Development 126, 4395-4404 (1999) 4395 Printed in Great Britain © The Company of Biologists Limited 1999 DEV2426

The role of the rhombic lip in avian development

Richard J. T. Wingate* and Mary E. Hatten Laboratory of Developmental Neurobiology, Rockefeller University, 1230 York Avenue, New York, NY 10021-10034, USA *Author for correspondence (e-mail: richard. wingate@kcl. ac. uk)

Accepted 3 August; published on WWW 27 September 1999

SUMMARY

We have used a combination of quail-chick fate-mapping neurons of the lateral pontine nucleus. DiI-labelling of techniques and dye labelling to investigate the development cerebellum explants reveals that external germinal layer of the avian cerebellum. Using Hoxa2 as a guide for the precursors have a characteristic unipolar morphology and microsurgical construction of quail-chick chimaeras, we undergo an orientated, active migration away from the show that the caudal boundary of the presumptive rhombic lip, which is apparently independent of either glial cerebellum at E6 maps to the caudal boundary of or axon guidance or ‘chain’ formation. rhombomere 1. By fate mapping the dorsoventral axis of rhombomere 1, we demonstrate that Key words: Granule cell, Lateral pontine nucleus, Quail-chick precursors are generated at the rhombic lip together with chimaera, Cell migration, Hoxa2

INTRODUCTION constriction, or isthmus, that separates these two embryonic vesicles is an important signalling centre and a number of The vertebrate neuraxis is patterned by dorsoventral and studies have demonstrated that genes induced at the isthmus anteroposterior molecular cues into regionally distinct regulate both cerebellum and development (Lumsden subdivisions each characterised by a different repertoire of and Krumlauf, 1996). However, it is only recently that the neuronal types. In the early neural tube, these complementary boundaries of the anlage have been investigated. By accurately patterning systems can be interpreted in terms of the constructing a fate map according to molecular anatomy, Millet orthogonal axes of a Cartesian co-ordinate system (Lumsden et al. (1996) showed that the posterior boundary of Otx2 and Krumlauf, 1996). However, for CNS structures generated expression indelibly marks the caudal limit of the presumptive later in development, these basic motifs are often obscured by midbrain and hence, by inference, the rostral boundary of the gross morphogenic changes. We have re-examined the cerebellar anlage. Accordingly, targeted mutation of this gene development of one such structure, the cerebellum, in the light leads to a rostral expansion of cerebellum (Acampora et al., of recent molecular evidence indicating that its development 1997). Conversely, a targeted mutation of Hoxa2, the most may be far simpler than previously presumed. In particular, we anterior Hox gene in the , results in a caudal have investigated the role of the rhombic lip, which lies at the enlargement of the cerebellum (Gavalas et al., 1997). Does interface between the cerebellum and roofplate, in the Hoxa2 define the caudal limit of presumptive cerebellum? At induction of cerebellar granule neurons. Over half the adult early embryonic stages, there are no structural landmarks brain consists of cerebellar granule cells, which receive mossy delimiting the region that lies between the limits of Hoxa2 and fibre input chiefly from the and participate in the relays Otx2 expression (Prince and Lumsden, 1994; Millet et al., that govern activity and hence cerebellar output. 1996). However, in the later embryo, this domain resolves to a Granule cell development involves two distinct phases of single defined neuromere designated as rhombomere (r) 1. proliferation and migration (Cajal, 1911). Of these, the second The role of dorsoventral patterning in cerebellar phase of cell division in the external granule cell layer (EGL) development, and in particularly the induction of granule cells, and subsequent, radial, inward migration to form the internal is currently unclear: recent molecular data indicate that granule granule cell layer has been extensively investigated (Hatten and cells are a dorsally derived cell population, but this conclusion Heintz, 1995). By comparison, the first phase of precursor is contradicted by fate-map data and classical neuroanatomy. induction leading to the formation of the EGL is poorly Key to understanding this apparent anomaly is determining the understood. The precise anteroposterior and dorsoventral derivatives and dorsoventral origins of the rhombic lip. Various origins of EGL precursors and their mode of migration over authors have argued that granule cells are born exclusively at the pial surface of the cerebellum remain key questions in the rhombic lip and subsequently migrate over the surface of developmental neurobiology. the cerebellum to form the EGL (Schaper,1897, adapted by The cerebellum arises in a region that initially encompasses Harkmark, 1954a; Miale and Sidman, 1961; Altman and Bayer, the boundary between the mesencephalon and the 1978, 1987, 1997; Hynes et al., 1986; Ryder and Cepko, 1994; (Hallonet et al., 1990; Millet et al., 1996). The Hatten and Heintz, 1995). This is supported by recent studies 4396 R. J. T. Wingate and M. E. Hatten

Fig. 1. (A) Dorsal view of the mesencephalon/metencephalon (mes/met) region of an E2 (stage 10) embryo triple-labelled by in situ hybridisation for Otx2 (blue), Hoxa2 (blue) and Gbx2 (red). Dark-grey shading in the accompanying schematic diagram shows the area from which r1 grafts were derived. (B) Four types of unilateral graft, of varying dorsoventral extent, are represented schematically as transverse sections of r1.

showing that a number of granule neuron markers are initially also expressed at the rhombic lip (Hatten et al., 1997). A targeted mutation of one of these genes, the pro-neural atonal homologue, Math-1, leads to a complete loss of granule cells (Ben-Arie et al., 1997). Similarly, dissociation studies show that all rhombic lip derivatives are apparently positive for a granule cell lineage-specific marker, RU49 (Alder et al., 1996). Both Math-1 and RU49 can be induced in presumptive cerebellum by the dorsalising factor, BMP (Alder et al.,1999), Fig. 2. The relationship between r1, r2 and the E6 cerebellum. (A) Ventricular view of an E3 chimaeric hindbrain showing the indicating a role for dorsoventral patterning in granule cell distribution of cells derived from an r2 graft (green) with respect to precursor determination. However, nearly all anatomical rhombomere boundaries (enhanced by Nomarski optics). Scale bars models suggest that the rhombic lip is a caudal structure and in A,C, 200 µm. (B) Lateral view a whole stage 19 chimaera hence under the influence of a very different set of showing r1 derivatives (brown) in the CNS and cranial nerves. Quail anteroposterior patterning cues (Altman and Bayer, 1978, neural crest cells from r1 can be seen in the ophthalmic lobe of the 1985; Avarez Otero et al., 1993; Ryder and Cepko, 1994). trigeminal ganglion (Vo) and ensheathing the oculomotor (III) and Moreover, a precise fate map of the dorsoventral axis of the trochlear (IV) nerves. (C) Ventricular view of an E6 chimaeric metencephalon at E2 shows that granule cells are derived from hindbrain showing the distribution of cells derived from an r2 graft all but the most ventral neural tube (Hallonet and Le Douarin, relative to Hoxa2 expression. (D) Schematic diagrams explaining the 1993). Therefore, whether the rhombic lip is the sole source of dissection and orientation of the flat-mounted cerebellum (cb) and hindbrain (hb): rhombic lip is shaded blue. In all subsequent figures, granule cells remains unclear and, significantly, the proposal anterior is to the top of the page. (E) At E6, the caudal limit of an r1 that rhombic lip is a dorsally derived structure has yet to be graft maps to the caudal limit of the cerebellum. Quail cells can be confirmed. seen in the trigeminal (V) but not the facial/vestibuloacoustic We have used a combination of techniques to assess the (VII/VIII) ganglia. (F) A ventricular view shows negligible caudal precise caudal boundary of the cerebellar anlage at E6 and the spread from r1. Ventral midline, (vml) is left, while the rhombic lip dorsoventral origins of the rhombic lip. These show that the (rl), fringe (f) and roofplate (rp) are to the right. (G) At E7, Hoxa2 is rhombic lip is a dorsally derived structure and the source of excluded from the cerebellum. (H) In the ventricular layer at E6, the EGL precursors. The caudal limit of presumptive cerebellar anterior limit Hoxa2 matches the r1/2 boundary identified by µ territory corresponds to the boundaries between r1 and r2, grafting. Scale bar for F, H, 300 m. suggesting that, at E6, the cerebellar anlage is derived from r1. EGL precursors have a distinctive morphology and mode of MATERIALS AND METHODS early migration. Perhaps surprisingly, granule cell precursors are generated alongside a population of ventrally migrating Quail-chick chimaeras neurons that give rise to the lateral pontine nucleus. This The metencephalon of E2 (Hamburger and Hamilton stages 10 to 11: suggests a possible lineage relationship between two Hamburger and Hamilton, 1951) Japanese quail embryos (Treslow functionally connected, but topographically and Farms, Chester Town, Maryland) was isolated in vitro and treated with phenotypically distinct neuronal populations. dispase (1 mg/ml in L15 for 15 minutes). Mesoderm and ectoderm Cerebellar rhombic lip derivatives 4397

Immunoresearch Laboratories, West Grove, Pennsylvania). Some were vibratome-sectioned at 100 µm. Brains fixed in Dent’s fixative were double-labelled with Q¢PN and the quail-specific neuronal antibody, QN (Tanaka et al., 1990), and FITC-conjugated and Cy3- conjugated secondary antibodies, respectively. BrdU-labelled cells were identified by a monoclonal antibody (Bio-Science Products AG, Emmenbrücke, CH) and a Cy3-conjugated secondary antibody. Quail explants were labelled with the endothelial and quail cell-specific monoclonal antibody, QH1 (Hybridoma Bank) and a FITC- conjugated secondary antibody. PF fixed tissue was cleared in PBS/glycerol (1:10) with 2. 5% 1,4-diazabicyclo[2. 2. 2]octane and 0. 02% sodium azide. Tissue treated in Dent’s fixative was dehydrated through a methanol series (50%, 70%, 100%, 100%) and cleared in Fig. 3. Proliferation at the E6 rhombic lip. (A) Decreasing densities 1:2 benzyl alcohol/benzyl benzoate. Some brains were processed for of quail cells in a chimaeric cerebellum outline the rhombic lip, in situ hybridisation using Hoxa2 digoxigenin-labelled riboprobes as fringe and roofplate. (B) BrdU incorporation (green) reveals s-phase previously described (Wingate and Lumsden, 1996) or triple-labelled nuclei within the rhombic lip and fringe. Scale bar, 100 µm. with Hoxa2 (digoxigenin), Otx2 (digoxigenin) and Gbx2 (FITC)- labelled riboprobes. were removed and grafts prepared from either the left or right side of Photomicrography and time-lapse imaging r1 (Fig. 1A) using flame-sharpened tungsten needles. Different graft Images were captured using a SPOT camera (Diagnostic instruments types were prepared according to their dorsoventral extent (Fig. 1B): Inc., Sterling Heights, Michigan) or a confocal, inverted microscope whole (0¡-180¡), dorsal (0¡-10 ¡ to -50¡), subdorsal (50¡- 180¡) and ventral (120¡-180¡). Host chicken eggs (Spafas Inc., Preston, Connecticut) were then windowed using sharp scissors and the prepared quail tissue orthotopically grafted into the neural tube of stage- matched hosts. Eggs were re-sealed with ‘Scotch’ tape and incubated for a further 1-6 days at 37¡C, before fixation in either 3.5% paraformaldehyde in 0.1 M PBS (PF) or Dent’s fixative (1:4 dimethylsulphoxide:methanol) overnight at 4¡C. Proliferation assay Chicken eggs were windowed at E3 and re-sealed to prevent the formation of blood vessels over the roof of the egg. At E6, the window was re-opened and approximately 6 µl of bromodeoxyuridine (BrdU: 15 mg/ml in sterile water) was injected via a glass micropipette into a vein within the chorioallantoic membrane. Eggs were re-sealed and re-incubated at 37°C for 30 minutes prior to fixation in PF. DiI-labelling of cerebellar explants Chicken or quail embryos were decapitated at E3-6 and the cerebellar plate rapidly dissected away from surrounding tissue. A fine glass rod coated with either 1,1′-didodecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate (DiIC12) or FM-DiI (Molecular Probes Inc., Eugene, Oregon) was used to focally label the rhombic lip by briefly applying its tip (1-2 seconds) to the ventricular surface. The isolated cerebellum was embedded in collagen (Vitrogen 100, Collagen Bio- Materials, Palo Alto, California), pial side downwards, in a culture dish with glass coverslip base. The gel was Fig. 4. Dorsal r1 grafts map to roofplate and rhombic lip at E6. (A) Graft-derived submerged in a physiological medium adapted from cells restricted only to the roofplate give rise to neural crest (arrow indicates Ard et al. (1985) and maintained at 37¡C and 5% CO2 trochlear nerve). (B) At high power, quail cells spread into the fringe but not for 1-2 days prior to fixation in PF. rhombic lip. (C) Graft-derived cells also line the trochlear nerve. (D) A graft occupying a larger dorsoventral extent of the neural tube gives rise to roofplate, Histochemistry neural crest, rhombic lip and ventrally migrating cells. (E) At high power, migrating Brains and cerebellar explants were processed for cells exit the rhombic lip at the pial surface. (F) By contrast, the pattern of immunohistochemistry using a standard whole-mount ventricular cell mixing is consistent with the random intermingling. Scale bar for protocol (Lumsden and Keynes, 1989). Chimaeras fixed B,C,E,F, 100 µm. (G) Migrating cells emerge perpendicular to the rhombic lip and in PF were labelled using the quail-cell-specific, converge distally. (H) No cells cross the dorsal midline (dml). Scale bar, 200 µm. perinuclear marker Q¢PN (Developmental Studies (I) In a flat mount of the chimaera in D, EGL precursors can be seen to accumulate Hybridoma Bank, Iowa City, Iowa) and FITC- or distal to the rhombic lip (arrow), while a separate population condense lateral to the peroxidase-conjugated secondary antibodies (Jackson ventral midline (asterisk). Scale bar, 500 µm. 4398 R. J. T. Wingate and M. E. Hatten

(MRC 600: BioRad, Welwyn, UK). Some images were enhanced by cerebellar rhombic lip within an E2 embryo for subsequent ‘seed-fill’ filtering for contiguous structures using VoxelView grafts of dorsoventrally defined quail neural tube sections. The software (Vital Images Inc., Minneapolis, Minnesota) running on a rhombic lip was designated as comprising the lateral edge of Silicon Graphics Indigo workstation. For time-lapse confocal the flattened cerebellum and hindbrain (blue shading in Fig. photomicroscopy, laser output was set at 10% of maximum. 2D). The cerebellar rhombic lip is distinguished by a fringe of Sequential photomicrographs were assembled into films using densely packed cells that lie in close contact to the roofplate TweenMachine software (Nick Didkovsky, Rockefeller University) and Voxel Animator (Vital Images). (Fig. 2F). At high power (Fig. 3A), a decreasing packing density of progressively more dorsal quail cells can be seen to determine the transition from neural tube precursor cells in the rhombic lip through the fringe population to the epithelial cells RESULTS of the roofplate. Analysis of short-term BrdU incorporation in E6 cerebellum revealed that both the lip and the fringe contain Determination of the caudal boundary of the a high density of cells in s-phase (Fig. 3B). cerebellum To assess the caudal limit of the cerebellum in terms of the Rhombic lip gives rise to migratory EGL precursors segmental organisation of the early neural tube, we fate Orthotopic grafts were used to determine both the mapped the r1/2 boundary to E6. Quail-chick chimaeras were dorsoventral origin of the cerebellar rhombic lip and its constructed at E2 by the orthotopic transplantation of sections derivatives. Restricted grafts (of variable rostrocaudal extent of neural tube from quail donor to chick host in ovo. At this within r1 territory) were made of the dorsalmost neural tube at age, r1 and r2 are still fused into a single vesicle, the E2 (Fig. 1B) and analysed at E6 (stage 27-30). The range of metencephalon, and are not divided by a recognisable derivatives varied systematically with the extent of graft (Table segmental constriction. Hence, we used gene expression to 1). All dorsal grafts gave rise to both roofplate cells and neural define r1 territory (Fig. 1A). From at least stage 5 (E. Bell and crest. In a few cases (n=3/20), there were no other derivatives R. J. T. W., unpublished observations), r1 is visible as a region apart from these two populations. These chimaeras were that is positive for Gbx2 (Shamim and Mason, 1998) lying characterised by donor cells that lay within the roofplate (Fig. between the expression domains of Otx2 (Millet et al., 1996) 4A), failing to spread into the neural tube (Fig. 4B) and and Hoxa2 (Prince and Lumsden, 1994). By these criteria, we contributing only to neural crest derivatives (Fig. 4C). estimated that r1 occupies the rostral two thirds of the If a dorsal graft was sufficiently large to encompass both metencephalon at stage 10-11. All grafts were unilateral and roofplate and rhombic lip (n=17/20) then two, distinct chimaeras that showed developmental asymmetry were additional populations of neural derivatives were always discarded (n=30/107). The distribution of quail cells was generated: a dorsal pial layer constituting the EGL and a assessed using the quail-specific nuclear marker, Q¢PN. population of ventral cells close to, but not abutting the midline We first confirmed that the chosen caudal boundary of (Table 1). In both cases, these derivatives must reach their final presumptive r1 grafts and the anterior boundary of r2 grafts positions by migration from a pool of precursors lying at the mapped to the r1/2 boundary. A number of chimaeras were interface between neural tube and roofplate; the rhombic lip. hence examined at E3 or E4 (n=13) prior to the loss of These migrations take place outside the ventricular layer. A hindbrain segmentation. Fig. 2A shows a flat-mounted high-magnification view of the pial side of the graft boundary hindbrain in which the anterior limit of an r2 graft, derived reveals streams of cell nuclei that appear to bud from the edge from Hoxa2 expression, corresponds to the emergent r1/2 of the grafted tissue (Fig. 4E). By contrast, within the boundary. Rostral to this boundary, r1 gives rise to neural crest ventricular layer of the same graft, neuroepithelial precursors derivatives in the ophthalmic lobe of the trigeminal ganglion break away from the body of the quail graft and move ventrally and surrounding the trochlear and oculomotor nerves (Fig. 2B). in a far more restricted manner (Fig. 4F), consistent with By the time that the presumptive cerebellum is distinct (E6), mixing as opposed to active migration (Wingate and Lumsden, overt rhombomere boundaries have disappeared. Hence to 1996). Larger grafts reveal an orderly organisation of migratory assess the r1/2 boundary, a number of chimaeras were streams reflected in their fasciculated appearance (Fig. 4G,H). processed for in situ hybridisation for Hoxa2 (n=13). At E6, the r1/2 graft boundary continues to map to the anterior limit of Hoxa2 expression (Fig. 2C). There appears to be no Table 1. Fates of quail cells from different types of r1 graft significant rostral migration of r2 cells into r1. Unilateral, (Fig. 1) in E6 chimaeras whole r1 grafts (n=7) also indicate that the r1/2 boundary maps Derivatives to the caudal edge of the cerebellar plate (Fig. 2E) and reveal Graft type Roofplate Neural crest EGL Ventral cells n that there is no contralateral cell migration between Whole + + + + 7 presumptive cerebellar hemispheres. Within the proliferative, Dorsal + + + + 17 ventricular layer (Fig. 2F), r1-derived territory has a sharp ++−−3 caudal boundary at the lateral angle of the fourth ventricle (the Subdorsal + − ++5 embryonic ‘rautenbreite’: His, 1890). The anterior limit of −−−+5 Hoxa2 expression lies at the caudal boundary of the cerebellum Ventral −−−+9 in whole mounts (Fig. 2G) and, in the ventricular layer (Fig. Graft types were scored for the presence (+) or absence (−) of derivatives in 2H), appears complementary to the caudal boundary of r1- four different tissues within r1; roofplate of the fourth ventricle, neural crest derived precursors (Fig. 2F). (in cranial nerve roots), EGL and ventral postmitotic neurons. Number of These grafts enabled us to define the regional origin of the chimaeras = n. Cerebellar rhombic lip derivatives 4399

Despite only a narrow isthmus of roofplate separating distribution of grafted precursor cells (brown label) from which hemispheres, graft-derived cells never crossed the dorsal they arose. Postmitotic cells have spread rostrocaudally to form midline (Fig. 4H). Streams emerge from all points along the a longitudinal domain directly adjacent to the floorplate in r1. curved edge of the cerebellar rhombic lip, perpendicular to its Ventrally derived postmitotic cells were never observed in interface with the roofplate and merge into a distal dorsal neural tube. accumulation of cells at the rostroventral edge of the anlage. In Fig. 4I (arrow), this accumulation of candidate EGL Rhombic lip derivatives have a distinct morphology precursor cells can clearly be seen when the chimaera in Fig. and migratory behaviour 4D is divided at the dorsal midline and flat-mounted (see Fig. We focally DiI-labelled the rhombic lip of E5 cerebellar 2D). No EGL was observed in embryos at stage 19 or younger explants to investigate the morphological heterogeneity of and distinctive pial migrations were first observed at stage 20. migrating cells and their mode of migration. In particular, we wished to determine whether the orderly and directed Rhombic lip also gives rise to ventral neurons streaming of precursors away from the lip reflected either a In addition to contributing to EGL, a small number of cells structured glial or axonal substratum, or the ‘chain’ migration follow a circumferential path and condense ventrally. In every exhibited in other granule cell populations (Wichterle et al., chimaera where a dorsal graft gave rise to EGL precursors, 1997). Explants were cultured for 24-48 hours and then cell ventral derivatives were also found (Fig. 4I, asterisk; Table 1). dispersal was analysed by confocal microscopy. In all cases, a The identity of these cells was investigated using a quail- variable number of small unipolar cells with a single leading specific neuronal marker QN (Tanaka et al., 1990). Fig. 5 process could be observed streaming away from the injection shows a confocal micrographic montage of a chimaera in site on the pial surface of the explant. In Fig. 7A, a labelled which quail cells derived from a dorsal graft were fluorescently population of cells has been reconstructed from a series of double-labelled with QN (red) and Q¢PN (green). The optical sections. Shown, inset, a single migrating cell has been perinuclear marker, Q¢PN, labels all graft derivatives. These digitally isolated from background label by a seed-fill consist principally of a carpet of EGL precursors overlying the algorithm. The majority of cells were oriented such that their cerebellum and a population of ventral cells that spread single process projected directly away from the rhombic lip. rostrocaudally to form a longitudinal nucleus with a We failed to label any underlying scaffold of orientated glial pronounced, medially displaced, caudal tail. The nucleus lies or axonal processes in any of the explants. Nor did we see any lateral to, but not abutting, the ventral midline. In addition, QN evidence of the homotypic binding that characterises chain reveals at least three populations of graft-derived axons. migration. To confirm that unipolar cell profiles were unique Ventral cells give rise to a longitudinal axon tract extending to rhombic lip-derivatives, we compared the pattern of rostrally into the midbrain and caudally into the hindbrain (Fig. labelling following dye application to the rhombic lip (Fig. 7B) 5a). Weakly labelled contralaterally projecting fibres are and to slightly more ventrally located precursors (Fig. 7C) visible at the midline (Fig. 5b). Arising from the body of the within a number of explants. The latter never gave rise to graft itself, ventral to the rhombic lip, a group of fasciculated migratory, unipolar cells. The dorsoventral specificity of their axons project first ventrally and then caudally towards the origin, orderly directional migration and restriction to the pial hindbrain (Fig. 5c). At higher magnification (inset), no specific surface made these cells strong candidates for EGL precursors association between graft-derived dorsoventrally projecting identified in quail-chick chimaeras. axons and pial migratory precursor cells could be seen. DiI injections also revealed a second population of larger, rounded cells deeper within the mantle layer (Fig. 7A, Contact with rhombic lip correlates with generation asterisk), which readily disperse rostrocaudally and exhibit an of migratory streams amoeboid appearance. Differences in the mode of migration of We transplanted sections of r1 that excluded the most dorsal labelled cells was revealed by confocal time-lapse segment of the neural tube (subdorsal grafts: Fig. 1B) to photomicroscopy. Larger, amoeboid cells show rapid determine whether contact with the rhombic lip is required to movement in various directions and apparent phagocytic generate EGL. In half of chimaeras receiving subdorsal r1 behaviour reminiscent of CNS macrophages (Fig. 7D). To grafts (n=5/10), quail precursor cells failed to spread far confirm their identity, we therefore labelled quail explants with enough dorsally to contact the roofplate (Fig. 6A). In these the quail-specific antibody, QH1, which specifically binds to cases, no migratory, graft-derived EGL precursors were visible cells with an endothelial and haemopoetic lineage (Pardanaud on the pial cerebellar surface in either flatmounts or coronal et al., 1987) and identifies macrophages and microglia sections (Fig. 6B). By contrast, if quail cells mixed sufficiently (Cuadros et al., 1993, 1997). In an explant cultured for 24 with host tissue to come into contact with the rhombic lip (Fig. hours, QH1 reveals blood vessels and two distinct, 6C, asterisk), then labelled cells were also present in the EGL rostrocaudally aligned ranks of macrophages (arrows). Close (Fig. 6C, arrow). Significantly, even where graft-derived cells to the rhombic lip, the outer rank of cells is disrupted at the abutted the roofplate, no neural crest was generated from any point where applications of DiI have been made (asterisks). subdorsal graft (Table 1). When rhombic lip derivatives labelled with a fixable form of Precursor cells derived from grafts of the ventral third of r1 DiI (FM-DiI) are counter-stained with QH1 (Fig. 7F), it is clear never spread to contact rhombic lip and do not give rise to EGL that a number of macrophages (initially outside the rhombic (n=9; Table 1). Donor and host cells in ventral neural tube (Fig. lip) have incorporated the dye (which, by contrast to DiI, is 6D, arrow) intermingled less readily than at graft interfaces in retained in vesicles rather than cell membranes). The size and dorsal neural tube (Fig. 6E, arrow). In Fig. 6D, the distribution distribution of double-labelled cells are reminiscent of the of ventral quail neurons (blue dots) is superimposed on the amoeboid population labelled by DiI injection in chick 4400 R. J. T. Wingate and M. E. Hatten

1985, 1997; Hallonet and Le Douarin, 1993; Ryder and Cepko, 1994). Moreover, the EGL originates at the interface between roofplate and neural tube: roofplate precursors (dorsal to the rhombic lip) give rise to neural crest but not EGL (Fig. 4B), while neuroepithelium ventral to the rhombic lip (Fig. 6A) produces neither neural crest nor EGL precursors. Despite apparent similarities in the origin and behaviour of neural crest and EGL precursors, we demonstrate that these two temporally distinct populations of dorsally derived migratory precursor cells do not share a common lineage. Following subdorsal grafts, in half of the resulting chimaeras, quail precursors reach the rhombic lip by random mixing with host cells where they generate EGL precursors (Fig. 6C), but never neural crest (Table 1). From this, we conclude that quail precursor cells only come into contact with the roofplate after the temporal window for neural crest production has closed. This suggests that granule cell precursors may be induced by local, lineage-independent mechanisms at the rhombic lip subsequent to E3. Such a model is consistent with the ongoing involvement of dorsalising factors such as BMPs in granule Fig. 5. Confocal micrograph of dorsal graft-derivatives fluorescently cell induction (Alder et al., 1999) well beyond their role in labelled by Q¢PN (green, nuclei) and QN (red, axons) at E6. early CNS patterning (Lumsden and Krumlauf, 1996) and Ventrally, graft-derived axons project rostrally, caudally (a) and neural crest production (Baker and Bronner-Fraser, 1997). contralaterally (b). Postmitotic cells close to the dorsal midline of the cerebellum give rise to fasciculated axons (c, shown at high-power inset) that project first ventrally and then turn caudally. Scale bar, 500 µm. explants. Double-labelled cells displaying a primitive, dendritic, microglial morphology (Cuadros et al., 1997) were also observed distant to the injection site (Fig. 7F, inset). This suggests that amoeboid macrophages phagocytose damaged cells within the injection site and subsequently develop into cerebellar microglia.

DISCUSSION

We have investigated the origins of precursors of cerebellar granule cells in the chick using a combination of fate-mapping techniques and direct cell-labelling in cerebellar explants. We have demonstrated that the caudal limit of the cerebellar anlage, at the time of EGL precursor generation (E6), maps to the r1/2 boundary as defined by the expression of Hoxa2. We also show that precursors of the EGL arise at the rhombic lip apparently in response to local, lineage-independent factors that operate at the interface between neural tube and roofplate. Generated alongside EGL precursors are neurons that migrate to ventral neural tube and which extend axons to the hindbrain, spinal cord and contralateral ventral r1. Finally, DiI-labelling Fig. 6. Labelled migratory cells arise at points of contact between in explants reveals that the rhombic lip-derived EGL precursors grafts and the roofplate. (A) Subdorsal graft where quail donor cells have a distinct, unipolar morphology and undergo an orderly fail to reach the rhombic lip and fail to give rise to EGL. Scale bar, migration diametrically away from the lip. This migration 50 µm. (B) In coronal section, absence of quail cells at the rhombic appears to be independent of glial or axonal substratum lip corresponds to an absence of label in the EGL (arrow). Non- guidance or chain migration. neuronal, elliptical, labelled nuclei mark the pia. (C) Where grafted cells reach the rhombic lip (asterisk), EGL precursors are present µ Induction of EGL precursors (arrow). Scale bar in B,C, 100 m. (D) Precursor cells derived from a ventral r1 graft show limited mixing with host tissue (arrow) and give Our grafting experiments reveal that EGL precursors are rise to a longitudinal distribution of postmitotic nuclei (superimposed generated initially at all points along the cerebellar rhombic lip blue circles) that remain ventral. (E) By comparison, dorsoventral which itself is derived from dorsal rather than caudal cerebellar mixing between precursor cells derived from a dorsal r1 graft and anlage (Miale and Sidman, 1961; Altman and Bayer, 1978, host tissue is extensive (arrow). Scale bar for D,E, 300 µm Cerebellar rhombic lip derivatives 4401

Fig. 7. Migrating cells labelled 24 hours after DiI application to rhombic lip of E5 cerebellar explants. (A) Unipolar migrating cells move over the pial surface away from the injection site. Rounded cells disperse more freely (asterisk). (Inset) A digitally isolated unipolar, migratory cell. Scale bar, 100 µm. (B) Migratory streams were labelled from the rhombic lip. (C) More ventral dye application did not label migrating cells. Scale bar for B,C, 200 µm. (D) Time-lapse sequence (40 minute intervals) shows the movement and apparent phagocytic behaviour of a single large, amoeboid cell (red) relative to two EGL precursors (green). Each image is a projection of 15 optical slices through a depth of 75 µm: each optical section is the average of 5 scans. (E) QH1, reveals two ranks of macrophages (arrows) in a (quail) cerebellum explant. Macrophages break rank and invade the rhombic lip at injection sites (asterisks). Scale bar, 200 µm. (F) In a 48 hour quail cerebellum explant, FM-DiI-labelled rhombic lip derivatives (red) intermingle with QH1- positive cells (green). A number of QH1- labelled macrophages (which initially lie outside the rhombic lip) have incorporated FM-DiI (asterisks). (Inset) A double-labelled microglial cell observed some distance from the injection site. Scale bar, 100 µm.

The identification of the rhombic lip as the source of granule cerebellar territory, which demonstrated that EGL precursors cells is consistent with in vitro (Alder et al., 1996) and arise from a broad dorsoventral extent of neural tube at E2 descriptive studies that identify a number of rhombic lip (Hallonet and Le Douarin, 1993). Only grafts of the entire markers as characteristic of a granule cell lineage (Zic-1; dorsal two thirds of the neural tube were sufficient to Aruga et al., 1994; Math-1: Akazawa, 1995; Ben-Arie et al., subsequently label all granule cells at E15, suggesting that 1996; RU49; Yang et al., 1996). Targeted mutation of such there is no spatially defined dorsal origin for EGL precursors. rhombic lip markers leads to either a partial (Zic-1: Aruga et We suggest that this anomaly can be explained by the large al., 1998) or complete loss of granule cells (Math-1: Ben-Arie degree of mixing between precursor cells of the dorsal neural et al., 1997). However, this simple model is apparently tube after grafts are performed at E2 (Fig. 6E; also, Clarke et contradicted by a previous dorsoventral fate map of E2 al., 1998). Such mixing implies that that mitotic progeny of even a relatively ventrally located precursor cell at E2 might reach the rhombic lip by E5. Hence, although the rhombic lip is a dorsally induced structure, at E2 only the ventral third of r1 is definitively excluded from contributing cells to the EGL (Fig. 6D; also, Hallonet and Le Douarin, 1993). This map of potential fates is shown schematically in Fig. 8A. It is unclear Fig. 8. (A) Schematic fate map of r1 shows the assignment of cells from different dorsoventral territories to tissues within the E6 hindbrain, cerebellum and neural crest. Cells from a broad dorsal territory at E2 (green) can contribute to the cerebellum and rhombic lip. However, at E6, only cells at the rhombic lip (dark green) generate EGL precursors. Neural crest derivatives and roofplate appear to arise from a separate dorsal population (red). (B) The direction of migration of the first EGL precursors is orientated directly away from all points along the rhombic lip. Previous models show EGL precursors arising from either (C) the caudolateral angle of the rhombic lip (Miale and Sidman, 1961), or (D) solely from the caudal rhombic lip (Hynes et al., 1986; Ryder and Cepko, 1994; Hatten and Heintz, 1995; Altman and Bayer, 1997). (E) Of previous models, Schaper’s (1897; adapted by Harkmark, 1954a) corresponds most closely to our results (B). 4402 R. J. T. Wingate and M. E. Hatten is whether such de facto lineage restriction from E2 reflects the suggests an affinity between precursors (Hausmann and early specification of ventral r1 to a non-cerebellar fate. Sievers, 1985) and might also be indicative of ‘chain’ migration. However, DiI-labelling did not reveal specific The migration of rhombic lip derivatives homotypic binding between migrating granule cells, We have shown that EGL precursors display a highly ordered correlating with the lack of such behaviour in vitro (Wichterle migration directed away from the rhombic lip (Fig. 8B) and et al., 1997). accumulate, initially, at a point furthest away from their origin. In the absence of a structural scaffold underlying the The initial direction of migration is perpendicular to the axis orientated migration of EGL precursor cells, a molecular of the rhombic lip and hence, in terms of major CNS axes, guidance system might be involved in the form of either varies systematically with the curve of the edge of the diffusible or substratum-bound cues. Of the various cerebellar plate. EGL precursors undergo active migration possibilities, a roofplate-derived chemorepellant, possibly (Hausmann and Sievers, 1985; Ryder and Cepko, 1994) as BMP itself, would provide a parsimonious explanation for the opposed to a ‘passive sprawl’ (Altman and Bayer, 1997). initial direction of cell migration but would not explain the Candidate granule precursor cells have a characteristic unipolar long-range, ventral targeting of derivatives. By contrast, a morphology reminiscent of both migrating cells visualised diffusible, ventrally derived chemoattractant seems unlikely deep in the EGL considerably later in development (Ryder and as a source of guidance at late stages, given the geometry of Cepko, 1994) and tangentially migrating cells in other CNS precursor migration (perpendicular to a curving rhombic lip), regions (O’Rourke et al., 1997). A second labelled population the changing surface of the cerebellum over time and the of larger, highly motile, amoeboid and apparently phagocytic considerable distances involved. Similarly, a co-ordinate cells are positive for the glial marker, QH1, suggesting that system of substratum molecular cues (either membrane- they are macrophages scavenging dying cells from around the bound or secreted) would have to be maintained against a labelling site (Cuadros et al., 1993, 1997). background of considerable cell mixing, rapid growth and Our direct observations of precursor migration help to morphogenic transformation. Alternatively, substratum resolve the question of how the EGL is initially assembled. guidance cues might be generated by precursors themselves Previous models have attempted to reconcile the proposed (Hatten et al., 1982), although such a mechanism would seem origin of its precursors at the rhombic lip (a caudal and medial unlikely to account for the apparent precision of migration structure) with the classical observation that the EGL appears and would not account for the initial trajectory of migration. first laterally (Herrick, 1894; Hanaway, 1967), ‘creeping Ultimately, a combination of such systems might be required, [rostromedially] over the cerebellar surface like a veil’ possibly varying in their contributions as migration paths (Bonnevie and Brodal, 1946). The majority of migration was are established and then maintained during cerebellar hence assumed to be caudolateral to rostromedial, in parallel morphogenesis. with the development of the layer as a whole. This model required that all cells were induced at a single point, the Identity and significance of ventrally migrating cells ‘caudolateral angle’ of the rhombic lip (Fig. 8C; Miale and Large-scale ventral migration of neurons from the dorsal Sidman, 1961). Other models have proposed that coverage is rhombencephalon was first proposed by His (1890) and has achieved by a caudorostral migration sometimes coupled with recently been directly demonstrated in chick using quail-chick a tangential, predominantly lateromedial dispersion (Fig. 8D: chimaera fate maps (Tan and Le Douarin, 1991) and lineage Hynes et al., 1986; Ryder and Cepko, 1994; Hatten and Heintz, analysis by retrovirus (Hemond and Glover, 1993) or dye 1995; Altman and Bayer, 1997). We show here that two injection (Clarke et al., 1998). However, only indirect evidence, different axes of dispersal are not required to explain the initial through ablation (Harkmark, 1954b) or birthdating studies lateromedial development of the EGL. Indeed, the origin of (Altman and Bayer, 1987), supports His’ proposal that the granule cells most closely conforms to their predicted point of rhombic lip itself is the origin of streams of migratory cells. production in the earliest model described – that of Schaper’s Our observations provide the first direct evidence that the (1897, adapted by Harkmark, 1954a; Fig. 8E). The orientation induction of a population of ventrally migrating neurons occurs of the ‘creep’ of the EGL is almost diametrically opposed to precisely at the rhombic lip. These neurons are always the actual direction of precursor migration and appears to result generated alongside EGL precursors suggesting either an from the initial preferential accumulation of actively migrating unusual lineage relationship between two very distinct cells distal (lateroventral) to their origin at the rhombic lip. We neuronal phenotypes or the coexistence of separate precursor predict that subsequent waves of migrating cells stop at pools within the cerebellar rhombic lip. successively more proximal points on the surface of the anlage, Ventral derivatives of the rhombic lip appear to condense perhaps inhibited from further migration by the accumulation in the position of the lateral pontine nucleus (Brodal et al., of older precursors. 1950; Harkmark, 1954b; Armstrong and Clarke, 1979; Tan The orderly migration of rhombic lip derivatives suggests a and Le Douarin, 1991). Their position is consistent with the precise guidance system that is apparently independent of classical description of the nucleus as a subpial condensation, structural substratum cues. Direct DiI-labelling of the rhombic which is lateral to the midline, becoming more medial at its lip fails to reveal an underlying glial or axonal (Hynes et al., caudal pole where it may merge with the medial pontine 1986) fibre scaffold that might direct migration. Similarly, nucleus (Brodal et al., 1950). Longitudinal axonal projections bundles of axons arising from neurons generated close to the from the neurons both rostrally and caudally confirm this rhombic lip (possibly of the fastigial nucleus) did not appear identification (Fig. 5). The lateral pontine nucleus also sends to be closely associated with migratory precursors. The mossy fibres ipsilaterally to the cerebellum (Brodal et al., fasciculated appearance of migratory streams (Fig. 4G,H) 1950; Harkmark, 1954b), although we failed to immunolabel Cerebellar rhombic lip derivatives 4403 these projections at these stages. Ipsilaterally projecting REFERENCES mossy fibres synapse on granule cells throughout both hemispheres (Armstrong and Clarke, 1979). Surprisingly, r1 Acampora, D., Avantaggiato, V., Tuorto, F. and Simeone, A. (1997). Genetic rhombic lip derivatives never contributed contralaterally to control of brain morphogenesis through Otx gene dosage requirement. Development 124, 3639-3650. the medial pontine nucleus whose origins from dorsal neural Akazawa, C., Ishibashi, M., Shimizu, C., Nakanishi, S. and Kageyama, R. tube have been described in chick (Harkmark, 1954b; Tan and (1995). A mammalian helix-loop-helix factor structurally related to the Le Douarin, 1991). Possibly, medial pontine neurons migrate product of Drosophila proneural gene atonal is a positive transcriptional both circumferentially and anteriorly into r1 territory from regulator expressed in the developing nervous system. J. Biol. Chem. 270, more caudal rhombomeres. 8730-8738. Alder, J., Cho, N. K. and Hatten, M. E. (1996). Embryonic precursor cells The shared origin of both granule cells and precerebellar from the rhombic lip are specified to a cerebellar granule neuron identity. nuclei at the rhombic lip might help to explain mechanisms Neuron 17, 389-399. underlying subsequent development of connectivity. We show Alder, J., Lee, K. J., Jessel, T. M. and Hatten, M. E. (1999). Generation of that lateral pontine neurons, which form long range synaptic cerebellar granule neurons in vivo by transplantation of BMP-treated neural progenitor cells. Nature Neurosci. 2, 535- 540. connections with cerebellar granule neurons, are derived from Altman, J. and Bayer, S. A. (1978). Prenatal development of the cerebellar the same spatially defined neuroepithelium as their axonal system in the rat. I. Cytogenesis and histogenesis of the deep nuclei and the target population. A common lineage within such a cortex of the cerebellum. J. Comp. Neurol. 179, 23-48. neuroepithelium might generate a shared repertoire of cell Altman, J. and Bayer, S. A. (1985). Embryonic development of the rat surface molecules that facilitate the subsequent development of cerebellum. I. Delineation of the cerebellar primordium and early cell movements. J. Comp. Neurol. 231, 1-26. appropriate long-range connectivity. Altman, J. and Bayer, S. A. (1987). Development of the precerebellar nuclei in the rat: I. The precerebellar neuroepithelium of the rhombencephalon. J. The r1/2 boundary and the caudal cerebellum Comp. Neurol. 257, 477-489. The first fate maps of the avian cerebellum revealed that the Altman, J. and Bayer, S. A. (1997). Development of the cerebellar system: In Relation to its Evolution, Structure and Functions. Boca Raton, FL: CRC anlage spans the midbrain/hindbrain isthmus at E2 (Hallonet Press. et al., 1990; Alvarez Otero et al., 1993). This indicated, perhaps Alvarez Otero, R., Sotelo, C. and Alvarado, M. R. (1993). Chick/quail surprisingly, that cerebellar boundaries did not map to any chimeras with partial cerebellar grafts: an analysis of the origin and definitive embryonic structures. Recently, a very precise fate migration of cerebellar cells. J. Comp. Neurol. 333, 597-615. map of the isthmus revealed that the constriction itself migrates Ard, M. D., Morest, D. K. and Hauger, S. H. (1985). Trophic interactions between the cochleovestibular ganglion of the chick embryo and its synaptic rostrally, after E2, only coinciding with the caudal limit of targets in culture. Neuroscience 16, 151-70. expression of the midbrain marker, Otx2¸ after E3 (Millet et Armstrong, R. C. and Clarke, P. G. (1979). Neuronal death and the al., 1996). Hence the rostral limit of the cerebellum lies at the development of the pontine nuclei and inferior olive in the chick. caudal edge of Otx2 expression: but the structural isthmic Neuroscience 4, 1635-1647. Aruga, J., Minowa, O., Yaginuma, H., Kuno, J., Nagai, T., Noda, T. and constriction, taken as a grafting landmark in previous fate Mikoshiba, K. (1998). Mouse Zic1 is involved in cerebellar development. maps, does not assume its definitive position until later in J. Neurosci. 18, 284-293. development. In a complementary experiment, we have used Aruga, J., Yokota, N., Hashimoto, M., Furuichi, T., Fukuda, M. and molecular data to define the initially cryptic anterior boundary Mikoshiba, K. (1994). A novel zinc finger protein, zic, is involved in of r2 and show, by fate mapping to E6, that cells from this neurogenesis, especially in the cell lineage of cerebellar granule cells. J. Neurochem. 63, 1880-1890. segment do not migrate rostrally into r1. At this age, the r1/2 Baker, C. V. and Bronner-Fraser, M. (1997). The origins of the neural crest. boundary maps to the caudal boundary of the cerebellum. A Part I: embryonic induction. Mech. Dev. 69, 3-11. previous fate map has indicated that r2 contributes to the Ben-Arie, N., Bellen, H. J., Armstrong, D. L., McCall, A. E., Gordadze, cerebellum at E10 (Marín and Puelles, 1996) suggesting either P. R., Guo, Q., Matzuk, M. M. and Zoghbi, H. Y. (1997). Math1 is essential for genesis of cerebellar granule neurons. Nature 390, 169- that an anterior migration takes place from E6-10, or that our 172. derivation of r2 territory by Hoxa2 expression (confirmed at Ben-Arie, N., McCall, A. E., Berkman, S., Eichele, G., Bellen, H. J. and E3 and E6) is more accurate than previous estimates. The latter Zoghbi, H. Y. (1996). Evolutionary conservation of sequence and explanation is supported by previous results indicating that expression of the bHLH protein Atonal suggests a conserved role in Hoxa2 is not expressed in the cerebellum at E10 (Wingate and neurogenesis. Hum. Mol. Genet. 5, 1207-1216. Bonnevie and Brodal, A. (1946). Hereditary hydrocephalus in the house Lumsden, 1996). However, an absence of r2 neurons from the mouse. IV. The development of cerebellar anomalies during foetal life with definitive avian cerebellum can only be confirmed by detailed notes on the normal development of the mouse cerebellum. Skr. Norske Vid.- analysis of older chimaeras. Considered alongside the Adak. I.M.N.Kl. (IV), 1-60. published isthmic fate map (to stage 20: Millet et al., 1996), Brodal, A., Kristiansen, A. and Jansen, J. (1950). Experimental demonstration of a pontine nucleus in birds. J. Comp. Neurol. 92, 23-70. our observations raises the possibility that the cerebellum is Cajal, S. R. y. (1911). Histologie du Système Nerveux de l’Hommme et des derived entirely from r1. Vertébrés. Paris: A. Maloine. Clarke, J. D., Erskine, L. and Lumsden, A. (1998). Differential progenitor We are grateful to Janet Alder, Esther Bell, Leah Toole and John dispersal and the spatial origin of early neurons can explain the Jacob for their technical advice and help. TweenMachine software predominance of single-phenotype clones in the chick hindbrain. Dev. Dyn. was written by Nick Didkovsky. We thank V. Prince, A. Simeone and 212, 14-26. A. Leutz for Hoxa2, Otx2 and Gbx2 probes, respectively. The Cuadros, M. A., Martin, C., Coltey, P., Almendros, A. and Navascues, J. monoclonal antibodies, QH1 (F. Dieterlen) and Q¢PN (B. Carlson and (1993). First appearance, distribution, and origin of macrophages in the early development of the avian central nervous system. J. Comp. Neurol. 330, 113- J. Carlson), were obtained from the Developmental Studies 129. Hybridoma Bank maintained by The University of Iowa under Cuadros, M. A., Rodriguez, R. J., Calvente, R., Almendros, A., Marin, T. contract N01-HD-7-3263 from the NICHD. The QN antibody was a J. and Navascues, J. (1997). Microglia development in the quail gift of H. Tanaka. R. J. T. W. was funded by an MRC Travel cerebellum. J. Comp. Neurol. 389, 390-401. Fellowship and a Wellcome Career Development Award. Gavalas, A., Davenne, M., Lumsden, A., Chambon, P. and Rijli, F. M. 4404 R. J. T. Wingate and M. E. Hatten

(1997). Role of Hoxa-2 in axon pathfinding and rostral hindbrain patterning. Lumsden, A. and Krumlauf, R. (1996). Patterning the vertebrate neuraxis. Development 124, 3693-3702. Science 274, 1109-1115. Hallonet, M. E. and Le Douarin, N. M. (1993). Tracing neuroepithelial cells Marín, F. and Puelles, L. (1995). Morphological fate of rhombomeres in of the mesencephalic and metencephalic alar plates during cerebellar quail/chick chimeras: a segmental analysis of hindbrain nuclei. Eur. J. ontogeny in quail-chick chimaeras. Eur. J. Neurosci. 5, 1145-1155. Neurosci. 7, 1714-1738. Hallonet, M. E., Teillet, M. A. and Le Douarin, N. M. (1990). A new Miale, I. L. and Sidman, R. L. (1961). An autographic analysis of approach to the development of the cerebellum provided by the quail-chick histogenesis in the mouse cerebellum. J. Exp. Neurol. 4, 277-296. marker system. Development 108, 19-31. Millet, S., Bloch, G. E., Simeone, A. and Alvarado, M. R. (1996). The caudal Hamburger, V. and Hamilton, H. (1951). A series of normal stages in the limit of Otx2 gene expression as a marker of the midbrain/hindbrain development of the chick embryo. J. Morphol. 88, 49-92. boundary: a study using in situ hybridisation and chick/quail homotopic Hanaway, J. (1967). Formation and differentiation of the external granular grafts. Development 122, 3785-3797. layer of the chick cerebellum. J. Comp. Neurol. 131, 1-14. O’Rourke, N. A., Chenn, A. and McConnell, S. K. (1997). Postmitotic Harkmark, W. (1954a). The rhombic lip and its derivatives in relation to the neurons migrate tangentially in the cortical ventricular zone. Development theory of neurobiotaxis. In Aspects of Cerebellar Anatomy, (eds. J. Jansen 124, 997-1005. and A. Brodal), pp. 265-284. Oslo: Johan Grundt Tanum Forlag. Pardanaud, L., Altmann, C., Kitos, P., Dieterlen, L. F. and Buck, C. A. Harkmark, W. (1954b). Cell migrations from the rhombic lip to the inferior (1987). Vasculogenesis in the early quail blastodisc as studied with a olive, the nucleus raphe and the pons. A morphological and experimental monoclonal antibody recognizing endothelial cells. Development 100, 339- investigation of chick embryos. J. Comp. Neurol. 100, 115-209. 349. Hatten, M. E., Alder, J., Zimmerman, K. and Heintz, N. (1997). Genes Prince, V. and Lumsden, A. (1994). Hoxa-2 expression in normal and involved in cerebellar cell specification and differentiation. Curr. Opin. transposed rhombomeres: independent regulation in the neural tube and Neurobiol. 7, 40-47. neural crest. Development 120, 911-923. Hatten, M. E., Furie, M. B. and Rifkin, D. B. (1982). Binding of developing Ryder, E. F. and Cepko, C. L. (1994). Migration patterns of clonally related mouse cerebellar cells to fibronectin: a possible mechanism for the granule cells and their progenitors in the developing chick cerebellum. formation of the external granular layer. J. Neurosci. 2, 1195-1206. Neuron 12, 1011-1028. Hatten, M. E. and Heintz, N. (1995). Mechanisms of neural patterning and Schaper, A. (1897). Die frühsten differenzierungsvorgänge im specification in the developing cerebellum. Annu. Rev. Neurosci. 18, 385- centralnervensystem. Arch. Entw. Mech. Org. 5, 81-132. 408. Shamim, H. and Mason, I. (1998). Expression of Gbx-2 during early Hausmann, B. and Sievers, J. (1985). Cerebellar external granule cells are development of the chick embryo. Mech. Dev. 76, 157-159. attached to the basal lamina from the onset of migration up to the end of Tan, K. and Le Douarin, N. M. (1991). Development of the nuclei and cell their proliferative activity. J. Comp. Neurol. 241, 50-62. migration in the medulla oblongata. Application of the quail-chick chimera Hemond, S. G. and Glover, J. C. (1993). Clonal patterns of cell proliferation, system. Anat. Embryol. 183, 321-343. migration, and dispersal in the brainstem of the chicken embryo. J. Neurosci. Tanaka, H., Kinutani, M., Agata, A., Takashima, Y. and Obata, K. (1990). 13, 1387-1402. Pathfinding during spinal tract formation in the chick-quail chimera Herrick, C. (1894). Architectonic of the cerebellum. J. Comp. Neurol. 1, 5- analysed by species-specific monoclonal antibodies. Development 110, 565- 14. 571. His, W. (1890). Die entwickelung des menschlichen rautenhirns vom ende des Wichterle, H., Garcia Verdugo, J. M. and Alvarez Buylla, A. (1997). Direct ersten bis zum beginn des dritten monats. I. Verlangertes Mark. Abh. kgl. evidence for homotypic, glia-independent neuronal migration. Neuron 18, sachs. Ges. Wissensch. math. Phys. Kl. 29, 1-74. 779-791. Hynes, R. O., Patel, R. and Miller, R. H. (1986). Migration of neuroblasts Wingate, R. J. and Lumsden, A. (1996). Persistence of rhombomeric along preexisting axonal tracts during prenatal cerebellar development. J. organisation in the postsegmental hindbrain. Development 122, 2143-2152. Neurosci. 6, 867-876. Yang, X. W., Zhong, R. and Heintz, N. (1996). Granule cell specification in Lumsden, A. and Keynes, R. (1989). Segmental patterns of neuronal the developing mouse brain as defined by expression of the zinc finger development in the chick hindbrain. Nature 337, 424-428. transcription factor RU49. Development 122, 555-566.