Microscopic examination of the ovaries collected from non-vitellogenic (NV) females showed undeveloped, small sized ovaries containing transparent stage I oocytes (Fig. 1A, D). In comparison, ovaries collected from 2-month old (Vit2) fish contained oocytes at the perinucleolar stage (stage II) and oocytes with cortical alveoli, yolk bodies (stage III) and a conspicuous germinal vesicle (Fig. 1B, E). Ovaries from 4-month old females (Vit4) contained stage III but also larger stage IV oocytes and oocytes undergoing germinal vesicle breakdown (GVBD; Fig. 1C, F), carrying yolk bodies with crystalline yolk accrue.

E2 levels increased in females according to their vitellogenic stage (Fig 2). The levels of E2 in plasma of vitellogenic females (Vit2 and Vit4) were similar to those of E2-treated males (p > 0.05). The E2 plasma levels of NV and of control males were significantly lower than those of vitellogenic females (Vit2 and Vit4) or E2-treated males (p < 0.01) and the plasma E2 levels of control males did not significantly differ from those of NV females (p > 0.05). Similar expression levels of esr1 were found by real-time PCR (Fig 2) in liver samples from Vit2 females, Vit4 females and E2-treated males (P > 0.05). The expression level of esr1 in these samples was significantly higher than those found in liver samples from NV females or control males (p < 0.05). The expression of esr1 was very low in the NV females and undetectable in the group of control males. Transcripts of vtg3 were detected only in vitellogenic females (Vit2 and Vit4) and E2-treated males. Expression levels of vtg3 were three fold higher in Vit4 females and E2-treated males than in Vit2 females (p < 0.01).

Only 16,252 transcripts out of 16,399 were left for further analyses after processing the intensity values of each spot on the arrays and after normalization. SAM analysis revealed 2,523 transcripts (see Additional file 1) differentially expressed in livers of vitellogenic females (Vit4 and Vit2), NV females, E2-treated males and control males. The top 20 most differentially expressed transcripts (Table 1) included transcripts of vtg3 and esr1 that were used as markers for vitellogenesis and E2 treatment, respectively. The list also included transcripts of vtg1, nothepsin (nots), ESTs similar to spectrin, nuclear envelope 1 (syne1; AW019847), EST similar to follistatin-like 1 (fstl1; AI884233), EST similar to nitric oxide synthase interacting protein (nosip; AI545274) and EST similar to glutamate receptor, ionotropic, kainate 1 (grik1; BM071695), that were reported previously to be regulated by E2. The transcript of vtg3 (Table 1; see Additional file 1) was the most up-regulated transcript in the livers of E2-treated males and Vit2 or Vit4 females when compared with control males or NV females, respectively. The abundance of vtg1 transcripts was much lower, but significantly higher in E2-treated males than in vitellogenic females. A similar higher expression pattern of transcripts in E2-treated males vs. vitellogenic females (Table 1) was found for nots, fstl1 and EST similar to lectin mannose- binding 1 precursor (lman1; AW175541). Several transcripts were more abundant in vitellogenic females (Vit2 or Vit4) than in E2- treated males, including the gene ankyrin (ank), transcripts of EST similar to ectonucleoside triphosphate diphosphohydrolase 4 (entpd4; AW344063) and an EST similar to family with sequence similarity 20C (fam20c) (Table 1).

Correspondence analysis (COA) was performed to study the associations between the 2,523 differentially expressed transcripts in the five tested groups (Fig 3). The results indicated close association and sharing of the same expression pattern of Vit2 with Vit4, and of NV females with control males. The group of E2-treated males did not show any association with the other groups. Most of the transcripts in the groups of E2-treated males and vitellogenic females (Vit2 and Vit4) were up-regulated in comparison with NV and control males (Fig 3).

Coupled Two-Way Clustering of the 2,523 differentially expressed transcripts was performed to reveal similarity in the expression patterns in the five tested groups. Three main clusters were revealed (Fig 4): 1) A small cluster of 34 transcripts with high expression in Vit4 females and low expression in other samples; 2) A cluster of 985 transcripts (Fig 4; subdivided into 2a, 2b, 2c and 2d) showing high expression patterns in control males and NV females, lower expression in the two groups of vitellogenic females (Vit2 and Vit4) and the lowest expression level in E2-treated males; and, 3) A cluster of 1,504 transcripts (Fig 4; subdivided into 3a, 3b and 3c) with high expression levels in E2-treated males, low expression levels in control males and NV females and interim expression levels in Vit2 and Vit4 females. Closer examination revealed that in some sub-clusters the expression pattern of Vit2 females was similar to that of NV females, while in others it was similar to Vit4 females. The expression patterns of Vit2 females were at an intermediate stage between those of the older vitellogenic females (Vit4 females) and NV females. This pattern corresponds with the expression levels of vtg3 in these groups mentioned previously (Fig 2).

Analyses of the molecular functions and biological processes were performed for all 2,523 transcripts corresponding to the differentially expressed transcripts. Only 607 genes were found as annotated among the 1,255 transcripts that were regulated in vitellogenic females (Vit2 and Vit4). In general, similar molecular functions and similar relative abundance (in percent) were found among the putatively E2 regulated genes and genes not regulated by E2 (Fig 8A, B). An exception includes genes associated with translational regulatory activity that were found only in the group putatively regulated by E2, indicating the prominent position of proteins synthesis during vitellogenesis. Analysis of the biological processes shows the occurrence of processes associated with reproduction only in the E2 regulated group and of immune system processes and biological adhesion only in the group of genes not regulated by E2 (Fig 8C, D). Other biological processes were similar among the putatively E2 regulated genes and those not regulated by E2. Some more detailed information on the number of transcripts and their descriptors that were associated with some selected functions, are shown in Table 2 and in Additional file 3, respectively.

Validation of the microarray results was performed by testing the relative expression of 16 genes (see Additional file 4) in the same RNA samples that were used for the chip hybridization, by real-time PCR. The 16 tested transcripts were: 1) genes highly expressed in vitellogenic females and known to be induced by E2 treatment [esr1, vtg1, vtg3, nots and cytochrome p450 2k6 (cyp2k6)], 2) genes that were highly expressed in males and known to be down-regulated by E2 (cyp1a1 and igf1), and 3) genes that showed significantly different expression levels in the tested groups [retinoic acid receptor alpha a (raraa), alcohol dehydrogenase 5 (adh5), alcohol dehydrogenase 8b (adh8b), igf1 and retinol dehydrogenase 10 (rdh10), retinol dehydrogenase 14 (rdh14), dehydrogenase/reductase (SDR family) member 10 (dhrs10), stearoyl-desaturase (sCd), fatty acid desaturase 2 (fads2) and steroidogenic acute regulatory protein (star)]. A very high correlation was found between the microarray and real-time PCR results, with regression coefficients (R² values) ranging from 0.9102 to 0.9340 (see Additional file 5).

The list of the 20 most differentially expressed hepatic transcripts includes genes known to be regulated by E2, genes that were not recorded previously as regulated by E2 or genes that were not previously reported to be expressed in the liver. Eight of the 20 most differentially expressed genes (Table 1), were previously reported to be regulated by E2, including: vtg1 and vtg3 25484950, nots 4251, syne1 52, fst1 53, nosip 54, grik1 55 and esr1 1549. Transcripts that were not associated previously with E2 regulation include reticulon1 (rtn1), entpd4 and lman1. A few transcripts (syne1, fstl1 and the fam20c) were not reported previously in hepatic cells. Proteins associated with cytoskeleton formation (syne1 and ank) showed higher up-regulation in vitellogenic females than after E2 treatment of males, supporting a putative role in the growth of the liver 56 and of the dramatic increase in protein synthesis and secretion by the endoplasmic reticulum 3 during vitellogenesis.

Zebrafish display eight different variant gene sequences 2527 for vtg genes but the array used here included probes only for vtg3 and vtg1. Since vtg1 shows high sequence similarity with vtg4, vtg5, vtg6 and vtg7 (all coding for VtgAo1), transcript levels for vtg1 may also reflect the expression of these genes. The higher expression levels of vtg3 compared with vtg1, may suggest their differential regulation by E2 stems from differences in the estrogen response elements (ERE's) in the promoter regions 42.

Numerous prominent and putatively regulated functions were revealed in this study to take place during vitellogenesis and include lipid metabolism and lipid binding, hormone and cholesterol binding, transcription factor activity, immune response, immune system processes and cell adhesion (Table 2 and see Additional file 3). The task of allocating processes and functional significance to genes that were putatively regulated by E2 and also to those that were putatively not regulated by E2, was faced with a general difficulty for zebrafish. Linking specific pathways or modes of function with genes by the general descriptors provided by Gene Ontology (GO) for zebrafish is problematic due to the incomplete annotation of the zebrafish genome and a deficiency in functional studies. Consequently, several GO terms rely on homology of putative functions described in higher vertebrate species.

Microarray profiling of liver samples revealed expression patterns characteristic of vitellogenic females of which only ~64% were found to be putatively regulated by E2. The repertoire of regulated genes implicates a wide range of functions especially those associated with protein synthesis, lipid metabolism, steroid biosynthesis, hormone binding and TF. Genes associated with the immune system and biological adhesion were among the genes that were up-regulated in vitellogenic females but not in E2-treated males, indicating that they were putatively not regulated by E2. E2-treated males expressed a large array of genes that were not associated with vitellogenesis. The study revealed several genes (rtn1, entpd4, lman1) that were not reported before as being regulated by E2. Also, the hepatic expression of several genes (e.g., syne1, fstl1, fam20c, star) is reported here for the first time for fish liver. In general, these results raise the question on the identity of the factors that regulate the pleiotropic expression of hepatic genes in vitellogenic females, in addition to E2.

Zebrafish were purchased from a local fish supplier (A&H Holdings, Israel LTD). All fish were maintained in 5-liter aquaria with UV treated, recycled and dechlorinated water and at ambient temperature of 25 ± 2°C with a light/dark cycle of 14/10 hr. The fish were fed twice a day, with shrimp nauplii (PGT, Eilat, Israel) in the morning and fish eggs in the afternoon. Non-vitellogenic females were kept under a light/dark cycle of 6/18 hr and fed twice a day with dry pellet food to avoid access to steroid compounds that maybe found in live food. For histological analysis the fish body or ovaries were fixed in Bouin's. Fixation, sectioning, and histological examination were performed according to 96. Paraffin sections of 4–7 µm were stained with hematoxylin and eosin. The terminology used by 97 for the zebrafish was adopted for this study. All fish were anaesthetized with Tricaine (Sigma-Aldrich, USA) before experimental procedures 98 and treatment of fish adhered with institutional regulations.

Two experiments were performed, one for evaluating the effect of E2 exposure in males and a second experiment for comparing gene expression in the liver of non-vitellogenic females with that of vitellogenic females. The experiments designs were as follows: 1) Four months old zebrafish were divided into three groups consisting of 8 fish in each group: i) adult spermeating males weighing 2.45 ± 0.207 g (N = 8) were exposed to E2 (Sigma-Aldrich, USA) by immersion for 48 hr (group E2-treated males). The concentration used was 5 μg/L (18 nM), as 3–4 ng/ml was determined to be the E2 natural concentration in the plasma of adult vitellogenic female ZF 99. A period of 48 hr of exposure was chosen as the highest expression level of esr1 was reached after 12 hr 100 and the expression of vtg stabilized after 48 hr of exposure to E2 and lasted for 17 days 50. The hepatosomatic index (HIS) of E2-treated males was 5.6 ± 0.6 after treatment. ii) control males weighing 2.26 ± 0.246 g (N = 8;) and a HIS of 3.2 ± 0.4; iii) vitellogenic females weighing 3.95 ± 0.303 g (N = 8) and showing a HIS of 5.9 ± 0.4 (group Vit4). Four replicate samples were prepared for each group and each replicate consisted of a pooled sample from livers of two fish. The samples were pooled after RNA extraction (see below). The fish in this experiment were kept with a light/dark cycle of 14/10 hr. The fish were fed twice a day, with shrimp nauplii (PGT, Eilat, Israel) in the morning and fish eggs in the afternoon. 2) In order to reveal the differences between vitellogenic and non-vitellogenic females, a second experiment was designed. One month old zebrafish were divided into two groups consisting of 32 fish in each group. Due to the small size of the liver, pooled samples from eight fish were prepared for each of the four replicates in the expression studies. The pooling of samples was done after total RNA extraction (see below). The groups consisted of: i) fish that were kept under a light/dark cycle of 14/10 hr for 5 weeks weighing 2.56 ± 0.426 g (N = 32) and a HIS of 5.6 ± 0.4, with ovaries in vitellogenic stage (group Vit2). Fish were kept at 25 ± 2°C and fed twice a day, with shrimp nauplii (PGT, Eilat, Israel) in the morning and with fish eggs in the afternoon. ii) fish that were kept under a light/dark cycle of 6/18 hr for the same time period and weighing 1.44 ± 0.391 g (N = 32) and showing a HIS of 4.6 ± 0.3, with non-vitellogenic ovaries (group NV). This group was fed only with dry pelleted food. In order to show the reversibility of this condition, some of the non-vitellogenic females were placed in tanks under the regular 10 hr light/14 hr dark photoperiod cycle. Fish were kept at 25 ± 2°C and were fed twice a day, with shrimp nauplii (PGT, Eilat, Israel) in the morning and with fish eggs in the afternoon. Ovaries collected after three weeks from these females were in the vitellogenic stage.

Determination of sex and of developmental stages of ovaries, were done by microscopic examination of the gonads. Oocyte stages were determined according to 97. The livers collected from all five groups were frozen instantly in liquid nitrogen and stored in -80°C until further use.

Measurements of E2 concentration in plasma was performed in similar designed separate experiments. Blood samples were collected using Micro-Hematocrit Tubes with Heparin (VWR, USA) from all tested groups and stored at -80°C. E2 concentrations were measured using Estradiol EIA Kit (Cayman, USA) according to the manufacture's protocol.

Total RNA was extracted from whole livers of zebrafish using Trizol reagent (GIBCO, USA) according to the manufacture's protocol followed by a clean-up and DNase treatement using RNeasy MiniElut Kit (Qiagene, Germany). After clean-up, 3 μl of the RNA samples were separated on 1.2% agarose gel to evaluate their quality and concentration. According to the gel picture, pooled samples from the livers of two fish were prepared for each group in Experiment 1 and of livers from eight fish for each group in the second experiment. Each RNA pool was quantified (A₂₆₀) and assessed for purity (A₂₆₀:A₂₈₀ ratio) using Gene Quant (Amersham, UK) and by visual inspection of the 3 μg RNA separated on a denaturing gel.

Zebrafish microarrays were prepared by the Kimmel Cancer Center at Thomas Jefferson University (TJU), Philadelphia, USA. The microarray is a single color system based on zebrafish oligonucleotide library from Compugen/Sigma Genosys and consists of 16,399 oligonucleotide probes (65 nt), representing 16,288 unique gene clusters. In order to minimize non-specific binding, CodeLink slides with a special coating were used. Also β-actin internal controls were used to monitor the labeling and hybridization quality.

Processed chips were scanned by using a Perkin Elmer ScanArray® XL 5000 Scanner, software version 3.1. Images were quantified by PerkinElmer (USA) Quant Array® Software 3.0. Quantization used the fixed circle method and outputs were total intensities. Microarray data were normalized across all arrays using quantile normalization of data in log base 2 scale 101. This method corrects background noise and non-specific hybridization.

Statistical analysis was performed using Significant Analysis of Microarray (SAM) software 102. For the multiclass analysis a false discovery rate (FDR) of 0.1% was used. For paired t-test comparisons between the different groups a FDR of 0.03% was used. Correspondence Analysis (COA) was performed using MultiExperiment Viewer (MeV) version 4.1. Clustering analysis was performed using Coupled Two Way Clustering (CTWC) algorithm 103105.

Real-time PCR was performed using the same RNA samples used for microarray hybridization (n = 16). For cDNA synthesis, 4 μg of total RNA were mixed with 0.1 µg of oligo-d(T) (Promega, USA), 4 µl of Bio-RT 5× buffer, 2 µM dNTP mix (Promega, USA), 200 U of Bio-RT (Bio-lab, Israel) and H₂O to reach a final volume of 20µl. After an incubation of 1 h at 37°C, 80 µl of H₂O were added to the reaction. The PCR mixture consisted of 0.5 ul of cDNA sample, 70 nM of each primer (see Additional file 4) and 12.5 μl of SYBR Green master mix (ABgene, UK), in a final volume of 25 μl.

Amplification was carried out in a GenAmp 5700 thermocycler (PE Applied Biosystems, USA) and according to the manufacture's protocol. Amplification was performed in triplicates and the results were analyzed with REST-384 version 2 106. The relative expression of the 16 tested genes (see Additional file 4) was calculated using zebrafish elongation factor 1 alpha (ef1a) as a reference gene. The gene ef1a was recently shown to be a suitable reference gene for tissue analysis, developmental and E2 exposure studies of zebrafish 41107.

The microarray annotations were updated using BlastX program against nr database of the GeneBank. Blast2GO software 108 was used for achieving GO annotations for the 2,523 differentially expressed genes found by SAM analysis (see Additional file 1 and Additional file 3). Combinations of gene lists were performed using Gene List Venn Diagrams software 109 (Fig. 5).