Multi-omics and pharmacological characterization of patient-derived glioma cell lines

Abstract

Glioblastoma (GBM) is the most common brain tumor and remains incurable. Primary GBM cultures are widely used tools for drug screening, but there is a lack of genomic and pharmacological characterization for these primary GBM cultures. Here, we collect 50 patient-derived glioma cell (PDGC) lines and characterize them by whole genome sequencing, RNA sequencing, and drug response screening. We identify three molecular subtypes among PDGCs: mesenchymal (MES), proneural (PN), and oxidative phosphorylation (OXPHOS). Drug response profiling reveals that PN subtype PDGCs are sensitive to tyrosine kinase inhibitors, whereas OXPHOS subtype PDGCs are sensitive to histone deacetylase inhibitors, oxidative phosphorylation inhibitors, and HMG-CoA reductase inhibitors. PN and OXPHOS subtype PDGCs stably form tumors in vivo upon intracranial transplantation into immunodeficient mice, whereas most MES subtype PDGCs fail to form tumors in vivo. In addition, PDGCs cultured by serum-free medium, especially long-passage PDGCs, carry MYC/MYCN amplification, which is rare in GBM patients. Our study provides a valuable resource for understanding primary glioma cell cultures and clinical translation and highlights the problems of serum-free PDGC culture systems that cannot be ignored.

Fig. 1. Definition of transcriptional subtypes for PDGCs.

a Schematic diagram showing the generation of PDGCs. GBM tissues from patients were dissected and digested into single-cell suspension, which was then cultured in a serum-free medium to obtain PDGCs. A total of 50 PDGCs were collected or generated in this study, of which 50 PDGCs underwent RNA-seq, 10 PDGCs underwent WES, and 49 PDGCs underwent WGS. b Oncoprint showing the genomic alterations of 50 PDGCs. HOMDEL: homozygous deletion, AMP: high-level amplification, INFRAME: in-frame insertion and deletion, MISSENSE: missense mutation, PROMOTER: mutation localized in promoters, TRUNC: truncation mutation, GAIN: low-level copy number gain, HETLOSS: heterozygous deletion. c Heatmap showing the expression levels of subtype-specific signatures. Gene expression values were normalized by z-score. The rows represent genes, and the columns represent PDGCs. d–f Barplots showing the enriched pathways by MES (d), PN (e), and OXPHOS (f) subtype. Bars were colored according to subtype identity. g, h Sankey plots showing subtype assignment (n = 45) change flow. The left columns represent subtypes defined by us and the right columns represent subtypes defined by Wang et al. (g) or by Neftel et al. (h). Source data are provided as a Source Data file.

Fig. 2. Subtype of cultured PDGCs and their original tissues.

a, b Sankey plots showing the change in subtype assignment between cultured PDGCs and their parental tissues using in-house data (n = 12) (a) and previously published data GSE119834 (n = 10) (b). c Uniform Manifold Approximation and Projection (UMAP) visualization of GBM tumor cells from tissue and cultured PDGCs, colored by cell type. d UMAP visualization of GBM tumor cells from tissues and cultured PDGCs, colored by subtype assignment. Tables next to the UMAP visualization summarizing the number of cells of each subtype. e The percentage of each subtype in GBM tumor cells from tissues and cultured PDGCs. Source data are provided as a Source Data file.

Fig. 3. Clinical, genomic, and tumorigenic comparisons of defined subtypes.

a Kaplan–Meier survival curves for GBM patients from the TCGA cohort with different subtypes. The P-value was calculated using the log-rank test. b Response to radiotherapy for defined subtypes. The y-axis represents the percentage of surviving cells upon radiation. P-values were calculated using the two-sided Wilcoxon rank-sum exact test. Boxplots show the median (center line), the upper and lower quantiles (box), and the range of the data (whiskers). c Frequency of genomic alterations of the indicated genes in PDGCs of different subtypes. Bars were colored by alteration types. AMP: high-level amplification, GAIN: low-level gain, HETLOSS: heterozygous deletion, HOMDEL: homozygous deletion, MISSENSE: missense mutation, TRUNC: truncation mutation. d Stacked barplots showing the tumorigenicity of PDGCs of different subtypes. For each cell line, 5 × 10⁵ cells were injected into the brains of nude mice. e 3D scatterplot showing the subtype score of PDGCs and their transplanted tumors’ cell culture. f Scatter plots showing the median survival time of nude mice after injection of PDGCs of different subtypes. MES (n = 2), PN (n = 6), OXPHOS (n = 8), other (n = 2). The lines were plotted as mean ± SEM. Source data are provided as a Source Data file.

Fig. 4. Drug responses of PDGCs with different subtypes.

a Schematic diagram showing the drug screening process. b Heatmap showing the drug response of PDGCs with different subtypes. The rows represent drugs, and the columns represent cell lines, colored according to the cell viability upon the treatment of corresponding drugs. The boxplots on the left side show the response of cell lines to specific drug treatments. c Drug (n = 118) response of PDGCs with different subtypes. P-values were calculated by a two-sided Wilcoxon rank-sum exact test. d, e Normalized cell viability of PDGCs with different subtypes upon treatment with the indicated drugs. Representative drugs that effectively inhibited the growth of PN and OXPHOS subtype PDGCs are listed in (d) and (e), respectively. MES (n = 26), PN (n = 30), OXPHOS (n = 26). P-values were calculated by a two-sided Wilcoxon rank-sum exact test. In b–e, boxplots show the median (center line), the upper and lower quantiles (box), and the range of the data (whiskers). Source data are provided as a Source Data file.

Fig. 5. Lovastatin inhibits the growth of OXPHOS subtype PDGCs by attenuating mitochondrial respiration.

a Cell viability of lovastatin-treated BNI423 supplemented with different concentrations of cholesterol (biological replicates n = 3). EtOH: ethyl alcohol. Data are presented as mean ± SD. b Cell viability of lovastatin-treated BNI423 supplemented with different concentrations of lanosterol (biological replicates n = 3). Data are presented as mean ± SD. c GSEA plot of OXPHOS subtype PDGCs treated with lovastatin compared to PDGCs without lovastatin treatment. GSEA is performed by R package clusterProfiler. P-value was calculated by the two-sided Kolmogorov–Smirnov test and adjusted by FDR. d Representative MitoTracker staining of BNI423 under different treatments for 16 h. The mitochondria were stained with MitoTracker (red), and the nucleus was stained with DAPI (blue). Scale bar, 10 μm. e Relative cellular ADP/ATP ratio of BNI423 under different treatments (biological replicates n = 3). The mitochondrial complex I inhibitor rotenone was used as a positive control. P-values were calculated by the two-sided Student’s t-test. Lov.: Lovastatin. Data are presented as mean ± SD. f Relative cellular ADP/ATP ratio of lovastatin-treated BNI423 with or without cholesterol supplementation (biological replicates n = 3). P-values were calculated using the two-sided Student’s t-test. Lov.: Lovastatin, Cho.: Cholesterol. Data are presented as mean ± SD. g Tumor size of mice with subcutaneously inoculated BNI17 (n = 6) treated with lovastatin or vehicle. P-values were calculated using the two-sided Student’s t-test. Data are presented as mean ± SEM. h, i Quantification of Ki67- (h) and cleaved Caspase-3-stained (i) brain tumor tissues. BNI423 were transplanted into the brains of nude mice. After five weeks, lovastatin was administered by brain infusion (42 μg per day per mouse) for three days, and the whole brains were harvested for histopathologic analysis. P-values were calculated using the two-sided Student’s t-test. Data are presented as mean ± SD. In (h), Vehicle (n = 9), Lovastatin (n = 9). In (I), Vehicle (n = 6), Lovastatin (n = 9). Source data are provided as a Source Data file.

Fig. 6. MYC/MYCN amplification in cultured PDGCs.

a MYC pathway expression score in PDGCs and matched GBM tissues. In-house data (n = 12), GSE119834 (n = 10). The P-value was calculated by a two-sided paired Wilcoxon rank-sum exact test. b Percentage of samples with MYC/MYCN amplification in long- (n = 25) and short-passage (n = 25) PDGCs. Cell lines cultured more than 10 passages were categorized as long-passage, otherwise, the cell lines were categorized as short-passage. The P-value was calculated by one-sided Fisher’s exact test, p = 0.036. c Percentage of samples with MYC/MYCN DNA amplification or with MYC/MYCN ecDNA amplification. The red bar indicates long-passage PDGCs, while the blue bar indicates short-passage PDGCs. Chr.: chromosomal DNA, ecDNA: extrachromosomal DNA. d The copy number of MYC. The blue bars represent the chromosomal copy number detected by WGS, the purple bars represent the ecDNA copy number detected by WGS, and the red bars represent the total copy number detected by ddPCR. e Representative FISH images showing the MYC/MYCN amplification in primary GBM tissues from patients CB1838, CB5304, and CB3145. The yellow arrow indicates cells with MYC/MYCN amplification. Scale bar, 20 μm. A total of 13 tissues were subject to FISH, of which 9 samples without MYC/MYCN amplification, 3 samples with MYC amplification, and 1 sample with MYCN amplification. f Heatmap showing the CNV of the indicated genes, colored by CNV status (red for amplification and blue for deletion). The rows represent cells isolated from GBM patients and the columns represent the genomic locations. CNV levels were inferred by R package infercnv. g Representative FISH images showing rare MYC-amplifying tumor cells in grades II, III, and IV glioma tissues. The yellow arrow indicates cells with MYC amplification. Scale bar, 20 μm. A total of 96 glioma tissues were tested, with grade II (n = 33), III (n = 24), and IV (n = 39). h Percentage of samples harboring more than three MYC amplification cells in grade II (n = 33), III (n = 24), and IV (n = 39) glioma tissues. Source data are provided as a Source Data file.