All strains used in this study are listed in Table 1. Bacterial strains were grown in glass test tubes (18 × 150 mm) at 35°C in cation adjusted Müeller-Hinton Broth II (MHB II, Sigma-Aldrich) with agitation at 210 revolutions per minute (RPM). Antibiotics and melittin were purchased from Sigma-Aldrich. Melittin used in this study was ≥85% pure and was derived from honey bee venom. Melittin can induce an IgE response in 1/3rd of patients sensitive to honeybee venom; however, it has been speculated this is due to additional compounds found in bee venom (Lee and Bae, 2016). Synthetically produced melittin is potentially more efficacious, less allergenic and will be evaluated in future studies. Tobramycin sulfate, gentamicin sulfate, and streptomycin sulfate were dissolved in autoclaved deionized water and filter sterilized using 0.22 μM filter membranes (Thomas Scientific). A stock solution of 1 mM of melittin was dissolved in dimethyl sulfoxide (DMSO). In a control experiment we determined that at the concentrations used DMSO has minimal effects on mature P. aeruginosa biofilms (Supplementary Figure S1).

Minimum inhibitory concentrations were determined as described previously (Andrews, 2001). Briefly, microdilutions of PAO1 P. aeruginosa were made in a 96-well plate in 10% (v/v) MHB II diluted in Dulbecco’s Phosphate Buffered Saline with magnesium and calcium (DPBS, Sigma-Aldrich). ∼1 × 10⁶ colony forming units/mL (CFUs/mL) were added and incubated for 24-h at 35°C with agitation at 150 RPM. MICs were chosen as the minimum concentration in which no turbidity greater than background was measured (absorbance at 595 nm) using a SpectraMax M5 microplate spectrophotometer system (Molecular Devices).

To measure biofilm antimicrobial susceptibility against the strains of P. aeruginosa listed in Table 1, the MBEC^TM assay was used (Innovotech) as previously described (Maiden et al., 2018a). Briefly, an overnight culture was washed and diluted to an OD₆₀₀ of 0.001 and seeded into a MBEC^TM plate and incubated for 24-h at 35°C with agitation at 150 RPM. After 24-h, the lid was then washed for 5-min to remove non-adherent cells. The lid was transferred to a 96-well treatment plate and incubated for the indicated time at 35°C without agitation. Following treatment, the MBEC^TM lid was washed and transferred to a black 96-well ViewPlate (PerkinElmer) filled with 40% (v/v) BacTiter-Glo^TM (Promega) diluted in DPBS to enumerate cell viability using luminescence by an EnVison Multilabel Plate Reader (PerkinElmer). The BacTiter-Glo^TM microbial cell viability assay is a luminescent assay that determines the number of viable cells present based on quantification of adenosine triphosphate concentration. A calibration curve was previously performed, and it was found to be r² = 0.9884 for luminescence versus CFUs/mL, indicating BacTiter-Glo^TM is an effective measurement of cell viability (Maiden et al., 2018a). Dose response curves (DRCs), checkerboard assays, and time killing curves were performed similarly. To test for antimicrobial susceptibility against the strains of S. aureus listed in Table 1, biofilms were formed using a standard 96-well ViewPlate as we found S. aureus did not form biofilms on MBEC^TM pegs at the air-liquid interface.

To study biofilm dispersal under static conditions, crystal violet staining was performed as previously described (Sambanthamoorthy et al., 2011). Briefly, biofilms were formed by PAO1 P. aeruginosa on MBEC^TM plates as described above and then stained with crystal violet following 6-h treatments.

24-h old biofilms were formed by POA1 P. aeruginosa in glass test tubes (18 × 150 mm) in 1 mL of 10% (v/v) MHB II at 35°C and agitated at 150 RPM as previously described (Maiden et al., 2018b). Cells were then washed in DPBS to remove non-adherent cells and treated with melittin and tobramycin for 2-h. Following treatment cells were washed in PBS (phosphate buffered solution without magnesium and calcium) and the biofilm was disrupted from the air-liquid interface using an autoclaved wooden stick. The cells were stained with TO-PRO-3 iodide, which fluoresces in cells that have compromised membranes by intercalating DNA. Single cell flow cytometry was performed on an LSR II (BD Biosciences) with excitation from the 640 mm laser.

Agarose hydrogels were made by dissolving 1 gm of agarose (Sigma-Aldrich) into 200 mL of Tris-acetate-EDTA buffer and heated to form a homogenous solution using a microwave. The 0.5% agarose solution was then allowed to cool and various treatments were added. The solution was then poured into 100 × 15 mm petri dishes (Thermo Fisher Scientific) and stored at 4°C overnight. Prior to treatment, a 4 mm biopsy punch (VWR) was used to create hydrogel wafers.

Wound surgery was performed on 8–9-week-old male and female SKH-1 mice (Charles River) as previously described (Agostinho Hunt et al., 2017; Hunt et al., 2017). 24-h old wounds were infected with ∼1 × 10⁹ Xen41 P. aeruginosa cells (PerkinElmer), which is a bioluminescent derivative of PAO1 that constitutively expresses the luxCDABE genes. Briefly, 24-h old biofilms were formed on sterilized polycarbonate membrane filters with a 0.2 μM pore size (Millipore Sigma) by diluting an overnight culture to an OD₆₀₀ of 0.001 and pipetting 100 μl on 4 membranes on a tryptic soy agar (TSA) plate. 24-h old biofilms were scraped using L-shaped spreaders (Sigma-Aldrich) from each membrane and re-suspended in 500 μl of DPBS. 20 μl of the biofilm-suspensions were inoculated into 24-h old wounds formed on the dorsal side of the mouse midway between the head and the base of the tail. 24-h later the biofilm was imaged using the In Vivo Imaging System (IVIS, Perkin Elmer). The biofilm was then treated by placing a 4 mm 0.5% agarose hydrogel on the wound for 4-h. The biofilm was imaged before and after treatment and total flux (photons/sec) was used to quantify bacterial susceptibility. Representative IVIS images are shown in Supplementary Figure S2.

This study was carried out in accordance with the Michigan State University Institutional Animal Care and Use Committee (application 03/18-036-00). MSU is accredited by the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC) and ensures compliance with federal regulatory requirements.

We previously demonstrated that the compounds triclosan and oxyclozanide synergize with tobramycin to reduce both Gram-negative and Gram-positive biofilms (Maiden et al., 2018a,b). During these studies, we used melittin as a positive control for cell permeabilization and observed that it had potent activity against P. aeruginosa biofilms. To further explore this finding, DRCs of melittin were performed alone and in combination with tobramycin, gentamicin or streptomycin against mature P. aeruginosa biofilms. Melittin was most effective at 100 μM, resulting in a ∼1.5-log₁₀ reduction in the number of cells within a biofilm compared to untreated controls after 6-h of treatment (Figure 1A). Tobramycin alone showed a maximal activity of ∼half-a-log₁₀ of cells within biofilms although this effect was observed at lower concentrations of tobramycin. However, when used together, enhancement was seen when 50 μM of melittin was combined with 50 μM of tobramycin, resulting in ∼1.5-log₁₀ cellular reduction compared to untreated controls, and 100 μM of melittin and tobramycin produced a ∼2-log₁₀ cellular reduction that was significantly more effective than melittin alone. DRCs were also performed with gentamicin or streptomycin combined with melittin (Figure 1B,C). Melittin alone and no treatment were re-plotted in each panel for comparison. The combination of gentamicin and melittin was significantly more effective than either treatment at 50 and 100 μM; however, streptomycin showed little enhancement of melittin at any concentration. This is not surprising as it is known that streptomycin has reduced activity against P. aeruginosa and is not used clinically (Gilbert, 2013). These data indicate that melittin alone is effective at reducing biofilms and has greater efficacy (defined as maximum cellular reduction at a given concentration) in combination with tobramycin or gentamicin.

Effective concentration 50 (EC₅₀) values were calculated to determine the potency of the combinations against P. aeruginosa PAO1 biofilms. The EC₅₀ value for melittin was decreased 2.8 and 4.6-fold when used in combination with tobramycin or gentamicin, from 46 to 16 (14–19 μM) and to 10.68 μM (9–11 μM), respectively (95% confidence intervals) (Table 2). An EC₅₀ value was ambiguous for the streptomycin and melittin combination because streptomycin failed to significantly potentiate melittin. EC₅₀ values for tobramycin, gentamicin or streptomycin when used alone could not be determined using the concentrations tested against cells growing as biofilms as they demonstrated decreased efficacy at higher concentrations. This could be due to the previously described paradoxical effects of aminoglycosides in which higher concentrations are less effective than lower concentrations (Lorian et al., 1979; Gilleland et al., 1989; Barclay and Begg, 2001).

Minimum inhibitory concentrations on PAO1 were performed to determine if melittin in combination with tobramycin, gentamicin, or streptomycin were more effective against planktonic cells. For planktonic cells, all aminoglycosides demonstrated lower MICs than melittin. In addition, the MIC values for each aminoglycoside did not change when used in combination with melittin, suggesting enhancement only occurs against PAO1 P. aeruginosa growing as biofilms (Supplementary Table S1).

Because gentamicin and tobramycin when combined with melittin exhibited similar enhancement of biofilm killing of P. aeruginosa PAO1, for the remainder of this work we studied the activity of the combination of tobramycin and melittin. Time-killing curves were performed to study the pharmacokinetic properties of melittin alone and in combination with tobramycin. To perform time-killing curves, the number of viable cells within the biofilms were determined by BacTiter-Glo^TM at 0, 2, 4, and 6-h. 50 μM melittin alone or in combination with 400 μM of tobramycin showed activity by 2-h, whereas tobramycin treatment alone was not effective at 6-h (Figure 2). By 6-h, the combination resulted in ∼1-log₁₀ cellular reduction whereas melittin alone resulted in ∼half-a-log₁₀ cellular reduction and tobramycin exhibited no significant killing. It is not surprising that tobramycin was ineffective as it is known to penetrate the biofilm poorly and can require up to 24-h to diffuse into biofilms (Tseng et al., 2013).

Checkerboard experiments were performed to determine the concentrations of melittin and tobramycin that was effective against P. aeruginosa biofilms. Melittin was effective alone at 100 μM, resulting in a ∼1.5-log₁₀ cellular reduction compared to untreated controls (Figure 3). The maximal effect was observed when 100 μM of melittin was combined with 400 μM of tobramycin, resulting in ∼2-log₁₀ cellular reduction compared to untreated controls. Melittin and tobramycin showed enhancement when used in combination between 12.5 and 50 μM of melittin and between 50 and 400 μM of tobramycin, resulting in a ∼1-log₁₀ cellular reduction within biofilms compared to either tobramycin or melittin alone (statistical significance is shown in Supplementary Table S2).

We tested the efficacy of tobramycin, melittin, and the combination to kill biofilm-growing bacteria from 7 P. aeruginosa CF clinical isolates and 4 S. aureus clinical isolates (strains are described Table 1). CF_110_N and CF_110_O were isolated longitudinally from the same patient 3 months apart and AMT0023_30 and 34 were isolated longitudinally from the same patient at 6 months and 8 years of age, respectively (De Soyza et al., 2013; Maiden et al., 2018a).

Melittin (50 μM) in combination with tobramycin (400 μM) significantly killed 7/7 P. aeruginosa CF isolates grown as biofilms compared to tobramycin treatment alone, resulting in a maximal ∼1.5-log₁₀ cellular reduction compared to untreated controls (CF_115_J) (Figure 4A). Importantly, the combination significantly enhanced killing of strain AMT0023_34 compared to tobramycin alone, which over expresses the RND-type MexXY-OpRM efflux pump, rendering it resistant to tobramycin (Mulcahy et al., 2010). In 5/7 isolates, the combination was significantly more effective than melittin alone, but this was not the case for strains CF_300_A and AMT023_34, suggesting killings of these biofilms is primarily driven by melittin.

Melittin (100 μM) alone was also effective at killing biofilms of 4/4 S. aureus isolates tested, resulting in a maximal ∼3-log₁₀ cellular reduction compared to controls (MSSA_Newman) (Figure 4B). The limit of detection in this assay is 10³. Interestingly, tobramycin (100 μM) was ineffective against each strain of S. aureus tested and resistance was observed, but the combination remained effective at killing S. aureus biofilms. For 3/4 of the S. aureus strains, the combination of tobramycin and melittin was similar to melittin alone, while a slight but significant enhancement was observed for MRSA_USA300. These results suggest that melittin and not tobramycin primarily drives killing of S. aureus biofilms.

The mechanism of action of AMPs against cells within biofilms is likely through cellular permeabilization (Melo et al., 2009). To test this hypothesis, we used the TO-PRO-3 dye which stains DNA in permeabilized cells with compromised membranes and analyzed treated P. aeruginosa biofilms by single cell analysis using flow cytometry. Tobramycin (400 μM) treatment alone did not increase the population of permeabilized cells compared to no treatment, but after 2-h of treatment, 100 μM of melittin significantly increased the population of permeabilized cells within biofilms to 78% compared to 6% for untreated controls (Figure 5A). A similar increase was observed for the combination treatment and this was not significantly different than melittin alone.

We also hypothesized that melittin in combination with tobramycin may cause increased biofilm dispersal. To test this hypothesis, we measured biofilm dispersal by staining the biofilm biomass after treatment using crystal violet. At all concentrations tested the combination resulted in significant biofilm dispersal compared to either melittin or tobramycin treatment alone, and 50 or 100 μM of melittin or tobramycin alone resulted in significant biofilm dispersal compared to untreated controls (Figure 5B). These data suggest that the mechanism of action of melittin alone and in combination with tobramycin may be biofilm dispersal and cellular permeabilization.

To determine if melittin and tobramycin were effective against Xen41 P. aeruginosa biofilms in vivo, we tested their activity using an IVIS murine wound model (Agostinho Hunt et al., 2017; Hunt et al., 2017). Xen41 P. aeruginosa is a bioluminescent derivative of PAO1 that constitutively expresses the luxCDABE genes. Mature biofilms colonizing an open wound were treated for 4-h using a hydrogel imbedded with either 100 μM of melittin or 400 μM of tobramycin alone and in combination. This combination was chosen because it caused the maximum amount of in vitro killing (Figure 3). Hydrogels containing melittin and tobramycin resulted in a significant 4.2-fold-reduction in biofilm bioluminescence compared to tobramycin treatment alone or animals treated with the control hydrogel (Figure 6). The combination also resulted in a maximal reduction in biofilm bioluminescence of 7.8 and 9.8-fold. Hydrogels containing only tobramycin resulted in 1.8-fold-reduction in biofilm bioluminescence after 4-h; however, this was not statistically significant compared to the control hydrogel. Surprisingly, hydrogels that contained only melittin exhibited no killing compared to biofilms treated with control hydrogels, suggesting that maximum efficacy requires both melittin and tobramycin.