An evolutionary and functional analysis of FoxL2 in rainbow trout gonad differentiation

Evolutionary analysis

The sequences of the FoxL2 orthologs from Tetraodon nigroviridis and Takifugu rubripes (pufferfish), Danio rerio (zebrafish) and mammals are those described by Cocquet et al.(2003). BLAST queries of Genbank with these open reading frames (ORFs) detected high-scoring matches at the forkhead domain, but similarity was surprisingly low outside this domain (putative paralogs). The rainbow trout (O. mykiss) FoxL2 sequences were obtained following BLAST searches against the ESTs available in GenBank. The partial sequences detected were then extended by PCR on cDNAs to obtain the full ORFs (Genbank AI: AY507927 for FoxL2a, AY507926 for FoxL2b).

For phylogenetic analysis of nucleotide and amino acid sequences, we used PhyML (Guindon & Gascuel 2003) and Tree-Puzzle (Schmidt et al. 2002) which implement fast heuristic tree-searching algorithms based on the maximum likelihood framework. Statistical support for alternative topologies was evaluated by the Kishino-Hasegawa test (Kishino & Hasegawa 1989) which has been implemented in Phylogenetic Analysis by Maximum Likelhood (PAML) (Yang 2002), Data Analysis in Molecular Biology and Evolution (DAMBE) (Xia 2001) and recently in Tree-Puzzle (Schmidt et al. 2002). Rate heterogeneity over sites was modeled by a discrete gamma distribution with the shape parameter α estimated from the sequences. The empirical Jones–Taylor–Thornton (JTT) substitution matrix (Jones et al. 1992) was used for analyzing amino acid sequences. For nucleotide sequences, the general time reversible (GTR) model was used with PhyML. As the GTR model has not yet been implemented in Tree-Puzzle, we used a more restrictive TN93 model (Tamura & Nei 1993) with one rate for transversions and two rates for A↔G and C↔T transitions.

Total RNA extraction and reverse transcription (RT)

Total RNA was isolated using TRIzol reagent (Invitrogen) as previously described (Govoroun et al. 2001b). Total RNA concentration was determined with an Agilent 2100 Bioanalyzer and the RNA 6000 LabChip kit (Agilent Technologies, Colorado Springs, CO, USA) according to the manufacturer’s instructions. For cDNA synthesis, 1 μg of RNA was denaturated in the presence of random hexamers (0.5 μg) for 5 min at 70 °C, and then chilled on ice for 5 min. RT was performed at 37 °C for 1 h using Moloney, Munine Leukemia Virus (M-MLV) reverse transcriptase (Promega) as described by the manufacturer.

Real-time PCR

Real-time PCR was carried out on an iCycler iQTM (BioRad, Hercules, CA, USA). Reactions were performed in 20 μl solution, with 300 nM of each primer, 5 μl of a 1/50 dilution of the RT reaction and the SYBER-Green PCR master Mix (Eurogentec, Seraing, Belgium) according to the manufacturer’s instructions. The first two incubation steps (50 °C for 10 min, 95 °C for 2 min), were followed by the thermal cycling protocol which included 10 min at 95 °C followed by 40 cycles of PCR (95 °C for 30 s, 60 °C for 1 min). For each primer set (see below) the efficiency of the PCR was measured in triplicate on serial dilutions of the same cDNA sample (pool of reverse transcribed RNA samples). Melting curve analysis was also performed for all genes to check the specificity and identity of the RT-PCR products. The relative amount of the target RNA was then determined using the I-Cycler IQ software (BioRad) by comparison with the corresponding standard curve for each sample run in duplicate. Each transcript level was then normalized on the basis of the expression values of the constitutive elongation factor 1α (EF1α). The primers used were: P450 aro-up, CTCTCCTCTCATACCT CAGGTT; P450 aro-dwn, AGAGGA ACTGCT GAGTATGAAT; FOXL2a-up, TGTGCTGGAT TTGTTTTTT GTT; FOXL2a-dwn, GTGTCGT GGACCATCA GGGCCA; FOXL2b-up, CGAG GAAGATTTA AACTACATG; FOXL2b-dwn, GAGGACGAG TCGGTTAGATCCA; EF1α-up, AGCGCAATC AGCCTGAGAGGTA and EF1α-dwn, GCTGG ACAAGCTGAAGGCTGAG

Histology

Gonads were fixed for 2 h in 0.13 M sodium cacodylate buffer (OP 320 mOsm; pH 7.4) containing 2.5% glutaraldehyde and 20 mM CaCl₂, then postfixed for 1 h with 1% osmium tetroxide and 1.5% potassium ferricyanide in the same buffer. After the postfixation, samples were washed in sodium cacodylate buffer. Subsequently, they were dehydrated in increasing concentrations of ethanol solutions (30%–100%), substituted in propylene oxide and embedded in epoxy resin. Semi-thin sections were cut, collected on glass slides and stained with toluidine blue.

FoxL2 orthologs and paralogs in fish

Pufferfish, T. nigroviridis, and zebrafish carry sequences highly similar to the mammalian FOXL2 ORFs and are considered to be conserved orthologs (Cocquet et al. 2003). Notably, they are without introns. Using BLAST searches, we have detected additional sequences with lower degrees of similarity and decided to analyze the highest scoring sequences, as they might be divergent paralogs. In fact, in the latter, conservation was essentially confined to the forkhead domain. We also isolated two FoxL2-related sequences from the rainbow trout (see Material and methods). One sequence was very close to well characterized FOXL2 orthologs and the other was more divergent. In order to show a potential evolutionary affiliation of these forkhead genes to the FoxL2 family, we gathered 13 well-known forkhead sequences from the zebrafish genome for phylogenetic analyses (as described in Kaestner et al. 2000). We also included the mammalian FOXL1 and FOXL2 sequences. The maximum likelihood trees from PhyML (Fig. 1a) and Tree-Puzzle (Fig. 1b) based on the amino acid sequences of the forkhead domains were largely congruent (see also online Fig. 1). The fully resolved PhyML tree (Fig. 1a) was significantly better (P=0.01) than the consensus tree from Tree-Puzzle (Fig. 1b) when evaluated by the Kishino–Hasegawa test, with lnL_PhyML =− 2025.20, and lnL_Tree-Puzzle =− 2051.26 (Δ lnL=26.06 and the variance of the difference, estimated by bootstrapping 10000 times, equals 11.13). The phylogenetic trees from nucleotide sequences were consistent with the tree from amino acid sequences, except that the tree from Tree-Puzzle had little resolution. These results suggest that Tree-Puzzle is limited in searching through the tree space and our interpretation below will be based on the PhyML trees (other details in online Fig. 2). The maximum likelihood trees from PhyML (Fig. 1a and online Fig. 2a) clustered all FOXL2 sequences together, including the highly diverged (presumed) paralogs. The murine Foxl1 sequence was weakly but consistently clustered with the FOXL2 sequences in the nucleotide-based tree (Fig. 1a and online Fig. 2a). The tree shows that T. nigroviridis, pufferfish, and rainbow trout have at least two FoxL2 paralogs, and justify our naming them as FoxL2a and FoxL2b genes. Recently, a forkhead gene from the ascidian Ciona intestinalis has been considered as the potential ortholog of FoxL2 based on a phylogenetic analysis (Mazet et al. 2003). It would be interesting to study the expression profile of the putative ancestor of Foxl2 in Ciona, which is a hermaphroditic organism.

Phylogenetic analysis of homologous sequences provides only limited support for paralogy, and extrinsic evidence of synteny has long been considered essential to support paralogy (Comings 1972). For this reason, we studied the conservation of synteny around FoxL2 in the completely sequenced pufferfish genome. We retrieved the genomic sequences containing the presumed FoxL2a and FoxL2b and searched for coding regions using the Genscan program (Burge & Karlin 1997). In one of the genomic contigs, this tool recognized an intronless ORF corresponding to FoxL2b. Moreover, both FoxL2a and FoxL2b were found to be linked to copies of the PI3K (phosphoinositide-3 kinase, catalytic sub-unit) and Hes (Hairy and enhancer of split-1) genes. Remarkably, this group of synteny has been conserved up to man (Fig. 2). This clearly shows that the highly diverged FoxL2b is indeed the paralog of FoxL2a. We next compared the FoxL2b ORF detected by Genscan in pufferfish with the intronless ORF containing the FoxL2-related forkhead of the close species Tetraodon. Only weak similarities were found in the 5′ region outside the forkhead domain (available for analysis). This shows that paralogs of FoxL2 in fish are evolving at very high rates, to a point that similarity is virtually undetectable by traditional means. This dissimilarity was particularly striking in the comparison with the rainbow trout FoxL2b sequence. This contrasts with the good conservation of many other paralogous pairs in fish (Robinson-Rechavi & Laudet 2001). It is conceivable that FoxL2 duplicates are degenerating. However, it is more likely that paralogs may have acquired novel functions. As FoxL2 is involved in fertility, the paralogs may contribute to reproductive isolation, which would explain their rapid evolution. A similar idea has been prompted to explain the divergence of the sex determining gene Sry in mammals (see Pamilo & O’Neill 1997 and references therein).

The analysis of the forkhead domains of the FoxL2 homologs, and also from other genes, provided two lines of evidence suggesting that selection favors a high isoelectric point (pI) in the forkhead domain. First, we detected a significant increase in pI in this domain with respect to those of the whole proteins (i.e. 10.48 vs 8.89 respectively) (Table 1). Secondly, the standard deviation of the pI is much smaller among the forkhead domains than among the whole sequences (i.e. 0.17 vs 1.26 respectively) (Table 1). The former finding can be explained by an increased usage of positively charged amino acids (Lys and Arg) in the DNA-binding domain with respect to the corresponding whole proteins (not shown). Indeed, clusters of positively charged amino acids can be important for DNA-binding as documented for the basic region in the leucine zipper (bZIP) gene family of transcription factors (Metallo et al. 1997). In the case of the forkhead domain, the positively charged amino acid residues may be important for both DNA binding and nuclear localization (Romanelli et al. 2003).

Expression of FoxL2a, FoxL2b and aromatase in the rainbow trout differentiating gonad

Quantitative real-time PCR analysis of the expression of the two rainbow trout FoxL2 paralogs in conditions of natural gonadal differentiation showed that their mRNA appears specifically during ovarian differentiation (Fig. 3). The steady state mRNA levels during testicular differentiation were barely detectable (from 10 to 100 times less than in female gonads). However, the kinetics of expression of these two FoxL2 paralogs differed substantially. Namely, FoxL2a was highly expressed since the very beginning of the free swimming period (d0) and remained expressed roughly at the same level from d0 to d111 (Fig. 3a). On the contrary, FoxL2b expression increased from d0 to d16, concomitantly with the first oocyte meiosis period (Fig. 4) and then dropped to levels similar to those detected at d0 (Fig. 3b). Female-specific expression of FoxL2 genes have also been found in other non-mammalian species. In the chicken, it is expressed in the two female gonads (ZW) at early stages of development, at least from day 5 to 8, whereas no expression is detected in male gonads at comparable stages. In the turtle, a species undergoing temperature-dependent sex determination, FoxL2 is expressed in the developing gonads at a higher level in female-promoting temperatures than in male-promoting temperatures (Loffler et al. 2003).

Sex determination in the rainbow trout is primarily governed by sex chromosomes. However, successful development of the gonads may be influenced by exposure to exogenous steroids during a critical period of gonadal differentiation. This may lead to reversal of the expected phenotypic sex, with the resulting population being fertile (Guiguen et al. 1999, Govoroun et al. 2001b). In our context, following EE2 treatment of males leading to an effective feminization (Fig. 4), the steady-state mRNA levels of FoxL2a and FoxL2b increased in the gonad, with important differences in their expression profiles. Specifically, FoxL2a expression was up-regulated very quickly (d7) to levels similar to those detected during natural ovarian differentiation (Fig. 3a). This up-regulation persisted after the end of the treatment (d63). We suggest that exogenous estrogens are able to rapidly trigger FoxL2a expression in a male background (mRNA is detected as early as d7). This also supports the notion that FoxL2 is involved in early ovarian development, in an evolutionary scale ranging from fish to man. Notice that in the case of the feminizing treatment, the gonad is forced to differentiate into an ovary and FoxL2a expression seems to be required for this process. FoxL2b was also clearly up-regulated when compared with the basal male levels (Fig. 3b). However, when compared with natural female differentiation, it only reached similar levels at d30 and higher levels at d63. After completion of the EE2 treatment, expression levels were similar to those of females. Our results suggest that FoxL2b is still functional in the rainbow trout as it has kept an intact ORF encoding a basic protein expressed in the female gonad (as its paralog) and is responsive to hormonal treatments. However, FoxL2b may have acquired a novel function, which explains its divergence in sequence and expression pattern. FoxL2b expression peaks at about d16, concomitantly with the onset of oocyte meiosis and may well be involved in this process. A possible neofunctionalization would be consistent with the maintenance of many duplicate genes in fish (Govoroun et al. 2001a).

The expression profile of aromatase, a key gene of estrogen synthesis, was also followed by real-time RT-PCR (Fig. 3c). In accordance with previous results, aromatase expression during natural differentiation was only detected in the female (Guiguen et al. 1999). The comparison of the aromatase expression profiles during natural differentiation with those of FoxL2a demonstrated a high correlation (R=0.98, P< 10⁻⁸) which suggests that both genes are close in a regulatory cascade, or that they respond to the same regulators. The feminizing EE2 treatment in males did not up-regulate aromatase expression, which suggests that the exogenous estrogens are enough to drive the feminization process, including early expression of FoxL2a (i.e. in absence of aromatase transcription). Accordingly, aromatase and FoxL2a mRNA expression profiles did not correlate during the feminizing treatment. After completion of the treatment (d63), expression increased slightly. On the contrary, the specific ovarian expression of aromatase during differentiation was very quickly down-regulated by masculinizing treatments in females, either with androgens (11βOHΔ4) or with an aromatase inhibitor (ATD), a situation that persisted after completion of the treatments (d91) as a functional testis had developed (Fig. 4). Following effective masculinization of females, we detected a decrease in the expression levels of both FoxL2 paralogs. These two masculinizing treatments had largely similar effects. The decrease of FoxL2a was detected as early as d7 and then reached a level similar to that of a normal male. Moreover, high and significant correlation coefficients between aromatase and FoxL2a expression were found for masculinizing treatments (for 11βOHΔ4: R=0.81, P<0.05; for ATD: R=0.88, P<0.02) when the gonad stops producing estrogens (as expected from the inhibition of aromatase transcription). This covariation may be explained by a concerted repression of both FoxL2a and aromatase, but is also compatible with the notion that estrogens stimulate FoxL2a transcription. Thus, a decrease in estrogen production due to aromatase repression or inhibition would induce FoxL2a down-regulation. After completion of the masculinizing treatments (d91), the expression levels of FoxL2a remained as low as in normal males. This is the result of the long-term transformation of the gonad into a functional testis. For FoxL2b there is no relevant repression detectable before d30. After completion of the treatments, expression levels remained low and similar to those in males. The masculinizing effects of androgens are probably triggered by the suppression of estrogen synthesis through inhibition of aromatase gene expression. The inhibition of aromatase following androgen treatment in females has already been reported in rainbow trout (Govoroun et al. 2001b). Here we have shown that an aromatase inhibitor can also induce a similar expression profile. These data taken together suggest that estrogens are key regulators of ovarian differentiation in fish and androgens would interfere with ovarian differentiation by inhibiting their synthesis.

Sex determination in birds and reptiles is thought to depend on the initiation of estrogen synthesis in the undifferentiated gonad, which would inhibit male differentiation and stimulate ovarian development. In the absence of this estrogenic signal, a testis would develop. However, further analyses are required to assess if the mere activation of aromatase is the initial trigger of gonadal differentiation (Gabriel et al. 2001). Here we have documented a similar process in fish. From a wider perspective, it has been demonstrated that female mutant mice homozygous for the targeted disruption of two estrogen receptors exhibit some degree of morphological gonadal sex reversal (Couse et al. 1999). This strongly suggests that, even in mammals, the estrogen response can also lead to a perturbation of gonad differentiation. Although early FOXL2 expression in mammals is thought to be steroid independent, it would be interesting to explore this connection at later stages of gonad development and in adults.