Multidimensional signatures in antimicrobial peptides Nannette Y. Yount*† and Michael R. Yeaman*†‡§ *Division of Infectious Diseases, Harbor–UCLA Medical Center, and †St. John’s Cardiovascular Research Center, Research and Education Institute at Harbor–UCLA, Torrance, CA 90502; and ‡David Geffen School of Medicine, University of California, Los Angeles, CA 90024 Communicated by H. Ronald Kaback, University of California, Los Angeles, CA, March 5, 2004 (received for review January 7, 2004) Conventional analyses distinguish between antimicrobial peptides by (iii) published antimicrobial activity, and (iv) up to 75 aa in length. differences in amino acid sequence. Yet structural paradigms common The initial dataset contained over 500 known antimicrobial pep- to broader classes of these molecules have not been established. The tides (see supporting information, which is published on the PNAS current analyses examined the potential conservation of structural web site). From this initial cohort, representatives of specific classes themes in antimicrobial peptides from evolutionarily diverse organ- [e.g., ␣-defensins, ␤-defensins, insect defensins, plant defensins, isms. Using proteomics, an antimicrobial peptide signature was dis- cysteine-stabilized (CS)-␣␤ peptides] were identified on the basis of covered to integrate stereospecific sequence patterns and a hallmark their being well characterized in the literature, having a known 3D three-dimensional motif. This striking multidimensional signature is structure, and being prototypic of their respective subclass. No conserved among disulfide-containing antimicrobial peptides span- cysteine-containing antimicrobial peptide subclasses were specifi- ning biological kingdoms, and it transcends motifs previously limited cally excluded from this algorithm. Thus, the primary dataset was to defined peptide subclasses. Experimental data validating this broadly encompassing of antimicrobial peptides with diverse se- model enabled the identification of previously unrecognized antimi- quence and secondary structures known at the time of this inves- crobial activity in peptides of known identity. The multidimensional tigation. Amino acid sequence data were analyzed, because not all signature model provides a unifying structural theme in broad classes nucleotide sequences have been characterized, and saturation of of antimicrobial peptides, will facilitate discovery of antimicrobial nucleotides occurs within nonmitochondrial sequences over evolu- peptides as yet unknown, and offers insights into the evolution of tionary time scales. molecular determinants in these and related host defense effector molecules. Proteomic Analysis of Structural Patterns. Primary sequences were analyzed for conserved motifs by multiple sequence alignment ature provides a context in which hosts across the phylogenetic [MSA (7, 8)]. Stereospecific sequence analyses used PROSITE to Nspectrum are confronted by potential microbial pathogens. In probe forward and reverse Swiss-Prot protein databases. Peptides turn, natural selection provides a corresponding requirement for with known structures [Protein Data Bank (PDB)] were used in rapid and effective molecular stratagems of host defense. Antimi- conformation analyses. Selection criteria for comparative nonan- crobial peptides represent a pivotal result of this coevolutionary timicrobial peptides followed a systematic algorithm paralleling that relationship. While higher organisms have evolved complex and used for the antimicrobial peptides examined. A Boolean argument adaptive immune systems, virtually all organisms rely on innate was used to identify appropriate controls from the group of immune mechanisms that involve antimicrobial peptides that are cysteine-containing peptides with known mature primary se- rapidly deployed to ward off microbial invasion. quences and 3D structures: chain length 10–75 residues [AND] Antimicrobial peptides may be categorized as those with or sequence containing a GXC motif [NOT] keywords antimicrobial, without disulfide bridges. Members of both classes share a number toxin, or protease inhibitor. Peptide fragments, nonnatural variants, of features that likely contribute to their antimicrobial actions, and duplicates were removed from search outcomes. A cohort of including (i) small size, typically Ͻ10 kDa; (ii) cationic charge, often Ͼ50 comparators with no evidence of antimicrobial activity in the ϩ ϩ ranging from 2to 7atpH7;and(iii) amphipathic stereoge- literature or to our knowledge were identified in the Protein Data ometry, conferring relatively polarized hydrophilic and hydropho- Bank. Specific comparators were identified to represent diverse IMMUNOLOGY bic facets (1). The general recurrence of such themes in antimicro- phylogenetic sources, including mammalian, plant, fungal, and bial peptides from diverse organisms suggests further structural prokaryotic. Three-dimensional alignments of peptides were car- relationships yet to be defined. The limited size of these polypep- ried out by using combinatorial extension (9), and quantified by rms tides places restrictions on the structural repertoire available to deviation (rmsd). As a control for alignment specificity, analyses meet these requirements. Despite this limitation, antimicrobial included a prototypic ␣-helical antimicrobial peptide that lacks peptides display a high degree of variability at nonconserved sites, disulfide-bridging (magainin, from Xenopus). with amino acid substitution rates on the order of those associated with positive selection (2). These observations are consistent with Assay for Antimicrobial Activity. Antimicrobial assays were per- the hypothesis that coevolutionary selective pressures drive host– formed by using a well established radial diffusion method modified pathogen interactions (3). to pH 5.5 or 7.5 (10). Peptides were obtained from commercial Amino acid sequence motifs have been identified in certain sources, with the exception of recombinant brazzein, kindly pro- antimicrobial peptide subclasses [e.g., cysteine array in mammalian vided by J. L. Markley and F. M. Assadi-Porter (University of defensins (4)], and prior studies suggested that structural themes Wisconsin) (11). Defensin HNP-1 was included in each assay as a may occur within distinct peptide subsets (5, 6). Yet comparatively control. A panel of microorganisms was tested: Gram-positive little is known about potentially more comprehensive relationships Staphylococcus aureus (ATCC 27217) and Bacillus subtilis (ATCC common to broader classes of antimicrobial peptides. The current 6633), Gram-negative Escherichia coli (strain ML-35), and the investigation integrated proteomic and microbiologic techniques to fungus Candida albicans (ATCC 36082). Logarithmic-phase organ- explore structure–function commonalities among antimicrobial isms were inoculated (106 colony-forming units͞ml) into buffered peptides from phylogenetically diverse organisms. Materials and Methods Abbreviations: CS, cysteine-stabilized; rmsd, rms deviation. Antimicrobial Peptide Dataset. Prototypic representatives from vir- §To whom correspondence should be addressed at: Division of Infectious Diseases, Harbor– tually all classes of disulfide-containing antimicrobial peptides were UCLA Medical Center, RB-2, Research and Education Institute, 1124 West Carson Street, included to generate a diverse primary dataset by using the follow- Torrance, CA 90502. E-mail: [email protected]. ing criteria: (i) mature primary sequence, (ii) cysteine-containing, © 2004 by The National Academy of Sciences of the USA www.pnas.org͞cgi͞doi͞10.1073͞pnas.0401567101 PNAS ͉ May 11, 2004 ͉ vol. 101 ͉ no. 19 ͉ 7363–7368 Downloaded by guest on October 2, 2021 Fig. 1. Enantiomeric sequence patterns in cysteine-containing antimicrobial peptides. Representatives are, in descending order (peptide name, GenBank GI accession code): ␣-defensins (HNP-3, 229858); ␤-defensins (mBD-8, 15826276); insect defensins (phormicin, 118432); insect CS-␣␤ peptides (drosomycin, 2780893); plant CS-␣␤ peptides (Ah-AMP-1, 6730111); crustacean CS-␣␤ peptides (MGD-1, 12084380); protegrins (protegrin-1, 3212589); big defensin (2493577); gaegurins (gaegurin-1, 1169813); tachyplesins (tachyplesin-1, 84665); polyphemusins (polyphemusin-1, 130777); mytilins (mytilin A, 6225740); gomesin (20664097); thanatin (6730068); and AFP-1 (1421258). Sequences are shown in their conventional dextromeric orientations unless indicated as levomeric (levo). The GXC or CXG–C motif is highlighted [glycine (G), orange; cysteine (C), yellow] within primary sequences corresponding to the ␥-core motif (outlined). Conserved cysteine residues beyond the ␥-core are shaded gray. agarose, and poured into plates. Peptides (10 ␮g) were introduced tionary distance of 2.6 billion years [estimated divergence of fungi into wells in the seeded matrix and incubated for3hat37°C. and plants from higher organisms (12)]. This dataset included Nutrient overlay medium was applied, and assays were incubated at prototypes of all major classes of disulfide-containing antimicrobial 37°C for bacteria or 30°C for fungi. After 24 h, zones of inhibition peptides, including distinct conformation groups such as defensin, were measured. Independent experiments were repeated a mini- CS-␣␤, ranabox, and ␤-hairpin. mum of two times. Conventional multiple sequence alignment (MSA; N- to C- terminal; dextromeric) revealed no clear consensus patterns among Results primary sequences of the antimicrobial peptide study set. However, Identification