Effect of targeted minimal sequence variations on the structure and biological activities of the human cathelicidin LL-37

2.1 Design of the LL-37 analogs

LL-37 analogs were designed by introducing the sequence of LL-37 and RL-37 in an ad hoc created Excel file capable of counting Arg, Lys, Asp, and Glu residues 4 or 5 positions from each other (i→ i + 3 or i→ i + 4). As these would be on the same side of a helix, they could result in intramolecular electrostatic interactions. LL-37 analogs were then designed by either mutating just one residue in a strategic position that potentially engages in a number of possible attraction interactions, or by interchanging pairs of oppositely charged residues. This can alter their capacity to form salt-bridges, with little or no effect on the overall charge and a minimal effect on physico-chemical parameters such as the overall hydrophobicity or helix amphipathicity (hydrophobic moment).

In particular, considering the sequence of LL-37, we identified the following numbered residues as potentially involved in salt bridging:

LLGDFFRKSK¹⁰E¹¹KIGK¹⁵E¹⁶FK¹⁸RIVQRIKDFLRNLVPRTES.

We therefore designed one analog in which K¹⁰ and E¹¹ were interchanged. In a second one E¹⁶ and K¹⁸ were interchanged. In a third analog, both these interchanges were effected. In the fourth analog, K¹⁵ was replaced by Gly, which occurs in this position in the RL-37 sequence (see Table 1).

To determine how these variations could affect salt bridging, the sequences of RL-37, LL-37 and the four analogs were submitted to the QUARK ab initio protein structure prediction server,21 and the resulting predicted structures visualized with the DS Visualizer (Accelrys Software) to determine which salt-bridging interaction could occur, in terms of oppositely charged residues spaced <4 Å apart.

2.2 Peptide synthesis and purification

All peptides were synthesized with a CEM Liberty automated microwave assisted peptide synthesizer, using Fmoc solid-phase chemistry and the Fmoc-Ser(tBu)-Novasyn-TGA resin (substitution 0.24 mmol/g, Novabiochem). All protected amino acids were from Novabiochem (Merck Millipore). Couplings were performed with a 4-fold excess of incoming amino acids in NMP (N-methyl pyrrolidone) and were activated with either PyBOP or HCTU. Double couplings were performed as appropriate, having used the Peptide Companion software (Coshi Soft) to predict difficult couplings. The deprotection of the Fmoc group was performed using a solution of 20% piperidine in DMF. On synthesis completion, peptides were cleaved from the resin using a version of the reagent K composed of 87% TFA (trifluoroacetic acid), 2% TIPS (triisopropylsilane), 3% water and 8% DODT (3,6-dioxa-1,8-octanedithiol) for 4 h at room temperature (22°C), under continuous stirring, and then precipitating in ice-cold TBME (methyl t-butyl ether).

The correct sequence of the crude peptides was confirmed by ESI-MS (Bruker Daltonics Esquire 4000) before purifying by RP-HPLC using a Phenomenex Jupiter Proteo C18 column (20 × 250 mm, 10 μm, 90 Å) using a gradient of 30%-60% acetonitrile in 60 min with a 8 mL/min flow. Peptide purity (> 95%) was confirmed by ESI–MS and analytical RP-HPLC. The peptides were lyophilized three times from 10 mM HCl suspensions to remove TFA counterions, and resuspended in water. The concentration of peptide stock solutions was quantified by (i) accurately weighing the peptide used to prepare solutions, (ii) spectrophotometric determination of the peptide bonds using ɛ₂₁₄ calculated as described22 and (iii) by measuring the differential absorbance at 215 nm and 225 nm,23 taking the mean value of the three.

The structure and biological activities of the peptides were compared to those of the parent peptide LL-37, its rhesus orthologue RL-37, and a dimeric version of LL-37 linked via an intermolecular S-S bridge involving an added C-terminal cysteine residues, to produce the covalent dimer [LL-37(C³⁸)]₂ (C-terminal dimer or CTD in the text). The preparation of these peptides has been described previously.12

2.3 Circular dichroism

CD spectroscopy was performed on a J-715 spectropolarimeter (Jasco) using 2-mm path length quartz cells and 20 μM peptides in water, 10 mM SPB (10 mM Na₂HPO₄/NaH₂PO₄, pH = 7.4), PIL buffer (Physiologic ion like; 130 mM NaCl, 24 mM NaHCO₃, 0.6 mM MgCl₂, 1.3 mM CaCl₂, and 3.9 mM KCl, pH 7.3), and at increasing concentrations of trifluoroethanol (TFE) in SPB. The helical content (percentage helix) was estimated from the molar ellipticity at 222 nm.24 CD spectra are the accumulation of at least three scans.

2.4 Antibacterial activity and membrane permeabilization

Antimicrobial activity was evaluated against Escherichia coli ATCC 25922 and Staphylococcus aureus ATCC 25923. MIC values were determined using the serial 2-fold dilution assay for each peptide using 20% Mueller-Hinton broth (MHB) in 10 mM SPB, pH 7.3 (as LL-37 is medium sensitive) in 96-well polypropylene microtitre plates (Sarstedt). Bacterial growth was monitored also in the absence of peptide as a control. Bacterial suspension (50 μL) with 5 × 10⁵ cells/mL in 20% MHB was added in each well containing 50 μL peptide solution in 20% MHB. The MIC was taken as the lowest concentration where no visible bacterial growth was observable after 24 h of incubation at 37°C. Bactericidal activity was determined by incubating 1 x 10⁶ cells/mL in 20% MHB with concentrations near the MIC for 60 min, serially diluting in PBS and then plating on MH agar. Colony counts were made after overnight incubation of the plates.

To determine the effect of peptides on bacterial growth kinetics, 10⁶ cells/mL were exposed to different peptide concentrations in 20% or 100% MHB (pH 7.2). Bacterial growth curves were obtained by monitoring the optical density at 600 nm for 4 h at 37°C, using a Tecan Sunrise microtiter plate reader. Quantification of growth inhibition (I, expressed in percentage) was calculated as %I = 1 – (A_P/A₀) × 100, where A₀ is the measured absorbance at 210 min in the absence of peptide and A_P the absorbance at a given peptide concentration at that time. The IC₅₀ value for growth inhibition was calculated as the mean peptide concentration that reduced growth to 50%, at 210 min.

2.5 Flow cytometry assays

Flow cytometric analyzes of membrane permeabilization were performed on a mid-exponential culture of S. aureus (10⁶ cells/mL) in 20% MHB in 10 mM SPB, treated with 2–8 μM peptide concentrations for 15–60 min at RT. Propidium iodide (PI, Sigma–Aldrich) or FITC-labeled dextran of 4, 10, 20, or 40 kDa (FD4, FD10, FD20, or FD40, Sigma–Aldrich) were then added to final concentrations of 10 μg/mL or 0.25 mg/mL, respectively.

Measurements were performed on a CYTOMICS^TM FC500 (Beckman Coulter Inc. Fullerton, California), equipped with an Argon laser (488 nm, 5 mV) and standard configuration with photomultiplier tube fluorescence detector for green (525 nm, FL1), orange (575 nm, FL2) or red (610 nm, FL3) filtered light. After acquisition of at least 10,000 events per each run, data were stored as list mode files and analyzed with the FCS Express V3 software.

2.6 Surface plasmon resonance studies

Inter-peptide interactions and peptide-membrane interactions were studied using an X100 instrument (Biacore, GE Lifesciences). Conditions for anchoring LL-37 molecules have been described previously13 and were used also for anchoring a selected analog. Briefly LL-37(K¹⁶E¹⁸), suspended in sodium acetate buffer (pH 5), was anchored to a CM5 chip following the protocol provided with the Biacore EDC/NHS coupling kit. The acidic conditions and peptide concentration used (50 μg/mL) were such that the peptide was most likely to deposit on the surface as monomeric molecules.13 The level of the substitution reached after immobilization was equivalent to 200 response units (RU). At these peptide concentrations, the anchored molecules on the chip should be well separated (∼4 nm apart), given that 100 RU corresponds to 0.1 ng/mm² of the surface.25 Peptide-peptide interaction studies were performed either in mildly basic conditions using HBS-EP (0.01 M HEPES pH 7.4, 0.15 M NaCl, 3 mM EDTA, 0.005% v/v Surfactant P20, GE Lifesciences), which favor helical structuring and oligomerization for LL-37. Free LL-37(K¹⁶E¹⁸) solutions with concentrations ranging from 4 to 64 μM were then flowed in successive cycles for 180 s of contact time and 300 of dissociation time across the surface of the LL-37(K¹⁶E¹⁸)-modified chip to collect sensorgrams. The analyte was then washed away after each cycle by injecting 50 mM NaOH for 60 s.

For studies with immobilized membrane, large unilamellar vesicles (LUVs) were prepared using 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC). Dry lipids were dissolved in a chloroform/methanol mixture (2:1 v/v), the solvent was removed by evaporation under a stream of nitrogen, and resulting cake vacuum-dried for 3 h. The dry lipid cake was re-suspended to a concentration of 1 mM in PBS by spinning the flask at a temperature above the transition temperature. The resulting multilamellar vesicle suspensions were then disrupted by several freeze-thaw cycles prior to extrusion through polycarbonate filters with 100 nm pores using a Mini-Extruder (Avanti Polar Lipids, Inc.). LUVs were used within a day to prepare the sensor chip surface. The surface of an L1 sensor chip was preconditioned with 40 mM of the nonionic detergent octylglucopyranoside (OGP). The DOPC liposomes were immobilized on the sensor surface by injecting three times for 10 min with flowrate of 5μl/min, until about 8000 RU was reached. Subsequently a 10 mM sodium hydroxide solution was injected (30 µL/min for 30 sec) to remove poorly anchored liposomes. BSA at a concentration of 0.1 mg/mL, was then injected at 5 µL/min for 60 sec for complete coverage of the nonspecific binding sites for polypeptides. This procedure results in integral liposomes on the chip surface, without fusion.26 To detect the binding of peptides with the liposomes, solutions with increasing peptide concentrations (0.5 to 16 μM) were injected at a constant flow rate of 10 µL/min for a contact time of 540 s, followed by a dissociation time of 1200 s with PBS buffer, to allow the peptides to return to bulk solution.

Sensorgrams for each peptide-peptide or peptide-liposomes interaction were obtained using BIAevaluation software and then elaborated using GraphPad.

2.7 Statistical analysis

Data obtained from repeated experiments were subjected to computer-assisted analysis using GraphPad Instat 3, and statistical significance was assumed at P ≤ .05 (ANOVA, Student-Newman-Keuls post test); EC₅₀ values were extrapolated by regression-correlation analysis performed by GraphPad Instat 3 from experimental concentration-effect curves (r² ≥0.9).

3.1 Design of the LL-37 analogs and potential effect on salt-bridge formation

The role of salt bridges in inducing LL-37 to fold as a stable helix, driving its oligomerization in aqueous solution, was studied by designing LL-37 analogs in which intramolecular electrostatic interactions (salt bridges) that could potentially stabilize the structure were altered. It is well known that ion pairs of oppositely charged residues spaced four or five positions apart in a helix can significantly stabilize its structure.27-29 The LL-37 variants were rationally designed to affect salt-bridging while causing minimal variations in the primary sequence, and were limited to either a single residue substitution (K¹⁵ → G in LL-37G¹⁵, see Table 1), to the inversion of a pair of Lys and Glu residues (e.g., K¹⁰E¹¹ → E¹⁰K¹¹ or E¹⁶K¹⁸ → K¹⁶E¹⁸) or to the inversion of both pairs of Lys and Glu residues (K¹⁰E¹¹E¹⁶K¹⁸ → E¹⁰K¹¹K¹⁶E¹⁸). Apart from LL-37(G¹⁵), which has a decreased charge and slightly altered hydrophobicity/amphipathicity characteristics, the other three have the same amino acid composition, charge and overall hydrophobicity/amphipathicity as the parent LL-37.

The inversion of Glu and Lys residues alters the possible salt bridges that can be formed with residues up or downstream. A first parameter taken into account was the ratio of oppositely charged residues in the peptides that could result in potential attractions, to that of similarly charged residues that could only result in repulsions (A/R, see Table 1). This is altered in favor of repulsions in the inversion variants, with respect to LL-37, and the overall number of potential attractions is decreased. For LL-37(G¹⁵), the A/R ratio is favorable but the overall number of possible attractions is in any case decreased. This preliminary assessment would thus suggest that all analogs could have decreased salt bridging.

Considering what salt-bridges may actually occur, several studies on helical peptides have indicated that the opposite charges on side-chains need to be spaced less than 4Å apart for these to be stabilizing.27, 30, 31 Furthermore, salt-bridges involving certain residues are reported to be more stabilizing than others.30-33 These considerations suggest that it is quite unlikely for all of the possible salt-bridges in LL-37 to occur simultaneously (e.g., a charged residue may not be able to form bridges in both N- and C-terminal directions). To try and assess what bridges could occur, the predicted helical structure of LL-37 and RL-37 were obtained using the Quark server34 (see Figure 1), and inter-charge distances measured using DS Visualizer.

The figure shows how LL-37 can form three salt bridges involving D⁴↔K8, R⁷↔E11, and E¹⁶↔R¹⁹. Furthermore, there is a noncanonical bridge between K²⁵ and E³⁶ involving a C-terminal loop in the peptide. For RL-37, only the R⁷↔E¹¹ bridge was observed, as well as a non canonical one involving K²⁵ and the C terminus. This study would seem to confirm that residues selected far sequence inversion (E¹¹ and E¹⁶⁾ may indeed be involved in stabilizing salt-bridges, while that selected for replacement (K¹⁵) may not.

The sequences of all the analogs were also submitted to the Quark server and visualized, and the predicted salt-bridges are shown in supplementary materials Supporting Information Figure S1. LL-37(E¹⁰K¹¹) and LL-37(K¹⁶E¹⁸) each show a disrupted bridge (respectively R⁷↔E¹¹ and E¹⁶↔R¹⁹), while in LL-37(E¹⁰K¹¹ K¹⁶E¹⁸) both are disrupted, although formation of a new, noncanonical E¹⁸-R¹⁹ bridge is observed. As expected, LL-37G¹⁵ does not present significant bridge disruption, although the pattern is subtly different from that of LL-37, with a bipartite bridge involving R⁷. Furthermore, all the analogs show the additional noncanonical bridge involving a residue in the N-terminal part of the helix (either K²⁵ or R²⁹) and E³⁶ in the C-terminal loop or the C-terminus itself. Overall, these models suggest that the dynamic nature of the helix can allow for compensatory salt-bridging to occur if specific salt bridges are disrupted.

The analysis indicates that all three inversion variants LL-37(E¹⁰K¹¹), LL-37(K¹⁶E¹⁸) and LL-37(E¹⁰K¹¹ K¹⁶E¹⁸) could have a destabilized salt bridging pattern that should result in lower propensity to adopt a helical structure in solution, which could in turn affect oligomerization. In other words, they might behave more like F-form peptides. The LL-37(G¹⁵) variant could instead maintain a similar structuring propensity to LL-37, although the region in the center of the peptide would be more flexible.

3.2 Structure stability

The secondary structure of the four LL-37 variants was characterized by CD spectroscopy and compared with that of the parent peptide (Figure 2). Spectra were collected in the presence of water, where LL-37 does not adopt a helical structure; in low salt concentration buffer (10 mM SPB, pH 7), where it starts to acquire this conformation, and a in buffer with plasma-like ionic concentrations (PIL), which strongly stabilizes it. They were also measured in the presence of increasing concentrations of TFE (up to 50% in SPB), a solvent known to induce helix formation in peptides.

All peptides were effectively random coil in pure water, as expected, and showed a very similar helical content in the presence of 50% TFE (see Table 1), suggesting that the slight variations in sequence did not intrinsically affect helix stability in the presence of this strongly helix-inducing solvent. Under these conditions, the determined helix content of ∼ 60% can be considered to be the maximum achievable, consistent with end effects in peptides of this length. The behavior at lower concentrations of TFE (e.g., 10%) is more varied (see Figure 2 and Supporting Information Figure S2 for spectra). The different extent of helix formation and the shape of the spectra suggest that altering the salt bridging pattern may result in different folding pathways. In any case, the analogs behaved more like A-form LL-37 than the F-form RL-37, suggesting either that the residual salt bridging was sufficient to stabilize the helix, or the dynamic secondary structure was capable of forming compensating bridges.

In 10 mM SPB, the CD spectrum for LL-37 indicates a helix content estimated at about 25%, suggesting that the peptide is in equilibrium between coil and helical forms (see Figure 2). The helix content is reduced to 15% for LL-37(K¹⁶E¹⁸) and LL-37(E¹⁰K¹¹K¹⁶E¹⁸), whereas it is unaltered for LL-37(E¹⁰K¹¹) (25%, see Table 1), suggesting that the central E¹⁶-R¹⁹ bridge in LL-37 is more relevant for structuring. The stability of the LL-37(G¹⁵) analog is also slightly decreased. On increasing the concentration of salt ions, in PIL buffer, the structural differences between LL-37 and its analogs are attenuated, as both inter- and intramolecular interactions are favored. Note that RL-37 is still random coil under these conditions.8, 10, 11 Taken together, these data would suggest that intramolecular salt bridges do contribute to helix formation in an aqueous environment, but are not the only relevant factors.

3.3 Surface plasmon resonance studies of peptide self-association

Surface plasmon resonance (SPR) is a useful technique for studying interactions between molecules, or between molecules and supported membranes. We have previously reported the use of a Biacore setup to observe self-assembly of free flowing LL-37 (analyte) with LL-37 molecules anchored on the Biacore sensor surface (ligand).13 Here we report the same experiment, but using its analog LL-37(K¹⁶E¹⁸) as both analyte and ligand, as it has the least stabilized structure in CD studies.

Peptide molecules were anchored to the carboxydextran moieties covering the surface of a CM5 chip by formation of amide bonds with the N-terminal amines. Anchoring conditions were chosen so that covalently bound peptide molecules were likely to be spread out and monomeric.13 Peptide association was then investigated by flowing a peptide solution over the chip surface, at increasing concentrations (4–64 μM). When increasing concentrations of LL-37 were flowed over the LL-37-modified chip surface under conditions that induce helix formation (HBS-EP buffer, pH 7.4, as verified by CD), a moderate, linear increase in response units was observed up to 16 μM. It then increased sharply, nonlinearly and reversibly, suggesting a high level of reversible aggregation onto the surface-bound LL-37 molecules (Figure 3A). This behavior was not observed under acidic conditions (acetate buffer, pH 5), in which the peptide remains unstructured.13

The sensorgrams obtained by flowing LL-37(K¹⁶E¹⁸) analyte over an LL-37(K¹⁶E¹⁸)-modified CM5 chip in HBS-EP buffer are shown for comparison in Figure 3B. Despite having a higher intrinsic loading than LL-37 (RU_max = 200 and 50 respectively), LL-37(K¹⁶E¹⁸) reaches a signal which is only a quarter that for LL-37 at the highest concentration, suggesting that it does have a decreased capacity for oligomerization. Nonetheless, the trend is the same as for LL-37, so that there is a significant jump in the signal at >16 μM, suggesting that self-aggregation may still occur, in a similar manner to the parent peptide, but to a reduced extent.

3.4 Interaction with model membranes

Sensorgrams for increasing concentrations of the LL-37 analogs flowed onto LUVs immobilized on an L1 chip are shown in Figure 4, with those for LL-37, its covalent dimer CTD12 and RL-37 also shown for comparison. All sensorgrams were obtained under the same conditions, and we chose to use DOPC LUVs as the membranes are neutral, and should not be affected by the peptide's charge or reflect transient electrostatic interactions. The conditions under which the chip was prepared each time, and the phospholipid composition of the LUVs, were such that these were likely to remain intact on anchoring.25

The sensorgrams for LL-37 and RL-37 indicate that the two peptides have different modes of interaction with the membrane. Response units increase gradually with concentration for RL-37, reaching a maximum of about 1500 RU, whereas they increase more slowly for LL-37 and then dramatically to over 4000 RU above a threshold concentration of 8 μM. This behavior is similar to that previously observed with both neutral DPPC LUVs, and anionic DPPC/DPPG (4:1) LUVs.13 These results suggest that both peptides can interact with and likely insert into the membrane, but that at higher concentrations only LL-37 undergoes significant self-association at the membrane, or inserts into it as an oligomer. The behavior of CTD would seem to confirm this, as its obligate dimeric structure shifts the threshold for the RU jump to 1 μM, and also results in a decreased off rate, which suggests membrane insertion. At >4 μM, saturation seaems to set in, as the signal no longer increases while the off rate does, suggesting a transient surface binding.

The behavior of the LL-37 analogs did not conform to the behavior of either RL-37 or LL-37, but was in fact somewhat surprising. It seems intermediate to that of the two reference peptides up to about 2 μM, but then the shape of the sensorgram changes dramatically, with a marked loss of RU during both injection and washing, going to negative values. This is consistent with a disruptive effect of the peptides on the LUVs, and consequent loss of material.

3.5 Antibacterial activity of the LL37 analogs

The capacity of the analogs to inhibit bacteria in vitro was evaluated by determining the MIC and the growth inhibition IC₅₀ values against two reference strains (Table 2). A potent antibacterial activity was observed against the gram-negative E. coli in 20% MHB, for all peptides with the exception of the covalently linked dimer CTD (likely due to its decreased capacity to penetrate the bacterial outer membrane12). It was not possible to observe significant differences between the activity of the variants and LL-37 or RL-37 under these conditions.

The effect on bacterial growth kinetics is a more sensitive method than MIC determination to observe activity differences. Furthermore, the antimicrobial activity of LL-37 is known to depend strongly on medium interactions, so that the difference in activity with respect to RL-37 is more marked in full medium.11 For this reason, the IC₅₀ values for inhibition of bacterial growth kinetics was determined for both 20% and 100% MHB (values in parentheses, Table 2). Under these conditions, variants were more active than LL-37, and activity was comparable to that of RL-37. This indicates that the medium sensitivity of LL-37 depends quite specifically on its structural characteristics.

The antimicrobial activity is more diversified against the Gram-positive S. aureus, with RL-37 showing a significantly more potent activity than LL-37 and its analogs, while CTD is effectively inactive, even in 20% MHB. It has been proposed that the A-form of LL-37 reduces its capacity to reach the bacterial membrane through peptidoglycan layer, and this is particularly evident for the obligate dimer CTD.11, 12 Among the analogs, only LL-37(K¹⁶E¹⁸) showed a slightly improved MIC value with respect to LL-37, and a correspondingly lower IC₅₀ value for S. aureus growth inhibition. The other analogs show a similar activity to that of LL-37, or a reduced activity in the case of LL-37(G¹⁵) (likely due to its lower charge of +5). These results suggests that while limited disruption of salt bridging in LL-37 can affect some of its functional characteristics as an A-form peptide (i.e., with respect to E. coli), it does not result in a complete switch to an F-form type of function.

3.6 Bacterial membrane permeabilization by LL-37 analogs

S. aureus cells were exposed to increasing peptide concentrations in the presence of propidium iodide (PI), to determine the capacity of these to permeabilize the bacterial membrane. The concentration dependent effects of RL-37, LL-37, and CTD are shown in Figure 5A, and RL-37 is more potent at permeabilizing the membrane than LL-37, with 3-fold lower PI₅₀ (concentration at which 50% of cells are PI+). CTD turns out to be the most potent in permeabilizing the bacterial membrane, with a PI₅₀ < 0.5 mM, apparently in contrast with its low antimicrobial potency. From this, one may deduce that it can cause membrane permeabilization at quite low concentrations, but that it is not sufficient to inactivate the bacterium, or that it is somehow able to recover, until it is overwhelmed at the significantly higher peptide concentrations required to overcome the sequestering effect of the peptidoglycan layer.

Permeabilization of the S. aureus membrane to PI by the LL-37 analogs is shown in Figure 5B, and while LL-37(G¹⁵) behaves similarly to LL-37, the concentration dependence of permeabilization for the three inversion analogs is different to both that of LL-37 or RL-37.

3.7 Analysis of bacterial membrane lesions using fluorescent probes

It has previously been possible to distinguish the type of lesion caused by LL-37 and RL-37 on model membranes using atomic force microscopy (AFM), which suggested that LL-37 formed discrete pores while RL-37 a more heterogeneous membrane disruption.10 In later studies, a flow cytometric method showed that LL-37 permeabilized the membrane less efficiently to PI than RL-37, but allowed passage of fluorescently labeled dextran particles of up to 40 kDA (9 nm) at higher concentrations, while RL-37 did not, irrespective of its concentration.12, 13 This is consistent with the formation of relatively large, discrete pores by LL-37, and more widespread but smaller heterogeneous lesions for RL-37, in accordance with a carpet-like detergent effect, and in agreement with the AFM experiments.

A similar experiment was performed with the LL-37 analogs, in comparison to RL-37, LL-37, and CTD, using all the peptides at the same concentrations (2–8 μm) and fluorescent probes of several different sizes, ranging from 1.6 nm diameter (PI) to 9 nm diameter (FITC-dextrans) (see Figure 6). The experiment confirmed that RL-37 only forms small lesions, allowing efficient entry of PI (populated LRQ, see Figure 6B) but not of larger probes (≥ 2.8 nm). This was observed not only at 2 μM peptide (not shown), but also at the relatively high peptide concentration of 8 μM (see Figure 6B). LL-37, on the other hand, permeabilized significantly less to PI at 2 μM (not shown), but at higher concentrations allowed efficient entry of both PI and FD40 dextran probe (populated URQ, see Figure 6C). With respect to the LL-37 analogs, the effect appears different to that of either LL-37 or RL-37. LL-37(G¹⁵) and (E¹⁰K¹¹) showed a behavior resembling that of LL-37 (Figure 6E,F), although they are less efficient in permeabilising to the larger FD40 probe. LL-37(K¹⁶E¹⁸) and (E¹⁰K¹¹K¹⁶E¹⁸) behave more like of RL-37 (Figure 6G,H).

Permeabilization to probes of different sizes, at the intermediate peptide concentration of 4 μM, is summarized in Table 3. RL-37 only permeabilizes to the smaller PI probe. LL-37 permeabilizes with decreasing efficiency up to a probe size of 9 nm, while the obligate dimer CTD permeabilizes efficiently to all probe sizes. All variants are less efficient than either RL-37 or LL-37 in permeabilizing to PI, but some are more efficient than RL-37 in permeabilising to the 2.8 nm probe.