Cerebellar granule cell progenitors proliferate postnatally in the upper part of the external granule cell layer (EGL) of the cerebellum. Postmitotic granule cells differentiate and migrate, tangentially in the EGL and then radially through the molecular and Purkinje cell layers. The molecular control of the transition between proliferation and differentiation in cerebellar granule cells is poorly understood. We show here that the transmembrane receptor Plexin-B2 is expressed by proliferating granule cell progenitors. To study Plexin-B2 function, we generated a targeted mutation of mouse Plexin-B2. Most Plexin-B2−/− mutants die at birth as a result of neural tube closure defects. Some mutants survive but their cerebellum cytoarchitecture is profoundly altered. This is correlated with a disorganization of the timing of granule cell proliferation and differentiation in the EGL. Many differentiated granule cells migrate inside the cerebellum and keep proliferating. These results reveal that Plexin-B2 controls the balance between proliferation and differentiation in granule cells.
The morphogenetic development of the cerebellum is based on a precisely orchestrated sequence of proliferation and differentiation of its different cell types. Cerebellar granule cells (GCs), by far the most numerous neurons in the cerebellum, play a key role in this process. In mice, granule cell progenitors (GCPs) are born from embryonic day 13 (E13) in rhombic lip (Wingate, 2005) and migrate tangentially over the cerebellar plate to form a secondary proliferating zone, the external granule cell layer (EGL). The proliferation of granule cell progenitors is controlled by several secreted proteins such as bone morphogenetic proteins (BMP), stromal cell-derived factor 1 (SDF-1), and Sonic Hedgehog (Shh) (Alder et al., 1999; Dahmane and Ruiz i Altaba, 1999; Wechsler-Reya and Scott, 1999; Klein et al., 2001; Blaess et al., 2006; Corrales et al., 2006). Postmitotic granule cells migrate tangentially in the EGL before descending along Bergmann glia to settle in the internal granule cell layer (IGL) (Yacubova and Komuro, 2003). The molecular mechanisms that control the switch from proliferation to differentiation in cerebellar granule cells is essentially unknown.
Plexins are a family of transmembrane proteins that have been originally characterized for their role in axon guidance. More recently, they were also shown to control cell migration, immune response, and blood vessel growth (Kruger et al., 2005). Plexins are a major component of the receptor complexes for semaphorins (Gherardi et al., 2004). In vertebrates, nine Plexin genes, distributed into four subclasses (A–D) have been identified (Tamagnone et al., 1999; Hartwig et al., 2005). Type-B Plexins (B1–B3) are widely expressed in developing and adult mice (Perala et al., 2005). In vitro experiments have suggested that B-Plexins can be activated by class 4 semaphorins (Kruger et al., 2005). Moreover, signaling downstream of B-Plexins, in particular Plexin-B1, is beginning to be well characterized in transfected cells and involves small GTPases (Oinuma et al., 2004b; Toyofuku et al., 2005). Although the physiological importance of A-, C- and D-Plexins in vivo has been primarily validated by the phenotypic analysis of knock-out mice (Pasterkamp et al., 2003; Gu et al., 2005; Suto et al., 2005; Yaron et al., 2005), the in vivo function of B-Plexins remains unknown. In the developing mouse nervous system, the highest levels of Plexin-B2 mRNAs are detectable in the ventricular zone of the telencephalon and in the EGL (Worz-Feld et al., 2004). To determine the function of Plexin-B2 in cerebellum development, we generated a targeted mutation of the mouse Plexin-B2 gene.
We report here that the Plexin-B2 mutation results in two different phenotypes, depending on the genetic background. In the C57BL/6 inbred strain, Plexin-B2−/− mutants develop a cephalic neural tube closure defect, accompanied by perinatal lethality. In the CD1 outbred strain, many Plexin-B2−/− mutants survive and display a profoundly altered layering and foliation of the cerebellum. These defects are correlated with a major disorganization of the EGL: the timing of granule cell proliferation and differentiation is abnormal, and cells expressing markers of differentiated granule cells migrate inside the cerebellum and keep proliferating for several more days.
Materials and Methods
Swiss and C57BL/6 mice (Janvier, Le Genest-St. Isle, France) were used for simple expression studies.
Postnatal day 0 (P0) to P5 mice were anesthetized on ice and, after P5, by inhalation of isofluorane Foren (Abbott Laboratories, Abbott Park, IL). The day of birth corresponds to P0. All animal procedures were performed in accordance with institutional guidelines.
Generation of Plexin-B2-deficient mice.
The Plexin-B2 gene was mutated by targeted trapping as described previously (Friedel et al., 2005). Briefly, the targeted trapping construct was constructed by flanking the placental alkaline phosphatase (PLAP) secretory trap cassette (Leighton et al., 2001) with 5′ and 3′ homology arms (5 and 3 kb, respectively), which were generated by PCR from genomic DNA of E14Tg2a.4 embryonic stem (ES) cells using the Expand High Fidelity PCR system (Roche, Indianapolis, IN). Correct homologous recombination resulted in the insertion of the secretory trap cassette between exons 16 and 17 (exon containing start codon is 1). Exons on the 5′ homology arm were sequenced to control for potential nonsense mutations.
The construct was electroporated into the feederless E14Tg2a.4 ES cell line, and drug-resistant clones were selected with 125 μg/ml G418 (catalog #11811-031; Invitrogen, Carlsbad, CA) for 10 d. A 1 kb 5′ external probe was used to screen for correct homologous recombination events by Southern blot analysis (112 of 118 clones positive). Before blastocyst injection, selected ES cell clones were confirmed by Southern blot analysis with an internal neo probe and by genomic PCR with a 3′ external primer.
Two mouse lines of different genetic background were established by continuously backcrossing heterozygous males to either C57BL/6 or CD1 females (Charles River Laboratories, Wilmington, MA).
PCR genotyping was performed with a common forward primer (5′-gccattgagaagctttgtcaggtg), a wild-type specific reverse primer (5′-gcaaacttctggatgaggctgaag), and a mutant specific reverse primer (5′-acttccggagcggatctcaaactc). The official allele name [Mouse Genome Informatics (The Jackson Laboratory, Bar Harbor, ME)] is Plxnb2tm1Matl.
Math1:GFP;Plexin-B2 mice were obtained by crossing Plexin-B2-deficient mice with transgenic mice expressing the green fluorescent protein (GFP) under the Math1 (mouse atonal homolog 1) promoter (Lumpkin et al., 2003).
Brains were collected as described previously (Marillat et al., 2002). Brain sections were incubated with antibodies against β-galactosidase (β-gal) (1:1000; Cappel, West Chester, PA), calcium-binding protein calbindin-D 28K (CaBP) (1:1000; Swant, Bellizona, Switzerland), phospho-histone-H3 (H3P) (1:1000; Cell Signaling Technology, Beverly, MA), Plexin-B2 (1:2000; gift from Dr. Havran, Scripps Research Institute, La Jolla, CA), GABAA receptor α6 subunit (α6) (1:1000; Chemicon, Temecula, CA), parvalbumin (1:1000; Sigma, St. Louis, MO), vesicular glutamate transporter 2 (vGlut2) (1:3000; Chemicon), semaphorin 6A (mSema6A) (1:200; R & D Systems, Minneapolis, MN), TAG-1 (transient axonal glycoprotein 1) (1:3000), glial fibrillary acid protein (GFAP) (1:400; Chemicon), laminin (1:500; Sigma), Pax6 (paired box gene 6) (polyclonal 1:1000; Chemicon), Pax6 (monoclonal 1:10; Developmental Studies Hybridoma Bank, University of Iowa, Iowa City, IA), nestin (1:500; Chemicon), retinoid orphan receptor α (RORα) (1:500; Santa Cruz Biotechnology, Santa Cruz, CA), activated-caspase-3 (1:250; Cell Signaling Technology), NCL-Ki67 (1:1000; Novocastra Laboratories, Newcastle upon Tyne, UK), Doublecortin (Dcx) (1:1000; Chemicon), and Zic1 (zinc finger protein of the cerebellum 1) (1:3000; gift from Dr. Rosalind Segal, Dana-Faber Institute, Boston, MA), followed by species-specific secondary antibodies (Jackson ImmunoResearch, West Grove, PA). Sections were counterstained with Hoechst 33258 (10 μg/ml; Sigma), mounted in Mowiol (Calbiochem, La Jolla, CA), and examined with a fluorescent microscope (DM6000; Leica, Nussloch, Germany) or a fluorescent confocal microscope (DM IRBE; Leica). Histochemical stainings for the β-gal and PLAP reporters were performed as described previously (Leighton et al., 2001). All comparisons and quantifications were done on sections of comparable mediolateral levels and within lobules IV to VIII.
Plexin-B2 antibody production.
For Western Blot analysis, we used a rabbit antiserum that had been raised against a N-terminal peptide (EYGSIPVDINKKIKQD) of Plexin-B2 (1:1000; a kind gift from Dr. Goodnow, Australian National University, Canberra, New South Wales, Australia). For immunohistochemistry, we used an Armenian hamster monoclonal antibody (clone 3E7) that had been raised against the extracellular domain of Plexin-B2 (1:2000; a kind gift from Dr. Havran).
Bromodeoxyuridine staining and quantification of cell proliferation.
P0, P11, P13, P15, or P27 mice were injected intraperitoneally with bromodeoxyuridine (BrdU) (15 mg/ml, 50 mg/kg body weight; Sigma) diluted in a saline solution. Animals were perfused 1 or 3 h (short pulse labeling) or 24 h after injection. Brain sections were incubated 30 min at 37°C in 2N HCl in PBS before immunohistochemistry with a rat anti-BrdU antibody (1:100; Harlan, Indianapolis, IN). To count BrdU-positive cells at P11, when proliferation rate is maximal, we realized 60× confocal microscopy acquisitions (stacks of 2 μm deep). At P15, when proliferation is lower, we counted the total number of BrdU-positive cells per section. To quantify the proportion of Ki67 and BrdU double-labeled cells at P15, we realized 60× confocal microscopy acquisitions (stacks of 2 μm deep) of superficial regions of cerebellum. To quantify H3P-positive cells at P0, we counted both the total number of positive cells per section (expressed in square millimeters) and the number of positive cells in the more superficial regions of cerebellum (expressed by millimeters of EGL), within 60× confocal microscopy fields (stacks of 2 μm; expressed in millimeters of EGL length), on 10 sagittal and nonadjacent sections.
In situ hybridization.
Antisense riboprobes were labeled with digoxigenin-11UTP (DIG) as described previously (Marillat et al., 2002) by in vitro transcription of cDNAs encoding mouse Math1 (a gift from M. Wassef, École Normale Supérieure, Paris, France) (de Diego et al., 2002), Barhl1 (BarH-like 1) (a gift from F. Qiu, University of Medicine and Dentistry of New Jersey, Piscataway, NJ) (Li et al., 2004), Shh, Smoothened, Gli1 (glioma-associated oncogene homolog 1) or Plexin-B2 [DNA fragment from mouse expressed sequence tagged homologous to amino acids 1479–1721 in hPlexin-B2 (KIAA0315)] (Cheng et al., 2001). Templates for class 4 semaphorin probes were synthesized by reverse transcription-PCR from cDNA of P0 mouse brain and cloned into pCRII-TOPO (Invitrogen). In situ hybridizations were performed as described previously (Marillat et al., 2002).
For all statistical analysis, the significance was calculated by ANOVA (Statview; Abacus Concepts, Calabasas, CA). Each quantitative value represents mean ± SEM.
Generation of mutation in the Plexin-B2 gene by targeted trapping
To study the role of transmembrane proteins in neural development, we applied a targeted mutagenesis strategy that uses a promoterless gene trap vector to efficiently target candidate genes (“targeted trapping”) (Friedel et al., 2005). The targeted trap of the mouse Plexin-B2 gene was generated with a targeting construct that led to the insertion of the secretory trap vector (Leighton et al., 2001) into the intron between exons 16 and 17 (supplemental Fig. 1, available at www.jneurosci.org as supplemental material). This insertion creates a fusion transcript of the first 16 exons of Plexin-B2 with the elements of the secretory trap vector, consisting of a transmembrane domain/β-geo (β-gal fused to neo) cassette, followed by an internal ribosomal entry sequence and a PLAP sequence (supplemental Fig. 1, available at www.jneurosci.org as supplemental material). Fusion proteins of the secretory trap vector are retained in intracellular compartments and cannot act on the cell surface (Mitchell et al., 2001). By Northern and Western blot analysis of Plexin-B2−/− mutants, no wild-type mRNA or proteins were detected, indicating efficient splicing of the splice acceptor of the secretory trap vector (supplemental Fig. 1, available at www.jneurosci.org as supplemental material). We therefore predict that the targeted trap of Plexin-B2 results in a functional null mutation.
Plexin-B2 mutation causes neural tube closure defects
We bred the Plexin-B2 targeted trap allele for four generations into the inbred C57BL/6 strain. In this genetic background, almost all homozygous mutant embryos were found with cephalic neural tube closure defects (supplemental Fig. 2, available at www.jneurosci.org as supplemental material), which led to exencephaly and neonatal lethality. This phenotype was caused by a failure of fusion of the neural head folds during a critical period at approximately E8.5. We used the β-gal reporter to study the expression of Plexin-B2 at this time point and observed a high expression in the ridges lining the head folds (supplemental Fig. 2, available at www.jneurosci.org as supplemental material). We also observed in some of the Plexin-B2−/− mutants additional closure defects of the posterior neuropore, ranging in severity from a curled tail to spina bifida aperta (supplemental Fig. 2, available at www.jneurosci.org as supplemental material). Heterozygous animals appeared to be fully viable, with no detectable phenotype. In parallel to breeding with the C57BL/6 strain, we also introduced the Plexin-B2 mutation into the CD1 outbred strain. Outbred strains, which are genetically heterogeneous populations, have higher physical strength and lower neonatal lethality than inbred strains. After one generation backcross to the CD1 genetic background, neural tube closure defects occurred with reduced frequency, and, after four generations of backcross to CD1, ∼30% of Plexin-B2−/− mutants were fully viable and fertile.
Plexin-B2 is expressed by proliferating granule cell progenitors
It has been reported previously that Plexin-B2 mRNA is highly expressed in the cerebellar EGL (Worz-Feld et al., 2004). To study the expression of Plexin-B2 in the developing cerebellum in more detail, we used in situ hybridization for Plexin-B2 mRNA, immunostaining for Plexin-B2, and also the genetic reporters β-gal and PLAP. In the embryonic mouse cerebellum, Plexin-B2 mRNA was first detected at E13 in the choroid plexus of the fourth ventricle and in the primordium of the EGL that contains the first granule cell progenitors migrating over the cerebellar plate from the rhombic lip (Fig. 1A). Later on (E14–E17), simultaneously with the extension of the EGL, Plexin-B2 mRNAs and β-gal were detected all along the expanding EGL at the surface of the cerebellum (Fig. 1B,C). At P0–P1, Plexin-B2 and β-gal were evenly expressed throughout the whole EGL (Fig. 1D,E), but β-gal was also detectable in the IGL. This slight difference between Plexin-B2 mRNA expression and β-gal reporter activity (also seen at later stages, see below) is most likely attributable to the perdurance of the β-gal protein after the normal extinction of the Plexin-B2 mRNA and protein. Such a phenomenon has been described before for other gene trap lines (for instance, see Kerjan et al., 2005) as a result of the stability of the reporter protein. At P7–P10, Plexin-B2 and β-gal were still highly expressed in the EGL but also at a low level in the IGL (Fig. 1F,G). The EGL becomes subdivided into two layers around birth, when the first granule cells leave the cell cycle: a superficial layer containing proliferating granule cell progenitors (upper EGL) and a deeper layer containing tangentially migrating postmitotic granule cells (lower EGL). On P10 sections, Plexin-B2 mRNA was confined to the upper EGL (Fig. 1G). Plexin-B2 protein expression was next studied with an anti-Plexin-B2 monoclonal antibody. In the E15 cerebellum, strong Plexin-B2 immunoreactivity was found in the EGL and choroid plexus (Fig. 1H). At P0, double staining for Plexin-B2 and the mitotic marker H3P revealed that Plexin-B2 was expressed by proliferating granule cell progenitors (Fig. 1I). At P10, Plexin-B2 protein was detected on proliferating cells in the upper EGL but not on postmitotic granule cells expressing TAG-1 in the lower EGL (Fig. 1J). In the adult cerebellum, low levels of Plexin-B2 mRNA could be detected in the IGL by radioactive in situ hybridization (data not shown) and by the β-gal and PLAP reporters (Fig. 2F,G).
In conclusion, our data show that Plexin-B2 is highly expressed by proliferating granule cell progenitors.
Plexin-B2 mutation causes cerebellar defects: the adult phenotype
Viable Plexin-B2−/− mutants in the CD1 background displayed no obvious behavioral or motor deficits, but the gross morphology of their cerebellum was severely altered. First, cerebella of adult Plexin-B2−/− mutants were often smaller than those of control littermates (65–100% of control size; data not shown). Second, major foliation defects were obvious with adult cerebella stained in toto for β-gal histochemistry (Fig. 2A,B) and on sagittal sections immunostained for α6, a marker of mature granule cells (Kerjan et al., 2005) (Fig. 2C,D). In Plexin-B2−/− mutants, the precentral (ce) and intercrural (i) fissures were absent, and lobules I–III and VI–VII were fused. The other fissures were less pronounced, and most folia had an irregular shape. We also observed in all Plexin-B2−/− cerebella a midline cleft of lobule X, possibly resulting from a fusion defect of the cerebellar plates (supplemental Fig. 3, available at www.jneurosci.org as supplemental material). Third, many ectopic clusters of granule cells expressing β-gal and α6 were found at the cerebellum surface (Fig. 2A–F,J). Fourth, the structure of the IGL and Purkinje cell layers were extremely perturbed, mainly in the rostral and caudal folia (Fig. 2D–G). The cerebellar cortex was fragmented into multiple groups or islands of Purkinje cells (identified by calbindin, CaBP, immunoreactivity) that were embedded in α6-positive granule cells. At the base of some folia, the IGL was almost completely absent, Purkinje cells were found in the white matter, and some granule cells even invaded the deep nuclei (Fig. 2H,I). In the less affected folia, Purkinje cells were well aligned, their morphology unaffected, and only a few ectopic granule cells were observed in the molecular layer (Fig. 2J).
Cerebellum circuitry in Plexin-B2 mutants
Although the Plexin-B2 mutation causes a severe disruption of the laminar architecture of the cerebellar cortex, we found that the overall arrangement of cerebellar neurons and axons was preserved. In the ectopic islands of Purkinje cells, molecular layer interneurons identified by parvalbumin immunostaining were detected at their usual position among Purkinje cell dendritic trees. and some typical basket cell “pinceaux” were observed at the base of Purkinje cell somata (Fig. 3A,B).
To examine precerebellar projections to the cerebellar cortex, we used immunostaining for the vesicular glutamate transporter vGlut2 (Kerjan et al., 2005), which labels climbing fibers that synapse on Purkinje cell dendrites and mossy fibers that contact granule cell dendrites in structures called glomeruli. vGlut2 staining on Plexin-B2−/− cerebellum showed that the cell-type specificity of these connections was maintained even in the ectopic islands: climbing fibers contacted Purkinje cells, and mossy fibers rosettes were only found next to granule cells (Fig. 3C,D). Moreover, in Plexin-B2−/− mutants, parallel fibers visualized by PLAP staining were confined to Purkinje cell dendrites in the ectopic islands and in the molecular layer in the less affected folia (Figs. 2G, 3E,F).
Bergmann glia, basal lamina, and granule cell migration
The presence of ectopic granule cell clusters at the cerebellar surface has been reported previously in several mouse models that displayed a loss of the subependymal basal lamina (Graus-Porta et al., 2001). This basal lamina provides an anchorage for the end feet of the radial Bergmann glia cells, which are a substrate for the radially migrating early postmitotic granule cells. We therefore tested for the integrity of the basal lamina by immunostaining for laminin and for the integrity of the radial glia network by immunostaining for GFAP and nestin. At E18, no major difference was observed in the organization of the radial glia palisade between Plexin-B2+/− and Plexin-B2−/− (data not shown). At P11 and in adults, the palisade of radial glia was perturbed in Plexin-B2−/− cerebellum with many fibers having curved and irregular courses. However, GFAP-labeled radial glia fibers still extended to the pial surface and formed characteristic end feet (Fig. 4A–D,G,H). In the cerebellum of P11 Plexin-B2−/− mice, Pax6-immunopositive granule cells were frequently observed apposed to Bergmann glia fibers, suggesting that radial migration was not altered (Fig. 4G,H). Accordingly, only a few ectopic cells remain close to the surface or in the molecular layer in the adult mutant cerebellum (Fig. 2). Last, the expression and distribution of the extracellular matrix protein laminin was identical in Plexin-B2+/− and Plexin-B2−/−, both at P15 and in adults (Fig. 4C–F). These results suggest that Plexin-B2 deficiency does not cause major granule cells migration defects.
Development of the Plexin-B2−/− phenotype
To better determine the origin of cerebellum defects, we next analyzed early cerebellum development in Plexin-B2−/− mutants. In all cases, the cerebella of heterozygous Plexin-B2+/− animals were identical to wild type. At E14, we found no major differences in the arrangement of the EGL between Plexin-B2−/− and Plexin-B2+/− cerebella (Fig. 5A–D). At E17, the cerebellar distribution of Pax6-expressing cells, which is very variable along the rostrocaudal and mediolateral axis even in wild-type, still appears comparable in Plexin-B2+/− and Plexin-B2−/− (Fig. 5E,F). However, the whole-mount observation of E17 cerebella from Math1:GFP;Plexin-B2+/− and Math1:GFP;Plexin-B2−/− showed that the cerebellum from homozygous mutants lacks a rostral fissure (Fig. 5G,H). Despite this slightly delayed foliation, the expression pattern of the transcription factor Math1 (also known as Atoh1 and HATH1) was similar to controls, with Math1-expressing cells only positioned at the cerebellar surface throughout its anteroposterior extent (Fig. 5I,M). This suggests that the initial tangential migration of granule cell progenitors was not primarily affected in Plexin-B2−/− mutants. However, the use of markers for early postmitotic granule cells revealed some important abnormalities. The two transmembrane proteins Sema6A and TAG-1 are expressed by postmitotic granule cells that migrate in the lower EGL (Kerjan et al., 2005). At P0, in Plexin-B2+/− and Plexin-B2−/− mice, Sema6A was expressed in a high-anterior to low-posterior gradient, whereas TAG-1 was homogeneously expressed (Fig. 5J,L,N,O). However, in Plexin-B2−/− mice, ectopic Sema6A- and TAG-1-expressing cells were also found deep inside the cerebellar parenchyma (Fig. 5N–P). Moreover, the upper and lower EGL often overlapped (Fig. 5P). This disorganization of the EGL architecture was also observed with immunostaining for the Zic1 transcription factor (Aruga et al., 2002) (supplemental Fig. 3, available at www.jneurosci.org as supplemental material). Likewise, in P0 Plexin-B2+/− mice, PLAP-expressing cells were only detected in the upper EGL above Purkinje cells that were arranged in a thick multilayer (Fig. 5Q,R). In Plexin-B2−/− mice, PLAP-expressing cells were also found in the EGL, but many of them were localized inside the cerebellar plate, intermingled with CaBP-positive Purkinje cell bodies (Fig. 5S,T, inset in S). This suggests that, in mice deficient for Plexin-B2, the organization of the EGL is perturbed and that some granule cells migrate prematurely into the cerebellar parenchyma.
Abnormal granule cell proliferation in Plexin-B2-deficient mice
In the newborn cerebellar cortex of Plexin-B2+/− mice, granule cell proliferation is mainly confined to the upper part of the EGL, as seen with the cell proliferation marker Ki67 (Fig. 6A) or BrdU pulse labeling (animals killed 3 h after BrdU injection) (Fig. 6C). We confirmed the identity of granule cells by costaining for Pax6 (Fig. 6C). Some proliferating cells (Pax6 negative) were also observed outside the EGL and most likely correspond to precursors for molecular interneurons and oligodendrocytes. In contrast, in Plexin-B2−/− mutants, the EGL was less compact, and proliferating Pax6-positive granule cells expressing Ki67 or labeled by BrdU and Pax6 were scattered throughout the entire EGL (Fig. 6B,D) but also outside the EGL, in the cerebellar parenchyma (Fig. 6B,D). We next asked whether the scrambled architecture of the EGL in Plexin-B2 mutants would correlate with a change in the overall proliferation rate. The quantification of the number of cells expressing the M-phase marker H3P in the EGL or total cerebellum (Fig. 6E,F) showed that, at P0, the number of mitotic cells was slightly lower in Plexin-B2−/− mice (9.2 ± 0.9 H3P-positive cells/mm of EGL and 89.8 ± 4.4 H3P-positive cells/mm2) than in Plexin-B2+/− (13.8 ± 2.0 H3P-positive cells/mm of EGL and 117.9 ± 8.4 H3P-positive cells/mm2; n = 10 intermediate rostrocaudal positions; see Materials and Methods; p < 0.01). These results suggest that the presence of proliferating and mitotic cells streaming away from the EGL in newborn Plexin-B2−/− cerebellum is not attributable to an initial excessive proliferation.
Interestingly, the difference in proliferation rate between mutant and control cerebella reversed at later stages of development. In P11 Plexin-B2−/− cerebellum, there was a small but not significant increase in the number of 3 h BrdU pulse-labeled cells in the EGL (Figs. 7A, 8I). This difference in proliferation rate had significantly increased at P13 (Fig. 7C,D) and P15 (Figs. 7G,H, 8J,K). Last, to test whether increased cell proliferation at P15 is attributable to GCPs reentering into the cell cycle instead of exiting it, we determined the percentage of cells double labeled with Ki67 (expressed by cells active in the cell cycle) and BrdU (animals killed 24 h after injection) or only labeled with BrdU (Chenn and Walsh, 2002). The GCPs that had differentiated or remained dormant will only express BrdU but not Ki67. In Plexin-B2−/− cerebella, 52.6 ± 2.1% (n = 2 animals) of BrdU-labeled cells also expressed Ki67 and thus reentered the cell cycle, showing no significant difference compared with Plexin-B2+/− cerebella (58.2 ± 4.6%; n = 2 animals; p > 0.2).
Sustained proliferation of migrating granule cells in Plexin-B2 mutants
In control P11 and P13 mice, the microtubule-associated protein Dcx is expressed by postmitotic granule cells migrating tangentially in the lower EGL or migrating radially in the molecular layer (Dcx) (Fig. 7A,C). In contrast, in Plexin-B2−/− mutants, cells of the upper and lower EGL were intermingled, and many Dcx-expressing cells reached the pial surface (Fig. 7B,D). Intriguingly, in mutant cerebella, Dcx-positive cells that were also positive for the proliferation markers Ki67 (Fig. 7B,D, inset in D), H3P (Fig. 7E,F), or BrdU (data not shown) were found in the molecular layer. Moreover, at P13, rings of Ki67-labeled cells were found at the periphery of Purkinje cells islands (Fig. 8A,B,E). These proliferating cells were granule cells because they expressed Pax6 or Math1:GFP (Fig. 8E and data not shown). If they were displaced granule cell progenitors similar to those found in the EGL of wild-type and Plexin-B2+/− mice, they should highly express mRNAs for the transcription factor Barhl1 (Li et al., 2004) (Fig. 8F) and Math1 (Fig. 8C). However, Barhl1 expression at the periphery of Purkinje cell islands was low, similar to IGL cells (Fig. 8G,H). In addition, Math1-expressing cells were never detected in the cerebellar parenchyma outside of the EGL (Fig. 8D) (see also Fig. 5M). This suggests that ectopically proliferating granule cells at the periphery or Purkinje cell islands may not be equivalent to granule cell progenitors found normally in the upper EGL. The increase in granule cell proliferation and the ectopic proliferation caused by Plexin-B2 deficiency were transient. No proliferating cells were detected in the cerebellar cortex of mice at P27 or older as determined by Ki67 staining (data not shown). Last, there was also no increased cell death in the P15 cerebellum of Plexin-B2-deficient mice as determined by anti-activated caspase-3 immunostaining (8.02 ± 0.6 positive cells/mm2 in Plexin-B2+/− and 11.7 ± 2.6 positive cells/mm2 in Plexin-B2−/−; n = 2 animals; p > 0.2), suggesting that the granule cells born from ectopic proliferation were viable.
Expression and function of class 4 semaphorins in the EGL
To get a better understanding of the mechanism of action of Plexin-B2 in EGL cells, we first studied the expression patterns of its putative semaphorin ligands. So far, only class 4 semaphorins and Sema5A have been shown to bind type-B Plexins (Perrot et al., 2002; Artigiani et al., 2004; Masuda et al., 2004). However, Sema5A does not bind to Plexin-B2 (Artigiani et al., 2004). There are seven known class 4 semaphorins in vertebrates, six of which (Sema4A, 4B, 4C, 4D, 4F, and 4G) were described in rodents. We have shown previously that Sema4D is only expressed by oligodendrocytes in the postnatal brain (Moreau-Fauvarque et al., 2003) and thus may not be a Plexin-B2 ligand in the EGL. We studied the expression pattern of the five other class 4 semaphorins at P10, the peak of granule cell proliferation. Sema4A was expressed at a low level in the Purkinje layer (supplemental Fig. 4, available at www.jneurosci.org as supplemental material). Sema4B was also expressed in the Purkinje cell layer but most likely by Bergman glia cells and not by Purkinje cells (supplemental Fig. 4, available at www.jneurosci.org as supplemental material, compare with Sema4G staining of Purkinje cells). Sema4C and Sema4F were weakly expressed mostly in the lower part of the EGL and in the IGL (supplemental Fig. 4, available at www.jneurosci.org as supplemental material). Last, Sema4G was expressed by Purkinje cells. We next analyzed cerebellum organization in mice single or double deficient for the class 4 semaphorins Sema4A (Kumanogoh et al., 2005) and Sema4D (Kumanogoh et al., 2002), which are all viable and have no obvious brain defect. We first stained sections from P30 cerebellum with Nissl or antibodies against CaBP, neuronal-specific nuclear protein, and α6. Neither cerebellar defects nor ectopic granule cells were observed in Sema4A, Sema4D, and Sema4A/Sema4D double knock-outs (supplemental Fig. 4, available at www.jneurosci.org as supplemental material) (data not shown).
Plexin-B2 controls the balance between proliferation and differentiation in granule cell progenitors
During CNS development, neurons are generated from proliferating progenitor cells located in the ventricular and subventricular zones (Ohnuma and Harris, 2003). Cerebellar granule cells that arise from displaced progenitors in the EGL are the most noticeable exception (Wingate, 2005). Like other neurons, cerebellar GCs are permanently postmitotic neurons that will not reenter the cell cycle once they leave the EGL. Thus, the sequence of progenitor proliferation, neuronal migration, and differentiation are precisely choreographed, and abnormal proliferation may lead to abnormal differentiation or migration and vice versa. However, the molecular mechanisms that trigger GC differentiation and simultaneously stop the proliferation of GCPs are almost completely unknown.
A plethora of transcription factors such as Hes (hairy and enhancer of split), Numb, Zic1, Zipro1 (zinc finger proliferation 1), Pax6, Math1 and NeuroD (Wingate, 2005) control either GCP proliferation or differentiation in the EGL. These two processes are also influenced by several diffusible factors such as Sonic Hedgehog, neurotrophins, and BMPs (Lindholm et al., 1997; Alder et al., 1999). Plexin-B2 function simultaneously influences both the proliferation of GCPs and the differentiation/migration of postmitotic GCs. In postnatal Plexin-B2−/− mice, staining with markers for the cells of the upper and lower EGL reveals that the two layers are primarily intermingled and that granule cell differentiation occurs prematurely. However, most Plexin-B2−/− GCs leave the EGL, migrate inside the cerebellum, and project parallel fibers on Purkinje cell dendrites. Thus, once initiated, granule cell differentiation seems to proceed normally in Plexin-B2−/− mutants. Last, in Plexin-B2−/− mutants, although radial migration from the EGL starts too early, GCs appear to follow radial glia fibers in vivo. However, the presence of ectopic GCs in the molecular layer suggests that some GCs were unable to complete their migration to the IGL. In addition, some ectopic GCs are localized in the deep nuclei, suggesting that they were unable to stop in the IGL. The variable penetrance of the Plexin-B2 phenotype between lobules and GCs cannot be simply explained by functional redundancy between type-B Plexins, because Plexin-B1 and Plexin-B3 are not expressed in the developing cerebellar cortex (A. Chédotal and G. Kerjan, unpublished data). Therefore, in absence of Plexin-B2, the basic migration machinery is likely functional, but its regulation is probably impaired.
In addition to perturbing the timing of granule cell differentiation, Plexin-B2 deficiency affects the timing of GCP proliferation, which, in Plexin-B2−/− animals, is first lower at birth and then higher from P11 on. This shifted GCP proliferation period probably explains the abnormal cerebellar foliation observed in Plexin-B2−/− mutants (Mares and Lodin, 1970; Millen et al., 1995). More importantly, proliferating cells that express markers of differentiated GCs are found in Plexin-B2−/− mutants within the cerebellum parenchyma and around ectopic Purkinje cells. This result suggests that the regulation of GCP proliferation and granule cell differentiation can be uncoupled. In other words, in the absence of Plexin-B2, GC differentiation and migration can start independently of cell cycle exit. This is surprising because recent studies suggested that the EGL was a mitogenic niche for GCPs and that cell cycle exit might be induced by moving GCPs away from the EGL (Choi et al., 2005). The abnormal positioning of Purkinje cells contributes to the disorganization of cerebellum architecture in Plexin-B2 mutants. Purkinje cells and GCs development is interdependent. Therefore, although Purkinje cells do not express Plexin-B2, their migration and differentiation are probably altered because of the disorganization of the EGL, the premature migration of granule cells, and the invasion of Purkinje cell bodies by dividing GCs before their final alignment into a monolayer (Fig. 9).
Proliferation of differentiated granule cells in Plexin-B2-deficient mice
Although the phenotype of Plexin-B2 mutant mice is somehow reminiscent of the phenotype described in mice deficient for CXCR4 (chemokine C-X-C receptor 4) or SDF-1 (Ma et al., 1998; Zou et al., 1998; Lu et al., 2001; Vilz et al., 2005), there are important differences that do not support the conclusion that SDF-1 signaling is perturbed. First, cerebellar defects appear much earlier in SDF-1/CXCR4 mutants. Second, in SDF-1/CXCR4 mutants, the invasion of the embryonic cerebellum by ectopic GCPs is massive, whereas it is more dispersed in Plexin-B2−/− mice. More importantly, in SDF-1/CXCR4 mutants but not in Plexin-B2−/− mutants, the ectopic and proliferating GCPs express markers of EGL progenitors such as Math1. Thus, GCs that proliferate outside the EGL in Plexin-B2−/− mice differ from EGL GCPs because they do not express high levels of Math1 and Barhl1 but express markers of differentiated granule cells such as TAG-1, Sema6A, and Dcx.
Shh is the strongest known mitogen for GCPs (Dahmane and Ruiz i Altaba, 1999; Wechsler-Reya and Scott, 1999). In Plexin-B2−/− mutants, the abnormal timing of GCP proliferation together with the increased proliferation at P13–P15 could be attributable to a deregulation of Shh signaling. However, we could not detect any modification of the expression of Shh, its receptors Patched and Smoothened, or of its effectors Gli1 and Nmyc (neuroblastoma myc-related oncogene) in Plexin-B2−/− cerebella (data not shown).
Impaired neuronal differentiation and perturbed cell proliferation, in particular activation of mitosis in differentiated neurons, were proposed to lead to cell death (Yang et al., 2001; Cicero and Herrup, 2005). In mice lacking cyclin D2 (Huard et al., 1999) or NeuroD (Miyata et al., 1999), there is an increased GC death. Last, in granule cell cultures from Pax6-deficient mice (Swanson et al., 2005) cells expressing the neuronal marker βIII-tubulin appear to proliferate and die. Our results, together with recent analysis of retinoblastoma conditional knock-out mice (MacPherson et al., 2003; Wu et al., 2003), show that sustained cell cycle activation in differentiated neurons does not necessarily trigger cell death. This absence of increased cell death together with the maintenance of an apparently correct cerebellar circuitry may explain the lack of noticeable behavioral defects in Plexin-B2−/− mice.
In summary, we propose that the abnormal morphogenesis in the Plexin-B2−/− mutant cerebellum most likely arises from a premature granule cell differentiation/migration and loss of feedback inhibition of proliferation. A disruption of the proliferation/differentiation balance in the neural tube could also explain neural tube closure defects observed in Plexin-B2−/− embryos, but this will have to be explored in more detail (for review, see Copp et al., 2003).
How could Plexin-B2 control granule cell differentiation?
In the EGL, Plexin-B2 is expressed by proliferating GCPs, and its downregulation coincides with the initiation of granule cell differentiation. In Plexin-B2-deficient mice, the differentiation of GCPs could start prematurely allowing cells to initiate their inward migration before the completion of their last division. Thus, a possible model of action for Plexin-B2 is that, after binding its ligand, expressed by GCPs themselves or by neighboring cells (e.g., meninges or Bergmann glia), Plexin-B2 may block the differentiation of GCPs, by preventing the appearance of neurites and blocking the migration machinery. Sema4D, a putative ligand for Plexin-B2 (Masuda et al., 2004), is not expressed in the developing cerebellar cortex (Moreau-Fauvarque et al., 2003). We show here that six class 4 semaphorins are expressed at variable levels in the postnatal cerebellar cortex. We also show that the cerebella organization in mice deficient for Sema4A, Sema4D, or both Sema4A and Sema4D was similar to that in wild-type mice. These results show that additional ligands are involved. Their identification will require generating mice deficient for several class 4 semaphorins, and other yet unknown ligands may also be involved. Two additional physiological partners could be the tyrosine kinase receptors MET and ErbB2, both of which bind to Plexin-B2 (Giordano et al., 2002; Swiercz et al., 2004). However, a partial loss-of-function MET mouse mutant only has a weak cerebella phenotype that does not resemble the Plexin-B2 phenotype (Ieraci et al., 2002), and ErbB2 is not expressed by GCPs in the EGL (Patten et al., 2003).
Cultured granule cells rapidly exit the cell cycle unless they are allowed to reaggregate, in which case their proliferation is maintained (Gao et al., 1991). Because Plexin-B2 is slightly homophilic (Hartwig et al., 2005), the absence of Plexin-B2 might affect proliferation of granule cells indirectly by affecting their aggregation. The expanded distribution of proliferating precursors cells in Plexin-B2−/− cerebella is consistent with this interpretation.
There is mounting evidence that proteins controlling cytoskeletal dynamics play a key role during neuronal proliferation, differentiation, and migration. This is well characterized for cyclin-dependent kinase 5 and its downstream effectors (Cicero and Herrup, 2005; Kawauchi et al., 2006) and the microtubule-associated proteins Dcx and doublecortin-like kinase, all of which influence GC development (Deuel et al., 2006; Koizumi et al., 2006; Shu et al., 2006). Like all plexins, Plexin-B2 should ultimately act on the cytoskeletal structures and dynamics, presumably through regulation of GTPases (Driessens et al., 2001; Aurandt et al., 2002; Perrot et al., 2002; Swiercz et al., 2002) or integrins (Oinuma et al., 2004a,b).
Expression studies have shown that, in the developing CNS, Plexin-B1 mRNA is also highly expressed in proliferative regions (Worz-Feld et al., 2004) and often coexpressed with Plexin-B2. Thus, Plexin-B1 and Plexin-B2 function in some proliferating cells may be redundant. This could be the case for neuronal progenitors in the ventricular zone of the neocortex that seem unaffected by Plexin-B2 deficiency (A. C., unpublished data). It will be important to determine whether the other B-Plexins, Plexin-B1 and Plexin-B3, also control the balance between proliferation and differentiation in vivo.
M.T.-L. and A.C. are co-senior authors.
This work was supported by National Institutes of Mental Health Grant RO1MH60612 (Identifying Brain Wiring Mechanisms by Gene Trapping) and from National Heart, Lung, and Blood Institute (NHLBI) Grant U01HL66600 (the NHLBI–Bay Area Functional Genomics Consortium) (M.T.-L. and S. K. McConnell). A.C. was supported by the Fondation pour la Recherche sur le Cerveau and the Association pour la Recherche sur le Cancer, R.H.F was supported by the Deutsche Forschungsgemeinschaft, U.S. was supported by a fellowship from the Dr. Mildred-Scheel-Stiftung für Krebsforschung, and C.S. was supported by Spanish Ministry of Education and Science Grant SAF2004-07990). We thank S. K. McConnell (Stanford University, Stanford, CA) for support and advice, W. Havran (Scripps Research Institute, La Jolla, CA), and C. Goodnow (Australian National University, Canberra, New South Wales, Australia) for Plexin-B2 antibodies, D. Karagogeos (University of Crete, Chania, Greece) for TAG-1 antibody, F. Qiu (University of Medicine and Dentistry of New Jersey, Piscataway, NJ) for Barhl1 cDNA, A. P. McMahon for Gli1 cDNA, M. Ruat for Shh and Smoothened, and M. Wassef (École Normale Supérieure, Paris, France) for Math1 cDNA. We also thank R. Schwartzmann and V. Georget (Service d'Imagerie IFR83, Université Paris 6, Paris, France) for their help with confocal and video microscopy and X. Lu (University of Virginia, Charlottesville, VA) for the design of the targeted trapping vector.
- Correspondence should be addressed to either of the following: Dr. Alain Chédotal, Centre National de la Recherche Scientifique, Unité Mixte de Recherche 7102, Université Paris 6, 9 Quai Saint Bernard, 75005 Paris, France, ; or Dr. Marc Tessier-Lavigne, Genentech Inc., 1 DNA Way, South San Francisco, CA 94080,