Some of the first data pointing to a possible genome duplication in fish came from Hox genes and Hox gene clusters. Hox genes encode DNA-binding proteins that specify cell fate along the anterior-posterior axis of bilaterian animal embryos and occur in one or more clusters of up to 13 genes. Whereas lobe-finned fish, amphibians, reptiles, birds and mammals have four clusters, extra Hox gene clusters have been discovered in zebrafish, Medaka, Nile tilapia and pufferfish 4. The observation that such distantly related species 5 all have seven or eight Hox gene clusters suggested the occurrence of an additional genome-duplication event in the ray-finned fish lineage before the divergence of most teleost (bony fish) species. More recent comparative genomic studies have turned up many more genes and gene clusters for which there are two copies in fish but one in other vertebrates 6. The findings that different paralogous pairs seem to have originated at about the same time, that different fish species seem to share ancient gene duplications, and that different paralogs are found on different linkage groups in the same order as other duplicated genes, all support the hypothesis that these genes arose through a large-scale gene-duplication event. It is worth noting, however, that some authors have argued that an ancestral whole-genome-duplication event was not responsible for the abundance of duplicated fish genes 7.

Additional evidence for a genome duplication in ray-finned fish was provided by analyzing the complete Takifugu genome, a draft sequence of which was published in 2002 2. Two recent studies identified duplicated genes in this genome and used phylogenetic trees to estimate the ages of these duplicates 89. Vandepoele et al. 8 constructed phylogenetic trees for all gene families containing between two and ten duplicated Takifugu genes, which amounts to a total of 3,077 families. For each gene family, the relative date of duplication events was determined to test whether gene duplications occurred before or after the split between fish and land vertebrates. To this end, neighbor-joining trees were created for each of the Takifugu gene families with homologous sequences from mouse and human. Absolute dating of duplication events was achieved through inference from linearized trees 10. In such trees - where branch length is directly proportional to time - the split between ray-finned fish and land vertebrates, dated at 450 Mya, was used as a calibration point for the dating of gene-duplication events. A major fraction (about one-third) of the duplicated genes in Takifugu could be ascribed to a large-scale gene-duplication event specific to the fish lineage, which was estimated to have occurred about 320 Mya (Figure 1).

A very similar approach was followed for the analysis of the Takifugu genome by Christoffels et al. 9, who obtained essentially the same result: by constructing linearized trees, the whole-genome-duplication event was estimated to have occurred approximately 350 Mya. To test whether the sudden increase in the number of duplicated genes in the Takifugu genome was the result of an entire-genome duplication rather than an increased rate of independent tandem-duplication events, both Vandepoele et al. 8 and Christoffels et al. 9 investigated the appearance of duplicated genes in duplicated blocks on chromosomes. Statistically significant regions of micro-colinearity were identified within the complete Takifugu genome, showing the same gene content and gene order. Indeed, both studies reported a large number of duplicated genes in so-called paralogons - homologous genomic segments that can be proved to have been created by duplication 11 - and concluded that most findings were congruent with a large-scale, probably whole-genome duplication event in a ray-finned ancestor that gave rise to the Takifugu and other fish lineages.

Because of the highly fragmented nature of the initial Takifugu genome assembly, it was difficult to prove that the large-scale gene-duplication event had indeed affected the whole genome. The recent release of the well-assembled Tetraodon genome 3 seems to have settled this issue in two ways. First, Jaillon et al. 3 analyzed the chromosomal distribution of ancient duplicates and observed that genes on one chromosome have a strong tendency to have duplicate copies on a single other chromosome. As would be expected from a whole-genome-duplication event, all chromosomes are involved. Second, by using a comparative approach in which they compared the Tetraodon genome with that of human, which has not undergone the genome-duplication event (Figure 1), Jaillon et al. 3 showed that almost every region in the human genome clearly corresponds to two regions in the Tetraodon genome. This type of comparative analysis (Figure 2) has proved very powerful for unveiling genome-duplication events. Recently, such an approach provided overwhelming evidence for the long-suggested 12, but contested (see, for example, 13) ancient whole-genome duplication in the yeast Saccharomyces cerevisiae, by comparing its genome with that of different relatives that diverged prior to the duplication events 1415.

A comparative analysis between the human and Tetraodon genomes has also allowed inference of the basic structure of the ancestral bony vertebrate genome, and the reconstruction of much of the evolutionary history of ancient and recent chromosomal rearrangements leading to the modern human karyotype. By matching up the genes on the Tetraodon chromosomes with homologs on human chromosomes, Jaillon et al. 3 inferred that the ancestor of both fish and land vertebrates had no more than 12 chromosomes, a number that has been previously suggested on the basis of linkage relationships between zebrafish, Medaka, and human 16. Comparison of the genomes of Tetraodon and human also showed that chromosome evolution in both lineages differed considerably. Whereas all but one of the ancestral Tetraodon chromosomes had not undergone interchromosomal exchange for 450 Mya, only one human chromosome was similarly undisturbed. A possible explanation for the difference in genome evolution might be the massive integration of transposable elements in the human genome, with an increased overall frequency of chromosome breaks as a result 3.

As mentioned above, on the basis of previous analyses of the Takifugu genome, the whole-genome-duplication event in fish is thought to have occurred somewhere between 300 and 350 Mya 89. An interesting question is whether this date correlates with a decisive period in the evolution of the fish. For instance, if the genome duplication had been responsible for the biological diversification and large number of ray-finned fish, as suggested previously 417, it must have occurred prior to the radiation of most fish lineages. The class Actinopterygii includes more than 23,500 species 18, of which the vast majority are teleosts or ray-finned fish. Interestingly, all older, more basal groups of ray-finned fish, namely Polypteriformes (bichirs), Acipenseriformes (sturgeons and paddlefish), Semionotiformes (gars), and Amiformes (bowfin), have only a few extant species (Figure 1). Most members of these basal actinopterygian lineages are considered to be 'living fossils', because their morphology has remained unchanged over very long evolutionary time periods.

In a recent study, Hoegg et al. 19 have tried to determine the timing of the duplication event in relation to the origin of lineages of teleost and 'nonteleost' fish by sequencing three nuclear genes - fzd8, sox11 and tyrosinase - from sturgeons, gars, bony tongues, and a tenpounder. For these three genes, two copies have been described previously in derived teleost model species, such as zebrafish and pufferfish, but only one orthologous copy has been found in tetrapods. The specific clustering of the genes in individual gene trees for these three genes and a dataset of concatenated genes support the hypothesis that the fish-specific genome-duplication event took place after the split of the Acipenseriformes and the Semionotiformes from the lineage leading to teleost fish, but before the divergence of Osteoglossiformes (bony tongues) and the other more derived groups of fish (Figure 1). This is in good agreement with the recent analyses of the Takifugu genome, as fossil data age the Semionotiformes at between 245 and 286 million years, whereas molecular estimates for the Amiiformes, which are of approximately the same age as the Semionotiformes, hint at a separation from the Teleostei stem lineage about 367-404 Mya. Likewise, molecular data suggest an age of 335 million years for the Osteoglossiformes 19. The inferred relative and absolute dates for the fish-specific genome duplication event seem to separate the species-poor branching lineages from the species-rich teleost lineages, providing evidence that the fish-specific genome duplication might be related causally to an increase in species and morphological diversity.

On the basis of isozyme studies, Werth and Windham 20 developed a model in which the 'reciprocal silencing' of genes in geographically separated populations would promote speciation. A few years ago, this idea was revived in a model called 'divergent resolution', in which the loss or silencing of gene duplicates was postulated to be more important for the evolution of species diversity than the acquisition of new functions by duplicated genes. Divergent resolution occurs when different copies of a duplicated gene are lost on different chromosomes in different populations, thereby creating genetic barriers for reproduction between them 2122. Divergent resolution and lineage-specific subfunction partitioning 17 can promote incompatibility among populations within a species, and thus might facilitate evolutionary radiation. Gene duplications might, therefore, bring about rapid speciation in populations fixed for different copies of a duplicated locus. The fish-specific genome duplication has created many duplicates that could be divergently resolved. Potentially, such genes have played a prominent role in the radiation of the teleosts. Further studies of the genes encoded in these fish genomes may shed light on how important the fish-specific whole-genome duplication has been in the evolution of the ray-finned fish.