Antimicrobial peptides are promising antibiotics as they possess strong antimicrobial activity and very broad spectra of activity. However, administration of an antibiotic with a very broad spectrum of activity disrupts normal microflora and increases the risks of other fatal infections. To solve the problem, we designed a novel antimicrobial peptide that is activated by virulent proteases of pathogenic organisms. We constructed a peptide composed of three domains, namely an antimicrobial peptide (lactoferricin) as the active center, a protective peptide (magainin intervening sequence) that suppresses antimicrobial activity, and a specific linker that joins these two components and is efficiently cleaved by virulent proteases. We utilized Candida albicans as a model organism that produces secreted aspartic proteases as a virulence attribute. We screened for a peptide sequence efficiently cleaved by secreted aspartic proteases isozymes and identified a GFIKAFPK peptide as the most favorable substrate. Subsequently, we chemically synthesized a peptide containing the GFIKAFPK sequence. The designed peptide possessed no antimicrobial activity until it was activated by secreted aspartic proteases isozymes. Furthermore, it demonstrated selective antimicrobial activity against C. albicans, but not against Saccharomyces cerevisiae. A designed peptide like the one described in this study may protect normal microflora, resulting in enhanced safety as a therapeutic.
The global concerns raised by drug-resistant microorganisms have encouraged exploration of novel antibiotics. Antimicrobial peptides (AMPs), which are evolutionarily conserved among many species for defense against pathogenic organisms, are promising therapeutic agents (1). Antimicrobial peptides have strong antimicrobial activity against a broad spectrum of microorganisms in vitro (2,3). They are also effective against pathogenic organisms that are resistant to conventional drugs (4). Therefore, AMPs have received much interest as a novel class of antibiotics.
Antimicrobial peptides are amphipathic peptides characterized by an overall positive charge and hydrophobicity. Antimicrobial peptides are electrostatically attracted to the negatively charged surface of the bacterial cytoplasmic membrane, and subsequent to the interaction with the membrane, they spontaneously oligomerize and form transmembrane pores that cause leakage of the cellular contents (5). Dissimilar to conventional antibiotics, acquisition of resistance against AMPs is very rare because the microorganism is killed by direct disruption of the microbial membrane (6,7).
In spite of their great potential, two major problems limit the development of AMPs as clinical therapeutics. One problem is the inherent instability and high metabolic turnover of AMPs (8). In general, peptides are rapidly metabolized by human cells and have a short half-life. Inversion of the stereochemistry is a promising way to resolve the rapid metabolic degradation of AMPs. Indeed, d-isomeric AMPs have excellent stability and a long plasma half-life (9). Importantly, they have similar activities compared with their corresponding l-amino acid enantiomers indicating that stereospecific recognition of AMPs is not required for cell disruption (10). d-Isomerization confers these favorable properties on AMPs, and some d-isomeric peptide drugs are currently undergoing clinical trials (11).
The other major problem is the exceptionally broad spectra of activity of AMPs. Broad-spectrum antibiotics disrupt the normal microflora and increase the risk of other fatal infections (12). Therefore, it is highly important to design therapeutics possessing temporally and spatially regulated antimicrobial activity. To solve the problem, we conceived a universal design of novel AMPs for the temporal and spatial regulation of their antimicrobial activity. Our design entailed inhibiting the antimicrobial activity until the peptide is cleaved by a particular protease (Figure 1A). We designed a compound composed of a positively charged AMP connected to a negatively charged protective peptide via a specific linker that is cleaved by a particular virulent protease. Selective killing of pathogenic microorganisms occurs following the release/activation of the protected AMP from the compound by particular virulent proteases (Figure 1B). In this study, we utilized Candida albicans as a model microorganism that possesses the secreted aspartic protease (Sap) family as an important virulence attribute (13,14).
Methods and Materials
Strains and media
Pichia pastoris strain GS115 [his4] (Invitrogen, Carlsbad, CA, USA) was used as a host for protein production. The C. albicans strain SC5314 (American Type Culture Collection) and Saccharomyces cerevisiae strain BY4741 [MATa, his3, leu2, met15, ura3] (Euroscarf) were used to measure the fungicidal activity of the AMPs. For protein production, P. pastoris transformants were precultivated in buffered glycerol-complex (BMGY) medium [1% (w/v) yeast extract, 2% (w/v) peptone, 1.34% (w/v) yeast nitrogen base (YNB) without amino acids, 4 × 10−5% (w/v) biotin, 1% (v/v) glycerol, and 100 mm potassium phosphate (pH 6.0)]. Precultivated cells were grown in buffered methanol-complex (BMMY) medium [1% (w/v) yeast extract, 2% (w/v) peptone, 1.34% (w/v) YNB without amino acids, 4 × 10−5% (w/v) biotin, 0.5% (v/v) methanol, and 100 mm potassium phosphate (pH 6.0)] for transcriptional induction. C. albicans and S. cerevisiae were grown in yeast peptone dextrose (YPD) medium [1% (w/v) yeast extract, 2% (w/v) peptone, and 2% (w/v) glucose] or YNB-bovine serum albumin (BSA) medium [0.17% (w/v) YNB without amino acids, 2% (w/v) glucose, and 0.2% BSA].
Production and purification of Sap isozymes
Secreted aspartic proteases isozymes were produced by P. pastoris according to the previously described method (15). In brief, the P. pastoris transformants were grown in BMGY media for 48 h at 30 °C. The culture medium was centrifuged, and the collected cells were resuspended in BMMY media for transcriptional induction. The supernatants of the BMMY media were mixed with an anti-FLAG M2 affinity gel (Sigma-Aldrich, St. Louis, MO, USA) and rotated for 1 h at 4 °C. The gel was washed with PBS buffer (pH 7.4) to remove non-specific proteins, and the protein concentration was quantified using the Protein Assay Bicinchoninate Kit (Nacalai Tesque, Kyoto, Japan).
Determination of substrate specificity of Sap isozymes
The substrate specificities of the Sap isozymes were determined using the FRETS-25Xaa library (Figure 2A; Peptide Institute, Osaka, Japan) as described in previous studies (15,16). In brief, the library contains 475 peptides, separated into 19 sublibraries. The FRETS-25Xaa sublibraries (2 mm) were dissolved in dimethyl sulfoxide (DMSO) and prepared by 200-fold dilution with 50 mm sodium citrate buffer at the optimum pH (pH 5, 4, and 3 for Sap1, Sap2, and Sap3, respectively). For the primary screening, purified Sap isozymes (final concentration, 5 nm) were mixed with each FRETS-25Xaa sublibrary (final concentration, 10 μm) in a total volume of 200 μL, and the increase in fluorescence was kinetically measured at an excitation wavelength (λex) of 355 nm and an emission wavelength (λem) of 460 nm. The relative degradation rate was calculated by the following equation: Relative degradation rate = [(Degradation rate of each sublibrary)/(Average degradation rate of all sublibraries)] − 1. For secondary screening, 10-fold-diluted FRETS-25K or FRETS-25R sublibraries were mixed with purified Sap isozymes in a manner identical to that described above. Then, the reaction solutions were subjected to liquid chromatography (LC)/mass spectrometry (MS) analysis.
LC analyses were performed using the Prominence HPLC system (Shimadzu, Kyoto, Japan) with a fluorescence detector (RF-10A; Shimadzu) at λex = 340 nm and λem = 440 nm. The reaction solutions were injected onto a reversed-phase MonoBis LC column (length, 100 mm; diameter, 3.2 mm; Kyoto Monotech, Kyoto, Japan) at a flow rate of 0.5 mL/min. The gradient was achieved by altering the ratio of the two eluents, namely A, 0.1% (v/v) trifluoroacetic acid and B, acetonitrile containing 0.1% (v/v) trifluoroacetic acid. The gradient was initiated with 5% B for 3 min, increased to 30% B for 20 min, further increased to 95% B within 3 min to wash, and then returned to the initial conditions. Chromatographic data collected by LC Solution software (Shimadzu) were used for quantitative analysis. To identify the peptide sequences, the target peptide peaks were fractionated and subjected to MS analysis using a LTQ Velos system (Thermo Fisher Scientific, Waltham, MA, USA). An electrospray ionization (ESI) voltage of 2.4 kV was applied directly to the flow using a microtee. The temperature of the ion transfer tube on the LTQ Velos ion trap was set to 300 °C. Subsequently, the peptide sequences were deduced from the theoretical molecular weights.
The peptide substrate DDAEAVGPEAFADEDLDE-GFIKAFPK-rrwqwrmkklG (using d-amino acids for the LF11 sequence as denoted by the lowercase letters) was synthesized for kinetic studies (Biologica, Aichi, Japan). Substrate solutions (100–1000 μm in 50 mm sodium citrate buffer) were mixed with each Sap isozyme (final concentration, 5 nm), and the reaction solutions were incubated for 24 min at 37 °C. Aliquots (5 μL) of the reaction solutions were subjected to LC/MS analysis to quantify the degradation rate, and the peptide peaks were quantified using UV absorption at 280 nm. The kinetic values KM and kcat were determined by fitting the data to the Michaelis–Menten equation.
Characterization of AMPs
All the peptides used in this study [AMP lactoferricin (LF11D), rrwqwrmkklG containing d-amino acids; the protective peptide magainin intervening sequence (MIS), DDAEAVGPEAFADEDLDE containing l-amino acids; the designed peptide, DDAEAVGPEAFADEDLDE-GFIKAFPK-RRWQWRMKKLG using l-amino acids or DDAEAVGPEAFADEDLDE-GFIKAFPK-rrwqwrmkklG using d-amino acids for the LF11 sequence] were synthesized by Biologica. To test their antimicrobial activity, 3.0 × 106 cells of S. cerevisiae were mixed with each peptide (0–25 μm) in 50 μL of 10 mm sodium citrate buffer (pH 3.0–7.0) and incubated for 1 h at 37 °C. Then, the solutions were serially diluted, and aliquots (3 μL) of the solutions were spotted onto YPD solid media. The sap isozymes were mixed with the peptides at their optimum pH and then mixed with S. cerevisiae to determine whether the Sap isozymes could activate the designed peptide. The antimicrobial activity of our peptide against vegetative cells of C. albicans was assessed by mixing the d-isomeric designed peptides (25 μm) and 3.0 × 106 cells of C. albicans or S. cerevisiae (as a control) in 50 μL of YNB-BSA media. To determine whether aspartic proteases activated the designed peptide, pepstatin A was added to be 1 μm. The mixture solutions were incubated for 3 h at 37 °C, and then, the number of colony forming units (CFUs) was measured by the serial spot assay described above.
SDS-PAGE and Western blotting
Sap2 production was detected by separating 10 μL of the YNB-BSA supernatant by SDS-PAGE on a 5–20% gradient gel. The protein bands were subjected to Western blotting and probed with an anti-Sap2 monoclonal antibody (Takara Bio, Otsu, Shiga, Japan).
The circular dichroism (CD) spectra were measured using a J720W spectropolarimeter (Jasco, Tokyo, Japan). The designed peptide or LF11D + MIS peptides (equimolar mixture) were dissolved in 10 mm sodium citrate buffer (pH 4.0 or 7.0) at a final concentration of 500 μm. To determine the secondary structure of the designed peptide in a membranous environment where AMPs function, SDS micelles were added to the solution at a final concentration of 200 mm. Spectra in the far-UV region (180–260 nm) were collected using a 1-mm quartz cell and averaged over 10 scans.
The mean values among groups were compared using one-way factorial anova, with p <0.05 being considered statistically significant. If the anova was significant, post hoc pairwise comparisons were performed using Tukey’s test, with p<0.05 being considered statistically significant.
Determination of substrate specificities: primary screening
We initially assessed the substrate specificities of the Sap isozymes to identify an appropriate linker sequence to join our AMP and the protective peptide in the design compound. We focused on Saps1–3 of the possible 10 Sap isozymes, because they are representative virulent proteases in the Sap family (13,17). The primary screening process entailed mixing the FRETS-25Xaa sublibraries with Saps1–3 and kinetically measuring their degradation rates by fluorescence. The average degradation rates of each sublibrary were calculated, and we found that Saps1–3 preferred the FRETS-25K and FRETS-25R sublibraries (Figure 2B). In addition, LC/MS analysis showed that almost all peptides were cleaved between Xaa and Ala as described in a previous study (15). These results indicated that Saps1–3 preferred basic amino acids at the P1 (Xaa) position.
Determination of substrate specificities: secondary screening
The FRETS-25K and FRETS-25R sublibraries were used to determine P2 (Yaa) and P3 (Zaa) preferences in the secondary screening. These sublibraries were mixed with Saps1–3, and the reaction solutions were subjected to LC/MS analysis. The peptide sequences of the four most intense peaks were identified by MS, and it was determined that Saps1–3 preferred Ile at the P2 position (Yaa) and Phe and Val at the P3 position (Zaa; Figure 2C). Among the peptides identified, GFIKAFPK was the most favorable substrate for Saps2–3 and GVIKAFPK for Sap1. We concluded that a GFIKAFPK peptide sequence was promising as a specific linker for our peptide because it was preferred by two of the three major Sap isozymes.
Design of an AMP activated by Sap isozymes
We designed an activity-regulated AMP composed of three domains, namely an AMP as the active center, a protective peptide that suppresses the activity of the AMP, and a specific linker that joins these two components and is efficiently cleaved by Sap isozymes (Figure 3A). We chose LF11 as a candidate AMP because it is generally produced by pepsin hydrolysis of lactoferrin, is safe for use in humans, and possesses strong antimicrobial activity against bacteria, viruses, and fungi, including C. albicans and S. cerevisiae (18). Lactoferricin has strong activity against C. albicans, and its minimal inhibitory concentration is 7.8 μg/mL (19). An MIS was utilized as a protective peptide to suppress the antimicrobial activity of LF11. Magainin intervening sequence functions in Xenopus laevis to temporarily suppress the antimicrobial activity of magainin (20), as it is highly negatively charged and able to suppress aberrant interactions of the positively charged magainin in the translation and maturation processes. Thus, we hypothesized that MIS could suppress the antimicrobial activity of LF11 until it is activated by Sap isozymes. Finally, we adopted the GFIKAFPK peptide sequence as the specific linker. These three components were chemically synthesized in tandem array, and the LF11 region was composed of d-isoform amino acids (LF11D) to promote metabolic stability.
The cleavage point of the Sap isozymes was identified by mixing our constructed peptide with Saps1–3 and analyzing the reaction solutions by LC/MS. As expected, the peptide designed in this study was broken into two fragments, and the cleavage point was identified to be between GFIK and AFPK by using MS (Figure S1). In addition, kinetic analysis of Sap2 against our designed peptide revealed that KM = 310 μm and kcat/KM = 5.37 × 104/M/s.
Characterization of fungicidal properties
The fungicidal activity of our peptide was investigated using S. cerevisiae (Figure 3B). S. cerevisiae do not possess any members of the SAP gene family, and therefore, we were able to evaluate the antimicrobial activity of the peptide without background Sap activity. First, the LF11D or MIS peptides were mixed with S. cerevisiae at various pH values and peptide concentrations to confirm their antimicrobial activities. As expected, LF11D showed fungicidal activity at all conditions, with the maximum activity at pH 4. Comparatively, MIS showed no activity under any of the conditions. Next, our compound was mixed with S. cerevisiae to investigate whether the protective peptide, MIS, successfully suppressed the antimicrobial activity of LF11D. Surprisingly, our peptide showed pH-dependent behavior (Figure 3B). Specifically, the antimicrobial activity of the designed peptide was as strong as that of LF11D at pH 3–4, while no activity was observed at pH 5–7. When the peptide was added to S. cerevisiae after 50% cleavage by Sap1, Sap2, or Sap3, the digested peptide showed antimicrobial properties similar to LF11D. Figure 3B demonstrates representative results obtained with Sap2.
Next, we synthesized our designed compound with l-amino acids for all the domains to determine whether the activation by low pH depended on d-isomerization of the LF11 region. The l-isomeric peptide showed a similar antimicrobial pattern as the d-isomeric one (Figure 3C). Thus, we concluded that the stereospecificity of the LF11 region did not affect the pH-dependent behavior. In addition, we assessed another version of our peptide that had an LKFFKA sequence for the specific linker. This peptide showed the same pH-dependent behavior, and thus, we ascertained that the activation by low pH does not depend on the specific linker sequence (Figure S2).
We hypothesized that the unexpected activation of our peptide in the acidic environment (pH 3–4) might be due to the extracellular proteases of S. cerevisiae. We subjected supernatants of the reaction solutions (pH 3 and 7) to LC/MS analysis and utilized a solution without S. cerevisiae as a control to test our hypothesis. Quantification of the residual amounts of the intact compound demonstrated no statistical difference between the samples (Figure 3D). This result indicated that activation of our peptide at low pH was not caused by unexpected digestion by the proteases of S. cerevisiae.
We hypothesized that our peptide changed its structure depending on ambient pH values and that this conformational transition led to activation. The peptide was dissolved in sodium citrate buffer (pH 4 and 7), and the CD spectra in the far-UV region were recorded (Figure 4A). The spectra in aqueous solution showed a sharp negative peak at 230 nm, indicating a poly-l-proline II helix-like conformation (21). The intensity of the peak at 230 nm at pH 4 was stronger than that at pH 7. To investigate the conformational differences in a membrane-like environment, where AMPs normally function, the CD spectra were collected in the presence of 200 mm SDS micelles. The negative peak at 230 nm significantly increased at pH 4 and decreased at pH 7. These results indicate that ambient pH values affect the secondary conformation of the peptide, especially in a membranous environment.
Next, we measured the CD spectra of an equimolar mixture of MIS and LF11D under identical conditions to investigate whether the MIS and LF11D peptides interacted with each other when not covalently bonded by a specific linker. The CD spectra obtained were different from that of the entire compound with no obvious secondary structure. This result indicates that the dynamics of the intensive peak at 230 nm of our peptide depends on the specific linker (Figure 4B).
Selective killing of C. albicans by our peptide
We investigated whether our peptide could selectively kill C. albicans. First, we confirmed that the LF11D and MIS peptides showed similar antimicrobial properties against C. albicans as against S. cerevisiae. As expected, the results obtained were similar to those obtained with S. cerevisiae, namely LF11D showed strong antimicrobial activity, but MIS showed none (Figures 3B and 5A). Next, C. albicans and S. cerevisiae were cultured in YNB-BSA media for 3 h, and C. albicans cells entered the vegetative state and secreted Sap2 (Figure 5B). Our peptide was then mixed with the vegetating cells in YNB-BSA, and the CFUs were measured after the reaction. Our peptide successfully showed selective antimicrobial activity against C. albicans, but not against S. cerevisiae (Figure 5C). The pH value of the YNB-BSA media was 5.6, and the liberated LF11D peptide was identified in the C. albicans supernatant by using MS (data not shown). In addition, a potent inhibitor against aspartic proteases, pepstatin A, inhibited activity of our peptide (Figure 5C). These results suggest that the peptide designed in this study is activated by Sap isozymes and selectively kills C. albicans.
The Sap family is a major virulence attribute of C. albicans. There are 10 types of Sap isozymes that play individual roles at certain stages of infection (22), and Saps1–3 are representative virulence factors (13,17). C. albicans produces these proteins mainly in the yeast form, uses them to damage epithelial tissues, and degrades host proteins for nutritional acquisition (17,23,24). Using the FRETS-25Xaa libraries, we revealed that Saps1–3 preferred positively charged or hydrophobic amino acids for P1 specificity, which is consistent with previous studies (15,25), and hydrophobic amino acids for P2–P3 specificity (Figure 2C). It has been previously reported that proteases with such a substrate specificity degrade AMPs that are rich in positively charged and hydrophobic amino acids (26). As expected, the l-isomeric peptide was quickly degraded when mixed with Sap2 (Figure S3). Thus, the peptide was synthesized using d-amino acids for the LF11 region to increase stability.
We measured the kinetic values of Sap2 against the designed peptide because Sap2 plays a major role among Sap isozymes (27). As a result, we found that the KM = 310 μm and kcat/KM = 5.37 × 104/M/s. The kinetic values did not exceed our expectations, but were equivalent to that of pepsin degrading fluorescent peptides (28). In this study, we screened a peptide substrate optimized for the P1–P3 specificity of Sap isozymes. Further screening aimed at optimizing the specificity will be able to improve the kinetic values.
Based on the substrate specificity of the Sap isozymes, we synthesized a peptide with a GFIKAFPK-specific linker. As expected, considering that Sap1, Sap2, and Sap3 have proteolytic activities at pH values of <6.5, <6.0, and <6.0, respectively (15), this peptide showed Sap-dependent antimicrobial activity above pH 5 (Figure 3B). Thus, we hypothesized the peptide will be able to temporally and spatially regulate its own antimicrobial activity depending on the Sap isozymes present in the physiological conditions.
To our surprise, our peptide was also activated by low pH (Figure 3B) and showed a strong activity, similar to that of LF11D at pH values of <4. We, therefore, investigated the possibility that our peptide was unfavorably cleaved by the proteases of S. cerevisiae. We quantified the amount of intact peptide in the reaction solution by LC analysis and found no degradation in the samples (Figure 3C). This result indicates that the activation at low pH was not caused by the extracellular proteases of S. cerevisiae.
The isoelectric point (pI) of our peptide was calculated to be 4.8 by using the compute Mw/pI tool (29). Thus, the designed peptide is positively charged when the ambient pH is <4.8. The cationic charge of the AMPs is important for interactions with the cytoplasmic membrane (30), and therefore, we suggest that the positive charge at low pH induces the activation of the designed peptide. This result demonstrates the possibility of designing AMPs that are more strictly regulated by decreasing the pI value via the addition of more acidic amino acids such as E and D.
Depending on their surrounding environments, AMPs undergo structural changes that affect their antimicrobial activity (18). For example, LF has an α-helical structure in intact lactoferrin (31), but once liberated from lactoferrin, it loses the α-helical structure and forms β-sheets. In a membrane-mimetic solvent, LF becomes an amphipathic molecule (32,33) that interacts with lipid tails via its hydrophobic cluster and is capable of disrupting membrane integrity. This type of conformational transition was also seen in our peptide. The CD spectra demonstrated that our peptide formed a poly-l-proline II helix-like conformation (21) in a membranous environment at pH values of <pI (Figure 4). We suggest this structure possessed the maximum amphipathicity and induced the activation of the peptide without proteases. In addition, vegetating C. albicans actively acidifies its surrounding environments (34); this can enhance the efficacy of our peptide, allowing it to be activated by both the Sap isozymes and the low pH.
Our peptide selectively killed C. albicans, but did not affect the survival of S. cerevisiae (Figure 5C). We propose that specific cleavage of our peptide by the Sap isozymes induced selective cell death, because pepstatin A inhibited activity of our peptide, and we identified liberated LF11D in the culture supernatant of C. albicans. Although S. cerevisiae has several extracellular proteases (35), the proteolytic activity of C. albicans is significantly greater than that of S. cerevisiae because of the Sap family. In vitro, C. albicans can grow using proteins as a sole nitrogen source, whereas in vivo it survives in the blood stream utilizing serum albumin (36). In addition, the specific linker of our peptide was optimized for Sap isozymes. Therefore, our peptide succeeded in discriminating between C. albicans and S. cerevisiae with the aid of Sap isozyme production.
It is possible that other proteases produced by host or other non-harmful microflora could also activate our peptide. Further characterization of antimicrobial specificity and optimization of the linker sequence will enable more efficient targeting to C. albicans. In addition, we should investigate whether secreted Sap isozymes are widely distributed to the physiological environment or not. Rapid diffusion of active Sap isozymes may lower the antimicrobial specificity of the designed peptide.
In conclusion, we designed an AMP that is activated by virulent proteases and is capable of selective killing C. albicans. Such smart AMPs do not disrupt the normal microflora and are, therefore, safer for use as antimicrobial therapeutics (37,38). The design utilized in this work has universal applicability for many pathogenic organisms that produce proteases as virulence attributes.
This work was supported by a research fellowship from the Japan Society for the Promotion of Science for Young Scientists (grant 22-101) and Regional Innovation Creation R&D Programs (grant 22R5005).