Three-dimensional NMR Structure of Hen Egg Gallin (Chicken Ovodefensin) Reveals a New Variation of the β-Defensin Fold*

Background: Ovodefensins are small peptides from eggs, related to avian antimicrobial defensins. Results: The first three-dimensional structure of ovodefensins (gallin) is solved, and its antimicrobial properties are screened. Conclusion: Gallin adopts a β-defensin fold, with significant variations. Its antibacterial spectrum was restricted to E. coli. Significance: The first structural features may be related to E. coli specificity and/or other yet unknown functions. Gallin is a 41-residue protein, first identified as a minor component of hen egg white and found to be antimicrobial against Escherichia coli. Gallin may participate in the protection of the embryo during its development in the egg. Its sequence is related to antimicrobial β-defensin peptides. In the present study, gallin was chemically synthesized 1) to further investigate its antimicrobial spectrum and 2) to solve its three-dimensional NMR structure and thus gain insight into structure-function relationships, a prerequisite to understanding its mode(s) of action. Antibacterial assays confirmed that gallin was active against Escherichia coli, but no additional antibacterial activity was observed against the other Gram-positive or Gram-negative bacteria tested. The three-dimensional structure of gallin, which is the first ovodefensin structure to have been solved to date, displays a new five-stranded arrangement. The gallin three-dimensional fold contains the three-stranded antiparallel β-sheet and the disulfide bridge array typical of vertebrate β-defensins. Gallin can therefore be unambiguously classified as a β-defensin. However, an additional short two-stranded β-sheet reveals that gallin and presumably the other ovodefensins form a new structural subfamily of β-defensins. Moreover, gallin and the other ovodefensins calculated by homology modeling exhibit atypical hydrophobic surface properties, compared with the already known vertebrate β-defensins. These specific structural features of gallin might be related to its restricted activity against E. coli and/or to other yet unknown functions. This work provides initial understanding of a critical sequence-structure-function relationship for the ovodefensin family.

There is currently a real explosion in antibiotic and multidrug resistance (1)(2)(3)(4). A highly promising approach to overcome this issue is to explore and exploit the huge diversity of innovative bioactive-engineered molecules provided by nature to fight off microbes. Among these natural products involved in the defense systems of living organisms, antimicrobial peptides, also called host defense peptides, may represent new clues to investigate (5)(6)(7)(8)(9)(10). These natural molecules of innate immunity ensure the defense of a multitude of organisms (plants, insects, and vertebrates) against the majority of pathogenic organisms. Most of them have a wide spectrum of activity, which covers the main bacterial and fungi species but also encapsulated viruses and protozoa (11). Among the host defense peptides, vertebrate defensins are widely studied (for a review, see Ref. 12). To date, these cysteine-rich cationic peptides have been classified in three groups, depending on the disulfide arrangement and the position of the six conserved cysteines: namely ␣-, ␤-, and cyclic -defensins (the last group being identified exclusively in some primates). Despite their divergence in disulfide pairing, ␣and ␤-defensins share the same three-dimensional fold, composed of a three-stranded antiparallel ␤-sheet. Whereas both ␣and ␤-defensins are found in mammals, only ␤-defensins have been to date identified in birds. It is presumed that ␣-defensins and ␤-defensins arose from a single ancestral ␤-defensin-like gene by gene duplication after the divergence of birds and mammals (13).
The sequencing of the chicken (Gallus gallus) genome revealed the presence of a cluster of 14 different genes on chromosome 3 coding for avian defensins and designated as AvBD1 to -14 (22), with specific tissue distribution (13,(22)(23)(24)(25). Proteomic analyses of the chicken egg revealed the presence of AvBDs, especially AvBD9, AvBD10, and AvBD11, identified in the egg yolk, eggshell, and eggshell/egg white/vitelline membrane, respectively (26 -29). In addition to these AvBDs, another cysteine-rich and cationic peptide of 41 residues (4731.68 Da), named gallin, was also detected in the hen egg white and vitelline membrane (27,30). This peptide was named gallin because of its sequence homology with meleagrin and cygnin, two peptides present in the egg whites of turkey (Meleagris) and swan (Cygnus), respectively (31,32). Two other related peptides, dBPS1 and dBPS2, were also found in duck egg white (33). Their cysteine spacing (CX 3-5 CX 3 CX 11 C 3-4 CC) provides evidence that they are related to defensins, but the gallin gene does not belong to the cluster of AvBD genes, and its primary structure differs from typical AvBD sequences (34). This new subfamily of defensins was therefore termed ovodefensins. In view of the slight difference in cysteine spacing, it has been proposed that ovodefensins and avian defensins may have diverged from a common ancestor (34). Interestingly, gallin is potentially coded by three different genes (34), which are clustered on chromosome 3 and separated from the cluster of AvBD genes. To date, the ovodefensin subfamily counts seven members, including the two sequences, taeniopygin-1 and taeniopygin-2, predicted from the zebra finch genome (34). Besides the classical members of the ovodefensin subfamily, it is noteworthy that the cysteine spacings of dBPS1 and taeniopygin-2 are slightly different and were first described as "related to ovodefensins" (34).
From a functional point of view, recombinant gallin was shown to be able to inhibit the growth of Escherichia coli (34), thus suggesting that this defensin belongs to the antimicrobial innate immune system. However, its potency against other bacteria, especially Gram-positive bacteria, is still unknown.
The purpose of the present work is therefore to further characterize the functional and structural features of gallin. This study successively describes the chemical synthesis of gallin and then its use for the screening of its antibacterial spectrum and for the resolution of its three-dimensional structure.

EXPERIMENTAL PROCEDURES
Reversed Phase High Performance Liquid Chromatography and Mass Spectrometry-HPLC analyses were carried out on either an Elite LaChrom system, composed of a Hitachi L-2130 pump, a Hitachi L-2455 diode array detector, and a Hitachi L-2200 autosampler, or on a LaChrom 7000 system, composed of a Merck-Hitachi L-7100 pump, a Merck-Hitachi L-7455 diode array detector, and a Merck-Hitachi D-7000 interface, which was also used for semipreparative purification. The machines were equipped with C18 reversed phase columns (Nucleosil), 300 Å, 5 m, 250 ϫ 4.6 mm for the analytical separations or 250 ϫ 10.5 mm for purification. Solvents A and B containing 0.1% TFA were H 2 O and MeCN, respectively. MS analyses were performed on an Autoflex MALDI-TOF instrument (Bruker Daltonics, Bremen, Germany) equipped with a 337-nm nitrogen laser and a gridless delayed extraction ion source. The instrument was used in reflector positive ion mode with a 150-ns delay and an accelerating voltage of 19 kV. Instrument control and external calibration were accomplished using Flex-Control software (Bruker). The observed m/z values correspond to the monoisotopic ions. The sample was co-crystallized with a solution of ␣-cyano-4-hydroxycinnamic acid as a matrix, using the dry droplet method. Synthetic gallin was further analyzed by MS and MS/MS on a nanoESI-Q-TOF Ultima Global mass spectrometer (Waters, Manchester, UK), as described previously (15). A multicharged precursor ion with an m/z value of 788.42 (charge state 6 (ϩ6)) was selected for the fragmentation (data not shown).
The time course of an analytical oxidative folding reaction was determined by taking, at intervals, aliquots from the reac-First Three-dimensional NMR Structure of Ovodefensins tion mixture, quenching the oxidation by acidifying the sample using TFA (final concentration 2%), and then analyzing the sample by analytical C18 reversed phase HPLC. Oxidized gallin was eluted earlier than its reduced form, 12  . E. coli CFT073 is a human uropathogenic strain whose genome has been sequenced (35).
Antibacterial activities of gallin were measured by a radial diffusion assay according to the method described by Lehrer et al. (36). Bacteria incubated overnight were diluted in trypticase soy broth or brain heart infusion at an absorbance of 0.02 and were incubated for 2.5-4 h, depending on the bacterial strain, at 37°C to obtain a mid-logarithmic phase culture. Bacterial concentrations were determined by plating serial 10-fold dilutions of bacterial suspension on trypticase soy agar plates and by counting colony-forming units (cfu) after 24 h of incubation at 37°C. A volume containing bacteria at 1 ϫ 10 7 cfu was centrifuged at 900 ϫ g for 5 min at 4°C, and bacteria were washed once with cold 10 mM sodium phosphate buffer (pH 7.4), resuspended in a small volume of cold sodium phosphate buffer, and mixed with 25 ml of previously autoclaved, warm (42°C) 10 mM phosphate buffer, containing 0.03% trypticase soy broth medium, 1% (w/v) low endosmosis-type agarose (Sigma-Aldrich), and 0.02% Tween 20. The agarose solution containing bacteria was poured into a Petri dish to form a 1-mm-deep uniform layer. A 2.5-mm-diameter gel punch was used to make 36 evenly spaced wells. In each well, 5 l of peptide dilutions or control solutions were added. MSI-94 (a linear amphipathic magainin variant displaying a broad antimicrobial spectrum) was used as a positive control. MSI-94 was a kind gift from Dr. Philippe Bulet (BioPark, Archamps, France). The peptides were allowed to diffuse in the bacteria-containing gel while the plates were incubated for 3 h at 37°C. The gel was then overlaid with 25 ml of agar consisting of a double strength (6% (w/v)) solution of trypticase soy broth containing 1% (w/v) agarose. After an overnight incubation at 37°C, the diameter of the clearing zone surrounding each well was measured. For each bacterial strain, three identical independent measurements of antibacterial activity were performed. The minimum inhibitory concentration (MIC) of each peptide was determined as described by Lehrer et al. (36). The best fit straight line was determined using linear regression with Excel Microsoft 2003 software. The MIC was calculated by finding the intersection of the line with the x axis, indicating the lowest peptide concentration at which no clear zone was obtained.
NMR Experiments-A standard set of two-dimensional 1 H NMR experiments (COSY, 80-ms TOCSY, and 160-ms NOESY) was performed on a 1 mM aqueous solution of the synthetic gallin (H 2 O/D 2 O (90:10) and 100% D 2 O) at pH 5.5 and at 293 K. All spectra were recorded on a 600-MHz VARIAN INOVA NMR spectrometer and processed with the NMRPipe/ NMRDraw software package (37,38). 1 H chemical shifts were assigned according to classical procedures, and NOE crosspeaks were assigned within the NMRView software (37). 13 C-1 H HSQC spectra were acquired in natural abundance, and 13 C chemical shifts were combined with 1 H chemical shifts to deduce 31 (,) angular restraints within the TALOS software (39).
Structure Calculations-Distance constraints were obtained from the volume integration of NOE correlations, using the NMRView software. The calculations were initiated using the default parameters of ARIA1.1 software (40) and a first set of easily assigned NOEs. At the end of each run, the new assignments proposed by ARIA were checked manually and introduced (or not) in the following calculation. Ambiguous intersulfur distances, an option assuming that a given half-cystine is part of a bridge without supposing a particular partner, were used in the first runs. With this "ambiguous disulfide bridges" protocol, each half-cystine is allowed to be linked to one of the five others, leading to 15 possibilities of pairing. During the calculations, each disulfide bridge floats freely, and the protocol is driven to the most compatible disulfide bridge array, under the influence of all the other NMR restraints. When enough experimental data had been introduced into the iterative ARIA process to converge to a specific disulfide bridge pairing, covalent bonds were added between sulfur atoms involved in each bridge. Final structures were displayed and analyzed using the MOLMOL program (41), and their quality was evaluated using the PROCHECK and PROMOTIF software packages (42,43).
Comparative Modeling-Comparative modeling, or homology modeling, was performed with the MODELLER software (44), taking gallin as the template, to build the three-dimensional structures of the four closest ovodefensins (Fig. 1a). The level of sequence identity between targets and template is Ͼ60% for meleagrin, cygnin, and dBPS2 from the egg white of turkey (Meleagris), swan (Cygnus), and duck, respectively (31)(32)(33), and is 41% for taeniopygin-1 from zebra finch (Taeniopygia guttata) (45). This makes it possible to follow the standard MODELLER protocol. We first checked for each sequence that no other protein with a known related structure displayed a greater sequence similarity. Then each ovodefensin was aligned with the gallin sequence, without needing to introduce any gap. When gaps are absent in the sequence alignment, the threedimensional model construction is straightforward, and the First Three-dimensional NMR Structure of Ovodefensins confidence in accuracy is high. Models were not built for dPBS1 and taeniopygin-2 ( Fig. 1a), which were considered too far from the template, gallin, to obtain reliable models and to precisely analyze the resulting molecular surfaces.

RESULTS
Because of the difficulties of purifying gallin present at very low concentration in egg white (26), a prerequisite to the functional and structural study of gallin was its chemical synthesis.
Synthesis and Folding of Gallin-Peptide elongation of gallin was carried out by solid-phase peptide synthesis using the polyethylene glycol dimethylacrylamide resin as a polymeric matrix facilitating the synthesis of "difficult sequences" (46). A Wang-type linker was attached to the aminomethyl polyethylene glycol dimethylacrylamide resin by using the commercially available Fmoc-Lys(Boc)-methylphenoxyproprionic acid. The completion of the reaction was checked by Kaiser's test. Elongation was then carried out by the Fmoc/tBu strategy. After TFA treatment, the reduced form of gallin was precipitated using cold tert-butyldimethyl ether and analyzed by HPLC and mass spectrometry.
The oxidative folding was optimized on the fully reduced HPLC-purified gallin. The folding kinetics was followed by analytical HPLC (Fig. 2). At room temperature in the presence of a redox system (GSSG/GSH) at pH 8.6, the reaction was shown to be complete in less than 30 min. Finally, the synthetic gallin was purified by semipreparative HPLC and lyophilized.
Antibacterial Activity of Gallin-The antimicrobial activities of gallin was evaluated using a radial diffusion assay against a panel of eight different bacterial strains, mostly composed of pathogenic bacteria. Inhibition zones in units were plotted as a function of log 10 (peptide concentration) for each bacterial strain, as illustrated on Fig. 3

First Three-dimensional NMR Structure of Ovodefensins
NMR Assignments-All protons were assigned except the NH of Ser 13 , Gly 27 , and Met 30 . Chemical shifts have been deposited in the BioMagResBank with the entry code 19729. Low field-shifted H␣ chemical shifts and a characteristic set of interstrand connectivities delineate the ␤-sheet. An unusual chemical shift was observed for NH of Cys 16 (4.38 ppm), outside of the already large range of chemical shifts listed in the BioMagResBank. In the three-dimensional structure, the NH of Cys 16 points toward Trp 22 , which explains the unusual chemical shift observed for Cys 16 .
Structure Calculations-NOE assignments were progressively introduced during the iterative process of ARIA and manually validated between each run until complete assignment of the NOESY map. In the first runs, the "ambiguous disulfide bridges" option was used, to prevent any speculation about the disulfide bridge array (Cys 6 -Cys 33 , Cys 12 -Cys 28 , and Cys 16 -Cys 32 for ␣-defensins; Cys 6 -Cys 32 , Cys 12 -Cys 28 , and Cys 16 -Cys 33 for ␤-defensins; or any other possibility). Then an intermediate run (500 initial structures, 250 structures refined in water, 50 best structures kept for analysis) was used to unambiguously define the cysteine pairing before further refinement. Among this set of 50 structures (already displaying a root mean square deviation calculated on C␣ atoms of 0.79Å), 40 structures (80%) unambiguously form the three disulfide bridges of ␤-defensins. In all of these 50 structures (50/50), the C 16 with exponents C 33 disulfide bridge typical of ␤-defensin is formed, whereas no structure (0/50) displays the C 6 with exponents C 33 or the C 16 with exponents C 32 disulfide bridge, both of which are expected in ␣-defensins. Further measurement of distances between the sulfur atoms of each half-cystine clearly discredited ␣-defensin pairing (Fig. 4). To switch ␤to ␣-defensins, a completed reversal of the central strand of the sheet is required, which is totally inconceivable without numerous violations of distance constraints. The determined disulfide bridge array was introduced in final runs of calculation. The last run, performed with 1000 initial structures, used a final list of 782 NOE-derived distance restraints (Table 2), divided into 331.2 intraresidue, 174.9 sequential, 84 medium range (2 Յ ͉i Ϫ j͉ Յ 4), and 191.9 long range (͉i Ϫ j͉ Ն 5) restraints, with an average of 20 restraints/residue. Among these NOE restraints, 727 are nonambiguous. In the last iteration, 200 structures were refined in a shell of water, 181 did not show any NOE violation greater than 0.1 Å, and the worst 19 structures only showed one violation above 0.1 Å. Among the best refined structures, 10 structures were selected, in agreement with all of the experimental    (47,48), such as defensins or such as host defense peptides in general, it could be particularly instructive to explore whether a priori unrelated proteins, with very low sequence identity, could match with the protein of interest. Structural databanks were explored to determine if other pro-teins adopt the five-stranded ␤-sheet fold described for gallin. A comparison of the three-dimensional structure of gallin with about 75,000 experimentally determined coordinate files available in the Protein Data Bank (interactive service authored by E. Krissinel and K. Henrick) did not reveal any other protein with such a five-stranded sheet arrangement. Thus, the gallin threedimensional fold can be considered a newly identified fold. The only proteins considered to be similar to gallin are ␤-defensins from humans, mice, and birds and one of the ␣-defensins (rabbit RK1). Comparable results were obtained by questioning the DaliLite version 3 software on the DALI server (49), where only ␣and ␤-defensin three-dimensional structures were found to be similar to gallin, with a better fit with ␤-defensin folds than with ␣-defensins. Not surprisingly, ␤-defensin-like peptides (␤-defensin disulfide bridges, but less structured) were not highlighted by our analysis. Thus, the gallin three-dimensional fold can be considered a newly described five-stranded fold and a new variation for the ␤-defensin fold.

DISCUSSION
Gallin is a chicken cysteine-rich cationic peptide previously identified in the chicken egg white and vitelline membrane by proteomic approaches (26,27). Its antibacterial activities against the Gram-negative bacteria E. coli were recently demonstrated (34). Moreover, gallin was described as being related to the ␤-defensin family due to the presence of six conserved cysteine residues (34). However, the spacing between these conserved cysteines, which is characteristic of ovodefensins, a group of peptides isolated from the egg white of different avian species, is different from that observed for other known ␤-defensins. Moreover, the disulfide bond pairing of ovodefensins is not known and must be determined to unambiguously classify ovodefensins as members of the ␣or ␤-defensin family. The lack of structural and functional data thus prompted us to define the three-dimensional structure of gallin in aqueous solution by NMR and to explore its antimicrobial activity against a series of Gram-positive and Gram-negative bacteria.

First Three-dimensional NMR Structure of Ovodefensins
This sheet is common to vertebrate ␣and ␤-defensins, which differ only by the cysteine pairing. Molecular modeling of gallin under NMR restraints converged only with a C 6 with exponents C 32 , C 12 with exponents C 28 , C 16 with exponents C 33 pairing, attesting that gallin can unambiguously be classified as a ␤-defensin. Some variations of the ␤-defensin fold have already been described, mainly the presence of an amino-terminal ␣-helix or an ␣-helical propensity in several structures, such as human hBD1-3 (PDB codes 1IJV, 1FD4, and 1KJ6 respectively), mouse mBD7-8 (PDB codes 1E4T and 1E4R), and king penguin AvBD103b (PDB code 1UT3), whereas other ␤-defensin structures do not contain this additional amino-terminal helical propensity (such as chicken AvBD2 and bovine bBD12, PDB codes 2LG5 and 1BNB, respectively). In the present study, the global fold of gallin does not contain any amino-terminal helix but contains an additional two-stranded parallel ␤-sheet (residues 2-3 and 24 -25) (Fig. 5). This observation reveals that the gallin structure adopts a new variation of the classical ␤-defensin fold. To our knowledge, this is the first structural proof that ovodefensins belong to the ␤-defensin family.
Structural Comparison with Avian Defensins-Only two three-dimensional structures of avian defensins are currently available (Fig. 1b): king penguin AvBD103b defensin, which our group determined in 2004 (PDB code 1UT3 (50)), and chicken AvBD2, which we recently solved (PDB code 2LG5 (16)). When these three structures are aligned, their three-stranded ␤-sheets can be easily superimposed (Fig. 6). However, they differ in their amino-terminal region; whereas gallin contains an additional two-stranded parallel ␤-sheet, AvBD103b has a high propensity to form an ␣-helix, as already observed for some mammal ␤-defensins. Chicken AvBD2 is shorter, lacking the amino-terminal residues, and consequently cannot form an additional amino-terminal structural element. The presence (or absence) of this amino-terminal helix cannot be related to antimicrobial properties, because both king penguin AvBD103b and chicken AvBD2 demonstrate large antimicrobial spectrum (21,51). However, the amino-terminal region could be involved in other yet unknown functions.
Sequence-Structure Relationships-The consensus sequence of ovodefensins, CX 3-5 CX 3 CX 11 CX 3-4 CC, is different from that observed for other known ␤-defensins found in mammals or birds. In particular, the spacing between the third and fourth cysteine has been pointed out (34). Fig. 1, b-e, illustrates multiple alignments of gallin with vertebrate ␣and ␤-defensin sequences with known three-dimensional structures. (i) Alignments with avian or mammal ␤-defensins clearly underline the particularities of the fourth cysteine; X 11 CX 3-4 for ovodefensins, contrasting with X 9 CX 5-6 for typical ␤-defensins, forces us to introduce gaps or to move this cysteine forward within the alignment (Fig. 1, b and c). This shift in sequence allows the two additional residues before the fourth cysteine to form an additional parallel sheet on the three-dimensional structure (Fig. 5) and is therefore directly related to the formation of the additional sheet. (ii) When the gallin sequence is compared with the five ␤-defensin-like peptides from various origins (Fig. 1d) (i.e. defensins with the typical disulfide bridges of ␤-defensins but lacking the complete three-dimensional structures), the consensus sequence almost fits (CX 3-5 CX 3 CX 11 CX 3-4 CC for the ovodefensin group versus CX 6 -7 CX 3-7 CX 7-13 CX 4 -6 CC for ␤-defensin-like peptides), particularly with regard to the spacing between the third and fourth cysteine. However, gallin is well structured and therefore does not belong to this ␤-defensin-like group. (iii) Finally, even if three-dimensional structures of ␣and ␤-defensins are globally superimposable, the primary structure of gallin is clearly different from ␣-defensin sequences (Fig. 1e). We conclude that ovodefensins form a new subfamily of ␤-defensins, with a specific consensus sequence, accounting for a variation in the ␤-defensin three-dimensional fold.
Surface Properties of Gallin-Gallin is a highly cationic molecule, with the six charged residues distributed over the surface of the molecule, resulting in an overall positive molecular surface (electrostatic potentials at the surface not shown). The representation of the hydrophobic properties at the molecular  MOLMOL software (41)). The five strands of ␤-sheets are numbered from the amino to the carboxyl terminus (␤1-␤5), and cysteine residues are numbered to clearly identify the disulfide bridges. FIGURE 6. Superimposition of gallin, chicken AvBD2, and king penguin AvBD103b in black, dark gray, and light gray, respectively. Whereas the common three-stranded ␤-sheet can be easily superimposed, each defensin differs in the amino-terminal region. MARCH 7, 2014 • VOLUME 289 • NUMBER 10 JOURNAL OF BIOLOGICAL CHEMISTRY 7217 surface is more instructive; indeed, hydrophobic and hydrophilic residues are not randomly distributed at the surface. The gallin surface displays two hydrophobic sides (Fig. 7, A and B) surrounded by a crown of hydrophilic residues (Asn 14 , Thr 15 , Ser 17 , Lys 18 , Ser 37 , Lys 39 , Lys 41 , Lys 4 , His 26 , and Lys 29 ). A basic residue, Lys 8 , points out of one face, in the middle of a rather hydrophobic region (Fig. 7A). This lysine is conserved in dBPS2, cygnin, and meleagrin sequences and could be essential to (yet unknown) biological function(s), and/or it could be essential to prevent oligomerization/aggregation.

First Three-dimensional NMR Structure of Ovodefensins
A comparison of gallin surface properties with those of other ovodefensins could give insight into structural properties of the ovodefensin subfamily. Thus, with gallin as a target, three-dimensional models were built for the four closest ovodefensins currently identified (dBPS2 from ducks, cygnin from black swans, and meleagrin from turkeys (displaying 62% of sequence identities with gallin)) as well as taeniopygin-1 predicted from the zebra finch genome and exhibiting 41% sequence identity with gallin (Fig. 1a).
The first three ovodefensins exhibit the same surface properties as gallin. In particular, the distribution of hydrophobic and hydrophilic residues at the surface, which could be related to function(s), is well conserved. On the contrary, the model built for the fourth ovodefensin, taeniopygin-1, leads to a mirror distribution of hydrophobic and hydrophilic residues at the surface of the three-dimensional models (Fig. 7, C and D). The crown of hydrophilic residues is preserved, ensuring a sort of "circular amphiphily." However, the substitution of Lys 8 (gallin numbering; Figs. 1a and 7A) by Pro in taeniopygin-1 makes this face hydrophobic (Fig. 7C). Two lysines, Lys 1 and Lys 22 , point outward on the opposite face and replace Leu 1 and Trp 22 in gallin (Fig. 7D). They could compensate for gallin Lys 8 for biological implications and/or for preventing aggregation of the protein.
Functional Data-In the present study, we observed that gallin has a potent antimicrobial activity against several E. coli strains, including an avian strain isolated from the oviduct of a hen with colibacillosis (BEN 3578). The observed anti-E. coli effect is in agreement with previous results showing that gallin at a concentration of 0.25 M was able to inhibit the growth of an E. coli strain different from those used in our study (34). In contrast, no activities against the other tested bacteria, including Gram-negative (S. enterica serovar Enteritidis and S. enterica serovar Typhimurium) and Gram-positive (S. aureus and L. monocytogenes) bacteria, were observed for gallin at a maximal concentration of 53 M. These results are unexpected because generally defensins possess a broad spectrum of antimicrobial activities, targeting both Gram-positive and Gram-negative bacteria (15,21,51). However, to our knowledge, no antimicrobial activity has been revealed for the other known members of the ovodefensin family to date. Indeed, when used at a concentration as high as 22 M, turkey meleagrin displayed no anti-E. coli (LE392 strain) activity (32). Likewise, the duck ovodefensins dBPS1 and dBPS2 at a concentration of 56 M possess no intrinsic antimicrobial properties, at least against the following Gram-positive and negative bacteria tested: S. aureus, B. subtilis, S. enterica serovar Enteritidis, and

First Three-dimensional NMR Structure of Ovodefensins
E. coli (52). The antimicrobial spectrum of the swan ovodefensin cygnin remains still undefined. Therefore, gallin is to date the only member of the ovodefensin family for which antimicrobial activity has been described, even if it is limited to E. coli. Gallin potentially participates in the antibacterial defense of the egg and the hen oviduct (reproductive organ involved in egg formation) against E. coli contamination. Nevertheless, this narrow antimicrobial activity raises questions, particularly whether this activity reflects the main function of this molecule. In fact, it is likely that gallin and, to a greater extent, all ovodefensins may have biological functions other than those related to antimicrobial defense. In this regard, potent anti-lipase activities have recently been found for the duck ovodefensins dBPS1 and dBPS2, suggesting that these molecules may play a role in the protection of lipids from lipasemediated hydrolysis (52). Interestingly, dBPS1/dBPS2 proteins have been predominantly detected in the oviduct, where the synthesis of the egg takes place, and also to a lesser extent in the gall bladder, where they are believed to exert potentially their antilipase activities (33,52). Similarly, maximum expression of the gallin genes is observed in the magnum (oviduct part responsible for the egg white protein synthesis) when compared with other oviduct parts (34). The biological activities of gallin need to be further investigated to highlight its function in the hen egg.

CONCLUSION
The gallin three-dimensional structure is the first ovodefensin structure to have been solved to date. No other protein with the same five-stranded arrangement could be found in structure databases. The gallin three-dimensional structure contains the typical ␤-sheet and the typical disulfide bridge array of ␤-defensins. Hence, gallin can be unambiguously classified as a ␤-defensin. However, gallin displays an additional small sheet, which can be directly related to the cysteine spacing in the sequence. Thus, gallin is the first member of a new structural subfamily of defensins. The apparent antibacterial selectivity of gallin toward E. coli strains, which may be related to the atypical hydrophobic properties of the gallin molecular surface, will need further investigation.
Finally, gallin is to date the only member of the ovodefensin family for which an antimicrobial activity has been described. The atypical structural features of gallin, compared with other defensins, could be related to other functions that remain to be characterized.