Visualization of two distinct classes of neurons by gad2 and zic1 promoter/enhancer elements in the dorsal hindbrain of developing zebrafish reveals neuronal connectivity related to the auditory and lateral line systems
Laboratory for Developmental Gene Regulation, RIKEN Brain Science Institute, Wako, Saitama, Japan
Neuroscience Center, University of North Carolina at Chapel Hill, 105 Mason Farm Road, Chapel Hill, NC 27599-7250
During development, diverse classes of neurons extend their axons to form synaptic connections with the correct targets. The optical transparency of zebrafish embryos and larvae makes it possible to visualize specific subsets of neurons in living animals by expressing green fluorescent proteins (GFP) under the control of a specific promoter/enhancer (Udvadia and Linney,2003). Recently, forward genetic screenings have been performed to study the development of particular neuronal cell types by taking advantage of transgenic zebrafish that express GFP in those neurons (Higashijima et al.,2000; Wada et al.,2005; Xiao et al.,2005). For example, the Islet1:GFP transgenic fish, in which the cranial motor neurons express GFP, has been used to investigate axon guidance and cell migration (Higashijima et al.,2000; Wada et al.,2005). Visualization of neuronal morphology is also particularly useful for classifying neuronal cell types according to their dendritic morphology and axonal projection patterns (Higashijima et al.,2004).
The alar plate of the vertebrate hindbrain receives all sensory modalities except for olfaction and vision. The dorsal-most part of the alar plate is primarily involved with processing sensory information from the inner ear and the lateral line system, although the latter is lost in terrestrial vertebrates (Butler and Hodos,1996). The sensory receptors in these systems are hair cells. Each type of sensory information enters adjacent but nonoverlapping regions of the dorsal hindbrain by means of the eighth nerve or the lateral line nerves (Piotrowski and Northcutt,1996; Raible and Kruse,2000). Various classes of interneurons have been described in this complex but phylogenetically conserved structure, based on their morphologies, connectivity, and physiological properties in adult animals from fish to mammals (Knudsen,1977; McCormick,1989; New et al.,1996; Cant and Benson,2003). However, the embryological origins and molecular profiles of these interneurons remain largely unknown. Recent studies in mice demonstrated the involvement of Wnt1 and Math1 (Atoh1) in the development of the auditory-related cochlear nuclei from a part of the lower rhombic lip, the dorsal-most portion of the hindbrain proliferative neuroepithelium (Wang et al.,2005; Farago et al.,2006). Although these studies revealed the origins and molecular profiles of the auditory nuclei, the fates of specific neuronal types remained obscure due to the difficulty in distinguishing them during development. The ability to visualize and identify specific classes of neuron in zebrafish should, therefore, be useful for providing such information.
The aim of this study was to characterize the different neuronal types in the dorsal hindbrain in terms of their molecular profiles and axonal projection patterns, and if possible, to relate the findings to those in adult animals. In this study, we identified two distinct classes of neurons in the dorsal hindbrain of developing zebrafish. We traced the projection patterns formed by these two classes of neurons by combining the Gal4VP16-UAS (UAS, upstream activating sequences) system (Köster and Fraser,2001b; Scheer and Camnos-Ortega,1999) with the transcriptional regulatory regions of gad2 (Martin et al.,1998) and zic1 (Grinblat et al.,1998; Nakata et al.,1998; Rohr et al.,1999; Köster and Fraser,2001a; Aruga et al.,2002). We found that the Gad1/2(+) neurons, visualized using the gad2 promoter/enhancer, projected to the contralateral hindbrain, while the zn-5(+)Lhx2/9(+) neurons, visualized using the zic1 promoter/enhancer, projected to the contralateral midbrain torus semicircularis (TS). Furthermore, we demonstrated that the basic helix–loop–helix (bHLH) factor Atoh1a is required for the development of zn-5(+)Lhx2/9(+) neurons but not for Gad1/2(+) neurons, revealing that it is specifically required for a particular class of projection neurons. From these projection patterns and cell positions suitable for receiving inputs from the periphery, we suggest that these two classes of neurons correspond to the projection neurons in the octaval and octavolateral nuclei described in adult fish and amphibians (Knudsen,1977; McCormick,1989; Will,1989; New et al.,1996; Edwards and Kelley,2001; Yamamoto and Ito,2005) and that they are involved in mediating hearing and lateral line information. These results reveal details about the development of neuronal circuits in the fish dorsal hindbrain and provide indications as to which aspects show phylogenetic conservation or divergence with respect to similar systems in mammals.
Two Distinct Classes of Neurons in the Dorsal Hindbrain Based on Molecular Profiles and Dependence on the bHLH Factor Atoh1a
As a first step toward characterizing the types of neuron in the dorsal hindbrain of the zebrafish, we used immunohistochemistry to localize markers previously reported to be expressed in the dorsal hindbrain. We have previously cloned the LIM homeobox 2 (lhx2) and LIM homeobox 9 (lhx9) that are expressed regionally in embryonic zebrafish brain (Ando et al.,2005). In the hindbrain, they are expressed in a segmental manner in the dorsolateral region (Ando et al.,2005; Seth et al.,2006). Because this expression pattern is similar to the immunostaining pattern of the zn-5 antibody, which recognizes the cell adhesion molecule DM-GRASP (Neurolin/Alcam) expressed by the commissural neurons located in the rhombomere boundaries (Fashena and Westerfield,1999; Trevarrow et al.,1990), we performed double immunostaining of the coronal sections of the hindbrain with the zn-5 and Lhx2/9 antibodies (Liem et al.,1997). We observed that the dorsolateral cluster of neurons expresses both the zn-5 antigen and Lhx2/9 (Fig. 1A) and designated this class of neurons as the zn-5(+)Lhx2/9(+) neurons. Glutamate decarboxylase 1 (gad1) and glutamate decarboxylase 2 (gad2) are also expressed in a segmental manner in the hindbrain, and it has been suggested that the Gad1/2(+) neurons might correspond to the zn-5(+)Lhx2/9(+) neurons (Martin et al.,1998). We examined this possibility by performing double immunostaining with the Lhx2/9 and Gad1/2 antibodies. Unexpectedly, we found that they marked separate neuronal populations; the Gad1/2(+) neurons were located medially to the zn-5(+)Lhx2/9(+) neurons with a sharp boundary between them (Fig. 1B). Thus, we identified two distinct classes of neurons discriminated by location and gene expression pattern (Fig. 1F).
The bHLH transcription factor Atonal homolog 1 (Atoh1) is expressed in the rhombic lip, a germinal zone in the dorsal hindbrain of vertebrates from zebrafish to mouse (Köster and Fraser,2001a; Adolf et al.,2004; Aruga,2004; Wang et al.,2005). We performed in situ hybridization analysis of serial coronal sections of the hindbrain with the atoh1a and lhx2/9 probes and found that these two genes are expressed with a partial overlap in the dorsolateral hindbrain (Fig. 1C–E). These overlapping expression patterns are similar to those of Math1 (Atoh1) and Lhx2/9 in the mouse spinal cord (Helms and Johnson,1998; Caspary and Anderson,2003). In the Math1 mutant mouse, the Lhx2/9(+) neurons in the spinal cord and hindbrain neurons derived from the rhombic lip fail to develop (Bermingham et al.,2001; Wang et al.,2005). Thus, we injected the antisense morpholino (MO) against atoh1a transcripts into zebrafish embryos to examine whether Atoh1a is required for the development of zn-5(+)Lhx2/9(+) and/or Gad1/2(+) neurons. In the atoh1a MO-injected embryos, similarly to the case of Math1 knockout in the mouse spinal cord (Bermingham et al.,2001), the expression of lateral Lhx2/9 (Fig. 1G,H,J,L; n = 10 from two independent experiments) and the zn-5 antigen (Fig. 1J,L; n = 20) was severely diminished. The expression of another bHLH factor atoh1b, which is expressed in a subset of atoh1a(+) cells (Adolf et al.,2004), was also abolished (Fig. 1K,M). In contrast, the Gad1/2(+) neurons remained apparently intact and formed neuropiles and commissural fascicles in the ventral hindbrain (Fig. 1G,H; n = 10). The medial Lhx2/9(+) clusters also remained intact (Fig. 1G,H). We counted the number of cells in the lateral Lhx2/9(+), medial Lhx2/9(+), and Gad1/2(+) domains (Fig. 1G–I). Only the lateral Lhx2/9(+) neurons, which also express the zn-5 antigen (Fig. 1A), were significantly affected (asterisk in Fig. 1I indicates P < 0.01, Student t-test), indicating that Atoh1a is specifically required for development of the zn-5(+)Lhx2/9(+) neurons but not for the Gad1/2(+) and medial Lhx2/9(+) neurons. These results confirm the presence of the two separate classes of neurons, which can be distinguished by the developmental requirement for Atoh1a, as well as by cell location and marker expression (Fig. 1F).
Injection of the atoh1b MO showed no effect on the expression of atoh1a (data not shown). Combined with the above-mentioned observation that injection of the atoh1a MO abolished atoh1b expression (Fig. 1K,M), these results suggest that atoh1a is an upstream regulator of atoh1b.
Gad1/2(+) Neurons Project to the Contralateral Hindbrain
Identification of the projection pattern of a neuron is particularly important for classifying it precisely. To visualize the Gad1/2(+) neurons, we isolated a zebrafish bacterial artificial chromosome (BAC) clone of gad2, which covered the 5′- and 3′-regions of 35 kb and 130 kb, respectively, from the translation initiation codon. We inserted a DNA fragment containing the Gal4VP16 and UAS:DsRed2 into the 5′-untranslated region (5′-UTR) of gad2 by homologous recombination in bacteria (Fig. 2A; Yang et al.,1997). Using the Gal4VP16-UAS system, genes of interest can be expressed by placing them downstream of upstream activating sequences (UAS), as shown by transient expression assay (Köster and Fraser,2001b). The modified BAC DNA and UAS:GFP plasmid DNA were injected into fertilized eggs, and the transient expression of GFP or DsRed2 was observed. Figure 2B,C shows a neuron positive for both GFP and Gad1/2, with an axon that crossed the midline and terminated in the contralateral hindbrain. In the lateral hindbrain, we visualized four other Gad1/2(+) neurons projecting similarly to the contralateral hindbrain (data not shown). Thus, these results suggest that the neurons of the lateral-most Gad1/2(+) cluster form the contralaterally projecting fascicles. It remains possible, however, that this Gad1/2(+) cluster might be heterogeneous and may include neurons with ipsilateral projections.
zic1 Promoter/Enhancer-Driven GFP Predominantly Labels the zn-5(+)Lhx2/9(+) Neurons
Although the zn-5 antibody is known to label commissural axons (Trevarrow et al.,1990), the precise target of the zn-5(+)Lhx2/9(+) neurons has not previously been identified. Because the zn-5 signal weakens in the commissural axons after they extend across the midline, the zn-5 antibody staining was insufficient to trace unambiguously the projection patterns of the zn-5(+)Lhx2/9(+) neurons (data not shown). Therefore, we expressed GFP in the zn-5(+)Lhx2/9(+) neurons with the aim of revealing their entire axonal projections, including terminals.
To visualize the zn-5(+)Lhx2/9(+) neurons, we used the promoter/enhancer elements of zic1, which is expressed in the dorsal neural tube in a variety of vertebrate species including zebrafish (Fig. 3A; Aruga et al.,1994; Grinblat et al.,1998; Nakata et al.,1998; Rohr et al.,1999; Köster and Fraser,2001a; Aruga,2004). We isolated a zebrafish P1-derived artificial chromosome (PAC) clone of zic1, which covers the 5′- and 3′-regions of 73 kb and 45 kb, respectively, and inserted a cDNA encoding Gal4VP16 into the 5′-UTR of zic1 by homologous recombination in bacteria (Fig. 3B; Yang et al.,1997). This modified PAC DNA and UAS:GFP plasmid DNA were injected into fertilized eggs, and the transient expression of GFP was observed. We found it difficult, however, to trace the axon of a single presumptive zn-5(+)Lhx2/9(+) neuron from the cell body to the axon terminal by transient expression assay, due to low signal intensity and overlap of GFP signals from multiple cells (data not shown). Therefore, we established a transgenic line Tg(zic1:Gal4VP16/UAS:GFP) to visualize the axonal fascicles rather than a single axon (Fig. 3C,D). The crossing of GFP(+) fish with the wild-type fish resulted in GFP-positivity in half of the progeny. This finding was consistent with the interpretation that the modified PAC and UAS:GFP plasmid cointegrated into the same locus.
Expression of GFP in the Tg(zic1:Gal4VP16/UAS:GFP) embryo was first detected at 13 hours postfertilization (hpf) in the prospective eye and dorsal brain regions (data not shown). At 26 hpf, GFP expression was observed in the eyes and dorsal neural tube from the telencephalon to the rostral spinal cord (Fig. 3C,D). The expression in the dorsal neural tube was consistent with that of endogenous zic1. No significant GFP expression was detected in the dorsal somites where zic1 is expressed (Rohr et al.,1999). We examined whether GFP was expressed in the zn-5(+)Lhx2/9(+) neurons by immunohistochemistry of coronal sections of hindbrain. As shown in Figure 3E,F, the lateral GFP(+) cells predominantly coexpressed Lhx2/9 (67% of GFP[+] cells) and the zn-5 antigen. Thus, the zn-5(+)Lhx2/9(+) neurons were predominantly labeled in the Tg(zic1:Gal4VP16/UAS:GFP) fish. We noticed that the expression of GFP was mosaic (for example, see Fig. 3E,F); the number of cells expressing GFP was variable among embryos derived from the same parents, and the expression level of GFP was variable among the cells of the individual embryos. Thus, each embryo had variable numbers of both bright and dim GFP(+) cells. This phenomenon was observed across generations.
The zn-5(+)Lhx2/9(+) Neurons Project Contralaterally to the Midbrain TS
In some of the 30-hpf Tg(zic1:Gal4VP16/UAS:GFP) embryos, due to the variegation in GFP expression patterns among different individuals, we observed a commissural axon of a single GFP(+) neuron turning rostrally after crossing the midline (Fig. 4A). By 3 days postfertilization (dpf), the GFP(+) axons formed a ladder-like structure, which consisted of many commissures and two longitudinal fascicles (Fig. 4B, arrowheads). In cross-sections, the GFP(+) commissural axons projected ventrolaterally after crossing the midline (Fig. 3F), then appeared to form the two longitudinal fascicles (Fig. 4B, arrowheads). Because the initial parts of these commissural axons are also immunostained by the zn-5 antibody (Figs. 1A, 3F), it is apparent that the axons of the zn-5(+)Lhx2/9(+) neurons contribute to the longitudinal fascicles (Fig. 3F, arrowheads).
By comparison with the positions of the reticulospinal neurons retrogradely labeled with tetramethylrhodamine–dextran (Metcalfe et al.,1986), we found that the majority of commissural fascicles were distributed from rhombomere (r)2 to r6 and that the caudal end of the longitudinal fascicle corresponded approximately to the caudal r6 border (Fig. 4C). Additionally, we observed a pair of GFP(+) patches in r4 around the initial axon segment of the Mauthner cell (open arrowheads in Fig. 4B,C), suggesting that these patches were the Mauthner axon caps formed by the spiral fiber neurons and/or passive hyperpolarizing potential neurons (Lorent et al.,2001), although we could not identify their cell body positions in this transgenic line.
It appeared that the GFP(+) longitudinal fascicles formed by the zn-5(+)Lhx2/9(+) neurons terminated in the lateral midbrain region ventral to the optic tectum (arrows in Fig. 4B,D). To demonstrate this directly, we applied 1,1′, di-octadecyl-3,3,3′,3′,-tetramethylindo-carbocyanine perchlorate (DiI) to one side of the hindbrain containing the zn-5(+)Lhx2/9(+) neurons and observed the contralateral side (Fig. 4E,F). After crossing the midline, the DiI-labeled axons segregated into dorsal and ventral fascicles. The ventral fascicle turned rostrally, joined the longitudinal fascicle and terminated in the GFP(+) region of the midbrain ventral to the optic tectum (Fig. 4E, arrow, compare with Fig. 4B). The dorsal fascicle advanced straight, then bifurcated and terminated in the contralateral hindbrain with extensive branches (open arrows in Fig. 4E,F). We also observed the retrogradely labeled cell bodies of GFP(+) neurons (asterisks in Fig. 4F). This observation might be explained by our finding that a small number of GFP(+) neurons coexpressed Gad1/2 (Fig. 3G) and that the Gad1/2(+) neurons projected to the contralateral hindbrain (Fig. 2). Thus, in the Tg(zic1:Gal4VP16/UAS:GFP) embryos the predominantly labeled neurons were zn-5(+)Lhx2/9(+) neurons with axons projecting to the contralateral midbrain.
The ascending projection to the midbrain formed by the zn-5(+)Lhx2/9(+) neurons likely corresponds to that from the octaval and octavolateral nuclei to the midbrain TS described in adult fish and amphibians (Knudsen,1977; McCormick,1989; Will,1989; New et al.,1996; Edwards and Kelley,2001; Yamamoto and Ito,2005). Therefore, we examined whether the target nucleus in the midbrain of the Tg(zic1:Gal4VP16/UAS:GFP) embryo corresponds to the TS. Because no markers for the TS were known in zebrafish, we searched for markers expressed in the midbrain and found that lhx9 was expressed in the adult TS as well as in the periventricular gray zone of the optic tectum (Fig. 4G). The expression of lhx2/9 in these regions has been also reported in Xenopus (Moreno et al.,2005). We used anti-Lhx2/9 antibody to immunostain 3-dpf Tg(zic1:Gal4VP16/UAS:GFP) sections of the midbrain that contained axon terminals of the GFP(+) longitudinal fascicles, and found labeled nuclei surrounding the white matter region containing the GFP(+) terminals (arrows in Fig. 4H,I). This pattern of lhx2/9 expression is essentially identical to that of lhx9 in adult fish (Fig. 4G), suggesting that the GFP(+) longitudinal fascicles were projecting to the developing midbrain TS.
To demonstrate further that the zn-5(+)Lhx2/9(+) neurons contribute to the longitudinal fascicles projecting to the TS, we genetically ablated the zn-5(+)Lhx2/9(+) neurons by injecting the atoh1a MO into the Tg(zic1:Gal4VP16/UAS:GFP) embryos. The GFP(+) commissural fascicles, which were predominantly distributed from r2 to r6 in the control embryos (Fig. 4C,J,L), became very thin in the atoh1a MO-injected embryos (Fig. 4K,M). Furthermore, the longitudinal fascicles were absent in 52 (96%) of 54 atoh1a MO-injected embryos (compare arrowheads in Fig. 4J–M). From these results, we conclude that the Atoh1a-dependent zn-5(+)Lhx2/9(+) neurons form longitudinal fascicles and project to the contralateral TS. Based on the descriptions of Wullimann et al. (1996) in adult zebrafish, we identified this longitudinal fascicle as the lateral longitudinal fascicle.
zn-5(+)Lhx2/9(+) and Gad1/2(+) Neurons May Receive Auditory and Lateral Line Information
In adult fish and amphibians, auditory and lateral line information enters the hindbrain octaval and octavolateral nuclei and is then transferred to the midbrain TS by projection neurons in the octaval and octavolateral nuclei (Knudsen,1977; Finger and Tong,1984; McCormick,1989; Will,1989; New et al.,1996; Edwards and Kelley,2001; Yamamoto and Ito,2005). Therefore, we examined the relationship between the lateral line inputs from the periphery and the position of the GFP(+) commissural neurons. The central nerves of the posterior lateral line (PLL) enter the hindbrain, then bifurcate and extend longitudinally in the dorsal hindbrain (Fig. 5A; Alexandre and Ghysen,1999), where the GFP(+) neurons were abundant (Figs. 4C, 5A,B). Most of the GFP(+) neurons were located dorsomedially to the axon terminals of the PLL nerves (Fig. 5B), and many GFP(+) neuropiles were observed overlapping the terminals of the PLL nerves (Fig. 5B, inset). This organization at the larval stage is similar to that of the adult zebrafish and goldfish (New et al.,1996; Wullimann et al.,1996). This result suggests that the GFP(+) neurons receive auditory and lateral line information, although their functional connectivity remains to be demonstrated. Because the GFP(+) neurons are widely distributed in the dorsal hindbrain, it is also possible that they are involved in other sensory systems, such as the vestibular system.
The dorsal hindbrain is the site of the first synapse within the central nervous system for the sensory inputs of the vestibular, auditory, and lateral line systems (Butler and Hodos,1996). Although various classes of neurons have been identified within this complex region in adult vertebrate animals (Knudsen,1977; McCormick,1989; New et al.,1996; Cant and Benson,2003), little is known about their embryological origins. Recent studies performed in mice showed that the auditory-related nuclei, that is, the cochlear nuclei, originate from a part of the lower rhombic lip, the dorsal-most portion of the hindbrain proliferative neuroepithelium, characterized by the expression of Wnt1 and/or Math1 (Atoh1; Wang et al.,2005; Farago et al.,2006). Although these studies revealed the origins and molecular profiles of the auditory nuclei, the difficulty in distinguishing particular neuronal types during development limited the resolution of the analysis.
We expected that the ability to visualize specific neuronal types in zebrafish might provide the opportunity to overcome this difficulty. In this study, we first identified two classes of neurons with distinct molecular profiles, the Gad1/2(+) and zn-5(+)Lhx2/9(+) neurons, in the dorsal hindbrain of developing zebrafish. We used the promoter/enhancer regions of gad2 and zic1 to visualize the morphologies of these neurons. Identifying their axonal targets provided support for their involvement in the auditory and lateral line systems and allowed comparison of these neurons with those classified in adult fish of other species (Knudsen,1977; Echteler,1984; Finger and Tong,1984; McCormick,1989; New et al.,1996; Yamamoto and Ito,2005). We established a transgenic line, Tg(zic1:Gal4VP16/UAS:GFP), which expressed GFP under the control of the zic1 promoter/enhancer in the dorsal hindbrain including the rhombic lip. In combination with the Atoh1a knockdown, this strategy allowed us to compare the embryonic origin of the dorsal hindbrain neurons in zebrafish projecting to the TS with those in mammals projecting to the inferior colliculus, a homologous structure to the TS.
We demonstrated that the zn-5(+)Lhx2/9(+) neurons project to the contralateral midbrain TS (Fig. 5C, left). Electrophysiological and retrograde tracer incorporation studies of the TS in various species of adult fish including goldfish, carp, and catfish have revealed a mediolateral segregation of the inputs of the auditory and lateral line sensory modalities (Fig. 5C, right; Knudsen,1977; Echteler,1984; Finger and Tong,1984). These medial and lateral parts of the TS receive projections predominantly from distinct nuclei of the octaval and octavolateral nuclei in the dorsal hindbrain. Among the five octaval nuclei that receive primary inputs from the inner ear, the dorsal region of the descending octaval nucleus (DON) and anterior octaval nucleus (AON) provide projections to the medial (auditory-related) part of the contralateral TS in goldfish, carp, and catfish (Fig. 5C, right; Echteler,1984; Finger and Tong,1984; Yamamoto and Ito,2005). Of the two octavolateral nuclei that receive primary lateral line inputs, the medial octavolateral nucleus (MON) provides projections to the lateral (lateral line-related) part of the contralateral TS in goldfish and catfish (Fig. 5C, right; Finger and Tong,1984; New et al.,1996). Thus, the projection neurons in the DON and AON transfer auditory information, and those in the MON transfer lateral line information, to distinct parts of the TS.
We suggest that the zn-5(+)Lhx2/9(+) neurons correspond to the developing projection neurons in at least one of these three nuclei and are, therefore, involved in auditory and lateral line information processing. This proposal is also supported by our observation that the position of the zn-5(+)Lhx2/9(+) neurons is appropriate for receiving primary inputs of the auditory and lateral line nerves, as demonstrated in Figure 5A,B for the posterior lateral line nerve. In the adult TS, we showed that lhx9 is a marker for both the auditory-related medial part and the lateral line-related lateral part of the TS (Fig. 4G). Because the GFP(+) axon terminals of the Tg(zic1:Gal4VP16/UAS:GFP) transgenic embryos are distributed along the entire mediolateral extent of the TS, as marked by the expression of Lhx2/9 (Fig. 4I), the zn-5(+)Lhx2/9(+) neurons appear to contribute to both the auditory- and lateral line-related nuclei (Fig. 5C). Because GFP expression is not maintained in the adult Tg(zic1:Gal4VP16/UAS:GFP) transgenic fish (data not shown), we could not examine whether the GFP(+) neurons contributed to the DON, AON, and MON at an adult stage. We did not go on to distinguish between these nuclei in the embryo by experiments such as local DiI application in the medial (auditory-related) or lateral (lateral line-related) part of the midbrain TS to retrogradely label neurons in the octaval and octavolateral nuclei. Thus, it remains to be determined whether the auditory-related and lateral line-related hindbrain nuclei are segregated in the embryo or are segregated later (Fig. 5C).
Our data also provide insights into the rhombomeric origin of the neurons projecting to the TS. We observed that the majority of commissural fascicles contributing to the lateral longitudinal fascicles in the Tg(zic1:Gal4VP16/UAS:GFP) transgenic fish were distributed from r2 to r6 (Fig. 4C). Thus, we suggest that the neurons (in the AON, DON, and/or MON) projecting to the TS originate mainly from r2 to r6. However, another lateral line-related nucleus, the caudal octavolateral nucleus, does not provide projections to the TS in bowfin and goldfish (McCormick,1989; McCormick and Hernandez,1996) and is expected to be located caudally to the MON. The origin of lateral line-related nuclei might, therefore, extend more caudally than r6.
The rhombomeric origins of the mouse auditory-related nuclei (i.e., the cochlear nuclei) have recently been demonstrated (Farago et al.,2006). On the basis of cytoarchitectonic features, the cochlear nuclei of mammals can be divided into three divisions: the anteroventral cochlear nucleus, the posteroventral cochlear nucleus, and the dorsal cochlear nucleus (Ryugo and Parks,2003). By using genetic fate mapping, Farago et al. (2006) proposed a model in which the cochlear nuclei derive from r2 to r5; the anteroventral cochlear nucleus mainly derives from r2 and r3, the posteroventral cochlear nucleus from r4, and the dorsal cochlear nucleus from r5. This finding suggests a more rostrocaudally restricted origin for the mouse cochlear nuclei (r2–r5) than for the zebrafish auditory- and lateral line-related nuclei (r2–r6 or more caudally). In mouse, the precerebellar nuclei, which provide afferents to the cerebellum, have been suggested to originate from the caudal part of the lower rhombic lip, from r6 to r8 (Farago et al.,2006). Neurons belonging to the mammalian hindbrain precerebellar nuclei migrate tangentially toward the ventral midline and some of them cross the midline and settle contralaterally (Altman and Bayer,1987; Kawauchi et al.,2006). We did not observe such migrating GFP(+) cells around the ventral midline in r2 to r8 of the Tg(zic1:Gal4VP16/UAS:GFP) embryos. Although the presence of tangentially migrating precerebellar neurons in fish remains to be determined, the caudal part of the zebrafish lower rhombic lip (r6–r8) might be involved in generating auditory- and lateral line-related neurons as well as precerebellar neurons.
We found that the development of the zn-5(+)Lhx2/9(+) neurons in zebrafish depended on the bHLH transcription factor Atoh1a. In the cochlear nuclei of the Math1 (Atoh1) mutant mouse, most of the ventral cochlear nuclei and some unidentified cells in the dorsal cochlear nucleus are absent (Wang et al.,2005). The result in zebrafish suggests that the development of neurons projecting to the inferior colliculus (Cant and Benson,2003), a mammalian homolog of the TS (Butler and Hodos,1996), might also be impaired in the Math1 mutant mouse.
We demonstrated that the Gad1/2(+) neurons project to the contralateral hindbrain (Figs. 2, 5C, left). There are reciprocal projections in the MON and DON of adult goldfish (Fig. 5C, right; McCormick,1989; New et al.,1996; Yamamoto and Ito,2005), although their transmitter phenotypes are unknown. Therefore, the developing Gad1/2(+) neurons might correspond to these classes of neurons. In the cochlear nuclei of guinea pig, there are also neurons that project to the contralateral cochlear nuclei and are immunostained by the anti-glycine antibody (Wenthold,1987). Thus, although the transmitters might be different between fish and mammals, the reciprocal inhibitory projections between the left and right sides of the auditory and lateral line systems might be phylogenetically conserved neuronal connections.
We found that the distributions of the Gad1/2(+) and zn-5(+)Lhx2/9(+) neurons are clearly segregated in the dorsal hindbrain of zebrafish embryos. In contrast, in the adult goldfish MON, neurons projecting to the contralateral TS and MON are intermingled (New et al.,1996). These two populations may be generated separately during development and then migrate to become intermingled by an adult stage.
We noticed that, in the Tg(zic1:Gal4VP16/UAS:GFP) fish, GFP was expressed in a mosaic manner (for example, see Fig. 3F). The expression levels of GFP in the cells of individual fish were also variable. Thus, each embryo had variable numbers of both bright and dim GFP(+) cells. In the hindbrain, GFP expression was more frequently and consistently observed in the anterior region than in the posterior region. With the availability of only a single transgenic line, it is not possible to attribute the cause of this phenomenon to the incomplete regulation of cis-elements in the transgene, integration site effect, or other mechanisms.
Although in this study we used the two transgenes only for visualization of the Gad1/2(+) and zn-5(+)Lhx2/9(+) neurons using the UAS:GFP plasmid, they could be used for further dissection of the developmental mechanisms of these two classes of commissural neurons. For example, the zn-5(+)Lhx2/9(+) neurons turned rostrally after crossing the midline, while the Gad1/2(+) neurons projected straight, suggesting that the axonal responsiveness to the guidance cues is different between the two types of neurons. Once the axon guidance receptors that are specifically expressed in one class of neurons are identified, it should be possible to test their role by ectopically expressing them in the other class of neurons. In addition, with the availability of genetically encoded probes that respond to or modify neuronal activity (Fetcho and Bhatt,2004; Hua et al.,2005), it will be possible to study how auditory and lateral line information is represented and processed in the hindbrain using the zebrafish as a model animal.
Zebrafish (Danio rerio) were maintained at the RIKEN BSI fish facility according to standard procedures (Westerfield,2000). Embryos and larvae were obtained from wild-type and transgenic Tg(zic1:Gal4VP16/UAS:GFP) fish (see below). For fluorescence observations, 0.002% 1-phenyl-2-thiourea was added to the fish water from 9 hpf to inhibit pigmentation. If necessary, embryos and larvae were anesthetized with 0.05% ethyl 3-aminobenzoate methanesulfonate (Tricaine). Developmental stages and neuroanatomical structures were identified according to Westerfield (2000) and Wullimann et al. (1996).
zic1 and gad2 complementary DNA (cDNA) were isolated from adult brain cDNA by polymerase chain reaction (PCR) using the following primer pairs: zic1F(5′-ATGCTNTTRGAYGCNGG-3′)/zic1R(5′-GCCATRTTCATRT-TCAT-3′) and gad2F(5′-TTCACCTATGAGGTGGCTCC-3′)/gad2R(5′-CTGC-AGGTCATAATGCTTGT-3′). Because the entire open reading frames of atoh1a and atoh1b are contained within a single exon, genomic DNA was used as a PCR template to isolate them. The primers used were atoh1aF(5′-CCAACAGAATGGATGGAATG-3′)/atoh1aR (5′-TACAAAGCCCAGATATGCGA-3′), and atoh1bF(5′-CTACACAAGCAATGACTGCA-3′)/atoh1bR(5′-GTGTGG-TAATCAGCGTCCTC-3′). These primers were designed based on the following GenBank accession numbers: zic1, AF052435; gad2, AF017265; atoh1a, NW_634984. For atoh1b, the open reading frame sequence was obtained by running BLAST search on the Ensemble zebrafish database (http://www.ensembl.org/Danio_rerio/index.html) with a partial cDNA sequence (Adolf et al.,2004).
Zebrafish PAC (Amemiya and Zon,1999) and BAC libraries (CHORI-211) were obtained from Incyte Genomics (St. Louis, MO) and BACPAC resources (Oakland, CA), respectively. Screening by filter hybridization was performed as described previously (Sassa et al.,2000). PAC and BAC DNA were purified using QIAGEN Plasmid Maxi Kit (QIAGEN, Valencia, CA).
Vector Construction and Homologous Recombination in Escherichia coli
We used an E. coli RecA-based homologous recombination system (Yang et al.,1997) to insert Gal4VP16 and Gal4VP16/UAS:DsRed2 fragments into the 5′-UTR of zic1 PAC and gad2 BAC clones, respectively. The Gal4VP16 (pCMV-GVP) and UAS:GFP (pUG) plasmids were kindly provided by S.E. Fraser, California Institute of Technology (Köster and Fraser,2001b). The pDsRed2-1 plasmid was obtained from Clontech (Mountain View, CA). The Gal4VP16/UAS:DsRed2 plasmid (pGVPUAS:DsRed2) was constructed from the plasmids described above. The pSV1.RecA plasmid and a protocol for recombination were kindly provided by N. Heintz, Rockefeller University (Yang et al.,1997). For zic1 PAC, the targeting vector pSV1.RecA-5′GVP3′ was constructed from the pCMV-GVP as a starting plasmid as described below. An approximately 800-bp genomic DNA fragment upstream from the translation initiation codon of zic1 was PCR amplified and cloned upstream of GVP to create the pCMV 5′GVP plasmid. Similarly, the same size fragment downstream from the translation initiation codon was PCR amplified and cloned downstream of GVP to create the pCMV5′GVP3′ plasmid. A 5′GVP3′ fragment was isolated and cloned into the pSV1.RecA plasmid, resulting in the pSV1.RecA-5′GVP3′ plasmid. For gad2 BAC, the targeting vector pSV1.RecA-5′GVPUAS:DsRed23′ was constructed in essentially the same manner except for the use of the pGVPUAS:DsRed2 instead of the pCMV-GVP as a starting plasmid. Homologous recombination was performed as described previously (Yang et al.,1997).
Injection Into One-Cell Stage Zebrafish Embryos
A Femtojet air pressure injector (Eppendorf, Germany) was used for all injection experiments. DNA and MOs were dissolved in deionized water, with 0.05% phenol red added to aid visualization. To generate the Tg(zic1:Gal4VP16/UAS:GFP) fish, the mixture of linearized zic1:Gal4VP16 PAC DNA and UAS:GFP circular plasmid DNA was injected at concentrations of 20 μg/ml and 1 μg/ml, respectively. Injected embryos were observed 1–3 days later using a fluorescent dissecting microscope and those which developed normally and expressed GFP were raised into adult fish. Thirty-one fish were screened by crossing among themselves or with wild-type fish, and two were identified as founders (F0). From these two, we were able to establish one line, while the other was lost due to infertility of F1 fish. For visualization of single Gad1/2(+) neurons, gad2:Gal4VP16/UAS:DsRed2 BAC circular DNA and UAS:GFP plasmid were injected at concentrations of 20 μg/ml and 5 μg/ml, respectively. Antisense and control five-base mismatch MOs against the translation initiation codon of atoh1a were obtained from Gene Tools (Philomath, OR). Their sequences were as follows. atoh1a: antisense MO, 5′-GTATCCGTGCTCATTCCATCCATTC-3′; control MO, 5′-GTATgCGTcCTgATTCaATCaATTC-3′ (mismatch bases are indicated by lower case). The MOs were injected at the concentration of 2 mg/ml. Control MO-injected embryos showed no differences from the wild-type embryos.
In Situ Hybridization and Immunohistochemistry
In situ hybridization was performed as previously described (Westerfield,2000). Cryosections (10–20 μm in thickness) were made at the level of the otic vesicle using a cryostat (HM500 M; Microm, Germany). For fluorescence detection, the Tyramid Signal Amplification Plus Dinitrophenyl system (NEN Life Science Products, Boston, MA) and Alexa 488-conjugated anti-dinitrophenyl antibody (1:500; Molecular Probes, Eugene, OR) were used. Antibody staining was performed according to standard methods (Westerfield,2000). The primary antibodies were used at the following dilutions: zn-5 (1:500, mouse monoclonal; Oregon Monoclonal Bank; Trevarrow et al.,1990; Fashena and Westerfield,1999), Lhx2/9 (1:5,000, rabbit polyclonal; kindly provided by T.M. Jessell, Columbia University; Liem et al.,1997), GFP (1:500, rabbit polyclonal; MBL, Japan), and Gad1/2 (1:500, mouse monoclonal; Affiniti, UK). We used three excitation wavelengths (with Alexa 488, 546, and 647) for observation of goat Alexa-conjugated anti-mouse or anti-rabbit IgG secondary antibodies (1:1,500, Molecular Probes). SYTOX Orange (1:10,000, Molecular Probes) was used for the fluorescent nuclear counterstaining.
Labeling of Axonal Projections
Labeling of reticulospinal neurons and lateral line nerves with tetramethylrhodamine–dextran (molecular weight, 3,000; Molecular Probes) was performed as described previously (Moens et al.,1996; Alexandre and Ghysen,1999). To label hindbrain projection neurons, DiI (Molecular Probes) at 2 mg/ml in N,N-dimethylformamide was applied by iontophoresis using a current injector (Intra 767; World Precision Instruments, Sarasota, FL) with a fixed-stage microscope (BX50WI; Olympus, Japan) under a ×63 water-immersion objective (Shoji et al.,1998). Dye-injected larvae were left for 3–12 hr before imaging to allow dye distribution.
Nonfluorescent samples were mounted in 75% glycerol and viewed with a differential interference microscope (Axioplan 2; Carl Zeiss, Germany) equipped with a charge-coupled device camera (DP50; Olympus, Japan). Fluorescent samples were observed using a confocal laser-scanning microscope (LSM510; Carl Zeiss, Germany). For the GFP- and Alexa 488-labeled samples, a 488-nm excitation beam and a 505- to 530-nm or a 505- to 550-nm bandpass filter were used. For the DsRed2, Alexa 546, tetramethylrhodamine–dextran, and DiI-labeled samples, a 543-nm excitation beam and a 560-nm longpass filter or a 565- to 615-nm bandpass filter were used. For the Alexa 647 samples, a 633-nm excitation beam and a 650- to 710-nm bandpass filter were used.
We thank N. Heintz for kindly providing the pSV1.RecA plasmid and a protocol for homologous recombination in Escherichia coli; S.E. Fraser for the Gal4VP16 and UAS plasmids; T.M. Jessell for the Lhx2/9 antibody; and members of the Okamoto laboratory for discussion. This research was supported by Grants-in-Aid (T.S. and H.O.) and a Special Coordination Fund (H.O.) from the Ministry of Education, Science, Technology, Sports and Culture of Japan, and by a grant for Core Research for Evolutional Science and Technology (H.O.) from the Japan Science and Technology Corporation, and by grants (H.O.) from RIKEN Brain Science Institute.