Segmental identity and cerebellar granule cell induction in rhombomere 1
© Eddison et al 2004
Received: 01 March 2004
Accepted: 15 June 2004
Published: 15 June 2004
Skip to main content
© Eddison et al 2004
Received: 01 March 2004
Accepted: 15 June 2004
Published: 15 June 2004
Cerebellar granule cell precursors are specifically generated within the hindbrain segment, rhombomere 1, which is bounded rostrally by the midbrain/hindbrain isthmus and caudally by the boundary of the Hoxa2 expression domain. While graded signals from the isthmus have a demonstrable patterning role within this region, the significance of segmental identity for neuronal specification within rhombomere 1 is unexplored. We examined the response of granule cell precursors to the overexpression of Hoxa2, which normally determines patterns of development specific to the hindbrain. How much does the development of the cerebellum, a midbrain/hindbrain structure, reflect its neuromeric origin as a hindbrain segment?
We show that a Gbx2-positive, Otx2-/Hoxa2-negative territory corresponding to rhombomere 1 forms prior to an identifiable isthmic organiser. Early global overexpression of Hoxa2 at embryonic day 0 has no effect on the expression of isthmic signalling molecules or the allocation of rhombomere 1 territory, but selectively results in the loss of granule cell markers at embryonic day 6 and the depletion of cell bodies from the external granule cell layer. By comparison the trochlear nucleus and locus coeruleus form normally in ventral rhombomere 1 under these conditions. Microsurgery, coupled with electroporation, to target Hoxa2 overexpression to rhombic lip precursors, reveals a profound, autonomous respecification of migration. Rhombic lip derivatives, normally destined to occupy the external granule cell layer, violate the cerebellar boundary to form a ventrolateral nucleus in a position comparable to that occupied by rhombic lip derived neurons in rhombomere 2.
Different overexpression strategies reveal that the recognition of migration cues by granule cell precursors is dependent on their identity as rhombomere 1 derivatives. Segmental patterning cues operate autonomously within the rhombic lip precursor pool. By contrast, a subset of coextensive nuclei is refractory to ectopic Hoxa2 and is presumably induced solely by isthmic organiser activity. Thus, graded (isthmic) and segmental mechanisms may operate exclusively of one another in the specification of different neuronal populations within rhombomere 1. The early designation of an Otx2-negative, Hoxa2-negative region, prior to the appearance of the isthmic organiser, is a key initial step in the specification of the cerebellum.
Subdivision along the rostrocaudal axis of the developing central nervous system assigns regional identity to neuronal precursor pools, influencing the fate of their progeny and generating structural diversity . Strategies for conveying fine-grain positional information vary along the length of the axis. For example, the hindbrain is a transiently rhombomeric structure organised by segmentally expressed Hox genes. By contrast, the development of the midbrain and forebrain, which lie within the Otx expression domain, is regulated by signalling centres generating graded patterning cues. At the transition point, where the hindbrain system of segmentation gives way to regional patterning by local gradients in the midbrain, the most anterior hindbrain segment, rhombomere 1 (r1), expresses neither Hox or Otx genes and displays aspects of both patterning systems. A number of distinct neuronal populations are generated in r1, including cerebellar granule cells , the most populous neuron in the vertebrate CNS. How graded and segmental cues interact in specifying r1 precursors is key to understanding the origins of the cerebellum, a major structure in the higher vertebrate brain.
A number of studies have examined different facets of r1 specification. The rostral boundary of r1 comprises the midbrain/hindbrain isthmus, an important signalling centre in the development of the midbrain and r1. Loss of isthmic signalling severely disrupts both midbrain and cerebellar development [3–5], while application of FGF8, the putative isthmic signalling molecule, suggests a direct involvement of the isthmic organiser in cerebellar induction [6, 7]. Moreover, the trochlear motor nucleus and the locus coeruleus, the latter of which is rostrocaudally coextensive with granule cell production, are born within r1 and are induced by isthmic FGF [5, 8, 9].
In addition to its isthmic identity, r1 also displays properties of a hindbrain segment. Rhombomeres 1 to 3 are defined by the expression of Gbx2: within this domain, r1 differs from the more caudal segments by the absence of Hox gene expression. In mice, a null mutation in Hoxa2, the only Hox gene expressed in adjacent r2, results in the caudal expansion of the cerebellum  and rostralisation of the second branchial arch to a first arch identity [11, 12]. This suggests, in the absence of any changes in isthmic signalling, that suppressing Hox gene expression in the hindbrain is sufficient to confer an r1 identity, as experimentally confirmed in zebrafish . Correspondingly, in chick, overexpression of Hoxa2 in r1 induces ectopic r2-like motor neurons , while in the first branchial arch, neural crest adopts second branchial arch characteristics [15, 16]. These apparent transformations support a model that r1, as r2, is a homeotic iteration within the meristic series that comprises the rhombencephalon.
To examine isthmic and segmental cues in cerebellar development, we have used a combination of retroviral and microsurgical techniques to assess patterning in r1, in particular with respect to the production of granule cell precursors from the rhombic lip. This population undergoes two distinct phases of migration, the first giving rise to the external granule cell layer (EGL), which covers the pial surface of the cerebellum, and a second in which granule cells migrate radially inwards to form the granule cell layer. We find that while nuclei whose induction is dependent on isthmic signalling remain unchanged, the formation of the external granule cell layer is inhibited by ectopic Hoxa2. This suggests that both segmental and isthmic-derived patterning mechanisms may operate independently within r1. Furthermore, microsurgically targeted misexpression of Hoxa2 in the rhombic lip reveals that the loss of granule cell from the cerebellum can be explained by a dramatic re-specification of tangential migration. A complementary grafting strategy demonstrates that these changes in the patterning of migration occur autonomously at the rhombic lip.
Given that the isthmus has an important role in patterning the midbrain and maintaining r1 identity, we first assessed the effect of global Hoxa2 overexpression on key isthmic genes. The RCASBP(B)Hoxa2 retrovirus was injected into E0 chick eggs. The expression of FGF8 (Fig. 1C), Engrailed (en)1 (Fig. 1D) at E3 and Wnt1 at E6 (data not shown) were examined at the midbrain/hindbrain boundary. The gross morphology of the neural tube in these embryos was normal and r1 territory formed appropriately. Isthmic genes were expressed in domains encompassing the isthmus that were indistinguishable from controls (injected with RCANBP(B)Hoxa2; data not shown). This suggests that the signalling properties of the isthmus are unaffected by ectopic Hoxa2 at the isthmus.
Having determined that isthmic function was likely to be unimpaired, we chose to examine the effect of ectopic Hoxa2 on the development of three definitive r1 neuronal populations: cerebellar granule cell precursors, trochlear motor neurons and the locus coeruleus. Of these, only the origin of cerebellar granule cells has been microsurgically fate-mapped to r1 in chick . Trochlear motor neurons and the catecholaminergic interneurons of the locus coeruleus can be allocated to r1 by the expression of their associated characteristic neuronal markers in relation to the molecular r1 boundaries defined above.
Cerebellar granule cell precursors are generated at the rhombic lip (E6 onwards), which comprises the interface between the neural tube and non-neural roofplate of the fourth ventricle. During their characteristic ventrolateral tangential migration to form the EGL, granule cell precursors express a number of distinct molecular markers, including a paired-homeodomain transcription factor, Pax6 , and the cell surface receptor, erbB4 . At E6, Pax6 labels cells condensing as the EGL in dorsal r1 (Fig. 2B). This dorsal expression is lost following Hoxa2 overexpression (Fig. 2C), along with a small domain of ventricular Pax6 expression (Figs 2B,2C, white arrows), just anterior to the normal limit of Hoxa2 expression (Fig. 2B, inset). Loss of dorsal Pax6 suggests that rhombic lip derivatives either fail to be generated or that they down-regulate Pax6 during their tangential migration. At E9, the EGL is strongly positive for erbB4, as illustrated in an embryo infected with the control virus RCANBP(B)Hoxa2 (Fig. 2D). Following overexpression of Hoxa2, erbB4 becomes restricted to small islands of superficial cerebellum (Fig. 2E, arrows) overlying a locally thickened cerebellum. As with Pax6, no ectopic expression of erbB4 was observed that might account for an apparent loss of granule cells. The depletion of this cell type was confirmed by cresyl violet staining of control RCAN- (Fig. 2F) and RCAS-infected (Fig. 2G) cerebellum. Strikingly, Hoxa2 overexpression at E2 results in a cerebellum displaying a clearly demarcated external cerebellar layer at E9 that is nevertheless devoid of cell bodies.
By contrast, in dorsal r1, Phox2a expression was induced throughout the presumptive cerebellum excluding the rhombic lip (Fig. 3D, asterisk). In transverse section, ectopic Phox2a expression in the cerebellum lies within the mantle, outside the cell-depleted EGL (Fig. 3E, asterisk). Dorsal r1 is the origin of ventrally migrating locus coeruleus neuroblasts, raising the possibility that Hoxa2 may have generated an ectopic locus coeruleus population within dorsal r1. However, in situ hybridisation for the enzyme tyrosine hydroxylase (TH), which is expressed at the earliest stages of locus coeruleus migration , demonstrates that ectopic Phox2a-positive cells are not catecholaminergic: in RCAS infected embryos, TH is expressed only within the ventral pool of locus coeruleus neurons (Fig. 3F).
The loss of granule cell markers and up-regulation of Phox2a in deeper cerebellar layers raise the possibility that the migration and/or fate of rhombic lip derivatives are specifically affected by ectopic Hoxa2. Correspondingly, a previous lineage study hypothesised that the locus coeruleus is derived from the EGL , which might favour an argument that Hoxa2 transfection biases production of rhombic lip derivatives towards a deeper-migrating, Phox2a-positive derivative.
Figure 4B shows a wholemount cerebellum from a control (uninfected) quail–chick chimaera at E6. Darkly stained graft-derived precursors occupy the narrow insertion of roofplate at the midline of the cerebellum. Where quail cells overlap the rhombic lip (more caudally), migrating EGL precursors are distributed across the surface of the cerebellum (black arrow). In Fig. 4C, the distribution of rhombic lip derivatives is profoundly changed when quail tissue electroporated with Hoxa2 is used as donor tissue (n = 6/7). No quail cells are seen in the presumptive cerebellum (dorsal r1). Instead, rhombic lip derivatives expressing high levels of Hoxa2 condense in ventrolateral r1 (white arrow).
The effect of exclusively targeting Hoxa2 to the rhombic lip is seen more clearly in transverse section. In control chimaeras, only a small population migrates into ventrolateral r1 (Fig. 4D), while the majority of migrating cells condense in the EGL. A few scattered cells lie within the pool of Phox2a-positive, presumptive locus coeruleus neurons. When Hoxa2-grafted r1 quail cells express Hoxa2, they condense in a dense cluster within ventrolateral r1 (Fig. 4E). The boundaries of their distribution are the same as the relatively small ventrolateral contribution in normal embryos. Despite the expanded rhombic lip contribution to ventral r1, the expression of Phox2a on control and operated sides of the brain is equivalent, suggesting that locus coeruleus formation is independent of changes in the population derived from the rhombic lip (Fig. 4E). Furthermore, while RCASBP(B)Hoxa2 induces a profound change in the migration path, transfected rhombic lip derivatives are unlikely to account for the Phox2a-expressing neurons in the cerebellum following a global virus transfection.
Previous studies have shown that the overexpression of vertebrate Hox genes in the hindbrain consistently results in transformations that can be described as homeotic, in that they recapitulate segmental characteristics of specific rhombomeres. Hoxa2 chiefly regulates the identity of neurons in r2: thus, r1 rhombic lip cells transfected with Hoxa2 might be expected to behave as r2 rhombic lip derivatives. In the absence of appropriate molecular markers for specific r2 populations, we used a grafting strategy to test this hypothesis.
We then grafted equivalent dorsal quail r2 heterotopically into r1, to directly compare the behaviour of r2 rhombic lip derivatives with that of r1 derivatives infected with Hoxa2 (Figs. 4, 5). In contrast to orthotopically grafted tissue, the distribution of r2 quail cells within r1, analysed at E6 (n = 6), more closely resembled that of normal r1 rhombic lip derivatives. Rhombic lip from r2 contributed to both the EGL and a ventrolateral nucleus (Fig. 6C). The difference in behaviour between r2 dorsal derivative within r2 (Fig. 6A) and r1 (Fig. 6B) can be attributed to the respecification of r2 precursors grafted into r1. In situ analysis of heterotopically grafted r2 demonstrates that, as early as E4 (n = 7), endogenous Hoxa2 expression is down-regulated in donor quail cells (Fig. 6D), presumably by signals from the isthmus [24, 25]. Thus, it is impossible within r1 to directly compare the migration path of rhombic lip derivatives from Hoxa2-infected r1 with those from r2 (Fig. 6A,6B). However, Hoxa2 overexpression within the cerebellar rhombic lip results in a rhombic lip migration path reminiscent of that in r2 (Fig. 6A,6B), supporting a conclusion that the transformation is homeotic. Furthermore, the maintenance of an r2-specific migration pattern is correlated with the maintenance of endogenous Hoxa2 expression.
We found that the initial allocation of r1 territory, the expression of isthmic signalling genes FGF8, Wnt1 and en1, the formation of the trochlear nucleus and the establishment of the locus coeruleus were unaffected by ectopic Hoxa2 expression within r1. By using retrovirus injection into pregastrula embryos, overexpression was induced before the establishment of either an Otx2-, Hoxa2-negative domain (at st.6) or isthmic organiser (st.8; ). Thus, ectopic Hoxa2 can be presumed to have no influence on the inductive interactions between midbrain (Otx2-positive) and hindbrain (Gbx2-positive) territory that establish isthmic signalling [3, 4]. While FGF8 can repress Hoxa2 in the hindbrain (Fig. 6B; [24, 25]), the converse is not found at the isthmus.
Isthmic signalling plays a role in regulating the development of r1 [3, 6, 8, 24] where FGF8 is sufficient for the induction the trochlear nucleus  and necessary for locus coeruleus specification . These processes, which are overtly linked to isthmic activity, are refractory to experimental Hoxa2 up-regulation, suggesting that a complex of isthmic induction pathways in r1 is independent of a set of patterning processes that rely on segmental identity cues.
By contrast, various aspects of r1 specification and patterning are profoundly influenced by ectopic Hoxa2. In a previous study, RCASBP(B)Hoxa2 was shown to induce an r2-specific motor neuron phenotype within ventral r1 , caudal to the trochlear nucleus, in a region normally devoid of any motor efferents. Ventral progenitors were thus presumed to undergo a homeotic transformation to a more caudal fate (Fig. 7).
Within dorsal r1, ectopic Hoxa2 results in a smaller cerebellum, consistent with the loss of the EGL due to the aberrant migration of granule cells to a ventrolateral target. Both the loss of specific granule cell markers and the induction of a pattern of nuclear condensation corresponding to that of r2 derivatives suggest that rhombic lip precursors have been caudalised in a homeotic manner. Furthermore, a combination of grafting and electroporation reveals that this effect is independent of changes to r1 as a whole. This suggests that segmental/rhombomeric identity programmes rhombic lip derivatives with region specific responses to generalised dorsoventral migration cues such as netrin [28, 29] and slit .
This transformation of rhombic lip migration pattern is independent of the formation of a cell-free layer corresponding to the EGL (Fig. 2F,2G), in distinct contrast to mouse models of granule cell depletion, such as the MATH1 mutant, where the EGL fails to form . This indicates that the transient presence of Hoxa2-positive rhombic lip derivatives migrating over the cerebellar anlage is apparently sufficient to trigger the allocation of space for an EGL, even if this layer is subsequently unoccupied.
Our observations cast doubt on lineage analyses implicating the rhombic lip as the origin of locus coeruleus neurons . Despite dramatic alterations in migration and condensation patterns of rhombic lip derivatives, the locus coeruleus always forms correctly. Furthermore, Hoxa2 induced, ecoptic Phox2a-positive neurons, which seemed ideal candidates for the 'lost' EGL population, have an as yet indeterminate origin. Our results correspond with observations indicating that locus coeruleus neurons are born earlier than EGL precursors  from a Mash1-positive precursor pool , which lies ventral to the Math1-positive rhombic lip .
Finally, the fate of other neuronal types born within dorsal r1, including the remainder of the cerebellum , remains to be determined. Ectopic dorsal Phox2a is seen in globally infected embryos but not when Hoxa2 is targeted to the rhombic lip (Fig. 4E). This points to further transformations of cerebellar neurons when Hoxa2 is overexpressed throughout dorsal r1 that are not secondary to the loss of granule cells. However separating autonomous and non-autonomous effects of ectopic Hoxa2 in dorsal r1 will require an accurate fate-map of the origins of each cerebellar neuronal population. With the exception of granule cells, the dorsoventral origins of each cerebellar neuronal type have not yet been established.
While midbrain/hindbrain patterning has focused on the role of the isthmic organiser in regional patterning [5, 24, 25], Hoxa2 overexpression reveals a dependence on segmental identity for the specification of cerebellar granule cells. It thus seems likely that the establishment of a specific Otx2-, Hoxa2-negative region at st.6 (Fig. 1A) is an important step in the specification of the cerebellum. Consistent with this hypothesis, the loss of Hoxa2 expression in r2 leads to a caudal expansion of the cerebellum , presumably in the absence of any changes in isthmic signalling cues. These results suggest that cerebellar granule cell induction is independent of isthmic signalling per se. Our results agree with suggestions that the effects of experimental manipulation of FGF on cerebellar induction act through modulation of Otx2 [5, 7], rather than a direct induction. Thus, the induction of ectopic cerebellar neurons by exogenous FGF application can be attributed to the down-regulation Otx2 in the midbrain [6, 7] or Hoxa2 in the hindbrain , recapitulating the molecular characteristics of r1. Correspondingly, loss of FGF signalling in the zebrafish ace mutant leads to an expansion of Otx2 within r1 territory, suppressing granule cell induction .
The establishment of r1 is therefore necessary for cerebellar induction and, by molecular criteria, precedes that of any detectable isthmic organiser. Whether the absence of both Otx and Hox genes is sufficient to potentiate cerebellar development in the absence of isthmic signalling is unclear. Complete suppression of Hox gene function results in a hindbrain that adopts a 'default' r1 phenotype in zebrafish , but whether later dorsal development is 'cerebellar' has not been determined. As in ontogeny, the appearance of r1 precedes, phylogenetically, that of both the isthmic organiser  and the cerebellum . The evolution of a cerebellum would therefore seem to require a modification of either extrinsic signalling cues or responses within r1, in addition to the simple allocation of r1 territory.
Our results suggest that rhombomere 1 displays patterning influences that are characteristic of both the isthmic organiser and the rhombomeric organisation of the hindbrain. Using Hoxa2 overexpression to repattern regional identity of r1, rhombic lip derivatives that give rise to granule cell precursors can be forced to undergo a radical reprogramming of their migration and fate. By contrast, isthmic signalling molecules and both the territorial extent and the establishment of a subset of r1 nuclei are refractory to segmental cues and thus are presumed to be patterned independently by the midbrain/hindbrain organiser. This supports an argument that granule cells are a product of the molecular identity of r1 and induced in an isthmus-independent manner.
Chick eggs (Hi-Sex White) were windowed with scissors at either E0 (6 hours of incubation) or E2 (stage [st.] 8+ to 10-, ). The anterior neural tube was injected with the retrovirus encoding Hoxa2, RCASBP(B)Hoxa2 (108 to 109 virions/ml) To control for the nonspecific effects of viral infection, a different group of embryos were injected with the retrovirus, RCANBP(B)Hoxa2, which lacks a splice acceptor site immediately 5' of Hoxa2 and hence fails to produce functional Hoxa2 protein . Eggs were sealed with tape and reincubated at 37°C for 4 to 7 days.
Donor quail embryos were windowed at E2 and plasmid containing the proviral sequences of RCASBP(B)Hoxa2 (1 mg/ml in water containing 0.0015% fast green) was injected into the neural tube at the level of r1/2. Cells were then electroporated (1–2 pulses, 200 milliseconds/10 V, square waveform) using electrodes placed ventral and dorsal to the donor r1. These and a control group of unelectroporated donor embryos were harvested and the anterior neural tube isolated in Tyrode's solution (137 mM NaCl, 2.7 mM KCl, 1.8 mM CaCl2, 3.5 mM NaH2PO4, 0.1 mM MgCl2, 5.5 mM D-glucose). The dorsal part of r1 was dissected away, after treatment with dispase (Roche, 1 mg/ml in L15, 10 minutes), using flame-sharpened tungsten wire . Tissue from both donor groups was orthotopically grafted into stage-matched host chick embryos, which were then sealed and reincubated for 4 or 5 days. In a complementary strategy, unelectroporated quail dorsal r1 was grafted into chick hosts that had been electroporated with RCASBP(B)Hoxa2. In a separate set of fate-mapping experiments, dorsal r2 from quail was grafted orthotopically into r2 or heterotopically into r1 of host chick embryos, which were reincubated for 2 to 5 days.
Virally infected, chimaeric embryos and controls were harvested at E6 in 4% paraformaldehyde in 0.01 M phosphate buffer (PFA) at 4°C. The cerebellum and hindbrain region were partially dissected and processed for in situ hybridisation as described elsewhere . Plasmid templates from a number of sources were used to generate digoxygenin- (dig-) and FITC-labelled antisense riboprobes: Hoxa2 , Phox2a (from P Goridis, INSERM, Marseilles), erbB4 , Pax6 , TH , en1 , Wnt1 (Dr Anthomy Graham, King's College London, UK), FGF8 , Gbx2  and Otx2 .
Quail–chick chimaeras were further stained using an antibody, Q¢PN (DSHB, Iowa University, Iowa, USA) to a quail-specific perinuclear marker and a peroxidase-conjugated secondary antibody. Embryonic brains were sectioned transversely at 50 μm on a vibratome, mounted on slides in glycerol/PBS (9:1) and photographed using an Axiocam mounted on a Zeiss Axiophot.
We thank Leigh Wilson and Jon Gilthorpe for their comments on the manuscript. The monoclonal antibody Q¢PN (B Carlson and J Carlson) was obtained from the Developmental Studies Hybridoma Bank maintained by the University of Iowa, Iowa City (USA), under contract N01-HD-7-3263 from the NICHD, Maryland, USA. This study was supported by the Medical Research Council (London, UK) and the Wellcome Trust (London, UK), of which RJTW was a Research Career Development Fellow.
This article is published under license to BioMed Central Ltd. This is an Open Access article: verbatim copying and redistribution of this article are permitted in all media for any purpose, provided this notice is preserved along with the article's original URL.