Establishing correct dorsal/ventral patterning in spinal cord development is a key process in producing a functional spinal cord. Discrete domains of spinal progenitors express a code of transcription factors, and then produce moto-, inter-, and sensory neurons. Studies in a variety of vertebrate species have described the extrinsic and intrinsic cues necessary for establishing the progenitor domains. For example, Dbx genes encode a family of homeodomain transcription factors that define an intermediate spinal progenitor domain (Lu et al.,1992). This gene family has multiple functions in spinal cord development. In Xenopus, Xdbx inhibits neurogenesis by regulating Xash3 expression at neural plate stages (Gershon et al.,2000), whereas in mouse and chick, studies have shown that Dbx1/2 are necessary in spinal cord development for the production of V0/V1 interneurons (Pierani et al.,2001), radial glia, astrocytes, and oligodendrocytes (Fogarty et al.,2005).
Multiple secreted signaling pathways establish and maintain gene expression profiles of spinal cord progenitor domains (Poh et al.,2002; Chesnutt et al.,2004). The effects of Hedgehog and retinoic acid (RA) signaling in regulating Dbx1/2 expression have been studied in mouse and chick (Pierani et al.,1999; Briscoe et al.,2001; Wijgerde et al.,2002; Novitch et al.,2003). Hedgehog is required for patterning the ventral spinal cord by either activating or repressing target genes by means of the Gli transcription factor family (Jacob and Briscoe,2003). Shh−/− mouse embryos at early stages of development exhibit ventrally expanded Dbx1 expression (Wijgerde et al.,2002), while later in development, Dbx1 and Dbx2 expression are localized to the ventral midline of the spinal cord (Pierani et al.,1999; Wijgerde et al.,2002). These studies indicated that the effect of reducing Shh activity changes during development of the spinal cord and that the ventral midline expression of Dbx1 and Dbx2 might be indirectly caused by the ventral expansion of more dorsal progenitor domains. Additionally, it has been suggested that low levels of Hedgehog induce Dbx expression in intermediate domains and high levels of Hedgehog repress Dbx more ventrally (Briscoe et al.,2001; Wijgerde et al.,2002). However, Hedgehog signaling does not appear to be absolutely required for Dbx expression, because cells lacking Smoothened function are still capable of expressing these genes, albeit in ectopic locations (Wijgerde et al.,2002). Indirect regulation of Dbx expression may occur by a secondary mechanism by which Class I and Class II homeodomain genes cross-repress each other to refine the borders between different domains. The Nkx6.2 and Nkx6.1 transcription factors have been shown to repress Dbx1 and Dbx2, respectively (Vallstedt et al.,2001). However, the expression of Nkx6.1/6.2 has not been examined in Dbx mouse mutants, leaving open the possibility of reciprocal repression between the two gene families.
It has also been suggested that RA signaling is required for regulating Dbx expression in the intermediate spinal cord. RA is secreted from the somitic mesoderm, neural tube cells, and notochord (Solomin et al.,1998; Berggren et al.,1999; Swindell et al.,1999; Molotkova et al.,2005; Maden,2006). Addition of RA to embryonic day (E) 10 neural explants induces the expression of Dbx genes (Pierani et al.,1999; Novitch et al.,2003), and blocking RA signaling from the somitic mesoderm decreases the number of Dbx-positive cells induced in vitro (Pierani et al.,1999). Furthermore, these studies established that the effects of Hedgehog and RA signaling on Dbx gene expression are independent. However, the endogenous source of RA in this process is not known, and it is not clear whether RA acts directly to induce Dbx expression or to counteract other signals, such as bone morphogenetic protein (BMP), that normally act to inhibit Dbx genes (Pierani et al.,1999; Novitch et al.,2003).
The main function of spinal progenitors is to produce various classes of postmitotic neurons. Lineage tracing in mouse reveals that Dbx1-positive cells predominantly give rise to Evx1/2+ interneurons in the V0 domain (Pierani et al.,2001). Overexpression of Dbx1 by means of electroporation causes an increase in Evx1/2+ cells with a concomitant loss of En1+ neurons and no effect on the production of other ventral neurons (Pierani et al.,2001). The opposite phenotype is observed in the Dbx1 mouse knockout, where there is a loss of Evx1/2+ interneurons with an expansion of En1+ neurons, indicating a V0 to V1 switch (Pierani et al.,2001). Further evidence suggests that Dbx2-positive progenitors give rise to En1+ neurons in the V1 domain (Pierani et al.,1999,2001). While this pathway of neuronal fate specification is clearly conserved between amniotes, it is not known whether the conservation extends to other vertebrates.
In zebrafish, three genes comprise the Dbx family: dbx1a, dbx1b, and dbx2 (Seo et al.,1999). Phylogenetic analysis suggests that dbx1a and dbx1b represent duplicate orthologs of the amniote Dbx1 gene. The zebrafish dbx2 gene is related, although not definitively orthologous, to amniote Dbx2 (Seo et al.,1999). Overexpression of dbx1a causes malformation of brain ventricles, resulting in a fused brain and defects in neuron clusters and axon projections in the forebrain (Hjorth et al.,2002). Knockdown of dbx1a by morpholino injection leads to altered hindbrain morphogenesis but does not affect patterning of this structure. Whereas these studies collectively reveal varied functions for Dbx genes throughout the central nervous system, detailed characterization of the establishment, regulation, maintenance, and progeny produced by Dbx-positive cells in the zebrafish spinal cord has not been described.
In the current study, we investigate Dbx regulation and function in zebrafish. First, we examine the expression of all three Dbx genes at multiple developmental stages and determine their spatial relationship with each other, as well as additional homeodomain genes. We show that blocking Hedgehog signaling causes a ventral expansion of all three Dbx genes and an increased number of two subclasses of interneurons, whereas RA signaling is necessary only for the maintenance of dbx1a expression. Lastly, we examine the consequences of dbx1a/1b loss-of-function and demonstrate that the progenitor marker nkx6.2 is dorsally expanded into dbx1a/1b territory, resulting in a loss of interneurons produced from the dbx1a/1b domain. Together, our data suggest that Dbx gene expression and function are evolutionarily conserved between zebrafish and amniotes, and that neither Hedgehog nor RA signaling is required for induction of Dbx expression in vivo.
Characterization of Dbx Gene Expression in Zebrafish Spinal Progenitors
Expression of dbx1a in the zebrafish spinal cord has been reported to begin around 13 hours postfertilization (hpf; Fjose et al.,1994). We were interested in comparing expression patterns of dbx1a (Fig. 1A–E), dbx1b (Fig. 1F–J), and dbx2 (Fig. 1K–O) at multiple time points during spinal cord development. Although their expression domain occupies a constant two- to three-cell-diameter stripe in the intermediate spinal cord, there are differences in the timing of expression between Dbx genes. Only dbx1a and dbx1b are expressed at 15 hpf (Fig. 1A,F,K), while dbx2 expression does not begin until 24 hpf (Fig. 1M) and does not span the entire rostral/caudal axis until 28 hpf (data not shown). By 48 hpf, dbx1a (Fig. 1D) and dbx1b (Fig. 1I) are still present, while dbx2 (Fig. 1N) is no longer expressed. The perdurance of dbx1a/1b expression at this late stage of spinal cord development indicates that dbx1a/1b may contribute not only to the production of neurons but also glia as observed in late stages of mouse spinal cord development (Fogarty et al.,2005). One day later in development at 72 hpf, dbx1b (Fig. 1J) is the only family member expressed. These data suggest that, although dbx1a/1b/2 occupy similar territory in the spinal cord, their regulation during development is quite different. Furthermore, we noticed that dbx1a is the only family member to be expressed in presumptive postmitotic neurons based on position and morphology (Fig. 1B,C; Fjose et al.,1994), similar to results from mouse where Dbx1 expression persists in neurons for a brief time (Pierani et al.,1999).
Next, we examined the relationship of the dbx1a domain to other spinal progenitor genes. Double in situ hybridization shows that the dorsal boundary of dbx1a expression is adjacent to pax3 (Fig. 2A–A″) and pax7 (Fig. 2B–B″). We found that dbx1a completely overlaps with dbx1b (Fig. 2C–C″) at 24 hpf and is expressed within the dbx2 domain (Fig. 2D–D″) at 28 hpf. We also found that the ventral boundary of dbx1a is adjacent to nkx6.1 (Fig. 2E–E″) and nkx6.2 (Fig. 2F–F″) and that there are multiple cell diameters between dbx1a and olig2 (Fig. 2G–G″) and nkx2.2 (Fig. 2H–H″) expression. These results illustrate that spinal progenitor gene expression patterns, in relation to the Dbx domain, are largely conserved between zebrafish and amniotes.
Hedgehog Signaling Is Required for Spatial Localization, but Not for Induction of Dbx Gene Expression
We were interested in determining if Dbx expression required Hedgehog signaling for either its induction or maintenance. We treated embryos with cyclopamine to block Smoothened function (Cooper et al.,1998) beginning at either 50% epiboly (5 hpf) or 18 hpf and then performed in situ hybridization for multiple progenitor markers at 24 or 28 hpf. To determine the effectiveness of our cyclopamine treatments, we examined ptc1 expression (Fig. 3A–B′), which is a direct target of Hedgehog (Marigo and Tabin,1996). In cyclopamine-treated embryos, detectable ptc1 expression was abolished (Fig. 3A′,B′), indicating that we had effectively blocked all Hedgehog signaling (Goodrich and Scott,1998). As another control, we analyzed the expression of spinal cord progenitor markers that are down-regulated in zebrafish when Hedgehog signaling is blocked. Consistent with previous studies, expression of nkx6.1 (Fig. 3C,C′; Cheesman et al.,2004) and nkx6.2 (Fig. 3E,E′; Guner and Karlstrom,2007) were abolished in embryos treated with cyclopamine at 50% epiboly, whereas dimethyl sulfoxide (DMSO) -treated embryos displayed normal expression of these markers. Interestingly, when we applied cyclopamine at 18 hpf and fixed at 24 hpf, low levels of nkx6.1 (Fig. 3D,D′) and nkx6.2 (Fig. 3F,F′) were present, indicating that Hedgehog is necessary not only for induction but also for maintenance of their expression.
We then examined expression of Dbx genes after cyclopamine application to determine whether Hedgehog was required for their induction (Fig. 3G,G′,I,I′,K,K′). All DMSO-treated embryos showed wild-type expression of dbx1a/1b/2 (Fig. 3G,I,K) whereas cyclopamine-treated embryos displayed a ventral expansion of all three genes (Fig. 3G′,I′,K′). Therefore, Hedgehog signaling is dispensable for the induction of these genes but it also prevents their expression in the ventral spinal cord. Our results differ from a previous study, which shows a more subtle ventral expansion of dbx1a in the smoothened mutant zebrafish (Guner and Karlstrom,2007). This difference is likely due to the presence of maternally supplied Smoothened protein in the mutant, allowing for Hedgehog signaling to partially repress dbx1a from the most ventral extent of the spinal cord. In our study the cyclopamine treatments block both maternal and zygotic Smoothened function.
We next asked if Hedgehog was required for the maintenance of the Dbx genes by applying cyclopamine at 18 hpf after the genes have been induced in the spinal cord (Fig. 3H,H′,J,J′,L,L′). Although dbx2 is not expressed throughout the entire spinal cord at this time of development, it is present at low levels in the very rostral spinal cord segments. We found that all three Dbx genes are still expressed and are slightly ventrally expanded, but not to the extent observed in the 50% epiboly–24 hpf treatment. Therefore, we concluded that Hedgehog function is not required for Dbx maintenance, but functions at later stages to restrict expression from the ventral spinal cord.
We were also interested in determining the state of the dorsal boundary of Dbx gene expression after cyclopamine treatment. Previous work has shown that Msx1/2 homeodomain proteins can repress Dbx gene expression (Timmer et al.,2002). We examined expression of a homologous gene, msxc, in zebrafish and found its expression was unaffected by blocking Hedgehog signaling at 50% epiboly (Fig. 3M,M′). This result is consistent with the unaffected dorsal boundary of dbx1a/1b expression (Fig. 3G′,I′), but not with the apparent ventral shift of the dbx2 dorsal boundary (Fig. 3K′), suggesting that other factors can also repress expression of dbx2. Of interest, we did observe ventral expansion of pax3 (Fig. 3O,O′), but not pax7 (Fig. 3Q,Q′), into the dbx1a/1b domain resulting in overlap that is normally not present (see Fig. 2A–A″). These data differ from those presented previously, which show that pax3 expression was unaffected in smoothened mutants (Guner and Karlstrom,2007). When cyclopamine was added at 18 hpf, we did not observe a shift in msxc (Fig. 3N,N′), pax3 (Fig. 3P,P′), or pax7 (Fig. 3R,R′), indicating that Hedgehog is not required for maintaining the expression domains of any of these genes. Cumulatively, these results suggest that Dbx expression does not directly require Hedgehog signaling, and the ventral Dbx boundary is set indirectly by Hedgehog signaling while the dorsal limit is set by another signal such as BMPs.
Effects on Dbx Gene Expression Lead to Changes in the Specification of Postmitotic Neurons
The results from blocking Hedgehog signaling indicate a general dorsalization of the spinal cord as observed by a loss of nkx6.1/6.2 and ventral expansion of dbx1a/1b/2 and pax3. We expect that the expansion of the Dbx progenitor domain would also produce increases in the postmitotic neurons produced from this domain. Previous studies in mouse have shown that En1+ (V1) interneurons are produced from the Dbx2 domain (Pierani et al.,1999,2001) and Pax2.1+ (V0/V1) interneurons are produced from the Dbx1/2 domain (Burrill et al.,1997). In zebrafish, dbx1a/1b/2 are expressed in overlapping domains (see Fig. 2C–C″, D–D″), whereas Dbx1 and Dbx2 in mouse occupy separate domains at equivalent developmental time. Therefore, when we blocked Hedgehog signaling from 50% epiboly, we hypothesized that the number of en1b+ and pax2.1+ interneurons would be increased (Fig. 4). We counted individual cells expressing each marker in the same five spinal cord segments of multiple DMSO-treated embryos and compared them with cyclopamine-treated embryos. We found that the average number of en1b+ (Fig. 4A,C,E) and pax2.1+ (Fig. 4B,D,F) cells was increased from 5.67 (n = 12) to 8.75 (n = 12; P < 0.0001) and 5.80 (n = 15) to 8.60 (n = 15) (P < 0.0001), respectively. Of interest, the additional neurons appeared to be located near their normal location rather than expanded ventrally, indicating that Hedgehog-independent factors may regulate their final position, or that not all Dbx-positive progenitors are competent to produce these neurons. We nonetheless conclude from these results that the increased number of en1b+ and pax2.1+ interneurons correlates with the expansion of dbx1a/1b/2 expression.
One possible cause of an increase in postmitotic neurons is that blocking Hedgehog signaling could generally affect neurogenesis in the spinal cord. To assay for changes in neurogenesis, we used transgenic zebrafish expressing green fluorescent protein (GFP) under the control of the pan-neuronal huC promoter (Park et al.,2000). At 24 hpf, the number of GFP-positive neurons per 12-μm section was not significantly affected (P > 0.4) in cyclopamine-treated embryos (6.0 ± 0.3 [SEM], n = 12) compared with DMSO-treated controls (6.4 ± 0.5 [SEM], n = 12; Fig. 4G,H). Thus, en1b+ and pax2.1+ interneurons were expanded independently from the total number of neurons. These results support our conclusion that expansion of the dbx1a/1b/2 domain directly causes an increase in specific classes of postmitotic neurons. Furthermore, we conclude that the dbx1a/1b/2 domain in zebrafish may produce en1b+ and pax2.1+ interneurons as it does in mouse.
RA Signaling Is Not Required for Induction of Dbx Gene Expression
Next, we were interested in examining the influence of another important signal, RA, on the establishment of the Dbx domain. Previous data have suggested that RA is required for Dbx expression in other vertebrates (Pierani et al.,1999; Novitch et al.,2003). To block RA signaling in embryos, we used 4-(diethylamino)-benzaldehyde (DEAB), which has been shown to inhibit retinaldehyde dehydrogenase and subsequently the synthesis of RA (Perz-Edwards et al.,2001; Hamade et al.,2006; Kopinke et al.,2006). Embryos were treated with DEAB at 50% epiboly to test whether RA is required for the induction of Dbx expression. To determine the effectiveness of the DEAB treatment, we performed two analyses. After application of 50 μM DEAB in DMSO to wild-type and Tg(isl1:GFP)rw0 embryos, we performed in situ hybridization using krox20 (Fig. 5A,B) and imaged cranial nuclei expressing GFP (Fig. 5C,D). DEAB-treated embryos exhibited loss of krox20 expression in rhombomere 5 (Fig. 5B), and loss of the vagal nuclei in the hindbrain of Tg(isl1:GFP)rw0 transgenics (Fig. 5D). Both of these results are consistent with previous studies and indicate that we effectively inhibited RA signaling (Begemann et al.,2004; Maves and Kimmel,2005).
Next, we examined the expression of the Dbx genes after DEAB treatment (Fig. 5E–P′). Because RA has previously been shown to induce Dbx expression in neural explants, we analyzed expression at both 18 hpf and 24 hpf. At 18 hpf, dbx1a and dbx1b expression was unaffected in DEAB embryos (Fig. 5E,F,I,J), suggesting that RA signaling is not required for inducing expression of these genes. However, we found that dbx2 expression was prematurely expressed at 18 hpf (Fig. 5M,N′). When we analyzed embryos at 24 hpf, we observed lack of dbx1a expression in DEAB-treated embryos (Fig. 5F,F), indicating that RA signaling may be required for the maintenance of this gene. In contrast, dbx1b and dbx2 expression were present in DEAB-treated embryos (Fig. 5K,L,O,P), although dbx2 expression spanned the entire rostral/caudal axis of the spinal cord unlike in DMSO-treated embryos (not shown). Collectively, these results suggest that RA signaling does play some role in regulating Dbx expression but is not required for the induction of these genes in vivo.
Although DEAB has been suggested to provide the most complete block of RA signaling possible in zebrafish (Begemann et al.,2004), our data contrasted with a previous study in chick using pharmacological agents to block RA receptors (Pierani et al.,1999). Therefore, we treated embryos with the pan-RA receptor inhibitor BMS493, which has also been used previously in zebrafish to block RA signaling (Grandel et al.,2002; Begemann et al.,2004). Embryos were soaked in 10 μm BMS493, beginning at 50% epiboly and fixed at either 18 or 24 hpf, and we analyzed mRNA expression of all three Dbx family members. At 18 hpf and 24 hpf, the dorsal/ventral position of dbx1a, dbx1b, and dbx2 expression was unaffected, while at 24 hpf, dbx2 was prematurely expressed throughout the entire rostral/caudal axis of the embryo (not shown). We did not observe a loss of dbx1a expression at 24 hpf or premature expression of dbx2 at 18 hpf after BMS493 treatment, which we believe is due to a less complete block of RA signaling compared with DEAB (Begemann et al.,2004). Nevertheless, the similar results obtained with antagonists of both ligand and receptors support the conclusion that RA is not required for Dbx induction.
dbx1a and dbx1b Function Are Required for Specification of Intermediate Spinal Progenitors
The overlapping expression of dbx1a and dbx1b indicate that both genes may act redundantly in spinal cord development. We were interested in determining how the loss of dbx1a/1b would affect spinal cord patterning because Class I and Class II homeodomain proteins often cross-repress each other to sharpen the boundaries between adjacent progenitor domains. Prior research has shown that, in the Nkx6.2 knockout mouse, Dbx1 expands ventrally, suggesting that these genes may mutually repress each other (Vallstedt et al.,2001). However, it has not been shown in any system if the loss of Dbx1 causes the dorsal expansion of Nkx6.2. We hypothesized that, if nkx6.2 and dbx1a/1b were cross-repressive partners in zebrafish, we would observe a dorsal expansion of nkx6.2 in dbx1a/1b morphants.
To test the combined function of dbx1a and dbx1b, we simultaneously knocked down zygotic transcription with splice-blocking morpholino oligonucleotides for both genes (Fig. 6). Reverse transcriptase-polymerase chain reaction (RT-PCR) showed partial block of splicing at a concentration of 7 ng/nl for dbx1a (Fig. 6A) and 2.5 ng/nl for dbx1b (Fig. 6B). Due to an early requirement for dbx1a/1b, embryos did not survive past gastrulation when we injected higher doses of morpholino; therefore, we analyzed a hypomorphic phenotype in the double morphants. Compared with 24 hpf stage-matched wild-type embryos (Fig. 6C), dbx1a/1b morphants have a shorter rostral/caudal axis (Fig. 6D). We also observed a defect in hindbrain morphogenesis similar to Hjorth et al. (2002; data not shown). As a control for morpholino specificity, we injected each morpholino individually and compared the resulting phenotypes to double morphants.
In the spinal cord, we observed a dorsal expansion of nkx6.2 expression in dbx1a/1b morphants, confirming that dbx1a/1b function is necessary to repress nkx6.2 from intermediate domains (Fig. 6E,F). Single dbx1a or dbx1b morphants had no change in nkx6.2 expression (not shown), indicating that the dorsal shift of nkx6.2 is specific to the simultaneous loss of both genes produced by the morpholinos. In contrast, we found that msxc expression was unaffected in dbx1a/1b morphants (Fig. 6G,H), consistent with previous data indicating that Dbx genes are unable to repress Msx expression (Timmer et al.,2002). We also did not observe any defects in pax3 expression (not shown), suggesting that other factors such as BMP and Wnt signaling may restrict the expression of dorsal progenitor markers (Chesnutt et al.,2004; Ille et al.,2006; Zechner et al.,2006).
Next, we were interested in determining how the loss of dbx1a/1b function affected the production of postmitotic interneurons (Fig. 7). As a control, we performed in situ hybridization at 24 hpf for isl1, a marker of primary motoneurons in the ventral spinal cord. As expected, the mean number of isl+ neurons within somite segments 8–18 did not significantly differ (P > 0.5) between wild-type (22.90, n = 10) and dbx1a/1b morphants (22.60, n = 10; Fig. 7A–C). We also examined en1b+ and pax2.1+ interneurons, which we hypothesized differentiate from the dbx1a/1b domain (Thaeron et al.,2000). en1b+ interneurons were significantly decreased (P < 0.0001) in dbx1a/1b morphants (11.44, n = 9) compared with wild-type (16.90, n = 10; Fig. 7D–F). pax2.1+ interneurons were similarly decreased (P < 0.001) in dbx1a/1b morphants (16.89, n = 9) compared with wild-type (27.50, n = 8; Fig. 7G–I). At 24 hpf, there was no significant change in the number of en1b+ interneurons in single dbx1a or dbx1b morphants (data not shown), again indicating that the phenotype is specific to the simultaneous loss of both genes.
As with our cyclopamine treatments, we were interested in determining if changes in the number of en1b+ and pax2.1+ interneurons after dbx1a/1b morpholino injections were secondary to a general effect on neurogenesis. We did not observe a change in the number of Hu-expressing cells per 12-μm section (P > 0.5) in dbx1a/1b morphants (7.2±0.3 [SEM], n = 12) compared with wild-type (7.3 ± 0.4 [SEM], n = 12). This result, combined with the consistent number of isl+ neurons, suggest that the effects on en1b+ and pax2.1+ interneurons were a direct consequence of eliminating dbx1a/1b function in progenitors.
Properly establishing and maintaining the Dbx progenitor domain is important for spinal cord development, but the mechanisms underlying regulation of Dbx expression and the function of these genes has not been well studied in zebrafish. Here, we have characterized the expression of dbx1a, dbx1b, and dbx2 throughout spinal cord development (Fig. 1). We found that the spatial relationship of the Dbx genes to other known progenitor markers is well conserved in zebrafish compared with amniotes (Fig. 2). Our data indicate that blocking Hedgehog signaling does not affect the induction or maintenance of dbx1a/1b/2 expression, but instead leads to expansion of these genes into the ventral spinal cord (Fig. 3). The patterning defects when Hedgehog was blocked were evident not only in the progenitor domain but also in postmitotic neurons (Fig. 4). We observed an increase in en1b+ and pax2.1+ interneurons in cyclopamine-treated embryos, indicating that (1) the Dbx domain produces en1b+ and pax2.1+ interneurons and (2) by blocking Hedgehog signaling, we have dorsalized the spinal cord. We determined that blocking RA signaling does not affect the induction of dbx1a and dbx1b but causes premature expression of dbx2 (Fig. 5). Also, it is possible that RA is specifically important for the maintenance of dbx1a expression. Lastly, we investigated the developmental consequence of knocking down dbx1a/1b gene function (Fig. 6). We found ventral spinal cord patterning defects demonstrated by the dorsal expansion of nkx6.2 into the dbx1a/1b domain, while dorsal progenitor expression was unaffected. Also, we observed decreases in the number of subtype-specific interneurons believed to differentiate from the dbx1a/1b domain (Fig. 7). Together our data indicate an evolutionarily conserved role for Dbx genes in spinal cord patterning, and lead to an open question regarding the nature of the inductive signal for Dbx expression.
Conservation of Dbx Gene Expression in Zebrafish
We have determined that expression of the Dbx genes during spinal cord development remains consistent from their onset when the spinal cord is being formed to the time when they are no longer present. The dbx1a and dbx1b genes most likely represent duplicate orthologs of the single Dbx1 gene in mouse and chick (Seo et al.,1999). Zebrafish dbx2 is less conserved in terms of onset of expression, although the territory it occupies is quite similar to, although a bit wider than, dbx1a and dbx1b. In fact, sequence comparison between the three genes indicates dbx2 is the most divergent (Seo et al.,1999). Recent reports in zebrafish have shown that many of the spinal progenitor genes have conserved spatial relationships (Cheesman et al.,2004; Lewis et al.,2005; Guner and Karlstrom,2007). We have demonstrated that the relationship of Dbx expression to other progenitor genes is evolutionarily conserved in zebrafish, which we find intriguing given the size difference of the zebrafish spinal cord compared with amniotes.
Hedgehog Signaling Is Not Required for Dbx Gene Induction
Previous studies have offered conflicting evidence as to whether Dbx expression absolutely requires a Hedgehog signal for induction (Pierani et al.,1999; Briscoe et al.,2001). However, cell-autonomous loss of all Hedgehog signaling by removal of Smoothened function leads to a shift in position, but not complete absence, of Dbx genes (Wijgerde et al.,2002). This experiment suggests that Hedgehog acts to set the position of Dbx-positive progenitors, but is not absolutely required for expression of these genes. Our data support this model, as cyclopamine-treated zebrafish embryos exhibit a ventral expansion of Dbx genes but not loss of expression (Fig. 8A). Although our controls indicate that we have effectively blocked Hedgehog signaling, we cannot rule out the possibility that very low remaining levels of Hedgehog induce Dbx in the normal domain. Further studies in embryos lacking maternal and zygotic Smoothened function could help resolve this issue. Regardless, the data suggest that Hedgehog may act through Nkx6 genes to indirectly limit the ventral boundary of Dbx expression. The difference between our results and previous studies with respect to the dorsal boundary of dbx1a/1b expression may be explained by variability in the timing or extent of signals such as BMP and Wnt, which could act to exclude Dbx from more dorsal spinal regions. In the mouse, loss of Smoothened function may allow endogenous BMP or Wnt signals to induce genes such as Msx1/2 and, thus, inhibit Dbx expression. In contrast, BMP or Wnt activity may not penetrate as far ventrally in zebrafish embryos at 24 hpf, resulting in preservation of the dorsal dbx1a/1b boundary. Interestingly, the dorsal boundary of dbx2 expression was shifted ventrally after cyclopamine treatment, suggesting that regulation of this later-expressed gene may be differentially sensitive to dorsal signals. The ventral shift observed in pax3 expression makes it a candidate for dbx2 regulation, and intriguingly, simultaneous knockout of Pax3 and Pax7 in mouse produces a phenotype consistent with expanded Dbx2 expression (Mansouri and Gruss,1998).
RA May Act Indirectly in Regulating Dbx Expression
Previous literature shows that application of RA to nascent neural tube explants can induce Dbx expression, whereas antagonists of RA added to the culture decreased the ability of adjacent tissues to induce Dbx1/2-positive cells (Pierani et al.,1999). In addition, expression of dominant-negative RA receptors in spinal progenitors leads to cell autonomous reduction of Dbx expression (Novitch et al.,2003). In contrast, our data suggest that RA signaling is not required for the induction of Dbx genes in the zebrafish spinal cord. One possible explanation for these results is that RA acts indirectly to antagonize dorsal BMP or Wnt signals, which themselves limit Dbx expression. This model is supported by the previous explant studies, in which addition of RA simultaneously led to induction of Dbx and repression of Pax7 (Pierani et al.,1999). If these dorsal signals are not playing an identical role in the zebrafish spinal cord before 24 hpf, then RA signaling would be unnecessary for Dbx induction. Our results also indicate that blocking RA signaling can lead to premature dbx2 expression in the spinal cord. The mechanism underlying this premature induction is not clear, but it is possible that RA can regulate the temporal progression of progenitor gene expression in the spinal cord. The differential effects of blocking RA signaling on the Dbx genes could be due to divergent upstream regulatory elements or compensatory mechanisms within the Dbx family. Additional time-dependent studies are necessary to further explore these possibilities.
Dbx Function Is Important for Intermediate Spinal Cord Development
Loss-of-function studies have been performed for dbx1a function in hindbrain development (Hjorth et al.,2002) but function of zebrafish Dbx genes in the spinal cord had not been examined. Our dbx1a/1b loss-of-function results indicate that similar phenotypes occur in the zebrafish as compared with the Dbx1 mouse knockout (Pierani et al.,2001). We also observed a dorsal expansion of nkx6.2 into the territory normally occupied by dbx1a/1b. This finding is the first evidence that dbx1 antagonizes nkx6.2 expression, suggesting that these genes are cross-repressive partners (Fig. 8B). We also determined that, in the absence of dbx1a/1b, there is a concomitant loss of interneurons produced from this progenitor domain (Fig. 8B). It would be interesting in the future to determine whether there is redundancy or compensation for the loss of dbx1a/1b function by dbx2. Furthermore, our analysis was limited to the 24 hpf time point, because we were investigating the patterning and differentiation function of dbx1a/1b. Later analysis, at 48 hpf for example, would be useful to determine whether this domain also contributes to the production of glia as observed in mouse (Fogarty et al.,2005).
Dbx Expression May Represent a “Default” Fate for Spinal Progenitors
The Dbx expression domain lies in a region where multiple signaling pathways are converging. In the intermediate spinal cord, Hedgehog concentrations are quite low (Poh et al.,2002; Jacob and Briscoe,2003) and it is clear that these low levels are necessary to prevent Nkx6 genes from being expressed in these cells. Low levels of BMP are also important to position the dorsal boundary of Dbx expression (Timmer et al.,2002), and RA is required in an unknown way to allow Dbx expression, possibly by antagonizing these dorsal signals (Pierani et al.,1999; Novitch et al.,2003). Cumulatively, these data suggest that Dbx expression may occur in the absence of any dorsalizing or ventralizing signals and, thus, may represent a “default” state of spinal progenitors. Interestingly, Dbx1-expressing progenitors uniquely produce the three major cell types of the spinal cord: neurons, astrocytes, and oligodendrocytes, and thus represent the most multipotent progenitor population in the tissue (Fogarty et al.,2005). Additional work is necessary to determine whether any inductive molecules are required to induce Dbx expression in vivo, or if instead, it is simply sufficient to remove all Hedgehog, Wnt, and BMP signaling.
Fish Strains and Staging
Embryos were obtained from natural spawning of wild-type (AB*) or Tg(isl1:GFP)rw0 (Higashijima et al.,2000) zebrafish lines. All developmental stages in this study are reported in hours postfertilization at 28.5°C, according to Kimmel et al., (1995).
The dbx1a and dbx1b splice-blocking morpholino antisense oligonucleotides were obtained from Gene Tools, LLC. Morpholino sequences are as follows: dbx1a, 5′-AAACACTTACCTTTCTTTGGTGACG-3′; dbx1b, 5′-CTGATGACGATGCTATAAAAAAAT-3′. Both Dbx morpholinos were simultaneously injected into one-cell stage wild-type zebrafish embryos at concentrations of 7 ng/nl and 2.5 ng/nl for dbx1a and dbx1b, respectively.
Seventy-five wild-type embryos and dbx1a or dbx1b morphants at 24 hpf were used for preparing RNA. Total RNA was isolated using Trizol (Invitrogen) reagent and was reverse transcribed by random hexamers using the Superscript First Strand Synthesis Kit (Invitrogen) following the manufacturer's protocol. PCR was performed for 30 cycles using an annealing temperature of 57°C (dbx1a) and 60°C (dbx1b), and reactions were visualized on 1% agarose gels in TAE. The spliced product for dbx1a is 191 bp and unspliced product is 243 bp. The spliced product in dbx1b is 194 bp, and the unspliced product is approximately 1.7 kb.
In Situ Hybridization
Probe synthesis and in situ hybridization were performed as described elsewhere (Oxtoby and Jowett,1993). Single and double in situ hybridizations were carried out using digoxigenin- or fluorescein-labeled antisense RNA probes (Jowett,2001) and visualized using BM Purple (Roche) and Fast Red (Roche). The following RNA probes were made by our lab from DNA plasmids: dbx1a, dbx1b, dbx2, msxc, and nkx6.2. We were given the following DNA plasmids: en1b (Higashijima et al.,2004), krox20 (Oxtoby and Jowett,1993), nkx2.2 (Barth and Wilson,1995), nkx6.1 (Cheesman et al.,2004), pax2.1 (Krauss et al.,1991), pax3 (Seo et al.,1998), pax7 (Seo et al.,1998), ptc1 (Concordet et al.,1996), and olig2 (Park et al.,2002). For whole-mount photography after all staining methods, embryos were de-yolked and mounted laterally or cross-sectioned on a cryostat. Whole-mount and cryosectioned embryos were imaged on a compound microscope.
A 10 mM stock solution of cyclopamine (Toronto Research Chemicals) dissolved in DMSO (Fisher) was diluted to 100 μM in embryo medium. Embryos were soaked in cyclopamine beginning at either 50% epiboly (5 hpf) or 18 hpf and fixed at 24 hpf or 28 hpf overnight in 4% paraformaldehyde.
A 10-mM stock solution of DEAB (Sigma-Aldrich) dissolved in DMSO was diluted to 50 μM in embryo medium. Embryos were soaked in DEAB beginning at either 50% epiboly and fixed at 18 hpf, 24 hpf, or 28 hpf overnight in 4% paraformaldehyde.
A stock solution of BMS493 (gift from V. Prince) dissolved in DMSO was diluted to 10 μM in embryo medium. Embryos were soaked in DEAB beginning at 50% epiboly and fixed at either 18 hpf or 24 hpf overnight in 4% paraformaldehyde.
Analysis of Hu-Positive Neurons
Postmitotic neurons were counted in 12 nonconsecutive 12-μm cryosections, either by direct GFP fluorescence using Tg(elavl3:EGFP)zf8 embryos (Park et al.,2000), or by immunofluorescence with anti-HuC/D antibody (Molecular Probes) used at 1:1,000. Staining of transgenic embryos with the antibody showed complete co-labeling of all GFP-positive neurons.
We thank the Appel, Eisen, and Chandrasekhar labs for providing us with DNA plasmids, Victoria Prince for providing BMS493, the University of Utah zebrafish facility staff for animal husbandry, and J. Bonner for helpful discussions. R.I.D. is supported by NIH, and S.L.G. was a predoctoral trainee supported by National Institutes of Health Genetics Training Grant.