The habenular nucleus influences a wide range of behaviors, from sleep and reward to fear and anxiety (Bianco and Wilson, 2009; Hikosaka, 2010; Hikosaka et al., 2008; Klemm, 2004; Lecourtier and Kelly, 2007; Okamoto et al., 2012; Sutherland, 1982), and has been a recent focus in research on addiction and depression (Aizawa et al., 2013; Hikosaka, 2010; Hikosaka et al., 2008; Viswanath et al., 2013). Despite increasing functional studies, our knowledge of its organization and neuronal properties is limited. Moreover, many vertebrates, including zebrafish, display prominent left–right (L–R) differences in this dorsal diencephalic structure (Concha and Wilson, 2001), further complicating our understanding of habenular organization, connectivity, and function.
The habenulae of mammals consist of bilaterally paired medial and lateral nuclei (Gurdjian, 1925) that have different afferent input (Herkenham and Nauta, 1977) and efferent connections (Herkenham and Nauta, 1979). The lateral nuclei project to a variety of areas in the midbrain and hindbrain, whereas the medial nuclei chiefly innervate the unpaired interpeduncular nucleus (IPN) in the ventral midbrain (Herkenham and Nauta, 1979). Axons from both nuclei contribute to the fasciculus retroflexus nerve bundles to form one of the most highly conserved conduction systems in the vertebrate brain (Herrick, 1948). Although it was initially suggested that zebrafish only possess a habenular region analogous to the medial nucleus of mammals (Concha and Wilson, 2001), further characterization of molecular and anatomical properties confirmed the presence of both a dorsal and ventral nucleus (Amo et al., 2010), which, respectively, correspond to the medial and lateral habenulae of mammals.
In several vertebrate species, the medial/dorsal and lateral/ventral nuclei have been further subdivided into discrete neuronal populations or subnuclei largely based on morphological criteria (Contestabile et al., 1987; Iwahori, 1977; Marburg, 1944; Quina et al., 2009; Wagner et al., 2014). Four subnuclei were initially recognized in the medial nucleus of the well-studied rat habenula by neuronal shape and packing, neuropil density, and neurotransmitter immunoreactivity (Andres et al., 1999; Contestabile et al., 1987; Geisler et al., 2003). Antibody labeling demonstrated that neurons located more dorsally produce substance P (Cuello et al., 1978), while those in more ventral regions are cholinergic (Eckenrode et al., 1987). More recent analyses of gene expression corroborate this major sub-division, as well as demonstrate that genes from the vesicular glutamate transporter (vglut) family are expressed throughout the habenulae (Aizawa et al., 2012; Quina et al., 2009). The dorsal region can be further divided into glutamatergic-only and substance P-expressing/glutamatergic neuronal populations by the localization of transcripts for vglut 1 and 2 and the substance P precursor, tachykinin 1 (tac1) (Aizawa et al., 2012; Quina et al., 2009). Immunolabeling indicates that the cholinergic neurons largely project to the central core of the IPN, whereas substance P-expressing neurons innervate peripheral subnuclei (Contestabile et al., 1987). Additional combinatorial patterns of gene expression reveal 5 distinct subnuclear regions in the medial habenula of the adult rat (Aizawa et al., 2012), and a recent study suggests a similar organization for the mouse (Wagner et al., 2014).
Discrete subnuclei identified in the dorsal habenulae of amphibians and fish show prominent differences in their size and organization between the left and right sides of the brain (refer to Concha and Wilson, 2001). For example, in Rana esculenta, the right dorsal habenula is a single nucleus, while the left contains lateral and medial subnuclei, which innervate the IPN via different routes (Braitenberg and Kemali, 1970; Kemali and Guglielmotti, 1984). Substance P neurons are present in the left and right dorsal habenulae; but only in the lateral subnucleus on the left side (Kemali and Guglielmotti, 1984).
In zebrafish, where the dorsal habenulae have emerged as a powerful genetic model to study the development and function of L–R asymmetry in the brain, subnuclear organization has been defined on the basis of patterns of gene and transgene expression. For example, combinatorial expression of the potassium channel tetramerization domain containing (kctd) genes kctd12.1, ktcd12.2, and ktcd8 revealed six molecularly distinct domains at the larval stage, with three differing in size between the left and right dorsal habenula, two unique to the left, and one unique to the right (Gamse et al., 2005). The transgenic line Tg(brn3a–hsp70:GFP), which was generated using regulatory sequences from the POU domain, class 4, transcription factor 1 (pou4f1) gene, labels neurons in the adult dorsal habenula in a mostly non-overlapping pattern with kctd12.1 expression (Aizawa et al., 2005). The brn3a and kctd12.1 domains have been proposed to correspond to the medial (dHbM) and lateral (dHbL) subnuclei of the dorsal habenulae, respectively (Aizawa et al., 2005). Integration into the neuronal pentraxin 2a (nptx2a) locus generated the Tg(nptx2:Gal4-VP16) driver line, which also activates reporter gene expression in a largely complementary pattern to Tg(brn3a–hsp70:GFP) in the dHbL of the adult brain (Agetsuma et al., 2010). Habenular axons emanating from the dHbL are thought to innervate the dorsal, intermediate, and most dorsal part of the ventral IPN, while those originating from dHbM neurons project to the intermediate and ventral IPN (Agetsuma et al., 2010; Aizawa et al., 2005; Gamse et al., 2005).
Transgenic lines have also been used to assess habenular L–R asymmetry. The Tg(brn3a–hsp70:GFP) labeled neuronal population is larger in the right habenula while, conversely, in nptx2:Gal4-VP16; UAS:DsRed2 double heterozygous fish more DsRed2-positive neurons are located in the left habenula. The size asymmetry between the dHbM and dHbL subnuclei has been attributed to a higher number of early born dHbL neurons in the left dorsal nucleus and later born dHbM neurons enriched on the right (Aizawa et al., 2007). Prolonging the early period of neurogenesis may also increase the dHbL subnuclei and decrease the dHbM, resulting in bilaterally symmetric dorsal habenulae (Doll et al., 2011).
The goal of this study was to determine how distinct neuronal populations defined by neurotransmitter phenotype correspond to the previously designated subnuclear regions of the dorsal habenulae. Through double labeling in the larval and adult brain, we provide a consolidated molecular map of the dorsal habenular subnuclei and demonstrate that L–R asymmetric cholinergic and peptidergic subregions only partially overlap with the previously proposed dHbL and dHbM subnuclear organization. By comparing the results in zebrafish with published studies on the medial habenulae of the adult rodent brain, we find that specific neurotransmitter-expressing domains and their spatial relationships are mostly conserved, even though L–R differences are not. Precise definition of subnuclei is important when monitoring neuronal activity and interpreting behavioral responses in efforts to understand how the habenular region of the brain regulates its diverse functions. Moreover, performing such analyses using the zebrafish model could shed light on the significance of directional asymmetry in an essential neural pathway.
MATERIALS AND METHODS
Zebrafish were raised and housed at 27°C on a 14/10 h light dark cycle. We used the wild-type AB strain (Walker, 1999) and the transgenic lines TgBAC(gng8:Eco.NfsB- 2A-CAAX-GFP)c375 (deCarvalho et al., 2013), Tg(vglut2a: loxP-DsRed-loxP-GFP)nns9 (Miyasaka et al., 2009), Tg(brn3a-hsp70:GFP)rw0110b (also known as Tg(pou4f1-hsp70l:GFP); Aizawa et al., 2005), Tg(nptx2:Gal4-VP16)rw0143a (also known as Tg(narp:Gal4-VP16); Agetsuma et al., 2010), Tg(UAS:DsRed2)rw0135 (Agetsuma et al., 2010), Tg(4xnrUAS:GFP)c354 (Akitake et al., 2008), and Tg(14xUAS:BGi-NLS-emGFP)y262 (H. Burgess, personal communication). For simplicity, we refer to these transgenic lines as Tg(gng8:GFP), Tg(vglut2a:DsRed), Tg(brn3a:GFP), Tg(nptx2:Gal4), Tg(UAS:DsRed2), Tg(4xUAS:GFP), and Tg(14xUAS:GFP). Maintenance of zebrafish and experimental procedures on larvae and adults were performed in accordance with the protocol approved by Carnegie Institutional Animal Care and Use Committee.
RNA in situ Hybridization and Immunofluorescence
Larvae and dissected adult brains were fixed in 4% paraformaldehyde (in PBS) at 4°C overnight. Single and double label colorimetric RNA in situ hybridization experiments were performed as described previously for whole larvae (Gamse et al., 2003) and modified for adult brain tissue (Gorelick et al., 2008). Single and double fluorescent RNA in situ hybridization (FISH) or FISH coupled with immunolabeling for green fluorescent protein (anti-GFP rabbit antibody; Torrey Pines TP401) was performed as in a prior study (deCarvalho et al., 2013).
To synthesize antisense RNA probe for amine oxidase, copper containing 1 (aoc1), a 1019 base-pair fragment was PCR-amplified from the cDNA clone MGC154101 (Thermo Scientific) using primers GTCACTGAATACATCGTTGGCCC and TTGTAGACTGTAGATGTAGTTCTGATC and sub-cloned into the pCRII-TOPO vector using the TOPO TA Cloning kit (Invitrogen). pCRII-TOPO-aoc1 was linearized with ApaI and RNA transcribed with SP6 RNA polymerase. For anoctamin 2 (ano2), an EcoRI to NotI fragment from the cDNA clone MGC171498 (Open Biosystems) was subcloned into the pBluescript SK(−) vector, linearized with EcoRI, and transcribed with T3 RNA polymerase. A zebrafish homolog to Drosophila muscleblind-like 3 (mbnl3) was obtained from a zebrafish kidney cDNA library and cloned into the EcoRI and XhoI sites of the pBK-CMV vector. pBK-CMV-mbnl3 was linearized with SalI and RNA transcribed with T7 RNA polymerase.
Methods for cloning and probe synthesis have been described previously for guanine nucleotide binding protein (G protein), gamma 8 (gng8; Thisse and Thisse, 2004), kctd12.1 (previously known as leftover; Gamse et al., 2003), kctd8 (previously known as dexter; Gamse et al., 2005), kctd12.2 (previously known as right on; Gamse et al., 2005), solute carrier family 18 (vesicular acetylcholine), member 3b (slc18a3b; commonly known as vachtb; Hong et al., 2013), somatostatin1.1 (sst1.1; Thisse and Thisse, 2004), and tachykinin1 (tac1; Hong et al.,2013).
Sectioning and Microscopy
Larval and adult brains were embedded in 4% low melt agarose (Lonza) in PBS (100 g/ml) and sectioned with a VT1000S vibratome (Leica Microsystems) from 50 to 250 µm as indicated. For labeling cell nuclei, sections were stained in 4′,6-diamidino-2-phenylindole (DAPI) (Life Technologies) in PBS with 0.1% Tween for 20 min and rinsed in PBS. Sections were mounted in either 50% glycerol (1:1 [v/v] with H2O), Aqua-Poly/Mount (Polysciences) or Prolong Gold (Invitrogen) anti-fade mounting media under cover slips.
Bright-field images were captured by an Axiocam HRc digital camera mounted on an Axioskop (Carl Zeiss). Fluorescent images were collected using a Leica SP5 confocal microscope and processed either with ImageJ (National Institutes of Health) or Imaris (Bitplane).
As summarized in Table 1, an increasing number of genes have been found to be expressed in the bilaterally paired dorsal and ventral habenulae of larval and adult zebrafish, and many of their homologs are transcribed in the corresponding medial and lateral habenular nuclei of rodents. For example, similar to expression of anoctamin 1 (ano1) in the medial habenula of the mouse (Quina et al., 2009), ano2 is transcribed in a bilaterally symmetric pattern throughout the dorsal nucleus of larval zebrafish (Fig. 1a), and is a useful marker for demarcating its boundaries. In zebrafish, the ventral nucleus was defined by its expression of the amine oxidase, copper containing 1 (aoc1) gene (Amo et al., 2010 and refer to Fig. 1b). We find that a zebrafish homolog of the muscleblind-like 3 (mbnl3) gene also exhibits bilaterally symmetric expression throughout the presumptive ventral nucleus (Fig. 1c,d), providing further evidence that the ventral habenulae have a distinct molecular identity. However, not all genes are transcribed exclusively in the dorsal or ventral nucleus. Transcripts for the guanine nucleotide binding protein (G protein), gamma 8 (gng8) gene, for instance, are distributed in a bilaterally symmetric pattern similar to ano2 (Fig. 1e,f) but are also found in a subset of neurons in both ventral nuclei (arrowheads in Fig. 1e). A transgenic line produced by integration of membrane-tagged green fluorescent protein gene into the gng8 locus (deCarvalho et al., 2013) labels the same pattern of dorsal and ventral habenular neurons (refer to Fig. 2a).
Table 1. Gene Expression in the Zebrafish Habenulae and Homology with Rodents
Additionally, a number of genes show L–R asymmetric expression in the dorsal habenular nuclei (e.g., Agetsuma et al., 2010; Aizawa et al., 2005; Amo et al., 2010; Gamse et al., 2005; Kuan et al., 2007; Taylor et al., 2011 and refer to Table 1). Members of the potassium channel tetramerization domain containing gene family were first discovered to exhibit L–R differences, with transcripts for kctd12.1 and kctd8 more extensive on the left and right sides, respectively (Gamse et al., 2005 and Fig. 1g,i). Double labeling confirms that expression of both genes is confined within the limits of the ano2-expressing dorsal habenular nuclei (Fig. 1h,j). kctd12.2 is not only expressed to a greater extent in the right dorsal nucleus (Fig. 1k), but also bilaterally in a subset of cells in the ventral habenular nuclei (Fig. 1l, arrowheads). Although gene expression patterns such as these appear to demarcate discrete subregions within the dorsal habenulae, how such asymmetric domains correlate with functional subnuclei is unclear.
Discrete Asymmetric Cholinergic and Peptidergic Subnuclei of the Larval Dorsal Habenulae
In mammals, as described above, habenular subnuclei have been identified on the basis of neurotransmitter phenotype and neuronal connectivity. Recent work by Hong et al. (2013) demonstrated that the zebrafish larval habenula contains glutamatergic and cholinergic neurons as early as 4 days post-fertilization (dpf). We determined the position of these neuronal groups more systematically using the gng8:GFP transgenic background or with respect to the asymmetric expression of kctd12.1. In Tg(gng8:GFP) larvae, neurons labeled with membrane-tagged GFP are found throughout the dorsal habenulae (Fig. 2a,b) and more sparsely in the ventral nucleus (arrowheads). Co-labeling with vglut2a:DsRed indicates that these neurons are all glutamatergic (Fig. 2c,d). In the ventral nucleus, vglut2a:DsRed expression is more widespread than gng8:GFP.
Expression of the vesicular acetylcholine transporter b gene (vachtb) demarcates the L–R asymmetric cholinergic regions of the dorsal habenulae as well as the bilaterally symmetric ventral nuclei (Hong et al., 2013). Double fluorescence RNA in situ hybridization demonstrates that vachtb and kctd12.1 have complementary patterns of expression (Fig. 2e,f); thus, the kctd12.1-positive region corresponds to the non-cholinergic territory of the dorsal habenulae. In mammals, this non-cholinergic region contains substance P-expressing neurons. Transcripts for the zebrafish substance P precursor tachykinin (tac1) are not detected in the habenular region as early as 4 dpf, although they are present elsewhere in the brain (data not shown). However, we find that the neuropeptide encoding gene somatostatin1.1 (sst1.1) is expressed at this stage, in a lateral subdomain of the dorsal habenula (Fig. 2g,h). The sst1.1 neuronal population is located within the non-cholinergic kctd12.1-expressing territory and is larger on the right side of the brain than the left (Fig. 2i,j). Analyses of the combinatorial gene expression patterns in 4 dpf larvae, therefore, indicate that neurons of the dorsal habenulae are all glutamatergic, with a portion of them possessing either a cholinergic or sst1.1 identity. In the larval brain, the non-overlapping cholinergic and somatostatin neuronal populations are significantly larger in the right dorsal habenula compared to the left.
Asymmetric Neurotransmitter-Specific Subpopulations Are Retained in the Adult Dorsal Habenulae
We next examined whether L–R asymmetric neuronal populations are maintained during the morphological changes that accompany formation of the adult zebrafish brain (refer to Amo et al., 2010). As in the larva, neurons throughout the dorsal and ventral habenular nuclei of adult zebrafish are glutamatergic as revealed by DsRed labeling from the vglut2a transgenic reporter (Fig. 3a). In the adult, the cholinergic territory is more extensive in the left habenula (Fig. 3b,c) compared to the larval stage (Fig. 2e,f). At the border, cholinergic and non-cholinergic cells are clearly distinct populations in the left dorsal nucleus, whereas some intermixing of vachtb and kctd12.1-expressing cells is observed in the right habenula (arrowhead, Fig. 3c). The sst1.1 and vachtb expression domains are also non-overlapping in the left habenula (Fig. 3d,e), although a subset of cells appears to express both genes in the right dorsal habenula (Fig. 3e, arrowhead). The neuropeptide encoding genes sst1.1 and tac1 are expressed by distinct neuronal populations within the kctd12.1-positive, non-cholinergic territory (Fig. 3f,g and Fig. 3h,i, respectively) (Fig. 3j,k). As in the larva, the sst1.1-expressing domain is larger in the right habenula (Fig. 3f,j). Overall, the cholinergic and peptidergic territories are preserved between the larval and adult stages; however, some neurons in the right habenula exhibit an overlap in sst1.1 and vachtb expression that was not detected at the larval stage.
Partial Correspondence of Neurotransmitter-Expressing Subdomains with Habenular Transgenic Reporters
Zebrafish transgenic lines have been used to define dorsal habenular subregions, to trace axonal connections to the IPN, and in behavioral studies following neuronal ablation or inactivation (Agetsuma et al., 2010; Aizawa et al., 2005; Lee et al., 2010). Notably, Tg(brn3a:GFP) was designated as labeling the dHbM subnucleus and Tg(nptx2:Gal4) as driving expression in the dHbL (Agetsuma et al., 2010; Aizawa et al., 2005; Okamoto et al., 2012). After establishing the gene expression patterns and neurotransmitter identity of habenular subregions, we sought to determine how this regionalization corresponds with habenular labeling by these transgenes.
We mated Tg(brn3a:GFP) and Tg(vglut2a:DsRed) carriers and observed that all GFP-positive cells in the habenulae of their larval progeny co-express DsRed (Fig. 4a,b), indicating that brn3a:GFP-positive cells are glutamatergic. In the dorsal habenulae of adults, the brn3a:GFP labeled region was found to overlap only partially with the cholinergic territory as defined by vachtb expression (Hong et al., 2013). We observed a similar pattern at the larval stage, with distinct neuronal populations that were either co-labeled, vachtb positive/brn3a:GFP negative (Fig. 4c, arrowhead) or vachtb negative/brn3a:GFP positive (Fig. 4c, arrow). Thus, the dorsal habenular subregion labeled by the brn3a:GFP transgene does not fully encompass a cholinergic subnucleus. Although the brn3a-GFP labeled neurons were designated as a distinct subnucleus in the adult habenula from that labeled by Tg(nptx2:Gal4); Tg(UAS:DsRed2) (Agetsuma et al., 2010), some neurons co-label with both fluorescent proteins in the larval dorsal habenula (Fig. 4d, arrowheads).
Consistent with their location in the kctd12.1 positive non-cholinergic territory, sst1.1-expressing neurons were not co-labeled by the brn3a:GFP transgene (Fig. 4e). These cells were found in a similar region of the dorsal habenula as neurons labeled by the nptx2:Gal4 and UAS:DsRed2 transgenes. Upon determining whether the sst1.1 neurons co-express nptx2:Gal4, we unexpectedly found that the extent of labeling varied significantly depending on the UAS-regulated transgenic reporter that was activated by this Gal4 driver (cf. Fig. 4f–h). However, despite the variability in labeling between lines, only a subset of sst1.1-expressing neurons was co-labeled by the nptx2:Gal4 driver for the three UAS-regulated reporters tested (Fig. 4i). Thus, expression driven by Tg(nptx2:Gal4) does not encompass a discrete somatostatin neuronal population.
Habenular Connectivity with the IPN
Although transgenic labeling in the dorsal habenulae does not appear to correspond to specific neuronal populations defined by neurotransmitter/neuropeptide expression, it is still useful for examining connectivity with the IPN. In zebrafish larvae and adults, the gng8:GFP transgene labels all dorsal habenular neurons (Fig. 2a) and their GFP-positive axons that terminate throughout the IPN (Fig. 5a,d,g,j). Because GFP is membrane-tagged in this line, discrete axonal densities can be recognized in a stereotypic pattern in sagittal sections of the adult IPN (Fig. 5a), with a minimum of two dorsal (D-i and D-ii) and three ventral clusters (V-i, V-ii, and V-iii). An intermediate (I), horseshoe-shaped zone of innervating fibers separates the dorsal and ventral axonal bundles. The stereotypic morphology of axonal bundles at the IPN is already apparent at 4 dpf (Fig. 5j).
Using Tg(brn3a:GFP) zebrafish, it had been previously shown that GFP labeled habenular axons predominantly terminate in the intermediate and ventral regions of the adult IPN (Aizawa et al., 2005) and axons labeled by Tg(nptx2:Gal4);Tg(UAS:DsRed2) innervate the dorsal, intermediate, and ventral IPN (Agetsuma et al., 2010). We corroborated these results and found that, more specifically, in the ventral IPN of adult zebrafish, the brn3a:GFP terminals primarily contribute to the V-i and V-iii axon bundles (Fig. 5b,e) and those labeled by the nptx2:Gal4 driver and UAS:DsRed2 reporter innervate the V-ii bundle (Fig. 5c,f).
By mating Tg(brn3a:GFP) and Tg(nptx2:Gal4);Tg(UAS:DsRed2) carriers, the relative extent of GFP and DsRed2 axon terminals along the dorsoventral IPN was examined in transgenic larvae. At 4 dpf, a largely similar pattern of innervation was found in IPN subregions as in the adult (Fig. 5h,k). Because neurons in the left habenula were more extensively labeled when the nptx2:Gal4 transgenic driver was used to drive 4xUAS:GFP (Fig. 4g,h), we compared the profiles of innervating axons at the IPN using the two fluorescent reporter lines. Habenular efferents were weakly labeled by the UAS:DsRed2 transgene in subregions of the dorsal and ventral IPN (Fig. 5h,k). Significantly more GFP-labeled axon terminals were observed in the D-i, D-ii and, I regions with the 4xUAS:GFP line (Fig. 5i,l), consistent with this reporter being activated to a greater extent in the left habenula (Fig. 4g,h).
Because brn3a:GFP labels many cholinergic neurons but not sst1.1-expressing neurons, and the axons of brn3a:GFP positive neurons predominantly innervate the D-ii, I, V-i, and V-iii regions of the IPN, we expect that these IPN regions receive cholinergic input. With only partial overlap between sst1.1- and nptx2:Gal4-expressing populations, it is difficult to infer the precise pattern of connectivity of this neuronal subgroup with certainty; however, efferents from the somatostatin neurons are expected to contribute to some of the same subregions of the IPN as the neuronal populations labeled by the nptx2 driver (Fig. 5k,l).
In the nervous system, it is generally assumed that discrete clusters of neurons exhibiting the same neurotransmitter phenotype fasciculate to innervate specific target regions and function together. In addition to their shared expression of particular neurotransmitters or neuropeptides, brain nuclei or their component subnuclei are defined by neuroanatomical features such as cell cluster size, density of cell packing, neuropil structure, or stereotypic position. The significance of nuclear designations is especially relevant when modern genetic methods are used to eliminate or modify brain regions for insights into their roles in physiology and behavior. These approaches have been applied to the study of the habenular nuclei in mammalian and non-mammalian systems (Agetsuma et al., 2010; Lee et al., 2010; Quina et al., 2009; Yamaguchi et al., 2013); however, detailed information about the precise subregions that are affected would allow more rigorous functional correlations.
The goal of this study was to generate a consolidated map of the habenular region in larval and adult zebrafish, by reexamining the expression of widely used genetic markers, describing the spatial expression patterns of newly identified genes, and determining how this information corresponds with neurotransmitter profiles and cell populations labeled by transgenic reporters. In addition, we further characterized the extent of L–R asymmetry in the dorsal habenular nuclei.
Designation of Habenular Subnuclei
The dorsal habenulae in the adult zebrafish brain have been subdivided into medial and lateral subnuclei on the basis of Tg(brn3a:GFP) labeling and by expression of kctd12.1 or directed by the Tg(nptx2:Gal4) driver (Agetsuma et al., 2010; Aizawa et al., 2005). The designation of the medial subdivision is further supported by experiments showing that brn3a:GFP positive and negative neurons have different birthdates, and that L–R differences in the timing of neurogenesis later contributes to the formation of asymmetric brn3a-labeled populations. Several lines of evidence, however, suggest that the regions defined by these transgenic tools may not represent discrete, functional subnuclei. Although initially defined as corresponding to distinct subnuclei, the kctd12.1 and brn3a domains do not appear to be completely segregated in the adult dorsal habenulae (Aizawa et al., 2005). In addition, the reported patterns of brn3a:GFP and nptx2:Gal4; UAS:DsRed2 labeled neurons fail to recapitulate endogenous gene expression, which appears more widespread in the habenular region for both genes (Agetsuma et al., 2010; Aizawa et al., 2005, 2007).
Partial overlap in labeling from the brn3a:GFP and nptx2:Gal4 transgenes in the larval habenula further suggests that expression of these markers does not correspond to discrete subnuclei. Although the brn3a:GFP population significantly overlaps with vachtb-expressing neurons in the larval dorsal habenula, a subset of cholinergic neurons are not labeled by this transgene. Similarly, the region of the dorsal habenulae labeled by the nptx2:Gal4 transgenic driver only partially overlaps with somatostatin1.1-expressing neurons. Together, these findings suggest that caution should be taken in interpreting brn3a:GFP and nptx2:Gal4; UAS:DsRed2 transgenic labeling as representative of functional habenular subnuclei at least for larval stages.
Distinct Neuronal Populations in the Zebrafish Dorsal Habenula Defined by Neurotransmitter Identity
The significance of subdomains of gene expression in the zebrafish habenulae is largely unknown. As a first step, we sought to determine how they might correlate with specific neuronal subgroups by examining the expression of zebrafish homologs for genes indicative of neurotransmitter identity. This approach avoids potential problems associated with using antibodies that are typically generated against mammalian antigens. We examined larval and adult stages, as some neurons are known to undergo changes in neurotransmitter phenotype over time (e.g., Landis and Keefe, 1983; Spitzer, 2012) and new neuronal subgroups may also arise.
The analysis of larval zebrafish revealed that cholinergic neurons are a distinct population from cells expressing kctd12.1, which encodes a protein shown to associate with gamma-aminobutyric B (GABAB) receptors to influence their pharmacology and signaling properties (Ivankova et al., 2013; Schwenk et al., 2010) and to promote formation of dense neuropil in the zebrafish dorsal habenula (Taylor et al., 2011). Instead, we found that the kctd12.1 domain contains somatostatin1.1-expressing peptidergic neurons.
All neurons of the zebrafish dorsal and ventral habenulae are glutamatergic. In the asymmetric dorsal habenulae of larval zebrafish, the left habenular nucleus is mainly composed of a glutamatergic-only subregion with additional small cholinergic and sst1.1 clusters. In contrast, the right nucleus largely consists of expanded cholinergic and sst1.1 subregions.
During maturation of the zebrafish brain to its adult form, the presumptive ventral nuclei become repositioned ventromedially (Amo et al., 2010), but the relative positions of subnuclear territories within the dorsal nuclei largely appear the same. Left–right differences in the size of neurotransmitter expressing populations, however, are less pronounced. Notably, at 4 dpf there is only a small cluster of cholinergic neurons in the left dorsal habenula, but by adulthood the proportion of neurons that express cholinergic genes is more similar between the left and right nucleus. In adults, both habenula also acquire a tac1-expressing subdomain (Hong et al., 2013), in addition to the cholinergic, sst1.1 and glutamatergic-only subregions.
Curiously, while the boundaries between these four neuronal populations are discrete in the left dorsal habenula, they are not as well demarcated on the right, due to intermixing of neurons expressing different neurotransmitters. It is the left habenula of zebrafish that receives preferential innervation from the parapineal organ, an accessory to the pineal organ, which in most individuals is asymmetrically positioned to the left of the pineal. The parapineal influences the gene expression and neuroanatomical properties of the left dorsal habenula that distinguish it from the right (Concha et al., 2003; Gamse et al., 2003). It is tempting to speculate that parapineal neurons may also serve to refine subnuclear organization, ensuring the formation of precise boundaries between peptidergic and cholinergic clusters in the left habenula. In certain frogs, the photoreceptive frontal organ sends asymmetric projections across the left habenular nucleus (Eldred et al., 1980; Guglielmotti and Cristino, 2006; Kemali and De Santis, 1983) and the left habenula contains distinct cholinergic and substance P subnuclei (Kemali and Guglielmotti, 1984; Marin et al., 1997). The right habenula also has cholinergic and substance P neurons (Kemali and Guglielmotti, 1984; Marin et al., 1997); however, as in zebrafish, the boundaries between these neuronal populations are not as clearly delineated as on the left. In mammals, where there is no structure analogous to the parapineal, deep pineal neurons project to both medial habenular nuclei (Korf et al., 1990), and both exhibit distinct cholinergic and substance P subnuclei (Cuello et al., 1978). Whether innervating parapineal or pineal neurons play a role in the refinement of the habenular map into discrete subnuclear compartments remains to be demonstrated experimentally.
From the analysis of neurotransmitter and neuropeptide producing neuronal populations, we present a schematic map of the dorsal habenulae of larval and adult zebrafish (Fig. 6a). It is important to stress that this is only a preliminary map, as it is highly likely that additional neuropeptides or transmitters are co-expressed with glutamatergic neurons and located in other subregions of the dorsal habenulae. For example, cells immunoreactive for thyrosine hydroxylase important in L-DOPA synthesis have been identified in the medial habenulae of alpaca (Marcos et al., 2013) and mRNA encoding orphanin FQ as well as this neuropeptide was localized to the rat medial habenulae (Neal et al., 1999). Indeed, as indicated in Table 1, genes encoding tachykinin 3a (Biran et al., 2012; Ogawa et al., 2012) and receptors for glutamate (Haug et al., 2013), serotonin (Norton et al., 2008), and hypocretin (Appelbaum et al., 2009) show enriched expression in subregions of the zebrafish dorsal habenulae, suggestive of other neuronal specializations.
Designating subnuclei in the zebrafish habenulae on the basis of neurotransmitter identity enables a more direct comparison with mammalian brains, in efforts to uncover the habenular subregions and circuitry modulating a wide variety of behaviors. In rodents, the medial habenulae are bilaterally symmetric and contain glutamatergic neurons (Barroso-Chinea et al., 2007; Qin and Luo, 2009), which have been subdivided into dorsal and ventral subnuclei depending on whether they use substance P or acetylcholine (Aizawa et al., 2012; Contestabile et al., 1987; Cuello et al., 1978; Lecourtier and Kelly, 2007; Quina et al., 2009). Recent work reveals an additional glutamatergic-only subnucleus, which is found medially to the substance P subnucleus (Aizawa et al., 2012). The location and organization of subnuclei designated by neurotransmitter phenotype in the dorsal habenulae of adult zebrafish are strikingly similar to that of the rodent medial habenulae. An exception is the presence of somatostatin neurons, which has not yet been described for the rodent medial habenula. Somatostatin-immunopositive fibers are found only in the lateral habenulae of the postnatal rat (Shiosaka et al., 1981). However, images from the Allen Brain Atlas indicate that the sst gene is transcribed in neurons in the medial habenula of the mouse, in the same region where cholinergic neurons are located. Further studies are needed to ascertain the location of these neurons relative to the cholinergic population, to determine whether there is a distinct somatostatin-producing subnucleus as in the zebrafish dorsal habenula.
Connectivity of Dorsal Habenular Subnuclei with the IPN
Neurons of the medial habenula of mammals project their axons through the fasciculus retroflexus to terminate at the interpeduncular nucleus (refer to Kappers et al., 1936; Sutherland, 1982). From the accumulations of terminating fibers, the structure of the IPN has been well defined in rodents as consisting of three unpaired (rostral, apical, and central) and four paired (intermediate, lateral, rostral lateral, and dorsal lateral) subdivisions in a consensus nomenclature put forward by Lenn and Hamill (1984).
Lesioning studies, histochemistry, and immunoreactivity have revealed the presence of a wide variety of neurotransmitters and neuropeptides in axons terminating within particular subdivisions of the IPN (reviewed by Morley, 1986).
Visualization of a transgenic line with expression of membrane-tagged GFP under the control of gng8 regulatory sequences (deCarvalho et al., 2013) provides a complete picture of innervation of the zebrafish IPN by dorsal habenular neurons. As with previous dye labeling and immunolabeling experiments (Aizawa et al., 2005; Gamse et al., 2005; Tomizawa et al., 2001), the axon terminals of habenular neurons can be readily distinguished as contributing to bundles in dorsal, intermediate, and ventral regions of the larval and adult IPN. However, because our understanding of the zebrafish is far from complete, it is premature to assign a nomenclature analogous to that proposed for mammals (Lenn and Hamill, 1984). We, therefore, designated IPN afferent input as consisting of two dorsal (D-i and D-ii) and three ventral (V-i, V-ii, and V-iii) axonal densities and an intermediate fiber bundle (I) that separates them on the basis of the stereotypic morphology of GFP labeled afferents (and refer to Fig. 11 in Agetsuma et al., 2010). Moreover, we find that the pattern of innervating fibers is already in place in the IPN of 4 dpf larvae. This simplified naming strategy must be modified as more information is obtained about the precise homology with mammalian IPN subdivisions and the distribution of synaptic terminals expressing different neurotransmitters.
Previous work had demonstrated that that brn3a:GFP and nptx2:Gal4 transgenic lines primarily label different neuronal populations in the dorsal habenulae of adult zebrafish, which innervate largely distinct dorsoventral regions of the IPN (Agetsuma et al., 2010). Overall, our findings support this connectivity map for the adult, with the potential for overlapping input in the D-ii and I IPN subregions (shown in Fig. 6b). In 4 dpf larvae, these transgenes label innervating fibers throughout the IPN, although they are more concentrated in certain subregions at this early stage. For example, brn3a:GFP labeled axons predominate in the V-i density and to a lesser extent in V-iii. Because the pattern of neuronal labeling in the larval habenula depends on the UAS reporter under nptx2:Gal4 control, the location of labeled axons also varied between lines. With the UAS:DsRed2 reporter that labeled neurons sparsely in both habenulae, axonal endings were enriched in the D-i and V-ii bundles, relative to the brn3a:GFP labeled terminals. However, when a significantly larger group of neurons in the left habenula was labeled by 4xUAS:GFP, innervating axons were primarily detected in the D-i, D-ii, and I regions of the IPN.
Differences in the pattern of IPN innervation between larval and adult stages, such as the strongly labeled V-ii terminal bundle present in nptx2:Gal4 transgenic adults or the increase in brn3a:GFP labeled terminals in the V-iii region, are likely the outcome of differential neurogenesis in subregions of the developing dorsal habenular nuclei and continued axonal outgrowth to target regions. Understanding how and when new neuronal populations arise, such as the expanded cholinergic cluster on the left or the bilateral tac1-expressing groups, will help clarify how the habenulo-interpeduncular connectivity map evolves over time.
We can predict the most likely sites of cholinergic and peptidergic afferent input to the IPN from the partial overlap between neurotransmitter expression and transgene labeling in dorsal habenular neurons. As depicted schematically in Figure 6c, cholinergic neurons are expected to terminate in the D-ii, I, and V IPN regions, owing to the majority of them being labeled by brn3a:GFP. Based on partial overlap with the nptx2:Gal4 driver and lack of co-expression with brn3a:GFP, the contribution of sst1.1-expressing neurons would be limited to the dorsal IPN, intermediate IPN, and V-ii region of the ventral IPN. A similar profile is expected for substance P neurons in the adult brain based on the location of tac1-expressing neurons within the brn3a:GFP negative region (Hong et al., 2013). Although the proposed connectivity map provides a useful guide, it is an oversimplification that rests on indirect inferences and will need to be validated by the production and analysis of transgenic lines that selectively label each neurotransmitter-expressing population of the dorsal habenulae.
L–R Asymmetry in IPN Connections
The value of having a framework to classify afferent input is that it helps resolve a discrepancy concerning how left and right habenular neurons connect with the IPN in the zebrafish brain. One model suggests that the left habenula primarily projects to the dorsal IPN and the right habenula innervates the ventral IPN in a laterotopic manner (Aizawa et al., 2005; Bianco et al., 2008). Results from dye labeling, immunolabeling, and transgenic labeling of habenular axons, however, indicate a more complicated scenario with differing proportions of neurons from the left and right sides of the brain contributing to both the dorsal and ventral IPN (Agetsuma et al., 2010; Beretta et al., 2012; Gamse et al., 2005; Kuan et al., 2007; Okamoto et al., 2012). For the right habenula of the larval brain, we hypothesize that the significantly larger cholinergic subnucleus innervates the ventral IPN. In the left habenula, the smaller cholinergic neuronal cluster would also project to the ventral IPN while the larger population of non-cholinergic neurons would innervate the dorsal IPN.
This revised model can explain two previous puzzling observations. First, when the parapineal is ablated and the left dorsal habenula adopts the molecular profile and neuroanatomical properties of the right dorsal habenula, the vast majority of habenular efferents project to the ventral IPN (Gamse et al., 2005; Kuan et al., 2007). These projections are presumed to emanate from the large cholinergic subnucleus that, following loss of the parapineal, is present in the left as well as the right dorsal habenula (K. Santhakumar, T. deCarvalho and M.E. Halpern, unpublished observations). However, a small axonal bundle remains at the dorsal IPN (Gamse et al., 2005) that may represent the efferents of peptidergic neurons present in both habenulae. Second, a study analyzing the projections of individual neurons described distinct axonal morphologies for cells labeled in the left dorsal habenula versus the right (Bianco et al., 2008). Axons with the left morphology (L-typical) exhibit a greater degree of arborization and innervate an extensive region of the IPN and those with the right morphology (R-typical) show less axonal branching and innervate a more dorsoventrally restricted region. Even though neurons with a L-typical arbor predominate in the left habenula and those with a R-typical arbor are the majority in the right habenula, neurons having the opposite axonal morphology were also observed (Bianco et al., 2008). We propose that the subset of neurons in the right habenula with L-typical arbors contain the sst-1 subpopulation that innervates the dorsal IPN; whereas neurons in the left habenula with R-typical arbors correspond, in part, to the small cholinergic group that projects ventrally. In this model, neurons from both sides of the brain that use the same neurotransmitter innervate the same regions of the IPN.
The neurotransmitter profile of the zebrafish dorsal habenular nuclei serves as a useful guide to understand habenular function, but also points to many outstanding questions. Knowledge of how subsets of habenular axons form connections at the appropriate subregion of the IPN is lacking. Selective expression of Neuropilin1a in only the left dorsal habenula allows neurons to respond to Semaphorin3b signaling and directs them to innervate the dorsal IPN (Kuan et al., 2007). However, more cues must be involved in steering habenular axons at the IPN and determining their precise connectivity along its dorsoventral axis. Once at the target, the interplay between habenular efferents and IPN neuronal populations is largely unknown. In larval zebrafish, IPN neurons can differ greatly in their morphology (Bianco et al., 2008) and cells with elaborate processes may receive diverse input from innervating habenular fibers. Although recent studies have focused on the modulatory role of dorsal habenular neurons in fear responses (Agetsuma et al., 2010; Lee et al., 2010), the exact neuronal subtypes involved remain unclear. Specific functions and microcircuits involving the distinct cholinergic and somatostatin-expressing subsets of neurons must also be determined. A greater understanding of the neurotransmitter identity, subnuclear organization, and connectivity of the zebrafish dorsal habenulae will ultimately reveal why the left and right nuclei differ and how this specialization might influence behavior.
The authors thank Harold Burgess and Hitoshi Okamoto for generously sharing transgenic lines, Lea Fortuno, Michelle Macurak, and Estela Monge for technical assistance, and members of the Halpern laboratory for helpful input.