The Purkinje neuron acts as a central regulator of spatially and functionally distinct cerebellar precursors

Summary

The prospective white matter (PWM) in the nascent cerebellum contains a transient germinal compartment that produces all postnatally-born GABAergic inhibitory interneurons and astrocytes. However, little is known regarding the molecular identity and developmental potential of resident progenitors, or key regulatory niche signals. Here we show that neural stem cell-like primary progenitors (Tnc^YFP-low CD133⁺) generate intermediate astrocyte (Tnc^YFP-low CD15⁺) precursors and GABAergic transient amplifying (Ptf1a⁺) cells. Interestingly, these lineally-related, but functionally divergent progenitors commonly respond to Sonic hedgehog (Shh), and blockade of reception in TNC^YFP-low cells attenuates proliferation in the PWM, reducing both intermediate progenitor classes. Furthermore, we show that Shh produced from distant Purkinje neurons maintains the PWM niche independent of its classical role in regulating granule cell precursor proliferation. Collectively, our study indicates Purkinje neurons maintain a bi-directional signaling axis, driving the production of spatially and functionally opposed inhibitory and excitatory interneurons important for motor learning and cognition.

Introduction

The embryonic neuroepithelium is the origin of neuronal diversity during genesis of the vertebrate nervous system (Alvarez-Buylla et al., 2001). Neural stem cells (NSCs) within specialized regions of the brain neuroepithelium are capable of generating secondary germinal zones where neurogenesis continues postnatally and in some regions throughout adulthood (Doetsch, 2003; Merkle et al., 2004). These secondary neurogenic zones not only contribute to regional neuronal diversity but also maintain a steady pool of new-born neurons (Doetsch, 2003). However, the establishment of these niches and their maintenance in distinct postnatal environments is less well understood.

The cerebellum provides an attractive system to study signaling mechanisms by which diverse cell types emerge from spatially and temporally defined niches. It has a relatively simple architecture, encompassing nine different neuronal subtypes that are stereotypically situated in distinct cortical layers: the superficial molecular layer (ML), Purkinje cell layer (PCL), internal granule cell layer (IGL), and the innermost white matter (WM) (Hatten and Heintz, 1995). Moreover, functional and anatomical evidence suggests that the cerebellum, in addition to a well-recognized role in motor learning and sensory control, contributes to cognitive function (Gillig and Sanders, 2010).

Neurogenesis in the cerebellum is initiated from two primary germinal neuroepithelia, the ventricular zone (VZ) of the fourth ventricle and the upper rhombic lip (RL). All excitatory glutamatergic neurons, including granule neurons, originate in the RL. The expansion of granule neurons is achieved through proliferation of granule cell precursors (GCPs) mediated by Shh during late embryogenesis and early postnatal life (Dahmane and Ruiz i Altaba, 1999; Lewis P.M., 2004; Wallace, 1999; Wechsler-Reya and Scott, 1999). In contrast, the VZ gives rise to a subset of inhibitory GABAergic interneurons including Purkinje neurons (PNs) and deep cerebellar nuclei (DCN) during embryogenesis (Sillitoe and Joyner, 2007). Recent studies demonstrated that a transventricular Shh signaling mechanism drives the expansion of GABAergic progenitors in the embryonic VZ (Huang X. et al., 2010).

The cerebellar VZ is also the source for precursors of the secondary germinal zone residing in the prospective white matter (PWM) through the first two weeks of postnatal life in mice (Maricich SM, 1999; Sudarov et al., 2011; Yuasa, 1996). Previous retroviral-mediated lineage analysis demonstrated the presence of dividing progenitors in the PWM, contributing to both GABAergic interneurons and astrocytes (Zhang and Goldman, 1996). These late-born interneurons, called stellate and basket cells, reside in the ML where they provide inhibitory input to PNs, the sole output neurons of the cerebellar cortex. More recent genetically-inducible fate mapping (GIFM) in mice showed that NSC-like ‘astroglial cells’ in the PWM robustly generate these late-born cell types (Silbereis et al., 2009).

Astroglia express prospective NSC markers such as Nestin, Sox2 and Leukocyte cluster of differentiation 15 (CD15), and may represent the common origin for different postnatally-born cerebellar cell types generated in the PWM (Silbereis et al., 2009). However, this niche also harbors a Prominin1 (CD133)-expressing population that is capable of self-renewal in culture and clonally generates neurons and glia when implanted into neonatal cerebella, and exhibits intrinsic region character (Klein et al., 2005; Lee et al., 2005). Therefore, it is unclear whether CD15 and CD133 mark common or distinct progenitors or if they are lineally related. Furthermore, their intrinsic developmental potentials in the PWM are unclear, as are the key regulatory signals and cellular mechanisms that support the remarkable neurogenic capacity of the PWM niche.

In this study, we describe functionally divergent PWM progenitor subpopulations with distinctive molecular signatures that commonly respond to Shh. We found that Ptf1a⁺ and Tnc^YFP-Low CD15⁺ cells, which are lineally related to Tnc^YFP-Low CD133⁺ primary progenitors, function as intermediate progenitors that, respectively, expand postnatal precursor pools of GABAergic interneurons and astrocytes. Accordingly, attenuation of Shh signaling in Tnc^YFP-Low cells significantly disrupts the PWM-based expansion of these precursor pools and their progeny. Furthermore, we found that Shh from distant PNs maintains the PWM niche, independent of, GCPs. Thus, PNs couple the generation of excitatory and inhibitory interneurons, facilitating their coordinated integration into the emerging cerebellar circuit permitting their opposing neuronal actions to ultimately feedback on PNs.

Results

Shh targets distinct progenitor populations resident in the neonatal PWM

To determine whether Shh signaling is active in the neurogenic PWM niche, we used Gli1^nlacZ mice to assess Shh pathway activity (Bai C.B., 2002). In addition to previously reported expression in GCPs and Bergmann glia (Corrales JD, 2004; Lewis P.M., 2004), we observed β-galactosidase (β-gal) expression in cells abundant throughout the lobular (but not DCN) PWM at P3 and P5, and observed a similar expression pattern for Gli1 mRNA, confirming activation of Shh signaling (Figures 1A–B). To ascertain the molecular identity of this Shh-responding population, β-gal⁺ cells were quantified in four regions of interest (ROI) that delineate domains of lobular PWM with greatest frequency of β-gal⁺ cells (Figure 1A, R1–4). Because no significant regional variations were observed, ROI measurements were combined to generate a single value.

Figure 1. Distinct progenitor populations in the neonatal PWM niche respond to Shh.

(A–H′) Sagittal sections of Gli1^nlacZ cerebella showing Gli expression, represented by either X-gal staining (A, A′), or Gli1 mRNA in situ hybridization (B) or β-gal antibody staining (C–H′), in the PWM. Gli1 expression in the PWM co-localizes with NSC/astroglia markers such as Sox2 (C), BLBP (D) and CD15 (F) as well as proliferative marker Ki67 (E) at P5. It is also expressed in Ptf1a⁺ (H, H′) but not Pax2⁺ (G, G′) GABAergic progenitors. (I–L) Ptf1a⁺ cells represent a proliferative population of GABAergic progenitors (I, J), and Ptf1a^creERT2-GIFM analysis reveal that Pax2⁺ progenitors (K) but not GFAP⁺ astrocytes (L) are derived from Ptf1a⁺ progenitors. Black arrowheads indicate X-gal⁺ cells (A′) or Gli1 mRNA expressing cells (B), and white arrowheads indicate double-labeled cells. PWM, presumptive white matter. Scale bars indicate 25 μm. See also Figure S1.

We found that most β-gal⁺ cells expressed NSC/ astroglial markers Sox2 (75±11.8% at P3 and 78±9.7% at P5, n=3) and BLBP (60±7.3% at P5, n=3), whereas fewer β-gal⁺ cells expressed cell cycle marker Ki67 (44±3%, n=3) or surface antigen CD15 (Figure 1C–F). In the early neonatal PWM, Shh-responding cells are numerous and represent ~one-half of total Sox2⁺ cells (54±13.2% at P3 or 47±6.6% at P5, n=3), but this signaling appears transient in nature because β-gal⁺ cells were not detected at P6 (not shown). It is important to note that neither β-gal expression nor concurrent cellular proliferation were observed in the cerebellar VZ (Figure S1A-A″), arguing against a contribution from that neuroepithelium in the postnatal period. These data indicate that PWM NSC-like astroglia actively respond to Shh in the early postnatal period.

In contrast, Pax2⁺ GABAergic progenitors, which delineate the PWM and are largely post-mitotic (Leto et al., 2009; Maricich SM, 1999; Weisheit et al., 2006), were negative for Shh signaling (Figure 1G, G′). However, many β-gal⁺ cells expressed Ptf1a (pancreatic transcription factor 1a) (31±4% at P3, n=3, Figure 1H, H′), which genetic studies have shown is required for GABA-lineage specification (Hoshino et al., 2005; Pascual et al., 2007). Following a 2-hour BrdU pulse we noted a large fraction of Ptf1a⁺ cells in S-phase that persisted at P6 (Figure 1I). This observation was surprising given that Ptf1a⁺ cells in the embryonic cerebellum are exclusively post-mitotic (Huang X. et al., 2010). To assess whether Ptf1a⁺ cells generate Pax2⁺ cells, genetically inducible fate mapping (GIFM) experiments were performed using a Ptf1a^CreERT2 knock-in driver (Pan et al., 2013) paired with R26R^eYFP mice, to which tamoxifen (TM) was administered on P1 and P2. Some Ptf1a-GIFM cells proliferate in the PWM at P7, but most are Pax2⁺ (Figure 1J, K), confirming that Ptf1a⁺ cells emerge upstream of and contribute considerably to neonatal Pax2⁺ pools. These data provide a novel cellular mechanism supporting the rapid neonatal expansion of GABAergic progenitors pools. Long-term Ptf1a-GIFM studies revealed exclusive marking of ML GABAergic interneurons at P30, with no labeling of astrocytes or other cell types at P7 or P30(Figures 1L and S1B, C).

Shh-responding cells establish progenitors of GABAergic interneurons and astrocytes

To characterize the developmental potential of Shh-responding PWM cells, we used GIFM with the Gli1^CreER mouse, which has been shown to efficiently label Gli1⁺ cells and their progeny 24 hours following TM administration (Ahn and Joyner, 2004). TM was administered to Gli1^CreER; R26R^eYFP mice on P1 and P2 (or on P3 and P4) and the fate of YFP⁺ cells was determined at P5, P7 and P30 (Figure 2A). Because P3, P4 TM administration yielded lower YFP-labeling in marker⁺ populations (not shown), due to the transient nature of Shh signaling in the PWM, administration at P1 and P2 only was used throughout the remainder of our study. We quantified Gli1-GIFM cells by measuring YFP-labeling in four PWM ROI delineated by adjacent NeuN⁺ IGL granule neurons (Figure S2A). Because substantial regional variations in cre activity were not observed, measurements were combined into a single value. Labeling of Bergmann glia, a persistent population of Shh-responding astrocytes (Corrales JD, 2004; Lewis P.M., 2004) indicated our labeling efficiency was between 18% and 37% at P5 or P7, respectively (n=3) (Figure S2G, H, and J). As expected, labeling of proliferative GCPs was also observed (Figure S2I).

Figure 2. PWM Shh-responding cells generate GABAergic interneurons and astrocytes.

(A) Schematics indicating regions used for analysis of Gli1-GIFM⁺ cells. (B–I) Gli1^CreER; R26R^eYFP mice were injected with TM on P1 and 2 and analyzed at P5 (B, C), P7 (D, E) and P30 (F–I). Some Gli1-GIFM⁺ cells highlighted by YFP expression are proliferative at P5 (B), and some express CD15 (C), Pax2 (D), or Ptf1a (E) in the PWM and contribute to both mature GABAergic interneurons (F, G) and astrocytes (H, I). (J, K) Quantitative analyses of Gli1-GIFM studies at P5 (J) and P7 (K) showing percentage of Gli1-GIFM⁺ PWM populations. Dotted line demarcates the PWM or ML as indicated, white arrowheads indicate double-labeled cells, and asterisks mark mature granule neurons. All the analyses were performed on 5 to 10 sections per brain and with three to four animals per genotype. Average ± SEM is presented in the pie charts. Abbreviations: IGL, internal granule cell layer; WM, white matter; ML, molecular layer. Scale bars indicate 25 μm. See also Figure S2

At P5 (n=5), the majority of PWM YFP⁺ cells express Sox2 (52±4.3%), while the remaining fractions express either Ptf1a (21±5.9%) or Pax2 (31±6.9%) (Figures 2J and S2B, D). We also found fractions of Gli1-GIFM cells that were dividing, based on a 2-hour pulse with S-phase marker BrdU⁺, or that expressed surface antigen CD15 (Figure 2B, C). At P7, we detected more extensive YFP-labeling of marker⁺ populations than at P5, likely from accumulation of these cell types over a 48-hour period. However, the percentage of Sox2⁺ Gli1-GIFM cells was lower (32±5.5%), suggesting progression to lineage commitment with Gli1-GIFM cells expressing Ptf1a (23±4.2%), Pax2 (19±4.7%) and GFAP (17.3±4.4%) (n=5) (Figures 2D, E, K and S2E, F). We noted that Gli1-GIFM⁺, Pax2⁺ cells were abundant only within lobular PWM, but not in the DCN/ periventricular WM (Figures 2D and S2B, C) where a subset of GABAergic DCN interneurons are born during perinatal stages (Leto K, 2006; Maricich SM, 1999).

To establish the terminal fate of Gli1-GIFM cells, we evaluated the Gli1^CreER; R26R^eYFP cerebellar cortex at P30 (n=5). Consistent with the above progenitor labeling at P5–P7, abundant Gli1-GIFM⁺ cells were found in the ML positive for GAD-67 and Parvalbumin, markers of basket and stellate cells (Figure 2F, G). However, Gli1-GIFM studies did highlight two spatially segregated populations of astrocytes residing in the IGL and WM (Figure 2H, I), which likely represent velate protoplasmic (bushy) and smooth protoplasmic astrocytes, respectively (Yamada and Watanabe, 2002). However, we did not find that Gli1-GIFM cells contribute to MBP⁺ oligodendrocytes (Figure S2K). Collectively, these findings indicate that Shh-responding PWM cells possess relatively broad developmental potential and contribute to distinct GABAergic interneuron and astrocyte subtypes.

Low TenascinC^YFP reporter expression marks PWM Shh-responding astroglia

To further characterize the molecular identity and function of Shh-responding cells during the early neonatal period, we generated inducible Tnc^YFP-CreER mice in which CreER and YFP are expressed as a bicistronic message from the endogenous TenascinC (Tnc) locus. Previous studies have shown that the extracellular matrix glycoprotein Tnc is expressed in glia occupying the neonatal cerebellar PWM (Yuasa, 1996). Our analyses indicated that Tnc^YFP-CreER expression faithfully recapitulates the reported Tnc expression pattern, and revealed no Tnc^YFP in the VZ at postnatal stages (Figure 3A–C and S3A, A′).

Figure 3. Tnc^YFP-low expression marks Shh-responding cells capable of self-renewal in vitro.

(A–C) Sagittal sections of Tnc^YFP-creER; Gli1^nlacZ cerebella at P3 showing Tnc expression as marked by YFP co-localizes with β-gal of Gli1 reporter (A), Sox2⁺ (B) and BLBP⁺ (C) astroglia markers. (D–F) Tnc^YFP-creER; Ai9^F/+ mice were received TM on P1, P2 and were analyzed at P30. Tnc-expressing cells contribute both to Parv⁺ GABAergic interneurons (D) and GFAP⁺ astrocytes (E, F) in the cerebellum. White arrowheads indicate double-labeled cells. (G, H) Histograms from FACS showing YFP fluorescence intensity of dissociated Tnc^YFP+ 2° neurospheres (G) or dissociated P4 Tnc^YFP-CreER cerebella (H). Note that dissociated Tnc^YFP-CreER cerebella resolves into three peaks; grey indicates cells that are negative for Tnc^YFP, blue indicates low intensity corresponding to the peak derived from dissociated 2° neurospheres (G, H), and red is high intensity. (I–J) A 1° neurosphere derived from Tnc^YFP-CreER P3 cerebella expresses YFP (I′) and NSC Sox2 (J). Bright-field view is shown in (I). (K–P) Tnc^YFP-low-derived 3° neurospheres express variety of NSC markers including BLBP (K), Nestin (L), Musashi (M), CD15 (N), Glast (O) and GFAP (P). Abbreviations: IGL, internal granule cell layer; ML, molecular layer; WM, white matter. Scale bars are 25 um. See also Figure S3.

Analysis with Gli1^nlacZ revealed that, similar to Sox2, ~3/4 of β-gal⁺ cells in the PWM express Tnc^YFP (74±7% at P3, n=3) (Figure 3A). Most Tnc^YFP+ cells, if not all, expressed Sox2 or BLBP at P3 and P5 (Figure 3B, C), but not GABA progenitor marker Pax2 and rarely GFAP (Figure S3B, C). We performed Tnc-GIFM with TM administered to Tnc^YFP-CreER; Ai9^F/+ mice at P1 and P2, and analysis at P21 revealed that Tnc^YFP+ cells, similar to Gli1⁺ PWM cells, contribute extensively to a diverse array of cerebellar cell types, including basket and stellate GABAergic interneurons (Parvalbumin⁺) and astrocytes (GFAP⁺) (Figure 3D–F).

We next measured the NSC attributes of Tnc^YFP+ astroglia in vitro, and found that when cultured under standard neurosphere conditions, cells dissociated from cerebella of P3, P4 or P5 Tnc^YFP-CreER mice consistently generated 1° neurospheres (Figure 3I–J), all of which were positive for Tnc^YFP (279/279 neurospheres). FACS was used to isolate Tnc^YFP+ cells from 1° neurospheres and from freshly dissociated neonatal cerebella (Figures 3G, H and S3D–F). From the latter, two separate populations were distinguishable by Tnc^YFP fluorescence intensity: a low population (blue peak) and a high population (red peak) (Figures 3H and S3F). The low population (hereafter referred to as Tnc^YFP-low) exhibited a fluorescence peak that corresponds to that of enriched Tnc^YFP+ sphere-forming cells (Figure 3G vs. H), indicating that sphere-forming cells are a distinctive subpopulation that can be readily isolated. Further analysis of Tnc^YFP-low neurospheres, which exhibited capacity for extended self-renewal (i.e., up to 5° spheres), also revealed robust expression of numerous NSC markers (Figure 3K–P). The Tnc^YFP-high cells likely represent Bergmann glia as they express higher level Tnc^YFP when compared to the PMW astroglia in vivo (Figure S3G, H) and the same has been shown at the mRNA level (Yuasa, 1996).

Shh signaling promotes self-renewal in TNC^YFP-low cells; addition of EGF and Shh agonist SAG (Chen et al., 2002) exerted a synergistic effect on 2° neurosphere formation, eliciting an ~49% (n=3) increase in 2° neurospheres versus EGF only, and that SAG alone could not induce neurosphere formation (Figure 4A). We performed 3° neurosphere formation and again measured a robust Shh-mediated increase in self-renewal that correlated with expression of Gli1 mRNA and was attenuated by Shh antagonist KAAD cyclopamine (n=3, Figure 4B, C). Furthermore, multi-lineage differentiation capacity was tested, which showed that 3° TNC^YFP-low neurospheres can generate an extensive spectrum of cerebellar cell types based on marker expression for GABA (GABA, GAD-67, Parvalbumin and Neurogranin) and Glutamatergic lineages (Glutamate, NeuN, Tbr2 and Zic), as well as astrocytes (GFAP) and oligodendrocytes (O4) (Figure 4D–M).

Figure 4. Shh promotes self-renewal of multi-potent Tnc^YFP-low astroglia, which include molecularly distinct subpopulations.

(A–B) 2° (A) and 3° (B) neurospheres were generated from Tnc^YFP-low cerebella and neurosphere formation were quantified following treatment with 25ng/ml EGF alone, 20nM SAG w/ EGF, 10nM KAAD cyclopamine w/ EGF or SAG alone for 10 days (n=3). SAG significantly augmented neurosphere-forming capacity of EGF. Additionally, neurosphere formation is dependent on Shh signaling as Smo inhibitor KAAD-cyclopamine inhibited neurosphere formation (B). (C) RT-PCR for Gli1 and GAPDH using total RNA from treatment groups in B. (D–M) Tnc^YFP-low-derived 2° neurospheres can differentiate into diverse array of cell types including glutamatergic neurons (D–G), GABAergic neurons (I–L), oligodendrocytes (H) and astrocytes (M) when exposed to serum or PDGF for 10 days. Glutamate (Glut), Tbr2, NeuN and Zic1 are markers of glutamatergic neurons; GABA, GAD-67, Neurograinin (Neuro) are markers of GABAergic neurons; O4 and GFAP are markers of oligodendrocytes and astrocytes, respectively. (N–O) Dot plot from FACS analysis showing distribution of CD133⁺ cells amongst all Tnc^YFP cells at P3 (N), and distribution of CD133⁺ and CD15⁺ cells in the Tnc^YFP-low population at P5 (O). Note that Tnc^YFP-low, CD133⁺ cells (boxed) represents < 0.5%. (P) Gli1 is expressed in both Tnc^YFP-low, CD133⁺ and Tnc^YFP-low, CD15⁺ subpopulations as indicated by RT-PCR. GAPDH was used as an internal control and e10.5 limb bud was used as a positive control for Gli1 expression. Error bars are presented as SEM, and p value relative to control samples indicated. Scale bars indicate 25 μm. See also Figure S4.

Distinct Shh-responding astroglial subpopulations are revealed by Tnc^YFP-low expression

Because Shh-responsive, Tnc^YFP-low cells display attributes of cerebellar NSCs, we tested them for expression of CD133, an established NSC marker in the postnatal cerebellum (Lee et al., 2005). We found that only a small fraction of Tnc^YFP-low cells express CD133 at P3 using flow cytometric analysis (<0.5%) (Figure 4N, boxed in Q2). Given that a subset of Shh-responding cells also express NSC-associated surface antigen CD15 (Figure 1F), the distribution of CD133 and CD15 was evaluated amongst Tnc^YFP-low cells (Figure 4N, blue population). This combinatorial marker analysis clearly distinguished astroglial subpopulations at P5 that are either 1) Tnc^YFP-low CD133⁺ (gold), 2) Tnc^YFP-low CD15⁺ (purple), or 3) Tnc^YFP-low CD133⁻ CD15⁻ (blue) (Figure 4O, <2.5%, <0.5% or >97%, of Tnc^YFP-low cells, respectively). Consistent with our above data we found that Tnc^YFP-low CD15⁺ cells exhibit Shh pathway activation as measured by Gli1 expression, but also found that Tnc^YFP-low CD133⁺ cells respond to Shh (Figure 4P).

We next evaluated the respective NSC attributes of the above three Tnc^YFP-low astroglial subpopulations in vitro, as well as the top 20% of Tnc^YFP-high cells. Consistent with previous studies (Lee et al., 2005), TNC^YFP-low CD133⁺ cells generated neurospheres, but, to our surprise, so did Tnc^YFP-low CD15⁺ cells, TNC^YFP-low CD133⁻ CD15⁻ cells, and Tnc^YFP-high cells, which expressed the same gamut of NSC markers (Figure S4A–L). Similar to previous studies, we observed progressively limited self-renewal with each passage (1/16–1/27 neurospheres/ cells plated)(Lee et al., 2005). However, when subjected to multi-lineage differentiation, all CD133⁻ astroglial subpopulations predominantly generated astrocytes and displayed very limited neurogenic potential with no apparent GABA or glutamatergic marker expression (Figure S4M–Q). These findings suggest that all CD133⁻ cells are restricted progenitors lacking the full repertoire of NSC characteristics. Collectively, our data indicate the PWM is comprised of multiple Shh-responding, functionally and antigenically divergent progenitors, and distinguishes previously unappreciated astroglial subpopulations.

Shh maintains PWM proliferation and GABAergic progenitor expansion

We next sought to determine the specific requirement for Shh signaling in Tnc^YFP-low cells in vivo during the early neonatal period. Because Shh signaling is required for the proliferation of radial glial cells in the cerebellar VZ during embryogenesis (Huang X. et al., 2010), we decided to use inducible Tnc^YFP-CreER mice to temporally ablate Shh-responsiveness in the PMW. To gauge Shh-responsiveness in the PWM following Smoothened (Smo) deletion with consecutive TM administrations on P1 and P2, we utilized the Gli1^nlacZ reporter strain and noted qualitatively fewer β-gal⁺ cells at P3 and P4 in PWM of Tnc^YFP-CreER; Gli1^nlacZ; Smo^F/− mice versus littermate controls (Figures 5A, B and S5A, B). At P5, Tnc^YFP-low cells were isolated from cerebella of wild-type TNC^YFP-CreER and TNC^YFP-CreER; Smo^F/− littermates using FACS, and RT-PCR for Gli1 mRNA confirmed attenuation of Shh signaling in mosaic mutant TNC^YFP-low cells (Figure S5C, D).

Figure 5. Shh signaling supports proliferation and expansion of GABAergic interneuron progenitors in PWM domains.

(A–L) Analysis of Tnc^YFP-CreER and Tnc^YFP-CreER; Smo^F/− cerebella from mice injected with TM on P1, P2 and analyzed at P3 (A, B), P5 (C–F), P7 (G–J) or P30 (K, L) with either the Gli1^nlacZ (AH) or Ai9^F/+ (I–L) reporter. Gli1 expression as detected by X-gal is reduced in PWM of P3 Tnc^YFP-CreER; Smo^F/− Gli1^nlacZ; Smo^F/− cerebella when compared to the control Tnc^YFP-CreER; Gli1^nlacZ (A, B). Similarly, Brdu⁺ Sox2⁺ (C–D′), Ptf1a⁺ (E, F), Pax2⁺ (G, H, I, J) and Parv⁺ (K, L) cells are also significantly reduced in the mutants. (M–Q) Quantitative analyses of BrdU⁺, Sox2⁺ cells over total Sox2⁺ cells (M), Ptf1a⁺ cells (N), Pax2⁺ cells (O), total Pax2⁺ cells (P) and Parv⁺ cells (Q) in Tnc^YFP-CreER (grey bar) and Tnc^YFP-CreER; Smo^F/− (open bar) cerebella. All the analyses were performed on 5 to 10 sections per brain and with three or four animals per genotype. Error bars are presented as SEM, and p value relative to control samples indicated. White dotted lines demarcate the PWM, white arrowheads indicate double-labeled cells. Scale bars indicate 25 μm. See also Figure S5.

We first focused our analysis on the extensive cellular proliferation occurring in the PWM at P5 following a 2 hour BrdU pulse, at which time presumably one-half of the GABA progenitor pool has been born (Weisheit et al., 2006). A large fraction of acutely dividing cells in normal cerebellum are Sox2⁺ (Figure 5C, C′), and consistent with our finding that Shh stimulation can enhance the sphere-forming capacity of Tnc^YFP-low cells (Figure 4A, B), there was a significant reduction in the fraction of dividing Sox2⁺ cells in the PWM of the Tnc^YFP-CreERT2; Smo^F/− cerebellum (ROI I: 28.92%, ROI II: 19.55%, ROI III: 18.24%, ROI IV: 29.73%, n=3) (Figure 5C–D′, M). Reductions in proliferation were not accompanied by increased cell death (Figure S5E).

Subsequent analyses were focused on Ptf1a⁺ and Pax2⁺ cells in PWM between P5–P7, by which time the majority (75%) of the latter have been born (Weisheit et al., 2006). Reductions were apparent in these populations at P5 and P7, respectively (ROI I: 35%, ROI II: 18%, ROI III: 44%, ROI IV: 69%, n=3 and ROI I: 31%, ROI II: 30%, ROI III: 23%, ROI IV: 29%, n=4) (Figure 5E–H, N, O). When total Pax2⁺ cell number was measured (versus the regional number above), it was considerably reduced (18.31%, n=4), precluding that a failure in migration of Pax2⁺ cells contributed to reductions in regional pools (Figure 5P).

Additionally, the Ai9^F/+ reporter strain was used to more closely approximate the contribution of the Tnc-lineage to PWM Pax2⁺ progenitor pools under normal conditions (Tnc^CreERT2; Ai9^F/+) and attenuated Shh signaling (Tnc^CreERT2; Ai9^F/+; Smo^F/−). Indeed, ablation of Smo disrupted the capacity of Tnc^YFP-low cells for generating GABAergic precursors; Tnc^YFP-CreER; Ai9^F/+; Smo^F/− cerebella had substantially reduced Pax2⁺, tdTomato⁺ (62.4%) and Ptf1a⁺, tdTomato⁺ (33%) double-positive cells at P7 (per mm² of PWM) (Figure 5I, J and S5F, G). The reduction in Pax2⁺, tdTomato⁺ cells is unlikely due to lower recombination efficiency of the reporter in mutant cells because similar numbers of tdTomato⁺, Tnc^YFP+ double-positive Bergmann glia are observed between control and mutant (89.8±1% versus 92.1±0.8%, Figure S5H). However, such extensive labeling of internal control cells likely over-estimates Smo deletion. Consistent with reduced Pax2⁺/ Ptf1a⁺ cells, we observed a substantial loss (30.6%) of basket and stellate GABAergic interneurons at P30 (per mm2 of ML) (Figure 5K, L, Q) Collectively, these data demonstrate the importance of Shh signaling for the establishment of a neurogenic niche in the PWM.

Shh promotes expansion of CD15⁺ astrocyte precursors in lobular PWM domains

Because Tnc^YFP-low CD15⁺ cells were overwhelmingly gliogenic in vitro (Figure S4) we sought to further evaluate their developmental potential in vivo. Although flow cytometry indicated that CD15⁺ cells represent a small fraction of Tnc^YFP-low cells at P5 (<0.5%) (Figure 4O), we found broad distribution of CD15 in the lobular PWM at P7. We first examined whether CD15 marks oligodendrocytes or their precursors at the latter stage. We did not find convincing co-expression with MBP, which was largely restricted to core PWM (DCN) near the 4th ventricle (Figure 6A, A′). Analysis of oligodendrocyte precursor cell marker Sox10 (Stolt et al., 2002) showed no definitive overlap with CD15 in neither core nor lobular PWM (Figure 6B, C), and revealed that Sox10⁺ cells persist in the PWM of Tnc^YFP-CreER; Smo^F/− mutants (Figure 6B, C versus D, E). These data are consistent with our Gli1-GIFM, which indicated Gli1⁺ cells do not generate oligodendrocytes in the postnatal period, and with previous transplantation and chick-quail chimeric studies that showed cerebellar OPCs largely arise from extracerebellar regions during embryogenesis (Grimaldi et al., 2009; Mecklenburg et al., 2011).

Figure 6. Shh signaling promotes expansion of CD15⁺ intermediate astrocyte precursors, lineally related to CD133⁺ PWM neuronal stem cells.

(A–I) Control Tnc^YFP-CreER (A–C, F, G) and mutant Tnc^YFP-CreER; Smo^F/ (D, E, H, I) mice were injected with TM on P1, P2 and cerebella were stained with CD15 (A–F, H), Sox10 (B–E) and GFAP (F–I) at P7. CD15 expression is broadly distributed in the core and lobular PWM (A, B) and does not represent oligodendrocytes as revealed by the lack of co-localization with MBP (A′) or Sox10 (C) expression. In Tnc^YFP-CreER; Smo^F/− mutants, only the lobular (B, D) but not the core (C, E) CD15 expression is drastically reduced, whereas Sox10 expression remains unaffected (C, E). Accordingly, the expression of astrocyte marker GFAP is reduced in the mutants (F–I). Red dotted lines in G and I indicate PWM. (J) Quantitative analysis of GFAP⁺ cells in Tnc^YFP-CreER (grey bar) and Tnc^YFP-CreER; Smo^F/− mutant (open bar) cerebella. All the analyses were performed on 5 to 10 sections per brain and with four animals per genotype. Error bars are presented as SEM, and p value relative to control samples indicated. (K–P) Lineage analysis of CD133⁺ cells in the cerebellum. CD133-CreER; Ai9^F/+ mice received TM on P1 and P2, analyzed at P7 (K–M) and P21 (N–P). CD133⁺ cells contribute broadly to diverse cell types including CD15⁺ precursors (K), GFAP⁺ astrocytes (L, N), Pax2⁺ progenitors (M) and Parv⁺ GABA interneurons (O, P). Asterisks indicate endothelial cells. Scale bars indicate 25 μm.

However, this analysis did reveal that expansion of CD15⁺ cells in lobular PWM is heavily dependent on Shh signaling: in Tnc^YFP-CreER; Smo^F/− mutants CD15⁺ cells fail to accumulate in these regions (Figure 6B versus D). Interestingly, we found frequent co-expression between CD15 and GFAP (Figure 6F), suggesting that Tnc^YFP-low CD15⁺ cells may be intermediate astrocyte precursors. These data demonstrate the importance of Shh activity we detected earlier in β-gal⁺, CD15⁺ cells in Gli1^nlacZ cerebella (Figure 1F) and in FACS-purified Tnc^YFP-low CD15⁺ cells at P5 (Figure 4P). Because we did not observe β-gal⁺ cells in the core PWM in Gli1^nlacZ cerebella (at either P3, P5 or P6), we were not surprised to find that CD15 expression there was largely unaffected in Tnc^YFP-CreER; Smo^F/− mutants (Figure 6C versus E). However, in lobular PWM at P7 the accumulation of GFAP⁺ cells was markedly reduced (37% overall, n=3) in Tnc^YFP-CreER; Smo^F/− mutants (Figure 6F versus H and G versus I, J), implicating Shh as a major driver in the generation of cerebellar astrocytes via a Tnc^YFP-Low CD15⁺ intermediate progenitor.

To assess whether CD15⁺ PWM cells descend from a CD133⁺ primary progenitor, CD133-GIFM with TM administered to CD133-CreER; Ai9^F/+ mice at P1 and P2 was performed. At P7 that PWM CD15⁺ cells have descended from a CD133⁺ parent population, similar to astrocytes and Pax2⁺ cells (Figure 6K–M). Additionally, at P21, CD133-GIFM cells generate many cerebellar cell types, including ML GABAergic interneurons and various astrocyte populations (Figure 6N, O). However, we never observed that CD133-GIFM cells contribute to Bergmann glia, which are born prior to E14.5 (Sudarov et al., 2011) (Figure 6P). These data support a lineage relationship between Tnc^YFP-low CD133⁺ primary progenitors with intermediate astrocyte precursors (Tnc^YFP-low CD15⁺) and transient amplifying GABAergic progenitors (Ptf1a⁺)

PN-derived Shh signals distantly to maintain the PWM niche

We previously reported that CSF-derived Shh promotes radial glial proliferation in the embryonic cerebellar VZ (Huang X. et al., 2010). Though it is unlikely this transventricular signaling could reach the lobular PWM, cerebellar PNs express Shh and extend axons deep into the cortex by the end of embryogenesis (Lewis P.M., 2004; Sillitoe et al., 2009). Our analysis with Shh^Cre; mT/mG mice clearly indicated PN axons descend through PWM at P3 (Figure 7A). These projections bring PNs into close proximity with otherwise distant Sox2⁺ cells (Figure 7B), suggesting a plausible means of disseminating Shh over this distance. Indeed, we were able to detect puncta of Shh ligand distributed throughout the lobular PWM at P3–P4, in a pattern similar to that observed in the EGL, while the strongest signal was concentrated around the soma of Shh-producing PNs (Figure 7C–D′).

Figure 7. Purkinje neuron-derived Shh signals distantly to maintain the PWM niche.

(A, B) Sagittal sections of P3 Shh^Cre; mT/mG cerebella showing PN cell body (Calbindin), PN axons (YFP) and astroglia (Sox2) in the PWM. (C–D′) Distribution of Shh protein (green) in PWM and EGL of P3 sagittal cerebellar sections. Nuclei are counterstained with To-pro3 iodide (blue). Note that Shh protein is concentrated around the PN soma (white arrowheads). (E, F) Shh signaling as revealed by β-galactosidase antibody staining is significantly reduced in L7-cre; Shh^F/−; Gli1^nlacZ mutant cerebella when compared to of control Shh^F/+; Gli1^nlacZ at P5. Inner white dotted line demarcates PWM, whereas outer dotted line indicates Bergmann glia. (G–J) In contrast to control, L7-cre; Shh^F/− mice show drastic reduction of proliferating Sox2⁺ cells (G, G′, I, I′) and Pax2+ progenitors (H, J) at P5 following a 2hr BrdU pulse. Staining for markers indicated. PWM is delineated by white dotted lines (G, G′, I, I′) and red dotted lines (H, J). (K–N) Quantitative analyses of BrdU⁺, Sox2⁺ cells (K), Pax2⁺ cells (L, M) and Parv⁺ cells (N) in L7-cre (grey bar) and L7-cre; Shh^F/− (open bar) cerebella. All the analyses were performed on 5 to 10 sections per brain and with three animals per genotype. Error bars are presented as SEM, and p value relative to control samples indicated. Red dotted lines demarcates the PWM. Scale bars indicate 25 μm. PWM, presumptive white matter. See also Figure S6.

To determine if Shh produced by PNs does indeed activate Shh signaling to maintain the PWM niche, Shh signaling in PMW cells was robustly blocked using the PN-specific L7-Cre transgenic driver to remove Shh function starting at e17.5, without affecting embryonic VZ function (Lewis P.M., 2004). We analyzed L7-Cre; Shh^F/−; Gli1^nlacZ cerebella and observed a dramatic loss of Shh signaling as evidenced by considerable loss of PWM β-gal⁺, Sox2⁺ (double-positive) cells compared to Gli1^nlacZ littermates (Figure 7E versus F). Accordingly, the fraction of S-phase Sox2⁺ cells was significantly reduced in L7-Cre; Shh^F/− PWM at P5 (ROI I: 25%, ROI II: 47%, ROI III: 29%, ROI IV: 28%, n=3) (Figure 7G′ versus I′, K). These results indicate that PN-derived Shh ligand strongly supports a proliferative compartment in the PWM.

We next evaluated the capacity of PN-derived Shh to promote neurogenesis by measuring local Pax2⁺ populations within the PWM. Indeed, these progenitors were considerably reduced in L7-Cre; Shh^F/− mice at P5 (ROI I: 25%, ROI II: 47%, ROI III: 29%, ROI IV: 28%, n=3) (Figure 7H versus J, L). This aberration was even more pronounced for total Pax2⁺ cell number, which was dramatically reduced (46.15%, n=3, Figure 7M). These changes were not accompanied by increases in cell death. In fact, decreases in cell death were evident, possibly due to the substantial loss of GABAergic progenitors (46.15%), some of which are known to normally undergo cell death in the PWM (Yamanaka et al., 2004).

PN-derived Shh supports the generation of multiple GABAergic interneuron subtypes

Birth-dating studies have shown that PNs are generated early, from E10.5 to E12.5, whereas GABAergic interneurons are generated during late embryogenesis through the first week of postnatal life (Altman J, 1997). Consistent with this observation, PN number is largely unaffected in L7-Cre; Shh^F/− mutants (Lewis P.M., 2004) whereas the number of Parvalbumin⁺ basket and stellate cells in the ML was significantly reduced at P30 (Figure 7N and S6A, B versus C, D). Interestingly, an earlier-born GABAergic interneuron subtype, Golgi cells, which reside in the IGL and selectively express Neurogranin, were also reduced (Figure S6E, F versus G, H, J). We did not find that Golgi cells were marked by our Gli1, Ptf1a, TNC, or CD133-GIFM experiments, likely owing to the proposed perinatal birth-date for this cell type (Sudarov et al., 2011). These results indicate that PN-derived Shh is a common driver for the generation of numerous GABAergic interneuron subtypes.

Postnatal expansion of GABAergic progenitors is independent of cerebellar growth

Because deletion of Shh from PNs substantially disrupts cortical expansion of the cerebellum (Lewis P.M., 2004), this approach could impair neurogenesis in the PWM indirectly via reduction of GCPs. To resolve this issue, we ablated Shh-responsiveness specifically from proliferative, Atoh1-expressing GCPs by TM administration at P1 and P2 (Figure 8A, A′) and analyzed local Pax2⁺ pools at P7. First, as expected, the Atoh1-Cre^ER; Smo^F/− EGL exhibited substantially less proliferation (Figure 8B, B′ versus D, D′). However, despite severe cerebellar hypoplasia (Figure 7C versus E), Pax2⁺ cell numbers were not decreased (n=3, Figure 8C, C′ versus E, E′ and F). Together our findings indicate that GABA neurogenesis in the PWM is not linked to expansion of GCPs and global cerebellar growth, and that it proceeds normally when the latter processes are disrupted.

Figure 8. Cerebellar hypoplasia does not affect GABA neurogenesis in the PWM niche.

A, E′) Analysis of Atoh1-creER^T2; R26R^eYF (A-A′), Atoh1-creER (B–C′) and Atoh1-creER; Smo^F/− (D–E′) cerebella from littermates injected with tamoxifen at P1, P2 and analyzed at P7. (A-A′) Sagittal section of Atoh1-creER^T2; R26R^eYF showing efficient recombination in the EGL. (B–D′) Number of Ki67⁺ cells in the EGL (B, B′, D, D′) but not Pax2⁺ progenitors (C, C′, E, E′) are drastically reduced in Atoh1-creER; Smo^F/− when compared to Atoh1-creER control. (F) Quantification of regional Pax2⁺ cells shows the development of GABAergic interneuron progenitors are not affected in Atoh1-creER; Smo^F/− mutants. Red dotted lines demarcate the PWM. Scale bars indicate 25 μm unless otherwise stated. Error bars are presented as SEM, and p value relative to control samples indicated. (G) Model of postnatal cerebellar neurogenesis. By sending Shh bi-directionally, Purkinje neurons (PN) simultaneously modulate the numbers of both excitatory and inhibitory interneurons, which originate in distant, spatially segregated progenitor domains. This model posits that the neonatal PWM niche is comprised of lineally-related, but molecularly and functionally divergent progenitor subpopulations, which descend from a Tnc^YFP-low CD133⁺ primary progenitor. Within this niche, we show that Ptf1a⁺ GABAergic progenitors and Tnc^YFP-low CD15⁺ astrocyte precursors are proliferative. This model was conceived based on clonal analysis of CD133⁺ PWM cells transplanted into neonatal cerebella (Lee et al., 2005) in combination with our own fate-mapping of Gli1⁺, Ptf1a⁺, Tnc⁺, and CD133⁺ cells. GCP denotes granule cell precursors.

Discussion

The work presented here demonstrates that the PWM niche contains molecularly distinct subpopulations with divergent developmental potentials (broad lineage, neurogenic and astrocytic). These progenitor subpopulations commonly respond to Shh, which robustly promotes the propagation of their respective progeny. Finally, we found that PN-Shh maintains a bi-directional signaling axis that simultaneously promotes EGL-based cortical expansion and maintenance of the PWM niche. These results, together with previous studies, demonstrate that PNs are a central signaling node, modulating both inhibitory and excitatory neuronal cell numbers. Perturbations to the generation of these functionally distinct subpopulations as the result of PN dysfunction may have important implications for human neurodevelopmental disorders rooted in excitatory/ inhibitory imbalance.

Shh signals to molecularly and functionally distinct populations of PWM progenitor cells

Previous studies have suggested that astroglia marked by human GFAP promoter activity in the postnatal cerebellum contribute to both neurons and astrocytes (Silbereis et al., 2009). Our studies suggest that these astroglia are the major Shh-responding population in the PWM, whereas the cerebellar VZ no longer responds to Shh and is mitotically inactive. These astroglia not only express markers characteristic of NSCs, but their descendents, analyzed through Gli1-GIFM analysis, contribute broadly to cerebellar architecture, generating distinct GABAergic interneuron as well as astrocyte subtypes. Shh-responding astroglia are heterogeneous as they can be sorted into two populations based on Tnc^YFP reporter levels. As shown previously for Tnc mRNA expression (Yuasa, 1996), Tnc^YFP is expressed in both PCL and PWM, but not in the postnatal cerebellar VZ (Figure S3A, A′). The majority of glial cells associated with Tnc^YFP-high likely represent Bergmann glia, as they are the most abundant glial population expressing strong Tnc. Indeed, the Tnc^YFP expression profile in cerebellar sections indicated that fluorescence intensity is much higher in Bergmann glia than in PWM astroglia (Figure S3G, H).

The TNC^YFP-low population can be further separated into three molecularly distinct subpopulations based on expression of NSC-associated surface antigens, CD133 and CD15. We found that The Tnc^YFP-low CD133⁺ cells are the most broadly potent Shh-responding population in the PWM. These primary progenitors and their descendent cells, analyzed through Gli1, Ptf1a, Tnc, and CD133-GIFM analyses, contribute broadly to cerebellar architecture, generating distinct GABAergic interneurons and astrocytes. Tnc^YFP-low CD133⁺ cells display the full repertoire of NSC attributes, similar to CD133⁺/ lineage⁻ NSCs (Lee et al., 2005). Unlike these previously reported cerebellar NSCs (Lee et al., 2005), we found that Shh signaling and EGF synergize to promote self-renewal of Tnc^YFP-low cells, similar to the synergism of Shh - EGF reported for postnatal SVZ NSCs (Palma et al., 2005).

Unlike Tnc^YFP-low CD133⁺ cells, both Tnc^YFP-low CD15⁺ and Tnc^YFP-low CD133⁻ CD15⁻ populations gave rise predominantly to astrocytes in culture. Collectively, our data (including CD133-GIFM) along with clonal analysis of CD133⁺/ lineage⁻ NSCs (Lee et al., 2005), suggest these latter populations represent stages along a progenitor lineage derived from more broadly potent Tnc^YFP-low CD133⁺ primary progenitors (Figure 8G). However, additional in vivo clonal analysis is required to demonstrate that intermediate astrocyte (Tnc^YFP-low CD15⁺) and GABAergic interneuron (Ptf1a⁺) precursors are derived from a common Gli1⁺ Tnc^YFP-low CD133⁺ astroglial primary progenitor.

The neurosphere-forming capacity of Tnc^YFP-low CD133⁻ populations was unanticipated but is not surprising, because restricted progenitors can also generate neurospheres in vitro (Kondo and Raff, 2000; Temple, 2001). Our work indicates that Tnc^YFP-low CD15⁺ cells are intermediate astrocyte precursors, and are functionally divergent from CD15⁺ cells in the EGL (Ashwell and Mai, 1997; Read et al., 2009). Given that CD15 marks NSCs in the embryonic and adult SVZ (Capela and Temple, 2002; Solter and Knowles, 1978), our observation that Tnc^YFP-low CD15⁺ cells are restricted progenitors was somewhat unexpected, providing insight into the function of CD15⁺ cells in other neurogenic niches.

Though Tnc^YFP cells display two distinct reporter levels, TNC^YFP-high Bergmann glial cells are spatially isolated from the proliferative PWM niche and lack CD133 expression. Additionally, Tnc^YFP, similar to Tnc mRNA (Yuasa, 1996), is not expressed in the postnatal cerebellar VZ (Figure S3A, A′). Thus, Tnc^YFP-low cells are the only relevant population in the context of neurogenesis/ gliogenesis occurring in the postnatal PWM. Unlike Bergmann glia, these Tnc^YFP-low cells respond to Shh only transiently, suggesting Shh signaling is actively inhibited after P6. Recently, it was shown that Wnt signaling is activated in the PWM at postnatal stages (Selvadurai and Mason, 2011) and when ectopically activated Wnt signaling disrupts self-renewal in neonatal CD133⁺/ lineage⁻ NSCs (Pei et al., 2012). Therefore, it is possible that Wnt is antagonistic to Shh activity in the PWM, analogous to that proposed elsewhere for cerebellar GCPs (Lorenz et al., 2011).

Ptf1a marks transient amplifying progenitors of GABAergic interneurons

Late–born GABAergic interneurons are estimated to represent ~90% of all inhibitory interneurons in the cerebellum, and the majority of their progenitors, marked by Pax2 expression, are generated during the first week of postnatal life (Weisheit et al., 2006). Recent studies have shown that Pax2 expression is upregulated at the final progenitor cell division (Leto et al., 2009; Weisheit et al., 2006), and thus, a mechanism must exist to account for the vast numbers of Pax2⁺ cells generated during this brief window. Although Pax2⁺ cells can be traced to astroglia (Silbereis et al., 2009), it is unlikely that proliferation of astroglia alone is sufficient to expand the vast Pax2⁺ pools. Our analyses demonstrated not only that Pax2⁺ cells descend from Ptf1a⁺ cells, but also that the latter are transient, intermediate progenitors that appear to drive the rapid expansion of Pax2⁺ pools. This role for Ptf1a is distinctly different from its previously reported function in GABA lineage allocation, in which it acts to potently suppress glutamatergic fate (Hoshino et al., 2005; Pascual et al., 2007).

Purkinje neurons are a bi-directional signaling center

PNs are recognized by elaborate dendritic arbors, through which Shh is likely disseminated to the EGL, driving transient proliferation of GCPs (Lewis P.M., 2004). Our present study indicates that Shh from PNs is also necessary for maintenance of the distant PWM niche, where we propose that it targets distinct progenitor populations, including Tnc^YFP-low CD133⁺ primary progenitors, Ptf1a⁺ GABA-intermediate and Tnc^YFP-low CD15⁺ astrocyte-intermediate progenitors (Figure 8G). While PNs are a common source of Shh ligand for both the EGL and PWM, our study indicated that direct perturbation of GCP proliferation did not impact GABAergic progenitor expansion, suggesting that cortical growth has little (if any) immediate impact on maintenance of the PWM.

Several recent studies suggested an atypical mode of Shh delivery to target sites, utilizing axons as a vehicle (Brownell et al., 2011; Garcia et al., 2010; Ihrie et al., 2011; Traiffort et al., 2001). This delivery mechanism was first noted in Drosophila in which Hh is transported along retinal axons to the brain (Huang and Kunes, 1996). Our study suggests a similar mechanism of Shh delivery from PNs to PWM targets, but extends this concept further by suggesting that Shh-producing neurons can promote neurogenesis along different axes. A bi-directional signaling mechanism has also been proposed for Brain-derived neurotrophic factor (BDNF) in neurons where it is transported along neuritic processes to axons and dendrites (von Bartheld et al., 1996). Interestingly, differential neuronal activity appears to modulate BDNF release from the axon and dendrite (Matsuda et al., 2009). Future studies are required to determine the mechanism of bidirectional Shh transport to the target fields.

One intriguing question raised by our study is why would PNs be charged with the complex task of regulating the simultaneous production of both inhibitory and excitatory interneurons? Some insight may be gained from the physiological function of PNs, the sole projection neurons in the cerebellar circuit. They integrate direct excitatory input from, in part, parallel fibers of granule neurons simultaneously with inhibitory input from basket and stellate cells (and Golgi cells indirectly), and relay cerebellar output via their projections to the DCN (Sillitoe and Joyner, 2007). Because the balance between inhibitory and excitatory neurons is critical for modulating precise neuronal output from PNs, we speculate that PNs represent a sentinel population governing the formation of neural circuitry by synchronizing the balanced generation of inhibitory and excitatory interneurons. Because the cerebellum has been linked to neurodevelopmental disorders associated with excitatory/ inhibitory imbalance, such as the autism spectrum disorders (ASD), in which neuropathologies associated with PNs have been widely noted (Catini et al., 2008; Palmen et al., 2004; Pierce and Courchesne, 2001; Tsai et al., 2012), this study provides an important contribution toward a greater understanding of neural network assembly

Experimental Procedures

Animals

All experiments were performed using young neonatal and adult animals, ages P1–P30 according to the NIH and VUMC Division of Animal Care. For details of genetics strains see SI.

Tamoxifen (Sigma) was dissolved to a final concentration of 2 mg/ ml in corn oil (Sigma). Postnatal CD133-CreER; Ai9, Gli1^CreER; R26R^eYFP, Ptf1a^CreER; R26R^eYFP, Tnc^YFP-CreERT2; Ai9, Tnc^YFP-CreERT2; Smo^F/−, Atoh1-CreER^T2; Smo^F/−, Atoh1-CreER^T2; R26R^eYFP and wild-type littermates received 50 ul of tamoxifen by consecutive intraperitoneal injections on P1 and P2, or P3 and P4 as specified.

BrdU (Roche) was dissolved in PBS to a final concentration of 10 mg/ ml and was administered by intraperitoneal injection. Tnc^YFP-CreERT2; Smo^F/− and wt littermates and L7-Cre; Shh^F/− and wt littermates received a 2-hour BrdU pulse on P5 prior to sacrifice.

RNA isolation and reverse transcription

Total RNA was purified from sorted cells using the RNeasy mini kit (Qiagen) and cell homogenization was performed using QIAshredder columns (Qiagen). cDNAs were synthesized with 300ng total RNA input for all samples tested using a High-Capacity cDNA reverse transcription kit (Applied BioSciences). PCR was performed with primers for Gli1 and GAPDH at 34 cycles with annealing temperature of 60°C and as previously described (Liu et al., 2013).

Tissue processing and Immunohistochemistry/Immunocytochemistry

For animals younger than P30, brains were dissected out and fixed in 4% paraformaldehyde for either 4–6 hours or O/N at 4° C. Animals between P21–P30 received 50 ul intraperitoneal injections of Ketamine and received ice-cold PBS via transcardial perfusion followed by 4% paraformaldehyde. Brains were collected and submersion fixed in 4% paraformaldehyde O/N at 4° C. These tissues were either processed for frozen embedding in OCT compound or processed for paraffin embedding. Frozen tissues were sectioned on a Leica cryostat at 10 um, paraffin embedded tissues were cut at 5 um.

X-gal staining and transcript detection

X-gal staining for β-galactosidase activity was performed on post-fixed, frozen sections according to standard protocols. The following cDNAs were used as templates for synthesizing digoxygenin-labeled riboprobes: Gli1.

Flow Cytometry, Neurosphere and Differentiation Assays

Cerebella were isolated at P3, P4 or P5 from wild-type Tnc^YFP-CreERT2 mice and were dissociated to single cell suspensions via microdissection and trituration in sterile PBS, then strained through 40 μm filters. Cells were resuspended in 1% BSA and stained on ice for 30 min with primary antibodies, washed in sterile PBS, and resuspended with DAPI to permit exclusion of non-viable cells. Flow cytometry and cell sorting were performed on a FACSAria cell sorter (BD) to purify Tnc^YFP-low-expressing, CD133⁺ or CD15⁺ cells. Tnc^YFP− cerebella and forward scatter (FSC) were used for gating, and post-sort analysis was performed to evaluate purity of sorted cells. These populations were then cultured under standard neurosphere conditions in NSC medium [Neurobasal medium (GIBCO) supplemented with EGF (25 ng/ ml) B27 (GIBCO), N2 (GIBCO), glutamine and Pen-Strep (Cellgro)] for 7–10 days at 37°C, 5% CO₂. For quantification of 2° and 3° neurosphere formation, SAG (20 nM) and KAAD-cyclopamine (4 μM) were used to manipulate Shh signaling. Induction of differentiation was performed on 1° and 2° neurospheres derived from FACS-purified, Tnc^YFP-low cells, which were attached to PDL-coated glass coverslips and cultured for 7–10 days in NSC medium (w/o EGF) containing either PDGF-aa (10 ng/ ml) or 10% serum.

Microscopy

Bright-field images were collected on an Olympus BX51 upright microscope or on a Leica M165 FC stereoscope. Detection of double-labeling was performed using either confocal imaging on a Leica TCS SP5 laser-scanning or with an Olympus fluorescent microscope outfitted with an Optigrid system (qioptiqimaging) for optical sectioning and Metamorph^™ software (Molecular Devices) for image acquisition.

Quantifications and statistical analyses

pMetamorph^™ (molecular devices) and ImageJ software were used to measure area (mm²) for regions of interest (ROI), for acquisition of cell counts, and for surface intensity measurements. For each developmental stage, four ROIs were used per section. For all immunohistochemical stainings cell counts were obtained from a minimum of five to ten sections per brain, and all cell counts were normalized to area. A minimum n=3 animals was required for all genotypes at each stage analyzed. Sections (5 um for paraffin, 10 um for frozen) were collected in serial from midline regions covering 200–400 um of tissue (depending on developmental stage) along the medial-lateral axis. For quantification of 2° and 3° neurospheres, single cells were plated at clonal density (1–2 cells/ mm²) on 60 mm gridded cell culture dishes (SARSTEDT) and neurospheres present within a 100 mm² area were counted. Statistical analyses were performed using Prism software (GraphPad).

1. Purkinje neurons signal bi-directionally to expand functionally divergent cell types

2. Nascent white matter harbors molecularly distinct, but lineally-related astroglia

3. Shh signaling maintains neurogenic niche in the nascent white matter

4. Ptf1a expression marks transiently amplifying GABAergic progenitor pools