Solution Structure of Spheniscin, a (cid:1) -Defensin from the Penguin Stomach* □ S

Recently two (cid:1) -defensins, named spheniscins, have been isolated from the stomach content of the king penguin ( Aptenodytes patagonicus ), which is capable of pre-serving food for several weeks during egg incubation (Thouzeau, Froget, Le C., Hoffmann, J. A., and Bulet, P. (2003) J. Biol. Chem. 278, 51053–51058). It has been proposed that, in combination with other antimicrobial peptides, spheniscins may be involved in this long term preservation of food in the bird’s stomach. To draw some structure/ function features, the three-dimensional structure in aqueous solution of the most abundant spheniscin (Sphe-2) was determined by two-dimensional NMR and molecular modeling techniques. The overall fold of Sphe-2 includes a three-stranded antiparallel (cid:1) -sheet stabilized by three disulfide bridges with a pairing typical of (cid:1)

During the final stage of egg incubation in king penguins (Aptenodytes patagonicus), the male can preserve undigested food in the stomach for several weeks (1). This ensures survival of the newly hatched chick in the event that the return of the foraging female from the sea is delayed. In accordance with the characterization of stress-induced bacteria (2), a previous study has demonstrated that numerous antimicrobial activities exist in preserved stomach contents (3). Two antimicrobial peptides have been isolated and fully characterized, namely spheniscin-1 and -2 (Sphe-1/pBD-1 and Sphe-2/pBD-2). 1 The two forms of spheniscins differ by a single residue, His 14 , in Sphe-1 versus Arg 14 in Sphe-2. A data bank search revealed that spheniscins are members of the well known defensin family.
The activity of a synthetic version of Sphe-2 is preferentially * 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. □ S The on-line version of this article (available at http://www.jbc.org) contains the following supporting information: 1) proton chemical shifts of Sphe-2 ( 1 The abbreviations used are: Sphe-1/pBD-1 and Sphe-2/pBD-2, spheniscin-1/penguin ␤-defensin-1 and spheniscin-2/penguin ␤-defensin-2, respectively; NOE, nuclear Overhauser effect; NOESY, NOE spectroscopy; TOCSY, total correlation spectroscopy; r.m.s.d., root mean square deviation; Gal, gallinacin; CHP, chicken heterophil peptide; GPV, gallopavin; THP, turkey heterophil peptide; HBD, human ␤-defensin; mBD, murine ␤-defensin; BNBD, bovine neutrophil ␤-defensin; MALDI-TOF-MS, matrix-assisted laser desorption ionization time-offlight mass spectrometry; MIC, minimal inhibitory concentration; RK, rabbit kidney; PDB, Protein Data Bank. directed against Gram-positive bacteria including pathogenic strains and is of bactericidal type. Interestingly, when an activity was recorded against Gram-negative strains, the effect of Sphe-2 is mainly of bacteriostatic type. This contrasts with the activity generally observed for ␤-defensins as most of them are effective against Gram-negative bacteria and yeast cells. In fact, the activity of spheniscin is closer to that of ␣-defensins that are microbicidal against Gram-positive and Gram-negative bacteria, yeast fungi, and several enveloped viruses (4).
In this study, we determined the three-dimensional structure of the king penguin ␤-defensin, Sphe-2, in aqueous solution by two-dimensional 1 H NMR spectroscopy and molecular modeling and compared this structure to those of the closest ␤-defensins from mammals. The global fold of Sphe-2, as of most ␤-defensins, includes a well defined three-stranded ␤-sheet and a short N-terminal domain with an ␣-helical propensity. However, differences were observed in the distribution of the charged and hydrophobic residues. While HBD-2 and BNBD-12 are amphipathic, HBD-3 is mainly hydrophilic. Sphe-2, although very cationic, is slightly less hydrophilic than HBD-3, and its surface displays a small hydrophobic patch that was not described in HBD-3 but apparently preserved in avian ␤-defensins according to comparative modeling studies. These results provide evidence that penguin ␤-defensin has some structural peculiarities that may explain the differences in the antimicrobial spectrum and activities between the bird and the mammalian ␤-defensins.

EXPERIMENTAL PROCEDURES
Peptide Synthesis and Purification-Synthetic Sphe-2 was obtained from Altergen Laboratory (Schiltigheim, France). The integrity, purity, and correct refolding were confirmed by MALDI-TOF-MS fingerprinting of a tryptic digest of 2 g of peptide following the experimental procedure reported previously (3). Briefly purified Sphe-2 (native and synthetic) was treated with trypsin (Roche Applied Science) at an enzyme/substrate ratio of 1:10 (w/w), and the digest was analyzed by MALDI-TOF-MS using ␣-cyano-4-hydroxycinnamic acid as matrix.
NMR Experiments-The NMR sample was prepared by dissolving 5.5 mg of synthetic Sphe-2 in 600 l of H 2 O/D 2 O (90:10) to obtain a final concentration of 2 mM. All 1 H NMR spectra were recorded on a Varian INOVA NMR spectrometer equipped with a z-axis field-gradient unit and operating at a proton frequency of 600 MHz. The pH was adjusted to 4.3, and the double quantum filtered correlation spectroscopy (20), clean TOCSY (21), and NOESY (22) experiments were performed at 293K. Clean TOCSY was acquired using a mixing time of 80 ms; NOESY spectra were performed with mixing times of 120, 160, and 300 ms. All spectra were referenced to the residual H 2 O signal set as the carrier frequency, 4.821 ppm at 293 K, and were processed with NMRPipe/NMRDraw softwares (23). The identification of amino acid spin systems and the sequential assignment were performed using the standard strategy described by Wü thrich (24). To identify the exchange rate of backbone amide protons, the sample was lyophilized and quickly dissolved in D 2 O. The exchange kinetic experiments were then monitored from one-dimensional spectra at different time intervals and from short TOCSY spectra.
Structure Calculations-Distance constraints were determined by volume integration of correlations observed in the NOESY spectrum recorded with a mixing time of 160 ms using the NMRView software (25). Since the disulfide pairing is known (3), covalent bonds were built between the sulfur atoms of the paired cysteine residues, Cys 5 -Cys 33 , Cys 12 -Cys 27 , and Cys 17 -Cys 34 . Structure calculations were carried out with ARIA 1.1 (26) implemented in the software CNS 1.1 (27). ARIA is a powerful approach using an iterative process to perform the assignment of ambiguous NOEs from structures calculated using a combination of NOEs assigned at the previous step and a set of ambiguous distance restraints treated as the sum of contributions from all possible assignments. Each ARIA run included nine iterations. In the last step, the 20 best structures were refined by molecular dynamics in explicit solvent to remove artifacts due to the simplification of the force field used in the previous steps. A table of chemical shifts and a first set of distance restraints deduced from easily assigned NOEs were introduced as input to ARIA, and the calculations were initiated. The "ambicutoff" p (26) for accepting or rejecting possible assignments was modified compared with the default values. In the standard ARIA protocol, p varies from 0.999 to 0.8 during the whole process. The variation was reduced from 0.999 to 0.9 to prevent the elimination of weak contributions to ambiguous cross-peaks that are essential for the formation of the secondary structure elements. At the end of each run, the new assignments proposed by ARIA were checked manually and introduced (or not) in the following run. Rejected restraints and residual violations were also analyzed, and assignments were corrected if required. This iterative process was repeated until complete assignment of the NOESY map was achieved. A last run of 100 structures was then performed with the final list of NOE-derived distance restraints in which no restraint can be rejected. The 10 structures with the lowest energy were considered as characteristic of the peptide structure.
The structures were displayed and analyzed using the MOLMOL (28) and SYBYL (SYBYL®6.5, TRIPOS Inc., St. Louis, MO) programs, and their quality was evaluated using the PROCHECK (29) and PROMO-TIF (30) softwares. The formation of hydrogen bonds was established according to distance criteria. Lipophilic and electrostatic potentials were calculated and represented using the MOLCAD option of SYBYL.
Salt Effect on the Antibacterial Activity of Sphe-2-The effect of sodium and magnesium chloride on the antibacterial activity of Sphe-2 was assessed using a liquid growth inhibition assay (31). Growth inhibition was estimated by measurement of the minimal inhibitory concentration (MIC) against selected strains, the Gram-negative Escherichia coli and the pathogenic Gram-positive Staphylococcus aureus. Briefly logarithmic phase bacterial cultures were diluted to an A 600 of 0.001 (approximately 10 5 colony-forming units/ml) in a broth supplemented or not with various concentration of salts. Diluted bacteria (90 l) were mixed with 10 l of either distilled water (control) or different concentrations of Sphe-2 ranging from 0.75 up to 100 M. Bacterial growth was measured after an overnight incubation at 30°C by measuring the change in the absorbance of the culture at 600 nm with a microplate reader. The MIC value corresponds to the interval of con- is the highest concentration tested at which the bacteria are growing and [b] is the lowest concentration that causes 100% inhibitory growth (32). To assay the salt effect on the bacterial growth, the poor broth medium (1% bactotryptone, 85 mM NaCl, pH 7.4) was supplemented with NaCl and MgCl 2 separately or in combination. The following final salt concentrations were assayed: (i) 160 and 480 mM in NaCl, (ii) 1 and 50 mM in MgCl 2 , and (iii) 160 mM NaCl plus 1 mM MgCl 2 . Osmolarity of the different media and of the stomach content was determined using an automatic milliosmometer (type 13/13DR, Roebling, Berlin, Germany).

RESULTS
NMR Data-After identification of the spin systems on the correlation spectroscopy and TOCSY spectra, sequence-specific assignments were obtained from the sequential connectivities observed on the amide-␣, amide-amide cross-peak region of the NOESY spectrum. The fingerprint region of the TOCSY and NOESY spectra are shown in Fig. 1 (see the complete maps in the supporting information). All protons were assigned except for the amide proton of Phe 2 , which was not observed in our experimental conditions; the arginine NH 2 ; the serine OH; and the N-terminal NH 3 ϩ groups, which are in very fast exchange with water. In addition, the protons of Phe residues could not be assigned unambiguously due to overlapping with the other aromatic protons (see the chemical shift table in the supporting information).
Low field shifted H␣ chemical shifts and a characteristic set of interstrand H␣(i)-H␣(j), NH(i)-NH(j), and H␣(i)-NH(j) connectivities delineate a triple-stranded ␤-sheet. The presence of medium range NOEs in the rest of the sequence suggests that it consists mainly of loops and turns. Strong sequential H␣/H␦ NOE cross-peaks for Xaa-Pro peptide bonds indicate a trans configuration for Pro 20 and Pro 23 . Twenty-one NH signals, exchanged before the acquisition of the first one-dimensional spectra (t ϭ 5 min), are described as "very fast exchanging protons." Four additional amide proton signals disappeared within the first 18 h. The residual amide protons, not exchanged after 18 h, are all located in the ␤-sheet (Phe 11 , Ala 13 , Ile 22 , Ile 24 , Cys 27 , Gln 32 , Cys 33 , Cys 34 , and Arg 35 ) except for the NH of Gly 10 and Val 31 ( Fig. 1) (see the short TOCSY spectra recorded in D 2 O to monitor the exchange data in the supporting information).
Structure Calculations-The NOE data set used in the final ARIA run included 594 distance restraints (Table I) involving 335.2 intraresidue, 139.7 sequential, 32 medium range (2 Յ ͉i Ϫ j͉ Յ 4), and 87.1 long range (͉i Ϫ j͉ Ն 5) restraints with an average of 16 restraints/residue. Among these restraints, 551 are non-ambiguous. Ten structures, in very good agreement with all the experimental data and the standard covalent geometry, were selected for further analysis. For these structures, no experimental distance constraint violation greater than 0.3 Å was observed and the root mean square deviation (r.m.s.d.) values, with respect to the standard geometry, are low. The Ramachandran plot exhibits 95% of the , angles of the 10 converged structures in the most favored and additional allowed regions according to the PROCHECK software nomenclature. The selected structures display small potential energy values. In particular negative van der Waals and electrostatic energy values are indicative of favorable non-bonded interactions.
Structure Description-The overall fold of Sphe-2 is typical of ␤-defensins including a twisted three-stranded antiparallel ␤-sheet with a (ϩ2X, Ϫ1) topology as defined by Richardson (33). Strand ␤ 1 (Phe 11 -Ala 13 ) is hydrogen-bonded to strand ␤ 3 (Val 31 -Arg 36 ), which in turn is hydrogen-bonded to strand ␤2 (Ser 21 -Cys 27 ) (Fig. 2). Residues Ile 24 and Gly 25 are both hydrogen-bonded to Cys 33 (Fig. 1) leading to a ␤-bulge in ␤2. The amide protons of all residues involved in hydrogen bonds exhibited a slow exchange rate in D 2 O (Fig. 1), and the two other residues with a slow NH exchange rate, Gly 10 and Val 31 , are embedded in the structure and particularly inaccessible to the solvent. Strands ␤1 and ␤2 are separated by a long loop, L1, including a type IV ␤-turn between residues Arg 18 and Ser 21 and a type I ␤-turn involving residues Ser 28 to Val 31 . Due to a low number of NOEs in this region, the N-terminal segment (Ser 1 -Gly 10 ) is less "well" defined (Fig. 2). In fact, a small FIG. 1. A, fingerprint region of the TOCSY (left) and NOESY (right) spectra of spheniscin. The sequential connectivity pattern is shown from residues 11 to 16. B, schematic diagram of the triple-stranded antiparallel ␤-sheet structure of Sphe-2 deduced from the analysis of the NMR data: observed backbone interstrand NOEs are indicated as arrows ("sup" means ambiguous NOE including the connectivities, "weak" indicates a weak NOE), slowly exchanging amide protons are in bold, and hydrogen bonds are indicated as dotted lines.
helical turn is formed on a majority of structures, but depending on the structure it was considered by PROMOTIF as an ␣-helix or as a succession of turns. In addition, a PROCHECK analysis showed that most residues in the segment from Phe 2 to Arg 8 lie in the helical region of the Ramachandran plot.
As evidenced by the r.m.s.d. calculations (Table I) and by a superposition of the backbone of the selected structures (Fig.  2), the ␤-sheet region of Sphe-2 is well defined with a pairwise r.m.s.d. on the C␣ atoms of 0.54 Å. The pairwise r.m.s.d. calculated for all C␣ backbone atoms is large (1.93 Å) due to the N-terminal segment (2.40 Å) and to the L1 loop (1.55 Å).
The tertiary structure of Sphe-2 is very compact. The hydrogen bonds between the three strands of the ␤-sheet and the three disulfide bridges contribute to the compactness of this molecule. Most residues are exposed to the solvent and exhibit a large variability in the position of their side chains. The only residues showing a low circular variance of their 1 , 2 angles are hydrophobic ones, Phe 11 (Fig. 3A). The other hydrophobic residues are scattered all over the rest of the molecule and are separated by hydrophilic residues. With 10 arginine residues over 38 amino acids and no anionic residue, Sphe-2 is a highly cationic molecule. The charges are spread all over the surface of the molecule, and the electrostatic potential calculated with SYBYL using Kollman charges (34) varied between Ϫ15 and ϩ380 kcal⅐mol Ϫ1 so that the surface of the molecule is almost entirely positively charged except for the C terminus (Fig. 3A).
Salt Sensitivity of Sphe-2-In control poor broth medium (85 mM NaCl, 216 mosM), Sphe-2 is active at a concentration below 6 M against the two microorganisms selected for this study, S. aureus (Gram-positive) and E. coli (Gram-negative) (Table II). An increase in the NaCl content to a concentration of 160 mM, representing an intermediate concentration between the values observed in seawater fishes and mammals (35), did not affect Sphe-2 activity. The osmolarity measured for this culture medium (348 mosM) was close to the value measured in the stomach content of three different penguins (324 Ϯ 23 mosM) and to the osmolarity of the plasma of sea water fishes (337 mosM) (35). When the concentration in sodium chloride was increased up to 480 mM, the efficacy of Sphe-2 against S. aureus decreased by a factor of 16 (MIC, 50 -100 M). In the control experiment with poor broth supplemented at 480 mM NaCl, E. coli did not grow properly.
In the presence of 1 mM MgCl 2 , the activity of Sphe-2 against S. aureus was unchanged, while at 50 mM a decrease by a factor of 8 was observed. In the presence of 1 and 50 mM of MgCl 2 , the growth of E. coli was altered by factors of 2 and 4, respectively. Combining monovalent ions (160 mM NaCl) and divalent ions (1 mM MgCl 2 ) in the range of the values found in seawater fishes led only to a decrease by a factor of 2 in the MIC values for the two bacterial strains tested. DISCUSSION We determined the three-dimensional structure of Sphe-2 in aqueous solution. Sphe-2 is a ␤-defensin isolated from the stomach contents of the king penguin and shows a wide array of antimicrobial activity against Gram-positive and Gram-negative bacteria, yeasts, and fungi (3). It displays all the structural characteristics of mammalian ␤-defensins, including a common array of disulfide bridges and the presence of a threestranded ␤-sheet as the main secondary structure element. The Solution Structure of Spheniscin second strand of the ␤-sheet includes a bulge found in all ␣and ␤-defensins of known structures. Despite a lack of NOEs in the N-terminal segment, a small helix or a short segment with helical features is found on most structures. Such a small helix or a helical turn, including the first cysteine residue, is observed in mammal ␤-defensins but not in ␣-defensins. The only exception is the ␤-defensin from bovine neutrophils, BNBD-12 (19), which displays a rather disordered N-terminal strand.
In addition to Sphe-2, a series of ␤-defensins has been isolated from chickens and turkey (10 -12). The sequences of these bird defensins are aligned with Sphe-2 in Fig. 4. Gal-3 from the chicken G. gallus and GPV-1 from the turkey M. gallopavo (12) share the strongest percentage of similarities to Sphe-2 with 50 and 47% of identity, respectively. The percentage of identity to Gal-1/CHP1 and Gal-1␣/CHP2 from chicken and to THP-1 and THP-2 from turkey (10, 11) is slightly lower (37-39%). Finally Gal-2 is more distant from Sphe-2 with 33% of identity. Besides the cysteine disulfide array, several residues are strictly conserved between Sphe-2 and the other avian ␤-defensins: a basic residue (Arg or Lys) at position 8 (Sphe-2 numbering), a glycine residue at position 10, a hydrophobic residue (Ile, Leu, or Val) at position 22, a glycine residue at position 25, a phenylalanine at position 30, and finally two hydrophobic residues at the C terminus. Among the four Gly-Xaa-Cys motifs observed in the Sphe-2 sequence, two are also conserved in the other avian ␤-defensins, one at the beginning of ␤1 (with glycine at position 10) and the second in ␤2 (with glycine at position 25). The only exception is GPV-1 where the second Gly-Xaa-Cys motif is replaced by an Ala-Xaa-Cys motif.
There was no previous structure available for bird defensins. However, assuming that other avian ␤-defensins adopt a structure similar to Sphe-2, we built structural models with MODELLER 6.0 (36) using the Sphe-2 structure as template (data not shown). We noticed that the hydrophobic patch observed on Sphe-2 surface (Fig. 3A) is well but not strictly preserved in most other avian defensins. In particular, the replacement of the doublet Phe 19 -Pro 20 of Sphe-2 by the sequence Ser-His in THP-2 and Gal-2 considerably reduces the hydrophobicity of this region. Another important feature concerns the electrostatic properties of the molecules. Even if all other avian ␤-defensins are cationic, they are less positively charged than Sphe-2 (global charge, ϩ10), and the positions of charged residues are rather variable. The differences observed in the hydrophobic and electrostatic properties of bird defensins are probably responsible for differences in the activity spectra of these molecules. However, the activity spectrum of bird defensins is poorly documented. There  Central column (panels 2 and 3), hydrophobic and hydrophilic potential areas, calculated with the MOLCAD option of SYBYL at the Connolly surfaces, are displayed in brown and blue, respectively. Green surfaces represent an intermediate hydrophobicity. A common scale is used for the three molecules. The orientation of panel 2 is given by the backbone representation. A 180°rotation with respect to a vertical axis is applied between panels 2 and 3. Right column (panels 4 and 5), electrostatic positive and negative areas are in red and blue, respectively. Intermediate areas are in green. A common scale is used for the three molecules (ϩ300, Ϫ300 kcal⅐mol Ϫ1 ). The orientation of panel 4 is given by the backbone representation (left column). A 180°rotation with respect to a vertical axis is applied between panels 4 and 5. term, terminus.

Effect of sodium and magnesium chloride on Sphe-2 activity
The MIC following the definition of Casteels and Temps (32) was determined using a liquid growth inhibition assay. Two strains were selected, the Gram-positive S. aureus and the Gram-negative E. coli SBS363. The osmolarity of the different culture media was recorded to compare these values with those observed for sea water (approximately 1000 mosM) and measured for the king penguin stomach content (324 Ϯ 23, mean Ϯ S.D., n ϭ 3). are no published data for Gal-3 and GPV-1 and only partial information on the antimicrobial activity of the others (10,11,37). Until more details become available, it will be particularly difficult to draw structure/activity relationships between Sphe-2 and the other avian ␤-defensins.
To investigate structure/activity relationships, we therefore searched in the Protein Data Bank (PDB) (38) for structural neighbors of Sphe-2 using DALI (39). Among the eight defensins (two ␣ and six ␤) available in the PDB, the best scores were obtained for three ␤-defensins, two human defensins, HBD-2 (PDB code 1fd3) and HBD-3 (PDB code 1kj6), and the bovine neutrophil defensin BNBD-12 (PDB code 1bnb). Interestingly a good score was also obtained for the ␣-defensin from rabbit kidney RK-1 (PDB code 1ews). Although structurally very close to Sphe-2, with a r.m.s.d. of 1.31 Å between the C␣ atoms of aligned residues (Table III), RK-1 displays only 28% of identity with Sphe-2 (Fig. 4B). The distribution of residues at the surface of Sphe-2 and RK-1 are very different (data not shown). The hydrophobic patch is not conserved, and with a global charge of ϩ1, RK-1 is not really cationic. RK-1 is active against E. coli at rather high concentrations (15-150 g/ml) but can also activate Ca 2ϩ channels in vitro (40). This means that, in addition to being antibacterial, it bears some toxin properties. The bovine ␤-defensin BNBD-12 displays 37% of sequence identity to Sphe-2 (Fig. 4B), and the r.m.s.d. is 1.33 Å for 28 aligned residues (Table III). BNBD-12 is active against E. coli and S. aureus at concentrations of the micromolar range (41). The human ␤-defensin HBD-2, although structurally related to Sphe-2 with a r.m.s.d. of 1.25 Å for 28 aligned residues (Table III), exhibits only a little sequence identity with Sphe-2 (18%) (Fig. 4B). In terms of efficacy, as far as the activity data can be compared, HBD-2 is less potent than Sphe-2 against Gram-negative (E. coli and Pseudomonas aeruginosa) and Gram-positive (S. aureus) bacteria but more active against the yeast Candida albicans. Interestingly, regarding the killing activity, Sphe-2 is a bactericide against S. aureus, while HBD-2 is only bacteriostatic (3,13,42). HBD-3 with 37% of sequence identity (Fig. 4B) presents a smaller r.m.s.d. at 0.94 Å for 28 aligned residues (Table III). Similarly, assuming that the data are comparable, Sphe-2 is as potent as HBD-3 on E. coli and P. aeruginosa, slightly more potent on S. aureus, and less efficient on C. albicans. In their modes of action, Sphe-2 and HBD-3 show similar levels of microbicidal activity against S. aureus, while against various Gram-negative strains Sphe-2 is less effective, having a bacteriostatic effect, and HBD-3 has a bactericidal effect (3,43). Thus in terms of antimicrobial activity, Sphe-2 appears to be closer to HBD-3 than to HBD-2.
The sequence alignment of Sphe-2 and the three mammal ␤-defensins discussed above (Fig. 4B) does not reveal many conserved residues apart from the cysteine residues and two Gly-Xaa-Cys motifs. Concerning the distribution of residues at the surface of the molecule, the hydrophobic potential calculated at the Connolly surface with the MOLCAD option of SYBYL, using the same hydrophobicity scale for Sphe-2, HBD-3, and HBD-2 (Fig. 3, central column), shows that HBD-3 is clearly the most hydrophilic molecule without any noticeable hydrophobic patch. Sphe-2 is hydrophilic but displays a hydrophobic patch including residues at the beginning of ␤2 and at the C terminus. HBD-2 and BNBD-12 exhibit an amphipathic structure (16,19). Electrostatic potentials calculated at the Connolly surface with the MOLCAD option of SYBYL (Fig. 3, right column), using the same scale for the three molecules (ϩ300, Ϫ300 kcal⅐mol Ϫ1 ) show that HBD-2, as expected from its total charge of ϩ7, has the lowest positively charged surface. These charges are concentrated on the hydrophilic face of the molecule suggesting a surfactant-like mechanism of action FIG. 4. A, sequence alignment of Sphe-2 and the other avian ␤-defensins. The secondary. id, identity structure elements of Sphe-2 are underlined and schematically represented at the top. In the consensus sequence, ϩ indicates a basic residue, and h stands for a hydrophobic residue. B, Sphe-2 and the bovine ␤-defensin BNBD-12, the human ␤-defensins HBD-3 and HBD-2, and the rabbit ␣-defensin RK-1. Conserved cysteines and GXC motifs are highlighted (magenta). The percentage of identity to Sphe-2 is given for each peptide. The asterisks ‫)ء(‬ indicate the exact position of residues 10, 20, and 30.

Scores of alignment of Sphe-2 with structurally related peptides
The structural alignments were restricted to residues 10 -37 because of the large structural variability of the N-terminal residues. In a best fitted superposition of the structures, two residues are considered to be structurally aligned when their distance is less than 2 Å. Solution Structure of Spheniscin (16). Sphe-2 (10 positive charges, no negative charge) and HBD-3 (13 positive charges versus two negative charges) are both highly positively charged, and the charges are spread all over their surface. The only exceptions concern the C termini COO Ϫ and, for HBD-3, the Glu 28 residue, delineating small negative areas.
Although the formation of a symmetrical dimer cannot be ruled out, our Sphe-2 NMR data do not indicate the presence of such a dimer. In contrast, native gel electrophoresis, NMR diffusion spectroscopy, and light scattering experiments demonstrate that HBD-3 forms a dimer (13). In the latter study, the authors proposed a possible model in which two monomers autoassociate by their ␤2 strand. The association is stabilized by intermolecular hydrogen bonds between the glutamine residues (Gln 29 ) situated at the center of ␤2 and by two intermolecular salt bridges between Glu 28 and Lys 32 . The glutamine residue is particularly well conserved among ␤-defensins but is not found in Sphe-2. The glutamate residue is found only in HBD-3, and there is no negatively charged residue favoring the formation of a possible salt bridge in Sphe-2. This lack of anionic residues in Sphe-2 makes the formation of a Sphe-2 dimer very unlikely due to the strong electrostatic repulsion between the monomers. This suggests that the high efficacy observed for Sphe-2 and its retention of killing of bacteria at physiological salt concentrations depend more on its high net positive charge than on a dimer formation.
This study demonstrates that the global fold of the penguin defensin Sphe-2 adopts all the common features of ␤-defensins. There is considerable sequence diversity among the ␤-defensins associated with a high variability in the distribution of the charged and hydrophobic residues at the surface of the molecule. Comparison with related defensins reveals that the high efficacy of Sphe-2 against bacteria can be correlated with the presence of a hydrophobic patch, which seems common to avian defensins, and with its high cationicity. In addition, we observed that the activity of Sphe-2 is preserved in salt conditions similar (judged through osmolarity measurement) to the ones in the stomach content of penguin. This favors the hypothesis that Sphe-2, retaining its efficacy in in vivo conditions, may regulate the bacterial flora in the penguin stomach, resulting in better food preservation. Finally Sphe-2 may represent a promising model for the structure-based drug design of salt-insensitive antimicrobial peptides to treat infections that result from the proliferation of bacteria and fungi in a salt-rich environment.