Transgenic technology for visualization and manipulation of the neural circuits controlling behavior in zebrafish




The vertebrate brain is innately equipped with neural circuits that make quick behavioral decisions possible. Elucidating these neural circuits, determining how their master plans are encoded in the genome, and revealing how they can be modified by postnatal experiences will facilitate our understanding of how nature and nurture interact to establish an animal's behavior. In this review, we explain how transgenic zebrafish can cast insights into the developmental mechanisms and functional roles of the neural circuits that directly and indirectly control visuomotor behavior, by taking as an example a transgenic line Tg(brn3a-hsp70:GFP) enabling visualization of the tectobulbar and habenulo-interpeduncular tracts. These insights emphasize the benefits of applying advanced transgenic technology in zebrafish to future research into this area.


Determining whether to bite or to escape is one of the most crucial decisions that animals must make every time they face an unknown or known object (Vallortigara & Rogers 2005). A quick decision is of vital importance. Any delays due to long, careful deliberation may lead to a fatal capture by predators or to a loss of food because the prey has had time to escape, even if the correct decision was made. Therefore, vertebrate brains are innately equipped with neural circuits that make such quick decisions possible and which free animals from the dilemmas of decision making. Elucidating these neural circuits, revealing how their master plans are encoded in the genome, and revealing how they can be modified by postnatal experiences will facilitate our understanding of how nature and nurture interact to establish animal behavior.

It is believed that animals have adapted to produce escape responses when a visual stimulus approaches from the dorso-posterior position. This indicates a risk of predation, and consequently activates the posterior tectum (Dean et al. 1989). In the rat superior colliculus, the uncrossed descending pathway originating from the posterior tectum is believed to be involved in escape or avoidance movements, whereas the crossed descending pathway, which mainly originates from the anterior tectum, is involved in orienting movements toward a visual target (Dean et al. 1989). Likewise, electrical microstimulation of the posterior tectum in goldfish which corresponds to a visual stimulus from above and behind evokes escape-like movements (Herrero et al. 1998).

The neural circuits responsible for regulating the bite or escape response in various animals possess not only an innate tendency to evoke various responses to visual inputs at different points along the anterior–posterior axis of the tectum, but also appear to have asymmetrical thresholds for triggering the final behavioral programs in the left and right sides of the brain. In zebrafish and chickens, the right eye is used when it is necessary to inhibit a premature response, in order to sustain viewing until a decision is reached. The left eye is used when it is necessary to watch a familiar scene (Bisazza et al. 1998; Miklosi & Andrew 1999). These findings suggest that the functional significance and neural mechanisms of cerebral lateralization are conserved in teleost fish and tetrapods.

The zebrafish is an animal model known for its amenability to various cellular, molecular and genetic manipulations because of its transparency during development and relatively short generation time. Recent advances in transgenic technology have enabled us to establish transgenic lines of zebrafish in which the neural circuits regulating visuomotor responses can be visualized or specifically manipulated. One good example is the stable transgenic line Tg(brn3a-hsp70:GFP), which expresses green fluorescent protein (GFP) under the control of the brn3a promoter (Aizawa et al. 2005; Sato et al. 2007). Brn3a is a POU-domain transcription factor expressed in specific central neurons and primary sensory neurons (He et al. 1989; Xiang et al. 1996). In wild-type embryos, brn3 a is mainly expressed in the retina and the tectum, as well as in the habenula and the cranial sensory ganglia (He et al. 1989; Xiang et al. 1996). The transgene brn3a-hsp70:GFP is composed of a regulatory region of the zebrafish brn3a gene containing the 4.9 kb upstream region, part of the first exon and the first intron, hsp70 as a basal promoter, and GFP as a reporter (Fig. 1a). Consistent with the expression of brn3a mRNA, the transgenic embryos primarily expressed GFP in the retina and the tectum, but also in the habenula, the pretectum and the torus semicircularis from about 34 h postfertilization (h.p.f.) (Fig. 1b,c).

Figure 1.

Green fluorescent protein (GFP) expression in Tg(brn3a-hsp70:GFP) embryos. (a) Structure of the zebrafish brn3a gene and the brn3a-hsp70:GFP transgene. A regulatory region of the brn3a gene contains a 4.9 kb upstream region, a non-coding region (brown bar), and an intron (yellow bar). The hsp70 promoter (red bar) and EGFP (green box) sequences were inserted downstream of the regulatory region. The translation start site is indicated as +1. S, Sac I. (b, c) Merged images of bright-field and GFP fluorescence images of Tg(brn3a-hsp70:GFP) embryos, shown in the frontal view (b) and the anterolateral view (c) at about 34 h postfertilization. (d) Ventral view of ipsilaterally descending tectobulbar tracts expressing GFP in the brain of a transgenic embryo at 7 days postfertilization. (e) A transverse section of the brain in (d). The dorsal branches of the tectobulbar tract elaborated around the ventral dendrites of the reticulospinal neurons. Reticulospinal neurons were retrogradely labeled with rhodamine-conjugated dextran (red, d, e).

Visualization of the tectobulbar projection

The optic tectum is one of the best model systems for investigating visuomotor transformation. It receives topographically ordered visual inputs from the retina in the superficial layers and sends motor outputs from the deeper layers to the premotor reticulospinal system in the hindbrain (Wurtz & Albano 1980). In teleost fish, the reticulospinal system is a segmentally organized motor center (Metcalfe et al. 1986) that develops within hindbrain segments, or rhombomeres (r) 0 to 8 (Moens & Prince 2002). Mauthner neurons in r4 initiate fast escape responses (Eaton et al. 1981), and their firing can be triggered by visual, auditory, or multimodal sensory stimuli (Eaton et al. 1991). Recent studies have shown that the majority of other reticulospinal neurons are also activated during escape behavior, suggesting coordinate functions of multiple reticulospinal neurons (Liu & Fetcho 1999; Gahtan et al. 2002; Nakayama & Oda 2004). However, little is known about how each reticulospinal neuron receives projections from the tectum.

Development of the stable transgenic line Tg(brn3a-hsp70:GFP), which expresses GFP under the control of the brn3a promoter (Aizawa et al. 2005; Sato et al. 2007) has allowed visualization of the optic tract projecting to the superficial layers in the contralateral tectum. It has also enabled visualization of the tectobulbar tract descending ipsilaterally from the deeper layers to the hindbrain reticulospinal neurons (Fig. 1d,e; Sato et al. 2007). To further investigate whether the GFP-positive tectal neurons project to the reticulospinal system with a defined spatial pattern, we established a single-neuron genetic labeling system in zebrafish (Sato et al. 2007). We co-injected three plasmids, pCS2:cre, brn3a-hsp70:loxP:DsRed2: loxP:Gal4VP16 and UAS:GFP into embryos at the 1- to 4-cell stage (Fig. 2a). This procedure stochastically induced the expression of Gal4VP16 under the control of the brn3a-hsp70 promoter after the DsRed sequence was excised by Cre recombinase, and resulted in the induction of GFP expression (Dong & Stuart 2004; Kimura et al. 2006). Neurons incorporating the brn3a-hsp70:loxP:DsRed2:loxP:Gal4VP16 plasmid in the injected embryos expressed DsRed, whereas only neurons that simultaneously incorporated all three plasmids expressed GFP (Fig. 2b arrowheads).

Figure 2.

Single-neuron genetic labeling system. (a) Plasmid constructs used for the single-neuron labeling system in combination with Cre/loxP and Gal4/VP16 systems. (b) Lateral view of a labeled neuron at the single-cell level in the tectum of an injected embryo at 7 days postfertilization (arrowhead).

Mapping the somatic locations in the tectum with respect to the final projection sites in the hindbrain revealed that the tectum is organized as a mosaic of subregions. Each subregion provides a different combination of descending projections to different hindbrain segments. Intriguingly, the probability of the neurons in the posterior tectum projecting to r2 was significantly higher than the corresponding probability of the neurons in the nonposterior tectum projecting to the same segment (Fig. 3a,d).

Figure 3.

Probabilities of neurons projecting to r2 in the ectopic expression of ephrinB2a and ephrinB2aΔC . (a) Soma locations for labeled neurons in wild-type embryos (colored circles, left). Wild-type neurons projecting to r2 (blue circles) and the other labeled neurons (gray circles) in the posterior tectum (middle) and the nonposterior tectum (right). (b) Soma locations for neurons expressing ephrinB2a (colored squares, left). ephrinB2a-expressing neurons projecting to r2 (blue squares) and the other labeled neurons (gray squares) in the posterior tectum (middle) and the nonposterior tectum (right). (c) Soma locations for neurons expressing ephrinB2aΔC (colored triangles, left). ephrinB2aΔC -expressing neurons projecting to r2 (blue triangles) and the other labeled neurons (gray triangles) in the posterior tectum (middle) and the nonposterior tectum (right). (d) Ratios of the number of neurons projecting to r2 to the total number of labeled neurons in the (bsl00005) posterior tectum and in the (□) nonposterior tectum for wild-type, ephrinB2a, and ephrinB2aΔC embryos. Pairwise comparisons of ratios with significant differences are indicated by ** for < 0.01 and *** for P < 0.001 (binomial test).

In zebrafish, ephrinB2a is expressed medially in the tectum at 36 h.p.f. (Chan et al. 2001), and is localized in the most posterior tectal cells at 3 days postfertilization (d.p.f.) (Wagle et al. 2004). This area of expression coincided approximately with the above-defined posterior region of the tectum, within which almost all of the neurons projecting to r2 were distributed. The molecules of the ephrin-Eph ligand-receptor family play a major role in the topographic projection of retinotectal neural connections (Flanagan & Vanderhaeghen 1998; O'Leary & McLaughlin 2005). To address the involvement of ephrinB2a in the specific projection from the posterior tectum to r2, we applied the single-neuron labeling system to the construct UAS:ephrinB2a:UAS:GFP to ectopically express ephrinB2a in tectal neurons concomitantly with GFP (Fig. 2a). The forced expression of ephrinB2a elevated the probability of the neurons projecting to r2 in the nonposterior tectum to the corresponding probability of wild-type neurons projecting to the posterior tectum (Fig. 3b,d; Sato et al. 2007). To further examine whether ephrinB2a functions through a reverse-signaling mechanism, we overexpressed ephrinB2aDC , a dominant-negative form of ephrinB2a lacking the cytoplasmic domain (Durbin et al. 1998), using the same system (Fig. 2a). This reduced the probability of the neurons projecting to r2 in the posterior to the level corresponding to the wild-type embryos in the nonposterior tectum (Fig. 3c,d; Sato et al. 2007).

These results clearly indicate that ephrinB2a is critically involved in increasing the probability of neurons projecting from the posterior tectum to r2 via a reverse signaling mechanism. The significance of this specific neural connection is not clear at present, although electrical microstimulation of the posterior tectum, corresponding to a visual stimulus from above and behind, evokes escape-like movements in goldfish (Herrero et al. 1998). In addition, an in vivo calcium imaging study in zebrafish has indicated that the reticulospinal neurons in r2, as well as the Mauthner cells and MiD2cs in r4 and r5, are activated during escape behavior (Gahtan et al. 2002). More substantial evidence will be obtained by examining how manipulating the activity of this neural connection influences escape behaviors, for example repression, hyperactivation or complete elimination. Such experiments are now possible by specifically expressing gene products such as inward rectifying potassium channels (Kir; Hua et al. 2005), photo-activated ion channels/transporters (Arenkiel et al. 2007), and a tetanus toxin C fragment (Yamamoto et al. 2003) under the control of UAS in the stable transgenic line brn3a-hsp70:loxP:DsRed2:loxP: Gal4VP16, if Cre can be activated in a tissue-specific manner in the tectum or its primordium. Crossing this line with another transgenic line expressing Cre under the midbrain-specific promoter, or activating Cre translation by midbrain-specific photoillumination to induce cleavage of 6-bromo-7-hydroxycoumarin-4-yl-(Bhc) caged cre mRNA will be useful for these purposes (Ando et al. 2001).

Visualization of anatomical asymmetry in the habenular projection

The habenulae are part of an evolutionarily highly conserved “conduction pathway” within the limbic system that connects the telencephalic nuclei to the midbrain region (Fig. 4a; Sutherland 1982). Each has distinct neural connectivity. The lateral habenula directly connects the globus pallidus and hypothalamus in the forebrain with the nuclei in the midbrain and the hindbrain, which are rich in monoaminergic (domaminergic and serotonergic) neurons. The medial habenula indirectly connects the septal nuclei with monoaminergic nuclei via the interpeduncular nucleus (IPN). Axons from the medial habenula terminate in the IPN in a characteristic pattern reminiscent of an electromagnetic coil (Herkenham & Nauta 1977; Herkenham & Nauta 1979). The medial and lateral habenulae can therefore differentially regulate monoaminergic activity under the influence of the basal ganglia and septohippocampal circuit. This activity can, in turn, affect the basal ganglia so that behavioral programs can be modified as necessary.

Figure 4.

Laterotopic projections from the habenular nuclei to the interpeduncular nucleus (IPN). (a) Schematic illustration of the lateral view of an adult zebrafish brain showing the telencephalo-habenulo-interpeduncular pathway, which connects the telencephalon with the ventral midbrain. (b) Dorsal view of a brain of a 4 days postfertilization Tg(brn3a-hsp70:GFP) larva showing green fluorescent protein (GFP) labeling of the entire length of the habenulo-interpeduncular projection (arrows). (c) Frontal view of an adult brain from a Tg(brn3a-hsp70:GFP) fish. (d) Schematic diagram showing the asymmetrical organization of the habenulae and their projection in the adult zebrafish brain. The lateral subnuclei are shown in red and the medial subnuclei shown in green. (e, f) Transverse sections through the habenulae (e) and IPN (f) of adult Tg(brn3a-hsp70:GFP) fish in which DiI was applied to the dIPN (f, red). Cbll, cerebellum; d, dorsal; FR, fasciculus retroflexus; Hb, habeunla; Hy, hypothalamus; IPN, interpeduncular nucleus; l, left; OB, olfactory bulb; PO, pineal organ; PP, parapineal organ; SM, stria medullaris; Tel, telencephalon; TeO, optic tectum; v, ventral.

A subpopulation of neurons in the part of the habenula that corresponds to the medial habenula in the mammalian brain could be visualized in the zebrafish transgenic line Tg(brn3a-hsp70:GFP) (Fig. 4b,c; Aizawa et al. 2005). To avoid describing further phylogenetic differences, we have tentatively termed this part of the zebrafish habenula simply as “habenula” for the purposes of this review. The GFP-positive neurons occupy the medial subregion of the habenula. Intriguingly, the expression domain is larger in the right habenula than in the left (Fig. 4c–e). Furthermore, the axons from the GFP-positive neurons selectively terminate in the ventral and intermediate IPN (Fig. 4d–f), and the vast majority of neurons that innervate the dorsal IPN are located in the lateral, GFP-negative subregion of the left habenula (Fig. 4d–f). A small number of neurons in the lateral subregion of the right habenula also innervate the dorsal IPN, and again these neurons are GFP negative (Fig. 4d–f).

These results indicate that the zebrafish habenular nuclei can be subdivided into at least two asymmetrical subnuclei populations that have different size ratios on the left and right sides of the brain and different targets in the IPN (Fig. 4f; Aizawa et al. 2005). The large medial habenular subnucleus on the right innervates the ventral and intermediate IPN together with axons from the small medial subnucleus on the left. Conversely, the large lateral subnucleus on the left is the predominant source of axons that innervate the dorsal IPN along with axons from the small lateral subnucleus on the right. We therefore concluded that left–right asymmetries in habenular subnuclear organization are converted to dorso-ventral asymmetries in habenular projections to the IPN, and named the characteristic pattern of this axonal projection “laterotopy” (Aizawa et al. 2005).

Birth date analyses of neurons in the medial and lateral subnuclei using the 5-bromo-2-deoxyuridine (BrdU) pulse-labeling technique was applied to the Tg(brn3a-hsp70:GFP) transgenic line, and showed that neural precursors for GFP-negative lateral subnuclei were born earlier than the neural precursors for GFP-positive medial subnuclei. Intriguingly, the neurogenetic activity of the lateral habenular subnuclei was greater on the left side at 32 h.p.f., when the number of BrdU-labeled cells peaked. In contrast, the number of pulse-labeled medial subnuclei cells peaked at 48 h.p.f., and was significantly greater on the right side at this stage. These data suggested that asymmetrical development of the habenular subnuclei might occur because neurons of the medial and lateral subnuclei are born at different stages during development (Aizawa et al. 2007).

The difference in the timing of neurogenesis may dictate the types of neurons and the orientation of habenular asymmetry. This hypothesis was further corroborated by hyperactivation and repression studies of Notch signaling during development and the subsequent examination of the expression patterns of lateral subnuclei (leftover) and medial subnuclei (righton) marker genes (Fig. 5a,b; Aizawa et al. 2007). To transiently repress neurogenetic efficiency in the embryonic dorsal diencephalon, Notch signaling was ubiquitously activated at a specific time point by utilizing the double transgenic zebrafish Tg(hsp70:gal4); Tg(UAS:myc-notch1a-intra), in which activated Notch (intracellular domain of Notch1a) was induced under the control of a heat shock promoter (Scheer et al. 2001). Repression of the lateral subnuclei neural precursors during early stages such as 28 h.p.f., 32 h.p.f. and 36 h.p.f. drastically reduced leftover expression on both sides of the habenulae in 56 h.p.f. embryos (Fig. 5c). Intriguingly, a large number of righton-positive cells were found in the left habenula of the 56 h.p.f. embryos upon induction of ectopic Notch at 28 h.p.f., 32 h.p.f. or 36 h.p.f. (Fig. 5d). This suggested that the neural stem cells bypassed the normal program, during which early-born leftover-expressing neural precursors are generated. Then, following the removal of neurogenetic repression, the stem cells started to produce cells expressing righton, the marker for late-born medial subnuclei precursors. Reduction of Notch signaling affected the neural specification for each subnucleus in the opposite manner. This was confirmed in the mind bomb (mib) mutant, in which Notch signaling was severely reduced due to a defect in the ubiquitin ligase essential for internalization of the Notch ligand, Delta (Itoh et al. 2003). All mib mutants showed a drastic increase of leftover expression in the right habenula, while righton expression was diminished on the right side at 56 h.p.f. (Fig. 5e,f).

Figure 5.

Genetic hyperactivation or repression of Notch signaling affects specification for the lateral and medial habenular subnuclei. Dorsal views of wild-type (a, b), Tg(hsp70:gal4)kca4 and Tg(UAS:myc-notch1a-intra)kca3 (c, d; heat shock was applied at 32 hours postfertilization [h.p.f.]), and mind bomb mutant embryos (e, f) at 56 h.p.f. showing lov (a, c, e), ron (b, d, f) mRNA expression in the habenulae.

Taken together, the genetic hyperactivation and repression of Notch signaling revealed that the timing of neurogenesis determines neuronal specificity for the medial and lateral habenular subnuclei. These results further support our model that the left-predominant neurogenesis of the lateral subnuclei during the early stages reduced the common pool of neural stem cells in the left habenula, and that this reduction could account for the right-predominant generation of late-born medial subnuclei precursors (Fig. 6).

Figure 6.

Neurogenetic model for the generation of left–right differences in the habenular subnuclei. At 18–24 hours postfertilization (h.p.f.), Nodal and its downstream genes (green in the top panels showing dorsal view of Tg[lefty1:GFP]) are activated in the left dorsal diencephalon. Neural precursors for the medial (green in the lower panel) and lateral subnuclei (red in the lower panel) are then generated, forming an asymmetrical organization in the habenular subnuclei as observed at stage 72 h.p.f. Altering neurogenesis during early development reduced the number of neural stem cells (light blue in the lower panel) on the left side, accounting for the right predominance of neurogenesis during late development on the right side.

We and others have also shown that unilateral and transient Nodal activation in the dorsal diencephalon specifies the orientation of habenular asymmetry. However, left–right differences still remained after bilateral Nodal activation and even in the absence of Nodal signaling (Concha et al. 2000; Concha et al. 2003; Aizawa et al. 2005). This suggests the presence of two distinct mechanisms for the establishment of habenular laterotopic projection (Fig. 6). Firstly, transient unilateral activation of Nodal in the left habenular primordium specifies the direction of laterality, and secondly, asymmetrical activation of Notch defines the extent of the structural differences between the left and right habenulae.

As described above, zebrafish mainly use the right eye while approaching a novel object, probably to inhibit a premature response and to sustain viewing until a decision is reached (Miklosi & Andrew 1999). This preferential eye use is inverted in the mutant fish in which the direction of habenular laterality is inverted (Barth et al. 2005). This suggests that the difference in the left and right habenulae sets up different thresholds for triggering behavioral programs. The extent of threshold differences between the left and right eye systems could be adjusted by variations in the level of Notch activity in the left and right habenular primordium among different species of animals.

Brain asymmetry is implicated in influencing the social behavior of animals of the same species. Concordance in laterality may allow the animals to coordinate their behavior when a group encounters the same situation (Vallortigara & Rogers 2005). Therefore, the first step in habenular lateralization, that is, unilateral and concordant Nodal activation in the left habenular primordium, may be involved in defining concordance in animal behavior. The second step, that is, differential Notch activation in the left and right habenulae, may be responsible for fine tuning the levels of concordance in animal behavior depending on how coordinated the social behaviors of individual species need to be.

Future prospects

It is not known how the habenulae on both sides are differentially involved in setting thresholds for triggering behavior. The dorsal conduction system relayed by the habenular nuclei, and the visuomotor transformation system in the tectum/superior colliculus may interact via the afferent projections from the midbrain dopaminergic neurons and the hindbrain serotonergic neurons into the tectum/superior colliculus, or indirectly via the basal ganglia (McHaffie et al. 2005). Further exploitation of modern genetic manipulation technologies as introduced in this review will help answer these questions, especially now a wider choice of techniques is available for tissue-specific transgene expression, for example, use of Tol2-mediated enhancer trap lines for the Gal4VP16/UAS gene expression system (Kawakami et al. 2004), or bacterial artificial chromosome or P1-derived artificial chromosome-clone mediated transgenesis (Kimura et al. 2006; Sassa et al. 2007).

Conflict of Interest

No conflict of interest has been declared by H. Okamoto, T. Sato or H. Aizawa.