The function of retinoic acid (RA), the major metabolic product of Vitamin A, has been studied for the majority of the last century. Vitamin A–deficient (VAD) embryos have multiple developmental abnormalities, including neural, heart, pharyngeal, limb, kidney, and eye defects (Evans,1928; Hale,1933; Wilson et al.,1953; Thompson et al.,1964). Conversely, addition of RA is teratogenic (Satre and Kochhar,1989). In adults, proper RA signaling is also important as aberrant RA signaling is associated with cancers, such as lung cancer and acute promyelocytic leukemia (reviewed in Soprano et al.,2004). Thus, appropriate RA signaling is required for the normal development of vertebrate embryos and the maintenance of healthy adult tissues.
Members of the superfamily of nuclear hormone receptors, the RA receptors (RARs) and retinoid X receptors (RXRs), transduce RA signaling. The RARs can bind all-trans RA and a photo-isomer, 9-cis RA (reviewed in Bastien and Rochette-Egly,2004; Soprano et al.,2004; Mark et al.,2006). The RXRs can only bind 9-cis RA. However, ligand binding does not always seem essential for RXR activity (Jones et al.,1995; Aranda and Pascual,2001). In addition, RXRs can interact with and function in the transcriptional activation of other nuclear hormone receptors (Aranda and Pascual,2001). In the non-ligand bound form, heterodimers of the RARs and RXRs are bound to DNA and repress transcription. Upon ligand binding, the RAR/RXR heterodimers undergo conformational changes and begin to activate transcription of RA signaling target genes (reviewed in Aranda and Pascual,2001; Soprano et al.,2004).
In all mammals examined, there are three orthologs of the RARs (alpha, beta, and gamma) and RXRs (alpha, beta and gamma). Moreover, the RARs each have multiple isoforms, which likely have different functions (Bastien and Rochette-Egly,2004). In mice, the functions of all of the RARs and RXRs have been studied through gene targeting (Kastner et al.,1994,1997; Lohnes et al.,1994; Mendelsohn et al.,1994). Importantly, all of the aberrant phenotypes observed in VAD embryos are also observed in mice lacking RARs or RXRs (Kastner et al.,1994,1997; Lohnes et al.,1994; Mendelsohn et al.,1994; reviewed in Mark et al.,2006). However, there is functional redundancy between the receptors as most of these phenotypes are only observed when more than one receptor is inactivated (reviewed in Mark et al.,2006). Thus, loss-of-function analysis indicates that combinations of RARs and RXRs are responsible for transduction of RA signaling.
Recent studies in zebrafish have demonstrated potent roles for RA signaling during development of the heart, limb, pancreas, hindbrain, branchial arches, and pharyngeal pouches (Begemann et al.,2001,2004; Grandel et al.,2002; Stafford and Prince,2002; Emoto et al.,2005; Keegan et al.,2005; Maves and Kimmel,2005; Kopinke et al.,2006; Stafford et al.,2006). Little is known about the expression, function, or even number of the individual zebrafish RARs and RXRs responsible for transducing RA signaling in these various embryonic contexts (Joore et al.,1994; Jones et al.,1995). The expression patterns of two RAR alpha (RARa) genes, raraa and rarab, and one RAR gamma (rarg) gene have been compared in a recent publication (Hale et al.,2006). Other recent reports indicate that at least rarab functions in specification of the pancreas (Stafford and Prince,2002; Stafford et al.,2006). With respect to the zebrafish RXRs, the expression patterns of four RXR genes, a rxr alpha (rxra), rxr betas (rxrba and rxrbb), and a rxr gamma (rxrg) have also been compared recently (Jones et al.,1995; Tallafuss et al.,2006). In vitro studies found that these RXRs can interact with mouse RARa to activate a RA reporter construct (Jones et al.,1995). However, nothing is known about the function of RXRs during zebrafish development.
Phylogenetic analysis using partial sequences of the family of nuclear hormone receptors indicates that teleost genomes contain more RAR and RXR genes than are found in mammalian genomes (Bertrand et al.,2004). We, therefore, hypothesized that there could be more zebrafish RARs and RXRs beyond those already identified. Here we report cloning of an additional RAR gene, rargb, and two additional RXR genes, rxraa and rxrgb. We have compared the expression patterns of these newly identified genes with those of the previously reported receptors (Joore et al.,1994; Jones et al.,1995; Hale et al.,2006; Tallafuss et al.,2006). As with other duplicated genes in the zebrafish, overlapping and distinct expression patterns are found, suggesting partial functional redundancy as well as divergence in the function of these duplicates (Force et al.,2004,2005). Although partial sequence of a pufferfish RARb has been reported (Bertrand et al.,2004; Hale et al.,2006), we have not found evidence of a zebrafish RARb. Thus, we have now assembled what is possibly the complete set of zebrafish RAR and RXR genes based on current available data, which consists of duplicates of both RARs (alpha and gamma) and all RXRs (alpha, beta, and gamma). This study provides a valuable reference for the expression patterns of receptors involved in the transduction of RA signaling and the foundation for future studies aimed at parsing out the functions of individual receptors in specific tissues.
Aa Amblyomma americanum Ce Caenorhabditis elegans Cp Celuca pugilator Ls Lymnaea stagnalis Mv Mustela vison Ol Oryzias latipes Sd Suberitis domuncula Sm Schistosoma mansoni Tc Tripedalia cystophora Ts Thais clavigera Am Ambystomma mexicanum Bf Branchiostoma floridae Cc Coturnix coturnix Cf Canis familiaris Cj Coturnix japonica Cp Cavia porcellus Tr Takifugu rubripes Gg Gallus gallus Hs Homo sapiens Ma Mesocricetus auratus Mm Mus musculus Ms Mus spretus Nv Notophthalmus viridescens Rn Rattus norvegicus Xl Xenopus laevis Dr Danio rerio
Other groups who have examined the evolution of nuclear hormone receptors have reported the partial sequences of RARg duplicates in the Fugu genome (Bertrand et al.,2004; Hale et al.,2006). This suggested to us that there might be two zebrafish RARg orthologs. Through BLAST searches, we found the sequence of a hypothetical second rarg gene (XM_693726), predicted from annotation of zebrafish genomic sequence. We then were able to confirm its expression through cloning of an 818-bp fragment from cDNA taken from early somitogenesis stages. We found that the predicted gene sequence is not entirely accurate: compared to our cDNA sequence, bases 742–846 are incorrect and disrupt part of the highly conserved ligand binding domain. The predicted full-length sequence from the combined hypothetical sequence and cloned cDNA fragment encodes a 490 amino acid (aa) protein. Sequence alignments indicate that 75% of the amino acid identities are conserved between the previously identified RARg and the new RARg. The majority of divergence in sequence occurs within the N-terminal “A” and C-terminal “F” domains (Fig. 1A; functional domains are reviewed in Bastien and Rochette-Egly,2004). Phylogenetic analysis with all the available full-length sequences of the RARs confirms that this new gene represents a duplicate zebrafish rarg, which we will now refer to as rargb, with the previously reported rarg paralog becoming rarga (Fig. 1B). In further support of these being different genes and not splice isoforms, rarga is located on chromosome 23 (in zebrafish genome release Zv_6), and the newly identified rargb is located on chromosome 6. Thus, we have cloned a second ortholog of RARg in zebrafish, rargb.
Next, we examined rargb's temporal and spatial expression with RT-PCR and in situ hybridization (ISH). Because we also wished to understand how the expression patterns of all the zebrafish RAR homologs relate to each other, we present this in comparison to the previously reported RARs (Joore et al.,1994; Hale et al.,2006). Using RT-PCR, we found that rargb, as well as raraa and rarab, were expressed maternally and zygotically (Fig. 1C; Joore et al.,1994; Hale et al.,2006). Although it was reported that rarga is maternally expressed (Hale et al.,2006), we could not detect maternal rarga expression or even zygotic expression at the shield stage (data not shown), but could detect its expression during somitogenesis (Fig. 1C).
rar Expression Patterns Through the Tailbud Stage
Using ISH, we found that rargb was ubiquitously expressed from the shield stage until the 80% epiboly stage (Fig. 2D,H, L). By comparison, we could not detect rarga until 80% epiboly, when it is expressed at low levels around the margin (Fig. 2C,G, and arrows in K). Not observing rarga expression until after the shield stage is consistent with our RT-PCR results (Fig. 1C). In addition, the lack of rarga expression through mid-gastrulation is similar to the single murine rarg, which is also not expressed during early development (Ruberte et al.,1990). Xenopus rarg is expressed at very low levels maternally (Escriva et al.,2006). Later, its expression around the blastopore during gastrulation is similar to the expression of zebrafish rarga during gastrulation (Fig. 2K) (Ellinger-Ziegelbauer and Dreyer,1991; Escriva et al.,2006). The previous report that zebrafish rarga is expressed maternally (Hale et al.,2006) could be due to differences in probe length or ISH protocols, which could allow for cross-reactivity between rarg genes. Our experiments lead us to conclude that the ubiquitous expression of rargb through early gastrulation is in contrast to the absence of rarga during these stages.
With respect to the rara genes, at 40% epiboly, we found that raraa was not expressed at high levels, but is restricted to the dorsal side (arrow in Fig. 2A), which has not been previously reported. This dorsal expression is similar to the single Xenopus rara, which is expressed in the dorsal lip (Sharpe,1992; Escriva et al.,2006). The mouse Rara is primarily ubiquitous through development (Escriva et al.,2006). Zebrafish raraa expression remains primarily on the dorsal side through 80% epiboly (Fig. 2E,I). As has been reported previously, rarab is ubiquitous until 80% epiboly, when it becomes highly expressed around the margin and excluded from the anterior (Fig. 2B, F, J; Stafford and Prince,2002; Hale et al.,2006). Thus, through gastrulation, all four of the rar genes have slightly overlapping, yet distinct temporal and spatial expression. Moreover, each rar paralog pair has one paralog (rarab and rargb) that is ubiquitously expressed through gastrulation. The other paralogs (raraa and rarga) exhibit lower and restricted expression, which appears to be conserved with the Xenopus orthologs.
rar Expression From the Tailbud Stage Through 24 Hours Post-Fertilization
At the tailbud (TB) stage, rargb is still expressed ubiquitously, but it is expressed at higher levels in the anterior and the tailbud (arrowhead and arrow, respectively, in Fig. 2P). High expression in the TB persists through 24 hr post-fertilization (hpf) (Fig. 2T,X). rargb is also expressed in the three streams of migrating cranial neural crest cells beginning around the 10-somite stage (Fig. 3G–I). Although expression of the Xenopus and mouse Rarg genes is not ubiquitous, their expression in the neural crest and the TB is similar to that of zebrafish rargb (Ruberte et al.,1990; Escriva et al.,2006).
For comparison, zebrafish rarga is expressed at high levels in the TB from the TB stage through 24 hpf (Fig. 2O,S,W). It is also expressed in the anterior lateral mesendoderm during the same time period (arrow in Fig. 2O and arrows in Fig. 3E). The Xenopus and mouse Rarg genes are also expressed in similar regions (Sharpe,1992; Escriva et al.,2006). At the 5-somite stage, rarga is also expressed in r4 of the neural tube (arrowhead in Fig. 3E). Through 24 hpf, its neural expression expands posteriorly into the posterior hindbrain and anterior spinal cord (Fig. 3F). Thus, after the TB stage, the expression patterns of rarga and rargb differ. However, because both share similarity to the single Xenopus and mouse Rarg genes (Ruberte et al.,1990; Escriva et al.,2006), the expression of the single ancestral vertebrate rarg was likely closer to the pattern in those vertebrates.
By the TB stage, both rara paralogs are expressed in the posterior of all germ layers of the embryos (Fig. 2M,N). This shared posterior expression pattern is reminiscent of the single Xenopus rara during late gastrulation and neurulation stages (Escriva et al.,2006). After this brief period of shared expression of the zebrafish rara paralogs, their expression patterns start to differ, with the later expression of zebrafish rarab being more similar to Xenopus rara.
By the 5-somite stage, raraa is expressed in the TB, hindbrain, and spinal cord (data not shown and Fig. 3A). The hindbrain expression of raraa is dynamic. At the 5-somite stage, it is directly adjacent to krox20, which is expressed in rhombomere 3 (r3) and r5 (Fig. 3A). By 24 hpf, raraa is expressed at high levels in the anterior spinal cord, but also has low levels of expression in r4 (Fig. 3B and Hale et al.,2006).
During somitogenesis, rarab is expressed highly throughout the spinal cord and the TB (Fig. 2R and Fig. 3C), which is similar to Xenopus rara (Escriva et al.,2006). Starting at about the 5-somite stage and continuing through the 10-somite stage, zebrafish rarab is expressed particularly highly in the r5/r6 regions of the hindbrain (Figs. 2R, 3C), a pattern not shared with Xenopus rara. Later, rarab expression expands more anteriorly in the hindbrain and the elevated r5/r6 expression is lost (Fig. 3D). RA signaling is required for r5/r6 specification through regulation of val expression (Hernandez et al.,2004). Given the upregulation of rarab in the r5/r6 region and the known role of RA signaling in specifying r5/r6, this expression pattern suggests that rarab may play a specific role in mediating the RA signaling in r5/r6.
In summary, from the TB stage through 24 hpf, all four zebrafish rar genes develop distinct expression patterns. Comparing the expression patterns of the rara and rarg paralogs with the single Xenopus and mouse orthologs reveals trends that hint at ancestral expression patterns of the single ortholog. The divergent expression patterns of the duplicated pairs in zebrafish indicate areas where functions may have been divided between the paralogs or where new functions may have been gained. The expression patterns also suggest that the RARs may have specific roles in mediating RA signaling in particular contexts, such as a putative role of rarab in controlling val expression in r5/r6 of the hindbrain or a potential function of rargb in the neural crest (Hernandez et al.,2004). Therefore, comparing the complete set of zebrafish rar genes reveals common trends in paralog pairs in addition to defining conserved and distinct expression patterns that suggest specific roles in mediating RA signaling.
Cloning of Zebrafish RXRa and RXRg Paralogs
RA signaling requires heterodimers of RAR and the RXR receptors. In zebrafish, four RXRs (RXRa, RXRba and RXRbb, and RXRg) have been reported previously (Jones et al.,1995; Tallafuss et al.,2006). By performing BLAST searches with the previously published RXRs, we were able to identify two additional sequences that encode RXRs. One hypothetical transcript (XM_678129), predicted from annotation of genomic sequences, would encode a 430-aa protein that shares 79% overall sequence conservation with the previously published zebrafish RXRa (Fig. 4A). The other transcript (zgc:92183), originally identified as a cDNA, encodes a 452-aa protein that shares 81% conservation with the previously published zebrafish RXRg (Fig. 4A). Phylogenetic analysis demonstrates that the new RXRs are indeed orthologous to other vertebrate rxra and rxrg genes and paralogous to previously characterized zebrafish rxra and rxrg genes (Fig. 4B). Genomic locations further indicate that the new sequences represent different genes than those previously identified. The new rxra gene is located on chromosome 21, and the previously cloned rxra is located on chromosome 5 (Tallafuss et al.,2006). We found that the previous rxrg maps to chromosome 2 (Ensemble release Zv_6), as reported by Tallafuss et al. (2006), while the new rxrg maps to chromosome 20. Thus, based on sequence, phylogenetic analysis, and genomic location, we now refer to the newly identified RXR paralogs as RXRaa (because phylogenetic analysis indicates it is more closely related to other vertebrate orthologs) and RXRgb, respectively. Overall, this analysis indicates that all three of the zebrafish RXRs have been duplicated.
rxr Expression Patterns Through 24 hpf
Next, we compared the expression patterns of all of the zebrafish rxr genes, assessing temporal expression with RT-PCR and spatial expression with ISH. We found that five of the rxrs, rxraa, -ab, -ba, -bb, and -g, were expressed maternally (Fig. 4C). The newly identified rxrgb was not expressed maternally (Fig. 4C, lane 6). Later during somitogenesis, all six are expressed (Fig. 4C).
Similar to our observations of RAR genes, ISH of the six RXR genes revealed conserved and divergent expression patterns of each paralog pair. Although we could detect maternal expression by RT-PCR (Fig. 4C), we could not detect specific expression above background levels of rxraa, rxrab, or rxrga by ISH from cleavage stages through gastrulation (data not shown). However, rxrba and rxrbb are expressed ubiquitously through 24 hpf (Fig. 5C,D,I,J,O,P). The single mouse Rxrb is also ubiquitously expressed during development, suggesting this may be an ancestral characteristic of the rxrb orthologs (Dolle et al.,1994).
Detectable rxraa expression begins at about the 5-somite stage with expression in two bilateral spots in the posterior hindbrain and anterior spinal cord (arrows in Fig. 6A). Expression then increases in this area through the 10- and 15-somite stages (Fig. 5A,G), and is found in the third neural crest stream slightly later at the 18-somite stage (arrows in Fig. 6B). Expression in the anterior hindbrain, the tailbud, and the neural crest is maintained through 24 hpf (Fig. 5M). By comparison, rxrab is found only in the tailbud beginning at the 3-somite stage and continuing through 24 hpf (Fig. 5B,H, and N; Tallafuss et al.,2006). Later, rxrab has been reported to be expressed in the pharynx and fin (Tallafuss et al.,2006). Both the rxraa and rxrab expression patterns differ from that of mouse Rxra, which is ubiquitous during early development and later expressed primarily in the skin (Dolle et al.,1994). Thus, zebrafish rxra paralogs share expression in the tailbud, but differ in their expression in the anterior spinal cord and neural crest cells.
Similar to the previous characterization (Tallafuss et al.,2006), we found that rxrga is expressed in the anterior spinal cord by the 5-somite stage (Figs. 5E and 6C). Spinal cord expression is maintained through 24 hpf (Fig. 5K,Q). At 24 hpf, rxrga is also expressed in the third stream of cranial neural crest, but we did not observe it in the migrating crest (Fig. 6D). The newly identified rxrgb is also expressed in the anterior spinal cord by the 5 somite stage (Fig. 5F). However, its expression is temporally shorter and restricted to the ventral spinal cord (Figs. 5F,L,R and 6E,F). The ventral expression in the anterior spinal cord is similar to that observed for the single mouse Rxrg, although the anterior-posterior position of the mouse Rxrg expression has not been reported (Dolle et al.,1994). After the 5-somite stage, rxrgb expression decreases (Fig. 5F,L,R). By 24 hpf, rxrgb expression is low and restricted to the medial ventral spinal cord (Fig. 6F). Thus, the zebrafish rxrg paralogs share overlapping expression in the anterior spinal cord. However, rxrgb expression in the ventral anterior spinal cord is more similar to the expression pattern of mouse Rxrg, suggesting it may be indicative of the ancestral expression pattern.
To understand the mechanisms by which RA influences development of zebrafish ectoderm, mesoderm, and endoderm, it is critical to identify all of the molecules responsible for RA signal transduction in zebrafish. Here, we have presented the cloning and expression of the RA receptor genes rargb, rxraa, and rxrgb and compared them to the previously characterized zebrafish rar and rxr genes (Joore et al.,1994; Jones et al.,1995; Hale et al.,2006; Tallafuss et al.,2006). Together, these 4 RARs and 6 RXRs may represent the complete set of zebrafish RA receptors. Each paralog pair exhibits both overlapping and specific expression patterns. Expression patterns of some individual paralogs more closely resemble the single Xenopus and mouse orthologs, which hints at expression patterns of the ancestral receptors and suggests that the zebrafish paralogs may have divided the ancestral function and/or acquired new functions (Force et al.,2004,2005). The overlapping expression patterns of rar and rxr genes also predict the potential pairings of specific receptors into heterodimers for RA signal transduction in specific tissues. For instance, three of the rxr genes (rxraa, rxrga, and rxrgb) are expressed specifically in the anterior spinal cord during somitogenesis stages, while only raraa has a similar, overlapping expression pattern. In addition, the primarily ubiquitous expression patterns of the rxrb paralogs and specific expression patterns of the rxra and rxrg paralogs predict that the RXRb receptors mediate RA transduction in all of the tissues that lack rxra and rxrg expression. Perhaps the RXRb receptors are also the only RXRs mediating interaction with other nuclear hormone receptors during development (Aranda and Pascual,2001; Bastien and Rochette-Egly,2004). Altogether, this work supplies the inspiration for a number of hypotheses regarding RAR and RXR function in zebrafish, providing an ideal starting point for many future experiments.
In Situ Hybridization and RT-PCR
In situ hybridization was performed as described in Oxtoby and Jowett (1993). Although reported previously (Joore et al.,1994; Jones et al.,1995; Hale et al.,2006; Tallafuss et al.,2006), all zebrafish rar and rxr genes were cloned independently. Partial sequences used for probes were cloned from mixed stage cDNA. Probe lengths are: raraa 1,343 bp, rarab 1,360 bp, rarga 1,502 bp, rargb 818 bp, rxraa 1,085 bp, rxrab 1,049 bp, rxrba 900 bp, rxrbb 999 bp, rxrga 1,217 bp, rxrgb 1,142 bp. Embryos were examined using Zeiss M2Bio and Axioplan microscopes and photographed using a Zeiss Axiocam. Zeiss AxioVision 3.0.6 software and Adobe Photoshop Creative Suite were used to process images.
Primer sequences for cloning and RT-PCR of these receptors are available upon request.
We thank H. Auman for comments on the manuscript. J.S.W. is supported by NIH F32 HL083591-01.