Carboxyamidated P1 (FFHHIFRGIVHVGKTIHRLVTG, MW 2,571) was obtained from the University of Texas Southwestern Medical Center. Fmoc solid-phase peptide synthesis was performed to produce the peptide in the unlabeled and ¹⁵N-labeled states, as described previously (Chekmenev et al., 2006a; Chekmenev et al., 2006b). The labeled amino acids (¹⁵N, 98%) were purchased from Cambridge Isotope Laboratories (Tewksbury, MA, United States). Purification of the crude peptide was performed by HPLC as previously reported to yield 98% pure peptide, as characterized by HPLC and mass spectrometry (Chekmenev et al., 2006b; Oludiran et al., 2019). Next, P1 was dissolved in dilute HCl to substitute chloride ions for trifluoroacetate ions. Excess chloride ions were removed by dialysis. The lyophilized peptide was dissolved in nanopure water prior to using it in the studies presented here. The concentrations of the stocks were determined by amino acid analysis performed at the Protein Chemistry Center at Texas A&M (College Station, TX). Metallation of P1 in a stoichiometric ratio with Cu²⁺ was achieved using copper chloride, as described previously (Libardo et al., 2017; Rai et al., 2019; Paredes et al., 2020; Comert et al., 2021; Fu et al., 2021). Reagents such as sodium hydroxide, buffers, salts, EDTA, chloroform, and copper chloride were purchased from Fisher Scientific (Hampton, NH). Calcein dye was obtained from Sigma-Aldrich (Saint Louis, MO). The Millipore MilliQ system (Sigma Aldrich, Saint Louis, MO) was used to obtain Nanopure water.

Lipids were acquired from Avanti Polar Lipids (Alabaster, AL). Non-deuterated lipids were dissolved in deuterated chloroform and concentrations quantified by solution NMR.

Vero E6 cells (ATCC, CRL-1586, Manassas, VA, United States) were cultured with Eagle’s minimum essential medium (EMEM, Quality Biological, 112–016,101CS, Gaithersburg, MD, United States) supplemented with 10% heat-inactivated fetal bovine serum (FBS, Gibco, Thermo Fisher Scientific, Waltham, MA), 10 U/mL penicillin and 10 μg/mL streptomycin antibiotics (Corning 30–002-CI, Corning, NY, United States), and 2 mM L-glutamine (Corning, 25–005-CI, Corning NY, United States). Vero cells (ATCC, CCL-81, Manassas, VA, United States) were cultured with Dulbecco’s Modified Eagle’s Medium (DMEM, Quality Biological, 112–013,101CS, Gaithersburg, MD, United States) supplemented with 5% heat-inactivated FBS, 10 U/mL penicillin and 10 μg/mL streptomycin antibiotics (Corning 30–002-CI, Corning, NY, United States), and 2 mM L-glutamine (Corning, 25–005-CI, Corning NY, United States). All cell lines were cultivated at 37°C and 5% CO₂.

Vero E6 cells were seeded at 1.5 × 10⁴ cells per well in 96-well plates and incubated at 37°C and 5% CO₂ for 24 h. Cells were then treated with 20 μg/mL or 10 μg/mL of P1 for 24 h and cell viability was determined using CellTiter-glo Luminescent Cell Viability Assay (Promega, G7570, Madison, WI, United States) following the manufacturer’s instructions and compared to vehicle control.

SARS-CoV-2 (Washington strain 2019-nCoV/USA-WA1/2020) was obtained from BEI Resources (NR-52281) and was used for all infections. Vero E6 cells were seeded at 1.5 × 10⁴ cells per well in 96-well plates 24 h prior to infection. P1 was dissolved in water and diluted in culture media to the indicated concentrations. For direct viral treatment, the virus was diluted to a multiplicity of infection (MOI) of 0.1 in media containing P1 at 10 μg/mL or vehicle control. For pre-treatment and post-treatment infections, cells were pretreated with P1 at 10 μg/mL or H₂O vehicle control for 1 h at 37°C and 5% CO₂. Pre-treatment was removed and cells were infected at an MOI of 0.1 in culture media for 1 h at 37°C and 5% CO₂. After infection, inoculum was removed and post-treatment, P1 at 10 μg/mL or H₂O vehicle, was applied. Treated inoculum was incubated at 37°C and 5% CO₂ for 1 h and then applied to cells for 1 h. After infection, inoculum was removed, and culture media was applied to the cells. For infections combing pre-treatment, direct viral treatment, and post-treatment with P1, the above steps with P1 were combined. For post-treatment only infections, cells were infected at an MOI of 0.1 for 1 h. After infection the inoculum was removed and media containing P1 at 10 μg/mL or H₂O vehicle control was added to the cells and incubated at 37°C and 5% CO₂. After 24 h post infection (hpi), viral supernatants were collected and used immediately for assays or stored at −80°C.

Vero cells were plated in 12-well plates at a density of 2 × 10⁵ cells per well and incubated for 24 h. Infection supernatants were serially diluted to 10⁻⁶ in culture media and overlaid on cells for 1 h. Cells were covered with Eagle’s Minimum Essential Medium (without phenol red (Quality Biological 115–073-101, Gaithersburg, MD, United States), supplemented with 5% FBS, 1% nonessential amino acids, 1 mM sodium pyruvate (VWR, 45,000–710, Dixon, CA, United States), 2 mM L-glutamine, 20 U/mL penicillin, and 20 μg/mL streptomycin) with 0.6% agarose (ThermoFisher, 16,500,100, Waltham, MA, United States). At 48 hpi, cells were fixed with 10% formaldehyde (ThermoFisher, F79P-4, Waltham, MA, United States) for 1 h. Medium was removed, wells were washed with diH2O and stained with a 1% crystal violet (ThermoFisher, C581-25, Waltham, MA, United States) and 20% ethanol solution (ThermoFisher, BP2818-4, Waltham, MA, United States).

The secondary structure of P1 interacting with the L_o and L_o/L_d viral envelope mimics was accomplished by CD in potassium phosphate buffer (5 mM, pH 7.4). LUVs were made from lipid films, as described previously (Chekmenev et al., 2006b; Paredes et al., 2020). Briefly, the films were made using DPPC, POPC, DPPE, POPE, and Chol (Avanti Polar Lipids, Alabaster, AL) mixed in chloroform in the amounts needed to form the L_o or L_o/L_d. The solvent was evaporated under N₂ gas and the films dried under vacuum overnight. The films were then hydrated with phosphate buffer (5 mM, pH 7.4), subjected to freeze-thaw cycles, and extruded using an extruder (Avanti Polar Lipids, Alabaster, AL) and 0.1 µm size membranes (Whatman, Florham Park, NJ). The vesicles were then diluted to 3 mM. Each spectrum was performed at fixed peptide concentration (20 µM). P1 was added to reach various P/L ratios, ranging from 1:20 to 1:200. CD spectra were acquired at 298 K on a Jasco J-1500 spectrometer (Jasco Analytical Instruments, Easton, MD). Experimental parameters included a wavelength range of 190–260 nm, scan speed of 100 nm/min, 1 nm bandwidth, and number of scans equal to four. The samples were run in duplicates. A blank containing buffer and lipids but no peptide was recorded and subtracted from the signal obtained for the peptide-containing sample, as needed to account for the background signal. Peptide helicity was obtained using the molar ellipticity measured at 222 nm and assuming an ellipticity of −32,000 deg·cm²/dmol for an ideal α-helix (Lee et al., 2007).

The permeabilization assays were performed as previously described (Mihailescu et al., 2019; Liu et al., 2023). Briefly, for each lipid composition, either L_o or L_o/L_d, a film was made in round bottom flask using DPPC, POPC, DPPE, POPE, and Chol (4 µmol total of lipids) mixed at a desired molar ratio in chloroform. After evaporating the organic phase under N₂ gas, the film was lyophilized overnight. Next, the film was hydrated with 300 µL of 80 mM calcein dye (pH 7.4) before undergoing three freeze-thaw cycles and extrusion through a mini extruder (Avanti Polar Lipids) using a 0.1 µm polycarbonate membrane. Free dye was removed using a size-exclusion column with a Sephadex G-50 resin stationary phase run with HEPES buffer (50 mM, 100 mM NaCl, 0.3 mM EDTA, pH 7.4). The exact lipid concentration of the LUVs was obtained using a Total Phosphorus Assay (Chen et al., 1956). They were then diluted to a final concentration of ∼35 μM, and distributed in a 96-well plate using 180 µL per well. Next, 20 µL of P1 solution was added to each well in concentrations needed to result in specific peptide-to-lipid ratios (P/L) in the wells to bracket P/L = 1:2 and 1:2048 based on the serial dilution of the P1 stock. Each P/L value was run in triplicate and at least three independent assays were run. Positive and negative controls featured 20 µL of 1% Triton-X detergent or 20 µL nanopure water instead of the peptide solution. Each plate was shaken for 1 hour at 45 °C (or 30 °C) and allowed to cool for 20 min before recording the fluorescence intensity using a BioTek SynergyH1 plate reader (BioTek, Winooski, VT). An excitation wavelength of 490 nm and an emission wavelength of 520 nm was used for calcein. Fractional peptide-induced leakage was obtained using the equation: %   l e a k a g e = I x − I 0 % I 100 % − I 0 % where I_x is the fluorescence intensity in a well with a particular peptide concentration, I_100% is the average intensity in the positive-control wells (100% leakage) and I_0%, is the average intensity in cells with the negative control (0% leakage). The fractional leakage was then plotted against lipid-to-peptide ratio (L/P) and fitted in GraphPad Prism using the Hill equation to identify the EC₅₀ or P/L at which half of the vesicles were lysed (Sani et al., 2014). The 95% confidence interval (CI) was obtained from the program.

Lipid films of samples featuring the L_o or L_o/L_d were made using the needed molar ratio of DPPC, POPC, DPPE, POPE, and Chol in chloroform. After evaporating the solvent and vacuuming overnight, the films were hydrated with potassium phosphate buffer (5 mM, pH 7.4), diluted, and extruded through a mini extruder (Avanti Polar Lipids) using 0.1 µm polycarbonate membrane (Whatman, Florhan Park, NJ). Piscidin was added in varying amounts to reach specific P/L ratios ranging from 1:20 to 1:200. To confirm the size of the LUVs made for Cryo-EM experiments, we used Dynamic Light Scattering (DLS). This involved transferring 540 µL of vesicles into 2 mL Eppendorf tubes. The particle size distribution of the samples was measured at room temperature using a Nicomp N3000 Submicron Particle Sizer (Particle Sizing Systems). Some peptide-containing samples were also tested. In this case, we used 60 µL of peptide solution to reach a specific P/L. For lipid-only samples, 60 µL of nanopure water were used instead of peptide solution.

Cryo-EM experiments were performed at the New York Structural Biology Center (NYSBC). Before cryo grids preparation, the samples were warmed up to 50 °C for 5–10 min. Then 4 μL of the solution was immediately added to freshly plasma cleaned (Gatan Solarus plasma cleaner, 75% argon/25% oxygen atmosphere at 15 Watts for 7 s) Au R1.2/1.3 300-mesh (UltrAuFoil^®) grids. Grids were blotted for 4 s or 5 s after a 3 s pre-blotting time, then plunge-frozen in liquid ethane using Vitrobot Mark IV with the chamber maintained at 20°C and 100% humidity. Cryopreserved grids were stored in liquid N₂ until use. Image collection was accomplished at ∼2 μm under focus on a Thermo Scientific™ Glacios™ Cryo-Transmission Electron Microscope (Cryo-TEM) operated at 200 kV equipped with a Falcon3 operated in linear mode. Data collection was performed in the automatic data collection software Leginon (Suloway et al., 2005) at 2.5 Å or 1.2 Å per pixel. The dose was fractionated over 30 raw frames and collected with the total electron dose around 31.0 e^–/Ų. Individual frames were aligned and dose-weighted with MotionCor2 (Zheng et al., 2017).

Oriented samples were made to characterize the structural and orientational features of P1 that had been exposed to the L_o/L_d viral envelope mimics, and to determine how P1 affects the phase behavior of these model membranes. Aligned phospholipid preparations were achieved using a protocol previously established (Chekmenev et al., 2006a; Perrin et al., 2014). Briefly, the samples were made using the L_o/L_d mixture as viral membrane mimics, i.e. 20:10:12:33:25 DPPC/DPPE/POPC/POPE/Chol. For ²H detection, 5 mg of DPPC-d62 was incorporated. For the sample containing P1, after the lipids were mixed in chloroform, P1 dissolved in trifluoroethanol was added at P/L = 1:100. About 2 mg of ¹⁵N-V₁₀G₁₃I₁₆ P1 was used. After co-dissolving all of the components, the solvent was evaporated using nitrogen gas and lyophilized overnight. The next day, the lipids were hydrated with 1 mL of Bis-Tris buffer (3 mM, pH 7.4). The mixture was incubated overnight at 40 °C and then spread on thin glass slides (dimensions 5.7 × 12 × 0.03 mm³ from Matsunami Trading Co., Japan). Each sample was then allowed to equilibrate in humidity chamber (>90% using a saturated potassium sulfate solution). The slides were stacked after hydrating at 42% by weight using ²H-depleted water. The stacked samples were placed in a glass cell (Vitrocom Inc., NJ), sealed with beeswax, and incubated at 45 °C until homogeneously hydrated.

Unoriented samples were prepared using a procedure similar to that used for oriented samples. Deuterated DPPC-d62 was weighed out and mixed in a round-bottom flask with appropriate amounts of non-deuterated lipids to form the L_o phase, i.e. 40:18:3:6:33 DPPC/DPPE/POPC/POPE/Chol. The solvent was evaporated, and the film dried on a lyophilizer overnight. The film was then hydrated with Bis-Tris buffer and underwent freeze-thaw cycles to ensure distribution of buffer. The resulting suspension was incubated overnight at 40°C. The suspension was then transferred to a small glass vial and lyophilized. The dry lipid was transferred to a sample tube and hydrated to 42% (v/w) with deuterium-depleted water, so that the final BisTris buffer concentration matched that of the oriented sample. To ensure complete removal of HOD that would give rise to ²H NMR signal, the sample was lyophilized at least 4 h, rehydrated to 42% with the ²H depleted water, and sealed. The sample was then incubated for 24 h before NMR analysis.

²H spectra were collected at the National High Magnetic Field Laboratory (NHMFL) on a 600 MHz wide-bore magnet equipped with a Bruker Avance I console where the ²H Larmor frequency is 92.12 MHz. Parameters for the quadrupolar echo pulse sequence included a 3.0 µs 90° pulse on ²H, and an echo time of 20 µs. The FIDs were recorded with a dwell time of 2 µs for an acquisition time of 4.096 ms. Samples were first run at 50°C to allow for annealing, then run from 10°C up to 50°C in steps of 5°C. A spectrum was also obtained at 42°C between the 40°C and 45°C spectra. Each sample was allowed to equilibrate at each temperature before 4,096 and 1,536 scans for the oriented and unoriented samples, respectively, were averaged with a recycle delay of 1 s. The data were processed in Topspin (Bruker, Billerica, MA) with 100 Hz of exponential linebroadening and referenced to D₂O at 0 kHz. The data from the unoriented L_o sample were de-Paked in NMRPipe (Delaglio et al., 1995) using the dePake macro detailed in Sani et al. (Sani et al., 2013). The FID was left-shifted to the echo maximum before applying a sinebell function, zero-filling, and dePakeing by weighted fast Fourier transform. Basic peak detection within NMRPipe was used to identify the position of resonances and yield the quadrupolar splittings.

¹H spectra were acquired at the NHMFL on a 600 MHz wide-bore magnet equipped with a Bruker Avance I console where the ¹H Larmor frequency is 600.13 MHz. A pulse sequence that suppresses the background signal from the probe was implemented (Wang et al., 2021). Samples packed in 3.2 mm rotors with caps containing O-rings (Revolution NMR, Fort Collings, CO) were spun at 10 kHz. Parameters included a 4.0 µs 90° pulse, and a recycle delay of 2 s. Samples were first run at 50°C to allow for annealing, then run from 10°C up to 55°C in steps of 5°C, with a spectrum at 42°C collected between the 40°C and 45°C spectra. The data, corresponding to 32 scans, were processed with no window function and referenced to the methylene signal at 1.30 ppm (Veatch et al., 2004; Polozov et al., 2008) TopSpin (Bruker, Billerica, MA) was used to process the data. Peak picking was used to determine the intensity of the methylene peak at 1.30 ppm. The peak width at half height was determined by measuring from the picked peak to the half-height on the left as the right side was convoluted in spectra.

2D SAMPI4 (Nevzorov and Opella, 2007) spectra were recorded on a 600 MHz wide bore Bruker instrument with Avance I console at the NHMFL where the ¹H and ¹⁵N frequencies were 600.13 and 60.82 MHz, respectively. The temperature was set to the indicated temperature ±0.1°C. For each spectrum, a recycle delay of 4 s and 32 t₁ increments with 2048 transients each were used. The pulse sequence used a contact time of 810 µs, cross-polarization field of 50.0 kHz, and ¹H decoupling of 62.5 kHz (using PWTPPM) (Fu, 2009). Data were processed using Topspin with Gaussian window function (LB = - 50 Hz, GB = 0.1) in the ¹⁵N dimension and (LB = - 100 Hz, GB = 0.1) in the ¹H-¹⁵N dimension. The ¹⁵N chemical shifts were referenced to a saturated solution of ammonium sulfate set at 26.8 ppm with respect to ammonia.

The potential for piscidins to exert an antiviral activity against SARS-CoV-2 was assessed using Vero E6 cells as an infection model. The Cell Titer Glo assay was used to determine cell viability 24 h post infection (hpi). For the initial assessment, we aimed at determining if the peptide could directly inhibit the ability of the virus to establish a productive infection in cells. For this purpose, the virus (Washington strain 2019-nCoV/USA-WA1/2020) was directly exposed to piscidin or vehicle control before the cells were infected. After 24 hpi, plaque assays were performed to determine if the peptide inhibited viral infection. As shown in Figure 1, P1 at 20.0, 10.0, and 1.0 μg/mL inhibited viral titers by 56.6%, 35.8%, and 18.9% respectively. This indicates that the piscidin peptide exerts a direct inhibitory effect on the ability of the virus to establish a productive infection.

Next, pre-treatment and post-treatment were used to determine if post-entry mechanisms were also inhibited by P1. As shown in Figure 2, pre- and post-treatment with 10.0 μg/mL of P1 reduced SARS-CoV-2 viral titer by 41.5% compared to the vehicle control. Similarly, post-treatment alone reduced viral titers by 40.0%. Combining pre-treatment, direct viral treatment, and post-treatment showed the highest inhibition at 78.1% compared to the vehicle control. This suggests that continual treatment with P1 is characterized by a concentration to reach 50% viral inhibition (EC₅₀) that is on the order of 5 μg/mL (2 µM). This is stronger than the direct viral treatment described above since the EC₅₀ in that case appeared to be on the order of 20.0 μg/mL (8 µM). Overall, these assays indicate that P1 can inhibit the infection potential of the virus in two ways, by directly disrupting the viral particles and indirectly influencing the post-entry infection mechanisms of the virus in cells.

The effectiveness of P1 was compared to that of other peptides, such as the copper-bound form of P1 (P1-Cu) and the piscidin isoforms P3 and TP4, to compare its effectiveness, as shown in Supplementary Figure S1. Brilacidin is a peptidomimetic which we have already published as exerting an antiviral activity against SARS-CoV-2 in this infection model and was used as a positive comparison control (Bakovic et al., 2021). In this case, the pre-treatment strategy was used on Vero cells and each peptide was added at a concentration of 10.0 μg/mL. No apparent cell toxicity was noted following peptide treatment as compared to the vehicle-alone control (data not shown). The data in Supplementary Figure S1 demonstrate that P1 and P3 peptide treatment resulted in a statistically significant decrease as compared to the vehicle-alone control, with P1 demonstrating greater inhibition (>90%) than P3. The extent of inhibition observed with P3, HR9 (histidine-rich nona-arginine) and TP4 was comparable. The conjugation of copper decreased the observed inhibitory potential of both P1 and P3 in a comparable manner. Cumulatively, the data support the antiviral potential of P1 and P3 piscidins against SARS-CoV-2 in cell culture. They also show that the peptides have a direct effect on the virus. Previously, P1 was demonstrated to be active in vivo on pseudorabies and porcine CoV (Lei et al., 2018; Hu et al., 2019). It is also inhibitory to HIV-1 in vitro (Wang, 2013). Since these viruses have a lipid envelope and piscidin is membrane active, we performed biophysical experiments on viral envelope mimics to gain further insight into the mechanism of direct inhibition.

As explained in the introduction, enveloped viruses acquire their lipids from host cells, but viral envelopes have lipid compositions that vary from virus to virus and differ from the host plasma membrane. P1 is active on viruses that feature a broad range of lipid compositions, especially with regard to Chol, a lipid that strengthens membranes. It was previously shown to be active on HIV-1 (Wang et al., 2010; Wang, 2013), which has a high Chol content in its envelope (Brügger et al., 2006). The reported EC₅₀ was 2.1 µM (5.4 μg/mL), which is similar to the EC₅₀ observed here against SARS-CoV-2, a virus that has a low Chol content (Saud et al., 2022). Based on their assembly and egress pathway, the low Chol level also applies to the porcine CoV and pseudorabies that P1 directly inhibits (Hogue et al., 2014; Schoeman and Fielding, 2019; Ghosh et al., 2020). Through our previous studies, we showed that P1 permeabilizes both anionic and zwitterionic L_d bilayers. Membranes tested contained lipids such as PC, PE, and Chol, which are relevant to the envelopes of the viruses that P1 inhibits. However, the effect of P1 has not been characterized on the L_o and L_o/L_d bilayers that are relevant to dangerous viruses such as HIV-1 and Influenza A (Brügger et al., 2006; Ivanova et al., 2015). To fill this gap and to extend previous studies beyond the L_d, we designed Chol containing membranes featuring the L_o and L_d.

In preparing bilayers containing the L_o and L_d, we opted for a membrane composition that best represented multiple viruses, including HIV-1 and Influenza A (Brügger et al., 2006; Ivanova et al., 2015). In addition to containing a large amount of Chol, their membranes have an outer leaflet that is enriched in zwitterionic lipids (Brügger et al., 2006). To represent the main lipid headgroups present in these viruses, we used phosphoethanolamine (PE) and phosphocholine (PC), which are also the dominant species in plasma membranes of host cells (Brügger et al., 2006; Ivanova et al., 2015; Lorent et al., 2020). In terms of acyl chains, we used saturated (palmitoyl) and unsaturated (oleoyl) fatty acids, as needed to be able to form both the L_o and L_o/L_d phases in the presence of Chol. Table 1 lists the specific lipids and the molar ratios used to generate L_o and L_o/L_d systems.

As a first step with the viral envelope mimics, we leveraged ²H static SS-NMR to characterize their phase behavior and confirm the presence of the L_o. To achieve high resolution, oriented lipid bilayers were made and DPPC-d62 was used as the ²H reporter. Figure 3A and Supplementary Figure S2 show ²H SS-NMR spectra collected at different temperatures to monitor the phase behavior. At the high temperature of 55°C, the methyl region of the ²H spectrum displays a single quadrupolar splitting (distance between the two black arrows at the top of Figure 3). This indicates that DPPC-d62 exists in a bilayer that is well mixed, and thus above its miscibility point. When the temperature is decreased to 45°C, multiple splittings start appearing, indicating the onset of demixing and the co-existence of the L_o and L_d. Essentially, the rate at which the DPPC-d62 exchanges between the L_o and the L_d phases is slow enough on the time scale of the ²H NMR experiments that their splittings are becoming resolved from each other (Veatch et al., 2004). The outer splittings, which are well resolved at 30 °C (green arrows in Figure 3) are 17.2 and 14.1 kHz, clearly corresponding to the sn-1 and sn-2 methyl groups of DPPC-d62 in the L_o phase (Veatch et al., 2004). The same chains in the L_d phase give rise to the inner splitting (8.4 kHz, gray arrows in Figure 3). As previously explained by the Gawrisch lab, the difference between the inner and outer splittings indicate that large L_o domains (>> 160 nm) are present, and thus the samples are experiencing large-scale demixing (Veatch et al., 2004). The onset temperature for this demixing is referred to as T_low. Using a similar approach, we also investigated the L_o system. As displayed in Supplementary Figure S3, two large quadrupolar splittings (e.g.,; 19.5 and 15.4 kHz at 30 °C) are detected, and therefore the L_o is present throughout the temperature range. Overall, the signals observed by solid-state NMR confirm that the chosen lipid compositions form the expected L_o/L_d or L_o phases.

In the presence of P1 at a peptide-to-lipid ratio (P/L) of 1:100 (Figure 3B), the large-scale demixing of the L_o/L_d occurs at 35°C ± 5°C, which is significantly lower than in the neat bilayer. Thus, P1 exerts a mixing effect on the L_o and L_d, forcing them to mix at a lower temperature than in the neat membrane. In parallel to these ²H NMR experiments, the viral envelope mimics were also subjected to magic angle spinning (MAS), the NMR method commonly used to detect the high-resolution ¹H signals of unoriented samples. As shown in Supplementary Figure S4, the intensity of the central band for the methylene signals from the lipid acyl chains keeps increasing throughout the temperature range, indicating that the miscibility point, T_mix, has not been reached by 55 °C. However, in the presence of P1 (Supplementary Figure S5), a plateau is achieved at T_mix = 40°C ± 5°C, showing that the peptide depresses the stability of the L_o, in agreement with the ²H data. We note that the ²H experiments detect large-scale demixing, explaining why its onset occurs at a lower temperature than T_mix (Veatch et al., 2004).

Next, the secondary structure of P1 was investigated in the presence of the viral envelope mimics. P1 is unstructured in aqueous solution while it adopts an α-helical structure bound to membranes (Chekmenev et al., 2006b; Perrin et al., 2014; Hayden et al., 2015; Paredes et al., 2020). CD spectroscopy readily resolves these two states, and thus provides a useful tool to determine the secondary structure of P1 when it is exposed to viral envelope mimics containing the L_o and L_o/L_d. Large unilamellar vesicles were made with a diameter of 0.1 µm to mimic the small size of virions (Lenard and Compans, 1974).

Figure 4A shows the CD data obtained at 45 °C when P1 was added to the L_o/L_d viral mimics at P/L = 1:200 and 1:120. Based on the ²H and ¹H data presented in Figure 3 and Supplementary Figure S5, the L_o and L_d are miscible at this temperature. The CD bands detected at 208 and 222 nm demonstrate that P1 is α-helical, and therefore bound to the viral envelope mimics. The binding is stronger at lower P/L values, with 98% of the peptide bound at P/L = 1:200. Similar results are obtained with the L_o sample, with 92% binding detected at P/L = 1:200 (Figure 4B). Below 45°C, the CD data became noisier, preventing a similar characterization. However, NMR data (below) were collected at 30°C, showing that the peptide binds to the L_o/L_d below T_low.

Next, we investigated whether P1 permeabilizes viral envelope mimics featuring the L_o/L_d and L_o. We previously used dye leakage assays to characterize the concentration-dependent permeabilization effects of P1 on various L_d membrane mixtures (Perrin et al., 2014; Hayden et al., 2015; Comert et al., 2019). It was found that P1 tends to be more membrane active when PC rather than PE headgroups are combined with anionic phosphatidylglycerol (PG), possibly because PE is more extensively hydrogen bonded and features stronger intrinsic negative curvature compared to PC (Murzyn et al., 2005). We also observed that P1 is active on zwitterionic membranes and Chol does not inhibit its membrane disruptive effects (Comert et al., 2019).

Figure 5 (red) shows the data obtained when we exposed L_o/L_d LUVs containing trapped calcein to increasing peptide concentrations. Based on the fitted dose-response curves, the effective lipid-to-peptide molar ratio (L/P) needed to reach 50% lysis (EC₅₀) is 1769 ± 220 at 45 °C. At 30°C, which is below T_low, the EC₅₀ became 865 ± 35, and thus remained low. These EC₅₀ values indicate that the peptide is more effective on the L_o/L_d than the previously studied 3:1 PC/PG (EC₅₀ = 22), 1:1 PE/PG (EC₅₀ = 10), PC (EC₅₀ = 166), and 4:1 PC/Chol (EC₅₀ = 130) lipid systems (the experimental error was indicated to be 20%) (Perrin et al., 2014; Hayden et al., 2015; Comert et al., 2019). We also tested P1 on the L_o phase alone. The data in Figure 5 (blue) show that the EC₅₀ decreased compared to the L_o/L_d mixture (EC₅₀ = 463 ± 31 and 206 ± 7 at 45°C and 30°C, respectively) but the peptide remained highly permeabilizing. Remarkably, even at 30°C, P1 is much more effective on the L_o/L_d than the zwitterionic 4:1 PC/Chol L_d. As indicated in the introduction, the interface between domains can be a place of higher susceptibility for disruption by membrane-active agents (Schön et al., 2008; Blicher et al., 2009; Apellániz et al., 2011; Hao et al., 2018; Nasr et al., 2020; Kumar et al., 2022; Utterström et al., 2022).

Previous studies with membrane-active HDPs, including P1, have showed that their binding to membranes induces thinning (Balla et al., 2004; Ouellet et al., 2006; Mihailescu et al., 2019). Knowing that P1 permeabilizes viral envelope mimics featuring a high Chol content, we employed Cryo-EM to investigate how thickness changes when these model membranes are exposed to the peptide. The L_o and L_o/L_d vesicles used for Cryo-EM were first characterized by DLS (Supplementary Figure S7) to confirm size homogeneity. Figure 6 shows the Cryo-EM images obtained for such LUVs before and after exposure to P1. The data indicate that the membrane is dramatically rearranged when the peptide is present. From a visual inspection of the micrographs, the LUVs become much larger and adopt multilamellar structures, which indicates that aggregation and fusion occur upon peptide addition.

To quantify the changes in membrane thickness and vesicle size that P1 induces to LUVs, we developed a new automated program to annotate the micrographs and characterize their curvature over a distance of 4 nm. This protocol was inspired by previous work done by the Heberle lab where Cryo-EM data collected on LUVs were analyzed to yield bilayer thickness and relative amounts of L_o and L_d (Heberle et al., 2020). In that case, the analysis was done manually. Here, a ML process was leveraged to process large data sets and give direct information on vesicle size since curvature and vesicle size are inversely related. Major benefits of this approach include that it is fast, and since all of the vesicles are analyzed, bias towards prioritizing some vesicles over others is avoided. Supplementary Figure S8 shows examples of vesicles that were annotated by this method.

As shown in Figure 7, a clear relationship exists between the curvature and P/L, with the curvature decreasing (i.e., the membrane becomes more flat as the vesicles become larger) when more peptide is added. This is observed whether the L_o/L_d or L_o system is used. A trend towards reduced membrane thickness is also detected as more peptide is added. For the L_o/L_d samples, the vesicles contain two types of domains, with the L_o phase being thicker due to higher Chol content. Since the analysis did not resolve the L_o from the L_d, the trends represent how the thickness of the overall bilayer, i.e., the mixed L_o and L_d phases, responded to P1. When the L_o is used alone, thinning occurs but less than when the L_o/L_d are mixed, indicating that the bilayer thinning effect of P1 occurs more readily when the peptide has the opportunity to interact with the L_d that it preferentially binds to (Comert et al., 2019). Overall, the statistical analysis indicates that the relationships associated with curvature and thickness are significant across the data sets: P1 decreases the curvature and thickness of the viral envelope mimics containing the ordered phase.

Given the strong remodeling effect of piscidin on viral envelope mimics detected by Cryo-EM, we sought to characterize the ability of the peptide to mix lipids from vesicles varying in lipid composition. Investigations by DSC were aimed at observing the effect of P1 on the melting transitions and mixing behavior of lipids commonly found in raft mixtures such as POPE, DPPC and Chol, all of which are used to make our L_o and L_o/L_d viral envelope mimics. Two types of experiments were performed. In the first type, POPE, DPPC and Chol were co-dissolved in chloroform to form a homogeneous lipid mixture that was then used to produce vesicles in water (Figure 8A). Taken separately, POPE and DPPC show the typical (gel-to-fluid) melting transitions at 25°C, and 41°C, respectively. However, when homogeneously mixed at a 1:1 molar ratio, they display a single transition peak at 30.8°C, broader than the two individual transition of the two lipids. The presence of Chol in proportion of 20% (molar fraction) broadened the transition of the mixture while shifting T_m to 29°C. Finally, addition of P1 to an aliquot of the same POPE/DPPC/Chol mixture in water further broadened the transition pointing to a strong lipid mixing and chain disordering effect due to P1. In a second approach, separate aliquots of POPE and DPPC vesicles were produced. We combined them with each other (Figure 8B) and with P1 in water (Figure 8C) just before the DSC measurements. We found that this type of investigation provides further insight into whether P1 has a differentiated effect on the two lipids. A few repeated scans were performed until the peak shapes stabilized (see also Supplementary Figure S9). As shown in Figure 8B, the POPE and DPPC melting peaks, while well separated at the beginning at the scans (T_m at ∼ 22°C and ∼39°C, respectively), eventually merged into a broad central peak (T_m ∼ 27°C), that was very similar to the transition peak for the pre-mixed POPE/DPPC (Figure 8A). However, when aliquots of the same POPE and DPPC vesicle suspensions in water were combined as above and P1 was added to this mixture immediately before the measurement, a much more dramatic mixing effect was observed (Figure 8C). By the third scan, the POPE peak completely disappears with no distinguishable mixed lipid peak. A remnant of the DPPC transition is still observed at 41°C. Overall, the data indicates that the P1 has an immediate effect on the melting and mixing of the two lipids, probably by promoting fusion of vesicles. Also, the POPE component appears to be more readily affected by P1. Not only does P1 cause “fluidization” and lipid mixing in membranes but it also displays affinities for certain components. In some of our previous studies, it was shown, by X-ray diffraction and microscopy in “raft” forming mixtures, that P1 has a preference to associate with fluid domains in membranes and/or causes Chol to segregate away from the P1-associated regions (Comert et al., 2019). Taken together, our DSC results indicate that P1 strongly affects the lipid landscape in mixed membranes such as those found in viruses. These changes in structure and organization are likely to directly affect the viral envelope lipids as well as affect viral protein distribution and function.

We previously used oriented sample solid-state NMR to solve the structure of P1 bound to various lipid mixtures, including 3:1 PC/PG, 1:1 PE/PG, and 4:1 PC/Chol (Perrin et al., 2014; Comert et al., 2019). Not only does this method provide the structure of the peptide but it also characterizes the tilt and helix rotation in the membrane (Ketchem et al., 1997; Opella et al., 2001; Opella et al., 2002; Nevzorov et al., 2004; Opella and Marassi, 2004). Since P1 interacts with viral envelope mimics as shown by the data presented above, we made oriented samples for NMR structural analysis of P1.

Figure 9 shows the data obtained for ¹⁵N-labeled V₁₀G₁₃I₁₆ P1 bound to the viral envelope mimics. Since the labeled positions are spread out along the backbone of the peptide, they report on the overall peptide structure, including G13 where the α-helix is kinked (Perrin et al., 2014). The data were collected at 30°C, which is below the T_low characterized by solid-state NMR. These spectra provide two orientational restraints per peptide plane: ¹⁵N chemical shifts (CSs) and ¹⁵N-¹H dipolar couplings (DCs). Both of these sets of values (CS near 70–80 ppm and DCs around 8–10 kHz) are consistent with the peptide lying almost parallel to the membrane surface. Based on previous studies of these three ¹⁵N-labeled sites (Perrin et al., 2014; Comert et al., 2019), the signals were assigned as follows: V₁₀: 80.5; G₁₃: 70.1; and I₁₆: 66.2 ppm. In comparison to the data obtained with the bacterial cell membrane mimic systems (PC/PG and PE/PG), the chemical shifts and DCs are comparable indicating similar tilt angles in the membrane, i.e., the peptide sits parallel to the membrane surface (Perrin et al., 2014; Comert et al., 2019). Hence, despite the high Chol content, P1 retains a strong ability to partition in these membranes. We also collected data at 45°C, which is above T_mix, and noticed that a new set of resonances appeared in proximity to G₁₃ and I₁₆ while the V₁₀ resonance became broader, and thus weaker (Supplementary Figure S10). This suggests that when the L_o and L_d are miscible based on the ²H data, the peptide adopts a minor conformational state, possibly reflecting that the peptide, which prefers interacting with the L_d, is now able to partition significantly in the L_o. Thus, its conformation or surrounding environment would be slightly different in the L_d than L_o. This could result in different CSs and DCs given the sensitivity of these restraints.