Introduction

Hox paralog group 2 (PG2) genes are critical in conferring regional identity on the embryonic hindbrain and pharyngeal arches in teleosts and tetrapods (Trainor and Krumlauf 2001). Phylogenetic reconstructions hypothesize that the Hox PG2 complement of ancestral osteichthyians was comprised of only two genes, Hoxa2 and Hoxb2 (Fig. 1). The evolutionary lineage comprising the sarcopterygians is presumed to have retained these two genes, as may have been the case for the actinopterygian stem lineage. A lineage-specific whole genome duplication event in the actinopterygian stem lineage before the origin of teleosts likely gave rise to four Hox PG2 genes (hoxa2a, hoxa2b, hoxb2a, and hoxb2b; Amores et al. 1998; Stellwag 1999). Subsequent differential loss of Hoxa2a and Hoxb2b in divergent teleost lineages resulted in variation observed among Hox PG2 genes of extant teleosts. The zebrafish (Danio rerio), a member of the superorder Ostariophysii, possesses only two Hox PG2 genes (hoxa2b and hoxb2a; Amores et al. 1998), whereas all the members of the superorder Acanthopterygii examined to date, including striped bass (Morone saxatilis), several pufferfishes, the Nile tilapia (Oreochromis niloticus), and the Japanese medaka (Oryzias latipes) each have three Hox PG2 genes (hoxa2a, hoxa2b, and hoxb2a; Fig. 1; Scemama et al. 2002; Amores et al. 2004; Jaillon et al. 2004; Santini and Bernardi 2005; Naruse et al. 2004).

Fig. 1
figure 1

Hox PG2 gene complement evolution in Osteichthyii. Phylogeny based on Steinke et al. (2006)

Full size image

Despite the evolutionary divergence in Hox PG2 gene complement among the osteichthyians, all orthologs have conserved anterior limits of hindbrain expression such that Hoxa2 and Hoxb2 genes are expressed with an anterior limit up to the rhombomere 1/2 (r1/r2) or r2/r3 boundaries, respectively. While the anterior expression boundaries of Hox PG2 genes in the hindbrain are highly conserved, there is documented divergence in expression patterns within the pharyngeal arches (Prince and Lumsden 1994; Vieille-Grosjean et al. 1997; Prince et al. 1998; Maconochie et al. 1999; Pasqualetti et al. 2000; Baltzinger et al. 2005; Scemama et al. 2002, 2006; Sham et al. 1993; Vesque et al. 1996; Yan et al. 1998). In tetrapods, Hoxa2 gene expression begins as neural crest cells migrate into the second pharyngeal arch (PA2) and continues during PA2 chondrogenesis (Vieille-Grosjean et al. 1997; Maconochie et al. 1999; Pasqualetti et al. 2000; Baltzinger et al. 2005). By contrast, tetrapod Hoxb2 expression in PA2 is either transient in the mouse and African clawed frog or absent in the chicken (Krumlauf, personal communication; Baltzinger et al. 2005; Vesque et al. 1996). Unlike tetrapods, the two Hox PG2 genes, hoxa2b and hoxb2a, of the zebrafish are both expressed in PA2 beginning when neural crest cells migrate into the arch and continuing into the chondrogenic phase of PA2 development (Hunter and Prince 2002).

Based on these expression analyses and the results of loss of function experiments, it has been hypothesized that the maintenance of Hox PG2 gene expression in PA2 during chondrogenesis is necessary for Hox gene-dependent specification of this arch (Hunter and Prince 2002; Baltzinger et al. 2005). We previously reported that in the perciform striped bass, all three Hox PG2 genes (hoxa2a, hoxa2b, and hoxb2a) are expressed in the hindbrain but that only the hoxa2 duplicates are expressed in the second and posterior pharyngeal arches (Scemama et al. 2002, 2006). Specifically, we demonstrated that hoxa2a expression alone was maintained in PA2 until the beginning of chondrogenesis suggesting that in striped bass, hoxa2a alone acted as the selector gene of the second arch identity. To understand whether this expression pattern was a common feature of teleosts with three Hox PG2 genes, we cloned the Hox PG2 gene complement from Nile tilapia (tilapia) and examined the expression patterns of these genes during embryogenesis.

Because striped bass and tilapia have a close phylogenetic affiliation, we hypothesized that the expression patterns of the three striped bass Hox PG2 genes would be conserved in tilapia (Nelson 2006). In this paper, we report the expression analysis of three tilapia Hox PG2 genes, hoxa2a, hoxa2b, and hoxb2a, as well as the cloning and early expression of the tilapia egr2 gene, which is used in this study as a marker to define the location of specific rhombomeres during early embryogenesis. In the course of these experiments, we determined that hoxb2a of tilapia is expressed in PA2, which prompted us to re-examine the expression of its ortholog in striped bass.

Materials and methods

Nile tilapia hoxa2a, hoxa2b, hoxb2a, and egr2 partial cDNA cloning

Tilapia hoxa2a, hoxa2b, and hoxb2a partial complementary deoxyribonucleic acid (cDNA) were generated by reverse transcriptase–polymerase chain reaction (RT-PCR) using total ribonucleic acid (RNA) isolated from stage 13 embryos (Fujimura and Okada 2007; Totally RNA®, Applied Biosystems, Foster City, CA). The primers for hoxa2a were initially designed to amplify a 537-bp partial cDNA fragment of striped bass hoxa2 transcripts in a previous study (Scemama et al. 2006): A2 exon 1: 5′-TCGCTTGCTGAGTGCCTGAC-3′ and A2 exon 2: 5′-GGTCTGTCGCTTGTGCTTC-3′. The primers for hoxa2b and hoxb2a were designed based on tilapia genomic sequences (Accession numbers AY757317 and AY757323) to amplify a 700- or a 650-bp fragment of the transcripts, respectively (tila2bF: 5′-ACTAACCCTGTCGGTGATGC-3′ and tila2bR: 5′-CCTCTGTAGCCTGCTCCAAC-3′; tilb2aF2: 5′-GCCCAGCAGCCATACACGAC-3′ and tilb2aR: 5′-TTGAATACGGAGAGGAGGCG-3′). The PCR products were cloned into the pCRII vector (Invitrogen, Carlsbad, CA), according to the manufacturer’s instructions. The inserts were sequenced using dideoxyterminator sequencing chemistry (Big Dye v. 3.0, Applied Biosystems), and the partial cDNA sequences are available in the GenBank database under the accession numbers EF547389, EF547390, and EF547391 for tilapia hoxa2a, hoxa2b, and hoxb2a, respectively. Egr2 primers were designed to amplify a 353-bp partial cDNA fragment of striped bass egr2 transcripts in a previous study (Scemama et al. 2006): Egr2F351: 5′-GGCTACCCTCTGCTTACAGTC-3′ and Egr2R: 5′-GGAGGTGGATTTTGGTGTGTC-3′.

Striped bass hoxb2a probe construction and labeling

A partial hoxb2a cDNA was amplified by RT-PCR from 48 hours postfertilization (hpf) striped bass embryonic RNA. The primers used for the PCR amplification were designed to amplify a 545-bp fragment encompassing the junction between hoxb2a exon1 and exon2: msab2aex1F: 5′-TCATCCCCCCTGCCTCATCC-3′ and msab2aex2R: 5′-TCAAACAGCCTCGCTCCTCC-3′. The amplification product was cloned into the pCRII® vector (Invitrogen), according to the manufacturer’s instructions. Insert orientation was determined using restriction digestion with the enzymes SmaI and SpeI. SmaI only cleaved the insert once and created two fragments of unequal length (195 and 350 bp), and SpeI only cleaved the vector once 36 bp 5′ of the insertion site. Two clones (pMsab2a-1 and pMsab2a-5) were selected such that each contained the insert in opposite orientations. Each clone was linearized with SpeI, and digoxigenin-labeled sense and antisense hoxb2a riboprobes were generated using T7 RNA polymerase with pMsab2a2-5 and pMsab2a-1, respectively (Maxiscript® transcription reaction kit, Applied Biosystems). Transcripts were purified using the Minielute® Purification Kit from Qiagen (Valencia, CA). The quality of the transcripts was assessed by denaturing agarose gel electrophoresis, and the transcripts were quantified by dot-blot analysis.

Whole-mount in situ hybridization

O. niloticus (Nile tilapia) brood fish were obtained from Southern Farm Tilapia (Castalia, NC). The species assignment of the brood fish was verified using morphological characteristics (Popma and Masser 1999) and restriction fragment length polymorphism analyses of PCR-amplified 18S ribosomal DNA (El-Serafy et al. 2007). The fish were raised at 28°C in 200 L aquaria arranged into a 2,600 L recirculating system. The exact time of fertilization was determined by observation of spawning, and the fertilized eggs were removed physically from the female’s mouth 30 min postfertilization. Tilapia embryos were allowed to develop at 28°C in Macdonald jars until sampled. Embryonic stages were determined according to Fujimura and Okada (2007). Embryos were euthanized by treatment with MS-222 (0.04%, w/v) and fixed for 7 days in 4% paraformaldehyde in phosphate-buffered saline (PBS) at 4°C, which are conditions that provided the best signal-to-noise ratio in the whole-mount in situ hybridization experiments. Embryos were then dechorionated and dehydrated through a series of solutions containing increasing concentrations of methanol to a final concentration of 100% and stored at 4°C until used. In situ hybridization experiments with tilapia and striped bass embryos were performed as previously described (Scemama et al. 2006) with antisense probes prepared from clones pOnia2a, pOnia2b, pMsab2a-1, pOnib2a, and pOniegr2 corresponding to hoxa2a, hoxa2b, hoxb2a, and egr2 antisense mRNA. Sense riboprobes were used in control experiments, to assess nonspecific hybridization. Identification of presumptive rhombomere 3 and 5 in tilapia embryos before the emergence of observable morphological landmarks was performed using the egr2 gene as a marker of r3 identity at the one-somite stage (McKay et al. 1994) and of r3 and r5 identity at the ten-somite stage (Fig. 3d). All hybridization experiments were performed in 48-well plates (Corning, New York, USA), and 133 ng of the probe was used per well, with each well containing up to five embryos in a final volume of 400 μL. Embryos up to stage 15 were deyolked and mounted between microscope slide and coverslip in 80% glycerol in PBS or in 1% low-melting-point agarose prepared in deionized water before examination and photography using a Leica DMR microscope equipped with a Nikon D1x digital camera. Older embryos were examined and photographed using either a Leica MZ75 dissecting microscope or an ML Digital Lab XLT (Microptics, Ashland, VA) equipped with a Nikon D1x digital camera. Images were processed in Adobe Photoshop, and the figures were prepared in Adobe Illustrator.

Results and discussion

Cloning and assignment of Nile tilapia Hox PG2 genes and egr2

Cloning of partial cDNAs obtained from RT-PCR amplification of 48 hpf tilapia embryonic RNA with Hox PG2 gene-specific primers generated three cDNA clones pOnia2a, pOnia2b, and pOnib2a. Basic Local Alignment Search Tool searches of the GenBank database at National Center for Biotechnology Information with sequences generated from the inserts contained in pOnia2a and pOnib2a were 99% similar to tilapia genomic sequences corresponding to hoxa2a (accession number AF533976) and hoxb2a (accession number AY757323), respectively. Based on these results, we concluded that pOnia2a and pOnib2a cDNA sequences corresponded to the tilapia hoxa2a and hoxb2a genes. The cDNA sequence from clone pOnia2b was 99% similar to the incomplete genomic sequence of tilapia hoxa2b (Genbank accession number AY757317) except for a region located immediately 3′ of the nucleotides encoding the conserved hexapeptide, as shown in Fig. 2a. This region of dissimilarity encompassed 60 nucleotides in pOnia2b and 30 nucleotides in the tilapia genomic sequence reported in AY757323. A comparison of the cDNA sequence from pOnia2b to the homologous PCR-amplified region of genomic DNA prepared from brood stock cultivated in our facility showed 100% sequence identity without any gaps other than the region corresponding to the genomically encoded intron (Fig. 2a). Comparison of the in silico-translated sequence from pOnia2b cDNA and its corresponding genomic sequence from our brood stock with orthologous sequences from zebrafish and striped bass (Fig. 2b) showed that the sequence directly downstream of the hexapeptide is more related to the amino acid sequences of zebrafish and striped bass hoxa2b orthologs than the amino acid sequence derived from in silico translation of the sequence reported in AY757323. The combined results of cDNA, genomic sequence, and in silico-translated sequence comparisons support the assignment of clone pOnia2b as a tilapia hoxa2b partial cDNA and suggest that the discrepancies observed between the pOnia2b-encoded cDNA and the genomic sequence published as Genbank accession number AY757323 may be either the result of tilapia interstrain divergence or a PCR, cloning, or sequencing artifact related to the generation of the AY757323 sequence. Overall, these sequence comparisons and analyses support the assignment of clones pOnia2a, pOnia2b, and pOnib2a as tilapia partial cDNAs for hoxa2a, hoxa2b, and hoxb2a genes, respectively.

Fig. 2
figure 2

Sequence analysis of Nile tilapia hoxa2b cDNA and genomic DNA. Panel a. Alignment of cDNA from clone pOnia2b and genomic DNA of the corresponding region with the incomplete genomic DNA sequence of Nile tilapia hoxa2b from GenBank accession number AY757317. Panel b. Amino acid alignment of Hoxa2b genes from zebrafish (Dre), striped bass (Msa), and Nile tilapia (Oni) from accession number EF547390 corresponding to the cDNA characterized in this study and accession number AY757317. These alignments were generated with ClustalX; identical residues are highlighted in yellow. Boxed regions indicate the sequences encoding (a) or corresponding (b) to the hexapeptide and the homeodomain. The blue boxes show the 60 nucleotide/20 amino acid region present in the cDNA and genomic sequence reported in this study. The red boxes show the 30 nucleotide/10 amino acid corresponding region in AY757317. The arrowheads indicate the position of the splice junctions. The shaded gray boxes show a 9 nucleotide/3 amino acid region missing in AY757317

Full size image

We also amplified and cloned a 351-bp tilapia egr2 partial cDNA (pOniegr2). The cDNA insert sequence of pOniegr2 was 99% similar to the striped bass egr2 gene (DQ383280) and 83% related to the zebrafish egr2 gene (NM130997), which supports its assignment as a tilapia egr2 partial cDNA.

Nile tilapia hoxa2a, hoxa2b, hoxb2a, and egr2 and striped bass hoxb2a gene expression

Given the close taxonomic relatedness of the tilapia with striped bass and the fact that both species have a full complement of three orthologous Hox PG2 genes, we hypothesized that the expression patterns of these three genes during embryonic development would be similar to one another. To test this hypothesis and to compare the expression patterns of Hox PG2 genes from divergent teleosts and tetrapods, we conducted whole-mount in situ hybridization experiments using Hox PG2 antisense RNA probes on tilapia embryos fixed beginning from their onset of expression and extending until the beginning of the larval stage. We analyzed the expression patterns of the three tilapia Hox PG2 genes relative to each other and to their orthologs in other teleosts (Fig. 6) and tetrapods.

Hindbrain expression

Tilapia hoxa2a, hoxa2b, and hoxb2a gene expression was first detectable as a faint, narrow band of staining corresponding to presumptive r3 of the hindbrain at embryonic stage 9 (one somite; 50% epiboly; 25 hpf, data not shown). The onset of expression for orthologous genes in the hindbrain compartment of striped bass and zebrafish are similar to those for tilapia both spatially and temporally (Fig. 6; Prince et al. 1998; Scemama et al. 2002, 2006). By embryonic stage 10 (ten somites; 75% epiboly; 30 hpf, Fig. 3a), tilapia hoxa2a hindbrain expression expanded to include the region encompassing r2 through r4, with expression being the most intense in r3, which is similar to the orthologous gene from striped bass at a comparable stage of development (Fig. 6; Scemama et al. 2006). By the end of the segmentation period (stage 12; 26 somites, 42 hpf, Fig. 3e) and during the pharyngula period, the expression of both tilapia and striped bass hoxa2a orthologs remained similar to one another and expanded into r7 (Fig. 6), which is indicative that these phylogenetically related species share common hoxa2a regulatory elements to control expression in the hindbrain.

Fig. 3
figure 3

Whole-mount in situ hybridization analysis of tilapia hoxa2a, hoxa2b, hoxb2a, and egr2 gene expression at 30, 42, 50, 66, and 72 hpf. Embryos are shown with the anterior end to the left; rhombomere numbers are indicated. Black arrowheads point to second arch neural crest cells, and white arrowheads point to posterior arch neural crest cells. In a, b, c, and d, 30-hpf embryos are shown with dorsal view toward the reader; egr2 was used as a marker of r3 and r5 in double in situ colocalization experiments; however, to avoid obscuring the weak expression signals observed at this early stage, the samples shown are from single in situ hybridization experiments only. The remaining panels show lateral views of hoxa2a, hoxa2b, and hoxb2a expression based on in situ hybridization using embryos ranging from 30 to 72 hpf (e, h, k, n hoxa2a; f, i, l, o hoxa2b; g, j, m, p hoxb2a, respectively). ov Otic vesicle. Scale bars equal 100 μm

Full size image

Tilapia hoxa2b was only faintly expressed in the hindbrain in r2 and r3 at stage 10 (ten somites, Fig. 3b), with expression expanding during the segmentation and pharyngula periods (66 hpf) to include r7 but not r6. This pattern of expression is somewhat divergent from the striped bass ortholog in which expression at the ten-somite stage extended more caudally to include r4 but during the later segmentation and pharyngula stages was restricted caudally to include only r5 (Fig. 6). A comparison of the hoxa2b hindbrain expression patterns among all the teleosts showed conservation of the r2–r5 expression domain and the absence of expression in r6 (Fig. 6), which indicates a conserved function for hoxa2b in the patterning of the anterior hindbrain.

In the mouse hindbrain, where Hoxa2 is also expressed in the r2 to r5 domain, knockout experiments have shown that Hoxa2 controls the segmentation of the anterior hindbrain and the axonal guidance of the Vth and VIIth cranial motor nerve axons out of r2 and r4 (Gavalas et al. 1997; Davenne et al. 1999; Barrow et al. 2000). The conservation of these exit points between mouse and zebrafish (Chandrasekhar 2004) along with the similar mouse Hoxa2 and teleost HoxA cluster gene(s) hindbrain expression in the r2 to r5 domain suggests that their role in Vth and VIIth branchiomotor axon guidance has been conserved between mammals and teleosts.

In addition to the patterning activity of Hoxa2 on branchiomotor nerves, recent experiments, performed using conditional Hoxa2 mouse mutants, have shown that Hoxa2 plays an important role in the establishment of the assembly of the somatosensory circuit of the trigeminal nerve by mediating arborization of the trigeminal axons with the rostral principal nucleus (PrV; Oury et al. 2006). Further, the authors demonstrated that the onset of arborization by the trigeminal axons into the PrV followed a rhombomere-specific pattern that correlated with the differential level and distribution of Hoxa2 transcripts, with the high-expression r3 domain corresponding to wiring by maxillary axons and low-expression r2 domain to mandibular axons. Like their mouse Hoxa2 ortholog (Gavalas et al. 1997), tilapia hoxa2a and hoxa2b are expressed at a higher level in r3 compared to r2, which is indicative that these genes are involved in patterning the trigeminal somatosensory map. Despite the fact that striped bass hoxa2a is expressed more strongly than hoxa2b in both r2 and r3, they are expressed at equal levels (Fig. 6), suggesting that tilapia and striped bass may have evolved divergent mechanisms to pattern the trigeminal circuit.

In zebrafish, Hox PG2 gene knockdown experiments did not show any effect on hindbrain segmentation or patterning of the branchiomotor neurons, a result that has been attributed to incomplete inactivation of Hox PG2 genes expression and the possible low threshold activity of these genes in the specification of Hox PG2 gene-dependent hindbrain structures (Hunter and Prince 2002; Ohnemus et al. 2001). However, in zebrafish, hoxa2b is expressed at a higher level in r2 than in r3 (Fig. 6; Prince et al. 1998; Hunter and Prince 2002), which is divergent from mouse and tilapia hoxA cluster gene(s) expression patterns, and suggests that the mechanism for the assembly of the somatosensory circuit of the trigeminal nerve is different in zebrafish. While it has been suggested that the afferent fibers from the trigeminal ganglion cells terminate in the primary sensory nucleus with a conserved somatotopic organization in tilapia, the cypriniform goldfish, and other vertebrates (Kerem et al. 2005; Puzdrowski 1988), the establishment of the trigeminal somatosensory map during embryonic development of the zebrafish, goldfish, and tilapia has not been examined. Further studies need to be conducted to clarify the role of differential r2/r3 Hoxa2 gene expression in the development of the somatotopic organization of the trigeminal pathway in divergent teleosts.

An examination of the expression patterns of the tilapia hoxb2a gene over the same developmental stages described for the HoxA cluster genes demonstrated that this gene was faintly expressed in r3–r4 at stage 10 (ten somites; Fig. 3c) with expression extending caudally to include r5 and increasing in intensity by stage 12 (26 somites) and extending even further caudally into r7 by the pharyngula period (stages 14–15/16; Fig. 3g,j,m). While the intensity of expression was relatively uniform among the rhombomeres before stage 14, the intensity of expression was elevated slightly in r3 and r4 during the pharyngula period. Similar expression patterns were observed for the striped bass and zebrafish hoxb2a genes (Figs. 5a–e and 6) except that tilapia hoxb2a expression in r7 was maintained throughout the pharyngula period. These observations suggest that the mechanisms specifying hoxb2a expression in the anterior hindbrain have been conserved in the three species, while those controlling expression in r7 and posteriorly have diverged in tilapia relative to striped bass and zebrafish. The conserved hoxb2a expression pattern in the anterior hindbrain of all three teleosts is similar to the pattern described for mouse Hoxb2 (Sham et al. 1993) and suggests that like their mouse ortholog, these genes specify the somatic motor component of the VIIth cranial nerve exiting r4 (Barrow and Capecchi 1996).

By the late pharyngula period (stage 15/16; 72 hpf), tilapia hoxa2b and hoxb2a expression receded to the dorsal aspect of the hindbrain (Fig. 3o,p), while hoxa2a continued to be expressed intensely in rhombomeres 2 to 7 (Fig. 3n). The hindbrain expression of all three Hox PG2 genes decreased in intensity toward the end of the pharyngula period (stage 16; 80 hpf, data not shown) and eventually became undetectable in this compartment by stage 17 (98 hpf), which is comparable to previous observations in striped bass and zebrafish (Fig. 6; Hunter and Prince 2002; Scemama et al. 2002, 2006).

The qualitative and quantitative differences in expression observed between tilapia and striped bass hoxa2b and hoxb2a orthologs are indicative of significant evolutionary divergence in the regulatory mechanisms controlling their expression and lead us to reject the hypothesis about similar Hox PG2 expression patterns between these two species. Of particular interest in this regard is the similarity in hindbrain expression patterns of the hoxa2b orthologs of striped bass and zebrafish relative to tilapia. This result argues that either the hindbrain regulatory elements controlling expression of the striped bass and zebrafish orthologs are more similar to each other than either is to those from tilapia or that striped bass and zebrafish hoxa2b transcriptional regulation has undergone convergent evolution leading to common expression patterns.

Neural crest and pharyngeal arch expression

The tilapia hoxa2a gene, like its striped bass ortholog, was the first of the Hox PG2 genes to be expressed in the migratory neural crest cells and pharyngeal arches, and its expression persisted in the pharyngeal arches longer than any of the other Hox PG2 genes. Tilapia hoxa2a expression was observed as early as developmental stage 10 (ten somites) in neural crest cells migrating out of r4 (30 hpf, Fig. 3a). After its initial expression in the neural crest cells, tilapia hoxa2a was detected in PA2 from the end of the segmentation period through the pharyngeal stage and well into the chondrogenic phase of PA2 development (stage 17–22, Fig. 4). In addition to its early expression in migratory neural crest cells and robust persistent expression in PA2, tilapia hoxa2a was also detected in the posterior pharyngeal arches beginning at the late segmentation period (stage 12; 42 hpf, 26 somites, Fig. 3e), extending through the pharyngula period (stages 14–16, 50–72 hpf, Fig. 3h,k,n) and into the larval period (stages 17–22, 98–192 hpf, Fig. 4a,d,g,j,m). The early onset and long-term maintenance of tilapia hoxa2a expression in migratory neural crest cells and the pharyngeal arches was consistent with the expression pattern of its orthologs in striped bass and tetrapods and its co-ortholog, hoxa2b, in zebrafish (Fig. 6). A unique aspect of tilapia hoxa2a expression not previously observed in the other teleosts is the extended persistence of expression well into the larval stage of development (stage 22) in the outermost perimeter of the opercle (a PA2 derivative) and in the gill filaments derived from PA3, PA4, and PA5 (Fig. 4a,d,g,j,m). Whole-mount alcian blue staining at these developmental stages revealed that the hoxa2a expression domain overlapped with nonskeletogenic cartilage structures previously described as cell-rich hyalin cartilage in the opercular valve and zellknorpel cartilage forming spicules in gill filaments (Benjamin 1990; Fig. 4c,f,i,l,o). It is notable that tilapia hoxa2a expression in the distal end of the gill filaments mirrors the pattern of expression observed for the zebrafish hoxa3 and hoxb3 genes. These zebrafish genes have been shown to be required for the budding of gill filaments (Hogan et al. 2004), suggesting that in tilapia, hoxa2a may act either independently or in concert with hoxa3 and hoxb3 genes in patterning these structures.

Fig. 4
figure 4

Tilapia larval whole-mount in situ hybridization analysis of hoxa2a expression and alcian blue cartilage staining. Larvae ranging from 98 to 192 hpf are shown in lateral (a–c) or ventral (d–o) views with anterior to the left. Black arrows point to the opercle (PA2-derived), and white arrows point to gill lamellae (posterior arch derived). g, h, and i are enlargements of d, e, and f, respectively. m, n, and o are enlargements of j, k, and l, respectively. Eyes were removed in ventral mounts to improve visualization of hybridization signal. In alcian blue-stained embryos, black arrows indicate opercular cartilage, while white arrows show characteristic gill filament cartilage spicules. Scale bars equal 300 μm

Full size image
Fig. 5
figure 5

Whole-mount in situ hybridization analysis of hoxb2a expression during striped bass embryonic development. Embryos are oriented with the anterior end to the left. a Embryos are positioned with the dorsal view toward the reader, while they are positioned laterally in b–f. Rhombomere numbers are indicated. Black arrowheads point to the second pharyngeal arch, while white arrowheads point to the posterior arches. ov Otic vesicle. Scale bars equal 100 μm

Full size image
Fig. 6
figure 6

Schematic representation of Hox paralog group 2 gene developmental expression patterns in striped bass, Nile tilapia, and zebrafish embryos. Embryos are represented in a lateral position with the anterior end to the left. Black, dark gray, and light gray color patterns represent high, medium, and low levels of gene expression based on the results from whole-mount in situ hybridization experiments, respectively. Rhombomeres and pharyngeal arches are labeled r1 to r7 and PA 1 to 7, respectively. ov Otic vesicle. Zebrafish expression patterns are taken from Prince et al. (1998) (segmentation period) and Hunter and Prince (2002; pharyngula and hatching periods). Striped bass expression patterns are taken from Scemama et al. (2006). * : striped bass hoxa2a expression becomes restricted to PA2 at 96 and 120 hpf

Full size image

The combined results of Hoxa2 gene expression and the phenotypes associated with Hoxa2 gene knockout/knockdown studies in tetrapods and zebrafish have lead several groups to hypothesize that persistence of Hox PG2 gene expression into the chrondrogenic phase of PA2 development is a determinate of selector gene function for this paralogous group (Hunter and Prince 2002; Baltzinger et al. 2005). Assuming that this hypothesis is correct and based on the persistent expression of tilapia hoxa2a in PA2, it is consistent to surmise that hoxa2a may serve as a selector gene of PA2 identity in this species and even in striped bass, wherein hoxa2a expression also perdures well into the chondrogenic phase of PA2 development. In this regard, it is notable that tilapia hoxa2b and hoxb2a gene expression was initiated later in development but terminated before the onset of the chondrogenic phase, which is indicative that neither of these two genes function directly as selector genes in the specification of PA2.

We observed that the tilapia hoxa2b gene was expressed at constant low levels in the second and posterior arches from the end of the segmentation period (Fig. 3f) until the beginning of the pharyngula period (Fig. 3i), with expression reduced to undetectable levels before the onset of chondrogenesis (Fig. 3l,o). The diminished persistence of tilapia hoxa2b expression in PA2, relative to the extended duration of expression for hoxa2a, was reminiscent of the temporal pattern of expression described for striped bass hoxa2b (Scemama et al. 2006) but significantly different from that of zebrafish hoxa2b, which perdures until the inception of chondrogenesis (Hunter and Prince 2002; Fig. 6). Despite the relatively low levels of tilapia hoxa2b expression overall, this gene was expressed at levels greater than hoxa2a in the posterior arches, and its expression in this region narrowed to defined stripes at 50 hpf (stage 14) before fading completely at 72 hpf (stage 16). We have observed a similar expression pattern for the striped bass hoxa2b gene in the posterior arches (Scemama et al. 2006). Given that tilapia and striped bass hoxa2a may be able to pattern PA2 independently of hoxa2b, it will be interesting to determine if hoxa2b has diverged from its zebrafish ortholog and whether it contributes to the specification of structures derived from the posterior arches instead of or in addition to playing a role in PA2 specification.

The expression of tilapia hoxb2a, which was readily detectable in both PA2 and the posterior arches from the end of the segmentation period (stage 12; 42 hpf; 26 somites) into the early pharyngula period (stage 14; 50 hpf; Fig. 3g,j), was unexpected in this developmental compartment because we previously failed to detect the expression of the tilapia ortholog in migratory or postmigratory neural crest cells resident in the pharyngeal arches of striped bass, a phylogenetically related member of the order Perciformes with the same complement of Hox PG2 genes as tilapia (Scemama et al. 2002). These results prompted us to reanalyze hoxb2a gene expression in striped bass embryos using a longer probe (hoxb2a-1, 545 bp) to enhance assay sensitivity. Using the new probe, we were able to detect the expression of striped bass hoxb2a in the cranial neural crest streams emanating from rhombomeres 4 and 6 and migrating into the second and posterior pharyngeal arches at the 12 somite stage (30 hpf, Fig. 5a). A comparison of striped bass and tilapia hoxb2a expression patterns revealed that the onset of expression in neural crest cells occurred earlier in striped bass than in tilapia, where its expression was only detected in postmigratory neural crest cells. Despite the earlier onset of expression of striped bass hoxb2a relative to its tilapia ortholog, both genes were expressed similarly in the second and posterior pharyngeal arches from the end of the segmentation period (42 hpf in tilapia, Fig. 3g; 39 hpf in striped bass, Fig. 5b) into the pharyngula period (50 hpf in tilapia, Fig. 3j; 48 hpf in striped bass, Fig. 5c). While tilapia hoxb2a expression became undetectable in the arches beyond the early pharyngula period (Fig. 3m,p), expression of the striped bass ortholog perdured until the end of the pharyngula period (Fig. 5d,e), with levels of expression that fluctuated in a manner not observed in either tilapia or zebrafish. Specifically, striped bass hoxb2a expression showed relatively high levels of expression at both the beginning and ends of the segmentation (20 somite stage, 39 hpf, Fig. 5b) and pharyngula periods (Fig. 5e), with a pronounced decrease in expression detected during the early pharyngula period (48 hpf, Fig. 5c). Importantly, although striped bass hoxb2a was expressed in the arches later than its tilapia ortholog, neither gene was expressed into the chondrogenic phase of PA2 development. These expression patterns contrast sharply with those of the zebrafish hoxb2a gene, the expression of which is restricted to PA2 and persists into the chondrogenic phase of PA2 development, wherein it exhibits functional redundancy with hoxa2b action as a selector gene (Hunter and Prince 2002). Assuming that Hox PG2 gene maintenance during PA2 chondrogenesis is a necessary determinant for selector gene activity, these results are indicative that neither of the hoxb2a genes from tilapia or striped bass serve as selector genes to specify PA2 and that instead hoxa2a alone acts to specify PA2 in these species. If this result is confirmed by functional studies in tilapia and striped bass using knockdown and misexpression experiments, it would provide evidence that the Hox PG2 genes from these two species function in a manner more similar to their tetrapod orthologs in which Hoxa2 serves as the sole selector gene for PA2 specification, while Hoxb2 is limited to the specification of the embryonic hindbrain. If tilapia and striped bass hoxa2a orthologs each act as sole selector genes to specify PA2 identity, as is the case for tetrapod Hoxa2, then it argues that the hoxa2 gene served as the ancestral selector gene of PA2 fate in osteichthyans and that the selector gene functional redundancy exhibited by the zebrafish hoxa2b and hoxb2a genes represents a derived state in the teleost lineage.