Mechanisms of bacterial membrane permeabilization by crotalicidin (Ctn) and its fragment Ctn(15–34), antimicrobial peptides from rattlesnake venom

Crotalicidin (Ctn), a cathelicidin-related peptide from the venom of a South American rattlesnake, possesses potent antimicrobial, antitumor, and antifungal properties. Previously, we have shown that its C-terminal fragment, Ctn(15–34), retains the antimicrobial and antitumor activities but is less toxic to healthy cells and has improved serum stability. Here, we investigated the mechanisms of action of Ctn and Ctn(15–34) against Gram-negative bacteria. Both peptides were bactericidal, killing ∼90% of Escherichia coli and Pseudomonas aeruginosa cells within 90–120 and 5–30 min, respectively. Studies of ζ potential at the bacterial cell membrane suggested that both peptides accumulate at and neutralize negative charges on the bacterial surface. Flow cytometry experiments confirmed that both peptides permeabilize the bacterial cell membrane but suggested slightly different mechanisms of action. Ctn(15–34) permeabilized the membrane immediately upon addition to the cells, whereas Ctn had a lag phase before inducing membrane damage and exhibited more complex cell-killing activity, probably because of two different modes of membrane permeabilization. Using surface plasmon resonance and leakage assays with model vesicles, we confirmed that Ctn(15–34) binds to and disrupts lipid membranes and also observed that Ctn(15–34) has a preference for vesicles that mimic bacterial or tumor cell membranes. Atomic force microscopy visualized the effect of these peptides on bacterial cells, and confocal microscopy confirmed their localization on the bacterial surface. Our studies shed light onto the antimicrobial mechanisms of Ctn and Ctn(15–34), suggesting Ctn(15–34) as a promising lead for development as an antibacterial/antitumor agent.


binds to and disrupts lipid membranes and also observed that Ctn(15-34) has a preference for vesicles that mimic bacterial or tumor cell membranes. Atomic force microscopy visualized the effect of these peptides on bacterial cells, and confocal microscopy confirmed their localization on the bacterial surface. Our studies shed light onto the antimicrobial mechanisms of Ctn and Ctn(15-34), suggesting Ctn(15-34) as a promising lead for development as an antibacterial/antitumor agent.
New antimicrobial drugs are urgently needed to address the growing challenge of bacterial resistance to existing antibiotics. Misuse of classical antibiotics has increased the number of superbugs and created a critical situation whereby previously controlled pathogens could in the future cause major morbidity or mortality (1,2). This alarming growth of multidrug-resistant pathogens has prompted an intensive search for anti-infective drugs with novel mechanisms of action (3,4). In particular, antimicrobial peptides (AMPs) 9 have emerged as promising alternatives due to their broad-spectrum activity (including superbugs), selectivity, and mechanisms of action that potentially hinder the development of resistance (5).
AMPs are ancient weapons of the host defense machinery, present in all life domains (6). Although they can act in several possible ways to accomplish microbial cell death (e.g. membrane disruption, apoptosis induction, and internal target inhi-bition) (7,8), an initial common step in the process is their recruitment onto the bacterial cell surface (9,10). Accordingly, most AMPs display fairly conserved structural and physicochemical properties, such as positive net charge, high content of hydrophobic amino acid residues, or amphipathic structure, all favoring interaction with and insertion into membranes (11,12). Cathelicidins are a large family of AMPs whose unifying feature is the presence of a conserved cathelin (cathepsin L inhibitor) domain at the N terminus. In contrast, their C-terminal domains contain a mature and active AMP and display high inter-and intraspecies diversity (13). Cathelicidins have been shown to be active against a broad range of targets, including bacteria, enveloped viruses, and fungi (14). In addition to causing direct pathogen killing, cathelicidins can modulate the immune response by assisting with pathogen clearance (15). Cathelicidins have been isolated from a wide range of organisms, including mammals (16 -18), birds (19), fish (20), frogs (21), and marine (22) and terrestrial snakes (23)(24)(25).

Synthetic peptides
The amino acid sequences in Table 1 were prepared in C-terminal amide form by Fmoc (N-(9-fluorenyl)methoxycarbonyl) solid-phase synthesis and purified to Ͼ95% purity (see Fig. S1 for HPLC and MS data). For the N-terminal rhodamine B (RhB)-labeled peptides, the two peaks observed in the chromatograms of purified compounds (Fig. S1a) are due to RhB atropisomerism. The overall hydrophobicity of unlabeled and RhB-labeled peptides can be compared by the percentage of acetonitrile (ACN) at which they elute in HPLC (Table 1). Values for unlabeled Ctn and Ctn(15-34) were 30.9 and 30.3% ACN, respectively, whereas RhB-labeled versions eluted at 34.7 and 34.2% ACN, respectively, underlining the increase in hydrophobicity brought about by RhB labeling.

Time-resolved peptide uptake and bacterial death
Changes in membrane permeabilization and peptide uptake were monitored by a time-resolved flow cytometry assay, which allows a more accurate study of the kinetics of peptide effect, compared with kinetic studies based on end-point sampling. We studied the membrane permeabilization of E. coli using SYTOX Green dye right after Ctn or Ctn (15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30)(31)(32)(33)(34) addition, monitoring the changes from negative to positive gates for 90 min. Fig. 3 (a and b) shows time-course results and negative and positive controls (for the whole acquisition, see Movies S1a and S1b for Ctn and Ctn(15-34), respectively). As detailed under "Experimental procedures," data from the FCA histograms were used to generate the kinetic curves on Fig. 3c. Data were fitted using the two-state kinetic model described previously for other AMPs (34), assuming that peptide-bacteria interaction consists of an initial binding step, followed by permeabilization of the membrane.
Although both peptides induce permeabilization of ϳ90% of bacteria after 90 min (Fig. 3c), the process differs between peptides at the early stages; based on the tendency of the experimental data, Ctn(15-34) seems to start killing bacteria right after its addition, and the permeabilization takes place in a single process. In contrast, Ctn shows an initial lag phase. Fitting of the kinetic data to a two-state model results in statistically poor residuals for some time intervals (Fig. S3), suggesting that for both peptides, permeabilization is more complex than as assumed by the model. An additional slow event leading to permeabilization is suggested because the data deviate from the fit at longer times. In contrast, the fit is quite acceptable at shorter times. k 0 , the membrane attachment rate constant, is higher for Ctn(15-34) than for Ctn (2 ϫ 10 Ϫ2 s Ϫ1 versus 1 ϫ 10 Ϫ3 s Ϫ1 ) as well as the permeabilization rate constant (1 ϫ 10 Ϫ3 s Ϫ1 versus 9 ϫ 10 Ϫ4 s Ϫ1 ). Cooperativity is not observed in either case (f ϳ 0), contrary to other AMPs (34).

Membrane-disruptive effect of Ctn and Ctn(15-34)
co-localizing with SYTOX Green (as shown in the fluorescence distribution plot; Fig. 4a). This suggests partial internalization and binding to DNA by the peptide, but the finding is not general for all bacteria. Images acquired in the normal confocal mode (Fig. S5) show untreated controls and bacteria treated with non-labeled peptides.

Table 3
Affinity of Ctn(15-34) for different lipid systems as followed by surface plasmon resonance P/L max (mol/mol) is the maximum binding of Ctn(15-34) for a given lipid system. P/L max was calculated by fitting the dose-response curves shown in Fig. 5b (right) with one-site specific binding with the Hill slope equation.

Stage 1: Initial peptide recruitment
There is wide agreement that AMPs are usually drawn to microbial surfaces by electrostatic attraction between cationic (Lys, Arg, and His) residues in the peptide and anionic components in the bacterial surface, such as LPS and LTA on Gram-negative or Gram-positive surfaces, respectively, and/or anionic phospholipids (PG and cardiolipin) in the plasma membrane (7). In the present case, the high cationicity of both Ctn and Ctn(15-34) (net charge ϩ16 and ϩ8, respectively) and their preferential action on Gram-negative bacteria suggest that both peptides might bind LPS, as proposed for other snake-derived cathelicidins, such as Pb-CATH (43) and HC-CATH (22). As shown in Fig. 5a, Ctn(15-34) is not highly efficient at neutralizing both LPS and LTA, but even so, the binding is higher for LPS than for LTA, in concordance with its preference for Gram-negative bacteria. The fact that both peptides show a preference for cardiolipin (see below in AFM results) also reinforces an electrostatically driven initial recruitment step.

Stage 3: Cell death by membrane disruption
Bacterial death results from membrane disruption and consequent loss of functionality. FCA data in Fig. 2 and Fig. S2 show a direct correlation between peptide uptake, membrane permeabilization, and viability loss. A mechanism dependent on the ability to target the membrane was also confirmed by imaging techniques. AFM studies showed important disturbances in the E. coli surface upon incubation with Ctn and Ctn(15-34) (Fig. 4b). The leakage of cytoplasmic content observed after peptide treatment is in agreement with a lytic mechanism of action, and the shrinkage observed in the middle region of the  (15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30)(31)(32)(33)(34) bacteria could be due to preferential accumulation of peptides in cardiolipin-rich domains, namely in the apical and septal area of the cell (44), as described previously for other AMPs (32). Confocal microscopy experiments found RhB peptides in the periphery of E. coli bacteria (Fig. 4a), probably located on the membrane, in tune with the membrane-disruption hypothesis, although they were also found internalized inside cells in some cases. Thus, we propose that both peptides are initially recruited and accumulate around the surface until a threshold concentration is reached, after which they begin to enter/cross the lipid bilayer, causing cell death.
Despite evidence suggesting a common membrane disruption mechanism for snake cathelicidins, how this process unfolds over time is not well studied, in part due to the limited information available by standard methods. To shed light on this issue, we employed time-resolved flow cytometry (34, 50), a high-resolution technique that allows monitoring peptide uptake and membrane permeabilization over time. This is particularly relevant to membrane-lytic peptides, for which binding and accumulation on the bacterial surface are early-onset events not easily monitored by end-point experiments. For instance, FCA kinetic curves (Fig. 3) reveal differences undetectable with the colony count method. Ctn (15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30)(31)(32)(33)(34) acts faster than Ctn due to faster binding and membrane permeabilization (Fig. 3). The slower stages in the action of Ctn cause a lag in its kinetic curve, although no cooperativity is detected, which demonstrates that common-sense reasoning that assigns lag phase directly to positive cooperativity (34,51,52) might be an oversimplification.
The mechanism of action of both Ctn and Ctn (15-34) includes steps whose kinetic curves deviate from the two-state model. Interestingly, the related NA-CATH cathelicidin seems to switch its mechanism of action from membrane disruption to pore-based lysis in a biphasic model, mimicking the E. coli membrane with an impact on kinetics similar to that described for Ctn (49). Given the high structural similarities between both peptides (an N-terminal ␣-helical segment followed by a disordered tail) (27,53), it is plausible that Ctn may act similarly, and such behavior would explain the complex cell death kinetic pattern here reported. This dual behavior might be tentatively related to the composite structure (␣-helix plus random tail) of both NA-CATH and Ctn, each moiety of the peptide embed-ding into the membrane in a different fashion, as suggested for NA-CATH (49). The fact that removal of the N-terminal helix of Ctn to give Ctn (15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30)(31)(32)(33)(34) results in a simplified killing mechanism corroborates our hypothesis.
In conclusion, by deploying a set of high-resolution methodologies, we have achieved accurate spatiotemporal characterization of events involved in the bactericidal action of Ctn and Ctn(15-34) on E. coli and P. aeruginosa. The study sheds light on the mechanism whereby the peptides accumulate and disrupt bacterial and model membranes and reveals slight differences in their behavior, including different kinetics of access to bacterial surfaces and different toxicity. By dissecting Ctn into its Ctn(15-34) fragment, we have produced a smaller, simpler, and more cost-efficient AMP displaying selectivity for bacterial-like, PS-rich membranes and a simplified mechanism of action. Taken together, our findings augur a promising role for Ctn (15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30)(31)(32)(33)(34) as an anti-infective lead and, in a broader sense, provide valuable insights into AMP lethal mechanisms that should be useful in the design and development of antimicrobials against resistant bacteria.

Viability and membrane permeabilization correlation studies
To determine bacterial viability and membrane permeabilization, bacterial suspensions at 10 8 cfu/ml in MHBII were centrifuged 10 min at 4000 ϫ g and diluted to 5 ϫ 10 5 cfu/ml in HEPES buffer. Incubation with 2-fold serial peptide dilutions was done at 37°C and 200 rpm, for 90 min for E. coli and 60 min for P. aeruginosa. Untreated bacteria were used as negative control. To estimate the percentage of permeabilized bacteria and percentage of viable bacteria, each sample was respectively evaluated by FCA and colony count methods using protocols described previously (42).

Membrane permeabilization induced by RhB-labeled peptides
Samples for FCA were prepared as described above. Data were acquired on a BD FACSAria III cell sorter (BD Biosciences) using blue (488-nm) and yellow-green (561-nm) lasers and emission filters BP530/30 (LP505) and BP586/15 to detect SYTOX Green and RhB, respectively. 5000 events corresponding to bacterial population as previously defined by forward scatter and side scatter were recorded. Experiments were performed in triplicate.
Estimated percentages of permeabilized bacteria and peptide uptake for each n were calculated by the following equations.
% permeabilized bacteria at n ϭ P n Ϫ P live P dead Ϫ P live ϫ 100 (Eq. 3) % peptide uptake at n ϭ U n Ϫ U live U dead Ϫ U live ϫ 100 The percentage of permeabilized bacteria was corrected using the equation derived from the calibration curve shown on Fig.  S8a, obtained by combining known percentages of dead and live bacteria.

Confocal microscopy
An E. coli suspension was prepared at 10 8 cfu/ml in MHBII, centrifuged for 10 min at 4000 ϫ g, and diluted to 10 7 cfu/ml on HEPES buffer. RhB-labeled or unlabeled peptides were added to the bacterial suspensions at a final concentration equal to their MBC, and the samples were incubated at 37°C and 200 rpm for 90 min. Untreated bacteria were used as a control. After incubation, samples were placed on ice, and SYTOX Green was added at 5 M final concentration. Samples were placed on an ibiTreatcoated 8-well -slide (Ibidi, Munich, Germany) before imaging. Acquisition was made on a confocal point-scanning Zeiss LSM 880 microscope (Carl Zeiss, Jena, Germany) equipped with an alpha Plan-Apochromat ϫ100 oil immersion objective (1.46 numerical aperture). The 488-nm line from an argon laser and the DPSS 561-20 laser were used to excite SYTOX Green and RhB, respectively. In the normal confocal mode, ϫ0.6 zoom images were recorded at 1024 ϫ 1024 resolution. In the Airyscan superresolution mode, 5.5ϫ zoom was applied to select one single bacteria cell, and the optimal resolution was selected. ZEN software was used for image acquisition and Airyscan image processing. Fiji software was used for background subtraction and brightness and contrast adjust (identically for compared image sets). At least 12 total images were acquired in three independent replicates.

Liposome preparation
POPC; POPS; POPG; POPE; E. coli polar lipid extract containing PE-phospholipids, PG-phospholipids, and cardiolipin in the proportion 67:23.2:9.8 (w/w/w); and octadecanoyl SM extracted from porcine brain were purchased from Avanti Polar Lipids (Alabaster, AL). Lipids were solubilized in spec-troscopy-grade chloroform and mixed as required to prepare defined lipid compositions. LUVs (100-nm diameter) or SUVs (50-nm diameter) were prepared in HEPES buffer by extrusion, as described previously (56). LUVs were used in fluorescence spectroscopy assays, whereas SUVs were used for SPR studies to facilitate bilayer deposition on the chip surface.

Vesicle leakage assay
Content leakage from POPC or POPC/POPS (80:20 molar ratio) vesicles was quantified by CF fluorescence dequenching using LUVs loaded with 50 mM CF as described previously (57). Briefly, 2-fold dilutions of peptides (starting with 10 M) were incubated for 20 min with LUVs (5 M of lipid concentration) in HEPES buffer. Fluorescence emission intensity (excitation at 494 nm and emission at 515 nm) was measured in an M1000 microplate reader (Tecan, Männedorf, Switzerland), and the leakage percentage was calculated as described previously (57).

Interactions with lipid bilayers
Peptide-membrane interactions were monitored by SPR at 25°C with a L1 biosensor chip in a Biacore 3000 instrument (GE Healthcare Australia, Parramatta, Australia). SUVs , or E. coli polar extract were prepared using HEPES buffer and deposited onto a L1 chip by injecting SUV suspensions with 0.5 mM lipid at a flow rate of 2 l/min for 40 min. Peptides were injected over the lipid bilayers at a flow rate of 5 l/min for 180 s (association phase), and dissociation was monitored for 600 s (58,59). HEPES was used as running buffer, and the chip was regenerated as before (60). All solutions were freshly prepared and filtered using a 0.22-m pore size filter. The response units were normalized to peptide/lipid ratio as described previously (58).