Structural Dissection of a Highly Knotted Peptide Reveals Minimal Motif with Antimicrobial Activity*

The increasing occurrence of bacterial resistance to antibiotics is driving a renewed interest on antimicrobial peptides, in the hope that understanding the structural features responsible for their activity will provide leads into new anti-infective drug candidates. Most chemical studies in this field have focused on linear peptides of various eukaryotic origins, rather than on structures with complex folding patterns found also in nature. We have undertaken the structural dissection of a highly knotted, cysteine-rich plant thionin, with the aim of defining a minimal, synthetically accessible, structure that preserves the bioactive properties of the parent peptide. Using efficient strategies for directed disulfide bond formation, we have prepared a substantially simplified (45% size reduction) version with undiminished antimicrobial activity against a representative panel of pathogens. Analysis by circular dichroism shows that the downsized peptide preserves the central double α-helix of the parent form as an essential bioactive motif. Membrane permeability and surface plasmon resonance studies confirm that the mechanism of action remains unchanged.

The increasing occurrence of bacterial resistance to antibiotics is driving a renewed interest on antimicrobial peptides, in the hope that understanding the structural features responsible for their activity will provide leads into new anti-infective drug candidates. Most chemical studies in this field have focused on linear peptides of various eukaryotic origins, rather than on structures with complex folding patterns found also in nature. We have undertaken the structural dissection of a highly knotted, cysteine-rich plant thionin, with the aim of defining a minimal, synthetically accessible, structure that preserves the bioactive properties of the parent peptide. Using efficient strategies for directed disulfide bond formation, we have prepared a substantially simplified (45% size reduction) version with undiminished antimicrobial activity against a representative panel of pathogens. Analysis by circular dichroism shows that the downsized peptide preserves the central double ␣-helix of the parent form as an essential bioactive motif. Membrane permeability and surface plasmon resonance studies confirm that the mechanism of action remains unchanged.
Antimicrobial peptides (1)(2)(3)(4) are important constituents of the innate immune system (5) of most organisms and have been postulated as one of the most ancient weapons devised by evolution to fight bacterial infections (3). Their swift mobilization in the early stages of microbial invasion fulfills a crucial role in host defense, before the onset of cell-mediated immune response, which occurs slowly compared with microbial proliferation (6,7). The increasingly high number of bacteria that are developing resistance to classical antibiotics (8 -10) drives much of the current interest on antimicrobial peptides, in the hope that understanding the structural features responsible for the activity of these natural products (11)(12)(13)(14) may provide useful leads into new anti-infective drug candidates. Such expectations are sustained on the relatively simple mechanism of action of antimicrobial peptides that, in contrast to classical antibiotics, involves simple disruption of microbial plasma membranes rather than complex intracellular targets and pathways. For pathogens to develop resistance to membraneacting peptides, substantial membrane redesign would be required, a non-affordable solution for most species.
Following earlier studies on linear antimicrobial peptides (15,16), we have recently focused our attention on plant thionins (17), the first family of antimicrobial peptides for which in vitro activity against plant pathogens (18) and a defensive role (19) were reported. Despite their early discovery, the potential of thionins and other plant antibiotic peptides (20) in medicine remains almost unexplored because most work on membrane active peptides has focused, until recently (21) on peptide families isolated from mammals, amphibians or insects (bbcm.univ.trieste.it/ϳtossi/pag1.htm contains a complete collection of gene-encoded antimicrobial peptides). Thionins are basic, 45-47-residue long, highly folded, disulfide-rich peptides classified into five structural types on the basis of amino acid sequence and cysteine pairing (22). Despite some inter-species diversity, all thionins share an almost identical folding pattern consisting of a short ␤-sheet and a pair of antiparallel ␣-helices connected through a ␤-turn (Fig. 1). Disulfide bonds play a crucial role in stabilizing thionin folding, providing a characteristic amphipathic distribution of residues that is thought to be responsible for their ability to strongly interact and disrupt microbial and model membranes (23).
There have been some attempts to decipher generally conserved structural motifs on antimicrobial peptides (26). We describe here our search for a minimal active structure for the thionin from Pyrularia pubera (PpTH), 1 a highly knotted peptide representative of this cysteine-stabilized family of membrane-active peptides. Our dissection of the native PpTH structure has led to several shortened versions, and eventually shown that the central antiparallel double helix of PpTH is fully active against representative plant pathogens. We have used surface plasmon resonance (SPR) to show that the affinity of this mini-thionin for artificial membranes is comparable with that of the native peptide, and demonstrated that the ability to bind model membranes correlates with antimicrobial activity. Experiments with the SYTOX Green dye have shown that all active peptides permeabilize in a similar way the plasma membrane of the fungal plant pathogen Fusarium oxysporum. We also show that a well defined secondary structure, prior to membrane binding, is required for the peptide to display any relevant antimicrobial activity. Our results are discussed in the light of recent advances in the understanding of the mechanism of action of antimicrobial peptides.

Chromatography and Mass Spectrometry
Analytical reversed-phase HPLC was done on a Shimadzu LC-2010A system using a Phenomenex Luna C8 column (3 m, 0.46 ϫ 5 cm). Separations were performed using a linear gradient (10 -40%) of buffer B into buffer A over 15 min at a flow rate of 1 ml/min. Buffer A was 0.045% trifluoroacetic acid in water. Buffer B was 0.036% trifluoroacetic acid in acetonitrile. Preparative HPLC runs were performed on a Shimadzu LC-8A instrument using a Phenomenex Luna C8 column (10 m, 2.1 ϫ 25 cm) eluted with the above gradient over 60 min at a flow rate of 25 ml/min. Mass spectrometric analysis of peptides was performed on a Voyager DE-STR instrument (Applied Biosystems, Foster City, CA) using ␣-cyano-4Ј-hydroxycinnamic acid as matrix.
Two-disulfide Peptides-PpTH-  and PpTHR-(7-32) (Table I), two 26-residue analogues, were prepared by stepwise disulfide formation (28) from linear precursors assembled by Fmoc solid phase methods on Rink amide-p-methylbenzhydrylamine resin as above, except that Cys residues (6 and 25) corresponding to positions 12 and 31 of the native sequence were protected with the acetamidomethyl (Acm) group. Deprotection and cleavage from the resin with trifluoroacetic acid/ triisopropylsilane/water/phenol/EDT (82.5:5:5:5:2.5, 2 h at room temperature) gave partially protected (Cys 6,25 (Acm)) derivatives that, after preparative HPLC purification, were dissolved at 15 M concentration in 0.1 M Tris, pH 8.0, and stirred under air until oxidation was shown to be complete by both the Ellman test and MALDI-TOF MS analysis of the major HPLC peak. The solution was then made up to 15% acetic acid and the Acm groups were removed simultaneously with Cys oxidation by dropwise addition of 4 eq of iodine (freshly prepared 5 mM solution in methanol). HPLC analysis showed the oxidation to be complete after 10 min. The reaction mixture was treated with sodium thiosulfate to stop the oxidation and directly loaded onto a preparative HPLC column. Fractions containing Ͼ95% (by analytical HPLC) of the expected double disulfide products were pooled and further characterized by amino acid analysis and MALDI-TOF MS.
used. The dithiol precursor of PpTH-(15-28) was air-oxidized at 10 M concentration in 0.1 M ammonium bicarbonate, pH 8.1, for 24 h. Oxidation was quenched by trifluoroacetic acid addition to pH 1.5-2. All peptides were purified by preparative HPLC to give products with amino acid analyses and MS consistent with theory.

Circular Dichroism
CD spectra in the 190 -260-nm range were acquired on a Jasco J715 spectropolarimeter purged with nitrogen (25 ml/min) and using a quartz 0.1-cm path length cell stabilized to Ϯ0.1°C of the desired temperature by a Peltier controller. Spectra of all peptides were recordered at 15 M concentration in 25 mM phosphate, pH 6.0, in the absence and presence of small unilamellar vesicles of dimyristoyl phosphatidylglycerol (DMPG) at a peptide/lipid ratio of 1:100. Small unilamellar vesicles were prepared by dissolving dry DMPG in chloroform/ methanol (2:1), removing the solvents in a rotary evaporator, and hydrating the residue (to 3 mM DMPG concentration) in 25 mM phosphate, pH 6.0, for 1 h at room temperature. The suspension was first mixed in a vortex shaker, then sonicated until clear.

Antimicrobial Activity
The antimicrobial tests were done in sterile 96-well microplates by mixing different amounts of the peptides dissolved in 66.7 l of sterile water with 33.3 l of bacterial suspension (final concentration 10 4 colony forming units/ml) in nutrient broth or TY medium (R. meliloti (29)), or 33.3 l of F. oxsyporum spore suspension (final concentration, 10 4 spores/ml) in potato dextrose. Microorganisms were incubated at 28°C with periodic agitation and growth was recorded 24 -48 h later by measuring absorbance at 490 nm in an enzyme-linked immunosorbent assay plate reader.

SYTOX Permeation Experiments
For SYTOX green uptake different amounts of the peptides dissolved in 25 l of sterile water were mixed with 12.5 l of F. oxsyporum spore suspension (final concentration 2 ϫ 10 4 spores/ml) in potato dextrose, the microplate was incubated at 28°C for 6 h, and 0.75 l of SYTOX Green (0.2 M final concentration) was added to the wells. After 18 h of additional incubation fungal hyphae fluorescence was viewed with a microscope (Zeiss AxioPhot, Germany) equipped with a B-2A filter set for fluorescence detection (excitation wavelength, 450 -490 nm; emission wavelength, 520 nm). Light and fluorescence microscopic images were taken with a video camera (SPOT33, Diagnostic Instrument Inc.) and analyzed using the SPOT33 2.2 software.

Surface Plasmon Resonance
Measurements were carried out on a BIAcore 3000 instrument (Biacore, Uppsala, Sweden) using a L1 sensor chip with a carboxymethylated dextran matrix modified with lipophilic units to capture lipid bilayer vesicles. Phosphate-buffered saline, pH 6.8, was used as run- ning buffer and regeneration and washing solutions were 10 mM NaOH and 40 mM N-octyl-␤-glucopyranoside, respectively. All solutions were freshly made, degassed, and filtered through a 0.22-m filter. Large unilamellar vesicles (100 nm) with a 3:1 DMPC/DMPG ratio were prepared by dissolving the dry lipids in chloroform/ methanol (2:1). After solvent removal by rotary evaporation, the lipids were suspended in phosphate-buffered saline, pH 6.8, by sonication, and liposomes of the desired size were obtained by sequentially extruding the resultant emulsion through 0.8-, 0.4-, 0.2-, and 0.1-m filters (5 passages through each filter). Prior to immobilization of the DMPC/DMPG large unilamellar vesicles on the L1 chip, an injection of 25 l of 40 mM N-octyl-␤-glucopyranoside at 5 l/min flow rate was performed to clean the surface. Liposomes were applied to the sensor surface at a flow rate of 2 l/min. Between each liposome injection, a 10 mM NaOH pulse was applied to remove multilamellar structures. Once the desired immobilization level (above 5000 response units) was reached, the active surface was stabilized by repeated injections of 10 mM NaOH and buffer until a stable baseline was obtained. Further details are provided as Supplementary Materials.
Peptides were dissolved in phosphate-buffered saline at concentrations in the 5-100 M range and injected (40 l) at 10 l/min flow rate on the liposome-loaded sensor chip. Peptide solution was next replaced by running buffer and the peptide-bilayer complex was allowed to dissociate for 5 min. Because peptides bind very tightly to the liposomes, no complete regeneration of the active surface was possible in this way and full removal of the immobilized bilayer had to be achieved by pulsing with 10 mM NaOH and 10 mM HCl (50 l each) followed by a 30-l wash with 40 mM N-octyl-␤-glucopyranoside at 5 l/min flow rate.
Sensorgrams for each peptide-lipid interaction were analyzed by curve fitting using numerical integration algorithms in the BIAevaluation 3.0 software package. Recorded sensorgrams at seven different concentrations were globally fitted using a two-state binding model, as poorer fitting was obtained using simpler models (e.g. 1:1 Langmuir binding). Equations used in the fitting process are detailed under Supplementary Materials.

RESULTS
Peptide Design and Synthesis-To elucidate which structural motifs are responsible for the antimicrobial activity of PpTH ( Fig. 1), several peptides reproducing different regions were designed (Table I). PpTH-(3-41) is a shortened version of PpTH, with both N-and C-terminal deletions but preserving the four disulfide arrangement. PpTHR-(3-41) is a PpTH-(3-41) analogue where the Asp 32 residue is mutated to Arg, a modification shown to significantly enhance the antimicrobial activity of full-length PpTH against Gram-negative bacteria (17). Other peptides were meant to reproduce secondary structure elements such as each one of the helical segments (PpTH- (7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18)(19) and PpTH-(24 -32)), the intervening loop (PpTH (15-28)), or the double antiparallel ␣-helix motif (PpTH-(7-32)), which given its amphipathic character, might be a key feature for antimicrobial activity. For all analogues containing the Asp residue at position 32 of the native sequence, the corresponding Arg replacement analogue was also synthesized to explore whether the positive effect of this mutation was maintained throughout the series. Disulfide pairings corresponding to native PpTH were preserved in all peptides except PpTH-(7-32)b, where the alternative connectivity was explored to evaluate the importance of disulfide bridges for the secondary structure and activity of this particular structure.
Circular Dichroism-CD spectra of PpTH-(3-41), PpTH-(7-32) and their Arg 32 analogues PpTHR-  and PpTHR-  in aqueous solution were almost identical to those of native PpTH and consistent with a mainly ␣-helical structure coexisting with some levels of ␤-type structure. Upon binding to anionic DMPG liposomes, the global pattern of CD signatures was maintained, with slight decreases in helical content for the larger PpTH-  and PpTHR-  peptides, whereas the smaller PpTH-(7-32) and PpTHR-(7-32) reinforced their helical character (Fig. 3, Table II). Temperature studies for these four peptides in the 5-85°C range showed the helical pattern mostly preserved throughout the interval, as for native thionin (17) (see Supplementary Materials). At elevated temperatures slight decreases in helical content were detected; globally, the temperature dependence pattern of molar ellipticity at 207 nm fitted with a non-cooperative thermal unfolding pathway.
Antimicrobial and Permeabilizing Activity-In vitro activities of the synthetic peptides against one Gram-positive and two Gram-negative bacteria, and a fungal plant pathogen are  shown in Table III. As previously described (17) PpTHR was significantly more active than PpTH against the Gram-negative bacteria, whereas activity against the Gram-positive organism and the fungus were comparable. The shortened PpTH-  and PpTHR-(3-41) analogues were undistinguishable from their full-size counterparts in activity against Clavibacter and Rhizobium but slightly less active against Fusarium and Xanthomonas. Again, the Arg 32 substitution increased activity against all Gram-negative bacteria. Substantially shorter PpTH-(7-32) and PpTHR-(7-32) peptides were only slightly less active than the full-length parent structures against the pathogens tested, whereas the misfolded PpTH-(7-32)b was completely inactive against Gram-negative and fungal pathogens and marginally active against Gram-positive bacteria. The remaining peptides did not inhibit the growth of Gramnegative bacteria nor F. oxysporum and only PpTHR-(24 -32) and PpTH- (7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18)(19) were slightly active against C. michiganensis.

DISCUSSION
Thionins are the first eukaryotic peptides for which antimicrobial activity against plant pathogens was demonstrated in vitro (18). Several other families of cystein-rich plant peptides have been characterized (22,36) including plant defensins, lipid transfer proteins, hevein, and knottin-type peptides, and more recently snakins (37). Thionins alter the permeability and fluidity of microbial membranes and promote ion leakage through the formation of cation-selective channels (38). In an attempt to understand the structural basis for thionin activity and to define the minimal structural motif with relevant antimicrobial properties, we designed peptides mimicking several elements of P. pubera thionin (PpTH) secondary structure as well as shortened versions of PpTH to evaluate the role of its Nand C-terminal sections on antimicrobial activity. Synthetic approaches to multiple disulfide peptides (39,40) were carefully adapted to obtain products with the desired native connectivity. For instance, in the oxidative procedure used for large PpTH analogues, the minimally denaturing conditions favor a folding process that drives formation of native-like disulfide connectivities. On the other hand, the substantially downsized two-disulfide thionins were not accessible by this procedure and required a chemoselective disulfide formation strategy.
Comparison of CD spectra, antimicrobial activities, and SY-TOX Green influx was illustrative, in the sense that only peptides with well defined secondary structures, both in buffer and upon binding to model membranes, had biological activities akin to native thionin. Moreover, all fully active peptides had almost identical CD curves, with high helical contents and minimal temperature dependences. A possible interpretation for a membrane activity that does not require conformational change upon membrane binding has been advanced (7). As the rigid thionin structure makes it unlikely for the peptide to modify its polar/non-polar residue distribution through environment-induced conformational changes, a fact borne out by CD data, a possible explanation could be that thionins have environment-dependent association models. Thus, in aqueous solution self-association would cause hydrophobic residues to be buried in the core of the oligomer whereas, on reaching the membrane, the organizational pattern would be reversed and oligomers with hydrophobic residues exposed to the lipid membrane moieties would be formed, whereas hydrophilic regions would face the newly generated channel.
In marked contrast with the above, peptides PpTH-(7-32)b and PpTH- (15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28), which are random coil in water, can be induced into helical conformation in the presence of DMPG and display minimal antimicrobial activity. Cys pairing in the misfolded PpTH-(7-32)b does not force it into the antiparallel double helix active conformation (see below), but it is reasonable to assume that in the presence of lipid bilayers it might be able (as its disulfide bonds do not geometrically prevent it) to adopt a conformation capable of disrupting membranes and killing microorganisms, although at much higher concentrations, as seen here for C. michiganensis. The other analogue, PpTH- (15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28), is shown by SPR studies to interact very weakly with membranes and is only marginally active as an antibiotic. All this would suggest that, for this particular group of antimicrobial peptides, biological activity and conformational plasticity are not associated in a simple way. This agrees with previous studies (7,41) showing poor correlation between peptide activity on either model membranes or intact bacteria, when the oligomerization state or volume of the peptide is crucial for the pore formation event (42).
Remarkably, the two shortened thionins, PpTH-(7-32) and PpTHR-(7-32), that we had designed to reproduce the antiparallel double ␣-helix core, retain the activity of full-length thionins against most tested organisms (C. michiganensis, R. meliloti, and F. oxysporum). The double helix folding, stabilized by the native-like Cys 6 -Cys 25 and Cys 10 -Cys 21 pairings, defines a characteristic amphipathic pattern and also contains residues shown to play a crucial role modulating thionin activity. Thus, iodination of Tyr 13 of PpTH leads to a loss of toxicity (43,44), suggesting that this residue is involved in the formation of transmembrane ion channels. Another sensitive position also present in our active, minimal versions of PpTH, is residue 32, the mutation of which (from Asp to Arg) significantly enhances activity against Gram-negative bacteria (17). The impact of this mutation is maintained along the PpTHR analogue series as long as the double helix folding pattern is preserved. Interestingly, even the tiny, 9-residue analogue PpTHR- (24 -32), with only one of the two helices, is able to inhibit C. michiganensis growth at micromolar concentrations.
SPR studies also support that the double ␣-helix core is the structural motif responsible for the ability of thionins to bind to a Obtained by numerical integration of SPR data using a two-state reaction model. k on 1 and k off 1 are respectively the association and dissociation kinetic constants of the first step of the interaction; k on 2 and k off 2 refer similarly to the second step. K is the thermodynamic affinity constant (see Supplementary information).
In conclusion, the present study shows that structure-guided deconstruction of a complex, highly knotted bioactive peptide such as PpTH can be used to reveal the main features responsible for its membrane-permeating, antibiotic activity. This information is not only useful in understanding the mechanism of action of this and similar families of antimicrobial peptides but can also be incorporated into the design of minimalist versions of such peptides with therapeutic potential.