Keywords

PCs as Project Managers of Cerebellar Cytoarchitecture and Connectivity

The cerebellum contains a limited number of cellular phenotypes , arranged in a highly conserved circuitry and identified by their morphological features, their reciprocal relationships, and the expression of distinctive neurochemical markers. The mouse is the main model system in which cerebellar ontogenesis has been studied extensively. Although the mammalian cerebellum is superficially homogeneous, it actually consists of several hundred distinct compartments, which form a complex, reproducible array of transverse zones and parasagittal stripes. Cerebellar architecture is built around multiple Purkinje cell subtypes [1,2,3,4,5,6] – most notably zebrin II/aldolase C [7, 8] – which form the transverse zone-and-parasagittal stripe scaffold upon which the adult cerebellum is built. For example, zone-and-stripe boundaries restrict the terminal fields of many cerebellar afferent projections (reviewed in [9]), interneuron neurites [10] and somata (reviewed in [11]), and glial gene expression profiles (e.g., 5′-nucleotidase [12]).

In the mouse, the general timeline of events that leads to cerebellar maturation from its embryonic anlage has been fully clarified [13,14,15,16,17,18]. Here we discuss some of the major features of cerebellar development, focusing on the ontogenesis of Purkinje cells (PCs), the sole projection neurons of the cerebellar cortex.

PC development is only partially characterized, despite the remarkable progress made in recent years (reviewed in [19]). Achieving a better understanding of PC cell fate specification and ontogenesis in general is important for a number of reasons. First, PCs orchestrate the early stages of cerebellar development, namely those that precede the massive proliferation of granule cell precursors in the external granular layer. Only later in embryogenesis, and especially after birth, do granule cells take control of cerebellar histogenesis and foliation, as they outnumber all other cerebellar cell types by several orders of magnitude.

Secondly, PCs actually control granule cell clonal expansion by releasing the extracellular morphogen/mitogen sonic hedgehog [20,21,22,23], with the result that the overall PC number heavily influences the final dimensions and organization of the cerebellum – and ultimately its function. The corollary is that defective PC genesis or migration impairs granule cell clonal expansion, and cerebellar foliation/PC migration failures result in a lissiform adult cerebellar cortex: e.g., the naturally occurring mouse mutant reeler (Relnrl: reviewed in [24]).

Thirdly, PCs guide the wiring of the cerebellum. Most afferent fiber systems invade the cerebellum at around embryonic day 13/14 (E13/14) in the mouse [25, 26] and terminate with a spatial organization that parallels the pattern of PC stripes [27]. PCs instruct afferent fibers, including olivocerebellar axons, which eventually establish a one-to-one contact with their target, as well as mossy fibers, which connect transiently with PCs and use PC-produced guidance cues prior to retracting and shaping their definitive synapses on granule cell dendrites (reviewed in [16, 28]). PC subtype organization is thought to play a key role in instructing circuit wiring into topographic maps: zone-and-stripe boundaries typically restrict the terminal fields of both cerebellar mossy fiber and climbing fiber afferent projections (reviewed in [9]), and interneuron neurites ([10], reviewed in [11]) and spontaneous and engineered mouse mutants with disrupted PC stripes have complementary alterations in the spatial arrangement of afferent terminals [29,30,31].

Cerebellar Anlagen and Germinal Zones

The cerebellum arises from a specialized region at the midbrain/hindbrain boundary [32,33,34]. In the mouse, at E8.5, the antagonistic interaction that takes place between homeobox genes Otx2 and Gbx2 defines the isthmic organizer region [35, 36], which controls the development of cerebellar structures via the secreted morphogens FGF8 and WNT1 [16, 37, 38]. At this stage, the cerebellar primordium consists of two distinct and symmetric bulges thought to grow and fuse on the midline, eventually giving rise to the vermis, flanked by the two hemispheres [15]. Importantly, however, homotopic and isochronic quail-chick grafting experiments have clearly shown that the caudal part of the early midbrain vesicle generates the rostral and medial part of the prospective cerebellum [32, 39,40,41,42]. Thus, the anterior part of the prospective cerebellar vermis, instead of resulting from fusion of lateral cerebellar plates (His, 1889), likely originates from the caudal and alar portion of the mesencephalic vesicle [39].

Once a low-resolution map has been drawn, cerebellar histogenesis begins, starting at E9. Around E9.5, two germinal neuroepithelia emerge in the cerebellar primordium, abutting the opening of the fourth ventricle: (1) the rhombic lip (RL), located at the outer aspect of the cerebellar plate, adjacent to the roof plate (RP, dorsal), and (2) the ventricular zone (VZ), lining the lumen of the fourth ventricle (ventral). These stem cell/progenitor compartments may be identified by the region-specific expression of two genes encoding basic helix-loop-helix transcription factors: pancreas transcription factor 1a (Ptf1a) in the VZ [43], and atonal homolog 1 (Atoh1) in the RL [44]. Cerebellar radial glial progenitors [45] expressing Ptf1a are fated to generate all GABAergic neurons of the cerebellum, including PCs and all inhibitory interneurons – cerebellar nuclear interneurons plus basket, stellate, Golgi, and Lugaro cells in the cerebellar cortex [43, 46, 47]. Homozygous mutations of PTF1A are associated with cerebellar agenesis in humans [48]. Conversely, all glutamatergic lineages – the large projection neurons of the cerebellar nuclei, unipolar brush cells, and granule cells – derive from Atoh1 + progenitors [49,50,51,52,53,54]: their development is exhaustively reviewed elsewhere [19].

Important genetic networks involved in the maintenance of the stem cell/progenitor pool and in cell fate specification are active in the VZ and/or RL between E10 and E13. The stem cell marker SOX2 is expressed in both neurogenic territories (VZ and RL ) and in the RP [55]. Its homolog SOX9 is largely co-expressed with SOX2 and may mediate termination of neurogenesis, thereby regulating a neurogenic-to-gliogenic fate switch in the mouse cerebellar primordium [55]. The target of Notch signaling, Hes5, is expressed in the VZ and RL, with a very sharp boundary and no expression in the RP. However, Hes1 expression levels are low to absent in the VZ and RL but present in the RP [56, 57]. Notch1 in the cerebellar primordium interferes with BMP2/BMP4 signal transduction causing downregulation of the BMP target Msx2.

As shown by birthdating studies, cerebellar projection neurons , (PCs in the cerebellar cortex and glutamatergic neurons in the cerebellar nuclei), are born first, at the outset of cerebellar neurogenesis, while both inhibitory and excitatory interneurons are generated perinatally [15, 58, 59]. Dividing VZ precursors delaminate into the cerebellar presumptive white matter, while those of the RL migrate below the pial surface where they form the rhombic lip migratory stream, initially containing nucleofugal neuron progenitors and, later, the granule cell precursors of the external granular layer. Postnatal neurogenesis continues in both regions through the third postnatal week, giving rise to GABAergic and glutamatergic interneurons, respectively [15, 17].

Establishment of Neurogenic Microdomains for GABAergic Progenitors

A schematic representation of microdomains present in the cerebellar VZ is provided in Fig. 1. All cerebellar GABAergic neurons originate in the VZ from Ptf1a + [43] and Ascl1 + [60] progenitors according to a two-step sequence [17, 19]. First, projection neurons (nucleo-olivary neurons and PCs) are generated from stem cells that give rise to fate-committed precursor populations. The nucleo-olivary neurons are generated between E10.5 and E12.5 in the mouse. Next, starting around E11 and through E13.5, mitotic PC progenitors exit the cell cycle and layer on top of the VZ to populate the nascent PC plate. The GABAergic interneurons are first born around E11 and sequentially generate all inhibitory local circuit neurons of the mature cerebellum.

Fig. 1
figure 1

A simplified representation of gene expression and cellular domains present in the E12.5 murine cerebellar primordium and giving rise to the mature cerebellar cortex. No reference is made here to cerebellar nuclei and their precurosrs. The drawing represents a sagittal section of the cerebellar anlage. The cerebellar primordium is bordered by the isthmic organizer (IO) rostrally and by the roof plate (RP) caudally. The choroid plexus (ChP), a roof plate derivative, is also shown. RP and ChP are non-neurogenic territories. The ventricular zone (VZ) is a mitotic cellular domain abutting the lumen of the fourth ventricle and giving rise to all GABAergic neurons of the cerebellar cortex. PTF1A is expressed by all GABAergic progenitors of the VZ, including PCs and interneurons. The PTF1A domain contains GSX1+ cells (mitotic interneuron progenitors) and OLIG2+ cells (mitotic PC progenitors). Both populations delaminate from the VZ (see text) giving rise to subventricular domains. The cortical transitory zone (CTZ) contains CORL2+ potmitotic PC precursors that subsequently migrate into the Purkinje cell plate (PCP), underneath the external granular layer (EGL). GSX1+ interneuron progenitors delaminate and give rise to PAX2+ interneuron precursors fated to populate the prospective white matter (not shown) before homing into the cortex. Both PAX2+ and CORL2+ domains are also positive for LHX1 and LHX5. Finally, all glutamatergic neurons of the cerebellum originate from the rhombic lip (RL, positive for the proneural gene Atoh1). Among them, granule cell precursors migrate tangentially beneath the pia mater and populate the prospective EGL

Full size image

The VZ is subdivided into mitotic progenitor domains abutting the ventricular lumen and corresponding postmitotic domains in the cerebellar primordium (an additional microdomain defines the rhombic lip) [61]. A microdomain positive for PTF1A contains two genetically defined progenitor cell types: OLIG2+ PC progenitors occupy a more caudal position and undergo their terminal mitosis between E11 and E13; GSX1+ progenitors are located more rostrally and medially. At E12.5, corresponding to the peak of PC neurogenesis, the c2 territory can be subdivided into a more caudal microdomain positive for CORL2, a selective marker of postmitotic PC precursors, and into a rostral/medial microdomain containing PAX2+ interneuron precursors. PC precursors, after leaving the cell cycle, start migrating and populate different regions of the cerebellar cortex according to their birthdate [15, 52]. Instead, actively proliferating interneuron progenitors (IPs), positive for GSX1, begin to delaminate from the VZ giving rise to PAX2+ interneuron progenitors and then migrate in successive waves to the nascent cerebellar nuclei or, with an inside-out progression, to the granular and molecular layers of the cerebellar cortex, where they acquire their definitive identities under the influence of instructive environmental cues ([62], reviewed in [63]).

The Regulation of PC Progenitor Specification and Commitment

At early stages (E11–12.5), a small number of GSX1+ interneuron precursors are found in the most rostral region of the VZ, while the majority of PC progenitors occupy more caudal regions of the VZ. Ablation of Olig2 has only a small effect [64] or no effect on PC number. However, a null mutation of both Olig2 and Olig1 produces a reduction of committed PC precursors [46]. As development proceeds, PC progenitors progressively become interneuron precursors, which spread from rostral (close to the isthmic organizer) to caudal, at the boundary between RL and RP. This temporal identity transition of cerebellar GABAergic neuron progenitors from PC progenitors to interneuron precursors is negatively regulated by OLIG2 and positively by GSX1 [46]. However, this view is challenged by the results of short- and long-term lineage tracing studies performed by other authors [64], suggesting that Olig2 + progenitors may not contribute importantly to the interneuron precursor lineage. Further analyses will be required to resolve this discrepancy: one possible scenario is that Gsx1 + progenitors affect the number of PC-committed Olig2 + precursors (or the maintenance of the PC-committed stem cell pool) through a paracrine, non-cell-autonomous mechanism.

The VZ subregion containing PC progenitors is also characterized by the strong expression of E-cadherin (encoded by Cdh1) and of the cell surface marker NEPH3, which is a direct downstream target gene of PTF1a [65]. When OLIG2+ PC progenitors exit the cell cycle, they activate the expression of Corl2 [66], which encodes a transcriptional repressor [67], and that of Lhx1 and Lhx5 [68], encoding LIM homeobox domain proteins. However, unlike CORL2, LHX1 and LHX5 label delaminating interneuron precursors as well as postmitotic PC precursors [46, 66]. Cells co-expressing LHX1/LHX5 [68] and CORL2 [66] are bona fide  differentiating VZ-born precursors committed to a PC fate.

Other PTF1A targets are expressed in the VZ in addition to those described above [69]. The Drosophila atonal homologs neurogenin 1 and neurogenin 2 are proneural genes encoding basic helix-loop-helix transcription factors. Neurog1 + progenitors give rise to inhibitory cortical interneurons and some PCs [70, 71], while Neurog2 is expressed mainly in the PC- and presumptive nucleo-olivary neuron lineages. NEUROG2 controls progenitor cell cycle progression, promotes cell cycle exit and differentiation, and spurs the cell-autonomous phase of PC precursor dendritogenesis. Nullisomy for Neurog2 causes a reduction in the overall PC number [72]. However, NEUROG1 and NEUROG2 are not required for the adoption of a PC fate (R. Hawkes, unpublished observation, and [72]). Interestingly, cell cycle analysis conducted by cumulative S-phase labeling on Neurog2 CreERT2 knockin mice has revealed for the first time that at the peak of PC neurogenesis (E12.5), dividing VZ progenitors cycle in ∼14 h, and their basal-to-apical oscillating motion is compatible with interkinetic nuclear migration, similar to what has been shown in other territories of the neural tube, but never before in the cerebellar primordium [72].

Ebf2 and PC Subtype Specification

Thus far we have treated PC development as though all PCs are the same. This is far from the case – indeed in the adult mouse cerebellum, multiple PC subtypes have been identified (e.g., zebrin II/aldolase C [7]; PLCβ3/4 [73]; HSP25 [74]: reviewed in [6]). The embryological origins of PC heterogeneity and pattern formation are only slowly coming into focus. PC subtype phenotype is cerebellum intrinsic and independent of neural activity (e.g., [75]) or afferent innervation [76, 77]. Cerebellar compartmentation appears to start at ~E10 in the VZ of the fourth ventricle but likely not sooner (e.g., [78,79,80,81]). The first stage likely occurs when PCs undergo terminal mitosis between E10 and E13 [58] in the Ptf1a-expressing progenitor domain of the VZ [43, 69]. Birthdating studies have identified two distinct PC populations: an early-born cohort (E10–E11.5) fated to become zebrin II+ and a late-born cohort (E11.5–E13) fated to become zebrin II− [82, 83]. However, individual PC stripes do not have a clonal origin [80]. There is also a direct correlation between PC birthdates and their adult stripe location, suggesting that both subtype specification and positional information (i.e., which zone or stripe the PC will occupy) may be acquired at this time (e.g., [82, 84,85,86]).

Several regulatory genes are implicated in PC progenitor development. Among them, Early B-cell factor 2 (Ebf2) [87] belongs to a family of atypical basic helix-loop-helix transcription factors that do not possess a basic domain and instead feature a unique DNA-binding domain. This family includes three transcriptional activators (EBF1–3) and one repressor (EBF4) (reviewed in [88, 89]). Ebf2 is expressed in a subset of late-born PC progenitors fated to populate zebrin II− parasagittal stripes, and in Ebf2 null mutants the cerebellum features a selective loss of zebrin II− PCs.

Upon cell cycle exit, late-born PC progenitors start expressing Ebf2 and migrate toward the PC plate. Posterior-born PCs migrate tangentially at first and then follow radial glial fibers, projecting their axons ventrally into the prospective white matter [90]. Conversely, anteriorly born PCs migrate radially into the PC plate, also following radial glial fibers, to populate anterior regions of the cerebellar cortex. Migration of this latter population is reelin (RELN) dependent and selectively delayed in Ebf2 null PCs, which accumulate before birth as an ectopic layer just above the VZ in the anterior third of the cerebellar anlage. A significant fraction of these PCs, many of which express neurogranin [91], dies by apoptosis [92, 93]. Ebf2 is required to support survival of late-born PCs at birth and accomplishes this by transactivating the insulin-like growth factor (Igf1) gene . In postnatal Ebf2 null cerebella, Igf1 expression is downregulated, with a resulting impairment of IGF-1 signal transduction [93]. Finally, some of the Ebf2 null PCs that survive lose their PC subtype specification features and trans-differentiate into zebrin II+ PCs – the only genetic manipulation thus far shown to subvert PC subtype specification [92]. In fact, Ebf2 acts to repress the zebrin II+ phenotype in late-born PCs [85]. Further studies, employing conditional mutants, are required to determine at which stage of postmitotic PC precursor development EBF2 acts to specify PC subtype. The results of genetic fate mapping experiments (GGC et al.: unpublished data) suggest that Ebf2 is expressed transiently in all PC progenitors, only to be restricted to late-born ones by the end of embryogenesis. The pathways that lead to further subtype specification (e.g., the HSP25+/− distinction within the zebrin II+ family: [74]) have not yet been explored.

Embryonic PC Cluster Formation

Newborn PCs migrate dorsally into the cerebellar anlage where they aggregate by ~E17 into a reproducible array of clusters that already contains multiple distinct molecular PC phenotypes ([6], reviewed in [83], e.g., Fig. 2 [94,95,96,97]). These clusters are the targets by which cerebellar afferents and many interneurons become topographically ordered (reviewed in [5, 28]).

Fig. 2
figure 2

From clusters to stripes. Embryonic clusters condense by migration from the cerebellar plate between E14 and E18 (mouse). At this stage numerous expression markers reveal that the PC population is already heterogeneous (exactly how many distinct phenotypes are present is not known, in part because of the paucity of double-labeling studies). The embryonic PC clusters also serve as a staging area to amass, organize, and restrict cerebellar afferents and interneurons. Starting perinatally signals via the RELN-Dab1 pathway trigger cluster dispersal into the adult cluster array by about P20. As for the embryonic clusters, the exact number of stripe phenotypes is not certain – at least ten may be identified based on expression data and mutant phenotypes. References to the lists of embryonic and adult cerebellar compartment markers may be found in [6]

Full size image

The mechanism that converts the PC plate into the elaborate array of embryonic PC clusters – >50 are recorded [97, 98] – is not well understood. As PCs migrate toward the cerebellar surface, the early-born (E10–E11.5: Ebf2 −; future zebrin II+) PC lamina interdigitates with the more superficial late-born (E11.5–E13; Ebf2 +; future zebrin II−) layer with the result that the stereotyped array of clusters emerges. Whether this migration is the mechanism that specifies cluster architecture or whether the clusters are already specified in the cerebellar plate (or are even preformed in the VZ, i.e., some version of the protomap model proposed by Rakic for the neocortex: reviewed in [99]) is not known. The cellular processes that guide cluster formation are not understood, but grafts of dissociated PCs also organize into discrete, ectopic zebrin II+/zebrin II− aggregates [100], pointing to cell-cell adhesion molecules as possible organizers: cadherins (reviewed in [101]) and integrins (e.g., [102]) are possible candidates. Also during this same period, the cerebellar anlage undergoes a 90° rotation, which converts the embryonic rostrocaudal axis into the mediolateral axis of the cerebellar primordium [81], so perhaps the adult stripe array ultimately derives from the anteroposterior patterning of dorsal rhombomere 1.

From Clusters to Zones and Stripes

Boundaries running from medial to lateral divide the cerebellar cortex into transverse zones . By combining different sources of evidence – molecular, genetic, and hodological – four highly conserved transverse boundaries, and hence five transverse zones, have been delineated in the adult mouse vermis (e.g., [103,104,105]): the anterior zone (AZ: ~ lobules I–V: reviewed in [104]), central zone anterior (CZa: ~lobule VI) and posterior (CZp), the posterior zone (PZ: ~lobules VIII to dorsal IX: reviewed in [105]), and the nodular zone (NZ). Each transverse zone is then further subdivided into a reproducible array of parasagittal stripes (e.g., revealed by using zebrin II/aldolase C [7, 8]: for zebrin II+/− stripes, these are labeled P+ and P−, e.g., zebrin II [106], phospholipase (PL) Cß3/Cß4 [73]), the small heat shock protein HSP25 [74], or L7/pcp2-lacZ transgene expression (reviewed in [6, 103]).

PC stripes are discontinuous across transverse boundaries so it seems plausible that the zones precede stripes in development, but whether transverse zones form prior to the PC clusters or at the same time is speculative. Transverse boundaries are certainly present in the embryonic cerebellum. The AZ/CZa boundary between lobules V and VI can be identified both in neonates and adults by the expression domains of numerous molecules (e.g., calbindin [103], reviewed in [107]) and is a developmental phenotype restriction boundary for several cerebellar mutations. (In some cases, the mutant phenotype is associated with defects in the AZ (e.g., [108]), lurcher Grid2Lc-J: [109], and cerebellar-deficient folia (Ctnna2cdf: [110]); in others – for example, the BETA2/NeuroD1 null [111] – it is the posterior cerebellar zones that are the most affected.) Finally, a granular layer lineage restriction boundary also lies in the anterior face of lobule VI, indicating that granule cells either side of the boundary derive from different lineages [112]. The CZa/CZp boundary [113] is a perinatal restriction boundary for FoxP2 [107], Gli [95], and HNK-1 expression [114]. The CZp/PZ boundary that separates lobule VII from lobule VIII is revealed in the perinatal cerebellum by FoxP2 [98, 107], PLCβ4 [115] and HSP25 [116] expression which is associated with a phenotypic abnormality in the lurcher (Grid Lc) mouse [109]. Finally, the most caudal transverse boundary in the adult mouse (PZ/NZ) lies near the base of the posterolateral fissure between lobules IX and X. A transverse boundary has also been located in the same area during development as a restriction boundary for the expression of En2 [95] and FoxP2 [96]. A granular layer transverse boundary in embryonic stem cell chimeras is also located at around the PZ/NZ boundary [112].

Starting at around E18, the embryonic clusters transform into adult stripes triggered by RELN signaling (reviewed in [117, 118]). Because PC dispersal and the associated development of cerebellar foliation occur almost entirely along the rostrocaudal axis, each cluster becomes stretched out into a long, narrow stripe.

RELN is secreted by both the external granular layer and glutamatergic projection neurons of the cerebellar nuclei [117] and binds to two PC receptors – the apolipoprotein E receptor 2 (Apoer2) and the very low-density lipoprotein receptor (VLDLR: [119, 120]). Both receptors are required for normal stripe formation, and if RELN is absent (e.g., the reeler mouse (Reln rl)), PC cluster dispersal is blocked, and the adult mouse retains the embryonic cluster morphology and is ataxic (reviewed in [24]). RELN binding induces Apoer2/Vldlr receptor clustering [121], which triggers a protein kinase cascade and tyrosine phosphorylation of the docking protein Disabled1 (DAB: [122,123,124,125,126]) by Fyn and Src [127, 128], leading eventually to a drop in mutual PC-PC adhesion, possibly via integrins. In parallel, DAB1 phosphorylation also activates Rac and Rho GTPases, which control actin filament assembly [129]. Together cytoskeletal and cell adhesion changes are thought to permit the embryonic PC clusters to disperse into stripes. That being said, it is not clear whether cluster dispersal requires the active migration of PCs or is the passive consequence of lobule formation.

However, the RELN pathway is not that straightforward. First, while expression mapping suggests that all PCs express both Apoer2 and Vldlr RELN receptors, mutations in individual receptors (Apoer2 −/− and Vldlr −/− nulls; Apoer2 +/−:Vldlr+/− double heterozygotes) result in specific partial reeler phenotypes with some clusters dispersing normally, while others remain ectopic ([130]; divergent roles are also seen in the developing cerebral cortex [131]). Similar behavior is seen in several naturally occurring mutants. For example, meander tail (mea2J [132]), rostral cerebellar malformation (Unc5crcm [108]), 1998), and cerebellar-deficient folia (Ctnna2cdf [110]) all display selective PC ectopias that are restricted to the zebrin II− phenotype (and because zebrin II− PCs are preferentially located in the AZ, it is the anterior vermis that is most severely affected). In a more complex model – the weaver (Kcnj6 wv) mouse – PC cluster dispersal failure is restricted to zebrin II+/HSP25+ stripes in the CZa/CZp [133]. The GIRK2 protein mutated in weaver [134] is expressed by all PCs so the molecular basis of the selective PC ectopias is unknown.

The relationship between the embryonic cluster topography and the zone and stripe pattern of the adult is not fully mapped. Because a few markers are expressed consistently in both clusters and stripes (FOXP2: [98], several are, for example, PLCß4 [115], an IP3R1 promoter-nls-lacZ transgene [135]), but others are only expressed in stripes at one stage or show very different expression patterns perinatally versus the adult, e.g., HSP25 (e.g., HSP25 [116], lysosomal acid phosphatase 2 [136]), there is limited evidence of the continuity of the cerebellar topographic map from perinate to adult. In theory, three relationships might occur: one embryonic cluster might form a single adult stripe; one cluster might split to yield more than one stripe; or several clusters might combine into a single stripe (Fig. 3). In fact, all three possibilities have been described. In several cases, the one cluster = one stripe model seems very likely (e.g., [83, 98, 135]). However, other examples are more complex. For example, the so-called P1− stripe in the AZ vermis clearly derives from three distinct embryonic clusters, which abut (as revealed by using PLCß4 expression [115]). An alternative – and perhaps better – description is that the apparently homogeneous P1− stripe in the adult (all zebrin II−/PLCß4+) actually comprises three distinct sub-stripes. The triplet structure is also seen in the afferent mossy fiber projections (where cuneocerebellar and spinocerebellar pathways terminate in different sub-stripes: [137, 138]) and in the expression of an L7/pcp2-lacZ transgene [103]. A similar covert heterogeneity is seen in ostensibly homogeneous zebrin II+ stripes when co-labeled for HSP25 [74]. Last, single clusters may give rise to multiple stripes. For example, inducible fate mapping with a Pcp2-CreER-IRES-hAP transgene showed three cluster pairs contribute to nine adult stripes [28].

Fig. 3
figure 3

Three models for the transformation of embryonic PC clusters into adult stripes: one embryonic cluster forms one adult stripe; one cluster splits to yield several stripes; or several clusters combine into a single stripe. All three models are found

Full size image

Finally, a striking feature of adult cerebellar topography is the high reproducibility between individuals and the concomitant low error rate (e.g., zebrin II+ PCs are very rarely seen in zebrin II− stripes). If stripes derive from clusters, and stripes have no errors, then either clusters have no errors (and migration from the VZ to the clusters is perfect) or errors that occur during cluster formation and dispersal are selectively eliminated. In this context it is interesting that many PCs – perhaps as many as a third – undergo cell death by apoptosis during the perinatal period [139]. This suggests the hypothesis that perinatal apoptosis eliminates those PCs that wind up in the wrong embryonic cluster. PC ectopia is not lethal per se: for example, PCs located ectopically may survive indefinitely. Rather, the hypothesis evokes a community effect, such that being in the wrong cluster leads to apoptosis. In support of the idea that apoptosis refines topography, studies of naturally occurring cell death in the cerebellum identified hot spots of PC apoptosis that correlate with stripe boundaries in the adult [140]. However, preliminary experiments do not support the hypothesis. Deleting the Bcl-2-/BH3-associated apoptotic protein BAX inhibits perinatal PC death (BAX is expressed in PCs perinatally [141] and Bax−/− mice have a 30% excess of PCs over controls (e.g., [142]). Nevertheless, the frequency of targeting errors was unaffected (RH and Y. Wang: unpublished data). Therefore, the remarkable reproducibility of the cerebellar map does not seem to result from perinatal error correction.

Conclusions

Early stages of PC development affect both susceptibility and outcome of several motor and cognitive disorders. Cerebellar development is protracted (from E7 to P30) and complex (at least two germinal zones, multiple migration pathways, etc.) so it is unsurprising that it represents a large target for developmental disruption. Spinocerebellar ataxia type 1 provides an example of this: transgenic mice in which expression of the expanded ATXN1 transgene is delayed until after the cerebellum has matured display a reduced disease phenotype, suggesting that mutant ATXN1 interacts with a pathway involved in PC development, likely by affecting RORa expression. Thus, compromising PC development appears to contribute to the severity of neurodegeneration [143]. Equally striking, recent evidence has linked PC development to the pathogenesis of autistic spectrum disorders (reviewed in [144]). In particular, selective deletion of the Tsc1 gene in the PC lineage from conditional knockout mice has been found to cause a decrease in PC number, increased spine density, and autistic-like alterations of social behavior [145]. One of many insults thought to trigger autism is maternal fever [146]. Possibly related to the putative role of the cerebellum in autism, we recently found that immune activation and fever in pregnant mice between E13 and E15 resulted in adult progeny with significantly wider zebrin II+/− stripes, greater numbers of PCs, poorer motor performance, and impaired social interactions in adolescence [147].

Finally, what are the prospects that early intervention might afford therapeutic advantages? While fast progress has been made in recent years, plenty remains to be learned in regard to the signals that instruct VZ progenitors to adopt PC versus GABAergic interneuron fate. To our knowledge, protocols aimed at producing PCs from ES/iPS cells in vitro are based on selection of early PC progenitors that express lineage-specific surface markers [148]. The identification of additional factors co-operating with PTF1a and OLIG2 in specifying the earliest PC progenitors should improve the efficiency of those protocols and make it possible to generate autologous PCs from iPS cells or via direct reprogramming. These short-range projection neurons produced in vitro may eventually constitute a source of cell replacement in patients affected by certain types of degenerative ataxias.