Design of self-processing antimicrobial peptides for plant protection


W.A. Powell, SUNY College of Environmental Science and Forestry, 1 Forestry Drive, Syracuse, NY 13210-2788, USA (e-mail:


Small antimicrobial peptides are excellent candidates for inclusion in self-processing proteins that could be used to confer pathogen resistance in transgenic plants. Antimicrobial peptides as small as 22 amino acids in length have been designed to incorporate the residual amino acids left from protein processing by the tobacco etch virus'(TEVs') NIa protease. Also, by minimizing the length of these peptides and the number of highly hydrophobic residues, haemolytic activity was reduced without affecting the peptide's antimicrobial activity.


There has been an extensive amount of research on antimicrobial peptides in the medical area to develop peptides as topical antimicrobial agents, antitumour agents, and for enhanced wound healing ( Maloy and Kari 1995). Also, significant progress is being made to develop these types of peptides for use in the area of plant protection ( Rao 1995). Previously, we designed magainin-like, synthetic peptides that demonstrated growth inhibitory effects on both fungal and bacterial plant pathogens ( Powell et al. 1995 ). Our current interests lie in designing genes that can deliver multiple resistance products that would enhance sustainable plant pathogen resistance in long-lived, woody plant species ( Powell and Maynard 1997). In this report, antimicrobial peptide designs are described that could be used in self-processing protein precursors. Transgenic plants containing genes encoding these precursor proteins could be made to release a ‘cocktail’ of antimicrobial peptides, possibly with antimicrobial enzymes, that could act synergistically against the pathogen. Using a single self-processing gene has an advantage over standard multiple-gene pyramiding because the gene products would be produced under the control of a single gene promoter, therefore their expression controlled as a unit. Also, when a single gene is used to deliver multiple antimicrobial products, there is no possibility of segregation during plant breeding as there is when using multiple genes for pathogen resistance. A single self-processing polyprotein product also simplifies the technical problems with designing and building large constructs with multiple promoters and terminators flanking the coding regions. One way to design this type of gene system is to utilize the self-processing protein mechanisms found in potyviruses. The NIa proteases from two of these viruses have been shown to process cloned gene products in vitro ( Marcos and Beachy 1994) and in vivo ( von Bodman et al. 1995 ; Ceriani et al. 1998 ). One potential problem with using this system is that the proteases need a recognition site for cleavage that leave amino acid sequences on the resulting protein or peptide products. On a large protein, these small amino acid ‘tails’ are usually trivial and often have little or no effect on the protein's activity. However, for small antimicrobial peptides, like our ESF designs ( Powell et al. 1995 ), these residual sequences would account for about a third of the peptide's total amino acid sequence and therefore could significantly effect the peptide's activity. Our hypothesis is that residual amino acid residues retained on a self-processed peptide can be incorporated into the overall peptide design so that it retains its high antimicrobial activity and low haemolytic activity. To test this hypothesis, we initially determined the lower size limit of active ESF peptides in order to keep gene design as compact as possible. Then selected amino acid changes assessed for their effect the haemolytic and antimicrobial activities of the peptides. Then, the self-processing NIa protease system from tobacco etch virus (TEV) was chosen as a model for modifying the ESF peptides. Using the known protease recognition sequence from this system ( Carrington and Dougherty 1988; Dougherty et al. 1988 ), it was determined if residual amino acid sequences from the protease left on the final peptide products would change their antimicrobial and haemolytic activities.

Materials and methods


As an antimicrobial peptide control (Ala8,13,15) Magainin 2 amide ( Chen et al. 1988 ) was purchased from Sigma. The ESF peptide designs were aided by using the Garnier-Robson and Chou-Fasman structural analysis, Eisenberg hydrophobicity (H), alpha moment (α), and beta moment (β) ( Eisenberg et al. 1984 ), pI, and charge functions of Protean computer software (DNASTAR, Madison, WI, USA). ESF peptides were synthesized (> 70% purity) by Genosys Biotechnologies (The Woodlands, TX, USA).

Fungal and bacterial culture

Two plant-pathogenic fungi were assayed. Cryphonectria parasitica (Murrill) Barr, strain EP42, was obtained from American Type Culture Collection (ATCC 38751) and maintained on potato dextrose agar supplemented with methionine and biotin (PDAmb) ( Anagnostakis 1982). Fusarium oxysporum Schlechtend: Fr. f.sp. lycopersici (Sacc.), strain 73 was obtained from Felice Cervone, Plant Biology Department, University of Rome, Italy and maintained on potato dextrose agar (PDA).

Two plant-pathogenic bacteria were assayed. Agrobacterium tumefaciens (wild-type strain Bo542) was provided by Eugene Nester ( Sciaky et al. 1978 ). Pseudomonas syringae pv syringae bacteria (Collection PSS34, streptomycin resistant) were provided by Thomas Burr (Department of Horticultural Sciences, Cornell University, New York State Agricultural Experiment Station, Geneva, New York). Stock cultures of Ps. syringae were grown on a specific, defined nutrient medium ( Atlas and Parks 1993). Ag. tumefaciens was maintained on PDA.

Determination of the minimal inhibitory concentration (MIC)

Conidial suspensions were aseptically prepared from agar plate cultures of Cryphonectria parasitica grown at 25 °C, 16-h photoperiod, or shaking liquid cultures for Fus. oxysporum grown at 25 °C, in the dark. Conidia were suspended from agar plates with 10 ml sterile 1% Tween 20 (Sigma). All culture suspensions were filtered through four layers of sterile cheesecloth, collected in sterile 15 ml Falcon tubes, centrifuged for 3 min at 1900 g and the supernatant was then poured off. The pellets were suspended in 10 ml sterile deionized water, centrifuged for 3 min at 1900 g in a IEC Centra-4B centrifuge at room temperature and then the supernatant was poured off. These ‘washed’ pellets were suspended in 5 ml sterile deionized water and stored on wet ice while the conidial concentration was determined using a haemacytometer (Fisher ultra plane, Neubauer ruling). Conidial suspensions were diluted to 1·0 × 104 conidia ml−1. Bacterial suspensions of Ag. tumefaciens and Ps. syringae were obtained from liquid cultures grown in potato dextrose broth (pH 5·2) (Difco). Cultures were grown at 25 °C on a Lab-Line orbital shaker (110 rev min−1) for 14 h. Bacterial cell suspensions with an O.D.600 between 0·7 and 0·85 were used for assay inoculations.

Dilutions (0, 0·625, 1·25, 2·5, 5, 10, 15, 20, 25, 50, 100, 150, 200 and 250 µmol l−1, except where noted) of peptides tested against fungi and bacteria were aseptically prepared in sterile double-distilled water. Media for fungal growth (PDAmb) was prepared with deionized distilled water, Difco potato dextrose broth (12 g l−1), Sigma d, l-methionine (0·05 g l−1), biotin (2 mg l−1), low melting agarose (20 g l−1) and adjusted to pH 5·2 (before the addition of the agarose). Aliquots (20 µl) of peptide solution were aseptically placed in Corning disposable sterile polystyrene ELISA plates (96-well, high binding) which had been pretreated with bovine serum albumin (BSA, Sigma) by rinsing each well with 200 µl of a 1·0-mg ml−1 BSA solution. Sterilized media (80 µl) was added to each well to make a 100-µl volume and allowed to solidify prior to inoculation with 10 µl of conidial or bacterial cell suspension. Inoculated microtitre plates were incubated in ambient light at room temperature in a moist chamber (plastic box containing wet paper towels and covered with clear plastic wrap). Plates were scored for growth by looking for mycelium for the fungi or colonies for the bacteria at 6 d. The plates were then photographed.

The MIC of the fungi and bacteria was the lowest peptide concentration that totally prevented growth. All tests were repeated four or more times. Controls without peptides were included with all tests. The highest concentration tested was 250 µmol l−1 except where noted. If germination or bacterial growth occurred at this concentration, the MIC is yet undetermined, but it is known to be greater than 250 µmol l−1.

Peptide haemolytic activity

Three millilitres of human blood cells (Carolina Biological Supply, Burlington, NC, USA) were gently mixed, poured into a sterile 15 ml polystyrene screw-cap tube and centrifuged 5 min, 850 g. The supernatant was poured off and the viscous pellet washed three additional times with 5 ml of chilled (4 °C) sterile isotonic phosphate-buffered saline (PBS) solution (amounts g l−1: NaCl, 8; KH2PO4, 0·2; Na2HPO4, 1·2; and KCl, 0·2, adjusted to pH 7·4, mixed for 60 min to stabilize pH). The washed cells were suspended in a final volume of 20 ml chilled, sterile PBS and the cells counted on a haemacytometer (Fisher ultra plane, Neubauer ruling). The blood cell suspension was maintained on wet ice and diluted with sterile PBS to 7·068 × 108 cells ml−1 for each assay. Aliquots of 20 µl of peptide were aseptically placed into 2·0 ml microfuge tubes. For each assay, 0·1% Triton X-100 was the positive, 100% lytic control and PBS was the negative, background (0% lysis) control. Aliquots of 180 µl diluted blood cell suspension were aseptically placed into each 2-ml tube and gently mixed three times with a wide mouth pipette tip. The peptide concentration tested was 250 µmol l−1. Tubes were incubated for 35 min at 37 °C with agitation (80 rev min−1). Immediately following incubation, the tubes were placed on ice for 5 min then centrifuged for 5 min at 1310 g. Aliquots of 100 µl of supernatant were carefully collected, placed into a sterile 1·5 ml microfuge tube, and diluted with 900 µl chilled, sterile PBS. All tubes were maintained on wet ice after dilution. Absorbance at 576 nm was measured on a Bausch and Lomb Spectronic 1001 spectrophotometer using a quartz cuvette. Three replicates were run for each peptide per assay. Two to five assays were performed for each peptide.

Statistical analyses

Data were analysed with SAS® software (SAS, Cary, NC, USA; 1994) using the general linear models (GLM) procedure. A multivariate analysis of variance ( manova) model with main effects was used. Tukey's Studentized Range Test (alpha = 0·01) was performed on all main effect means generated from minimum inhibitory concentration data and haemolysis data. Means were compared among replicates within each peptide treatment (within treatment variance) and among means between peptide treatments (between treatment variance). Tukey's Studentized Range Test controls the maximum experiment-wise error rate and is applicable to unequal sample sizes.


To determine if size effects the ESF peptide designs, peptides were synthesized containing 22 (ESF22A and B), 20 (ESF1B), 15 (ESF15), 13 (ESF13A) and nine (ESF9A) amino acids ( Table 1). These were compared with ESF12 (18 amino acids), ESF17 (17 amino acids), and a control (Ala8,13,18) magainin 2 which have previously been studied ( Powell et al. 1995 ). MICs were determined as described in the methods section and reported as an average of four or more repeats ( Table 2). Average MICs within a column containing the same superscript letter were not significantly different as determined by the Tukey's Studentized Range Test. Haemolytic bioassays ( Table 3) were performed with human red blood cells and report the average lysis with respect to the triton X-100 control.

Table 1.  Antimicrobial peptides
PeptideAA sequencea *chargeH
  • *

    Alpha moment,

  • †hydrophobicity.

ESF peptides
ESF15SRAAGLAARLARLAL0·47+ 2·91+ 0·06
ESF13ASRAAGLAARLARL0·46+ 2·91− 0·01
ESF9ASRLARLAAR0·62+ 2·91− 0·23
Control peptide
(Ala8,13,18) magainin 2 GIGKFLHAAKKFAKAFVAEIMNS0·46+ 3·07+ 0·20
Table 2.  Average minimum inhibitory concentration (MIC), in micromoles per litre (µmol l−1), of ESF peptides and control for four plant pathogens, Fusarium oxysporum, Cryphonectria parasitica, Agrobacterium tumefaciens and Pseudomonas syringae. The same superscript letter within a column indicates no significant difference in the average MIC as determined by Tukey's Studentized Range Test
PeptideNo. AAFusarium
  1. *Highest concentration tested for these peptides was 100 µmol l −1 instead of 250 µmol l−1 used for the rest of the peptides.

  2. a, b, c, d Average MICs within a column containing the same superscript letter were not significantly different as determined by the Tukey's Studentized Range Test.

ESF22A222·8 d11 b6·3c, d88a, b
ESF22B225·0c, d13 b3·8 d50a, b
ESF1B201·9 d6·3 b3·1 d100a, b
ESF12187·9c, d8·8 b25 c150a, b
ESF17 *17> 100> 100> 100> 100
ESF15 *1535 a36 a50 b23 b
ESF13A *13> 100> 100> 100> 100
ESF9A *9> 100> 100> 100> 100
(Ala8,13,18) magainin 2 2313b, c13 b100 a250 a
r2 0·9770·9490·9830·847
Table 3.  Haemolytic activity, as a percentage of haemolysis caused by 0·1% triton X-100, for 250 µmol l−1 of ESF peptides and controls. The same superscript letter within the haemolysis column indicates no significant difference in the average MIC as determined by Tukey's Studentized Range Test
PeptideNo. AA#Leu or ValHaemolysis *
  1. * Tukey's Studentized Range Test grouping (A–F) of lysis by peptide, r2= 0·997.

  2. a, b, c, d Average MICs within a column containing the same superscript letter were not significantly different as determined by the Tukey's Studentized Range Test.

ESF22A22570 b
ESF22B22420 c
ESF1B20673 b
ESF620516c, d
(Ala8,13,18) magainin 2 23N/A10 d,e
PBS buffer  0·32f
0·1% Triton-X 100  100 a

The results of the MIC bioassays indicate a significant loss of antimicrobial activity when reducing the ESF peptide size from 18 to 17 amino acids for all four of the plant pathogens tested. Interestingly, antimicrobial activity of the 15 amino acid derivative was significantly higher than that of the 17 amino acid form, but still significantly lower than the 18 amino acid form. The one exception to these results was with the bacterial assay using Ps. syringae, in which the 15 amino acid form's MIC was not statistically different than the 18 amino acid form. In this assay it was still more active than the 17 amino acid peptide. The only difference between ESF17 and ESF15 is the loss of an alanine from the amino and carboxyl ends of the ESF17 peptide to form ESF15 ( Table 1). The decrease in activity of ESF17 and regaining of some activity in ESF15 as the size suggests that the smaller peptide's activity could be due to a different mechanism as has been reported for other synthetic peptides ( Bessalle et al. 1993 ). There was little difference in the MICs of the peptides varying in size between 18 and 22 amino acids, with the exception of ESF12 in the Ag. tumefaciens bioassays where ESF12 was significantly less active ( Table 2). The relative consistency of antimicrobial activity among the different peptide sequences indicate some flexibility in design within this size range without loss of antimicrobial activity.

All the ESF peptides which were 18 amino acids in length or larger had similar MICs as the (Ala8,13,18) magainin 2 ( Chen et al. 1988 ) control in the fungal and Ps. syringae bioassays. In the Ag. tumefaciens bioassays, the ESF peptides were significantly more inhibitory than (Ala8,13,18) magainin 2. ESF15 was less active than (Ala8,13,18) magainin 2 in the fungal bioassays but more active than (Ala8,13,18) magainin 2 in the bacterial bioassays.

Differences in peptide size did cause a statistically significant change in haemolytic activity ( Table 3). ESF peptides of 18 amino acids or less were not significantly different in haemolytic activity than the PBS buffer control. ESF peptides of 20 amino acids or larger, including the (Ala8,13,18) magainin 2 control, were significantly more haemolytic than the PBS buffer control. Therefore, size effects haemolytic activity in general. However, by examining other differences among ESF1B, ESF6 and ESF5, which were all 20 amino acids in length, other influencing factors could be discerned. The most consistent factor among these three peptides was the number of highly hydrophobic amino acids in the design. By manipulating these amino acid sequences larger peptides can be designed with lower haemolytic activities than smaller peptides, as demonstrated by comparing ESF22B and ESF1B ( Table 3).

ESF22A and ESF22B were designed to contain amino acid residues of the TEV NIa protease's recognition sequence that would be retained by the resulting peptide if digested with this protease. The antimicrobial activity of both these peptides grouped with the most active peptides tested ( Table 2). ESF22A had a significantly higher haemolytic activity than ESF22B.


The primary hypothesis tested was that the residual amino acids retained on a self-processed peptide could be incorporated into a peptide design so that it retains a high antimicrobial activity yet poses a low haemolytic activity. Consideration of both these activities is important to target the peptides to microbial pathogens while having a minimal effect on higher organisms that might consume plant products containing these peptides. In testing this hypothesis, we first determined the lower size limit of active ESF peptide before adding the residual amino acid sequences which would be left after cleavage by the self-processing NIa protease system from TEV ( Carrington and Dougherty 1988; Dougherty et al. 1988 ). The lower size limit for antimicrobial activity of the ESF peptide design was determined to be 18 amino acids. The exception was ESF15, which had measurable antimicrobial activity but was only 15 amino acids in length. The antimicrobial activity of ESF15 is likely due to a different mechanism than the larger ESF peptides. Changes in activity due to size reductions have been reported with other synthetic peptide designs ( Bessalle et al. 1993 ). However, there were no significant structural differences in ESF15 predicted using the Garnier-Robson and Chou-Fasman structural analysis functions of Protean computer software (DNASTAR).

After determining a minimal size that would retain antimicrobial activity, a peptide sequence was designed that would prevent amino terminal methionine cleavage during peptide expression in transgenic plants. ESF1B was designed by replacing the alanine at position two in ESF1 ( Powell et al. 1995 ), with a valine ( Table 1). This change did not significantly change the MIC as compared with the other most active ESF peptides in this study ( Table 2). However, ESF1B did have one of the highest haemolytic activities of the peptides tested ( Table 3). The haemolytic activity of ESF1B was compared with two other 20-amino acid peptides, ESF5 and ESF6. ESF1B had the highest haemolytic activity, followed by ESF6 and ESF5. The hydrophobicity, alpha moment or overall charge ( Table 1) of the three peptides does not appear to be associated with these haemolytic differences. The only discernible difference in these three peptide designs that follow the change in haemolytic activity was the total number of highly hydrophobic residues. ESF1B has six (1 Val, 5 Leu) as compared with ESF6 that has five (5 Leu) and ESF5 which has four (4 Leu). In other studies of synthetic antimicrobial peptides, increases in hydrophobicity have been associated with increases in haemolytic activity ( Dathe et al. 1997 ; Weiprecht et al. 1997 ). However, our data show that in some cases a more hydrophobic peptide of equal size can be less haemolytic. For example, ESF6 is more hydrophobic than ESF1B, yet it is less haemolytic. ESF6 is more hydrophobic because it contains more alanines (H = 0·62) in place of some hydrophilic amino acids present in ESF1B. But ESF1B has a greater number of highly hydrophobic amino acids such as valine (H = 1·10) even though its overall hydrophobicity is less then ESF6. Therefore, from these results we find it is more important to limit the number of highly hydrophobic amino acids than the overall hydrophobicity when trying to reduce haemolytic activity in antimicrobial peptide designs.

ESF1B had a relatively high haemolytic activity (73% as lytic as the 0·1% triton X-100 control) when tested at a concentration of 250 µmol l−1. However, when the concentration of ESF1B was reduced to 25 µmol l−1, the relative haemolytic activity was not significantly different from PBS buffer control. This concentration (25 µmol l−1) is approximately a magnitude higher than ESF1B's average MIC (1·9–6·3 µmol l−1 for the fungal pathogens) and therefore indicates this might still be a useful antimicrobial peptide in some applications.

To employ antimicrobial peptides in self-processing polypeptide systems in plants, they must be designed to be active despite containing residual amino acids left by the processing reaction. As a model system, we chose to design peptides containing residues from NIa protease recognition site Glu-X-X-Tyr-X-Gln/(Gly or Ser)( Carrington and Dougherty 1988; Dougherty et al. 1988 ). This recognition sequence has the advantage of placing a glutamic acid on the hydrophobic side of the amphipathic alpha helix, similar to the natural magainins. This configuration has been suggested to influence the magainin's differential lytic activity between animal and bacterial cells, possibly due to interaction with cholesterol ( Tytler et al. 1995 ). Our first peptide design to incorporate these residues was ESF22A ( Table 1). The design incorporated all the amino acids of ESF12 except the amino terminal methionine and alanine. In addition, amino acids were added to the carboxyl end that fit the recognition site pattern while still maintaining the amphipathic alpha helical structure of the overall peptide. ESF22A retained a high antimicrobial activity ( Table 2), indicating that a NIa protease recognition site could be incorporated with no significant effect on this activity. However, the haemolytic activity of ESF22 was among the highest activities of the peptides tested in this study ( Table 3). To reduce this activity, ESF22B ( Table 1) was designed by changing the eighth and fifteenth amino acids from an alanine and leucine, respectively, to arginines. This increased the positive charge, decreased the overall hydrophobicity, and lowered the number of highly hydrophobic amino acids. With these changes, ESF22B retained its antimicrobial activity to all of the plant pathogens tested and its haemolytic activity decreased significantly ( Tables 2 and 3). This demonstrated that if a new peptide design has a high haemolytic activity, small changes in its sequence could be made to reduce the haemolytic activity without compromising its antimicrobial activity.

From these experiments, we conclude that small antimicrobial peptides can be designed that incorporate self-processing recognition site residues. When designing these peptides, there is flexibility in the size and amino acid sequence used with respect to antimicrobial activity. However, to minimize haemolytic activity, the size and number of highly hydrophobic amino acids should be kept to the minimum needed to maintain antimicrobial activity. Other factors, as yet unidentified, must also contribute to the haemolytic activity because the control peptide (Ala8,13,18) magainin 2, had a longer amino acid sequence and also had more highly hydrophobic amino acids than the ESF peptides and yet had only a moderate haemolytic activity. These factors could not be identified in this study due to many differences in the amino acid sequence. Our next step is to design a gene encoding an antimicrobial peptide precursor using the ESF22 peptide design and test its ability to be processed in vitro and in vivo by the TEV NIa protease.


This research was funded in part by the USDA Cooperative State Research Service, McIntire-Stennis Program, and the New York Chapter of the American Chestnut Foundation.