OLIG2 mediates a rare targetable stem cell fate transition in sonic hedgehog medulloblastoma

Abstract

Functional cellular heterogeneity in tumours often underlies incomplete response to therapy and relapse. Previously, we demonstrated that the growth of the paediatric brain malignancy, sonic hedgehog subgroup medulloblastoma, is rooted in a dysregulated developmental hierarchy, the apex of which is defined by characteristically quiescent SOX2⁺ stem-like cells. Integrating gene expression and chromatin accessibility patterns in distinct cellular compartments, we identify the transcription factor Olig2 as regulating the stem cell fate transition from quiescence to activation, driving the generation of downstream neoplastic progenitors. Inactivation of Olig2 blocks stem cell activation and tumour output. Targeting this rare OLIG2-driven proliferative programme with a small molecule inhibitor, CT-179, dramatically attenuates early tumour formation and tumour regrowth post-therapy, and significantly increases median survival in vivo. We demonstrate that targeting transition from quiescence to proliferation at the level of the tumorigenic cell could be a pivotal medulloblastoma treatment strategy.

Fig. 1. Distinct growth, transcriptomic and chromatin accessibility profiles characterise SHH-MB stem cells.

a Visualisation of the approximate lineage hierarchy in SHH-MB based on quantitative analysis of proliferation dynamics (for details, see Supplementary Note); based on data from Vanner et al.; created in BioRender.com. b Unsupervised clustering using Uniform Manifold Approximation and Projection (UMAP) performed on 6739 single cells from four mouse medulloblastoma primary Ptc tumours; OPC oligodendrocyte progenitor cells. c Scatter plot of the cell cycle phases in SOX2⁺ cells computationally extracted from the scRNAseq data of 4 Ptc mouse tumours. d Genomic annotation distribution calculated using cis-regulatory element annotation system (CEAS) of the genome (left), of the catalogue of chromatin accessible regions identified through ATACseq of SOX2⁺ and SOX2⁻ subpopulations (centre) and of the top 10% most variable chromatin accessible regions across SOX2⁺ and SOX2⁻ subpopulations (right). e Relative normalised chromatin accessibility between SOX2⁺ and SOX2⁻ subpopulations for each of the regions significantly correlated to the Sox2 promoter calculated using C3D (‘Methods’); n = 4 biologically independent samples; the box of the box and whiskers plots extends from the 25th to 75th percentiles; the centre line is plotted at the median; whiskers indicate min. and max. values; two-tailed unpaired t-test; *p < 0.05, **p < 0.01, ***p < 0.001; L–R p-values: 0.0025, 0.0009, 0.0085, 0.0353, 0.0432, 0.0245, and 0.0280. f Gene Ontology Biological Process (GO BP) enrichment analysis of the genes associated with significantly differentially accessible regions across SOX2⁺ and SOX2^– subpopulations. The nodes represent gene sets, and the edges represent mutual overlap. Clusters, representing overlapping gene sets, were enriched in either SOX2⁺ population-associated gene sets (green nodes), SOX2^– population-associated gene sets (grey nodes), or a mixture of both; n = 4 biologically independent samples. g HOMER motif analysis of uniquely accessible regions in the SOX2⁺ and SOX2^– subpopulations; AT2=AT2G33550; n = 4 biologically independent samples; motif enrichment is calculated using cumulative binomial distributions. Source data are provided as a Source Data file.

Fig. 2. Olig2 is a regulator of stem cell proliferation.

a HOMER motif analysis of uniquely accessible regions in the cycling (MKI67⁺) and non-cycling (MKI67⁻) fractions of the stem (SOX2⁺) cells; n = 2 biologically independent samples; motif enrichment is calculated using cumulative binomial distributions. b Overlap of OLIG2 binding peaks in mNSCs (data obtained from Mateo et al.) with ATACseq peaks of SOX2^+/− (top pie chart) as well as SOX2⁺/MKI67⁺ and SOX2⁺/MKI67^– (bottom pie chart) tumour fractions. c Gene Ontology Biological Process (GO BP) analysis of genes associated with the overlaps in (b). d, e Immunocytochemistry of Ptch1^+/−;Trp53^−/− mouse tumour cells (d), quantification of EdU+ cells after a 1-hour pulse averaged across three biological replicates (e); n = 3, error bars denote mean ± SEM; two-tailed unpaired t-test; ns p ≥ 0.05, ****p < 0.0001; p > 0.9999 (EdU⁺ vs. OLIG2⁺), p < 0.0001 (EdU⁺ vs. EdU⁺/OLIG2⁺), p < 0.0001 (OLIG2⁺ vs. EdU⁺/OLIG2⁺). f Phase object confluence of adherently grown Ptch1^+/−;Trp53^−/− CRISPR-Cas9 control or Olig2-KO mouse tumour cells measured across 14 days using live cell imaging; data shown are representative of three independent experiments; error bars denote mean ± SEM; two-tailed unpaired t-test performed at mid-exponential phase time point (144 hours); p = 0.0021 (control vs. Olig2-KO). g Limiting dilution analysis (LDA) performed on Ptch1^+/−;Trp53^−/− CRISPR-Cas9 control or Olig2-KO cells; n = 3 biological replicates; individual replicates were assessed by goodness of fit testing using extreme limiting dilution analysis (ELDA) software (see Methods); error bars denote mean ± SEM; unpaired t-test; two-tailed p-value; p = 0.0182. h Tumour volumes in NOD-scid-gamma (NSG) mice subcutaneously engrafted in both flanks with either Ptch1^+/−;Trp53^−/− CRISPR-Cas9 control or Olig2-KO cells. n = 4 mice (8 flanks) for control and n = 3 (6 flanks) for Olig2-KO condition; error bars denote mean ± SEM; chi-square test; p = 0.0157; f–h: *p < 0.05, **p < 0.01. Source data are provided as a Source Data file.

Fig. 3. Olig2 plays a key role in stem cell fate transition.

a Phase object confluence of adherently grown Ptch1^+/−;Trp53^−/− CRISPR-Cas9 control or Olig2-KO mouse tumour cells measured across 9 days using live cell imaging. The cells were grown in either proliferation media, modified quiescence-inducing media containing BMP4 or modified media control for 3 days and then washed and re-exposed to proliferation media; data shown are representative of three independent experiments; error bars denote mean ± SEM; two-tailed unpaired t-test performed at final time point (212 h); two-tailed p-values; ns p ≥ 0.05, ****p < 0.0001; p = 0.4693 (Olig2-KO + veh vs. control + veh), p = 0.4967 (Olig2-KO + MM vs. control + MM), p < 0.0001 (Olig2-KO + MM + wash vs. control + MM + wash). b, c Expression of endogenous SOX2 and OLIG2 proteins detected by immunofluorescence in the external granule layer (EGL) at P14 in Ptc WT and Ptc mice (b) and quantification of the SOX2⁺/OLIG2⁺ cells as a fraction of the SOX2⁺ cells and OLIG2⁺ cells respectively (c); n = 3 biological replicates; error bars denote mean ± SEM; two-tailed unpaired t-test; *p < 0.05; p = 0.0177. d–k scRNAseq analysis on CGNP-like (SOX2⁺ or DCX⁺) Math1-Cre;SmoM2 mouse tumour cells at P7; unsupervised clustering using Uniform Manifold Approximation and Projection (UMAP) performed on ~5000 single cells (d), dot plot of the top genes most differentially expressed in each cluster (e), graphical demonstration of dimension separation strategy of neoplastic CGNP-like cells (f), construction of a neuronal differentiation trajectory by Monocle2 and the expression of Sox2, Olig2, Dcx, Neurod1, and Stmn2 across the trajectory (g), construction of a cell cycle pseudotime by slingshot and the expression of cell cycle markers across the cell cycle pseudotime (Pcna, Mki67, and Top2a) (h), heatmap of the differentially expressed genes across the neurogenesis trajectory and the cell cycle pseudotime (i), comparison of the neurogenesis and cell cycle trajectories; gene names associated with each heat map are listed in Supplementary Data 2 (j), violin plots of cell cycle pseudotime and Mki67 expression between OLIG2⁺ vs. OLIG2^– cells; centre line, median; box bounds, upper and lower quartiles; whiskers, 1.5x interquartile range; points, outliers; n = 1096/3996 (OLIG2⁺/OLIG2⁻); two-tailed unpaired t-test (k). Source data are provided as a Source Data file.

Fig. 4. Inhibiting OLIG2 enriches for less proliferative and a more potent sphere-forming stem cell population.

a Structure of the OLIG2 inhibitor, CT-179. b, c Immunocytochemistry of Ptch1^+/−;Trp53^−/− mouse tumour cells (b), quantification of EdU incorporation in OLIG2⁺ cells after a 1-hour pulse (c); data shown are representative of three independent experiments; error bars denote mean ± SEM; two-tailed unpaired t-test; ***p < 0.001; p = 0.0001 (c). d Proportion of Ptch1^+/−;Trp53^−/− mouse tumour cells in each phase of the cell cycle as measured through FACS analysis of propidium iodide (PI) staining and analysed using FlowJo software; data shown are representative of three independent experiments; error bars denote mean ± SEM; two-tailed unpaired t-test; *p < 0.05; p = 0.0143 (G0/G1), p = 0.0157 (S), p = 0.0210 (G2/M). e Percentage confluence of Ptch1^+/−;Trp53^−/− mouse tumour cells transfected with a doxycycline (DOX)-inducible OLIG2 overexpression (O/E) construct or empty vector control upon treatment with CT-179 or vehicle for 10 days; data shown are representative of three independent experiments; error bars denote mean ± SEM; two-tailed unpaired t-test; *p < 0.05; p = 0.0404 (O/E vector control vs. OLIG2 O/E), p = 0.0484 (OLIG2 O/E vs. OLIG2 O/E + CT-179). f–h Immunocytochemistry of Ptch1^+/−;Trp53^−/− mouse tumour cells treated with IC₉₀ of CT-179 (276.1nM) (f), quantification of SOX2⁺ cells (g), quantification of the fluorescence intensity of DAPI, SOX2 and OLIG2 measured using ImageJ software (h); data shown in are representative of three independent experiments; scale bar: 80µM (f); g, h: error bars denote mean ± SEM; two-tailed unpaired t-test; ns p ≥ 0.05, **p < 0.01, ***p < 0.001; p = 0.0001 (SOX2) (g); p = 0.4551 (DAPI), p = 0.0003 (SOX2), p = 0.0035 (OLIG2) (h). i, j Secondary limiting dilution analysis (LDA) performed on Ptch1^+/−;Trp53^−/− mouse tumour cells pre-treated with vehicle, IC₁₀ (129.7 nM), IC₅₀ (189.2 nM) or IC₉₀ (276.1nM) dose of CT-179 for 24 hours; data show the percentage of sphere-forming capacity (i) and a quantification of the spheres size (j); data shown are representative of three independent experiments; line at the median; two-tailed unpaired t-test; ns p ≥ 0.05, *p < 0.05, ****p < 0.0001; p = 0.0724 (IC₁₀ vs. vehicle), p = 0.0259 (IC₅₀ vs. vehicle), p = 0.0259 (IC₉₀ vs. vehicle) (i); p = 0.3791 (IC₁₀ vs. vehicle), p < 0.0001 (IC₅₀ vs. vehicle), p < 0.0001 (IC₉₀ vs. vehicle) (j). k, l Cell confluence measured using high-throughput live-cell imaging of Ptch1^+/−;Trp53^−/− mouse tumour cells pre-treated with vehicle or IC₅₀ (189.2 nM) CT-179 for 24 hours, immediately followed by a dose-response assay to AraC (k) and Vismodegib (GDC) (l). Source data are provided as a Source Data file.

Fig. 5. OLIG2 inhibition blocks SOX2⁺ MB stem cell activation.

a Volcano plot comparing fold change (x-axis) and false discovery rate (y-axis) obtained from EdgeR analysis of the differentially expressed genes between IC₅₀ (189.2 nM) CT-179 or vehicle treatment of Ptch1^+/−;Trp53^−/− (PTCP53 304) mouse tumour cells for 6 days. b Relative fold change expression of select genes in vehicle and CT-179 treated cells from dataset shown in (a); n = 3 biological replicates, error bars denote mean ± SEM; two-tailed unpaired t-test; ns p ≥ 0.05, *p < 0.05, **p < 0.01. c, d Representative immunofluorescence (IF) images of endogenous SOX2, OLIG2, and EdU in early (c) or late (d) neoplastic lesions of mice sacrificed after 7 days of continuous EdU label in the presence of either vehicle (dH₂O) or CT-179 daily injections. Scale bar: 50 µM for 20X and 25 µM for 40X. e Quantification of the fraction of SOX2 ⁺ cells and of EdU incorporation in (c) relative to DAPI⁺ cells. n = 4 for CT-179, n = 3 for vehicle; error bars denote mean ± SEM; two-tailed unpaired t-test. unpaired t-test; *p < 0.05, ****p < 0.0001; p = 0.0240 (SOX2⁺), p = 0.0081 (OLIG2⁺/EdU⁺). f Quantification of the fraction of SOX2⁺ or SOX2⁺ and OLIG2⁺ cells that have incorporated EdU in (d). n = 3 for CT-179, n = 3 for vehicle; error bars denote mean ± SEM; two-tailed unpaired t-test; **p < 0.01, ***p < 0.001; p = 0.0006 (EdU⁺), p < 0.0001 (EdU⁺). g Representative immunofluorescence (IF) images of endogenous SOX2, EdU, and BrdU in early neoplastic lesions in mice sacrificed after 4 days of EdU label followed by an 8-day chase period of which 7 days were treatment with either vehicle (dH₂O) or CT-179 by daily IP injections, followed by a single BrdU pulse. Scale bar: 50 µM for 20X and 25µM for 40X. SOX2⁺/EdU⁺ cells are indicated with white arrows in the 40x image, while the yellow arrow indicates a triple positive cell (SOX2⁺/EdU⁺/BrdU⁺). h Quantification of the fraction of SOX2⁺ cells that have incorporated EdU or BrdU as represented in (g). For the quantification, 8 independent lesions were evaluated from 3 mouse brains for each treatment condition (vehicle or CT-179); error bars denote mean ± SEM; two-tailed unpaired t-test; ns p ≥ 0.05, **p < 0.01; p = 0.0034 (EdU⁺), p = 0.9926 (OLIG2⁺/BrdU⁺). Source data are provided as a Source Data file.

Fig. 6. Targeting OLIG2 abrogates tumour initiation and relapse in vivo.

a Treatment strategy and H&E staining of Ptc mouse hindbrains treated with 20 mg/kg CT-179 or vehicle daily by IP injection from P3–P21 and sacrificed at P22. Lesion is indicated by the black arrow. n = 4 for CT-179, n = 4 for vehicle. b Treatment strategy and Kaplan–Meier survival curve of Ptc mice treated with 20 mg/kg CT-179 (Vehicle: n = 12; CT-179: n = 8). The dotted lines represent the duration of treatment of vehicle or CT-179 by daily IP injection (P3–P28). Significance was estimated using the log-rank (Mantel–Cox) test. Chi square = 5.425, p = 0.0198. c Sagittal MRI images of the brains of Ptc mice taken at age 3.5 months. The mice were treated with MCT (vehicle for GDC) from P56-P64 followed by CT-179 or vehicle from P65–P93. Tumours are indicated by the yellow arrows or yellow dotted lines. d Phase object confluence of adherently grown Ptch1^+/−;Trp53^−/− mouse tumour cells measured across 14 days using live cell imaging. Cells were treated with vehicle control, IC₃₀ CT-179 (163.5 nM), 50 µM Vismodegib (GDC-0449), or a combination of CT-179 and Vismodegib; data shown are representative of three independent experiments; error bars denote mean ± SEM; two-tailed unpaired t-test; ns p ≥ 0.05, *p < 0.05, ***p < 0.001; p = 0.7291 (water vs. CT-179 at 172 h), p = 0.0365 (water vs. GDC at 172 h), p = 0.0277 (water vs. GDC + CT-179 at 172 h), p = 0.8717 (water vs. CT-179 at 184 h), p = 0.0614 (water vs. GDC at 184 h), p = 0.0101 (water vs. GDC + CT-179 at 184 h), p = 0.6868 (water vs. CT-179 at 220 h), p = 0.8109 (water vs. GDC at 220 h), p = 0.0002 (water vs. GDC + CT-179 at 220 h). e Treatment strategy (top) and H&E staining of Ptc mouse hindbrains treated with Vismodegib from P28–P35 followed by CT-179 or vehicle from P35–P61. Lesion is indicated by the black arrows. n = 4 for CT-179, n = 6 for vehicle. f Treatment strategy (below) and Kaplan–Meier survival curve of Ptc mice treated GDC followed by vehicle or 20 mg/kg CT-179. n = 16 for CT-179, n = 16 for vehicle. The dotted lines represent the durations of GDC debulking (daily gavage) (P56–P63) followed immediately by treatment with vehicle or CT-179 by daily IP injection (P64–P92). Significance was estimated using the log-rank (Mantel–Cox) test. Chi square = 5.452, p-value = 0.0195. Source data are provided as a Source Data file.

Fig. 7. OLIG2 inhibition prevents tumour progression.

Model of proposed role of Olig2 and OLIG2 inhibition in tumour initiation and relapse; created in BioRender.com.