To identify novel defensin genes in the chicken, all five groups of known defensin-like peptide sequences from plants, invertebrates, and vertebrates were first queried individually against the translated chicken nonredundant (NR), EST, high throughput genomic sequence (HTGS), and whole-genome shortgun sequence (WGS) databases in the GenBank by using the TBLASTN program21. All potential hits were then examined manually for the presence of the characteristic cysteine motifs. For every novel defensin identified, additional iterative BLAST searches were performed until no more novel sequences could be found. In addition to three known chicken β-defensins (Gal 1–3) 2223, nine novel putative sequences, namely Gal 4–12, have been found in the EST database with at least two hits for each, and such sequences have also been confirmed in genomic sequences (Table 1). Because of the fact that mammalian defensins tend to form clusters 12141518, all chicken HTGS and WGS sequences containing defensin sequences were also retrieved from GenBank, translated into six open reading frames, and manually curated. As a result, an additional putative β-defensin, Gal13, was identified in several genomic clones (Table 1). The open reading frame of Gal13 was predicted by GENSCAN 24 and confirmed by directly sequencing of RT-PCR product amplified from chicken kidney.

No other sequence containing β-defensin-like six-cysteine motif has been found in NR, EST or genomic databases, suggesting that 13 Gal genes constitute the entire repertoire of the β-defensin family encoded in the chicken genome. Although it is highly unlikely, we could not rule out the possibility that additional defensin-related genes with distant homology might be uncovered in the chicken by different computational search methods such as the use of Hidden Markov models 1215. It is noted that none of other groups of defensins have been discovered in the chicken, indicating that plant, invertebrate, α-, and θ-defensins are absent in the chicken lineage.

Similar to Gal 1–3, 10 novel β-defensins, deduced from either EST or genomic sequences, vary from 63 to 104 amino acid residues in length. Alignment of these peptides revealed a conservation of the signal sequence at the N-terminus and the characteristic six-cysteine defensin motif at the C-terminus (Figure 1). Consistent with the fact that all β-defensins are a group of secreted molecules in response to infections, the signal sequences of all chicken defensins are hydrophobic and rich in leucines. In addition, the mature C-terminal sequences are all positively charged due to the presence of excess arginines and lysines. Interestingly, Gal11 contains two tandem, but highly divergent, copies of the six-cysteine motif at the C-terminus, and is the only defensin having such sequences. Functional significance for existence of such two defensin motifs remains to be studied.

Phylogenetic analysis of vertebrate β-defensins showed that chicken defensins clustered with various different groups of mammalian β-defensins (Figure 2). However, the bootstrap support for these patterns was very weak (less than 50% in all cases). The clustering of certain chicken β-defensins with mammalian homologues suggests that major subfamilies of β-defensins arose before the last common ancestor of birds and mammals, estimated to have occurred about 310 million years ago 25. This in turn implies that some duplication of β-defensin genes must have taken place before the divergence of birds and mammals. The apparent lack of α-defensins in the chicken and other non-mammalian species (G. Zhang, unpublished data) suggests that α-defensins may have evolved after mammals diverged from other vertebrates.

Comparison of the numbers of synonymous and nonsynonymous nucleotide substitutions provides a powerful test of the hypothesis that positive Darwinian selection has acted to favor changes at the amino acid level 26. This approach has previously been applied to both α- and β-defensins of mammals and has revealed positive selection acting on the mature defensin but not on other regions of the gene 1617. In the comparison of the chicken β-defensin sequences, synonymous sites were saturated with changes or nearly so, making it impossible to test the hypothesis of positive selection in every case. In pairwise comparisons among all sequences, mean p_S in the propeptide region was 0.551 ± 0.036 (S.E.), while mean p_N was 0.369 ± 0.040. In the mature defensin region, mean p_S was 0.673 ± 0.027, while mean p_N was 0.534 ± 0.051. Mean p_N in the mature defensin was significantly greater than that in the propeptide (z-test; P < 0.05), indicating lesser functional constraint on the amino acid sequence of the former. The high mean p_S shows that chicken β-defensin genes have not duplicated recently, unlike β-defensin genes of the bovine 16. In the comparison between the most closely related pair of sequences (Gal6 and Gal7), mean p_S in the mature defensin was 0.221 ± 0.082, while mean p_N was 0.331 ± 0.076. While these values are not significantly different at the 5% level, the fact that p_N was higher than p_S suggested that positive selection may have acted to diversify the mature defensin region between these two genes.

Searching through HTGS database led to identification of two overlapping bacterial artificial chromosome (BAC) sequences, TAM31-54I5 (accession no. AC110874) and CH261-162O9 (accession no. AC146292), both of which were sequenced and deposited earlier by one of us (J.F. Chen). Alignment of these two sequences allowed to re-order three DNA fragments in AC110874 and to construct a continuous, gap-free genomic contig that includes 11 Gal genes except for Gal4 and Gal5. Later search of chicken WGS sequences released on February 29, 2004 confirmed the order of the genomic contig that we assembled and also revealed the locations of two remaining genes, Gal4 and Gal5, both of which reside on a WGS (accession no. AADN01058096) that overlaps with AC110874 (Figure 3). The position and orientation of each Gal gene were obtained by comparing its cDNA with the assembled DNA sequence. As shown in Figure 3, all 13 Gal genes were clustered densely within a distance of 86.0 Kb on the genome. It was also confirmed by aligning such a contig with the chicken genome assembly, in which 13 Gal genes are located on six WGS contigs (Table 1) of chromosome 3 that are only ~3.3 Mb from the distal end. Consistent with this, the Gal gene cluster was physically mapped to the tip of chicken chromosome 3 at the region of q3.5-q3.7 by fluorescence in situ hybridization (FISH) using the TAM31-54I5 BAC DNA as probe (Figure 4).

Comparing the cDNA with genomic sequences also revealed the structure of each Gal gene. Unlike most mammalian β-defensin genes, which primarily consist of two exons and one intron, the Gal genes were found to be composed of four short exons separated by three introns with variable lengths ranging from 117 bp to 3,322 bp (Table 1). Gal12 is an exception, in which the last two exons have been fused together. While the first exon of the Gal genes encodes 5'-untranslated region (UTR) and the majority of the last exon encodes 3'-UTR as well as a few C-terminal amino acids, two internal exons resemble mammalian β-defensin genes in that one exon encodes the signal and pro-sequence and the other encodes the mature sequence with six-cysteine motif 19272829. Apparently, the first two and the last two exons of the Gal genes have joined together during the evolution as a result of exon shuffling, which occurred in many other evolutionarily conserved gene families 30, including invertebrate defensins 5. The fusion of defensin exons in mammals is presumably adaptive because it allows a faster mobilization of such host defense molecules to better cope with invading microbes.

It has been shown that Gal1 and Gal2 are expressed in bone marrow and lung, while Gal3 is more preferentially expressed in bone marrow, tongue, trachea, and bursa of Fabricius 23. To study the tissue expression patterns of novel Gal genes that we identified, RT-PCR was performed with a panel of 32 different chicken tissues. Similar to Gal 1–3, Gal 4–7 are highly restricted to bone marrow cells with Gal5 also expressed in tongue, trachea, lung, and brain at lower levels (Figure 5). By contrast, the six remaining genes, Gal 8–13, were not found in bone marrow, but instead in liver, kidney, testicle, ovary, and male and female reproductive tracts (Figure 5). These results clearly suggested that all chicken β-defensin genes can be divided into two subgroups. Seven genes (Gal 1–7) are predominantly expressed in bone marrow and the respiratory tract, whereas the other six genes (Gal 8–13) are more restricted to liver and the urogenital tract. However, the functional significance and transcriptional regulatory mechanisms of these genes during inflammation and infection remain to be investigated.

To study the origin and evolution of mammalian defensins, a comparative analysis of β-defensin gene clusters in the chicken, mouse, and human was performed by employing additional, more phylogenetically conserved gene markers surrounding the defensin clusters. As shown in Figure 6, two genes, CTSB (Cathepsin B, accession no. NP_680093) and a human EST sequence (accession no. BE072524) immediately located centromeric to chicken defensins, were also found to be conserved in the defensin gene clusters on human chromosome 8p22 and mouse chromosome 14C3. Similarly, another gene, HARL2754 (accession no. XP_372011) that is 6-Kb telemetric to Gal4 is also conserved in another defensin cluster in human (8p23) or mouse (8A1.3) (Figure 6).

These results strongly suggested that all vertebrate β-defensins are evolved from a single gene. This conclusion is further supported by the fact that there are three highly similar β-defensin-like sequences present in the largely finished zebrafish genome (G. Zhang, unpublished data). In addition, a group of homologous β-defensin-like sequences, namely crotamine and myotoxins, have been found in several Crotalus snakes 31, which are presumably derived from a single ancestral gene. The appearance of multiple β-defensin gene clusters on different chromosomal regions in mammalian species 12 is apparently a result of rapid gene duplication, positive diversifying selection, and chromosomal translocation following divergence of mammals from other vertebrate lineages.

In addition to the structural conservation between β-defensin-like sequences in the rattlesnake and mammals 32, a growing body of evidence suggests that their functions appear to be largely conserved in that both are capable of interacting negatively-charged lipid membranes followed by formation of ion channels or pores 323334. It is noteworthy that the conservation of Cathepsin B (CTSB) adjacent to β-defensins is perhaps not surprising, given the recent finding that cathepsins are involved in the cleavage and inactivation of β-defensins 35.

To identify novel defensins in the chicken, all known cysteine-containing defensin-like peptide sequences discovered in plants, invertebrates, birds, and mammals were individually queried against the translated chicken NR, EST, HTGS, and WGS databases in the GenBank by using the TBLASTN program 21 with default settings on the NCBI web site 36. All potential hits were then examined for the presence of the characteristic defensin motif. For every novel defensin identified, additional iterative BLAST searches were performed until no more novel sequences could be revealed. Because mammalian defensins tend to form clusters 12141518, all chicken genomic sequences containing defensin sequences were also retrieved from the GenBank and translated into six open reading frames and curated manually for the presence of the defensin motif in order to discover potential sequences with distant homology.

Multiple sequence alignment was constructed by using the ClustalW program (version 1.82) 37. A phylogenetic tree of amino acid sequences of mature β-defensins was constructed by the neighbor-joining method 38. So that a comparable data set would be used for all pairwise comparisons, any site at which the alignment postulated a gap in any sequence was excluded from the analysis. To maximize the number of sites available for analysis, certain sequences with large deletions were excluded from the analysis. Because the sequences were very short (25 aligned sites), no correction for multiple hits was applied. The reliability of clustering patterns within the tree was assessed by bootstrapping; 1000 bootstrap pseudo-samples were used. The proportion of synonymous nucleotide differences per synonymous site (p_S) and the proportion of nonsynonymous nucleotide differences per nonsynonymous site (p_N) were estimated by the method of Nei and Gojobori 26. Again, no correction for multiple hits was applied because a small number of sites were examined.

To generate a continuous defensin gene cluster, the HTGS and WGS sequences containing the putative defensin genes were retrieved from the GenBank, aligned to generate a longer contig, which was confirmed later by searching through the assembled chicken genome released on February 29, 2004, by using the BLAT program 39 under the UCSC Genome Browser web site 40. The relative positions, orientations, and structural organizations of individual genes were determined by comparing its cDNA sequence to the continuous genomic contig that we assembled.

Fluorescence in situ hybridization (FISH) was used for chromosomal assignment of the chicken β-defensin gene cluster by using the BAC clone TAM31-54I4 as probe, which harbors 11 Gal genes. Metaphase chromosome speads were prepared from mitogen-stimulated chicken splenocyte culture as we described 4142. The BAC clone was labeled by nick translation with biotin 16-dUTP (Roche Diagnostics), hybridized to metaphase chromosome DNA, followed by detection with FITC-labeled avidin (Roche Diagnostics) and staining with propidium iodide to simultaneously induce the R-banding.

Total RNA was extracted with Trizol (Invitrogen) from a total of 32 different tissues from healthy, 2-month-old chickens (see Figure 5). A total of 4 μg RNA from each tissue were reverse transcribed with random hexamers and Superscript II reverse transcriptase by using a first-strand cDNA synthesis kit (Invitrogen) according to the instructions. The subsequent PCR was carried out with 1/40 of the first-strand cDNA and gene-specific primers for each β-defensin and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) as described 2843. Every pair of primers were designed to locate on different exons to aid in distinguishing PCR products amplified from cDNA vs. genomic DNA (Table 2). The PCR program used was: 94°C denaturation for 2 min, followed by different cycles of 94°C denaturation for 20 sec, 55°C annealing for 20 sec, and 72°C extension for 40 sec, followed by a final extension at 72°C for 5 min. The number of PCR cycle was optimized for each gene to ensure linear amplification (Table 2). A half of the PCR products were analyzed by electrophoresis on 1.2% agarose gels containing 0.5 μg/ml ethidium bromide. The specificity of each PCR product was confirmed by cloning of the PCR product into T/A cloning vector, followed by sequencing of the recombinant plasmid.

Abbreviations: Gal, Gallinacin; NR, nonredundant; EST, expressed sequence tag; HTGS, high throughput genomic sequence; WGS, whole-genome shortgun sequence; BAC, bacterial artificial chromosome; FISH, fluorescence in situ hybridization; UTR, untranslated region; GAPDH, glyceraldehyde-3-phosphate dehydrogenase.

YX carried out the tissue collection, RT-PCR analysis of tissue expression patterns, and drafted the manuscript. ALH carried out the phylogenetic and molecular evolutionary analyses. JA and YM carried out the fluorescence in situ hybridization. JFC carried out the sequencing of two chicken defensin-containing BAC clones. DSN participated in tissue collection and preparation. GZ conceived of the study, carried out all computational analyses and annotation, drafted the manuscript, and participated in its design and coordination. All authors read and approved the final manuscript.