Redesigned Spider Peptide with Improved Antimicrobial and

Jul 25, 2017 - ... cells and elucidating its antimicrobial and anticancer mode of action. ... For a more comprehensive list of citations to this artic...
0 downloads 0 Views 2MB Size
Download PDF
Articles pubs.acs.org/acschemicalbiology

Redesigned Spider Peptide with Improved Antimicrobial and Anticancer Properties Sónia Troeira Henriques,* Nicole Lawrence, Stephanie Chaousis, Anjaneya S. Ravipati, Olivier Cheneval, Aurélie H. Benfield, Alysha G. Elliott, Angela Maria Kavanagh, Matthew A. Cooper, Lai Yue Chan, Yen-Hua Huang, and David J. Craik Institute for Molecular Bioscience, The University of Queensland, Brisbane, Queensland 4072, Australia S Supporting Information *

ABSTRACT: Gomesin, a disulfide-rich antimicrobial peptide produced by the Brazilian spider Acanthoscurria gomesiana, has been shown to be potent against Gram-negative bacteria and to possess selective anticancer properties against melanoma cells. In a recent study, a backbone cyclized analogue of gomesin was shown to be as active but more stable than its native form. In the current study, we were interested in improving the antimicrobial properties of the cyclic gomesin, understanding its selectivity toward melanoma cells and elucidating its antimicrobial and anticancer mode of action. Rationally designed analogues of cyclic gomesin were examined for their antimicrobial potency, selectivity toward cancer cells, membrane-binding affinity, and ability to disrupt cell and model membranes. We improved the activity of cyclic gomesin by ∼10-fold against tested Gram-negative and Gram-positive bacteria without increasing toxicity to human red blood cells. In addition, we showed that gomesin and its analogues are more toxic toward melanoma and leukemia cells than toward red blood cells and act by selectively targeting and disrupting cancer cell membranes. Preference toward some cancer types is likely dependent on their different cell membrane properties. Our findings highlight the potential of peptides as antimicrobial and anticancer leads and the importance of selectively targeting cancer cell membranes for drug development.

A

tally different in their membrane composition, architecture, and structure. Bacterial cell membranes are comprised of negatively charged phospholipids and a surface-exposed cell wall enriched with either anionic peptidoglycan, in the case of Gram-positive bacteria, or lipopolysaccharide (LPS), in the case of Gramnegative bacteria. In contrast, the surface of mammalian cells is composed mainly of neutral lipids, hence cationic AMPs preferentially target negative microbial surfaces over neutral mammalian cell surfaces.8 After targeting bacterial membranes, AMPs are thought to exert their activity by membrane destabilization.4 Several models of membrane destabilization have been proposed and include permeabilization through formation of toroidal pores, barrel-stave pores, or via disruption by carpeting.9 Membrane destabilization without disruption, such as formation of domains that affect proteins in the cell wall,10 has also been proposed. AMPs can alternatively cross the cell membrane without disrupting it to affect the function of an intracellular protein.11 Gomesin (Gm), an AMP produced in the hemocytes of the Brazilian spider Acanthoscurria gomesiana, is active against many diverse bacteria and fungi.12,13 It is highly positively charged

global increase in the number of bacteria resistant to multiple antibiotics and the lack of novel antimicrobials being developed is resulting in alarming outbreaks of difficult to treat bacterial infections, posing a great threat to human health.1,2 Thus, tackling antibiotic resistance and identifying novel antimicrobials has become a high priority for the World Health Organization (February 2017). In general, conventional antibiotics act by interfering with enzymes involved in nucleic acid metabolism, cell wall biosynthesis, or protein synthesis;3 thus, identifying antimicrobials that use alternative mechanisms would be helpful in addressing infections caused by multiresistant bacteria. Antimicrobial peptides (AMPs) are produced in all organisms as part of their host defense mechanisms. Owing to their ability to attack a wide range of microorganisms, AMPs have been regarded as endogenous antibiotics, and as potential leads for the design of novel antibiotics.4−6 Furthermore, the coadministration of AMPs with traditional antibiotics with distinct antibacterial mechanisms has the potential to work in a synergistic way to target multidrug resistant microbes.1 Many AMPs target the anionic cell membrane of microbes via positively charged and hydrophobic residues that are arranged in an amphipathic structure (see the Antimicrobial Peptide Database, http://aps.unmc.edu/AP/main.php).7 These properties facilitate their selective targeting and insertion into bacterial cells.4 Mammalian and microbial cells are fundamen© 2017 American Chemical Society

Received: June 1, 2017 Accepted: July 25, 2017 Published: July 25, 2017 2324

DOI: 10.1021/acschembio.7b00459 ACS Chem. Biol. 2017, 12, 2324−2334

Articles

ACS Chemical Biology

membranes, and toxicity to human blood cells were compared. In summary, we have created an analogue, [G1K,K8R]cGm, that is 10 times more active against tested bacteria than Gm or cGm without being more toxic toward red blood cells. In addition, we have demonstrated that cGm analogues are more toxic toward melanoma and chronic myeloid leukemia cells than against other tested cancer types. Selectivity toward certain cancer cells is likely mediated by differences in cell membrane properties.

and contains 18 amino acid residues, two disulfide bonds in a laddered connectivity, and two post-translational modifications (i.e., a Pyr at the N-terminus and amidation of the Cterminus).12 Gm adopts a well-defined β-hairpin-like structure,14 and the disulfide bonds have been reported to be important for the maintenance of this structure, along with resistance to serum proteases and antimicrobial activity.15,16 Recent studies suggest the antimicrobial activity of Gm is, at least in part, dependent on its ability to bind to and disrupt negatively charged membranes.17 For example, Mattei and colleagues compared the membrane-binding ability of a set of Gm analogues and showed a direct correlation between the overall hydrophobicity and the extent of membrane binding and permeabilization.18 In addition to antimicrobial activity, Gm has been demonstrated to have cytotoxic activity against melanoma cells, but not against other cancer cell lines. The ability to induce cell membrane permeabilization was suggested to be the mechanism by which Gm induced melanoma cell death.19 In a recent study, a head-to-tail cyclic analogue of Gm was synthesized (cyclic Gm, cGm), determined to have a β-hairpinlike structure (Figure 1A), to be active against bacteria and

■

RESULTS AND DISCUSSION Design of Novel Cyclic Gomesin Analogues. To improve the antimicrobial properties of cGm and examine its mode of action and selectivity for melanoma cells, a set of cGm analogues was designed (Figure 1B), and their properties were compared to identify structural features important for peptide activity. We were particularly interested in examining the importance of the cyclic backbone, disulfide bonds, and amphipathic surface (Figure 1A,B), and given the likelihood of these features being important for activity and/or stability, we designed analogues to examine each of these properties. Previous studies have demonstrated that replacing either Tyr with Trp and/or Lys with Arg improves membrane-binding affinity and antimicrobial activity of AMPs,21 without increased toxicity. The cGm analogues [Y7W]cGm, [Y14W]cGm, [K8R]cGm, and [Y7W,K8R,Y14W]cGm were designed on this basis (Figure 1B). To examine the effect of charge on activity and membrane binding, analogues with a decrease (+4) or an increase (+7) in their overall charges compared to cGm (+6) were designed: [R4A,R18A]cGm and [G1K,K8R]cGm, respectively. To evaluate whether introducing an extra aromatic residue would increase the overall hydrophobicity and the antimicrobial activity without increasing the toxicity, the analogue [L5W]cGm, in which Leu5 was replaced with Trp, was designed. Previous studies have shown that the motif D-Pro−LPro adopts a rigid turn structure, increasing the β-hairpin stability and the activity against tested yeast.22 To examine whether this strategy would improve the antimicrobial activity of cGm, we have designed the analogue [DPLP]cGm, in which the motif D-Pro−L-Pro was added in a turn. As selenium has been reported to have antimicrobial properties,23 [C/U]cGman analogue identical to cGm but with the four Cys replaced with Seleno-Cys (Sec)was also designed. The analogues [G1K,L5Y,K8R]cGm and [C/ U,G1K,L5Y,K8R]cGm differ in that they have Cys and Sec residues, respectively, and were designed to have an increased overall charge and potentially improved interaction with membranes. Specifically, Gly1 was replaced with a Lys, Lys8 with an Arg, and Leu5 with a Tyr. The roles of backbone cyclization and disulfide bonds on the activity of cGm were studied by comparing cGm with native Gm, and with the reduced form of native Gm (Gmred), respectively. All of the designed analogues were successfully synthesized, and their observed masses were identical to those calculated for fully oxidized peptides, confirming the formation of two disulfide bonds (Table S1). Retention times obtained by reversed-phase high-performance liquid chromatography (RPHPLC) were similar for all analogues (Table S1), and analysis of 1D NMR spectra revealed that all analogues had a welldispersed amide region, similar to that of cGm, suggesting that they had a native-like tertiary fold. The exception to this was Gmred, which showed broader resonance peaks as a result of not

Figure 1. Structure and sequence of cGm and analogues. (A) Threedimensional (3D) structure (BMRB ID: 17986; left panel), sequence (middle panel), and surface (right panel) representation of cGm. The 3D structure shows the two antiparallel β-sheets and the two parallel disulfide bonds (shown in yellow) formed between Cys2-Cys15 and Cys6-Cys11 with a ladder configuration. As visible in the surface representation, positively charged residues (shown in blue) and Tyr residues (shown in green) are located in the same region of the peptide. (B) Amino acid sequence of cyclic Gomesin (cGm) and analogues used in this study. Backbone cyclization is shown with the dashed line (all the peptides, but Gm and Gmred, are backbone cyclized). Disulfide bonds are shown in yellow. cGmred is the reduced native gomesin. Mutated residues of cGm analogues are in red. Z is Nterminal pyroglutamic acid. U is SelenoCys, and * is C-terminal amidation. Net charge of the peptide at pH 7.4 is shown.

melanoma cells, and more stable in serum than native Gm.20 In the study presented herein, we were interested in improving the antimicrobial activity of cGm and in further examining selectivity of cGm toward melanoma cells. A set of cGm analogues was synthesized (Figure 1B), and their antimicrobial activity, anticancer properties, ability to interact with model 2325

DOI: 10.1021/acschembio.7b00459 ACS Chem. Biol. 2017, 12, 2324−2334

Articles

ACS Chemical Biology

Figure 2. Structure and serum stability of cGm and its analogues. (A) NMR α-proton secondary chemical shifts (Δδ) of Gm, cGm and its three analogues. The sequence of cGm is shown, and the four Cys residues are indicated by gray bars. The location of the two β-sheets (three or more consecutive residues with Δδ > 0.1 ppm) are indicated by gray arrows. (B) Stability of Gm, cGm, and selected analogues in 25% (v/v) human serum for 24 h. Samples were analyzed by analytical HPLC and the percentage of peptide remaining calculated by comparing the respective peptide peak height in the chromatogram.

Figure 3. Hemolysis induced by cGm and its analogues. Peptides were incubated with 0.25% (v/v) human red blood cells for 1 h at 37 °C. (A) Dose response curves obtained with cGm and its analogues. For the sake of simplicity, only Gm, cGm, reduced Gm, [R4A,R18A]cGm, and [G1K,K8R]Gm are represented, but all the other analogues showed identical behavior and a concentration required to induce hemolysis in 50% of the RBCs (HC50) > 64 μM. Melittin, a peptide known to have hemolytic properties, was included as a positive control. Data points are mean ± SD of three replicates. (B) Percentage of hemolysis obtained with 64 μM of peptide.

Cyclic Gomesin Analogues Have Potent Antimicrobial Activity. The antimicrobial activities of the cGm analogues were tested against Escherichia coli ATCC 25922 and Staphylococcus aureus ATCC 25923, which were used as representatives of Gram-negative and Gram-positive bacteria, respectively (Table 1). All the peptides were more active against E. coli than S. aureus, with potency in the range of 0.5−8 μM for E. coli and 2−32 μM for S. aureus. cGm and both native and reduced Gm were similarly potent against tested bacteria, suggesting that the cyclized head-to-tail backbone and disulfide bonds are not required for the activity of the peptide. A comparison of cGm analogues revealed [R4A,R18A]cGm as the least active against E. coli and [G1K,K8R]cGm as the most potent against both E. coli and S. aureus (Table 1), which suggests that net charge is important for the antimicrobial properties of cGm. Replacement of Cys with Sec had only a weak effect on antimicrobial activity, as shown by a comparison of cGm and [C/U]cGm, and of [G1K,L5Y,K8R]cGm and [C/ U,G1K,L5Y,K8R]cGm. All other mutations, such as replacement of Tyr with Trp, Lys with Arg, and Leu with Trp or the introduction of D-Pro−L-Pro motif, resulted in minor improvements in antimicrobial potency compared to cGm. To examine whether cGm and its analogues are active against other bacteria, a selection of analogues were tested for their

having formed disulfide bonds. Moreover, analysis of the Hα proton secondary chemical shifts derived from 2D NMR H1− H1 NOESY and TOCSY of selected analogues (i.e., [R4A,R18A]cGm, [G1K,K8R]cGm and [C/U]cGm) were found to be similar to those of Gm and cGm (Figure 2A), confirming that the analogues possessed similar folds and 3D structures. Designed Cyclic Gomesin Analogues Have High Stability and Low Hemolytic Activity. The stability of selected analogues in 25% (v/v) human serum was compared to that of cGm. Specifically, [R4A,R18A]cGm, [G1K,K8R]cGm, [C/U]cGm, [L5W]cGm, [G1K,L5Y,K8R]cGm, and [C/ U,G1K,L5Y,K8R]cGm were tested. All peptides were stable for at least 24 h (Figure 2B), including analogues with a larger proportion of positively charged amino acid residues, demonstrating that the introduced modifications did not decrease stability. Thus, all the cGm analogues were subsequently compared for their activity. Hemolytic properties of the analogues were tested against human red blood cells (hRBCs). The results indicate that cGm and its analogues have similar hemolytic properties and induce 20−50% of hemolysis at 64 μM when tested with 0.25% (v/v) of hRBCs (Figure 3). 2326

DOI: 10.1021/acschembio.7b00459 ACS Chem. Biol. 2017, 12, 2324−2334

Articles

ACS Chemical Biology Table 1. Activity of Gm Analogues against Bacteria and Fungi MIC (μM)

peptide

E. coli ATCC 25922b

S. aureus ATCC 25923b

Gm Gmred cGm [Y7W]cGm [Y14W]cGm [K8R]cGm [Y7W,K8R,Y14W]cGm [R4A,R18A]cGm [G1K,K8R]cGm [C/U]cGm [L5W]cGm [DPLP]cGm [G1K,L5Y,K8R]cGm [C/U,G1K,L5Y,K8R]cGm

4 4 4 1−2 1−2 1−2 1 8 0.5−1 4−8 2−4 4 0.5−1 1−2

32 32 32 16 16 16 16 32 2 16 8−16 16 8 8

K. pneumoniae ATCC 700603b

A. baumannii ATCC 19606b

P. aeruginosa ATCC 27853b

H. pylori ATCC 43504c

C. albicans ATCC 90028d

C. neoformans ATCC 208821d

32

4−8

8

>80

8−16

0.5−1

0.06

4−8

1−2

1−4

4−8

0.5−1

0.5−1 80

MIC (minimum inhibitory concentration) is the lowest peptide concentration in μM showing no growth. When a concentration range is shown, it reflects variations between replicates. bAntibacterial activity was determined in triplicate with 5 × 105 cells/mL using a microbroth dilution method in Mueller Hinton Broth (MHB). cActivity against H. pylori was examined by colony growth in 24-well agar plates inoculated with 104 cells per well and incubated with peptide for 48 h. Melittin was included as a control and showed no activity up to 80 μM. dAntifungal activity was determined in duplicate with 5 × 103 cells/mL in Yeast Nitrogen Base Broth. MIC is the lowest concentration with ≥85% growth inhibition after incubation with peptide for 18 h, as determined by OD600 for C. neoformans and OD530 for C. albicans. a

Figure 4. Binding to and permeabilization of bacteria induced by cGm analogues. (A) The binding affinity of [R4A,R18A]cGm and [G1K,K8R]cGm to LTA and LPS was examined by LAL assay. The peptide concentration required to neutralize 50% of LTA or LPS (KD) was calculated by nonlinear fit of the dose responses. Data points are mean of three replicates and error bars represent standard errors. (B) Permeabilization of bacterial cells induced by cGm analogues. E. coli ATCC 29522 or S. aureus ATCC 29523 (107 cells/mL in PBS) were incubated with [R4A,R18A]cGm or [G1K,K8R]cGm for 1 h at 37 °C. Cells with their membrane permeabilized were identified by screening of fluorescent cells upon the addition of SYTOX green by flow cytometry (excitation at 488 nm and emission at 530/30 nm). Data are mean ± SEM of three independent experiments. PC50 is the peptide concentration in μM required to induce permeabilization of 50% of the bacteria cells.

the least active; thus, these analogues were included in subsequent studies, to represent the most active and the least active cGm analogues, respectively. Overall, the antimicrobial results demonstrate that it is possible to design a cGm analogue with improved antimicrobial potency without increasing hemolytic potency. Cyclic Gomesin and Analogues Target and Disrupt Microbial Membranes. Gram-negative bacteria have an outer membrane with LPS in the outer leaflet, whereas Gram-positive bacteria have a thick peptidoglycan layer covered with lipoteichoic acid (LTA).24 Both LPS and LTA are negatively charged molecules that are surface exposed and present an initial barrier for AMPs when encountering bacteria. Thus, to

potency against a panel of bacteria and fungi (Table 1), including the Gram-negative bacteria Helicobacter pylori, Klebsiella pneumoniae, Acinetobacter baumannii, and Pseudomonas aeruginosa and the yeast Candida albicans and Cryptococcus neoformans. All peptides were active against all bacteria and yeast, except H. pylori, revealing a broad spectrum of activity for cGm and its analogues. None of the peptides, including the control melittin, showed activity against H. pylori, suggesting this bacterium has additional resistance to the peptides compared to the other bacteria. For all other bacteria and fungi tested, the peptide potency followed the same general trend as that observed for E. coli and S. aureus, with [G1K,K8R]cGm being the most active and [R4A,R18A]cGm 2327

DOI: 10.1021/acschembio.7b00459 ACS Chem. Biol. 2017, 12, 2324−2334

Articles

ACS Chemical Biology

Figure 5. Binding to and leakage of model membranes. Panels A, B, and C show binding of cGm and its analogues to model membranes studied by surface plasmon resonance. Peptide samples were injected for 180 s (association phase) and followed for 600 s (dissociation phase) over a particular lipid bilayer deposited onto an L1 chip surface. The left panel shows sensorgrams obtained upon injection of 32 μM of peptide over a specific lipid bilayer. The right panels show the dose−response binding obtained with the respective peptide injection at the end of the association phase (t = 170 s). Signal of sensorgrams and dose−response curves were normalized to peptide-to-lipid ratios (P/L mol/mol) by converting response units (RU) into moles of peptide and normalized for the amount of lipid deposited onto the lipid surface (1 RU = 1 pg/mm2 of peptide or lipid). (A) Binding of cGm to various lipid systems. (B) Comparison of [R4A,R18A]cGm and [G1K,K8R]cGm with cGm in their binding to POPC and POPC/POPG (80:20). (C) Effect of Cys residues versus Sec residues in their binding to model membranes: cGm compared with [C/U]cGm and [G1K,L5Y,K8R]cGm with [C/U,G1K,L5Y,K8R]cGm, in terms of their binding to POPC and POPC/POPG (80:20). (D) Membrane leakage induced by cGm and its analogues. Vesicles composed of POPC or POPC/POPG (80:20), encapsulating carboxyfluorescein (CF), were incubated at 10 μM lipid with 2-fold dilutions of peptides (up to 5 μM) for 20 min. The percentage of permeabilized vesicles was calculated by measuring CF fluorescence emission (λexcitation = 485 nm, λemission = 520 nm). The percentage of vesicles permeabilized was plotted as a function of P/L, mol/mol in solution, and the curves fitted with sigmoidal dose−response binding. L50 is the P/L required to induce leakage in 50% of the vesicles. Data are mean ± SD of three replicates.

coli CGSC 5167a mutant with a shorter LPS, thus easier to penetrate and more susceptible to AMPs that act on the microbial surface.21 These experiments indicate that all the tested peptides were 10- to 100-fold more active against the strain with shorter LPS (CGSC 5167) than the representative E. coli (ATCC 25922, Table 1), suggesting that they act by disrupting the microbial cell membrane after penetrating the outer protective layer. To further confirm that the antibacterial mode of action of cGm, and its analogues, is dependent on the ability to target and disrupt bacterial cell membranes, the percentage of bacteria cells that become permeabilized upon treatment (at 37 °C for 1

examine whether the antimicrobial mode of action of cGm includes the ability to target the microbial cell surface, we compared the ability of the analogues [R4A,R18A]cGm and [G1K,K8R]cGm (the peptides with the lowest and highest antimicrobial efficiency, respectively) to bind to LPS and LTA using a Limulus Amebocyte Lysate (LAL) assay. This assay detects the presence of non-neutralized LPS or LTA.21 Figure 4A shows that both peptides neutralize LPS and LTA, but that [G1K,K8R]cGm is more efficient. This suggests that cGm analogues are effective in targeting bacterial surfaces. The ability to target the microbial cell membrane was further examined by testing the activity of selected analogues against E. 2328

DOI: 10.1021/acschembio.7b00459 ACS Chem. Biol. 2017, 12, 2324−2334

Articles

ACS Chemical Biology

A set of cGm analogues, namely [R4A,R18A]cGm, [G1K,K8R]cGm, [C/U]cGm, [G1K,L5Y,K8R]cGm, and [C/ U,G1K,L5Y,K8R]cGm, was tested against POPC and POPC/ POPG (80:20), and all had weak affinity for neutral membranes and improved affinity for membranes containing the negatively charged POPG (Figure 5B,C) under physiological conditions (pH and salt concentration). Despite the fact that [R4A,R18A]cGm and [G1K,K8R]cGm differ in their overall charge, both were observed to have a high affinity for POPC/POPG (80:20; Figure 5B) and POPC/POPG (50:50) model membranes (data not shown). Dose−response curves and sensorgrams obtained with 32 μM [G1K,K8R]cGm indicate that the peptide has a higher affinity at lower concentrations and slower dissociation rate (kd) from the membrane than [R4A,R18A]cGm and cGm (see Figure 5B and Table S2). Interestingly, the presence of Sec instead of Cys decreased the affinity for negatively charged membranes, as shown by a comparison of cGm and [C/U]cGm and of [G1K,L5Y,K8R]cGm and [C/U,G1K,L5Y,K8R]cGm (Figure 5C and Table S2). To examine whether the peptides could disrupt lipid model membranes, large unilamellar vesicles (LUVs) encapsulating carboxyfluorescein (CF) were incubated with cGm or its analogues. LUVs form stable membranes without significant curvature and are therefore a good model to mimic the properties of the lipid bilayer in cell membranes.25 Dose− response curves (Figure 5D) show that although none of the peptides induced dye leakage from neutral POPC vesicles, they disrupted negatively charged POPC/POPG (80:20) vesicles in a dose-dependent manner, with an efficiency that follows the trend: [G1K,K8R]cGm > cGm > [C/U]cGm > [R4A,R18A]cGm. The ability to bind to and disrupt negatively charged, but not neutral, membranes, is consistent with the cGm peptides being toxic to bacteria, which have negatively charged surfaces, and nontoxic to RBCs, with neutral membrane surfaces. Furthermore, a comparison of cGm analogues in terms of their efficiency to disrupt negatively charged model membranes follows the same trend as their antimicrobial efficiency (see Figure 5D and Table 1). Overall, these results suggest cGm and its analogues exert their antimicrobial activity by targeting and disrupting the negatively charged microbial membrane. A similar resistance to physiological salt concentration, preference for negatively charged membranes, and an antimicrobial mode-of-action dependent on membrane-disruption have also been proposed for theta-defensins.26,27 These peptides also possess a cyclic backbone, a positive overall charge (e.g., BTD-2 has +6), and a structure comprising two antiparallel β-sheets stabilized by three parallel disulfide bonds.26 Together, innate host defense peptides including theta-defensins, from primate leukocytes, and beta-haripin peptides produced by invertebrates and mammals (e.g., Gm from spider hemocytes (A. gomesiana), protegrin-1 from porcine neutrophils, and tachyplesin-1 from horseshoe crab hemocytes (Tachypleus tridentatus))13 have a similar antimicrobial mode of action that involves targeting and disrupting anionic microbial membranes. Cyclic Gm and Its Analogues Are Toxic toward Melanoma and Leukemia Cells. Gm and cGm have been shown to be active against melanoma cells;19,20 however, selectivity toward this type of cancer is not fully understood. To gain insights into the anticancer mechanism, and confirm the selectivity, Gm, cGm, and analogues were tested against a panel of cancer and noncancer cell lines (Table 2). All peptides were

h in phosphate buffered saline (PBS)) with increasing concentrations of peptide was determined using flow cytometry and SYTOX green. This dye can only penetrate bacterial cells with compromised membranes to become fluorescent once bound to nucleic acids. Thus, the proportion of fluorescent cells correlates with the proportion of cells with permeabilized membranes. [R4A,R18A]cGm and [G1K,K8R]cGm were compared in terms of their ability to disrupt the membrane of E. coli and S. aureus (Figure 4B) and shown to permeabilize bacterial cells under physiological conditions (pH and salt concentration). [G1K,K8R]cGm permeabilized both E. coli and S. aureus with greater efficiency than [R4A,R18A]cGm, and both peptides disrupted the cell membrane of E. coli with greater efficiency than that of S. aureus. The trend in the concentration required to permeabilize 50% of the bacterial cells (PC50, see values in Figure 4B) correlated with the trend for antimicrobial activity. Altogether, these results suggest the antimicrobial mode of action is dependent on the ability to bind to and permeabilize bacterial cell membranes, and that differences in killing efficiency across the tested bacteria and fungi are likely to be related to specific properties of the microbial membranes. Cyclic Gomesin and Its Analogues Selectively Bind to and Disrupt Negatively Charged Lipid Bilayers. cGm shares key properties with peptides that can bind to lipid membranes, namely an amphipathic structure with surfaceexposed positively charged and hydrophobic residues on the same side of the molecule (see Figure 1A). To examine whether the ability to bind to lipid membranes plays a role in the antimicrobial activity of cGm, the affinity of cGm and its analogues to bind to membranes composed of different lipid compositions was measured using surface plasmon resonance (SPR). In particular, binding to deposited lipid bilayers composed of pure 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC; a zwitterionic phospholipid that forms neutral, fluid lipid bilayers at 25 °C) was compared with binding to membranes comprised of a mixture of POPC, sphingomyelin (SM), and cholesterol (Chol) (POPC/SM/Chol (27:33:40) molar ratio), which forms bilayers with liquid-ordered properties, and to those comprised of a mixture of POPC and the zwitterionic 1-palmitoyl-2-oleoyl-sn-glycero-phospho-L-ethanolamine (POPE; POPC/POPE (80:20) molar ratio). Dose− response curves and sensorgrams (Figure 5A and Table S2) revealed a weak affinity with rapid dissociation from all the zwitterionic lipid systems tested (i.e., POPC; POPC/POPE and POPC/Chol/SM). The potential effect of negatively charged phospholipids, akin to those exposed on the outer leaflet of bacteria and cancer cell membranes, was examined by testing the membrane-binding affinity to mixtures of POPC with the negatively charged 1palmitoyl-2-oleoyl-sn-glycero-3-phospho-L-glycerol (POPG; POPC/POPG at 80:20 and 50:50 molar ratios) or with 1palmitoyl-2-oleoyl-sn-glycero-3-phospho- L -serine (POPS; POPC/POPS (80:20) molar ratio). cGm had a high affinity for membranes with negatively charged lipids, which was further improved by increasing the percentage of POPG (see POPC vs POPC/POPG (80:20) vs POPC/POPG (50:50), Figure 5A). Additionally, the peptide had a greater affinity for POPS than for POPGas demonstrated by a comparison of binding to POPC/POPG (80:20) and POPC/POPS (80:20; see Figure 5A and Table S2). Sensorgrams show a fast association and a slow dissociation from membranes containing negatively charged phospholipids. 2329

DOI: 10.1021/acschembio.7b00459 ACS Chem. Biol. 2017, 12, 2324−2334

Articles

ACS Chemical Biology

Table 2. Cytotoxicity of Gomesin and It Analogues against Cultured Cell Lines and Peripheral Blood Mononuclear Cellsa CC50 (μM) peptide Gm Gmred cGm [Y7W]cGm [Y14W]cGm [K8R]cGm [Y7W,K8R,Y14W]cGm [R4A,R18A]cGm [G1K,K8R]cGm [C/U]cGm [L5W]cGm [DPLP]cGm [G1K,L5Y,K8R]cGm [C/U,G1K,L5Y,K8R]cGm

CRL-1739 67.0 >64 39.8 23.3 30.3 29.4 31.6 >64 19.5 51.0 18.3 22.5 26.5 46.2

± 5.0 ± ± ± ± ±

1.0 0.4 1.1 1.1 1.3

± ± ± ± ± ±

0.5 4.6 1.4 3.3 2.4 7.6

MM96L 3.7 8.6 2.9 5.1 4.0 5.0 5.4 10.3 1.7 4.6 4.0 3.0 2.0 2.3

± ± ± ± ± ± ± ± ± ± ± ± ± ±

0.2 0.8 0.1 0.3 0.1 0.3 0.3 1.1 0.1 0.2 0.1 0.1 0.1 0.2

HFF-1 49.9 58.2 71.7 39.7 68.4 34.5 31.3 51.2 9.0 15.6 25.0 14.9 14.7 30.8

± ± ± ± ± ± ± ± ± ± ± ± ± ±

HeLa

16.6 3.0 9.2 3.8 9.6 3.6 2.6 4.0 0.5 3.6 2.9 0.8 1.5 2.0

54.1 >64 >64 >64 50.4 39.4 41.0 >64 44.2 >64 17.1 51.5 20.9 36.2

MCF-7

± 5.0 15.3 ± 1.4 ± 2.4 ± 2.6 ± 4.2 ± 2.2 ± ± ± ±

0.7 3.9 1.4 2.8

15.1 48.4 6.7 37.5

± ± ± ±

1.4 0.7 0.7 3.3

15.2 ± 1.9 29.0 ± 3.7

K-562 3.8 2.7 2.7 3.9 2.7 3.1 3.9 11.5 2.1 1.4 1.0 3.9 1.3 6.4

± ± ± ± ± ± ± ± ± ± ± ± ± ±

0.3 0.2 0.1 0.2 0.1 0.1 0.1 0.6 0.2 0.2 0.1 0.2 0.1 0.6

HL-60

PBMCs 13.4 ± 1.8

12.7 ± 1.1

10.8 ± 0.8

34.1 ± 3.9 9.0 ± 1.1 38.5 ± 4.8

30.8 ± 2.5 5.1 ± 0.3 15.5 ± 0.8

15.7 ± 1.1 33.3 ± 7.4

14.1 ± 0.9 10.4 ± 1.0 9.5 ± 0.5

Cytotoxic concentration required to kill 50% of the cells (CC50) and respective SD is shown in μM and was obtained by fitting dose−response curves with a sigmoidal curve. Cells were incubated with peptides for 18−24 h and cell death detected by resazurin. Each experiment was conducted at least three times. Noncancerous cells are in bold font. Tested cells are CRL-1739, human gastric adenocarcinoma cells; MM96L, human skin melanoma cells; HFF-1, human foreskin fibroblast normal cells; HeLa, human cervical adenocarcinoma cells; MCF-7, human breast adenocarcinoma cells; K-562, human bone marrow chronic myelogenous leukemia cells; HL-60, human peripheral blood acute promyelocytic leukemia; and PBMCs, peripheral blood mononuclear cells. a

Figure 6. Anticancer mode of action of cGm and its analogues. (A) Permeabilization of K-562 and HL-60 induced by cGm analogues was followed by incubation of cells in suspension (5 × 104 cells/100 μL in PBS). Cells were treated with [R4A,R18A]cGm or [G1K,K8R]cGm at different concentrations for 1 h at 37 °C. Cells with their membranes permeabilized were identified by the addition of SYTOX green and screening by flow cytometry (excitation at 488 nm and emission at 530/30 nm). Data are mean ± SD of three independent experiments. PC50 is the peptide concentration in micromolarity required to induce permeabilization of 50% of the cells. (B) Toxicity of cGm and its analogues on K-562 and HL-60 cells (5 × 103 cells/100 μL in medium), either treated or untreated with 30 mM of sodium chlorate for 24 h before incubating with 2-fold dilutions of cGm and a selection of analogues. The percentage of cell death was determined using a resazurin assay in three independent experiments. The concentration required to induce 50% cell death (CC50) ± SD was determined by fitting the dose−response curve with a sigmoidal curve. Statistical analysis was performed using a two-way ANOVA with Tukey’s multiple comparisons test by comparison of CC50 obtained against untreated K-562 with all the other conditions, for each peptide (n.s.: p ≥ 0.05; *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001).

active against the epithelial cancer cell line MM96L and the chronic myeloid leukemia (CML) K-562 cell line and less toxic to the gastric cancer cell line CRL-1739, the breast cancer cell line MCF-7, the cervical cancer cell line HeLa, or the noncancerous epithelial cell line HFF-1. To examine whether these peptides kill other white blood cells, along with the K-562 leukemia cells, a selection of cGm analogues was tested against HL-60, a model cell line of acute promyelocytic leukemia, and human primary peripheral blood mononuclear cells (PBMCs). The cGm peptides were less toxic to HL-60 and PBMCs compared to K-562 (Table 2). The toxicity of cGm analogues toward mammalian cells follows the

same trend as that observed for antimicrobial activity, with the [G1K,K8R]cGm being the most active and [R4A,R18A]cGm the least active, suggesting that overall charge is important for the toxicity of cGm against cancer and healthy cells, and via a mechanism similar to that used to kill bacteria. However, this does not explain the differences in toxicity between the different cell lines. Cytotoxicity of Cyclic Gomesin Peptides Correlates with Their Ability to Disrupt Cell Membranes in Cancer Cells. To examine whether cGm and its analogues also kill mammalian cells by binding and permeabilizing cell membranes, the ability of [G1K,K8R]cGm and [R4A,R18A]cGm to 2330

DOI: 10.1021/acschembio.7b00459 ACS Chem. Biol. 2017, 12, 2324−2334

Articles

ACS Chemical Biology

heparan sulfate and chondroitin sulfate glycan chains, for 24 h prior to adding peptide.33,35 The toxicity of a selection of cGm analogues (Figure 6B) shows identical CC50’s against sodium chlorate-treated and untreated K-562 cells, in agreement with these cells being heparan sulfate deficient. On the other hand, HL-60 became more susceptible to cGm following treatment with sodium chlorate, and the CC50 for sodium chlorate treated HL-60 was similar to that of untreated K-562 cells (see Figure 6B). These results suggest that the absence of HSPG on the surface of K-562 cells makes them more susceptible to cGm. The results presented herein suggest that toxicity of cGm and its analogues toward mammalian cells is due to their ability to disrupt the cell membrane and that the peptide’s overall positive charge and the ability to selectively target more negatively charged cancer cell membranes are important in doing so. The results also support the hypothesis that cell membrane composition varies across different cancer cell types; therefore, each cancer should be considered individually for generating therapeutic leads. Additional cell membrane properties are likely to be important for specific anticancer activity of peptides; for instance, the lack of HPSG in the K-562 cell line makes it more susceptible to cGm and its analogues than HL60. Conclusion. In summary, we created analogues with higher antimicrobial activity than cGm without increasing hemolytic properties. In particular, the analogue [G1K,K8R]cGm has a 10-fold higher potency against both Gram-positive and Gramnegative bacteria than native Gm or cGm. We demonstrated that the antimicrobial mode of action is dependent on the ability to bind and disrupt the bacterial membrane. In addition, we demonstrated that cGm and its analogues are also toxic to cancerous mammalian cell lines in a manner that is cell-type-dependent, with the greatest toxicity observed toward melanoma and chronic myeloid leukemia cells. The mode of action seems to be dependent on the ability to target and disrupt cell membranes, while specific membrane properties of the cancer cell type dictate the level of cytotoxicity. Overall, we have demonstrated that Gm, originally identified as an AMP, can be redesigned to improve antimicrobial activity, without increasing hemolytic properties, and additionally be more toxic toward some types of cancer cells.

permeabilize K-562 and HL-60 cells (Figure 6A) was tested using SYTOX green and flow cytometry, as described above for bacteria cells. Both peptides were observed to permeabilize K562 and HL-60 cells, and the peptide concentration required to permeabilize 50% of the cells correlated with the concentration required to induce toxicity, with [G1K,K8R]cGm being more efficient than [R4A,R18A]cGm. This correlation suggests the peptides kill these leukemia cell lines via a mechanism dependent upon cell membrane permeabilization. The ability of cGm and analogues to kill cells by membrane permeabilization was also confirmed for melanoma and cervical cancer cells (i.e., MM96L, HeLa), as shown following cell membrane permeability to calcium ions in a fluorescence-based assay using a Fluorescence Imaging Plate Reader (FLIPRTETRA) and the FLIPR calcium 4 assay kit. HeLa and MM96L cells do not express voltage-gated calcium channels (CaV), thus calcium influx in these cells is independent of CaV activity and, instead, related to membrane permeabilization. By contrast, neuroblastoma cells SH-SY5Y show CaV dependent Ca2+ influx due to endogenously expressed CaV28 and were included in this study as a control (for more details on experimental design and methods, see Supporting Information and Figure S1). Briefly, the concentration required to detect calcium influx is peptidedependent, with [G1K,K8R]cGm being the most potent and [R4A,R18A]cGm the least potent (see Figures S1B,C). Furthermore, the concentrations at which influx is detected correlated with concentrations that induced cell death. This calcium influx response is likely to result from cell membrane permeabilization, which eventually leads to cell death. Cell Membrane Differences Drive Selectivity of Cyclic Gomesin for Certain Cancer Types. The surface of cancer cells is more negatively charged than healthy cells, which provides an explanation as to why some cationic AMPs are reported to have anticancer properties.29 In particular, membranes of healthy cells are asymmetric in their lipid distribution, and negatively charged phosphatidylserine (PS)− phospholipids are restricted to the inner leaflet, whereas membranes in cancer cells lose their asymmetry and the anionic PS−phospholipids become surface exposed.30 Toxicity of cGm and its analogues toward cancer cells is in agreement with their higher affinity for PS-containing lipid membranes, compared to neutral membranes (see Figure 5A). Nevertheless, that does not explain the higher toxicity of cGm and its analogues toward MM96L and K-562 cells, compared to other cancer cell lines. It is worth mentioning that in blood there is always a portion (∼15−25%) of PBMCs in programmed cell death that present negatively charged PS−phospholipids at the cell surface as detected by studies conducted with annexin V.31,32 This would explain why PBMCs are more susceptible to positively charged cGm and analogues used in the current study than RBCs. Heparan sulfate proteoglycans (HSPGs) are negatively charged glycoproteins found at the cell surface and extracellular matrix and comprise a core protein to which heparan sulfate glycosaminoglycan chains are covalently attached. HSPGs have been proposed to decrease the toxicity of peptides by sequestering the lytic peptides from the phospholipid bilayer, thus impeding their ability to lyse the cell membrane.33 K-562 cells are deficient in heparan sulfate,34 whereas HL-60 cells are not. As both cell lines are leukemia cells, we have examined whether the increased toxicity of cGM analogues toward K-562 cells, compared with HL-60, correlates with a lack of cell surface HSPG in K-562 cells. Thus, cells were pretreated with sodium chlorate, to prevent sulfonation of

■

MATERIALS AND METHODS

Peptide Synthesis, Folding, and Purification. Gm was synthesized using Fmoc solid-phase chemistry and folded as previously described.20 Gmred was obtained as Gm, except the folding step was excluded and 5 mM tris(2-carboxyethyl)phosphine hydrochloride was added. cGm mutants were synthesized, cyclized, and folded using Fmoc solid-phase chemistry, as per the methodology developed by Cheneval et al.36 Peptides with Sec were prepared with the FmocSec(Mob)−OH (ChemPep Inc.) residue. The mass of the folded/ reduced peptides was confirmed by ESI-MS. Folding of the synthesized peptides was confirmed using 1H NMR spectroscopy on a Bruker Avance 600 MHz spectrometer, as previously described.20 The spectra were compared with that obtained with native Gm. All peptides used were ≥95% pure, as confirmed by analytical RP-HPLC. Peptide sample concentration was determined by absorbance at 280 nm (see Table S1). Serum Stability Assay. Peptide stability in 25% (v/v) human serum (Sigma-Aldrich, human male AB plasma) was examined, as previously described with the native Gm and cGm.20 Antimicrobial Assay. The antimicrobial activity of Gm and its analogues against E. coli ATCC 25922, S. aureus ATCC 25923, E. coli CGSC 5167, K. pneumoniae ATCC 700603, A. baumannii ATCC 19606, and P. aeruginosa ATCC27853 was evaluated by bacterial 2331

DOI: 10.1021/acschembio.7b00459 ACS Chem. Biol. 2017, 12, 2324−2334

Articles

ACS Chemical Biology

(supplemented with 1% (v/v) penicillin/streptomycin and 10% (v/ v) FBS (HeLa and MCF-7) or 15% (v/v) FBS (HFF-1)). CRL-1739 cells were grown in DMEM/F12 medium (supplemented with 2 mM L-glutamine 1% penicillin/streptomycin and 10% (v/v) FBS). Cells were grown in T175 flasks and incubated in a humidified atmosphere (5% CO2, 37 °C) and split by dilution after reaching confluence. Cell Viability Studies. Toxicity of Gm and its analogues against cell lines was tested using resazurin dye, as previously described.21 Adherent cells (i.e., HeLa, MM96L, HFF-1, and AGS) were seeded in 96-well plates (2.5 × 103 cells/well) the day before the assay. Suspension cells (K-562, HL-60 and PBMCs), with or without pretreatment with 30 mM of sodium chlorate for 24 h, were dispensed onto 96-well plates (5 × 103 cells/well) on the day of the assay. Suspension Cells Permeabilization Followed Using SYTOX Green and Flow Cytometry. K-562 and HL-60 were washed and diluted in PBS (5 × 105 cells/mL). Cell suspensions (100 μL) were incubated with 2-fold dilutions of peptide samples (64−0.125 μM) for 1 h at 37 °C with continuous shaking. SYTOX green (Invitrogen; 2 μM) was added to each sample before reading in the flow cytometer (BD FACSCanto II cell analyzer). Measurements and percentages of permeabilized cells were determined as previously described.41 Adherent Cells Permeabilization Followed by Calcium Influx Using Fluorescence Plate Reader. The ability of cGm and its analogues to induce permeability to calcium ions was conducted using a Fluorescence Imaging Plate Reader (FLIPRTETRA) and the FLIPR calcium 4 assay kit using HeLa, MM96L, and SH-SY5Y cells. HeLa and MM96L cells were seeded at 16 000 cells/mL and SH-SY5Y were seeded at 50 000 cells/mL in a 96-well plate 24 h before the assay. Cells were washed and 100 μL of physiological salt solution (PSS; 140 mM NaCl, 11.5 mM glucose, 5.9 mM KCl, 1.4 mM MgCl2, 1.2 mM NaH2PO4, 5 mM NaHCO3, 1.8 mM CaCl2, 10 mM HEPES, pH 7.4) and pretreated with 50 μL of calcium4 dye for 30 min. The peptides were solubilized in PSS and added at different concentrations to the cells and the fluorescence signal (excitation 470−495 nm and emission 515−575 nm) followed over time. PSS was used as the blank and the detergent Triton X-100 (TX-100) as a membrane-permeabilizing control.

growth inhibition using a microtiter broth dilution method, as previously described.13,21 Activity against H. pylori ATCC 43504 was examined by colony growth in agar plates in an anaerobiose box with microaerophilic atmosphere generator sachets (Biomerieux Australia Pty Ltd.) at 37 °C. Peptides with a final concentration of 80, 20, 5, or 1.25 μM were mixed with bacterial suspension to 5 × 105 cells/mL and 20 μL of the mixture spread onto previously prepared ultralow attachment 24-well plates containing selective H. pylori agar in each well prepared in-house (Columbia blood agar base containing 5% (v/v) lysed horse blood and HP Dent Supplement (SR0147, Oxoid, Thermo Scientific)). The bacteria inoculum was 104 cells/well. Agar plates were incubated for 48 h at 37 °C in a microaerophilic atmosphere. The MIC was defined as the lowest concentration showing no visible colony growth or