Neuronal subtype specification in the cerebellum and dorsal hindbrain


Author to whom all correspondence should be addressed.
Web site:


In the nervous system, there are hundreds to thousands of neuronal cell types that have morphologically, physiologically, and histochemically different characteristics and this diversity may enable us to elicit higher brain function. A better understanding of the molecular machinery by which neuron subtype specification occurs is thus one of the most important issues in brain science. The dorsal hindbrain, including the cerebellum, is a good model system to study this issue because a variety of types of neurons are produced from this region. Recently developed genetic lineage-tracing methods in addition to gene-transfer technologies have clarified a fate map of neurons produced from the dorsal hindbrain and accelerated our understanding of the molecular machinery of neuronal subtype specification in the nervous system.


The dorsal hindbrain, or alar plate of the hindbrain, represents a good model system to investigate the molecular machinery to specify neuronal subtypes, as many different types of neurons are generated from this region. In mammals and avians, the hindbrain consists of eight rhombomeres (r1–r8). Previous grafting studies as well as detailed precision anatomical studies have suggested that the alar plate of r1 produces all cerebellar neurons, the alar plate of the middle hindbrain (r2–r5) generates neurons in the cochlear nucleus, and the alar plate of the caudal hindbrain (r6–r8) produces precerebellar neurons, although some slight differences exist between rodents and birds (Altman & Bayer 1987; Tan & Le Douarin 1991; Cambronero & Puelles 2000; Cramer et al. 2000; Farago et al. 2006). Furthermore, recently developed genetic lineage-tracing methods in addition to gene-transfer technologies have clarified a fate map of neurons produced from the dorsal hindbrain and accelerated our understanding of the molecular machinery of neuronal subtype specification in the dorsal hindbrain.

Neuron subtype specification in the cerebellum

There are three major regions in the cerebellum: cortex, white matter and nuclei. The cerebellar cortex includes several types of glutamatergic excitatory and GABAergic inhibitory neurons. Glutamatergic neurons are comprised of granule cells and unipolar brush cells (UBCs), while the GABAergic population includes Purkinje, Golgi, Lugaro, stellate, basket, and candelabrum cells. Cerebellar nuclei (CN) comprise three major types of neurons: large glutamatergic projection neurons (CN-Glu neurons), mid-sized GABAergic inhibitory projection neurons (CN-GABA-ION neurons) and small GABAergic interneurons (CN-GABA interneurons). CN-GABA-ION neurons extend their axons to the inferior olivary nucleus (ION) (Carletti & Rossi 2008), while CN-Glu neurons send their axons to nuclei outside the cerebellum, including the red nucleus and the thalamus. These neurons mutually regulate each other’s activity to achieve proper cerebellar function.

During development, the neuroepithelium of the alar plate of rhombomere 1 (r1) generates all types of cerebellar neurons (Millet et al. 1996; Wingate & Hatten 1999; Chizhikov & Millen 2003; Zervas et al. 2004). The dorsal-most part of the neuroepithelium, the roof plate, of r1 does not generate neurons but produces cells of the choroid plexus (Chizhikov et al. 2006). Cerebellar neuron-producing neuroepithelium can be divided into two regions; the rhombic lip (RL) and the ventricular zone (VZ). These two regions can be morphologically discriminated by a notch located on their border.

Although studies on the cerebellum have a very long history (Cajal 1909), the molecular machinery through which cerebellar neuron subtype specification occurs is still unclear. In 1997, Ben-Arie et al. reported that a basic-helix-loop-helix type (bHLH) transcription factor, Atoh1 (also called Math1), is expressed in the rhombic lip and is involved in cerebellar granule cell generation (Ben-Arie et al. 1997). At that time, it was believed that only granule cells are generated from the rhombic lip, while the remaining neurons are produced from the cerebellar VZ. Although several studies had clarified the specification machinery for granule cells, how other types of cerebellar neurons developed remained elusive until three breakthrough papers were published in 2005.

During the generation of some transgenic mice, our group obtained a mutant line “cerebelless” in which homozygotes exhibited uncoordinated locomotion, ataxic gait and tremors. Surprisingly, adult mice were found to lack the entire cerebellar cortex. In this mutant, no GABAergic inhibitory neurons are produced from the cerebellar primordium. In contrast, glutamatergic populations are initially generated, but glutamatergic granule cells are lost secondarily at postnatal stages. Eventually, the entire cerebellar cortex is lost in this mutant (Hoshino et al. 2005). The responsible gene was identified as pancreatic transcription factor 1a (Ptf1a), a bHLH transcription factor known to participate in pancreatic development. This gene is expressed in the neuroepithelium of the VZ but not in the RL and its expression is lost in the cerebelless mutants. Cre-loxP recombination-based lineage tracing analysis revealed that most types of cerebellar GABAergic neurons (Purkinje, Golgi, basket, stellate cells and GABAergic neurons in the CN) are derived from Ptf1a-expressing neuroepithelial cells in the VZ, but glutamatergic neurons, such as granule cells and CN-Glu neurons, are not. Loss of Ptf1a expression in cerebelless as well as in Ptf1a-knock out mice resulted in inhibition of the production of GABAergic neurons in the cerebellar primordium. Furthermore, ectopic introduction of Ptf1a by means of in utero electroporation resulted in the abnormal production of neurons with GABAergic characteristics from the dorsal telencephalon: the dorsal telencephalon only produces glutamatergic neurons under normal conditions. In addition, Pascual et al. (2007) reported that in the Ptf1a-null mutants, the fate of neurons produced from the VZ is changed to that of granule cells. Moreover, a recent genetic fate mapping study using Ascl1CreER knock-in mice showed that other cerebellar GABAergic neurons, such as Lugaro and candelabrum cells, are also derived from the cerebellar VZ (Sudarov et al. 2011). These observations suggested that Ptf1a, expressed in the cerebellar VZ, determines GABAergic neuronal fate in the cerebellum. PTF1A was also identified as a causative gene for a human disease that exhibits permanent neonatal diabetes mellitus and cerebellar agenesis (Sellick et al. 2004).

Just after our report, Zoghbi’s and Fishell’s groups reported a molecular fate map of the derivatives of Atoh1-expressing neuroepithelial cells in the cerebellar RL (Machold & Fishell 2005; Wang et al. 2005). They showed that not only granule cells but also, at least some DCN neurons are derived from the RL, although they did not discriminate between neuron types in the CN. In their studies, development of RL-derived CN neurons was shown to be disrupted in the Atoh1 mutants. As GABAergic but not glutamatergic CN neurons were found to be derived from Ptf1a-expressing neuroepithelial cells in the VZ (Hoshino et al. 2005), this suggests that cerebellar glutamatergic neurons such as granule cells and CN-Glu neurons are derived from the RL. Accordingly, unipolar brush cells, which are glutamatergic, were also shown to emerge from the RL (Englund et al. 2006).

Together, these studies indicate the presence of two molecularly defined neuroepithelial areas in the cerebellum, the Atoh1-expressing RL and the Ptf1a-expressing VZ, which generate glutamatergic and GABAergic neurons, respectively. Each bHLH transcription factor is involved in producing the corresponding neuronal subtype in the cerebellum. These facts suggest a model in which regionalization of the cerebellar neuroepithelium into two distinct regions, the VZ and the RL, is mediated by the two bHLH transcription factors, Atoh1 and Ptf1a (Hoshino 2006). During embryonic development, the ventral cerebellar neuroepithelium expresses Ptf1a, leading to cerebellar VZ characteristics and the ability to generate GABAergic neurons, while the dorsal cerebellar neuroepithelium expresses Atoh1 and becomes the cerebellar RL, producing glutamatergic neurons. In the telencephalon, similar regionalization by bHLH transcription factors takes place. Glutamatergic neurons emerge from dorsal neuroepithelium expressing Neurogenin 1/2 (Ngn 1/2) and GABAergic neurons are produced from ventral neuroepithelium expressing Ascl1 (also called Mash1) (Wilson & Rubenstein 2000).

How are these neuroepithelial areas formed? In general, the roof plate can affect the dorsal structure of the neural tube (Lee et al. 2000; Millonig et al. 2000). Chizhikov et al. (2006) revealed that the roof plate plays an important role in the formation of the cerebellar dorso-ventral domain formation by analyzing cerebellar mutants that lack the roof plate. Moreover, it has been suggested that bone morphogenetic proteins (BMPs) secreted from the roof plate as well as Notch signaling are involved in the formation of the RL and the VZ (Machold et al. 2007). An in vitro study that induced Purkinje cells from ES cells suggested that loss of sonic hedgehog signaling may provide dorso-ventral spatial information of the cerebellar VZ to the cerebellar neuroepithelium, which eventually leads to the expression of Ptf1a (Muguruma et al. 2010).

Although some clarification of the machinery governing GABAergic and glutamatergic neuronal subtype specification by transcription factors has been provided, molecular mechanisms to specify each GABAergic (e.g. Purkinje, Golgi, basket, stellate cells and CN-ION, CN-interneurons) or glutamatergic (e.g. granule, unipolar brush cells and CN-Glu neurons) subtype remain unclear. Birthdating studies using 3H-thymidine and BrdU (Chan-Palay et al. 1977; Batini et al. 1992; De Zeeuw & Berrebi 1995; Sultan et al. 2003; Leto et al. 2006) as well as adenovirus (Hashimoto & Mikoshiba 2003) have revealed that each type of neuron is generated at distinct developmental stages.

With regard to GABAergic neurons, Purkinje cells are produced early (embryonic day (E) 10.5–13.5 in mice), Golgi cells a little later (E13.5–postonatal day (P) 0) and stellate/basket cells mainly perinatally. The newest study by Sudarov et al. revealed that candelabrum cells are generated around P0, while GABAergic CN neurons arise at early stages (E10.5–11.5). As to glutamatergic neurons, in addition to the above studies, molecule-based lineage tracing analyses (Machold & Fishell 2005; Wang et al. 2005; Englund et al. 2006) have clarified that CN-Glu neurons leave the cerebellar RL at early stages (E10.5–12.5) and granule cells and unipolar brush cells at middle to late stages (granule cell; approximately E12.5∼perinatal, ubc: approximately E12.5-E18.5). In addition, somatic recombination-based clonal analyses suggested that Purkinje, Golgi and basket/stellate cells as well as some CN neurons (probably GABAergic) belong to the same lineage (Mathis et al. 1997; Mathis & Nicolas 2003). These data indicate that some temporal information in the neuroepithelium may be involved in specification of neuronal types in the RL and VZ, respectively. However, the underlying molecular mechanisms have not yet been clarified.

Some scientists have attempted to divide the structure of the cerebellar primordium into several domains (Fig. 1). Chizhikov et al. (2006) defined four cellular populations (denoted c1–c4 domains) in the cerebellar primordium via the expression of a few transcription factors. c1 corresponds to the Atoh1-expressing RL and c2 is located just above the Ptf1a-expressing VZ (denoted pc2), indicating that c2 cells mainly consist of GABAergic inhibitory neurons. Although c3 and c4 express Lmx1a and Lhx1/5 respectively, their neuronal subtypes are still unknown. This domain structure is disrupted when the roof plate was removed (Chizhikov et al. 2006). Furthermore, at the early neurogenesis stage (e.g. E12.5 in mice), Minaki et al. (2008) subdivided the c2 domain into dorsally (c2d) and ventrally (c2v) located subdomains that express corl2 and Pax2, respectively. While corl2 is exclusively expressed in immature and mature Purkinje cells (Minaki et al. 2008), Pax2 is expressed in GABAergic interneurons (e.g. Golgi, stellate, basket, CN-GABA neurons) in the cerebellum (Maricich & Herrup 1999; Weisheit et al. 2006). They also subdivided the Ptf1a-expressing neuroepithelial domain (pc2) into pc2d and pc2v, which strongly and weakly express E-cadherin, respectively. From the positions of the neuroepithelial and neuronal subdomains, they suggested that the pc2d neuroepithelial subdomain produces cells in the c2d domain, which give rise to Purkinje cells, while the pc2v subdomain generates cells in the c2v that become GABAergic interneurons (Mizuhara et al. 2010). As development proceeds, pc2d and pc2v subdomains contract and expand, respectively, and by E14.5 in mice, the Ptf1a-expressing pc2 domain comprises only the pc2v subdomain, which expresses E-cadherin weakly. This correlates with the fact that, at E14.5, Ptf1a-expressing neuroepithelium does not produce Purkinje cells but Pax2-positive interneurons (Maricich & Herrup 1999; Hashimoto & Mikoshiba 2003). The expression of several other transcription factors in the cerebellar VZ during development have also been reported. For example, Zordan et al. (2008) described the expression patterns of proneural bHLH transcription factors, such as Ngn1, Ngn2 and Ascl1 in the cerebellar VZ although their function in cerebellar development is still unclear. It has also been reported that Pax2-positive neurons, but not Purkinje cells, are reduced in the Ascl1-null cerebellum (Grimaldi et al. 2009), while Purkinje cells are reduced in Ngn1-null mice (Lundell et al. 2009), suggesting that these bHLH transcription factors play distinct roles in cerebellar development.

Figure 1.

 Domain structure of the cerebellar primordium. The c1 domain, expressing Atoh1, corresponds to the rhombic lip that produces all types of glutamatergic neurons in the cerebellum. The pc2 is the Ptf1a-expressing neuroepithelial domain that generates all types of GABAergic cerebellar neurons. At early neurogenesis stages, such as E12.5, the pc2 domain can be subdivided into pc2d and pc2v subdomains, which expresses E-cadherin strongly and weakly, respectively. The c2 domain, expressing Lhx1/5, consists of immature GABAergic neurons putatively generated from pc2 neuroepithelial domain. This domain can also be subdivided into two subdomains, c2d and c2v, corresponding to pc2d and pc2v, respectively. The c2d subdomain consists of corl2-expressing neurons or Puarkinje cells, whereas the c2v subdomain includes Pax2-positive cerebellar GABAergic interneurons. Although c3 and c4 domains are Lmx1a- and Lhx1/5-positive, respectively, cell types which make up these domains are unknown. The roof plate (RP) is located most dorsally, and plays prominent roles in organizing this cerebellar domain structure.

In addition, several transcription factors have been reported to participate in the development of specific types of cerebellar neurons. Double knockout of Lhx1 and Lhx5 as well as the targeted disruption of their cofactor Ldb1 resulted in a lack of Purkinje cell production in the cerebellum although Pax2-positive interneurons did not seem to be affected. Because Lhx1 and Lhx5 are expressed in post-mitotic cells, this suggests that Lhx1, Lhx5 and Ldb1 are post-mitotically involved in Purkinje cell specification (Zhao et al. 2007). In addition, in the cyclin D2 KO mice, the progenitor pool of GABAergic interneurons is precociously exhausted and progenitor numbers are significantly reduced, leading to a remarkable decrease in the number of late-born interneurons, such as stellate cells (Huard et al. 1999; Leto et al. 2011).

From the RL, several types of glutamatergic neurons, such as CN-Glu neurons, granule cells and unipolar brush cells, are generated (Machold & Fishell 2005; Wang et al. 2005; Englund et al. 2006). CN-Glu neurons leave the RL early during neurogenesis. Some transcription factors, such as Tbr1, Irx3, Meis2, Lhx2 and Lhx9 are expressed in post-mitotic progenitors of CN-Glu neurons, but their roles have not been clarified (Morales & Hatten 2006). Other molecules, such as Zic1 (Aruga et al. 1998), have been reported to play important roles in the migration, maturation and survival of granule cells, but the molecular machinery underlying the specification of granule cell identity is unknown. Although unipolar brush cells strongly express Tbr2, its function remains elusive.

Heterotopic and heterochronic transplantation studies have also provided important clues to understanding cerebellar development (Carletti & Rossi 2008). When tissues from embryonic and postnatal cerebella were mixed and transplanted to the fourth ventricle of an adult mouse, the postnatal-derived cells differentiated only into interneurons such as granule, basket and stellate cells, but not projection neurons, such as Purkinje cells, whereas the embryonic-derived cells were capable of becoming all types of cerebellar neurons (Jankovski et al. 1996). It has also been shown that dissociated cells taken from cerebellar primordium at early neurogenesis stages can differentiate into all major types of cerebellar neurons, while those from postnatal cerebellum differentiated only to Pax2-positive interneurons (Carletti et al. 2002). These findings suggest that the differentiation competence of cerebellar progenitors becomes restricted as development proceeds. However, the molecular mechanisms underlying this fate restriction process have not yet been clarified. Interestingly, Leto et al. (2006) suggested that Pax2-positive interneurons, such as Golgi, stellate, basket cells and CN-GABA interneurons are derived from the same progenitor pool. Leto et al. (2009) also clarified that, after leaving the VZ, progenitors for GABAergic interneurons continue to proliferate in the prospective white matter during late embryonic and postnatal development. Their grafting studies showed that terminal commitment does not occur while precursors are still proliferating but occur postmitotically according to host-specific information, suggesting an instructive cue provided by the microenvironment of the prospective white matter.

Neuron subtype specification in the cochlear nucleus

Sounds received in the ear are transmitted via the auditory nerve to the cochlear nucleus (CoN) of the mammalian hindbrain, where the auditory information is properly processed and relayed to the brain. The CoN is a very complex cell assembly that can be divided into two subregions, the ventral and dorsal cochlear nuclei (VCoN and DCoN), which differ in structure and feature.

Because of its importance in sound perception, the CoN has been intensely studied from anatomical, physiological and histochemical points of view (Osen 1969; Ryugo & Willard 1985; Hackney et al. 1990). Histological observations have deduced that a portion of neurons generated from the dorsal hindbrain neuroepithelia migrate tangentially to give rise to CoN neurons (Pierce 1967; Ivanova & Yuasa 1998). More directly, genetic fate mapping studies using transgenic mice confirmed that many CoN cells are derived from the dorsal region of the hindbrain neuroepithelia where the Wnt1 promoter is active (Farago et al. 2006; Nichols & Bruce 2006). As to the rostro-caudal axis, the origins of CoN neurons seem to differ between avians and mammals. Grafting studies revealed that avian CoN neurons are derived from a broader part of the hindbrain (r3–r8), (Tan & Le Douarin 1991; Cambronero & Puelles 2000; Cramer et al. 2000), while mouse genetic studies have suggested a more rostral and narrower origin (r2–r5) (Farago et al. 2006).

Very sophisticated genetic fate mapping studies were carried out by Farago et al. (2006) using an FLP-FRT and Cre-loxP-based dual lineage tracing system. In addition to showing that CoN neurons are derived from r2 to r5, they also revealed that neurons in the anterior part of the VCoN (aVCoN), the posterior part of the VCoN (pVCoN) and the DCoN, generally tend to be generated from rostral (∼r2, 3), middle (∼r3, 4), and caudal (∼r4, 5) parts of the CoN neuron-producing hindbrain (r2–5), respectively, with some overlap.

The CoN contains a variety of neurons that have distinct features (Osen 1969; Ryugo & Willard 1985; Hackney et al. 1990). For example, the DCoN includes Golgi, molecular layer (ML)-stellate, cartwheel, tuberculo-ventral, unipolar brush, giant, and fusiform cells, while the VCoN is comprised of octopus, globular-bushy, spherical-bushy, T-stellate and D-stellate cells.

In the neuroepithelium of the middle hindbrain (r2–r5), distinct transcription factors are expressed constituting several molecularly-defined domains (Fig. 2, upper panel). Using Cre-LoxP-based genetic fate mapping studies, our group identified the origins of inhibitory and excitatory neurons of the cochlear nucleus; inhibitory (glycinergic and GABAergic) and excitatory (glutamatergic) neurons are derived from Ptf1a- and Atoh1-expressing neuroepithelial regions, respectively (Fujiyama et al. 2009) and their development is dependent on the corresponding bHLH proteins.

Figure 2.

 In the middle (r2–5, upper panel) and caudal (r6–8, lower panel) hindbrain, several transcription factors are expressed within the dorsal neuroepithelium during embryonic development. The dorsal-most part, the roof plate (RP), expresses Lmx1a. Other than the roof plate, the dorsal neuroepithelium can be divided into six domains (dP1–dP6) according to the expression pattern of indicated transcription factors. (upper panel) As to the neurons in the cochlear nucleus, glutamatergic excitatory neurons and GABAergic/glycinergic inhibitory neurons are derived from dP1 and dP3/dP4, respectively. (lower panel) While mossy fiber (MF) neurons are derived from the dP1 domain expressing Atoh1, climbing fiber (CF) neurons are generated from the dP4 domain expressing Ptf1a and Olig3. Ngn1 is expressed in r7 and r8 but not in r6.

Subtype specification of precerebellar neurons

There are two types of precerebellar afferent systems; mossy fiber (MF) and climbing fiber (CF) systems. MF neurons are located in several nuclei throughout the brain stem and extend their glutamatergic projections to granule cells conveying peripheral and cortical information to the cerebellum. Four major nuclei containing MF neurons are the pontine gray nucleus (PGN), the reticulotegmental nucleus (RTN), the lateral reticular nucleus (LRN) and the external cuneate nucleus (ECN) in the hindbrain (Altman & Bayer 1987). Some MF neurons are also located in the spinal trigeminal nucleus (Sp5) in the hindbrain and Clarke’s column in the spinal cord. In contrast, CF neurons reside exclusively in the inferior olive nucleus (ION), which receives input from the cerebral cortex, the red nucleus, spinal cord and other brain stem nuclei, and sends glutamatergic projections to Purkinje cells (Ruigrok et al. 1995). Both types of precerebellar neurons also send branch axons to the neurons in the cerebellar nucleus. These precerebellar systems are thought to transmit both external and internal information to the cerebellar cortex to modulate cerebellar function, including regulation of movement.

Previous birthdating studies in mice revealed that CF neurons are generated at relatively early neurogenesis stages (E9.5–11.5) while MF neurons are produced at slightly later stages (E10.5–16.5) (Pierce 1973). Along the rostrocaudal axis, both MF and CF neurons in the hindbrain are generated from the caudal hindbrain, around rhombomeres 6–8, as suggested by avian grafting studies as well as mammalian fate map analyses (Ambrosiani et al. 1996; Cambronero & Puelles 2000; Farago et al. 2006; Kawauchi et al. 2006). By contrast, MF neurons in the Clarke’s nucleus are generated in the spinal cord (Bermingham et al. 2001). Classic anatomical and immunohistochemical studies have suggested that these precerebellar nuclei neurons in the hindbrain emerge from the dorsal part of the hindbrain and migrate tangentially or circumferentially to their final loci (Bloch-Gallego et al. 1999; Yee et al. 1999; Kyriakopoulou et al. 2002). However, they take slightly different paths from each other; MF and CF neurons move extramurally and intramurally, respectively. Introduction of a GFP-expressing vector into the embryonic dorsal hindbrain allowed the dramatic visualization of migrating precerebellar nuclei neurons during development (Kawauchi et al. 2006; Okada et al. 2007).

Many groups have reported transcription factors that are expressed within the dorsal neuroepithelium of the caudal (r6–8) hindbrain during embryonic development, trying to define domains along the dorsoventral axis. The dorsal-most part expressing Lmx1a corresponds to the roof plate, which gives rise to the choroid plexus (Chizhikov et al. 2006). Other than the roof plate, the dorsal neuroepithelium can be divided into six domains (dP1–dP6) according to the expression pattern of the transcription factors, such as Atoh1, Ngn1, Ascl1, Ptf1a and Olig3 (Fig. 2, lower panel). As for the precerebellar nuclei neurons, a series of studies have tried to clarify the precise origins of MF and CF neurons by genetic lineage tracing methods.

By analyzing genetically engineered mice that express lacZ or Cre recombinase under the control of the endogeous or exogenous Atoh1 promoter, MF neurons of PGN, RTN, LRN and ECN were shown to emerge from the Atoh1-expressing neuroepithelial domain (dP1, Ben-Arie et al. 2000; Rodriguez & Dymecki 2000; Landsberg et al. 2005; Wang et al. 2005). Targeted disruption of the Atoh1 gene resulted in loss of these MF neurons, suggesting an involvement of Atoh1 in the MF neuron development.

Atoh1 regulates the expression of the transcription factor Barhl1 (Mbh2) that is expressed in MF neurons. Loss of Barhl1 expression resulted in a decrease of MF neurons, leading to a decrease in the size of MF precerebellar nuclei (Li et al. 2004). In addition, Flora et al. (2007) reported that Tcf4, an E-protein, interacts with Atoh1 and regulates differentiation of a specific subset (PGN, RTN) of MF neurons.

Landsberg et al. also performed lineage trace analysis by using two variants of FLP (Flippase recombinase) with different recombinase activities that were expressed under the control of the Wnt-1 promoter whose strength is the highest at the dorsal-most part and gradually decreases ventrally. They demonstrated that CF neurons are derived from the neuroepithelial region where Wnt-1 is very weakly expressed, whereas MF neurons emerge from the strong Wnt1-expressing region (Landsberg et al. 2005). In addition, Nichols & Bruce (2006) generated transgenic mice carrying a Wnt-1-enhancer/lacZ transgene and observed that MF neurons but not CF neurons were labeled by β-gal in those mice. These findings suggested that CF neurons are generated from the neuroepithelial region ventral to the Atoh1-expressing domain.

By Cre-loxP-based lineage trace analysis, our group showed that all CF neurons in the ION are derived from the Ptf1a-expressing neuroepithelial region (Yamada et al. 2007). Loss of the Ptf1a gene resulted in the fate change of some CF neurons to MF neurons, suggesting that Ptf1a plays a critical role in fate determination of CF neurons. We also showed an involvement of Ptf1a in migration, differentiation and survival of CF neurons. It has been reported that both MF neurons and CF neurons are derived from the Olig3-expressing neuroepithelial region that broadly expands within the dorsal hindbrain (Storm et al. 2009) by Cre-loxP-based linage tracing. Targeted disruption of the Olig3 gene caused the disorganized development of MF neurons and complete loss of CF neurons (Liu et al. 2008; Storm et al. 2009). Moreover, the ectopic co-expression of Olig3 and Ptf1a induced cells expressing a CF neuron marker in chick embryos (Storm et al. 2009). These findings suggest that CF neurons emerge from the Ptf1a/Olig3-expressing neuroepithelial domain (dP4) and that Ptf1a and Olig3 are cooperatively involved in the development of CF neurons. The domain structure of the dorsal neuroepithelium in the caudal hindbrain region is shown in the lower panel of Figure 2.


Various types of neurons are generated from the dorsal hindbrain. As described above, the dorsal neuroepithelium of the rostral hindbrain (r1) produces all types of cerebellar neurons, while the dorsal regions of the caudal hindbrain (r6–r8) generate neurons that include the precerebellar system neurons, such as MF and CF neurons. In addition, the dorsal part of the middle hindbrain (r2–r5 in mice) produces neurons of the cochlear nucleus, where auditory information is processed and relayed to the brain.

There are dorso-ventral domain structures defined by several transcription factors, which are longitudinally expressed throughout the hindbrain. In particular, two bHLH transcription factors, Atoh1 and Ptf1a seem to play important roles in specifying distinct neuronal subtypes. These two proteins are expressed in different neuroepithelial regions throughout the hindbrain (Fig. 3). In both the rostral (r1) and middle hindbrain (r2–r5 in mice), Atoh1 and Ptf1a participate in generating excitatory and inhibitory neurons, respectively. However, this rule is not applicable to the caudal hindbrain. The Ptf1a neuroepithelial domain in the caudal hindbrain (r6–r8) produces not only inhibitory neurons (local circuit neurons) but also glutamatergic neurons (CF neurons) (Yamada et al. 2007), while the Atoh1 domain generates glutamatergic MF neurons. This raises the possibility that the rostral/middle (r1–r5) and caudal (r6–r8) hindbrain subregions have distinct characteristics. Overall, throughout the hindbrain regions, transcription factors, such as Atoh1 and Ptf1a, seem to define neuroepithelial domains along the dorsoventral axis and participate in specifying distinct neuronal subtypes according to the rostrocaudal spatial information (Fig. 3).

Figure 3.

 Neurons produced from the dorsal hindbrain. Atoh1 and Ptf1a are expressed in distinct neuroepithelial regions throughout the rhombomeres 1–8 (r1–8). Each number represents the rhombomeric number. Upper side is dorsal, lower is ventral. Left side is rostral, right side is caudal. Neuronal subtypes generated from the dorsal neuroepithelium of the rostral, middle and caudal hindbrain regions are shown.

In the spinal cord, Atoh1 and Ptf1a are also expressed in the dorsal neuroepithelium in a non-overlapping manner, defining specific neuroepithelial domains, although Ptf1a seems to be transiently expressed in immature postmitotic neurons. Ptf1a is involved in producing inhibitory neurons in the dorsal spinal cord (Glasgow et al. 2005). Atoh1 is known to participate in generating some types of neurons in the dorsal spinal cord, including commissural neurons (Helms & Johnson 1998; Bermingham et al. 2001; Miesegaes et al. 2009). These neurons in the Atoh1-lineage are believed to include glutamatergic populations, although the neurotransmitter subtypes have not been well studied. Thus, from the viewpoint of transcription factors and neurotransmitter subtypes, the dorsal spinal cord has similar characteristics of those of rostral (r1) and middle (r2–r5) hindbrain but not caudal hindbrain (r6–r8), which emphasizes the complexity of the hindbrain.


I thank Mayumi Yamada, Yusuke Seto, Kei Hori and Tomoyuki Fujiyama for discussion and figure construction and Ruth Yu for reading the manuscript.