Lower rhombic lip-derived cells undergo transmedian tangential migration followed by radial migration in the chick embryo brainstem

Migration behaviour and fate of cells originated from the lower rhombic lip (LRL) was examined in the chick embryo hindbrain. LRL‐derived cells tangentially migrate along the pial surface of the brainstem and form a transient subpial migratory stream. In the initial stages of migration, LRL‐derived cells appose each other or axon‐like processes, which is indicative of mode of homophilic chain migration and/or axophilic migration. Some LRL‐derived cells relocate rostroventrally towards the pontine region, although the majority of them migrate circumferentially to the ventral medulla oblongata. Depending on the stage of generation, LRL‐derived cells undergo transmedian migration; late‐generated LRL‐derived cells preferentially colonize the contralateral brainstem compared with early generated cells. Thus, latecomer neuron precursors may migrate past their predecessors in the migratory stream. When LRL‐derived cells leave the subpial migratory stream, they change their migratory direction to a radial one and relocate inwardly, with a profile that resembles a tangential‐to‐radial change seen in cerebellar granule cell precursors. After they enter the parenchymal region of the brainstem, they exhibited morphological differentiation, and some differentiate into excitatory neurons. The present results suggest that LRL‐derived cells migrate across boundaries such as midline or rhombomere, which may facilitate to build up cellular and functional architectures of the hindbrain.

Keywords: EGFP; glutamate; in ovo electroporation; precerebellar nuclei; raphe

• Abbreviations
• AEMS anterior extramural migratory stream
• BrdU 5‐bromo‐2′‐deoxyuridine
• CLSM confocal laser scanning microscope
• ECN external cuneate nucleus
• EGFP enhanced green fluorescent protein
• LRL lower rhombic lip
• PB phosphate buffer
• PEMS posterior extramural migratory stream

The lower rhombic lip (LRL) is a dorsolateral edge of the developing brainstem that continues to the tela chorioidea ([15]), and which generates neurons in the precerebellar nuclei and probably also raphe nuclei ([10]; [12]; [1], [2], [3]; [7], [8]; [35]). LRL‐derived cells tangentially migrate in the superficial part of the brainstem, forming two distinct transient subpial migratory streams ([4], [5]; [34]). One is a marginal migratory stream (or posterior extramural migratory stream, PEMS; [4]) observed in the caudal medulla oblongata, which generates the lateral reticular and external cuneate (ECN) nuclei. The other is a pontine migratory stream (or anterior extramural migratory stream, AEMS, [5]), which leads to the basal pons. Histological analyses suggested the transmedian migration of cells in the PEMS ([12]; [1]; [16]). Recently, in vitro flat whole‐mount cultures of the rat hindbrain allowed for the observation of the migration of LRL‐derived cells across the midline ([46]).

Although extensive morphological studies indicated the mode of neuronal migration and putative molecular mechanisms that regulate this migration, migratory behaviour of individual cell and fate of LRL‐derived cells are poorly understood. Especially, transmedian migration is not substantiated in vivo (or ovo), except in quail‐chick chimeric experiment ([44]), which suggests that several brainstem nuclei are composed of cells with various sites of origin including the contralateral side. A detailed description of migratory cell morphology is apparently required for an understanding of the migration pattern and cellular differentiation of LRL‐derived cells in vivo.

In the present study, we examined migration and fate of LRL‐derived cells in the chick hindbrain. The in ovo electroporation of enhanced green fluorescent protein (EGFP) gene was used to label LRL cells of the chick embryo, and following cell migration and cellular differentiation were then observed. Our results provide in ovo direct evidence of the transmedian migration of LRL‐derived cells across the ventral midline, and show that subpial migrating cells exhibit morphological changes from tangential unipolar to radial T‐shaped profiles. In addition, these cells were found to differentiate into glutamatergic neurons.

Fertilized chicken eggs that were purchased locally were incubated at 38 °C. The stages of development of the chick embryos were determined according to [11]). All procedures were approved by the Institutional Animal Care Ethics Committee of the Shimane Medical University and the Okazaki National Research Institutes.

The in ovo electroporation method was previously described ([27]). Briefly, a pGAP‐GFP plasmid, which encodes a fusion protein of EGFP and N‐terminal 20 amino acid sequence (membrane targeting signal) of GAP43, was used as a reporter gene. The gene product associates with the cell membrane ([27]) and a Golgi staining‐like profile of labelled cells can be observed using this method ([43]). Eggs were windowed at E4–5 (HH stage 22–27). A plasmid solution was injected into the fourth ventricle via a micropipette, and electrodes were laterally placed on the LRL (anode) and the opposite side of the embryo (cathode). A rectangular pulse at 35 V for 50 ms was delivered twice by a pulse generator (SEN‐7203, Nihon Koden, Tokyo, Japan). The window was sealed with adhesive tape and the eggs were returned for further incubation.

The EGFP transfected hindbrain was fixed at E6–9 (HH stages 28–34), 2–5 days after electroporation, by immersion in 4% paraformaldehyde in phosphate buffer (PB; pH 7.2, 0.1 m) at 4 °C for 3–5 h, and then stored overnight in PB containing 20% sucrose. After the fixed tissues were observed under an epifluorescent microscope (ECLIPS; Nikon, Tokyo, Japan) to check if transfection had occurred, they were embedded in 5% agar and cut into 70‐µm‐thick sections using a Vibratome. The sections were lightly stained with Hoechst 33258 (0.1 µg/mL, 10–30 min) to observe nuclear distribution or cytoarchitecture. After a brief wash, they were mounted on gelatin‐coated glass slides, coverslipped with phosphate‐buffered saline–glycerol or polyvinyl alcohol and observed under an epifluorescent or a confocal laser scanning microscope (CLSM; Leica, Heidelberg, Germany). To track the profile of the labelled migratory cells, eight images of the optical sections at regular intervals were taken by a CLSM and digitally merged onto one plate. EGFP‐labelled cells in some specimens were immunohistochemically visualized. The sections were incubated with rabbit anti‐EGFP polyclonal antibody (diluted at 1 : 2000; [43]; a gift from N. Tamamaki, Kyoto University) overnight, and subsequently processed by the ABC method (ABC Elite kit, Vector Laboratory, Burlingame, CA, USA) or incubated with anti‐rabbit IgG conjugated to FITC (Cappel, Durham, NC, USA) or to Alexa Fluor 546 (Molecular Probes, Eugene, OR, USA). To study the differentiation of LRL‐derived cells, free‐floating sections of EGFP‐transfected LRL were incubated with rabbit anti‐serotonin antiserum (1 : 5000; [37]; a gift from N. Okado, Tsukuba University) or with anti‐glutamate antibody (rabbit polyclonal antibody 1 : 2000; Arnell, NY, USA or mouse IgM monoclonal antibody, 1 : 2; [42]; a gift from Dr C. Takayama, Hokkaido University) overnight. They were subsequently incubated with species‐specific secondary antibody conjugated to Alexa Fluor 546 or 594. These sections were mounted onto gelatin‐coated glass slides and observed as above‐mentioned.

The formation and general morphology of the migratory stream was examined in the E5–10 chick embryos. The embryo brain was fixed for a few hours with Bodian‐II fixative (70% ethanol, 5% acetic acid and 5% formalin), 6 h after the administration of 5‐bromo‐2′‐deoxyuridine [BrdU, 100–300 µL of 5 mg/mL in phosphate‐buffered saline; Sigma, St. Louis, MO, USA]. The fixed brain was dehydrated through a graded ethanol series, cleared with xylene and embedded in paraffin. Six‐micrometer serial sections were coronally cut and mounted on albumin‐coated glass slides. Every eighth section was stained with haematoxylin and eosin or with cresyl violet. For immunohistochemistry, the sections were irradiated with microwaves for 5 min in 0.05 m citric acid (pH 6.0) over 90 °C. After washing, they were incubated with primary antibody overnight. These sections were subsequently processed using the ABC method and visualized with diaminobenzidine as previously described ([31]). The primary antibodies used were anti‐BrdU mouse monoclonal antibody (1 : 2000; PharMingen, San Diego, CA, USA), anti‐NeuN mouse monoclonal antibody (1 : 2000; Chemicon, Temacula, CA, USA) and anti‐Pax6 antiserum (1 : 2000; [18]; a gift from N. Osumi, Tohoku University).

Bright‐field and epifluorescent photomicrographs were taken by a digital camera (DC200, Leica; DM70, Olympus, Tokyo, Japan). All digital images including CLSM pictures were imported into Photoshop 5 (Adobe Systems), and all figures were constructed after the contrast and brightness were appropriately adjusted.

We first examined the formation of the marginal migratory stream through Pax6 expression, as well as Nissl staining, in the medulla oblongata, as Pax6 is expressed in the cells of the marginal migratory stream ([16]). The medulla oblongata of the E5 chick was composed of a three‐layered structure, namely the ventricular zone, mantle layer and marginal zone. No marginal migratory stream had formed by this stage (not shown). In the E6 medulla oblongata, although the three‐layered structure of the medulla oblongata looked almost the same as that of the E5 medulla oblongata (Fig. 1A), the base of the LRL had cell strands that extended ventromedially along the subpial surface (inset in Fig. 1B). These cells were immunoreactive to Pax6 (inset in Fig. 1C). This was the first sign of the onset of formation of the marginal migratory stream. The remaining portion of the marginal zone was still composed of only fibrous structures.

Graph: 1 Developmental appearance of the subpial marginal migratory stream in the medulla oblongata. (A–C) E6 (HH stage 29). (D–I) E8 (HH stage 33). All are coronal sections. (A) E6 medulla oblongata with Nissl staining. No subpial cell strand has formed in the ventral midline region. (B and C) Higher magnification pictures of the lower rhombic lip (LRL) and its adjacent areas stained by the Nissl method (B) and Pax6 immunohistochemistry (C). Note that cell stream formation starts from the bottom of the LRL (arrows) and is magnified in insets. (D and E) E8 medulla oblongata, stained by the Nissl method (D) and Pax6 immunohistochemistry (E). Note that Pax6‐positive cells are seen along the pial surface. The areas pointed by arrows in D and E are magnified in inset, and the boxed areas in D and E are magnified in G and F, respectively. Arrowheads in insets indicate subpial migratory cells. (F–I) Higher magnification pictures of the ventral midline region stained with anti‐Pax6 antibody (F), by the Nissl method (G), anti‐NeuN antibody (H) and anti‐5‐bromo‐2′‐deoxyuridine (BrdU) antibody after a 6‐h pulse of BrdU (I). Note that cells in the marginal migratory stream (arrows in G) were NeuN‐positive (H) but did not incorporate BrdU (I). Scale bars, 500 µm (A, D and E); 200 µm (B and C); 50 µm (F–I); 20 µm (inset of B–E).

To study the migration and differentiation of LRL‐derived cells, focal transfection of the LRL with the reporter gene encoding EGFP was carried out on E4 and E5 chick embryos (HH stage 22–27), and labelled cell morphology and distribution was sequentially analysed by E9 (HH stage 35). This method enabled us to label cells in a very limited and targeted region of the CNS and study their migration and morphological change in ovo as well as in vitro ([27]; [41]; [13]; [46]).

When the plasmid had been transfected into the LRL at E4 and the brainstem of the embryo was examined on E6 (Fig. 2A and F), EGFP‐labelled cells were seen just beneath the pia mater and slightly deep to the pia (Fig. 2C, D and F). They were mostly unipolar or asymmetrical bipolar in shape and were aligned along the pial surface (Fig. 2B–E and G). Putative leading processes extended in a ventral or ventromedial direction in parallel to the pial surface. The length of the leading processes varied, sometimes up to 500 µm (Fig. 2G). They also occasionally possessed a growth cone at their tip (Fig. 2H) and short trailing processes behind the cell bodies. Cell bodies of unipolar or asymmetrically bipolar cells frequently made a contact with the leading process of other migratory cells (Fig. 2D) or axon‐like fibres (arrows in Fig. 2E). Labelled perikarya were ipsilaterally restricted to the site of plasmid transfection, although labelled processes extended to the contralateral side across the midline (Fig. 2B). This ipsilateral localization of labelled cell bodies was maintained in the E7 brainstem following the E5 transfection (data not shown).

Graph: 2 Initial dispersion of LRL‐derived cells to form the migratory stream. Two cases (A–E and F–H) with restricted EGFP transfection in the LRL. (A) A whole‐mount dorsal view of the E6 brainstem electroporated with the EGFP gene on E4. Note the restricted expression of EGFP (arrows). Arrowheads indicate the posterior margin of the cerebellar anlage. (B) A whole‐mount ventral view of the same brainstem as in A. An axon (arrow) labelled with EGFP is observed to cross the ventral midline (broken line). (C) A coronal vibratome section cut through a site of EGFP transfection. Many unipolar cells expressing EGFP are streaming along, in parallel to the pial surface from the transfection site. (D and E) Higher magnification pictures of migrating LRL‐derived cells expressing EGFP. They are apposed to each other (arrows in D) or axon‐like fibres (arrows in E).  (F) Another example of restricted EGFP transfection in the LRL (arrow). (G) Migrating LRL‐derived cells in the superficial part of the ventrolateral brainstem. Arrows indicate a cell with an asymmetrical bipolar shape and a long leading process whose tip is magnified in G. (H) A higher magnification picture of the growth cone at the tip of a leading process. BS, brainstem. Scale bars, 200 µm (A); 100 µm (B, C, F and G); 20 µm (D, E and H).

Graph: 3 Transmedian migration of LRL‐derived cells. (A) Whole‐mount dorsolateral view of the brainstem showing a restricted EGFP transfection of the LRL (arrow). (B) EGFP‐expressing cells on the brainstem subpial surface. Cells at which the arrow points are magnified in C. (C) Subpially located cells that are rostroventrally orientated (arrows). (D–F) Distribution of EGFP‐expressing LRL‐derived cells in the brainstem coronal sections arranged in a rostral (D) to caudal (F) manner. EGFP‐expression is immunohistochemically shown with anti‐GFP antibody. Each dot indicates a labelled cell, and the arrow in F indicates the site of transfection. The level of the section in (D) is at about 800 µm rostral to the site of transfection. Note the presence of several labelled cells in the contralateral brainstem (arrowhead in F). Scale bars, 1000 µm (A); 500 µm (B and D–F); 100 µm (C).

We next examined whether there is any correlation between the stage of plasmid transfection and the appearance of EGFP‐labelled cells in the contralateral brainstem. When the plasmid was transfected at E5 and the brainstem was examined at E8 or E9, more than 70% of the embryos contained labelled cells in the contralateral brainstem (Fig. 4B, 14 among 17 cases). In contrast, in cases of E4 electroporation, approximately 75% of embryos contained labelled cells exclusively in the ipsilateral brainstem of the age‐matched animals (Fig. 4A, 26 among 34 cases). Thus, only one‐quarter of the embryos contained labelled cells in the contralateral brainstem. Later labelling yielded more labelled cells in the contralateral brainstem than early labelling. Although EGFP expression in the LRL lasted to E5 onwards after E4 transfection, the resultant distribution of contralaterally located EGFP‐positive cells following E4 transfection was different from that following E5 transfection. It is possible that EGFP gene transfected on E4 was diluted in the ventricular zone by cell proliferation and thus E8 or E9 brainstem that was subjected to electroporation on E4 did not show similar phenotypes (contralaterally located labelled cells) to those cases where electroporation occurred on E5.

Graph: 4 Transmedian migration of late generated cells. Distribution of EGFP‐expressing cells in the E9 brainstem followed by EGFP transfection on E4 (A) or E5 (B). EGFP expression is immunohistochemically shown with anti‐GFP antibody followed by the ABC method. Each dot indicates a labelled cell. Note that transfection on E5 resulted in more labelled cells in the contralateral brainstem (B). Although EGFP transfection widely occurred in the lateral brainstem, no EGFP‐expressing cells were encountered in the area of contralateral brainstem subjected to EGFP transfection on E4 (A). Scale bar, 500 µm.

Labelled cells were also distributed in raphe nuclei, inferior olivary nucleus, pontine nuclei, the reticular formation of the ipsilateral brainstem and the reticular formation in the contralateral brainstem of the E9 embryo (Figs 3 and 4), although the boundaries of these nuclei were not clear exactly.

The subpial marginal migratory stream of the E8 or E9 brainstem contained asymmetrical bipolar cells crossing the midline just beneath the pia mater (Fig. 5B). In addition to tangentially orientated cells, inwardly orientated radial cells were observed at these stages (Fig. 5C–F). These cells inwardly extended thicker leading processes and left thin trailing processes behind them (Fig. 5E). Continuity between the trailing process behind the perikaryon and tangential fibres in the subpial region was sometimes observed, and exhibited a T‐shaped profile for radial cells (Fig. 5F). Under closer examination of EGFP‐labelled cells in the subpial region, some tangentially orientated cells were observed extending radial processes with various lengths toward the inner part of the brainstem (Fig. 5H and I). In addition, there were some inwardly orientated cells whose perikarya were still positioned in the subpial region (Fig. 5J). Such radially orientated cells or cells extending radial processes were observed all the way to the midline from the LRL, although the midline region most frequently contained such cells. A limited number of labelled cells in the raphe region and the reticular formation possessed more processes that resembled those of dendrites in the E9 brainstem (Fig. 5G). It is likely that terminal or morphological differentiation had been initiated by this point.

Graph: 5 Morphological differentiation and orientation change from a tangential to radial direction. (A–G) An E9 caudal medulla oblongata subjected to EGFP transfection on E5. (A) Site of EGFP transfection. (B) A tangentially migrating cell crossing the ventral midline. (C) A low‐magnification picture of the ventral midline region. EGFP‐expressing cells are bilaterally observed along the floor plate (FP). The cell indicated by an arrow is magnified in E. (D) Bipolar cells observed along the contralateral floor plate. (E) Radially orientated cells with asymmetrically bipolar shape, whose trailing process is continuous with the tangential fibres (arrows) in the submarginal area. (F) An inverted T‐shaped cell with a tangential fibre in the marginal zone (arrows). (G) Morphologically differentiating cells showing dendritic arborization in the E9 raphe nuclei. (H–J) The presumed sequence of orientation change from a tangential to radial direction. Arrows indicate inwardly orientated radial processes. Within the marginal migratory stream, a tangentially orientated cell extends a short radial process (H), which becomes longer (I), and the perikaryon subsequently becomes radially orientated (J). Scale bars, 500 µm (A); 100 µm (C); 50 µm (B, D and G–J); 20 µm (F).

LRL‐derived cells may have colonized the ventromedial brainstem including the raphe and inferior olivary nuclei (see above). The differentiation of LRL‐derived cells was examined in the E9 or E8 brainstem. Because a portion of LRL‐derived cells settled in the raphe nuclei, which possess serotonergic neurons, we first examined whether they express serotonin or not. No EGFP‐positive cells were immunoreactive for serotonin by E9 (Fig. 6A–C; 7 cases). Next, we labelled EGFP‐positive cells with anti‐glutamate antibody. Although sections treated with the antibody exhibited homogenous staining, a limited number of EGFP‐expressing cells showed glutamate‐like immunoreactivity in the raphe nuclei, inferior olive and reticular formation of the E9 brainstem (Fig. 6D–F; 5 cases). Although it was difficult to precisely enumerate the percentage of glutamate‐like immunoreactive cells among the EGFP‐positive cell population in this experiment as the staining pattern is usually diffuse and not limited on the labelled cells, less than one‐quarter of the EGFP‐expressing cells showed glutamate‐like immunoreactivity in five cases (embryos).

Graph: 6 Possible differentiation of LRL‐derived cells into excitatory neurons. (A–C) EGFP‐expressing LRL‐derived cells show no serotonin‐like immunoreactivity. (D–F) Glutamate‐like immunoreactivity in the EGFP‐expressing LRL‐derived cells. The arrow indicates a double‐labelled cell. Scale bar, 50 µm.

Neural cell migration occurs during the formation of the vertebrate CNS. During cortical layer formation, neuron precursors undergo radial migration from their sites of origin, and processes of radial glia provide a scaffold for radial migration ([24]; [32], [33]; [14]). In addition to radial migration, neural cells migrate in parallel to the ventricular or pial surface in several regions of the CNS ([1], [4], [5]; [17]; [20]; [23]; [36]; [45]). Tangentially migrating cells are relocated at relatively long distances, sometimes greater than 1 mm. Although these two modes of migration are quite distinct from each other, cerebellar granule cells are known to utilize both modes to reach their destination ([26]; [34]).

The present experiment clearly demonstrated migration pattern of LRL‐derived cells in ovo, which was summarized in Fig. 7, and differentiation of these cells into glutamatergic excitatory neurons. These phenomena were firstly demonstrated in vivo and described in detail in this study. Individual profiles of migrating LRL‐derived cells have not been observed up until now.

Graph: ig. 7. FSchematic diagram of lower rhombic lip (LRL)‐derived cell migration. LRL‐derived cells sequentially leave their site of origin, and tangentially migrate to form a subpial migratory stream (A). Late generated cells migrate past early generated cells in the migratory stream (B). They change their polarity from a tangential to radial direction (C), and subsequently migrate radially into the parenchymal region of the brainstem (D). 4V, fourth ventricle; FP, floor plate.

LRL‐derived cells start migrating around E6, and the subpial cell strand becomes continuous at the ventral midline by E8. During the early stages of the LRL‐derived cell migration, EGFP‐labelled cells are apposed to each other and axon‐like fibres (Fig. 7A). This was observed in the foetal mouse brainstem by electron microscopy ([28], [29]). Thus, LRL‐derived immature neurons may relocate in a chain and/or neurophilic migration manner, and neuronal components may guide this tangential migration. In addition, tangentially migrating cells bear short trailing processes, which probably are representative of the locomotion type migration ([25]).

The present results clearly demonstrated that some LRL‐derived cells migrate across the midline and colonize the contralateral brainstem. Because contralateral labelling occurs more frequently in cases with E5 transfection of the LRL than in cases with E4 transfection, latecomer neuron precursors (labelled on E5 in this experiment) migrate past their predecessors (labelled on E4) in the migratory stream (Fig. 7B). This migration pattern is somewhat similar to that of the inside‐out pattern of the cerebral cortex.

The mechanisms regulating transmedian migration of LRL‐derived cells are, in part, similar to those responsible for ventral commissure formation in the spinal cord. Netrin‐1 attracts LRL‐derived cells to the ventral midline region ([39], [38]; [19]; [49]), and the combination of TAG‐1/axonin‐1 and Nr‐CAM may guide cells when they cross the ventral midline ([40]; [6]; [21]). The expression of these molecules may be spaciotemporally regulated in relocating LRL‐derived cells, and the majority of early generated cells may be devoid of signals required for transmedian migration.

The present results show that LRL‐derived cells undergo tangential‐to‐radial change in migration direction of cells upon leaving the migratory stream (Fig. 7C and D). Radially orientated cells maintain their trailing processes, which are continuous with tangential fibres in the submarginal zone, which results in a T‐shaped cell profile (Fig. 7D). This observation indicates that radially migrating cells are undergoing somal translocation ([25]). The sequence of polarity change and the resulting profile (Fig. 5) is similar to those of cerebellar granule cells in the external granular layer. Morphological similarity suggests that these distinct cell lineages share similar mechanisms of movement during polarity or migration direction changes. Common molecules are expressed in the external granular layer and in PEMS, such as Tag‐1/axonin‐1, Pax6 and PACAP (pituitary adenylate cyclase‐activating polypeptide). Among them, Pax6 is required for the polarity change of cerebellar granule cells ([47]). Both radial fibres of Bergmann glia and parallel fibres provide morphological cues for radial migration in the cerebellum ([24]; [32]; [26]; [14]; [30]). The terrain over which cells radially migrate may be similar between the superficial part of the cerebellum and that of the brainstem; parallel fibres correspond to tangential fibres in the submarginal zone of the brainstem, and Bergmann fibres correspond to radial fibres of the floor plate. These similarities in molecular and geometrical environments result in the resemblance of the developmental features between these cellular groups.

In the rodent brainstem, apart from the PEMS, LRL‐derived cells form AEMS, through which cells migrate towards the pontine flexure and form pontine nuclei ([1], [4], [5]). Pontine nuclei of the chick embryo are smaller, but rostrocaudally longer, than those of the mouse embryo ([12]; [44]; [22]) and, therefore, a distinct migratory stream may not be obvious in the chick embryo. A quail‐chick chimera experiment found that chick pontine nuclear neurons originate from several rhombomeres ([44]; [22]). Because a small number of EGFP‐labelled cells were positioned approximately 1 mm rostral to the site of transfection (Fig. 3), inter‐rhombomeric migration ([22]) likely occurred. LRL‐derived cells migrate across boundaries, such as the rhombomere and midline.

Labelled cells were located in the raphe nuclei, reticular formation, inferior olive and caudal part of the pontine nuclei, as well as in the migratory stream, but not in the dorsolateral brainstem corresponding to the ECN of rodents. The survival period after electroporation may not have been long enough for some cell groups, which settle in those regions farthest from the LRL.

The settling of the LRL‐derived cells in raphe nuclei was suggested to occur in the rat brainstem ([7], [8]) and chick embryo ([44]). Tan and Le Douarin hypothesized that raphe nuclei are composed of cells of two different origins; one type originates from the local ventricular zone and the other from the alar plate (alar raphe). The present study supports them both; raphe nuclei contain serotonin‐positive/EGFP‐negative cells and serotonin‐negative/EGFP‐positive (LRL‐derived) cells. LRL‐derived cells were glutamate‐like immunoreactive, and are probably excitatory neurons. The differentiation of serotonergic neurons requires specific signals in the dorsoventral and anteroposterior axes, such as sonic hedgehog (Shh), fibrobrast growth factor 4, Nkx2.2 and Pet‐1 ([48]; [9]). The LRL is located in an area corresponding to the dorsal‐most region in the dorsoventral axis, and cells in this region may be least exposed to ventral signals such as Shh. Therefore, LRL‐derived cell differentiation into excitatory neuron but not serotonergic neurons supports the previously made hypothesis. Long‐distance tangential migration of LRL‐derived cells enables mixing of cells of different origin, and such cellular mixture may be required to build up functional organization of the CNS.

We wish to express our appreciation to Nobuo Okado (Tsukuba University), Nobuaki Tamamaki (Kyoto University), Noriko Osumi (Tohoku University) and Chitoshi Takayama (Hokkaido University) for providing antibodies, and also to Hirohide Takebayashi and Seiji Hitoshi (National Institutes for Physiological Sciences) for critical reading of the manuscript. This study was supported by the Brain Science Foundation, Narishige Neuroscience Foundation and by Grant‐in‐Aid for Scientific Research (12680775, 14580770), and for Priority Areas (10156228), provided by the Ministry of Education Culture, Sports, Science and Technology.