Several of the genes included in the β-defensin cluster of the chicken have been highly conserved both in the zebra finch and in the chicken since they split around 100 million years ago. However, new avian β-defensin genes have been acquired through duplications along the two ancestral lineages, especially in the zebra finch genome. After the duplications, the new genes have evolved under relaxed purifying selection (although some amino acids show evidence of positive selection) compared to the non-duplicated genes.
The zebra finch genome contains a cluster of 22 β-defensin genes, of which ten were found to have orthologous genes in the chicken genome, originating from ancient gene duplications before the G-P split. The remaining 12 genes have evolved by a series of more recent gene duplication events. The long branches and low basal resolution in the phylogenies (figure 2 and 3) suggests that the different orthologous genes have occurred by ancient gene duplication events and then been conserved along the evolutionary lineages of the chicken and the zebra finch, i.e., in the region of 100 million years. That β-defensin genes arose by ancient gene duplications has been suggested by earlier studies [5, 17, 25]. The functions of the avian β-defensins in chicken are increasingly well understood, both in terms of where in the body they are being expressed but also how effective they are against different pathogens (for summary see [16, 33]). The overall patterns of tissue specific expression of zebra finch AvBDs do coincide to large extent to that found in chicken (for detailed comparison see van Dijk 2008  and table 2). Similar to chicken, AvBD2, 8, 9, and 10 was expressed in the liver, AvBD9 and 10 in testis and skin, AvBD2 and9 in spleen. The discrepancies are AvBD2 and 13 that in the zebra finch that also was found to be expressed in the skin, AvBD7 and 10 was additionally found in the spleen (table 2).
Within a small region of 56 Kbp in the zebra finch genome a series of more recent gene duplications have occurred. Within this short physical distance on chromosome 3, there are two AvDB genes in the chicken genome. These two genes have undergone several duplication events and can now comprise 12 different AvBD genes in the zebra finch genome. That these duplicated genes have not lost their function is indicated by several factors.
Firstly, our data on expressed genes together with a similar study on expressed genes in the brain of zebra finch  identify four or possible five of the duplicated genes (with two and three expressed genes from each of the duplication clusters). In this study, we identified the genes AvBD123 (expressed in spleen) and 125 (expressed in the testis) that are a result of a duplication event (Table 2). The third expressed gene belonging to duplication cluster identified in this study was AvBD115 (also with expression in the spleen). AvBD115 was together with AvBD117α/β also found to be expressed in the separate study of . In their study the expressed sequence tag, FE7232851, was identical with AvBD115 and ESTs, DV954612.1 and FE728335.1, are likely alleles of AvBD 117α/β as it only differed by three SNPs, two of which have caused nonsynonomous substitutions at position 34 and 56 (figure 4) changing Threonine (T) to Serine (S, both amino acids with uncharged polar R groups) and FE728335.1 had an additional SNP causing a nonsynonomous substitution on position 41 changing an Phenylalanine (F) to Tyrosine (Y). It should be noted that a blast search of the EST against the zebra finch genome yielded AvDB 117α and β as the closest genes. A possible reason that some of the identified genes were not found to be expressed in the investigated zebra finches could be due to tissue specific expression in organs not investigated by us. In fact, in chickens several of the β-defensins are expressed in tissues not evaluated in our study. These are manly tissues associated with internal organs such as bursa, the intestines, lungs but also bone marrow and leucocytes . Further, if no infection or immune response associated with the investigated genes have been initiated then we would not expect to see the gene being expressed.
Secondly, several of the genes share identical signal peptides, although the mature peptides have evolved new amino acid sequences (figure 2, 3 and 4). Thirdly, within cluster B the amino acids under positive selection (Table 3) are similar to the sites known to be under selection in the functional chicken AvBD genes (Table 3 and figure 2 and 3, ).
Several models have been put forward about how new genes and gene functions might arise through gene duplication. Initial models (different versions of the neo-functionalization model) [35, 36] suggested that after duplication a gene is free from selection pressure and can thereby accumulate mutations that could lead to new functions. However, the main problem with these models is that they do not explain why the original (neutral) duplication is not lost by drift in the population [37, 38]. If the original gene have evolved to conduct more than one function then the two functions can be divided between the two new genes thus escaping adaptive conflicts between the two functions (sub-functionalization models) [35, 39, 40]. Even here, the duplicated genes are released from selection until one of the functions has been lost, although this time might be shorter than in the neo-functionalization model, as in this case mutations have to cause losses of a function, instead of gaining new specialised ones . However, a model (the innovation, amplification, duplication, IAD model), that maintains duplicated genes under continuous selection has recently been proposed by Bergthorsson and co-workers . First, innovation, in which the original gene product has both a primary function and a secondary function which have neither deleterious nor beneficial properties. With a change in ecological niche, or for immune genes, the encounter of a new pathogen fauna, the secondary function becomes beneficial and an increase of the activity is selected for. In the second step, amplification, an increase of the secondary activity is achieved by gene duplication and thus selected for. Lastly, divergence, being freed from its original function the duplicated gene can now evolve improvements in what previously was a secondary function, thereby diverging from its original gene. In the present dataset there are several indications that the IAD model might be able to explain how the AvBD genes have duplicated and diverged in the zebra finch genome, as discussed below.
Individual AvBD genes have been shown to have activity against several different groups of pathogens [16, 41]. The effectiveness against the different groups of pathogens might, however, differ between the different AvBD genes [41, 42, 43] thus depending on the environment having primary and secondary function against different pathogens. In two cases in the zebra finch genome very recent events of gene duplication can be observed, in one case the mature peptides are identical (i.e. AvBD117 α and β) and in the second case (AvBD121 and 122) the mature peptide is highly similar. If having multiple copies of an AvBD gene increases the gene expression, as found when investigating gene-copy variation of human β-defensins . Then the effectiveness against certain pathogens might increases simply due to an increase in dosage, as seen in vitro . Of interest would be to investigate whether the expression of the AvBD117 α and β is associated with an increased expression, in relation to other single copy genes. Once duplicated, the genes should start to diverge in order to be optimized for the function that was formerly only a secondary function. In the cluster of duplicated genes in the zebra finch genome selection has been relaxed (Figure 6a and 6b), compared to the genes that have not been duplicated. It should also be mentioned that in the case of the more recent duplication event in the chicken genome (AvBD6 and 7, figure 2) strong positive selection has acted upon the genes since the duplication event (pairwise comparison, ML, ω = 1.8). As defensins have been found to have broader functions such as signalling to the innate immune system [12, 13, 44], it might also be that the diversification of the duplicated genes in the zebra finch genome is associated with selection of other functions than direct disruption of pathogen cell membranes. In the chicken genome the expression patterns of the duplicated genes AvBD6 and 7 show similar tissue expression pattern , suggesting that they have retained similar functions and both have kept strong antimicrobial activity in vitro [16, 43].
It can not be excluded that our finding that AvBD14 has been lost somewhere along the zebra finch lineage is an artefact from a misassemble of the zebra finch genome. It may also be that our method for finding β-defensin genes was not general enough to find this gene. Furthermore, AvBD14 is yet to be annotated in the chicken genome (although the coding sequence for this gene was deposited in GenBank in 2006 there is still no publication describing the finding of this gene). However, an ortholog gene to AvBD 14 has been identified in the Turkey genome indicating that the chicken AvBD14 is not an assembly artefact (David Burt, pers. comm.)
We have shown here, together with previous studies on mammals and frogs, that antimicrobial peptides such as β-defensins are evolving through gene duplication and positive selection in vertebrates [45, 46]. This stand in contrast the counterpart of AMPs in the immune system in insects (i.e. Drosophila spp.) where studies several times failed to detect positive selection, although duplications seems to have been common [47, 48, 49]. As insects lack the adaptive immune system, selection might work on other mechanisms that regulate the gene expression instead of making the peptides more specific . To fully understand the different evolutionary pressures imposed on the immune system of different organisms, we call for studies that examine not only the selective forces acting upon the peptide itself but also studies examining the selection on regulatory elements associated with the peptide expression.
Studying the evolution of immune genes is associated with additional challenges compared to other genes as they might lose functionality by not evolving as pathogens constantly are evolving in order to evade the function of the immune genes. However, as shown in this study, the steady increase in the availability of genomic data makes comparative studies possible in order to understand the underlying mechanisms of evolution and diversification of genes in the immune system.