ZIC1 is a context-dependent medulloblastoma driver in the rhombic lip

Abstract

Transcription factors are frequent cancer driver genes, exhibiting noted specificity based on the precise cell of origin. We demonstrate that ZIC1 exhibits loss-of-function (LOF) somatic events in group 4 (G4) medulloblastoma through recurrent point mutations, subchromosomal deletions and mono-allelic epigenetic repression (60% of G4 medulloblastoma). In contrast, highly similar SHH medulloblastoma exhibits distinct and diametrically opposed gain-of-function mutations and copy number gains (20% of SHH medulloblastoma). Overexpression of ZIC1 suppresses the growth of group 3 medulloblastoma models, whereas it promotes the proliferation of SHH medulloblastoma precursor cells. SHH medulloblastoma ZIC1 mutants show increased activity versus wild-type ZIC1, whereas G4 medulloblastoma ZIC1 mutants exhibit LOF phenotypes. Distinct ZIC1 mutations affect cells of the rhombic lip in diametrically opposed ways, suggesting that ZIC1 is a critical developmental transcriptional regulator in both the normal and transformed rhombic lip and identifying ZIC1 as an exquisitely context-dependent driver gene in medulloblastoma.

Fig. 1. Characterization of subgroup-specific chromatin landscape of medulloblastoma.

a, Summary of the newly generated and public datasets. Number within the bracket indicates the number of tumors with previously published data. b, Hierarchical clustering plots generated using the top 10,000 variable H3K27ac and H3K27me3 ChIP–seq peaks. c, Schematic representation summarizing different types of ChIP–seq peaks used in downstream analysis. Subgroup-specific peaks were defined by identifying peaks that (1) exhibit subgroup enrichment in ChIP–seq read counts or (2) are recurrently present only for specific subgroups even if the average ChIP–seq read count is not strongly subgroup enriched on average. d, Number of subgroup-specific peaks for each subgroup in the H3K27ac cohort. After batch correction, peaks annotated as subgroup enriched for ChIP–seq reads or subgroup recurrent were characterized separately. e, Number of subgroup-specific H3K27me3 peaks using the same annotations/criteria as d. f, Number of peaks and proportion of genome covered by H3K27ac and H3K27me3 peaks across the medulloblastoma subgroups. P values were calculated by the tailed Mann–Whitney U test. Biological sample size for H3K27ac—G3/G4/SHH/WNT = 27/47/39/10 and H3K27me3—G3/G4/SHH/WNT = 14/24/22/3. Center of box, median. Bounds of box, 25% and 75% percentile. Whiskers show minimum and maximum values within the 1.5× interquartile range. g, Schematic representation summarizing how high-confidence enhancer–promoter interactions were identified from HiChIP and ChIP–seq data. Adjusted P values were calculated using Pearson correlation between target gene transcript and enhancer H3K27ac read levels, which was corrected for multiple testing. h, Summary of distance distribution for high-confidence enhancer–promoter interactions. Proportion of SCLs (g; Methods) over a total number of loops is depicted as overlapping Venn diagrams. Double asterisk (**) indicates a significant correlation (P.adj < 0.1).

Fig. 2. ZIC1 is recurrently mutated and repressed by H3K27me3 in G4 medulloblastoma.

a, Overlap between genes regulated by subgroup-specific H3K27me3 peaks in G3, G4 medulloblastoma and genes recurrently mutated in each subgroup. b, Ranking of SEs across medulloblastoma subgroups, showcasing the number of total SEs identified (in gray) as well as the proportion of subgroup-enriched SEs in pie charts. c, ZIC1 and ZIC4 expression patterns across medulloblastoma subgroups. Biological sample size—G3/G4/SHH/WNT = 72/122/93/24. P values from two-tailed Mann–Whitney U test. Center of box, median. Bounds of box, 25% and 75% percentile. Whiskers show minimum and maximum values within 1.5× interquartile range. d, Sequencing depth normalized bigwig tracks showcasing recurrent (n ≥ 3 per subgroup) ZIC1 and ZIC4 chromatin states across four subgroups. e, Summary of chromatin states observed at the ZIC1 promoter across all samples in the ChIP–seq libraries with both H3K27ac and H3K27me3 modifications. f, Expression levels of ZIC1 and ZIC4 in G3/G4 medulloblastoma samples that harbor both H3K27ac and H3K27me3 (AM) or just H3K27ac (A) peaks on the ZIC1 promoter. Biological sample size for G4—AM/A = 6/18 (24 total) and G3—AM/A = 3/11 (14 total). P values from two-tailed Mann–Whitney U test. Same whisker box plot parameters as c. g, Density plot summarizing H3K27ac versus H3K27me3 signal at H3K27ac and H3K27me3 peaks. Correlations between H3K27ac and H3K27me3 were calculated by Pearson correlation on merged peak coordinates. h, Method for inferring heterozygous SNPs using H3K27ac and H3K27me3; two mutually exclusive histone modification marks. i, Distribution of inferred heterozygous SNPs across H3K27ac and H3K27me3 libraries of four G4 samples and one SHH sample with H3K27ac and H3K27me3 peaks on the ZIC1 promoter.

Fig. 3. ZIC1/ZIC4 exhibit mono-allelic expression patterns in G3 and G4.

a, SEs that are recurrently (n ≥ 3 for G4 and n ≥ 2 for others) mono-allelic across different medulloblastoma subgroups. SEs that harbor SNPs (phased and pooled for each allele) that are heterozygous in WGS but homozygous (normalized allelic frequency ≥ 0.9) in H3K27ac ChIP–seq reads (same SNPs) from the same sample were defined as mono-allelic. Dot plots above each SE show differences in pooled allelic frequencies for heterozygous SNPs (allele A–B) in (1) H3K27ac reads from the SE (left) and (2) RNA-seq reads from the SE target gene (right). Matching samples are connected by lines between SE and RNA. b, Allelic frequency summary for heterozygous germline SNPs for ZIC1 and ZIC4 transcripts in RNA-seq within the validation cohort (251 samples with both WGS and RNA-seq data). Adjusted P values from two-tailed pairwise Fisher’s exact test. c, Whisker box plots summarizing ZIC1 and ZIC4 expression cross the medulloblastoma subgroups, but G3 and G4 are divided according to mono-allelic (mono) versus bi-allelic (bi) expression of ZIC1 or ZIC4. Biological sample size: G3_bi/G3_mono = 19/8, G4_bi/G4_mono = 24/44 and SHH/WNT = 93/24. P values from two-tailed Mann–Whitney U test. Center of box, median. Bounds of box, 25% and 75% percentile. Whiskers show minimum and maximum values within the 1.5× interquartile range. d, Mutational landscape of ZIC1 in G4 and SHH. e, Allelic frequency distribution for ZIC1 mutations in G4 (n = 3) and SHH (n = 2) samples from the assembled validation cohort. f, ZIC1 VAF obtained from published medulloblastoma RNA-seq data. P value from two-tailed Mann–Whitney U test.

Fig. 4. ZIC1/ZIC4 mono-allelic and bi-allelic G3/G4 medulloblastomas enrich for distinct mutations.

a, Whisker box plot of normalized PRDM6 transcript counts in bi-allelic versus mono-allelic ZIC1/ZIC4 G4 samples. PRDM6 transcription occurs exclusively in the context of single allele inactivation of ZIC1/ZIC4. b, Oncoplot showcasing mutation status of previously published recurrently mutated genes in mono-allelic and bi-allelic G4 samples. Each column represents different samples. Each row represents different genes that are recurrently mutated in medulloblastoma. Distinct types of mutations for a gene in each patient are depicted with different size/colored bars. c, Oncoplot showcasing mutation status of previously published recurrently mutated genes in mono-allelic and bi-allelic G3 samples. d, Sample distribution summary and two-tailed Fisher’s exact test outputs for the significance of enrichment for chromatin modifier mutations in ZIC1/ZIC4 mono-allelic G4 and G3, as well as KBTBD4 mutation in ZIC1/ZIC4 bi-allelic G4. e, Summary of different proportions of G4 medulloblastoma samples exhibiting transcriptional repression within the chromatin (H3K27me3) data or RNA (mono-allelic expression) data. f, Sequencing depth normalized bigwig tracks for H3K27ac and H3K27me3 in one G4 sample with bi-allelic ZIC1/ZIC4 SE and two G4 samples with mono-allelic ZIC1/ZIC4 locus SE. Not all G4 samples with mono-allelic ZIC1/ZIC4 SE harbor H3K27me3 peak on the locus. g,h, Allelic frequencies for heterozygous SNPs in WGS, H3K27ac and H3K27me3 ChIP–seq data in the two mono-allelic G4 samples: MDT-AP-1168, where H3K27me3 is observed, and MDT-AP-2673, where H3K27me3 is absent on the ZIC1/ZIC4 locus.

Fig. 5. ZIC1/ZIC4 locus exhibits distinct genomic rearrangements in G3/G4 and SHH medulloblastoma.

a, CNA track for medulloblastoma samples exhibiting ZIC1/ZIC4 locus copy gain/loss. b, GISTIC output for SHH medulloblastoma, highlighting 2p24.3 (MYCN), 2q14.2 (GLI2) and 3q23 (ZIC1/ZIC4) gain. FDR, false discovery rate. c, CNA summary for the ZIC1/ZIC4 locus per medulloblastoma subgroups. Adjusted P values from two-tailed pairwise Fisher’s exact test. d, Chr3q and ZIC1/ZIC4 focal copy deletion frequency across three subtypes of G4 medulloblastoma. P values from two-tailed Fisher’s exact test and Hochberg correction. e, Breakdown of chromatin repression of a single allele of ZIC1/ZIC4 locus across three subtypes of G4 medulloblastoma. P values were calculated by two-tailed Fisher’s exact test followed by Hochberg multiple correction. f,g. Allelic frequencies for heterozygous germline SNPs across normal tumor DNA and tumor RNA from a representative G4 sample with (f) chr3q deletion and (g) epigenetic suppression of the ZIC1/ZIC4 locus. h, Whisker box plots for ZIC1 and ZIC4 expression in SHH medulloblastoma tumors with chr3 copy gain versus neutral. Expression values from RNA-seq data with matching SNP6 array data. P values were calculated from the two-tailed Mann–Whitney U test. Center of box, median. Bounds of box, 25% and 75% percentile. Whiskers show minimum and maximum values within the 1.5× interquartile range. i, Whisker box plots for ZIC1 and ZIC4 expression in G4γ medulloblastoma with chr3q copy loss versus copy neutral. Expression values from expression array data with matching SNP6 array data. Same statistical test and whisker box plot parameters as h. j, Breakdown of ZIC1/ZIC4 allelic expression pattern, ZIC1/ZIC4 CNA and ZIC1 SNVs in medulloblastoma samples with both RNA-seq and WGS data available, as well as harboring heterozygous germline SNPs in ZIC1/ZIC4 exons.

Fig. 6. ZIC1/ZIC4 reduces G3 medulloblastoma cell proliferation both in vitro and in vivo.

a, 3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium (MTS) cell proliferation assay results (mean ± s.d.) for D425. Three biological replicates. P values from two-tailed Welch t-test. b, Cell proliferation assay results for D283. P values from two-tailed Welch t-test. Data points show mean ± s.d. Five biological replicates. Center of box, median. Bounds of box, 25% and 75% percentile. Whiskers show minimum and maximum values within the 1.5× interquartile range. c, Western blot validation of ZIC1/ZIC4 overexpression in D283 and D425. d, Pathway analysis for ZIC1/ZIC4 versus EV (BFP) overexpressing D425 (RNA-seq, biological n = 3). e,f. Representative images (e) and whisker box plots (f) summarizing BLI signals in BFP versus ZIC1/ZIC4 overexpressing D425-injected mice. P values were calculated by two-tailed Welch t-test. Same whisker box plot parameters as b. g, Survival curves for BFP versus ZIC1/ZIC4-transduced D425-injected mice. P values from two-tailed log-rank test. h, Normalized bigwig tracks showcasing chromatin state of ZIC1/ZIC4 locus in patient-derived G3 xenograft line MB051. i, Allelic frequency of heterozygous SNP rs6766244 on coding exon of ZIC4 from MB051 RNA-seq and H3K27me3 ChIP–seq counts, and Sanger sequencing result from the tumor DNA for the same SNP. j, ZIC1/ZIC4-normalized counts from RNA-seq in MB051 (biological n = 3 for EV and ZIC1/ZIC4 constructs). Mean ± s.d. k,l. Representative images (k) and whisker box plots (l) summarizing BLI signals in BFP versus ZIC1/ZIC4 overexpressing MB051-injected mice. P values were calculated by two-tailed Welch t-test. Same whisker box plot parameters as b. m, Survival curves for BFP versus ZIC1/ZIC4-transduced MB051-injected mice. P values from two-tailed log-rank test. H3, histone 3; EV, empty vector. Source data

Fig. 7. ZIC1 mutations from G4 and SHH medulloblastoma are functionally distinct.

a, AlphaFold2 predicted structure of ZIC1. Mutant constructs generated and used in the study are summarized in the structure. b, Proliferation assay for D425 G3 cell line transduced with ZIC1 mutant constructs and mCherry EV. Three technical replicates for each construct. Mean ± s.d. P values from two-tailed Welch t-test. c, Schematic representation for the cell competition assay using D283. d, Cell competition assay results using D283 transduced with ZIC1 mutant constructs and mCherry EV. Three technical replicates for each construct. Mean ± s.d. P values from two-tailed Welch t-test. e, Representative western blot visualization of exogenous ZIC1 expression in D283 transduced with FLAG-ZIC1 constructs. f, Whisker box plots showing exogenous ZIC1 expression in D283 transduced with FLAG-ZIC1 constructs. Signals were normalized by transduction efficiency and GAPDH levels. Center of box—median. Bounds of box—25% and 75% percentile. Whiskers show minimum and maximum values within the 1.5× interquartile range. P values from two-tailed Welch t-test. g, Representative cycloheximide chase results for WT and mutant ZIC1 constructs in D283. h, Comparison of ZIC1 protein level across varying exposure times to cycloheximide for WT (n = 2), G4 medulloblastoma mutant (n = 4) and SHH medulloblastoma ZIC1 mutant (n = 4) constructs. n, biological replicates. Mean ± s.d. P values from two-tailed Welch t-test. i, Number of DEG (DESeq2 output) for ZIC1 constructs when compared against EV or WT ZIC1. Q value cutoff of 0.05. j, Volcano plot summarizing differentially expressed genes between WT ZIC1 and EV. k, Distribution of normalized reads from FLAG ChIP–seq peaks from FLAG-tagged WT versus G4 medulloblastoma mutant ZIC1-transduced D283. DEG, differentially expressed genes. Source data

Fig. 8. ZIC1 is a GOF driver oncogene in SHH medulloblastoma.

a, Schematic representation summarizing GNP 5-ethynyl-2'-deoxyuridine (EdU) proliferation assay. FACS, fluorescence activate cell sorting. b, Summary of EdU proliferation assay for GNP transduced with ZIC1 mutant constructs and mCherry EV. GNPs enriched from multiple mouse cerebellums were used to generate biological triplicates for each construct. Mean ± s.d. as error bars. P values were calculated by two-way ANOVA. c, Representative results from two independent replicates from running cycloheximide (CHX) chase on GNP transduced with WT ZIC1 construct, two G4 medulloblastoma ZIC1 mutant constructs and two SHH medulloblastoma ZIC1 mutant constructs. d, Comparison of ZIC1 protein level from GNP across varying exposure times to cycloheximide for WT (n = 2), G4 medulloblastoma mutant (n = 4) and SHH medulloblastoma ZIC1 mutant (n = 4) constructs. n, biological replicates. Mean ± s.d. P values were calculated by two-tailed Welch t-test. e, RNA-seq-derived volcano plot summarizing DEG (DESeq2 output) between ZIC1 (mch⁺ ZIC1⁺) and EV (mch⁺) transduced granule cells. Two biological replicates were generated for bulk granule cells and sorted GNPs (biological n = 4). Q value cutoff of 0.05. f, Normalized RNA-seq counts for Gli2 transcript in EV (mch⁺) and ZIC1 (mch⁺ ZIC1⁺) transduced GNPs. Adjusted P value from differential expression was calculated from DESeq2 differential expression analysis. g, Top ten pathways upregulated with ZIC1 overexpression in bulk granule cells and GNPs. h,i, Summary of normal rhombic lip development (h) as well as epigenetic and genetic events (i) that lead to ZIC1 LOF in G3 and G4 medulloblastoma and ZIC1 GOF in SHH medulloblastoma. ANOVA, analysis of variance; NSC, neural stem cell. Source data

Extended Data Fig. 1. Medulloblastoma exhibits subgroup-specific master transcription factors (TFs) and chromatin landscape.

a, Saturation analysis for H3K27ac and H3K27me3 peak identification. For each number of samples shown on the x axis, a subset of total cohort of ChIP–seq samples corresponding to this number was randomly selected. Number of non-overlapping peaks identified from this subset were recorded for each iteration of random sampling. Average and standard deviation for 10 iterations were plotted for each number up to total cohort size. Number of peaks identified starts to plateau toward the end of the curve, suggesting that addition of new samples will likely lead to diminishing returns. b, Annotation for typical enhancers, super-enhancers (SE) and H3K27me3 peaks that are classified as (1) all peaks found in the subgroup, (2) subgroup-enriched peaks (defined in Fig. 1c) and (3) subgroup-recurrent peaks (defined in Fig. 1c). P values were calculated by performing two-tailed chi-square test on H3K27me3 peaks. Standardized residuals for chi-square tests performed on H3K27me3 peak distributions were also calculated. c. Strategy used to define core regulatory circuit (CRC) score for each transcription factor for each subgroup. In degree (number of TFs that target the TF of interest) and out degree (number of TF promoters targeted by the TF of interest) were calculated for each TF to identify subgroup-specific and pan-subgroup core TFs. d, Heatmap summarizing pan subgroup and subgroup-specific core TFs crucial for shaping core circuitry landscape for each subgroup. e, Top 5 subgroup-specific master transcription factors identified for each subgroup according to CRC score. f, Number of genes assigned for each enhancer across enhancer–promoter interactions identified using HiChIP and 27ac ChIP–seq data. g, Proportion of enhancers that target the closest genes for SHH, G3 and G4 subgroups.

Extended Data Fig. 2. Overlap between recurrent copy number deletions and subgroup enriched/recurrent H3K27me3 peaks for group 3 (G3)/group 4 (G4) medulloblastoma.

a, Venn diagram depicting overlap between subgroup-enriched H3K27me3 peaks with recurrently mutated genes in WNT, SHH as well as genes recurrently affected by focal deletion (<12 Mb) in all 4 subgroups (Supplementary Table 13). b, BCOR mutation pattern identified in SHH medulloblastoma. c, Breakdown of BCOR H3K27me3 pattern in SHH medulloblastoma. Highly female-enriched pattern is observed, suggesting that X inactivation may have a role in the observed chromatin phenomenon. d, Showcase of recurrent deletion of 2q37.3 locus identified in G3 and G4. MIR4786 locus exhibits a G3/G4-enriched copy loss pattern (Supplementary Table 13). e, Representative H3K27me3 ChIP–seq signal patterns for all subgroups on BCOR and MIR4786 locus, which exhibit SHH-enriched and G3/G4-enriched H3K27me3 signal, respectively (Supplementary Tables 12 and 13). f, Read depth normalized 27ac bigwig tracks for a representative sample from each subgroup. Bidirectional promoters regulating ZIC1 and ZIC4 transcription are regulated by a common super-enhancer identified across all subgroups. g, H3K27ac signal strength of SE overlapping ZIC1/4 promoter across MB subgroups. Biological sample size: G3/G4/SHH/WNT = 27/47/39/10. Center of box—median. Bounds of box—25% and 75% percentile. Whiskers show minimum and maximum values within the 1.5× interquartile range. P values from two-tailed Mann–Whitney U test. h, ZIC1- and ZIC4-normalized transcript count levels in ChIP cohort samples with matching H3K27ac, H3K27me3 and RNA-seq data (N = 58). Biological sample size: G3/G4/SHH/WNT = 13/24/18/3. Box plot parameters same as g. P values from two-tailed Wilcoxon rank-sum test. i, Allelic frequencies for the inferred heterozygous single-nucleotide polymorphisms (from Fig. 2h,i) in 2 G4 samples with matching WGS data.

Extended Data Fig. 3. ZIC1/4 locus is regulated by multiple super-enhancers (SE) that are recurrently epigenetically repressed on single alleles.

a, Allelic frequencies for heterozygous SNPs present in both H3K27ac ChIP–seq reads on ZIC1/4 SE as well as RNA-seq reads on ZIC1/4 exons. Identical schematic to dot plots from Fig. 3a, but only the exact match heterozygous SNPs identified in both H3K27ac ChIP–seq and RNA-seq data were used. Matching samples are connected by lines between SE and RNA columns. Y axis shows difference in pooled allelic frequency between SNPs from the two different alleles. ZIC1/4 RNA and SE exhibit bias for the same alleles from the heterozygous single-nucleotide polymorphisms (SNPs), suggesting that the monoallelic SE drives monoallelic expression. b, Correlation between H3K27ac reads on two SEs that target ZIC1/4 locus (from Extended Data Fig. 2g), SE2954 and SE2957, and ZIC1/ZIC4 transcript levels in group 3 (G3) and group 4 (G4) medulloblastoma. P values generated from two-tailed Spearman correlation analysis. c, ZIC1/4 targeting SEs, their interaction maps with ZIC1/4 locus and frequency of their monoallelic status in G3 and G4 medulloblastoma. SE directly on top of ZIC1/4 genes (SE2957) was monoallelic in 9 out of 19 samples in G4 and 3 out of 7 samples in G3. SEs upstream (SE2954) and downstream (SE2958) of ZIC1/4 locus are also recurrently monoallelic and were identified as high-confidence enhancer–promoter interactions with HiChIP, H3K27ac ChIP–seq and RNA-seq data. While most samples harbored SE2957, a smaller proportion of G3 and G4 samples harbored SE2954 and SE2958. d, Allelic frequency distribution of heterozygous germline SNPs for ZIC1 and ZIC4 transcripts in RNA-seq within the validation cohort (total of 251 samples with both WGS and RNA-seq data). A total of 190 samples contain heterozygous SNPs within ZIC1/4 exons in both normal control and tumor DNA.

Extended Data Fig. 4. Genetic and transcriptional patterns associated with biallelic and monoallelic status of ZIC1/4 across medulloblastoma.

a, Volcano plot summarizing differentially expressed genes between ZIC1/4 monoallelic and biallelic group 4 (G4) samples. Q value threshold of 0.01 and log₂(fold change) threshold of 2 were used. b, Oncoplot summarizing the mutational landscape of SHH tumors with or without ZIC1 mutations. U1 snRNA mutations were always mutated together (RNU1-2, RNVU1-18) with ZIC1. c, Whisker box plot summarizing neuronal differentiation score for group 3 (G4) and G4 medulloblastoma tumors. Previously published 39 G3/G4 neuronal differentiation signature genes (Supplementary Table 14) were used to calculate the overall differentiation score for each tumor. Biological sample size: G3/G4 = 72/122. P value was calculated by two-tailed Mann–Whitney U test. Center of box—median. Bounds of box—25% and 75% percentile. Whiskers show minimum and maximum values within the 1.5× interquartile range. d, Scatter plot showing expression level of ZIC1 across G3 and G4 medulloblastoma tumors vs. differentiation score in the same tumors. e, Hierarchical clustering of G3/G4 samples by top 10,000 variable genes from transcriptome. ZIC1/4 monoallelic G3/G4 samples do not form distinct clusters from the biallelic samples. f, Hierarchical clustering of G3/G4 samples by expression level of the neuronal differentiation signature genes from c. ZIC1/4 monoallelic G3/G4 samples do not form distinct clusters from the biallelic samples. g, Frequency of somatic mutations on super-enhancer (SE) on top of ZIC1/4 locus (SE2957) across WNT, SHH, G3 ZIC1/4 biallelic, monoallelic, G4 ZIC1/4 biallelic and monoallelic samples. h, Breakdown of somatic mutation patterns on SE2957 for all subgroups.

Extended Data Fig. 5. MB051 exhibits similar transcriptional changes as D425 upon ZIC1/4 overexpression in vivo.

a, Immunofluorescence showing tumor cells (GFP+) and ZIC1 protein level (Alexa Fluor 555), both separately and merged, for BFP (empty vector) or ZIC1/4-transduced MB051 patient-derived group 3 (G3) medulloblastoma xenograft intracranially injected into NOD SCID γ (NSG) mice. One biological replicate for BFP-transduced MB051, and two biological replicates for ZIC1/4-transduced MB051. Two fields of views captured for BFP, and four fields of views captured for ZIC1/4-transduced MB051 (3 for one biological replicate and 1 for another). All views exhibited identical observations. b, Top 10 pathways upregulated in D425 in vitro upon overexpression of ZIC1/4 compared to BFP empty vector. c, Top 10 pathways upregulated in MB051 in vivo upon overexpression of ZIC1/4 compared to BFP empty vector. d, Pathway analysis depicting commonly upregulated pathways between D425 in vitro and MB051 in vivo. While there was a small overlap, neuronal differentiation pathway emerged as a commonly upregulated pathway between two different models. e, Top 10 pathways downregulated in D425 in vitro upon overexpression of ZIC1/4 compared to BFP empty vector. f, Top 10 pathways downregulated in MB051 in vivo upon overexpression of ZIC1/4 compared to BFP empty vector.

Extended Data Fig. 6. ZIC1/4 overexpression does not result in morphological differences for MB051 at the H&E level.

Representative H&E results at various magnifications generated from injecting MB051 into NOD SCID γ (NSG) mice. Magnifications are shown on the left side of the panels. MB051 was transduced with BFP (empty vector) or ZIC1/4 overexpression construct prior to injection. Minimal morphological differences were observable between the different constructs. One biological replicate for BFP-transduced MB051, and two biological replicates for ZIC1/4-transduced MB051. Three fields of views captured for BFP and each biological replicate of ZIC1/4-transduced MB051. Twenty-one fields of views for BFP-transduced MB051, 20 fields of views for one replicate of ZIC1/4-transduced MB051 and 27 fields of views for the other replicates. Images were captured at varying magnifications ranging from ×2, ×10, and ×40.

Extended Data Fig. 7. Group 4 (G4) and SHH medulloblastoma ZIC1 mutant overexpression result in distinct transcriptional changes in group 3 (G3) cells.

a, ZIC1 transcript levels (qRT-PCR) across the biological and technical replicates of G3 cell lines transduced with ZIC1 constructs. Primers used are in Supplementary Table 1. b, Volcano plot summarizing genes differentially expressed in G4 medulloblastoma mutant vs. wild-type (WT) ZIC1 and SHH medulloblastoma mutant vs. WT ZIC1-transduced G3 medulloblastoma cells (D425 and D283). Genes that are upregulated with WT ZIC1 compared to empty vectors are highlighted in purple. P adjusted threshold of 0.05 was used. c, Heatmap showcasing expression pattern of all WT ZIC1-induced genes across all ZIC1 mutation construct overexpressing cells. G4 medulloblastoma ZIC1 mutants exhibit reduced upregulation of the ZIC1 target genes, whereas SHH medulloblastoma ZIC1 mutants exhibit augmented upregulation of these genes. d, Pathway analysis of genes upregulated with WT ZIC1 construct compared to empty vector. e, Pathway analysis of genes that are downregulated with G4 medulloblastoma ZIC1 mutant compared to WT ZIC1. f, Pathway analysis of genes upregulated by SHH medulloblastoma ZIC1 mutant compared to WT ZIC1. g, Number of ChIP–seq peaks identified from Flag-tagged ZIC1 ChIP–seq in D283 cells transduced with WT ZIC1 or G4 medulloblastoma ZIC1 mutant. Two biological replicates were generated for each arm, using different constructs for the G4 medulloblastoma ZIC1 mutants.

Extended Data Fig. 8. ZIC1 regulates Gli2 and cell cycle pathway genes in granule cells.

a, Number of ChIP–seq peaks identified from Flag-tagged ZIC1 ChIP–seq in granule neuron progenitor (GNP) cells transduced with wild-type (WT) ZIC1 or group 4 (G4) medulloblastoma ZIC1 mutant. Two biological replicates were generated for WT ZIC1 and three for G4 medulloblastoma ZIC1 mutants. b, Distribution of normalized reads for WT vs. G4 medulloblastoma mutant Flag-tagged ZIC1-transduced GNP cells across peaks identified from FLAG ChIP–seq. c, Schematic summarizing the RNA-seq libraries generated from mouse granule lineage cells. d, Top 10 pathways downregulated by ZIC1 overexpression compared to empty vector in bulk granule cells and GNPs. e, Expression level of GLI2 across different medulloblastoma molecular subgroups. Plot was generated using the RNA-seq cohort used in the study (N = 311). GLI2 exhibits a highly SHH medulloblastoma-specific expression pattern. Center of box—median. Bounds of box—25% and 75% percentile. Whiskers show minimum and maximum values within the 1.5× interquartile range. P values calculated by two-tailed Mann–Whitney U test. f, Zic1/2 ChIP–seq track demonstrating presence of peaks on the Gli2 promoter in 2 immunoprecipitation replicates but not in input (data for f–h from GSE60731). g, Volcano plot summarizing genes differentially expressed by knocking down Zic1 from mouse GNP. P adjusted threshold = 0.05. h, Normalized counts of Gli2 transcript in control shRNA and Zic1 shRNA treated GNP. Biological sample size = 2 for each arm. P adjusted value was obtained from DESeq2 differential expression analysis.

Extended Data Fig. 9. ZIC1/4 are expressed throughout the rhombic lip, particularly in the rhombic lip ventricular zone (RL-VZ) and rhombic lip subventricular zone (RL-SVZ).

a, Breakdown of glutamatergic neuronal cell lineage from developing human cerebellum (panel a–c from ref. data). RL-SVZ cell populations were further subdivided according to expression pattern of KI67, EOMES and ATOH1. b, Violin plots summarizing expression level of ZIC1, ZIC4, KI67 and other transcription factors critical for rhombic lip development throughout distinct glutamatergic lineage cell types. c, Feature plot summarizing expression levels for 12 developmental transcription factors across the developing human rhombic lip. d, Bulk RNA-seq quantification of ZIC1 and ZIC4 transcript levels across human rhombic lip regions isolated by laser capture microdissection (LCM; ref. data). Center of box—median. Bounds of box—25% and 75% percentile. Whiskers show minimum and maximum values within the 1.5× interquartile range. P values from two-tailed Mann–Whitney U test. e, RNA-scope visualization of ZIC1 and ZIC4 expression pattern across different regions of the rhombic lip in developing human cerebellum (11–19 postconception weeks). High expression level of both transcripts is observed across all regions, particularly in the RL-VZ and RL-SVZ. Biological sample size of 1 for 11, 14, 17 and 19 post-conception weeks (PCW). f, Immunofluorescence result showcasing ZIC1 protein expression pattern across different regions of the rhombic lip in developing human cerebellum (11–17 postconception weeks). Biological sample size of 1 for 11, 14 and 17 PCW. Three different sections were used for each sample. Representative images are shown. g, Violin plots summarizing expression level of ZIC1 transcript across different cells of the developing cerebellum (ref. data).