Tbx2b is required for the development of the parapineal organ

The formation of an organ requires specification of multiple cell types within a primordium, followed by appropriate morphogenetic movements that shape the mature organ. In the case of asymmetrically shaped and positioned organs, morphogenesis must also be regulated along the left-right (L-R) axis. The zebrafish presents a unique opportunity to study the formation of an asymmetric region in the brain, namely the pineal complex, which is found in the dorsal diencephalon (epithalamus). The pineal complex includes the melatonin-secreting pineal organ and the left-sided parapineal organ(Butler and Hodos, 1996; Borg et al., 1983). In adult fish, the pineal organ is attached to the brain via a stalk, which emerges just to the left of the midline (Liang et al., 2000). The parapineal organ is situated within the brain to the left of the pineal stalk (Gamse et al., 2002). Adjacent to the pineal complex are the bilateral habenular nuclei (habenulae). Along with the medial forebrain bundle, the habenulae and their efferent connections are a principal pathway connecting the limbic forebrain with the midbrain. Habenular axons travel via the fasciculi retroflexus (FR) to the interpeduncular nucleus (IPN) of the midbrain; this habenulo-interpeduncular connection is highly conserved throughout the vertebrate lineage(Sutherland, 1982). In teleosts, including zebrafish, axons from the left versus right habenula innervate different dorsoventral regions of the IPN, in contrast to other vertebrates, in which the two habenulae exhibit similar connections to the IPN(Kuan et al., 2007a).

In zebrafish, the asymmetric location of the parapineal organ imposes laterality on the flanking habenulae, which exhibit L-R differences in their neuropil density (Concha et al.,2000) and in the size and expression of genes such as leftover (lov; also known as kctd12.1 - ZFIN)(Gamse et al., 2003). In mutants in which the parapineal is on the right side of the brain, structural and gene expression differences in the habenulae are also L-R reversed(Gamse et al., 2003; Concha et al., 2000). Selective destruction of the parapineal by laser ablation reveals its instructive role on the left habenula. In the absence of the parapineal, both habenulae develop with equivalent size, patterns of gene expression, and neuropil density characteristic of the right habenula(Gamse et al., 2003; Concha et al., 2003). This abnormal development of the habenulae also affects efferent projections to the IPN. In wild-type (WT) embryos, the left habenular neurons project to both the dorsal and ventral regions of the IPN(Gamse et al., 2005; Aizawa et al., 2005). However,in parapineal-ablated larvae, left habenular neurons no longer project dorsally, but instead innervate the ventral IPN, characteristic of right habenular neurons (Gamse et al.,2005). Therefore, the development of the parapineal is important for asymmetry of the dorsal diencephalon as well as for the formation of diencephalic connectivity to other brain regions.

The parapineal organ is morphologically distinguishable between 28 and 32 hours post-fertilization (hpf) as a cluster of cells adjacent to the anterior left edge of the pineal anlage (Gamse et al., 2003; Concha et al.,2003). Studies of fixed samples suggest that over the course of the next 3 days, the parapineal becomes located more posteriorly and ventrally relative to the pineal organ (Gamse et al., 2002). A single parapineal organ is detected in 95% of WT zebrafish larvae on the left side of the brain, and only very rarely(<0.3%) are bilateral or no parapineal organs observed(Gamse et al., 2003; Concha et al., 2000). Precursors of the parapineal and pineal organ are thought to derive from a common pool of cells, as shown by lineage labeling of the pineal complex anlage at 22-24 hpf (Concha et al.,2003). The genes that regulate how parapineal cells are specified from this precursor pool, and that are responsible for their ultimate left-sided localization, are unknown.

To identify genes involved in asymmetric development of the epithalamus, we performed a genetic screen for mutations that disrupt L-R differences of lov expression in the habenulae. Here, we report the isolation and characterization of the from beyond (fby) mutation that produces homozygous embryos with reduced lov expression in the left habenula. We identify fby as a lesion in the T-box2b(tbx2b) gene that disrupts parapineal development. Time-lapse analysis reveals details of the morphogenetic process, showing that parapineal cells migrate leftward away from the pineal complex anlage as a cluster. By contrast, the parapineal cells of fby mutants are fewer in number and are scattered near the midline of the brain. We propose that Tbx2b assigns cells to the parapineal lineage and regulates their asymmetric migration.

Zebrafish were raised at 28.5°C on a 14/10 hour light/dark cycle and staged according to hpf or days post-fertilization (dpf). The wild-type strains AB (Walker, 1999) and WIK (Rauch et al., 1997), the transgenic lines Tg(foxd3:GFP)^fkg17 and Tg(foxd3:GFP)^fkg3(Gilmour et al., 2002), the ethylnitrosourea-induced mutations knypek^m119 and trilobite^m747(Solnica-Krezel et al., 1996),and fby^c144 were used.

ENU mutagenesis was performed by placing male zebrafish in 3 mM ENU for three 1-hour treatments with 48 hours between treatments. F1 fish were generated by crossing ENU-mutagenized males to AB females. F2 fish were generated by intercrossing F1 males and females. F3 larvae were generated from F2 intercrosses and at 4 dpf were fixed for in situ hybridization with the lov probe. Putative mutations showing reversed or bilaterally symmetric expression of lov in the habenular nuclei were re-identified in F2 heterozygotes and lines established through outcrosses to AB fish.

Heterozygous fish (AB background) carrying the fby^c144mutation were crossed with wild-type WIK fish. F2 mutant larvae were used for meiotic mapping with simple sequence length polymorphism (SSLP) markers, as previously described (Bahary et al.,2004). For identification of the lesion in the tbx2bgene, cDNA was prepared using Superscript II reverse transcriptase(Invitrogen) from mRNA isolated from 4-day-old AB and fby mutant larvae using Trizol (Invitrogen). For sequencing, cDNA was amplified using primers that flanked the tbx2b open reading frame and Pfu DNA polymerase, and subcloned into the pCRII-Topo vector (Invitrogen). Clones were sequenced using an ABI 3730xl sequencer and sequence data were analyzed with Sequencher software (GeneTools).

To genotype larvae, DNA was extracted by boiling larvae for 10 minutes in 25 μl of Embryo Lysis Buffer (Bahary et al., 2004). Proteinase K (Roche Applied Bioscience) was added to a final concentration of 0.5 mg/ml and samples were incubated at 55°C for 2 hours, then boiled for 10 minutes to deactivate the proteinase K. The fby mutation introduces an MseI restriction site at bp 134 of exon 3. Primers flanking this site were used to amplify genomic DNA(forward primer, 5′-TGTGACGAGCACTAATGTCTTCCTC-3′; reverse primer,5′-GCAAAAAGCATCGCAGAACG-3′). The PCR product was digested with MseI (NEB) at 37°C for 1 hour, then run on a 3% agarose gel.

Antisense morpholino (MO) sequences for tbx2b were derived from the translation initiation site and the splice donor site of the third exon as described [Tbx2b-ATG and Tbx2b-SP (Gross and Dowling, 2005)]. The MO stock solution (10 mg/ml) was diluted to 6 ng/nl in distilled H₂O, and embryos (1- to 2-cell stage) were pressure-injected with ∼1 nl to deliver 6 ng of MO per embryo. Embryos from the transgenic line Tg(foxd3:GFP)^fkg17 were used so that the location of the parapineal could be readily assessed by fluorescence in live larvae at 3-4 dpf.

Whole-mount RNA in situ hybridization was performed as described previously(Gamse et al., 2003), using reagents from Roche Applied Bioscience. RNA probes were labeled using fluorescein-UTP or digoxygenin-UTP. To synthesize antisense RNA probes,pBS-otx5 (Gamse et al.,2002), znot (flh)(Talbot et al., 1995), pBS II SK-znr1 (cyc) (Sampath et al., 1998), and pBK-CMV-leftover(Gamse et al., 2003) were linearized with EcoRI and transcribed with T7 RNA polymerase;pCRII-tbx2b, zPitx2 (Tsukui et al., 1999) and pBK-CMV-right on(Gamse et al., 2005) with BamHI and T7 RNA polymerase; pCRII-dexter(Gamse et al., 2005) with XhoI and SP6 RNA polymerase; pBS II SK-ATV (lft1)(Thisse and Thisse, 1999) with NotI and T7 RNA polymerase; pBSK+ southpaw(Long et al., 2003) with SpeI and T7 polymerase; pGEM-fzd7a(El-Messaoudi and Renucci,2001) with ApaI and SP6 polymerase; pBS II SK-wnt11 (Makita et al.,1998) with HindIII and T3 polymerase; and pBS-gfi1 (Dufourcq et al.,2004) with SacII and T3 RNA polymerase. Embryos were incubated at 70°C with probe and hybridization solution containing 50%formamide (60% for cyc). Hybridized probes were detected using alkaline phosphatase-conjugated antibodies and visualized by 4-nitro blue tetrazolium (NBT) and 5-bromo-4-chloro-3-indolyl-phosphate (BCIP) staining for single labeling, or NBT/BCIP followed by iodonitrotetrazolium (INT) and BCIP staining for double labeling. For sagittal sections, larvae were embedded in OCT (Sakura) and sectioned on a Leica cryostat. All in situ data were collected on a Leica DM6000B microscope with a 5× or 20×objective.

For whole-mount immunohistochemistry with rabbit or mouse antibodies, 4-dpf larvae were fixed overnight in 4% paraformaldehyde and stored at -20°C in methanol. Samples were permeabilized by treatment with 10 μg/ml proteinase K, refixed in 4% paraformaldehyde, and blocked in PBS with 0.1% Triton X-100,10% sheep serum, 1% DMSO, 1% BSA (PBSTrS). For antibody labeling, rabbit anti-Lov or anti-Ron (1:500) (Gamse et al., 2005), rabbit anti-GFP (1:1000, Torrey Pines Biolabs), mouse anti-opsin (1:500) (Adamus et al.,1991) or mouse anti-Zpr1 (1:250, Zebrafish International Resource Center, Eugene, OR) were used. Larvae were incubated overnight in primary antibody diluted in PBSTrS. Primary antibody was detected using goat anti-rabbit or goat anti-mouse secondary antibodies conjugated to the Alexa 568 or Alexa 488 fluorophores (1:350, Molecular Probes).

To simultaneously detect Lov and Ron, chicken anti-Lov polyclonal antibody was generated. A bacterially expressed C-terminal fragment of Lov was generated and purified as described previously(Gamse et al., 2005), injected into hens, and total IgY was isolated from egg yolks (Aves Labs). For Lov/Ron double labeling, larvae were fixed in 100% Prefer fixative (Anatech) overnight and stored in PBS containing 0.1% Tween 20 for up to 2 weeks. Samples were permeabilized, refixed and blocked as above, then incubated overnight with chicken anti-Lov (1:500 dilution of total IgY) and rabbit anti-Ron (1:500)antibodies simultaneously in PBSTrS, and primary antibodies detected with goat anti-chicken:Alexa 568 and goat anti-rabbit:Alexa 488 (1:350, Molecular Probes). All immunofluorescence data were collected on a Zeiss/Perkin Elmer spinning disk confocal microscope with a 40× oil-immersion objective and analyzed with Volocity software (Improvision).

For time-lapse imaging, wild-type or fby mutant embryos carrying the transgene (foxd3:GFP)^fkg3 were either uninjected or injected with a 6 ng/nl solution of tbx2b splice MO. Embryos were mounted in a 0.8% solution of low-melting-point agarose containing 0.003%1-phenyl-2-thiourea (PTU) and anesthetized with 0.04% Tricaine. Images were collected on a Zeiss/Perkin Elmer spinning disk confocal microscope with a 40× oil-immersion objective every 15 minutes from 24-48 hpf and analyzed using Volocity software.

Ablation of a subset of parapineal cells was performed using Tg(flh:GFP)^c161 or Tg(foxd3:GFP)^fkg331-hpf embryos, mounted dorsal-side-up in 0.8% agarose in 35 mm Petri dishes. Parapineal cells were visualized by GFP fluorescence and ∼5 cells were ablated by 5-10 pulses/cell from a 440 nm laser beam (Photonic Instruments)focused through a 40× water-immersion objective mounted on a Leica DM6000B microscope. An equivalent number of cells contralateral to the parapineal were ablated in the controls.

We screened 4-dpf larvae to identify ENU-induced mutations that altered expression of the K⁺ channel tetramerization domain-related (KCTD)gene leftover (lov). In the habenulae, lov is typically expressed at high levels and in a broader domain on the left(Gamse et al., 2003)(Fig. 1A). The from beyond (fby) mutation reproducibly caused reduced expression of lov in the left habenula (Fig. 1B). In WT larvae, the left habenula is larger and contains more neuropil than the right (Gamse et al.,2003; Concha et al.,2000), whereas in fby mutants the size and neuropil content of the left habenula was similar to the right (data not shown). Intercrosses between carriers from the original F2 family revealed that fby segregates as a simple mendelian recessive mutation (104/438,24%).

fby mutant larvae developed with normal overall body morphology,formed swim bladders and fed on schedule, but most larvae died at ∼2 weeks of age. Rarely, mutants survived to adulthood (4/101, 4%). A morphological phenotype of reduced otic vesicle size was also detected, with 30% of mutant larvae showing fused otoliths in one or both otic vesicles at 4 dpf, and 100%by 7 dpf (data not shown).

Reduced lov expression in the left habenula was associated with other alterations in the usual L-R asymmetric pattern of the habenulae. In WT larvae, a smaller area of the left than of the right habenula expresses the KCTD genes right on (ron; also known as kctd12.2 -ZFIN) and dexter (dex; kctd8 - ZFIN)(Gamse et al., 2005)(Fig. 1C,E). By contrast, the left habenula of fby mutant larvae had an expanded region of ron- and dex-expressing cells similar to the right habenula(Fig. 1D,F).

In addition to asymmetric gene expression, the left and right habenulae differ in their axonal projections to the IPN(Kuan et al., 2007b; Gamse et al., 2005; Aizawa et al., 2005). In WT larvae, Lov⁺ axons project to the dorsal and ventral regions of the IPN, whereas Ron⁺ axons project to the ventral IPN only(Gamse et al., 2005)(Fig. 1G,I-K). In fbymutant larvae, the dorsal/ventral (D/V) pattern of L-R habenula projections was altered. Fewer Lov⁺ axons projected dorsally and more projected ventrally (Fig. 1H,L), and an increased number of Ron⁺ axons were detected ventrally(Fig. 1M,N). These data are consistent with left habenular efferents adopting a projection pattern characteristic of the right habenula.

The abnormal laterality in the brains of fby mutants prompted an examination of asymmetric Nodal signaling, a key early determinant of L-R axis formation (Shen, 2007). In WT embryos, two Nodal genes, southpaw (spaw) and cyclops (cyc; also known as ndr2 - ZFIN), are expressed in the zebrafish left lateral plate mesoderm (LPM), as are the Nodal-induced genes lefty1 (lft1) and pitx2(Sampath et al., 1998; Rebagliati et al., 1998b; Rebagliati et al., 1998a; Long et al., 2003; Liang et al., 2000; Concha et al., 2000; Bisgrove et al., 2000) (see Fig. 2A,B). cyc, lft1and pitx2 were also expressed in the left side of the epithalamus(Fig. 2C-E). In fbymutants, expression of spaw, cyc, lft1 and pitx2 was indistinguishable from that in WT siblings in both the left LPM and left epithalamus (Fig. 2F-J). The right-sided position of the pancreas and rightward looping of the heart tube,which are influenced by Nodal signaling in the left LPM(Yan et al., 1999), were also unaffected in fby mutants (data not shown).

Previous studies have shown that the left-sided parapineal organ is required for the asymmetric identity of the habenulae, and therefore for L-R habenular neurons to adopt D/V differences in their projection to the IPN(Gamse et al., 2003; Gamse et al., 2005; Concha et al., 2003; Aizawa et al., 2005). To determine if the altered laterality of the habenulae in fby mutants was due to disrupted development of the parapineal, we examined pineal complex formation.

The pineal complex adopted an abnormal morphology in fby mutants. In 24-hpf WT larvae, the otx5 gene is expressed throughout the pineal complex anlage (Gamse et al.,2002) (Fig. 3A). By 34 hpf, the otx5-expressing parapineal is revealed as a morphologically distinct group of cells emerging from the left anterior border of the pineal complex anlage (Gamse et al., 2002) (Fig. 3C). By 4 dpf, parapineal cells are found in a more posterior and ventral position adjacent to the left habenula(Gamse et al., 2003; Concha et al., 2003)(Fig. 3E,G). In fbymutants, the pineal complex anlage appeared similar to WT at 24 hpf(Fig. 3B), but by 34 hpf a distinct parapineal organ was not found(Fig. 3D). At 4 dpf, individual otx5-expressing cells were ectopically located below the pineal organ(Fig. 3F,H).

A number of molecular tools allow pineal and parapineal cells to be distinguished in WT larvae. The growth factor independent 1(gfi1) gene is selectively expressed at 48 hpf in a tight cluster of parapineal cells on the left side of the brain (10±1 cells, n=23) (Dufourcq et al.,2004) (see Fig. 4A,C and Table 1)and are never detected in the pineal. We also examined pineal-specific markers, in the context of the transgenic line Tg(foxd3:GFP)^fkg17 that labels both pineal and parapineal cells with GFP (Gilmour et al.,2002; Concha et al.,2003) (see Fig. S1 in the supplementary material). The entire pineal complex of WT larvae consisted of 60±10 GFP-positive cells(n=13, Table 1). Red-green cone cells labeled by the Zpr1 antibody were found in the pineal but were absent from the parapineal (Fig. 4E). Almost half of the GFP-positive pineal cells were identified in confocal sections as red-green cone cells (44%, n=13, Table 1). Rod cells in the dorsal midline of the pineal organ were labeled by the anti-rhodopsin antibody(Concha et al., 2000). Parapineal cells were negative for rhodopsin expression(Fig. 4G,I, Table 1).

In fby mutants, some ectopic cells located ventral to the pineal organ were labeled by gfi1 at 4 dpf(Fig. 4B). However, there were far fewer gfi1-expressing cells in mutants than in WT larvae(Table 1), and these were scattered near the midline of the brain, rather than tightly clustered on the left as in WT (Fig. 4D). Some of the ectopic cells in fby mutants sent out projections reminiscent of WT parapineal axonal projections to the left habenula(Fig. 4, compare I with J). In mutants, projections from the ectopic cells did not appear to extend to either habenula, as assessed by GFP expression(Fig. 4J and see Fig. S1 in the supplementary material). Other ectopically positioned cells expressed the pineal cone-cell marker Zpr1, but not rhodopsin(Fig. 4F,H,J). These data indicate that the ectopic cell population consisted of a mixture of pineal and parapineal cells. The few parapineal cells that did develop in fbymutants failed to organize into a coherent structure.

Migration of a group of cells, such as the lateral line primordium,requires polarization of the cell cluster (reviewed by Ghysen and Dambly-Chaudiere,2007). The failure of parapineal migration in fby mutants could therefore be a secondary consequence of specifying too few parapineal cells to form a polarized group. To test whether parapineal migration and cohesion depend on cell number, ∼5 cells were ablated at 31 hpf, prior to their movement away from the pineal complex anlage, leaving a number of cells comparable to fby mutants (Fig. 5A-F). The placement of the remaining parapineal cells was assayed at 55 hpf, when they have moved leftward, as well as posteriorly and ventrally, and begin to express gfi1. We found that the remaining parapineal cells formed a coherent and correctly positioned, albeit smaller,parapineal organ, similar to controls (Fig. 5G-L).

The fby mutation was mapped to a region of chromosome 15,approximately 100 kb telomeric to the SSLP marker z22430(Fig. 6A). Examination of the zebrafish genome sequence (Hubbard et al.,2007) revealed two genes, tbx2b and tbx4, in this region. The tbx2b gene was a likely candidate based on previous reports of expression in the pineal anlage(Ruvinsky et al., 2000; Dheen et al., 1999) and the observation that antisense morpholino (MO)-treated embryos developed abnormal otic vesicles similar to those seen in fby mutants(Gross and Dowling, 2005). Sequencing of tbx2b cDNA from fby mutants at 4 dpf revealed a T-to-A transversion introducing a premature stop codon(Fig. 6B). This lesion lies within the T-box sequence and is predicted to prevent translation of 9 of the 20 DNA-contacting residues (Sinha et al.,2000). Two different MOs that blocked either tbx2btranslation or removal of the third intron(Gross and Dowling, 2005)effectively phenocopied the fby mutation(Fig. 6C,D). Microinjection of tbx2b MOs into fby mutants did not cause a further reduction in the number of gfi1-expressing cells (data not shown), supporting the premise that the fby mutation eliminates tbx2bfunction.

Expression of tbx2b in the cells that give rise to the medially located pineal complex anlage was present at the six-somite stage, in two clusters of cells at each lateral edge of the neural plate(Fig. 7A). By the ten-somite stage, when the lateral edges of the neural plate meet to form the neural rod,a single dorsal midline domain of tbx2b expression was detected(Fig. 7B)(Ruvinsky et al., 2000). At 24 hpf, expression of tbx2b was found in many cells of the pineal complex anlage, and was most intense in the midline(Ruvinsky et al., 2000; Dheen et al., 1999; Cau and Wilson, 2003)(Fig. 7C). At 4 dpf, tbx2b expression was maintained in the parapineal organ and the pineal stalk (Fig. 7D). In fby mutants, tbx2b RNA expression was still detected in the pineal complex anlage at a similar level to that in WT(Fig. 7E,F), indicating that the fby mutation did not cause nonsense-mediated decay.

The tbx2b expression pattern resembled that of flh, a homeobox transcription factor also expressed in the pineal complex anlage(Masai et al., 1997), although the two expression patterns did not completely overlap. During somitogenesis,many flh-positive cells expressed tbx2b(Fig. 7G). However, some medial cells of the pineal complex anlage expressed tbx2b only(Fig. 7H). By 24 hpf, the pineal complex anlage expressed both flh and tbx2b, except at its most lateral edges, where only flh expression was detected(Fig. 7I,J).

Previous reports indicated that during neural plate formation in zebrafish, tbx2b expression is regulated by the Wnt receptor frizzled 7a (fzd7a) (Fong et al.,2005), and that in the developing mouse hindbrain, Tbx2regulates expression of Wnt planar cell polarity (PCP) genes including Fzd7 (Song et al.,2006). However, we find that disrupting the Wnt PCP signaling pathway, by mutation of the proteoglycan knypek (glypican)(Topczewski et al., 2001) or van gogh-like 2 (strabismus; also known as trilobite in zebrafish) (Jessen et al., 2002), has no effect on tbx2b expression,parapineal specification or migration (see Fig. S2A-I in the supplementary material). In addition, neither fzd7a nor other PCP genes (wnt11,wnt5, fzd7b, trilobite) are expressed in the epithalamus (see Fig. S2J-Q in the supplementary material; data not shown).

In fby/tbx2b mutants, the absence of a distinct left-sided parapineal organ and the development of parapineal cells in an ectopic ventral location suggested that parapineal cells failed to move to the left side of the brain. Although studies have inferred that parapineal precursors migrate out from the anlage to the left side of the brain(Gamse et al., 2003; Concha et al., 2003), these cellular movements have not been directly documented in live embryos. We performed time-lapse analysis using foxd3:gfp transgenic embryos,recording cell locations between 24 and 48 hpf (see Movie 1 in the supplementary material). One or two highly protrusive cells extending multiple short and wide filipodia were detected at the anterior left edge of the pineal complex anlage at 31 hpf (Fig. 8A). Over the next 8 hours, these cells migrated in a leftward and slightly caudal direction (Fig. 8D,G). They were very closely followed by more cells that originated from the same anterior left region of the pineal complex anlage(Fig. 8J). By 48 hpf, the cells stopped moving, they coalesced into a tightly packed cluster and extended highly dynamic, thin processes (Fig. 8M,P).

In contrast to the robust migration of parapineal cells to the left side of the brain, leftward migrating cells were not observed in tbx2bMO-treated or fby homozygous mutant embryos (see Movies 2, 3 in the supplementary material). At 33 hpf, one or two highly protrusive cells were seen at the anterior end of the pineal complex anlage(Fig. 8B,C). They were joined by an additional two or three highly protrusive cells over the next 9 hours(Fig. 8E,F,H,I,K,L). However,these cells did not migrate to the left of the pineal, but remained close to the midline near the anterior end of the pineal complex anlage. Although in an ectopic location, these cells still extended long thin processes by 48 hours(Fig. 8N,O,Q,R). The time-lapse analyses indicated that tbx2b is required for the leftward migration of parapineal cells and their morphogenesis into a coherent cluster.

In this study, we examined the specification and migration of cells that form the parapineal organ using the fby mutation, which disrupts the Tbx2b transcription factor. In fby mutants, fewer parapineal cells develop, and these cells fail to migrate to the left side of the brain to coalesce into the parapineal organ. Our findings have therefore identified a genetic regulator of a distinct subpopulation of cells in the pineal complex anlage that is important for their specification and migration.

The tbx2b gene is expressed in many cells of the pineal complex anlage, but only a dozen cells contribute to the parapineal organ. How does a widely expressed transcription factor play a role in the development of a limited number of parapineal cells? One likely model is that different cell types in the pineal complex anlage are specified by different repertoires of transcription factors that interact with tbx2b. Modulation of enhancer-binding activity by interaction with other transcription factors(particularly Lim, Gata and homeodomain proteins) is a common feature of Tbx family members (Showell et al.,2004; Naiche et al.,2005). The tbx2b gene is co-expressed with the homeobox gene flh in most of the pineal complex anlage at eight somites. However, a small symmetric domain in the anterior medial region of the anlage expresses tbx2b but not flh. We propose that cells from this anterior medial domain will be specified as parapineal, whereas cells expressing both flh and tbx2b will generate the pineal organ. This model predicts that the fby mutation causes parapineal precursors to be respecified to a pineal fate. In support of this model, we find that in addition to reduced numbers of gfi1-expressing parapineal cells, red-green cone cells typically restricted to the pineal are instead situated in an ectopic position ventral to the pineal organ.

The absence of a tbx2b-exclusive domain in the pineal complex anlage of fby mutants affects parapineal specification; however, a few cells still adopt parapineal identity as assessed by gfi1expression. The action of tbx2b in combination with additional transcription factors during parapineal and pineal specification will be directly tested by conditional misexpression in the epithalamus using tissue-specific promoters.

A trivial explanation for the formation of a few gfi1-expressing cells in fby mutants is that the mutation is hypomorphic; however,the c144 ENU lesion produces a premature stop codon that is predicted to abolish Tbx2b binding to DNA. Moreover, depletion of tbx2b in WT or fby mutant embryos by two highly effective MOs phenocopied all aspects of the fby mutation, but did not eliminate gfi1-expressing cells completely. It is unlikely that a second zebrafish Tbx gene partially compensates for the loss of tbx2b,because the closest relatives of tbx2b (tbx2a, tbx3, tbx4and tbx5) do not appear to be expressed in the pineal primordium(Ruvinsky et al., 2000; Dheen et al., 1999).

In addition to expression in the pineal complex, tbx2b is expressed in the chordoneural hinge (CNH), a structure in the embryonic tailbud (Ruvinsky et al.,2000; Dheen et al.,1999). Expression of the Nodal gene southpaw in the CNH is required for its subsequent asymmetric expression in the left LPM, where it regulates left-sided expression of another Nodal gene, cyc, in the LPM and epithalamus (Gourronc et al.,2007). Loss of spaw activity results in the absence of cyc expression in the left epithalamus(Long et al., 2003) and in L-R randomization of parapineal location(Gamse et al., 2005). However, fby mutant embryos develop with normal left-sided expression of Nodal genes, indicating that tbx2b activity in the CNH is not required for initiation of asymmetric spaw expression. This suggests that tbx2b acts in the pineal complex anlage to regulate specification of the parapineal organ in a pathway that is independent of Nodal signaling.

In WT embryos, migration of parapineal precursor cells from the pineal complex anlage begins with the emergence of a single cell from the anterior left region, followed by other closely apposed cells. This sequence suggests a leader-follower mechanism for directional migration, in which cells at the leading edge coordinate the behavior of the group. Such cellular behavior has been best described in zebrafish for the migration of the lateral line primordium, where leader cells expressing cxcr4b at the anterior end of the primordium organize the migration of the rest of the cells(Haas and Gilmour, 2006). The first cells to emerge from the pineal complex anlage may be leader cells that direct the left-sided movement of the parapineal organ. In fbymutants, a reduced number of cells develop with parapineal characteristics,including gfi1 expression and anteriorward axonal projections, and they do not migrate to the left side of the epithalamus; rather, at 48 hpf they are found in the midline, at the anterior edge of the pineal complex. The failure of parapineal cells to migrate in fby mutants could be due to the reduced number of parapineal precursor cells that are specified, the inability of mutant cells to move, or a loss of leader cell function. We believe that the latter hypothesis is correct. First, reducing the total number of parapineal precursor cells by laser ablation to a number similar to that of the scattered cells in the fby mutant, still allows a cohesive organ to form on the left side of the brain. Second, fbymutant parapineal cells still move posteriorly and ventrally, and thus are not immobile. Therefore, we speculate that tbx2b is required for the specification or maintenance of leader cells. Testing this hypothesis will require the development of methods to identify the leader cells and specifically destroy them as they emerge from the pineal complex.

One mechanism invoked to mediate tbx2b regulation of migration is the Wnt-PCP pathway. Although tbx2b plays a role in PCP in other contexts (Fong et al., 2005; Song et al., 2006), we find no role for PCP in parapineal migration.

Delayed parapineal migration away from the midline until after 48 hpf has been observed in masterblind (mbl; also known as axin1) and RNA-rescued one-eyed pinhead (Roep)mutants in a significant percentage of larvae(Gamse et al., 2002; Carl et al., 2007). However, by 96 hpf, the parapineal cells of mbl and Roep mutants have achieved their correct left-sided position, whereas the parapineal cells of fby mutants remain scattered close to the midline. The data suggest that mbl and oep regulate the timing or speed of migration,but are not required for leader cell formation.

Fish, frog, lizard and bird embryos all develop a pineal and a pineal accessory organ (although in birds the parapineal does not persist in the adult), whereas mammals seem to form only the pineal gland(Hill, 1892; Butler and Hodos, 1996; Braitenberg and Kemali, 1970; Borg et al., 1983). The development of a pineal complex that includes both a pineal and an accessory organ is phylogenetically correlated with the expression of Tbx homeobox genes in the pineal complex anlage. Such expression is observed for tbx2bin zebrafish (Ruvinsky et al.,2000; Dheen et al.,1999) (this study), and for the closely related Tbx3 gene in chick (Gibson-Brown et al.,1998) and Xenopus(Takabatake et al., 2000). However, in the mouse, neither Tbx2 nor Tbx3 is expressed in the developing diencephalon (Chapman et al., 1996). We propose that the pineal accessory organ of non-mammalian vertebrates is specified by Tbx2/3 function.

In lizards and frogs, the pineal accessory organ (known as the parietal eye or frontal organ, respectively) is photoreceptive, expresses certain opsin family members and is thought to transmit diurnal information(Engbretson et al., 1981; Eldred et al., 1980; Su et al., 2006; Blackshaw and Snyder, 1997). In mammals, which do not receive direct light input to the pineal, the major source of diurnal information is stimulation of the melanopsin-expressing retinal ganglion cells (Hattar et al.,2002). The acquisition of photic input to the pineal via the eyes in mammals rendered a photoreceptive pineal accessory organ unnecessary. Loss of Tbx2/3 expression in the pineal complex anlage might be the mechanism that prevents formation of a pineal accessory organ in mammals, by altered specification and disrupted migration of precursors. How a transcription factor regulates the number and behavior of parapineal cells will be revealed by identification of its downstream target genes.

We thank Jeff Gross and John Dowling for tbx2b morpholino reagents; Paul Hargrave for the rhodopsin antibody; Lila Solnica-Krezel for knypek mutants and PCP gene clones; Jason Jesson for trilobite mutants; Chris Wright and Bruce Appel for valuable discussions; and Lijun Bian, Erin Booton and Lea Fortuno for technical help. Support was provided by a zebrafish initiative funded by the Vanderbilt University Academic Venture Capital Fund and a grant from the Whitehall Foundation to J.T.G., and a N.I.H. grant (HD042215) to M.E.H.