In Vivo Crispr Screening Identifies Novel Vulnerabilities in ETO2/GLIS2-rearranged Leukemia

Background: CRISPR screening has allowed for massively parallel interrogation of the genome for vulnerabilities in a variety of cancer models. However, these experiments are largely performed in vitro due to the scale and conditions required to effectively conduct these screens. Furthermore, there still remain genetic subtypes of acute leukemia which are poorly represented among existing cell lines. We sought to determine the feasibility of conducting screens with focused libraries in patient-derived leukemia xenografts.

Methods: We utilized a patient-derived acute megakaryoblastic leukemia (AMKL) bearing an ETO2/GLIS2 fusion, representing an aggressive subtype of pediatric acute leukemia with limited comparable cell lines (Gruber TA, et al. Cancer Cell 2012.) Based on its transplantability in NOD.scid.Il2Rγc^null-SGM3 (NSGS) mice vs culture in vitro, we performed all subsequent experiments in vivo. We first sought to establish stable Cas9 expression, which we successfully achieved by lentiviral transduction followed by FACS-enriched transplant, followed by a second round of FACS enrichment with limiting dilution transplant. As judged by mCherry flow cytometry, we achieved highly pure (>90%) Cas9-expressing populations, which we verified by western blotting. We further verified functionality of these cells by verifying cutting at the safe harbor locus AAVS1.

In order to identify molecular vulnerabilities in this system, we identified the top 100 differentially expressed genes (DEG) between ETO2/GLIS2-rearranged leukemias versus other leukemia PDXs which we previously characterized. We included eight transcription factors with motifs enriched in our DEG analysis, as well as eight pan-essential positive controls (such as MYC, BRD4). Four guide RNAs (gRNA) per gene were selected from the Broad Brunello genome-wide library (Doench JG, et al. Nat Biotechnol 2016) as well as 50 "safe targeting" control guides (Morgens DW, et al. Nat Commun 2017) directed against non-essential genomic regions for a total of 518 guides. Guides were ordered as pooled oligonucleotides, cloned into a lentiviral gRNA backbone using standard techniques, then used to prepare lentivirus via transient transfection of 293T cells followed by titration to achieve an MOI of 0.3. With this MOI and an approximate engraftment frequency of 2% following intrafemoral injection, we transplanted 8.7e7 cells across eight NSGS mice to achieve adequate starting representation of our pooled library. Mice were sacrificed at humane endpoints, and spleens and bone marrow underwent PCR amplification for gRNA sequences followed by Illumina deep sequencing, and analysis with MAGeCK-VISPR (Wei L, et al. Genome Biol 2015).

Results: One mouse per cohort was sacrificed at 24 hours post-transplant and sequenced for gRNA representation, which confirmed that library representation was largely preserved with our approach. At the time of sacrifice, spleen and bone marrow samples were independently sequenced, but no major bias in guide dropouts was observed between these compartments. Subsequent analyses were performed on the marrow alone. We observed the dropout of 38 genes with a false discovery rate < 0.25, including all positive control gRNAs as well as targets previously described to be essential through our own orthogonal experiments (BCL2) or previously published findings (GLIS2, GATA3) (Thirant C et al. Cancer Cell 2017). Among transcription factors found to be essential, we identified several which had not previously been strongly associated with acute leukemia (FOXD4L1, ZNF740).

Conclusion: Patient-derived xenografts represent a unique platform for targeted functional genomic interrogation of acute leukemias. These screens can be performed with a tractable number of animals and modest sequencing requirements. Additional experiments are ongoing to identify the mechanistic implications of the vulnerabilities described here in ETO2/GLIS2-rearranged AMKL.

No relevant conflicts of interest to declare.