Targeting the non-coding genome and temozolomide signature enables CRISPR-mediated glioma oncolysis

Abstract

Glioblastoma (GBM) is the most common lethal primary brain cancer in adults. Despite treatment regimens including surgical resection, radiotherapy, and temozolomide (TMZ) chemotherapy, growth of residual tumor leads to therapy resistance and death. At recurrence, a quarter to a third of all gliomas have hypermutated genomes, with mutational burdens orders of magnitude greater than in normal tissue. Here, we quantified the mutational landscape progression in a patient's primary and recurrent GBM, and we uncovered Cas9-targetable repeat elements. We show that CRISPR-mediated targeting of highly repetitive loci enables rapid elimination of GBM cells, an approach we term "genome shredding." Importantly, in the patient's recurrent GBM, we identified unique repeat sequences with TMZ mutational signature and demonstrated that their CRISPR targeting enables cancer-specific cell ablation. "Cancer shredding" leverages the non-coding genome and therapy-induced mutational signatures for targeted GBM cell depletion and provides an innovative paradigm to develop treatments for hypermutated glioma.

Keywords: CP: Cancer; CRISPR-Cas9; cancer shredding; genome shredding; glioblastoma; hypermutated glioma.

Figure 1.. Glioma mutational landscape reveals non-coding genome as key source of variation

(A) Schematic of GBM progression with frontline therapy including radiation, TMZ chemotherapy, and surgical resection. Highlighted is CRISPR-Cas “genome/cancer shredding,” a strategy for recurrent tumors that induces DNA damage and cell elimination through targeting of repetitive sequences in the genome. MMR, mismatch repair. MGMT-p, O-6-methylguanine-DNA methyltransferase promoter. sgCIDE, sgRNA targeting repeat loci. (B) Quantification of protein-coding and non-coding variants (mutations) in a patient’s primary and recurrent GBM compared to the patient’s native genome. Variants were detected using Mutect2 based on whole-genome sequencing. Blood was used for the patient’s native control. (C) Venn diagram showing the patient’s variants that are reported in ClinVar (NCBI) as risk factor, likely pathogenic, or pathogenic. Each section of the Venn diagram reports the number of coding and non-coding mutations (coding/non-coding). (D) Venn diagram of the patient’s collection of all possible sgRNAs, termed “sgRNA-ome.” The right panel quantifies the number of sgRNAs with the indicated amount of repeat targets (20-mer sequences, 5′-NGG-3′ PAM). (E) Violin plot of the repeat sgRNA-ome target counts in the human genome (hg38). Only sgRNAs with at least 10 target sites are represented. (F) Target distribution of highly repetitive sgRNAs, termed “sgCIDE,” between coding and non-coding genome, showing repeat sgRNA mainly target the non-coding genome. The inlay shows a zoom-in, highlighting that there are two discrete key classes of sgCIDEs. (G) Circos plots of select sgCIDEs indicating the location of target sites across the genome (Hamming distance = 0). Each line of the inner circle represents one target locus. The outer circle highlights the chromosomes.

Figure 2.. CRISPR-Cas genome shredding enables fast and efficient cell elimination

(A) Diagram of a competitive proliferation assay to quantify cell depletion rates using flow cytometry. (B) Shown are the sgCIDE sequence and count of fully complementary (Hamming distance = 0) loci in the human genome. (C) Competitive proliferation assay in Cas9-expressing U251 and LN229 GBM cells. Wild-type cells not expressing Cas9 were used for normalization. Cell lines were stably transduced with the indicated guide RNAs inducing genome shredding (sgCIDE), targeting an essential gene (sgRPA1), or representing a non-targeting control (sgNT). Changes in ratios of sgRNA-transduced cells (mNeonGreen+) were monitored by flow cytometry over 7 days. Error bars indicate SD (n = 3). (D) Comet assay to assess genomic degradation at single-cell resolution. (E) Comet assay demonstrating rapid and robust genome shredding at the single-cell level. Cas9-expressing U251 and LN229 cells were transduced with the indicated sgRNAs, and genome fragmentation was assessed at 24 h post transduction. Select representative images are shown. (F) Quantification of genome fragmentation observed in comet assay in (E). Error bars indicate SD. Significance was determined using an unpaired, two-tailed t test. (G) Genome shredding abolishes target cell proliferation. Growth curves showing the number of mNeonGreen+ cells detected by live-cell imaging of Cas9-expressing U251 transduced with the indicated sgRNAs. Quantification indicates the average count per condition across n = 4 images.

Figure 3.. Genome shredding is cell-state agnostic and hard to escape

(A) Quantification of cell viability (CellTiter-Glo) after treatment of TMZ-sensitive and -resistant glioblastoma cells with TMZ (concentration gradient) or transduction with sgNTs and sgCIDEs (viral dilution gradient, 1:X). (B) Schematic of cell-cycle analysis using flow cytometry. PI, propidium iodide. (C) Cell-cycle analysis of TMZ- and sgCIDE-treated GBM cells. Cas9-expressing U251 cells were treated with TMZ (50 μM) or the indicated sgRNAs, and cell-cycle profiles analyzed after PI staining. Shown are select representative flow cytometry plots. (D) Quantification of Sub-G1 fraction in cell-cycle profiles shown partially in (C). Cas9-expressing U251 and T98G cells were treated with DMSO, TMZ (50 μM), or the indicated sgRNAs and stained with PI after 1–5 days. Error bars indicate SD (n = 3). (E) Quantification of cell viability 5 days after treating glioblastoma stem-like cells (GSCs) with TMZ (50 μM) or all-in-one lentiviral vectors (pCF827) to deliver Cas9 and sgRNAs (sgNT-1/2/3, sgCIDE-1/2/3). Error bars indicate SD (n = 4).

Figure 4.. Rare escapee clones can efficiently be retreated

(A) Schematic showing retreatment of GBM cells that have escaped a first cycle of genome shredding. During initial treatment (colony formation assay), GBM cells are transduced with a lentiviral vector expressing Cas9 and a puromycin selection marker (pCF226), followed by a vector expressing an sgRNA and mNeonGreen (pCF821). Rare “escapee” cells that survive this regimen (Puro^R, mNeonGreen+) are expanded and treated a second time. Retreatment consists either of a vector expressing only an sgRNA and mCherry (pCF820) or an all-in-one vector expressing an sgRNA and Cas9 linked to an mCherry marker (pCF826). (B) Colony formation assay in Cas9-expressing U251 and T98G. (C) Quantification of cell depletion of the indicated U251-Cas9 escapee clones and controls treated with a vector (pCF826) expressing the indicated sgCIDE or sgNT-1 and Cas9 linked to mCherry. sgC1, sgCIDE-1. sgC2, sgCIDE-2. Error bars indicate SD (n = 3). Note, for three out of four escapee clones, re-expression of Cas9 alone was sufficient for cell depletion, indicating that the previously introduced sgCIDEs were still active in those clones.

Figure 5.. Genome shredding efficiently ablates GBM cells in vivo

(A) Representation of experimental design. (B) Sub-G1 assay confirming that concentrated sgCIDE-1 virus efficiently transduces and destroys Cas9-expressing LN229 cells in vitro. Error bars indicate SD (n = 3). (C) In vivo treatment of mice harboring Cas9-expressing LN229 GBM intracranial xenografts with sgRNA virus. Mouse brains were analyzed when tumor-related symptoms were observed. The percentage of mCherry+ (tumor cells) and mNeonGreen+ (sgRNA virus) cells from mice treated with 5 μL concentrated sgNT-1 or sgCIDE-1 virus (four mice each) and 15 μL concentrated sgNT-1 or sgCIDE-1 virus (six and five mice, respectively) were quantified by flow cytometry. Error bars indicate SD. Significance was determined using the Mann-Whitney test (*p < 0.05; **p < 0.01). (D) Kaplan-Meier curves of mice harboring Cas9-expressing LN229 GBM intracranial xenografts treated with 5 μL or 15 μL concentrated sgNT-1 or sgCIDE-1 virus. Significance was determined using the log rank test. (E) Flow cytometry analysis of wild-type (WT) Cas9 and TiCas9-edited HEK293T cells encoding GFP (HEK-RT1), demonstrating efficient inducible editing with TiCas9, though not fully matching WT Cas9. Error bars indicate SD (n = 3). (F) Competitive proliferation assay in TiCas9-expressing U251 GBM cells. Untreated cells (pre-treatment) were used for normalization. Error bars indicate SD (n = 3). (G) Kaplan-Meier curves of mice harboring TiCas9-sgRNA-expressing U251 GBM intracranial xenografts. Tumor-bearing mice were injected with tamoxifen for 5 consecutive days to induce TiCas9 expression 1 week after tumor implantation. Significance was determined using the log rank test.

Figure 6.. Minimal requirements for efficient cell ablation

(A) Competitive proliferation assays to evaluate CRISPR target requirements for efficient cell ablation by genome shredding in various eukaryotic cells. (B) Target site occurrences for the indicated sgRNAs in the human (hg38), mouse (mm10), and chicken (galGal6) genomes. Quantified are the number of targets with full complementarity (HD 0) and one mismatch (HD 1). HD, Hamming distance (mismatches). (C) Competitive proliferation assays in Cas9-expressing human (U251) and mouse (GL261) GBM cells and chicken fibroblasts (DF1). Wild-type cells not expressing Cas9 were used for normalization. Shown is the relative ratio of sgRNA-expressing (mNeonGreen+) cells at day 7 post transduction. Error bars indicate SD (n = 3).

Figure 7.. Targeting TMZ signature mutations enables selective elimination of recurrent GBM cells

(A) Diagram showing identification of cancer-specific sgRNAs and assessment of cancer shredding in a patient-derived cell line (PDCL) of the recurrent GBM. (B) Mutational signature of the patient’s GBMs. Indicated is the frequency of variants in the primary and recurrent GBM after normalization to the patient’s native genome. Note, the recurrent tumor shows a C>T mutational signature (TMZ signature) characteristic of hypermutated GBM after TMZ treatment. (C) Venn diagram showing the sgRNA-ome of the patient’s recurrent GBM, determined in both the patient’s recurrent GBM tissue and in two technical replicates of the recurrent GBM PDCL. Numbers indicate cancer-specific sgRNAs, defined as those with at least one mismatch in the PAM-proximal eight nucleotides compared to the patient’s native and reference genomes. (D) Identification of mismatch location of cancer repeat sgRNAs when compared to the patient’s native and reference genomes. (E) Enrichment and depletion of sgRNAs from a CRISPR screen in NHA control and SF11411 PDCL cells. The sgRNAs were ranked based on the log(2) fold change (FC) of their representation from day 28 (D28) post transduction compared to day 1 (T0). Specific subsets of sgRNAs are highlighted. (F) Quantification of cancer-specific sgRNA enrichment and depletion in NHA control versus SF11411 PDCL cells using MAGeCK RRA. Shown is the log(2)FC of normalized sgRNA representation between the two cell lines at the end of the screen (D28). (G) Significantly depleted (FDR < 0.05, MAGeCK RRA gene level) sgRNAs enabling TMZ signature targeted cancer shredding. Note, cancer repeat sgRNAs showed the strongest cancer-specific depletion. (H) Potential use cases of CRISPR cancer shredding. Targetable mutations include therapy-induced TMZ signature mutations. sgHYP, hypermutated cancer-specific sgRNAs. sgTMZ, TMZ signature mutations-specific sgRNAs. sgREC, recurrent cancer-specific sgRNAs.