Regulation of iro3 expression in the zebrafish spinal cord

Authors


Abstract

Specification of spinal cord neurons is regulated by several different transcription factors. In this study, we analyze expression and regulation of the transcription factor iro3 in zebrafish spinal cord. In addition to its broad expression in the progenitor domain of intermediate spinal cord, iro3 is also expressed in postmitotic ventral neurons, starting at early somitogenesis stages. Initially, this expression is only in two primary motoneurons, CaP and VaP, but by 24 hr postfertilization, iro3 is expressed by all classes of zebrafish spinal motoneurons as well as by a ventral interneuron called VeLD. iro3 expression in the progenitor domain of intermediate spinal cord is regulated independently from its expression in ventral neurons. Hedgehog (Hh) signaling is unnecessary for iro3 expression in intermediate spinal cord, but it is required to repress iro3 expression in the progenitor domain of ventral spinal cord. We also show that the basic helix-loop-helix transcription factor Olig2 is required for repression of iro3 expression in the progenitor domain of ventral spinal cord. We discuss our findings in the context of previous studies, suggesting that iro3 represses formation of motoneurons and promotes formation of interneurons. Developmental Dynamics 232:140–148, 2005. © 2004 Wiley-Liss, Inc.

INTRODUCTION

Zebrafish Iro3 (called Irx3 in amniotes, Xiro3 in frogs, and previously called Ziro3 in zebrafish; http://zfin.org/) is a member of the Iroquois family of homeoproteins, which have been described in vertebrates, flies, and nematodes (Gomez-Skarmeta and Modolell, 2002 and references therein). iroquois genes were first identified in Drosophila melanogaster as encoding neural prepattern factors that regulate the expression of proneural genes (Gomez-Skarmeta and Modolell, 1996; Leyns et al., 1996). Later, a vertebrate homologue, Xiro3, was shown to specify neural tissue and induce a neural precursor state (Bellefroid et al., 1998; de la Calle-Mustienes et al., 2002). More recently, Iro3 was shown to be required for correct organizer function in zebrafish (Kudoh and Dawid, 2001).

In addition to these early roles in embryonic development, Iro3 and its orthologues may also have a later role in patterning the vertebrate spinal cord. In all vertebrates examined so far, orthologues of iro3 are expressed in a broad domain in the so-called intermediate spinal cord (Gomez-Skarmeta and Modolell, 2002), a region located just dorsal and ventral of the sulcus limitans. Research conducted in amniotes suggests that expression of irx3 is crucial for the correct specification of interneurons that arise from this spinal cord region (Briscoe et al., 2000). Irx3 is one of several transcription factors that have been implicated in specifying different neuronal populations in the developing spinal cord. Opposing gradients of Hedgehog (Hh) signals emanating from notochord and ventral spinal cord and Bone Morphogenetic Protein (BMP) signals emanating from dorsal spinal cord and non-neural ectoderm are thought to produce several progenitor domains at distinct dorsoventral positions within the spinal cord (Briscoe and Ericson, 2001, and references therein; Goulding et al., 2002, and references therein; Timmer et al., 2002). Each domain can be identified by expression of a distinct combination of transcription factors (Goulding et al., 2002, and references therein). In the ventral spinal cord, these transcription factors have been divided into two groups that respond differently to Hh signaling (Briscoe et al., 2000). Class I transcription factors are repressed by Hh signals, whereas Class II transcription factors are induced by Hh signals. Each progenitor domain boundary is associated with the ventral limit of expression of a Class I transcription factor and the dorsal limit of expression of a Class II transcription factor. Class I and Class II transcription factors that share an expression boundary mutually repress one another and, in doing so, sharpen and maintain the expression domain boundary (Kessaris et al., 2001, and references therein). Irx3 is a Class I transcription factor: in amniote embryos, Hh signals repress irx3 expression in the ventral spinal cord and, hence, determine the ventral limit of irx3 spinal cord expression (Briscoe et al., 2000; Persson et al., 2002; Wijgerde et al., 2002; Meyer and Roelink, 2003). The Class II partner of irx3 is olig2, which is expressed in the spinal cord ventral to the irx3 ventral boundary (Mizuguchi et al., 2001; Novitch et al., 2001; Fig. 1). Ectopic expression experiments have shown that, in amniote embryos, olig2 can repress expression of irx3 and vice versa (Mizuguchi et al., 2001; Novitch et al., 2001). This olig2/irx3 expression boundary demarcates motoneuron and interneuron precursors: motoneurons develop from the more ventral olig2-expressing domain and a particular class of ventral interneurons, called V2 interneurons, develop more dorsally, from the most ventral region of the irx3 expression domain (Briscoe et al., 2000; Mizuguchi et al., 2001; Novitch et al., 2001; Fig. 1). Ectopic expression of irx3 in chick embryos prevents cells that would otherwise develop as motoneurons from adopting this fate choice—instead they develop as V2 interneurons (Briscoe et al., 2000). The corresponding loss of function experiment has not yet been reported, but when olig2 is ectopically expressed, the resulting down-regulation of irx3 expression correlates with a dorsal expansion of motoneurons (Novitch et al., 2001).

Figure 1.

irx3 and olig2 have complementary patterns of expression in amniote spinal cord. Schematic cross-section of amniote spinal cord showing irx3 and olig2 expression in the progenitor domain (central region of diagram). irx3 expression is shown in yellow, and olig2 expression is in red. The ventral limit of irx3 expression is the dorsal limit of olig2 expression. Lateral ovals represent different populations of postmitotic neurons. Dorsal interneurons are shown in different shades of blue and ventral neurons in different shades of green, with the exception of the V2 domain, which is fuchsia and the motoneuron (MN), domain which is gray. MNs develop from the olig2 expression domain, and V2 interneurons develop from the most ventral part of the irx3 expression domain. The dorsal limit of the irx3 domain has not been characterized precisely. FP, floor plate; RP, roof plate.

In contrast to these results in amniote embryos and to results from frogs which suggest that Xiro3 induces a neuronal precursor state and prevents neuronal differentiation (Bellefroid et al., 1998; de la Calle-Mustienes et al., 2002), analysis in zebrafish has suggested that iro3 may be expressed in differentiating primary motoneurons (Tan et al., 1999). Unlike amniotes, zebrafish have two distinct populations of motoneurons: primary motoneurons (PMNs) are born earlier and are larger than secondary motoneurons (SMNs) and may be specific to anamniote vertebrates such as fish and amphibians (Kimmel and Westerfield, 1990; Kimmel et al., 1994). All the evidence to date suggests that both PMNs and SMNs in zebrafish are specified by the same mechanisms that specify motoneurons in amniotes (Beattie et al., 1997; Chen et al., 2001; Lewis and Eisen, 2001). However, the suggestion that iro3 may be expressed in zebrafish motoneurons (Tan et al., 1999) prompted us to analyze iro3 expression and regulation in more detail in the zebrafish spinal cord to investigate whether zebrafish Iro3 functions differently to its orthologues.

RESULTS

iro3 Is Expressed by Motoneurons and VeLD Interneurons

We used in situ hybridization to determine what types of neurons in the zebrafish spinal cord express iro3. Consistent with an earlier report (Tan et al., 1999), we observed spinal cord expression of iro3 from early somitogenesis stages. At six to eight somites, iro3 is expressed in a broad stripe in the intermediate spinal cord as well as in a few individual cells in anterior ventral spinal cord (Fig. 2A). The number of ventral cells expressing iro3 increases with time such that, by 10–12 somites, there is a discontinuous row of cells expressing iro3 along most of the trunk spinal cord (Fig. 2B). At both this stage and 18–20 somites, these ventral iro3-expressing cells are regularly spaced, with one or two cells expressing iro3 per somite length (Fig. 2B–D). Cross-sectional analysis revealed that the broad intermediate band of iro3 expression extends mediolaterally throughout the spinal cord, although expression is usually weaker more laterally, whereas the ventral iro3-expressing cells are exclusively lateral, suggesting that they might be postmitotic neurons (Fig. 2G,I). To determine whether these ventral cells are PMNs, we stained 18- to 20-somite stage embryos with both Islet antibody and iro3 riboprobe. At this stage, the only cells that express Islet in the ventral spinal cord are PMNs. We found that all of the ventral iro3-expressing cells also express Islet protein. We observed a regular alternating pattern of PMNs that only express Islet and PMNs that express both Islet and iro3 (Fig. 2C,F), suggesting that iro3 is only expressed by a subset of PMNs.

Figure 2.

iro3 is expressed in CaP and VaP motoneurons. iro3 in situ hybridization (A,B,D,E,G,H) and Islet antibody staining + iro3 in situ hybridization (C,F,I). iro3 staining is blue and cytoplasmic; Islet antibody staining is brown and nuclear. Lateral views of the trunk at 6–8 somites (A), 10–12 somites (B), 18–20 somites (CF), and cross-sections through the trunk at 18–20 somites (GI). A–D,F,G,I: iro3 expression in wild-type (wt) embryos. E,H: iro3 expression in mib mutants. At all of these stages, in wild-type embryos, iro3 is expressed in a broad domain in intermediate spinal cord and in a subset of primary motoneurons (CaPs and VaPs). Arrows indicate iro3-expressing primary motoneurons (all of the iro3-expressing PMNs are indicated in A, C, and F, but only examples are indicated in B, D, E, and G–I); stars indicate primary motoneurons that do not yet express iro3 (MiPs and RoPs, brown only cells). The dotted lines in B demarcate somite boundaries. F is a close up of a two-somite length of the embryo shown in C. In mib mutants, the broad intermediate domain of iro3 expression is lost but additional cells in the primary motoneuron domain express iro3. Scale bar in A = 50 μm in A–E,G–I, 25 μm in F.

In zebrafish, there are three different PMN subtypes: Rostral Primary (RoP), Middle Primary (MiP), and Caudal Primary (CaP). One PMN of each subtype forms per spinal hemisegment, with the exception that approximately half the hemisegments initially have two CaP-like PMNs, one of which is called Variable Primary (VaP) and usually dies (Lewis and Eisen, 2003, and references therein). Each PMN subtype is uniquely identifiable by its soma position relative to overlying somites and by its axon trajectory. For example, MiP and RoP somata are initially adjacent to overlying somite boundaries and CaP somata are initially adjacent to overlying somite middles. CaP axons project into ventral myotome, MiP axons project into dorsal myotome, and RoP axons project into the region of myotome between the CaP and MiP projections.

By comparing the locations of the ventral, postmitotic, iro3-positive cells with somite boundaries, we determined that at 18–20 somites at least the vast majority of these cells are CaP and VaP MNs (e.g., Fig. 2B). We further confirmed that CaPs express iro3 at this stage by dye-labeling individual CaPs in live embryos at 18–20 somites and confirming that the labeled cells expressed the iro3 riboprobe (5/5 CaPs in three live embryos).

These experiments suggested that iro3 might be a PMN subtype-specific marker, expressed in CaPs and VaPs but not in RoPs or MiPs. However, in a few wild-type embryos at 18–20 somites, we saw additional ventral cells expressing iro3 in very anterior spinal cord, suggesting that iro3 may start to be expressed in additional cell types slightly later in development. We, therefore, examined iro3 expression at later stages. At 24 hr postfertilization (h), we observed small groups of at least three to four iro3-expressing ventral spinal cord cells in the vicinity of overlying somite boundaries. These cells were nearly continuous with the intermediate broad stripe of iro3 expression (Fig. 3C star and Fig. 4A), in contrast to the more distinct domains of iro3 expression at earlier stages (e.g., Fig. 2B). To determine whether any of these cells were MiPs, RoPs, SMNs, or ventral interneurons, we dye-labeled individual neurons in 13 live embryos and determined whether individually identified cells expressed iro3. Based on their soma positions and axon trajectories, we found that all of the dye-labeled RoPs (n = 5; Fig. 3A, arrow), CaPs (n = 7; Fig. 3C, arrow, and 3D), MiPs (n = 7; Fig. 3E,F), SMNs (n = 6; Fig. 3B), and individually identifiable VeLD interneurons (n = 8; Fig. 3G,H) expressed iro3 at this stage. This finding suggests that iro3 expression is not specific to a particular subtype of PMN or even to PMNs but is more generally expressed by postmitotic neurons in ventral spinal cord.

Figure 3.

iro3 is expressed by all motoneuron subtypes and by VeLD interneurons. Lateral trunk views of wild-type embryos at 24 hours postfertilization. Individual neurons were labeled with a mixture of fluorescein dextran and rhodamine dextran and then embryos with labeled neurons were processed for iro3 in situ hybridization (blue) and anti-fluorescein antibody staining (red). A–H: Shown are brightfield images (A–C,E,G) and fluorescent images (D,F,H). A: A labeled RoP that expresses iro3. B: A labeled SMN that expresses iro3. C,D: The same labeled CaP that expresses iro3. At this stage of development, CaP cell bodies have often moved dorsally, so that, like this example, they are now much closer to somite boundaries (Myers et al., 1986). E,F: The same labeled MiP that expresses iro3. G,H: The same labeled VeLD that expresses iro3. In both A and C, the PMN is the more ventral of the two labeled cells and is indicated with white arrows; the more dorsal cells were inadvertently labeled by the pipette en route to the PMN. A,B: The labeled neuron was identified after the staining procedures by comparing its location with the previously recorded locations and identities of labeled cells in that embryo (see Experimental Procedures section). C: The white star indicates a small group of iro3-expressing ventral cells near the somite boundary (see text; this finding can be seen more clearly in Fig. 4A). Scale bar = 50 μm in A (applies to A–H).

Figure 4.

iro3 expression is regulated by Hedgehog (Hh) signals and Olig2. Lateral trunk views (A,C,E,G,I,K) and cross-sections through the trunk (B,D,F,H,J,L) showing iro3 in situ hybridization at 24 hours postfertilization (A–D) and 18–20 somites (E–L). A,B: Wild-type (WT) embryos. C,D: syu mutants. E,F: smu mutants. G,H: Wild-type embryos injected with hh-morpholino antisense oligonucleotides (MO), a severely affected embryo (G) and a slightly less severely affected embryo (H). I,J: Wild-type embryos injected with hh RNA. K,L: Wild-type embryos injected with olig2-MO. Dotted white lines in A demarcate somite boundaries. Scale bar = 50 μm in A (applies to A–L).

iro3 Is Expressed in Intermediate Spinal Cord Progenitor Cells

The cells that form the earliest-developing (“primary”; Kimmel and Westerfield, 1990) neurons in zebrafish spinal cord are selected by a process of lateral inhibition. Delta/Notch signaling is required in this process to maintain a population of proliferative neural precursors that will eventually differentiate into later-developing neurons and glia (Lewis, 1996, and references therein). For example, in the absence of Delta/Notch signaling, excess PMNs form in zebrafish embryos at the expense of later-developing SMNs (Appel and Eisen, 1998).

To further confirm that iro3 is expressed in PMNs and to investigate whether it is expressed in any other early-developing neurons, we analyzed iro3 expression in mutants which lack Delta/Notch signaling. The mind bomb (mib) locus encodes a ubiquitin ligase that is required for efficient Delta/Notch signaling and in mib mutants at least the vast majority of neural precursor cells precociously differentiate as early-developing neurons (Jiang et al., 1996; Schier et al., 1996; Itoh et al., 2003; Park and Appel, 2003). When we examined iro3 expression in mib mutants, we found that most of the broad band of iro3 expression in intermediate spinal cord was lost, consistent with this expression being in neuronal or glia precursor cells, whereas the number of iro3-expressing cells in ventral spinal cord was dramatically increased, consistent with these cells being postmitotic PMNs (n = 19; compare Fig. 2D,E and Fig. 2G,H). Although this result does not address whether iro3 is expressed in later-developing neurons, it shows that of all the neuronal populations that are known to be expanded in mib mutant spinal cords (e.g., lhx1a-expressing, lhx5-expressing, vsx1-expressing and Pax2-expressing interneurons and Rohon–Beard sensory neurons; Appel et al., 2001; Itoh et al., 2003; K.E. Lewis, J. Bates, and J.S. Eisen, unpublished data) only ventral neurons express iro3.

iro3 Expression Is Regulated by Hh Signals

Experiments in amniotes suggest that the ventral limit of irx3 expression in spinal cord is determined by Hh signaling. Hh signals from the notochord and floor plate are reported to repress iro3 expression in the most ventral regions of the spinal cord, including the motoneuron precursor domain (Briscoe et al., 2000; Persson et al., 2002; Wijgerde et al., 2002; Meyer and Roelink, 2003). The role of Hh signaling in regulating interneuron specification in the spinal cord has not yet been addressed in zebrafish, but previous analysis has shown that, as in other vertebrates, Hh signals are present in the zebrafish embryonic midline, and these signals are required to induce MNs in the ventral spinal cord (Eisen, 1999, and references therein; Lewis and Eisen, 2001).

We examined the role of Hh signaling in regulating iro3 expression in the zebrafish spinal cord. For these studies, we analyzed iro3 expression in embryos with mutations in genes encoding different components of the Hh signaling pathway, as well as in embryos injected with morpholino antisense oligonucleotides (MOs) to the three different hh genes expressed in the zebrafish embryonic midline (shh, twhh, and ehh; Krauss et al., 1993; Ekker et al., 1995; Currie and Ingham, 1996). In all cases, we observed a ventral expansion of iro3 expression in the spinal cord progenitor domain. The degree to which this expression occurred depended on the extent to which Hh signaling was reduced. Compared with wild-type embryos (Fig. 4A,B), in syu mutants lacking only Shh signaling (Schauerte et al., 1998), there was a slight ventral expansion of iro3 expression (n = 85/85; Fig. 4C,D). In smu mutants (n = 91/91; Fig. 4E,F), which lack nearly all Hh signaling because of a mutation in Smoothened, an essential component of the Hh signaling pathway (Chen et al., 2001; Lewis and Eisen, 2001; Varga et al., 2001) and in most of the embryos injected with MOs against all three hh genes (n = 76/83; Fig. 4G,H), iro3 expression expanded throughout most of the ventral spinal cord, often extending all the way to the medial floor plate (e.g., Fig. 4F). In contrast to these loss-of-function results, overexpression of hh RNA caused a loss of most iro3 expression in the spinal cord progenitor domain, although expression was still observed in the more ventral postmitotic neurons, which at this stage are predominantly PMNs (n = 162/183; Fig. 4I,J). Consistent with earlier reports, we did not observe any significant increase in the number of PMNs in these embryos (see discussion in Lewis and Eisen, 2003, and references therein). These results demonstrate that Hh signals are required for correct iro3 expression in the spinal cord progenitor domain but suggest that expression of iro3 in PMNs is regulated by an independent mechanism.

Olig2 Is Required for Ventral Repression of iro3 Spinal Cord Expression

In amniotes, irx3 can repress expression of olig2 (Novitch et al., 2001). We, therefore, attempted to determine whether iro3 is sufficient or required to repress olig2 expression in zebrafish, by injecting both full-length iro3 and a dominant negative iro3 construct, VP-iro3 (Kudoh and Dawid, 2001). Consistent with earlier studies that overexpressed iro3 (Kudoh and Dawid, 2001) or related genes iro1 and iro7 in zebrafish (Itoh et al., 2002), we found that even mosaic ectopic expression of very small amounts of iro3 or VP-iro3 RNA produced embryos that died or had very disturbed morphology, probably because of defects in gastrulation. To follow the mosaic expression of our constructs, we coinjected embryos with lacZ RNA. In the rare case in which we obtained an embryo with reasonably normal morphology and X-GAL staining in the spinal cord, we saw no effect on olig2 expression from either construct (data not shown). However, it is possible that, in these cases, the injected RNA was present at too low a concentration to have an effect.

We also attempted to knock-down iro3 function with a morpholino. Similar to our results with the dominant negative iro3 construct VP-iro3, we found that injections of iro3-MO caused severe gastrulation defects (2.5 mg/ml; n = 146/146). When we injected increasingly lower concentrations of iro3-MO the gastrulation defects became less severe and at low concentrations most of the embryos had normal morphology. However, when we analyzed iro3-MO–injected embryos with a normal or slightly abnormal morphology by in situ hybridization, we saw no effect on olig2 expression (0.6 mg/ml, n = 44/44; 1.25 mg/ml, n = 54/54). However, as with the dominant-negative iro3 construct, it is possible that, in these cases, the morpholino was present at too low a concentration to have an effect.

In amniote embryos, in addition to irx3 repressing expression of olig2, olig2 regulates expression of irx3. Ectopic expression of olig2 in chick spinal cord represses irx3 expression (Mizuguchi et al., 2001; Novitch et al., 2001) and in mouse mutants that lack olig2 (Lu et al., 2002) or olig1 and olig2 (Zhou and Anderson, 2002), irx3 expression expands ventrally within the spinal cord. Zebrafish olig2 is expressed in ventral spinal cord, as in amniote embryos (Park et al., 2002). To test whether Olig2 is required for normal repression of iro3 expression in the progenitor domain of ventral spinal cord in zebrafish, we knocked-down Olig2 function using MOs. In most olig2-MO–injected embryos, we observed a clear ventral expansion of iro3 expression (n = 96/138; Fig. 4K,L); in the most severe cases, iro3 expression extended to the floor plate.

olig2 expression requires Hh signals (Park et al., 2002), raising the possibility that Hh signaling regulates iro3 expression by means of Olig2. We therefore analyzed olig2 expression in embryos with reduced or ectopic Hh signaling and compared this with iro3 expression. Consistent with an earlier report (Park et al., 2002), we found that, at 20–24 h, olig2 expression is reduced in syu mutants (n = 78/78; Fig. 5B,F,I) and absent in smu mutants (n = 78/78; Fig. 5C). This finding correlates with our observation that the ventral expansion of iro3 expression in the spinal cord progenitor domain is less severe in syu mutants than in smu mutants (Fig. 4C–F). We also overexpressed hh RNA at concentrations that severely reduce iro3 expression in the spinal cord progenitor domain (Figs. 4I,J, 5J) and analyzed olig2 expression. In the majority of injected embryos, we observed a dramatic expansion of olig2 expression (n = 206/227; Fig. 5D,G,J). All of these results are consistent with the hypothesis that Hh signaling regulates iro3 expression in the spinal cord progenitor domain by means of Olig2.

Figure 5.

olig2 expression is regulated by Hh signals. Lateral trunk views (A–D) and cross-sections through the trunk (E–J) at 22–24 hours postfertilization showing olig2 single in situ hybridization (A–G) and olig2 (blue) and iro3 (red) double in situ hybridization (H–J). A,E,H: Wild-type (wt) embryos. B,F,I: syu mutants. C: smu mutant. D,G,J: Wild-type embryos injected with hh RNA. J: Most iro3 expression has been lost from the intermediate spinal cord, but iro3 is still expressed in motoneurons (example indicated with an arrow; see also Fig. 4I,J). Scale bar = 50 μm in A (applies to A–J).

DISCUSSION

In this study, we describe expression of iro3 in the zebrafish spinal cord and show how this expression is regulated. iro3 is expressed broadly in neural progenitors in intermediate spinal cord, but in addition, it is expressed in several ventral spinal cord neurons, including primary and secondary motoneurons and VeLD interneurons. Initially, this ventral spinal cord expression of iro3 is PMN subtype-specific, as it is confined to CaPs and VaPs. However, at slightly later stages iro3 is also expressed by MiPs, RoPS, SMNs, and VeLDs. CaP and VaP are the first PMNs to extend axons (Eisen et al., 1986; Myers et al., 1986), suggesting that this dynamic expression of iro3 may reflect temporal differences in the development of these ventral spinal cord neurons.

Our analysis shows that iro3 expression in the spinal cord progenitor domain does not require Hh signaling, as even in embryos in which all Hh signals have been “knocked down” by hh-MOs, iro3 expression remains. However, Hh signals are required for correct spatial regulation of iro3 expression in this domain, as they repress iro3 expression in the progenitor domain of ventral spinal cord. With progressively more severe loss of Hh signaling, we observed more significant ventral expansion of iro3 expression. This finding suggests that there is a specific concentration of Hh signals that represses iro3 expression in the spinal cord progenitor domain. This repression may act through Olig2 as olig2 expression requires Hh signals (Park et al., 2002 and this study), and our results demonstrate that Olig2 is also required to repress iro3 expression in the ventral spinal cord progenitor domain. We also show that, when Hh signaling is reduced to different extents, there is a strict correlation between the effects on iro3 expression and olig2 expression. In smu mutants, which lack olig2 expression, iro3 expression extends throughout most of the ventral spinal cord progenitor domain, whereas in syu mutants, which retain some olig2 expression, the ventral expansion of iro3 expression in the spinal cord progenitor domain is less pronounced. In addition, when Hh signaling is increased olig2 expression expands dorsally and iro3 expression in the spinal cord progenitor domain is lost. All of these observations are consistent with the hypothesis that Hh signaling regulates iro3 expression in the spinal cord progenitor domain by means of Olig2.

In zebrafish, as in other vertebrates, motoneuron specification requires Hh signals (Beattie et al., 1997; Chen et al., 2001; Lewis and Eisen, 2001) and Olig2 function (Park et al., 2002). Thus, we were unable to test whether Hh signals are also required for iro3 expression in motoneurons, independently of their role in motoneuron specification. However, the ability of hh overexpression to repress iro3 expression in the spinal cord progenitor domain while leaving iro3 expression in ventral motoneurons unaffected suggests that these two different domains of iro3 expression are regulated by independent mechanisms.

Given that irx3 is thought to be responsible for instructing neural precursor cells to develop into interneurons rather than motoneurons (Briscoe et al., 2000), it was initially surprising to us that, in zebrafish, iro3 is expressed in motoneurons. However, irx3 expression has also been reported in cells in the motoneuron domain in both chick and mouse embryos (Bosse et al., 1997; Houweling et al., 2001; Novitch et al., 2001), although to our knowledge double-labeling experiments have not been conducted to confirm that these cells are motoneurons, and these observations have not been generally discussed in the literature. The observation that motoneurons express iro3 suggests that iro3 can only repress motoneuron fate during a very specific time period, which probably corresponds to a time when motoneurons have not yet become postmitotic. This timing would be consistent with iro3 acting mainly through repression of olig2, as olig2 is predominantly expressed in motoneuron progenitors (Mizuguchi et al., 2001; Novitch et al., 2001; Park et al., 2002).

The expression of iro3 in motoneurons may simply be a consequence of the down-regulation of olig2 in these cells: iro3 may be expressed everywhere in the spinal cord that olig2 does not repress its expression. However, if this were the case, we would expect iro3 to be expressed indiscriminantly in postmitotic neurons throughout the dorsoventral extent of its expression domain. Our results in mib mutants argue against this model, as they suggest that neuronal expression of iro3 is restricted to specific populations of neurons. Of all the neuronal populations that are expanded in mib mutant spinal cords (e.g., lhx1a-expressing, lhx5-expressing, vsx1-expressing, and Pax2-expressing interneurons and Rohon–Beard sensory neurons; Appel et al., 2001; Itoh et al., 2003; K.E Lewis, J. Bates, and J.S. Eisen, unpublished data), only PMNs express iro3. This finding suggests that iro3 expression in motoneurons is regulated by a more specific mechanism than simply down-regulation of olig2 in these cells. Although expression alone is not proof of function, these observations raise the interesting possibility that iro3 may have a specific function in cells after they have been specified as motoneurons.

iro3 and irx3 expression in motoneurons in fish and amniotes is also surprising given the evidence from studies in frogs that Xiro3 induces a neural progenitor state and prevents neuronal differentiation (Bellefroid et al., 1998; de la Calle-Mustienes et al., 2002). This paradox may be resolved if the role of Iro3 and its orthologues in preventing neuronal differentiation, like its role in inhibiting motoneuron fate, is limited to a very precise temporal window. Alternatively, it is possible that this function of Xiro3 is species-specific. If this is the case, we predict that frog motoneurons will not express Xiro3; to our knowledge, this question has not yet been addressed. However, our results in zebrafish are consistent with Iro3 inducing a neural progenitor state and preventing neuronal differentiation in the intermediate spinal cord domain, as iro3 expression is lost in this domain when cells precociously differentiate in mib mutants. More studies, therefore, are needed to determine whether iro3 and its orthologues have distinct functions in different vertebrates and to establish what function, if any, iro3 has in postmitotic ventral spinal cord neurons.

EXPERIMENTAL PROCEDURES

Propagation and Identification of Wild-Type and Mutant Zebrafish Embryos

Zebrafish (Danio rerio) embryos were obtained from natural spawnings of wild-types (AB) or crosses of identified carriers heterozygous for specific mutations. Fish were maintained in the University of Oregon Zebrafish Facility on a 14-hr light/10-hr dark cycle at 28.5°C and embryos staged according to Kimmel et al. (1995) by number of somites or hours postfertilization at 28.5°C (h). Mutant embryos were identified by morphology at 24 h or, in the case of mind bomb mutants, by their gene expression phenotype (19 of 83 embryos had a distinct gene expression phenotype).

Mutant Alleles Used in This Study

The strongest available mutant alleles of each locus were used for this study. mind bombta52b (mib) is a mutation in a ubiquitin ligase, which is necessary for efficient Notch signaling (Itoh et al., 2003); sonic yout4 (syu) is a deletion of the sonic hedgehog (shh) locus (Schauerte et al., 1998); slow muscle omittedb641 (smu) is a mutation in the smoothened gene that encodes an essential component of the Hh signaling pathway (Varga et al., 2001). smu mutants have a more severe loss of Hh signaling than syu mutants, but they still retain some Hh signaling, probably due to maternal provision of smu RNA (Chen et al., 2001; Lewis and Eisen, 2001; Varga et al., 2001).

In Situ RNA Hybridization and Antibody Staining

In situ RNA hybridization was performed as previously described in Concordet et al. (1996). iro3 probe was synthesized from a ziro3 construct in Bluescript (Tan et al., 1999), but see the discussion of this construct in Kudoh and David (2001). olig2 probe was synthesized as in Park et al. (2002). islet2 probe was synthesized as in Appel et al. (1995). In all cases, these probes were detected with nitroblue tetrazolium chloride/5-bromo-4-chloro-3-indolyl-phosphate, toluidine salt (NBT/BCIP) or Sigma Fast Red tablets. Islet antibodies originally isolated by the Jessell lab were obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the NICHD and maintained by the University of Iowa, Department of Biological Sciences (Iowa City, IA). A 1:1 mixture of antibodies 39.4D5 and 40.2D6 was used with both antibodies at a final concentration of 1/300. In cases in which Islet antibody staining and iro3 in situ RNA hybridization were performed on the same embryos, antibody staining was done first as in Lewis and Eisen (2004) and was detected with DAB.

Specimens were analyzed using a Zeiss Axioplan microscope and photographed with Kodak Ektachrome 64T or 164T film. Fluorescent images were photographed with Kodak P1600 Ektachrome film. Images were scanned on a Nikon LS-1000 35-mm film scanner and processed using Adobe Photoshop software.

Morpholino Injections

MOs were obtained from Gene Tools and injected into one- to two-cell wild-type embryos as in Lewis and Eisen (2001). A combination of MOs against all three hh genes was used, with each MO at a final concentration of 1 mg/ml. We exploited the fact that Hh signals are required for PMN induction, and even small amounts of Hh signal are sufficient for some PMNs to form, to assess the degree to which Hh signaling had been “knocked down” in each experiment (Lewis and Eisen, 2001). For each injection experiment, some embryos were processed for islet2 RNA in situ hybridization to determine how many, if any, PMNs remained. We only analyzed experiments in which PMN numbers were severely reduced. olig2-MO was used at a concentration of 1–1.5 mg/ml. olig2-MO has also been reported to reduce PMN numbers (Park et al., 2002). We therefore checked whether this MO was working by processing some olig2-MO–injected embryos for islet2 expression. The olig2-MO was less effective and more variable at reducing PMN numbers than the hh-MOs, but 41/61 of the olig2-MO–injected embryos that we analyzed for islet2 expression had reduced numbers of PMNs (data not shown). iro3-MO was injected at concentrations of 0.6, 1.25, and 2.5 mg/ml. Embryos with very severe gastrulation defects were not processed further, but embryos with more subtle gastrulation defects or normal morphology were processed for olig2 in situ hybridization. MO sequences were as follows: twhh-MO-2, TCCATGACGTTTGAATTATCTCTT (Nasevicius and Ekker, 2000); ehh-MO, CGCCGCCGCCGTGGAGAGTCTCAT (Lewis and Eisen, 2001); shh-MO, CAGCACTCTCGTCAAAAGCCGCATT (Nasevicius and Ekker, 2000); olig2-MO, CGTTCAGTGCGCTCTCAGCTTCTCG (Park et al., 2002); iro3-MO, AGCTGTGGGAAAGACATTGTTGTGG.

RNA Injections

We injected both shh and twhh RNA into one- to two-cell stage embryos as in Lewis et al. (1999). These experiments gave identical results and, therefore, are considered together in our analysis. We also injected a full-length iro3 construct, pCS2-iro3, and a dominant negative iro3 construct, VP-iro3 (Kudoh and Dawid, 2001) into 4- to 32-cell stage embryos. In these cases, a mixture of iro3 RNA was injected in combination with lacZ RNA so that we could follow the mosaic distribution of injected RNAs. A range of concentrations were attempted for each construct (0.005–0.01 mg/ml pCS2-iro3; 0.005–0.012 mg/ml VP-iro3). Embryos were fixed in 2% paraformaldehyde (PFA) with 0.02% Igepal and processed for X-GAL staining and then re-fixed overnight in 4% PFA and processed for in situ hybridization with an olig2 riboprobe.

Intracellular Labels

Individual PMNs, SMNs, and VeLD interneurons were labeled in live embryos with a mixture of fixable rhodamine dextran and fluorescein dextran as in Eisen et al. (1989). At 24 h, all of these neurons can be readily identified by their unique cell body positions and axonal morphologies within the ventral neural tube (Lewis and Eisen, 2003); CaPs are readily identified by their axon trajectories as early as 17 h (Eisen et al., 1986). We labeled 5 CaPs in three 18–19 h embryos and 33 individually identified neurons in thirteen 24 h embryos. In all cases, the positions and morphologies of the labeled cells were recorded so that the neurons could be recognized by their cell body positions at later stages, even if their axons were no longer visible. The embryos were then fixed overnight in 4% PFA and processed for in situ hybridization with an iro3 riboprobe detected in blue with NBT/BCIP, followed by an anti-fluorescein antibody staining detected with fast red. In all cases, the cell bodies of the labeled cells could be identified by their red color using brightfield microscopy, and in many but not all cases, the axons could also be seen either by brightfield or fluorescence microscopy.

Acknowledgements

We thank Stavros Diamantakis, Amanda Lewis, Ellie Melançon, Chapell Miller, and the staff of the UO Zebrafish Facility for fish husbandry; the staff of the UO Histology Facility, and Flavio Zolessi for help with sectioning; Igor Dawid and Zhiyuan Gong for constructs; Bruce Appel for constructs and olig2-MO; and Estelle Hirsinger, Lisa Maves, and David Rivers for comments on previous versions of this manuscript. J.S. Eisen received funding from the NIH, and K.E. Lewis received a Royal Society University Research Fellowship.

Ancillary