Phox2b expression in the taste centers of fish

MATERIALS AND METHODS Animals

Fertilized zebrafish eggs were obtained from natural spawning of wild-type lines according to the The Zebrafish Book (Westerfield, 1995). Adult fish were maintained at 28 C with a lighting schedule of 14-hours light and 10hours dark. Embryos were collected within 3 hours of spawning, rinsed, placed into 100-mm Petri dishes containing embryo medium, and allowed to develop at 28.5 C to the required stage. To reduce pigmentation, embryos were transferred to embryo medium containing 0.003% phenylthiourea (PTU) starting from 24 hours post fertilization (hpf). Adult goldfish were obtained from local commercial sources.

Transgenic mouse lines used in this work were the Crereporter line ROSA stoploxYFP (Srinivas et al., 2001) and the Phox2b::Cre(BAC), generated with the bacterial artificial chromosome (BAC) CHORI # RP24-95M11 (D' Autr eaux et al., 2011). This BAC construct recapitulates endogenous expression of Phox2b, as ascertained by comparing the pattern of immunofluorescence with the anti-Phox2b antibodies and anti-GFP antibodies in Phox2b::Cre;ROSAlox stopYFP embryos at developmental stages embryonic (E)8.5, E9.5, E10.5, E11.5, E12.5, E13.5, and E18.5 (D'Autr eaux et al., 2011).

All experimental procedures were approved by the ethical committee in charge of the E ´cole normale sup erieure.

Tissue preparation for immunostaining and in situ hybridization

Zebrafish embryos were collected at the required stage and killed with 0.04% MS222 (ethyl-m-aminobenzoate methanosulfonate; Sigma, St. Louis, MO) in embryo medium. Adult zebrafish and goldfish were anesthetized by using 0.04% MS222 and decapitated when spontaneous respiration ceased; the brains were dissected out in phosphate-buffered saline (PBS; pH 7.4). Embryos and brains were fixed in 4% paraformaldehyde (PFA) in PBS overnight at 4 C, cryoprotected overnight in 20% sucrose in PBS, and then embedded in OCT. Sections (12 lm) were cut in the transverse plane on a cryostat.

Staged mouse embryos (day of vaginal plug was E0.5) were deeply anesthetized on ice, decapitated, and subsequently treated in the same way.

A total of 17 mouse embryos, 5 adult goldfish, 8 adult zebrafish, and 30 larval zebrafish were used for this study.

Western blots

E14.5 mouse embryo and adult goldfish and zebrafish tissues were lysed in cold RIPA buffer (3 ml/g of tissue) containing protease and phosphatase inhibitors (Complete Protease and PhosSTOP Phosphatase Inhibitor Cocktail Tablets; Roche, Penzberg, Germany), by using the Tissuelyser (Qiagen, Chatsworth, CA). The samples were then centrifuged at 12,000 rpm for 20 minutes at 4 C to get rid of the cellular debris. The supernatants were stored at À80 C.

Lysates were denatured in LDS sample loading buffer (Thermo Scientific Pierce, Waltham, MA) containing 1.3 mM b-mercaptoethanol, at 90 C for 5 minutes. Samples were electrophoresed on precast 12% polyacrylamide gels (NuPAGE sodium dodecyl sulfate-polyacrylamide gel electrophoresis [SDS-PAGE] gels; Invitrogen, Carlsbad, CA) and transferred on Hybond-C Extra nitrocellulose membrane (Amersham, Arlington Heights, IL). Membranes were washed in 20 mM Tris-HCl, pH 7.5, 130 mM NaCl, 0.1% Tween (TBS-T), blocked with 5% skim milk in TBS-T, and incubated overnight at 4 C with rabbit antimouse Phox2b antibody (1:2,500) diluted in 1% skim milk in TBS-T. After being washed three times with TBS-T, the membranes were incubated for 2 hours at room temperature with horseradish peroxidase (HRP)-conjugated goat anti-rabbit antibody (1:5,000; Jackson ImmunoResearch, West Grove, PA) in 1% skim milk in TBS-T. After three washes with TBS-T, the membranes were developed with enhanced chemiluminescence (ECL) substrate (SuperSignal West Femto Chemiluminescent Substrate; Thermo Scientific Pierce ) for 2 minutes. The ECL signal was recorded by using the ImageQuant imager LAS 4000 mini (GE Healthcare Life Sciences, Buckinghamshire, UK).

Immunostaining

The sections were incubated with the primary antibody solution overnight at 4 C in PBS/0.1% Triton X-100/20% fetal calf serum (FCS) and revealed for fluorescent observation by using DyLight 488-or Alexa Fluor 488-, Cy3-, and Cy5-labeled secondary antibodies of the appropriate specificity (Jackson ImmunoResearch), for 2 hours at room temperature. In some experiments, following the immunofluorescence protocol, sections were incubated with NeuroTrace 640/660 (cat. no. N-21483; Molecular Probes, Eugene, OR) diluted 1:100 in PBS/0.1% Triton X-100 for 20 minutes at room temperature. Sections were washed twice in PBS/0.1% Triton X-100 for 10 minutes and then in PBS for 2 hours.

Primary antibodies

See Tables 1 and2 for details on primary and secondary antibodies.

An anti-Phox2b rabbit polyclonal antiserum (1:500) (Pattyn et al., 1997) was raised against the 14-amino acid C-terminal sequence of the Phox2b protein (coupled to bovine serum albumin [BSA]), and its specificity was confirmed in mouse by the perfect match between in situ hybridization and immunohistochemistry (Pattyn et al., 1997), and the absence of reactivity on Phox2b knockout animals (J.-F.B., unpublished data), and in rats by suppression of reactivity with the original peptide (Kang et al., 2007). The guinea pig polyclonal antiserum against Phox2b (1:500) (Dubreuil et al., 2009) was obtained against the same peptide, and its specificity was demonstrated in the same way. Because the last 14 amino acids of zebrafish Phox2b differ from the mouse sequence by three residues (Fig. 1A), we tested whether the antibody still recognized specifically zebrafish Phox2b by comparing its pattern of immunoreactivity with that revealed by the zebrafish cRNA probe, which matched perfectly (Figs. 1B-D, 2). Finally, the antibody recognizes a band of the expected molecular weight in mouse hindbrain and zebrafish and goldfish medulla but not in mouse cortex (Fig. 1E).

The mouse anti-Nestin monoclonal (rat-401) antibody (1:300; CA556309; BD Pharmingen, San Diego, CA) was produced in mouse, by using amino acid residues 544-776 as immunogen (Hockfield and McKay, 1985). It recognizes a single band of %198-kDa molecular weight in homogenates of developing neural tube cells by western blot and labels neuroepithelial progenitor cells by immunohistochemical analysis (Hockfield and McKay, 1985). Immunohistochemistry and western blot analysis on neural tube of E15 rats showed that the Nestin antibody only recognizes radial glial cells and dividing neural stem cells in the embryonic but not adult rat CNS (Lendahl et al., 1990). Finally, in mouse, it recognizes a single band in cell lysates of neuroprogenitors from wild-type but not Nestin knockout animals, and immunoreactivity is seen in neuroprogenitor cultures only in cells from wild type but not Nestin knockout animals (Park et al., 2009).

The Islet-1 antibodies (1:100; a mixture in equal parts of 40.2D6 and 39.4D5; Developmental Studies Hybridoma Bank, Iowa City, IA) were raised against the C-terminal portion of rat Islet-1. Their staining pattern in zebrafish neural tube coincides with the sum of the patterns revealed by in situ hybridization with probes for Islet-1 and Islet-2 (Lewis and Eisen, 2003), and is eliminated by injection of one-cell embryos with a mixture of spliceblocking morpholinos directed against Islet-1 and Islet-2 (Hutchinson and Eisen, 2006). They label all hindbrain motoneurons, independently indentified by retrograde labeling (Chandrasekhar, 1997).

The rabbit anti-Olig3 antibody (1:20,000) (Muller et al., 2005) was raised against the affinity-purified fusion of the full-length protein with a HIS-Tag. The staining pattern detected in the spinal cord was indistinguishable from that obtained by in situ hybridization with an Olig3 probe (Zechner et al., 2007) and was abrogated in Olig3 knockout embryos (L.R. Hernandez-Miranda and C. Birchmeier, personnal communication).

Chicken anti-GFP (1:500; GFP-1020; Aves, Tigard, OR) was raised by using purified recombinant enhanced green fluorescent protein (EGFP) as immunogen. This anti-EGFP recognizes the YFP molecules expressed from the Rosa locus in ROSA stoploxYFP mice after recombination with a Cre, according to the expected pattern of the Cre allele, and gives no signal in mice that do not express Cre (D'Autr eaux et al., 2011).

The goat-anti ChAT antibody (1:50; AB114P; Chemicon, Temecula,CA) raised against human placental ChAT, has already been used to describe cholinergic neurons in zebrafish (Clemente et al., 2004;Mueller et al., 2004) and goldfish (Giraldez-Perez et al., 2009) brain. This antibody recognized a single band of 70 kDa in lysate of 293 The guinea pig anti-Phox2b antibody was used only for immunostaining on mouse in Figure 5A. cells transfected with an expression vector containing the sequence of zebrafish ChAT (Volkmann et al., 2010).

Western blot analysis of brain extracts of rat, dogfish, sturgeon, trout (Anad on et al., 2000), and goldfish (Giraldez-Perez et al., 2009) showed that this antibody recognized similar bands between 68 and 72 kDa.

The mouse monoclonal zrf-1 antibody (1:2,500; Zebrafish International Resource Center, Eugene, OR) was raised against a basal lamina and cytoskeleton preparation from adult zebrafish hindbrains and spinal cords (Trevarrow et al., 1990). It was later found to recognize glial fibrillary acidic protein (GFAP) by western blot analysis (Marcus and Easter, 1995). The staining pattern obtained in zebrafish larvae matches that obtained with EGFP expressed under the control of Gfap regulatory sequences in the Tg-gfap:GFP transgenic line (Kim et al., 2008).

The mouse NeuN antibody (1:200) (clone A60; NEUronal Nuclei) specifically recognizes the DNA-binding, neuron-specific protein NeuN, present in most CNS and PNS neuronal cell types (Mullen et al., 1992). In the goldfish optic tectum it was shown to recognize neurons but not non-neuronal cells (astrocytes, oligodendrocytes, endothelial cells, and microglia) (King et al., 2004).

For the experiments on zebrafish we used the anti-Phox2b raised in rabbit, the anti-ChAT, anti-Islet-1, zrf-1, and NeuN antibodies. For the experiments on mouse we used the anti-Phox2b raised in guinea pig, the anti-Olig3, anti-GFP, and anti-Nestin antibodies. For experiments on goldfish we used the anti-Phox2b raised in rabbit, the anti-ChAT and NeuN antibodies.

In situ hybridization

For in situ hybridization, sections were briefly washed with PBS, followed by treatment with RIPA buffer (150 mM NaCl, 0.1% SDS, 1 mM EDTA, pH 8.5, 50 mM Tris-HCl, pH 8, 0.5% sodium deoxycholate, 1% NP-40) for 20 minutes. The sections were postfixed with 4% paraformaldehyde in PBS and washed three times in PBS. Then they were treated with triethanolamine buffer (100 mM triethanolamine, 0.2% acetic acid) containing 0.25% acetic anhydride. The samples were washed three times in PBS and then prehybridized with prewarmed (70 C) hybridization buffer (50% formamide, 5X SSC, 5X Denhardt's solution, 500 lg/ml salmon sperm, 250 lg/ml yeast RNA) for at least 1 hour. The slides were incubated overnight 70 C with 250 ll hybridization buffer containing 200-400 ng/ml of digoxigenin (DIG; Roche) RNA probe, in a humidified chamber (5X SSC, 50% formamide). The slides were washed twice for 1 hour in prewarmed post-hybridization solution (50% formamide, 2X SSC, 0.1% Tween) at 70 C, then washed three times with buffer B1 (100 mM maleic acid, 150 mM NaCl, 0.1% Tween), followed by a 1-hour incubation with B2 solution (10% FCS in B1).

Alkaline phosphatase-conjugated anti-DIG antibodies were diluted 1:2,000 in solution B2 and incubated at 4 C overnight. The following day the samples were washed twice in B1 and incubated in solution B3 (100 mM NaCl, 50 mM MgCl 2 , 100 mM Tris-HCl, pH 9.5, 0.1% Tween) for 30 minutes before the alkaline phosphatase substrate BCIP-NBT (Liquid Substrate Solution; Sigma) was added. Tween (0.1%) was added to the substrate solution to obtain a bluish color of the precipitate. The color reaction was allowed to develop in the dark and stopped with several washes in PBS 0.1% Tween (PBT); then the slides were rinsed briefly in H 2 O, dried, and mounted in Aquatex (Merck, Darmstadt, Germany).

The probes used for zebrafish were the following: Phox2b, Olig4, and a mixture of VGlut2.1 and VGlut2.2 to detect glutamatergic neurons (Higashijima et al., 2004).

Cloning of goldfish VGluT2

Two degenerate oligonucleotide primers were designed, based on two stretches of nine amino acids in VGlut2 shared among mouse and zebrafish VGluT2: GGI-GTI-CA(ACGT)-TA(TC)-CC(ACGT)-GC(ACGT)-TG(TC)-CA(TC)-GG (I standing for inosine) (corresponding to the amino acid sequence GVTYPACHG) and GG(GA)-TCI-GCC-CA(ACGT)-GG(TC)-TG(TC)-TT(TC)-TC(ACGT)-CC (corresponding to the amino acid sequence GEKQPWADP). An 891-bp DNA fragment was amplified by polymerase chain reaction (PCR) from whole brain-derived cDNA and subcloned in pGEMT-easy (Promega, Madison, WI). Orthology with Dr-VGlut2.2 was established by aligning nucleotide sequences using ClustalW2 (Larkin et al., 2007) from the EMBL-European Bioinformatics Institute (http://www. ebi.ac.uk/Tools/msa/clustalw2) and Bayesian inference analysis. The latter was performed on a conserved block of 896 nucleotides with MrBayes (Ronquist et al., 2003). Four chains were run for 2,000,000 generations with the following parameters: General Time Reversible model, four gamma-shaped rate variation categories with a proportion of invariant sites, sampling frequency 250, burn in 2000. The Sialin genes were used as an outgroup. The tree is unrooted and designed as radial logarithmic with TreeIllustrator version 0.52 (Trookens et al., 2005).

The plasmid was digested with NcoI, transcribed with the Sp6 polymerase (Roche) for synthesis of the antisense probe, digested with NdeI, and transcribed with T7 polymerase for synthesis of the sense probe.

The sequence of Ca-VGluT2 has been deposited in Genebank under the accession number JQ435859.

Analysis and imaging

Images were taken with a Leica (Nussloch, Germany) microscope CTR5000, equipped with a Leica DFC420C camera for brightfield, and a Hamamatsu (Hamamatsu City, Japan) Orca ER camera for fluorescent observation. Digital images were processed with Adobe (San Jose, CA) Photoshop for color merging, and, when necessary, for adjustment of brightness and contrast of the entire image. In the case of goldfish, because of the large size of the brain, the photomerge function was used to assemble several fields into one picture.

RESULTS

Expression of Phox2b in the visceromotor and viscerosensory columns of adult zebrafish

We first analyzed Phox2b expression in the medulla of adult zebrafish. Phox2b þ neurons could be divided into two large groups. The first group included all visceral motoneurons (general [or preganglionic parasympathetic neurons] and special [or branchiomotor neurons] hereafter collectively designated as BM/VM neurons), as identified by their position, compared with previously published maps (Wullimann et al., 1996). These were, from rostral to caudal: the trigeminal motor nucleus (with its two divisions, ventral and dorsal [Fig. 2A]), the facial motor nucleus (Fig. 2B), and the juxtaventricular column formed by the visceromotor neurons (Fig. 2C,D). BM/VM neurons coexpressed ChAT (Fig. 2E and inset), in agreement with their cholinergic phenotype. In the isthmus and midbrain, the trochlear and oculomotor nuclei expressed Phox2b (Fig. 3A,B and insets), as they do in mouse (Pattyn et al., 1997;and see Discussion).

The second large group of Phox2b þ neurons was situated in the viscerosensory column, i.e., the facial (Fig. 2C) and vagal lobes (Fig. 2D,E) and the general viscerosensory zone or nCC (Fig. 4D). Coexpression of NeuN indicated that the vast majority of Phox2b þ cells in the taste lobes were neurons (Fig. 2F-H). Expression in these lobes could be further subdivided based on spatial arrangement: cells packed in a dense superficial layer (black arrowheads) overlaid cells situated more deeply, scattered in no overt pattern (magenta arrowheads, Fig. 2C,D). Most Phox2b þ neurons in the facial and vagal lobes coexpressed Vglut2 (Fig. 4 and data not shown), an established marker for glutamatergic neurons in fish and mammals (Higashijima et al., 2004 and references therein), showing that they correspond to glutamatergic neurons, as is the case in the rodent nTS (Kang et al., 2007).

Thus, in fish as in mammals, Phox2b is expressed in BM/VM neurons and in the glutamatergic neurons of the viscerosensory column. However, the latter has a strikingly different structure from the mammalian nTS. To understand the source of this difference, we next com-pared the formation of the viscerosensory column in zebrafish and mouse.

Formation of the visceromotor and viscerosensory columns compared between mouse and zebrafish

In mice, glutamatergic nTS precursors are born between E9.5 and E12.5 (Pattyn et al., 2006) from a neuroepithelial domain termed dA3, located in the dorsal alar plate, which expresses the bHLH transcription factor Olig3 (Storm et al., 2009) (Fig. 5A). Neuronal precursors emerge from dA3 and migrate tangentially (Fig. 5C,D), i.e., perpendicularly to radial glia, to settle in more ventral position (Fig. 5E-H). The dorsal motor nucleus of the vagus nerve concomitantly migrates dorsally, from the ventralmost pMNv (for ''progenitor of visceral motoneurons'') progenitor domain (Briscoe et al., 1999) to a level just ventral to the nTS (Dauger et al., 2003) (Fig. 5E,F). There, vagal motoneuronal precursors reestablish contacts with the ventricle, through processes (Fig. 5G,H and insets) that have been described earlier, although they were thought at the time to be ''remnants'' of an original ventricular attachment (Rinaman and Levitt, 1993).

In zebrafish, at 24 hpf, Phox2b was only detected in the paramedian domain of the neuroepithelium (Fig. 6A and insets), likely corresponding to the pMNv domain of the mouse hindbrain, where visceral motoneurons are born. Accordingly, early postmitotic cells coexpressed Islet-1, an early marker of motoneurons (Ericson et al., 1992) (Fig. 6A). The dorsal alar plate was devoid of Phox2b expression at caudal levels but expressed high levels of Olig4, the orthologue of mouse Olig3 (Tiso et al., 2009) (Fig. 6A,B). At 33 hours the cell arrangement was basically the same, except that postmitotic precursors of vagal motoneurons had accumulated (Fig. 6C,D). At 48 hpf, a chain of Phox2b-positive neurons, roughly parallel to the lateral edge of the hindbrain, emerged from the Olig4-positive dorsal neuroepithelial domain (Fig. 6E,F). At the same stage, VGluT2 was switched on in a column overlapping the Phox2b þ cells, as judged from adjacent sections (Fig. 6E,G) and as previously described (Higashijima et al., 2004). Thus, this dorsolateral column of Phox2b þ cells, by its neurotransmitter identity [VGluT2 þ ] and region of origin [Olig4 þ domain of the neuroepithelium] corresponds to the anlage of the viscerosensory column. Double labeling for Phox2b and Zrf-1, a marker of zebrafish radial glia, showed that the migration of the viscerosensory neuron precursors was aligned with the radial glia (Fig. 6H-J).

At 3 days post fertilization (dpf) the arrangement of the dorsal Phox2b þ cells was basically unchanged, except that a superficial layer of Phox2b À cells was no longer present, bringing the column of Phox2b þ cells just beneath the pia (Fig. 6K), whereas motoneurons tended to occupy a wider area (Fig. 6K) and Olig4 expression had considerably decreased, most likely reflecting the narrowing of the progenitor zone (Fig. 6N). The pattern was basically unchanged at 7 dpf (Fig. 6L,O). At 1 month post fertilization, the number of Phox2b þ cells had vastly increased (Fig. 6M) and, consistent with a continuous production, Olig4 expression persisted in the neuroepithelium (Fig. 6P). The motoneurons, expressing ChAT (Fig. 6M), were lined up dorsoventrally along the ventricular surface. At none of the developmental stages examined (i.e., up to 7 dpf) could we distinguish the vagal from the glossopharyngeal and facial lobes.

Expression of Phox2b in the facial and vagal lobes of Carassius auratus

We next examined expression of Phox2b by immunohistochemistry in the viscerosensory column of adult goldfish. Phox2b immunoreactivity was detected in many cells throughout the vagal, glossopharyngeal, and facial lobes (Fig. 7A). As in zebrafish, the vast majority of Phox2b þ cells coexpressed NeuN and were therefore neuronal (Fig. 7B-D). On closer inspection, four of the five neuronal layers identified in the ''vagal sensory zone'' of Carassius (Morita and Finger, 1985) contained Phox2b þ neurons: layers III (where they were by far the most abundant) and V (which project to the secondary gustatory zone) as well as VII and IX (which project [together with V] to the ''vagal motor zone'' (Fig. 8A,B). Layer XI contained only occasional Phox2b þ neurons (green arrowhead in Fig. 8B). Layer XIV, where motoneurons of the vagal lobe reside (projecting their axons into layer XV), also contained Phox2b þ cells (Fig. 8B) whose moto-neuronal identity was confirmed by coexpression of ChAT (Fig. 9A andinset). The smaller intrinsic neurons described in the superficial region of layer XIV (Morita et al., 1983) were Phox2b-negative (white arrows in Fig. 8B).

We cloned goldfish Vglut2.2 (Fig. 1F,G) and verified that it was massively expressed in layer III (Fig. 9B), most likely in Phox2b þ neurons, showing that they correspond to the principal relay neurons, which are glutamatergic (Ikenaga et al., 2009). Sandwiched between the sensory and motor zones, the fiber layers XII and XIII (which contain, respectively, primary afferent fibers and efferent fibers to the secondary gustatory center) were largely devoid of Phox2b þ cells (Fig. 8A,B). The facial lobe also contained many Phox2b þ neurons (Fig. 9C), organized in one dense superficial layer with a seemingly random pattern underneath, matching their appearance by the Golgi impregnation method in C. carassius (Morita et al., 1983). The superficial neurons and at least some of the deep ones expressed Vglut2 (Fig. 9D). Ventrally to the facial lobe, the facial motoneurons coexpressed ChAT and Phox2b (Fig. 9C, inset). Caudally to the vagal lobes, Phox2b þ neurons were found in medial and paramedian groups of cells corresponding to the lateral and medial subnuclei of the commissural nucleus of Cajal, i.e., the general viscerosensory zone and the area postrema (Morita and Finger, 1987b) (Fig. 10).

DISCUSSION

Homology of the viscerosensory nuclei in aquatic and terrestrial vertebrates

The taste lobes of fish have long been considered homologous to the gustatory portion of the nTS of mammals on hodological grounds, i.e., based on their status as projection sites for gustatory afferents of the facial, (Tiso et al., 2009). H-J: Transverse sections at 48 hpf double-stained with anti-Phox2b and an antibody against radial glia (Zrf-1), as indicated. Note that the migration of the vagal lobe precursors is aligned with the radial glia. K-M: Transverse sections at 3 dpf, 7 dpf, and 1 month, immunostained for Phox2b (green) and Islet-1 or ChAT (magenta). Insets: Closeups of the motoneurons. N-P: Transverse sections adjacent to K-M respectively, hybridized with Olig4. Scale bar ¼ 50 lm in A (applies to A-D); 100 lm in O (applies to E-O) and P (applies to M-P); 25 lm in closeup from A (also applies to C; 20 lm in insets of K (applies to K-M).

[Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.] glossopharyngeal, and vagal nerves (e.g., Herrick, 1899). In fact, homology was first proposed before the gustatory nucleus of mammals was properly identified (as is clear from Herrick, 1905, p. 381), a curious case of an a priori statement of homology. The widespread expression of Phox2b in both the nTS and fish taste lobes that we document here provides a strong additional argument, and the first molecular one, for this homology, as well as the homology of the nCC with the general viscerosensory zone of the nTS. Herrick (1899, p. 216) cautioned that ''if we should attempt to draw up a detailed comparison, the various elements would doubtless not be exactly equivalent in the two groups of animals.'' It is indeed doubtful that each class of interneurons identified in the goldfish vagal lobe by its position in different layers, its morphology, and its connections (Morita et al., 1983) has a homologous counterpart in the nonlaminated mammalian nTS. However, many excitatory interneurons of the goldfish taste lobes, including putative synapomorphic types, have recruited the same transcription factor, Phox2b, which makes them likely ''sister cell types'' (Arendt, 2008), produced from an ancestral relay taste neuron.

Contrasted topology of the viscerosensory nuclei in cyprinid fish and mammals

In his landmark monograph on Menidia, Herrick (1899, p. 213) tried to reconcile the homology of the teleost and mammalian viscerosensory columns with their blatantly different form and position. He argued that the vagal lobes were not ''dorsal'' as they appeared, because they were ventral to the roof plate, and that their inordinate development had crowded out (laterally and ventrally) a topologically-Herrick says ''morphologically''-more dorsal structure, the somatosensory column, or SpV. This theoretical distortion allowed Herrick to fit the viscerosensory column of fish into the intermediodorsal position assigned by the columnar model of the hindbrain inherited from Gaskell (reviewed in Nieuwenhuys et al., 1998). Implicitly, this interpretation displaced ventrolaterally the junction of the tela chorioidea (''beneath'' which the lobes were supposed to develop [Herrick, 1905, p. 384]) with the pia mater (which covers the SpV, and made the entire surface of the vagal lobe ventricular), an intriguing notion, with no further history to our knowledge.

Our data in zebrafish show that the topology of the vagal lobe is in fact irreconcilable with that of the mammalian nTS, the reason lying in a different migration pattern from a similar origin. In mammals, the intermediodorsal position of the nTS arises by a tangential migration from the dA3 neuropithelial domain of the alar plate, in a ventral direction (Fig. 5A-D). On the second axis of the tube, radial (or ventriculo-pial), this tangential migration maintains the nTS close to the ventricle, with the tractus solitarius at its pial border (Fig. 5H). In fish, the Phox2b þ , Vglut2 þ interneurons of the taste lobes are born in the same dA3 domain, marked by Olig4, but their migration is radial, so that they remain dorsal (Fig. 6J); they become even more so once their dorsal neighbors apparently migrate out in a ventral direction, in ways that we have not analyzed. The latter could correspond to the barlh2 þ cells described at a similar position (Kinkhabwala et al., 2011) and could be homologous to dA1-derived precerebellar neurons in mouse (Bermingham et al., 2001), also known to express Barhl1/2 and to settle ventrally. In addition, Phox2b þ neurons migrate to various extentssome remaining ventricular, some ending up pial, and others everywhere in between-so that, viewed in transverse sections, the cells form a column rather than a nucleus, lined up along the radial axis. This cellular arrangement in the developing hindbrain of zebrafish has been described for other neuron types (Higashijima et al., 2004;Kinkhabwala et al., 2011). Note that, relative to body axes, these radial columns of neurons assume a dorsoventral appearance at the hindbrain level due to a ventral deflection of the radial axis in the alar plate (Fig. 6H )-and, at more rostral levels, of the ventricular surface itself, where it becomes T-shaped (e.g., Lyons et al., 2003). This distortion is transient, and later on, the fish viscerosensory column shifts from a lateral to a more dorsal position (in a Cartesian space).

Our data on the formation of the zebrafish vagal lobe are compatible with previous descriptions made in the goldfish (Lamb and Kiyohara, 2005) where proliferation was said to occur ''mainly at the tip of the vagal lobe'' forming ''cohorts that are older at the base of the lobe and progressively younger dorsally.'' This mode of growth was likened by the authors to the ''tangential morphogenetic pattern'' of the fish optic tectum (Nguyen et al., 1999). Topologically, however, i.e., relative to the axes of the developing tube rather than the shape of the adult structure, this growth, at least in zebrafish, is not tangential but radial. From this perspective, the main differences between zebrafish and goldfish would be the much higher number of cells produced in the latter, feeding the bulge of the lobe, as well as lamination, whose mechanism remains to be explored (Finger, 2009). One can speculate that this mode of migration, resulting in a two-dimensional sheet of cells, was key in achieving the fine viscerotopic representation of the oropharynx in the vagal lobe (Morita and Finger, 1985). This representation, together with an equivalent one in the motor nucleus (see below) allows for pointby-point projections between the two (Morita and Finger, 1985), the anatomic basis for the reflex system, whereby local taste sensation triggers local muscle contraction in the palatal organ, enabling food sorting (Finger, 2008). Whether this topology and the mode of migration that brings it about is primitive, or a derived feature of teleosts, will require the study of additional fish clades.

The facial lobe also contains a topological representation of the periphery-a somatotopic representation of the body surface or ''pisciculus'' (Puzdrowski, 1987)-but it differs markedly from the vagal lobe in that it is a medial, unpaired structure. Our data do not elucidate the developmental events leading to this configuration.

Migratory behavior of visceromotor neurons

In contrast to sensory neurons, the migratory behavior of BM/VM neurons is globally conserved in mammals and fish: they are born ventrally from the pMNv domain and migrate tangentially in a dorsal direction (Dauger et al., 2003;Guthrie, 2007;Ohata et al., 2009; this study) to settle dorsal to somatic motoneurons, in the intermedioventral position described by the columnar model of the hindbrain. This migration keeps them close to the ventricle, with which they reestablish direct contacts, at least in rodents. However, the fine settling pattern of motoneurons also diverges between fish and mammals. In the former, and particularly in goldfish, the special visceromotoneurons form an arrayed layer along the dorsoventral axis-again, sheet-like-rather than a compact nucleus, as in the nucleus ambiguus of mammals.

TABLE 1 .

The Journal of Comparative Neurology | Research in Systems Neuroscience