Nucleus Rostrolateralis: An Expansion of the Epithalamus in Some Actinopterygii


  • William M. Saidel

    Corresponding author
    1. Center for Computational and Integrative Biology, Rutgers University, Camden, New Jersey
    • Department of Biology, Rutgers University, Camden, New Jersey
    Search for more papers by this author

Correspondence to: William Saidel, Department of Biology and Center for Computational and Integrative Biology, Rutgers University, Camden, NJ 08102-1411. E-mail:


The diencephalic nucleus rostrolateralis (RL) in the African butterfly fish (Pantodon buchholzi) is a brain nucleus identified in fewer than a dozen of the ∼25,000 species of actinopterygian fishes. Located in the rostrolateral diencephalon, this nucleus in Pantodon receives direct and indirect visual input from the superior visual field. Its lack of precedent or consistent phylogenetic expression creates a difficulty in interpreting the functional role of this nucleus within the visual system. By tracing experiments, RL was found to be afferent to nucleus interpeduncularis (IP) and the target of cells from the subpallium of the telencephalon. RL is a component of a descending telencephalic pathway involved in at least one behavior at the intersection of limbic and somatic activities—feeding. The parallelism between the ventral telencephalon—RL—IP and the limbic/striatal—habenula—IP pathway (the dorsal diencephalic conduction system, DDCS) suggests that RL is a component within the DDCS. Moreover, the hodological connections of RL suggest that RL is likely a hypertrophy of the lateral habenula. Anat Rec, 296:1594–1602, 2013. © 2013 Wiley Periodicals, Inc.

Abbreviations used
4th V

4th ventricle


anterior commissure


commissura ansulata




cytochrome oxidase


dorsal diencephalic conduction system


1,1′-dioctadecyl-3,3,3',3'-tetramethy-lindocarbocyanine perchlorate


entopeduncular nucleus


fasciculus retroflexus


habenula nucleus


nucleus interpeduncularis


dorsal division of IP


ventral division of IP

lat hab

lateral habenula nucleus

med hab

medial habenula nucleus


optic chiasm

olf b

olfactory bulb


optic tectum


dorsal optic tract


nucleus rostrolateralis


cell layer of NR


neuropil of NR




tractus RL


nucleus ventromedialis.

The habenula (hab) and epiphysis define the epithalamic nodes of the dorsal diencephalic conduction system (DDCS; Sutherland, 1982; Klemm, 2004; Beretta et al., 2012). This pathway is phylogenetically ancient. Its components are found in the most primitive of vertebrates, for example, Eptatretus or hagfish (Wicht and Northcutt, 1992, 1998) and Petromyzon, the lamprey (Yaňez and Anadón, 1994) as well as every descendent branch of the vertebrate phylogeny (Butler and Hodos, 2005; Aizawa et al., 2011; Stephenson-Jones et al., 2012). It is involved in a range of behaviors in which an internal state initiates or modifies one of many behaviors (Sutherland, 1982; Klemm, 2004; Hikosaka, 2010). This pathway links forebrain limbic and motor systems to the nucleus interpeduncularis (IP) in the midbrain and eventually to the motor system (Sutherland, 1982; Klemm, 2004; Hikosaka et al., 2008). A defining characteristic of the DDCS is the monosynaptic relationship between the hab and IP.

In the teleost Pantodon buchholzi, the African butterflyfish, and a number of other species (Saidel and Butler, 1997a; Butler and Saidel, 2003), a medially positioned nucleus just anterior to the hab named nucleus rostrolateralis (RL) has been identified to date in sporadic clusters of related fishes totaling less than a dozen genera (Fig. 1). These clusters have been found as a limited distribution throughout the taxonomic subdivisions of the Actinopterygii (Nelson, 2006). In Pantodon, the nucleus is especially large and distinctive, but the limited information about it in this species or any other in which it has been identified restricts an understanding of its functional role. RL in Pantodon is located just medial to the lateral optic tract, lateral to the anterior aspect of nucleus ventromedialis (VM), and just anterior to the hab. It has been described as cerebellar-like (Bell, 2002) with an oval-shaped neuropil (RLn) and a medial layer of neurons (RLc) (Fig. 2A). The proximal dendrite(s) of an RL neuron spans nearly the entire short axis in the neuropil, with secondary dendrites parallel to the long axis, not unlike the geometry of a Purkinje cell. These dendrites receive afferent input from the contralateral ventral retina (Saidel and Butler, 1991), from periventricular cells of the dorsomedial optic tectum bilaterally (Butler and Saidel, 1992; Saidel and Butler, 1997a), and the nucleus is reciprocally connected with the ventral telencephalon (Rink and Wullimann, 2004).

Figure 1.

Distribution of the known presence of RL in the teleostei. Despite the large numbers of orders (38) and families (446) in the division teleostei, only three clusters at the level of family or subfamily have been found to date (modified after Fig. 10 of Butler and Saidel, 2003). Not illustrated is the presence of RL in the ancestral Lepisosteus (Class Actinopterygii, order Semionotiformes). Numbers of orders and families are from Nelson (2006).

Figure 2.

RL and IP of Pantodon buchholzi. A: RL comprises a neuropil (RLn) and a neuron layer (RLc); RL is located in the rostral diencephalon, medial to the dorsal optic tract (OTd) and lateral to the rostral aspects of VM. This section is at the level of the optic chiasm (OC). Dashed white line: midline of section. Scale: 0.25 mm. B: IP is a midline nucleus comprising dorsal (IPd) and ventral (IPv) divisions of its neuropil, which is surrounded by its neurons. Rostrally, it is just dorsal to the tectobulbar tract and just caudal to the CA. Caudally it is bordered by the median raphe nucleus. Scale: 0.5 mm.

An extensive, horizontal falciform process divides the retina into distinct and separate dorsal and ventral retinal regions (Schwartz, 1971; Saidel and Braford, 1985) with the latter providing RL with its retinal input (Saidel and Butler, 1991). Because the habitat of this fish is immediately subjacent to the water surface, the ventral retinal region views the water surface and through it (or the “aerial” visual field), while the dorsal region views into the water column (the “aquatic” visual field). This fish is a surface feeder. It targets and eats prey imaged only in its dorsal visual field (or by the ventral retina). In addition, periventricular cells in both tectal lobes topographically corresponding to the ventral retina are also afferent to RL.

However, other than a direct and indirect mapping of the dorsal visual field into RL, little can be inferred about the functional role of this distinctive nucleus in any of the fishes in which it has been identified except that metabolically it is quite active (Saidel and Butler, 1997a). Knowing its efferent target would further define the system within which it acts. To identify that target, the lipophilic dye, DiI, was used to reciprocally study the pathway in which RL is contained.


Experimental Animal

The experimental work of this study was performed with the African butterflyfish, P. buchholzi (subdivision Osteoglossomorpha, Order Osteoglossiformes, Family Osteoglossidae, see Fig. 1). Standard lengths of the individuals used ranged from 5 to 8 cm. The original research reported here was performed according to guidelines established by the Rutgers University Animal Care and Facilities Committee (protocol 92-059).

Tissue Fixation and DiI Tracing Procedures

DiI (1,1′-Dioctadecyl-3,3,3′,3′-tetramethy-lindocarbocyanine perchlorate, Invitrogen, CA; Godement et al. 1987; Honig and Hume 1989) was used as a lipophilic tract tracer. Individual fish were heavily anesthetized to level 3 (DeTolla et al., 1995), and perfused with saline followed by 4% paraformaldehyde in 0.1 M phosphate buffer, pH 7.4. Brains were allowed to remain in situ for 1–2 days prior to removal. RL forms a small, ∼140 µm long, concave surface at the rostral diencephalon. To expose this surface, a small wedge of solidified gelatin was inserted between the caudal telencephalic and the rostral diencephalic surfaces which increased the space by about 1 mm, and thereby permitted visualization of RL. An 80 micron deep hole was made in RL with a shaped #000 insect pin and a tiny crystal of DiI was inserted in that hole with another pin whose tip had been flattened. For confirmation of the efferent pathway from RL to IP, DiI was placed in IP to label the pathway in the reverse direction. The brain was transversely bisected about 0.5 mm behind the commissura ansulata (CA) on the ventral surface of the brain, thereby exposing IP as a round neuropil at the midline. Similar to the RL study, a pin was inserted in the rostral half of the brain about 50–100 µm into IP to form a pinhole and a crystal of DiI was inserted into the hole. The analysis reported here is based on results from 11 specimens in which DiI was deposited in RL (Figs. 2A and 3) and 12 in which DiI was placed in its discovered target, IP (Figs. 2B and 3). Diffusion times ranged from 2 to 4 weeks.

Figure 3.

Telencephalic and mesencephalic connections of DiI-labeled RL. All views are sagittal except B. A: The proximity of RL of Pantodon to the hab is seen in this Klüver-Berrera stained section. The locations of C and E are indicated by the boxes. Scale: 2 mm. B: In this transverse image of rostral IP, dorsal IP is heavily labeled bilaterally. Note that anterior dorsal IP is positioned just above the tecto-bulbar tract. Scale: 0.5 mm. C: The most rostral neuron afferent to RL seen in the subpallial telencephalon (one of several) is located as indicated in D. Scale: 25 μm. D: Low power of ventral telencephalon. Box identifies location of the cell illustrated in C. Scale: 0.5 mm. E: IP is labeled after DiI in RL. The fr is intensely labeled as well. The two arrowheads in this sagittal view indicate the location of the transverse illustration of B (from another case). Scale: 0.5 mm.

Histological Procedures: DiI

Brains labeled with DiI were sectioned with a vibratome at 35–70 microns. Sections were floated onto a slide in buffer, covered with buffer, coverslipped, and visualized wet using fluorescent imaging with the standard DiI filter set.

Cytochrome Oxidase Cytochemistry

One additional brain was fixed and histochemically stained for cytochrome oxidase activity (COX) because RL is characteristically identified by this staining (Saidel and Butler, 1997b). The brain was perfused and fixed with 2% paraformaldehyde in 0.1 M phosphate buffer, pH 7.4, frozen sectioned at 40 microns, and reacted for COX staining as free-floating sections (Wong-Riley, 1979) prior to mounting, dehydration, and coverslipped.


Both fluorescent and transmitted light images were captured digitally by a Nikon D-90 DSLR mounted on a Zeiss Axioplan microscope equipped with fluorescence and differential interference contrast imaging. Images were imported into Adobe Photoshop© or NIH's ImageJ (Rasband, 1997–2012) to enhance the contrast or superimpose multiple images. Final composite figures were assembled in Photoshop.


Connections of RL

A small crystal of DiI placed in RL led to pathways extending caudad to the IP (Fig. 3B,E) and rostrad to the ventral telencephalon (Fig. 3C,D). Efferent axons of RL neurons continued in a loose arrangement into the neuropil of the ventral hab but exited as a distinct bundle within the fasciculus retroflexus (fr). The labeled axons in fr continued through the diencephalon and pretectum in a characteristic position until terminating in the midbrain IP (Fig. 3B,D). The neuropil of IP has dorsal and ventral divisions both divided by the midline (Fig. 2B). Terminal branches of axons from RL extended throughout the neuropil of dorsal IP bilaterally (Fig. 3B). Although RL receives input from only one region of the visual field, no hint of a topographic division within the dorsal IP was found in the position of axon terminals. The dorsal division of IP in Pantodon was distinctly more heavily labeled than the ventral division. In the ventral division of IP, the few fibers present were oriented along the mediolateral axis unlike the more heterogeneous pattern in the dorsal division. A limited number of cells were also labeled at the ventral and ventrolateral margin of the labeled neuropil.

A fiber bundle exited RL rostrally and continued into the telencephalon in the direction of the anterior commissure (ac). DiI labeled neurons were detected in two loci around the ac. A small number of neurons were labeled in the subpallium rostral to the ac (Fig. 3C). Others were located directly below the ac in or about the anterior region of the preoptic nucleus (see discussion).

Connections of IP

To confirm this pathway, DiI was placed in IP, allowed to retrogradely diffuse from IP for up to 4 weeks before the brain was sectioned. Neurons were labeled in the cell layer of RL (Fig. 4). Labeled RLc neurons measured between 9.5 and 12 microns in diameter. They possessed primary dendrites extending into the neuropil of RL perpendicularly to the long axis of the nucleus similar to Golgi-stained RL neurons (personal observation) and the axons of these neurons left the medial edge of RL (Fig. 4C) to form a diffuse tract (tractus RL, tRL) that passed through the hab (Fig. 5B) and into the fr. At the caudal end of RL closest to the hab, the neurons in RLc are positioned as a mass of cells with little neuropil (Fig. 4D). At this locus, cells labeled more frequently and in larger numbers than in the clearly defined cell layer of the nucleus (Figs. 2A and 4C,E) seen in more rostral transverse sections. Since RL is just rostral to the hab, the possibility of DiI contamination to it was considered. Amo et al. (2010) point out that the ventral hab in zebrafish is afferent to the raphe, while the dorsal hab is primarily afferent to IP. Since neither raphe neurons nor the neuropils around them were stained nor the adjacent neurons of the anterior ventral hab, it is unlikely that the results reported here include such contamination.

Figure 4.

Neuron labeling in RL after DiI applied to IP. A: A DiI labeled RL is imaged in transmitted light to illustrate the location of the cell shown in B. Scale: 100 µm. B: A cell labeled from nu IP Note the axon emerging from the cell and joining the (tractus RL, tRL) to the left. Scale: 25 µm. C: In this figure three RL neurons are well labeled by DiI. Arrowheads point to their exiting axons which enter the tRL. Scale: 20 µm. D: This image is a tangential section through the caudal tip of RLc. Many cells (bordered by the white arrowheads) labeled after DiI in IP. Note the intense fluorescence to the left of RLc, which is the location of the tRL. Scale: 50 µm. E: In this section, the arrowheads identify 13 labeled cells in RLc. (This figure is a cumulative image of 4 superimposed images from different foci within one 40 μm section.) Scale: 50 µm. Note: For all parts of this figure, the midline is to the left and lateral is to the right.

Figure 5.

The role of the hab between RL and IP. A: The lateral habenula (lat hab) is intensely labeled following DiI application to IP whereas the medial hab (med hab) is not. The asterix identifies a small corner of the contralateral lateral hab indicating a bilateral connection from the IP to both lat hab. The tRL flares loosely before coalescing into a discrete bundle (B). Scale: 100 µm. B: In this more caudal section, the tRL coalesces into the fascicle of the RL–IP connection (arrow) that passes through the neuropil of the hab and into the fr. Scale as in A. In both images, white lines outline the margins of the diencephalon with medial to the left.

Dense labeling of neurons in the dorsal hab was expected and observed as it was known that the dorsal hab is reciprocally connected with IP (Fig. 5A; Yaňez and Anadón, 1996; Amo et al., 2010). The fascicle from RL to IP was identified in the same material as it coursed through the neuropil of the hab (Fig. 5B).

Relationship Between RL and IP as Indicated by Correlated COX Staining

Both RL and IP densely stain compared to neighboring areas after COX histochemistry. This consistency in the density of COX staining is (Fig. 6A,B) not unlike that seen in some serially connected mammalian sensory systems (Wong-Riley, 1979, Wong-Riley and Welt, 1980). The high levels of tonic neuronal activity are indicated by this stain, thereby suggesting a direct and ongoing physiological relationship between RL and IP.

Figure 6.

Both RL (A) and both parts of IP (B) are intensely labeled following staining for COX. Midline is indicated by the white dashed line. Scale: A: 0.25 mm; B: 0.5 mm.

Figure 7.

Telencephalic neurons connect with RL. A: The box in this sagittal telencephalic section at the level of the ac is imaged in transmitted light to illustrate the location of labeled preoptic neurons ventral to the ac. Scale: 250 μm. B: The arrowheads identify two neurons beneath the ac projecting to RL. Scale: 25 μm


The goal of this study was to identify the circuit in which RL participates. It appears to be a component of the DDCS but the significant visual input puts it apart from the classical view of this system. In addition, RL shows some differentiation in its telencephalic afferents from the hab proper.

Nucleus Rostrolateralis as Epithalamic or Thalamic

The connections observed in this study raise a question about RL. Although originally described as a dorsal thalamic nucleus (Butler and Saidel, 1991) by its location, by its connections it may be defined as epithalamic (as suggested in Butler and Hodos, 2005, p. 414). Given the close association with the hab, the same efferent pathway and the same efferent target, one can only speculate that RL may be a hypertrophy of the lateral hab. Although innervating both portions of IP in Pantodon, efferents from RL preferentially innervate dorsal IP (Figs. 2 and 3), as do efferents from the lateral hab in other teleosts, for example, zebrafish, Danio (Aizawa et al., 2005; Gamse et al., 2005), trout, Oncorhynchus (Yaňez and Anadón, 1996), and goldfish, Carassius (Villani et al., 1996). Alternatively, RL may share with the lateral hab the same embryonic ventricular zone from which the neurons of the lateral hab arose. This question remains to be resolved.

Nucleus Rostrolateralis and the Dorsal Diencephalic Conduction System

RL is presynaptic, like its well-recognized neighbor, the hab (Herkenham and Nauta, 1979; Contestabile and Flumerfelt, 1981; Villani et al., 1996; Bianco and Wilson, 2009), to the IP bilaterally. The axons of RL neurons ascend to enter the neuropil of the ventral hab (Figs. 4C and 5B) and join its output as a component of the fr before terminating in IP. Therefore, RL satisfies a connectivity that defines it as a component of the DDCS. Moreover, the intense COX staining of mitochondria in both RL and IP (Fig. 6) combined with the tract connecting the two provides additional support for a functional association between the two nuclei.

In mammals, this system connects limbic and striatal forebrain areas to the midbrain (Sutherland, 1982; Hikosaka et al., 2008; Bianco and Wilson, 2009). Its main components, the septal and basal motor areas of the telencephalon, the hab of the diencephalon, and the IP of the mesencephalon, are components of an axial system in every vertebrate species examined (e.g., Sutherland, 1982; Butler and Hodos, 2005; Bianco and Wilson, 2009). The hab also receives an input from the entopeduncular nucleus (EP) of the telencephalon (Villani et al., 1996; Yaňez and Anadón, 1996; Giuliani et al., 2002). RL does not. In mammals, the EP is part of the striatal circuitry, but a nucleus with the same name in teleostean fishes is not homologous to EP of mammals (Wullimann and Mueller, 2004) and does not provide input to RL. However, similar to the telencephalic-habenula-IP system, RL also receives a telencephalic input (Fig. 7), and its efferent pathway terminates in IP, thus supporting the hypothesis that in species of fish with RL, that it is a component of the DDCS.

One would expect that, physiologically, the DDCS is involved with some aspect(s) of movement (Hikosaka et al., 2008). Neurophysiological recordings in IP showing a temporal contingency based on head orientation movements associated with specific navigational directions support this inference (Sharp et al., 2006; Clark et al., 2009).

The DDCS is a Phylogenetically Ancient Pathway

In mammals, the DDCS has been described as a core brain pathway serving multiple functions (Sutherland, 1982; Klemm, 2004; Hikosaka, 2010). These functions are affective and mostly involved with individual and species survival behaviors including ingestion, drinking, stress, sex and mating, and of a general affective nature involving aversion and/or reward. In most cases, these behaviors are responses to disturbances of one or more homeostatic states. Sufficient neuroanatomical evidence exists to suggest that the DDCS in bony fishes exists and includes RL when present. It is a second (after the medial forebrain bundle) descending, axial telencephalic pathway associated with limbic-associated motor behaviors, and therefore, the DCCS must be an ancestral pathway of the Vertebrata.

The Role of the DDCS may Increase Behavioral Success

Such a pathway would be important during survival activities. For one behavior of fishes in general and for Pantodon in particular, success of predatory acts in the wild is probabilistic. With RL as a component of the DDCS, the physiological consequences of the DDCS may enhance the fish's success in the face of a less than certain probability and convey a selective advantage (see Bromberg-Martin and Hikosaka, 2011). Interestingly, a recent study in zebrafish suggests that survival-related flight from a visually detected predator may be modified by experience-related changes in the hab-IP component of this circuit (Agetsuma et al., 2010).

Pantodon, a genus of the most basal subdivision of the Teleostei (Fig. 1; Nelson, 2006), is a visual feeder of surface prey. Along with Pantodon, other bony fishes with RL (eg, Anableps, Gambusia, Danio) are also at or near surface feeders, but surface feeding is not obligatory to possess a RL (e.g., Rasbora, Xiphophorus). In Pantodon, RL receives direct and indirect visual input about the dorsal visual field, which, in its behavior, is the only visual area involved with predation. A lack of topography between RL terminals within the IP suggests that the RL–IP connection is not associated with a computation for motor precision. Perhaps the visual input to RL provides a useful stimulus to the DDCS to increase the excitability of its motor system prior to feeding and therefore, enhances its probability of successfully capturing its prey. For other fishes with an RL and without the detailed retinal features, the visual input may be less direct but still function in a similar manner. The DDCS including RL in the Actinopterygii may well act as a bridge between an affective state and a specific somatic motor behavior mediating an increased excitability of a motor response subsequent to a specific visual input. One might call this an increase in motivation.


The author is indebted to Dr. Ann Butler of the Krasnow Institute, George Mason University, Fairfax, VA, for years of insightful discussions about this system.