Structure-activity Relationship Study to Develop Peptide Amphiphiles as Species-Specific Antimicrobials

Abstract

Antimicrobial peptide amphiphiles (PAs) are a promising class of molecules that can disrupt the bacterial membrane or act as drug nanocarriers. In this study, we prepared 33 PAs to establish supramolecular structure-activity relationships. We studied the morphology and activity of the nanostructures against different Gram-positive and Gram-negative bacterial strains (such as Staphylococcus aureus, Escherichia coli, Pseudomonas aeruginosa and Acinetobacter baumannii). Next, we used principal component analysis (PCA) to determine the key contributors to activity. We found that for S. aureus, the zeta potential was the major contributor to the activity while Gram-negative bacteria were more influenced by the partition coefficient (LogP) with the following order P. aeruginosa > E. coli > A. baumannii. We also performed a study of the mechanism of action of selected PAs on the bacterial membrane assessing the membrane permeability and depolarization, changes in zeta potential and overall integrity. We studied the toxicity of the nanostructures against mammalian cells. Finally, we performed an in vivo study using the wax moth larvae to determine the therapeutic efficacy of the active PAs. This study shows cationic PA nanostructures can be an intriguing platform for the development of nanoantibacterials.

Graphical Abstract

Antimicrobial peptide amphiphiles nanostructures were studied regarding the supramolecular structure-activity and the key structure contributors for activity. For Staphylococcus aureus the zeta potential was more important while Gram-negative bacteria were more influenced by LogP. Mechanism of action and an in vivo study was also performed. We showed these nanostructures can be an intriguing platform for the development of novel nanoantibacterials.

Introduction

The deleterious consequences of infectious diseases (from pneumonia and fever to cancer and death) are a major health burden worldwide.^[1] Especially concerning is the increasing number of antimicrobial resistance (AMR) observed in various pathogenic strains. Around 4.95 million deaths were associated with bacterial AMR in 2019, including 1.27 million directly attributable to bacterial AMR. The top six bacterial species responsible for human deaths are: Escherichia coli, Staphylococcus aureus, Klebsiella pneumoniae, Streptococcus pneumoniae, Acinetobacter baumannii, and Pseudomonas aeruginosa.^[2] Unfortunately, there is an imbalance between the escalating need for new therapeutic options and the limited supply of antimicrobial agents under development.^[3] This stems from the many scientific and economic challenges associated with securing clinical approval which have halted antibiotic development^[4], resulting in a low number of drugs reaching the market. Moreover, newly approved antibiotics (from 2017-2020) are based mainly on previously known molecules.^[4] It is clear there is a need of antimicrobial agents with novel structures.

Antimicrobial peptide amphiphiles (PAs) are inspired by antimicrobial peptides. PAs possess properties that make them attractive platforms for the development of antibacterials including: 1) broad-spectrum activity against a wide range of pathogens;^[5] 2) the ability to disrupt the bacteria membrane, which reduces the likelihood of resistance development;^[6] 3) a rapid killing profile, critical for infections where a fast acting therapy is needed^[7]; 4) due to their membrane-targeting mechanism, PAs may act synergistically with other antimicrobials enhancing overall efficacy;^[8] 5) PA nanostructures can be used as delivery systems allowing to “tune” the PK/PD profile of the encapsulated drugs.^{[8, 9]}

However, there are challenges associated with developing PAs as antibiotics. Their toxicity towards human cells, especially red blood cells, needs to be addressed to fully harness their potential for clinical use.^[10] More importantly, the lack of understanding of the key determinants for activity and the high complexity of PA-nanostructures hinder their development as antimicrobials. PAs can form nanostructures of various shapes, sizes and surface properties.^[11] These nanostructures are highly dynamic and their physiochemical and biological properties can be influenced by environmental cues such as pH, proteins, counterions, etc. Based on the molecular composition of the PAs segments their interactions with membrane components can change. Thus, the large number of molecular factors and the unpredictable interplay among them make it difficult to establish supramolecular structure activity relationship (SSAR) of antimicrobial PAs.

Although PAs are already under investigation for their antimicrobial properties^{[12, 13]}, there is a knowledge gap about how these molecules can be engineered to generate structural specificity for different bacterial strains. Among the benefits of achieving a specific treatment, we can mention preservation of beneficial bacteria, lower risk of toxicity/side effects, a focused treatment and improved patient outcomes.^[14]

In this report, we evaluated the activity of 33 PAs against both Gram-positive and Gram-negative bacteria using minimum inhibitory concentration (MIC) and minimum bactericidal concentration (MBC) assays. Then, we performed a principal component analysis (PCA) based on the zeta potential and the calculated LogP value of the PA molecules to find a link between these parameters and bacterial specificity. Next, the bacterial membrane changes induced by PA treatment were evaluated by membrane permeability (PI uptake) and depolarization (DiSC 3(5)) tests. The direct visualization of membrane alterations was obtained by transmission electron microscopy (TEM) and scanning electron microscopy (SEM). The toxicity of the most active PAs towards HEK- 293T cells was tested. Finally, an in vivo study was performed using a larvae model to determine the toxicity and therapeutic efficacy of the best PAs.

Results and Discussion

Design and Characterization of PAs

We designed and synthesized 33 PAs with different alkyl tails and peptide sequences. A diagram with a generic PA structure and the different structural portions modified are shown in Figure 1. The cationic segment combined positively charged amino acids: lysine, K (pK_a value ≈ 10.7), arginine, R (pK_a value ≈ 12.1), and histidine, H (pK_a value ≈ 6.04). We also explored a cationic peptoid segment (PAs 11, 26-28). Peptoids are isomers of peptide with the side chain attached to the amide nitrogen instead of the α-carbon, resulting in increased flexibility and stability. Peptoid modifications have been reported to increase the selectivity between bacteria and human erythrocytes.^{[15, 16]} Further, Ryge and colaborators reported a series of lysine-peptoid hybrids that are active against several clinically relevant bacteria such as MRSA and P. aeruginosa.^[17] To modulate the internal cohesion of the nanostructures and to vary the lipophilicity of the PAs we used hydrophobic amino acids such as alanine, A; valine, V ; phenylalanine, F; tryptophan, W; and tyrosine, Y.^[18]

Figure 1.

General PA structural diagram containing the main modifications performed in this study. The position of the residues (and peptoid component) along the peptide sequence was modified in some PAs (see Table 1). Different residues (besides the most used one which is lysine, K) were introduced along different positions in the sequence. One, two or three tail unsaturations were introduced in selected PAs. The lipidic tail length contains 14 to 18 carbons. The lipidic tail was alternatively coupled to the amine in the side chain of first lysine (instead of the regular N-terminal position) in PAs 29-31 (see Table 1). R¹ can also be a lysine side chain. PA 11 is an exception to this diagram because contains it two peptoid Lys side chains in the first two residues of the peptide (see Table 1).

For the hydrophobic core of the nanostructures, we explored alkyl tails with different lengths (14 to 18 carbons) and odd or even number of carbons. Alkyl chain with an odd number of carbons have been reported to have membrane-disruptive properties^[19] and increased antifungal activity.^[20] Furthermore, since unsaturated alkyl tails have been shown to exhibit more potent bactericidal, higher biological stability, and selectivity to bacteria^{[21, 22]} some PAs (PA 7, PA 10, PA 32, and PA 33) contained such modification.

Previous studies have demonstrated that positive charges (which can display electrostatic interaction with biological membranes),^[23] hydrophobicity (hydrophobic interaction with lipid membranes),^[24] and morphology (general interaction with membrane)^{[25, 26]} are critical contributors to the antimicrobial activity of PAs. Table 1 lists the zeta potential, calculated LogP values, and morphology observed by TEM. Only PAs 12-15 formed fibrous shapes (due to the presence of β-sheet promoting residues) while the rest formed spherical micelles (shown in Table 1, selected TEM in Figure 2 and Figure S2-S6 in supporting information).

Table 1.

PAs sequences, properties, and morphology (by TEM).

PAs	Structure	Zeta potential (mV)	LogPa	Morphology
PA 1b	[C16:1]K4H	20.6±0.9	1.52	Micelles
PA 2	[C16]K4	9.96±0.6	2.88	Micelles
PA 3	[C15]K4	12.3±0.2	2.38	Micelles
PA 4	[C14]K4	14.6±1.6	1.87	Micelles
PA 5	[C16]K4H	11.2±1.0	2.01	Micelles
PA 6	[C15]K4H	5.89±1.2	1.50	Micelles
PA 7b	[C18:3]K5	23.1±1.4	0.61	Micelles
PA 8	[C15]K5	33.9±3.5	1.32	Micelles
PA 9	[C17]K5	35.6±3.3	2.33	Micelles
PA 10b	[C18:1]K5	42.5±1.1	2.35	Micelles
PA 11c	[C18]K3K2(peptoid)	38.3±0.9	1.78	Micelles
PA 12d	[C16]V2A2K3	34.8±0.8	3.24	Nanofibers
PA 13d	[C16]V4K5	31.2±0.9	2.69	Nanofibers
PA 14d	[C14]V3A3K4	33.9±0.6	0.83	Nanofibers
PA 15d	[C16]V3A3K4	34.0±0.2	1.84	Nanofibers
PA 16	[C18]K2RK	11.0±1.7	3.44	Micelles
PA 17	[C18]K2RK2	2.53±0.3	2.39	Micelles
PA 18	[C17]K2RK2	26.4±0.2	1.88	Micelles
PA 19	[C16]K2RK2	28.5±0.4	1.38	Micelles
PA 20	[C17]K2RK	27.6±1.5	2.94	Micelles
PA 21	[C16]K2RK	32.5±3.2	2.43	Micelles
PA 22	[C14]FK4	31.3±1.7	2.77	Micelles
PA 23	[C14]WK4	47.3±1.8	2.92	Micelles
PA 24	[C14]YK4	22.5±2.5	2.29	Micelles
PA 25e	[C14](2-Nal)K4	43.7±2.8	3.95	Worm-like micelles
PA 26c	[C18]K2KpeptoidK2	39.1±1.4	3.09	Micelles
PA 27c	[C17]K2KpeptoidK2	38.9±0.1	2.58	Micelles
PA 28c	[C16]K2KpeptoidK2	28.1±1.1	2.08	Micelles
PA 29f	[C16]K-K5	37.4±1.3	0.84	Micelles
PA 30f	[C14]K-K5	28.7±2.4	−0.17	Micelles
PA 31f	[C17]K-K5	43.8±1.7	1.35	Micelles
PA 32b	[C18:2]K5	51.4±0.7	1.86	Micelles
PA 33b	[C16:1]K5	47.3±1.9	1.34	Micelles

^aLogP values obtained from https://molinspiration.com/cgi/properties (Molinspiration Cheminformatics©).

^bPAs with unsaturations in the tail. See SI for structures and location of unsaturations.

^cPAs containing one or two peptoid lysine side chain.

^dPAs with β-sheet forming sequence.

^e2-Nal stands for a 2-naphthyl-L-alanine residue.

^fReversed PAs (lipid tail bound to the side chain of the first lysine).

Figure 2.

TEM images of selected PAs: odd number alkyl tail PA 3 and PA 9, peptoid hybrid PA 26, and reversed PA 29. Scale bar 100 nm. PAs were dissolved in water at 1 mg/mL, annealed at 80 °C for 30 mins, and aged overnight before testing.

2. Antimicrobial Activity Studies

2.1. PCA Indicates the Contributors of Broad-Spectrum Antimicrobial PAs

As shown in Table 2, the PAs exhibit broad-spectrum activity. They can be classified as highly active (MIC ≤ 16 μg/mL), intermediate (MIC 32 μg/mL) and low active (MIC ≥ 64 μg/mL). The PAs with highest activity against P. aeruginosa were 5 (PA 9, PA 18-19, PA 22, and PA 23). For E. coli K12, 16 PAs (PA 2-3, PA 5-PA 6, PA 9, PA 17-23, and PA 25-29) were promising. Meanwhile, 17 structures were highly active against A. baumannii (PA 3-6, PA 9-11, PA 18-19, PA 22-23, PA 25, PA 27-30, and PA 33). For the Gram-positive MRSA 8 PAs displayed high activity (PA 3, PA 9, PA 18-19, PA 22-23, PA 25, and PA 26). These 8 PAs were tested against another Gram-positive bacteria, a vancomycin-resistant strain (VRE, patient isolated) of Enterococcus faecium and exhibited similar MICs.

Table 2.

MIC and MBC values of PAs against three Gram-negative strains and two Gram-positive strains.

		Gram-negative						Gram-positive
		Pa a		Ec b		Ab c		Sa d		Ef e
	Sequence	MIC	MBC	MIC	MBC	MIC	MBC	MIC	MBC	MIC	MBC
Gent.		0.5	-	0.5	-	0.5	-	0.5	-	-	-
PA 1f	[C16:1]K4H	n.a.	n.a.	128	128	128	n.a.	128	n.a.	-	-
PA 2	[C16]K4	32	128	16	64	32	128	32	128	-	-
PA 3	[C15]K4	32	n.a.	16	32	16	64	16	128	16	n.a.
PA 4	[C14]K4	128	n.a.	32	128	16	128	32	128	-	-
PA 5	[C16]K4H	64	128	16	128	16	64	32	64	-	-
PA 6	[C15]K4H	64	128	16	32	16	64	32	64	-	-
PA 7f	[C18:3]K5	64	n.a.	32	64	32	64	64	64	-	-
PA 8	[C15]K5	128	128	64	128	64	128	64	128	-	-
PA 9	[C17]K5	4	32	4	16	4	16	8	32	4	256
PA 10 f	[C18:1]K5	128	n.a.	32	128	16	128	32	128	-	-
PA 11 g	[C18]K3K2(peptoid)	32	128	32	128	16	64	32	128	-	-
PA 12h	[C16]V2A2K3	128	n.a.	128	n.a.	64	128	128	n.a.	-	-
PA 13h	[C16]V4K5	n.a.	n.a.	64	n.a.	128	n.a.	128	n.a.	-	-
PA 14h	[C14]V3A3K4	128	n.a.	32	n.a.	32	128	64	n.a.	-	-
PA 15h	[C16]V3A3K4	n.a.	n.a.	64	n.a.	64	128	64	n.a.	-	-
PA 16	[C18]K2RK	64	n.a.	32	128	32	128	32	128	-	-
PA 17	[C18]K2RK2	32	128	16	64	32	64	32	128	-	-
PA 18	[C17]K2RK2	16	64	8	32	16	64	16	64	8	64
PA 19	[C16]K2RK2	16	64	8	32	8	32	16	128	8	32
PA 20	[C17]K2RK	64	n.a.	32	n.a.	32	n.a.	32	64	-	-
PA 21	[C16]K2RK	64	128	16	64	32	128	32	128	-	-
PA 22	[C14]FK4	16	64	16	64	16	128	16	128	16	256
PA 23	[C14]WK4	16	128	16	64	16	64	16	128	16	256
PA 24	[C14]YK4	64	n.a.	32	64	32	128	32	64	-	-
PA 25 i	[C14](2-Nal)K4	32	128	16	64	16	128	16	64	16	256
PA 26 g	[C18]K2KpeptoidK2	32	128	8	32	32	n.a.	16	128	16	n.a.
PA 27 g	[C17]K2KpeptoidK2	32	128	8	32	16	128	32	128	-	-
PA 28 g	[C16]K2KpeptoidK2	32	128	16	64	16	128	32	n.a.	-	-
PA 29 j	[C16]K-K5	32	128	16	128	16	128	32	128	-	-
PA 30 j	[C14]K-K5	32	128	32	128	16	n.a.	32	128	-	-
PA 31j	[C17]K-K5	32	128	32	n.a.	32	n.a.	32	128	-	-
PA 32f	[C18:2]K5	64	n.a.	32	128	32	n.a.	32	n.a.	-	-
PA 33 f	[C16:1]K5	64	n.a.	32	128	16	128	32	n.a.	-	-

^aPseudomonas aeruginosa ATCC27853.

^bEscherichia coli K12.

^cAcinetobacter baumannii strain recovered from patient.

^dStaphylococcus aureus JE2 (methicillin-resistant).

^eEnterococcus faecium VRE recovered from patient (strain tested against 8 of the most active PAs only).

^fPAs with unsaturations in the tail.

^gPAs containing one or two peptoid lysine side chain.

^hPAs containing a sequence that can form β-sheets.

ⁱ2-Nal stands for a 2-naphthyl-L-alanine residue.

^jReversed PAs (lipid tail bound to the side chain of the first lysine). Gent.: gentamycin. PAs with high activity (MIC of 4 to 16 μg.mL⁻¹) are in bold green. MIC (μg.mL⁻¹), MBC (μg.mL⁻¹). n.a.: no activity within the tested concentrations of 128 to 2 μg.mL⁻¹. All PA assemblies tested were prepared in water at pH 7.

Interestingly, we observed that peptides with odd number of carbons display good activity, but the potency is dependent on the peptide sequence. For PA 2 (C₁₆), PA 3 (C₁₅) and PA 4 (C₁₄), all displaying 4 lysines, the PA 3 has better antimicrobial activity with 15 carbons). For the PAs containing 5 lysines, we observed that both PA 8 and PA 9, with 15 and 17 carbons respectively, have activity, but PA 9 displayed higher potency. Finally, the PAs containing histidine and 4 lysines (PAs 5 and 6) and 15 or 16 carbons did not show any difference in activity (Table 2).

Comparing our results with previously reported data of our group, we observed a reduction in the activity when inserting unsaturations in the alkyl tails of PAs. For instance, the PA previous reported C₁₈K₅ presents MIC ranging from 4-16 μg/mL for the same bacterial strains tested.^[12] On the opposite, PA 10, PA 32 and PA 7 containing 1, 2 and 3 unsaturations respectively displayed MIC values from 16 to 64 μg/mL (Table 2). Although we are uncertain why this happens, the rigidity of the double bonds may affect the dynamics of the PA, which may affect membrane insertion and activity.

Our theory is that the cationic PAs interact electrostatically with slightly negatively charged biological membranes (i.e. bacteria and cancer cell lines),^[9] disrupting them. Although not fully understood, our hypothesis is the PAs nanostructures act as a monomer’s reservoir, disassembling and then promoting intercalation into the membrane by its hydrophobic tail.^[12] This hypothesis can be supported by another study which showed inversed relationship between critical aggregation concentration (CAC) and MIC of peptoids.^[27] The PAs described herein can exert a bacteriostatic or bactericidal effect in a concentration-dependent manner. The lower the MIC presented by a PA, the lower the concentration required to achieve bactericidal activity (MBC). We believe PAs with high membrane disrupting potency will lead to structures with a smaller gap between bacteriostatic and bactericidal effect (membrane damage should kill the pathogen).

Although, the varying composition of the bacteria membrane of different pathogens makes the systematic study of structure-activity relationship (SAR) of PAs hard, identifying the major structural contributors to activity against specific organisms is key for the development of PA nano antibacterials (kill the pathogen) or nanocarriers (deliver a molecule, the PAs may have activity or be inert).^{[28, 29, 30]} Thus, we used principal component analysis (PCA) to analyse the relationship between the physicochemical properties of PAs and antimicrobial activity aiming to establish a guidance to design species-specific antibiotics.

In Figure 3a-d, we present the PCA results of MRSA JE2, E. coli K12, A. baumannii, and P. aeruginosa, respectively. In the PCA analysis we included MIC, zeta potential, and LogP values to calculate the principal components (PC1 and PC2). The PC1 and PC2 scores are linear combinations of each of the variables in the dataset. Around 70% of the variance in the dataset can be explained by PC1 and PC2 in this analysis. The loadings plot visually represents the relationships between the variables and the principal components. If the loading value is closer to 1.0 or −1.0 (longest arrow), we can conclude that the respective variable has greater contribution to the variance (i.e. the observed result).

Figure 3.

PCA 2D plots of a) MRSA JE2 b) E. coli K12 c) A. baumannii and d) P. aeruginosa, determined by MIC value, zeta potential, and calculated LogP of PAs. Green dots represent PAs with high antimicrobial activity (MIC 4-16 μg.mL⁻¹), blue dots intermediate activity (MIC 32 μg.mL⁻¹) and red dots low activity (MIC 64-128 μg.mL⁻¹).

For MRSA JE2 the zeta potential is the major contributor having a loading value of −0.910 (PC2), while the LogP possess a loading of 0.732 (PC1) (Figure 3a). Therefore, for this bacteria the zeta potential of the PAs is more important than their LogP.

The three Gram-negative strains are shown in Figure 3b-d. For E. coli K12, LogP (PC2, loading −0.905) is more important than zeta potential (PC1, loading −0.738). For A. baumannii, PC1 is represented by a combination of LogP and zeta potential with similar loadings, −0.768 and −0.705. This is an indication that both contribute in a similar manner in the differentiation of the groups (high, intermediate and low activity). For P. aeruginosa, LogP (PC2, loading 0.924) is more important than zeta potential (PC1, loading 0.789).

In summary, opposite to MRSA JE2, the activity against Gram-negative bacteria correlates better with LogP than zeta potential, with the following order P. aeruginosa > E. coli > A. baumannii. Meanwhile, the preference of zeta potential of these three Gram-negative strains are similar with an order of P. aeruginosa > A. baumannii > E. coli K12. Although among Gram-negative species LogP is the major contributor, for P. aeruginosa the activity is also dependent on zeta potential (according to the PCA loadings). We suppose that for this specie there is a tighter balance for LogP and zeta potential which is reflected by the lower number of active PAs.

2.2. Structure-Activity Relationship Study

We combined the PCA scores to determine key parameters for antimicrobial activity. The first general rule is that PA with fibrous shape (PA 12- PA 15) did not display good activity regardless of their zeta potential and LogP value, as previously suggested.^[12] The antimicrobial activity of PAs 12-15 was lower when compared to the PAs forming spherical micelles (Table 2) and this may be explained by the strong intermolecular cohesion of fibrous nanostructures, which disturbs the release of individual PAs to act on the membrane.^[31] An analysis of the results for each bacteria is detailed below.

1). MRSA JE2:

While a LogP value > 1.32 is necessary for PAs to be active, PCA suggests that the dominant contributor for PA’s activity is the zeta potential (ranging from 12.3±0.2 to 47.3±1.8 mV). Notably, there is a much wider window of zeta potential for PAs to be active for MRSA JE2 comparing to the other strains in this study. MRSA JE2 does not have lipopolysaccharide (LPS, which is found in Gram-negative bacteria only) resulting in lower density of negative charges on the cell surface.^[32] This may reduce the required number of net charge of cationic PAs to disrupt bacteria membrane. This hypothesis can be reinforced by the greater alteration of MRSA JE2 zeta potential when compared to E. coli K12 and A. baumannii caused by the PAs (shown in section 2.3). Further, peptidoglycan (PG) is the major component of Gram-positive bacteria cell wall, which contains glycan chains connected by short peptides.^[33] The large presence of glycans and peptides in the MRSA JE2 cell wall can provide more sites for H-bonding within the peptide sequence of PAs. This may also explain why a larger range of zeta potential was found among the active PAs against MRSA JE2.

2). E. coli K12:

Based on the PCA results, a LogP larger than 1.84 (PA 15) is required for activity. Of all the strains analysed, E. coli K12 is the one that requires higher LogP value. One possible explanation may be linked to the fact that E. coli K12 presents higher density of LPS in the outer membrane (approximately 2-3x10⁶ molecules of LPS).^[34] In comparation, Acinetobacter species have been reported to display lower amounts of LPS (A. nosocomialis 1.74x10⁵ and A. baumannii 6.9x10⁴).^{[35, 36]} The higher levels of LPS in the outer membrane produces more hydrophobic points and requires the PA to have more hydrophobicity for a good interaction. Interestingly, even though E. coli presents higher amounts of LPS in the outer membrane, it has been reported that the zeta potential is generally lower compared to A. baumannii.^{[37, 38]} Our results agree with this observation since the major contributor for E. coli is LogP alone.

3). A. baumannii:

the PCA data indicated that zeta potential and LogP values are equally critical for good activity. Based on this study, if the PA meets a threshold of zeta potential > 20 mV and LogP > 1.32 is needed for activity against A. baumannii. If a PA does not fit the requirement, a ratio of zeta potential/LogP < 10 also provides activity.

A difference between E. coli K12 and A. baumannii is the structural composition of LPS. A. baumannii mainly produces hepta-acylated lipid A (a component of LPS) while E. coli K12 has predominant hexa-acylated lipid A.^[39] Further, wild type A. baumannii lipid A also presents an additional hydroxyl group on the 2’-linked secondary acyl chain.^[40] These two structural features create additional hydrogen bonding sites compared to the LPS of E.coli K12. For the PAs on this study, the net charge (zeta potential) is related to the amino acid residues, which means a higher zeta potential is related to a larger cationic peptide sequence. The larger peptide sequence introduces more coulombic attractions and hydrogen bonding between the PAs and the lipid A of A. baumannii in the LPS. This explanation does not apply to the fiber-like nanostructures because they are more stable, and that prevents (or decreases) the insertion process. In addition, as mentioned before, usually the zeta potential is higher in A. baumannii wild type strains. Thus, we suppose that due to these differences in LPS, both zeta potential and LogP are critical for the antimicrobial activity as pointed out by the PCA analysis.

4). P. aeruginosa:

a combination of zeta potential > 26 mV and LogP > 1.38 is necessary to have good activity against P. aeruginosa. However, PAs with some modifications (cationic amino acid alteration and unsaturated alkyl tail) do not show activity even if they fall into the required range. Meanwhile, PAs with peptoid and K-side chain alkylation can keep the intermediate activity. One important feature for P. aeruginosa is that its outer membrane is highly packed and ordered due to an asymmetric pattern, which restricts the fluctuation and movement of the phospholipid layer.^[41] Thus, active PA, such as PA 9 (C₁₇K₅) possesses 1) a high affinity towards bacteria membrane by sufficient electrostatic interaction and hydrogen bonds; and 2) a lipid tail displays a good flexibility to insert, adjust and interact better with the bilayers. Inactive PA 20 and PA 21 have one K residue less than the active PA 18 and PA 19. The lack of one K residue results in reduced hydrogen bonding between PA backbone and membrane, thus interfering with their activity.^[41] Similar rationale can be applied to PAs with peptoid (reduced hydrogen bonding in the backbone) and K-side chain alkyl tail (spacer between peptide backbone and alkyl tail), which have reduced activity than the unmodified counterparts, but still display activity. A summary of the results described in this section can be found on Figure 8.

Figure 8.

Summary of the major contributors for antimicrobial activity according to each bacteria studied. Image partially created with Biorender^™.

In conclusion, zeta potential and LogP values can be a preliminary filter for active PA scanning. However, to ameliorate the SAR, more parameters should be included, such as PA-membrane affinity, molecular area, and PA flexibility. Further, it is important to note that we did not include specific parameters to better describe small changes in the structure such as unsaturation and position of the alkyl tail (even though LogP is subtly affected by these changes) which may help distinguish between groups of active and inactive molecules. Given the fact that the PAs studied here contain a lipidic portion, unnatural amino acids or peptoids, traditional peptide properties calculator software may not be able to process such molecules, making harder to obtain appropriate data.

2.3. Various Membrane Alterations on Different Bacteria Strains Caused by PAs

To understand the action on Gram-negative bacteria (E. coli K12 and A. baumannii) and one Gram-positive bacteria (MRSA JE2) membranes, we assessed the membrane permeability, depolarization, and bacteria zeta potential after treatment. A key parameter is understanding changes in membrane permeability upon treatment. Propidium iodide (PI) is a cell viability assay. PI can intercalate between DNA base pairs but only when the membrane is damaged (it cannot cross the membrane of live cells).^[42] Membrane potential is another important characteristic when considering activity. DiSC 3(5) (3,3’-Dipropylthiadicarbocyanine iodide) is a voltage-sensitive dye which is distributed inside and outside of bacterial cells at standard membrane potential. Upon treatment by depolarizing agents, the dye is released into the external medium, increasing its fluorescence intensity.^[43] Finally, the net charge on the bacterial surface is negative and is balanced by counter ions of opposite charge from the surrounding media, therefore we measured the zeta potential of bacteria after treatment.^[44]

For MRSA JE2 (Figure 4a-c), the PI assay (Figure 4a) shows a significant change in fluorescence for PA 3 at 2X MIC and PA 9 and 26 at both tested concentrations, indicating that membrane permeability has been affected. From the DiSC 3(5) assay (Figure 4b), a significant change on the membrane potential can be observed from all PAs with the order of PA 26 > PA 9 > PA 3. Notably, the zeta potential of these PA nanostructures follows the same pattern. PA 3 micelles own the lowest positive charge (indicated by zeta potential in Table 1), which may explain the lowest ability to interact and change the bacteria membrane potential. For MRSA JE2, as shown by the PCA analysis, the zeta potential of the nanostructures is the main contributor to PA activity. In this context, PA 9 micelles present the higher potency of these three analyzed structures (MIC: 8 μg/mL), which is in line with a high value for the zeta potential (35.6 mV). A similar alternation of MRSA JE2 zeta potential is caused PA 9 and PA 26 (Figure 4c). The control, daptomycin, induces (slightly) the zeta potential to be more negative, which might be the result of daptomycin-Ca²⁺ complexes^[45] since removal of Ca²⁺ may decrease the zeta potential. These observations highlight that variances in membrane composition among different strains have significant impact on the interacting property of PAs.

Figure 4.

Antimicrobial mode of action study on MRSA JE2, E. coli K12 and A. baumannii using PA 3, PA 9, and PA 26 (plus controls untreated (UTD), Polymyxin B (PMB) and Daptomycin (DAPTO) at 0.5,1, 2, 4 or 10X the respective MIC value. a) Membrane permeability of PAs 3, 9 and 26 (0.5, 1 and 2X MIC) tested by PI uptake measurements against MRSA JE2. b) Membrane depolarization of PAs 3, 9 and 26 (0.5, 1 and 2X MIC) tested by DiSC 3(5) assay against MRSA. c) Zeta potential of MRSA JE2 altered by PAs 9 and 26 treatment (10X MIC) determined by DLS. d) Membrane permeability of PAs 3, 9 and 26 (1 and 2X MIC) tested by PI uptake against E. coli K12. e) Membrane depolarization of PAs 3, 9 and 26 (1 and 2X MIC) tested by DiSC 3(5) assay against E. coli K12. f) Zeta potential of E. coli K12 altered by PAs 9 and 26 (10X MIC) determined by DLS. g) Membrane permeability of PAs 3, 9 and 26 (1 and 2X MIC) tested by PI uptake against A. baumannii. h) Membrane depolarization of PAs 3, 9 and 26 (1 and 2X MIC) tested by DiSC 3(5) assay against A. baumannii. i) Zeta potential of A. baumannii altered by PAs 9 and 26 (10X MIC) determined by DLS. Fluorescence for PI and DiSC 3(5) assays was recorded after 2 hours of treatment.

For E. coli K12 (Figure 4d-f) a significant change in membrane permeability can be observed for all PAs in the PI assay (Figure 4d). As shown in Figure 4e, the three PAs can significantly induce the depolarization of the bacteria membrane, in the similar level of PMB. Interestingly, PA 9 and PA 26 alter the zeta potential of the bacteria significantly (Figure 4f), from a negative value to a positive value, which is also reflected in the activity of these two PAs, which are the most active against E. coli K12 in the data set (MIC: 4 and 8 μg/mL for PA 9 and PA 26, respectively). Meanwhile, PMB does not show a notable change on the bacteria zeta potential.

For A. baumannii, as shown in Figure 4g, only PA 26 and PMB presented a significant increase in PI intensity. Meanwhile, there is no dose dependence for all these samples from 0.5 MIC to 2 X MIC. Then, for the DiSC 3 (5) assay, only PA 3 was not able to change the membrane potential (Figure 4h). The order of activity was PA 26 > PMB > PA 9. For the zeta potential, only PA 26 was able to change the zeta potential of bacteria significantly (Figure 4i). This case is interesting due to its lack of correlation with the MIC values of these PAs. Among the three PAs analyzed, PA 26 exhibits the most significant increase in membrane permeability, alteration of membrane potential, and alteration of bacterial zeta potential. Surprisingly, PA 26 also has the highest MIC value, indicating the lowest antimicrobial activity. The MIC values are 32, 16, and 4 μg/mL for PA 26, PA 3, and PA 9, respectively. This fact might be resulting of different factors, for instance, the killing kinetics of each PA against this specific strain (killing rate differences at various time frames). Therefore, due to the different incubation times of the assays (MIC assays are evaluated after 18 hours of treatment and the PI/DiSC 3 (5) after 2 hours), the antimicrobial potency may not correlate well.

These observations indicated that each PA, as well as PMB, has a different acting profile with different Gram-negative strains. Based on the data above, having both a higher hydrophobicity/longer alkyl tail (LogP > 3 and 18 carbons) and surface charge (zeta potential > 39 mV) are essential to disrupt the membrane of various Gram-negative bacteria.

ONPG is a substrate of β-galactosidase (located in the cytoplasmatic space) and used as a probe to monitor the permeability of the inner/outer membrane of E.coli.^[46] If the permeability increases, ONPG can access the cytoplasmatic space and be cleaved by β-galactosidase, generating a fluorescent signal. For E. coli K12, the ONPG study was performed using various concentration of sample and monitored for 180 mins (Figure 5a-d). All the samples exhibited an increase in absorbance over time while there was no significant dose-dependent response from 0.5X to 2X MIC values. The order of membrane permeability was PMB > PA 26 > PA 3 > PA 9. Note that PA 26 (C₁₈K₂K_peptoidK₂) has the highest LogP value while the other two have similar values. Generally, the hydrophobic tail of PA inserts into the lipid bilayers and disrupt the inner membrane. Thus, the longer alkyl chain (stronger hydrophobicity) of PA 26 may explains its higher ability to promote membrane disruption.

Figure 5.

Antimicrobial mode of action study on E. coli K12 using PAs 3, 9 and 26. a-d) ONPG assays using 0.5, 1 and 2X MIC values of PA 3, PA 9, PA 26, and PMB, respectively. e-g) SEM images of bacteria incubated without (e) and with PA 9 at 2X MIC for 1 hour (f,g). Yellow arrows indicate bacterial membrane damage by PA9 (in “f” some bubbling is observed). All treatments were incubated for 1 hour at 37°C before performing the SEM assay imaging.

As the Gram-positive bacteria has a thicker outer membrane than Gram-negative bacteria, we performed SEM and TEM to study the membrane damage caused by PA 9. A clear increase on the roughness of bacteria surface can be observed from SEM (yellow arrows in Figure 6b). A mixture of dead bacteria (debris) and bacteria with integral membrane can be observed when the bacteria (10⁸ CFU/mL) were treated with 2X MIC value. The bacteria with integral membrane in this assay are due to the inoculum concentration used to obtain good quality SEM images, which is much higher than the standard inoculum (10⁵ CFU/mL) used for MIC determination assays. TEM (Figure 6c-d) shows various membrane alterations including completed disruption (dark blue), membrane detachment (bright blue), and total loss of membrane (light blue). Meanwhile, due to the membrane disruption, other alterations such as cytoplasmatic content leaking (green arrow), bubbling (pink arrow), and mesosomes (purple arrow) can be observed. It is worth noting that the peptoid hybrid PA 26 shows good membrane alteration (permeability and potential) to all tested strains, though only possesses moderate MIC values. This feature might make PA 26 a good nanocarrier adjuvant in antibacterial therapy.

Figure 6.

Antimicrobial mode of action study on MRSA JE2 using PA 9 at 2X MIC. a-b) SEM imaging for untreated and PA 9 treatment, respectively. Yellow arrows indicate membrane damage (bubbling and disruption). c-d) TEM images of bacteria incubated without and with PA 9, respectively. Arrows show membrane alterations such as: completed disruption (dark blue arrow), membrane detachment (bright blue arrow) and total membrane loss (light blue arrow). Also, other alterations like cytoplasmatic content leaking (green arrow), bubbling (pink arrow), and mesosomes (purple arrow) can be observed. All treatments were incubated for 1 hour at 37°C before performing the assays.

3. In Vitro Toxicity and In Vivo Activity

Next, we studied the cytotoxicity of selected PA nanostructures towards HEK-293T (human embryonic kidney) cells. As shown in Figure 7a-b, the tested PAs have an IC₅₀ ranging from 16.4 to higher than 256.0 μg/mL. The PAs modified with peptoid and odd number alkyl tail exhibit larger therapeutic index than the other PAs. Other have reported that peptoids possess lower cytotoxicity than their peptide counterparts against mammalian cells.^{[47, 48]}

Figure 7.

a-b) Cytotoxicity of HEK-293T cells and therapeutic index; c-d) In vivo studies using G. mellonella model. MRSA JE2 bacterial inoculum: 1.5 X 10⁸ CFU/mL. Larvae were injected with bacterial inoculum (10 μL) at day 0. Drugs (10 μL, one single dose) were injected 1h after the bacterial administration. PBS was used as negative control. VCM: vancomycin, used as antimicrobial control. UTD: no injection group (untreated). No infection group is also no injection group (for infection assay).

To further evaluate the acute toxicity and antimicrobial activity of our most potent PAs in vivo, we used a Galleria mellonella model. G. mellonella (also known as wax moth larvae) is used to study pathogenesis, antimicrobial activity and toxicity of drug candidates.^[49] This insect presents an innate, cellular and humoral, immune system that shares some similarities to vertebrate immune responses.^[50] The cellular immune response has phagocytic^[51], encapsulation^[52] and clotting^[50] capacities through cells called hemocytes (found in the hemolymph). Meanwhile, the humoral response is achieved by soluble effector molecules including complement-like proteins, enzymes, and antimicrobial peptides^[53] to manage infections.^[54]

Figure 7c presents the acute toxicity of the PA nanostructures and the results for the MRSA JE2 infection assays. Regarding toxicity, PA 9 and PA 26, at 50 mg/kg, exhibit a good safety profile (80% and 100% survival, respectively at day 4), comparable to the PBS-treated group (~ 80% survival). PA 3 showed higher toxicity at the same concentration (0% survival on day 4). Therefore, we selected PA 9 and PA 26 to assess the antimicrobial activity. Both PAs displayed high percentages of survival up to the last day analyzed, indicating a promising in vivo antimicrobial profile (Figure 7d) against MRSA JE2. PA 9 (50 mg/kg) was able to protect around 80% of the infected larvae from death (day 4). Further, PA 26 showed better antimicrobial effect than PA 9 (90% survival), comparable to vancomycin (VCM, 90% survival). All the subjects in the PBS-treated group died until day 3. Overall, the survival rates indicate that at 50 mg/kg PA 3 and PA 9 are non-toxic and provide similar antimicrobial protection as VCM.

Conclusion

In this study we designed and synthesized a set of PAs that self-assembled into supramolecular nanostructures (micelles and fibers) and displayed activity against Gram-positive and Gram-negative bacteria. Using the MIC obtained, zeta potential and LogP of each PA we could analyze and differentiate the data obtained against each bacterium by PCA. According to this analysis the charge (zeta potential) plays a major role for MRSA, while hydrophobicity (LogP) is more important for P. aeruginosa and E. coli K12 while for A. baumannii both LogP and zeta potential have similar importance. Furthermore, we studied the PAs effect on bacterial membrane permeability, depolarization, and overall charge. The in vitro toxicity was determined against HEK-293 cells and the therapeutic index calculated. Finally, an in vivo model using wax moth larvae showed good antimicrobial effect as measured by survival rates.

More studies are needed to correlate the membrane composition of each bacterium to the PAs activity, but our results are an important step on understanding how to design more active PAs. It is worth highlighting that different PAs possess promising TI against specific bacteria (i.e., PA 9 vs MRSA or PA 27 vs E. coli K12 or A. baumannii) while PA 26 (with a peptoid motif) possesses an intriguing ability to work as a nanocarrier.

Materials and methods

1. Materials

Solvents were purchased from Fisher Chemical. Fmoc-amino acids and coupling reagents were obtained from P3Bio (Louisville, KY). Fmoc-Rink Amide AM resin (0.61 mmol/g) was purchased from CreoSaulus (Louisville, KY, USA). Materials for bacteria culture were purchased from BD Difco^™, while materials for cell culture were purchased from Thermo Scientific (Waltham, MA, USA). A. baumannii and Enterococcus faecium VRE (patient isolated), were provided by Dr. Paul Fey’s Lab from the Department of Pathology and Microbiology at UNMC.

2. PA synthesis, purification, and self-assembly preparation

PAs were prepared using standard Fmoc solid phase peptide synthesis (SPPS). Building blocks such as amino acids were coupled to resin using HBTU/DIPEA in DCM/DMF solvent systems. Fmoc was deprotected using 20% piperidine and the final compound was cleaved from the resin using 95% TFA for 4 hours. The crude PA was purified using prep HPLC (0.1% TFA in acetonitrile and H₂O). Final PA product was confirmed by MALDI-TOF and analytical HPLC. The PA was lyophilized after purification and stored in the freezer. PA was dissolved in PBS at 1 mg/mL, annealed at 80 °C for 30 mins, and aged overnight to form nanostructures before testing. For biological assays, the PA was diluted with PBS or media to targeted concentrations before use.

3. PA characterization

Transmission Electron Microscope (TEM): Briefly, a given PA was loaded on the grid and stained with negative stain (NanoVan) and dried. The resulting PA thin film was visualized with an FEI Tecnai G2 Spirit transmission electron microscope (120 kV). Images were acquired using an AMT digital imaging system.^[55]

Dynamic light scattering (DLS): Zeta Potential was determined using a Zetasizer NanoZS (Malvern Instruments, Westborough, MA, USA) at 25 °C. At least 10 runs for each measurement, and three measurements were taken for each sample.

4. Principal Component Analysis (PCA)

PCA was performed using GraphPad Prism 10.0.2 software for Windows (GraphPad Software Inc., La Jolla, CA, USA). Eigenvalues were used as the method for principal component selection and outlier removal was performed for optimized analysis of the data.

5. Evaluation of antimicrobial activity of PAs

All bacterial cultures were made by the direct colony suspension method. The minimum inhibitory concentration (MIC) values were determined using the broth microdilution method according to the Clinical & Laboratory Standards Institute (CLSI) guidelines.^[56] The MIC determination assay was determined with a final bacteria density of 1.5$\times$10⁵ CFU/mL. Pure media was used as media control and PBS/media as untreated control. The plates were incubated for 18 h at 37°C. A 0.1% 2,3,5-Triphenyl-tetrazolium chloride (TTC) solution was added to the wells to allow better visualization (red color indicates bacterial growth). The MIC was defined as the lowest drug concentration that showed no visible bacterial growth. Absorbance readings were measured with a 470 nm filter using a multiscan FC microplate photometer (Thermo Fisher Scientific).

6. PA treated bacteria membrane studies

Inner Membrane Permeabilization Assay:

It was performed by measuring the l-galactosidase release in the media using o-nitrophenyl-β-d-galactopyranoside (ONPG) as substrate and an adapted method from Balakrishna et al.^[57] Polymyxin B (PMB), a potent membrane-active peptide, was used as positive control. The o-nitrophenol was quantified by measuring absorbance at 415 nm at various time intervals at 37 °C. Wells containing bacteria suspension, ONPG and PBS were used as negative control. The experiment was performed in triplicate.

PI uptake:

Bacteria were washed and resuspended in 5 mM HEPES and 20 mM glucose (pH 7.4) at 0.3 McFarland turbidity standard. A final concentration of 7.5 μg/mL propidium iodide (PI) was added to the samples and incubated for 2 hours at 37°C. Following, the fluorescence intensity was recorded at an excitation wavelength of 535 nm and an emission wavelength of 615 nm using a fluorescence plate reader (SpectraMax^®, Molecular devices).

DiSC 3 (5) membrane potential assay:

Bacterial suspension of 0.5 McFarland was prepared in HEPES buffer (5 mM HEPES and 20 mM glucose, pH 7.4). A final concentration of 1 μM of 3,3′ -dipropylthiadicarbocyanine iodide ([DiSC3(5)]) was added to the samples and incubated for 2 hours at 37°C. Following, the fluorescence intensity was recorded at an excitation wavelength of 622 nm and an emission wavelength of 670 nm using a fluorescence plate reader (SpectraMax^®, Molecular devices).

Zeta potential:

The MRSA JE2 and A. baumannii were cultured 18-20 hours before the assays. The bacteria were resuspended in PBS and adjusted to 5 X 10⁷ CFU/mL, and treated with samples (polymyxin B, daptomycin, PA alone, PA-LJC nanostructures) for 1 hour before the measurements.

SEM (Scanning Electron Microscopy):

20 mL of bacteria at 10⁸ CFU/mL were treated with 2 X MIC values of PAs for 1 hour. The resulting bacteria suspension was washed and fixed. Samples for SEM imaging were fixed by immersion in a solution of 2% glutaraldehyde, 2% paraformaldehyde in a 0.1M Sorenson’s phosphate buffer (pH 6.2) for a minimum of 24 h at 4°C. Samples were then washed three times with phosphate buffer to clear excess fixative. They were post-fixed in a 1% aqueous solution of osmium tetroxide for 30 minutes to aid in conductivity. Subsequently, samples were dehydrated in a graded ethanol series (50, 70, 90, 95, 100% x3). Following dehydration, samples were dried at the critical point and attached to aluminum SEM stubs with double-sided carbon tape. Silver paste was applied to increase conductivity. The following day, samples were coated with ~50nm Gold-Palladium alloy in a Hummer VI Sputter Coater (Anatech USA) and imaged at 30kV in a FEI Quanta 200 SEM operating in high vacuum mode.

TEM (Transmission Electron Microscopy):

Bacteria preparation was performed following the same procedure as SEM. Samples for TEM imaging were fixed by immersion in a solution of 2% glutaraldehyde, 2% paraformaldehyde in a 0.1M Sorenson’s phosphate buffer (pH 7.2) for a minimum of 24 h at 4°C. Samples were then washed three times with Phosphate Buffer to clear excess fixative. During processing, samples were post-fixed in a 1% aqueous solution of osmium tetroxide for 30 minutes. Subsequently, samples were dehydrated in a graded ethanol series (50, 70, 90, 95, 100%) and propylene oxide was used as a transition solvent between the ethanol and Embed 812 resin. Samples were allowed to sit overnight in a 50:50 propylene oxide: resin solution until all the propylene oxide had evaporated. Samples were then incubated in fresh resin for 2 hours at room temperature before final embedding. Polymerization took place at 65°C for 24 hours. Thin sections (90nm) made with Leica UC7 Ultracut ultramicrotome were placed on 200 mesh copper grids, post stained with 2% Uranyl Acetate followed by Reynolds Lead Citrate, and examined on a Tecnai G2 Spirit TWIN (FEI) operating at an accelerating voltage of 80kV. Images were acquired digitally with an AMT digital imaging system.

7. In vitro cytotoxicity study

Cytotoxicity was performed against Human Embryonic Kidney (HEK)-293t cell line donated by Dr. David Oupicky Lab (Center for Drug Delivery and Nanomedicine of UNMC). The cells were cultured in DMEM high glucose, with 10% fetal bovine serum and 1% penicillin/streptomycin. The assay was carried out in sterile 96-well flat-bottomed polystyrene microtiter plates. Plates contained 100 μL of cell suspension in each well (5000 cells/well) were preincubated for 24 h at 37 °C in a humidified environment with 5% CO₂. Samples to be tested were 2-fold diluted and added to test plates in triplicate to get final concentrations ranging from 4 to 256 μg/mL. The test plates containing the compounds and the control (cells in culture medium) were incubated for 24 h and further incubated again with 50 μL of XTT (sodium 3′-[1-(phenylaminocarbonyl)-3,4-tetrazolium]-bis (4-methoxy-6-nitro) benzene sulfonic acid hydrate) labeling mixture for 4 hours. The absorbance of the solution was determined at 450 nm using a multiwell plate reader AccuSkan, MultiSkan FC (Thermo Fisher Scientific) and the IC₅₀ values were calculated from the triplicates dose response curves obtained.

8. In vivo study

Galleria mellonella were purchased from Bestbait (https://www.bestbait.com) and maintained in wood chips in the dark. All in vivo assays were followed by the protocol published by Ignasiak and co-workers,^[58] and our previous report.^[12]