Introduction

During cerebellar development, granule cell precursors (GCPs) drive much of the massive postnatal expansion and foliation of the cerebellum (Roussel and Hatten, 2011; Sudarov and Joyner, 2007). These progenitors, defined by the expression of Atoh1, originate from the rhombic lip germinal area, migrate to form a transient external granular layer (EGL) cell population on the surface of the cerebellum, and then undergo extensive expansion through cell division, before migrating inward and differentiating into mature granule neurons (Espinosa and Luo, 2008; Machold and Fishell, 2005; Wingate and Hatten, 1999). Traditionally considered a homogeneous population, GCPs have more recently been shown to exhibit molecular and functional heterogeneity (Consalez et al., 2021; Espinosa and Luo, 2008; Li et al., 2013; Wojcinski et al., 2019). However, the biological significance of the diversity remains poorly understood.

Nestin (NES), a type VI intermediate filament protein, is commonly used to label neural progenitors derived from the ventricular zone (VZ), including the VZ of the cerebellum, its second germinal area (Buffo and Rossi, 2013; Milosevic and Goldman, 2004). Indeed, in the postnatal cerebellum under normal conditions, Nes expression is restricted to neural and glial progenitors in the white matter and Bergmann glial layer referred to as NEPs (Nes-expressing progenitors) and not detected in the EGL (Wojcinski et al., 2017; Pakula et al., 2025). However, one set of studies (Li et al., 2013, 2016) identified a rare population of Nes positive (Nes⁺) cells under the EGL that exhibit GCP-like behavior and suggested that a subset of ventricular zone derived progenitors may transiently acquire granule progenitor features, even in the absence of pathological conditions. Another study also provided evident for Atoh1-expressing cells arise from the cerebellar VZ (Khouri-Farah et al., 2022). Thus, although most studies assume that Nes expression is restricted to non-rhombic lip lineages in the developing cerebellum, emerging evidence indicates that rare Nes⁺ cells might exist near the EGL, some of which potentially correspond to a small GCP subpopulation. Whether such cells exist during normal development and how they contribute to cerebellar lineages or cancer remains unclear.

This question is particularly important given that GCPs are the cell of origin for the sonic hedgehog (SHH) subtype of medulloblastoma (SHH MB), the most common pediatric malignant brain tumor) (Roussel and Hatten, 2011; Smith et al., 2022). In mouse models, GCPs are susceptible to MB formation when mutations result in the SHH pathway remaining constitutively active. Yet, it is unclear whether all GCPs share equivalent tumorigenic potential, or whether specific subpopulations exhibit unique vulnerabilities, perhaps distinguished by their spatial localization or molecular identity.

Here, we identified and characterized a rare subpopulation of Nes-expressing (Nes⁺) GCPs within the EGL during normal postnatal cerebellar development. These cells are specifically enriched in the posterior EGL at postnatal day 1 (P1), co-express the rhombic lip derivative marker Atoh1 and exhibit a lower proliferation rate compared to Nes negative (Nes^-) GCPs. Single cell RNA-sequencing (scRNA-seq) analysis provided additional evidence that Nes⁺ GCPs are more quiescent and less differentiated than Nes^- GCPs and more closely related to GCP-like cells found in SHH MB (Tan et al., 2018; Lao et al., 2026). Using lineage tracing and SHH MB mouse models, we demonstrated that Nes⁺ GCPs are competent to give rise to SHH MB with a higher frequency than the general GCP population when targeted to express an activated form of the SHH receptor Smoothened (SMO). Furthermore, tumors arising from the Nes-lineage form preferentially in the posterior and lateral region of the cerebellum, corresponding to where SHH MB is commonly seen in mouse and human. Our findings therefore reveal a novel GCP subpopulation defined by their spatial and molecular identity, that could represent the main cell-of-origin of SHH MB.

Results

A rare subpopulation of Nes-expressing granule cell precursors is spatially restricted to the posterior cerebellum

To investigate whether Nes is expressed in GCPs within the EGL, we analyzed cerebella from Nes-CFP transgenic mice at P1 expressing CFP from promoter/enhancer sequences from the rat Nes gene (Encinas et al., 2006). Immunofluorescence staining for CFP, BARHL1 (a GCP marker), and DAPI revealed a rare population of CFP⁺ BARHL1⁺ double-positive cells located specifically within the EGL (Figure 1A-D). Higher magnification images indicated that the double-positive cells were largely absent in anterior regions of the cerebellum (Figure 1A’–D’) and highly enriched in the posterior portion of the EGL (Figure 1A’’–D’’), highlighting a spatial restriction. Quantification of serial sagittal sections across the medial-lateral axis confirmed this posterior enrichment (Figure 1E, F). Indeed, double-positive CFP⁺ BARHL1⁺ cells were almost exclusively detected in posterior lobules (folds of the cerebellum) (square bracket in Figure 1A-D). Based on systematic cell counts across half of the cerebellum (n=5 mice), there were an average of 18,189 ± 181 CFP⁺ GCPs per cerebellum at P1. These results identify a distinct posterior enriched Nes⁺ (CFP+) GCP subpopulation within the EGL (Figure 1E-F). We next examined whether these cells co-express SOX2, a marker of neural progenitors and co-expressed with Nes (CFP) in the ventricular layer, molecular layer and white matter (Wojcinski et al., 2017; Bayin et al., 2021). In the posterior EGL, a subset of CFP⁺ BARHL1⁺ cells were found to overlap with SOX2⁺ cells (Figure 1G-L). Quantification revealed that while a majority of CFP⁺ BARHL1⁺ cells were SOX2⁻ (69%), a notable proportion co-expressed SOX2 (20%) (Figure 1K-L), indicating molecular heterogeneity within the Nes⁺ GCP population. Interestingly, 11% of BARHL1⁺ cells where SOX2⁺ CFP^-.

Figure 1.

We next evaluated the proliferative status of the CFP⁺ BARHL1⁺ cells by administering a 1-hour pulse of EdU before euthanasia and stained sections for CFP, EdU and BARHL1 (Figure 1M). CFP⁺ EdU⁺ cells were observed in the posterior EGL (BARHL1⁺), confirming that some Nes-expressing GCPs are proliferative (Figure 1M’). Interestingly, quantification revealed that CFP⁺ BARHL1⁺ GCPs had a significantly lower percentage of EdU⁺ cells than their CFP⁻ BARHL1⁺ counterparts in the posterior lobules of the medial cerebellum (Figure 1N; p = 0.0149), suggesting a distinct proliferative behavior of Nes⁺ GCPs compared to the bulk of the population. Together, these data reveal a new, spatially restricted subpopulation of Nes-expressing GCPs in the posterior cerebellum that is distinct from the previously characterized CFP⁺ cells in Nes-CFP mice situated below the EGL since they express BARHL1 and they exhibit reduced proliferation compared to other GCPs.

Nes-expressing GCPs co-express Atoh1 and can be isolated by flow cytometry

We next asked whether Nes⁺ GCPs (CFP⁺) express the rhombic lip marker Atoh1 by generating Nes-CFP; Atoh1-GFP (Rose et al., 2009) double transgenic mice, allowing us to label and isolate GCPs expressing the two fluorescent markers. Using flow cytometry analysis on cerebellar cells from Atoh1-GFP; Nes-CFP P1 mice, we identified four distinct cell populations: CFP⁻ GFP⁻, CFP⁺ GFP⁻, CFP⁻ GFP⁺, and CFP⁺ GFP⁺ (Figure 2A). Quantification of each cell population revealed that the double-positive population represents only a small fraction of total cerebellar cells at P1 (1.19%), whereas most cells were double-negative or single-positive for either GFP (Atoh1⁺) or CFP (Nes⁺ NEPs) (Figure 2B). This result confirms that the Nes⁺ GCP subpopulation is rare. To validate co-expression of Nes and Atoh1 RNA, we sorted the four populations and performed RT-qPCR on each group. Expression of Atoh1, Nes, Gfp, and Cfp was measured in the single-or double-positive populations (GFP⁺ and/or CFP⁺) and compared to the double-negative cells (Figure 2C, Figure 2–Source data 1). As expected, GFP⁺ cells showed strong Atoh1 and Gfp expression but minimal Nes, whereas CFP⁺ cells expressed high levels of Nes and Cfp, but low Atoh1, indicating that the two single fluorescent populations were distinct. In contrast, the double-positive population co-expressed both Atoh1 and Nes transcripts, along with their respective fluorescent reporters (Figure 2D, Figure 2–Source data 1). Compared to double-negative cells, Atoh1 and Nes were significantly higher in the double-positive fraction, supporting the identity of the cells as a previously unrecognized rare population of GCPs at P1 that expresses both the GCP marker Atoh1 and ventricular zone marker Nes.

Figure 2.

Nes-expressing GCPs are competent to form SHH MB

To determine whether Nes⁺ GCPs have tumorigenic potential, we tested their ability to form SHH MB. Two mouse models were generated that expresses Flp recombinase from either Atoh1 or Nes promoter/enhancer sequences (Rose et al., 2009; Lao et al., 2012), along with two conditional alleles in the Rosa26 (R26) locus referred to as SmoM2 mice (R26^{FSF-GFPcre/LSL-SmoM2}). SmoM2 mice express Cre in Flp-expressing cells (R26^FSF-GFPCre where FSF refers to frt-stop-frt) (Lao et al., 2012) and an activated form of SMO (SMOM2) in Cre expressing cells (R26^LSL-SmoM2 where LSL refers to lox-stop-lox) (Mao et al., 2006) (Figure 3A). Tamoxifen was injected at P0 to induce Flp activity which induced expression of a GFPcre fusion protein which then induces expression of SmoM2 and acts as a fluorescent reporter. A conditional loss-of-function allele of Trp53 (Marino et al., 2000) was included in some mice to enhanced tumor penetrance (Figure 3B). Kaplan-Meier survival analysis revealed that Atoh1-FlpoER; R26^{/FSF-GFPcre/ LSL-SmoM2} (referred to as Atoh1-SmoM2) and Nes-FlpoER; R26 ^{FSF-Cre/ LSL-SmoM2} (referred to as Nes-SmoM2) mice both initiated tumorigenesis within 150 days upon constitutive SHH pathway activation, although Atoh1-SmoM2 mice showed a significantly shorter survival than Nes-SmoM2 mice (median survival 128.5 days vs. >150 days, p=0.0475)(Figure 3B). In the absence of Trp53 in GCPs (Atoh1-SmoM2; Trp53^fl/fl and Nes-SmoM2; Trp53^fl/fl), both genotypes exhibited similar tumor incidence and shorter survival (medial survival 63.5 days for both) and showed that loss of Trp53 strongly enhances tumor aggressiveness (Figure 3B). Histological analysis of tumors at P35, prior to overt morbidity, revealed distinct spatial patterns of lesion development based on genotype. In Atoh1-SmoM2; P53^fl/fl mice, GFP⁺ lesions were distributed broadly across the surface of the cerebellum, consistent with the widespread presence of Atoh1⁺ GCPs throughout the EGL at P1 (Figure 3C-F). However, the largest lesions were detected along the surface of the posterior lobules mainly in the lateral cerebellum. In contrast, tumors arising in Nes-SmoM2; P53^fl/fl mice exhibited more spatial restriction, with GFP⁺ lesions predominantly localized to the posterior and lateral cerebellum, corresponding to the more limited distribution of Nes⁺ GCPs (Figure G-J). These findings are similar to our previous finding that mouse SmoM2-stimulated SHH MB tumors induced with Atoh1 transgenes only arises from the lateral cerebellum (Tan et al., 2018). At end-stage, medial and lateral sections from both models had large tumors that extended into the midline (Figure 3 K-N). Together, these results demonstrate that the rare Nes⁺ GCPs are competent to form SHH-driven MB and do so with temporal spatial dynamics that are distinct from bulk Atoh1⁺ GCPs.

Figure 3.

Nes-FlpoER labels a smaller population of GCPs than Atoh1-FlpoER

To compare the efficiency and specificity of recombination between Atoh1- and Nes-driven tumor models, we analyzed Atoh1-FlpoER; R26^{LSL-tdTom/FSF-GFPcre}(Atoh1-FlpoER; tdTom reporter) and Nes-FlpoER; R26^{LSL-tdTom/FSF-GFPcre} (Nes-FlpoER; tdTom reporter) (Daigle et al., 2018) mice at P2 following tamoxifen induction at P0. Immunofluorescence for RFP (tdTomato) revealed that the Nes-FlpoER line labeled fewer cells in the EGL than the Atoh1-FlpoER line, with recombination largely restricted to the posterior cerebellum (Figure 4A–H). Quantification of RFP⁺ cells across sagittal sections showed an average of approximately 2,920 ± 4102 cells per cerebellum in the Nes-FlpoER model compared to 25,480 ± 953 cells in the Atoh1-FlpoER model, representing an approximate ten-fold higher targeting efficiency using Atoh1-FlpoER (Figure 4I–J). These results confirm that the Nes-FlpoER driver targets a rare and spatially restricted subpopulation of GCPs within the developing cerebellum.

Figure 4.

Nes-FlpoER labels Atoh1-expressing cells capable of forming SHH MB

To validate that the targeted population corresponds to Nes⁺ GCPs and to reassess their ability to initiate MB, we developed an intersectional dual-recombinase model (Nes-FlpoER; Atoh1^FSF-Cre/+; R26^LSL-SmoM2/+) (Ruffault et al., 2015). In this model, oncogenic activation of SHH signaling occurs only in GCPs co-expressing Atoh1 and Nes. Following tamoxifen induction at P0, these mice developed SHH MB with a latency and penetrance that seems comparable to those observed in the single-driver models (Figure 4–figure supplement 1A). Immunofluorescence analysis of Atoh1^FSF-Cre/+; Nes-FlpoER; R26^LSLtdTom/+ cerebella collected at P4 confirmed RFP labeling within the EGL that displayed a pattern consistent with the spatial distribution of Nes⁺ GCPs (Figure 4–figure supplement 1B–E). Together, these findings validate that the Nes-FlpoER model effectively and selectively targets a minor subset of GCPs that co-express Nes and Atoh1 and demonstrate that this restricted population is fully competent to initiate SHH-driven MB. In contrast, the Nes-CreER transgene used in many publications to model SHH MB does not match the endogenous Nes expression, since it targets many more GCPs in the EGL and throughout the anterior and posterior cerebellum compared to our Nes-FlpoER transgene, as shown by analyzing Nes-CreER; R26^LSL-tdTom/+ mice (Figure 4–figure supplement 2).

Nes-expressing GCPs exhibit a transcriptome associated with rhombic lip cells and quiescent stem cell identity

To characterize the molecular identity of the Nes⁺ GCP population we have identified, single-cell RNA-seq analysis was performed on five samples of P1 cerebella. Four independent samples were enriched for posterior GCPs using a Percoll gradient method and isolating the tissues from the posterior cerebellar region of individual Nes-CFP mice (Figure 5–figure supplement 1A-C). A fifth sample contained CFP⁺ GFP⁺ cells isolated by FACS (fluorescence activated cell sorting) from the cerebella of 3 Atoh1-GFP; Nes-CFP mice. After filtering out low-quality cells (see Materials and Methods), the integration of 27,037 cells (22,961 posterior GCP-enriched samples and 4,076 CFP⁺ GFP⁺ cells) revealed 29 transcriptionally distinct clusters, including 12 GCP clusters expressing Barhl1 (c0, c1, c2, c3, c7, c8, c9, c10, c13, c16 and c23) and 10 clusters with less cells expressing Nes (c4, c12, c14, c15, c17, c18, c19, c20, c21 and c27) that correspond to the ventricular zone-derivative NEPs (Figure 5A–C, Figure 5-source data 1; Figure 5-figure supplement 1D; Figure 5-figure supplement 1-source data 1). Additional clusters 11 and 22 represent unipolar brush cells, and five additional clusters likely represent degraded cells (5), fibroblasts (22), oligodendrocytes (24), microglia (25) and endothelial cells (26).

Figure 5.

The clusters containing GCPs were re-clustered producing 15 clusters, of which one transcriptionally distinct population, cluster 10, co-expressed Nes and Atoh1 and the mouse line reporters Cfp and Gfp (Figure 5D-F, Figure 5-figure supplement 1E; Figure 5-Source data 2). This cluster also expressed the proliferation marker Mki67, the SHH target gene Gli1 and posterior GCP marker Otx2 (Figure 5F-I; Figure-figure supplement 1E; Figure 5-Source data 2). Cell cycle phase analysis showed that cells at all phases were present in cluster 10 (G1, G2M, S) (Figure 5J, Figure 5-Source data 3). Three neighboring clusters (1, 3, and 6) appeared to display a close transcriptional similarity to cluster 10, including the expression of Otx2 and Tlx3 which confirms these GCPs are localized to the posterior cerebellum (lobules 8-10) that contains the anatomical domain of Nes⁺ GCPs (Figure 5I and Figure 5–figure supplement 1E,F; Figure 5-Source data 2) (Divya et al., 2016; El Nagar et al., 2018).

Based on the similarity of the clusters, we compared the transcriptional profile of cluster 10 to the three posterior GCP clusters (Figure 5K). Comparison of cluster 10 (741 cells) with GCP clusters 1, 3, and 6 (4,066 cells) using the FindMarkers approach identified 1,586 genes upregulated in cluster 10 and 3,130 downregulated (Figure 5-Source data 4) and using the more stringent Pseudobulk (Libra) analysis 1,983 upregulated and 1,876 downregulated genes (Figure 5L, Figure 5-Source data 5). Genes enriched in cluster 10 included Sox2, Nes, Wls, and Wnt1, while Neurod1, and Rbfox3 were preferentially expressed in the other GCP clusters (1, 3, 6), indicating that cluster 10 represents a less differentiated, more stem-like GCP state, potentially positioned upstream in the lineage hierarchy.

Gene Ontology revealed that cluster 10 is enriched for biological processes linked to WNT signaling whereas genes associated with neuronal differentiation, synapse formation, and morphogenesis were downregulated (Figure 5M,N, Figure 5-Source data 6, 7). This enrichment of WNT-associated pathways and up regulation of Wls and Wnt1 (Figure 5L), together with the fact that differentiation pathways are down regulated, suggests a transcriptional resemblance to rhombic lip progenitors. This molecular profile supports their identity as a transient, developmentally plastic population within the GCP lineage related to the rhombic lip.

Nes-expressing GCPs are transcriptionally closer to GCP-like cells in SHH MB tumors than other GCPs

To further assess the relationship between the Nes-expressing normal GCPs and the GCP-like cells in SHH MB tumors, we integrated the cells of clusters 10, 1, 3 and 6 from our GCP cells (Figure 5K) with the GCP-like tumor cells from SmoM2 early tumors in a dataset we reported in Lao et al., 2026 (Lao et al., 2026) (Figure 6A-B, Figure 6- - data 1). We then performed pseudobulk differential expression analyses comparing GCP-like tumor cells with three GCP populations: all GCPs in the four clusters, Nes^- GCPs and Nes⁺ GCPs (Figure 6C-H).

Figure 6.

Comparison of all GCPs with SmoM2 GCP-like tumor cells identified 2,680 and 2,668 significantly differentially downregulated and upregulated genes, respectively (Figure 6C-D, Figure 6-Source data 2). Comparison between Nes- GCPs and tumor cells identified 3,617 significantly downregulated genes and 4,083 significantly upregulated genes (Figure 6E-F, Figure 6-Source data 3). In contract, a comparison between Nes⁺ GCPs and tumor GCP-like cells yielded substantially fewer differentially expressed genes, with 1,724 and 1,619 significant downregulated and upregulated genes, respectively (Figure 6G-H, Figure 6–Source data 4). Thus, among the three GCP population analyzed, Nes-expressing GCPs displayed the smallest transcriptional difference from the SmoM2 tumor cells. These data indicate that the Nes-expressing GCP population in the normal cerebellum is transcriptionally closer to SHH MB tumor cells than the remainder of the GCPs (clusters 1, 3, 6).

Discussion

In this study, we identify a rare subpopulation of Nes-expressing GCPs in the normal developing cerebellum and showed that these cells are spatially restricted, molecularly distinct from other GCPs, and highly competent to form SHH MB. Although GCPs have long been considered the cells of origin of SHH MB, our findings refine this view by demonstrating that not all GCPs are equivalent in their propensity to form tumors. Instead, a minor Nes-expressing subset of GCPs within the EGL displays unique biological properties, including preferential posterior location, reduced proliferation, enhanced progenitor-like transcriptional characteristics and a high potential to form SHH MB. Together, these results reveal a new aspect of the heterogeneity within the GCP population and suggest that this heterogeneity is functionally relevant to both normal and tumoral cerebellar development.

An important feature of the Nes-expressing GCP population is its restricted localization to the posterior cerebellum. Previous work showed that mouse SHH MB preferentially arises from the posterior and lateral mouse cerebellum under conditional activation of SHH signaling and also human SHH MBs shows a similar anatomical preference (Teo et al., 2013; Wefers et al., 2014; Perreault et al., 2014; Zhao et al., 2017; Tan et al., 2018). Why only the laterally located Nes⁺ GCP give rise to tumors remains to be determined. Our data nevertheless support the idea that the regional enrichment of Nes-expressing GCPs may contribute to the known spatial preference of SHH MB formation. It is also interesting that CFP⁺ (Nes⁺) BARHL1⁺ cells are detected in the cochlear region (Figure 1B), a region where more than 30% of SHH MB arise from in human (Grammel et al., 2012).

Our analysis further showed that the majority of Nes-expressing GCPs do not express the neural stem cell marker SOX2, with only a minority of them showing co-expression of CFP (Nes), BARHL1 and SOX2. This is an important distinction as SOX2⁺ progenitors were proposed as candidate cells of the origin in SHH MB (Selvadurai et al., 2020). Our findings suggest that the Nes-expressing GCP population identified here is not equivalent to the previously described SOX2⁺ progenitor population and may instead represent an additional and distinct cell of origin of SHH MB. It will therefore be important to directly compare the tumorigenic potential of Nes⁺ SOX2⁺ and Nes⁺ SOX2^- GCPs. Furthermore, the presence of a smaller SOX2⁺ fraction within the Nes-expressing GCPs suggests an internal heterogeneity consistent with the possibility that these cells have distinct developmental or functional states with respect to tumor formation.

One of the most significant findings of this study is that a GCP population targeted at a ten-fold lower frequency than the broader Atoh1⁺ GCP pool nevertheless gives rise to SHH MB with a similar penetrance. This observation argues that tumor susceptibility is not determined simply by the number of targeted GCPs, but rather by the intrinsic biological properties of specific GCP subtypes. This result provides direct functional evidence that developmental heterogeneity within the GCP population translates into differential tumorigenic potential.

Our findings also help clarify prior work based on Nes-expression driven tumor models. Earlier studies using a Nes-CreER transgene concluded that the Nes-lineage cells can give rise to SHH MB (Li et al., 2013), but the precise identity of the targeted population remained difficult to resolve because the Nes-CreER transgene used is broadly expressed throughout the EGL and in multiple progenitor populations during development (Figure 4–figure supplement 2). In contrast, the Nes-FlpoER transgene strategy used here labels much fewer cells and a spatially restricted subset of cells within the EGL. Furthermore, our intersectional genetic experiments provide additional support that the tumor-competent population corresponds specifically to cells that co-express Nes and Atoh1. Thus, our model likely provides a more precise readout and definition of the GCP subset of interest within the broader Nes-lineage.

The Nes gene might represent more than a simple marker of a tumor-prone GCP population and could also contribute directly to its tumorigenic behavior. Although Nes is commonly used as a marker of immature neural states, previous work in SHH MB showed that NES can functionally promote tumorigenesis by maintaining SHH signaling (Li et al., 2016). Nes-expression was shown to increase during tumor formation, and Nes-depletion found to reduce proliferation, promote differentiation and suppress tumor growth through regulation of GLI3 processing. These findings raise the possibility that Nes is not simply labeling a permissive progenitor population but may itself support the oncogenic SHH pathway activity. In this context, our data suggest an important nuance: that although initially Nes-expressing GCPs in the medial and lateral cerebellum form lesions, ultimately only laterally located Nes-expressing GCPs form tumors. One interesting question is whether this finding is related to the level of NES protein present in the cells, perhaps being higher in lateral cells. Nes-expressing GCPs therefore might not possess a greater ability to initiate tumor formation, but rather a stronger capacity to remain in a tumorigenic state once transformation has been engaged. It will be important to address this idea directly in future studies using genetic loss-of-function approaches within the Nes-expressing GCP population.

Our scRNA-seq characterization of Nes⁺ GCPs further supports their identity as a progenitor or stem-like cells. Nes-expressing GCPs exhibit lower EdU incorporation in vivo and, and in the scRNA-seq data they segregate as a distinct GCP cluster with features suggestive of reduced differentiation and increased developmental plasticity. Compared with neighboring posterior GCP clusters, Nes-expressing GCP cluster cells retains GCP markers while showing enrichment for progenitor-associated genes and pathways, including components linked to WNT pathways (including Wls) (Yeung and Goldowitz, 2017) and a relative reduction of neuronal differentiation. This combination of reduced proliferative activity and reduced differentiation suggests that these cells may exhibit quiescent and stem-like features reminiscent of cancer stem cells (Chen et al., 2016). Indeed, these features may enhance their competence to maintain the oncogenic state.

An additional observation supporting the relevance of this population is its relationship to tumor cells at the transcriptomic level. Relative to other GCP clusters, Nes-expressing GCPs show fewer transcriptional difference from SHH MB GCP-like cells, suggesting that they may lie closer to the tumor cells than do other GCPs. This result implies that their developmental program may require fewer transcriptional changes to transit into a tumorigenic state under constitutive SHH activation. These observations fit well with our developmental findings, as a progenitor population that is relatively undifferentiated and still relatively plastic might be expected to transform more easily that other GCPs.

In summary, our study identifies a rare Nes-expressing GCP population as a distinct population with the cells located in the lateral developing cerebellum being highly tumor competent. These findings refine the view of GCPs as the cell of origin of SHH MB by showing that tumor susceptibility is linked to the heterogeneity within the GCP population. Finally, our findings support a model in which SHH MB arises preferentially from rare Nes-expressing GCPs defined by their spatial, molecular and functional properties.

Materials and Methods

Key resources table

Animal model and Breeding

All the mouse experiments were performed according to protocols approved by the Institutional Animal Care and Use Committee of Memorial Sloan Kettering Cancer Center (MSKCC) (protocol no. 07-01-001). Animals were housed in temperature-and humidity-controlled rooms on a 12-hour light/dark cycle and given access to standard laboratory mouse chow and water ad libitum. Both sexes were used randomly in this study. Experimenters were blinded for genotypes whenever possible. The following mouse line were used: Nes-CFP (JAX #034387) (Encinas et al., 2006), Atoh1-GFP (JAX#013593) (Rose et al., 2009), Atoh1-FlpoER (JAX #040091) (Tan et al., 2018; Wojcinski et al., 2019), SmoM2 (JAX #005130) (Mao et al., 2006), R26^FSF-GFPcre (MASTR, JAX #018903) (Lao et al., 2012), Trp53^flox (JAX # 008462) (Marino et al., 2000), Nes-FlpoER (JAX #040094) (Lao et al., 2012), R26^LSL-tdTom ( Ai75, JAX #025106) (Daigle et al., 2018), Nes-CreER (JAX #016261)(Battiste et al., 2007) and Atoh1^FRT-Cre was obtained from Christo Goridis (Ruffault et al., 2015).

Tamoxifen administration: Genetic recombination in mouse lines was induced by injection of tamoxifen. Tamoxifen (T5648, Sigma-Aldrich) was dissolved in corn oil (C8267, Sigma-Aldrich) at 20mg/ml and injected subcutaneously into the back of P0 mice at a dose of 100mg per g of body weight.

EdU administration: 5-ethynyl-2’-deoxyuridine (EdU) (900584, Sigma-Aldrich) stock was dissolved in sterile phosphate-buffered saline (PBS) at 10 mg/mL and a dose of 5 μg/g was injected intraperitoneally into animals 1 hour prior to euthanasia.

Tissue processing and immunostaining

For immunocytochemistry, animals younger than P8 were decapitated and then the whole head was fixed in 4% paraformaldehyde for 48 hours (hr) at 4°C, cryoprotected in 30% sucrose in phosphate-buffered saline (PBS) until they sank and then frozen in Cryo-OCT (Tissue-Tek). Older animals were anesthetized and then perfused with cold PBS followed by 4% paraformaldehyde prior to brain dissection. Frozen brains were sectioned sagittally at 12 μm or 14 μm and slides stored at −80°C.

For immunofluorescence staining, slides were left to warm to room temperature (RT) for 1 hr and washed 3 times in PBS. Primary antibodies diluted in the blocking buffer were placed on slides for overnight incubation at RT. The blocking solution was composed of 4% Bovine serum albumin (BSA) (A2153, Sigma Aldrich), 3% donkey serum ((D9663, Sigma Aldrich) and 2% Triton X100 (BP151, Fisher Scientific) dilatated in PBS. Slides were then washed 3 times in PBS and incubated with fluorophore-conjugated secondary antibodies diluted in the blocking buffer for 2 hr at RT. Nuclei were counterstained with DAPI (D9542, Sigma-Aldrich, 1:3000) and the slides were mounted with Fluoro-Gel (17985, Electron Microscopy Sciences). Primary antibodies used are described in Key Resources Table and secondary antibodies were Alexa Fluor-conjugated secondary antibodies (1:1000). EdU was detected using a Click-it EdU (C10340, Invitrogen) assay with Sulfo-Cyanine5 azide (A3330, Lumiprobe Corporation) following the manufacturer’s protocol.

Image acquisition

Images were collected with a DM6000 Leica microscope, a NanoZoomer Digital Pathology microscope (Hamamatsu Photonics), or an LSM880 confocal microscope (Zeiss). Images were processed using NDP.view2 software, ImageJ software (NIH, Bethesda, MA, USA) and Photoshop (Adobe).

Flow cytometry and Cell sorter

Cerebella were dissected into cold Hank’s Buffered Salt Solution (HBSS) (14170112, ThermoFisher) and then dissociated into Papain (LK003150, Worthington) at 37°C for 30 min. After dissociation, cells are centrifuged 5 minutes (min) at 400 g. Papain was inactivated with Ovomucoid protease inhibitor. After centrifugation, cells were finally washed in PBS filtered using a 40 μm mesh cell strainer (CLS431750, Sigma-Aldrich). After 5 min of centrifugation at 400g, the pellet was resuspended in PBS, 0.5% BSA for downstream experimentation.

RT qPCR

The 4 populations of cells (GFP^- CFP^-, GFP⁺ CFP^-, GFP^- CFP⁺ and GFP⁺ CFP⁺) from Nes-CFP; Atoh1-GFP mice at P1 were sorted and frozen in TRI-reagent (Sigma-Aldrich). Total RNA was extracted using the mRNeasy kit (Qiagen), and cDNA was synthesized using SuperScript IV reverse transcriptase (Invitrogen). For the real-time PCR, 20ng of cDNA was used per reaction and amplified using StepOnePlus™ Real-Time PCR System. The reactions used a Step One Plus apparatus and software. The fold change was determined using the formula 2^-ΔΔCT, where ΔΔCT= ΔCT_sample – ΔCT_{GFP-CFP- sample} with ΔCT= CT_gene - CT_Gapdh. Statistics were determined using the Mann-Whitney test. Sequence of primer pairs used are listed in the Key resources Table.

Single-cell RNA sequence sample preparation and analysis

Single Cell RNA-seq sample preparation: Nes-CFP and Atoh1-GFP; Nes-CFP mice were euthanized at P1 and the cerebella were dissected in cold HBSS. For Nes-CFP cerebella, only the posterior part was kept, posterior to the primary fissure. Posterior cerebella or whole cerebella (Atoh1-GFP; Nes-CFP mice) were dissociated in Papain for 30 minutes at 37°C before being washed out in Ovalbumin Tumors were triturated by pipetting up and down using a P1000 and filtered through a 40µm cell strainer. Once the cells were dissociated, Nes-CFP cell pellets were washed with PBS. For each sample (4 samples, 1 cerebellum each), cells were counted using 0.2% Trypan on a hemocytometer and diluted before loading onto a 10X chip, targeting 10,000 viable cells per sample. Atoh1-GFP; Nes-CFP cell pellets were grouped together (3 cerebella for 1 sample) and the double positive population CFP⁺ GFP⁺ were sorted using BD FACS aria. Transgene negative, Nes-CFP and Atoh1-GFP cerebella were used as controls to set up the FACS settings. The scRNA-Seq of cell suspensions for the 5 samples was performed on a Chromium instrument (10X genomics) following the user guide manual for 3’ v3.1. Final libraries were sequenced on an Illumina NovaSeq6000 by the IGO core of MSKCC.

scRNA-seq data analysis: The Cell Ranger Single Cell software suite (10x Genomics) was used to align reads and generate feature-barcode matrices. The Genome Reference Consortium Mouse Build 38 (GRCm38, Gencode annotation mm10) was used as the reference genome. Raw reads were treated using the Cell Ranger count program with the default parameters. Seurat v5.0.2 package was used to generate a UMI (unique molecular identifier) count matrix from the Cell Ranger output38. Genes expressed in less than 10 cells were removed for further analyses. Cells with low transcriptomic complexity were filtered out by computing the ratio of detected genes to unique molecular identifiers (log10GenesPerUMI). Only cells with a log10GenesPerUMI > 0.80 were retained for downstream analyses to exclude low-quality or low-complexity cells. Furthermore, cells with less than 500 UMIs, more than 0.20 mitoRatio, and less than 250 genes detected were considered low quality/outliers and discarded from the datasets. Normalization was performed on individual samples using the NormalizeData function with default parameters. The normalized data was scaled by ScaleData function. Samples were then integrated using IntegrateData. The first 40 dimensions were used for the FindNeighbors function, and clusters were identified using the FindClusters function with a resolution of 0.8. Data were projected onto the 2D space using the FindUMAP or RunUMAP function with 40 dimensions. Cluster markers and further differential gene expression analyses were performed using normalized counts (NormalizeData) in the RNA assay. Cluster markers were identified using the FindAllMarkers and comparing markers generated to the literature. To refine clustering further, the SubsetData function was used to create a new Seurat object, and the above clustering was reiterated. Differential gene expression analyses were performed using the FindMarkers function in the Seurat package. Genes with an adjusted P value (Padj) < 0.05 were considered significantly up or downregulated. Results were visualized by Feature plot, Violin plot and Umap plot using Seurat package, and Volcano plot using the EnhancedVolcano package on Bioconductor39. GSEA analysis on GO and Hallmark terms were performed using the enrichGO, clusterProfiler packages on Bioconductor40.

Quantification and statistical analysis

At least 3 mice of each genotype were used for all quantifications, and the numbers for each experiment are stated in the figure legends. A statistical method was not used to predetermine each sample size – the sample sizes are similar to previously reported work.

Statistical analyses

All statistical analyses used to quantify data were performed using Prism software (GraphPad). Statistical significance was defined as p ≤ 0.05. The specific statistical tests applied are detailed in the corresponding figure legends. Data, presented as scatter plots, are shown as mean ± standard error of the mean (SEM).

Figure supplements

Figure 4 – figure supplement 1.

Figure 4 – figure supplement 2.

Figure 5 – figure supplement 1.

Data Availability

Sequencing data are deposited in GEO (token inoxueqgzrqntgt) under accession code GSE328459. All other experimental data in this study are included in the main text, figures and graphs and Source data files in the manuscript. For the research, no new materials were generated.

Acknowledgements

We thank members of the Joyner lab for interesting discussions, in particular Reeti Mayur Sanghrajka for insightful discussions and advice at the beginning of the project. We thank Christo Goridis for the Atoh1^FRT-Cre mice and the Memorial Sloan Kettering Cancer Center Flow Cytometry Core and the Integrated Genomics Operations Core Facility for technical support.

Additional information

Funding

NCI (R01CA192176), Alexandra L Joyner

NINDS (R01NS092096), Alexandra Joyner

Cycle for Survival, Alexandra L Joyner

NCI Cancer Center Support Grant (CCSG, P30 CA08748), Alexandra L Joyner

François Wallace Monahan Fellowship, Salsabiel El Nagar

Ethics

There are no ethical considerations since no human data were used.

Funding

HHS | NIH | National Cancer Institute (NCI) (R01CA192176)

• Alexandra L Joyner

HHS | NIH | National Institute of Neurological Disorders and Stroke (NINDS) (R01NS092096)

• Alexandra L Joyner

Cycle for Survival

• Alexandra L Joyner

HHS | NIH | National Cancer Institute (NCI) (CCSG P30 CA08748)

• Alexandra L Joyner

Additional files

Supplementary file 1. Figure 2–Source data 1. RT-qPCR raw data and computation of ΔCT and 2-ΔΔCT values for comparing GFP^- CFP⁺, GFP⁺ CFP^- and GFP⁺ CFP⁺ populations to GFP^- CFP^-population. Related to Figure 2C, D. Figure 5-Source data 1. Gene expression in each cluster of the integrated 5 sample scRNA-seq data. Related to Figure 5A and Figure 5 supplement 1C. Figure 5-Source data 2. Gene expression in each GCP cluster. Related to Figure 5D-I and Figure 5 supplement 1E. Figure 5-Source data 3. Cell cycle phase assignment for each cell of GCP clusters. Related to Figure 5J. Figure 5 - Source data 4. ScRNA-seq significant differential gene expression analysis in cluster 10 compared to Clusters 1, 3 and 6 using FindMarkers approach. Figure 5-Source data 5. ScRNA-seq significant differential gene expression analysis in cluster 10 compared to Clusters 1, 3 and 6. Related to Figure 5L. Figure 5-Source data 6. Gene Ontology (GO) analysis of genes upregulated in cluster 10 compared to clusters 1, 3 and 6. Related to Figure 5M. Figure 5-Source data 7. Gene Ontology (GO) analysis of genes downregulated in cluster 10 compared to clusters 1, 3 and 6. Related to Figure 5N. Figure 6-Source data 1. Gene expression in each cluster of integrated GCP clusters 1, 3, 6 and 10 and SmoM2 GCP-like tumor cells (Lao et al., 2026). Related to Figure 6A. Figure 6-Source data 2. ScRNA-seq significant differential gene expression analysis of GCP clusters 1, 3, 6 and 10 compared to SmoM2 GCP-like tumor cells. Related to Figure 6D. Figure 6-Source data 3. ScRNA-seq significant differential gene expression analysis of Nes-GCPs compared to SmoM2 GCP-like tumor cells. Related to Figure 6F. Figure 6-Source data 4. ScRNA-seq significant differential gene expression analysis of Nes⁺ GCPs compared to SmoM2 GCP-like tumor cells. Related to Figure 6H.