Mesendoderm and left-right brain, heart and gut development are differentially regulated by pitx2 isoforms

The three primary tissue layers, endoderm, mesoderm and ectoderm, are defined during vertebrate gastrulation. Of the three, the origin and patterning of the most internal layer, the endoderm, is least understood. At late blastula stages during zebrafish embryogenesis, the blastoderm begins to thin and spread over the large yolk cell in a process termed epiboly. Marginal cells located around the entire circumference of the blastoderm are fated to become both mesoderm and endoderm during gastrulation (Kimmel et al., 1990). As determined by fate mapping, the mesoderm domain extends more cell-diameter lengths from the blastoderm margin than the endoderm; the cells closest to the yolk correspond to both presumptive mesoderm and endoderm (Warga and Nusslein-Volhard, 1999). During gastrulation, dorsal cells in the embryonic shield involute to form the notochord and the mesendoderm of the prechordal plate (Shih and Fraser, 1995). The endoderm component of the prechordal plate later contributes to the internal organs, and the mesoderm forms the hatching gland and head mesoderm. Signals from the shield and subsequent midline, including the prechordal plate, are responsible for patterning the embryo along the anterior-posterior, dorsal-ventral and left-right axes (Chen et al., 1997; Schier and Talbot, 1998).

Pitx2 is a member of the bicoid-class of paired homeodomain transcription factors that is mutated in patients with Reiger syndrome (Semina et al., 1996). Pitx2 has also been implicated in late stages of left-right development and is expressed bilaterally in anterior mesendoderm and on the left side in the lateral plate mesoderm, which forms the cardiac and visceral primordia in mouse (Meno et al., 1998; Piedra et al., 1998; Ryan et al., 1998; Yoshioka et al., 1998), chick (Logan et al., 1998; Piedra et al., 1998; Ryan et al., 1998; St Amand et al., 1998; Yoshioka et al., 1998), frog (Campione et al., 1999; Ryan et al., 1998) and zebrafish (Campione et al., 1999). Isoforms of human pitx2 are expressed by alternative splicing and the use of two promoters (Arakawa et al., 1998; Semina et al., 1996), generating proteins identical throughout their homeodomains and carboxy termini but different in their amino termini. Previous developmental studies have not made distinctions among pitx2 isoforms, and neither the developmental upstream regulatory pathways nor downstream functions of pitx2 isoforms are known. Here we identify pitx2 isoforms in zebrafish and examine their differential expression and functions during development of the mesendoderm, prechordal plate and left-right asymmetry.

Mesendoderm patterning appears to involve complex inductive and antagonistic interactions. In zebrafish, mutations in cyclops (cyc) (Hatta et al., 1991), squint (sqt) (Heisenberg and Nusslein-Volhard, 1997) and one-eyed pinhead (oep) (Hammerschmidt et al., 1996; Schier et al., 1996) lead to cyclopia and demonstrate a requirement for these genes in formation of the prechordal plate. Cyc and Sqt are members of the nodal-related transforming growth factor β (TGFβ) cell-cell signaling molecule family (Feldman et al., 1998; Rebagliati et al., 1998b; Sampath et al., 1998), while oep encodes a novel EGF-CFC transmembrane protein that appears to mediate both Cyc and Sqt signaling (Gritsman et al., 1999; Zhang et al., 1998). Conversely, other members of the TGFβ family, the zebrafish lefty-related genes (lft1 and lft2), are expressed in the same mesendoderm precursors in zebrafish but are antagonistic to nodal-related signaling (Bisgrove et al., 1999; Meno et al., 1999; Thisse and Thisse, 1999). Ectopic expression of either lft1 or lft2 in zebrafish also produces cyclopia, suggesting that a balance between Lft and Cyc/Sqt/Oep signaling is required for proper prechordal plate development and ventral patterning of the central nervous system (CNS).

The interactions of pitx2 isoforms with genes implicated in mesendoderm development are examined here. The isoform expressed in early mesendoderm, pitx2c, is regulated by TGFβ signaling that controls prechordal plate development, including cyc and oep. In addition, increased expression of pitx2c induces cyc expression, suggesting a novel feedback mechanism on cyc by Pitx2c. Early ectopic expression of pitx2a or pitx2c disrupts gastrulation with the isoforms regulating their own and cross-regulating each other’s expression.

In addition to early roles in mesendoderm patterning, cyc, lft1 and lft2 are expressed asymmetrically in zebrafish during late somite stages in the left dorsal diencephalon and left lateral plate mesoderm (Bisgrove et al., 1999; Rebagliati et al., 1998a; Sampath et al., 1998; Thisse and Thisse, 1999). Surprisingly, pitx2 isoforms are differentially expressed in these left primordia. Targeted ectopic expression assays in Xenopus embryos show that both Pitx2a and Pitx2c are competent to randomize left-right development; however, Pitx2a has a greater impact on cardiac asymmetry than Pitx2c. Taken together, these results show differential regulation and distinct functions of pitx2 isoforms during both early mesendoderm development and later left-right development.

Zebrafish embryos were obtained by natural spawning and maintained at 28.5°C in system water (Westerfield, 1995). Fish heterozygous for cyclops (cyc^b229 and cyc^tf219) (Brand et al., 1996; Hatta et al., 1991), iguana (igu^ts294e) (Brand et al., 1996), no tail (ntl^b195) (Halpern et al., 1993), one-eyed pinhead (oep^m134) (Schier et al., 1996), schmalspur (sur^ty68b) (Brand et al., 1996) and spadetail (spt^b104) (Kimmel et al., 1989) were bred to obtain homozygous mutant embryos.

A 500 bp fragment with identity to pitx2 from other species was amplified by degenerate primer-based PCR using 24 hours post-fertilization (hpf) embryo cDNA as a template. Degenerate primers were designed using the CODEHOP program (Rose et al., 1998): 5′-CCC GGG AGG AGA TCG CNG TNT GGA C-3′ and 5′-TGC AGG TGT CCC GGT ACA CRT ANG GNG G-3′. Total RNA was isolated from zebrafish embryos using Trizol (BRL) and further purified using Qiagen RNeasy minicolumns. cDNA templates for PCR were reverse-transcribed from RNA using oligo d(T) and Superscript II reverse transcriptase (BRL). The PCR product was amplified with Platinum Taq (BRL) and cloned into the vector pCRII using the TOPO TA cloning kit (Invitrogen). This fragment was used to screen a late somite cDNA library (a generous gift from David Grunwald, University of Utah) at reduced stringency: 30% formamide, 5× SSPE (Sigma), 5× Denhardt’s solution, 0.5% SDS, 100 μg/ml yeast RNA (Sigma) at 42°C. Membranes were washed at a final stringency of 2× SSPE/0.1% SDS at 50°C. Fifteen positive clones were subjected to restriction enzyme analysis and sequencing and placed into two groups corresponding to pitx2a and pitx2c.

Total RNA from staged embryos was isolated as above, and 5 μg from each stage were treated with glyoxal (Ausubel et al., 1987) for 1 hour at 50°C. The RNA was fractionated in a 1% agarose gel in 1× TAE and transferred with 20× SSC. Isoform specific probes for pitx2a (5′-pitx2a) and pitx2c (5′-pitx2c) were generated from the 5′ untranslated regions (UTRs). The 400 bp amplification fragment corresponding to the 5′ UTR from pitx2a was cloned using 5′ RACE-PCR (BRL) with the following oligonucleotides for reverse transcription (RT) and PCR: RT, 5′-TTC TCT AAT TGA GCA CAC GTT GAT GC-3′; PCR, 5′-CAC GTT GAT GCA AGT TTG CGG-3′. A 5′-UTR probe to pitx2c was generated by subcloning a 400 bp BamHI restriction fragment from the full-length pitx2c cDNA into pBSIIKS− (Stratagene). DNA probes were labeled by random priming in the presence of [α-³²P]dCTP using the Decaprime kit (Ambion). Hybridizations were carried out using Quik Hyb (Stratagene) following the manufacturer’s instructions.

A full-length pitx2a probe was used to detect all mRNAs expressed from the pitx2 gene by in situ hybridization. Isoform-specific probes were the same as used above for northern hybridizations. Probes generated from cDNAs corresponding to gsc (Stachel et al., 1993), no tail (Schulte-Merker et al., 1992), cyc (Rebagliati et al., 1998a) and lft2 (Bisgrove et al., 1999) were used as molecular markers. Probes were produced by transcription from linear templates using the Maxiscript kit (Ambion) in the presence of either digoxigenin-11-UTP or fluorescein-12-UTP (BMB). Free nucleotides were removed using Bio-Gel P-6 Micro Bio Spin columns (BioRad).

Embryos were fixed in 4% paraformaldehyde in sucrose buffer (Westerfield, 1995) overnight at 4°C. In situ hybridizations were carried out at 70°C (Essner et al., 1996). Anti-digoxigenin-or anti-fluorescein-alkaline phosphatase (AP)-conjugated antibodies (BMB) were used at a 1:2000 dilution for detection of digoxigenin- and fluorescein-labeled probes, respectively. For double in situ hybridizations, probes were hybridized simultaneously and detected using sequential alkaline phosphatase reactions. After development of the first probe with NBT and BCIP (Sigma), the anti-digoxigenin antibody was removed by incubating the embryos in 100 mM glycine, pH 2.2 for 20 minutes followed by three washes in PBS with 0.1% Tween 20 (PBST). Embryos were refixed overnight in 4% paraformaldehyde in PBS. An anti-fluorescein-AP antibody against the second probe was absorbed to the embryos and detected with Fast Red (Sigma). Embryos were refixed, cleared in either 70% glycerol in PBS or 2:1 benzylbenzoate:benzyl alcohol and photographed with a Leica MZ12 stereoscope.

For construction of expression vectors, the coding sequences of pitx2a, pitx2c and gsc were PCR amplified with Pfu polymerase (Stratagene). The PCR products were restriction-digested and directionally ligated into pCS2+, a vector that allows both the production of in vitro synthesized transcripts and in vivo DNA expression in embryos from a CMV promoter and enhancer (Rupp et al., 1994; Turner and Weintraub, 1994). The cloned cDNA fragments were sequenced to confirm identity and integrity.

For injection of synthetic mRNAs into early zebrafish embryos, linear templates from pCS2+/pitx2a and pitx2c were produced for transcription. The templates were transcribed in vitro with SP6 RNA polymerase using the Message Machine transcription kit (Ambion). The concentrations of the RNAs were adjusted to inject approximately 5 pg of RNA for pitx2a and pitx2c and 1 pg for gsc per zebrafish embryo. Embryos were injected at the 1-to 2-cell stage.

For targeted left-right asymmetric expression of zebrafish pitx2a and pitx2c in Xenopus embryos, either pCS2+/pitx2a or pitx2c DNA was injected at the future margin of either the ventral left or ventral right blastomere of 4-cell embryos (stage 3) (Nieuwkoop and Faber, 1967). Approximately 15-25 pg of either pitx2 DNA was coinjected with 15-25 pg of pEGFP-N1 DNA (Clonetech), and the correct right or left localization of GFP expression was confirmed using epifluorescence illumination of living embryos between stages 43-46. Correctly targeted embryos that did not display dorsoventral or anteroposterior defects were scored for heart and gut morphologies at stage 46.

For antibody staining, Xenopus embryos were fixed at stage 46 in 4% paraformaldehyde/PBS overnight at 4°C and subsequently rinsed and stored in PBS. Embryos were dehydrated with a series of methanol washes and placed at −20°C in 100% methanol overnight. After rehydration, embryos were washed with PBST and blocked in 5% sheep serum and 2% bovine serum albumin (BSA) in PBST at room temperature. Monoclonal antibodies against bovine cardiac troponin T (DSHB University of Iowa) were diluted 1:1 with blocking solution and absorbed to embryos overnight at 4°C. Embryos were washed with 2% BSA in PBST at room temperature. Fluorescein-conjugated, goat anti-mouse IgG2a (Southern Biotechnology Associates, Inc.) secondary antibodies were diluted 1:100 in blocking solution and incubated with embryos overnight at 4°C. Embryos were washed with 2% BSA in PBST and cleared in 70% glycerol containing Slow Fade (Molecular Probes). Embryos were visualized at 5× magnification using epifluorescence on a Leica DMR microscope and photographed using a DAGE color video camera.

To understand the conservation of pitx2 during vertebrate evolution and its involvement in head and asymmetric organ development, we used degenerate RT-PCR to clone pitx2 from late somite stage zebrafish embryos. A 500 bp fragment with identity to pitx2 from other vertebrate species was cloned and used to screen a 20-28 hpf stage library. Two classes of clones corresponding to two different isoforms of pitx2 from humans and mice were recovered and designated zebrafish pitx2a and pitx2c (GenBank accession numbers: AF132447 and AF132446). These isoforms are identical in their 3′-UTRs and most of their coding sequences; however, they differ in their 5′-UTRs and N-terminal coding sequences (Fig. 1). In cultured human cell lines, different pitx2 isoforms arise from utilization of both different promoters and alternative splicing (Arakawa et al., 1998; Semina et al., 1996). The human pitx2a and pitx2b isoforms represent transcription from the same promoter and differ by alternative splicing of exon 2, while the pitx2c isoform is generated from a separate promoter. While the genomic structure of the zebrafish pitx2 gene is not known, the encoded zebrafish Pitx2a protein displays 60-73% identity to Pitx2a from Xenopus (Campione et al., 1999), mice (Gage and Camper, 1997) and humans (Arakawa et al., 1998) over 15 N-terminal amino acids unique to Pitx2a (Fig. 1B). The zebrafish Pitx2c protein shows a lower amino acid identity (38-50%) to that of Xenopus (Campione et al., 1999), chicken (St Amand et al., 1998), mice (Arakawa et al., 1998) and humans (Arakawa et al., 1998) over a 61-amino-acid region unique to Pitx2c (Fig. 1C). The remainder of the zebrafish Pitx2 protein, common to all isoforms and containing the bicoid-like homeodomain, displays approximately 95% identity to Pitx2 from other species (Fig. 1D). The high degree of structural conservation among pitx2 isoforms suggests conserved functional roles during development.

To define the potential roles of the zebrafish pitx2 isoforms during early development, pitx2 expression patterns were determined by in situ hybridization. Hybridizations were carried out with both a full-length pitx2 probe in order to detect all pitx2 transcripts and unique 5′-UTR probes to characterize isoform-specific expression (Fig. 2A). Expression of pitx2 was observed during dome stage to 60% epiboly in a ring of marginal cells around the entire circumference of the embryo (Fig. 2Aa-d). A similar hybridization pattern was observed with both the pitx2c-specific probe and the full-length pitx2 probe. Interestingly, the marginal cells expressing pitx2c were found only 4 cell-diameter lengths from the margin during dome to 40% epiboly (data not shown). These cells give rise to both endoderm and mesoderm while cells further from the margin give rise to mesoderm only (Warga and Nusslein-Volhard, 1999). By 60% epiboly pitx2c-expressing cells had a deep location in the blastoderm next to the yolk cell, suggesting an endodermal phenotype (data not shown). At 40% epiboly an increased number of cells on the dorsal side accumulated pitx2c transcripts. During gastrulation, dorsal marginal cells involute and form the prechordal plate and notochord (Kimmel et al., 1995). pitx2c was detected in both the anterior (polster) and posterior prechordal plate mesendoderm at 90% epiboly (Fig. 2Ae-h). During these early phases of pitx2 expression, pitx2a was not detected (Fig. 2Ac,g). However, at bud stage following gastrulation, pitx2a was observed in the polster and pitx2c in both the polster and posterior prechordal plate mesendoderm (Fig. 2Ai-l). At 4-to 6-somite stages, pitx2a and pitx2c expression was detected in the polster (Fig. 2Am-p). Additionally, weak pitx2c expression remained in the posterior prechordal plate (Fig. 2Ap). As detected with the full-length pitx2 probe, pitx2a transcripts were also observed in lateral head mesendodermal cells extending from the polster (Fig. 2An,o).

To further examine the temporal onset of expression of each isoform, northern blot hybridizations were performed (Fig. 2B). In agreement with in situ hybridization, pitx2c was detected during dome stage while pitx2a was not observed until bud stage. Furthermore, pitx2a showed strong expression during heart looping stages (30-36 hpf) and at 48 hpf. The differential onset of pitx2 isoform expression suggests that pitx2c plays an earlier role than pitx2a and that these isoforms have distinct roles during early anterior mesendoderm development.

To understand the regulation of pitx2c and its involvement in early mesendoderm development, the expression of pitx2 was examined in mutant embryos defective in early mesendoderm development. In mutants in which pitx2 expression was changed, the alterations occurred in approximately one quarter of the embryo population analyzed from intercrosses of identified heterozygotes (Table 1). Two mutants, spadetail (spt) and oep, displayed altered pitx2 expression at 50% epiboly. spt corresponds to a member of the T-box class of transcription factors (Griffin et al., 1998), and mutant embryos display defects in specification of trunk and tail mesoderm but no obvious head defects (Ho and Kane, 1990). At 40% epiboly cells fated to become the shield in spt mutant embryos do not undergo compaction and exhibit defects in dorsal convergence (Warga and Nusslein-Volhard, 1998). In spt mutant embryos at 50% epiboly, pitx2 expression was absent from superficial dorsal cells but still expressed in non-dorsal marginal cells (Fig. 3A,B).

At 50% epiboly, pitx2 expression was completely absent in oep mutant embryos (Fig. 3C), which display strong defects in anterior mesendoderm specification. The zebrafish oep gene encodes a novel EGF-CFC transmembrane molecule that is implicated in mediating cell-cell signaling by the TGFβ family members Cyc and Sqt (Gritsman et al., 1999; Zhang et al., 1998). In contrast to the loss of pitx2c expression in oep mutants, pitx2 expression was normal in cyc (cyc^b229) mutant embryos during these early stages (Table 1). This indicates the oep-dependent expression of pitx2c at late blastula stages does not occur solely through the cyc pathway; it might be regulated in conjunction with another member of the nodal family such as sqt. pitx2 expression was also normal during late blastula stages in other mutant embryos defective in mesendoderm development and dorsal specification such as no tail (ntl), iguana (igu) and schmalspur (sur), suggesting the altered pitx2c expression in spt and oep is not merely due to alterations in cell fate. Together, these results implicate spt and oep in the specific regulation of both pitx2c expression and early mesendoderm development.

To examine the regulation of pitx2c during early prechordal plate development, pitx2 expression was examined at 80-90% epiboly in igu, spt, sur, cyc and oep mutant embryos (Table 1). With the exception of spt, all these mutant embryos display varying degrees of cyclopia due to defects in the specification of ventral cell fates in the CNS (Brand et al., 1996). In igu mutant embryos, pitx2 expression appeared in a narrower domain, indicating reduced prechordal plate mesendoderm (Fig. 3D,E). This suggests in addition to the strong CNS defects resulting in posterior cyclopia (Brand et al., 1996), igu mutant embryos also have reduced prechordal plate mesendoderm. In agreement with the early reduction in pitx2 expression on the dorsal side of the embryo, spt mutant embryos displayed a loss of pitx2 expression in the posterior domain of the prechordal plate, but maintained expression in the polster (Fig. 3F). In sur mutant embryos, which display posterior cyclopia similar to igu mutant embryos, pitx2 expression was absent except for a small domain in the polster (Fig. 3G). In contrast to the normal expression of pitx2 at 40% epiboly in cyc mutant embryos, pitx2 expression was absent at 90% epiboly in cyc mutants except in a very small number of cells at the anterior end of its normal expression domain (Fig. 3H). This effect was observed with both a deletion (cyc^b229) and a point mutation (cyc^tf219) allele of cyc (Table 1). As in oep embryos at 50% epiboly, pitx2 expression was also absent from the prechordal plate at 80% epiboly (Fig. 3I). In contrast, ntl mutant embryos, which display defects in notochord morphogenesis, showed no alteration of pitx2 expression in the prechordal plate domain at shield stage (Table 1). Taken together, pitx2c expression is regulated either directly or indirectly in the prechordal plate mesendoderm by igu, spt, sur, cyc and oep.

pitx2 isoforms have distinct expression patterns during early mesendoderm development. To test whether this corresponds to functional differences in the activity of pitx2 isoforms, pitx2a and pitx2c were ectopically expressed in zebrafish embryos by injection of synthetic mRNA and hybridized with mesendodermal markers. cyc is expressed in the mesendoderm during late blastula stages and the embryonic midline during late zebrafish gastrulation (Rebagliati et al., 1998a; Sampath et al., 1998). At 40% epiboly, cyc was expressed in a ring of marginal cells and concentrated dorsally (Fig. 4A), similar to the expression of pitx2c. Ectopic expression of pitx2a downregulated cyc expression in marginal cells (Table 2, Fig. 4B). However, injection of pitx2c RNA upregulated cyc expression in both marginal and animal cap cells (Fig. 4C), suggesting that Pitx2a and Pitx2c have opposite effects on cyc expression. The ectopic activation of cyc by Pitx2c was also observed at 80% epiboly (data not shown).

Like pitx2, goosecoid (gsc) is a member of the bicoid-class of transcription factors and is expressed in dorsal marginal cells during late blastula stages in zebrafish embryos, corresponding to the future embryonic shield or organizer region (Fig. 4D; Stachel et al., 1993). During late gastrulation, gsc expression is restricted to the prechordal plate mesendoderm. gsc expression domains are conserved in other vertebrate species (Blum et al., 1992; Cho et al., 1991; Izpisua-Belmonte et al., 1993), and disruption of gsc in mice by homologous recombination results in prechordal plate and craniofacial defects (Belo et al., 1998; Rivera-Perez et al., 1995; Yamada et al., 1995). In Xenopus, ectopic gsc expression dorsalizes embryos and can induce a secondary embryonic axis (Cho et al., 1991). In embryos injected with pitx2a or pitx2c RNA, the number of gsc expressing cells on the dorsal side was reduced (Table 2, Fig. 4E,F), suggesting a reduction in dorsal development. However, unlike Pitx2a, Pitx2c produced ectopic patches of gsc expression in other marginal cells (Fig. 4F). This effect of ectopic Pitx2c on gsc correlates with the ectopic activation of cyc by Pitx2c (Fig. 4C).

Ectopic expression of either pitx2a or pitx2c disrupted gastrulation and led to an accumulation of cells at the blastoderm margin by late gastrulation (Fig. 4G-I). This phenotype was also associated with slowed epiboly relative to control embryos. To examine the nature of this perturbation, injected embryos were hybridized with lft2. At 80% epiboly, lft2 is expressed in prechordal plate mesendoderm and floor plate precursors (Fig. 4J; Bisgrove et al., 1999). Injection of pitx2a and pitx2c RNA had differing effects on lft2 expression (Table 2, Fig. 4K,L). In pitx2a-injected embryos, lft2 expression was shortened along the anteroposterior axis and was not observed in cells at the margin (Fig. 4K). In contrast, lft2 expression in pitx2c-injected embryos was observed in both the presumed midline and regions of cell accumulation at the blastoderm margin (Fig. 4L), indicating differences in the regulation of lft2 by Pitx2a and Pitx2c may exist. Based on in situ hybridization analysis with lft2 (Fig. 4J-L) and cyc (data not shown), the excess cells located at the blastoderm margin during gastrulation resulting from pitx2a- and pitx2c-injection were not necessarily coincident with the dorsal side of the embryo. Ectopic expression of either pitx2 isoform may increase a cohesive cell behavior in groups of cells, possibly leading to a delay in epiboly and an increased number of cells that fail to migrate from the blastoderm margin. However, the fate of these cells may differ in embryos injected with pitx2a and pitx2c.

Injected embryos were further analyzed for the effects of pitx2 isoforms on mesoderm patterning by hybridization with ntl. At 80-90% epiboly, ntl is expressed in two domains corresponding to the notochord on the dorsal side and to the mesoderm undergoing involution at the blastoderm margin (Fig. 4M; Schulte-Merker et al., 1992). In pitx2a- or pitx2c-injected embryos, which displayed slowed epiboly, ntl expression was severely reduced in the notochord domain compared to uninjected embryos at 80-90% epiboly (Fig. 4M-O). Additionally, the ring of ntl expression at the blastoderm margin was disrupted in many injected embryos, suggesting that mesoderm specification was disorganized in injected embryos. Correlating with disrupted cell movements and reduced ntl expression during gastrulation, many pitx2a- and pitx2c-injected embryos at 28 hpf displayed severe axis defects, including notochord loss and reduced dorsoanterior structures (Table 2, Fig. 4P-R). Consistent with pitx2 isoforms being involved in distinct regulatory pathways, pitx2c-injected embryos displayed a higher proportion of embryos with reduced head development (Table 2).

Analysis of injected embryos indicated distinct differences in the activities of Pitx2a and Pitx2c on mesendoderm marker gene expression. To further explore these differences, embryos injected with pitx2a or pitx2c were examined at late blastula stages for their ability to cross-regulate each other’s expression. Additionally, the effects of ectopic pitx2 expression were compared to the effects of ectopic gsc expression. Embryos injected with pitx2a displayed ectopic activation of pitx2a in both marginal and animal pole cells at 30% epiboly (Fig. 5A,B), suggesting a pitx2a autoregulatory loop. In contrast, pitx2a-injected embryos displayed no induction of pitx2c (Fig. 5E,F). pitx2c injected embryos displayed ectopic expression of pitx2a and itself, but only in the marginal zone and not in the animal pole cells (Fig. 5C,G). gsc-injected embryos displayed activation of pitx2a and pitx2c in both marginal and animal pole cells (Fig. 5D,H), suggesting Gsc positively regulates both pitx2a and pitx2c expression. Together with the early pitx2c and late pitx2a expression in the prechordal plate mesendoderm, we suggest that Gsc regulates pitx2 isoforms during both early dorsal specification and later prechordal plate development.

At 22-25 somites of development, pitx2 was expressed asymmetrically on the left side in the dorsal diencephalon and developing gut and heart fields (Fig. 6A). pitx2 was also expressed at these stages in the trunk in presumed Rohon-Beard neurons (Fig. 6A), symmetrically at the base of the diencephalon (Fig. 6D), and in superficial cells on the yolk that likely contribute to the hatching gland (Fig. 6A). In situ hybridizations with isoform-specific probes displayed striking differences between pitx2a and pitx2c expression in asymmetric domains. pitx2c, but not pitx2a, was expressed asymmetrically in the dorsal diencephalon (Fig. 6A-F). The expression in the dorsal diencephalon most likely corresponds to the future habenulae nucleus, a structure known to exhibit both structural and functional differences between the left and right side in vertebrate organisms (Morgan, 1991). In the developing gut, only left-sided expression of pitx2c was detected (Fig. 6A-C,G-I).

Conversely, pitx2a, but not pitx2c, displayed left-sided, asymmetric expression in the developing heart field (Fig. 6A-C,J-L). This expression was strongest in the anterior regions of the heart field. In zebrafish, anterior regions of the heart field include the future atrium (Stainier and Fishman, 1992), opposite to the orientation in other vertebrate species. As the yolk is absorbed, the heart comes to lie in its normal orientation with the atrium to the posterior. To further mark different regions of the heart primordia, embryos were hybridized with lft2 (red), which also has left-sided expression in the heart field during late somite stages (Fig. 6J-L) (Bisgrove et al., 1999). The strongest asymmetric expression of pitx2a in the developing heart overlapped with lft2 in the anterior domain of lft2 asymmetric expression. Weaker expression of pitx2a was also observed in left/lateral regions of the lft2 domain. The absence of a complete overlap of pitx2a and lft2 domains suggests that pitx2a and lft2 may have different zones of activity in the development of heart asymmetry. Together, these results indicate that different organ primordia utilize distinct pitx2 isoforms during left-right development.

pitx2a was expressed asymmetrically in the heart field, while pitx2c transcripts were found in both the left diencephalon and developing gut (Fig. 6). To examine the potentially distinct roles of pitx2 isoforms in laterality and heart and gut morphogenesis, targeted left or right expression of pitx2a or pitx2c was carried out in Xenopus embryos. Due to the indeterminate orientations of the early cleavage planes (Helde et al., 1994), injections targeted to the prospective left or right sides are not as readily feasible in zebrafish. Xenopus allows targeted injection of molecules into either the left or right sides and serves as a superb assay system for candidate genes in left-right development (Hyatt and Yost, 1998; Ramsdell and Yost, 1999). Zebrafish pitx2a or pitx2c DNA expression constructs were injected into the ventral left or ventral right blastomere of 4-cell Xenopus embryos, and the effects of ectopic expression on heart and gut morphogenesis were observed on day 5 of development (stage 46).

Normally at stage 46, the heart outflow tract, the conus truncus, extends anteriorly from the right side of the ventricle and loops to the left of the embryo, as shown in uninjected control embryos stained with anti-cardiac troponin T (Fig. 7A). Injection of either pitx2 isoform on the left or right side of the embryo was able to reverse heart laterality (Table 3), with the conus extending from the anterior-left side of the ventricle and looping to the right (Fig. 7B). However, right-sided injection of pitx2a, the isoform expressed asymmetrically in the developing zebrafish heart, produced approximately double the number of heart reversals as compared to pitx2c injections (P<0.1). Furthermore, injection of either isoform into either side sometimes resulted in malrotation of the heart, with both the atrium and ventricle displaced lateral to their normal position (Table 3, Fig. 7C). This effect was observed at a greater frequency in the right-side injected specimens than in the left-injected, and it did not appear to correlate with the laterality phenotype since the incidence of reversals among the malrotated hearts was no higher than that observed among all experimental hearts (data not shown). Although both Pitx2a and Pitx2c altered heart morphogenesis, their competence to affect cardiac laterality was distinct.

The laterality and morphogenesis of the gut was also affected by injection of pitx2a and pitx2c (Fig. 7D-F, Table 3). At stage 46 in Xenopus, the gut extends caudally from the esophagus along the right lateral wall of the abdominal cavity, traverses the cavity posteriorly, and forms a large counterclockwise coil on the left side of the embryo (Fig. 7D; Chalmers and Slack, 1998). Injection of either pitx2 isoform on either side of the embryo was able to reverse gut morphology, resulting in the major intestinal coil being located on the right side of the abdominal cavity (Table 3, Fig. 7E). In addition, many of the injected embryos displayed disorganized coiling of the gut (Fig. 7F). Both dysmorphologies were more frequently observed with right-sided injections. In addition, a higher proportion of embryos had both reversed hearts and reversed guts when pitx2 isoforms were injected on the right rather than on the left side of the embryo (Table 3). These results further demonstrate the ability of pitx2 isoforms to affect morphogenesis and laterality of the internal organs.

We have isolated two isoforms of pitx2 from zebrafish, which show strong conservation to pitx2a and pitx2c from other vertebrate species. pitx2c is expressed during early mesendoderm specification and throughout prechordal plate mesendoderm during gastrulation. In contrast, pitx2a expression is first expressed after gastrulation in only the anterior prechordal plate (polster). Analysis of expression patterns in zebrafish embryos indicates that pitx2 isoforms are under distinct regulation, both spatially and temporally, and may have distinct functions during mesendoderm development. Fig. 8 summarizes the genetic pathways identified by mutant analysis and ectopic expression experiments.

During late blastula stages, pitx2c is expressed at the blastoderm margin in cells fated to become both mesoderm and endoderm (Warga and Nusslein-Volhard, 1999). As development proceeds through gastrulation, cells expressing pitx2c have characteristics of early endoderm (Fig. 2 and data not shown), including a flattened appearance and a deep location in the blastoderm next to the yolk cell (Warga and Nusslein-Volhard, 1999). Specification of the early endoderm has been shown to require TGFβ signaling, whereas mesoderm specification is thought to require both TGFβ and Fibroblast Growth Factor (FGF) signaling pathways (Rodaway et al., 1999). Consistent with a requirement for TGFβ signaling, the expression pattern of pitx2c during late blastula stages is similar to those of members of the TGFβ family, including cyc (Rebagliati et al., 1998a), sqt (Erter et al., 1998; Rebagliati et al., 1998a), lft1 (Bisgrove et al., 1999; Thisse and Thisse, 1999), lft2 (Bisgrove et al., 1999) and a modulator of nodal signaling, oep (Gritsman et al., 1999).

Several lines of evidence show that pitx2c is regulated by TGFβ signaling pathways (Fig. 8). The analysis of pitx2c expression in oep mutants indicates that oep acts as an upstream regulator of pitx2c expression. Homozygous mutant oep embryos display severe defects in mesendoderm formation (Schier et al., 1996; Solnica-Krezel et al., 1996). Oep, an EGF-CFC transmembrane protein (Zhang et al., 1998), may directly modulate Cyc and Sqt signaling (Gritsman et al., 1999), and together with Cyc and Sqt could positively regulate pitx2c expression. Contrary to this hypothesis, pitx2c expression is normal during late blastula stages in cyc mutant embryos (Table 1). This probably reflects an overlapping function of cyc and sqt during this time (Erter et al., 1998; Rebagliati et al., 1998a). In support of a partially redundant function, cyc/sqt double mutant embryos have a more severe reduction in mesendoderm, similar to oep mutant embryos (Gritsman et al., 1999). Furthermore, pitx2c expression is absent later in cyc mutants, after sqt expression is no longer present. In Xenopus, ectopic expression of the nodal-related gene Xnr1, a homolog of cyc, dramatically upregulates pitx2 in animal cap assays (Campione et al., 1999). In both mice and zebrafish, lft, another member of the TGFβ family, antagonizes the activity of cyc (nodal) signaling (Bisgrove et al., 1999; Meno et al., 1999). Ectopic expression of lft1 in zebrafish eliminates the expression of pitx2 at 50% epiboly (Bisgrove et al., 1999). Together these data suggest that pitx2 expression in the mesendoderm is regulated by a balance between the activities of the mutually antagonistic TGFβ signaling molecules, Cyc and Lft.

During early dorsoventral patterning of the mesendoderm, Pitx2 and Gsc, both members of the bicoid-homeobox class of transcription factors, have opposing activities and regulate the expression of one another. gsc expression leads to dorsal development (data not shown) and upregulates pitx2a and pitx2c at late blastula stages. This agrees with studies using Xenopus animal cap assays where expression of gsc upregulated pitx2 (Campione et al., 1999), although in these studies different pitx2 isoforms were not distinguished. Given its activity to regulate pitx2a and pitx2c expression during late blastula stages, Gsc may normally function at different stages of prechordal plate development to affect the expression of both pitx2 isoforms. Early in dorsal mesendoderm formation and later throughout the prechordal plate, Gsc may positively upregulate pitx2c (Fig. 8A). pitx2a is not expressed until after gastrulation and could be repressed by an unknown inhibitor (Fig. 8B). Later, when the inhibition is absent in the polster, Gsc could positively regulate the transcription of pitx2a (Fig. 8B). Our data also shows Pitx2a and Pitx2c negatively affect gsc expression and promote a ventralized phenotype. The opposing activities of Gsc and Pitx2 isoforms during dorsal mesendoderm development suggest that a balance between the activities of Gsc, Pitx2a and Pitx2c is required for the anterior and posterior identity of the prechordal plate mesendoderm at multiple stages of development.

Both Pitx2a and Pitx2c regulate changes in cell behavior. Ectopic expression of either isoform disrupts gastrulation and leads to excess cells at the blastoderm margin (Fig. 4). This phenotype could be a result of increased cellular adhesion in cells overexpressing Pitx2 isoforms. Ectopic expression by RNA injection at the 1-to 2-cell stage leads to mosaic expression in zebrafish, with patches of cells expressing the RNA in random locations within the blastoderm. In pitx2-injected embryos, the random location of the accumulation of cells at the blastoderm margin with respect to the dorsoventral axis is consistent with the mosaic nature of ectopic expression experiments. This accumulation during gastrulation most likely reflects a cell-autonomous effect of expressing Pitx2a or Pitx2c in these cells, possibly by causing cells to interact with one another rather than migrate to their normal positions.

During late blastula stages in zebrafish, presumptive mesendodermal cells in the dorsal marginal zone flatten and maximize cellular contact with one another (Warga and Nusslein-Volhard, 1999), likely occurring through an increase in cellular adhesion among these cells. This cohesive behavior is disrupted in spt mutant embryos (Warga and Nusslein-Volhard, 1998). In addition, paraxial protocadherin (papc), encoding a cell adhesion molecule, is a downstream target of spt (Yamamoto et al., 1998). Consistent with a role for Pitx2 isoforms in mediating changes in cellular adhesion, our data show that spt regulates the early dorsal-marginal expression of pitx2c (Fig. 3). pitx2 may have an intermediate role between spt and changes in cell behavior mediated through papc and other cell adhesion molecules. Similar alterations in cell behavior could elicit morphological asymmetries during left-right development.

The observation that ectopic Pitx2a and Pitx2c both led to the accumulation of cells during gastrulation implies that these isoforms have some similar regulatory functions (Fig. 4). The phenotypes observed at 28 hpf after injection of either pitx2a or pitx2c RNA also support this conclusion, and the loss of notochord and anterior structures could be a direct result of a failure of axial mesendodermal cells to migrate to their normal positions during gastrulation. However, Pitx2a and Pitx2c also appear to have distinct functions in regulating cyc, lft2 and gsc (Fig. 4). The observation of both similar and opposing effects of ectopic Pitx2a and Pitx2c suggests that the ability of pitx2 isoforms to differentially affect gene transcription may be dependent on the context of the promoters in these target genes.

The maintenance of pitx2c expression during late gastrulation by sur and spt defines anterior and posterior domains in the prechordal plate mesendoderm (Figs 3, 8A). The absence of pitx2c expression during gastrulation in the posterior prechordal plate in sur mutant embryos correlates with the later observation of posterior cyclopia (Brand et al., 1996) and indicates that pitx2c expression is regulated differently in the anterior versus the posterior regions of the prechordal plate. Similar to sur mutant embryos, spt mutant embryos also lack pitx2c expression in the posterior prechordal plate. In contrast, spt mutants form normal notochord and head structures (Warga and Nusslein-Volhard, 1998), suggesting that spt and pitx2c expression in the posterior prechordal plate are not required for either head or midline development. spt and sur mutant embryos also have defects of the left-right patterning of the heart (Chen et al., 1997). The loss of pitx2c expression in the posterior prechordal plate of spt and sur mutant embryos may relate to a specific role for pitx2c and the posterior prechordal plate in left-right axis determination or maintenance.

During late gastrulation, maintenance of pitx2c expression in the polster and posterior prechordal plate requires the activities of cyc and oep (Figs 3, 8A). The alteration of pitx2c expression in cyc and oep mutants could reflect a secondary effect, due to either an absence of cells in the prechordal plate mesendoderm or a change in cell fate. In cyc mutants at least some cells maintain a prechordal plate identity; both gsc and forkhead-domain 2 (fkd2) expression are present although reduced at 90% epiboly (Thisse et al., 1994; Warga and Nusslein-Volhard, 1999). The almost complete lack of pitx2c expression in cyc mutant embryos at 90% epiboly suggests that pitx2c expression is more directly regulated by Cyc signaling than gsc or fkd2 expression and may be a proximal downstream target. While oep mutants have a severe reduction in anterior mesendoderm, which may account for the loss of pitx2c expression in this mutant at late gastrula stages, the earlier loss of pitx2c expression in oep mutants suggests more direct regulation of pitx2c by Oep. Based on our mutant analysis, we suggest that Cyc and Oep signaling function together in the regulation of pitx2c expression.

The pitx2c isoform acts in a feedback loop with cyc and lft2. In contrast to Pitx2a, ectopic Pitx2c can upregulate cyc and lft2 expression during mesendoderm development (Figs 4, 8). Differences also exist between the ability of Pitx2a and Pitx2c to regulate each other and gsc (Figs 4, 5). Thus, pitx2a and pitx2c have divergent activities, which may be reflected in their different expression domains during both mesendoderm and left-right development.

The distinct patterns of pitx2 isoform expression during asymmetric organ development imply the left-sided identity of different organ primordia is regulated by distinct mechanisms. In early zebrafish embryos, a balance between Cyc and Lft activities appears to regulate the expression of pitx2c, and pitx2c itself can feed back to induce both cyc and lft2 expression. A similar mechanism may regulate pitx2c expression in the dorsal diencephalon. lft1 (Bisgrove et al., 1999; Thisse and Thisse, 1999), lft2 (Bisgrove et al., 1999) and cyc (Rebagliati et al., 1998a) are also expressed at late somite stages in the left dorsal diencephalon and, together with pitx2c, may maintain each others’ expression in this tissue. However, in the heart and gut primordia a different mechanism exists to induce pitx2a and pitx2c.

pitx2a asymmetric expression does not entirely overlap with lft2 in the left heart primordium (Fig. 6), while cyc appears to be broadly expressed in the left lateral plate mesoderm (Rebagliati et al., 1998a), suggesting that different mechanisms may regulate lft2, cyc and pitx2 isoforms on the left side of the embryo. Furthermore, in early embryos, ectopic Pitx2a does not positively regulate either lft2 or cyc expression and may function distinctly from Cyc and Lft2 later in the heart field. While both pitx2 isoforms were equivalent in their ability to affect laterality in the gut in Xenopus embryos, Pitx2a had a stronger effect on heart asymmetry than Pitx2c, correlating with the normal expression of pitx2a in the zebrafish heart field. In addition, lft1 and lft2 are not expressed in the presumptive gut (Bisgrove et al., 1999), while pitx2c (this paper) and cyc (Rebagliati et al., 1998a) expression appear to overlap in this region. Based on their expression domains and our results in early zebrafish embryos, we hypothesize a regulatory loop between cyc and pitx2c in the gut primordium.

Our results indicate that overexpression of both Pitx2a and Pitx2c by DNA injection can perturb laterality in the developing heart and gut in Xenopus. However, Pitx2a has a stronger effect on the heart than Pitx2c (Table 3). Similar laterality phenotypes in Xenopus were previously observed as a result of targeted injection of Xenopus Pitx2 RNA on the right side of the embryo (Campione et al., 1999; Ryan et al., 1998). However, these reports did not distinguish between different pitx2 isoforms. Additionally, we found that pitx2 isoforms affect left-right development when injected on either the left or right side of the embryo. As expected, right-sided injections, which would place pitx2 expression ectopically on the right side of the embryo during asymmetric organ development, resulted in a higher frequency of laterality disturbances than left side injections, which serve to overexpress pitx2 on its normal side of expression. Ectopic expression of pitx2 on the right side or increased expression of pitx2 on the left side may also affect left-right development at an earlier developmental stage. Our observation that pitx2 isoforms can alter the expression of other genes implicated in earlier steps of left-right development in zebrafish, including cyc (Chen et al., 1996) and ntl (Danos and Yost, 1996), was surprising. These early disturbances may account for a certain frequency of Xenopus embryos exhibiting perturbed left-right morphogenesis following unilateral pitx2 injection.

Ectopic expression of either pitx2 isoform during early zebrafish development disrupted cell movements during gastrulation, possibly resulting from increased cell adhesion among blastomeres. This raises the intriguing possibility that pitx2 isoforms may function during asymmetric organ development in a similar manner. Although the cellular mechanisms that are downstream of pitx2 are unknown, pitx2 isoforms may function in laterality decisions by regulating cell adhesion and cell migration differences between the left and right sides of organ primordia.

We thank Brent W. Bisgrove, Maura McGrail and Molly Wagner for their critical reading of and thoughtful comments on the manuscript. We also thank Steve Bleyl for providing the pCS2+/gsc construct. We are grateful to Monica Biran for her expert technical assistance with zebrafish, Sharon E. Johnson and David J. Grunwald for their advice and assistance, and Dan Weeks for suggesting the use of the Troponin T antibody. We would also like to thank the MGH/Harvard and Tubingen zebrafish groups for kindly providing mutant fish stocks. This research was supported by funds from the Huntsman Cancer Institute, a grant from NIH/HLBI, and an American Heart Association Established Investigatorship to H.J.Y.