Selective reciprocity in antimicrobial activity versuscytotoxicity of hBD-2 and crotamineNannette Y. Younta,b, Deborah Kupferwassera,b, Alberto Spisnic, Stephen M. Dutzd, Zachary H. Ramjand,Shantanu Sharmad, Alan J. Waringe, and Michael R. Yeamana,b,e,1

aDivision of Infectious Diseases, LAC-Harbor University of California, Los Angeles Medical Center, Torrance, CA 90509; bSt. John’s Cardiovascular ResearchCenter, Los Angeles Biomedical Research Institute at Harbor–University of California, Los Angeles, Torrance, CA 90502; cDepartment of ExperimentalMedicine, Section of Chemistry and Structural Biology, University of Parma, 43100 Parma, Italy; dDepartment of Chemistry and Center for MacromolecularModeling and Materials Design, California State Polytechnic University, Pomona, CA 91768; and eDepartment of Medicine, David Geffen School of Medicineat University of California, Los Angeles, CA 90024

Edited by H. Ronald Kaback, University of California, Los Angeles, CA, and approved June 24, 2009 (received for review April 23, 2009)

Recent discoveries suggest cysteine-stabilized toxins and antimi-crobial peptides have structure–activity parallels derived by com-mon ancestry. Here, human antimicrobial peptide hBD-2 and rat-tlesnake venom-toxin crotamine were compared in phylogeny, 3Dstructure, target cell specificity, and mechanisms of action. Resultsindicate a striking degree of structural and phylogenetic congru-ence. Importantly, these polypeptides also exhibited functionalreciprocity: (i) they exerted highly similar antimicrobial pH optimaand spectra; (ii) both altered membrane potential consistent withion channel-perturbing activities; and (iii) both peptides inducedphosphatidylserine accessibility in eukaryotic cells. However, theNav channel-inhibitor tetrodotoxin antagonized hBD-2 mecha-nisms, but not those of crotamine. As crotamine targets eukaryoticion channels, computational docking was used to compare hBD-2versus crotamine interactions with prototypic bacterial, fungal, ormammalian Kv channels. Models support direct interactions of eachpeptide with Kv channels. However, while crotamine localized toocclude Kv channels in eukaryotic but not prokaryotic cells, hBD-2interacted with prokaryotic and eukaryotic Kv channels but did notocclude either. Together, these results support the hypothesis thatantimicrobial and cytotoxic polypeptides have ancestral structure-function homology, but evolved to preferentially target respectivemicrobial versus mammalian ion channels via residue-specific in-teractions. These insights may accelerate development of anti-infective or therapeutic peptides that selectively target microbialor abnormal host cells.

channel � defensin � host defense � toxin

Cysteine-stabilized antimicrobial polypeptides are thought tofunction as a first line of host defense. One of the most

well-characterized groups of such molecules is the �-defensinsfrom myeloid and epithelial tissues of mammalian and avianspecies. The �-defensins contain a highly conserved cysteine-array that affords structural rigidity and promotes hypervari-ability for accelerated evolutionary adaptation.

Several parallels suggest structural, functional, and evolutionarycommonalities between defensins and other host defense and/oroffense molecules, such as toxins. Like antimicrobial peptides, manytoxins are small, cysteine-stabilized, and cationic (1, 2), and sharea striking degree of conservation with host defense peptides thatcontain a �-core motif (3). One group of toxins with particularlyclose homology to �-defensins is the crotamine-myotoxin familyfrom South American rattlesnakes (4, 5). As in �-defensins, toxinexpression is often targeted for mucosal or extracorporeal secretion(6). Such toxins typically induce rapid paralysis resulting in localmyonecrosis in skeletal muscle. Crotamine-like toxins are thoughtto induce such effects through targeting of ion channels and alteringmembrane transport or conductivity (5).

Understanding structural and mechanistic reciprocity is of directrelevance to immunologic roles and potential therapeutic develop-ment of antimicrobial and cytotoxic peptide-based therapies. Priorstudies have noted overall diversity in primary structures of �-de-

fensin and crotamine-like toxin families, but conserved cysteine-array and �-core motifs (4). In the present study, evolutionary,structural, and mechanistic investigations were performed to ad-dress the hypothesis that reciprocal relationships exist betweenantimicrobial and cytotoxic host defense peptides.

ResultsPhylogenetic, structural, mechanistic, and target interactioncomparisons between hBD-2 and crotamine were investigatedthrough complementary methods.

Phylogenetic Analysis. Crotamine and related serpentine toxinsformed a monophyletic group with robust statistical confidence(Fig. 1). A single subclade in this group was represented by crotasin,a �-defensin from the nonvenomous somatic tissues of Crotalusdurissus terrificus (7). A sister group was comprised of 2 crotamine-like proteins (CLP) from the venom of the bearded dragon Pogonabarbata (6). The next most closely related sequences are �-defensinsfrom various avian species. Notably, the serpentine and aviansequences form a unified clade within the Sauropsida. Thesefindings suggest a divergence of peptides optimized for antimicro-bial versus cytotoxic functions appeared concomitant with thedivergence of synapsids (mammals) from sauropsids (aves/reptiles)(SI Text and Figs. S1 and S2).

Structural and Biophysical Comparison. Three-dimensional (3D)alignment between hBD-2 and crotamine revealed a striking degreeof identity (Fig. 2A). The greatest degree of 3D alignment (RMSD�2) occurs between respective �-helical and �-core regions. Thus,evolutionary selective pressures have favored conservation of 3Dstructure in the face of limited sequence identities (28%) of the 2peptides (2). Despite striking conformational homology, biophys-ical analyses revealed significant physicochemical differences be-tween hBD-2 and crotamine that likely relate to differences intarget preference. For example, the solvent accessible surface areaof hBD-2 (�70% lacking charged residues; Fig. 2B) is markedlymore hydrophobic than that of crotamine.

Antimicrobial Activity. Crotamine antimicrobial activities paralleledthose of hBD-2 at both experimental pH values against most

Author contributions: N.Y.Y. and M.R.Y. designed research; N.Y.Y., D.K., S.M.D., Z.H.R., S.S.,A.J.W., and M.R.Y. performed research; N.Y.Y., A.S., S.S., and M.R.Y. contributed newreagents/analytic tools; N.Y.Y., S.M.D., Z.H.R., S.S., A.J.W., and M.R.Y. analyzed data; andN.Y.Y., A.S., S.S., A.J.W., and M.R.Y. wrote the paper.

Conflict of interest statement: M.R.Y. is a shareholder of NovaDigm Therapeutics, Inc., andhas received research funding from Pfizer, Inc., Amgen, Inc., Cubist Pharmaceuticals, andNovozymes Pharmaceuticals. None of these entities provided support for the currentstudies.

This article is a PNAS Direct Submission.

1To whom correspondence should be addressed. E-mail: MRYeaman@ucla.edu.

This article contains supporting information online at www.pnas.org/cgi/content/full/0904465106/DCSupplemental.

www.pnas.org�cgi�doi�10.1073�pnas.0904465106 PNAS Early Edition � 1 of 6

IMM

UN

OLO

GY

Dow

nloa

ded

by g

uest

on

Aug

ust 3

, 202

0

organisms (P � 0.05; Fig. 3). The only exception to this observationwas for the prokaryote Staphylococcus aureus at pH 7.5, wherehBD-2 had significantly greater efficacy (P � 0.05). Notably, bothcrotamine and hBD-2 had marked activity against the eukaryoticpathogen Candida albicans at pH 7.5 or 5.5. Generally, hBD-2 hadgreater efficacy at pH 7.5, except against C. albicans, where pH hadno discernable impact.

Cytotoxic Properties. To assess relative cytotoxicities of crotamineversus hBD-2, flow cytometry was used to compare membraneelectrophysiology (��), permeabilization, and phosphatidylserineaccessibility in bacteria and 2 eukaryotic cell systems.

Membrane Energetics. Membrane potential (��) was evaluatedusing 3,3-dipentyloxacarbocyanine (DiOC5), a charged lipophilicdye that emits a fluorescent signal proportional to ��. The validityof this method to assess channel activity approaches the reliabilityof classical patch clamp methods, as borne out by a compellingbody of evidence (8–10). Overall, hBD-2 and crotamine altered�� in several target organisms versus untreated controls (Fig. 4A–D, Figs. S3 and S4, and SI Text). In Escherichia coli, hBD-2caused cell membrane hyperpolarization, as evidenced by in-creases in the overall percentage of cells polarized, as well as

BNBD12_BOSTABNBD8_BOSTA

BNBD3_BOSTATAP_BOSTA

LAP_BOSTABD402_BOSTASBD2_OVIAR

SBD1_OVIARBD1_CAPHIBD2_CAPHI

EBD_BOSTANBD4_BUBBUBNBD5_BOSTA

BNBD4_BOSTAMBD8_MUSMU

RBD3_RATNORBD5_RATNO

MBD4_MUSMUBD2_MACMUBD2_PANTRBD2_HOMSABD1_SUSSC

BD1_EQUCABD1_CHILA

BD103_CANFABD3_SUSSCBD7_GALGA

BD_ANAPLBD2_GALGAGAL2_GALGABD5_GALGAGAL9_GALGA

GAL1_GALGATHP1_MELGA

CROTASIN_CRODUCLPPOGL3_POGBA

CLPPOGL2_POGBA

MYO3_CRODUCROT1_CRODU

CROT3_CRODUMYO_CRODUMYOA_CROVIVMYO2_CROVIC

MYO3_CROVIVMYO1_CROVIC

BD4_GALGAGAL12_GALGABD8_GALGA

Artiodactylββ-defensins

Rodent β-defensins

Primate β-defensins

Mammalian β-defensins

Avian β-defensins

Serpentine Toxins

Avian β-defensins

SynapsidaSauropsida

Reptilian Toxins

Fig. 1. Phylogenetic parallels between crotamine and hBD-2. Neighbor-joiningtree visualization (36) of the hBD-2/crotamine polypeptide family. Branch signif-icance was validated by bootstrap analysis.

E

HBD-2 Crotamine

A

B

C

D

180o

90o

B

HBD-2 CROA

Fig. 2. Structure and biophysical comparison of crotamine and hBD-2. (A) 3Dalignment of crotamine/hBD-2 was performed by combinatorial extension (34).Coloration is per secondary structure schema: �-Sheet (blue); turn (gray); �-helix(red),molecularvisualizationMOLMOL(42). (B)Biophysicalparallels incrotamineand hBD-2. (A) �-Core domain, in red, hBD-2 and crotamine; (B) partial transpar-ency of the �-core domain (red); (C) Kyte-Doolittle hydrophathy plot overlay:Brown, most hydrophobic; green, intermediate; blue, most hydrophilic; (D)charge overlay: Blue, basic (Arg, Lys); red, acidic (Asp, Glu); (E) visualization ofcationic (blue) and hydrophobic (Kyte-Doolittle coloration, as above) residueswithin the �1-�2 loop for hBD-2 and �2-�3 loop for crotamine. Data are forcrotamine (1H5O) and hBD-2 (1FD3) as visualized using University of California,San Francisco (UCSF) Chimera (43).

2 of 6 � www.pnas.org�cgi�doi�10.1073�pnas.0904465106 Yount et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 3

, 202

0

their mean channel f luorescence (P � 0.05 versus control).Neither peptide had a detectable affect on membrane �� of S.aureus (Fig. 4B). In C. albicans, both hBD-2 and crotamine

caused marked increases in membrane potential (respective 2.2-and 2.8-fold increases in mean channel f luorescence; P � 0.01;Fig. 4C). Importantly, addition of the Nav channel inhibitortetrodotoxin (TTX) mitigated hBD-2-induced hyperpolarizationof E. coli, but had little impact on activities of hBD-2 onmembrane potential in other cells.

Membrane Permeabilization. Peptide permeabilization of cells wasmeasured using the intercalating dye propidium iodide (PI). Thisfluorescent probe enters permeabilized cells and binds to double-stranded nucleic acids, but is excluded from cells with normalmembrane integrity. Defensin hBD-2 caused marked permeabili-zation of E. coli and S. aureus (P � 0.01; Fig. 4 A and B).Interestingly, crotamine did not increase E. coli permeability, butsignificantly increased the ratio of permeated S. aureus cells (P �0.05; Fig. 4 A and B). Moreover, in the eukaryote C. albicans,hBD-2 or crotamine caused significant permeabilization (respective99- or 95-fold increases in percent of cells permeabilized; P � 0.01;Fig. 4C). In human umbilical vein endothelial cells (HUVECs),neither hBD-2 nor crotamine caused significant increases in cellpermeability versus controls (Fig. 4D). As with membrane ener-getics, the Nav channel blocker TTX revealed peptide-specificconsequences on target cell permeability. In bacteria, TTX causeda complete abrogation of cell permeabilization by hBD-2 (nosignificant difference from control). Conversely, TTX enhancedpermeabilization of S. aureus by crotamine (P � 0.05), but did notdo so for E. coli. In C. albicans, TTX only modestly increasedpermeabilization by hBD-2 or crotamine.

Fig. 4. Reciprocal activities of crotamine and hBD-2 on distinct prokaryotic and eukaryotic cells. Comparative flow cytometric analysis of membrane energetics(DiOC5),membranepermeabilization(PI),andphosphatidylserineaccessibility (annexinV) inAE.coli, (B)S.aureus, (C)C.albicans, and(D)HUVECs inresponsetopeptideexposure for 1 h. Lower left quadrant depicts basal fluorescence. Cells above the quadrant threshold are considered positive for permeability and phosphatidylserineaccess. For cellular energetics, movement relative to the control cell population indicates either hyper- or hypopolarization. Percent of cells gating positive are indicatedon each panel, with mean channel fluorescence provided in parentheses. Data were generated using FCS Express V3 software (De Novo Software).

Fig. 3. Comparative antimicrobial efficacies and pH optima of crotamine andhBD-2. Antimicrobial activity of crotamine and hBD-2 were determined usingradialdiffusionagainstapanelofGram-positiveandGram-negativebacteriaandfungi. Antimicrobial activity was measured at pH 5.5 and 7.5. Data are displayedas the zone of inhibition. Coloration: Complete clearing (blue); partial clearing(red). †, P � 0.05 vs. same peptide at alternate pH; *, P � 0.05 vs. alternate peptideat same pH.

Yount et al. PNAS Early Edition � 3 of 6

IMM

UN

OLO

GY

Dow

nloa

ded

by g

uest

on

Aug

ust 3

, 202

0

Phosphatidylserine (PS) Accessibility. One of the earliest markers ofcellular transition to an apoptotic state is the translocation of PSfrom the inner to the outer leaflet of the plasma membrane (11).There, PS is accessible to staining by fluorescent-labeled annexin V,a phospholipid-binding protein with specificity for PS. In bacteria,neither hBD-2 nor crotamine caused significant PS accessibility(Fig. 4 A and B). Likewise, hBD-2 did not lead to increased PSaccessibility in C. albicans in the presence or absence of TTX.However, crotamine did induce a substantial increase in PS acces-sibility in C. albicans, which was not affected by TTX (Fig. 4C).Interestingly, both study peptides influenced PS accessibility inHUVECs. In these cells, hBD-2 or crotamine alone caused signif-icant increases in PS accessibility (30-fold and 24-fold, respectively;P � 0.01; Fig. 4D). However, consistent with other mechanisticresults, TTX inhibited this effect of hBD-2, but not of crotamine.

Predicted Interactions with Kv Channels. Docking studies were per-formed to compare crotamine versus hBD-2 interactions withprokaryotic and eukaryotic Kv channels, including this model of theC. albicans Kv channel.

Crotamine. Crotamine preferentially targeted eukaryotic over pro-karyotic Kv channels [interaction energies (E; kcal mol�1): Kv1.2,�1119; CaKv, �618; KcsA, �355]. Moreover, crotamine interactedwith the eukaryotic Kv1.2 and CAKv channels so as to completelyocclude their apertures (Fig. 5). For Kv1.2, crotamine residues Arg31and Tyr1 participated in electrostatic and hydrogen-bonding inter-actions with Asp375 and Tyr373 of the channel pore. A similarinteraction was observed for crotamine and the fungal CAKvchannel, although the opposite facet of the peptide was involved inthis interaction. In this association, peptide residues Lys27 and Trp32occupied the channel via electrostatic and hydrogen bonding interac-tions with aspartic acid residue Asp452 of the channel pore. In contrast,the predicted interaction between crotamine and the prokaryoticKcsA channel did not occlude the channel pore (Fig. 5).

hBD-2. While its binding energies were similar, the orientation andlocalization of hBD-2 were distinguishable from those of crotamineand eukaryotic Kv channels at the molecular level. Notably, thehBD-2 backbone did not occlude either eukaryotic channel pore.Rather, hBD-2 targeted the outer region of these channels throughresidues of its �1-�2 loop (Arg22 and Phe19; Figs. 2 and 5). In theprokaryotic KcsA model, hBD-2 did not interpose the channelaperture, but localized to partially occlude this pore. Furthermore,the interaction between hBD-2 and KcsA was mediated by Arg22and Arg23 from the �1-�2 loop of hBD-2 (Figs. 2 and 5).

DiscussionPrior reports have noted structural homology among certainantimicrobial peptide and toxin classes (4, 12). However, struc-tural, mechanistic, and evolutionary relatedness among thesemolecules is less clear. If such reciprocal relationships exist, theycould provide vital insights into molecular origins of innateimmunity, overcome previous barriers to development of non-toxic peptide therapeutics, and reveal agents for targeting ab-normal host cells.

Structure-function reciprocity of hBD-2 and crotamine waspredicted based upon congruent conformational homology, pro-pensity for membrane disruption, proclivity for cationic charge atneutral pH, and affinities for electronegative microbial surfacetargets. Supporting these predictions, the current data demonstratehBD-2 and crotamine to have parallel yet distinct structure-activityrelationships: (i) Charge and hydrophobic topologies consistentwith conserved antimicrobial and cytotoxic functions; (ii) similarantimicrobial spectra and pH optima; (iii) prokaryotic (hBD-2) oreukaryotic (crotamine) preference with respect to membrane dis-ruptive activities; (iv) an ability to evoke asymmetric membranephosphatidylserine expression consistent with cytotoxic injuries

leading to programmed cell death; (v) a preferential trophism ofcrotamine for eukaryotic (CAKv or Kv1.2) and hBD-2 for prokary-otic (KcsA) channel targets, respectively; and (vi) antagonism ofhBD-2 for TTX-sensitive ion channels. Together, these findingssubstantiate the hypothesis that overall 3D structural homologyprovides a framework for reciprocal antimicrobial and cytotoxicpropensities of hBD-2 and crotamine, respectively. Moreover, these

B

C D

E

F

I

J

B

C

F

H

E

1Y

373Y

2K

375D

31R375D

10H

374G

22R375D 4D

350R

351D

23R

19F

374G

7K

452D

14K

452D

452D

27K

452D

35K

22R

452D 452D

25K

36K452D

452D

19F

31R

78Y 31R

79G

32W80D

1Y

84V

Cro

tam

ine

hBD

-2C

rota

min

ehB

D-2

Cro

tam

ine

Kv1.2

CA

KK

csAJ

E

J

A

G

H

A

D

G

I

A B

C D

hBD

-2 80D

23R

H

41P

80D

22R82Y

22R

79G

GF

I

Fig. 5. Computational modeling of predicted interactions between crotamineor hBD-2 with Kv1.2, CaKv, and KcsA Kv channels. Residue-specific interactionsbetweencrotamine (green; A–D)orhBD-2 (orange; I–L) andrespective tetramericKv channels (gray) backbones are visualized. Predicted ligand/receptor complexesare shown as ribbon models in top view orientations (E and J). Side chaincoloration is per backbone of the specific peptide and atomic elements thereof.Molecular interactions were visualized with Visual Molecular Dynamics (VMD)software (44).

4 of 6 � www.pnas.org�cgi�doi�10.1073�pnas.0904465106 Yount et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 3

, 202

0

results point to residue-specific interactions as potentially critical todifferentiating antimicrobial versus cytotoxic proclivities of thesepeptides.

Kv, NaKv, and Nav channels represent an evolutionary continuumof ion-selective channels bridging prokaryotes to eukaryotes (13–17). TTX, a specific inhibitor of Nav channel function (17), was usedto probe potential differential effects of crotamine and hBD-2 ontarget cell Nav-related functions. Generally, TTX mitigated effectsof hBD-2, but failed to inhibit crotamine functions, and in somecases enhanced crotamine action. This pattern of results suggestshBD-2 and crotamine preferentially target Nav, NaKv, or otherTTX-sensitive targets in eukaryotes, and orthologous targets inprokaryotes (13, 14). Whether such mechanisms may be direct orindirect or exclusively involve Nav or NaKv channels is not yet clear.It is conceivable that hBD-2 antagonizes cell functions distinguish-able from or in addition to TTX-sensitive Nav or NaKv channels.For example, hBD-2 and crotamine may perturb lipid membraneintegrity in relation to ion-channel function in certain target cells.Investigations to determine how hBD-2 or other host defensepeptides may antagonize voltage-sensitive ion channels or relatedtargets are ongoing. The observation that defensin-like peptidesmay interact with Nav channels is not entirely unprecedented.Rogachevskii et al. suggested an interaction between human de-fensin NP-1 and the slow Nav channels of rat ganglial neurons (18).Likewise, plant �-thionin defensins inhibit the sodium current incultured GH3 cells (19). If so, such interactions may begin to explaintoxicity in animal models wherein bolus systemic administration ofdefensin-like peptides leads to neurotoxic- and cytotoxic-like ef-fects (20–22).

Mechanistic findings above suggested biologically distinct inter-actions of hBD-2 or crotamine with Nav versus Kv channels.Therefore, as a complement to these studies, quantitative modelingwas used to compare hBD-2 versus crotamine interactions withprototypic Kv channels of prokaryote, fungal, and mammalianorganisms. Potential interactions between peptides and Nav chan-nels were not studied, as no NMR- or X-ray-validated structuremodels for these channels were available. Relative docking affini-ties may reflect the relative electrostatic attractions of cationiccrotamine and hBD-2 peptides to anionic Kv channel surfaces(most electronegative to least: Kv1.2 � CAKv � KcsA) or largermolecular areas of eukaryotic versus prokaryotic channels. How-ever, the finding that the peptides were distinct in TTX antagonismsuggests that specific stereogeometric interactions influence rela-tive targeting of hBD-2 and crotamine for specific ion channels.

Computational docking analyses suggested a preferential reci-procity of hBD-2 and crotamine for interactions with prokaryoticversus eukaryotic Kv channels, respectively. In these analyses,crotamine localizes to the inner-pore domain through a classicalcationic-aromatic (e.g., Lys27 and Trp32) functional dyad. Consis-tent with this theme is the fact that Lys27 occurs at the ‘‘X’’ positionof the GXnC element of the �-core domain in many defensin-liketoxins and is present within crotamine (GKMDC). However, a Lysresidue is absent from the GXnC element of the hBD-2 �-core motif(GTC). The current results also indicate that crotamine perturbseukaryotic Kv channels as do the charybdotoxin (23), kaliotoxin(24), and cobatoxin (25) family of scorpion toxins (26). Basicresidues in such toxins form a cationic facet integrating Lys or Argresidues of their �-core GXnC motif (2, 3). In turn, this cationicfacet binds to 4 highly conserved Asp452 residues symmetricallydistributed on the extracellular P-loops of Kv tetramers (27). Inaddition, a highly-conserved aromatic residue, typically Phe or Tyr,also participates in this pore-occluding complex, forming the clas-sical cationic-aromatic residue dyad (25). Collectively, the currentdata support the hypothesis that residue-specific differences inpeptides that have conserved overall 3D homology may contributeto target ion-channel preferences.

Many defensins exert considerable cytotoxicity in cell and animalmodels (21, 22). The present data suggest that such toxicities may

be attributable in-part to structure-mechanism correlates parallel-ing those of crotamine-like toxins. Crotamine causes rapid paralysisand myonecrosis following prey envenomation. Early studies sug-gested an interaction with voltage-dependent Nav channels pre-dominant in fast-twitch muscles of mammals (28). However, recentstudies by Rizzi et al. (29), using expressed Nav1.1–1.6 �-subunits,did not find an interaction between crotamine and Nav channels.Further support for the interaction of defensin-like toxins withKv ion channels derives from recent structural mapping of the Kvchannel surface and, in some cases, toxin/channel complexes(23–25). Together, these facts support the concept that hBD-2and crotamine may preferentially but not exclusively targetrespective Nav, NaKv, or other TTX-sensitive ion channels—versus Kv channels—contributing to their net cytotoxic effects.Ongoing studies are designed to assess the potential direct and/orindirect interactions through which such peptides may differen-tially target and antagonize these or other channels structurallyor mechanistically.

The present findings may have significant implications for de-velopment of polypeptide anti-infectives or other therapeutics. Inthe past decade, a sizeable investment has been made in exploitingmicrobicidal activities of CS-stabilized peptides such as defensins toaddress the mounting resistance of many important human patho-gens to conventional antibiotics. However, for systemic use, toxicitynot unlike that induced by cytotoxins has been a significant barrierto such advances (20–22). For example, the current data suggestthat crotamine and hBD-2 initiate apoptotic pathways in eukaryotictarget cells (Candida, HUVECS). Such events have been linked topeptide perturbation of ion channel functions in fungal as well asmammalian cells (11). Identification of molecular determinantsthat differentiate cytotoxicity from antimicrobial activity may en-able optimization of molecules for therapeutic efficacy withoutconcomitant host toxicity. Alternatively, engineering peptides tohave selective host cell toxicity may offer approaches to prevent ortreat cancer, autoimmune, or other diseases. Thus, a clearer un-derstanding of the ancient molecular features that both unite anddistinguish host defense, toxin, and venom peptides may aid indevelopment of anti-infectives or other therapeutics to address 21stcentury medical challenges.

Materials and MethodsMicroorganisms. Microorganisms representing Gram-positive (S. aureus; ATCC27217; and Bacillus subtilis; ATCC 6633); Gram-negative (E. coli; ML-35); andfungal (C. albicans; ATCC 36082) human pathogens were studied. Microorgan-isms were cultured overnight in brain heart infusion (BHI) broth (Difco) at 37 °C(bacteria) or 30 °C (fungi). Cells were sonicated and adjusted to 106 CFU/mL.

Endothelial Cells. Studies using HUVECs were conducted in accordance withNational Institutes of Health (NIH) and institutional guidelines for human sub-jects. HUVECs were harvested as described previously (30). Cells were cultured toconfluency (M-199 medium; Invitrogen; 10% FBS; Gemini Bio-Products; 2 mML-glutamine, penicillin, and streptomycin; Irvine Scientific), detached with 0.1%trypsin EDTA, washed, and enumerated (30).

Molecules. TTX (Sigma). hBD-2 (Peptides International). Crotamine was enrichedfrom yellow venom of Crotalus durissus terrificus as described previously (5).Crotamine purification was achieved by RP-HPLC on a C18 column (Vydac) equil-ibrated with 0.01% triflouroacetic acid and eluted with a 0–40% gradient ofwater:acetonitrile. Crotamine identity and purity were authenticated usingMALDI-TOF spectrometry.

Radial Diffusion Antimicrobial Assay. Antimicrobial assays were performed usinga radial diffusion method (3). As pH can influence peptide antimicrobial efficacy,the assays were conducted at pH 5.5 or 7.5 (31, 32). These conditions reflect therelevant contexts in which antimicrobial peptides often function (e.g., intracel-lular phagolysosome, pH 5.5, or extracellular milieu, pH 7.5) and which caninfluence peptide activities. Organisms were inoculated (106 CFU/mL) into buff-ered agarose (10 mM Pipes, pH 7.5, or 10 mM MES, pH 5.5); peptides (10 �g/well)were aliquoted into wells in the seeded matrix and incubated for 3 h at 37 °C.

Yount et al. PNAS Early Edition � 5 of 6

IMM

UN

OLO

GY

Dow

nloa

ded

by g

uest

on

Aug

ust 3

, 202

0

Zones of inhibition were measured 24 h later. Independent experiments wererepeated a minimum of 2 times.

Flow Cytometry. Multicolor flow cytometry was used to assess the comparativemechanisms of hBD-2 and crotamine versus prokaryotic (E. coli, S. aureus) oreukaryotic (C. albicans, HUVECs) target cells. The fluorophores used were asfollows: Membrane permeabilization, PI (Ex535nm/Em620nm; Sigma); transmem-brane potential, DiOC5 (Ex484nm/Em500nm; Invitrogen); phosphatidylserine acces-sibility, annexin V (allophycocyanin conjugate; Ex650nm/Em660nm; Invitrogen). Forexperiments, 105 cells were incubated with peptide (20 �g/mL) or peptide withTTX (50 nM; subinhibitory concentration) in 100 �L 10 mM Pipes, pH 7.5, for 15min or 1 h with shaking at 30 °C (C. albicans) or 37 °C (bacteria, HUVECs). Pipes isa zwitterionic organic-based buffer that is not absorbed through cell membranesand is nontoxic to study cells as assayed (31–33). Cells were stained for 10 min atroom temperature by adding 900 �L stain buffer (PI, 5.0 �g/mL; DiOC5, 0.5 �M;annexinV,2.5�L/mL in50mMK� MEM).Control cellswereexposedtoSDS(0.5%;Sigma), CCCP (100 �M; Sigma), or K� MEM buffer (Sigma) alone. Flow cytometrywas performed using a FACSCalibur instrument (Becton Dickinson) in 10 mM K�

MEM, pH 7.2. Fluorescence of a minimum of 5 � 103 cells was acquired forstatistical analysis.

Bioinformatics. Structural superimpositions and root mean squared deviation(RMSD) calculations were carried out using combinatorial extension [http://cl.sdsc.edu./ce; (34)]. Sequences for phylogenetic analyses were identified initerative BLASTp searches using �-defensin and toxin sequences. Sequences werealigned with CLUSTALW (35), and phylogenetic trees were constructed using theneighbor-joining method (36).

Computational Modeling and Protein Docking. A Kv channel model from C.albicans was generated using homology modeling [Phyre; 3D-PSSM foldingserver; (37)]. The highest scoring template was the mammalian shaker-familyKv1.2 channel (95% estimated precision; E-value, 4.5 � 10�5; PDB code, 2A79).Channel regions spanning S2 and S4–S6 were derived from C. albicans [residues271–293 and 365–494, respectively (gi:68486701)], which includes all of the

P-loop domains used for docking studies. The monomeric C. albicans Kv channelwas then assembled using MUSTANG (38) implemented in YASARA (39). Theresulting tetramer (CAK) was energetically minimized with conjugant gradientsfor 5,000 steps using NAMD (40). The structural coordinates for mammalianKv1.2–2.1 (2R9R), S. lividans KcsA (1BL8), crotamine (1H5O), and HBD-2 (1FD3)were obtained from the Protein Data Bank (www.pdb.org).

Computational models for Kv channel-peptide complexes were generatedusing RosettaDock (www.rosettacommons.org) implemented in CAPRI (41). Inbrief, the docking method used a 2-step process: (i) rigid-body Monte Carlosearches and (ii) parallel optimization of backbone displacement and side-chainconformations using Monte Carlo minimization. The Kv channel domains avail-able fordockingwererestrictedtoextracellular regions.The initial searchyielded�2 � 104 decoys for each ligand (crotamine or hBD-2). For each of the top 50ranked conformers, the �-carbon RMSD of the decoy was compared against eachmember of the conformational set in the second search. Iterative refinementresulted in 8 top-scoring conformations per ligand. Interaction sites were rankedby binding energy, and the energy contributions per residue (5 Å radius) tabu-lated. Ligand-protein residue pairs were then ranked based on total energycontribution and orientation-dependent hydrogen bonding. The 8 most favor-able docking complexes were evaluated as potential binding sites.

Statistical Analyses. Experiments were performed a minimum of 2 independenttimes on different days. Unpaired Student’s t test was used to compare differ-ences in data exhibiting normal distributions; data exhibiting discontinuousdistributions were analyzed using the standard nonparametric Kolmogorov-Smirnoff methodology. P values � 0.05 were considered significant.

ACKNOWLEDGMENTS. We thank Trang Phan and Scott G. Filler (Division ofInfectious Diseases, Harbor–UCLA Medical Center, and the General Clinical Re-search Center at Harbor–UCLA Medical Center) for providing HUVECs for thesestudies and H. Ronald Kaback, Terry J. Smith, Robert I. Lehrer, Eric P. Brass, andJohn E. Edwards, Jr., for helpful discussions. This work was supported by NationalInstitutes of Health, National Institute of Allergy and Infectious Diseases Grants5R01AI39001 and 5R01AI48031 to M.R.Y.

1. Yeaman MR, Yount NY (2003) Mechanisms of antimicrobial peptide action and resis-tance. Pharmacol Rev 55:27–55.

2. Yeaman MR, Yount NY (2007) Unifying themes in host defence effector polypeptides.Nat Rev 5:727–740.

3. Yount NY, Yeaman MR (2004) Multidimensional signatures in antimicrobial peptides.Proc Natl Acad Sci USA 101:7363–7368.

4. Nicastro G, et al. (2003) Solution structure of crotamine, a Na� channel affecting toxinfrom Crotalus durissus terrificus venom. Eur J Biochem 270:1969–1979.

5. Mancin AC, et al. (1998) The analgesic activity of crotamine, a neurotoxin from Crotalusdurissus terrificus (South American rattlesnake) venom: A biochemical and pharma-cological study. Toxicon 36:1927–1937.

6. Fry BG, et al. (2006) Early evolution of the venom system in lizards and snakes. Nature439:584–588.

7. Radis-Baptista G, et al. (2004) Identification of crotasin, a crotamine-related gene ofCrotalus durissus terrificus. Toxicon 43:751–759.

8. Baxter DF, et al. (2002) A novel membrane potential-sensitive fluorescent dye improvescell-based assays for ion channels. J Biomol Screen 7:79–85.

9. Dorn A, et al. (2005) Evaluation of a high-throughput fluorescence assay method forHERG potassium channel inhibition. J Biomol Screen 10:339–347.

10. Slack M, Kirchhoff C, Moller C, Winkler D, Netzer R (2006) Identification of novel Kv1.3blockers using a fluorescent cell-based ion channel assay. J Biomol Screen 11:57–64.

11. Andres MT, Viejo-Diaz M, Fierro JF (2008) Human lactoferrin induces apoptosis-like celldeath in Candida albicans: Critical role of K� channel-mediated K� efflux. AntimicrobAgents Chemother 52:4081–4088.

12. Froy O, Gurevitz M (2004) Arthropod defensins illuminate the divergence of scorpionneurotoxins. J Pept Sci 10:714–718.

13. Ren D, et al. (2001) A prokaryotic voltage-gated sodium channel. Science 294:2372–2375.

14. Alam A, Jiang Y (2009) High-resolution structure of the open NaK channel. Nat StructMol Biol 16:30–34.

15. Gordienko DV, Tsukahara H (1994) Tetrodotoxin-blockable depolarization-activatedNa� currents in a cultured endothelial cell line derived from rat interlobar arter andhuman umbilical vein. Pflugers Arch 428:91–93.

16. Walsh KB, Wolf MB, Fan J (1998) Voltage-gated sodium channels in cardiac microvas-cular endothelial cells. Am J Physiol 274:H506–H512.

17. Lewis RJ, Garcia ML (2003) Therapeutic potential of venom peptides. Nat Rev Drug Dis2:790–802.

18. Rogachevskii IV, et al. (2000) The defensin receptor: A possible mechanism responsiblefor reduced excitability of the neuronal sensory membrane. Dokl Biol Sci 375:595–598.

19. Kushmerick C, de Souza Castro M, Santos Cruz J, Bloch C Jr, Beirao PS (1998) Functionaland structural features of gamma-zeathionins, a new class of sodium channel blockers.FEBS Lett 440:302–306.

20. Gordon YJ, Romanowski EG, McDermott AM (2005) A review of antimicrobial peptidesand their therapeutic potential as anti-infective drugs. Curr Eye Res 30:505–515.

21. Zasloff M (2002) Antimicrobial peptides of multicellular organisms. Nature 415:389–395.

22. Ganz T, Oren A, Lehrer RI (1992) Defensins: Microbicidal and cytotoxic peptides ofmammalian host defense cells. Med Microbiol Immunol 181:99–105.

23. Yu L, et al. (2005) Nuclear magnetic resonance structural studies of a potassiumchannel-charybdotoxin complex. Biochemistry 44:15834–15841.

24. Zachariae U, et al. (2008) The molecular mechanism of toxin-induced conformationalchanges in a potassium channel: Relation to C-type inactivation. Structure 16:747–754.

25. Jouirou B, et al. (2004) Cobatoxin 1 from Centruroides noxius scorpion venom: Chem-ical synthesis, three-dimensional structure in solution, pharmacology and docking onK� channels. Biochem J 377:37–49.

26. Zhu S, Bosmans F, Tytgat J (2004) Adaptive evolution of scorpion sodium channeltoxins. J Mol Evol 58:145–153.

27. Zhou Y, Morais-Cabral JH, Kaufman A, MacKinnon R (2001) Chemistry of ion coordi-nation and hydration revealed by a K� channel-Fab complex at 2.0 A resolution.Nature 414:43–48.

28. Chang CC, Tseng KH (1978) Effect of crotamine, a toxin of South American rattlesnakevenom, on the sodium channel of murine skeletal muscle. Br J Pharmacol 63:551–559.

29. Rizzi CT, et al. (2007) Crotamine inhibits preferentially fast-twitching muscles but isinactive on sodium channels. Toxicon 50:553–562.

30. Filler SG, Swerdloff JN, Hobbs C, Luckett PM (1995) Penetration and damage ofendothelial cells by Candida albicans. Infect Immun 63:976–983.

31. Tang YQ, Yeaman MR, Selsted ME (2002) Antimicrobial peptides from human platelets.Infect Immun 70:6524–6533.

32. Tang YQ, et al. (1999) A cyclic antimicrobial peptide produced in primate leukocytes bythe ligation of two truncated alpha-defensins. Science 286:498–502.

33. Good NE, et al. (1966) Hydrogen ion buffers for biological research. Biochemistry5:467–477.

34. Shindyalov IN, Bourne PE (1998) Protein structure alignment by incremental combi-natorial extension (CE) of the optimal path. Protein Eng 11:739–747.

35. Higgins DG, Sharp PM (1988) CLUSTAL: A package for performing multiple sequencealignment on a microcomputer. Gene 73:237–244.

36. Saitou N, Nei M (1987) The neighbor-joining method: A new method for reconstructingphylogenetic trees. Mol Biol Evol 4:406–425.

37. Kelley LA, MacCallum RM, Sternberg MJ (2000) Enhanced genome annotation usingstructural profiles in the program 3D-PSSM. J Mol Biol 299:499–520.

38. Konagurthu AS, Whisstock JC, Stuckey PJ, Lesk AM (2006) MUSTANG: A multiplestructural alignment algorithm. Proteins 64:559–574.

39. Krieger E, Koraimann G, Vriend G (2002) Increasing the precision of comparativemodels with YASARA NOVA—a self-parameterizing force field. Proteins 47:393–402.

40. Phillips JC, et al. (2005) Scalable molecular dynamics with NAMD. J Comput Chem26:1781–1802.

41. Gray JJ, et al. (2003) Protein-protein docking predictions for the CAPRI experiment.Proteins 52:118–122.

42. Koradi R, Billeter M, Wuthrich K (1996) MOLMOL: A program for display and analysisof macromolecular structures. J Mol Graph 14:51–55, 29–32.

43. Pettersen EF, et al. (2004) UCSF Chimera—a visualization system for exploratoryresearch and analysis. J Comput Chem 25:1605–1612.

44. Humphrey W, Dalke A, Schulten K (1996) VMD: Visual molecular dynamics. J Mol Graph14:33–38, 27–28.

6 of 6 � www.pnas.org�cgi�doi�10.1073�pnas.0904465106 Yount et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 3

, 202

0