at GUV rims were normalized to the final intensity of each rim, and the FITC intensities of GUV lumen were normalized to the bulk intensity outside the GUV.The data shown in the graphs were described as mean ± SD of at least two independent experiments.The number of samples for each analysis was specified in the figure legends.

Figure 1 .
Figure 1.Freestanding planar lipid membrane array for analyzing R 9 -induced membrane reorganization.a) Experimental setup for real-time monitoring of lateral membrane reorganization.Multiple lipid membranes are created in hexagonal holes of a transmission electron microscopy (TEM) grid when immersed in an aqueous buffer.The membrane array is observed in real-time through an inverted fluorescence microscope.The detailed experimental procedure is illustrated in Figure S1 (Supporting Information).b) Phase-separated lipid membrane containing anionic saturated phospholipid, DPPS.The lipid membrane composed of unsaturated phospholipids (DOPC), saturated phospholipids (DPPC), and cholesterol undergoes phase separation into a liquid-disordered (L d ) phase (bright) and liquid-ordered (L o ) phase (dark) at 24 °C.When cationic nona-arginine (R 9) is introduced, it electrostatically binds to the surface of lipid membranes, specifically interacting with anionic phospholipids.The heterogeneous membrane is formed using a lipid oil mixture of DOPC/DPPC/DPPS + TR-DHPE (60/30/10 + 0.5 mol%), and cholesterol is added via MCD.The prepared FITC-R 9 solution was injected at a concentration of 3 μm into the aqueous solution above the lipid membranes.The images were obtained ≈1 min after the FITC-R 9 addition.

Figure 2 .
Figure 2. R 9 -induced membrane reorganization on heterogeneous membranes.a-c) Membrane dynamics with different lipid compositions upon addition of 50 μm R 9 .No change in phase behavior was observed with neutral PC membranes a), whereas both phase separation and mixing were found with anionic DPPS-containing, phase-separated membranes (b-c).d) Normalized area of L o domains over time.The normalized area (A R ) is defined as A R = A/A 0 , where A and A 0 indicate the area of L o domains at each time point and the initial area of L o domains before the addition of R 9 .Data represent mean ± SD (n = 4-11).e) Fraction of phase-separated membranes as a function of temperature.The apparent miscibility transition temperature (T m ) was determined for bare DOPC/DPPC/DPPS/cholesterol membranes (30.1 °C) and after 30 min incubation with 50 μm R 9 (28.2 °C) by fitting the counted fractions to a linear function.Data were obtained from n = 40-92 planar membranes.f) Left: Normalized area of L o domains over time.R 9 concentrations increased from 5 to 50 μm every 30 min.Data represent mean ± SD (n = 2).Right: Phase-separated membranes at each R 9 concentration at each time.L o domain dissolution was accelerated with increasing R 9 concentrations.The membranes were formed using a lipid oil mixture of DOPC/DPPC + TR-DHPE (60/40 + 0.5 mol%) a), DOPC/DPPC/DPPS + TR-DHPE (60/30/10 + 0.5 mol%) b-f), DOPC/DPPC/DPPS + TR-DHPE (75/20/5 + 0.5 mol%) c), and cholesterol was added via MCD.Imaging was performed at 24 °C.Scale bars, 50 μm.

Figure 5 .
Figure 5. Real-time tracking of FITC-R 9 entry into a DOPC/DPPC/DPPS/cholesterol GUV.a) Time-dependent FITC-R 9 accumulations at the rim and in the lumen of the phase-separated GUV.The FITC fluorescence intensities of L o and L d phase rims (solid black and red squares, respectively) were normalized to the final intensity of L o phase rim.Data represent mean ± SD (n = 5).The intensities of the lumen (empty green square) were normalized to the bulk intensity outside the GUV.b) Confocal fluorescence microscopy images of the phase-separated GUV at each time point.At (1), the binding of FITC-R 9 to the outer surface of the GUV reached equilibrium.Between (1) and (2), only the intensity of L d phase rim increased, indicating that the FITC-R 9 would be continuously translocated through the L d domain.From (2), the intensity of L o phase rim began to increase as the FITC-R 9 in the lumen bound to the inner surface of the L o domain.At (3), the intensity of L d phase rim reached a plateau, but that of the lumen began to increase rapidly.At (4), the intensity of L o phase rim also reached a plateau, but that of the lumen consistently increased from (4) to(5).The images in the last row are intensity mappings of the GUV lumen (white dashed squares in merged channels).Membrane composition: DOPC/DPPC/DPPS/cholesterol/biotin-cap-PE + TR-DHPE (33/25/10/30/2 + 0.2 mol%).Imaging was performed at 20 °C.Scale bar, 5 μm.

Figure 6 .
Figure 6.The potential mechanism for accelerating direct translocation of R 9 into laterally heterogeneous membranes with membrane reconstitution and permeabilization.Cationic R 9 binds to anionic L o phase domains due to their strong electrostatic interactions.R 9 binding actively induces lipid rearrangements, allowing phase transition from L o to L d phase for membrane fluidization.As a more loosely packed state, the reconstituted L d phase domain becomes a route for accelerating the direct translocation of R 9 across the phase-separated membranes.
L d phase induced by R 9 plays a substantial role in facilitating the penetration of the peptide.