Olig3 regulates early cerebellar development

The mature cerebellum controls motor skill precision and participates in other sophisticated brain functions that include learning, cognition, and speech. Different types of GABAergic and glutamatergic cerebellar neurons originate in temporal order from two progenitor niches, the ventricular zone and rhombic lip, which express the transcription factors Ptf1a and Atoh1, respectively. However, the molecular machinery required to specify the distinct neuronal types emanating from these progenitor zones is still unclear. Here, we uncover the transcription factor Olig3 as a major determinant in generating the earliest neuronal derivatives emanating from both progenitor zones in mice. In the rhombic lip, Olig3 regulates progenitor cell proliferation. In the ventricular zone, Olig3 safeguards Purkinje cell specification by curtailing the expression of Pax2, a transcription factor that suppresses the Purkinje cell differentiation program. Our work thus defines Olig3 as a key factor in early cerebellar development.

In this study, we sought to identify bHLH factors that contribute to the development of distinct cerebellar neuron types. We report here that Olig3 is crucial for generating the earliest rhombic lip and ventricular zone neuronal derivatives. Our lineage-tracing studies illustrate that the majority of DCN neurons, EGL, granule cells, as well as Purkinje cells emerge from Olig3+ progenitor cells. In contrast, few inhibitory interneurons had a history of Olig3 expression. Ablation of Olig3 results in severe cerebellar hypoplasia. In particular, we show that in Olig3 mutant mice, most DCN neurons as well as half of the EGL cells, granule cells and Purkinje cells are not formed. In contrast, supernumerary inhibitory interneurons develop in Olig3 mutant animals. Mechanistically, we show that Olig3 cell-autonomously suppresses the development of inhibitory interneurons in the ventricular zone. Olig3 is first expressed in ventricular zone progenitor cells and transiently retained in newborn Purkinje cells to curtail the expression of Pax2, a gene that we found to suppress the Purkinje cell differentiation program. We also show that Olig3 and its close family member Olig2 specify complementary Purkinje cell populations. Altogether, our data provide new insights into the molecular machinery that secures the correct development of cerebellar neurons.

Results
Olig3 is expressed in rhombic lip and ventricular zone progenitor cells during early cerebellar development About 130 bHLH transcription factors have been found in humans and 117 in mice (Skinner et al., 2010;Stevens et al., 2008). To identify candidate bHLH factors that contribute to the generation of early versus late derivatives from the rhombic lip and ventricular zone, we first analyzed the expression patterns of 110 bHLH transcription factors annotated in the Human Genome Organization (HuGO; https://www.genenames.org/data/genegroup/#!/group/420) throughout mouse cerebellar development using publically available data from the Allen Developing Mouse Brain Atlas (https:// developingmouse.brain-map.org). We found that 51 bHLH genes were expressed during cerebellar development, of which 27 were seen in progenitor niches (rhombic lip, ventricular zone, and/or EGL), and the remainder in postmitotic regions ( Figure 1A; Table 1 and Figure 1-figure supplement 1). In particular, 9/27 genes displayed differential spatial-temporal expression patterns in the rhombic lip (Atoh1 and Olig3 between E11.5-E13.5), EGL (Atoh1 and Neurod1 between E13.5-birth), and ventricular zone (Ptf1a, Ascl1, Olig3, and Olig2 between E11.5-E13.5; Ptf1a, Neurod6, Neurog1 and Neurog2 between E13.5-E18.5). The remaining (18/27) factors appeared to belong to either a common set of transcription factors expressed in all progenitor niches or they maintained their expression in a particular niche throughout cerebellar development (Table 1). Of particular interest was the expression pattern of Olig3, which has not been previously reported to have a function in cerebellar development. However, it is known to participate in the specification of defined hindbrain and spinal cord neurons (Hernandez-Miranda et al., 2017a;Liu et al., 2008;Müller et al., 2005;Storm et al., 2009;Zechner et al., 2007).
In the ventricular zone, Olig3+ cells were first seen at E11.5. Their numbers peaked by E12.5 and became rare by E14.5 ( Figure 1B; Figure 1-figure supplement 2A,B). Most Olig3+ cells (59%) in the ventricular zone co-expressed the progenitor marker Sox2 at E11.5 and E12.5, but this co-  Table 1. (B) Left, a sagittal section of the cerebellum stained against Olig3 (red) and counterstained with DAPI (blue) at E11.5. Right, quantification of Olig3+ cells in the rhombic lip (RL) and ventricular zone (VZ) between E10.5 and E14.5. See also Figure 1-figure supplement 2A. The rhombic lip and ventricular zone domains were defined in this study according to the expression of Atoh1 and Ptf1a, respectively (see Figure 1-figure supplement 2B). (C) Left, magnifications of the rhombic lip and ventricular zone of a sagittal cerebellar section stained against Olig3 (red) and Sox2 (blue) at E11.5. The boxed area displayed in the ventricular zone is illustrated to the right of the main photograph. Right, quantification of the proportion of Olig3+ cells co-expressing Sox2 between E10.5 and E14.5. Dots in the graphs represent the mean of individual analyzed animals. (D) Left, immunofluorescence characterization of rhombic lip progenitor cells stained against Olig3 (red) and Atoh1 (green) at E11.5. Right, quantification of the proportion of Olig3+ rhombic lip cells co-expressing Atoh1 between E10.5 and E14.5. (E) Left, immunofluorescence characterization of ventricular zone progenitor cells stained against Olig3 (red), Ptf1a (green), and the Purkinje cell marker Foxp2 (blue) at E11.5. The boxed area on the micrograph is illustrated to the right of the main photograph. Right, quantification of the proportion of Olig3+ ventricular zone cells co-expressing Ptf1a or Foxp2. The mean and SD are plotted in all graphs. n = 4 mice per age. Photomicrographs were acquired using the automatic tile scan modus (10% overlap between tiles) of the Zeiss LSM700 confocal microscope. The online version of this article includes the following source data and figure supplement(s) for figure 1: Source data 1. Source data for Figure 1.    localization, as well as the total number of Olig3+ cells, declined by E13.5 ( Figure 1C). Furthermore, about one-third of the BrdU+ cells in the ventricular zone co-expressed Olig3 (Figure 1-figure supplement 2D). Lastly, 52% of the Olig3+ ventricular zone cells co-expressed Ptf1a, while 41% coexpressed the postmitotic Purkinje cell marker Foxp2 ( Figure 1E). This indicates that whereas most Olig3+ cells in the ventricular zone are progenitors (Olig3+/Ptf1a+/Foxp2-), Olig3 is transiently retained in early-born Purkinje cells (Olig3+/Ptf1a-/Foxp2+). We conclude that Olig3 is expressed in rhombic lip and ventricular zone progenitor cells during the generation of their earliest neuronal derivatives.

Cerebellar hypoplasia and loss of early-born cerebellar neurons in Olig3 mutant mice
We next analyzed the consequences of Olig3 ablation on cerebellar development. At birth (P0), the cerebella of Olig3 null (Olig3 -/-) mutant mice were drastically reduced in volume when compared to control (Olig3 +/-) littermates ( Figure 3A (asterisks in A'). (B) A coronal three-dimensional reconstruction of the cerebellum. DCN nuclei are marked with asterisks. The pontine nuclei (Pn) and their axons (arrowhead), which develop from Olig3+ progenitor cells in the medulla oblongata, are labeled with Tomato. (C) A sagittal section stained against Tomato and DAPI (blue). Other known Olig3 derivatives, such as the thalamus (Th) including its projections to the cortex (Cx), pontine nuclei (Pn), inferior olive (Io) and many neurons in the medulla oblongata (Mo) are marked with Tomato. A magnification of the cerebellum is displayed with (C') or without (C'') DAPI. The external granule cell layer (EGL), Purkinje cell layer (PL), and DCN neurons are labeled with Tomato. (D) Left, schematic display of DCN nuclei positive for Brn2 (dentatus and interpositus) and Tbr1 (fastigii). Right, sagittal cerebellar sections stained against Tomato and Brn2 or Tbr1 (green). The boxed areas are illustrated to the bottom of the main photographs displaying individual and merged fluorescent signals. Asterisks mark double positive cells. (E-H) Analysis of cerebellar neurons with a history of Olig3 (bgal+) expression. Olig3 creERT2/+ ;Mapt nLacZ/+ mice were recombined with tamoxifen (TAM) at different embryonic stages and analyzed at E19. See Figure 2-figure supplement 1B for a description of the experiment. Sagittal cerebellar sections from these mice were stained against bgal (red) and markers for the EGL and granule cells (Pax6, blue in E), DCN neurons (Tbr1, blue in F), Purkinje cells (Foxp2, blue in G), and inhibitory interneurons (Pax2, blue in H). Double-positive (bgal+/marker+) cells were quantified at E19. The boxed areas on the micrographs are illustrated to the bottom of the main photographs displaying individual and merged fluorescent signals. The mean and SD are plotted in all graphs, and the dots represent the mean of individual animals. n = 3 mice per age. Significance was obtained using one-way ANOVA followed by post hoc Tukey's test, see Table 2 for statistical details. Photomicrographs were acquired using the automatic tile scan modus (10% overlap between tiles) of the Zeiss LSM700 confocal microscope. The main microphotograph displayed in G was mounted on a black frame to maintain figure panel proportions. The online version of this article includes the following source data and figure supplement(s) for figure 2: Source data 1. Source data for Figure 2.  supplement 1C). Together these data show that Olig3 is critically involved in cerebellar development and the generation of DCN neurons, EGL cells (including their granule cell derivatives) and Purkinje cells.
To define the developmental onset of the cerebellar deficiencies seen in Olig3 mutant mice, we carried out a short-term lineage-tracing experiment using a knock-in mouse strain that expresses GFP from the Olig3 locus (Olig3 GFP ) (Müller et al., 2005). We compared control (Olig3 GFP/+ ) and Olig3 mutant (Olig3 GFP/GFP ) mice at E13.5. In control mice at this stage, DCN neurons have already migrated away from the rhombic lip and accumulated in the nuclear transitory zone, EGL cells have formed their characteristic subpial layer, and Purkinje cells have completed their specification (Figure 3-figure supplement 2). Compared to control littermates, E13.5 Olig3 mutant mice showed a severe reduction in the number of DCN and EGL cells but no significant reduction in the number of Purkinje cells (Figure 3-figure supplement 2). Therefore, while the deficits observed in DCN neurons and EGL cells arise early during cerebellar development in Olig3 mutant mice, the reduction in Purkinje cell numbers occurs at a later developmental stage than E13.5 (see below). Purkinje cells (E, in red) and Pax2 + inhibitory interneurons (F, in red) in newborn control and Olig3 -/mice. All cerebellar sagittal sections were counterstained with DAPI (blue). The mean and SD are plotted in all graphs, and the dots represent the mean of individual animals. n = 4 mice per genotype in A, B, E, and F; n = 3 mice per genotype in C and D. Two-tailed t-tests were performed to determine statistical significance. See Table 2 for statistical details. Photographs in A and B were acquired with a conventional bright-field microscope and photomicrographs in C-F were acquired using the automatic tile scan modus (10% overlap between tiles) of the Zeiss LSM700 confocal microscope. The online version of this article includes the following source data and figure supplement(s) for figure 3: Source data 1. Source data for Figure 3.      We next analyzed whether Atoh1 and/or Ptf1a progenitor cell numbers changed in Olig3 mutant mice at early embryonic stages. Analysis of Olig3 mutant mice at E11.5 and E12.5 revealed reduced numbers of Atoh1+ cells in the rhombic lip, but no change in the number of Ptf1a+ cells in the ventricular zone (Figure 3-figure supplement 3A,B). We then asked whether the ablation of Olig3 might affect the proliferation of rhombic lip and ventricular zone progenitor cells and/or their viability. In the rhombic lip, the number of proliferative (BrdU+) cells was reduced in Olig3 mutant animals ( Figure 3-figure supplement 3C). However, no change in the number of Tunel+ apoptotic bodies (puncta) were seen at any of the analyzed embryonic stages (Figure 3-figure supplement 3E). In the ventricular zone of Olig3 mutant animals neither the number of BrdU+ or Tunel+ apoptotic bodies changed (Figure 3-figure supplement 3D,E). Thus, the mutation of Olig3 impairs progenitor proliferation in the rhombic lip but not in the ventricular zone, illustrating that Olig3's function in the ventricular zone differs from that in the rhombic lip.

Ablation of Olig3 misspecifies Purkinje cells that transform into inhibitory interneurons
We next compared the development of Foxp2+ Purkinje cells and Pax2+ inhibitory interneurons in Olig3 mutant mice. In wildtype and heterozygous Olig3 GFP/+ mice, Pax2+ cells first appeared at E13.5 in a rostral domain of the ventricular zone that lacked expression of Olig3, and by E14.5 occupied most of the ventricular zone ( Figure 4A). The spread of Pax2+ cells from rostral to caudal coincided with the receding of Olig3+ cells (schematically displayed in Figure 4B). In sharp contrast to wildtype and heterozygous Olig3 GFP/+ mice, we found supernumerary Pax2+ cells in Olig3 GFP/GFP mutant mice from E13.5 to P0 ( Figure 4C; Figure 4-figure supplements 1 and 2; quantified in Figure 4D). Many of the supernumerary Pax2+ cells co-expressed GFP at E13.5 and E14.5 (see magnifications in Figure 4C and Figure 4-figure supplement 2A). Thus, the ablation of Olig3 derepresses Pax2 in the early developing cerebellum.

Olig3 cell-autonomously curtails Pax2 expression to secure Purkinje cell differentiation
To experimentally assess whether Pax2 cell-autonomously suppresses Foxp2 expression to induce an inhibitory interneuron differentiation program, we electroporated in utero a Pax2-IRES-GFP-expressing vector in the ventricular zone of wildtype animals at E12.5. This is a timepoint during which Foxp2+ cells are abundant and Pax2+ cells are absent (see Figure 5A for a schematic display of the experimental conditions). Electroporated pCAG-Pax2-IRES-GFP (Pax2 overexpressing) and pCAG-GFP + Empty-IRES-GFP (control) embryos were analyzed at E14.5. In pCAG-Pax2-ires-GFP electroporated mice, no GFP+/Pax2+ cell co-expressed Foxp2, whereas in pCAG-GFP + Empty-IRES-GFP electroporated mice about 72% of the GFP+ cells were also Foxp2+ ( Figure 5B). Calcium-binding proteins are characteristic of inhibitory interneurons in the central nervous system, among which we found Parvalbumin expression to coincide with the onset of cerebellar interneuron specification and to be virtually absent in Foxp2+ cells at E14.5 ( Figure 5-figure supplement 1A). Next, we evaluated whether Pax2 electroporated cells acquire an inhibitory interneuron identity using the same conditions described above. In pCAG-Pax2-ires-GFP electroporated mice, a third of the Parvalbumin + cells co-expressed GFP, whereas in pCAG-GFP + Empty-IRES-GFP electroporated mice about 4% of the GFP+ cells were also Parvalbumin+ ( Figure 5-figure supplement 1B). We conclude that Pax2 is an efficient suppressor of Foxp2 expression and that its expression seems to induce a differentiation program characteristic of inhibitory interneurons.
To assess whether Olig3 cell-autonomously suppresses Pax2 expression, we forced the ectopic expression of Olig3 in the ventricular zone of wildtype mice at E14, a timepoint when Olig3 expression is almost absent and Pax2+ cells initiate their specification. Electroporated Olig3-IRES-GFP (Olig3 overexpressing) and pCAG-GFP + Empty-IRES-GFP (control) embryos were analyzed at E15.5. The proportion of GFP+ cells that co-expressed Pax2 was greatly reduced in the ventricular zone of Olig3-overexpressing embryos when compared to control electroporated mice ( Figure 5C, Figure 5-figure supplement 2). Thus, expression of Olig3 is sufficient to cell-autonomously suppress Pax2 expression. We conclude that during early development, Olig3 in the ventricular zone suppresses Pax2 in newborn Purkinje cells to prevent their misspecification and secure their identity.

Olig3 and Olig2 specify complementary Purkinje cell populations
Our analysis of bHLH factors expressed throughout cerebellar development showed that in addition to Olig3 and Ptf1a, Ascl1 and Olig2 are also expressed in the ventricular zone during Purkinje cell generation ( Table 1). While ablation of Ascl1 does not interfere with Purkinje cell development (Grimaldi et al., 2009;Sudarov et al., 2011), the exact role of Olig2 in the generation of GABAergic derivatives is unclear (Ju et al., 2016;Seto et al., 2014). In order to clarify the function of Olig2 we analyzed Olig2 null mutant mice and found a reduction in Purkinje cell numbers and a modest increase in inhibitory interneurons ( Figure 6-figure supplement 1A,B). We next carried out a longterm lineage-tracing experiment using Olig2 cre and Mapt nLacZ alleles (Olig2 cre/+ ;Mapt nLacZ/+ mice) to determine the contribution of Olig2 to cerebellar GABAergic neurons. This showed that while roughly half of the Foxp2+ Purkinje cell population had a history of Olig2 expression, few inhibitory interneurons were generated from Olig2+ progenitors ( Figure 6-figure supplement 1C). We thus conclude that the phenotypes of Olig3 and Olig2 mutant mice partially overlap (summarized in Figure 6A; Figure 6-figure supplement 1D).

Figure 4 continued
Olig3 expression becomes extinguished. (C) Immunofluorescence characterization of Foxp2+ (blue) Purkinje cells and Pax2+ (red) inhibitory interneurons in E13.5 control (Olig3 GFP/+ ) and Olig3 mutant (Olig3 GFP/GFP ) mice. All cerebellar sagittal sections were stained against GFP (green). Boxed areas are magnified underneath the main photographs. Note that GFP+ and Foxp2+ cells ectopically express Pax2 in Olig3 GFP/GFP mice. See Quantification of Pax2+ cells in Olig3 GFP/+ and Olig3 GFP/GFP mice at the indicated stages. (E) Quantification of the proportion of Foxp2+ Purkinje cells co-expressing Pax2+ in Olig3 GFP/+ and Olig3 GFP/GFP mice at the indicated stages. (F) Immunofluorescence characterization and quantification of the proportion of Pax2+ (blue) inhibitory interneurons co-expressing bgal (red) in E19 control (Olig3 creERT2/+ ;Mapt nLacZ/+ ) and Olig3 mutant (Olig3 creERT2/GFP ; Mapt nLacZ/+ ) mice that were recombined at E10.5. (G) Schema illustrating the above findings. In control mice, the ventricular zone generates two sets of GABAergic neurons: Foxp2+ Purkinje cells (E11.5-E13.5) and Pax2+ inhibitory interneurons (E14.5-P0). In Olig3 mutant mice, about half of the Foxp2+ cells are misspecified and co-expressed Pax2. These cells subsequently undergo a fate shift and transform into inhibitory interneurons. The mean and SD are plotted in all graphs, and the dots represent the mean of individual animals. n = 4 mice per genotype in D and E; n = 4 control mice and n = 3 Olig3 mutant mice in F. Significance was determined using a one-way ANOVA followed by post hoc Tukey's (in D and E) or two-tailed t-test (in F) analyses, see Table 2 for statistical details. Photomicrographs were acquired using the automatic tile scan modus (10% overlap between tiles) of the Zeiss spinning disk confocal microscope (in A and C) and the Zeiss LSM700 confocal microscope (in F). The online version of this article includes the following source data and figure supplement(s) for figure 4: Source data 1. Source data for Figure 4.    and Olig3 expressing vectors were electroporated at E12.5 and E14.0, respectively. Electroporated embryos were harvested at the indicated stages. Electroporated plasmids are shown. Right, a representative cerebellar section stained with GFP (green) and DAPI (blue) of an E14.5 mouse that was electroporated with control plasmids at E12.5. (B) Analysis of E14.5 wildtype mice that were electroporated at E12.5 with control (pCAG-GFP + Empty-IRES-GFP) or Pax2-overexpression (pCAG-Pax2-IRES-GFP) plasmids. Left, representative analyzed cells in the cerebellum of electroporated embryos that were stained against Pax2 (red), GFP (green), and Foxp2 (blue). Right, quantification of the proportion of GFP+ cells coexpressing Foxp2 in electroporated control (Pax2-; n = 3) and Pax2-overexpressing (Pax2+; n = 7) mice. (C) Analysis of E15.5 wildtype mice that were electroporated at E14.0 with control (pCAG-GFP + Empty-IRES-GFP) or Olig3-overexpression (pCAG-Olig3-IRES-GFP) plasmids. Left, representative analyzed cells in the cerebellum of electroporated embryos that were stained against Pax2 (red), GFP (green) and Olig3 (blue). Right, quantification of the proportion of GFP+ cells co-expressing Pax2 in electroporated control (Olig3-; n = 5) and Olig3-overexpressing (Olig3+; n = 5) mice. See Figure 5 continued on next page Olig3 and Olig2 are known to specify non-overlapping neuron populations during the development of the hindbrain and spinal cord (Takebayashi et al., 2002;Takebayashi et al., 2000). To determine whether Olig3 and Olig2 mark complementary ventricular zone progenitor cells that specify distinct Foxp2+ Purkinje cells, we stained the cerebella of E10.5-E14.5 wildtype embryos with antibodies against these two factors. We observed that roughly 58% of Olig3+ cells in the ventricular zone co-expressed Olig2 but not Foxp2, while the remaining 42% of the Olig3+ cells co-expressed Foxp2 but not Olig2 ( Figure 6B). Notably, we observed no Olig3-/Olig2+ cells that co-expressed Foxp2, illustrating that differentiated Purkinje cells retain Olig3 but not Olig2 expression.
We then asked whether ablation of Olig3 might affect the expression of Olig2 in ventricular zone progenitors, and vice versa we assessed whether mutation of Olig2 might compromise the expression of Olig3. Mutation of Olig3 did not affect the expression of Olig2, and neither did mutation of Olig2 affect the expression of Olig3 in progenitor cells of the ventricular zone ( Figure 6C and D; Figure 6-figure supplement 1E,F). These data demonstrate that the expression of Olig3 and Olig2 in ventricular zone progenitor cells is independent of the other factor. Next, we assessed whether mutation of Olig2 might also de-repress Pax2 in newborn Purkinje cells in a similar manner as the ablation of Olig3. Indeed, we observed numerous Foxp2+/Pax2+ misspecified cells in Olig2 mutant animals at E13.5 ( Figure 6E, quantified in Figure 6F), but unlike in Olig3 mutant embryos, these cells were only located in the rostral-most part of the ventricular zone (compare insets in Figure 6E). These data demonstrate that Olig2 specifically suppresses Pax2 in rostrally generated Purkinje cells, while Olig3 has a broader function in the suppression of Pax2 in most of the Purkinje cell population. We therefore conclude that Olig3 and Olig2 complementarily contribute to the correct specification of Purkinje cells by curtailing the expression of Pax2 (schematically displayed in Figure 6G).

Discussion
bHLH transcription factors are highly conserved in evolution and function as principal regulators of cell differentiation and neuronal specification (Atchley and Fitch, 1997;Baker and Brown, 2018;Ben-Arie et al., 2000;Bertrand et al., 2002;Dennis et al., 2019;Dokucu et al., 1996;Jones, 2004;Sommer et al., 1996). In this study, we sought to identify bHLH factors that regulate the specification of distinct cerebellar neuron types. We report here that Olig3 is a key player in cerebellar development and the generation of its earliest neuronal derivatives. Ablation of Olig3 results in pronounced cerebellar hypoplasia at birth and the massive loss of DCN neurons, EGL cells including their granule cell derivatives, and Purkinje cells. These deficits are accompanied by an increase in the number of inhibitory interneurons. Our data illustrate that Olig3 regulates progenitor cell proliferation in the rhombic lip, whereas in the ventricular zone Olig3 cell-autonomously suppresses the development of inhibitory interneurons by curtailing the expression of Pax2. We demonstrate that Pax2 acts as an effective suppressor of the Purkinje cells differentiation program. In addition, we show that Olig3 and its close family member Olig2 specify complementary Purkinje cell populations.
Here, we show that Olig3 is critically involved in the generation of EGL cells as well as DCN neurons. Earlier studies revealed that these rhombic lip derivatives depend on Atoh1 for their development, as loss of Atoh1 results in the severe reduction of EGL cells and impairs the development of DCN neurons (Ben-Arie et al., 1997;Gazit et al., 2004;Jensen, 2004;Machold and Fishell, 2005;Machold et al., 2011;Wang et al., 2005;Yamada et al., 2014). Our long-term lineage-tracing studies demonstrated that most EGL and DCN cells have a history of Olig3 expression, ablation of which massively reduced their cell numbers. In the early rhombic lip (E10.5-E13.5), we found that most  The mean and SD are plotted in all graphs, and the dots represent the mean of individual animals. Significance was determined using two-tailed t-tests, see Table 2 for statistical details. Photomicrographs were manually acquired using a Leica SPL confocal microscope. The online version of this article includes the following source data and figure supplement(s) for figure 5: Source data 1. Source data for Figure 5.    = 4), and Olig3 -/-(n = 4) mutant mice at E13.5. (G) Schematic summary explaining the function of Olig3 and Olig2 in the ventricular zone during the specification of Purkinje cells. Induction of Olig3 and Olig2, in committed Ptf1a+ progenitor cells, curtails the expression of Pax2 to allow for the correct specification of Purkinje cells. Olig2 predominantly operates in the rostral ventricular zone, whereas Olig3 has a broader function and is transiently retained in newborn Purkinje cells. The suppression of Pax2 is critical for Purkinje cell development, as it can override the Purkinje cell differentiation program. The mean and SD are plotted in all graphs, and the dots represent the mean of individual animals. Significance was determined using a one-way ANOVA followed by post hoc Tukey (in F) or two-tailed t-test (in C and D) analyses, see Table 2 for statistical details. Photomicrographs were acquired using the automatic tile scan modus (10% overlap between tiles) of the Zeiss spinning disk confocal microscope (in C-E) and the Zeiss LSM700 confocal microscope (in B). Figure 6 continued on next page proliferative progenitor cells (Sox2+/BrdU+) co-expressed Olig3 and a third of them co-expressed Atoh1 (Olig3+/Atoh1+ cells). This temporal window overlaps with the generation of DCN cells (Fink, 2006;Sekerková et al., 2004;Wang et al., 2005;Yamada et al., 2014), which are the most reduced neuron type in Olig3 mutant mice (this study). Thus, Olig3 is essential for DCN neuron development. Ablation of Olig3 reduced the number of BrdU+ (proliferative) progenitor cells in the rhombic lip, and consequently decreased the number of Atoh1+ cells. This impairment led to smaller numbers of EGL cells and, therefore, to fewer differentiated granule cells. The severe loss of EGL cells and their granule cell derivatives seems to largely account for the pronounced cerebellar hypoplasia observed in Olig3 mutant mice, and has also been observed after the loss of EGL cells in other studies (Ben-Arie et al., 1997). One should note, however, that the reduced numbers of granule cells in Olig3 mutant mice might be mainly independent of Olig3 and due to the reduction of instructive signals emanating from Purkinje cells, which are severely affected in these mutant mice. Indeed, available evidence shows that Purkinje cells regulate EGL proliferation and the differentiation of granule cells via sonic hedgehog signaling (Dahmane and Ruiz i Altaba, 1999;Wallace, 1999;Wechsler-Reya and Scott, 1999).
During the development of rhombomeres 2-7, there exists a dorsal progenitor domain (called dA1) that also co-expresses Olig3 and Atoh1 (Hernandez-Miranda et al., 2017a;Liu et al., 2008;Storm et al., 2009). This domain generates the mossy fiber precerebellar (pontine, lateral reticular, external cuneate) nuclei. Like in the rhombic lip, ablation of Olig3 greatly reduces the number of Atoh1+ cells in this area and their derivatives (Liu et al., 2008;Storm et al., 2009). This phenomenon also occurs in the spinal cord when Olig3 is ablated (Müller et al., 2005). Thus, Olig3 has a conserved function in the proliferation of Atoh1+ progenitor cells.
We also show that the ablation of Olig3 results in the development of supernumerary inhibitory interneurons. Both Purkinje cells and inhibitory interneurons depend on Ptf1a for their development (Hashimoto and Mikoshiba, 2003;Hoshino et al., 2005;Leto et al., 2006;Yamada et al., 2014). We predominantly found expression of Olig3 in the ventricular zone between E11.5 and E13.5, the temporal window during which Purkinje cells are specified. In the ventricular zone, about half of the Olig3+ cells co-express Ptf1a, while the rest co-express the Purkinje cell marker Foxp2. This shows that Olig3 expression is initiated in progenitors and transiently retained in newborn Purkinje cells. Ablation of Olig3 neither impaired the number of Ptf1a+ cells nor their proliferation. Strikingly, around half of newborn Purkinje cells erroneously co-expressed Pax2 in Olig3 mutant mice at E13.5, illustrating that ablation of Olig3 misspecifies newborn Purkinje cells. In Olig3 mutant mice, the number of misspecified cells declined over time and became rare by P0. This decline correlated with a parallel increase in the number of inhibitory interneurons. Thus, the primary function of Olig3 in the ventricular zone is to secure the development of Purkinje cells by cell-autonomously suppressing an alternative program that specifies inhibitory interneurons. In this context, our functional data demonstrate that forced expression of Olig3, during the temporal generation of inhibitory interneurons, is sufficient to curtail Pax2 expression. Furthermore, our functional data demonstrate that Pax2 acts as an effective suppressor of Foxp2 and the Purkinje differentiation cell program. In agreement with our findings, it was previously shown that supernumerary inhibitory neurons become specified at the expense of excitatory neurons in the hindbrain and spinal cord of Olig3 mutant mice (Müller et al., 2005;Storm et al., 2009;Zechner et al., 2007). Interestingly, there is a unique progenitor domain in rhombomere 7 (called dA4) that co-expresses Olig3 and Ptf1a, which generates the pre-cerebellar climbing fiber neurons of the inferior olive (reviewed in Hernandez-Miranda et al., 2017a). In the absence of Olig3, inferior olive neurons and many spinal cord excitatory neurons seem to change their fate and erroneously adopt an inhibitory interneuron identity (Liu et al., 2008;Müller et al., 2005;Storm et al., 2009). This suggests that inhibitory interneurons are the default neuronal type generated from the brainstem, spinal cord and the ventricular zone of the cerebellum.  Based on short-term lineage-tracing experiments, Seto et al., 2014 postulated a 'temporal identity transition' model in which Olig2+ Purkinje cell progenitors transition into inhibitory interneuron progenitors . From this model, one would expect that inhibitory interneurons would have a history of Olig2 expression. In keeping with observations made by Ju et al., 2016, our long-term lineage-tracing experiments using Olig2 cre and Olig3 creERT2 mice showed that Pax2+ inhibitory interneurons rarely have a history of Olig2 or Olig3 expression. This casts doubt on the 'temporal identity transition' model as both factors are abundantly expressed in Ptf1a+ progenitors during the specification of Purkinje cells (this work and Seto et al., 2014). Our data unambiguously show that neither Olig2 nor Olig3 control the transition of early (Purkinje) to late (inhibitory interneuron) ventricular zone progenitor cells. Rather, our work demonstrates that these factors are essential for the correct specification of Purkinje cells by curtailing an inhibitory interneuron transcriptional program.
Development of the central nervous system is characterized by molecular 'grids' of combinatorial transcription factor expression that single out distinct progenitor domains. It is from here that the enormous diversity of neuron types is generated (reviewed in Alaynick et al., 2011;Hernandez-Miranda et al., 2017a;Hernández-Miranda et al., 2010;Jessell, 2000). Here, we show that cerebellar DCN neurons and internal granule cells develop from Olig3+/Atoh1+ rhombic lip progenitor cells, whereas Purkinje cells derive from Olig3+/Ptf1a+ ventricular zone progenitors. In the mature cerebellum, DCN neurons and granule cells receive input from brainstem precerebellar mossy fiber neurons that originate from progenitor cells that co-express Olig3 and Atoh1, whilst Purkinje cells receive input from climbing fiber neurons that emerge from progenitors that co-express Olig3 and Ptf1a (Liu et al., 2008;Storm et al., 2009). The question of how these progenitor cells, located at such distant positions, acquire similar molecular signatures to specify both targets and inputs that in turn form functional cerebellar circuits remains to be elucidated.

Continued on next page
All animal experimental procedures were done in accordance to the guidance and policies of the Charite Universitatsmedizin, Berlin, Germany; Max-Delbrü ck-Center for Molecular Medicine, Berlin, Germany; and the Institute of Neuroscience, Lobachevsky University of Nizhny Novgorod, Russian Federation. Mouse strains used for this study were: Olig3creERT2 (Storm et al., 2009), Olig3GFP (Müller et al., 2005, MaptnLacZ (Hippenmeyer et al., 2005), Rosa26lsl-tdT (Madisen et al., 2010), and Olig2cre (Dessaud et al., 2007). All strains were maintained in a mixed genetic background. For tamoxifen treatment, pregnant dams were treated with tamoxifen (Sigma-Aldrich; 20 mg/ml dissolved in sunflower oil) as described previously (Hernandez-Miranda et al., 2017b;Storm et al., 2009). Tamoxifen delays labor in rodents and humans (Lizen et al., 2015). Therefore, offspring from tamoxifen-treated dams were delivered by caesarean section at E19.

Histology and cell quantifications
Immunofluorescence and tissue processing were performed as previously described (Hernández-Miranda et al., 2011). Briefly, mouse tissue (E10.5-P0) was fixed in 4% paraformaldehyde (PFA), made in phosphate buffered saline (PBS), for 3 hr at 4˚C. After fixation, brains were cryoprotected overnight in 30% sucrose in PBS, embedded and frozen in Tissue-Tek OCT (Sakura Finetek), and sectioned at 20 mm using a cryostat. Sections were washed in PBS and blocked in PBS containing 5% normal goat serum (Sigma-Aldrich) (v/v) and 0.1% Triton X-100 (v/v) (Sigma-Aldrich) at room temperature for 1 hr. They were subsequently incubated in primary antibodies at room temperature overnight. After incubation in primary antibodies, sections were washed in PBS and then incubated in secondary antibodies for 2 hr at room temperature. Primary and secondary antibodies used in this study are displayed in the Key Resources Table. For a 45 min BrdU pulse labeling, BrdU (Sigma-Aldrich) was diluted to a concentration of 16 mg/ml in saline solution and injected intraperitoneally.
Cell quantifications were performed in a non-blind manner on non-consecutive 20-mm-thick brain sections encompassing the complete lateral-medial cerebellar axis. On average six to ten sections per animal were used for quantifications. E12.5 whole-mount embryos were analysed for b-gal activity with X-gal (0.6 mg/ml; Merck Millipore, B4252) in PBS buffer containing 4 mM potassium ferricyanide, 4 mM potassium ferrocyanide, 0.02% NP-40 and 2 mM MgCl2 as previously described (Comai and Tajbakhsh, 2014). For the estimation of the cerebellar volume and area, consecutive 20-mm-thick sagittal sections were collected encompassing the whole cerebellum and stained with Nissl. Roughly 32-35 sections of the cerebellum were obtained per animal (four animals/genotype). The area of every section was measured using ImageJ; NIH, version 1.34 n. Estimation of the total volume of the cerebellum was obtained by application of Cavalieri's method (West, 2012). Fluorescence images were acquired using: (i) a Zeiss LSM 700 confocal microscope using the automatic tile scan modus (10% overlap between tiles) and assembled using ZEN2012, (ii) a Zeiss spinning disk confocal microscope using the automatic tile scan modus (10% overlap between tiles) and assembled using ZEN2012, and (iii) a Leica SPL confocal microscope. Photographs obtained with the Leica SPL confocal microscope were manually acquired and these were assembled using Image J. Unless otherwise specified all photomicrographs were acquired in a non-blind manner.

In utero electroporation
In utero electroporation was performed as previously described (Saito and Nakatsuji, 2001). Briefly, DNA plasmids were mixed with Fast Green and injected into the fourth ventricle of embryonic brains from outside the uterus with a glass micropipette. Holding the embryo in utero with forceps-type electrodes (NEPA GENE), 50 ms of 40 V electronic pulses were delivered five times at intervals of 950 ms with a square electroporator (Nepa Gene, CUY21). The plasmids used in this study are displayed in the Key Resource table. The primer sequences to clone to clone the mouse Olig3 (NM_053008.3) and Pax2 (NM_011037.5) genes are displayed in in the Key Resource table. The electroporated plasmid DNA mixtures were as follows: (i) for the control experiment, pCAG-GFP (0.5 mg ml À1 ) + pCAG-Empty-IRES-GFP (0.5 mg ml À1 ); (ii) for the Olig3 overexpression experiment, pCAG-Olig3-IRES-GFP (0.5 mg ml À1 ); and (iii) for the Pax2 overexpression experiment, pCAG-Pax2-IRES-GFP (0.5 mg ml À1 ).

Statistics
Statistical analyses were performed using Prism 8 (GraphPad). Data are plotted in scatter dot plots or column dot plots with means and standard deviations (SD) displayed. The statistical significance between group means was tested by one-way ANOVA, followed by Tukey's post hoc test (for multiple-comparison tests), or two-tailed t-test (for pair comparison tests). Degrees of Freedom as well as F and t values are provided in Table 2. No statistical method was used to pre-determine the sample size. No randomization or blinding was used for in vivo studies.