Transgenic zebrafish expressing fluorescent proteins in central nervous system neurons


*Author to whom all correspondence should be addressed.


Zebrafish is a powerful model system for investigations of vertebrate neural development. The animal has also become an important model for studies of neuronal function. Both in developmental and functional studies, transgenic zebrafish expressing fluorescent proteins in central nervous system neurons have been playing important roles. We review here the methods for producing transgenic zebrafish. Recent advances in transposon- or bacterial artificial chromosome-based transgenesis greatly facilitate the creation of useful lines. We also present our study on alx-positive neurons to reveal how transgenic zebrafish expressing fluorescent proteins in a specific class of neurons can be used to investigate their development and function.


Zebrafish is a widely used model organism for studying vertebrate development. One of its major advantages is the almost complete transparency of the embryos, which allows researchers to look at individual cells during development. Zebrafish is also amenable to genetics due to its relatively short generation time (2–3 months). These two features make this organism well suited for expressing green fluorescent protein (GFP) or other fluorescent proteins using transgenic techniques. Such transgenic fish are particularly useful when fluorescent proteins are expressed in specific classes of neurons in the central nervous system (CNS). This is because there are a large number of neuronal classes in the CNS, and distinguishing specific types from others is necessary for studying development and function of the nervous system.

This review has two aims. The first is to introduce the basic and innovated methods of generating transgenic zebrafish. The second is to introduce the power of transgenic fish that express fluorescent proteins in a particular class of neurons. Thus, I will first describe the basic method for producing transgenic zebrafish. The generation of the Tg[isl1:GFP] transgenic fish (the first stable transgenic fish that expresses GFP in a particular type of CNS neurons) is described as an example of a very useful transgenic zebrafish that was generated by basic methods. Then, the transposon- and bacterial artificial chromosome (BAC)-mediated transgenic methods that have greatly facilitated the generation of transgenic zebrafish are described. Finally, to illustrate the power of GFP-expressing transgenic fish, our study on alx-expressing neurons is described. The present study reveals that transgenic zebrafish can be used not only for developmental study but also for investigating the physiological roles of neurons in the functional neural circuits.

Basic method for generating transgenic zebrafish

The first transgenic zebrafish was made by Stuart et al. (1988), who showed that that DNA injected into the cytoplasm of fertilized zebrafish embryos could integrate into the fish genome and be inherited in the germline. Several reports followed (e.g. Stuart et al. 1990; Bayer & Campos-Ortega 1992). Transgenic zebrafish of these early days, however, suffered from inconsistent expression of a transgene; in effect, variegated expression or no expression. The exact reason for this is unknown, but the use of promoter/enhancer derived from other organisms might have been responsible. In 1997, two groups succeeded in generating transgenic zebrafish that reliably expressed GFP in specific tissues (Higashijima et al. 1997; Long et al. 1997). These two reports marked the practical start of GFP-expressing transgenic zebrafish.

The basic procedure for making transgenic zebrafish with tissue specific GFP expression is as follows. First, researchers make a DNA construct, and inject it into the cytoplasm of one-cell-stage zebrafish embryos. The embryos are then examined to see whether or not the promoter/enhancer in the construct is capable of driving GFP expression in expected tissues. In this transient expression assay in which DNA-injected embryos themselves are analyzed, GFP expression occurs in a mosaic manner. This is because only a subset of cells retain the foreign DNA at the time when GFP expression starts. Once a positive construct is found, DNA-injected embryos are raised to adulthood. At certain probability, the foreign DNA integrates into the genome of germ cell precursors. Thus, some adults become positive founders: they produce fluorescent embryos. The frequency of obtaining positive founders with the simple DNA injection method has been 5–20%.

Islet1-GFP transgenic fish

The first transgenic fish to express GFP in CNS neurons was generated by the author in collaboration with Hitoshi Okamoto (Higashijima et al. 2000). Our strategy was to use the islet1 (isl1) gene to drive gene expression in motor neurons. Isl1 is a member of the LIM/homeobox gene family and is expressed in all postmitotic motor neurons (Ericson et al. 1992; Korzh et al. 1993; Inoue et al. 1994). Approximately 100 kb of genomic DNA flanking the isl1 gene was cloned, and the DNA fragment that was capable of driving GFP expression in cranial motor neurons was identified. In the transgenic fish generated with the construct, GFP was expressed in the majority of cranial motor neurons (Fig. 1). The intensity of GFP fluorescence by the cranial motor neurons in the transgenic fish (Tg[isl1:GFP]) was sufficiently high to enable clear visualization of the cell bodies, main axons, and the peripheral branches. Using the Isl1-GFP fish, the development of cranial motor neurons was characterized (Higashijima et al. 2000).

Figure 1.

Green fluorescent protein (GFP) expression in cranial motor neurons in the Tg[isl1:GFP] embryos. (A) Lateral view of the head region of a 34 h embryo. (B) Dorsal view of the midbrain and hindbrain of a 42 h embryo.

The generation of the Tg[isl1:GFP] transgenic line and the characterization of the GFP-expressing neurons set the stage for future applications of the transgenic fish. The Tg[isl1:GFP] line has widely been used by many researchers for various studies. Of particular significance was the large-scale mutagenesis screening carried out by Hitoshi Okamoto's group. Using the transgenic fish line as the parental strain, they identified a large number of mutations that impaired the development of cranial motor neurons (Wada et al. 2005; Tanaka et al. 2007). Subsequent molecular cloning and cellular analyses revealed several important aspects of cranial motoneuron development, including the crucial role of planar-cell-polarity (PCP) pathway for the caudal migration of the facial motor neurons (Wada et al. 2005; Wada et al. 2006), and semapholin-plexin pathway for axonal pathfinding of the jaw-innervating motor axons (Tanaka et al. 2007). In these studies, the ability to directly observe the GFP-labeled cells in live animals was crucial to carry out the experiments. These studies demonstrate the power of GFP-expressing transgenic fish for studies of neural development.

Transposon-mediated, and BAC-mediated transgenesis

The Tg[isl1:GFP] transgenic fish was made by the basic method; in effect, a simple DNA-injection. For the past several years, new techniques that have further facilitated the generation of useful lines have been used. Among them are the transposon-based and the BAC-mediated transgenesis.

When making transgenic animals, the higher efficiency of the transgenesis is beneficial. Toward this end, several efforts have been put to increase the efficiency of the transgenesis in zebrafish (Thermes et al. 2002; Davidson et al. 2003). Currently, the most popular one is the Tol2-based transposon system. As excellent reviews about this system have already been published (Kawakami 2005; Kawakami 2007), I briefly describe it here. The Tol2 element was originally identified in the genome of the medaka fish (Koga et al. 1996). Subsequently, Koichi Kawakami developed the Tol2-medeated transgenic system in zebrafish (Kawakami et al. 2000). In this system, a DNA construct is made on a Tol2-based plasmid vector that contains inverted repeats for Tol2. The DNA construct is then injected into one-cell-stage embryos with synthetic mRNA for the Tol2 transposase. With the activity of the transposase, integration of the injected DNA into the genome occurs much more frequently than that with the simple DNA injection. The frequency of obtaining positive founders reaches about 50%. This high frequency has greatly alleviated the labor of making transgenic fish.

Another new technique that has become popular is the BAC-mediated transgenesis. This technique does not improve the efficiency, but greatly reduces the labor of making DNA constructs. When generating transgenic fish, we need to obtain regulatory elements that control gene expression in desired tissues. Often, however, the regulatory elements are scattered over a large length of the genome DNA. For example, the cis-regulatory elements for the isl1 gene expression is scattered over a 100 kb of the genome (Higashijima et al. 2000; Uemura et al. 2005). In such cases, building DNA constructs for proper gene expression was laborious with standard molecular manipulations, since researchers had to make a lot of constructs, and had to test the enhancer activity of each construct. If we can place reporter genes into BACs, the labor-intense steps can largely be bypassed. This can now be easily carried out, thanks to the highly efficient homologous recombination techniques that were originally developed by the Stewart group (Zhang et al. 1998), and modified by the Copeland group (Lee et al. 2001). The core of the techniques lies in the introduction of the phage-derived recombinases (RecET in the case of Stewart's original, and Exo and Beta in the Copeland's modified one) in Escherichia coli. Because of the powerful activities of these enzymes, homologous recombination in E. coli can be achieved with homology arm sequences that are as short as 50 base pairs (see Fig. 2A). This means that homology arms can be included in polymerase chain reaction (PCR) primers when preparing targeting DNA, and thus, no subcloning step is necessary (reviewed in Copeland et al. 2001; Muyrers et al. 2001). It is possible that researchers make DNA constructs within a week after obtaining bacteria that harbor BACs. The technique has been used by several zebrafish researchers, and a number of useful transgenic zebrafish have been established (e.g. Shin et al. 2003; Kimura et al. 2006).

Figure 2.

The alx:GFP bacterial artificial chromosome (BAC) construct and an Tg[alx:GFP] transgenic fish. (A) Construction of the alx:GFP BAC. Template DNA was polymerase chain reaction (PCR)-amplified from a plasmid containing green fluorescent protein (GFP), Bovine Growth Hormone (BGH) poly(A) (pA) and a kanamycin resistance gene (Kmr). Each primer contained 50 bp of the alx-derived sequences that served as homology arms for homologous recombination (top drawing). After homologous recombination, GFP was inserted between the transcription start site and the translation start site (middle drawing). The bottom drawing shows the structure of the alx:GFP BAC used for generating transgenic fish. (B) A Tg[alx:GFP] transgenic fish at 3 dpf. (C) Confocal image of a Tg[alx:GFP] transgenic fish at 56 h postfertilization (hpf). Several optical sections were stacked, and montages were made using the stacked images.

Because of the long size of BACs (usually, over 100 kb), the transposon-based technique cannot be applied. Therefore, we cannot expect the efficiency that is achieved by the Tol2-based method. The efficiency in our hand has been 1–5%. Nonetheless, it is still a practical level. When working with genes whose regulatory elements are scattered, a benefit of being liberated from laborious subcloning steps is quite significant.

Studies on alx-positive neurons

We applied the BAC-mediated transgenic zebrafish technique to study development and function of alx-positive neurons. alx is a zebrafish homolog of mammalian Chx10, a homeobox gene known to be expressed in a subset of developing spinal interneurons located in the ventral region of the cord. Studies of the differentiation of spinal cord have identified a number of transcription factors that are expressed in a small subset of developing interneurons, and Chx10 is such a transcription factor. It is generally assumed that expression of these transcription factors are linked with some fundamental characters of spinal interneurons. The nature of these transcription factor positive cells, however, is largely unknown. This is because investigating how interneurons develop and function in mammals is extremely difficult due to the enormous complexity of their spinal cord. Our main aim on the study of alx-positive neurons was to reveal the properties and physiological functions of these cells in zebrafish. These, in turn, could provide insights into the properties and functions of corresponding cells in mammals.

By using the BAC-mediated transgenic technique, we generated Tg[alx:GFP] transgenic fish (Fig. 2B), in which each alx-expressing neurons (hereafter, called alx cells or alx neurons) in the spinal cord was clearly visualized in live zebrafish embryos/larvae (Fig. 2C). To examine the morphology of alx neurons more clearly, we made another transgenic fish using the DNA construct shown in Figure 3(A). In this construct, DsRed is sandwiched by two loxP sites. The resulting transgenic fish, Tg[alx:loxP-DsRed-loxP-GFP] expressed DsRed in alx cells. By stochastically expressing Cre protein with a transient expression system in the stable Tg[alx:loxP-DsRed-loxP-GFP] transgenic fish, we could obtain fish in which a small, random subset of alx cells expressed GFP instead of DsRed, as Cre-mediated recombination flipped DsRed out. An example of this is shown in Figure 3A, which shows that alx neurons are ipsilateral descending interneurons: a primary axon that first extends ventrally, turns caudally in ventral spinal cord, and then descends on the ipsilateral side of the spinal cord. We also confirmed this result by filling individual GFP-labeled alx neurons with rhodamine-dextran using the electroporation method (Fig. 3B). We next examined neurotransmitter phenotype of alx neurons by carrying out dual in situ hybridization either with vglut2 (a marker for glutamatergic neurons) or glyt2 (a marker for glycinergic neurons), and found that all of the alx neurons were positive for vglut2 (Kimura et al. 2006). The results strongly suggested that alx neurons were glutamatergic excitatory neurons.

Figure 3.

Single cell morphology of alx neurons. (A) The top panel shows the structure of the construct for generating Tg[alx:loxP-DsRed-loxP-GFP] transgenic fish. The picture panel shows a confocal image of a 48 h postfertilization (hpf)Tg[alx:loxP-DsRed-loxP-GFP] transgenic fish injected with a Cre-containing plasmid. (B) Lateral view of a Tg[alx:GFP] transgenic fish in which three green fluorescent protein (GFP)-positive cells were loaded with rhodamine-dextran by electroporation.

Previous calcium imaging studies in larval zebrafish have shown that ipsilateral descending neurons were active in escape behavior (Ritter et al. 2001). Studies in frog tadpoles and adult lamprey have also shown that ipsilateral descending neurons were important for swimming behavior (Grillner et al. 1991; Roberts et al. 1998). In either case, the suggested role of the descending neurons is to drive activity of motoneruons. Thus, we hypothesized that alx neurons play roles during locomotion such as escapes and swimming, by providing direct excitation onto motoneurons. To test this, we carried out whole-cell recordings using the patch-clamp technique from GFP cells in the Tg[alx:GFP] fish. The results indicated that alx neurons were indeed active during locomoation. Some were active during escape or fast swimming (Fig. 4A), while others were active during slow swimming (Fig. 4B). Next, we examined whether alx neurons provided direct excitation onto motoneurons. By using paired recordings between alx neurons and motoneurons, we found that alx neurons made frequent monosynaptic excitatory connections onto motoneurons. An example is shown in Figure 4(C). Consistent with dual in situ hybridization data, the connections were found to be glutamatergic (Kimura et al. 2006). Taken together, these results indicated that alx neurons are premotor interneurons that provide direct excitations onto motoneurons during locomotion such as escapes and swimming.

Figure 4.

Activity of the alx neurons during fictive locomotion. Embryos were paralyzed with alpha-bungarotoxin. The experiments were carried out in the Tg[alx:GFP] embryos at 52–62 h postfertilization (hpf). (A) and (B) are simultaneous recordings between alx neurons and a ventral root (VR), while (C) is a simultaneous recording between alx neurons and a motoneurons (moto). Patch electrodes were targeted to green fluorescent protein (GFP)-positive cells to record from alx neurons. Fictive locomotion was elicited by applying a brief electrical stimulation near the ear (contralateral side) at the time point indicated by the arrowhead. In (A) and (B), the top panel for each figure shows the image after electrophysiological recording, while the middle and bottom panels show the whole cell recordings from the alx neuron and the ventral root recording, respectively. The asterisk in (A) shows a cell that was labeled on an earlier attempt. The cell is unrelated to the recording shown in (A). In (A) and (C), the alx neurons only fired at the initial phase of the locomotion episode, indicating that these neurons were only active during the escape or fast swimming. In contrast, the alx neuron in (B) fired during slow swimming. Note that frequency of the rhythmic VR activities is faster during the early phase of the locomotion episode. (Cc) The current-evoked spikes in the alx neuron lead to Excitatory Postsynaptic Potential (EPSPs) in the motoneuron. Five superimposed traces of EPSPs in the motoneuron are shown. The traces show short, constant latency EPSPs. Only one spike in the alx neuron is shown for simplicity. Traces of EPSPs in the motoneuron are aligned with the peak of the spikes of the alx neuron.

Thus, we succeeded in identifying the important components of fish locomotor circuits in the spinal cord. Furthermore, we succeeded in tracing their roots to those cells that express a transcription factor, alx, in the developmental stage. This has a great effect when placed in a comparative context. We previously revealed the features of cells that express En1, another transcription factor expressed in a small subset of developing spinal interneurons (Higashijima et al. 2004). Parallel studies in frog tadpoles and mammals have shown that key features of En1-positive neurons are widely conserved between aquatic vertebrates and mammals (Saueressig et al. 1999; Li et al. 2004; Gosgnach et al. 2006). These studies indicate that the expression of transcription factor during development can provide a bridge between cell types across vertebrate species. The features and function of Chx10-positive neurons in mammals are not understood well, but based on our results, we would predict that these neurons are premotor interneurons that provide excitation onto ipsilaterally located motoneurons during mammalian locomotion.

Future prospects

Transgenic zebrafish expressing fluorescent proteins in specific classes of neurons have proven powerful tools for investigating development and function of these neurons. The transposon- and BAC-mediated transgeniesis techniques have been facilitating the generation of useful transgenic zebrafish. For example, we have been making a number of transgenic fish expressing GFP or DsRed in specific classes of neurons. Such transgenic fish undoubtedly contribute to the studies toward the understanding of development and function of neuronal circuits. In addition to GFP and DsRed, a wide variety of fluorescent proteins (reviewed in Miyawaki 2005) or other functional proteins can also be expressed. For example, transgenic zebrafish expressing recently developed light-activated channel or pomp that can activate or inhibit neurons (Li et al. 2005; Zhang et al. 2007) could be extremely powerful tools for the investigation of the function of neurons. The ability to generate transgenic zebrafish with a relative easiness will facilitate this kind of exciting studies in the future.


Thanks to Yukiko Kimura, Hitoshi Okamoto, Hironori Wada, and Hideomi Tanaka for discussions.