Integrative multi‐omic analysis reveals neurodevelopmental gene dysregulation in CIC ‐knockout and IDH1 ‐mutant cells

Abstract Capicua (CIC)'s transcriptional repressor function is implicated in neurodevelopment and in oligodendroglioma (ODG) aetiology. However, CIC's role in these contexts remains obscure, primarily from our currently limited knowledge regarding its biological functions. Moreover, CIC mutations in ODG invariably co‐occur with a neomorphic IDH1/2 mutation, yet the functional relationship between these two genetic events is unknown. Here, we analysed models derived from an E6/E7/hTERT‐immortalized (i.e. p53‐ and RB‐deficient) normal human astrocyte cell line. To examine the consequences of CIC loss, we compared transcriptomic and epigenomic profiles between CIC wild‐type and knockout cell lines, with and without mutant IDH1 expression. Our analyses revealed dysregulation of neurodevelopmental genes in association with CIC loss. CIC ChIP‐seq was also performed to expand upon the currently limited ensemble of known CIC target genes. Among the newly identified direct CIC target genes were EPHA2 and ID1, whose functions are linked to neurodevelopment and the tumourigenicity of in vivo glioma tumour models. NFIA, a known mediator of gliogenesis, was discovered to be uniquely overexpressed in CIC‐knockout cells expressing mutant IDH1‐R132H protein. These results identify neurodevelopment and specific genes within this context as candidate targets through which CIC alterations may contribute to the progression of IDH‐mutant gliomas. © 2021 The Authors. The Journal of Pathology published by John Wiley & Sons, Ltd on behalf of The Pathological Society of Great Britain and Ireland.


Introduction
Capicua (CIC) functions downstream of receptor tyrosine kinase (RTK) signalling through a mechanism called default repression: in the absence of RTK signals, CIC maintains its transcriptional repressor activity, whereas induction of RTK signalling results in inactivation of CIC and subsequent de-repression of its target genes [1,2]. To date, CIC has been implicated in a broad range of physiological processes, including lung alveolarization [3], bile acid homeostasis [4], and T-cell development [5,6]. Furthermore, CIC activity appears to be important in neurodevelopment, as its dysfunction has been linked to a spectrum of neuro-behavioural syndromes [7], neurodegeneration [8], and altered lineage specification of neural stem cells (NSCs) [9][10][11]. Among brain tumours, CIC is mutated almost exclusively and at high frequency (~50-80%) in oligodendroglioma (ODG) [12][13][14][15]. Defining molecular characteristics of ODG include IDH1/2 mutation and single copy deletion of chromosome arms 1p and 19q [16]. CIC resides within the portion of 19q that is lost. The observation that the remaining copy of CIC frequently harbours a somatic mutation that either truncates the protein or results in a loss of DNA binding activity supports the notion that CIC may have a tumour suppressor role in ODG. Moreover, the invariable co-occurrence of CIC and IDH1/2 mutations in ODG is compatible with the notion that a functional relationship exists between these two alterations in conferring a selective advantage to cells in ODG progression. However, both CIC's putative tumour suppressor role and its connection with mutant IDH proteins remain poorly understood.
Recent studies have identified several chromatin modifier proteins to be interactors with CIC, indicating a linkage between the epigenome and CIC's function [2,11]. Additionally, the IDH1/IDH2 mutations characteristic of ODG tumours where CIC mutations are found result in the overproduction of 2-hydroxyglutarate (2-HG), which has the downstream consequence of widespread hypermethylation of CpG sites and histone tail residues [17][18][19]. Considering the emerging link between CIC and chromatin modifiers, we posited that CIC and IDH mutations may functionally collaborate to dysregulate the transcriptome and/or the epigenome. We thus analysed genome-wide profiles of RNA expression, DNA methylation, and selected histone modifications in CIC-wild type (CIC-WT) and CIC-knockout (CIC-KO) cell lines, in the presence and absence of IDH1-R132H expression, seeking new insight into how CIC loss and mutant IDH protein expression might interact to promote ODG.

Ethics statement
The work presented here was approved by the UBC BC Cancer Research Ethics Board (H19-030103, H08-02838).

Whole transcriptome library construction and sequencing
Details are presented in Supplementary materials and methods.

Differential expression analysis
Raw read counts were mapped onto Ensembl 75 gene annotations using JAGuaR [21]. DESeq2 [22] v.1.8.2 was used to conduct independent differential expression analyses between each CIC-KO line and its CIC-WT counterpart, and between the CIC-WT (IDH1-R132H) and CIC-WT (IDH1-WT) lines. Differential expression analysis results are presented in supplementary material, Table S1, including whether each differentially expressed (DE) gene was also identified as DE in CIC-mutant ODGs compared with CIC-WT ODGs [20]. DE genes were considered statistically significant if they met a Benjamini-Hochberg-adjusted P value (q-value) of 0.05 and were considered CIC-associated if they met this threshold and had concordant directionality.

Functional enrichment analysis
CIC-associated DE genes within each IDH1 context were submitted separately for pathway enrichment analysis using Metascape [23]. For IDH1-associated DE genes (n = 6044), those with a fold-change ≥ 2 were submitted (n = 2722), since Metascape restricts gene lists for pathway enrichment analysis to 3000 genes.

CIC ChIP-seq analysis
Methods for CIC chromatin immunoprecipitation and sequencing, including processing of CIC ChIP-seq data and peak calling, are described in detail in Supplementary materials and methods. CIC peaks derived from our dataset were assessed for overlap (≥1 bp) with CIC peaks from a published CIC ChIP-seq dataset [2]. The 150 most significant peaks in our dataset were deemed high-confidence CIC peaks based on the inflection point at which the presence of reproducibly identified peaks increased in relation to peak rank by statistical significance (supplementary material, Figure S2).
Genomic features of high-confidence CIC peaks were obtained using ChIPseeker [24]. De novo motif enrichment analysis was conducted on high-confidence CIC peaks centred on their summits (the coordinate at which fold-enrichment of ChIP read coverage relative to its matched control was greatest) using HOMER [25] v.4.9.1, with the size parameter set to 200 bp as recommended for identifying primary and co-enriched motifs for transcription factors (TFs) (http://homer.ucsd.edu/ homer/ngs/peakMotifs.html).

Histone modification ChIP-seq analysis
Methods for histone modification ChIP, library construction and sequencing, and processing of histone modification ChIP-seq data, including the identification of peaks and enhancer regions, are described in Supplementary materials and methods.
CIC-associated and IDH1-associated differentially enriched (DER) peaks were identified using DESeq2, analogously to the identification of DE genes. DER peaks were required to meet a q-value threshold of 0.05 and a fold-change ≥ 2 to be considered significant, and additionally required directional concordance between both CIC-KO replicate cell lines for CICassociated DER peaks. DER peaks were annotated with their associated genomic feature and nearest gene using ChIPseeker.
Enhancers were labelled as DER if an overlap was present with at least one H3K4me1 or H3K27ac DER peak. The nearest genes associated with DER enhancers were considered to be putative targets of such enhancers. De novo motif analysis was performed on downregulated and upregulated DER enhancers using HOMER, with the size parameter set to 500 bp as recommended for histone marked regions (http://homer.ucsd.edu/homer/ngs/ peakMotifs.html) and otherwise default parameters.

Differential methylation analysis
Methods for whole genome bisulphite sequencing and data processing, including differential methylation analysis, are described in Supplementary materials and methods. Differentially methylated regions (DMRs) were identified using Defiant [26] with default parameters. DMRs between replicate CIC-KO cell lines were assessed for both overlap (≥1 bp) and concordant directionality, and were considered to be CIC-associated if they met these criteria.

Results
Transcriptome and epigenome profiles of CIC-KO cell lines expressing IDH1-WT and IDH1-R132H To investigate the effects of CIC loss and IDH1-R132H expression on transcriptomes and epigenomes, we The target site of the CRISPR-Cas9 sgRNA used to generate CIC-KO cell lines is located at the first commonly shared exon between the two CIC isoforms (CIC-L: long isoform; CIC-S: short isoform). (B) The experimental models used in this study: from immortalized human astrocyte cell lines expressing either IDH1-WT or IDH1-R132H, two independent CIC-KO cell lines were obtained using CRISPR-Cas9. (C) The -omic datasets and analytical workflow involved in characterizing the transcriptomic and epigenomic consequences of CIC loss and mutant IDH1 expression (also see Materials and methods section). From each differential analysis, CIC-associated and IDH1-associated alterations were identified by comparing CIC-KO cells with their CIC-WT parental counterparts, and CIC-WT (IDH1-R132H) cells with CIC-WT (IDH1-WT) cells, respectively. Since there were two CIC-KO cell lines in each IDH1 context, we considered CIC-associated alterations as those that consistently appeared in both CIC-KO cell lines (i.e. the intersection of the Venn diagrams) within each respective IDH1 background. DE, differentially expressed; DER, differentially enriched; DMR, differentially methylated region.
Multi-omic study of CIC-knockout and IDH1-mutant cells 299 performed whole-transcriptome sequencing (RNAseq), whole-genome bisulphite sequencing (WGBS), and chromatin immunoprecipitation followed by DNA sequencing (ChIP-seq) for six different histone modifications (H3K4me1, H3K4me3, H3K27ac, H3K27me3, H3K9me3, H3K36me3) to analyse CIC- In each IDH1 context, two independent CIC-KO lines were derived and are individually referred to as CIC-KO1 and CIC-KO2 throughout this study. For each -omic dataset, we identified alterations present in CIC-KO cells relative to CIC-WT cells (i.e. CIC-associated alterations) in both IDH1-WT and mutant contexts. In a similar manner, we identified IDH1-associated differences by compar- We also conducted CIC ChIP-seq to identify CIC binding sites in the IDH1-WT line. The experimental models, datasets, and analytical approaches used in this study are outlined in Figure 1.
CIC knockout and mutant IDH1 expression have overlapping consequences at the level of differentially expressed genes and pathways We conducted differential expression analyses to compare gene expression levels between each CIC-KO cell line and its parental CIC-WT counterpart (see Materials and methods). The number of up-and down-regulated protein-coding genes obtained from each CIC-KO versus CIC-WT comparison is presented in Figure 2A. . Other genes whose promoters were previously confirmed to be bound by CIC in HEK cells (ETV1, DUSP4, GPR3, SPRY4, SHC3, and SHC4 [20]) were also generally upregulated (supplementary material, Figure S3 and Table S1). We also found that 66 and 46 CIC-associated DE genes identified in IDH1-WT and IDH1-R132H cell lines, respectively, overlapped with genes previously found to be DE in CIC-WT versus CIC-deficient primary ODGs [20] (supplementary material, Table S1). This relatively low overlap is consistent with CIC-associated DE genes being largely context-specific, as previously demonstrated [30].
To glean insights into CIC-associated transcriptional alterations at the level of biological pathways, we conducted functional enrichment analyses of DE genes (see Materials and methods). Consistent with previous associations made between CIC and central nervous system (CNS) development [9][10][11], pathways related to neuron differentiation and synapse formation were among the most significantly enriched terms for CIC-associated DE genes in both IDH1 backgrounds ( Figure 2D and supplementary material, Table S2). These same terms were among the top enriched pathways for IDH1-associated DE genes, indicating an overlap between the consequences of CIC-KO and IDH1-R132H expression at the level of biological processes in addition to the overlap at the level of DE genes noted previously.
CIC ChIP-seq identifies neurodevelopmental genes as potentially novel direct target genes of CIC To help identify candidate direct CIC targets among the list of CIC-associated DE genes, we performed CIC ChIP-seq on the CIC-WT (IDH1-WT) cell line. We focused on the 150 most significant CIC peaks based on the overlap of these peaks with an independently generated CIC ChIPseq peak set [2] (supplementary material, Figure S2 and Table S3; Materials and methods). As expected, among these 150 peaks were those in close proximity to the transcriptional start sites (TSSs) of known CIC target genes, such as DUSP4, ETV4, ETV5, GPR3, SPRY4, and PLK3 ( Figure 3A and supplementary material, Table S3). Moreover, the most significantly enriched motif (p < 1e-49) within these 150 peaks contained the known CIC consensus binding site [2,31] ( Figure 3B). Thus, we refer to these 150 reproducibly identified peaks as high-confidence CIC peaks. Consistent with published results [2], over half of the high-confidence CIC peaks were located in introns and intergenic regions ( Figure 3C), indicating that CIC might have regulatory roles at regions other than TSSs.
In addition to known CIC target genes, high-confidence peaks were found at the promoters of FOS, FOSL1, MAFF, MAFG, EPHA2, ID1, and RUNX1 ( Figure 3A). Of these genes, EPHA2, MAFF, MAFG, and RUNX1 were significantly (q < 0.05) upregulated in at least one CIC-KO cell line compared with its CIC-WT counterpart ( Figure 3D). FOSL1 and ID1 also generally exhibited increased transcript levels in CIC-KO cells compared with CIC-WT cells, although the differences were not statistically significant ( Figure 3D). These observations support the notion that these genes may be previously unexplored direct targets of CIC-mediated transcriptional repression.

CIC-KO is associated with dysregulation of enhancers near neurodevelopmental genes
To assess the effects of CIC loss and IDH1-R132H expression on histone modification patterns, we used

SD Lee et al
ChIP-seq to profile H3K4me1, H3K4me3, H3K27ac, H3K27me3, H3K9me3, and H3K36me3 in our cell line models. Consistent with previous reports associating mutant IDH and increased histone methylation (e.g. ref 19), the cell lines expressing IDH1-R132H displayed greater genomic enrichment of H3K4me1, H3K4me3, Figure 2 Legend on next page.
Multi-omic study of CIC-knockout and IDH1-mutant cells 301 and H3K27me3 compared with those expressing IDH1-WT (supplementary material, Figure S4). To identify specific changes in chromatin state [i.e. differentially enriched (DER) peaks] associated with different CIC and IDH1 states, we compared read counts within peaks across mutant and wild-type states (supplementary material, Figure S5A). Similar to the RNA-seq analysis ( Figure 2C), a considerable proportion of CIC-associated DER peaks were also IDH1-associated, while the overlap between CIC-associated DER peaks in the two IDH1 backgrounds was comparatively small (supplementary material, Figure S5B). This indicates that the majority of changes in the chromatin landscape attributed to CIC loss may be additionally influenced by IDH1-R132H expression. CIC-associated DER H3K4me1, H3K4me3, H3K27me3, and H3K27ac peaks were predominantly located more than 10 kb away from a TSS and at introns and intergenic regions ( Figure 4A), indicating that CIC loss may primarily affect distal regulatory elements. A similar result was observed regarding IDH1-associated DER peaks ( Figure 4A). The observation that most DER peaks were not proximal to TSSs, together with the presence of candidate CIC target genes whose functions are linked to enhancer activity such as FOS, FOSL1, and ETV5 [32][33][34], led us to investigate chromatin state changes at enhancer regions. Approximately 36% and 23% of CIC-associated H3K4me1 DER peaks and 62% and 58% of CIC-associated H3K27ac DER peaks were found at enhancer regions in the IDH1-WT and IDH1-R132H contexts, respectively ( Figure 4B; Materials and methods). The H3K4me1 and H3K27ac coverage profiles at enhancers overlapping a DER H3K4me1 and/or H3K27ac peak (henceforth referred to as DER enhancers) confirmed the differences in mean coverage between CIC-WT cells and their CIC-KO counterparts, primarily around the centre of enhancer regions ( Figure 4C).
Motif enrichment analysis within CIC-associated DER enhancers revealed that motifs related to the AP-1 complex, which encompasses the CIC candidate targets FOS and FOSL1, were the most significantly enriched across all DER enhancers ( Figure 4D). Motifs that most closely resembled ETS-family transcription factors (TFs; ERG and ETS:E-box), to which ETV1/4/5 belong, also emerged. Furthermore, motifs matching those of additional candidate CIC targets, namely RUNX1, MAFB, and BACH1, were among the top five most significantly enriched ( Figure 4D). Together, these results illustrate that CIC loss may indirectly lead to enhancer disruption through the de-repression of ETV genes and other putative direct CIC targets.
Consistent with the association of H3K4me1 and H3K27ac with active enhancers, there was a clear positive correlation between both H3K4me1 and H3K27ac enrichment at enhancers and the expression of their closest genes ( Figure 5A). Putative gene targets of such enhancers included CDH8, TMEM108, PDGFRA, and NFIA in the IDH1-WT model. In the IDH1-R132H model, prominent DER enhancer-associated genes included NRG1, EPHA4, ETV1, and NFIA. Expression of PDGFRA, a glioblastoma (GBM)-associated gene [35], was observed to be lower in comparisons of both CIC-KO to CIC-WT cells and IDH1-R132H to IDH1-WT cells, indicating that both CIC loss and the expression of IDH1-R132H independently resulted in reduced expression of PDGFRA ( Figure 5B). Interestingly, an enhancer within PDGFRA displayed a lower enrichment of active marks (H3K4me1 and H3K27ac) in the same cell lines that exhibited lower PDGFRA expression, suggesting that the loss in PDGFRA expression was due to the inactivation of this enhancer ( Figure 5C). An intragenic enhancer within NFIA, a regulator of both gliogenesis and gliomagenesis [36,37], exhibited a reduction of H3K27ac in CIC-KO cells compared with their CIC-WT counterparts and in CIC-WT (IDH1-R132H) compared with CIC-WT (IDH1-WT) cells, again illustrating the independent impacts of CIC-KO and IDH1-R132H expression in reducing the enrichment of active marks at the same enhancer. Intriguingly, however, CIC-KO cells that also expressed IDH1-R132H displayed a gain of H3K27ac, which we interpret as a potential reactivation of this enhancer ( Figure 5C). This intriguing pattern of enhancer dysregulation was also accompanied by concurrent changes in NFIA gene expression ( Figure 5B). In summary, these examples highlight the impacts of CIC-KO and IDH1-R132H on enhancers at genes associated with CNS tumours and, regarding NFIA, illustrate an interesting case of an enhancer whose activity is differentially regulated by CIC loss in an IDH1-dependent fashion.

Analysis of differentially methylated regions identifies CIC and IDH1-associated changes in the DNA methylation landscape
To investigate the effects of CIC loss or IDH1-R132H on DNA methylation, we generated WGBS libraries from

302
SD Lee et al Multi-omic study of CIC-knockout and IDH1-mutant cells 303 our cell line models. In agreement with IDH1-R132H expression being linked to global hypermethylation [19], the majority of the top 10 000 most variably methylated CpG sites exhibited hypermethylation in cells expressing IDH1-R132H ( Figure 6A). Moreover, mean genome-wide CpG methylation, as well as methylation within CpG  Multi-omic study of CIC-knockout and IDH1-mutant cells 305 islands and CpG shores, was significantly higher (p < 0.0005) in all cell lines expressing IDH1-R132H compared with those expressing IDH1-WT ( Figure 6B). To identify regions of DNA methylation that were affected by CIC loss or IDH1-R132H, we conducted a differentially methylated region (DMR) analysis using Defiant [26] ( Unsurprisingly, many more IDH1-associated DMRs were identified, totalling 83 572. Strikingly, CIC-associated DMRs almost exclusively involved increased DNA methylation, as was expected and observed with IDH1-associated DMRs ( Figure 6C). Considering CIC's established role as a transcriptional repressor, the prominence of hypermethylated DMRs relative to hypomethylated DMRs in CIC-KO cells was unexpected. However, we found no association between CIC binding sites and these DMRs (supplementary material, Figure S6), and also found no correlation between DMRs and the expression of their closest genes (supplementary material, Figure S7A,B). These observations illustrate that CIC-associated DMRs likely arose independently of CIC binding and appeared to have minimal impact on gene expression.

Discussion
Recent studies have indicated a regulatory role for CIC in neurodevelopment, in which CIC loss resulted in neural maturation defects [11], the promotion of EGFindependent NSC proliferation [10], and the expansion of NSCs and oligodendrocyte precursor cells (OPCs) [9]. These studies support the notion that functional CIC may be important in the maintenance of NSC quiescence, and that CIC loss can promote increased proliferation and partial commitment towards the oligodendrocyte lineage. Moreover, recurrent CIC mutations are found in gliomas (exclusively in those possessing an IDH1/2 mutation), implicating CIC as a potential

306
SD Lee et al tumour suppressor in this cancer type. In this study, we characterized the effects of CIC-KO on global gene expression, histone modification profiles, and DNA methylation patterns in IDH1-WT and IDH1-R132H backgrounds, including analyses of reproducible CIC binding sites. In IDH1-WT cells, we identified novel or previously unexplored candidate CIC target genes, including RUNX1, ID1, and EPHA2. The finding that CIC may directly regulate RUNX1, a gene linked to leukaemogenesis [38], may provide mechanistic insight into the reported associations between CIC loss and altered Tcell development and T-cell acute lymphoblastic leukaemia (T-ALL) onset in mice [5,6,29]. RUNX1 also appears to have a pro-neurogenic role, since its expression was found to correlate with the survival and proliferation of adult neural precursor cells [39]. ID1 has demonstrated roles in GBM tumour progression [40] and the control of NSC quiescence during regenerative neurogenesis [41]. EPHA2 overexpression was observed to promote glioma stem cell (GSC) invasiveness in vivo and promote neurosphere formation in vitro [42]. Notably, while ID1 was upregulated in all CIC-KO lines, RUNX1 and EPHA2 were upregulated specifically in CIC-KO (IDH1-R132H) cells ( Figure 3D), suggesting some interplay between CIC loss and mutant IDH1 to promote the activation of RUNX1 and EPHA2. Thus, our identification of RUNX1, ID1, and EPHA2 as candidate CIC targets, and their increased expression in CIC-KO cell lines, reveals potential mechanistic insights underpinning the link between CIC and the modulation of neural stem cell fate and ODG. Notably, our CIC ChIP-seq experiment was only performed in CIC-WT (IDH1-WT) cells. It is possible that CIC binding may be influenced by the epigenomic consequences of neomorphic IDH1/2. Nevertheless, the increased expression of RUNX1 and EPHA2 observed in CIC-KO (IDH1-R132H) indicates that these genes are potentially direct or indirect targets of CIC in an IDH-mutant context.
Based on our data, the consequences of CIC loss on histone modifications appeared to largely impact enhancers whose differential histone modification profiles may be ascribed to the differential expression of direct CIC target genes. Supporting this notion, we observed a significant enrichment of motifs matching those of known and candidate CIC targets such as ETV1/4/5, RUNX1, MAFF/G, and FOS/FOSL1 at DER enhancers ( Figure 4D). DER enhancer-associated genes, including PDGFRA [34] and NFIA [35,36], have been linked to the tumourigenicity of glioma models and/or neural progenitor cell fate decisions. Ablation of PDGFRA was shown to lead to precocious differentiation of OPCs in the developing spinal cord [43]. The relationship between CIC and PDGFRA dysregulation may therefore be of relevance in the context of CIC's function in neural cell fate specification and warrants further investigation. Interestingly, our results were in contrast to published observations of increased PDGFRA expression in IDH-mutant gliomas and glioma cell lines [44].
The pattern of dysregulation of a genic enhancer in NFIA and its expression showcased a striking phenomenon in which the expression of IDH1-R132H appeared to have reversed the effect of CIC loss. Furthermore, NFIA was found to be upregulated in CIC-mutant cells compared with CIC-wild type cells in a single-cell gene expression analysis of a primary ODG [45], indicating that the increased NFIA expression that we observed in CIC-KO (IDH1-R132H) cells may be relevant in a primary tumour context. Together with NFIA's demonstrated role in gliomagenesis, our finding supports the view that NFIA dysregulation may partially underlie the synergistic relationship between CIC loss and mutant IDH1 in ODG pathology.
The NHA cell line model used in this study presents some caveats. Firstly, it exhibits impaired p53 and RB function as a product of its immortalization [46]. Notably, TP53 mutations are not found in ODGs but rather in IDH-mutant astrocytomas and are mutually exclusive with CIC mutations. It would thus also be of importance to study the consequences of CIC loss in a p53-proficient background, in regard to CIC's role both in neurodevelopment and in ODG biology. Secondly, the IDH1-R132H cells overexpress an IDH1-R132H construct, which has been shown to promote some distinct metabolic features in short-term culture compared with cells that endogenously express mutant IDH1 [47]. The fact that most of our comparisons were made within IDH1-WT or IDH1-R132H cells (comparing CIC-KO with CIC-WT), along with comparisons to CICassociated changes in primary ODGs, nevertheless helps to distinguish possible effects of artefacts associated with IDH1-R132H overexpression. Thirdly, the NHA model is astrocytic in origin; however, while ODGs have traditionally been hypothesized to arise from an oligodendrocytic origin due to their histology, recent singlecell studies have shown that ODGs and astrocytomas share a common cellular hierarchy [45]. Therefore, our results are relevant to CIC's role in ODG as well as to its role in neurodevelopment.
Neomorphic IDH1/2 mutations and 1p/19q codeletions are the defining features of ODG, while CIC alterations are found in~50-80% of these primary tumours [12,13]. These mutational frequencies imply an order of events in which CIC mutations occur after the IDH1/2 mutation and 1p/19q co-deletion. The neomorphic IDH1 mutation and consequent DNA hypermethylation have been demonstrated to affect the neural developmental hierarchy, specifically in blocking differentiation [48,49]. Integrating our results and the emerging evidence for CIC being an important mediator of neural/glial cell fate specification, we conceptualize a model in which CIC loss is presumed to lead to the genesis of OPC-like cells and their expansion is further enabled due to the de-differentiating influence of neomorphic mutant IDH1 expression. This amplification of self-renewing cells could provide more opportunities for cancer-promoting mutations to arise. Overall, our work provides a rationale for future research to examine the functional relationship between CIC loss and

Data availability statement
The data that support the findings of this study are being made openly available in the Gene Expression Omnibus under the project title Transcriptomic and Epigenomic Profiles of CIC-knockout and IDH1mutant cells at https://www.ncbi.nlm.nih.gov/geo/query/ acc.cgi?acc=GSE189861 (GEO accession GSE189861). Codes used to generate the analyses presented in the paper are publicly available in Github (https://github.com/sdlee94/ multi-omic-analysis-of-CIC-KO-and-IDH1-R132H-cells).

SUPPLEMENTARY MATERIAL ONLINE
Supplementary materials and methods Figure S1. Confirmation of CIC and IDH1 status in cell line models Figure S2. Number of reproducibly identified CIC peaks versus MACS2 q-value significance Figure S3. Known CIC target genes are overexpressed in CIC-KO cells Figure S4. Comparison of peaks across all cell lines for each histone modification Figure S5. Summary of DER peaks Figure S6. CIC binding is not associated with differential methylation Figure S7. CIC-associated differential methylation is not associated with differential gene expression Table S1. DESeq2 differential expression analysis results for CIC-and IDH1-associated DE genes Table S2. Metascape pathway enrichment analysis results for CIC-and IDH1-associated DE genes Table S3. ChIPseeker annotations of high-confidence CIC peaks Multi-omic study of CIC-knockout and IDH1-mutant cells 309