Nkx6 proteins specify one zebrafish primary motoneuron subtype by regulating late islet1 expression

Motoneurons of the vertebrate central nervous system have precise subtype identities characterized by soma location, axon trajectory, target muscle innervation and combinatorial transcription factor expression(Lewis and Eisen, 2003; Shirasaki and Pfaff, 2002). Specification of motoneuron subtypes has been well-studied in mouse and chick embryos (Jurata et al., 2000; Pfaff and Kintner, 1998; Tsalik et al., 2003); however,nothing is currently known about the molecular mechanisms that establish very fine-grained motoneuron patterning, such as the segmentally reiterated,individually identified primary motoneurons (PMNs) of the zebrafish spinal cord.

We focus our attention on two PMN subtypes, CaP and MiP, that are present in a segmentally repeated, alternating pattern(Lewis and Eisen, 2003). CaPs project axons that innervate ventral muscle and express the LIM homeobox gene islet2, whereas MiPs project axons that innervate dorsal muscle and express the LIM homeobox gene islet1(Lewis and Eisen, 2003). The expression of islet genes is dynamic in these cells(Fig. 1A); both CaP and MiP express islet1 around the time they exit the cell cycle(Appel et al., 1995; Inoue et al., 1994; Korzh et al., 1993; Tokumoto et al., 1995). MiPs then transiently downregulate islet1 expression and reinitiate it prior to axogenesis (Appel et al.,1995). Thus, MiPs express islet1 in two distinct phases -an early phase and a late phase. In contrast to MiPs, CaPs initiate expression of islet2 while they still express islet1, and then downregulate expression of islet1(Appel et al., 1995; Inoue et al., 1994; Korzh et al., 1993; Tokumoto et al., 1995). Thus,CaPs have an early phase of islet1 expression and a later phase of islet2 expression. The end result of these dynamic changes in islet gene expression is that by the time of axon extension, MiPs express exclusively islet1 and CaPs express exclusively islet2.

Transplantation of single CaPs and MiPs revealed that their subtypes are initially labile and responsive to environmental signals, but become committed shortly before axogenesis (Eisen,1991), around the time that the alternating pattern of islet2 and islet1 expression is established(Appel et al., 1995). Thus, it was surprising to learn that either Islet1 or Islet2 protein is sufficient to specify both CaP and MiP subtypes, suggesting that the differences between these PMNs might be regulated by factors upstream of the islet genes(Hutchinson and Eisen, 2006). Homeodomain transcription factors expressed in the motoneuron progenitor (pMN)domain, such as Nkx6.1 (Cheesman et al.,2004), are good candidates for performing this function.

The Nkx6 transcription factor family is important in formation of mouse spinal motoneurons. One family member, Nkx6.1, is expressed in the spinal cord pMN domain and is required for formation of a proportion of all spinal motoneuron subtypes (Sander et al.,2000). Another family member, Nkx6.2, is expressed dorsal to the pMN domain and is negatively regulated by Nkx6.1. In the absence of Nkx6.1 function, Nkx6.2 expression spreads ventrally into the pMN domain and Nkx6.2 partially substitutes for Nkx6.1 during motoneuron formation(Vallstedt et al., 2001). Mice deficient in both Nkx6.1 and Nkx6.2 lose nearly all their spinal motoneurons(Vallstedt et al., 2001). As in mouse, zebrafish Nkx6.1 is expressed in the pMN domain and is also important for formation of some zebrafish motoneurons(Cheesman et al., 2004). Nkx6.1-deficient zebrafish lack secondary motoneurons (SMNs), a type of spinal motoneuron that develops later than PMNs(Myers, 1985), but have normal PMNs, raising the possibility that additional Nkx6 proteins might regulate zebrafish PMN formation.

Here, we provide evidence for a novel role of Nkx6 proteins in motoneuron subtype specification. We show that zebrafish have at least three nkx6 genes, two of which are expressed in the pMN domain and transiently in postmitotic PMNs. Nkx6 proteins are not required for PMN formation because early islet1 expression is normal in embryos lacking Nkx6 proteins. In Nkx6-deficient embryos, CaPs have normal islet2 expression and project normal ventral axons out of the spinal cord; however, MiPs fail to initiate the second phase of islet1expression and do not form their subtype-specific, dorsal peripheral axons,instead adopting a more interneuron-like morphology by projecting an axon within the spinal cord. We suggest that Nkx6 proteins control MiP subtype specification at least in part by regulating the second phase of islet1 expression and that this late expression of islet1 is required to promote the MiP subtype and to suppress interneuron development.

Wild-type (AB) and Tg(islet:GFP)(Higashijima et al., 2000)embryos were reared and staged by hours post-fertilization at 28.5°C(hpf), and gross morphology (Kimmel et al., 1995).

A search of the zebrafish genome revealed an nkx6.2 gene highly similar to zebrafish nkx6.1(Cheesman et al., 2004). A forward primer to an nkx6.2-specific region(5′-CGGCTTCAAGGCTCATTC-3′) and a reverse homeobox primer(5′-CCATTTAGTTCTTCTGTTCTG-3′) isolated a full-length cDNA clone from a 14- to 19-hpf library (Appel and Eisen, 1998). These primers did not amplify control nkx6.1 cDNA. The isolated clone contains an open reading frame and UTR sequences. The DNA sequence predicts a 279 amino acid protein containing both a homeodomain and an NK decapeptide, which defines the nkx gene family. The GenBank accession number for the complete zebrafish nkx6.2 mRNA is DQ416765. We isolated a fragment of nkx6.3 from first-strand cDNA(forward primer, 5′-AGTCCAACATCTCAGGATCC-3′; reverse primer,5′-TCCACTCATACCCTCCATC-3′) and based on this clone (GenBank accession DQ415639) and genomic clones, compiled a putative full-length sequence.

Zebrafish RNA in situ hybridization was performed as described previously(Appel and Eisen, 1998). nkx6.1, islet1 and islet2 mRNA probes were described previously (Appel et al., 1995; Cheesman et al., 2004), as were pax2a (Thaeron et al.,2000), evx1 (Thaeron et al., 2000), eng1b(Higashijima et al., 2004) and chx10 (Kimura et al.,2006). The nkx6.2 mRNA probe was made from full-length nkx6.2 DNA. nkx6.3 expression was determined using a probe against a portion of the nkx6.3 gene that excluded the homeodomain. In the nkx6.3 anti-GFP double-label experiment, the RNA in situ hybridization was performed first, followed by antibody detection.

The following primary antibodies (Abs) were used: polyclonal rabbit anti-Nkx6.1 (1:1200; gift of O. Madsen, Hagedorn Research Institute, Gentofte,Denmark), monoclonal mouse anti-Islet [1:200; antibody recognizes both Islet1 and Islet2 proteins (Korzh et al.,1993); 39.4D5, Developmental Studies Hybridoma Bank], monoclonal mouse zn1 (Trevarrow et al.,1990), monoclonal mouse znp1(Trevarrow et al., 1990),monoclonal anti-GFP (1:200; Clonetech, JL-8,), and polyclonal rabbit anti-GABA(1:1000; Sigma). Secondary antibodies from Molecular Probes were used: goat anti-mouse Alexa-488 (1:1000) and goat anti-rabbit Alexa-568 (1:1000). One secondary antibody from Jackson Labs was also used: goat anti-rabbit Cy5(1:200). Embryos were fixed for 3.5-4 hours in 4% paraformaldehyde (PFA) and 1×Fix Buffer (Westerfield, 2000) at 4°C, blocked in 1×PBS/5%NGS/4 mg/ml BSA/0.5% Triton X-100 for 1 hour at room temperature, and incubated in primary antibody (diluted in blocking solution) overnight at 4°C. Embryos were then washed at room temperature for 1.5 hours in PBS containing 0.1% Tween 20, incubated in secondary antibody (diluted in blocking solution) for 4 hours at room temperature, and then washed for 1.5 hours at room temperature in PBS containing 0.1% Tween 20.

The Nkx6.2 conceptual amino acid translation revealed two potential start methionines; we do not know if one or both of them are functional start sites. As they are close together, GeneTools (Corvallis, Oregon) designed a translation-blocking MO just upstream of the first methionine(5′-GGTGCGCCGGAGCCACAGGACAAAC-3′) on the premise that it would interfere with initiation of translation at either start site. We also utilized a splice-blocking MO (5′-CGCGCAAAACTCACCCGCACAGGGA-3′)that began at position 386 of the nkx6.2 open reading frame and ended in the first intron, blocking the splice-donor site of exon 1. Both nkx6.2 MOs produced similar phenotypes. Several nanoliters of 5 mg/ml nkx6.2 MO, diluted in 0.2 M KCl, were injected with Phenol Red into the yolk cell of two-cell-stage embryos; for double MO injections, we added 2.5 mg/ml nkx6.1 MO (see Cheesman et al., 2004). All our MO injections worked efficiently, with 84.3% (27/32) of nkx6.1 MO-injected, 90.9% (30/33) of nkx6.2MO-injected, and 76.1% (35/46) of nkx6 double MO-injected embryos exhibiting a MiP axon phenotype.

Images of zebrafish embryos were captured on a Zeiss Axioplan equipped with a digital camera, on a Bio-Rad Radiance, or on a Zeiss Pascal confocal microscope. In a few cases, an Olympus IX81 microscope with an FV300 confocal was used to capture images. The brightness and contrast of images was adjusted in Photoshop (Adobe).

All observations of CaP and MiP motoneurons were made in the spinal cord adjacent to somites 8-12. To quantify MiP axons, we counted the number of MiP axons in 28-hpf control, nkx6 single and double MO-injected embryos,as well as nkx6 MOs and islet1 RNA co-injected embryos labeled with zn1 and znp1 Abs. Axons were scored as MiPs if they projected posterior and dorsal to the zn1-labeled CaP soma. The percentage of MiP axons remaining in experimental embryos was calculated relative to controls. The percentage of segments with cells labeled with islet1 RNA in the MiP position was calculated from cells counted in segments 8-12 of 18-hpf control, nkx6.1, nkx6.2 and nkx6 double MO-injected embryos. For experiments in which individual MiPs were dye-labeled, we recognized these cells by their position and used protocols described by Eisen et al.(Eisen et al., 1989). GABA-positive interneurons were counted as described by Hutchinson and Eisen(Hutchinson and Eisen, 2006). The number of cells labeled with chx10 or eng1b riboprobes was counted in segments 1-11 on each side of 24-hpf control and nkx6double MO-injected embryos. The average number of positive cells per segment in experimental embryos was compared with that in control embryos.

Velocity software was used to generate 3D images of individually labeled MiPs. A velocity classifier based on intensity was used to generate a threshold for red and green images of Nkx6 and Islet Abs, followed by Velocity image arithmetic to identify co-labeled cells. The co-labeled cells were then pseudo-colored blue.

Studies implicating two Nkx6 homologs in mouse motoneuron formation(Vallstedt et al., 2001) and showing that zebrafish Nkx6.1 is unnecessary for PMN formation(Cheesman et al., 2004) raised the possibility of additional nkx6 genes in zebrafish. We isolated zebrafish nkx6.2 and nkx6.3 based on publicly available genomic DNA sequences. Phylogenetic analysis indicated that zebrafish nkx6 genes are most closely related to other vertebrate nkx6genes, distinct from nkx2 family members (see Fig. S1A in the supplementary material). Alignment of the three zebrafish Nkx6 predicted proteins revealed that Nkx6.1 and Nkx6.2 are more similar to each other than to Nkx6.3 (see Fig. S1B in the supplementary material). Similar to mouse Nkx6.3 (Alanentalo et al.,2006), zebrafish nkx6.3 expression is restricted to a group of cells in the hindbrain and is not found in the spinal cord (see Fig. S1C in the supplementary material), and thus it cannot be involved in PMN development.

nkx6.1 and nkx6.2 are dynamically expressed in the zebrafish ventral spinal cord. We previously showed that nkx6.1 is expressed in the ventral spinal cord, both in the floor plate and in several longitudinal rows of cells dorsal to the floor plate(Cheesman et al., 2004). The number of nkx6.1-expressing cell rows dorsal to the floor plate(Cheesman et al., 2004) is approximately the same as the number of olig2-expressing cell rows dorsal to the floor plate (see Park et al., 2002). Thus, like olig2, nkx6.1 is expressed in the pMN domain (Cheesman et al.,2004). However, it is also expressed in additional cells ventral to the pMN domain, and might also be in some cells dorsal to the pMN domain. At 10 hpf, just before PMN progenitors begin leaving the cell cycle, nkx6.1 was expressed in more pMN domain cells than nkx6.2(Fig. 1B,C). At 14 hpf,expression of nkx6.1 and nkx6.2 were almost indistinguishable and cross-sections revealed that both genes were confined to neurectoderm (data not shown). At 18 hpf, nkx6.1 and nkx6.2expression was no longer apparent in many of the ventrally located cells(Fig. 1D,E). We previously showed that the ventral cells lacking nkx6.1 at this stage are PMNs(Cheesman et al., 2004),suggesting that expression of both nkx6 genes is extinguished from PMNs by 17-19 hpf, the time when they undergo axogenesis. Both nkx6.1and nkx6.2 continued to be expressed in the ventral spinal cord through at least 24 hpf (see Fig. S2 in the supplementary material).

Nkx6 proteins are differentially expressed in CaPs and MiPs, as shown by our double-label experiments with an Islet monoclonal antibody that recognizes both Islet1 and Islet2 proteins (Korzh et al., 1993), and an Nkx6 polyclonal antibody that recognizes both Nkx6.1 (Cheesman et al., 2004)and Nkx6.2 (see Fig. S2 in the supplementary material). At 14 hpf, both CaPs and MiPs expressed Islet and Nkx6 (Fig. 1F). By 16 hpf, CaPs had extinguished expression of Nkx6 and expressed only Islet (Fig. 1G-I). By contrast, MiPs downregulated expression of Islet1 protein, similar to their downregulation of islet1 mRNA(Appel et al., 1995), but continued to express Nkx6 (Fig. 1G). By 17 hpf, MiPs initiated a second phase of Islet1 protein expression and continued to express Nkx6(Fig. 1H). MiPs expressed Nkx6 until 18 hpf, after which time it was downregulated(Fig. 1I). Nkx6 proteins are expressed in PMN progenitors (Cheesman et al., 2004), thus these proteins are present at the right time to be involved in PMN formation. The differential expression of Nkx6 proteins in postmitotic MiPs and CaPs suggests that Nkx6 might also participate in specification of MiP and CaP subtypes.

We previously showed that Nkx6.1 is unnecessary for PMN formation(Cheesman et al., 2004). To ascertain whether PMNs form in the absence of Nkx6.2, or in the absence of both Nkx6.1 and Nkx6.2, we injected embryos with morpholino antisense oligonucleotides (MOs). Early expression of islet1 was normal in the absence of Nkx6.1, Nkx6.2, or both Nkx6 proteins(Fig. 2A-D), suggesting that PMNs formed normally. Thus, although Nkx6 proteins are expressed in the pMN domain, they appear unnecessary for PMN formation.

Because CaPs and MiPs continue to express Nkx6 proteins after subtype commitment, we asked whether Nkx6 proteins have a role in subtype specification. In the absence of Nkx6.1, Nkx6.2, or both Nkx6 proteins, CaPs expressed islet2 normally (Fig. 2E-H) and developed normal, ventrally projecting axons(Fig. 3A-D). Thus, Nkx6 proteins appear unnecessary for CaP subtype specification or axon pathfinding.

In contrast to CaPs, MiPs were severely affected by the lack of Nkx6 proteins. In the absence of either Nkx6.1 or Nkx6.2, some MiPs failed to initiate the late phase of islet1 expression(Fig. 2I-K). In the absence of both Nkx6 proteins, very few MiPs expressed islet1 at 18 hpf(Fig. 2I,L). In single nkx6 MO-injected embryos, many MiPs had dorsal axons that were truncated or excessively branched and some MiPs failed to form dorsal axons(Fig. 3A-C). MiP axons were present in only 61% of segments in nkx6.1 MO-injected embryos, as compared with 87% of segments in nkx6.2 MO-injected embryos and 98%of segments in control embryos (n=83 segments in nine nkx6.1MO-injected embryos; n=168 in 17 nkx6.2 MO-injected embryos; n=100 in ten control embryos). MiP axons were most severely affected in embryos injected with both nkx6.1 and nkx6.2 MOs, as indicated by the presence of MiP dorsal axons in only 40% of segments(n=240 in 24 nkx6 MO-injected embryos). The MiP axons that remained in nkx6 MO-injected embryos were severely truncated(Fig. 3D). Thus, Nkx6 proteins are required for proper MiP subtype specification and axon pathfinding.

The loss of MiP axons in embryos lacking both Nkx6 proteins correlated with the failure of MiPs to initiate the late phase of islet1 expression. Therefore, we reasoned that MiPs did not form normal axons because this process requires Islet1 protein. To test this hypothesis, we co-injected nkx6.1 and nkx6.2 MOs together with islet1 RNA(Fig. 3E,F). MiP axons were present in 98% of segments in control embryos (n=100 in ten embryos),in 40% of segments in nkx6 double MO-injected embryos (n=240 in 24 embryos), and in 74% of segments in embryos co-injected with islet1 RNA and nkx6 MOs (n=270 in 27 embryos). These data provide evidence that Islet1 is sufficient for the projection of normal MiP axons even in the absence of Nkx6 proteins, and suggest that Nkx6 proteins regulate MiP specification at least in part by regulating expression of islet1.

The failure to maintain islet1 expression and the absence of dorsal axons in MiPs lacking Nkx6 proteins led us to consider whether MiPs developed as interneurons in the absence of late islet1 expression,as they do in the absence of early islet1 expression(Hutchinson and Eisen, 2006). We labeled individual cells in the MiP position in nkx6 double MO-injected embryos by intracellular dye iontophoresis(Eisen et al., 1989). As a negative control, we labeled VeLD interneurons, which are located adjacent to MiPs (Eisen, 1991), and found that VeLDs are normal in nkx6 double MO-injected embryos (12 VeLDs in 11 embryos; data not shown). Dye-labeling of MiPs in nkx6 double MO-injected embryos revealed that these cells had a range of phenotypes(Fig. 4B-F). MiPs normally extend a short ventral axon out of the spinal cord before they form their subtype-specific dorsal axon; the ventral axon is later retracted(Eisen et al., 1989). In the absence of Nkx6 proteins, many MiPs projected a normal ventral axon(Fig. 4C,D), but often failed to retract it. Some MiPs had a dorsal axon, but often it was excessively branched (Fig. 4B) or truncated(Fig. 4C). Surprisingly, many MiPs initiated motoneuron development by projecting a normal ventral axon, but then projected an interneuron-like axon within the spinal cord instead of a dorsal motor axon (Fig. 4D). Finally, some MiPs did not project either ventral or dorsal motor axons, but only extended an interneuron-like axon within the spinal cord(Fig. 4F). However, this interneuron-like axon was excessively branched compared with axons of any of the previously described types of ventral spinal interneurons(Lewis and Eisen, 2003). As an additional control, we antibody-stained embryos in which we had individually dye-labeled MiPs to verify that dye-labeled MiPs lacking dorsal axons correlated with segments lacking dorsal MiP axons as revealed by zn1 and znp1 Ab labeling (Fig. 4G). The results showed that in the absence of Nkx6 proteins, MiPs formed but they failed to extend their normal dorsal axon and instead adopted a more interneuron-like morphology, often developing a `hybrid' phenotype in which they displayed morphological characteristics of both motoneurons and interneurons.

To ascertain whether the absence of late Islet1 expression causes MiPs to take on the molecular characteristics of one or more specific types of interneurons, we characterized the expression of several different interneuron markers in nkx6 double MO-injected embryos and assayed whether individual MiPs labeled with fluorescent dye expressed any of these markers. The GABA Ab labels several types of ventral interneurons that are derived from the pMN domain (see Hutchinson and Eisen,2006), but the number of GABA-positive cells did not increase in nkx6 double MO-injected embryos(Fig. 5A,B,G), and labeled MiPs were not GABA positive (Fig. 4G). Thus, we next examined the expression of several transcription factors that are expressed by interneurons in specific dorsoventral spinal cord positions. pax2a and evx1(Thaeron et al., 2000) are both expressed in interneurons in the dorsal and midregion of the spinal cord;the evx1 expression domain is just ventral to the pax2aexpression domain. Expression of both pax2a and evx1 were was in nkx6 double MO-injected embryos (data not shown). By contrast, chx10 and eng1b, which are both expressed more ventrally than pax2a and evx1, had altered expression patterns in nkx6 double MO-injected embryos. Expression of chx10, which is probably in CiD interneurons (Kimura et al., 2006), was downregulated in embryos lacking Nkx6(Fig. 5C,D,G), whereas expression of eng1b, which is specifically in CiA interneurons(Higashijima et al., 2004), was expanded in nkx6 double MO-injected embryos(Fig. 5E-G). These results are similar to the expansion of En1 (an eng1b ortholog) and reduction of Chx10 in mice lacking Nkx6.1 and Nkx6.2(Vallstedt et al., 2001). Despite the changes in eng1b and chx10 expression,dye-labeled MiPs in nkx6 double MO-injected embryos did not express either of these genes at 28 hpf (data not shown). Therefore, until additional markers of specific types of interneurons become available, we will remain unable to ascribe a specific type of interneuron identity to MiPs that develop in the absence of Nkx6 proteins.

Our studies reveal dynamic differences in Nkx6 expression and function in zebrafish PMN subtypes and provide three key findings. First, we show that although two zebrafish Nkx6 proteins are expressed in PMN progenitors, they are unnecessary for PMN formation. This contrasts with mouse, in which Nkx6 proteins are required for formation of nearly all spinal motoneurons(Vallstedt et al., 2001). It is possible that in zebrafish there are additional nkx6.1 or nkx6.2 orthologs that we have not discovered. This seems unlikely as the Nkx6.1 antibody we used (Cheesman et al., 2004) recognizes both Nkx6.1 and Nkx6.2 proteins (see Fig. S2 in the supplementary material); however, we cannot eliminate the possibility of additional nkx6 orthologs until the complete sequence of the zebrafish genome is available. The few spinal motoneurons that form in Nkx6 double-mutant mice appear to develop early, similar to zebrafish PMNs. Thus,it would be interesting to learn whether early-born motoneurons in mouse have other features that resemble zebrafish PMNs and distinguish them from later-developing motoneurons.

Second, we show that the MiP subtype requires Nkx6 proteins, whereas they appear dispensable for the CaP subtype. In mouse, Nkx6 proteins are required for formation of most motoneurons(Vallstedt et al., 2001) (but see above), but differential function of these proteins in distinct spinal motoneuron subtypes has not been reported. It will be important in future studies to identify the factors that regulate CaP subtype specification, as well as to learn whether Nkx6 proteins have any role in the specification of mammalian motoneuron subtypes.

Finally, we show that Nkx6 proteins prevent MiPs from developing an interneuron-like axon by regulating late islet1 expression. We have previously shown that the early phase of Islet1 expression promotes PMN formation and inhibits interneuron development(Hutchinson and Eisen, 2006). Here, we suggest that the second phase of Islet1 expression promotes the MiP subtype and also inhibits interneuron axon development. However, in the absence of only the late phase of Islet1 expression, MiPs often adopt a hybrid motoneuron/interneuron identity, rather than developing as interneurons, as they do in the absence of both phases of Islet1 expression.

The ability of zebrafish Nkx6 proteins to prevent MiPs from developing an interneuron-like axon is reminiscent of the ability of mouse Nkx6 proteins to inhibit pMN-domain cells from expressing interneuron-specific transcription factors (Vallstedt et al.,2001). Interestingly, Vallestedt and colleagues reported that in the absence of Nkx6 proteins, motoneurons transiently express Evx1,although whether these cells developed interneuron-like axons was not reported. We did not observe ectopic evx1 expression; however, it is possible that evx1 might be expressed early and transiently in MiPs in the absence of Nkx6 proteins. Mouse Nkx6 proteins apparently act within motoneuron progenitors. By contrast, our data raises the possibility that in MiPs, Nkx6 proteins function later, perhaps after the cell has become postmitotic. It will be important to address the timing of Nkx6 function further in future studies.

It is interesting that in the absence of Nkx6 proteins, MiPs become more interneuron-like, rather than becoming more like CaP. MiPs can develop as CaPs when they are transplanted to the CaP spinal cord position several hours before axogenesis (Appel et al.,1995; Eisen, 1991). However, when MiPs are in their normal spinal cord position, they are unlikely to transform into another PMN subtype because specification of MiP and CaP subtypes requires positional signals derived from the overlying somites(Lewis and Eisen, 2004). Our data support a model in which interneuron formation is continuously suppressed during motoneuron development. It will be important to determine whether interneuron specification similarly requires continuous suppression of motoneuron development.

We thank Bret Pearson, Chris Doe and Monte Westerfield for comments on earlier versions of the manuscript; members of the Eisen laboratory for their support; Richard Dorsky for help with spinal interneuron markers; Ole Madsen for the Nkx6 antibody; The University of Oregon Zebrafish Facility staff for animal husbandry; and Chris Rodesch and the Core Imaging Facility at The University of Utah for help with confocal microscopy. The Islet monoclonal antibody developed by Thomas Jessell and colleagues was 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 52242. Supported by NIH grants NS23915, HD22486 and GM007413,and NSF grant DGE9972830.