Homodimeric (cid:1) -Defensins from Rhesus macaque Leukocytes ISOLATION, SYNTHESIS, ANTIMICROBIAL ACTIVITIES, AND BACTERIAL BINDING PROPERTIES OF THE CYCLIC PEPTIDES*

Rhesus (cid:1) -defensin 1 (RTD-1) is a unique tridisulfide, cyclic antimicrobial peptide formed by the ligation of two 9-residue sequences derived from heterodimeric splicing of similar 76-amino acid, (cid:2) -defensin-related precursors, termed RTD1a and RTD1b (Tang, Y. Q., Yuan, J., Osapay, G., Osapay, K., Tran, D., Miller, C. J., Ouellette, A. J., and Selsted, M. E. (1999) Science 286, The structures of RTD-2 and RTD-3 were predicted to exist if homodimeric splicing of the RTD1a and RTD1b occurs in vivo . Western blotting disclosed the presence of puta-tive (cid:1) -defensins, distinct from RTD-1, in leukocyte extracts. Two new (cid:1) -defensins, RTD-2 and RTD-3, were purified by reverse-phase high performance liquid chromatography and characterized by amino acid analysis, matrix-assisted laser desorption/ionization time-of-flight mass spectroscopy, and comparison to the synthetic standards. RTD-2 and RTD-3 are the predicted homodimeric splicing products of RTD1b and RTD1a, respectively. The cellular abundances of RTD-1, acid -hydroxysuccinimidyl free termini and sequence on pmol of each peptide. Natural peptides com- pared with synthetic versions by RP-HPLC, AU-PAGE, amino acid analysis, and MALDI-TOF mass

The characterization of the host defense components of Rhesus macaque granulocytes disclosed two distinct subfamilies of ␣-defensins (24) and a new tridisulfide peptide termed rhesus -defensin 1 (RTD-1) 1 (25). RTD-1 is a macrocyclic 18-amino acid antimicrobial peptide formed by the ligation of two 9-residue sequences derived from similar 76-amino acid, ␣-defensinrelated precursors, termed RTD1a and RTD1b (25). RTD-1 shares some structural similarities with the pig neutrophil protegrins and the horseshoe crab tachyplesins (25)(26)(27). The cyclic structure of RTD-1 is an important determinant for microbicidal potency and resistance to the inhibitory effect of physiologic sodium chloride. The antimicrobial potency of synthetic acyclic RTD-1 was 3-fold lower than that of the native peptide under low salt conditions, and the acyclic peptide was completely inhibited by physiologic NaCl (25). Those studies suggested that the cyclic conformation of RTD-1 confers saltinsensitive microbicidal activity that may be critical for antimicrobial function in the extracellular milieu.
The isolation of RTD-1 revealed the existence of a novel post-translational pathway for the production of head-to-tail cyclized peptides in primates. Based on the heterodimeric splicing model that produces RTD-1 from RTD1a-and RTD1bderived nonapeptides (25), we hypothesized that two additional -defensins, termed RTD-2 and RTD-3, would be produced by the homodimeric splicing of RTD1b and RTD1a, respectively. Here we report the isolation of RTD-2 and RTD-3 from circulating leukocytes, as well as the synthesis and antimicrobial and bacterial binding properties of the three rhesus -defensins.

EXPERIMENTAL PROCEDURES
Peptide Synthesis, Disulfide Formation, and Cyclization-Peptide synthesis was performed essentially as described for RTD-1 (25). Peptide sequences corresponding to open-chain versions of RTD-2 and -3 (see Fig. 1) were assembled at 0.2 mmol scale on Fmoc (9-fluorenylmethoxycarbonyl)-Arg(2,2,4,6,7-pentamethyldihydro-benzofuran-5-sulfonyl) polyethylene glycol-polystyrene resin using a Milligen 9050 automated synthesizer. Arg, Cys, and Thr side chains were protected with 2,2,4,6,7-pentamethyldihydro-benzofuran-5-sulfonyl, triphenylmethyl, and tert-butyl groups, respectively. All amino acids except cysteine were coupled with O-(7-azabenzotriazol-1-yl)-1,13,3-tetramethyluronium hexafluorophosphate/N,N-diisopropylethylamine activation. Cysteine residues were coupled as the pre-formed pentafluorophenyl ester derivative. RTD-2 was assembled with double coupling at every cycle. RTD-3 was assembled with double coupling of Thr and Ile residues. Following * This work was supported by National Institutes of Health Grant AI22931 and a grant from Large Scale Biology, Inc. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Linear synthetic RTD-2 and -3 were purified by preparative C 18 RP-HPLC on a 25 ϫ 100-mm DeltaPak C 18 cartridge (Waters, MA) developed with a 0.25%/min gradient of water-acetonitrile containing 0.1% trifluoroacetic acid. Aliquots from eluant fractions were analyzed by matrix-assisted laser desorption/ionization time-of-flight mass spectroscopy (MALDI-TOF MS), and those containing reduced/linear peptides were pooled and concentrated 10-fold by centrifugal evaporation. The peptide solutions were diluted to 100 -200 g/ml in 17.4 mM ammonium acetate, pH 8.0, and stirred vigorously in an open container for 18 h at 22°C. Peptide folding and oxidation were monitored by C 18 RP-HPLC and MALDI-TOF MS. The acyclic versions of RTD-2 and -3 were then purified by preparative C 18 RP-HPLC as described above. Purity was confirmed by analytical C 18 RP-HPLC and acid-urea PAGE on 12.5% polyacrylamide gels (29). For MALDI-TOF MS, peptide solutions were mixed with an equal volume of 10 mg/ml ␣-cyano-4-hydroxycinnamic acid in 50/50 water-acetonitrile containing 0.1% trifluoroacetic acid and analyzed on a Voyager DE-RP mass spectrometer (PerSeptive Biosystems) (25).
Antibody Production-Rabbit anti-RTD-3 antibody was produced as described previously for the preparation of anti-RTD-1 antibody (25). Briefly, acyclic RTD-3 (3.5 mg) was conjugated to ovalbumin (3.5 mg) with 0.1% glutaraldehyde in 7 ml of 100 mM sodium phosphate, pH 7.4, and stirred for 18 h at 22°C. The reaction was quenched with 300 mM glycine, and the peptide/ovalbumin conjugate was dialyzed exhaustively against water. Two New Zealand White rabbits were repeatedly immunized using standard procedures until the anti-RTD-3 antiserum titer was 1:10,000, as determined by enzyme-linked immunosorbent assay. IgG-enriched preparations were obtained by chromatography on a DEAE Econo-Pac column according to the manufacturer's protocol (Bio-Rad).
Purified RTD-1-3 were characterized by MALDI-TOF MS, amino acid analysis, and AU-PAGE. Cysteine content was determined by comparing the masses of the native peptides with those obtained following reduction of disulfides with 1,4-dithiothreitol and alkylation with iodoacetamide (31). The amino acid compositions of RTD-1-3 were determined on 6 N HCl hydrolysates (2 h, 150°C) as 6-aminoquinolyl-N-hydroxysuccinimidyl carbamate derivatives (32). The absence of free N termini in natural RTD-2 and -3 was determined by Edman sequence analysis on 20 -50 pmol of each peptide. Natural peptides were compared with synthetic versions by RP-HPLC, AU-PAGE, amino acid analysis, and MALDI-TOF mass spectroscopy.
Antimicrobial Activities of RTD-1-3-The antimicrobial activities of synthetic RTD-1-3 against bacteria (Staphylococcus aureus 502a and Escherichia coli ML35) and fungi (Candida albicans 16820 and Cryptococcus neoformans 271A) were determined in an agar diffusion assay as described previously (33). Briefly, 10-l wells were bored in a 9-cm 2 plate of agarose, buffered with 10 mM PIPES, pH 7.4, containing 5 mM glucose, and seeded with 1 ϫ 10 6 mid-log phase cells. Five-l aliquots of each peptide, dissolved in 0.01% HOAc at 10 -300 g/ml, were added to each well. After incubation at 37°C for 2 h, the seeded agar was overlaid with molten agarose containing 6% trypticase soy broth (for bacteria) or Sabouraud dextrose broth (for fungi). Plates were incubated at 37°C for 18 -24 h, and antimicrobial activity was determined by measuring the diameter of clearing around each well.
The microbicidal activities of each peptide were determined by incubating 2 ϫ 10 6 colony-forming units/ml with peptides (0.5-12 g/ml) in 50 l of low salt diluent, 10 mM PIPES buffer containing 5 mM glucose, pH 7.4, or the same diluent supplemented with 25-150 mM NaCl. After 2 h of incubation at 37°C, the cell suspensions were diluted 1:50 with 10 mM sodium phosphate buffer, pH 7.4, and exponentially spread with an Autoplate 400 (Spiral Biotech) onto trypticase soy agar (bacteria) or Sabouraud dextrose agar (fungi). After incubation at 37°C for 18 -48 h, colonies were counted and cell survival was expressed as colony-forming units/ml.
Binding of RTD-1-3 to E. coli ML35 was evaluated by incubating 2 ϫ 10 6 log-phase bacteria with increasing peptide concentrations (0.5-8 g/ml final) in 1 ml of 10 mM PIPES, pH 7.4, containing 5 mM glucose for 2 h at 37°C. The incubation mixtures were centrifuged at 20,000 -25,000 ϫ g for 10 min at 22°C, and supernatant samples were analyzed for RTD content by RP-HPLC. Binding of each RTD to E. coli was determined by subtraction of the quantity of supernatant peptide from total added to each tube, and comparing this to control incubations lacking bacteria.

RESULTS
Peptide Synthesis, Disulfide Formation, and Cyclization-RTD-2 and RTD-3 are cyclic analogs of RTD-1 predicted to be produced by homodimeric splicing of nonapeptides from RTD1b and RTD1a, respectively (Fig. 1). Linear RTD-2 and -3 were synthesized and purified by preparative RP-HPLC. Disulfide bond formation proceeded efficiently in room air, giving Ͼ90% yield of monomeric, tridisulfide peptide as determined by quantitative RP-HPLC and MALDI-TOF MS. The yields of the subsequent peptide cyclization steps were 92% for RTD-2 and 64% for RTD-3. The cyclic peptides were purified by C 18 RP-HPLC and characterized by AU-PAGE, amino acid analysis, and MALDI-TOF MS. RTD-2 (9.2 mg) and RTD-3 (3.2 mg) preparations were more than 99% pure, and were indistinguishable from the natural peptides (see below).
Isolation of Natural RTD-1-3-Synthetic RTD-1, -2, and -3 had unique R F values on acid-urea PAGE because of their differing arginine contents (Fig. 2). Acid extracts of rhesus macaque leukocytes contained a band that co-migrated with synthetic RTD-1 on AU-PAGE and Western blots. RTD-1 and two additional immunopositive bands that co-migrated with RTD-2 and RTD-3 synthetic standards were detected in leukocyte extracts with anti-RTD-1 and anti-RTD-3 antibodies (Fig.  2). These data strongly suggested the presence of RTD-2 and RTD-3 in leukocyte extracts.
RTD-1-3 were isolated from leukocyte extracts by RP-HPLC (Fig. 3). Peptides with masses of RTD-1, -2, and -3 were detected in three peaks following the initial chromatographic step (Fig. 3A), and the RP-HPLC elution times precisely matched those of the respective synthetic peptides. Each -defensin was purified to homogeneity (Fig. 3B), and their identities were confirmed as described below.
Characterization of RTD-2 and -3-Automated Edman degradation of 20 -50 pmol of purified RTD-2 and RTD-3 yielded no amino acid signal, consistent with the -defensin cyclic structure. The molecular masses of natural RTD-2 and RTD-3, determined by MALDI-TOF MS, matched the calculated values of the predicted sequences (Fig. 1). The cysteine content of purified RTD-1-3 was determined by comparing the molecular masses of native peptides with those that had been reduced and alkylated. Carboxamidomethylated RTD-1, -2, and -3 had molecular masses of 2430.5 atomic mass units (2430.7 ϭ theoretical), 2436.9 atomic mass units (2437.7 ϭ theoretical), and 2424.5 atomic mass units (2423.6 ϭ theoretical), respectively, consistent with the complete alkylation of 6 cysteine residues in each -defensin (Table I).
The compositions of natural RTD-2 and -3, determined by amino acid analysis of peptide hydrolysates (Table I), were consistent with those of the corresponding structures shown in Fig. 1 (B and C), and that of purified RTD-1, which was as reported previously (25). In a previous study, synthetic RTD-1 was biochemically and functionally equivalent to the natural peptide (25). Synthetic RTD-2 and -3 were also indistinguishable from the natural isolates by amino acid analysis, MALDI-TOF MS, AU-PAGE, and analytical RP-HPLC (Fig. 4).
Antimicrobial Activities of RTD-1-3-An agar diffusion assay was utilized to assess the combined microbicidal and microbistatic activities of RTD-1-3 against Staphylococcus aureus 502a, Escherichia coli ML35, and yeast forms of Candida albicans 16820 and Cryptococcus neoformans 271A. The antimicrobial activities of the three -defensins were equivalent against S. aureus, C. albicans, and C. neoformans. RTD-2 was 2-3-fold less active than RTD-1 and RTD-3 against E. coli (Fig. 5). The microbicidal potencies of RTD-1, -2, and -3 were determined in liquid-phase assays (see "Experimental Procedures"). The bactericidal and fungicidal activities of the three peptides were similar. However, approximately twice as much peptide was required to kill fungal suspensions as was needed to kill the same number of bacterial cells (Fig. 6). RTD-2-mediated killing of E. coli showed a steep dose dependence similar to that of RTD-1 and -3, but approximately twice as much RTD-2 was required to achieve the microbicidal activities of RTD-1 and -3. However, at concentrations of 2 g/ml or higher, all three -defensins reduced the viability of E. coli ML35 by at least 99.9%.
Unlike most ␣and ␤-defensins, the microbicidal activity of RTD-1 is relatively unaffected by physiologic NaCl (25). The microbicidal activities of 10 g/ml of RTD-2 and RTD-3 were determined using S. aureus 502a and E. coli ML35 as test organisms in the presence of 0 -150 mM NaCl. As with RTD-1, the potent microbicidal activity of RTD-2 and RTD-3 was unaffected by NaCl at all salt concentrations tested (data not shown). The activities of -defensins were also compared with those of other potent antibacterial peptides. Microbicidal assays were performed in parallel with indolicidin (a linear, tryptophan-rich, bovine cathelicidin; Refs. 34 and 35) and mouse enteric ␣-defensins (cryptdins) 3 and 4. On a molar basis, RTD-1-3 were each 2-4 times as potent as indolicidin against S. aureus 502a, and each -defensin was equipotent to the cryptdins against S. aureus (data not shown). A (29:1:2) mixture of synthetic RTD-1, -2, and -3 was prepared and used in microbicidal assays against the four test organisms. The peptide mixture exhibited microbicidal potencies that were identical to that of RTD-1 (data not shown), indicating that synergistic microbicidal interactions do not occur under these assay conditions.
Experiments were performed to determine whether the difference activities of RTD-1-3 against E. coli ML35 are the result of differential binding of the peptides to the bacterium. As shown in Fig. 7, the binding of the three -defensins to E. coli cells was essentially identical. For all three -defensins, maximal peptide binding by 2 ϫ 10 6 bacteria was ϳ2 g. The corresponding peptide concentrations (0 -2 g/ml) are associated with dose-dependent killing of bacterial and fungal targets (Fig. 6). However, despite the equal binding by RTD-1-3 to E. coli cells, the colicidal activity of RTD-2 is substantially lower than that of RTD-1 and -3 (Fig. 6). Thus, other peptidecell interactions influence the relative microbicidal potencies of the three -defensins.

DISCUSSION
Macrocyclic peptides composed entirely of L-amino acids are relatively rare biomolecules, nearly all known examples of which have been isolated from plants. More than 30 macrocyclic peptides, collectively termed cyclotides, have been identified in plants of the Rubiaceae and Violaceae families (36). The mature active peptides typically contain 30 amino acids including 6 cysteines connected in a 1-4, 2-5, 3-6 motif. The structures of two cyclotides (circulin A and cycloviolacin O1), determined by NMR, contain several ␤-strands constrained by a cysteine knot (36,37). Peptide folding kinetics of acyclic permutants indicated that the intact cysteine knot is critical for native conformation. Biosynthesis of plant cyclotides involves the post-translational ligation and splicing of the polypeptide precursors (38,39). The observation that some cyclotides possess antimicrobial activities suggests that these peptides may have a role in plant defense (40,41).
RTD-1 was the first macrocyclic antimicrobial peptide discovered in animals, disclosing the capacity of animal cells to produce this circular motif (25). The fact that RTD-1 is produced by the heterodimeric ligation of two nonidentical nonapeptides encoded by distinct genes suggested that homodimers might be formed similarly. The isolation of RTD-2 and RTD-3 confirmed this hypothesis. Moreover, the biosyn- FIG. 3. Purification of RTD-1-3. A, acid extracts of 4 ϫ 10 7 leukocytes were chromatographed on a C 18 reverse-phase column using a 0.5%/min water-acetonitrile gradient containing 0.1% trifluoroacetic acid. Numbered peaks contained peptides with molecular masses shown, including masses consistent with RTD-1, -2, and -3. B, purified RTD-1-3 were chromatographed on a C 18 column using a 1%/ min water-acetonitrile gradient containing 0.1% trifluoroacetic acid.  Fig. 1 (B and C). b Determined by MALDI-TOF mass spectroscopy. c Calculated based on amino acid compositions of linear peptides. d Differences between a and c, accounted for by loss of 18 atomic mass units resulting from cyclization and loss of 6 atomic mass units because of disulfide formation. thesis of three unique -defensins suggests that the cellular machinery responsible for peptide cyclization may be involved in post-translational modification of other gene products (25).
The cyclic structures of RTD-1-3 endow the peptides with resistance to exoproteinases, and this may be advantageous in a protease-rich inflammatory milieu. Furthermore, peptide cyclization confers other properties to RTD-1 (25) that are absent in an acyclic analog, because acyclic RTD-1 is substantially less active than the native peptide against S. aureus and E. coli (25). RTD-1-3, but not the respective acyclic analogs, maintain their staphylocidal and colicidal activities in the presence of physiologic sodium chloride (Ref. 25 and data not shown).
The relative yields of RTD-1, -2, and -3 obtained from leukocyte extracts indicated that RTD-1 is 10-fold more abundant than RTD-2 and -3 combined. This suggests a strong preference for production of RTD-1 by heterodimeric splicing of the RTD-1 precursors, RTD1a and RTD1b. We speculate that distinct elements in the two precursors may direct the assembly of heterodimeric intermediates prior to peptide splicing (25).
Despite differences in cationicity, RTD-1 (ϩ5), RTD-2 (ϩ6), and RTD-3 (ϩ4) possess similar antimicrobial potencies against four organisms tested in this study, and were nearly identical against S. aureus, C. albicans, and C. neoformans. The most cationic peptide, RTD-2, was slightly less active against E. coli than RTD-1 and RTD-3. This was somewhat surprising because increased cationicity typically correlates with greater antimicrobial activities and increased spectrum of activity (42,43). However, the activity of RTD-2 against E. coli was nearly equivalent to those of RTD-1 and RTD-3 when longer incubation times (4 or 6 h) were used in the agar diffusion assay (data not shown). This suggests that the kinetics of peptide-bacteria interactions differs among the three -defensins under these assay conditions.
The binding of peptide to E. coli ML35 was equivalent for all three -defensins (Fig. 7), indicating that difference in bactericidal activities of RTD-1-3 (Figs. 5 and 6) is the result of subsequent interactions of peptides with bacterial cells. We speculate that the increased electrostatic interaction of the more cationic RTD-2 with components of the E. coli cell envelope adversely affects the antibacterial activity. An alternative explanation is that quaternary interactions occur once the peptide is bound to the target cell, and that those interactions are less favored in the case of the highly cationic RTD-2.
The studies presented here demonstrate that RTD-1-3 are broad spectrum microbicides that kill bacterial and fungal targets in a salt-insensitive manner. Under the conditions of the in vitro assay system, synergistic antimicrobial activity was not apparent. However, it is possible that the activities of -defensins, individually or in combination, may be substantially altered by the microenvironment in which they function in vivo.
Addendum-As the current studies were being prepared for publication, we became aware of a report by Leonova et al. (44), who described the isolation of RTD-1-3 from rhesus macaque bone marrow, the structures of which are identical to those reported here. Further, the solution structure of synthetic RTD-1 was recently reported by Trabi et al. (45), fundamentally confirming the model proposed previously (25).