DDDR-47. Development of in vivo CRISPR interference screens in diffuse midline glioma xenografts to identify genetic vulnerabilities controlling responses to radiation therapy

Diffuse midline gliomas (DMGs) have uniformly poor survival outcomes despite initial responses to radiation therapy (RT). Functional genomic and high throughput drug screens have identified promising therapeutic targets for DMGs, but these efforts have largely been performed in cell cultures, thereby missing targets that are selectively vulnerable within in vivo tumor environments. Here, we develop a genome-wide CRISPR interference (CRISPRi) platform to screen essential and radiation sensitizing genetic targets in DMG xenografts. We demonstrated the feasibility of genome-wide in vivo screens by introducing a lentiviral library of 22,580 inert barcodes prior to orthotopic transplantation of SJ-DIPG7 cells. Individual tumors yielded only 55-77% of this library after 3-4 weeks of intracranial growth. However, in silico pooling of two or more tumors enabled >88% genome-wide coverage. We then performed in vivo CRISPRi screens using a dual-sgRNA genome-wide library comprising 21,554 gene-targeting vectors and 1,026 non-targeting controls. SJ-DIPG7 stably expressing dCas9-Zim3 were infected with CRISPRi libraries and intracranially transplanted to immunocompromised mice. Animals were treated with control or RT (2 Gy x 5 fractions), a dose that extended survival from SJ-DIPG7 xenografts but not sufficient for cure. After in silico pooling of 2-3 mice, 93% of barcoded sgRNAs were recovered and we identified 340 genes that diminished in vivo growth upon knockdown (mean log2 fold-change -5.58;FDR 2%). These genes included 145 known essential genes from DepMap, supporting the validity of our method, and were enriched for proteasome components. We also identified 108 radiation sensitizing targets (mean log fold-change -6.02;FDR 7%) including DNA repair (TOPBP1, PPP4C), metabolism (CNDP1, LDHA, OGDH), and neuronal function (LSAMP, EFNA1). We validated the essential phenotype of the proteasome component PSMA3. Our work establishes the feasibility of genome-wide in vivo functional genomics for systematically identifying therapeutic vulnerabilities of DMGs, enabling the discovery of combination treatments for patients living with DMG.