Transgenic tools to characterize neuronal properties of discrete populations of zebrafish neurons

Neuronal diversity is a fundamental feature of the nervous system. Elucidating how this diversity arises is one of the central aims in developmental neurobiology. For the past 15 years, it has been shown that differential gene expression during development forms the basis of this diversity (Goulding et al., 2002; Jessell, 2000). Thus, knowing gene expression patterns in neuronal cells is of fundamental importance. Transgenic animals expressing fluorescent proteins in a subset of neuronal cells are powerful tools, because they allow researchers to see expression patterns of particular genes in live animals. The zebrafish possesses a strong advantage over other vertebrate model organisms because its body is transparent during development.

Here, we have generated a wide variety of transgenic zebrafish, in which reporter genes are expressed in a specific subset of neurons or neuronal progenitors. We had three specific aims in this research. The first aim was to expand a repertoire of transgenic fish that express reporter genes in neurons having a particular neurotransmitter phenotype. Because neurotransmitter properties are fundamental to neurons, such transgenic fish would be of general importance. We have previously generated transgenic fish that express GFP or RFP in glutamatergic and glycinergic neurons (McLean et al., 2007; Satou et al., 2012). These lines have been effectively used in many studies (e.g. Kinkhabwala et al., 2011). Here, we have expanded our repertoire by generating lines that are linked to a GABAergic phenotype, or lines expressing Gal4 in glutamatergic and glycinergic neurons. The second aim was to reveal domain organization in the developing dorsal spinal cord. In the vertebrate spinal cord, the first step towards generating neuronal diversity is the subdivision of neuronal progenitor cells into distinct domains along the dorsoventral axis. In amniotes, 11 distinct domains (five in the ventral and six in the dorsal spinal cord) are formed (Zhuang and Sockanathan, 2006). In zebrafish, domain organization in the ventral spinal cord is conserved (Goulding, 2009). In the dorsal spinal cord, however, domain organization is not well understood. We sought to determine the domain organization of the dorsal spinal cord by generating transgenic fish that expressed fluorescent proteins under the promoter/enhancer of genes that are known to be expressed in a domain-like manner in amniotes. The third aim was to investigate transmitter phenotypes of neurons that are produced from each domain of the dorsal spinal cord.

We successfully expanded our repertoire of transgenic fish that label particular neurotransmitter phenotypes. We found that dorsal spinal cord domain organization is largely conserved in zebrafish. We also determined transmitter properties of neurons that are produced from some of the domains.

All the transgenic fish (Table 1) were generated using the BAC (bacterial artificial chromosome) homologous recombination technique with either the I-SceI-mediated method (Kimura et al., 2006) or the Tol2-mediated method (Suster et al., 2009). The BACs used were: zK145P24 for vglut2a (slc17a6b - Zebrafish Information Network), zC24M22 for gad1b, zC247L22 for atoh1a, zK171N3 for neurog1, zK97C6 for gsx2, zC271B14 for gsx1, zC163N18 for dbx2 and zC246F16 for dmrt3a. The glyt2 (slc6a5 - Zebrafish Information Network) BAC was isolated from a BAC library with a hybridization-based screening. The GFP and lRl-GFP (loxP-DsRed-loxP-GFP) constructs for homologous recombination were described previously (Kimura et al., 2006). The Gal4 (Gal4FF) (Asakawa et al., 2008) and Cre constructs were described by Kimura et al. (Kimura et al., 2013) and Satou et al. (Satou et al., 2012), respectively. The lRl-Gal4 construct was prepared by replacing GFP with Gal4 in the lRl-GFP. For the vglut2a:lRl-Gal4 and glyt2:lRl-Gal4 construct, mCherry was used as the RFP, whereas for the neurog1:lRl-GFP and gsx2:lRl-GFP constructs, dTomato was used. neurog1 is expressed in neuronal progenitors and also in Rohon-Beard (RB) neurons. An RB neuron enhancer was found to be located within 1 kb upstream. This 1-kb sequence was eliminated in the neurog1:lRl-GFP construct. Tg[gad1b:GFP], Tg[neurog1:GFP], Tg[gsx2:GFP] and Tg[glyt2:Gal4] lines were generated by injecting Cre mRNA into embryos of Tg[gad1b:lRl-GFP], Tg[neurog1:lRl-GFP], Tg[gsx2:lRl-GFP] and Tg[glyt2:lRl-Gal4] lines, respectively.

Images were taken using Zeiss 510 or Nikon A1 confocal microscopes. For the time-lapse imaging, animals were embedded in agarose every time for imaging (after imaging, animals were removed from the agarose).

Glutamate, glycine and GABA are three major neurotransmitters in the central nervous system (CNS). For glutamatergic, glycinergic and GABAergic neurons, vglut, glyt2 and gad genes, respectively, can be used as markers. We have previously generated Tg[vglut2a:GFP], Tg[vglut2a:RFP] and Tg[glyt2:GFP] lines (McLean et al., 2007; Satou et al., 2012). To expand our repertoire, we generated Tg[glyt2:RFP] and Tg[gad1b:GFP] and Tg[gad1b:RFP] transgenic fish. We compared the fluorescent protein expression patterns in each line by crossing a GFP line to an RFP line of another neurotransmitter. The spinal cord at 3.5 days post-fertilization (dpf) was examined. Fig. 1A-C shows cross sections of Tg[vglut2a:GFP] and Tg[glyt2:RFP] (Fig. 1A), Tg[vglut2a:GFP] and Tg[gad1b:RFP] (Fig. 1B) and Tg[glyt2:GFP] and Rg[gad1b:RFP] (Fig. 1C). In each of the panels, GFP and RFP are expressed in distinctive cells. Thus, each transgenic fish successfully labeled discrete neuronal populations that have a particular neurotransmitter phenotype.

Our RFP lines were created based on the lRl-GFP (loxP-RFP-loxP-GFP) cassette (Table 1). This allowed us to induce GFP expression in the subpopulation of glutamatergic, glycinergic and GABAergic neurons by crossing the lines to Cre driver lines. Two examples of such labeling are shown in Fig. 1D,E. In Fig. 1D, GFP is expressed in glycinergic neurons that derive from the dbx1b-positive p0 progenitor domain (V0 neurons). In Fig. 1E, GFP is expressed in GABAergic neurons that derive from the gsx1-positive pd4/pd5 progenitor domains (for clarification of the pd4/pd5 progenitor domains, see below). Neurons in which Cre-mediated recombination did not occur continued to express RFP (Fig. 1D,E).

The Gal4-UAS system is a powerful tool to express reporter/effector genes, such as genetically encoded calcium indicators and optogenetic tools. It would be useful to have lines expressing Gal4 in neurons having particular neurotransmitter phenotypes. Thus, we generated Tg[vglut2a:Gal4] and Tg[glyt2:Gal4], and these lines were capable of driving reporter gene expression in glutamatergic and glycinergic neurons, respectively, with high expression levels (supplementary material Fig. S1).

Finally, we sought to combine the Gal4-UAS system with the Cre-loxP system. We created Tg[vglut2a:lRl-Gal4] and Tg[glyt2:lRl-Gal4]. Fig. 1F,G shows examples of the inducible expression of GFP in V0 neurons. In Fig. 1F, glutamatergic V0 neurons express GFP, whereas in Fig. 1G, glycinergic V0 neurons express GFP. Using this system, researchers will be able to drive expression of various reporter/effector genes in specific populations of neurons having a particular neurotransmitter phenotype.

In summary, we have successfully expanded our repertoire of transgenic fish that are associated with neurotransmitter phenotypes by creating GFP, RFP (conditional GFP), Gal4 and conditional Gal4 lines.

In amniotes, progenitors in the developing dorsal spinal cord are divided into six distinct domains (pd1-pd6) that are defined by the expression of transcription factors (Zhuang and Sockanathan, 2006). In zebrafish, domain organization in the developing dorsal spinal cord has not been well characterized. To examine domain organization, we generated transgenic fish that express fluorescent proteins under the promoter/enhancer of genes that are known to be expressed in particular progenitor domains in amniotes. Genes used include atoh1a, neurog1, gsx2, gsx1 and dbx2. We examined the expression domain spatial profiles of these transcription factors (represented by GFP or RFP expression). In addition to the lines listed in Table 1, Tg[dbx1b:RFP], in which RFP is expressed in the p0 domain (Satou et al., 2012), was also used for comparison.

Examinations were made using embryos at 36-48 hours post-fertilization (hpf), when extensive neurogenesis occurs (Lyons et al., 2003). atoh1a was expressed in the most dorsal progenitor domain (Fig. 2A). neurog1 had two expression domains: a dorsal thin neurog1 expression domain and a ventral wide neurog1 expression domain (Fig. 2A,B, short and long lines, respectively), and we only consider the former in this study. The dorsal boundary of this thin domain was adjacent to the atoh1a expression domain (Fig. 2A). The gsx2 expression domain was thin (Fig. 2B,C), and this domain was bordered dorsally by the dorsal neurog1 domain (Fig. 2B) and ventrally the gsx1 expression domain (Fig. 2C). dbx2 was expressed widely in the middle region of the spinal cord (Fig. 2D-G). In some sections, the dbx2 expression domain was bordered dorsally by the gsx1 expression domain (Fig. 2D), whereas in other sections, the dorsal-most dbx2-positive progenitors were also positive for gsx1 (Fig. 2E). This indicates that gsx1-positive progenitors can be divided into two populations: one population that is negative for dbx2, and another population that is positive for dbx2. The dbx1-positive progenitors (p0 progenitors) were all positive for dbx2 (Fig. 2F). The ventral boundary of the dbx2 expression domain was more ventral than that of the dbx1 expression domain (Fig. 2F). The dbx2-positive cells were found to be negative for the V2 neuron marker chx10 (vsx2 - Zebrafish Information Network) (Kimura et al., 2008), suggesting that the ventral boundary of the dbx2 expression domain corresponds to the p1-p2 boundary (Fig. 2D-G). There was a gap between the gsx1 and dbx1 expression domains (Fig. 2G), indicating that a domain exists between the gsx1 expression domain and the p0 domain.

Collectively, the results are summarized in Fig. 2H,I, revealing the existence of six progenitor domains (pd1 to pd6), as in amniotes. The pd1 domain is defined by atoh1a expression, and the pd2 domain corresponds to the neurog1 dorsal expression domain. The pd3 domain is defined by gsx2 expression. The gsx1 expression domain consists of two cell populations: dbx2-negative cells and dbx2-positive cells. The former corresponds to the pd4 domain, and the latter corresponds to the pd5 domain. The pd6 domain is defined by those cells that are positive for dbx2 and negative for gsx1 and dbx1.

Expression patterns of atoh1a, neurog1, gsx2, gsx1, dbx2 and dbx1 are consistent with those of the corresponding genes in mice (Alaynick et al., 2011), with only some minor differences. In mice, gsx2 is reported to be expressed at a low level in the pd4 and pd5 domains (Kriks et al., 2005). In zebrafish, gsx2 expression, as revealed by GFP expression, is restricted to the pd3 domain (Fig. 2B,C). In addition, the pd4 and pd5 domains, which are defined by gsx1 expression, might not form a sharp dorsoventral boundary, because in some sections only cells belonging to the pd4 domain (dbx2 negative) are observed (Fig. 2D), whereas in other sections cells belonging to the pd5 domain (dbx2 positive) are predominantly present (Fig. 2F).

We sought to determine the neurotransmitter properties of neurons that derive from each domain. Fluorescent protein expression in the transgenic lines persisted in their progeny. This allowed us to examine neurotransmitter properties of progeny neurons by crossing the transgenic lines with the lines associated with neurotransmitter properties. Examinations were made at 3.5 dpf.

We first examined dI1 neurons (neurons that derive from the pd1 progenitors) using Tg[atoh1a:GFP] fish. Virtually, all of the GFP-labeled dI1 neurons were positive for vglut2a:RFP (Fig. 3A, arrows and arrowheads). None of GFP-labeled dI1 neurons was positive for glyt2:RFP or gad1b:RFP (Fig. 3B,C). The results indicate that dI1 neurons are essentially glutamatergic. We noted that some of dI1 neurons were located in a more ventral region (Fig. 3A, arrows) and our time-lapse imaging showed that some dI1 neurons migrated ventrally (supplementary material Movie 1).

Because we do not have transgenic fish that specifically label pd2 progenitors, we next examined dI3 neurons using Tg[gsx2:GFP] (Fig. 3D-F). The results showed that glutamatergic neurons (Fig. 3D, arrowheads) were present among dI3 neurons, but that glycinergic and GABAergic neurons were not present (Fig. 3E,F).

We then examined dI4 and dI5 neurons using Tg[gsx1:GFP] (Fig. 3G-I). The results showed that glutamatergic (Fig. 3G, arrowheads), glycinergic (Fig. 3H, arrowhead), and GABAergic neurons (Fig. 3I, arrowhead) were all present among dI4/5 neurons.

We do not have transgenic fish that specifically label pd6 progenitors. We sought to examine neurotransmitter properties of dI6 neurons using a gene that is known to be expressed in post-mitotic dI6 neurons in mice. A recent report shows that Dmrt3 is expressed in inhibitory dI6 neurons (Andersson et al., 2012), and we expected this was also the case in zebrafish. To explore whether or not dmrt3 is expressed in inhibitory dI6 neurons, we generated Tg[dmrt3a:RFP] and Tg[dmrt3a:Gal4]. We first examined whether dmrt3a was expressed in dI6 neurons using Tg[dmrt3a:RFP]. As shown in supplementary material Fig. S2, all of the RFP-labeled neurons were positive for dbx2:GFP, but none of them was positive for gsx1:GFP and dbx1b:GFP, indicating that dmrt3a:RFP neurons do indeed derive from the pd6 progenitors (Fig. 2H). We then examined the neurotransmitter properties of dmrt3a-positive neurons using Tg[dmrt3a:Gal4] and Tg[UAS:GFP]. As shown in Fig. 3J-L, all of the GFP-labeled neurons were positive for glyt2:RFP (Fig. 3K, arrowhead), and none of them was positive for vglut2a:RFP or gad1b:RFP, indicating that dmrt3a marks glycinergic dI6 neurons. Thus, glycinergic neurons are present among dI6 neurons. It remains unknown whether or not glutamatergic and GABAergic neurons are also present among dI6 neurons.

The results described in this section are summarized in Fig. 3M. Transmitter properties of neurons that derive from dorsal progenitor domains are generally consistent with those in mice (Alaynick et al., 2011).

We determined domain organization in the dorsal spinal cord. We also systematically determined neurotransmitter properties of neurons that are produced from several domains. This forms the basis for future detailed analyses of the neurons produced by each domain. We have generated a wide variety of transgenic zebrafish, including those that are linked to neurotransmitter properties. Although we focused on the spinal cord in the present study, reporter gene expression in these transgenic fish occurs widely in the CNS, including hindbrain, midbrain and forebrain (supplementary material Fig. S3). Therefore, the newly generated transgenic fish should serve as powerful tools for the investigation of neuronal structures and functions in many CNS regions.

We thank Shunji Fujioka for helping to generate Tg[dbx2:GFP].