INTRODUCTION

The Iroquois (Iro) genes were discovered in Drosophila for their role in formation of sensory bristles in the dorsal mesothorax or notum of the fly (Dambly-Chaudiere and Leyns, 1992; Leyns et al., 1996). Further studies showed that the Iro locus encodes factors essential for the expression of proneural genes in the achaete-scute complex that are necessary for determination of sensory organ precursors (Gomez-Skarmeta et al., 1996). Drosophila has three Iro genes, araucan (ara), caupolican (caup) and mirror (mirr), and together they form the Iroquois complex (Iro-C). Molecular characterization of the Iro genes in Drosophila has allowed the identification of homologs in C. elegans and several vertebrates, including Xenopus, mouse, zebrafish, chick and human (Bao et al., 1999; Bellefroid et al., 1998; Bosse et al., 2000; Bosse et al., 1997; Christoffels et al., 2000; Cohen et al., 2000; Funayama et al., 1999; Gomez-Skarmeta et al., 1998; Goriely et al., 1999; Kudoh and Dawid, 2001; Ogura et al., 2001; Peters et al., 2000; Tan et al., 1999; Wang et al., 2001). The Iro genes encode proteins that show a strong similarity in their homeodomain and all contain a characteristic motif named the Iro box. Based on these features, the Iro products constitute a unique class of proteins within the TALE super-class of atypical homeodomain proteins (Burglin, 1997).

Analysis of Iro function in many developmental contexts and different model systems has now defined a broader role for these genes during development. During early development, these genes appear to have a role in defining the identity of large territories. In Drosophila their early expression defines dorsal eye, head and mesothorax territories (Cavodeassi et al., 1999; Cavodeassi et al., 2000; Diez del Corral et al., 1999). Later, the Iro genes have a role in the subdivision of such large territories into subdomains. For example, while they have an early role in defining the entire notum, later the Iro genes are required to specify the identity of the lateral notum where they are essential for expression of proneural genes and sensory bristle formation (Diez del Corral et al., 1999).

Analysis of Iro function in Xenopus has shown that these genes have similar roles in vertebrate development. The early expression of Xiro1 and Xiro2 in the dorsal ectoderm at the beginning of gastrulation and the effects of ectopic expression of Xiro1 are consistent with an early role in establishment of neural fate in a large territory of the dorsal ectoderm (Gomez-Skarmeta et al., 2001). At a later stage, expression of Xiro1 and Xiro2 becomes restricted to two stripes within the neural plate that extend caudally from midbrain-hindbrain boundary (MHB). Expression of Xiro1 and Xiro2 in this restricted domain along with Xiro3 suggests a late role in determining the expression of the proneural gene Xenopus achaete-scute homolog 3 (XASH3) in a specific subdomain of the neural tube where neuronal precursors may be generated (Bellefroid et al., 1998; Gomez-Skarmeta et al., 1998). These studies suggest that vertebrate Iro genes function to establish cell fate in the neural plate in a manner that is similar to Drosophila. However, loss-of-function studies have not defined how Iro genes contribute to development of specific territories in the neurectoderm during development.

In this study, we have examined roles of two zebrafish Iro genes: iro1 and a novel Iro family member, iro7. We focus on two related issues, the role of these genes in neurogenesis and their role in determining the development of a large territory in the neurectoderm. We characterized their ability to induce expression of the proneural gene, neurogenin1 (ngn1) and examined how a knockdown of these genes affects the development of tissues within an anteroposterior compartment defined by their early expression. Finally, by exploiting repressor or activator fusions, we determine how these homeodomain proteins affect transcription of target genes.

MATERIALS AND METHODS

Zebrafish maintenance and mutants

Zebrafish were raised and maintained under standard conditions. To collect maternal zygotic headless^m881 mutant embryos, heterozygous males and homozygous females were crossed (Kim et al., 2000). In this study, we also used no isthmus (noi^tu29a) and acerebellar (ace^ti282a) mutants (Brand et al., 1996; Lun and Brand, 1998; Reifers et al., 1998).

Identification of iro1 and iro7

iro1 was cloned during a random in situ based screen. iro7 was initially identified as an EST (fc24a10) as an unknown Iro family gene. To obtain the 5′ region of iro7, 5′ RACE was performed using a tailbud cDNA library made by the SMART RACE cDNA amplification kit (Clontech). Sequences were deposited in GenBank under Accession Numbers AF414133 (iro7) and AF414134 (iro1). Sequence alignment was analyzed by J. Hein’s method with PAM250 residue weight table using DNASTAR software. iro7 was mapped on the LN54 radiation hybrid panel (Hukriede et al., 1999) using the primers 5′-AAATCTGACGAGGAGGATGAGGAAGAAGAG-3′ and 5′-TTCATTGACTTTGTTTGAGAAGGTCGTGTG-3′.

Whole mount in situ hybridization, antibody or β-galactosidase staining

For iro1, full-length cDNA was used as a template for making RNA probe (XhoI/T7). For iro7, the 3′ region of a cDNA containing approximately 750 bp was used for making RNA probe (SalI/SP6). Zebrafish gbx1 was found by EST search (fj77a06) and the coding fragment was subcloned into pCRIITOPO for RNA probe synthesis (NotI/SP6). Other plasmids that have been used to make in situ probes have been published previously: otx2 (Li et al., 1994; Mori et al., 1994), pax2.1 (pax2a – Zebrafish Information Network) (Krauss et al., 1991), hoxb1b (Alexandre et al., 1996), ngn1 (neurod3 – Zebrafish Information Network) (Blader et al., 1997; Kim et al., 1997), fkd6 (foxd3 – Zebrafish Information Network) (Odenthal and Nusslein-Volhard, 1998), krox20 (egr2 – Zebrafish Information Network) (Oxtoby and Jowett, 1993), huC (elavl3 – Zebrafish Information Network) (Good, 1995; Kim et al., 1996) and gata2 (Detrich et al., 1995). Double in situ using digoxigenin- and fluorescein-labeled RNA probes and antibody staining were performed as described (Itoh and Chitnis, 2001; Jowett, 2001). To detect β-galactosidase activity, embryos co-injected with various synthesized mRNA, were fixed in 4% paraformaldehyde overnight at 4°C and stained by either X-gal or salmon-β-D-galactoside (Biosynth).

Constructs

iro1 and iro7 cDNA fragments encoding full-length protein were subcloned into the pCS2+ vector. To generate En-iro7HD and VP16-iro7HD, we amplified a fragment by PCR with primers 5′-CCGCTCGAGCCGTATCACCAAGCTCTCCTCGGA-3′ and 5′-GCTCTAGATTTTCCTTTGGACGCCCAGCT-3′. The amplified fragment was digested with XhoI and XbaI, and subcloned into pCS2-En or pCS2-VP16 (Kawahara et al., 2000). To make ΔN-iro1 and ΔN-iro7 constructs that lacked the morpholino antisense oligo (MO1 or MO7)-binding site, fragments were amplified by PCR using primers: 5′-CGGGATCCATGGAGGGAAGCTCGGACAACAGCGCA-3′ and 5′-GCTCTAGAAGAAATTGTCTTCAAAGCGCGTTGTG-3′ for the ΔN-iro1 construct, 5′-CGGGATCCAACTTCTTCATGGACAGAAACATCAACATG-3′ and 5′-CGTCTAGAAGTTGACTTTGTTTGAGAAGGTCGTGTGT-3′ for the ΔN-iro7 construct, and they were subcloned in the BamHI/XbaI sites of the pCS2+ vector.

mRNA and morpholino antisense oligo injection

For microinjection of mRNA, constructs were linearized and transcribed with SP6 RNA polymerase using the mMessage mMachine Kit (Ambion). For injection of wild type iro1, iro7, En-iro7HD and VP-iro7HD mRNA, we injected those mRNA into embryos at the16-64 cells stage to prevent gastrulation defects.

Morpholinos (Gene Tools) were resuspended in DEPC water and stored at –20°C. The sequences of the morpholinos used were 5′-GCGTGGAGAGGACGGCATTACACCC-3′ for iro1 and 5′-GCAAACCCCGTTGATGAAGCAGGCA-3′ for iro7. The oligos were injected into one- to two-cell stage embryos.

In vitro translation

Iro1 and Iro7 protein were synthesized in the presence or absence of morpholino for iro1 or iro7 using TNT coupled reticulocyte lysate systems (Promega). Proteins were made from CS2+ iro1 and CS2+ iro7 plasmid (0.5 μg each) and labeled with [³⁵S] methionine. After the translation reaction was complete, reaction mixtures were subject to SDS-PAGE. The dried gel was exposed to X-ray film.

RESULTS

Cloning of zebrafish iro1 and iro7

We identified two zebrafish Iro genes, iro1 and iro7 in a zebrafish EST database and in an in situ-based screen for genes with interesting expression patterns (Kudoh et al., 2001). A full-length 1.9 kb cDNA encoding 419 amino acids of iro1 was obtained from the plasmid library used for the in situ-based screen. 5′ RACE was performed with a tailbud stage library to obtain the full coding sequence for iro7. This yielded a 1.3 kb cDNA encoding 314 amino acids of iro7.

The two uncharacterized Iro genes were identified as iro1 and iro7 based on a comparison of their sequences with previously identified members of the Iro family. Comparison of iro1 with other members of family indicates that iro1 is the Irx1 ortholog with overall amino acid similarity of 47.6% and 44.0%, to Xiro1 and mouse Irx1, respectively. Zebrafish iro1 was independently characterized by another group that came to the same conclusion (Wang et al., 2001). However, iro7 has a sequence that is very divergent from the six previously described Irx orthologs and so has been designated as iro7. It has also recently been independently characterized by Lecaudy et al. (Lecaudy et al., 2001).

iro7 may be a novel paralogue of iro1 and iro3 in zebrafish

Analysis of the human and mouse genome has suggested that there are a total of six Iroquois (Irx) family members in mammals (Ogura et al., 2001; Peters et al., 2000). The six murine genes are in two tightly linked complexes with three genes in each cluster: Irx1, Irx2 and Irx4 are on chromosome 13 and form the IrxA cluster, while their respective paralogues Irx3, Irx5 and Irx6, are on chromosome 8 and form the IrxB cluster (Peters et al., 2000).

Phylogenetic analysis of the six murine Irx genes was facilitated by comparing their amino acid sequence in two relatively conserved domains, the N-terminal and homeodomain region [Fig. 1A (shown in blue), Fig. 1B] and the Iro box domain [Fig. 1A (shown in green), Fig. 1C]. We compared these two regions in the six murine Irx genes with corresponding domains of zebrafish iro1, iro3, iro5 and iro7 (Fig. 1B,C). We found that zebrafish iro1, iro3 and iro5 have most similarity (Fig. 1D, shown in red) with their mouse orthologs Irx1, Irx3 and Irx5, while the next best similarity (Fig. 1D, shown in blue) is with mouse paralogs Irx3, Irx1 and Irx2, respectively. However, iro7 is less similar to other members except in the homeodomain (Fig. 1B, broken line) and Iro box (Fig. 1C).

In zebrafish, map locations of iro1, iro3 and iro5 are consistent with the genomic organization described in mouse. Zebrafish iro1 maps to Linkage Group (LG) 19, while zebrafish iro3 and iro5 map together on LG7, suggesting that, like Irx3 and Irx5 in mouse, the later are members of one cluster (Wang et al., 2001). Interestingly, iro7 does not map to either LG19 or to LG7, but to LG23, close to Z5526 on the LN54 radiation hybrid panel where it is not clear if iro7 is a part of an additional Iro gene cluster or whether it reflects a break up of an extant cluster that might have occurred during teleost evolution.

Expression patterns of iro1 and iro7

Expression of iro1 and iro7 begins around the dome stage. Whole-mount in situ hybridization shows that iro1 is almost undetectable at the dome stage, while in some embryos iro7 is widely expressed at low levels (Fig. 2A,J). By the shield stage, their expression becomes more clearly defined and both genes are expressed in a similar pattern in two distinct domains of the embryo. They are expressed in the dorsal epiblast where their expression includes the prospective neurectoderm (Lecaudey et al., 2001; Wang et al., 2001) and adjacent to the lateral margin in the hypoblast where their expression is excluded from the shield (Fig. 2B,K).

At 75% epiboly, although expression of iro1 and iro7 is transiently retained at the anterior edge of the neurectoderm (data not shown), it is lost from much of the rostral neurectoderm and it becomes prominent in the prospective midbrain-hindbrain region (Fig. 2C,L). The limits of this expression domain were defined by comparison with genes expressed in various compartments of the forebrain, midbrain and hindbrain at 80% epiboly. At this stage the domain of iro1 and iro7 expression overlaps at its rostral edge with the most caudal expression of otx2, a marker of prospective forebrain and midbrain. pax2.1, a marker of the prospective midbrain-hindbrain boundary (MHB) is expressed within the expression domains of both Iro genes (Fig. 3A,B,E,F) and their expression overlaps with gbx1 in the prospective hindbrain beginning at rhombomere 1 (Fig. 3C,G). The caudal limit of the iro1 and iro7 expression is defined by hoxb1b, whose expression identifies neurectoderm caudal to rhombomere 4 (McClintock et al., 2001). Its expression abuts iro1 and overlaps slightly with iro7, whose expression extends slightly more caudally than iro1 at this stage (Fig. 3D,H). These comparisons show that by 80% epiboly, expression of iro1 and iro7 defines a compartment of the neurectoderm that extends from the midbrain to rhombomere 4 in the hindbrain.

By tailbud stage, differences in expression of the two genes become more apparent. While both iro1 and iro7 continue to be expressed in the prospective midbrain-hindbrain region, iro1 expression expands at its lateral margins caudally into domains where peripheral ganglia will develop; in addition, its expression begins in the caudal neurectoderm (Fig. 2D,M). By early somitogenesis, a gap in expression at the midbrain-hindbrain boundary (MHB) splits midbrain-hindbrain expression of iro1 and iro7 into two subdomains (Fig. 2E-G,N-P). Both genes are expressed in a subdomain rostral to pax2.1 in the midbrain. Expression of iro1 just caudal to pax2.1 identifies a caudal subdomain in rhombomere 1 (Fig. 3K,O). Comparison with krox20, which is expressed in rhombomeres 3 and 5, reveals a caudal subdomain of iro7 expression in rhombomeres 3 and 4 (Fig. 3P), and expression of iro1 in the caudal neurectoderm beginning at rhombomere 5 (Fig. 3L). As somitogenesis continues, expression of iro1 remains over a broad area as its expression extends into the developing caudal neural tube, while iro7 becomes progressively restricted (Fig. 2F-I,O-R).

Consistent with a role for Iro genes in controlling the expression of vertebrate proneural genes, iro1 and iro7 are expressed in partially overlapping patterns that cover many domains of ngn1 expression in the neurectoderm at the tailbud stage (Fig. 3I,M). iro1 and iro7 expression extends laterally outside the neurectoderm to include domains of ngn1 expression in developing trigeminal placodes (Fig. 3I,M, arrowhead). Just medial to this domain, expression of iro1 and iro7 overlaps with expression of fkd6, a marker of premigratory neural crest cells (Odenthal and Nusslein-Volhard, 1998) (Fig. 3J,N). In the caudal neurectoderm, expression of iro1, but not iro7, overlaps with ngn1, where this proneural gene defines longitudinal proneuronal domains where early neurons differentiate (Fig. 3I).

The size of the iro1 and iro7 expression domain is expanded by exaggerated Wnt signaling

iro1 and iro7 are expressed in a caudal compartment of the anterior neurectoderm that includes the prospective MHB domain, the adjacent neural crest and trigeminal neurons. In maternal zygotic (MZ) hdl mutants, failure to repress Wnt target genes adequately in the anterior neurectoderm leads to exaggerated Wnt signaling and a rostral expansion of the trigeminal neurons, the MHB domain and cranial neural crest, identified by expression of ngn1, pax2.1 and fkd6, respectively (Kim et al., 2000) (Fig. 4C,D). Examination of MZ hdl mutants reveals expansion of the iro1 and iro7 expression domain (Fig. 4A, B), which becomes evident by 75% epiboly (data not shown). It is likely that this expansion predominantly reflects a shift in the rostral boundary of iro1 and iro7 expression, as the expression of hindbrain markers krox20 (Fig. 4D) and gbx1 (data not shown) is not appreciably altered in hdl mutants. A similar expansion of iro1 and iro7 expression was observed in masterblind (mbl) mutants (data not shown) that are also characterized by exaggerated Wnt signaling, in this case due to a mutation in axin, which normally promotes degradation of β-catenin, an effector of Wnt signaling (Heisenberg et al., 2001). These observations suggest that the size of the iro1 and iro7 expression domain is determined by the level of Wnt signaling in the anterior neurectoderm.

iro7 is essential for determination of trigeminal neurons

Embryos were injected with antisense morpholino oligos complementary to the N-terminal of iro1 (MO1) and iro7 (MO7) to examine the function of these Iro genes. To demonstrate that both MO1 and MO7 were effective at knocking-down translation of respective Iro transcripts, an in vitro translation assay was done to compare translation of iro1 and iro7 from plasmids (mixture of 0.5 μg each) in the presence of different amounts of the morpholinos. Gels comparing expression of ³⁵S-labeled Iro1 and Iro7 in the presence of increasing concentrations (from 12.5 ng to 12.5 μg) of either MO1 or MO7 showed that each morpholino specifically inhibits translation of the targeted Iro gene in a dose-dependent manner (Fig. 5J).

Examination of ngn1 expression in injected embryos at the one-somite stage revealed that 10 ng MO1 has no effect on expression of ngn1 but 10 ng MO7, directed against iro7, leads to a loss of ngn1 expression in the developing trigeminal ganglia (95%, n=22) (Fig. 5A-D). MO7 also affected expression of ngn1 in developing neurons within the neural plate, including those in rhombomeres 2 and 4; however, the effects on CNS neurons have not yet been fully examined. To show that effects of MO7 are specifically due to its ability to knockdown iro7 function, MO7 was injected with mRNA encoding an engineered form of iro7 (ΔNiro7) in which nine N-terminal amino acids were deleted to prevent hybridization with MO7. Embryos were first injected with MO7 at the one-cell stage so that both sides received the morpholino, and then in addition, one side was injected at the two-cell stage with ΔNiro7 and β-galactosidase mRNA (to detect distribution of ΔNiro7transcripts). Embryos injected in such a manner revealed that co-expression of ΔNiro7 with MO7 led to a recovery of ngn1 expression in the trigeminal placode (Fig. 5D). To show that reduced ngn1 expression in MO7-injected embryos prevents formation of trigeminal ganglia, MO1- and MO7-injected embryos were examined at 24 hpf with an antibody to acetylated α-tubulin that identifies differentiating neurons. Only MO7-injected embryos showed a loss of neurons in the trigeminal ganglia (83%, n=12) confirming that iro7 has an essential role in determination of these neurons (Fig. 5E-G).

iro7 can induce expression of ngn1 as either an activator or repressor

As a morpholino directed against iro7 reduced expression of ngn1 in the trigeminal ganglia, we examined if ectopic expression of iro1 and iro7 promotes expression of this proneural gene. Initially it was difficult to interpret the effects of ectopic iro1 and iro7 expression because widespread expression of those mRNA caused severe gastrulation defects. This problem was overcome by injecting single cells relatively late in development to restrict the domain of ectopic expression. Embryos injected with 50 pg iro1 or iro7 mRNA at the 16-to 64-cell stage and assayed at the tailbud stage revealed that both Iro genes could induce expression of ngn1 in the neurectoderm ectoderm (Fig. 5D, data not shown) and the ventral ectoderm where this proneural gene is normally not expressed (iro1: 95%, n=19; iro7: 100%, n=24) (Fig. 5L,M).

To determine if iro1 and iro7 induce ngn1 expression by acting as activators or repressors, we made engineered forms of iro7, expected to exclusively repress or activate Iro target genes. Plasmids encoding chimeric transcription factors (En-iro7HD and VP-iro7HD) were made by combining domains encoding the repressor domain of Engrailed (En) or the activation domain of VP16 (VP) with a fragment of the iro7 homeodomain (iro7-HD) (Fig. 5K) (Conlon et al., 1996; Kessler, 1997). Though the chimeric constructs contained the iro7 homeodomain, it was expected that the fusion proteins would bind target sequences for other Iro family members because of over 86% similarity between Iro genes in the homeodomain.

The effects of En-iro7HD and VP-iro7HD on ngn1 expression were surprising: both repressor and activator forms of iro7 were capable of inducing ectopic ngn1 expression and the pattern of ectopic ngn1 induced was unique in each case. Like iro1 and iro7, 50 pg of En-iro7HD mRNA effectively induced ngn1 expression in the ventral ectoderm (96%, n=27) (Fig. 5N). However, in contrast to the widespread or patchy expression induced by iro1 and iro7, respectively (Fig. 5L,M), ngn1 expression induced by En-iro7HD in the ventral ectoderm was typically in a discrete salt-and-pepper pattern (Fig. 5N). Injection of 50 pg of VP-iro7HD mRNA had very different effects. It was ineffective at inducing ectopic ngn1 in the ventral ectoderm but was able to induce widespread ectopic ngn1 expression within the neurectoderm (100%, n=24) (Fig. 5O). These data suggest that both activator and repressor forms of iro7 can induce ectopic ngn1 but they may achieve this by slightly different mechanisms in the ventral and dorsal ectoderm.

Previous studies have shown that Xiro1 can function as a repressor to inhibit BMP expression and neuralize the ectoderm (Gomez-Skarmeta et al., 2001). One possibility is that iro1, iro7 and the repressor En-iro7HD induce ngn1 expression in the ventral ectoderm by inhibiting BMP signaling and neuralizing the ectoderm. Consistent with this possibility iro1, iro7 and En-iro7HD can inhibit expression of a BMP target gene, gata2 in the ventral ectoderm (iro1: 85%, n=39; iro7: 93%, n=30; En-iro7HD: 77%, n=30) (Fig. 5 P-S).

iro1 and iro7 are necessary but not sufficient for determination of neural crest fate

The expression of iro1 and iro7 in a compartment that defines where trigeminal neurons, the neural crest and the MHB domain are located raised the possibility that these Iro genes not only have a role in determination of trigeminal neurons but they also regulate development of adjacent tissues in this compartment. To test this hypothesis, we examined the effects of iro1 and iro7 morpholinos on development of the cranial neural crest and the MHB domain.

Embryos injected with 10 ng MO1 showed a small decrease in fkd6 expression, while injection of 10 ng MO7 resulted in little change; however, when 5 ng of MO1 and MO7 each were simultaneously injected, there was a clear reduction of fkd6 expression in the neural crest (88%, n=25) (Fig. 6B-D). The specificity of this effect was revealed by the observation that the morpholinos never affected axial expression of fkd6 (Fig. 6D). In addition, the reduction in fkd6 expression induced by MO1 and MO7 was suppressed by co-injection of modified iro1 and iro7 transcripts (ΔNiro1/7) that lacked the N terminal sequences that the morpholinos were targeted against (Fig. 6E, arrowhead). These observations suggest that the overlapping expression of iro1 and iro7 determines the fate of fkd6-expressing neural crest cells in a partially redundant manner. Embryos injected with 50 pg of wild-type iro1 or iro7 mRNA, however, did not show much ectopic fkd6 expression suggesting that while expression of iro1 and iro7 is necessary for fkd6 expression, it is not sufficient to induce its expression (Fig. 6H,I). Activator and repressor forms of iro7 revealed that they have opposite effects on neural crest formation: 50 pg of En-iro7HD mRNA induced a small increase in fkd6 expression (Fig. 6J, arrowhead), while 50 pg of VP-iro7HD mRNA reduced expression of this neural crest marker (82%, n=33) (Fig. 6K). These observations suggest iro1 and iro7 act as repressors to determine formation of the cranial neural crest.

iro1 and iro7 are essential for formation of the Midbrain-Hindbrain boundary and establishment of the isthmic organizer

To examine the role of iro1 and iro7 in establishment of the MHB domain and function of the isthmic organizer, we examined the effects of morpholinos, MO1 and MO7, on expression of genes that identify this domain at 24 hpf. Injection of 10 ng of MO1 or MO7 alone had subtle effects on expression of eng2, pax2.1, fgf8 and wnt1, while injection of 5 ng each of MO1 and MO7 resulted in significant reduction or absence of their expression at the MHB domain (eng2: 69%, n=13; pax2.1: 87%, n=15; fgf8: 82%, n=11; wnt1: 92%, n=12) (Fig. 7A-D). Injection of MO1 alone also resulted in some decrease in eng2 expression (Fig. 7A). Co-injection of MO1 and MO7 not only resulted in a loss of pax2.1, fgf8 and wnt1 expression in the MHB domain it also altered their expression in the forebrain and hindbrain. The morpholinos expanded fgf8 expression in the forebrain and made pax2.1 expression in the forebrain and otic vesicles more prominent (Fig. 7B,C). In part, the effects on pax2.1 expression in the forebrain and otic vesicles may be accounted for by the slight delay in development caused by the injection of morpholinos, as at a slightly earlier stage of development expression of pax2.1 is normally prominent in these domains. MO1 and MO7 injection also disrupted segmental wnt1 expression in rhombomeres (Fig. 7D).

To address how early iro1 and iro7 work together in MHB formation, we examined embryos at the tailbud stage. We found injection of MO1 and MO7 together leads to a specific reduction of pax2.1 expression (67%, n=12) (Fig. 7E) without causing a change in wnt1 and fgf8 expression (data not shown). The reduction in pax2.1 expression induced by MO1 and MO7 was suppressed by co-injection of ΔNiro1/7 (Fig. 7G, arrowhead). However, injection of 50 pg iro1 mRNA resulted in ectopic expression of both pax2.1 and fgf8 at the tailbud stage (pax2.1: 100%, n=14; fgf8: 95%, n=19), while it had little effect on wnt1 expression at this stage (Fig. 7H). Ectopic expression of 50 pg iro7 mRNA also induced fgf8 expression (73%, n=26), however, it had no effect on expression of pax2.1 and in some embryos it reduced expression of wnt1 (Fig. 7I).

As with injection of iro1 mRNA, injection of 50 pg of En-iro7HD mRNA initiates ectopic expression of pax2.1 and fgf8, but not wnt1 (pax2.1: 94%, n=17; fgf8: 82%, n=17) (Fig. 7J). By contrast, injection of 50 pg VP-iro7HD mimics the combined effects of the two morpholinos. It prevents formation of the isthmus or the constriction between midbrain and hindbrain (Fig. 7L), and inhibits expression of pax2.1, fgf8 and wnt1, genes that mark the isthmic organizer at 24 hpf (pax2.1: 100%, n=28; fgf8: 89%, n=19; wnt1: 83%, n=6) (Fig. 7K,M and data not shown). These data suggest that iro1 and iro7 are likely to function as repressors in initiating establishment of the MHB domain and the isthmic organizer.

Expansion of the MHB domain and adjacent tissues in hdl mutants is dependent on the function of iro1 and iro7

Expansion of the MHB domain and the adjacent trigeminal ganglia and neural crest correlates with early expansion of iro1 and iro7 in hdl and mbl mutants, and establishment of these tissues appears normally dependent on iro1 and iro7 function (Fig. 4A-E). These observations suggest that expansion of the MHB domain and adjacent tissues is due to the early expansion of iro1 and iro7 gene expression in mutants with exaggerated Wnt signaling. To test this hypothesis, we examined expression of pax2.1 and fkd6 in MZ hdl mutants injected with MO1 and MO7, and ngn1 in mutants injected with MO7 alone. The anterior expanded expression of pax2.1 (100%, n=19), fkd6 (95%, n=21) and ngn1 (100%, n=16) was inhibited in morpholino-injected embryos, supporting the hypothesis that expansion of the MHB domain and adjacent tissues is dependent on expanded expression of iro1 and iro7 in mutants with exaggerated Wnt signaling (Fig. 5H,I, Fig. 6F,G, Fig. 7F).

DISCUSSION

Iro genes and primary neurogenesis

We have identified two zebrafish Iro genes, iro1 and a novel zebrafish Iro family member, iro7, that are expressed during gastrulation in a dorsal compartment of the ectoderm. This compartment includes the prospective MHB domain, the adjacent cranial neural crest and neurons of the trigeminal ganglia. Our initial motivation in this study was to examine the role of these Iro genes in neurogenesis and in formation of the trigeminal neurons. Ectopic expression of both iro1 and iro7 led to ectopic expression of ngn1 and a knockdown of iro7 function led to a clear reduction in ngn1 expression and differentiation of trigeminal ganglia. Together, these observations suggest that iro7 has an essential role in determining the fate of trigeminal neurons and that its ectopic expression accounts in part for the expanded distribution of these neurons in hdl mutants. The knockdown of iro7 also affected distribution of ngn1 in the midbrain-hindbrain region; however, these effects have not been analyzed at this stage. The knockdown of iro1 function had little effect on ngn1 expression in the caudal neurectoderm (future spinal cord). This suggests that iro1 function is redundant in this domain and that other unidentified Iro genes may be able to compensate for its loss.

iro1 and iro7 are essential for development of an anteroposterior territory

An unexpected finding was the observation that a knockdown of iro1 and iro7 function not only affected ngn1 expression in the trigeminal placode it also affected formation of adjacent neural crest cells and the MHB domain. Each of these tissues is a derivative of a different ectodermal compartment, epidermal, neural crest and neural, respectively, whose individual fates are determined by a number of signaling pathways that determine dorsoventral fate including BMP signaling (Chitnis, 1999; Marchant et al., 1998; Nguyen et al., 1998). All three domains, however, are contained within the anteroposterior compartment of the ectoderm where iro1 and iro7 are initially expressed. Together, these observations suggest that iro1 and iro7 are not just involved in regulating neurogenesis but are also essential for normal development of an anteroposterior compartment of the dorsal ectoderm. This conclusion is consistent with the emerging view that Iro genes, both in the fly and vertebrates are required at early stages of development to define large territories (Cavodeassi et al., 2001). However, this is the first loss-of-function study to define how Iro genes contribute to development of a large territory in the ectoderm during early vertebrate development.

Patterning of the neurectoderm by Wnt signaling mediated by Iro genes

Wnt signaling patterns the neurectoderm along the anteroposterior axis (Patapoutian and Reichardt, 2000). During early gastrulation, regulation of Wnt signaling plays an essential role in establishing forebrain, eye, midbrain and MHB territories in the anterior neuroectoderm (Bally-Cuif et al., 1995; Glinka et al., 1998; Heisenberg et al., 2001; Kim et al., 2000). Genes that determine the fate of the most rostral tissues in the anterior neurectoderm are dependent on mechanisms that repress Wnt signaling, while genes expressed in relatively caudal domains are dependent on relatively high levels of Wnt signaling. Ineffective repression of Wnt target genes in hdl mutants or reduced destruction of a Wnt effector in mbl mutants leads to exaggerated Wnt signaling in the anterior neurectoderm (Heisenberg et al., 2001; Kim et al., 2000). Increased Wnt signaling in hdl and mbl mutants is accompanied by a rostral expansion of iro1 and iro7 expression, suggesting that, as recently reported for Xiro1 in Xenopus (Gomez-Skarmeta et al., 2001), Wnt signals regulate the size of the territory where these Iro genes are expressed. Loss-of-function studies in wild-type and hdl mutant backgrounds suggest that the territory of iro1 and iro7 expression not only defines the region within which trigeminal neurons, neural crest and the MHB domain are formed but the function of these Iro genes is essential for the development of these tissues. These observations suggest that Wnt signaling defines the identity of a caudal compartment of the anterior neurectoderm through the function of iro1 and iro7.

iro1 and iro7 are essential for establishment of the isthmic organizer

The isthmus is a specialized tissue with secondary organizer properties formed at the boundary between the midbrain and hindbrain. It eventually becomes the source of Wnt and FGF signals, and is essential for normal anteroposterior patterning of the adjacent midbrain and anterior hindbrain (Rhinn and Brand, 2001; Wurst and Bally-Cuif, 2001). Inhibition of iro1 and iro7 function with morpholinos leads to loss of the isthmus and patterning defects that suggest iro1 and iro7 have an essential role in establishing a functional isthmic organizer.

Interactions between pax2.1, wnt1 and fgf8 play an early role in establishing and maintaining the isthmic organizer at the boundary of gbx2 and otx2 expression domains (Bally-Cuif et al., 1995; Lun and Brand, 1998; Reifers et al., 1998; Schwarz et al., 1997; Wurst and Bally-Cuif, 2001). While it is not yet clear how iro1 and iro7 regulate formation of the isthmic organizer, our data suggests that they have a relatively early role, as they are expressed in the midbrain-hindbrain domain before otx2 and gbx1. They may also function by a mechanism that is independent of otx2 and gbx1, because rostral expansion of Iro genes in hdl mutants is not accompanied by any noticeable change in otx2 and gbx1 expression (data not shown).

In contrast, reduced iro1 and iro7 function results in loss of pax2.1, the earliest marker described so far for the MHB domain in zebrafish. Furthermore, while ectopic expression of iro1 induces expression of pax2.1 and fgf8, expression of iro1 and iro7 remains unaffected in mutants where the function of pax2.1 (noi) and fgf8 (ace) is lost (data not shown). Together, these observations suggest a relatively early role for Iro genes in initiation of pax2.1 expression and establishment of the MHB domain. While iro1 and iro7 are required to initiate pax2.1 expression, they are not required for the initial expression of wnt1 or fgf8. This suggests that wnt1 and fgf8 expression in the MHB may be established independently by alternative pathways, as has been suggested by previous studies (Lun and Brand, 1998; Rhinn and Brand, 2001; Wurst and Bally-Cuif, 2001).

Does iro7 act as a repressor or an activator?

Iroquois homeoproteins have been suggested to act as either activators or repressors in different experimental model systems and developmental contexts. In Drosophila neurogenesis, ara and caup can bind to the promoter of the achaete-scute proneural genes and function as activators (Gomez-Skarmeta et al., 1996). However, Xiro1 functions as a repressor to inhibit BMP expression and neuralize the ectoderm (Gomez-Skarmeta et al., 2001). Our study suggests that in the context of neural crest and MHB formation, iro1 and iro7 are likely to function as repressors, because formation of these structures is inhibited by the combination of morpholinos and by the VP-iro7HD fusion.

In the context of neurogenesis and induction of ngn1 expression, the interpretation is complicated by the observation that both the activator and repressor forms of iro7 induce expression of ngn1. However, the repressor form is more effective at inducing ngn1 expression in the ventral ectoderm, while the activator form only induces widespread ngn1 expression dorsally in the neurectoderm. Furthermore, the repressor form of iro7 induces ngn1 expression in a salt-and-pepper pattern, while iro1 and VP-iro7HD induce ngn1 in a relatively broad domain. One potential explanation for these differences is that Iro genes can act as both activators and repressors and induce ngn1 expression by different mechanisms in the ventral and dorsal ectoderm. As repressors, Iro genes may indirectly induce ngn1 expression by inhibiting expression of BMPs and neuralizing the ectoderm. This possibility is supported by the ability of iro1, iro7 and En-iro7HD to suppress expression of gata2, a BMP target gene. ngn1 induced in such a manner might more easily be regulated be lateral inhibition and eventually acquire a salt-and-pepper pattern (Ma et al., 1996). However, in the dorsal ectoderm, which is already neuralized, Iro genes may function as activators to directly induce expression of ngn1 in a much broader domain. Such a scenario would explain why En-iro7HD induces ngn1 expression a salt-and-pepper pattern in the ventral ectoderm and VP-iro7HD induces ngn1 in a broad domain in the neurectoderm. Wild-type iro1 may function both as a repressor and activator to induce broad expression of ngn1 in the ventral ectoderm: as a repressor, it could neuralize the ventral ectoderm; as an activator, it could induce broad expression of ngn1 in this domain.

Unresolved issues

This study has explored the role of iro1 and iro7 in neurogenesis and defined a new role for Iro genes in establishment of an ectodermal compartment following Wnt signaling in vertebrate development. However, many questions remain unanswered. Although ectopic expression of Iro genes can promote ngn1 expression, endogenous ngn1 expression is only observed in restricted subdomains of the normal Iro expression domain, suggesting that additional factors regulate ngn1 expression. Furthermore, when Iro genes induce ngn1 expression in a broad domain, they inhibit differentiation of neurons, suggesting that Iro genes also induce expression of factors that prevent differentiation [(Bellefroid et al., 1998; Gomez-Skarmeta et al., 1998) and data not shown]. Clarification of the mechanisms that inhibit ngn1 expression and neuronal differentiation will be necessary to better understand how the Iro genes pattern early neurogenesis.

We have demonstrated that a knockdown of iro1 and iro7 prevents formation of the isthmic organizer at the MHB and it affects patterning in the forebrain and hindbrain. These effects may reflect functions for iro1 and iro7 independent of their role in MHB formation, as previous studies have shown that elimination of the isthmic organizer does not affect forebrain and hindbrain development in a similar manner (Lun and Brand, 1998; Reifers et al., 1998). Further characterization of the changes in forebrain and hindbrain patterning are likely to reveal a relatively early role in the forebrain and a late role in the hindbrain when iro1 and iro7 are expressed together in these domains during development. Finally, while loss of iro1 and iro7 function prevents formation of the neural crest and the isthmic organizer, ectopic expression of these genes is not sufficient for the formation of these tissues. Identification of factors that work together with Iro genes to determine neural crest fate and MHB identity also remain important directions for future studies.