Identifying a high confidence group of genes that increase with GNP differentiation.

A) Schematic representation of the spatial distribution of stages of GNP development represented in the P7 mouse cerebellum. B) Schematic representation of the temporal distribution of the stages of GNP development. Stages represented in the RNASeq data are indicated by vertical bars. Representative sagittal slices of the cerebellum from mice harboring a Math1-GFP reporter, injected with EDU (1.5 h, red) and DAPI (blue) at P1, P7, and P14. C) A Sankey diagram describes how genes are partitioned in order to identify high confidence differentiation genes. Briefly, genes are segregated by the following steps: 1, excluded genes by non-expression; 2, excluded genes by contamination from non granule neuron lineage identified through scRNAseq; 3, excluded by minimal change in expression during development; 4, segregated genes by the time of minimum and maximum expression (highest P14 to P6 differentiation genes, highest P0-7 proliferation genes, highest at E15 embryonic genes); 5 excluded genes by low statistical confidence; 6, exclude genes that increase prior to P0; 7, separate by fold-change (>2 fold log2 fold change or 1-2 log2 fold change).

H3K27me3 modification near the promoter of genes at P7 is associated with genes that increase later with differentiation.

A) Enrichment of chromatin features near promoters of proliferating P7 GNP whose expression increases during differentiation. Enrichment is calculated for a given modification as the number of differentiation genes over either all genes (black) or genes expressed at any point during the granule neuron lineage (green). The log2 value of this fraction is then shown on the figure. The p-value is calculated using the hypergeometric test and the p-values are shown as the size of the dot. B) A parallel sets diagram that shows correlation between H3K27me3 binding and other modification. The disconnected bar graphs shows the number of genes that are bound or within the long upper tail for each modification. The red colour that spans the bar segments shows the genes that contain the next modification.

Ezh2 cKO leads to a depletion of the inner EGL

A) Schematic of granule neuron (GN) development in a P7 cerebellum sagittal slice. Inner external grandule layer (iEGL), outer external granule layer (oEGL); molecular layer (ML), internal granule layer (IGL). B) WT and C) Ezh2 cKO 40x confocal micrographs of the deep sulcus between lobule V and VI in the control and in the cKO, showing DAPI (red) and p27 (green; a cell cycle arrest marker in GNPs). In the mutant, the boundary between the inner EGL and the ML and the ML with the IGL is ill-defined. The cyan arrow indicates the position of the inner EGL, there are fewer p27 labeled cells sitting within the EGL. The cellular density is reduced in both the inner EGL and IGL. D) Quantification of the boundaries adjacent to the ML in WT and Ezh2 cKO by collapsing 2D images into 1D line segments. The line segments for anti-p27 immunofluorescence span from the pial boundary to the IGL and show a hump of high average fluorescence at the inner EGL and the IGL. In the Ezh2 cKO the iEGL p27 signal extends further into the ML then in the WT. The blue arrows show the approximate position of the transition between the iEGL and the ML. E) The average fluorescence intensity difference between mutant (n = 7) and WT (n = 4) shows where along the line segment from the pia to the IGL is the Ezh2 cKO most different from the WT. Positive values are associated with p27 value being higher in the Ezh2 cKO and is highest at the boundary of the inner EGL and the ML and to a lesser extent at the boundary between the ML and IGL. The WT shows higher signal within the inner EGL itself where the values of E are negative. F) Quantification of each line segment at the point where the differences between Ezh2 cKO and WT are largest as indicated by the hashed lines seen in panel D and E. The left panel shows quantification at position where the WT had had more P27 staining then to the Ezh2 cKO and the right panel shows the opposite. The distribution of the values at the 2 positions in right and left panel for the 266 WT and 315 Ezh2 cKO line segments was shown as a violin plot (a histogram that is mirrored). The dots correspond to individual mice, with multiple slides analyzed for each mouse. P-values are calculated using a T-test showing a significant increase of P27 labeling within the boundary of the EGL and ML. G) Segmentation into inner and outer EGL, ML, IGL and deep white matter using panoramic images of the entire cerebellum, labeled with DAPI, P27, NeuN. Anti-p27 over the entire P7 cerebellar section, for WT i) and Ezh2 cKO ii). H) Quantification of the area for each layer in WT and Ezh2 cKO shows no significant difference, indicating the overall structure of the cerebellum is not dramatically altered and the phenotype is limited to the early migration out of the inner EGL. I) Quantification of the anti-p27 fluorescence intensity by layer shows a decrease in the fluorescence in Ezh2 cKO versus WT, for the inner EGL (p-value 0.028, t-test) and IGL (p-value 0.013, t-test). The decrease in p27 abundance suggests that the total number of neurons within the inner EGL and the IGL density of cells within those areas is reduced.

Ezh2 cKO leads to premature process extension and NeuN expression.

GNPs cultured for 24 in Shh, were labeled with DAPI (nucleus) and for MAP2 (neuronal processes) (A,C), then segmented with Neuronctyo 2 (B,D). Segmented images showed increased process extension in Ezh2 cKO (D for cKO versus WT in B). F) Mean process length comparison between Ezh2 cKO and WT shows cells in Ezh2 cKO mice had significantly longer processes (n = 5). E) Quantification of the number of cells with no process, a process less than 5, 10, 50, 100 pixels, or greater then 100 pixels, showed a significant increase in the number of processes longer then 50 pixels. P-value determined by T-test.

Transcriptional regulation of known H3K27me3 effector protein complexes during granule neuron (GN) development.

A) Rank order plot of genes that change expression during GN development shows Ezh2 as the gene with the largest decrease in RNA abundance. The y-axis indicates the combined fold change for both datasets. B) RNA abundance over time for Ezh2. C) Schematic of granule neuron (GN) development in a P7 cerebellum sagittal slice. A P7 cerebellar slice stained with Anti-Ezh2 (left), DAPI (middle), and NeuN (right) shows reduced Ezh2 staining intensity in the postmitotic cells of the iEGL and EGL. D) Schematic of GNP culture without Shh, which causes GNPs to immediately exit cell cycle. Ezh2 protein abundance was measured after 6 h and 24, splitting cells into proliferating (high Math1>GFP) cells and non-proliferating (low Math1>GFP) cells (n=6). The decrease in Ezh2 shortly after cell cycle suggests that the decrease in Ezh2 is linked to cell cycle exit but does not show a causal relationship.

In medulloblastoma cells GNP differentiation genes are transcriptionally repressed and H3K27me3 modified.

A) H3K27me3 quantified over the promoters of GNP differentiation genes in GNPs (x-axis) and MB (y-axis). Genes are color-coded based upon H3K27me3-marked versus not marked in GNP and MB. Genes whose promoters show more H3K27me3 modification in GNP are shown in red, while those higher in MB are blue and those modified in both are purple. Duller colors (dark tones) were used if higher relative expression in MB than in GNPs. Higher saturation colors indicate gene expressed relatively higher in GNPs than in MB. The colour legend is seen bellow panel A. B) Transcript levels of differentiation genes in P7 GNPs, x-axis) versus in Ptch1+/- derived MB cells (y-axis). Colors are inherited from panel A with dark tones used for genes in MB (2 fold higher expression and adjusted p-value <0.05) and high saturation colours when higher expression in GNPs. C) Sankey diagram showing how the H3K27me3 modified differentiation genes are expressed in MB. The second column shows the differentiation genes in GNP separated by H3K27me3 modification. The third column compares MB H3K27me3 ChIP, showing of 447 H3K27me3 modified GNP differentiation genes, 336 (75%) of those genes are also H3K27me3 modified in MB. The fourth column shows differential expression between P7 GNPs and MB. Even when the H3K27me3 is not present in MB cells only 8% of the differentiation genes show an increase in expression in MB compared to P7 GNPs. The fifth column shows the final classification.

Human SHH MBs with High EZH2 are associated with worse outcome and show lower expression of GNP differentiation genes.

A) Labeled dot histogram of EZH2 transcript abundance from 223 human SHH MB with the top (orange) and bottom (brown) 30% labeled. Each human tumor is represented as a dot with the symbols reflecting the most recent SHH MB classification (9). Note the high number of patients with gamma-class SHH MB with low EZH2 expression. These patients are infants with a typically good prognosis. B) Comparison of average gene expression between the human MB samples from the 30% highest (x-axis) and lowest EZH2 (y-axis) expressing tumours. The color labels show more highly expressed genes that are 1.5 fold higher (adjusted p < 0.05) in tumours that were low in EZH2 were labelled orange, while those that are higher in EZH2 high tumours are brown. The GNP differentiation genes (red) show considerable overlap with the genes highest in EZH2 low tumours (orange), while the GNP proliferation genes (green) show overlap with the genes that are highest in EZH2 high tumours (brown). C) Quantification of the extensive overlap between GNP differentiation genes and genes highly expressed in EZH2 low human MB tumours. The categories of genes based on the GNP time course (y-axis) are quantified using a bar graph that shows the intersection of genes for each GNP group to either genes that are highly expressed when Ezh2 is low (EZH2low) or genes that are highly expressed when Ezh2 is high (EZH2high). Enrichment is calculated as the number of GNP group genes that intersect (∩) with EZH2low genes over the number that intersect with EZH2high genes as described in the equation. The p-value is determined by the hypergeometric test. D) Kaplan-Meyer curves show a significant 10 year survival difference for SHH MB patients with high versus low EZH2 transcript.

Treatment of MB cells with Ezh2 and CDK4/6 inhibitors leads to differentiation.

A) Schematic of treatment and imaging protocol. MB cells were treated using the Ezh2 inhibitor UNC1999 with or without the CDK4/6 inhibitor palbociclib. B) Quantification of immunofluorescence in treated MB cells (n = 16), derived from Ptch1+/- mice, stained for NeuN, pRb (Ser807/811) and DAPI to categorize the cells as in G1/2 (green), G0 (yellow) and differentiating (G0 with high NeuN (red)). Note an increase in fraction of cells scored as differentiating with combined Palbociclib and UNC1999 treatment. C) Schematic of treatments with inhibitors followed by drug wash out. D) Immunofluorescence images of fields of MB cells stained for NeuN (red), pRb (green), DAPI (blue) and phalloidin (grey) after treatment for 72h with 1 uM Palbociclib (i-iii) combined with 1 uM Palbociclib or 5 uM UNC1999 (iv-vi) followed by washout as per C. Dashed boxes in i and iv show regions enlarged in ii and v respectively, with the individual colour channels shown in separate panes. Note fewer pRB positive cells with UNC1999 and Palbociclib (iv) than with Palbociclib alone (i). E) Quantification of fractions of cells in G0, G1/2, and G0 and differentiating as described in B (n = 8). Note that the MB cells treated with both UNC1999 and Palbociclib showed persistent cell cycle exit even after washout. P-values in B and E calculated using ANOVA with post-hoc Tukey test.

Antibodies and concentrations used for ChIP

Mice alleles used

Antibodies used for immunofluorescence