Zebrafish homologue irx1a is required for the differentiation of serotonergic neurons



Serotonergic (5HT) neurons produce neurotransmitter serotonin, which modulates various neuronal circuits. The specification and differentiation of 5HT neurons require both extrinsic signals such as Shh and Fgf, as well as intrinsic transcription factors such as nkx2.2, mash1, phox2b, Gata2, and pet1. In this study, we show that iroquois homeodomain factor irx1a, but not irx1b, is expressed in the 5HT neurons. Knockdown of irx1a by antisense morpholino nucleotides reveals that it is a critical determinant for the differentiation of 5HT neurons in the hindbrain. However, irx1a morphants do not show a reduction of the progenitors of 5HT neurons. Hence, irx1a is not required for the initial specification but it is required for the complete differentiation of 5HT neurons. Developmental Dynamics 236:2661–2667, 2007. © 2007 Wiley-Liss, Inc.


The serotonergic (5HT) system is highly evolutionarily conserved in both invertebrates and vertebrates (Weigar,1997; Tierney,2001). Dysfunction of the serotonergic system is associated with many disorders, including anxiety, depression, aggression, schizophrenia, Alzheimer's disease, and Parkinson's disease (Onodera et al.,1994; Lucki,1998; Davidson et al.,2000; Nelson and Chiavegatto,2001; Hendricks et al.,2003). Despite the clinical importance of the central serotonergic system, the mechanisms that control its specification and differentiation during development are poorly understood.

Vertebrate 5HT neurons are present in the raphe nuclei and localized along the midline of the hindbrain (Jacobs and Azmitia,1992). In rodents, 5HT neurons first appear in the mantle layer of the hindbrain adjacent to the floorplate and midbrain–hindbrain boundary (MHB; Lidov and Molliver,1982; Wallace and Lauder,1983). Subsequent 5HT populations develop in more caudal positions, with many sending ascending innervations to modulate spinal sensory and motoneurons (Hynes and Rosenthal,1999). The regional restriction of 5HT induction is believed to be resulted from signals from the floorplate and isthmus organizer, particularly Sonic Hedgehog and fibroblast growth factor (Ye et al.,1998). Recent studies have started to identify the transcription determinants of 5HT neurons. The homeobox gene Nkx2.2 is expressed in the progenitors of both 5HT and branchiomotor visceromotor neurons (pMNv), and phox2b acts as a binary switch in the selection of 5HT or motoneuron fate (Pattyn et al.,2003). The specification of 5HT cell fate is determined by the switch of homeobox gene code from Nkx2.2+/Nkx2.9+/Phox2b+ to Nkx2.2+/Nkx2.9/Phox2bC.-c.H.. Loss-of-function experiments have demonstrated that Nkx2.2 has a prominent role for 5HT neuron production. Moreover, ETS domain factor Pet-1, Lim homeodomain containing factor Lmx1b, basic helix–loop–helix (bHLH) factor Mash1 and C4 type zinc-finger domain factor Gata2 are all required for 5HT neuron differentiation (Ding et al.,2003; Hendricks et al.,2003; Pattyn et al.,2004). Another zinc-finger domain gene Gata3 is also expressed in the 5HT progenitors and chimeras derived from wild-type and Gata3-null stem cells show that Gata3 mutant cells have a reduced contribution to the caudal, but not the rostral 5HT nuclei (van Doorninck et al.,1999).

The Iroquois (Iro/Irx) family belongs to a group of highly conserved homeodomain transcriptional regulators known as the three amino acid loop extension (TALE) homeodomain proteins, which have been identified in both invertebrate and vertebrate species serving various functions ranging from early functions in specifying neuroectoderm to late functions in subdividing the neural plate (see review by Cavodeassi et al.,2001). Genetic studies in Drosophila have revealed that a high level of Iroquois gene expression is required for the expression of the proneural bHLH genes in the dorsal mesothorax or notum (Leyns et al.,1996; Gómez-Skarmeta and Modolell,1996). Six Iroquois family of genes are identified in both human and mouse, and they are organized as two clusters (Bosse et al.,2000). Irx4 and Irx5 have been shown to play distinct roles in mammalian heart development (Bruneau et al.,2001; Costantini et al.,2005) and Iro1/Irx1 homologues have been demonstrated to play multiple roles in neural development. In Xenopus, Xiro1 is important in neural plate specification, coordinating cell cycle exit, primary neuron formation, and specification of cranial placode fields (Gómez-Skarmeta et al.,1998; de la Calle-Mustienes et al.,2002; Glavic et al.,2004). In zebrafish, a total of 11 Iroquois family of genes have been identified (Dildrop and Rüther,2004; Feijoo et al.,2004), and the iro1/irx1 homologue irx1b and a divergent irx1 paralogue irx7 have been shown to play both specific and redundant roles in hindbrain neural patterning. Ectopic expressions of both irx1b and irx7 have been shown to be able to induce ectopic expression of ngn1 (Itoh et al.,2002). Loss-of-function studies through morpholino knockdown have revealed specific and divergent roles of these two genes. Knockdown of irx7 suppresses the expression of ngn1 in trigeminal placode and anterior hindbrain, but irx1b morphant does not exhibit any alterations in ngn1 expression (Itoh et al.,2002; Lecaudey et al.,2004). In irx7 morphant, the proneural clusters that give rise to reticulospinal and motor neurons from rhombomere (r) 2 to r4 are underdeveloped (Lecaudey et al.,2004). Furthermore, double knockdown of irx1b and irx7 revealed that they share partially redundant functions in the formation of midbrain–hindbrain boundary and neural crest determination (Itoh et al.,2002). Another iro1/irx1 homologue irx1a has been shown to be important for normal heart function (Joseph,2004) and propagation of retinal neurogenesis (Cheng et al.,2006). However, the role of irx1a in hindbrain patterning and/or differentiation is still not elucidated.

Here, we characterize the contribution of irx1a to the differentiation of 5HT neurons. We show that zebrafish irx1a is the unique iro1/irx1 homologue that is expressed in the 5HT progenitors. We also provide evidence that it is required for 5HT induction but not the initiation of differentiation.


irx1a Is Expressed in the Hindbrain During Embryogenesis

Irx1 homologues are expressed extensively in the central nervous system in mouse, frog, and zebrafish (Bosse et al.,1997; Gómez-Skarmeta et al.,1998; Cohen et al.,2000; Cheng et al.,2001; Wang et al.,2001) and irx1a is transiently expressed in the cerebellum and hindbrain by the end of gastrulation (Cheng et al.,2001; Joseph,2004). We extended these observations by undertaking a detailed analysis of irx1a from early embryogenesis to 2 days postfertilization (dpf). At the beginning of somitogenesis, irx1a is distinctly expressed in the prospective midbrain and cerebellum (Cheng et al.,2001, and data not shown). Beside these 2 domains, a weak irx1a expression domain can be detected in the posterior hindbrain at the five-somite stage (Fig. 1A,C). By 20 hours postfertilizationn (hpf), irx1a is expressed in the rostral rhombomere and the caudal limit of irx1a expression in the rostral hindbrain is r1 (Cheng et al.,2001). By 24 hpf, irx1a expression persists in the rostral hindbrain and extends caudally (Fig. 1B,E). At this stage, irx1a is restricted in the ventral rhombomere but the expression is extended to the dorsal hindbrain by 36 hpf (Fig. 1C,F), especially in the dorsal commissural neurons. By 48 hpf, irx1a is expressed in a more dynamic manner and bilateral longitudinal expression of irx1a is observed in the ventral hindbrain. To determine more accurately the position of irx1a expression in the hindbrain, we compared the expression of irx1a with the floorplate marker shh by double color in situ hybridization. In the ventral hindbrain, shh is expressed in the floorplate and the ventral irx1a expression domain is just adjacent to the shh-expressing floorplate domain (data not shown). By comparing with the neurofilament staining, reticulospinal neurons are located dorsal to the irx1a ventral expression domain and ventral reticulospinal axons lie just dorsal to the irx1a ventral expression domain (Fig. 1G).

Figure 1.

Developmental expression of irx1a in the zebrafish brain. A–F: Dorsal view of flat-mounted (A–C) and a lateral view (D–F) of the zebrafish brain region. A,D:irx1a is expressed in the prospective midbrain and the anterior hindbrain regions but not in the midbrain–hindbrain boundary at the five-somite stage (5 s). Note that irx1a is also weakly expressed in the posterior hindbrain region (open arrowheads). B,E:irx1a is expressed in the tectum, cerebellum, rostral hindbrain region, and acoustic ganglion at 24 hpf. C,F: By 36 hpf, expression of irx1a can be detected in the lateral line. In the hindbrain, irx1a expression is extended to the caudal part of hindbrain and in the commissural neurons (black arrowheads). G: Dorsal view of the hindbrain region of flat-mounted 48 hours postfertilization (hpf) embryo; double in situ hybridizaton and immunostaining staining of irx1a (dark blue) and 3A10 (brown). irx1a is not expressed in the reticulospinal neurons, and the Mauthner axons (white arrow) lie just dorsal to the ventral irx1a expression domain. Asterisks (*) indicate the position of midbrain–hindbrain boundary. ac, acoustic ganglion; ce, cerebellum; ey, eye; la, lateral line; Mn, Mauthner neuron; M2, Mi2cm; M3, Mi3cm; ov, otic vesicle; r1, rhombomere 1; te, tectum; tg, tegmentum.

irx1a Is Expressed in Serotonergic Neurons

In rostral hindbrain, serotonergic neurons represent the ventral-most localized neurons (Pattyn et al.,2003) and previous study showed that 5HT reactivity is in close apposition to the ventral dendrite of the Mauthner cell (McLean and Fetcho,2004). By comparing with the 5HT immunostaining, the irx1a-expressing bilateral longitudinal columns in the rostral hindbrain corresponded to the serotonergic neurons (Fig. 2A,B,E,F). In mouse, Lmx1b and Gata3 are also expressed in the serotonergic neurons (van Doorninck et al.,1999; Ding et al.,2003; Craven et al.,2004). To confirm the irx1a ventral expression domain demarcates serotonergic neurons, we have compared the expression of irx1a with these serotonergic neuron markers. Overlap expression of irx1a, Gata3, Lmx1b1, and 5HT can be detected in the rostral–ventral irx1a-expressing bilateral longitudinal region (Fig. 2C,D,G,H). In contrast to other serotonergic neurons markers, Lmx1b1 is also expressed in the floor plate (Fig. 2I–L). Moreover, similar to the mouse Gata3 homologue (van Doorninck et al.,1999), expression of Gata3 in the ventral hindbrain is broader than the 5HT immunoreactivity as well as that of irx1a, suggesting that it may be expressed in more than one kind of neuron.

Figure 2.

Expression of irx1a in hindbrain serotonergic neurons. A–H: Dorsal view of flat-mounted (A–D) and a lateral view (E–H) of a 48 hpf zebrafish brain region. A,B,E,F:irx1a is expressed in two rows of cells in rostral ventral hindbrain (black brackets, A,E) and have the same pattern as serotonergic neurons (B,F). C,D,G,H: The serotonergic markers Lmx1b1 (C,G) and Gata3 (D,H) are also expressed in the rostral ventral hindbrain. I–L: Transverse section of zebrafish hindbrain at rhombomere 1 position which are indicated in E–H, respectively. J: The serotonergic neurons are located in the ventral hindbrain (dashed lines) and irx1a is expressed in the same region. K: Lmx1b1 is expressed in the serotonergic neurons as well as in the floor plate. On the other hand, Gata3 is expressed in the serotonergic neurons but not the floorplate, and note that the ventral expression domain of Gata3 is boarder than the serotonergic neurons. M,P: Transverse section of zebrafish hindbrain at the otic vesicle region, which is indicated in E,F, respectively. The ventral expression domain of irx1a is on the boundary of notochord (dashed line circles in M) and is overlapped with serotonergic neurons (P). N,O: The other iro1 homologue irx1b (N) nor irx5a (O) are not expressed in this domain (N).

In zebrafish, there are at least two closely related iro/irx1 paralogues, irx1a and irx1b (Cheng et al.,2001; Wang et al.,2001; Itoh et al.,2002), and one diverse paralogue irx7 (Lecaudey et al.,2001; Itoh et al.,2002). In early developmental stages, all of the iro1/irx1 homologues are expressed in the prospective hindbrain region (Wang et al.,2001; Cheng et al.,2001; Itoh et al.,2002). In later stages, although all irx1a, irx1b, and irx7 are expressed in the rostral hindbrain, only irx1a is expressed in the caudal ventral hindbrain (Fig. 2M,N, and data not shown). This irx1a-expressing caudal ventral domain is adjacent to the notochord and is reminiscent of the ventral differentiated serotonergic neurons (Fig. 2P). However, none of the other Iroquois family of genes is expressed in this region (Fig. 2O and Lecaudey et al.,2005). This expression pattern suggests that irx1a may play a unique role in serotonergic neuron specification and/or differentiation.

irx1a Is Required for the Induction of Serotonergic Neurons in Hindbrain

As irx1a is expressed in serotonergic neurons, we tested whether knockdown of irx1a had any effect on serotonergic neuron differentiation. Embryos with inactivation of irx1a show remarkable reduction of hindbrain serotonergic neurons (Fig. 3A,B,E,F). By 48 hpf, the number of 5HT neurons in hindbrain is 39.2 ± 4.8 (n = 12) in control morpholino-injected embryos and is significantly reduced (t-test; P ≤ 0.0005) in irx1a morphants (16.6 ± 6.2; n = 17). To show the specificity of the observed defects by irx1a inactivation, irx1a mRNA was co-injected with irx1a morpholino to rescue the irx1a-MO–induced phenotypes. Embryos injected with irx1a MO (Fig. 4B) show reduction of serotonergic neurons, whereas embryos co-injected with irx1a MO plus irx1a mRNA (Fig. 4C) show normal differentiation of serotonergic neuron comparable to the control MO-injected embryos (Fig. 4A). Tryptophan hydroxylases (tph) are rate limiting enzymes that catalyze the hydrozylation of tryptophan to 5HT. At least three tph genes have been identified, and raphe tph (tphR) is only expressed in the raphe serotonergic neurons (Teraoka et al.,2004). Similar to the 5HT immunostaining result, tphR is also remarkably reduced in irx1a morphant (Fig. 3C,G) and the expression of tphR and 5HT in the caudal hindbrain is completely abolished. Failure of neuron differentiation may cause ectopic cell death, and the irx1a morphant also shows increase of cell death in hindbrain, especially in the rostral region (Fig. 3D,H). Nevertheless, 5HT-positive and tphR-positive neurons can still be observed at the rostral hindbrain in all of the irx1a morphants. One possibility is that there is partial redundancy between different irx genes as irx1b is also expressed in rostral hindbrain but not caudal ventral hindbrain. Previous study has shown that irx1b and the diverse paralogue irx7 act redundantly to specify the MHB (Itoh et al.,2002). Functional compensation of different Iroquois genes is also observed in mouse loss-of-function studies (Lebel et al.,2003). Another possibility is that irx1a is only required for the differentiation of the serotonergic neurons but not the commitment of the serotonergic neuron lineage. This finding is consistent with our previous observation on the function of Irx5 in retinal bipolar cell differentiation in mouse (Cheng et al.,2005). Loss-of-function study of Irx5 in mouse also indicated that it is required for type 2 and type 3 cone-off bipolar cells differentiation but not specification.

Figure 3.

Suppression of serotonergic (5HT) neuron formation by irx1a inactivation. A–D,I–K: Embryos are injected with control morpholino (MO). E–H,M–P: Embryos are injected with irx1a MO. A,E: Lateral view of 48 hours postfertilization (hpf) rostral hindbrain region. B,C,F,G: Dorsal view, dashed lines demarcate the eye, and asterisks (*) indicate the position of midbrain–hindbrain boundary. The differentiation of serotonergic neurons is examined by the expression of 5HT and tphR. The number of differentiated serotonergic neurons in the hindbrain is largely reduced in the irx1a morphant (E–G) in comparison to control MO-injected embryos (A–C). D,H: Acridine orange fluorescent staining (lateral view of 36 hpf embryos, anterior to the left) showed that irx1a morphant (H) has ectopic cell death at rostral ventral hindbrain. I–N: Dorsal view of hindbrain region. Hindbrain segmentation is visualized by znp-1 staining. I,L: The segmentation is not altered in irx1a morphant. J,K,M,N: Mauthner neurons in rhombomere 4 (J,M, black arrows indicate the crossing of Mauthner axons) and cranial facial motor neurons (K,N) are also differentiated and patterned normally in irx1a morphants, except a mild reduction of the motor neurons in some of the morphants.

Figure 4.

Rescue of irx1a morphant phenotypes by co-injection of irx1a mRNA. A–C: Dorsal view of a flat-mounted 48 hours postfertilization (hpf) zebrafish brain region. Embryos injected with irx1a morpholino (MO, B) show reduction of serotonergic neurons, whereas embryos co-injected with irx1a MO plus irx1a mRNA (C) show normal differentiation comparable to the control MO-injected embryos (A).

Previous study showed that inactivation of irx7 resulted in mis-patterning in r4/r5 (Lecaudey et al.,2004). We wonder whether the defects in serotonergic neurons in irx1a morphants could be a result of general malformation in the hindbrain. However, irx1a morphant has intact hindbrain segmentation (Fig. 3I,L) and normal expression pattern of r3/r5 marker krox-20 (data not shown). Mauthner neurons are differentiated normally in r4, and the axons cross the midline in a normal manner (Fig. 3J,M). The number of motor neurons in the VIth motor nuclei is slightly reduced in the morphant, but the patterning of the cranial motor neurons is grossly unaffected (Fig. 3K,N). These data suggested that the hindbrain anteroposterior patterning of the irx1a morphant is grossly normal and that the loss of 5HT neurons is not due to the patterning defect in the morphant.

Normal Expression of the Ventral Markers in the irx1a Morphant

Previous studies showed that shh and the downstream transcription factor Nkx2.2 are essential for serotonergic neuron specification and differentiation (Ye et al.,1998; Pattyn et al.,2003). Loss of the differentiated serotonergic neurons in irx1a morphant raised the possibility that irx1a may regulate the expression of shh and/or Nkx2.2. However, in situ hybridization results showed that there was no major difference in the expression pattern of shh between the control and irx1a morphant by 30 hpf (Fig. 5A,E,I,M). In addition, expression of Nkx2.2 in irx1a morphant in the ventricular zone of the ventral hindbrain is also indistinguishable from that of the control embryos (Fig. 5B,F,J,N). We next examined the early markers of serotonergic neurons Lmx1b1 and Gata3. Similarly, ventral expression of both Lmx1b1 (Fig. 5C,G,K,O) and Gata3 (Fig. 5D,H,L,P) in the ventral domain can be detected in the irx1a morphants. Taken together, these results indicate that irx1a does not regulate either shh or Nkx2.2 expression, and the dorsoventral patterning in irx1a morphant is grossly normal.

Figure 5.

Early hindbrain development in the ventral domain is not affected in irx1a morphant. A–P: Lateral view (A–D,I–L) and dorsal view (E–H,M–P) of 30 hours postfertilization (hpf) zebrafish brain region. Control embryos (A–H) are not different from irx1a morphants (I–P) in floorplate expression of shh (A,E,I,M), ventral hindbrain expression of Nkx2.2 (B,F,J,N), Lmx1b1 (C,G,K,O), and Gata3 (D,H,L,P). Asterisks (*) indicate the position of the midbrain–hindbrain boundary.

Multiple lines of evidence suggest that irx1a performs late roles in serotonergic neuron differentiation. First, the early differentiated rostral 5HT neurons can still be detected in the morphants. Second, all of the 5HT neuron progenitors are derived from the Nkx2.2-positive cells in the ventral hindbrain (Hendricks et al.,1999), and the expression of Nkx2.2 is normal in the irx1a morphant. Third, expression of other early markers Lmx1b1 and Gata3 are grossly unchanged after irx1a inactivation. However, because both Lmx1b1 and Gata3 are also expressed in other cell types in the ventral hindbrain, we cannot rule out the possibility that they may also have subtle alterations in expression in the irx1a morphants.

In conclusion, this article is the first report of the dynamic expression of irx1a in the hindbrain during serotonergic neurons differentiation. We also demonstrated that irx1a is required for serotonergic neurons differentiation, but not the commitment of the progenitors.


In Situ Hybridization

Whole-mount in situ hybridization was performed as previously described (Cheng et al.,2001). Digoxigenin-labeled (Roche) antisense RNA probes for Gata3 (Neave et al.,1995), irx1a (Cheng et al.2001), irx1b, irx5a (Wang et al.,2001), irx7 (Itoh et al.,2002), Lmx1b1, Nkx2.2, shh, tphR (Teraoka et al.,2004) were synthesized by in vitro transcription. For cryosection, the whole brain was sectioned in 14-μm coronal serials.

Fluorescent and Colorimetric Immunostaining

Immunostaining was performed as previously described (Cheng et al.,2006). Primary antibodies used were as follows: mouse anti-neurofilament 3A10 (DSHB, 1:200); mouse anti-Islet1/2 (DSHB, 1:100); mouse anti-znp1 (University of Oregon, 1:200); rabbit anti-5HT (Diasorin, 1:500). Secondary antibodies used were AlexaFluor 488-conjugated and AlexaFluor 633-conjugated goat anti-rabbit IgG (Molecular Probes). Fluorescent images were captured with a Zeiss confocal microscope (Zeiss, Germany). Colorimetric staining was performed by the reaction of diaminobenzidine and H2O2.

Morpholino and Capped mRNA Injections

Standard control, two antisense morpholino oligonucleotides specific for irx1a (Cheng et al.,2006) were obtained from GeneTools. Capped RNA was transcribed with SP6 RNA polymerase using the mMessage mMachine Kit (Ambion). The co-injection experiment was performed as previously described. Embryos co-injected with 4 ng of irx1a MO and 0.16 ng of irx1a mRNA generated the highest percentage of rescued embryos.

Acridine Orange Staining

Apoptotic corpses in live embryos were stained with 5 mg/ml of the vital dye acridine orange (acridinium chloride hemi-[zinc chloride], Sigma) in embryo medium for 10 min, and then washed for 3 times of 5 min each before anesthetized with 0.016 M tricaine and mounted in agarose for confocal imaging.


S.H.C. was funded by a grant from the Research Grants Council of the Hong Kong Special Administrative Region, China, and C.-c.H. was funded by grants from the Canadian Institutes of Health Research.