Regulatory interactions specifying Kolmer-Agduhr interneurons

Motoneurons and different interneuronal subtypes are specified in the ventral spinal cord in response to different concentrations of the morphogen sonic hedgehog (Shh). Close to the source of Shh (notochord, floor plate) V3 interneurons form, while, at a further distance, motoneurons and V2, V1 and V0 interneurons differentiate. The decreasing concentrations of Shh at a distance from the sources are interpreted by transcription factors, the expression of which is either repressed (class 1 transcription factors) or induced (class 2 transcription factors) by Shh (Briscoe and Ericson, 2001; Ingham and McMahon, 2001; Jessell, 2000).

The cells adjacent to the medial floor plate express the homeobox transcription factor genes nkx2.2 and nkx2.9, which are induced by Shh (Cheesman et al., 2004; Guner and Karlstrom, 2007; Schäfer et al., 2005; Schäfer et al., 2007; Xu et al., 2006). In this region of the zebrafish spinal cord (the lateral floor plate), nkx2.9 and two related nkx2.2 genes, nkx2.2a and nkx2.2b, are expressed (Barth and Wilson, 1995; Schäfer et al., 2005; Schäfer et al., 2007; Strahle et al., 2004; Xu et al., 2006). Knockout of Nkx2.2 in the mouse abolishes the formation of V3 interneurons (Briscoe et al., 1999), whereas inactivation of the related transcription factor Nkx2.9 does not result in a phenotype in the mouse spinal cord (Pabst et al., 2003).

The zebrafish lateral floor plate is the origin of two distinct neuronal types: V3 interneurons and GABAergic Kolmer-Agduhr′ (KA′) cells (Fig. 1A,B) (Bernhardt et al., 1992; Schäfer et al., 2007). The KA′ cells of the lateral floor plate and the more dorsally located KA′ cells form a special class of neurons that stay in contact with the ventricular lumen (Martin et al., 1998) (see also Fig. 1A,B). Whereas the ventral KA′ cells develop from the lateral floor plate, the more dorsal KA′ cells are derived from olig2-expressing motoneuron precursors (Park et al., 2004). At least some of the KA′ cells are part of the neuronal network that controls spontaneous swimming movement (Wyart et al., 2009).

Although all these cell types depend on Shh signaling (Pinheiro et al., 2004; Schäfer et al., 2007), it is less clear how the Shh signals are interpreted to trigger the differentiation of KA interneurons. The zinc-finger transcription factors Gli1 (detour) and Gli2 (you-too) are the immediate mediators of the Shh signal in the spinal cord (Karlstrom et al., 1999; Karlstrom et al., 2003). The primary targets of the Gli factors appear to be nkx2.2a, nkx2.2b and nkx2.9 (Cheesman et al., 2004; Guner and Karlstrom, 2007; Xu et al., 2006). As in mammals, zebrafish V3 interneurons express the leucine zipper/PAS domain transcription factor Sim1. In addition, a number of other transcription factors, such as the C4 zinc-finger transcription factors Gata2 and Gata3 and the basic helix-loop-helix transcription factor Tal2, are expressed in the lateral floor plate, but also in other cells at more dorsal aspects of the spinal cord (Batista et al., 2008; Pinheiro et al., 2004; Schäfer et al., 2007). The precise functional relationships of these factors with respect to the different neuronal subtypes are not understood. Morpholino knockdown of nkx2.2a and nkx2.2b does not abolish tal2 expression in the zebrafish spinal cord (Schäfer et al., 2007). In the lateral floor plate, cells that co-express either tal2 and nkx2.2b or foxa2 and nkx2.2b can be distinguished (Schäfer et al., 2007). Although not affecting the tal2-positive cells, double knockdown of nkx2.2a and nkx2.2b abolishes the foxa2/nkx2.2b co-expressing cells. These latter cells seem to differ also in their dependence on hedgehog (Hh) signaling from tal2-positive cells. Moderate reduction of Shh signaling abolishes the foxa2-expressing cells, whereas complete removal of Shh signaling is necessary to prevent the differentiation of tal2-positive cells. The foxa2/nkx2.2b co-expressing and tal2/nkx2.2b co-expressing cells were suggested to represent proliferating and post-mitotic cells, respectively (Schäfer et al., 2007).

We investigated here the functional relationships of transcription factors expressed in the lateral floor plate (V3 interneurons, KA′) and KA′ cells. Our data show that nkx2.9 cooperates with nkx2.2a and nkx2.2b in the specification of V3 and KA′ cells. We provide evidence for distinct hierarchies of regulatory genes that are involved in the terminal differentiation of the different neuronal cell types. The transcription factor genes tal2, gata2 and gata3 are expressed in both KA′ and KA′ cells, even though they are relevant for the differentiation of GABAergic character in only one of these cell types. Thus, the regulatory relationships differ in the two cell types, suggesting that the regulatory architecture might be dependent on the lineage of the cells and not just on the expression of particular transcription factors.

The wild-type zebrafish were derived from an intercross between the AB line and the wtOX line. Fish were bred and embryos staged as described (Kimmel et al., 1995; Westerfield, 1993).

We carried out in situ hybridization as described (Oxtoby and Jowett, 1993), following, in the case of double stainings, the instructions of the suppliers of the reagents (Roche, PerkinElmer). For details of probes, see Table S1 in the supplementary material.

Counts of expressing cells were derived from the entire trunk and tail on both sides of the spinal cord. In colocalization studies, cells were counted on both sides over a distance of five somites in the spinal cord above the yolk extension.

The morpholinos (Table 1) were resuspended in 0.1% Phenol Red and injected at 0.5 mM (single and double injection) or 0.25 mM (triple injection). As the penetrance of the effects on GABA immunoreactivity and gad67 expression reached almost 100%, we believe that the degree of knockdown is sufficient. RT-PCR showed that the gata2 splice morpholino efficiently blocks splicing of gata2 mRNA at 48 hours post-fertilization (hpf).

The ventral spinal cord of the zebrafish embryo is composed of a number of distinct neurons. In the lateral floor plate, GABAergic KA′ cells, V3 interneurons and progenitor cells (P_LF) are intermingled (Bernhardt et al., 1992; Lewis and Eisen, 2003; Schäfer et al., 2007) (Fig. 1A,B). At a slightly more dorsal position, motoneurons and progenitor cells (P_MN) are found (Fig. 1A,B). The latter give rise to the GABAergic KA′ cells and the ventral longitudinal descending interneurons (VeLD) (Bernhardt et al., 1992; Lewis and Eisen, 2003) (Fig. 1A,B).

The homeobox genes nkx2.2a, nkx2.2b and nkx2.9 share high sequence similarity and are expressed in the ventral neural tube in overlapping patterns (Barth and Wilson, 1995; Schäfer et al., 2005; Schäfer et al., 2007; Xu et al., 2006). Duplicated genes in the zebrafish genome have frequently diversified in expression, adopting specialized functions in addition to sharing redundant roles (Hadzhiev et al., 2007; Yan et al., 2005). We first compared the pattern of expression by analyzing stage-matched embryos hybridized to probes complementary to the three Nkx2 mRNAs.

In the brain of the 24-hour-old embryo, nkx2.2a was expressed in the hypothalamus, the regions flanking the zona limitans intrathalamica (prethalamus, thalamus) and the tegmentum of the midbrain (Fig. 2A). By contrast, nkx2.2b and nkx2.9 showed a gap in the diencephalon, being expressed more in thalamic and prethalamic areas, respectively (Fig. 2D,G). All three genes were transcribed in the lateral floor plate of the hindbrain and spinal cord (Fig. 2B,C,E,F,H,I; data not shown). Whereas nkx2.2b was expressed in a continuous band in the lateral floor plate (Fig. 2E), nkx2.9 transcripts were detectable at high levels in individual cells or in clusters of two to three cells separated by cells with no or very low expression (Fig. 2H). Similarly, nkx2.2a was expressed at different levels in the cells of the lateral floor plate, even though gaps expressed the gene at low, but always detectable, levels (Fig. 2B, arrowhead). Thus, the three genes have overlapping, but not identical, patterns of expression. The medial floor plate did not express any of the three genes (Fig. 2C,F,I), consistent with previous findings (Barth and Wilson, 1995; Guner and Karlstrom, 2007; Schäfer et al., 2005; Xu et al., 2006).

In mammals, Nkx2.2 plays a crucial role in the development of V3 interneurons (Briscoe et al., 1999). However, previous knockdown experiments in zebrafish (Schäfer et al., 2007), in which the translation of nkx2.2a and nkx2.2b was blocked, did not affect expression of the bHLH transcription factor tal2, a neuronal marker (Pinheiro et al., 2004; Schäfer et al., 2007).

We tested whether nkx2.9 is required for the expression of tal2 in the lateral floor plate by injecting an antisense morpholino (Mo-nkx2.9) directed to the translation start site (Table 1). Knockdown of nkx2.9 translation resulted in ∼50% reduction of tal2-positive cells in the lateral floor plate (Fig. 3A,B; Fig. 4). More dorsally located cells expressing tal2 were not affected, showing that this effect is specific for the lateral floor plate (Fig. 3A,B). When Mo-nkx2.9 was co-injected together with a morpholino directed against either nkx2.2a (Mo-nkx2.2a) or nkx2.2b (Mo-nkx2.2b) mRNA, a further reduction of lateral floor plate, but not dorsal, tal2-expressing cells was noted (Fig. 3C-F; Fig. 4). This indicates that the two closely related nkx2.2 genes also play a role in the specification of tal2-expressing cells. This result appears to contradict a previous report, in which morpholinos directed against nkx2.2a and nkx2.2b mRNA did not result in a reduction of tal2-positive cells (Schäfer et al., 2007). We repeated these experiments (Fig. 3G,H; Fig. 4) and found that double knockdown of nkx2.2a and nkx2.2b indeed does not lead to loss of tal2-positive cells. When all three Nkx2 genes were knocked down, tal2-expressing cells were completely lost in the lateral floor plate, whereas more dorsally located cells were unaffected (Fig. 3I,J; Fig. 4). This shows that nkx2.2a, nkx2.2b and nkx2.9 are required for tal2 expression in the lateral floor plate. Moreover, these results indicate a partially redundant function of the three Nkx2 genes: nkx2.9 function appears to be more crucial, whereas the other two genes contribute to the development of tal2-positive interneurons; however, this function is only detectable when nkx2.9 is knocked down.

The specificity of the knockdowns was controlled by injection of morpholinos carrying five mismatches (Table 1). Neither individual nor combined injection of these control morpholinos abolished the formation of tal2-positive cells (Fig. 3A,C,E,G,I; Fig. 4). The specificity of the effect is further underscored by the observation that the co-injection of Mo-nkx2.2a and Mo-nkx2.2b did not cause an effect and that the effect of Mo-nkx2.9 only became fully penetrant when the other two morpholinos were co-injected.

We next examined whether the lateral floor plate cells would be transformed into more dorsal cells in triple-knockdown embryos. The expression of the homeo/paired box transcription factors pax6.1 and pax6.2 (also known as pax6a and pax6b, respectively – Zebrafish Information Network) is detectable immediately dorsal to the lateral floor plate (see Fig. S1G,I in the supplementary material). The expression domains of pax6.1 and pax6.2 were not expanded ventrally into the lateral floor plate in triple-knockdown embryos (see Fig. S1G-J in the supplementary material). By contrast, when triple-knockdown embryos were stained with antisense probe directed against the bHLH factor olig2 mRNA, we noted an expansion of the olig2 domain of expression into the lateral floor plate (see Fig. S1K,L in the supplementary material). Thus, the Nkx2 genes appear to be required to suppress the expression of the motoneuron and oligodendrocyte marker olig2 in the lateral floor plate.

We next assessed whether knockdown of the three Nkx2 genes abolishes the expression of other neuronal marker genes expressed in the lateral floor plate. The mRNA of the zinc-finger transcription factor Gata2 can be detected in KA′ cells in the lateral floor plate and in V2b/VeLD and KA′ cells in more dorsal regions (Batista et al., 2008; Detrich et al., 1995) (see Fig. 6). Triple-knockdown embryos (24 hpf) did not express gata2 in the lateral floor plate (control, 46±2 gata2-positive cells, n=10 embryos; triple-knockdown embryos, 1±1 gata2-positive cell, n=10 embryos). More dorsally located cells expressing the gata2 gene were unaffected by the knockdown of the Nkx2 genes (Fig. 5D; for lateral views see Fig. S1A,B in the supplementary material). Injection of a cocktail of the three mismatch morpholinos did not affect the pattern of gata2 expression in the spinal cord (Fig. 5A).

The leucine zipper/PAS transcription factor gene single-minded homolog 1 (sim1) is expressed in scattered cells in the lateral floor plate of zebrafish embryos (Schäfer et al., 2007) and is a marker for V3 interneurons in the murine spinal cord (Fan et al., 1996). In contrast to that of tal2 and gata2, sim1 expression is not detectable until the second day of development, and so embryos were analyzed at 48 hpf. sim1 mRNA was not detectable (Fig. 5B,E; for a lateral view see Fig. S1C,D in the supplementary material) in triple-knockdown embryos, and injection of the mismatch morpholinos did not affect the pattern of expression (control, 82±10 sim1-positive cells, n=16 embryos; triple-knockdown embryos, 1±1 sim1-positive cell, n=12 embryos). sim1 expression was not affected in nkx2.2a/nkx2.2b double-knockdown or nkx2.9 single-knockdown embryos (see Fig. S2A,B in the supplementary material; data not shown). This suggests that the three Nkx2 genes are required for the specification of sim1-expressing cells. In the mouse spinal cord, V3 interneurons are excitatory and express the vesicular glutamate transporter Vglut2.1 (Slc17a6 – Mouse Genome Informatics) (Zhang et al., 2008). To assess whether sim1 cells correspond to zebrafish V3 interneurons, we carried out co-expression studies at 36 hpf. From 88 sim1-expressing cells (n=6 embryos), 72 cells expressed vglut2.1 at 36 hpf (see Fig. S3A-C in the supplementary material). Thus sim1-expressing cells appear to correspond mostly to V3 interneurons. Also, the expression of vglut2.1 in the lateral floor plate was abolished in Nkx2 triple-knockdown embryos (see Fig. S2E,F in the supplementary material).

In addition to sim1-expressing cells, the lateral floor plate is also the origin of KA′ interneurons, which are characterized by synthesis of the neurotransmitter GABA (Bernhardt et al., 1992; Martin et al., 1998). GABA immunoreactivity was depleted in the lateral floor plate in triple-knockdown embryos (triple-knockdown embryos, 2±2 GABA-positive cells, n=10 embryos; Fig. 5C,F; see Fig. S1E,F in the supplementary material). Embryos injected with the mixture of control morpholinos showed a normal pattern of GABA-synthesizing cells (control, 54±8 GABA-positive cells, n=17 embryos). Hence, Nkx2 genes are required for the differentiation of GABAergic KA′ neurons.

In contrast to nkx2.2b, expression of nkx2.9 in the lateral floor plate is not continuous but is interrupted by cells that express nkx2.9 at very low levels or not at all (Fig. 2H). Therefore, we examined whether nkx2.9-positive cells co-express tal2 mRNA by double in situ hybridization with nkx2.9 and tal2 probes. In two-thirds of nkx2.9-expressing cells, we also detected expression of tal2 mRNA (66%, 111 cells, n=7 embryos; Fig. 6A-C).

We next assessed whether the tal2-expressing cells were GABAergic by staining embryos (24 hpf) hybridized to the tal2 antisense probe with an anti-GABA antibody. Not only the ventral tal2-positive cells in the lateral floor plate, but also the more dorsally located tal2-positive cells synthesized GABA (100%, 101 cells, n=10 embryos; Fig. 5D-F). Thus, tal2-expressing cells in the lateral floor plate are identical to the GABAergic KA′ cells. The more dorsally located tal2/GABA-positive cells correspond to KA′ cells (Batista et al., 2008; Park et al., 2004). This was confirmed by transverse sectioning of tal2-stained embryos: more dorsally located tal2-positive cells were in contact with the ventricle, in agreement with their identity as KA′ cells (data not shown).

Next, we mapped the expression of the zinc-finger transcription factor gene gata2 relative to tal2-expressing cells. Almost all gata2-expressing cells in the lateral floor plate also expressed tal2 mRNA (94%, 94 cells examined, n=5 embryos), whereas only a fraction of gata2-expressing cells in more dorsal aspects of the spinal cord co-labeled with tal2 mRNA (Fig. 6G-I). gata2 expression was lower in the row 2 cells corresponding to KA′ cells than in the cells of the lateral floor plate. As with gata2, mRNA of the related factor gata3 was co-expressed in the tal2-expressing cells in the lateral floor plate (96%, 85 cells, n=5 embryos) and also in a number of cells at a more dorsal location in the spinal cord (Fig. 6J-L). Since the tal2-positive cells also expressed GABA (Fig. 6D-F), this suggests that gata3 and tal2 are co-expressed in both the ventral KA′ cells and the more dorsal KA′ cells.

We next examined the relationship between sim1-expressing V3 cells and tal2 expression. Only a quarter of sim1-positive cells expressed tal2 in double-labeled 36 hpf embryos (28 of 113 cells, n=8 embryos; Fig. 6M-O). Thus, GABAergic KA′ and V3 interneurons only partly share marker expression in the 36 hpf embryo.

We next investigated the epistatic relationships between the different factors expressed in the GABAergic cells in the lateral floor plate. As a marker for these cells, we used the expression of the GABA-synthesizing enzyme glutamic acid decarboxylase 67 gene (gad67; also known as gad1 – Zebrafish Information Network). First, we verified that expression of gad67 in the lateral floor plate is also dependent on the activity of Nkx2 genes. Triple knockdown of nkx2.2a. nkx2.2b and nkx2.9 abolished expression of gad67 in the lateral floor plate completely (n=16 embryos; Fig. 7A,B), confirming the immunohistochemical results (Fig. 5C,F).

Since tal2 is expressed in cells that produce GABA, we tested whether tal2 is required for expression of gad67. tal2 knockdown embryos lacked expression of gad67 in the lateral floor plate, whereas more dorsally located gad67-expressing cells were unaffected by the knockdown of tal2 (n=18 embryos; Fig. 7C,D). Control embryos that were injected with a five-mismatch morpholino showed normal gad67 expression. This suggests that tal2 acts upstream of gad67 in KA′ cells of the lateral floor plate but seems to be dispensable for gad67 expression in KA′ cells.

The bHLH factor Olig2 controls motoneuron and oligodendrocyte development (Park et al., 2004). To assess whether olig2 is required for the development of gad67-positive cells, olig2 translation was knocked down by injection of a previously employed morpholino (Zannino and Appel, 2009). Knockdown of olig2 abolished gad67 expression in KA′ cells, but gad67 expression was still present in the lateral floor plate (n=15 embryos; Fig. 7E,F). We also tested the effect of olig2 knockdown on tal2 expression. As with gad67, expression of tal2 was abolished in dorsally located KA′ cells but not in the KA′ cells of the lateral floor plate (data not shown). This suggests that tal2 expression and/or KA′ cell differentiation are under the control of olig2. tal2 activity is, however, not required for the expression of gad67 in KA′ cells. Injection of the mismatch control morpholino did not affect gad67 or tal2 expression (Fig. 7E; data not shown). Interestingly, knockdown of olig2 abolished gata3 expression in KA′ cells but did not affect the expression in KA′ cells nor presumably VeLD/V2b cells (see Fig. S2K,L in the supplementary material). When the tal2 and olig2 morpholinos were injected together, gad67 expression at the location of both KA′ and KA′ cells was abolished (n=25 embryos; Fig. 7G,H; see Fig. S2I,J in the supplementary material). Taken together, these data strongly suggest that gad67 expression and/or the differentiation of KA′ and KA′ cells are specified by distinct mechanisms.

The triple knockdown of the Nkx2 genes suggested that they are required for gata2 expression in the lateral floor plate. Since gata3 is expressed in a pattern overlapping with that of gata2 (Fig. 6G-L), we tested whether gata3 is also dependent on the Nkx2 genes. Triple-knockdown embryos lacked gata3 expression in the lateral floor plate, whereas expression in more dorsal aspects in KA′ and VeLD/V2b cells was not affected (n=18 embryos; Fig. 8A,B). Thus, gata3 expression, like that of gata2, depends on the Nkx2 genes in lateral floor plate cells.

Since tal2 expression overlaps with that of gata2 and gata3, we tested whether tal2 is required for their expression (Fig. 8C-F). Knockdown of tal2 abolished neither gata3 (Fig. 8C,D) nor gata2 expression in the lateral floor plate (Fig. 8E,F), even though the same injection led to loss of gad67 expression (Fig. 7C,D; data not shown). Thus, tal2 could act downstream of gata2 and gata3 or in parallel pathways.

We also analyzed whether sim1 expression depends on tal2. sim1 expression was not affected in tal2 morphants (data not shown), suggesting that tal2 is not required for sim1 expression, even though tal2 expression is detectable in a fraction of sim1-positive cells at 36 hpf.

Since knockdown of tal2 did not have an effect on gata2 and gata3 expression, we hypothesized that gata2 and gata3 act upstream of tal2. To test this, we injected morpholinos against gata2 and stained embryos with tal2 or gad67 probes. Knockdown of gata2 abolished expression of tal2 in the lateral floor plate, whereas tal2 expression was unaffected in more dorsal positions corresponding to KA′ and V2b/VeLD cells (n=20 embryos; Fig. 9A,B; see Fig. S2G,H in the supplementary material). Knockdown of gata2 abolished gad67 expression in the lateral floor plate in the same manner as it abolished tal2 expression (n=15 embryos; Fig. 9C,D). Expression of gad67 in KA′ cells was unaffected, as was tal2 expression, in the gata2 morphants (Fig. 9C,D). These results are in line with the notion that gata2 acts upstream of tal2 and gad67 in KA′ cells. As gad67 expression depends also on tal2 function (Fig. 7C,D), this suggests that gata2 acts directly or indirectly through tal2 on gad67 expression. Interestingly, gata2 morphants also showed loss of gata3 expression in the lateral floorplate, suggesting that gata3 is directly or indirectly under the control of gata2 in KA′ cells (see Fig. S2M,N in the supplementary material).

Next, we analyzed the role of gata3 in the control of tal2 and gad67 transcription. In gata3 morphants, expression of tal2 and gad67 was abolished in KA′ cells, but their expression in KA′ cells was unaffected. Hence, gata2 and gata3 morphants present complementary patterns of activity: gata2 is required for the expression of gad67 and tal2 in, and/or for the differentiation of, KA′ cells, whereas gata3 is necessary for the control of gad67 and tal2 expression in, and/or for the differentiation of, KA′ cells. Thus, the related genes gata2 and gata3 do not act redundantly but have specific functions in the two cell types. tal2/gad67-expressing V2b/VeLD cells, which are distinguished from KA′ cells by their pial location and characteristic cell shape, are not affected by the knockdown of gata3 (data not shown).

Since V3 interneurons depend on Nkx2 gene function (Fig. 5B,E), we tested whether gata2 or gata3 could be employed in the specification of these cells. However, 48-hour-old gata2 and gata3 morphants formed sim1- and vglut2.1-expressing cells normally (see Fig. S2C,D in the supplementary material; data not shown), suggesting that gata2 and gata3 are not required. Thus, the regulatory mechanisms controlling V3 interneuron differentiation downstream of Nkx2 genes differ from those of KA′ cells.

Differentiation of V3 and KA′ interneurons in the lateral floor plate of the zebrafish spinal cord relies on nkx2.9 and on the two related genes nkx2.2a and nkx2.2b. In the mouse, Nkx2.2 plays a crucial role in the specification of V3 interneurons (Briscoe et al., 1999). However, simultaneous knockdown of nkx2.2a and nkx2.2b in the zebrafish does not abolish the differentiation of V3 interneurons nor of the KA′ cells marked by tal2 and gad67 expression (Schäfer et al., 2007) (this study). Only when nkx2.2a and nkx2.2b were knocked down together with nkx2.9 did we observed a complete loss of V3 and KA′ interneurons. By contrast, knockout of Nkx2.9 in the mouse does not lead to loss of the P3/V3 compartment of the spinal cord (Pabst et al., 2003): Nkx2.9^–/– mice have a rather mild phenotype, with defects in the spinal accessory nerve and with lower penetrance in the vagal and glossopharyngeal nerves (Pabst et al., 2003). Whereas Nkx2.2 is essential for the differentiation of the P3/V3 compartment in the spinal cord of the mouse, nkx2.9 plays a crucial role in this differentiation process in the zebrafish, suggesting that the relative importance of nkx2.2 and nkx2.9 has changed during evolution. This might reflect an independent drift of function in the two vertebrate lineages upon duplication of a common ancestral gene (Hadzhiev et al., 2007; Yan et al., 2005).

Although there are neurons contacting the cerebrospinal fluid in the mammalian CNS (Stoeckel et al., 2003), it is not clear whether the mammalian spinal cord has KA cells. Tal2 is not expressed in the mouse spinal cord (Pinheiro et al., 2004). KA cells might thus be associated with the specification of the neuronal network characteristic of anamniotes that underlies the swimming movement of their free-living embryos and larvae (Wyart et al., 2009). Interestingly, V3 neurons in the mammalian spinal cord are involved in the control of the regularity and robustness of locomotor rhythms during walking (Zhang et al., 2008).

The nkx2.2a, nkx2.2b and nkx2.9 genes act in a partially non-redundant manner. Knockdown of nkx2.9 resulted in a 50% reduction in tal2-positive cells, suggesting that nkx2.2a and nkx2.2b cannot compensate totally for the loss of nkx2.9 function. Although sim1- and tal2-expressing cells in the lateral floor plate were unaffected, Schäfer et al. noted that nkx2.2b/foxa2-positive cells do not form in nkx2.2a/nkx2.2b double morphants (Schäfer et al., 2007). This also suggests specialized functions of the three genes in the lateral floor plate. In this context, it might be of note that not all lateral floor plate cells express nkx2.9 and nkx2.2a with equal intensity. Expression of nkx2.2b, however, appears to be present in all cells at 24 hpf in line with a specific function that is different from those of the other two Nkx2 genes. Schäfer et al. suggested that nkx2.2b/foxa2 co-expressing cells continue to proliferate (Schäfer et al., 2007). By contrast, nkx2.9-positive cells might be post-mitotic precursors that differentiate into tal2- and sim1-expressing neurons. Consistent with this notion, tal2-expressing cells do not express foxa2 and have exited the cell cycle (Schäfer et al., 2007).

The expression of nkx2.2a, nkx2.2b and nkx2.9 is dependent on Hh signals emitted from the adjacent medial floor plate and from the underlying notochord (Barth and Wilson, 1995; Guner and Karlstrom, 2007; Schäfer et al., 2005; Schäfer et al., 2007; Xu et al., 2006). nkx2.9 expression is driven by a conserved Shh-dependent enhancer, which binds the Hh transducer Gli, suggesting that nkx2.9 is a direct target of the Hh signaling cascade (Xu et al., 2006). There is also evidence that nkx2.2a in zebrafish and Nkx2.2 in mouse are direct targets of Hh signaling, with Gli binding sites present in their regulatory regions (Vokes et al., 2007; Xu et al., 2006).

Knockdown of nkx2.2a, nkx2.2b and nkx2.9 abolishes sim1, gata2, gata3, tal2 and gad67 expression in the lateral floor plate, suggesting that the Nkx2 genes act upstream of these genes (Fig. 10). Ectopic olig2 expression (see Fig. S1K,L in the supplementary material) and islet1 expression (L.Y., unpublished) was observed in Nkx2 triple-knockdown embryos, suggesting that lateral floor plate cells take up a motoneuronal fate. It remains to be determined whether any of the affected genes is a direct target of the Nkx2 genes. The observed effects might also be explained by loss of cell identity due to a ventral shift of dorsal cells. Important in this context, however, is the fact that we did not observe activation of pax6.1 or pax6.2 in the cells immediately adjacent to the medial floor plate.

Knockdown of gata2, but not gata3, abolished expression of tal2 and gad67 in the lateral floor plate, suggesting that gata2 acts upstream of tal2 and gad67 in KA′ cells (Fig. 10). This was confirmed by the observation that knockdown of tal2 did not affect gata2 or gata3 expression in the lateral floor plate. Removal of tal2 activity abolished, however, the expression of gad67 in the lateral floor plate, placing tal2 upstream of gad67 (Fig. 10). Since gata2, tal2 and gad67 are expressed in the same cells, these interactions could be direct. Indeed, the upstream sequence of the tal2 gene contains a cluster of Gata2 and Gata3 binding sites (L.Y., unpublished). We cannot exclude, however, additional mediators that act in parallel or in series with Gata2.

Like KA′ cells, V3 interneurons depend on the nkx2.2a, nkx2.2b and nkx2.9 genes. We noted co-expression of tal2 and sim1 in 25% of cells at 36 hpf, in agreement with previous findings (Schäfer et al., 2007). However, neither tal2, gata2 nor gata3 knockdown abolished sim1-expressing cells in the 48-hour-old spinal cord (L.Y., unpublished). Although both V3 and KA′ cells depend on the three Nkx2 genes, the two cell types employ different downstream regulators for further differentiation (Fig. 10). It remains to be elucidated whether the minor fraction of cells (25%) that co-express sim1 and tal2 in the lateral floor plate at 36 hpf represent a transitory state in the switch from one differentiation program to the other or an as yet uncharacterized distinct cell type.

The ventral half of the zebrafish spinal cord contains two other GABAergic inhibitory interneuron classes: VeLD/V2b and KA′ (Bernhardt et al., 1992; Park et al., 2004; Batista et al., 2008). The KA′ cells, which are located in the immediate vicinity of the motoneurons, depend on olig2 function and express gata2, gata3, tal2 and gad67 (Batista et al., 2008) (Fig. 10). As shown by analysis of olig2:gfp transgene expression, KA′ interneurons appear to be derived from progenitors in the motoneuron domain (Park et al., 2004). Differentiation of GABAergic KA′ cells requires gata3 function. However, knockdown of gata3 does not seem to abolish gad67 in VeLD/V2b cells (L.Y., unpublished), suggesting that the functional connections of the regulatory genes differ in KA′ and VeLD/V2b cells. gata3 expression is dependent on olig2 in KA′ cells, indicating that olig2 acts upstream of gata3 on gad67 expression in KA′ cells (Fig. 10).

Whereas knockdown of olig2 abolished KA′ cells, ventrally located KA′ cells were slightly increased in abundance. This is in agreement with previous findings that showed that the olig2-dependent motoneuron domain produces the Notch ligand jagged 2, which maintains the proliferating precursors in the lateral floor plate cells via activation of Notch5 (Notch3 – Zebrafish Information Network) (Yeo and Chitnis, 2007).

Although gata2 and tal2 are both expressed in KA′ neurons, their knockdown does not affect the differentiation of these neurons, suggesting that the two genes are not required for gad67 expression in KA′ cells. This is in striking contrast to KA′ cells, in which gata2 and tal2 are instrumental for gad67 expression and/or cell differentiation. KA′ and KA′ cells also differ with respect to their dependence on olig2 expression and their origin in the spinal cord: whereas KA′ cells are derivatives of the lateral floor plate, the KA′ cells originate from the motoneuron compartment (Park et al., 2004). Moreover, for gad67 expression and/or the differentiation of KA′ cells, gata3 and olig2, but not tal2 and gata2, are required, even though the latter are expressed in KA′ cells. Conversely, in KA′ cells, gata2 and tal2 are responsible for gad67 expression (Fig. 10). Our data imply that gad67 relies on different regulatory elements for its expression in KA′ versus KA′ cells. Our data also suggest that the cis-regulatory elements that mediate activation via the Gata2/Tal2 pathway must be silenced in KA′ cells, otherwise the Tal2/Gata2 and Olig2/Gata3 pairs would act redundantly. This implies that the lineage and epigenetic history, and not the expression state of transcription factors, determine which regulatory network controls expression of the differentiation marker gad67.

We thank N. Borel and her fish house team, M. Rastegar for help with microscopy, T. Dickmeis and O. Armant for comments and the E. Davidson lab. for BioTapestry. This work was supported by the Helmholtz Association and the European Commission (ZF-MODELS, EUTRACC and NeuroXsys).