PDX models depend on KRAS and NPM1 for in vivo growth. (A) Experimental procedure for CRISPR/Cas9 in vivo screens performed with PDX models. Serially transplantable AML PDX models were established from primary patient AML cells and lentivirally transduced to express a split version of Cas9 together with a single-guide RNA (sgRNA) library (see supplemental Figure 1 for constructs). Transgenic cells were enriched by flow cytometry (Cas9–green fluorescent protein [GFP]) and puromycin selection (sgRNA library). Except for the input control aliquot, cells were injected into groups of mice and recovered from the mice at advanced leukemia stage (output). Next- generation sequencing (NGS) was performed and analyzed using the DepMap_CHRONOS, Lin et al,¹⁷ MAGeCK algorithm to compare sgRNA distribution between input and output. (B) CRISPR/Cas9 in vivo dropout screens were performed in 5 PDX AML models using the library of 34 genes recurrently mutated in AML; gene essentiality scores were calculated using the DepMap_CHRONOS algorithm (see supplemental Figure 2 for quality controls). (C) Experimental procedure for competitive in vivo assays for single-hit validation. sgRNAs targeting either KRAS or NPM1 or nontargeting (NT) sgRNAs (n = 3 per gene) were cloned into the sgRNA construct together with the appropriate fluorochromes and transduced into Cas9-GFP–expressing PDX cells. After puromycin selection, 3 subpopulations (KRAS KO, NPM1 KO, and NT sgRNA) were mixed at a 1:1:1 ratio as an input. Three replicate mixtures, each containing different sgRNAs, were transplanted into one mouse each (9 different sgRNAs per experiment in 3 replicate mice) and recovered at advanced disease stage (output). The distribution of the subpopulations was analyzed by flow cytometry (see supplemental Figure 3 for the step-by-step analysis and supplemental Figures 4 and 5 for quality controls). Blue fluorescent protein (BFP). (D) Representative flow cytometry plots for KRAS KO1 and NT-1 in AML-661, using Boolean gating. (E and F) Quantitative summaries of the knockout effects for NPM1 (E) and KRAS (F) in all PDX models studied. Each dot represents the percentage of gene of interest KO population from a single mouse, with related sgRNAs linked by a dotted line. Bar plots indicate mean, minimum, and maximum. The results of a 2-tailed paired t-test are shown if they were significant: ∗P < .05, ∗∗P < .01, and ∗∗∗P < .001.