A key role for the HLH transcription factor EBF2^COE2,O/E-3 in Purkinje neuron migration and cerebellar cortical topography

Purkinje cells (PCs) are the focus of all cerebellar inputs and the sole output neuron of the cerebellar cortex. They are born in the cerebellar ventricular zone (VZ) between 11 and 13 days of mouse embryonic development(E11-E13) (Sotelo, 2004),shortly after the birth of deep cerebellar nuclei (DCN) neuron progenitors. After exiting the cell cycle, PC progenitors migrate along radial glial fibers away from the VZ and into an intermediate domain known as the cortical transitory zone (CTZ), while DCN neurons occupy a more dorsal domain termed the nuclear transitory zone (NTZ). From the CTZ, PC precursors migrate around NTZ neurons, forming a reproducible array of clusters in the subpial zone. From as early as E13.5, these clusters can be defined through differential expression patterns (reviewed by Armstrong and Hawkes, 2000; Herrup and Kuemerle, 1997). From E18 onwards, these clusters disperse rostrocaudally under the influence of reelin/dab1 signaling(Gallagher et al., 1998; Heckroth et al., 1989; Jensen et al., 2004; Jensen et al., 2002) to form the PC monolayer of the adult.

The cerebellar cortex is highly compartmentalized: it is first divided into four transverse zones, and then each zone is further subdivided rostrocaudally into parasagittal stripes (Ozol et al.,1999; Sillitoe and Hawkes,2002). Accordingly, the organization of afferent and efferent terminal fields is known to match closely the striped distribution of several molecular markers (Ji and Hawkes,1994; Ji and Hawkes,1995; Voogd et al.,2003; Voogd and Ruigrok,2004). The most extensively studied marker of parasagittal PC stripes is Zebrin II (ZII) (Brochu et al.,1990; Hawkes,1992), the respiratory enzyme aldolase c(Ahn et al., 1994; Hawkes and Herrup, 1995; Walther et al., 1998). Alternating stripes of ZII-positive and -negative PCs extend through the vermis and hemispheres, dividing the cerebellar cortex into a pattern of parasagittally oriented stripes, which is consistent between individuals and across species (Sillitoe et al.,2004).

Our findings implicate a transcription factor, Ebf2^COE2,O/E-3,in cerebellar cortical development. Collier/Olf/Ebf genes (reviewed by Dubois and Vincent, 2001; Garel et al., 1997; Hagman et al., 1993; Malgaretti et al., 1997; Wang and Reed, 1993; Wang et al., 1997), referred to here as Ebf genes, encode phylogenetically conserved HLH transcription factors originally characterized for their roles in the immune system(Hagman et al., 1993), and subsequently implicated in various aspects of neural development, including neuronal differentiation (Dubois et al.,1998; Pozzoli et al.,2001), migration(Garcia-Dominguez et al.,2003; Garel et al.,2000), and axon fasciculation and guidance(Garel et al., 1999; Garel et al., 2002; Prasad et al., 1998). One member of this family, Ebf2, is not essential for completion of embryogenesis, and its mutation leads to a combination of neuroendocrine,olfactory, and peripheral nerve abnormalities(Corradi et al., 2003; Wang et al., 2004). Outside the nervous system, Ebf2 regulates bone development and homeostasis non-cell-autonomously, by antagonizing the terminal differentiation of osteoclasts (Kieslinger et al.,2005). In the present study, we address the role of Ebf2in cerebellar cortex formation. Ebf2 regulates (1) the migration and survival of a specific PC precursor subpopulation during embryonic and postnatal development; (2) the establishment of a parasagittally striped molecular pattern in the cerebellar vermis; and (3) the determination of the ZII-negative phenotype in surviving PCs.

All experiments described in this paper were conducted in agreement with the stipulations of the San Raffaele Scientific Institute Animal Care and Use Committee, University of Calgary guidelines, and the Guide to the Care and Use of Experimental Animals as outlined by the Canadian Council for Animal Care.

The targeting construct, described in Corradi et al.(2003), contains a lacZ cDNA. lacZ staining distribution in CNS development is in full agreement with the results of in situ studies(Garel et al., 1997; Malgaretti et al., 1997)(present paper). All experiments were carried out on F₁ hybrids obtained by crossing Ebf2^+/- pure-bred FVB/N(N₉) females with Ebf2^+/- pure-bred C57BL/6J males. All studies were conducted using coisogenic control littermates. This hybrid strain was chosen to obviate the low fertility and poor maternal behavior of C57BL/6J heterozygous mothers.

Postnatal mice were anesthetized with Avertin (Sigma) and perfused with 0.9% NaCl followed by 4% paraformaldehyde (PFA). Embryos were fixed overnight by immersion with either 4% PFA or Carnoy's solution (6:3:1 solution of 100%ethanol, chloroform and acetic acid). Tissues fixed with 4% PFA were rinsed three times in 1× PBS, cryoprotected in 30% sucrose overnight, embedded in OCT (Bioptica), and stored at -80°C, before sectioning on a cryotome(20 μm); tissues fixed with Carnoy's solution were placed in 96% ethanol overnight, dehydrated in an ethanol series, embedded in paraffin. Sections of 5-7 μm were cut using a microtome. lacZ staining was conducted as described (Corradi et al.,2003).

Cryosections or paraffin-embedded sections were immunostained with the following antibodies: rabbit or mouse polyclonal anti-calbindin (1:1000,Swant); rabbit polyclonal anti-beta-galactosidase (1:700, Abcam); rabbit antineurogranin (1:1500, Chemicon); rabbit monoclonal anti-active Caspase 3(1:200, BD Pharmingen); rabbit polyclonal anti-sphingosine kinase 1a (1:500),a gift of N. Terada (Terada et al.,2004); mouse monoclonal anti-ZII(Brochu et al., 1990), used in spent hybridoma supernatant diluted 1:5000; mouse monoclonal anti-bromodeoxyuridine (1:100, Becton Dickinson); rabbit polyclonal anti-HSP25(1:500, Stressgen Biotechnologies); rabbit polyclonal anti-phosphorylated-histone 3 (1:500, Upstate Biotechnology); rabbit polyclonal anti-Ki67 (1:800, Novo Castra Laboratories) goat polyclonal anti-EphA4 (1:300, R&D Systems). For immunohistochemistry, sections were immunostained as recommended (ABC Elite kit, Vector Laboratories), dehydrated and mounted in DPX (BDH-Merck). For anti-Ki67 and anti-phosphorylated-histone3 immunohistochemistry, a high-temperature antigen retrieval procedure was required. For dual immunofluorescence, cryosections were rinsed three times in 1× PBS, preincubated in 15% goat serum, 0.2% triton X-100, 1× PBS and incubated overnight at 4°C with the two primary antibodies. Sections were then washed 6×10 min in PBS and incubated for 2 h at room temperature with the two secondary antibodies (Alexa 546 anti-mouse Ig,1:1000, and Alexa 488 anti-rabbit Ig, 1:1000). In controls either primary antibody was replaced with normal serum. Wholemount immunocytochemistry was performed as described (Sillitoe and Hawkes, 2002).

Digoxygenin-labeled riboprobes were transcribed from plasmids containing Ebf1, Ebf2, Ebf3, Reelin (A. Bulfone), RORα (B. A. Hamilton), p57 (L. Muzio and A. Mallamaci), Semaphorin3A(J.M. Solowska) and Protocadherin10 (S. Hirano) c-DNAs. In situ hybridizations were performed as described by Pringle and Richardson(www.ucl.ac.uk/∼ucbzwdr/doubleinsituprotocol.htm).

Paraffin-embedded P15 and P60 wt (n=3) and null (n=3)cerebella were sagittally sectioned. Complete series of 7 μm sections spanning the entire cerebellum were immunostained with an anti-CaBP antibody to visualize PC bodies. Morphometric evaluation of PC numbers was carried out using Neurolucida software (MicroBrightField) connected to a Nikon E-800 microscope via a color CCD camera. The number of PCs and the length of the PC layer were estimated from sagittal sections of the cerebellar vermis from wild-type (wt) and Ebf2 null mice (for details, see Buffo et al., 1997). For each animal, three sections close to the midline were analysed. Statistical analysis was conducted by using a two-tailed t-test with equal variance.

The count of apoptotic bodies (active Caspase 3-positive) at P0 was performed in triplicate. Using a Zeiss Axioplan upright microscope (5×objective) and a digital camera, pictures were taken of each complete series of frontal sections (wt versus mutant) stained with anti-active Caspase 3;pictures spanning each section were merged with Photoshop 8 to produce high-definition screen images. Positive cell bodies were counted on screen,marking each with the paintbrush tool of Adobe Photoshop 8. Statistical analysis of mean apoptotic cell numbers/section was conducted by using a two-tailed t-test with equal variance.

Pregnant females were given a single intraperitoneal injection with 100μg/kg BrdU (Sigma) at gestational ages E11.5 and E12.5, and killed 2 h later. Embryos were dissected in cold 1× PBS, fixed overnight with Carnoy's solution and paraffin embedded. Anti-BrdU immunohistochemistry was performed as described (Garel et al.,1997). Cell counts were performed in triplicate as described(Takahashi et al., 1993).

The Ebf2 null cerebellum appears reduced in size in the horizontal, frontal and sagittal planes(Fig. 1B,D,F, respectively,controls in A,C,E). A PC count conducted on P15(Fig. 1G) and P60 (not shown)sagittal sections revealed a 38% reduction in the total number of PCs compared to the wt cerebellum. Notably, the number of PCs/mm of PC layer length is not significantly changed in mutants. Cerebellar dysmorphogenesis differentially affects cortical lobules. The anterior vermis is clearly hypotrophic, most strikingly at the level of the lobules II, III and IV-V (arrowheads in Fig. 1B,D,F), whereas lobule VIa is normal and its size is not obviously affected by the Ebf2mutation. However, lobules VIb, VII and VIII are reduced in size (arrows in F). Posteriorly, lobule IX is malformed, while X is normal. The hemispheres are also malformed (see Fig. 1B,D, controls in A,C). These morphogenetic alterations are accompanied by obvious motor deficits (see also Corradi et al., 2003) and behavioral abnormalities (Rotarod test, not shown). These results prompted us to conduct further studies of cerebellar development and function in these mutants.

During mouse embryonic development, PCs are born in the VZ between E11 and E13, when the bulk of PC proliferation is completed(Hashimoto and Mikoshiba,2004; Sotelo,2004), then move into a postmitotic domain known as the cortical transitory zone (CTZ). Because Ebf genes encode structurally similar proteins that could play redundant roles in development, we compared the distribution of Ebf1-Ebf3 mRNAs in the cerebellar primordium at E11.5(Garel et al., 1997) and E12.5, when the peak of PC progenitor proliferation occurs, and then analyzed subsequent stages of embryogenesis and postnatal development.

As shown by in situ hybridizations on adjacent 7 μm sections, at E12.5(Fig. 2A-C), the Ebf2expression profile differs from those of Ebf1 and Ebf3 in two respects: first, Ebf2 expression is restricted to a thin subventricular layer within the CTZ (Fig. 2B) while Ebf1 and Ebf3 are expressed throughout the CTZ; second, Ebf2 expression spans the midline of the metencephalic plate (arrow in Fig. 2B), suggesting that it may play a role in the formation of medial territories of the cerebellar plate. In addition at this stage, Ebf2is strongly expressed in the nuclear transitory zone (NTZ), which harbors presumptive DCN neurons. Thus, in early cerebellar development Ebf2expression overlaps partially with the distributions of Ebf1 and Ebf3, as well as exhibiting some features of its own.

At E17.5, the expression domains of Ebf1-Ebf3 in the cerebellar cortex are nested: Ebf1 is expressed from the midline through the lateral edge of the developing PC layer; Ebf3 is excluded from the most lateral domains and Ebf2 labels an even more medially confined PC subset (Fig. 2D-F). For comparison, we analyzed the expression of RORα, a universal PC marker, in adjacent sections (Fig. 2G) and found a similar distribution as compared to Ebf1,allowing the latter to be considered as a new bona-fide PC-specific marker.

In the postnatal cerebellum (P3), Ebf2 is expressed in a PC subset(Fig. 2I) and is distributed in discontinuous parasagittal and transverse domains. Unlike Ebf2, Ebf1is present in all PCs (Fig. 2H), while Ebf3 is expressed in most PCs and in the external granular layer (Fig. 2J).

Having scored a decrease in the total number of PCs in the mutant cerebellum, we analyzed markers of DNA synthesis, mitosis, and cell cycle exit in the VZ of the cerebellar primordium. Initially we measured BrdU incorporation 2 h after a single pulse(Takahashi et al., 1993)(Fig. 3A,B) at E11.5 and E12.5(Fig. 3I). Moreover we counted progenitors positive for phosphohistone 3, a marker that selectively labels metaphase chromosomes (Paulson and Taylor,1982) (Fig. 3C,D). We also tested the distribution of Ki67, which interacts with members of the heterochromatin protein 1 family, present in all active phases of the cell cycle (Scholzen et al., 2002; Scholzen and Gerdes, 2000)(Fig. 3E,F). Finally, we analyzed by in situ hybridization the expression of the Kip2 gene,encoding p57, a cyclin kinase inhibitor of the Cip/Kip family that regulates cell cycle progression during development(Hong et al., 1998)(Fig. 3G,H). Our results do not reveal any significant changes in the null VZ, indicating that Ebf2does not control PC progenitor proliferation or cell-cycle exit. This is in keeping with observations independently made by others in chick embryos(Garcia-Dominguez et al.,2003).

We then tackled the possible role of Ebf2 in the control of PC migration. During normal development, PCs migrate along radial glial fibers,skirting the DCN neuron clusters in the nuclear transitory zone (NTZ) to reach their final destination beneath the granule cell progenitors located in the external granular layer (reviewed in Sotelo, 2004). Importantly, in the mutant cerebellum, radial glia are normally developed, as revealed by immunostaining for several specific markers: BLBP, RC2, GLAST and GFAP (not shown).

In order for PC migration to be analyzed in the mutant, PCs were immunolabeled at E15.5 for Calbindin (CaBP), a calcium-binding protein selectively expressed in PCs at the onset of their migration into the cortex. Our results indicate that in the mutant a large number of CaBP-positive PCs accumulate in the anterior CTZ (Fig. 4B,D), while CaBP transcript levels are unchanged, as assayed by real-time PCR (not shown). Studies of Ebf1-Ebf3and RORα transcript distribution at the same stage confirmed these findings, as detailed in Fig. S1 in the supplementary material. Importantly the expression of Reelin, the best-characterized guidance cue in PC migration (Jensen et al.,2002), is unchanged in DCN neurons and in the EGL of Ebf2mutants (see Fig. S1I,J in the supplementary material).

Our findings raised questions as to whether delayed-migrating PCs undergo cell death. At birth, in the null cerebellum the number of apoptotic bodies positive for active caspase 3 is significantly increased (P=0.00024)(Fig. 5A). Many dying cells are located near the origin of their migratory path(Fig. 5B,F), and colocalize with delayed-migrating lacZ-positive cells(Fig. 5G), while no clustering of dead cells is observed at the same location in the wt(Fig. 5E). By double immunofluorescence for active Caspase 3 and CaBP we found that 51% of the apoptotic cells in the vicinity of the VZ(Fig. 5B) were also positive for CaBP (example in Fig. 5C,D). However, some cell debris immunoreactive for active Caspase 3 was CaBP-negative (arrowheads in Fig. 5D). The DCN are already well developed at birth in the wt and null alike, and no increase in the number of Caspase 3-positive cells could be found in the null. Finally, postmigratory, caspase-3/lacZ-double-positive PCs are observed in the Ebf2 null cerebellar cortex, both in the vermis(Fig. 5I) and in hemispheres(Fig. 5L,M).

We then asked if the apoptotic PCs belong to any established PC subpopulation. Between E16.5 and P10, a subset of maturing PCs express neurogranin II and are fated to become ZII-negative during the third week of life (Larouche et al., 2006). E18.5 and P4 cerebella of Ebf2 null and wt mice were stained for neurogranin. While in the wt, neurogranin labels arrays of PCs located in the cerebellar cortex (Fig. 6A), in the null mutant E18.5 cerebellum, neurogranin labels mostly ectopic,delayed-migrating cells near the VZ (Fig. 6B). By P4, while the wt cerebellum exhibits an ordered array of neurogranin-positive PC clusters (Fig. 6C,E), no neurogranin immunoreactivity is detectable in the null mutant cerebellum (Fig. 6D,F). This suggests that neurogranin-positive PC precursors fail to migrate, and subsequently die between E18.5 and P4. Importantly, in the heterozygous cerebellum, neurogranin is coexpressed with beta-galactosidase (not shown),suggesting a cell-autonomous effect of Ebf2 on the development/survival of neurogranin-positive PCs. Likewise, cerebella were stained for HSP25, another PC marker expressed in parasagittal stripes in the vermis (Armstrong et al.,2001b). Hardly any immunostaining for HSP25 can be observed at P3 in the anterior vermis of Ebf2 null mice(Fig. 6H). The loss of two independent striped markers, neurogranin and HSP25, in the vermis is suggestive of PC loss rather than transcriptional downregulation.

The degree of PC loss was also tested by in situ hybridization and immunohistochemistry on P3 frontal sections, again focusing on the vermis. We found that the number of PCs positive for Ebf1 and Ebf3 is profoundly reduced in the anterior zone(Fig. 7C,L). In addition, as markers of PC subtypes, we examined the cerebellar distribuiton of striped markers such as Semaphorin 3A (Sema3A)(Solowska et al., 2002), that is widely coexpressed with Ebf2, and the receptor tyrosine kinase EphA4 (Rogers et al., 1999),showing that EphA4-positive stripes are considerably reduced in size(asterisks in Fig. 7I, control in H), and that a large number of Sema3A-expressing PCs are lost in the mutant (arrows in Fig. 7Fand O, controls in E and N).

In parallel, we performed in situ hybridizations of P6 frontal sections to evaluate the distribution of two markers of cerebellar hemispheres, protocadherin 10 (Pcdh10)(Hirano et al., 1999) and Sema3A (L.C. and G.G.C., unpublished observations). Cerebellar hemispheres exhibit complex folding defects in the Ebf2 null mutant. Our results indicate that the number of Pcdh10-expressing PCs is hardly or not reduced, whereas the number of Sema3A-expressing PCs is sharply decreased in the hemispheres (supplemental Fig. S2). Overall, our result speak for a massive reduction in the number of a molecularly defined PC subset in the Ebf2 null cerebellum.

Finally, we examined the consequences of the Ebf2 mutation on the development of cerebellar topography, as defined by the distribution of well-established PC striped markers. The striped organization of the cerebellar cortex is best seen in wholemount cerebella(Fig. 8A), which reveal a remarkable array of reproducible parasagittal stripes. The mouse cerebellum comprises four transverse zones that extend across both the vermis and the hemispheres (Ozol et al.,1999): the anterior zone (AZ: approximately lobules I-V), the central zone (CZ: approximately lobules VI and VII), the posterior zone (PZ:approximately lobules VIII and dorsal IX), and the nodular zone (NZ:approximately dorsal lobule IX and lobule X). Each zone includes several lobules, and interdigitates at their boundaries, but zonal boundaries do not coincide with lobule boundaries (Hawkes and Eisenman, 1997; Ozol et al., 1999; Eisenman,2000).

We analyzed the distribution of the Ebf2-lacZ transgene in adult Ebf2^+/- cerebella, which develop normally(Fig. 8). In the NZ, flocculus and paraflocculus, there is no transgene expression. In the PZ (e.g. lobule VIII), the transgene is expressed in the vermis by two pairs of PC stripes,and an additional array of broad expression stripes lies in the hemispheres(l. paramedianus). More rostrally, in the CZ (e.g. lobules VIa,b) and in its hemispheric extension, expression is either absent or weak. The ansiform lobule is negative. Finally, in the AZ (lobules I-V) most PCs in the vermis express the transgene (e.g. lobule III), but narrow transgene-negative stripes can be seen (e.g. arrows in lobule III).

We systematically compared the expression of Ebf2-lacZ in the heterozygote to the expression of ZII(Brochu et al., 1990)(Fig. 8B-H). In general, the transgene is expressed at high levels in the ZII-immunonegative PC population,as shown for the posterior hemispheres (the hemispheric PZ: Fig. 8B,D,E), posterior vermis(both PZ, alternating stripes, and NZ, all ZII-immunopositive/lacZ-negative: Fig. 8C,F), anterior hemisphere(hemispheric AZ: Fig. 8G) and anterior vermis (the AZ: Fig. 8H). Rare exceptions to this rule were also found(Fig. 8F).

To determine the consequences of the Ebf2 null mutation for the pattern of parasagittal and transverse compartments, wt and null P24 cerebella were immunostained with antibodies against ZII, HSP25, and sphingosine kinase 1a (SPHK1a). In the Ebf2 null the pattern of lobulation is perturbed,due in large measure to the perinatal loss of PCs, but stripes of transgene expression are visible and related to those in the heterozygote(Fig. 9A). The pattern of expression in the posterior cerebellum (the NZ and PZ) is barely altered, and the changes are probably not significant. The expression in the CZ (e.g. lobule VIII) tends to be indistinct and weak, but is probably not patterned differently from the heterozygote. Expression in the central zone is unaffected in the null mutation. Significant changes are seen in the AZ as a result of heavy PC loss perinatally: the vermis is dramatically narrowed and the broad paravermian stripes in the +/- are much closer to the midline (e.g. compare lobule III in Fig. 8Awith Fig. 9A). In the hemispheres, the pattern of striped transgene expression in the AZ and PZ,with no transgene expression in the NZ and CZ, is maintained unaltered.

While the pattern of transgene expression appears hardly affected by the null mutation (except where it is distorted due to PC loss), the expression of ZII is radically changed. Fig. 9B-G, compare sections through the heterozygote (B-G) and homozygote (B′-G′), immunoperoxidase-stained for ZII, and Fig. 9I-P present similar data from wholemount immunostained cerebella. The NZ (lobule X: wholemount views in Fig. 9K,L) is comprised entirely of ZII+/Ebf2-lacZ PCs in the heterozygote, and is unaffected in the null. However, in the PZ, where ZII is normally expressed in regular parasagittal stripes, in the null cerebellum striped expression is replaced by a uniformly ZII-immunopositive domain [lobule IX(Fig. 9B,B′), VIII(Fig. 9C,C′) and caudal VII (Fig. 9D,D′)]. The same applies in the posterior hemisphere (compare wholemounts in Fig. 9K with 9L). As for the NZ, all PCs in the CZ are ZII-immunoreactive, and this is unaffected in the null mutant cerebellum (not shown). In both the NZ and the CZ, although ZII expression does not reveal a striped topography, the expression of other markers does. In particular, the small heat-shock protein HSP25 is expressed in both zones in parasagittal stripes (Armstrong et al.,2000): this pattern is unaltered in the Ebf2 null mutant cerebellum (not shown). In the AZ, an alternating pattern of ZII expression can still be defined (lobule IV (Fig. 9E,E′), III (Fig. 9F,F′), I (Fig. 9G,G′), but the spacing between the ZII+ stripes is significantly reduced (e.g. in lobule III from ∼1050 μm in the +/- to 700 μm in the null; see also wholemounts in Fig. 9M-P).

The effect of the Ebf2 deletion is not restricted to ZII expression. The striped expression of sphingosine kinase 1a (SPHK1a), which is co-expressed in the same stripe arrays as ZII(Terada et al., 2004), is similarly affected (Fig. 9H′: control in 9H), strongly suggesting that the effect is not mediated via the direct control of ZII gene expression by EBF2, but rather reflects a wider alteration in the specification of PC subtypes. The effect of the null mutation on cerebellar patterning in the anterior cerebellum can be explained by the loss of a large fraction of the ZII-immunonegative PC population. However, the effect in the posterior cerebellum clearly suggests that Ebf2 is required for a subset of ZII-positive PCs to switch and acquire the ZII-negative phenotype. A large subset of Ebf2+ (i.e. beta-galactosidase-positive) PCs, normally fated to become ZII-negative during the third week of life, fail to switch, and remain ZII positive instead. For example, `transdifferentiated' PCs are plentiful in the paramedian lobe of the homozygous mutant (e.g. Fig. 10A-C,D-F); conversely, in Ebf2 heterozygous littermates,no double-labeled cells are seen (Fig. 10G). These results confirm that the cerebellar stripe topography is profoundly disrupted in Ebf2 mutants.

Our analysis of the Ebf2 null mutant has revealed a selective disruption of cerebellar corticogenesis. During embryogenesis, Ebf2is expressed robustly in postmitotic PC precursors, in domains that are also positive for its close homologs Ebf1 and Ebf3. The fact that the cerebellar cortex is malformed in the Ebf2 null mutant despite coexpression of Ebf1 and Ebf3 during development suggests that EBF2, despite the tight structural similarity to EBF1 and EBF3, may actually exert unique and distinct functions in PC development.

The Ebf2 null mutation does not overtly affect cell proliferation or cell cycle exit, nor does it appear to significantly disrupt the development of cerebellar radial glia. Conversely, the mutant cerebellum displays a clear defect in the migration of a subset of PC precursors. The migration abnormalities observed in Ebf2 null mutants are probably not secondary to a defect in early stages of the canonical Reelin pathway, and reveal the existence of a Reelinindependent network presiding over PC precursor migration.

The effects of nullisomy for Ebf2 on PC migration become obvious by embryonic day 15. At birth, numerous delayed-migrating PCs accumulate in the proximity of the VZ, due to an arrest in their migration prior to reaching the cortex, and die by apoptosis. Dying cells colocalize with delayed-migrating lacZ-positive neurons, suggesting a cell-autonomous role of Ebf2 in regulating PC migration/survival. In addition, many PCs that do reach the null mutant cortex also die, possibly due to a lack of intrinsic survival mechanisms; alternatively, mislocalized PCs may fail to connect properly to their postsynaptic targets. Importantly, no marker misexpression has been scored in mutant DCN labeled with Sema3A, Ebf3,Cdh8 or Pcdh10 riboprobes (not shown).

The considerable increase in perinatal PC death observed in the mutant is in keeping with the 38% PC loss scored at P15 and P60 in null cerebella. A large share of the apoptotic bodies observed at birth are clearly positive for the PC-specific marker CaBP, while others are not, potentially representing dying PCs that have ceased expressing CaBP(Vig et al., 1998), PCs that do not yet express CaBP, or possibly other unrelated neural cell types.

A high percentage of apoptotic cells are represented by delayedmigrating,neurogranin-positive PCs, that are lost between E18.5 and P4. Neurogranin is an abundant thyroid hormone-dependent protein kinase C substrate, that regulates Ca²⁺ signaling modulating Ca²⁺/calmodulin availability, and lessens the extent to which calcium-calmodulin-dependent enzymes become or stay activated(Gerendasy, 1999). Neurogranin-positive PCs, in which the Ca²⁺ buffering strength of calmodulin is attenuated, may fail to respond to specific Ca²⁺-dependent developmental signals such as retinoic acid, thyroid hormone or (potentially) Notch signaling(McKenzie et al., 2005). Incidentally, the gross abnormalities observed in Ebf2 null cerebella are reminiscent of, albeit more severe than, those observed in hypothyroid mice (Koibuchi and Chin,2000).

Neurogranin is not the only subtype-specific marker lost in the mutant cerebellum: during early postnatal development HSP25, normally expressed in six narrow parasagittal stripes in the vermis of the anterior lobe(Armstrong et al., 2001b), is lost in the anterior lobe of the Ebf2 null mutant. PCs lost around birth are positive for generic PC markers, and for genes with established roles in axon guidance, such as Sema3A and EphA4, that are expressed in parasagittal stripes.

PCs positive for neurogranin during late development are fated to become ZII-negative after birth (Larouche et al.,2006). The fact that in the Ebf2 null cerebellum neurogranin-positive PCs are selectively lost by P4 translates into a loss and underrepresentation of ZII-negative PCs in the mature cerebellum. The same applies to other parasagittally striped markers coexpressed with ZII, such as SPHK1a. The development of cerebellar patterning extends from before E10 to P20. Evidence from multiple routes concurs that most major elements of the fundamental cerebellar topography are established at the time of PC birth,between E11 and E13, in the VZ of the 4th ventricle (reviewed by Armstrong and Hawkes, 2000; Herrup and Kuemerle, 1997; Oberdick et al., 1998). From E13.5 onwards, molecular markers reveal evidence of PC compartmentation [e.g. the restricted distribution of the OMP transgene(Nunzi, 1999)]. Incidentally,the OMP promoter contains two binding sites for OLF1/EBF transcription factors(Kudrycki et al., 1993). Thus, Ebf2 is one of the earliest-known functional links to the molecular compartmentalization of the adult cerebellar cortex. PC migration defects restricted to the ZII-negative population have been previously described, e.g. in the cerebellar-deficient folia mutant (cdf)(Beierbach et al., 2001; Park et al., 2002). Selective PC ectopia has also been described within a ZII-positive population in the weaver mouse (Armstrong and Hawkes,2001). However, the Ebf2 null mouse reveals the first evidence of the mechanisms through which the prospective ZII phenotype is determined in postmitotic PC precursors. Ebf2 expression in the wt cerebellum is restricted to ZII-negative PCs, and in the Ebf2 null cerebellum ZII-negative PC stripes are selectively disrupted, with profound consequences on cerebellar topography. In the vermis, lobules VIII and VII,which normally display alternating ZII-positive/-negative stripes(Fig. 9C and D, respectively),now contain exclusively ZII-positive PCs(Fig. 9D′), or an unpatterned admixture of ZII-positive and -negative cells(Fig. 9C′). Likewise, in the anterior zone, where neurogranin-positive PCs are lost, the characteristic stripes of ZII-positive PCs are much more narrowly spaced than normal. Taken together, these abnormalities show that the Ebf2 null mutation results in hypotrophy, defective lobulation and an incomplete failure of PC dispersal.

Furthermore, the paramedian region of the hemispheres, that expresses Ebf2 (Fig. 8A and Fig. 10G), contains an overwhelming majority of `transdifferentiated'(ZII-positive/beta-galactosidase-positive) PCs(Fig. 10A-C,D-F), at odds with the mutually exclusive expression of beta-gal and ZII in the heterozygous control (Fig. 10G). This is the first time that any manipulation has been shown to cause PCs to switch their striped molecular phenotype via transdifferentiation (e.g. Armstrong and Hawkes, 2000; Leclerc et al., 1988; Wassef et al., 1990). Previous attempts to alter cerebellar topography surgically(Armstrong et al., 2001a; Leclerc et al., 1988) or genetically (Armstrong and Hawkes,2001; Edwards et al.,1994; Eisenman et al.,1998; Gallagher et al.,1998) have proven ineffectual: the gross morphology may be affected but the underlying topography is normal. These data suggest a model for PC phenotype determination: while the default phenotype is ZII-positive,the presence of Ebf2 - which is expressed from at least E11.5 - affects the development of Ebf2-expressing PCs fated to acquire the ZII-negative phenotype. When Ebf2 is deleted then some PCs fated to become ZII-negative fail to migrate and die, and others default to a ZII-positive phenotype.

What could be the functional relevance of the observed alterations in cerebellar circuits documented in the present paper? The cerebellar striped markers' distribution has been found to reflect the ordered arrangement of the afferent and efferent terminal fields (Ji and Hawkes, 1994; Ji and Hawkes, 1995; Voogd et al.,2003; Voogd and Ruigrok,2004). Preliminary results (F.R. and S.G., unpublished) clearly indicate that climbing fiber terminals spread laterally in the Ebf2^-/- molecular layer across ZII^+/-boundaries, a finding that faithfully mimics the effect of olivocerebellar denervation (Rossi et al.,1991). Further studies are in progress to characterize these events and their molecular basis in due detail.

To date, our results allow us to propose Ebf2 as a factor regulating short-range cerebellar patterning, organizing orderly molecular domains and intraparenchymal boundaries. The abnormalities scored in the Ebf2 null cerebellum strengthen the hypothesis that EBF transcription factors are key players in a phase of development that links neuronal differentiation, morphogenetic movements, cell survival and the early functional activation of fledging neuronal circuits.

Image analysis was carried out in Alembic, a microscopy laboratory established by the San Raffaele Scientific Institute and the Vita-Salute San Raffele University. We thank Cairine Logan and Ray Turner for their comments on the MS. GGC was supported equally by the Italian Telethon Foundation and by Fondazione Mariani. Additional support came from the Italian University Ministry (MIUR) through FIRB grants RBNE01WY7P, RBNE015242 and RBLA03AF28. R.H. was supported by the Canadian Institutes of Health Research.