Rational design of hyperstable antibacterial peptides for food preservation

Analysis of natural peptides for the rational design of antibacterial properties

We started with the analysis of two natural peptides: HVBBI-a β-hairpin BBI found in skin secretions of Chinese bamboo odorous frog, Huia versabilis³⁶ and SFTI-a bicyclic trypsin inhibitor from sunflower seeds³⁷. Common to these peptides is a central trypsin inhibitory loop (hereafter termed simply as ‘loop’), flanked by cysteine residues that form a disulfide bridge (Fig. 1a). In addition to the loop, there are additional N- and C-terminal tail segments attached to the loop cysteines (hereafter termed as ‘tail’). The two peptides are structurally similar but differ in the nature of amino acids-both within the loop and the adjoining tail region. In case of HVBBI, the tail segments are composed of a cluster of hydrophobic residues and notably, a lysine residue at the C-terminal. Similarly, the terminal segments in SFTI, albeit a little shorter, also possess a couple of hydrophobic residues and an arginine at the N-terminus (Fig. 1b). The peptides exist as a hairpin due to disulfide-bond mediated cyclization of the loop region. Several studies have highlighted the importance of disulfide bond for high thermostability of these peptides and their resistance to proteolytic degradation³³. Due to their net positive charge, these peptides are attracted specifically toward bacterial phospholipid membrane, which is negatively charged due to 23% phosphatidylglycerol (PG) content³⁸. Interestingly, HVBBI, but not SFTI, displays a moderate antibacterial activity (Table 1, box A).

There are also some important differences. SFTI is a bicyclic peptide formed through a C–N-terminal cyclization, in addition to the disulfide bond (Fig. 1c). The residues comprising the loop region within the two peptides are also slightly different. In HVBBI, the loop sequence is WTKSIPPRP. This loop is responsible for the inhibitory activity of the peptide against extracellular and intracellular serine proteases¹⁶. The core sequence, TKS is conserved in trypsin inhibitory loops³⁹. SFTI has a shorter loop sequence TKSIPPI, which also retains the core inhibitory sequence, TKS. Nevertheless, these differences in the loop sequences result in different trypsin inhibitory activities (K_i); SFTI shows about 7 times higher K_i (5.5 × 10⁻⁸ M) compared to the K_i of HVBBI (37 × 10⁻⁸ M) (Table 1). Likewise, we see that the tail segments of both peptides show differences in sequence length and residue composition. HVBBI has a slightly higher hydrophobic residue composition compared to SFTI. Although the C-terminal lysine residue observed in HVBBI is missing in SFTI, the function associated with it, i.e., specific binding to bacterial membranes, could in principle be taken over by the arginine near the N-terminal end of SFTI. The tail segment—consisting of cationic residues together with hydrophobic segments—are also common in natural cationic antibacterial peptides¹⁷. Previous studies on cationic peptides have attributed their antibacterial, membrane destabilizing activity to the presence of basic (cationic) residues and a hydrophobic region (Fig. 1d)¹⁷. These peptides are thought to assume an amphiphilic structure, wherein the positively charged ends interact with the negatively charged polar head-groups, while the hydrophobic core interacts with the lipid acyl chains. From this, it follows that a greater number of cationic residues and/or hydrophobic residues would enable a stronger interaction between peptides and the phospholipid bilayer. Notably, the net positive charge on HVBBI is also higher compared to the net charge on SFTI (Table 1, box A).

We set out to understand how the aforementioned differences in HVBBI and SFTI affect their antibacterial activities. It is important to mention here that antimicrobial peptides are known to demonstrate significantly different antimicrobial activities or minimum inhibitory concentrations (MICs) against different bacterial species, even when they are closely related. Therefore, we use Gram-positive bacterium, Micrococcus luteus—one of the widespread food spoilage bacteria—as a simple model organism to compare the efficiency (MIC values) of the various peptides designed in this study. HVBBI shows an MIC value of 150 µg/mL against M. luteus, whereas SFTI shows no detectable inhibitory activity within the assayed concentration range (Table 1, box A). Furthermore, removing the tail segment from HVBBI peptide drastically compromises the antibacterial activity (HVBBI-loop, Table 1, box A). As expected for SFTI-loop, no antibacterial activity was observed upon removing the tail. Interestingly, for both SFTI and HVBBI, the removal of tail segments affected the trypsin inhibitory constant by 2 orders of magnitude (compare K_i values in Table 1, box A). This suggests that apart from its role in antibacterial activity, the tail segment also influences the anti-trypsin activity of the peptide.

Rational design of peptides with better antimicrobial properties requires an understanding of how the sequence and structure of the peptides may be related to their bioactivity. From the above discussion and previous studies on cationic antimicrobial peptides^40,41,42,43, we reasoned that increasing the hydrophobic core and the positive charge would result in better antibacterial activity. Interestingly, besides small peptides like HVBBI and SFTI, BBIs are usually found as 8, 16, and 24 kDa proteins with a minimal repeating 4 kDa unit⁴⁴. In the larger BBIs, the loop is flanked by two beta strands on either side, which are linked by one or more disulfide bridges that further enhance the stability. In the case of short peptides, the beta strands (tail segment) can be designed with higher hydrophobic composition and/or greater cationic charge for improved antibacterial activity through side-chain mutations or insertion of additional residues.

To test our hypothesis, we first designed a chimeric peptide, HSEP1—derived from HVBBI and SFTI, wherein the tail of SFTI was interchanged with that from HVBBI (Fig. 2a). SFTI-loop has about three times higher binding affinity for trypsin compared to HVBBI-loop (Table 1, box A). At the same time, HVBBI has a higher antibacterial activity, which may be attributed to its longer tail region. We reasoned that replacing the SFTI-tail with the HVBBI-tail, in principle, yields a peptide with a better overall antibacterial activity. Table 1 (box B) shows that the trypsin inhibitory constant obtained for the new peptide—HSEP1 was 23 × 10⁻⁸ M, which is similar to the value obtained for HVBBI. However, the antimicrobial activity of the chimeric peptide HSEP1 (MIC: 75 µg/mL) also showed improvement over that of HVBBI (150 µg/mL).

It is interesting to compare HSEP1 and HVBBI sequences in the context of their measured antibacterial activities. The two peptides differ only in the loop region. If one assumes that the antibacterial activity of the peptide was entirely dependent on the tail segments, the MIC values for both peptides should be similar. This however is not the case, as HSEP1 shows enhanced antibacterial effect compared to HVBBI (Table 1). This is despite the lower basicity (charge +2) of HSEP1 compared to that of HVBBI (+3); this leads to an interesting possibility of the loop region influencing the overall antibacterial activity of the peptide. It appears that the loop and the tail segments both influence the antibacterial as well as the anti-trypsin activities of the peptides; the exact factors governing this influence need to be understood. In summary, supplementing the SFTI-loop with HVBBI-tail resulted in a peptide, HSEP1, with better antimicrobial efficacy against M. luteus.

Improving upon sequence designs for better antimicrobial efficacy

We next sought to improve the HSEP1 activity by varying physiochemical properties such as hydrophobicity and net charge. First, we increased the hydrophobicity of HSEP1. It is understood that the hydrophobic residues aid in the interaction of the peptides with the acyl chain of the lipids. Moreover, mutual aggregation of peptides through these hydrophobic regions has been suggested to aid the membrane destabilization process⁴⁵. Of all hydrophobic residues, aromatic phenylalanine residue is particularly interesting as it has been reported to enhance mutual aggregation of the peptides on membrane surfaces¹⁶. Therefore, we introduced additional phenylalanine residues in both the N-tail (note, in HSPE2 that a Phe is inserted between Ile3 and Gly4 of HSEP1) and C-tail regions (note, in HSEP2 a Gly-Phe inserted between Val15 and Lys16 of HSEP1) to boost the hydrophobic character of the HSEP1 peptide (Table 1) (HSEP1: SVIGCTKSIPPICFVK and HSEP2: SVIFGCTKSIPPICFVGFK). Notice that in the C-tail region, we introduced an extra glycine residue as well. As bulky residues like phenylalanine introduce conformational restrictions that may hinder efficient peptide aggregation, an extra glycine was added to offset some of the backbone rigidity due to the two additional phenylalanine residues. Indeed, the redesigned peptide, HSEP2, exhibits significant antibacterial activity (6.25 µg/mL). Notably, the incorporation of additional phenylalanine residues affects the trypsin inhibitory constant of HSEP2 (54 × 10⁻⁸ M) only marginally compared to that of HSEP1 (23 × 10⁻⁸ M) (Table 1, box B).

We next sought to improve HSEP2 further by enhancing the net positive charge. We tested the variant HSEP3 wherein we introduced a basic residue (arginine) at the N-terminal end of HSEP2 to enhance its cationic character. This peptide exhibited the best antibacterial activity (1.25 µg/mL). Comparison of the sequence and activities of the three designed peptides (see HSEP1, HSEP2, and HSEP3 in Table 1, box B) suggests that the tail segment—common to all three peptides and including the hydrophobic core and terminal basic residues—is critical for the antibacterial activity of the peptide.

Hyperstability of HSEP3

The presence of a disulfide bridge imparts thermostability to the peptides—a desirable property for food preservation application. We determined the stability of the designed peptides at high temperature. HSPE3 exhibited ~30% decrease in trypsin inhibitory activity within the first 30 min of incubation at 95 °C (Fig. 2b) and retained 50% of its activity after heating at 95 °C for 200 min. We also tested the effect of temperature on the antibacterial activity of HSEP3. HSEP3 did not show any detectable reduction in the MIC upon dry or moist heat treatment (Supplementary Table 2). For testing pH stability, the peptides were incubated at three different pH [at pH 2.5, pH 5, and pH 9] (Fig. 2c, d and e). HSEP3 retained more than 80% of its trypsin inhibitory activity at all three pH values (Fig. 2e).

Characterization of designed peptides and their mode of action

The designed peptides fall under the class of cationic peptides that cause membrane destabilization. We performed localization studies on one of the peptides, HSEP2, using confocal laser-scanning microscopy. We first designed FITC tagged HSEP2 (FITC-HSEP2) and checked that FITC-tag had no effect on the antibacterial activity of the peptide (Table 1). For microscopy studies, we used B. cereus in addition to M. luteus. We verified that HSEP2 and HSEP3 show significant MIC values against B. cereus (HSEP2: 75 µg/mL, HSEP3: 12.25 µg/mL; Table 2). Confocal images of cells depict the bacterial cells (B. cereus and M. luteus) in fluorescent green color, indicating that the peptide indeed targets the bacterial cell membrane (Fig. 3). Moreover, live-dead staining shows that the peptides increase membrane permeability. In this assay, a cell with a damaged membrane takes up PI and fluoresces red; the live cells bind SYTO 9 and fluoresce green. HSEP3-treated cells appeared red indicating membrane permeability and cell death (Fig. 4a, b, c and Supplementary Figure 1).

We further monitored the membrane destabilization event over time through a simple propidium iodide (PI) uptake assay using B. cereus cells. Herein, we measured the total fluorescence emitted by dead cells over a time-course of 2 h. HSEP3-treated B. cereus cells become permeable to PI with increased fluorescence within the first 5 min (Fig. 4d), with signal saturation in 45 minutes. Moreover, the PI fluorescence also showed a dose dependency. At a concentration of 6.25 and 3.125 μg/mL of HSEP3, the fluorescence intensity was at the basal level. At 12.5 μg/mL, fluorescence increased significantly; this correlates with the MIC for HSEP3 peptide against B. cereus. A further increase in concentration resulted in increased PI uptake; however, after 50 μg/mL concentration, the PI fluorescence reached a plateau.

Finally, we observed changes in bacterial cell membrane morphology on treatment with peptides, as evident from scanning electron microscopy (SEM) imaging. Membrane surface of cells treated with peptides had a corrugated appearance and intracellular content leakage was observed (Fig. 5). In contrast, the negative control cells had a smooth appearance and had no signs of cell damage or intracellular leakage. Although the SEM images revealed membrane corrugations and cell leakage, the resolution was insufficient to discern membrane damage and pore formation on the cells. Transmission electron microscopy (TEM) was further carried out to obtain high-resolution micrographs of bacterial cells treated with HSEP3. Untreated cells showed normal cell shape with an undamaged inner and outer membrane ultrastructure. Cells were characterized by an increased electron density for the cytosol. The intracellular region exhibited a highly heterogeneous electron density, and no apparent damage or cytoplasmic leakage was observed in case of the untreated cells. In contrast, TEM micrographs of HSEP3-treated bacterial cells show distinctive cytoplasm-devoid zones, homogenous electron density, and loss of membrane integrity with clearly visible pores on the outer membrane (Fig. 6).

Simulation studies on lipid-peptide-water ternary system

To gain atomic-level picture of the peptide interaction with bacterial phospholipid membrane, we performed all-atom MD simulations on a system consisting of HSEP3 peptides in association with POPE-POPG bilayer. POPE-POPG bilayer was used previously as a model for a negatively charged bacterial membrane in studies on cationic peptides^46,47. Prior to peptide-membrane simulations, we performed simulations on POPE-POPG bilayer of different sizes—with 200, 400, and 800 lipids (see “Methods”)—to ensure that the bilayer properties such as bilayer thickness (D_P-P) and area-per-lipid (A_L) converged to stable and expected values (Supplementary Table 1). The calculated ensemble-averaged values of A_L and D_P-P, and K_A (isothermal area compressibility modulus) from the simulations are tabulated in Supplementary Table 1 and are close to previous reported values^48,49,50,51.

Since we obtained similar equilibrium properties for all three systems sizes, we chose the median 400-lipid system, for peptide-membrane simulations. We simulated a membrane-peptide-water system for 1.5 µs wherein the peptides were already bound to the upper leaflet of the membrane (Supplementary Figure 2). The ensemble-average values of A_L obtained for the peptide-bound membrane were 60.87 ± 0.75 Ų, and the calculated K_A of about 230 mN/m. The average bilayer thickness (D_P-P) was calculated to be 3.83 ± 0.18 nm. Comparison of these values with those calculated from a 300 ns control simulation of peptide-free membrane indicates that binding of HSEP3 to the membrane causes a decrease in the average membrane thickness by ~3 Å (D_P-P = 4.12 ± 0.03 nm for peptide-free membrane). However, simulations did not indicate a significant change in A_L values calculated for the peptide-free (A_L = 59.84 ± 0.64 Ų) and peptide-bound membrane (A_L = 60.87 ± 0.75 Ų). Previous studies have reported significant changes in the area-per-lipid values in case of membranes bound to peptide⁵². However, this modulatory effect of peptide binding on the membrane property varies with the peptide concentration used and is therefore expected to increase in simulations with higher peptide concentrations.

Visual analysis of the simulation trajectory revealed that HSEP3 peptides undergo aggregation on the upper leaflet surface of the bilayer, which seems to disrupt lipid packing and cause membrane-surface deformation. The distribution for the average bilayer thickness shows that the peptide-bound membrane has a lower average bilayer thickness with a higher variance (0.024 nm) compared to peptide-free membrane system (0.0021 nm) (Fig. 7). Moreover, we observed a drastic increase in water permeation and water insertion defects due to membrane deformations in case of the peptide-bound membrane, indicating higher water permeability (Fig. 8). Similar observations have been noted previously for other membrane destabilizing molecules such as Maculatin⁵³. We also calculated the lipid acyl-chain order parameter, S_CD for the peptide-bound and free membrane. Order parameters measure the chain ordering and higher chain ordering is correlated with greater lipid packing and bilayer thickness. The S_CD plots (Supplementary Figure 3) also indicate differences in the order parameters values for the acyl-chain carbon atoms of the peptide-bound membrane, compared with the corresponding values for the peptide-free membrane. Based on these results, it can be concluded that interaction of the membrane with the peptide disrupts proper lipid packing and subsequently leads to a decrease in membrane thickness. The aforesaid effect on lipid acyl-chain ordering leads to formation of transient defects in the membrane structure that become accessible to water molecules.

The designed HSEP3 peptide belongs to the class of cationic peptides as they possess a net positive charge. Higher positive charge enhances the binding of the peptides to the membrane. Analysis of the trajectory reveals that the peptides form on an average of 3–4 hydrogen bonds per peptide through lysine and arginine residues—amounting to roughly 6–8 kcal/mol of binding energy per peptide. The typical interactions formed between the peptide and membrane lipids are depicted in Fig. 9. This apart, the peptides comprise a number of hydrophobic residues; we have already demonstrated that the hydrophobic residues are essential for effective antimicrobial activity.

Our simulations show that HSEP3 exerts its membrane destabilizing activity by aggregating on membrane surface, leading to bilayer thinning and increased water permeation into the bilayer. For the effective number of peptides used in the simulations, we did not observe any membrane-insertion events in case of HSEP3 within the timescale of these simulations. There is experimental evidence that the exact mechanism of membrane disruption is dependent on lipid:peptide ratio and membrane composition⁵⁴. Recent work by Tieleman and coworkers indicates the absence of pore-forming membrane-insertion events in long simulations using the CHARMM force-field⁵⁵. Nonetheless, the simulation results support the ‘carpet model’ as a plausible membrane disruption mechanism employed by HSEP3. According to this model, the peptides bind and accumulate on the membrane surface like a ‘carpet’ and at high concentrations, the peptides disrupt the bilayer by causing membrane defects.

Investigating the role of the ‘loop’ in antibacterial function

The peptides have high inhibitory activity against trypsin owing to the loop region. Our observations in previous sections suggest that the peptides cause membrane disruptions leading to cell death, but the role of the loop region in the overall antimicrobial effect is unclear. Fluorescence experiments show that the peptides are internalized by the cells, which further indicates they may inhibit intracellular serine proteases. To evaluate this, we measured trypsin activity of bacterial cell lysate with and without peptide treatment. While the latter showed significant trypsin activity, peptide-treated lysates exhibited dose-dependent attenuation of intracellular trypsin activity (Supplementary Figure 4). This raises an interesting possibility; the peptides, in addition to binding and localizing at the membrane, enter the cytoplasm and inhibit intracellular trypsin. However, it is still not clear whether the trypsin inhibition per se plays any role in bacterial inhibition. Therefore, we next probed the link between trypsin inhibition exerted by the loop and the antibacterial efficacy of the peptide.

To this end, we sought to abolish the anti-trypsin activity through a lysine mutation at the TKS motif within the loop and examine the antibacterial activity of the mutant peptide. For our experiments, we created two mutants of HSEP2 (not HSPE3, as HSEP2 was the best peptide design at the time these experiments were conceived); one with a lysine deletion (HSEP2-ΔK8) and the other with a lysine to glycine mutation (HSEP2-K8G). As anticipated, both peptides failed to show any trypsin inhibition. However, in terms of antibacterial property, the two peptides behave differently. HSEP2-ΔK8 did not show any antibacterial activity, while HSEP2-K8G retained significant antibacterial activity, although less effective than HSEP2 (Table 1, box C). We also compared the fluorescence (PI uptake) for bacterial cells treated with peptides—HSEP2 and its mutants (HSEP2-ΔK8 and HSEP2-K8G). Results obtained agree with the antibacterial activity of these peptides; HSEP2-treated cells gave the highest fluorescence intensity indicating membrane destabilization and cell death. HSEP2-ΔK8 and HSEP2-K8G treated cells gave minimal fluorescence intensity, albeit higher in case of HSEP2-K8G, indicating marginal cell death (Fig. 4e and Supplementary Figure 5). These effects may be rationalized as follows. Both the deletion (HSEP2-ΔK8) and point mutant (HSEP2-K8G) variants have decreased net charge (by +1). Indeed, this is expected to affect the antibacterial efficacy of both peptide variants. More importantly, the deletion of lysine leads to a significantly perturbed loop structure due to a shortened length, which likely abrogates peptide-trypsin binding. In comparison, a significant antimicrobial activity seen for HSEP2-K8G could be because the mutation of lysine to glycine (in HSEP2-K8G) does not alter the loop structure significantly.

To delineate whether the TKS lysine contributes to antibacterial activity via charge effects, we also tested a HSEP2 peptide variant, HSEP2-ΔK19 wherein the C-terminal lysine was deleted without altering K8 (Table 1, box C). This peptide variant retains the overall net charge of +1, as HSEP2-K8G. HSEP2-ΔK19 demonstrated similar antibacterial activity (25 μg/mL) as HSEP2-K8G (37.5 μg/mL). While it is difficult to precisely pinpoint the roles, it appears that both K8 and the terminal lysine K19 contribute to antibacterial activity through charge effects that directly aid peptide binding to bacterial membrane. Indeed, HSEP3 with an additional arginine residue showed a better MIC value (1.25 μg/mL). We further redesigned the superior HSEP3 peptide such that it loses anti-trypsin activity, without altering the net charge and hydrophobicity. Particularly, we replaced the loop sequence (TSKIPPI) with a non-specific sequence (YRRF) that is expected to abrogate the anti-trypsin activity and at the same time compensate for the charge and hydrophobicity. We tested the new peptide, HSEP3-∆T_L,C_L⁺ for its anti-trypsin and antibacterial activity [Note: T_L stands for Trypsin Loop and C_L⁺ stands for the compensated charge]. As expected, it did not show any anti-trypsin activity, but still possessed good antibacterial activity (Table 1). Interestingly, the MIC value for HSEP3-∆T_L,C_L⁺ (3.125 μg/mL) was better than that of HSEP2-ΔK8, HSEP2-K8G, and HSEP2- ΔK19 peptides. The results show that the residues introduced in place of the loop sequence can, to an extent, substitute for its loss via charge and hydrophobicity effects. However, trypsin inhibition appears important, given the complete loss in antibacterial activity by HSEP2-ΔK8. Overall, it appears that the loop contributes to antibacterial activity through physiochemical (charge and hydrophobicity) effects causing membrane disruption, and intracellular trypsin inhibitory effects, the former being a dominant factor. The interplay of the two factors appear to govern the overall activity of the peptides.

We investigated this further by rephrasing the problem statement as follows. Of the total amount of peptide required for antibacterial inhibition, what fraction is sufficient for membrane destabilization, so that the remaining fraction can be assumed to be available for intracellular trypsin inhibition? Toward this, we made the following assumption. Out of the total amount of peptide required for inhibition, say x (equivalent to the MIC value), only a fraction of the peptides (a/x) is sufficient for membrane destabilization effects, while the rest of the peptide fraction (y = 1−a/x) is free to diffuse into the cytoplasm through the damaged membrane and inhibit intracellular trypsin. Under this assumption, we hypothesized that if the peptide fraction, y, is replaced by a different peptide that possesses only trypsin inhibitory activity, i.e., HSEP2-ΔHR (denoting that the Hydrophobic Region is deleted) then the effective antibacterial efficacy of the resulting cocktail of peptides should be the same as that observed for the total peptide (x). To test this hypothesis, we prepared cocktails of HSEP3 and HSEP2-ΔHR in ratios [HSEP3:HSEP2-ΔHR] of 0:4, 1:3, 2:2, 3:1, and 4:0 (for peptide sequences and properties, see Tables 1 and 3). Note that HSEP2-ΔHR does not possess any detectable antibacterial activity but retains reasonably high trypsin inhibitory activity. Here, 0:4 and 4:0 ratio cocktails represent pure HSEP2-ΔHR and HSEP3, respectively; and 1:3, 2:2, and 3:1 are the test cocktails in varying ratios. We tested the antibacterial activity of these cocktails against both B. cereus and M. luteus. The results tabulated in Table 3 show that out of the three tested cocktails, the one with ratio 3:1 achieves the same activity as observed with pure HSEP3 preparations (i.e., 4:0). The observed trend was the same for both bacterial species. From this, it is evident that about 75% of the total peptide is sufficient for membrane destabilization, while the rest 25% is involved in trypsin inhibition. Further, we tested another set of cocktails prepared from HSEP3 and HSEP2-ΔHR-ΔK8, which represents the inactive form of HSEP2-ΔHR without any trypsin inhibitory activity (as the lysine K8 is mutated here). As expected, the results demonstrate an increase in MIC values against both B. cereus and M. luteus, at this 3:1 ratio.

These results pose important implications to the design of synergistic cocktails with antibacterial properties, wherein the surplus fraction of the peptide (y = 1−a/x) may be replaced by another antibacterial molecule with a different mode of action, for instance, the inhibition of another key bacterial function. Previous studies on synergistic action of drugs have concluded that the combination of two drugs often leads to better activity for the combined preparation as compared to the individual activities of parent molecules^56,57. Further studies are required that test cocktails of HSEP3 in combination with other antimicrobials; such studies may facilitate the identification of effective combinations that possess high antibacterial activity.

Design of HSEP3 variants: modulating peptide properties and efficacy

We know that HSEP3-membrane interactions are mediated chiefly via polar bonds, weak van-der-Waals forces, and stronger hydrophobic interactions. Thus, modulating the properties of the peptide, such as number of cationic and polar amino acids and hydrophobicity, may in turn alter the antibacterial activity of the peptide; particularly, the introduction of arginine residue and/or hydrophobic residues such as phenylalanine and isoleucine. We created a small set of four HSEP3 variants, which were mutated to have (a) an increased hydrophobicity [HSEP3a (HSEP3-Val3Phe)], (b) higher polar residues [HSEP3b (HSEP3-Phe19Thr)], and (c) increased cationic charge [HSEP3c: Lys inserted between Ser10 and Ile11 of HSPE3]. We also designed a peptide variant, HSEP3d with mutations that introduced arginine (for greater H-bonding propensity) in place of lysine (HSEP3-Lys9Arg, HSEP3-Lys20Arg), tryptophan (for more hydrophobicity) in lieu of phenylalanine and valine (HSEP3-Val3Trp, HSEP2-Phe19Trp), and a bulkier tyrosine (for additional stacking interactions) in place of serine (HSEP3-Ser10Tyr). Out of these set of peptides, HSEP3c and HSEP3d demonstrated the best activity against M. luteus (3.125 µg/mL), whereas HSEP3b was reasonably effective (6.25 µg/mL). All peptides showed reasonable trypsin inhibitory activity. We also tested these peptides against a broad range of bacteria and observed that some of these peptide variants (HSEP3b and HSEP3c) also show efficacy against other microorganisms such as Listeria and Pectobacterium (Table 2).

Cytotoxicity and hemolytic assays

We tested the safety of selected peptides by cytotoxicity assay on two different human cells lines, human retinal pigment epithelial cells and intestinal epithelial cell lines. The results show that for HSEP2, the cell viability remained >80% for a peptide concentration range of 0–160 µg/mL. At a concentration of 200 µg/mL, the cell viability was >70%. HSEP3 too showed similar cell viabilities as HSEP2 for the assayed peptide concentrations (Fig. 10a and Supplementary Figure 6).

We also performed hemolytic assay on both peptides to check the effect of peptides on human RBCs. Figure 10b shows the results of the assay; we observed low hemolytic activity of the peptides toward RBCs. At a peptide concentration ranging from 0 to 200 μg/mL, both HSEP2 and HSEP3 showed <5% hemolytic activity. These results show that the peptides are relatively safe to use.

Employing designed peptides for food preservation

In the previous sections, we have demonstrated that the designed peptides have high stability and antibacterial activity. These and similar peptides can in principle be employed for various applications, including topical medical applications, surface sterilization, food preservation, etc. Here, we focus on applications to food preservation.

For assessing the antibacterial efficacy of the peptides, we tested some selected peptides against a range of important spoilage bacteria some of which also cause food-borne illness. Four Gram-positive and three Gram-negative bacteria were selected for these assays (see “Methods”). The MIC values for all tested peptides are tabulated in Table 2. Comparing the MIC values of all peptides across different bacteria, we conclude that HSEP3 demonstrates the best antibacterial efficacy and inhibits a wider spectrum. For a maximum concentration up to 100 µg/mL, HSEP3 can kill 5 out of 7 bacterial species that were tested. Other peptides such as HSEP3b and HSEP3c also show reasonable activity (<=50 µg/mL) against important food pathogens such as P. cartovorum, L. monocytogenes, and B. cereus.

Further, we demonstrate the applicability of the best peptide in this study—HSEP3 as an antibacterial preservative. Since rice is a major staple food consumed all over the globe, we sought to test the efficacy of HSEP3 in preventing spoilage of cooked rice samples. Briefly, rice samples treated with HSEP3 were inoculated separately with B. cereus, M. luteus, and L. monocytogenes cells. The test samples were incubated for up to 6 days along with negative control samples (without HSEP3). Samples were aseptically obtained for testing every 24 h and the level of bacterial proliferation was monitored by (a) culturing samples on agar media to obtain the CFU values and (b) qualitative cell viability using resazurin dye (see “Methods” for details). We found no growth in the samples up to six days (Supplementary Figure 7). The control samples on the other hand, demonstrated a steady increase in CFU values over the assay duration.

Notably, Nisin—a peptide-based preservative used in food industry is relatively inadequate on accounts of both pH and temperature stability. It shows highest activity at pH 3 with rapid loss of activity as pH increases. The temperature stability of Nisin is also dependent on the ambient pH; it can be autoclaved without much loss of activity at pH 3, but at pH 7 it loses 90% of its activity, independent of the ambient temperature⁶. This severely limits applicability of Nisin. HSEP3 has comparatively better stability against high temperature and wide range of pH (Fig. 2 and Supplementary Table 2). Hence, HSEP3 is compatible with other methods of preservation such as thermal sterilization, acid regulation, etc. When HSEP3 and Nisin were compared for the antibacterial activity against B. cereus, HSEP3 showed better efficacy (Supplementary Table 1). We also compared HSEP3 and Nisin for their ability to inhibit B. cereus spore germination. HSEP3 demonstrated ~16-fold higher activity compared to the Nisin (Supplementary Table 2). To summarize, we demonstrate the potential of HSEP3 as an excellent alternative to current antimicrobial preservative. It demonstrates better stability and antimicrobial spectrum and therefore could be employed in cases where other preservatives such as Nisin are ineffective.

This work reports our attempt toward the rational design of antibacterial peptides—notably, with multiple properties like broad pH-resistance, exceptional thermostability, and anti-proteolytic activity—that enables their wider application to a variety of areas. Out of the 12 peptide variants designed in this study, 7 peptides show good MIC (<10 µg/mL) value against at least one bacterial sp., M. luteus (Tables 1 and 2). The best peptide design (HSEP3) shows reasonable antibacterial activity (<100 µg/mL) against five different bacterial species, while retaining stability against temperature, pH, and proteolysis. This demonstrates the advantage of rational design over a purely combinatorial approach. Typically, combinatorial methods require designing and screening several thousands to a million candidates to obtain good hits. Rational methods employing fundamental knowledge about various properties of the system can enable narrowing down to good hits within a few design iterations from a limited number of starting candidates. A combination of rational and combinatorial design strategies has been increasingly successful in designing proteins with novel functions^58,59,60,61. A similar strategy can also be applied to design of peptides tailored for the application.

Our study focuses on the role of the loop in antibacterial action and the synergism between its trypsin inhibitory function and the membrane destabilizing action of the peptide tail. We show that (a) the peptides also inhibit intracellular trypsin and (b) both membrane destabilization and intracellular trypsin inhibitory activity of the peptide contribute to its overall antimicrobial activity. This means that it may be possible to redesign the ‘loop’ region against other essential intracellular proteins, which should further boost the inhibitory activity as well as the bactericidal spectrum of the peptides. While the loop serves as a determinant of intracellular target stability, the terminal region of the peptides can be redesigned to achieve maximal membrane destabilizing activity. Such strategies can be employed to expand the pool of functional antibacterial molecules. This finding is also of relevance to the design of synergistic combinations to obtain effective antibacterial peptide cocktails.

The insights obtained through the aforesaid studies are of broad importance to the design of antimicrobials for clinical and biomedical applications as well. In addition to the development of novel alternatives to current antimicrobials, synergistic combinations of new and existing peptides may pave way to more effective treatment of resistant infections. Moreover, by using combination of peptides with different modes of antibacterial action, the chance of developing bacterial resistance is also lowered. A case in point is the development of a superior version of the antibiotic—Vancomycin—that combines three different inhibition mechanisms, resulting in a molecule that has much better efficacy⁶².

Although in principle, the peptide designs may be applicable to a wide range of applications, a separate assessment of their efficacy is necessary for specific applications. For this study, we have tested the best peptide design (HSEP3) for its applicability as a food preservative and demonstrate that HSEP3 could be a potential alternative to current antibacterial preservatives like Nisin. However, one compelling argument against the application of trypsin inhibitor-based antimicrobial peptides could be its activity against digestive serine proteases. We reason that these peptides should be safe for human and animal consumption, if the inhibitor concentration required for effective food preservation is lower compared to the concentration that cause perceptible anti-nutritive effects during digestion—a concept that is similar to the Therapeutic Index used to assess systemic drugs. In relation to this, the studies on USFDA-approved soybean Bowman–Birk inhibitor concentrate (S-BBIC) (which is the same class of inhibitors as our peptides) are roughly indicative of the safety of trypsin inhibitor-based molecules for animal and human consumption. S-BBIC has been tested for its efficacy against various carcinogenic conditions in animal and human trials⁶³. For these studies, a concentration of up to 10 mg/kg body weight of humans has been recommended as a safe dose. This value corresponds to roughly 500 µg/mL of peptide concentration which is more than 3-fold of the maximum MIC values obtained for our peptides (150 µg/mL). Although these values are rough estimates based on S-BBIC data, they suggest that HSEP3 could be an effective and safe candidate as a food preservative. However, animal studies are necessary to obtain a more reliable assessment of the potential of HSEP3.

Studies are currently underway to test the suitability of these peptides for other applications and develop better variants employing more robust design approaches. In addition, the molecular details underlying the difference in antibacterial efficacy against different bacteria need to be investigated in order to rationally develop peptides with broad-spectrum antibacterial activity.