Rp58 is essential for the growth and patterning of the cerebellum and for glutamatergic and GABAergic neuron development

The structure and organization of the cerebellum are crucial for the proper execution of its functions in motor control and coordination and cognition. The adult mammalian cerebellar cortex contains three main cell layers (Goldowitz and Hamre, 1998; Herrup and Kuemerle, 1997; Sotelo, 2004), from outside to inside: the molecular layer, the Purkinje layer (PL) and the internal granule layer (IGL). These layers are generated throughout embryonic to postnatal stages from two main germinative zones: the ventricular zone (VZ) and the upper rhombic lip (URL) (Fig. 1A). The VZ gives rise to GABAergic neurons including Purkinje neurons (PNs), whereas the URL gives rise to all glutamatergic neurons including granule neurons (Butts et al., 2011; Carletti and Rossi, 2008). In the VZ, Ptf1a is essential for the generation of cerebellar GABAergic neurons (Hoshino et al., 2005). URL cells express Atoh1 (Math1) and give rise first to deep nuclei projection (DNP) neurons and later generate granule neuron progenitors (GNPs) (Machold and Fishell, 2005; Wang et al., 2005) (reviewed by Butts et al., 2011; Carletti and Rossi, 2008). DNP neurons are generated from URL cells that migrate along the subpial stream to form the nuclear transitory zone (NTZ) and sequentially express Pax6 and Tbr1 as they differentiate (Fink et al., 2006). Following their specification, GNPs migrate tangentially to cover the surface of the cerebellum to form the external germinal layer (EGL), where they undergo amplification during the early postnatal weeks (Chedotal, 2010). When GNPs become postmitotic they move inwards and differentiate in the IGL. The transcriptional program leading to correct differentiation of VZ and URL progenitors into postmitotic neurons, which includes regulation by Ptf1a, Atoh1 and Pax6, is still poorly understood.

Cerebellum hypoplasia is likely to result from abnormal progenitor behavior. Deletion of the distal end of human chromosome 1 (1q) is associated with cerebellar vermis hypoplasia (e.g. Boland et al., 2007; Hill et al., 2007; van Bon et al., 2008). The gene or genes contributing to this phenotype are still unknown, but a critical region containing the gene RP58 (also known as ZNF238) has recently been defined (Boland et al., 2007; Hill et al., 2007; van Bon et al., 2008). Rp58 is required for neocortex development (Okado et al., 2009; Xiang et al., 2012), acting on neuronal differentiation partly via its direct regulation of neurogenin 2 (Ngn2) and Neurod1 (Xiang et al., 2012). Our study demonstrates that Rp58 is crucial for cerebellum growth and organization and identifies Rp58 as an essential regulator of the early development of both GABAergic and glutamatergic cerebellar neurons.

Mice carrying loxP-flanked Rp58 alleles (Rp58^fl/fl) were described previously (Xiang et al., 2012). Rp58 conditional knockouts (Rp58 cKO) with Nestin-Cre (Tronche et al., 1999) were obtained from crossing Rp58^fl/+;Nestin-Cre and Rp58^fl/fl. Rp58^+/– heterozygotes (Xiang et al., 2012) were crossed to generate Rp58^–/– homozygotes. Rp58^fl/fl;Atoh1-CreER mice were bred with Rp58^fl/fl to generate Atoh1-CreER-driven mutants with tamoxifen treatment. For proliferation assays, BrdU was injected at 40 mg/kg body weight into pregnant dams 30 minutes before embryos were collected. All experimental procedures were approved by the Wistar Institute Institutional Animal Care and Use Committee (IACUC).

Embryos were dissected and fixed in 4% paraformaldehyde (PFA) on ice for up to 2 hours or overnight. Postnatal mice were perfused using 4% PFA, then post-fixed in PFA for 1 hour. Immunohistochemistry was performed as described (Xiang et al., 2012) and antibodies are listed in supplementary material Table S1.

ISH was performed as previously described (Tiveron et al., 1996). DIG-labeled antisense riboprobes were generated from linearized plasmids: Ascl1 (Lo et al., 1991), Atoh1 (Helms and Johnson, 1998), Corl2 (Minaki et al., 2008), cyclin D2 (Ciemerych et al., 2002), Neurod1 (Allen Brain Atlas, http://mouse.brain-map.org/experiment/show/75650865), Ngn1 (Ma et al., 1996), Ngn2 (Gradwohl et al., 1996), Pax2 (Nornes et al., 1990), Ptf1a (Pascual et al., 2007), Rp58 (Tatard et al., 2010) and Zic1 (Aruga et al., 1996).

These analyses were performed on cerebellar sagittal sections.

Although the cerebellum is very different in size between control and Rp58 mutants, the extent of the mediolateral axis is similar. For each cerebellum at various stages (mostly E14.5 to P2), we collected a similar number of sections/slide, which made it easier to select sections at comparable levels for Rp58 mutant and control.

Area measurements were carried out on sagittal cerebellar sections after Nissl staining. At least three sections/cerebellum located around the vermis area were analyzed with NIH ImageJ software. For each genotype, three embryos from three different litters were used.

Cell counting was performed after processing sections for immunohistochemistry. For quantification on total cerebellum, for each staining all positive cells were counted on all the serial sections/slide/cerebellum. For quantification at the vermis level, at least three sections around the vermis area per embryo were chosen and all cells positive for the specific staining were counted. In each case, we performed the quantification on three embryos of each genotype from different litters.

Student’s t-test was applied to determine the significance of the differences between controls and mutants.

Total RNA was extracted from whole E14.5 cerebellum in TRIzol reagent (Invitrogen) and reverse transcribed into cDNAs with Superscript reverse transcriptase II (Invitrogen). Real-time PCR was performed with SYBR Green probe on an ABI Prism 7000 according to the manufacturer’s parameters. Gene expression was assessed in three embryos of each genotype from three different litters with normalization to the expression of the housekeeping gene Gapdh. Primers are listed in supplementary material Table S2.

At E12.5, Rp58 is expressed in the VZ, the URL and the subpial stream (Fig. 1B). At E14.5, Rp58 expression is strong in deep cerebellar nuclei (DCN) neurons (Fig. 1C). At this stage, expression in the VZ is still present but reduces at E16.5 as Rp58 becomes expressed in progenitor cells, just beside the VZ (Fig. 1D). In the EGL, Rp58 is detected at E14.5 and at E16.5 starts to show a gradient of expression that is higher in the inner than outer EGL (Fig. 1D, supplementary material Fig. S1). Therefore, the Rp58 expression pattern suggests a role in the generation and/or differentiation of cells originating from the VZ and URL.

To decipher the role of Rp58 in cerebellum development, we used a recently generated conditional Rp58 knockout (Rp58 cKO) model (Xiang et al., 2012). Ubiquitous loss of Rp58 (Rp58 KO) causes late embryonic/early postnatal lethality (Okado et al., 2009; Xiang et al., 2012), whereas Rp58 cKO mice, obtained by crossing Rp58^fl/fl with the Nestin-Cre line, die at ∼3 weeks and display defects in neocortex development (Xiang et al., 2012). Compared with control littermates at P18, the gross morphology of mutant brains showed a much smaller cerebellum, no vermis and a decreased number of lobules (Fig. 1E,F).

The size of the Rp58 cKO cerebellum is much reduced at P2 (Fig. 1G-J) and this decrease is more pronounced at P8 (Fig. 1K-N), with the most drastic reduction being at the level of the vermis (Fig. 1K,L). At P2, cerebellar lobules became clearly discernible in the control (Fig. 1G,I) but not in the Rp58 cKO (Fig. 1H,J). At P8, few lobules were present in the mutant (Fig. 1Q,R), suggesting a delay in establishing the primary fissures. However, the single lobules in the Rp58 cKO were underdeveloped or absent in the vermis (Fig. 1L,P). Loss of Rp58 in the cerebellum also led to defective lamination, which was clearly observable at P8 (Fig. 1P,R).

Examination of cell populations in the Rp58 cKO cerebellar cortex revealed defects in its organization and lamination. The mutant PNs migrate but do not form a monolayer as in the normal cerebellum (Fig. 1S,T). Besides a reduced number of postnatal PNs in the mutant (supplementary material Fig. S2), their maturation is defective as the dendrites fail to develop (Fig. 1U,V). The Bergmann glia fibers are disorganized (Fig. 1W,X). Defective PN differentiation might also be due to the abnormal glial scaffold, as this has been shown to be important for PN dendritogenesis (e.g. Lordkipanidze and Dunaevsky, 2005). Similar to Rp58, expression of Zic1, which is normally detected in GNPs as they start to differentiate and in mature IGL neurons (Aruga et al., 1998), is highly reduced (Fig. 1Y-BB). Additionally, the Zic1 gradient observed in the control is absent from the mutant EGL (Fig. 1AA,BB), further indicating defects in GNP differentiation. These results demonstrate that Rp58 is essential for cerebellum development, acting on various cell types including the PNs and GNPs.

To investigate whether the reduction of cerebellum size is an early event in the Rp58 KO phenotype, we quantified the cerebellar area at the vermis level of E16.5 and E18.5 embryos. The Rp58 KO cerebellar area is less than half that of the control (supplementary material Fig. S3), suggesting that Rp58 is required for the generation of embryonic cerebellar cells.

The loss of embryonic cerebellar area prompted us to investigate the role of Rp58 in the generation of glutamatergic and GABAergic cerebellar neurons. First, we examined the expression of Atoh1, Pax6 and Tbr1 at E14.5. Atoh1 is detected in the URL and in the developing EGL with no apparent changes in expression levels between control and mutant brain (Fig. 2A-D,E-H). Pax6⁺ cells are detected in the URL, NTZ and VZ (Fig. 2I-L, supplementary material Fig. S4) and their EGL number is reduced by more than half in the mutant (Fig. 2Q). As they undergo differentiation in the URL at ∼E10.5, DNP progenitors downregulate Pax6 expression and begin to express Tbr1 (Fink et al., 2006). Pax6⁺ cells are detected in greater numbers in the mutant NTZ (supplementary material Fig. S4), suggesting that these cells fail to progress toward timed differentiation. Indeed, the number of Tbr1⁺ cells is severely decreased in the Rp58 KO compared with the control cerebellum (Fig. 2M-P,R).

Given the expression pattern of Atoh1 and Rp58 during glutamatergic neuron differentiation, our results suggest that Rp58, which was recently identified as a potential Atoh1 direct target (Klisch et al., 2011), is required downstream of Atoh1 for glutamatergic differentiation.

GABAergic neurons are produced from the VZ, with PNs being produced between E10.5 and E12.5 and other GABAergic interneurons between E13.5 and P7. Ptf1a is essential for GABAergic differentiation (Hoshino et al., 2005; Pascual et al., 2007). At E14.5, we observe a slight decrease in Ptf1a expression in the lateral part of the mutant cerebellum, with no apparent change of expression in the medial part (Fig. 2S-V). Corl2 (Skor2 – Mouse Genome Informatics) expression, which labels PNs as they become postmitotic (Minaki et al., 2008), is also slightly reduced in the E14.5 and E16.5 Rp58 mutant (Fig. 2AA-HH). The localization of the Corl2 expression domain indicates that the migration of PNs occurs normally following loss of Rp58 (Fig. 2EE-HH). Pax2 expression, which labels inhibitory interneurons generated from the Ptf1a⁺ population (Maricich and Herrup, 1999; Weisheit et al., 2006), is decreased at E14.5 (Fig. 2W-Z). This defect becomes stronger at E16.5 (supplementary material Fig. S5). Quantification of Pax2⁺ cells at E16.5 indicates a 20% decrease in the number of these neurons in the Rp58 KO (Fig. 2II,JJ).

These results suggest that Rp58 is also necessary, most likely downstream of Ptf1a, for GABAergic neuron development.

Because the development of the main cerebellar cortical cell types is tightly linked, we sought to delete Rp58 in a subset of cells to test its requirement in a spatiotemporal manner. To examine the defects induced by loss of Rp58 in embryonic GNPs, we chose the Atoh1-CreER line to ablate Rp58 in newly generated GNPs through administration of tamoxifen at E14.5 and E16.5 (supplementary material Fig. S6) (Machold and Fishell, 2005).

The effects of this deletion on cerebellum development were first examined at P16 (Fig. 3A-J). Nissl staining of control and Rp58 conditional mutant (Rp58^fl/fl;Atoh1-CreER) sections revealed a hypoplastic cerebellum. The size reduction at the vermis level is not as severe as that observed in the Nestin-Cre-driven Rp58 mutant and suggests that Rp58 might function prior to E14.5 in GNPs and/or in other cell lineages during vermis formation/growth. In addition to growth defects, the foliation pattern is also disrupted in the Rp58^fl/fl;Atoh1-CreER cerebellum, with the more drastic defects observed in folia I-III and VI-VIII (Fig. 3A,B). The five cardinal lobes are produced by four fissures that occur during late embryogenesis, followed by secondary fissures that establish the ten lobules observed in mice (Sillitoe and Joyner, 2007). Secondary fissures in the anterobasal lobe that establish lobules III and I-II and fissures that divide the central lobe into lobules VI to VIII failed to occur in the Rp58^fl/fl;Atoh1-CreER cerebellum (Fig. 3A,B). Furthermore, a dramatic reduction of the IGL layer was observed, indicating substantial loss of granule neurons in the Rp58^fl/fl;Atoh1-CreER cerebellum (Fig. 3A,B), which was also confirmed by a reduction in Zic1 expression (Fig. 3E,F).

Nissl staining of E19.5 sections showed that the Rp58^fl/fl;Atoh1-CreER cerebellum is smaller and presents a defective foliation pattern (Fig. 3K,L, supplementary material Fig. S7). A foliation defect is observed in a secondary division of the primary lobes into lobules: the fissure in the anterobasal lobe does not form in the mutant. Quantification of the cerebellar surface at the vermis level showed a 30% decrease in the mutant compared with control (Fig. 3W). We examined whether loss of Rp58 in Atoh1⁺ cells affects early GNP differentiation and observed that Zic1 expression is quickly reduced following Rp58 ablation (Fig. 3M-R). In addition, the size of the EGL is reduced by 20% (Fig. 3Q,R,X), suggesting a defect in GNP pool amplification.

We also explored the impact of Rp58 loss in embryonic GNPs on other cerebellar cortical cells. In the Rp58^fl/fl;Atoh1-CreER cerebella, the PNs fail to fully differentiate and align in a monolayer, in contrast to the control brain (Fig. 3G,H). Similarly, the Bergmann glia fibers are disorganized, showing defects in lamination (Fig. 3I,J). Examination of lobule X indicates that PNs still exhibit localization and differentiation defects despite a visible IGL (supplementary material Fig. S8).

These results show that Rp58 in embryonic GNPs plays a crucial role in the correct development of the cerebellum.

To examine whether proliferation defects are responsible for the reduction in the generation of glutamatergic and GABAergic neurons, we analyzed proliferation in control and KO E14.5 cerebella. BrdU incorporation was much reduced in the mutant EGL but not significantly modified in the VZ or URL (Fig. 4A,B). Proliferation defects were also observed in the Rp58^fl/fl;Atoh1-CreER EGL: phosphohistone H3 (PH3) staining, which labels mitotic cells, was strongly reduced in the mutant EGL (Fig. 3S,T,Y), together with loss of cyclin D2 expression (Fig. 3U,V). By contrast, detection of apoptotic cells by TUNEL assay or cleaved caspase 3 immunostaining did not reveal any significant difference between control and Rp58 mutant cerebella at E14.5 or P2 (supplementary material Fig. S9).

Our results suggest that defects in cell proliferation are involved in the reduction of neurons following Rp58 deletion. Furthermore, although we did not detect defects in cell apoptosis at the stages tested, we cannot exclude the possibility that apoptosis is also involved in the drastic Rp58 cKO cerebellar hypoplasia.

Rp58 directly regulates neocortical Ngn2 and Neurod1 expression (Xiang et al., 2012). We examined whether neurogenic gene expression was altered in the Rp58 KO. Ngn2 expression in the VZ is reduced but is maintained in progenitor cells as they undergo neurogenesis (Fig. 4D). A similar result is observed for Ascl1 (Fig. 4E), which is involved in specification of GABAergic neurons in the cerebellum (Grimaldi et al., 2009; Sudarov et al., 2011). Ngn1 expression is almost undetectable in the mutant GABAergic progenitors (Fig. 4C), and expression of Neurod1, which is required for GNP differentiation (Pan et al., 2009), is strongly reduced in differentiating EGL cells and also in the VZ (Fig. 4E). These results, as confirmed by quantitative RT-PCR (Fig. 4F), suggest that Rp58 acts on neuronal differentiation, and thus on cerebellum growth and patterning, by regulating, directly and/or indirectly, the transcriptional program of early cerebellar glutamatergic and GABAergic progenitors.

Our study establishes the transcription factor Rp58 as a novel essential regulator of cerebellum development. Its loss leads to cerebellar hypoplasia, defective lamination and a defective foliation pattern. Our results demonstrate that Rp58 is required for both early glutamatergic and GABAergic neuronal differentiation, placing this gene at the core of cerebellar development and neuronal differentiation. The precise mechanisms by which Rp58 orchestrates neuronal differentiation and cerebellar growth, in particular the neuronal subtype-specific transcriptional program that it controls, remain to be refined in future studies. Since human RP58 and mouse Rp58 are highly conserved and syntenic at 1qter, deletion of which causes human cerebellar hypoplasia, we propose that RP58 regulates the size of the human cerebellum.

We thank Drs Rachel Brewster, Douglas Epstein, Ellen Heber-Katz, Paul Lieberman and Sharmistha Pal for comments on the manuscript; Drs Jun Aruga, Douglas Epstein, Judith Grinspan, François Guillemot, Jane Johnson, Yuichi Ono and Piotr Sicinski for reagents; and the Wistar Microscopy Facility (James Hayden and Frederick Keeney) for the images shown in Fig. 1E,F (NIH Core Grant CA010815).