Fiber connections of the central nucleus of semicircular torus in cyprinids

Tract-tracing studies of nonfixed brains

Carbocyanine dye (DiI) experiments of fixed brains

The DiI was used for tract-tracing experiments on the saccular and lagenar nerves because of technical difficulties of in vivo and in vitro injections. One carp and one goldfish were used for DiI application experiments to the saccular nerve, and one carp and one goldfish for lagenar nerve experiments. The fish were perfused as in in vivo specimens. The brains were removed from the skull and postfixed in a fresh solution of the same fixative for 2 days. Small crystals of DiI were inserted into the targeted nerve and the insertion site was covered by 5% gelatin (Sigma type A) to avoid translocation of the crystals. The brains were maintained in the dark in the fixative for 2 weeks at 37°C. The brains were then embedded in 8% gelatin (Sigma, type A), fixed for 3 days, and cut into 80-μm frontal sections on a microslicer (D.S.K., DTK-3000). Sections were mounted on slides with 50% sucrose in 0.1 M PB (pH 7.4), coverslipped, and observed with an AX80 microscope (Olympus, Lake Success, NY) equipped with epifluorescence illumination system.

Cytoarchitecture

In addition to tract-tracing materials, Nissl- and Bodian-stained materials available in our laboratory (Ito, 1978) were used to analyze the cytoarchitecture of the cyprinid brain.

Nomenclature

For the octavolateral areas in the rhombencephalon and mesencephalon we adopted the nomenclature on goldfish of McCormick and Braford (1994) and McCormick and Hernandez (1996). We used the terminology of Ito and Yoshimoto (1990) on the carp brain for the subdivisions of lateral valvular nucleus. For diencephalic structures we basically used the nomenclature of Braford and Northcutt (1983) on the goldfish brain. Other terms are also introduced where necessary.

Cytoarchitecture of the preglomerular complex

The PG has been subdivided into several parts in the goldfish (Peter and Gill, 1975; Braford and Northcutt, 1983). The subdivisions, unfortunately, do not agree with cytoarchitectonic as well as hodological data as studied in the present study. Therefore, we first describe the organization of PG, based on the analysis of Nissl- and Bodian-stained materials in conjunction with tract-tracing results, for clarity of the following descriptions.

The organization and cytoarchitecture of the PG are similar in carp (Fig. 1) and goldfish. The most rostral preglomerular structure is the anterior preglomerular nucleus (PGa) of Braford and Northcutt (1983), which occupies a position ventrolateral to the optic tract and torodiencephalic tract. The PGa is composed of small and round neurons (about 5–6 × 5–8 μm in carp) stained darkly in Nissl-stained materials (Fig. 1E). Slightly caudal to the rostral pole of the PGa appears the PGl. For some rostrocaudal distance the PGa and PGl are seen in the same frontal planes, with the PGl located lateral to the PGa (Fig. 1A). Caudally, preglomerular tertiary gustatory nucleus (previously termed nucleus glomerulosus) and medial preglomerular nucleus (PGm) appears ventral to the PGl (Fig. 1B). Our cytoarchitectonic analysis clearly indicates that the PGl can be further divided into rostrolateral (PGlr; Fig. 1A–C) and caudomedial (PGlc; Fig. 1C,D) regions. The PGlc is composed of small and round neurons (about 5–6 × 5–8 μm: carp, Fig. 1F), which are darkly stained, similar to PGa neurons. Neurons in the PGlr possess totally different morphological features. They are medium-sized and usually more elongate (about 7–12 × 8–15 μm: carp), and are lightly stained (Fig. 1F,G). Caudally, the PGa shifts ventrally and becomes smaller in frontal sections. At these levels emerging PGlc and PGm are dorsally and ventrally adjacent to the PGa, respectively. The PGm is composed of small neurons like those in the PGa and PGlc (Fig. 1H). The boundaries between the three parts composed of small neurons are quite difficult to determine. Further caudally, the commissural preglomerular nucleus appears caudomedial to the PGlc.

Tract-tracing studies revealed that the PGa receives abundant fibers from the TSc (Fig. 2A,B). The terminal zone of TSc fibers continued caudally from the PGa to the PGlc and a medial zone of PGm (Fig. 2C,D), although the terminal zone became small in frontal sections at junction levels between the auditory preglomerular parts (Fig. 2B). More caudally, the terminal zone became larger again in the PGlc and the medial zone of PGm (Fig. 2C,D), where the former replaces the rostrolateral region or PGlr.

Tract-tracing studies

Biocytin and BDA were well transported both anterogradely and retrogradely. The data obtained by the two tracers were identical qualitatively, although the superiority of biocytin to BDA in anterograde transport and the superiority of BDA than biocytin in retrograde transport were noticed. Studies in in vivo and in vitro brains yielded almost identical results. Therefore, we present together the results obtained by the two tracers either in vivo or in vitro. Results obtained in the carp and goldfish were virtually the same, indicating similar organization of auditory pathways through the brain stem to the diencephalon in these closely related fishes. The following descriptions thus apply to both species unless otherwise noted.

Injections to the central nucleus (TSc).

In three carp and three goldfish cases injections were confined within the TSc (Figs. 2F,G, 3A). Tracers injected to the TSc did not diffuse into subjacent TSvl, and labeled structures were seen almost throughout the TSc (Figs. 2E–I, 3B). In the TSvl only sparsely labeled cells and terminals were present.

Labeled fibers in the TSc continued rostromedially to the medial pretoral nucleus of Finger and Tong (1984). Labeled terminals as well as cells were seen in the nucleus (Figs. 2C,D, 3C). Some labeled fibers crossed the midline through the posterior commissure to terminate in the contralateral medial pretoral nucleus (Fig. 2B–D). Occasionally labeled cells and terminals were seen in the contralateral TSc (Fig. 2E,F). In the midbrain tegmentum a few neurons and terminals were labeled lateral to the lateral lemniscus (Fig. 2F). Only sparsely labeled terminals but no cells were seen in the ipsilateral optic tectum, mainly in the stratum album centrale (Fig. 2A–I). Tectal injections do not label cells in the TSc (Yamamoto and Ito, unpubl. obs.).

Many labeled fibers entered the lateral lemniscus and ran caudally (Fig. 2E). At isthmic levels labeled ovoid or multipolar neurons were seen in or medially adjacent to the lateral lemniscus with ipsilateral dominance (Figs. 2I, 4A). These cells appear to correspond to neurons of isthmic reticular nucleus (Bell, 1981; Kozloski and Crawford, 1998; also see McCormick, 1999). At similar levels labeled fibers were seen to cross the midline to reach labeled cells in the contralateral nucleus. Some of the labeled fibers in the lateral lemniscus crossed the midline at levels of the rhombencephalon (Fig. 2L). Labeled neurons were seen in the anterior octaval nucleus, mainly contralateral to the injection site (Figs. 2K, 4B). In the goldfish, however, labeled neurons were seen in the nucleus only after large injections that affected both the TSc and TSvl.

Most of the labeled fibers in the lateral lemniscus could be traced to the SO and DO, which were the main sources to the TSc (Fig. 2K–N). Slightly larger numbers of labeled neurons were present in the ipsilateral SO than the contralateral nucleus. Three divisions of SO as reported by McCormick and Hernandez (1996) in goldfish were labeled also in carp: 1) dorsal division with Purkinje cell-like large neurons or crest cells that possess a dorsal dendrite extending to a medial zone of cerebellar crest, and a ventral dendrite that reached the intermediate division; 2) ventral division with large fusiform or occasional multipolar neurons near or rostral to the crossing fibers of lateral lemniscus; and 3) intermediate division with spherical and medium-sized neurons (Fig. 4C). Most of the fusiform neurons in the ventral division were oriented dorsoventrally, extending dorsal dendrites into the intermediate division and ventral dendrites toward the labeled torobulbar fibers running dorsally to terminate in the SO. Occasionally horizontally oriented neurons were also present. Rostrally, similar fusiform neurons were seen to the level of a group of granular neurons capping the rostral pole of the SO or the granule population of McCormick and Hernandez (1996). Rostrally located fusiform neurons were regarded as a part of the ventral division of SO. A possible topographical organization of the SO-TSc pathway was also noted. When the injection sites were in the rostral TSc, labeled cells were mainly present in a medial zone of the SO (Fig. 4C). In contrast, injections to middle to caudal TSc resulted in labeled cells in a lateral zone of the SO (Fig. 4D).

Most of the labeled cells in the DO were medium-sized and ovoid (Fig. 5). Occasionally large fusiform neurons of crest cell type were also seen (Fig. 5A). Many labeled DO neurons were seen dorsally and laterally adjacent to the facial sensory root (Fig. 5), i.e., restricted in the DOdm and DOi of McCormick and Braford (1994). At middle levels of the nucleus, labeled fusiform cells in the DOdm were also seen near the midline ventral to the cerebellar crest in the carp. The axis of these neurons was horizontal and extended dendrites that cross the midline to the contralateral nucleus in addition to dendrites in the ipsilateral side (Fig. 5E). The vast majority of somata and dendrites of the labeled DOi neurons were restricted in the medial part of this zone. Labeled neurons in the DOdm also tended to aggregate in the ventromedial portion of the zone or close to the facial sensory root. The DOdm and medial part of DOi, where labeled neurons and dendrites were seen, will be referred to as medial DO (DOm) for convenience when we refer to these zones cumulatively. A few labeled cells were seen in a more lateral part of the DOi (Fig. 5B,F). No labeled neurons were encountered in the lateralmost region of DOi or in the ventral zone of the DO of McCormick and Braford (1994). Labeled neurons in the DOm were abundant in the contralateral nucleus, which is different from the laterality of the SO neurons. The DOm-TSc pathway might also be organized topographically. When the injection sites were in the rostral TSc, labeled cells mainly appeared caudal and ventral portions of the DOm (Fig. 5A–D). In contrast, injections to middle to caudal TSc resulted in labeled cells mainly in rostral and dorsal portions (Fig. 5F).

Another population of labeled fibers originated from the injected TSc, coursed ventromedially, and descended to the rhombencephalon (torobulbar tract: Figs. 2F–N, 3A, double arrowhead). Labeled fibers in the torobulbar tract gave rise to a terminal zone in the isthmus bordered medially by the lateral lemniscus and dorsally by the isthmic nucleus (Fig. 2G,H). This area corresponds to the perilemniscular nucleus of Wullimann and Northcutt (1988), a cerebellopetal nucleus identified in sunfish and goldfish. Dense terminals were seen only in a dorsal portion of this nucleus (Fig. 6A), where labeled neurons were also present (Fig. 2G,H). Some of the fibers running close to the perilemniscular nucleus coursed dorsally to terminate in the lateral valvular nucleus (Fig. 2H,I). Terminals were restricted to a lateral portion of the nucleus, a zone that appears a caudal extension of the anterior part of lateral valvular nucleus of Ito and Yoshimoto (1990). Terminals in the lateral valvular nucleus showed a peculiar cup-shaped morphology (Fig. 6B), which characterizes terminals in this nucleus (Yoshimoto and Ito, 1993).

At the level of the secondary gustatory nucleus a compact bundle of labeled fibers departed the torobulbar tract proper and turned dorsomedially (Fig. 2I: small arrow). This bundle then traveled caudally through an area ventrolateral to the granule population of McCormick and Hernandez (1996), providing terminals to the population of granular cells (Figs. 2J,K, 6C). The labeled fibers continued further caudally and entered the SO from its rostral aspect to terminate in the nucleus (Fig. 2L).

Other labeled fibers in the torobulbar tract proper continued further caudally and gave off terminals in the reticular formation (Fig. 2J,K). Labeled neurons were seen in the rhombencephalic reticular formation dorsomedial to the torobulbar tract and ventromedial to the lateral lemniscus (Figs. 2J,M,N, 6D). They were larger than the fusiform neurons in the ventral division of SO. Each neuron had a ventrolateral dendrite extending toward the labeled fibers of torobulbar tract (Fig. 6D) and a dorsomedial dendrite (destination was unclear). Similarly located neurons were also encountered contralaterally.

At the level of the SO most fibers in the torobulbar tract turned dorsally to terminate densely in the SO (Fig. 2L,M). Terminals were seen in the intermediate and ventral divisions of SO and crest cells of the dorsal division can also receive fibers on their ventral dendrites. The terminal zone continued sparsely into a restricted ventral region of the DOm (i.e., ventral to most neurons afferent to the torus; Fig. 2N). It was not clear whether the terminal zone is a caudal extension of the dendritic field from the ventral division of SO. Terminals in the SO showed the same topography as labeled cells in the SO (Fig. 4C,D). Terminals were also seen in the contralateral SO. However, terminating fibers in the contralateral SO could be traced back to retrogradely labeled neurons of the contralateral DOm, indicating that DOm axons projecting to the TSc send collaterals to the SO. Similar collateral axon terminals from the DOm may be also present in the ipsilateral SO, but this could not be confirmed definitively due to numerous labeled fibers. A few fibers in the torobulbar tract continued further caudally to terminate in the ipsilateral reticular formation (Fig. 2M,N).

A third population of labeled fibers originated from the injected TSc (Fig. 2F,G, arrowhead), the majority of which projected to the diencephalon (torodiencephalic tract). The labeled fibers coursed rostrally through the ventrolateral margin of the semicircular torus (Figs. 2E, 3A,B, arrowhead), some of which turned medially and gave terminals in the PGlc (Figs. 2C,D, 7A), while not in the PGlr. Labeled fibers also traveled ventrally to terminate in a medial zone of the PGm (Fig. 7A,D,E) and in the anterior tuberal nucleus. More rostrally, the terminal zone in the PGlc and PGm continued to the PGa (Fig. 7E). Labeled terminals in the preglomerular nuclei are quite large and showed strange morphologies (Fig. 7B,C, described in more detail below). Other medially oriented fibers reached a terminal zone in the central posterior thalamic nucleus of Braford and Northcutt (1983) or a caudal portion of the dorsomedial thalamic nucleus (Figs. 2B,C, 8A). Large labeled neurons were also present in a lateral zone of the nucleus (Figs. 2C, 8B). These neurons were multipolar and the dendrites could be classified into three populations: lateral, medial, and rostral. The lateral dendrites approached labeled fibers and terminals of the torodiencephalic tract (Fig. 8B). The medial dendrites reached the caudomedial zone of the central posterior thalamic nucleus, where smaller unlabeled cells and labeled terminals were present. The rostral dendrites were extremely long, coursed rostroventrally, and finally entered the anterior tuberal nucleus, where numerous labeled terminals were seen (Figs. 2B,C, 8C).

Other labeled fibers in the torodiencephalic tract continued further rostrally through an area dorsal to the PGlr (Fig. 2B), then shifted ventromedially, and provided abundant labeled fibers with large swellings in the PGa (Figs. 2A,B, 8D,E). The swellings were quite large but were regarded as terminals for the following reasons: 1) they were smaller (about 3 μm: carp) than the somata of PGa neurons (5–6 × 5–8 μm: carp); and 2) their peculiar morphology was unlike that of somata (Fig. 8E). Very frequently, each swelling appeared like a donut with an unlabeled hole (about 1 μm or less in diameter: carp) or a pair of semicircular structures sandwiching a narrow unlabeled slit (about 1 μm or less in width and 2–3 μm in length: carp). Thus, the labeled structures are most likely presynaptic terminals surrounding dendrites of the target PGa neurons and thus exhibit different morphologies, depending on the cutting angles. Terminals in the PGlc and a medial zone of PGm also showed similar morphologies (Fig. 7B,C). Some of the labeled fibers continued medial to the PGa (Fig. 2A), and then turned caudally to terminate densely in the anterior tuberal nucleus (Figs. 2B–D, 8C). Some fibers crossed the midline (Fig. 2A) to terminate in the contralateral anterior tuberal nucleus (Fig. 2B–D) and TSc (Fig. 2E). Small numbers of labeled neurons were seen bilaterally in the anterior tuberal nucleus (Figs. 2B,C, 8C).

A few large labeled neurons (presumed eurydendroid cells) were present in the contralateral cerebellum close to the granular eminence (Fig. 2K). In one goldfish case, labeled terminals were seen in the supracommissural and ventral parts of the ventral telencephalic area (not charted).

Injections to the DOm.

In three carp and one goldfish cases tracers were injected into the DOm with minimal diffusion to neighboring structures (Figs. 9A, 10D). In these cases, labeled terminals, spherical cells, and crest cells were present in the contralateral DOm (Fig. 10D), indicating reciprocal connections between the nuclei on both sides. Labeled fibers from the DOm ascended through the bilateral lateral lemniscus (Fig. 10B,C). Some of the labeled fibers turned dorsally and terminated in the SO with ipsilateral dominance (Figs. 9B, 10C). Terminals were seen in intermediate and ventral divisions, but crest cells can receive this projection on their ventral dendrites. A few lightly labeled cells were also present in the intermediate and ventral divisions of SO (Figs. 9B, 10C). At the same level a few labeled cells were seen in the bilateral descending trigeminal nucleus and magnocellular octaval nucleus (Fig. 10C). Labeled neurons in the magnocellular octaval nucleus were smaller than gigantic cells that give rise to descending spinal projections (Yamamoto and Ito, unpubl. obs.). Along the lateral lemniscus occasional labeled terminals were seen on both sides and a few labeled fusiform cells were also present in the terminal field at the levels of granule population of McCormick and Hernandez (1996) (Fig. 10B). They were regarded as fusiform neurons of the ventral division of SO (see Discussion). At mesencephalic levels labeled fibers finally shifted laterally and terminated exclusively in the TSc with contralateral dominance (Figs. 9C, 10A). Terminal distribution in the TSc was not uniform. Terminals were present only in the deep or ventral zone of this nucleus, forming a terminal layer (Figs. 9C, 10A).

Injections to SO and medial cerebellar crest.

In two carp cases the injection was restricted within the SO (excluding the ventral division) and did not involve the overlying cerebellar crest (Figs. 11A, 12F). In these cases labeled spherical and ventral fusiform cells (but not crest cells) as well as terminals were seen in the contralateral SO, indicating reciprocal connections between the bilateral nuclei (Fig. 12F). Crest cells may also receive commissural projections on their ventral dendrites. Many labeled ovoid cells as well as a few crest cells were observed in the DOm with slight ipsilateral dominance (Fig. 12G). The distribution and morphology of labeled DOm neurons were similar to those labeled by TSc injections (Fig. 11B, cf. Fig. 5). Terminals were also present in the bilateral DOm (Fig. 12G). A few small cells were labeled bilaterally in the medial nucleus of the rhombencephalic octavolateral area (Fig. 12F,G). Labeled ascending fibers coursed through the bilateral lateral lemniscus (Fig. 12C–E). Occasionally labeled terminals as well as fusiform cells were seen in or near the lateral lemniscus at the level of granule population (Fig. 12E), similar to DOm injection cases. Sometimes the ventral dendrites of the labeled cells could be followed to reach a reticular formation zone where torobulbar fibers and terminals were labeled after TSc injections. They were regarded as the rostral continuation of fusiform neurons of ventral division of SO (see Discussion). A small fraction of labeled fibers coursed rostrally along the ventrolateral margin of the medulla oblongata and terminated sparsely in a dorsal zone of the perilemniscular nucleus (Fig. 12C). Labeled terminals were seen medial to the lateral lemniscus on both sides or in the isthmic reticular nucleus (Fig. 12C). A few labeled fibers were seen in the preeminential nucleus. Labeled fibers along the lateral lemniscus reached the torus and terminated in the TSc with a slight ipsilateral dominance (Figs. 11C, 12B,C). No terminals were encountered in the TSvl. In contrast to DOm injection cases, labeled fibers terminated in both dorsal and ventral zones of the TSc (Fig. 11C). Also different from the DOm injection cases, labeled fibers continued rostral to the TS and terminated in the medial pretoral nucleus (Figs. 11D, 12A). Some of the fibers crossed the midline through the posterior commissure, and fewer terminals were seen in the contralateral nucleus (Fig. 12A). Labeled cells were observed in the ipsilateral medial pretoral nucleus (Figs. 11D, 12A). There were no labeled cells in the TSc.

In one goldfish case the injection site was confined within a medial zone of the cerebellar crest overlying the SO (Fig. 13). This injection labeled densely many crest cells of SO, while only a few crest cells of the DO were faintly labeled. In this case labeled cells were seen in the granule population of McCormick and Hernandez (1996) (Fig. 13). A few labeled cells continued from the granule population to the rostrolateral part of the granular layer of the caudal lobe. Another group of labeled cells was seen in a lateral zone of the granular eminence (Fig. 13). Labeled granular cells were scattered between the granule population and granular eminence. Also, a few labeled cells continued rostroventrally from the granular eminence to the preeminential nucleus. Labeled neurons in the above-mentioned structures were dominant in the side ipsilateral to the injection. In contrast to cases with injection limited to the SO proper, very few labeled terminals were seen in the contralateral SO. A labeled crest cell was seen in the contralateral SO, which most likely became labeled directly from its dorsal dendrite invading the cerebellar crest of the injected side. Few terminals were present in the DOm. Labeled axons of the crest cells exited the SO laterally and then traveled ventrally along the lateral margin of the facial sensory root. Most axons ascended through the lateral lemniscus to terminate in the TSc with slight ipsilateral dominance. Terminals were seen in both dorsal and ventral zones of the TSc, similar to injection cases to the SO proper. Labeled terminals were also seen in the medial pretoral nucleus with ipsilateral dominance. No labeled neurons were seen in the nucleus, which is distinct from SO injection cases.

In other carp and goldfish cases, injected tracers involved both the cerebellar crest and SO proper ventral to the crest. The results of these cases were in accordance with the specific injection cases to the SO proper and medial cerebellar crest described above.

Injections to the saccular, lagenar, and posterior lateral line nerves.

The in vivo case including both saccular and lagenar nerves in a carp resulted in only a few labeled fibers and was not analyzed further. Selective DiI insertions to the saccular and lagenar nerve in the carp labeled the targeted nerve without affecting other branches of the octaval nerve. Labeled terminals were seen only in the ipsilateral brain. The most dense terminal zone of the saccular nerve was an area dorsolateral to the facial sensory root (Fig. 14A,B). Labeled fibers also ascended further dorsomedially and terminated in more dorsal and rostral zones of DOdm. Transneuronally labeled cells including crest cells were occasionally seen in and close to the terminal field. The saccular terminal zone thus coincided almost perfectly with the somatic and dendritic zones of DOm neurons afferent to the TSc (cf. Fig. 5). The lagenar nerve fibers terminated laterally adjacent to the saccular terminal zone (Fig. 14C,D). These two terminal zones may partially overlap each other. A few terminals were also seen in the anterior octaval nucleus subsequent to injections of either nerve. Basically the same pattern of connections was seen in goldfish, confirming a previous study (McCormick and Braford, 1994). Some labeled fibers proceeded further rostrally to terminate in the granular eminence and in or ventrolaterally adjacent to the granule population.

After tracer injections to the posterior lateral line nerve, most of the labeled fibers terminated in the medial nucleus of the rhombencephalic octavolateral area. Some of the labeled fibers coursed ventromedially toward the DO and terminated in a lateral zone of the DOi, and terminals may partially overlap with saccular and lagenar terminal zones (Fig. 14, asterisk). Most labeled terminals appeared lateral to the somatic and dendritic zones of most DO neurons afferent to the TSc (Fig. 14E,F, cf. Fig. 5). However, a fraction of DO neurons may fall within the overlapping zone of saccular, lagenar, and lateral line terminals (Fig. 5B,F, arrowheads). Labeled lateral line terminals were not seen in the DOdm, unlike that reported in a previous study of goldfish (Puzdrowski, 1989).

Injections to the medial pretoral nucleus.

In addition to the medial pretoral nucleus, ventrally adjacent structures were slightly affected by these injections. We thus focus the following descriptions on labeled structures in auditory-related nuclei, which represent presumed connections of the medial pretoral nucleus.

Labeled terminals and cells were seen in the superficial zone of the TSc (ipsilateral dominance: Fig. 15A), confirming reciprocally the results of the TSc injection study.

Three to four compact bundles of labeled fibers emerged from the injection site, which were close to but distinct from the lateral lemniscus. All but one bundle descended through the rhombencephalon, assuming a position ventral to the lateral lemniscus and finally turned dorsally to terminate in the bilateral SO (ipsilateral dominance: Fig. 15B). All three types of SO neurons were also labeled in the bilateral nuclei. These observations agree with the results of SO injection experiments. Some fibers contributed sparse terminals in the ventralmost region of DOm. One of the compact labeled bundles took a more dorsal course toward the granule population of McCormick and Hernandez (1996), where numerous labeled terminals of mossy fiber type were seen (Fig. 15C). Labeled fibers in the granule population continued dorsolaterally to supply terminals in a lateral zone of the granular eminence, a zone where labeled cells were seen after injections to the medial cerebellar crest. Terminals in the eminence were large, with quite bizarre morphologies (Fig. 15D,E). Fewer terminals were seen in the granule population and granular eminence contralateral to the injection. Labeled cells as well as terminals were seen in the dorsal region of the perilemniscular nucleus with ipsilateral dominance.

Labeled terminals were seen in the ipsilateral PGlr restricted in the ventralmost zone (Fig. 15F). Labeled cells and terminals were seen in the anterior tuberal nucleus and central posterior thalamic nucleus. Labeled terminals were also seen in the PGa, a medial zone of PGm, and PGlc ipsilateral to the injected side. These terminals showed similar morphologies to those labeled by auditory toral injections. The possibility thus cannot be excluded that some TSc neurons send fibers to the medial pretoral nucleus and collaterals to these preglomerular portions.

Toropetal zone of the descending octaval nucleus or DOm

The sacculus is regarded as a specialized and major auditory receptor organ in cyprinids, although other otolithic organs might also contribute to hearing (Popper, 1983; McCormick, 1999). In the carp DOm, the distribution of the somata and dendrites retrogradely labeled by TSc injections coincided almost perfectly with the saccular terminal zone (see Figs. 5, 14). These results are in accordance with previous studies in goldfish (McCormick and Braford, 1994; McCormick and Hernandez, 1996). The DOm of the present study corresponds to the DOdm and DOi of these authors. Bilateral DOm are reciprocally connected in carp, as has been reported in goldfish (McCormick and Hernandez, 1996) and in a batrachoidid (Edds-Walton, 1998).

The present study suggests that the lagenar and lateral line nerves terminate mainly lateral to the saccular terminal zone, in accordance with previous studies in goldfish (McCormick and Braford, 1994; Puzdrowski, 1989). Terminal zones of the three nerves may overlap, where a small number of toropetal neurons were found lateral to the major neuronal population (present study). Thus, the cyprinid DOm may also relay minor auditory inputs from the lagena and lateral line inputs to the torus. The laterally located neurons of the cyprinid DOm may correspond to those of dorsolateral division of DO in a batrachoidid (Bass et al., 2000), which also receive octaval and lateral line inputs and project to the TSc. The present findings of labeled cells in the descending trigeminal nucleus and magnocellular octaval nucleus after tracer injections to the DOm apparently suggest convergent somatosensory and vestibular inputs to the DOm. An alternative explanation, however, is the slight tracer spread to more lateral portions of DO, which is a probable vestibular and lateral line zone (McCormick and Braford, 1994; present study) and is more likely to have such afferent connections.

The DOdm or its homolog is well developed in “hearing specialists” (Finger and Tong, 1984; McCormick and Braford, 1993; 1994; McCormick, 1997) but is present also in “hearing generalists” (McCormick, 1992; O'Marra and McCormick, 1999; for review, see McCormick, 1999). Spherical and crest cells are present in the DOdm in both goldfish and carp (McCormick and Braford, 1994; McCormick and Hernandez, 1996; present study). However, the present study suggests a small difference in the nuclear organization between carp and goldfish. In carp, DOdm neurons near the midline extend their dendrites to the contralateral nucleus in addition to those in the ipsilateral nucleus. A convergence of bilateral saccular inputs in the DOdm is also known in mormyrids (Kozloski and Crawford, 1998; McCormick, 1999).

Secondary octaval population (SO)

Based on tract-tracing experiments a secondary octaval nucleus was identified in a pioneering study on auditory pathways and was termed the medial auditory nucleus in catfish (Finger and Tong, 1984). However, subsequent studies suggest that the nucleus is in fact composed of primary and secondary nuclei that correspond to DOdm and SO of cyprinids, respectively (McCormick and Braford, 1993; McCormick and Hernandez, 1996). In goldfish the SO is composed of three divisions, dorsal (crest cells), intermediate (spherical), and ventral (fusiform), as revealed by a detailed cytoarchitectonic as well as a hodological study by McCormick and Hernandez (1996). Similarly, three divisions are present also in other “hearing specialists” (carp: present study; catfish: McCormick and Hernandez, 1996; mormyrid: Kozloski and Crawford, 1998; clupeid; McCormick, 1997; also see McCormick, 1999). The intermediate division may be lacking in a batrachoidid (midshipman: Bass et al., 2000) and a percomorph (cichlid: O'Marra and McCormick, 1999), i.e., in “hearing generalists.” Thus, the presence of the intermediate division may be linked with the specialization in the peripheral auditory system and with the well-developed DOdm. Studies in other teleost groups as well as in nonteleost fishes, however, are important to understand the evolution of the SO more precisely.

A previous study in goldfish suggests that the SO receives inputs from the bilateral DOm and sends fibers to the TSc with slight ipsilateral dominance (McCormick and Hernandez, 1996). Specific injections to the SO and DOm of the present study confirmed these connections. In addition, the present study indicates that the bilateral SO are reciprocally connected. The catfish medial auditory nucleus is known to project to the medial pretoral nucleus (Finger and Tong, 1984). The present study indicates that only the secondary nucleus or SO projects to the medial pretoral nucleus in cyprinids. The medial auditory-medial pretoral projection in catfish may originate from the neuronal component that corresponds to the SO in cyprinids.

Labeled cells were found in the medial nucleus of the octavolateral area and in a lateral zone of DOi in the present SO injection experiment. These results, however, may be due to tracer uptake from commissural fibers of these nuclei that course dorsal and ventral to the SO (McCormick and Hernandez, 1996; Yamamoto and Ito, unpubl. obs.).

In the present study we injected tracer specifically to a medial zone of cerebellar crest overlying the SO, an experiment that has not been performed in teleosts so far as we know. Such injections clarified efferents of crest cells of the SO. The crest cells gave rise to only ascending projections. In contrast, injections to the SO proper resulted in labeling descending and commissural fibers as well. These results suggest functional differentiation of the three divisions of SO.

Anterior octaval nucleus

In the present study toral injections labeled neurons in the anterior octaval nucleus, where sparse projections of saccular and lagenar nerves were found. Thus, the anterior octaval nucleus may serve as a minor auditory relay nucleus to the torus. This nucleus may also relay inputs from other inner ear endorgans (McCormick and Braford, 1994). Auditory toral injections did not label cells in the anterior octaval nucleus in goldfish (McCormick and Hernandez, 1996) and a marine catfish, Arius (McCormick, 2001), although these toral projections are present in carp (Echteler, 1984; present study) and a different catfish, Ictalurus (Finger and Tong, 1984). In the present study, only large injections including the TSc and TSvl labeled cells in the anterior octaval nucleus in goldfish. In batrachoidids the anterior octaval nucleus appears to send no or few fibers to the auditory torus (Edds-Walton, 1998; Bass et al., 2000, 2001). The anterior octaval nucleus is known to project to the torus in mormyrids (Bell, 1981; Kozloski and Crawford, 1998), Xenomystus (Braford et al., 1993), and a percomorph rockfish (Murakami et al., 1986a), while not in a different species of percomorph (cichlid: O'Marra and McCormick, 1999). Thus, the anterior octaval nucleus-toral pathway is present in most teleosts with varying degrees of development.

Toropetal neurons along the lateral lemniscus

Toropetal neurons have been observed along the lateral lemniscus in various teleosts (e.g., Echteler, 1984; Murakami et al., 1986a; McCormick and Hernandez, 1996; Kozloski and Crawford, 1998; present study; also see McCormick, 1999). These neurons appear to represent multiple populations. The present study adds new hodological data on these neurons and suggests they may be classified into four populations or nuclei as follows.

Most of the toropetal neurons found along the lemniscus in the rhombencephalon may be SO neurons, as judged from the similar morphology, connections, and continuous distribution with the ventral division of SO. The ventral division of the SO has been suggested to continue rostrally to the dorsal and intermediate divisions (McCormick and Hernandez, 1996). In the present study, TSc and SO injections labeled fusiform neurons along the lemniscus rostral to the SO. These neurons possess dorsal and ventral dendrites similar to neurons of the ventral division of SO. Tracer injections to the DOm labeled similar fusiform neurons as well as terminals along the lateral lemniscus, suggesting that the rostrally located fusiform neurons receive auditory inputs and project back to the DOm as do SO neurons. Although we do not have direct evidence that neurons labeled by toral injections and those by SO or DOm injections are the same neuronal population, it is likely that rostrally located fusiform neurons afferent to the TSc belong to the SO.

Large toropetal neurons were seen in the rhombencephalon ventromedial to the lateral lemniscus, i.e., in the medial reticular formation. These labeled cells were usually larger than ventral fusiform neurons of the SO, and their axes were oriented horizontally (see Fig. 6D). These neurons are hence regarded as reticular neurons, although they have dendrites extending to the labeled toromedullary tract, as do ventral fusiform neurons of the SO.

The third toropetal population is found at the isthmic level in or medially adjacent to the lateral lemniscus. These neurons are ovoid, form a more compact population than more caudal fusiform neurons, and are not labeled by SO or DOm injections. The neuronal group probably corresponds with isthmic reticular nucleus in mormyrids (Bell, 1981; Kozloski and Crawford, 1998). Labeled terminals were seen close to these toropetal reticular neurons after SO injections in the present study. The isthmic reticular nucleus in cyprinids is hence probably an auditory nucleus.

A fourth toropetal population is the perilemniscular nucleus, which is laterally adjacent to the lateral lemniscus and is known as cerebellopetal in goldfish and green sunfish (Wullimann and Northcutt, 1988, 1989). Perilemniscular neurons were labeled in the present materials of TSc injections, in which there was no tracer spread to the overlying cerebellar valvula. We thus conclude that this nucleus projects to the TSc, as has been reported in a marine catfish (McCormick, 2001). Interestingly, toropetal cells and torofugal terminals were restricted in the dorsal zone of the nucleus in the present study, suggesting the presence of subnuclear connectional differences.

Lateral valvular nucleus

Labeled terminals were observed in the lateral valvular nucleus after TSc injections in the present study. This connection has not been reported in previous studies in cyprinids. In a mormyrid, auditory toral fibers are known to terminate in the isthmic granular nucleus (Kozloski and Crawford, 1998), which is a probable homolog of the lateral valvular nucleus (Meek and Nieuwenhuys, 1998). Both superficial and deep zones of the torus (i.e., presumed lateral line and auditory zones) project to the lateral valvular nucleus in a percomorph tilapia (Yang et al., 2004). It is also known in a mormyrid that the isthmic granular nucleus (or ganglion isthmi) receives fibers also from an electrosensory toral nucleus (Haugede-Carre, 1979). Thus, auditory as well as lateral line inputs appear to reach the lateral valvular nucleus, which in turn sends fibers to the cerebellum (Meek and Nieuwenhuys, 1998). Interestingly, toral fibers form large terminals surrounding the somata of target neurons in the lateral valvular nucleus (referred to as “cup-shaped” in cyprinids or “end-bulb” in mormyrids), suggesting that these terminals exert strong influences on the activity of target neurons.

Central nucleus of the semicircular torus (TSc)

The present TSc injection materials resulted in sparse labeled cells and terminals in the TSvl, suggesting that the toral nuclei are connectionally quite independent. It remained unclear whether the few labeled cells in the TSvl send axons and/or dendrites to the TSc in the present study. Toral neurons with dendrites extending in both auditory and lateral line nuclei have been reported in the TSvl of a nonelectrosensory “hearing generalist,” a batrachoidid (toadfish: Edds-Walton and Fay, 2003). The organization of the TSc has also been studied in detail in batrachoidids based on toral injection materials (Bass et al., 2000, 2001). There are differences in the degree of development and the organization of the TS between nonelectrosensory cyprinids (McCormick and Hernandez, 1996; present study) and batrachoidids. In batrachoidids the TSc occupies only a thin periventricular layer of the torus and appears less well developed than in cyprinids. Some neurons of the batrachoidid TSc extend dendrites to the TSvl, suggesting convergence of auditory and lateral line inputs on these neurons (Bass et al., 2000; Weeg and Bass, 2001), which was not observed in cyprinids (present study). Also different from the cyprinids, the TSc and TSvl are reciprocally connected in batrachoidids (Bass et al., 2000; Weeg and Bass, 2001). Detailed analyses on the organization of the TS in other teleosts are required to correlate the differences to the degree of specialization in the peripheral auditory system.

The present study suggests laminar organization of the TSc with respect to auditory afferent fibers. The DOm projects only to the deep (or ventral) zone of TSc, while the SO to both superficial and deep zones, which can be a morphological substrate important for auditory processing in the toral nucleus. In goldfish the lateral line nucleus (TSvl) is present ventral to the TSc (McCormick and Hernandez, 1996). Therefore, the cyprinid TS as a whole may be organized in a laminar fashion. Laminar organization of the torus is also known in other nonelectrosensory teleosts (Ito, 1974). In some electrosensory fishes the electrosensory portion adds to the torus as nuclear masses rather than as additional layers (see Meek and Nieuwenhuys, 1998), and the reason for the organizational differences is an interesting question to be explored.

The present toral injection experiments suggest that the rostrocaudal axis of the TSc is reflected to the caudoventral to rostrodorsal axis of DOm and to the mediolateral axis of the SO. It has been reported that the auditory torus of carp is tonotopically organized (Echteler, 1985b). The axis of tonotopy is rostromedial to caudolateral (rostromedial: high frequency, caudolateral: low frequency), an axis that appears to correspond to the rostrocaudal topography of the TSc suggested in the present study. To establish firmly the presence and pattern of topographical relationship and to correlate it to the tonotopy of the torus, however, we need to inject tracers into different portions of TSc more systematically.

Diencephalic ascending connections of the TSc

The present study indicates that the TSc projects to the central posterior thalamic nucleus and to the anterior tuberal nucleus. These results basically agree with previous studies (carp: Echteler, 1984; batrachoidid: Bass et al., 2000, 2001). In a mormyrid, Kozloski and Crawford (1998) found an auditory toral projection to the posteroventral nucleus of Bell (1981). The homolog of the posteroventral nucleus remains uncertain in cyprinids. Toral projections to the ventromedial thalamic nucleus have been reported in carp and rockfish (Echteler, 1984; Murakami et al., 1986a), but were not observed in the present study.

Auditory toral projections to the PG were found in the present study. Such toral projections have not been reported in a different otophysan catfish (Striedter, 1991), although they have been observed in some teleosts (mormyrid: Kozloski and Crawford, 1998; batrachoidids: Bass et al., 2000, 2001). The auditory preglomerular part has been identified as the PGl in batrachoidids. The present study indicates that the PGa, a medial zone of PGm, and the caudomedial region of the PGl or PGlc receive auditory toral fibers in cyprinids. Toral terminals were seen as a continuum in these preglomerular parts that are composed of neurons with a similar morphology. Auditory toral fibers were not seen in the rostrolateral region of the PGl or PGlr, which clearly characterizes the region and confirms the heterogeneity of PGl pointed out based on nonexperimental materials in goldfish (Braford and Northcutt, 1983). Auditory toral terminals in the three auditory toral recipient parts of the PG share peculiar shapes. Each terminal appears like a donut or a pair of semicircles, presumably encircling a dendrite of the target preglomerular neuron to form a powerful synapse. The three auditory toral recipient parts of the PG also share reciprocal connections with the medial part of dorsal telencephalic area (Yamamoto and Ito, 2005; accompanying article). These preglomerular parts might in fact constitute a single neuronal population. Further experimental data on the connections of each part, however, are required to test this hypothesis.

Connections of the PG are apparently quite diverse among different teleosts, which may represent actual species differences or be caused by inconsistent terminologies. We thus discuss some major points concerning toral connections of the PG. As discussed above, the PGl of batrachoidids receives fibers from the auditory torus, while the PGl of Striedter (1992) or PGlr of the present study does not. Instead, PGl or PGlr in otophysans receives fibers from the lateral line portion of TS (Striedter, 1992; Yamamoto and Ito, unpubl. obs.). A preglomerular region receiving lateral line toral fibers may be lacking in batrachoidids (Weeg and Bass, 2000). In a percomorph rockfish, toral fibers terminate in a part of the PG (PGm of Yoshimoto et al., 1998), which is composed of a small-celled part and a medium-celled part (Murakami et al., 1986a). Specific injection experiments to the TSc and TSvl are required to clarify the toro-preglomerular connectional patterns in percomorphs. The organization and terminology of the PG in mormyrids appear quite distinct from those in cyprinids and acanthopterygians (Meek and Nieuwenhuys, 1998), and homology with other teleosts is quite difficult to establish at present. Clearly, detailed hodological data on the PG in various lineages are important to understand the evolution of this poorly understood complex and then to consider consistent terminology across lineages in the future.

Descending auditory pathways and their relay nuclei

The present study clarified various indirect descending pathways from diencephalic auditory centers via the torus and medial pretoral nucleus to the SO, which in turn projects to the DOm. These findings suggest that feedback control is important for auditory information processing in cyprinids. The significance of descending inputs on the rhombencephalic auditory neurons, in particular crest cells of the SO, is an interesting question to be studied.

Diencephalo-toral connections

The anterior tuberal nucleus and central posterior thalamic nucleus are the diencephalic sources to the cyprinid TSc (Echteler, 1984; present study). Both nuclei receive fibers from the TSc (Echteler, 1984; present study) as well as from the medial pretoral nucleus (present study) and thus presumably form auditory feedback circuits to the torus. Toropetal cells were found only in the central posterior thalamic nucleus in a catfish (Striedter, 1991) and in neither nuclei in batrachoidids (Bass et al., 2000, 2001) and mormyrid (Kozloski and Crawford, 1998), suggesting species differences in the diencephalo-toral connections.

In the present study the morphology of toropetal neurons in the central posterior thalamic nucleus was revealed in detail. They are quite large and possess three types of dendrites. The lateral dendrites are surrounded densely by torodiencephalic fibers and terminals, from which the dendrites may receive inputs. The medial dendrites extend to the smaller-celled portion of the central posterior nucleus and may receive inputs from torodiencephalic fibers terminating there or from other pathways. The rostral dendrites are very long. Surprisingly, the dendrites invade another auditory structure, the anterior tuberal nucleus. The rostral dendrites may receive inputs from torodiencephalic fibers terminating there or from other systems including intrinsic neurons. The dendritic morphology suggests integration of various types of auditory information by the large neurons.

Central nucleus of the torus (TSc)

Injections to the TSc labeled terminals in the SO, but tracer injections to the SO did not label TSc neurons in the present study. Injections involving the SO did not label neurons in the TSc in another study of goldfish (McCormick and Hernandez, 1996). The lack of labeled cells in the torus may be explained by failure to label existing toro-SO neurons or because TSc experiments led to labeling en-passant medial pretoral fibers to the SO. Auditory toral projections to the medial auditory nucleus, which includes secondary auditory neurons, is known in a different otophysan catfish (Finger and Tong, 1984).

The present toral injections also labeled terminals in the granule population, as reported in goldfish (McCormick and Hernandez, 1996). Echteler (1984) also refers to toral terminals in “granule cells immediately adjacent to the IVth ventricle” in carp, which most likely correspond to the granule population of McCormick and Hernandez (1996). This connection may be absent in a catfish (Ictalurus: Finger and Tong, 1984) but is present in a marine catfish (Arius: McCormick, 2001). It is unclear whether a homologous granule population is present in non-otophysans.

Medial pretoral nucleus

The medial pretoral nucleus was first described in catfish (Finger and Tong, 1984) and has been identified in other otophysans (Striedter, 1992; McCormick and Hernandez, 1996; present study) and in a batrachoidid (Bass et al., 2000). The nucleus is a periventricular nucleus in the mesodiencephalic boundary region, and has been sometimes regarded as a part of the torus (Striedter, 1991). This nucleus is connected reciprocally with the auditory torus (catfish: Finger and Tong, 1984; batrachoidid: Bass et al., 2000; cyprinids: present study) and with the SO (present study), indicating that the nucleus forms auditory feedback circuits to the mesencephalic and rhombencephalic auditory structures.

The present study suggests that the medial pretoral nucleus projects to the granule population and lateral part of the granular eminence, both of which in turn send fibers specifically to a part of the cerebellar crest overlying the SO. Thus, the pathways probably influence the activity of crest cells of the SO. Such descending pathways have not been reported in cyprinids or in other teleosts. Although our pretoral injections affected more ventral structures to some extent and reciprocal confirmation injections were not performed, the descending projections are likely efferents of the medial pretoral nucleus for two main reasons: 1) The medial pretoral nucleus appears an important origin of auditory feedback circuits, so it is not strange that the nucleus gives rise to descending pathways specific to the crest cells of the SO via different routes; 2) There appear to be no auditory structures ventral to the medial pretoral nucleus.

The medial pretoral nucleus receives afferents from the central posterior thalamic nucleus and anterior tuberal nucleus and projects to the ventral region of PGlr in otophysans (Striedter, 1991; present study). The medial pretoral nucleus appears reciprocally connected with the perilemniscular nucleus and may also project to the PGa, the medial zone of PGm, and PGlc in cyprinids (present study). The precise connectional pattern of medial pretoral nucleus, however, must be determined by more specific injections and reciprocal confirmation experiments.

Granule population and granular eminence

These populations of granular neurons presumably relay descending pretoral inputs to the crest cells of SO, as discussed above. The granule population corresponds to granule cells associated with medial auditory nucleus of the catfish, which became labeled after tracer injections to the medial auditory nucleus (Finger and Tong, 1984), probably due to involvement of the medial cerebellar crest in tracer injections.

Presumed pretoral fibers in these populations of granular neurons form large terminals with peculiar morphologies, the significance of which remains to be determined. Only a restricted lateral region of the granular eminence receives pretoral fibers, where descending fibers originate at the part of the cerebellar crest overlying the SO. The lateral region of the eminence thus may be specialized for the control of crest cells of the SO among several types of crest cells in the rhombencephalic octavolateral area. Labeled terminals were also seen in the granule population after tracer injections to the SO, and this population of granular cells may also receive ascending auditory fibers.