Differential transcriptional modulation of duplicated fatty acid-binding protein genes by dietary fatty acids in zebrafish (Danio rerio): evidence for subfunctionalization or neofunctionalization of duplicated genes

Background In the Duplication-Degeneration-Complementation (DDC) model, subfunctionalization and neofunctionalization have been proposed as important processes driving the retention of duplicated genes in the genome. These processes are thought to occur by gain or loss of regulatory elements in the promoters of duplicated genes. We tested the DDC model by determining the transcriptional induction of fatty acid-binding proteins (Fabps) genes by dietary fatty acids (FAs) in zebrafish. We chose zebrafish for this study for two reasons: extensive bioinformatics resources are available for zebrafish at zfin.org and zebrafish contains many duplicated genes owing to a whole genome duplication event that occurred early in the ray-finned fish lineage approximately 230-400 million years ago. Adult zebrafish were fed diets containing either fish oil (12% lipid, rich in highly unsaturated fatty acid), sunflower oil (12% lipid, rich in linoleic acid), linseed oil (12% lipid, rich in linolenic acid), or low fat (4% lipid, low fat diet) for 10 weeks. FA profiles and the steady-state levels of fabp mRNA and heterogeneous nuclear RNA in intestine, liver, muscle and brain of zebrafish were determined. Result FA profiles assayed by gas chromatography differed in the intestine, brain, muscle and liver depending on diet. The steady-state level of mRNA for three sets of duplicated genes, fabp1a/fabp1b.1/fabp1b.2, fabp7a/fabp7b, and fabp11a/fabp11b, was determined by reverse transcription, quantitative polymerase chain reaction (RT-qPCR). In brain, the steady-state level of fabp7b mRNAs was induced in fish fed the linoleic acid-rich diet; in intestine, the transcript level of fabp1b.1 and fabp7b were elevated in fish fed the linolenic acid-rich diet; in liver, the level of fabp7a mRNAs was elevated in fish fed the low fat diet; and in muscle, the level of fabp7a and fabp11a mRNAs were elevated in fish fed the linolenic acid-rich or the low fat diets. In all cases, induction of the steady-state level of fabp mRNAs by dietary FAs correlated with induced levels of hnRNA for a given fabp gene. As such, up-regulation of the steady-state level of fabp mRNAs by FAs occurred at the level of initiation of transcription. None of the sister duplicates of these fabp genes exhibited an increase in their steady-state transcript levels in a specific tissue following feeding zebrafish any of the four experimental diets. Conclusion Differential induction of only one of the sister pair of duplicated fabp genes by FAs provides evidence to support the DDC model for retention of duplicated genes in the zebrafish genome by either subfunctionalization or neofunctionalization.


Background
In his seminal book, 'Evolution by Gene Duplication', Susumu Ohno [1] 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 [1] 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 [2]. 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. [3] 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 [1], 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. [3] that subfunctionalization or neofunctionalization results in the retention of duplicated genes in the genome by investigating the expression of duplicated copies of fatty acidbinding 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][5][6][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][9][10][11][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. [13] 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 rayfinned fishes by the WGD event mentioned above [4][5][6][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 [14] 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 [15]. 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][11][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 [13], 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 [20]. 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][22][23]]. Based on the zebrafish fabp10 sequence [24], Her et al., [22] 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) [23].
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][28][29][30][31][32][33][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. [4] 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% [6]. 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.

The steady-state level of fabp mRNAs does not differ between sexes of zebrafish
Several authors have argued that lipid metabolism is influenced by the sex of an organism. For example, longchain fatty acids are cleared from plasma by the liver more rapidly in female rats than male rats [35]. Bass et al. [36] reported that female rats have a higher intracellular concentration of FABP3 than male rats. We, therefore, assayed by reverse transcription, quantitative polymerase chain reaction (RT-qPCR) the steady-state level of mRNA encoded by each fabp gene in intestine, liver, muscle and brain of male and female zebrafish reared on Tetramin ® (Tetrawerke, Melle, Germany) flake diet. No difference in the steady-state level of fabp mRNAs in a specific tissue (i.e., intestine, liver, muscle or brain) was observed between male and female zebrafish (data not shown).

Effect of diet on the FA composition in different tissues of zebrafish
Zebrafish readily ate all four diets and no differences were observed in their feed intake during the 10 week feeding period. Composition and abbreviated names for the four diets are shown in Table 1. The weight of fish increased from 0.26 ± 0.02 g at the beginning of the feeding trial (~22 weeks of age) to 0.45 ± 0.01 g at the end of the feeding trial (~32 weeks of age) for all diets. At the end of 10 weeks, each diet resulted in a different FA profile for specific tissues. Fig 1 shows the composition of four major FAs in intestine, liver, muscle and brain. For detailed composition of FAs in each tissue see additional file 1: Tables S1-S4. Intestine, liver, muscle and brain of fish fed HD and LND had higher proportions of n-3 fatty acids, due largely to elevated levels of linolenic acid (18:3 n-3), eicosapentaenoic acid (20:5 n-3), and docosahexaenoic acid (22:6 n-3), and reduced proportions of n-6 fatty acids. In contrast, intestine, muscle, and brain of fish fed LD and LFD exhibit higher percentages of n-6 fatty acids, linoleic acid (18:2 n-6) and arachidonic acid (20:4 n-6), and reduced proportions of n-3 fatty acids. The percentages of other fatty acids varied among tissues within each experimental group and also among the experimental groups (see Additional file 1: Tables S1-S4).

Effect of diet on the steady-state level of fabp1a/fabp1b.1/ fabp1b.2 mRNAs in different tissues
No modulation in the steady-state level of fabp1a mRNAs was observed in the intestine ( Fig. 2A) regardless of diet fed to the fish. In the intestine of fish fed LND, the steadystate level of fabp1b.1 mRNAs was higher than those in the intestine of fish fed the other three diets (Fig. 2B). The level of fabp1a and fabp1b.1 mRNAs was below that quantifiable by RT-qPCR in liver, muscle and brain of zebrafish. fabp1b.2 mRNA was detectable in intestine and brain of fish fed the four diets, but the levels in liver and muscle were below the quantifiable range in RT-qPCR assays (data not shown). Furthermore, dietary FAs did not change the steady-state level of fabp1b.2 transcripts in the intestine and brain.

Effect of diet on the steady-state level of fabp7a and fabp7b mRNAs in different tissues
The steady-state level of fabp7a mRNAs was not modulated in intestine (Fig. 3A) or brain (Fig. 3G) of fish fed any of the experimental diets. In contrast, the steady-state level of fabp7b mRNAs was higher in intestine of fish fed LND compared to fabp7b mRNAs in the intestine of fish fed LD and LFD, but not different in intestine of fish fed HD (Fig. 3B). Also, the steady-state level of fabp7b mRNAs was elevated (10-fold) in the brain of fish fed LD compared to the level of fabp7b transcripts in brain of fish fed LFD and LND, but was not different in brain of fish fed HD (Fig. 3H).
The steady-state level of fabp7a mRNAs was 4-fold higher in the liver of fish fed LFD compared to the transcript levels in the liver of fish fed one of the three other diets (Fig.  3C). The steady-state level of fabp7a mRNAs was 3-fold higher in muscle of fish fed LND compared to mRNAs levels for this gene in muscle of fish fed LD, but was not different in the muscle of fish fed HD and LFD (Fig. 3E). No difference was observed in the steady-state level of fabp7b mRNAs in liver of fish fed any of the four experimental diets (Fig. 3D). Finally, fabp7b mRNAs were not quantifiable by RT-qPCR in muscle (Fig. 3F).

Effect of diet on steady-state level of fabp11a and fabp11b mRNAs in different tissues
No difference was observed in the steady-state level of fabp11a mRNAs in the intestine, liver and brain of fish fed any of the four diets ( Fig. 4A, 4B, 4D). The steady-state level of fabp11a mRNAs was, however, 2-fold higher in muscle of fish fed LFD compared to fabp11a transcript levels in muscle of fish fed the other three diets (Fig. 4C). The level of fabp11b mRNAs was not quantifiable by RT-qPCR in liver, muscle and intestine of zebrafish owing to their low abundance. The steady-state level of fabp11b mRNAs weas not changed in the brain of zebrafish fed any of the experimental diets (data not shown).

Modulation of the steady-state level of fabp mRNAs is due to up-regulation of transcriptional initiation
Heterogeneous RNA (hnRNA) for a given gene is rapidly processed to a mature mRNA transcript by removal of intervening sequences and the addition of a 5' cap and a poly(A) tail [37]. As such, the level of hnRNA for a given gene indirectly correlates with the rate of initiation of transcription for that gene. To test whether the observed increase in the steady-state level of fabp mRNAs by dietary FAs was due to increased transcriptional initiation, the steady-state level of hnRNA for a specific fabp gene was  assayed by RT-qPCR in different tissues of zebrafish fed the four experimental diets.
In the intestine of fish fed LND, the level of fabp1b.1 hnRNA was higher than in the intestine of fish fed one of the other three diets (Fig. 5A). The steady-state level of fabp7a hnRNA was unchanged in the liver of fish fed any of the four experimental diets (Fig. 5B). The level of fabp7a hnRNA was higher in muscle of fish fed LND compared to fabp7a hnRNA in muscle of fish fed one of the other three diets (Fig. 5C). fabp7b hnRNA was elevated in the intestine of fish fed LND compared to fabp7b hnRNA in the intestine of fish fed one of the other three diets (Fig. 5D). fabp7b hnRNA was 6-fold higher in the brain of fish fed LD compared to fish fed the rest of the treatment diets (Fig. 5E), while the steady-state level of fabp11a hnRNA was 2-fold higher in muscle of fish fed LFD compared to hnRNA transcript levels in muscle of fish fed HD and LD, but was unchanged in muscle of fish fed LND (Fig. 5F).  [55][56][57]. Dietary longchain FAs are known to activate these nuclear receptors [58][59][60][61][62]. Once activated, these nuclear receptors form heterodimers with retinoic acid receptors (RAR) or retinoid X receptors (RXR) (e.g., PPAR-RXR or PPAR-RAR), which in turn bind to response elements in FABP genes and, thereby, stimulate initiation of transcription [63]. In part, our results showing the increased transcription of fabp genes by dietary FAs are consistent with this mechanism of induction (Figs. 2, 3, 4 and 5). However, none of the fabp genes were up-regulated in any of the tissues assayed in zebrafish fed HD (Table 2). Indeed, the steady-state level of fabp1a, fabp1b.1 and fabp11a mRNAs was lowest in all of the tissues assayed in fish fed HD (Figs. 2 and 4). Also, the steady-state level of Fabp7 mRNAs in rat brain [64], Fabp5 mRNAs in mouse liver [65], and the steadystate level of fabp3 and fabp10 mRNAs in the muscle of Atlantic salmon [47] were not elevated in animals fed diets rich in fish oil. Similarly, Liu et al. [66] showed that PC12 cells from rats exposed to eicosapentaenoic acid and docosahexaenoic acid, both abundant in the HD diet fed zebrafish, did not change the level of Fabp5 transcripts. Based on the results reported here and previous studies [47, [64][65][66], we propose that, besides up-regulation of fabp genes by FAs [67], there may be other mechanisms of transcriptional regulation of the fabp genes, such as repression of transcriptional initiation. Berger et al. [65] suggested that the observed down-regulation of Fabp5 mRNA levels in the liver of mice fed a fish oil diet was mediated via transforming growth factor, beta 1 (TGF1). Although a few reports [67][68][69] emphasize the role of TGF1 in the transcriptional regulation of Fabp genes, the exact mechanism remains unknown.   levels of Fabp1 mRNAs in their intestine than those weaned to a low-fat diet [70,71]. Also, feeding rats a diet containing 6% sunflower oil (rich in linoleic acid) resulted in the greater induction of Fabp1 in intestine compared to rats fed a diet containing 3% sunflower oil [25].

Effect of diet on FA profiles in tissues of zebrafish
Since FAs, such as oleic acid, linoleic acid and linolenic acid, induce the expression of the mammalian FABP1 gene [25,26,70,71], we anticipated that the steady-state level of mRNAs for all the duplicated copies of zebrafish fabp1 (fabp1a, fabp1b.1, and fabp1b.2), might be modulated by the different FAs in one or more of the experimental diets. However, in this study, we only observed up-regulation of the steady-state level of fabp1b.1 mRNAs in the intestine of fish fed LND (Fig. 2) at the level of transcriptional initiation (Fig. 5). The steady-state level of fabp1a (Fig. 2) and fabp1b.2 mRNAs remained unchanged in the different tissues of zebrafish fed experimental diets. We conclude, therefore, that the fabp1a and fabp1b.2 genes, but not the fabp1b.1 gene, no longer respond to dietary FAs owing to loss of cis-regulatory elements in their respective promoters. In their study of cis-regulatory elements after a WGD event in yeast, Papp et al. [72] concluded that the total number of cis-regulatory elements remained unchanged over time between duplicated genes. They further suggest that the number of shared cis-regulatory elements between duplicated genes decreased. The difference in the mode of gene regulation reported here for the duplicated fabp1s may be due to loss of cis-regulatory elements in the fabp1a and fabp1b.2 genes, a mechanism consistent with the process of subfunctionalization.
In contrast, zebrafish fabp1s exhibit a divergent pattern in their tissue-specific distribution of transcripts when compared to their mammalian ortholog. Mammalian Fabp1 transcripts were abundant in the intestine, liver and were detected at lower levels in kidney, ovary, and lung [12,73].  [3], where the combined tissue-specific expression of duplicated zebrafish genes equals the tissue-specific expression of their mammalian ortholog, as described for the duplicated hoxb1a and hoxb1b [74], mitfa and mitfb [75,76], and sox11a and sox11b [77] genes. Comparative functional analysis of cis-regulatory elements of Fabp1 genes will be necessary to determine if the dupli-cated fabp1 genes are retained in the zebrafish genome due to either subfunctionalization or neofunctionalization.

Tissue-specific transcriptional modulation of fabp7a and fabp7b mRNAs by dietary FAs
In this study, duplicated copies of zebrafish fabp7, fabp7a and fabp7b, exhibited distinct patterns of up-regulation by dietary FAs in different tissues for steady-state levels of both mRNA and hnRNA (Figs. 3 and 5). The steady-state level of fabp7a mRNAs in muscle and liver were elevated in zebrafish fed LND and LFD, whereas the steady-state level of fabp7b mRNAs was modulated in brain and intestine of the fish fed LD and LND.
Mammalian Fabp7 transcripts are detected in brain, retina, spinal cord and mammary gland [8,12]. Zebrafish fabp7a transcripts are detected in brain, spinal cord, retina, testis, liver, intestine, and muscle [18] and zebrafish fabp7b transcripts are detected in brain, retina, testis, liver, intestine, skin, and swim bladder [18]. In addition, the steady-state level of zebrafish fabp7b transcripts, but not fabp7a transcripts, was up-regulated in the brain of zebrafish fed LD (Fig 3).
Since the duplicated copies of the zebrafish fabp7, fabp7a and fabp7b, show a different tissue-specific pattern of expression compared to their mammalian ortholog and exhibit a different mode of gene regulation by dietary FAs from each other, it is possible that either subfunctionalization or neofunctionalization may account for the retention of both duplicated fabp7 genes in the zebrafish genome.

Transcriptional modulation of fabp11a, but not fabp11b, by dietary FAs
Previously, we reported that the duplicated copies of the zebrafish fabp11 gene and the tetrapod FABP4, FABP5, FABP8 and FABP9 genes are derived from a common ancestral gene [33]. To date, the fabp11 gene has only been identified in fishes [33,78]. This study shows that both the steady-state level of fabp11a mRNAs (Fig. 4) and hnRNAs (Fig. 5) was only elevated in muscle of fish fed LFD, but not in muscle of fish fed one of the other three diets. Although fabp11a and fabp11b exhibit different tissue-specific patterns of expression in embryos, larvae and adult zebrafish [29,33] and differential regulation by dietary FAs, we are unable at this time to resolve whether these duplicated genes were retained in the genome by either subfunctionalization or neofunctionalization as no ortholog of this gene has been identified thus far in other species, such as birds or mammals.

Conclusion
In this study, we show that dietary FAs change the FA profile in intestine, liver, muscle and brain of zebrafish. The tissue-specific changes in FA content modulated the steady-state level of mRNAs for only one sister duplicate of three pairs of duplicated genes of the fabp multigene family in zebrafish (Table 2). Furthermore, changes in fabp hnRNA directly correlated with changes in the steadystate levels of fabp mRNAs, suggesting that the affects of FAs on these fabp genes occurred at the site of transcriptional initiation ( Table 2). These findings indicate that the retention of duplicated fabp genes in the zebrafish genome is most likely the result of either subfunctionalization or neofunctionalization. To distinguish between these processes as outlined in the DDC model [3] will require functional analysis of the fabp promoters to identify ciselements responsible for transcriptional induction by FAs.

Diets and fish husbandry
Four isoproteic (41% crude protein) diets differing in FA composition and lipid content were formulated (Table 1). In addition to National Research Council recommendations on the nutritional requirements of warm-water fishes [79], we were guided by the results of previous dietary studies in zebrafish [43-45, 80,81] in the formulation of the diets used in this study. All dry ingredients were mixed in a Hobart mixer. Butyl hydroxy anisole was dissolved in the lipid source and choline chloride was dissolved in distilled water. Both of these solutions were then added to dry ingredients to make a wet dough. The wet dough was scrubbed through an 800 m screen and the resulting wet particles were freeze-dried and stored at -20°C. The particle size was 600-800 m in diameter, a size interval for feeding appropriate to the age and weight of the fish used in this study. Mean weight of fish at the start of the feeding trial was 0.26 g. Highly unsaturated FA-rich diet (HD), linoleic acid-rich diet (LD), linolenic acid-rich diet (LND) contained 12% lipid, whereas the low fat diet (LFD) was composed of 4% lipid ( Table 1). The FA composition of the four diets is shown in Table 3. The concentration of linoleic acid (18:2 n-6) was highest in LD (44.82%), while the concentration of linolenic acid (18:3 n-3) was highest in LND (33.53%). Among the four diets, the concentration of eicosapentaenoic acid (20:5 n-3) and docosahexaenoic acid (22:6 n-3) was highest in HD (6.14% and 20.94%, respectively).
One hundred and eighty wild-type inbred zebrafish siblings (mean wt. 0.26 g), bred from 2 males and 2 females (obtained from a local aquarium supply store) to control for genetic variance, were randomly assigned to 12 aquaria (35 l) with 15 fish per tank at approximately 20 weeks of age. Each dietary treatment had three tank replicates to control for environmental variance. Fishes were acclimatized for 15 days and fed Tetramin ® (Tetrawerke, Melle, Germany) twice a day to satiation. The Tetramin diet was discontinued and then fish were fed one of the four diets twice a day at 0900 h and 1900 h to satiation for 10 weeks. A constant water temperature (28°C), a light/ dark cycle (14/10 h) and other water parameters were maintained according to Westerfield [82]. Feeding was stopped 24 h prior to sampling of tissue. Fish were individually euthanized by immersion in 0.2% MS-222, weighed and dissected to collect intestine, liver, muscle and brain tissues for RNA and FA analysis. Animal husbandry and protocols for this experiment were reviewed by the Animal Care Committee of Dalhousie University in accordance with the Canadian Council on Animal Care.

RNA isolation, cDNA synthesis and RT-qPCR
Total RNA was isolated from intestine, liver, muscle and brain of zebrafish using the standard TRIzol method (Invitrogen, Carlsbad, California, USA). The quality and quantity of extracted RNA was assessed by agarose gelelectrophoresis and spectrophotometry at 260 nm, respectively. Messenger RNA in 2 g of total RNA was converted to cDNA using an oligo(dT) primer according to the manufacturer's instructions for the Omniscript RT kit (Qiagen, Mississauga, Ontario, Canada). cDNA was synthesized from heterogeneous nuclear RNA (hnRNA) using random hexamers. Primer pairs for quantification of mRNA and hnRNA encoded by different fabp genes and the annealing temperature for primer pairs for each fabp gene are shown in Table 4. For assay of gene-specific hnRNA, one primer was based on an intronic sequence, while the other was based on an exonic sequence (see Table 4 for details). Primers for the amplification of elongation factor 1 alpha (ef1) mRNA by RT-qPCR are based on a previous study [83].
The target sequence for each gene was quantified to generate a standard curve of known copy number. Amplification of cDNA samples and DNA standards was carried out  Table 4), and 30 s of elongation at 72°C. Following the PCR cycles, the melting temperature of the PCR product was determined to assess its purity. Fluorescence was measured following each cycle. The copy numbers of mRNA and hnRNA for each fabp gene were determined using the standard curves as explained by Bustin et al. [84]. As negative controls, the reverse transcriptase was omitted from cDNA synthesis reactions for each sample and these controls were subjected to RT-qPCR. To determine the relative steady-state level of fabp mRNA and hnRNA in each tissue, the absolute copy number of fabp mRNA and hnRNA transcripts was divided by the copy number of ef1 mRNA and hnRNA transcripts in each sample.

Authors' contributions
SK and JMW conceived and designed the research; SK conducted the experimental work and statistical analysis; SPL provided expertise in the design of diets and fatty acid analyses; ED-W assisted in design and interpretation of