In his seminal book, 'Evolution by Gene Duplication', Susumu Ohno  argued that one of the major mechanisms that facilitate the increasing complexity in the evolution of life is duplication of genes and whole genomes. Models to explain retention of duplicated genes in eukaryotic genomes have undergone a development of thought since Ohno proposed his model almost 40 years ago. Ohno  argued that most duplicated genes are lost from the genome owing to nonfunctionalization, a claim which has been validated by empirical evidence from Lynch and Conery . Nonfunctionalization is a process where deleterious mutations accumulate in the coding region of a gene giving rise to either a dysfunctional protein or no protein product. Duplicated genes might, however, be retained in the genome owing to mutations in the coding region that led to a novel function for the protein product of a gene, a process Ohno termed 'neofunctionalization'. This model by Ohno for the preservation of duplicated genes came to be known as the 'classical model'. Data primarily derived from genome sequencing projects over the past decade suggest that a much higher proportion of gene duplicates is preserved in the eukaryotic genome than predicted by Ohno's "classical model". To explain this apparent "conundrum", Force et al.  proposed the Duplication-Degeneration-Complementation (DDC) model in which subfunctionalization serves as an alternative mechanism, but not to the exclusion of neofunctionalization, for the preservation of duplicated genes. According to the DDC model, duplicated genes are retained in the genome either by subfunctionalization, where the functions of the ancestral gene are sub-divided between sister duplicate genes, or by neofunctionalization, where one of the duplicates acquires a new function. In this model, both processes occur by either loss or gain of cis -acting regulatory elements in the promoters of the duplicated genes. As with the "classical model" of Ohno , the DDC model proposes that most duplicated genes are lost from the genome (i.e., nonfunctionalization) owing to an accumulation of deleterious mutations in coding or control regions leading to functional decay.
We chose to test the DDC model of Force et al.  that subfunctionalization or neofunctionalization results in the retention of duplicated genes in the genome by investigating the expression of duplicated copies of fatty acid-binding protein (Fabp) genes, members of the multigene family of intracellular lipid-binding protein (iLBP) genes, in zebrafish for two reasons. First, bioinformatic resources for zebrafish are readily available, including linkage maps, extensive expressed sequence tags (EST) http://www.zfin.org and an almost complete genome sequence database http://www.ensembl.org/Danio_rerio. Second, and most importantly, owing to a whole genome duplication (WGD) event, an event that occurred early in the ray-finned fish radiation about 230-400 million years ago [4–7], we predicted that many members of the iLBP multigene family might exist as duplicated copies. This prediction has proved to be correct (see below).
The iLBPs are encoded by a highly conserved multigene family, which consists of the fatty acid- (FABP), cellular retinol- (CRBP) and cellular retinoic acid-binding protein (CRABP) genes [reviewed in [8–12]]. Currently, 17 paralogous iLBP genes have been identified in animals, but no member of this multigene family has been identified in plants and fungi. Schaap et al.  have suggested, therefore, that the first iLBP gene emerged after the divergence of animals from plants ~930-1000 million years ago. This ancestral iLBP gene presumably underwent a series of duplications followed by sequence divergence, giving rise to the diversity of the extant iLBP multigene family. This multigene family has been further augmented in ray-finned fishes by the WGD event mentioned above [4–7].
Originally, iLBP genes and their proteins were named according to the initial tissue of isolation, e.g., liver-type FABP (L-FABP), brain-type FABP (B-FABP), etc. Owing to some confusion with this earlier nomenclature, here we use the nomenclature proposed by Hertzel and Bernlohr  in which numerals distinguish the different FABP proteins and their genes, (e.g., FABP1 (liver-type), FABP7 (brain-type)).
FABP1, the first FABP isolated, was described almost 40 years ago . Although extensive studies in mammals have focused on the tissue distribution and binding activities of FABPs, the regulation and evolution of their genes, and mice FABP gene knock-outs [8, 10–12], the precise physiological roles of FABPs remain unclear. However, accumulated data have provided evidence that FABPs play an important role in uptake, sequestering and transport of fatty acids and interaction with other transport and enzyme systems. Indirect evidence suggests other putative physiological functions for FABPs, such as: (i) transport of fatty acids to the nucleus to regulate gene transcription via activation of the nuclear receptors, the peroxisome proliferator-activated receptors (PPAR) [see e.g., [12, 16]]; (ii) essential functions in early development, especially neural growth and differentiation [see [17, 18] and references therein]; and (iii) a role in human diseases [10, 12].
Although the coding sequence and structure of the FABP genes has been well conserved over millions of years, each FABP gene exhibits a distinct, yet sometimes overlapping pattern of tissue-specific expression with other FABP genes. If all FABP genes in this multigene family arose from a single ancestral gene as proposed , the regulatory elements in the promoters must have evolved different functions in the subsequent duplicated genes giving rise to the complexity of the spatio-temporal expression of this multigene family. Regulatory elements in the promoters of some mammalian FABP genes and one insect have been identified [8, 19]. As such, cis -acting elements that determine the spatio-temporal expression of these genes, with a few exceptions, are not well defined. From sequence data, it appears that mammalian iLBP gene-promoters consist of a modular structure similar to many eukaryotic promoters, comprised commonly of a TATA box with proximal and distal regulatory elements . Currently, our understanding of the regulatory elements that control the expression of the FABP genes is modest and limited primarily to mammals [but see [21–23]]. Based on the zebrafish fabp10 sequence , Her et al.,  cloned its promoter and by functional analysis, identified a 435 bp regulatory element that is sufficient to modulate the liver regional expression in transgenic zebrafish. How this cis- acting element functions is not known. In addition to functional promoter studies, fatty acids and peroxisome proliferators have been shown to induce the transcription of some FABP genes in mammals [25, 26] via activation of PPAR, or other unknown transcription factors, that bind to a fatty acid (FA) response element (FARE) .
To date, we have characterized 11 zebrafish FABP genes with respect to their cDNA sequence, gene structure, chromosome location, conserved gene synteny with their mammalian orthologs, and their spatio-temporal patterns of expression in embryos, larvae and adults [17, 18, 24, 27–34]. Based on phylogenetic analyses and conserved gene synteny with their single-copy, mammalian orthologs, eight (four pairs) out of 11 of the extant members of the zebrafish FABP genes arose as a result of the ray-finned fish-specific WGD event [[18, 32, 33] and unpublished data]. One pair of genes, fabp1b.1 and fabp1b.2, are tandemly-arrayed on chromosome 8 separated by ~4 kb of DNA [unpublished data]. This duplication, which was subsequent to the WGD but is not yet dated, is presumably the result of unequal crossing-over between homologous chromosomes during meiosis. The number of duplicated FABP genes (63%) retained in the zebrafish genome owing to the WGD event in the ray-finned fishes lineage is remarkable as Postlethwaith et al.  estimate that only 20% of all the duplicated genes following WGD have been retained in the zebrafish genome. Other estimates for retention of duplicated genes in the zebrafish genome from the WGD are 14-30% . Three zebrafish FABP genes, fabp2, fabp3 and fabp6 exist as single copies (a duplicate of fabp10 has recently been identified by us [unpublished data]). Following the WGD event, the sister duplicates of these genes have presumably been lost by accumulation of mutations leading to functional decay.
In mammals, FAs up-regulate the expression of some FABP genes as evidenced by an increase in mRNA and protein levels [25, 26]. We hypothesized that (i) zebrafish fabp genes might be up-regulated by dietary FAs, and (ii) sister duplicated fabp genes might be differentially modulated by dietary FAs. We show by assaying steady-state mRNA and heterogeneous nuclear RNA (hnRNA) levels for three sets of duplicated fapb genes, fabp1a/fabp1b.1/fabp1b.2, fabp7a/fabp7b and fabp11a/fabp11b, that dietary FAs modulate the transcriptional initiation of only one of the duplicated fabp genes in specific tissues of zebrafish. This result provides compelling evidence that these duplicated fabp genes were retained in the genome by either subfunctionalization or neofunctionalization owing to the divergence of cis -acting regulatory elements in the fabp gene promoters.