A broad-spectrum bactericidal lipopeptide with anti-biofilm properties

Previous studies of the oligoacyllysyl (OAK) series acyl-lysyl-lysyl-aminoacyl-lysine-amide, suggested their utility towards generating robust linear lipopeptide-like alternatives to antibiotics, although to date, none exhibited potent broad-spectrum bactericidal activity. To follow up on this premise, we produced a new analog (C14KKc12K) and investigated its properties in various media. Mechanistic studies suggest that C14KKc12K uses a non-specific membrane-disruptive mode of action for rapidly reducing viability of Gram-negative bacteria (GNB) similarly to polymyxin B (PMB), a cyclic lipopeptide used as last resort antibiotic. Indeed, C14KKc12K displayed similar affinity for lipopolysaccharides and induced cell permeabilization associated with rapid massive membrane depolarization. Unlike PMB however, C14KKc12K was also bactericidal to Gram-positive bacteria (GPB) at or near the minimal inhibitory concentration (MIC), as assessed against a multispecies panel of >50 strains, displaying MIC50 at 3 and 6 µM, respectively for GPB and GNB. C14KKc12K retained activity in human saliva, reducing the viability of cultivable oral microflora by >99% within two minutes of exposure, albeit at higher concentrations, which, nonetheless, were similar to the commercial gold standard, chlorhexidine. This equipotent bactericidal activity was also observed in pre-formed biofilms of Streptococcus mutans, a major periodontal pathogen. Such compounds therefore, may be useful for eradication of challenging poly-microbial infections.


Results
As part of its biophysical characterization, the purified synthetic lipopeptide was first subjected to light scattering measurements in order to assess its potential for self-assembly in an aqueous environment, as compared with two analogs, C 12 KKc 12 K and C 16 KKc 12 K. Expectedly, the light scattered by these analogs at low concentrations and up to ~10 µM, displayed proportional amplitudes, where C 14 KKc 12 K was intermediate between C 12 KKc 12 K (lowest) and C 16 KKc 12 K (highest) (Fig. 1b) and started to deviate from linearity at a concentration range where both C 12 KKc 12 K and C 16 KKc 12 K were previously found to aggregate 24,41 . At ~20 µM however, C 14 KKc 12 K revealed a sharp divergence since its light scattering pattern exhibited significantly higher amplitudes, likely to reflect its capacity to form supramolecular structures of larger sizes 42 . We also assessed the aggregation tendency in a more complex medium such as the supernatant of centrifuged saliva and found it to display an overall similar trend to that in PBS, only starting at somewhat lower concentrations (i.e., the respective critical aggregation concentrations values were 8 ± 1 and 23 ± 3 µM). Often, such self-assembly is deleterious to antibacterial potency of AMPs and OAKs 24,32,33 . Though, in the case of C 12 KKc 12 K and C 16 KKc 12 K, we previously used their respective unsaturated N-terminal acyls to show that the less hydrophobic counterparts (i.e., C 12ω7 KKc 12 K and C 16ω7 KKc 12 K) 24,43 displayed higher potencies, even as aggregates. The proposed explanation for this discrepancy was that the unsaturated analogs formed different structures where the self-assembled lipopeptides were not as tightly bound 24,43 , thereby gaining the ability to dis-aggregate upon interaction with bacteria and the ability to exert their antibacterial activity, unlike the saturated counterparts. Consequently, we next verified the analogs propensities to disaggregate in presence of bacteria by monitoring their light scattering amplitudes. Figure 1c illustrates the rapidly reduced light scattered by C 14 KKc 12 K, indicating its greater tendency to disaggregate in presence of bacteria. While these findings raise interesting questions (such as whether higher bacterial concentrations and/or longer incubation periods will increase the OAK disassembly), the fact that C 14 KKc 12 K demonstrated a tendency for rapid disaggregation in presence of bacteria supports the rapid activity expected in certain applications (as exemplified later in this section). Collectively therefore, the data suggest that the OAK's tendency for self-assembly in a biological milieu at high concentrations, might not interfere with its antibacterial activity.
Next, we characterized its antibacterial activity by determining the MIC against a multispecies panel of 54 representative bacteria, including various medically relevant strains. Table 1 summarizes the MIC values obtained with 30 GPB (five species) and 24 GNB (seven species). With one exception (one of the Acinetobacter baumannii clinical isolates), the data reveal that C 14 KKc 12 K was active on most tested bacteria, although GPB appear generally more sensitive (MIC 50 3 and 6 µM, respectively). Replacing LB with cation adjusted Mueller Hinton Broth resulted in essentially similar outcome. For instance, MICs of S. aureus 29213 or E. coli 25922 were 3 and 6 µM, respectively. Also note that C 14 KKc 12 K displayed an unchanged MIC value (3 µM) on both the wild type (AG100) and its efflux deletion-mutant (AG100A), unlike its analogs 18 (this issue will be elaborated in the Discussion section).
To investigate the mode of action, we used standard methodologies for MAC characterization, including assessment of membrane damages and determination of time-kill kinetics over both Escherichia coli and Streptococcus mutans, respectively representing GNB and GPB. Figure 2a shows the OAK's ability to affect viability of E. coli, reflecting a rapid bactericidal mode of action (e.g., >2-log-unit reduction within 2 hours exposure at ≥MIC).
To assess potential membrane damages, we used an assay capable of differentiating permeability changes in bacterial membranes by testing the leakage of small organic molecules. The assay employs the engineered E. coli strain, ML-35p, which is constitutive for cytoplasmic β-galactosidase, lacks lactose permease, and expresses a plasmid-encoded periplasmic β-lactamase 44 . The chromogenic β-galactosidase substrate ONPG is used to assess permeation of the CM, while OM permeability is assessed using nitrocefin, a chromogenic β-lactamase substrate. The data summarized in Fig. 2b suggest that both membranes were permeabilized at the MIC value (3 µM) albeit the OM appears more susceptible. To validate the CM damages at low concentrations, bacteria were subjected to another permeability assay, this time monitoring cytoplasmic access to the DNA binder, ethidium bromide (Fig. 2c). The fact that ethidium bromide accumulated in E. coli provides confirming evidence for the CM permeabilization at the MIC value. The inset shows two representative kinetic curves illustrating ethidium bromide's rapid accumulation in bacterial cytoplasm. Combined, these findings support the view that bacterial death has resulted from the OAK's capacity to abruptly disrupt both membranes. A similar mode of action was attributed to the natural bacteria-derived 11-residue cyclic lipopeptide, polymyxin B (PMB) 45 , an effect believed to stem from its high-affinity interaction with LPS. When compared, PMB and C 14 KKc 12 K exhibited a similar binding affinity to LPS originating from E. coli or Pseudomonas aeruginosa, as determined by their abilities to displace the binding Scientific RepoRts | 7: 2198 | DOI:10.1038/s41598-017-02373-0 of dansyl-PMB (Fig. 3). Under similar conditions 46 the analogs C 8 , C 10 and C 12 exhibited a binding affinity that was lower than that of C 14 KKc 12 K but increased with increasing hydrophobicity 40 . This is further addressed below. Figure 4a shows the OAK's ability to affect viability of S. mutans, reflecting, again, the OAK's rapid bactericidal mode of action at low micromolar concentrations (e.g., ~2 log units reduction within 2 hours exposure to 1.56 µM). Membrane damages were evident from the rapid and massive leakage of protons and cytoplasmic accumulation of ethidium bromide ( Fig. 4b and c, respectively).
To determine the applicative potential of C 14 KKc 12 K, we next aimed to exploit apparent OAK advantageous properties such as protease stability 38,39 and rapid bactericidal mode of action over a broad spectrum of bacteria. Hence, we performed a preliminary assessment for OAK's ability to affect multiple bacterial species in saliva, given that the list of susceptible strains (Materials & Methods section) included various bacterial species known to promote periodontal diseases 47, 48 whose treatment with an orally active antimicrobial has been the standard clinical approach. For this purpose, we elected to use S. mutans, a prototypical oral pathogen. Unlike other salivary floating bacteria, S. mutans can adhere to the oral cavity surfaces (particularly to teeth) and promote biofilm formation, which disturbs the balanced oral microbiome by pH reduction, culminating in dental complications such as caries and periodontal inflammations [49][50][51] .
Data shown in Fig. 5a confirms the OAK's ability to maintain a rapid bactericidal activity in a complex medium, such as the supernatant from centrifuged human saliva 52 , albeit at the expense of higher doses compared with BHI medium. Furthermore, using whole saliva instead, the OAK has also maintained the capacity for rapid killing (within 2 minutes) of the multi-organism oral microflora, as efficiently as chlorhexidine (Fig. 5b). Chlorhexidine (CHX) is a cationic polybiguanide used since the 1970's 53 as a mouth-wash formulation (1-2 mM) to treat oral inflammations. CHX represents the gold standard reference in the field 54 despite a few shortcomings such as dental/tissue discoloration and/or negative effects on taste 55,56 . For these reasons its application is limited to short periods of time (about a week), and is usually employed as part of pre-or post-surgical interventions.
Moreover, noteworthy is the fact that, when compared at high concentrations (e.g., ≥0.5 mM) C 14 KKc 12 K and CHX were (again) equipotent in their ability to affect viability of S. mutans in a preformed biofilm, killing >99% of the initial massive inoculum (10 9 CFU/ml) within one hour of exposure (Fig. 5d).

Discussion
Antibacterial membrane-active lipopeptides are currently gaining extensive interest for their potential to affect critical bacterial processes ranging from communication 8,9 to antibiotic functions, in mono-and combination-therapy 17 . For example, PMB and daptomycin are two naturally occurring bactericidal cyclic lipopeptides, clinically used against GNB and GPB, respectively, with MIC values ranging between 0.12 and 8 µg/ml (PMB) 58 or 0.015 and 32 µg/ml (daptomycin) 59 . In both cases, the acyl moiety plays a critical role 45,60 .
Acyl conjugation was also shown to enhance the antimicrobial properties of various AMPs and synthetic mimics alike 31,61,62 . For instance, acylated derivatives of dermaseptin, a broad spectrum amphibian AMP, have drastically influenced its activity spectrum, enabling its conversion to specific activity on either GPB or GNB, depending on the acyl selected 63,64 . Other interesting studies reported synthetic ultra-short lipopeptides (3-5 residues) with potent activities against plant-related pathogenic fungi 65 and bacteria 66 , though at higher doses 67 .
The present study provides evidence for the capacity of equivalent lipopeptides (generated via the OAK approach) to yield small molecules susceptible to be useful as simple investigation tools to help clarifying debated mechanistic aspects and potentially useful in biomedical applications. Namely, the findings reported herein, establish C 14 KKc 12 K as the shortest broad-spectrum antibacterial OAK known hitherto. The data also represent a quite remarkable yet ill-understood counter-intuitive outcome regarding the relationships between the hydrophobicity and self-assembly of this particular sequence. As expected from their respective elution time in HPLC using a hydrophobic column, C 14 KKc 12 K possesses an intermediate hydrophobicity value compared with C 12 KKc 12 K and C 16 KKc 12 K (Table 2). However, their tendencies for self-assembly, which presumably also depend on hydrophobicity, did not increase correspondingly, as evidenced by light scattering measurements (Fig. 1). Its sharp premature deviancy from linearity (as evidenced by high intensity) suggests that C 14 KKc 12 K forms drastically different aggregates in terms of three-dimensional organization. In contrast, our findings point to a dramatic increase in antibiotic efficiency since C 14 KKc 12 K achieved not only the lowest MIC value for this analog series (e.g., C 14 KKc 12 K MIC over S. mutans is 0.78 versus 1.56 µM for C 12 KKc 12 K or C 16 KKc 12 K) but has also upgraded its mode of action from bacteriostatic to bactericidal, including on GNB (Table 2). In this respect, the study expands and reinforces previous findings that illustrated the potential of N-terminal acyl conjugation to the core structure of AMPs 63, 68, 69 or OAKs as a potent and versatile strategy for optimizing the hydrophobic/cationic balance required for antimicrobial properties.
The data also hint to a relationship between antibacterial potency and bacterial efflux function. Although so far, OAK properties were often rationalized strictly in terms of interactions with bacterial membranes, it is now clear that efflux 70,71 represents another decisive factor whose contribution to the mode of action must be accounted for. Previously, we proposed that related but borderline-hydrophobic analogs (e.g., C 12ω7 KKc 12 K 18 or C 10 KKc 12 K 40 ) are substrates of the AcrAB-TolC system, the resistance nodulation division (RND) efflux pump present in E. coli and various enteric bacteria. The fact that C 12ω7 KKc 12 K was potently active on isogenic mutant strains (whose efflux-pump components were deleted) supports the view that inactivity on normal GNB stems from the OAK's rapid extrusion by these pumps 18 . Additional support for this view comes from the fact that bactericidal OAKs are also prone to deeper insertion within the CM 17 . In this respect, the fact that C 14 KKc 12 K is equally potent on both the wild type (AG100) and its deletion-mutant strain (AG100A) reinforces this hypothesis. Combined, our findings hint to a scenario implicating a simultaneous/competitive attraction of the OAK molecules in the periplasm, to the CM anionic phospholipids and the membrane-embedded efflux pumps. Consequently, borderline-hydrophobic OAKs would be more susceptible to extrusion, whereas outright-hydrophobic OAKs (e.g., C 14 KKc 12 K) are more likely to escape extrusion due to their tighter/deeper anchoring within the CM.
Considering the OAK properties in the context of a series of analogs (as outlined in Table 2), the resulting perspective provides an overall rational picture describing a bell-shaped continuum of effects that exacerbate (possibly through accumulation) with increasing hydrophobicity. Thus, the least hydrophobic among the tested lipopeptides (C 8 KKc 12 K) was clearly devoid of growth inhibitory activity against GNB but managed to induce mild damages, mostly to the OM 40 without affecting bacterial proliferation. While these effects are readily reparable, they may cause some delay in initial bacterial doubling time upon minor hydrophobicity increase (as observed with C 10 KKc 12 K) where they start affecting the CM as well, as expressed by partial depolarization 40 . Additional increase in hydrophobicity (to yield C 12 KKc 12 K) further intensifies these effects to cause a variety of non-reparable damages to both GNB and GPB as expressed by a full-fledged growth inhibitory activity (i.e., MIC). These damages are nonetheless not severe enough to jeopardize bacterial viability, hence often leading to a bacteriostatic mode of action, as observed previously 40 and in this study. Owing to its optimal hydrophobicity level, C 14 KKc 12 K was able to induce the most severe damages that culminated in rapid bactericidal outcomes. In contrast, excess hydrophobicity (e.g., in C 16 KKc 12 K 24 ) leads to the formation of tight peptide aggregates, which in turn, inverse the MAC potency-trend by limiting OAK's availability for optimal interactions with bacterial targets.   24 Trend reversal (reduced potency) due to excess hydrophobicity as selfassembly approaches the critical aggregation concentration In conclusion, our results support the view that N-terminal acyl-manipulations of the core sequence KKc 12 K, exhibit straightforward structure-activity relationships. Thus, simply by switching the N-terminal acyl, the OAK properties became tunable, gradually evolving from lack of "visible" antibiotic activity on to exerting bacteriostatic activity over GPB only, and ultimately, exercising broad-spectrum bactericidal activity. The fact that C 14 KKc 12 K was equally potent on both wild type and efflux mutant strains suggests that stronger anchoring within the CM enables hydrophobic MACs to escape extrusion by RND pumps, thereby providing a rational for the observed increased potency.
Besides their potential role as investigation tools, such compounds may be useful in treating infections involving multiple microbial populations, such as oral mucositis. Our findings may have relevance to various biofilm-associated micro-environmental niches that hamper drug efficacy in infections or industry related issues. Future studies might clarify this issue. Interestingly, a 35-residue-long MAC currently in phase 2 clinical trials, C16G2 72, 73 displayed specific antistreptococcal bactericidal properties in saliva. In this respect, the OAK platform might present advantages in the capacity to generate superior anti-biofilm candidates, including in terms of biological robustness, simplicity and production costs.

Materials and Methods
Peptide synthesis. OAKs were synthesized in-house (433 A Peptide Synthesizer; Applied Biosystems, Foster City, CA, USA) by the solid-phase method using 9-fluorenylmethyloxycarbonyl (Fmoc) active-ester chemistry on 4-methylbenzhydrylamine (MBHA) resin. OAKs were then deprotected and cleaved from the resin using trifluoroacetic acid:H 2 O (95:5) and purified to >95% chromatographic homogeneity by reverse phase high performance liquid chromatography (RP-HPLC) using C 18 column (Vydac), a flow rate of 2 ml/min and a linear acetonitrile gradient of 1%/min (Alliance; Waters, Milford, MA, USA). Peaks identity was verified by mass-spectrometry (Xevo G2 Tof; Waters, Milford, MA, USA). Purified OAKs were then lyophilized and kept as dry powder at −20 °C.

Organization in solution.
To assess the OAK's self-assembly in solution, serial two-fold dilutions of the OAK (initial concentration of 200 µM) were prepared in phosphate buffered saline (PBS; 10 mM Na 2 HPO 4 , 154 mM NaCl, pH = 7.4) and incubated for 2 hr at room temperature (RT). Light scattering at a 90° angle was measured through a 1 nm slit while holding both excitation and emission at 400 nm (Spectrophotometer Fluorolog-3 FL3-22; Horiba Jobin Yvon, Edison, NJ, USA).
To evaluate the disassembly of these aggregates upon bacterial exposure, an OAK solution (200 µM) was incubated (2 hr in PBS, at RT) after which, bacteria were added (10 5 CFU/ml) and the light scattering evolution of these suspensions was monitored as described above.
Bacteria. Gram Culture conditions. Staphylococci, Bacilli, Escherichia, Pseudomonas, Acinetobacter, Salmonella and Klebsiella species were grown in Luria Bertani broth (LB; 5 gr/l NaCl, 5 gr/l yeast extract, 10 gr/l tryptone). Enterococci and E. coli ML-35p were grown in Tryptic Soy Broth (TSB). Listeria and Streptococci were grown in Brain Heart Infusion (BHI). All bacteria were grown over-night at 37 °C with shaking. S. mutans 35668 plated on BHI agar for enumeration was grown for 48 hr. S. mutans UA159 was grown in 5% CO 2 enriched atmosphere. Fusobacterium and Porphyromonas were grown in Wilkins-Chalgren growth medium and an anaerobic atmosphere.
Minimal inhibitory concentration (MIC) was determined using the microdilution assay. Mid-log-phase bacteria at 5 × 10 5 CFU/ml were incubated in a 96-well plate with serial two-fold dilutions of the tested compound for 18-24 hr at 37 °C (final volume of 200 µl). O.D. at 620 nm was measured (Synergy HT, BioTek Instruments, Winooski, VT, USA), and the MIC was determined as the lowest concentration for which no increase in O.D. was detected.
Cytoplasmic membrane permeation to ethidium bromide (EtBr) was evaluated as follows: Mid-log-phase bacteria at 1 × 10 8 CFU/ml were centrifuged for 5 min at 15,000 g. Pellet was washed twice with PBS containing 0.5% glucose (pH = 7.4), suspended in the same buffer and incubated for 10 min at 37 °C with shaking. 180 µl of the bacterial suspension were mixed in a 96-well plate with 25 µl of the tested compound and EtBr (EtBr final concentration = 1 µg/ml) and fluorescence was recorded immediately (excitation: 535 nm, emission: 590 nm) for up to 30 min at 37 °C with shaking (Synergy HT, BioTek Instruments, Winooski, VT, USA).
Outer and cytoplasmic membrane permeation in Gram-negative bacteria was assessed using the engineered E. coli ML-35p by monitoring the chromogenic hydrolysis of two indicators: ortho-nitrophenyl-β-galactoside (ONPG) and nitrocefin. Mid-log-phase E. coli ML-35p were centrifuged for 5 min at 15,000 g and the supernatant was removed. Pellet was washed three times with sodium phosphate buffer (SPB; 10 mM NaH 2 PO 4 , pH = 7) and suspended in the same buffer (O.D. at 620 nm = 1). Bacteria were then 10-fold diluted into SPB containing 3% TSB. 100 µl of bacterial suspension were placed in a 96-well plate with 100 µl of the tested compound and 20 µl of either ONPG (final concentration = 2.5 µM) or nitrocefin (final concentration = 25 µM). Hydrolysis of ONPG and nitrocefin was monitored immediately by recording the absorbance at 420 nm and 486 nm respectively, for 30 min at 37 °C with shaking (Synergy HT, BioTek Instruments, Winooski, VT, USA).
Cytoplasmic membrane depolarization was assessed by monitoring the displacement of the membrane binding fluorescent dye DiSC 3 (5) (3,3′-dipropylthiadicarbocyanine iodide) as follows. Mid-log-phase bacteria at 5 × 10 8 CFU/ml were centrifuged for 5 min at 15,000 g. Pellet was washed twice with 5 mM 4-(2-hydroxyethyl) piperazine-1-ethanesulfonic acid (HEPES) containing 5 mM glucose (pH = 7.2) and suspended in the same buffer. Bacteria were diluted 10-fold in the same buffer, DiSC 3 (5) was added to a final concentration of 4 µM and samples were incubated at RT in the dark for 1 hr. KCl was added to a final concentration of 100 µM and incubation continued for an additional hour. 180 µl of bacterial suspension were placed in a 96-well plate and fluorescence was recorded until signal stabilization (excitation: 620 nm, emission: 680 nm). Then, 20 µl of the tested compound were mixed into the wells and fluorescence was recorded immediately for up to 30 min at 37 °C with shaking (Synergy HT, BioTek Instruments, Winooski, VT, USA).
Dansyl-polymyxin displacement assay was assessed by displacement of dansyl-polymyxin B bound to lipopolysaccharide (LPS) as follows. Polymyxin B sulfate was covalently attached to dansyl chloride and mono-dansyl Polymyxin B (DPMB) was purified by RP-HPLC. 180 µl of 5 mM HEPES containing 3 µg/ml LPS (from E. coli or P. aeruginosa) and 2 µM mono-DPMB were incubated in a 96-well plate with 20 µl of the tested compound for 1.5 hr at RT and fluorescence (excitation: 340 nm, emission: 485 nm) was measured immediately (Synergy HT, BioTek Instruments, Winooski, VT, USA).
Bactericidal kinetics in saliva was assessed as follows. Whole saliva was pooled from two-three healthy volunteers at a time, after obtaining their informed consent. To test for specific activity against S. mutans, saliva was centrifuged for 5 min at 15,000 g and 4 °C, and the supernatant was spiked with 1 × 10 7 CFU/ml of mid-log-phase S. mutans 35668. Otherwise, activity against natural oral microflora was also performed on whole saliva. Saliva was mixed with the tested compound, incubated at 37 °C with shaking, and aliquots were taken at t = 0, 2, 5 and 10 min. Aliquots were subjected to serial 10-fold dilutions in saline and plated on BHI agar for enumeration.
Anti-biofilm activity was evaluated against established biofilms in 96-well plates as follows. 200 µl of mid-log-phase S. mutans 35668 at 5 × 10 5 CFU/ml in BHI containing 2% sucrose were placed in each well, and plates were incubated for 24 hr at 37 °C without shaking. Unattached cells were removed by decanting the plates and biofilms were washed three times with milliQ water. Solutions of the tested compounds were placed on the biofilms for up to 3 hr at 37 °C without shaking. Plates were decanted in 0.5 hr intervals and biofilms were washed three times with milliQ water to remove any compound residues. Biofilms were scraped from the bottom of the wells, suspended in BHI and sonicated for 5 min in a sonication bath, subjected to serial 10-fold dilutions in saline and plated on BHI agar for enumeration.