The current study correlates the locations of reticular and motor neurons in the adult goldfish hindbrain to test for retention of the underlying segmental framework throughout ontogeny. The known rhombomeric origin of reticulospinal neurons in embryonic and larval cyprinids served as a reference frame for identifying rhombomeres r1–r8 in the adult goldfish. This provided a framework on which we placed cranial nerve motoneurons and octavolateralis efferent neurons, whose embryonic segmental origins are likewise well documented. The resultant 3D pattern allowed us to infer the shapes of adult rhombomeric territories in the basal portion of the adult brain and assign precise segmental locations for specific brainstem nuclei, including those that are characterized by substantial longitudinal migration.
Reticulospinal neurons in adult goldfish serve as intrinsic neuronal markers of hindbrain segments
The discovery and description of identified nerve cells and equivalent sets of neurons is one of the major contributions of comparative neuroscience to a fundamental understanding of the vertebrate brain (Bullock, 1978, 1984, 2000; Buchanan, 2001). Reticular neurons, present in all vertebrates, form one such equivalence set (Baron van Hoevell, 1911), within which clusters of reticulospinal neurons form specific subsets (Zottoli et al., 2007). In general, the vertebrate reticular formation divides into median, medial, and lateral longitudinal zones, with further rostrocaudal subdivisions in the hindbrain based on anatomical particularities (for review see Nieuwenhuys and Pouwels, 1983; Cruce and Newman, 1984; Wullimann et al., 1996). The medial reticular zone has been further divided into superior, middle, and inferior reticular nuclei that have been used to describe the location of retrogradely labeled reticulospinal neurons in adult teleost fishes (see, e.g., Prasada et al., 1987, 1993; Bosch and Roberts, 1994, 2001; Becker et al., 1997). In addition, these zones have also been used independently as a positional reference for cranial motor nuclei (see, e.g., Luiten and van der Pers, 1977; Oka et al., 1986a) and afferent projections (see, e.g., Oka et al., 1986b; Puzdrowski, 1987-1989).
The discovery of distinct segmental clusters of reticulospinal neurons in the medial rhombencephalic zone of larval zebrafish using whole-mount techniques led to a re-evaluation of the three major divisions in teleosts (i.e., superior, middle, and inferior) as rostral (Ro), middle (Mi), and caudal (Ca) subgroups, each with multiple subdivisions (Kimmel, 1982; Kimmel et al., 1982). Linking the segmental neuronal pattern to the earlier described but largely overlooked rhombomeric pattern of embryonic neuroepithelium helped to establish the current model of hindbrain segmentation. Accordingly, the rhombencephalic medial reticular zone has eight segments (Ro 1–3 [r1–r3]; Mi 1–3 [r4–r6]; Ca 1,2 [r7,8]) that have been identified in embryonic, larval, and juvenile cyprinid fish (Trevarrow et al., 1990; Suwa et al., 1996). Seven of these segments are clearly distinguishable (r1–r7), whereas a much larger but less well-defined eighth appears to be composed of at least four segments (Hanneman et al., 1988; Bass et al., 2008; Ma et al., 2009). A comparison of the present data with those from earlier studies in cyprinids (Lee and Eaton, 1991; Lee et al., 1993) and amphibians (Straka et al., 2001) indicates that somata of identifiable subsets of reticulospinal neurons reside near the centers of segments, whereas the terminations of some of their dendritic fields approximate segmental boundaries. Further analysis of the 3D organization of commissural axon tracts, radial glia, and blood vessels in adult fish brains should allow refinement of the exact configuration of segmental centers and borders known from embryonic stages (Marcus and Easter, 1995; Yoshida and Colman, 2000; Ulrich et al., 2011).
The segments in the adult goldfish hindbrain adopt the 3D form of a series of rostrocaudally and mediolaterally slanted parallelepipeds
Visualization of adult goldfish brain sections in all three planes (i.e., transverse, horizontal, and parasagittal) illustrated the orientation of dendritic trees of backfilled reticulospinal neurons within each hindbrain segment. The horizontal view highlights dendrites in each segment that project caudolaterally, and the sagittal view demonstrates another set of dendrites that projects ventrorostrally. Accordingly, the dendrites approximately fill a 3D parallelogram (i.e., a parallelepiped) that is slanted such that, 1) the most rostral portion of a segment is located ventrally and 2) the lateral portion is more caudal than the medial portion (Fig. 2; Supp. Info. Movie A). In the parasagittal plane, the caudal to rostral slant is most dramatic in the rostral segments (r1–r5), whereas the caudal segments (r6–r8) appear to be more vertically oriented. Therefore, the slanted orientation of a specific segment must be taken into account when assigning a morphological structure to that segment. This is particularly important when linking neuronal identities to specific rhombomeres on transverse sections, because neuronal subgroups known to be located in adjacent segments can occur in the same cross-section (see Figs. 1C, 6D–F). The difficulty in defining specific structures as part of a particular segment can be compounded by sections that are not cut perpendicular to an axis. Furthermore, brain flexures that secondarily change the shape of the brain during development and thus the orientation of neurons or entire nuclei within individual segments further complicate assignment of structures to a particular segment. For example, unlike that of cyprinids, the trout hindbrain has a shallow, U-shaped flexure in the rostral hindbrain where neuronal processes slant rostrally from dorsal to ventral, whereas neuronal processes in the caudal hindbrain appear to slant in the opposite direction (Salmo fario; Beccari, 1922, Fig. 50).
The location of adult motor nuclei relative to the reticulospinal framework shows that the nuclei retain the positions established by embryonic origin or by embryonic longitudinal migration
Previous studies of the adult fish brain have located cranial motor nuclei with respect to superior, middle, and inferior reticular nuclei of the medial reticular zone (e.g., Luiten, 1976). The ability to identify hindbrain segments in the adult goldfish allows cranial nerve motoneuronal and octavolateralis efferent populations (Gilland and Baker, 2005), their dendritic and axonal projections, and their nerve root exits to be placed within a segmental scaffold. Cranial motoneuron somata and their dendrites appear either to be confined to one hindbrain segment (nVr,nVc, nVIr, nVIc, nIX, nX) or to extend over multiple segmental boundaries (nVII; Fig. 8G). Cranial motor nuclei nVII and nIX, as well as octavolateral efferent neurons (REN, CEN, Fig. 8G), are displaced caudally from their segments of origin, which results in sharp lateral bends as their axon bundles turn to exit nerve roots in the original segments. The axonal genu of nIX motoneurons is at the level of the REN in r6, whereas that of the nVII neurons is at the level of the Mauthner cell in r4. Otic efferent neurons that project to the inner ear of the goldfish exit the brain at the r4/r5 border as part of the VIIIth cranial nerve, but their somata are displaced caudally, forming a rostral (REN) and a caudal (CEN) nucleus in segments r6 and r7, respectively (for larval zebrafish see Sapède et al., 2005). In addition, lateral line efferent neurons (not labeled in this study) are intermingled with inner ear efferents in the two nuclei, and their axons have been traced through two pathways that ultimately enter the posterior lateral line nerve; one pathway exits the brain in r6 while the other pathway exits the brain in r4 and projects caudally outside of the brain to join the posterior lateral line nerve (Zottoli and van Horne, 1983).
The location of cranial motor nuclei within the hindbrain scaffold agrees in general with work on larval trout and larval zebrafish. Beccari (1922, Fig. 50) provides a diagrammatic representation of the larval trout (Salmo fario) hindbrain in parasagittal sections on which he draws reticular neurons and cranial motoneurons stained with a Cajal reduced silver method. His dashed lines approximate the center of segments r2–r4, r6, and r8 (Beccari's labels A–E, respectively). If lines are added to mark the center of r5 and r7, the positions of the cranial motor nuclei are nearly identical to what is described here. Our results also agree, in general, with the position of motor nuclei and migratory patterns described for larval zebrafish (Chandrasekhar et al., 1997; Higashijima et al., 2000; Gilland and Baker, 2005), with a few minor differences. Specifically, in larval zebrafish, the somata of nVII appear to be confined to r6 and the rostral portions of r7, whereas, in the adult goldfish, somata of nVII neurons extend from the center of r6, through segment r7, and possibly into the rostral portion of r8. In addition, the somata of nIX neurons in larval zebrafish have been reported in segment r7, whereas they are located in the rostral portion of r8 in the adult goldfish.
Conservation of hindbrain organization during ontogeny offers the possibility to postulate the developmental origin and physiological function of other neuronal phenotypes
The configuration of reticulospinal neurons in the adult goldfish hindbrain clearly demonstrates the conservation of segmental organization during ontogeny. The reticulospinal scaffold provides robust landmarks to link numerous morphophysiological phenotypes described in the literature with the hindbrain segmental framework. The location of neurons based on their neurochemistry within the reticulospinal framework has recently met that goal for larval zebrafish; the immunoreactivity for tyrosine hydroxylase and serotonin has been specifically correlated with elements of the sensorimotor circuitry involved in the startle response known to originate from specific hindbrain segments (Kinkhabwala et al., 2011; Koyama et al., 2011). In addition, the development of a highly detailed “template coordinate system” for reticulospinal neurons in larval zebrafish (Kamali et al., 2009) will assist in the placement and identification of individual reticulospinal neurons on a hindbrain scaffold when comparable studies of adult neurons are undertaken. The distribution of neurons and their processes containing specific transmitters, enzymes, or related chemicals reported for adult teleost fishes can be aligned to the hindbrain scaffold. In goldfish, these functional phenotypes include neurons containing choline acetyltransferase (Danielson et al., 1988; Zottoli and Danielson, 1989; Giraldez-Perez et al., 2009), nitric oxide (Giraldez-Perez et al., 2009), and substance P (Sharma et al., 1989), among others. In each case, specific neuronal landmarks allow correlation with adult segmental regions described here. Segmental evaluation of the reticular pattern in other well-studied teleosts (e.g., midshipman [Brantley and Bass, 1988] and three spined stickleback [Ekström, 1987]) will allow estimation of the broader taxonomic generality of such correlations and should identify unique variations that are certain to exist in such a diverse group.
The most important test for the usefulness of an adult reticular scaffold will be to see whether detailed segmental homologies can be established with the reticular formation nuclei described for nonteleosts, especially elasmobranchs, basal actinopterygians, and tetrapods. Such studies would reveal the generality, or lack thereof, for segmental reticular anatomy as a primitive vertebrate feature and would perfectly complement the large number of detailed reports on brainstem cholinergic neurons, which have been mapped with segmental interpretations for lamprey (Pombal et al., 2001), bichir (López et al., 2013), gar and bowfin (Morona et al., 2013), lungfish (López et al., 2012), and other critical comparative species (see references in López et al., 2012, 2013).
Because their ontogenies are quite similar, the segmental organization of reticulospinal and vestibulospinal neurons in adult cyprinids is likely to be mirrored in the hindbrains of adult amniotes as well. Mapping the neuronal contributions of individual rhombomeres using segmental expression reporters (e.g., r1, Moreno-Bravo et al., 2013; r4, Chen et al., 2012) or of entire brain regions by charting expression of large numbers of transciption factors (Gray, 2013) shows that rhombomeric patterning is likely to prevail in the adult reticular and vestibular anatomy throughout amniotes. It is likely that clusters of reticulospinal neurons, homologous both in developmental history and functional connectivity to those in cyprinids, will be found in adult mammals. For example, about 60 large reticulospinal neurons are found in the ventrocaudal pontine reticular formation in mammals (Davis et al., 1982; Newman, 1985a,b; Lingenhöhl and Friauf, 1994; Yeomens and Frankland, 1996; Martin et al., 2011). These neurons, likely located between r4 and r6 in adult mammalian brains, compose a network that integrates sensory information to initiate startle responses. Their similarity in development and function to reticular neurons in cyprinids indicate that visualization of the adult hindbrain scaffold will provide a useful tool to link the massive amount of anatomical, immunohistochemical, and physiological information already available in the literature (Nieuwenhuys, 2011).
As a practical outcome of the present study, the delineated 3D shape of the goldfish hindbrain segments (parallelepipeds) can be used to map neuronal phenotypes segmentally onto the rhombomeric framework in an adult brain. For example, a vast body of information exists on neuronal identities responsible for vestibulomotor behavior in the adult goldfish (Straka and Baker, 2011, 2013). Accordingly, vestibulo-ocular neurons, retrogradely labeled from the oculomotor nucleus, are located within the limits of the fish anterior octaval nucleus in the rostral hindbrain (Gilland et al., 2004; Straka and Baker, 2013). In the 3D reconstruction in the current study, this nucleus coincides with r1–r3; this segmental identification was facilitated by the overlapping position of the caudal portion of this vestibular subgroup with trigeminal motoneurons in r2 and r3. The r1–r3 hindbrain location of oculomotor nucleus-projecting vestibulo-ocular neurons along with the termination of anterior semicircular canal afferent fibers in this octavolateral region makes r1–r3 a dominant relay area for vertical angular vestibulo-ocular reflexes in adult teleosts (Straka and Baker, 2013). The identical segmental location of this particular population of vestibulo-ocular neurons in adult, embryonic, and larval goldfish (Suwa et al., 1996) suggests a positional persistence of the circuitry during ontogeny. This remarkable stability with respect to hindbrain segmental organization in goldfish also includes known oculomotor-projecting vestibulo-ocular subgroups in more caudal rhombomeres (Torres et al., 1992; Suwa et al., 1999) and likely applies for vestibulomotor pathways in general (Straka and Baker, 2013). The spatiotemporal functionality of the vestibulo-ocular reflex depends not only on vestibular afferent inputs but also on reciprocal commissural inhibitory connections (Straka and Dieringer, 2004). These vestibular commissural neurons, retrogradely labeled from the contralateral vestibular nuclei, form a rostral and a caudal population within the goldfish octavolateral area (Gilland et al., 2004; Straka and Baker, 2013). A comparison of the somal location of these neurons with the established 3D outline of hindbrain segments indicates that the rostral subgroup aligns with r2 and r3 and the caudal subgroup with r5–r7, separated by a zone (r4) devoid of vestibular commissural neurons. The segmental positions of identified subgroups of vestibulo-ocular and vestibular commissural neurons are thus nearly identical to those described for adult frog (Straka et al., 2001, 2003). This suggests a conserved segmental stability of the basic circuitry underlying vestibulomotor behavior. This example of successfully placing specific neuronal subpopulations within segmental domains illustrates the potential to align a large number of other identified neuronal phenotypes in a similar manner in adult goldfish.