Supplementary Figure S2: A. Counts of small deletions (<30kb) are compared in Sig3+ and Sig3- LMS tumors (analyzed by OncoPanel v3 and v 3.1, n=166). B. Counts of small deletions (<30kb) are significantly higher in tumors with deleterious alterations in POLH, POLD1, FANCM or XRCC1, all of which are implicated in DNA double strand break repair during different phases of the cell cycle (**** = p < 0.0001; n.s.= not significant)

Supplementary Table S5: Molecular features of patient-derived LMS models from Champions Oncology used in in vivo validation studies (five LMS PDX), including clinical and morphological features of LMS tumors from which they were derived.

Supplementary Figure S10: Effect of peposertib (100 mg/kg bid), low-dose pegylated liposomal doxorubicin (3 mg/kg qw) and combination on additional LMS PDX. A. CTG-1182. B. CTG-1005. C. CTG-1079.

Supplementary Figure S6 Representative cell culture images of LMS04 cultures untreated or treated with 2nM doxorubicin, 400 nM peposertib, or the combination of both, utilized for computationally assisted cell culture confluence analysis (See Supplementary Figure S7).

Supplementary Figure S9: Cell proliferation assay (BrdU incorporation) demonstrates that combination of DNA-PK inhibition with low-dose doxorubicin reduces LMS03 and LMS05 cell proliferation compared to either drug alone. Peposertib and AZD7648 are structurally unrelated DNA-PK inhibitors. Cells were treated for 7 days with each drug and the combinations at the indicated doses; BrdU incorporation over 24h was measured at day 7 with a luminescence-based ELISA assay (Roche)

Supplementary Figure S8: Synergistic effects of peposertib and low-dose doxorubicin combinations in LMS cells. Cell proliferation assay (BrdU incorporation) demonstrates that addition of low-dose doxorubicin to DNA-PKi treatment results in synergistic effects on LMS03, LMS04 and LMS05 cell proliferation. Cells were treated for 7 days with the drug combinations at the indicated doses; BrdU incorporation over 24h was measured at day 7 with a luminescence-based ELISA assay (Roche). Synergy scores and heatmap / surface plots were generated using SynergyFinder (RRID:SCR_019318)(35).

Supplementary Figure S4: PRKDC and RPA2 are essential for LMS cells, as demonstrated by cell morphology after shRNAmediated knockdown under puromycin selection at day 7 in LMS04 and LMS05 cells (quantification is provided in Figure 2C).

Supplementary Figure S3: Mutated genes, fraction and counts of non-structural mutations in each gene (right) in clinical samples from 251 LMS tumors sequenced with OncoPanel. A. Alterations in TP53, RB1 and ATRX were detected in 73%, 73% and 31% of LMS, respectively, including single copy deletions and low amplitude gains; B. Alterations in genes mutated in >4% of samples, excluding single copy deletions and low amplitude gains. Deleterious mutations impacting BRCA2 or BRCA1 were demonstrated in 49% of LMS and were most often heterozygous deletions, whereas 8% of LMS had identifiable BRCA1/BRCA2 intragenic mutations or homozygous deletions. As with other targeted NGS platforms, these findings underestimate the true frequency of TP53 and RB1 mutations, given challenges detecting inactivating structural variants, particularly in large genes such as RB1.

Supplementary Figure S7 Computationally assisted analysis of cell culture confluence of representative cell culture images of LMS04 cultures untreated or treated with 2nM doxorubicin, 400nM peposertib, or the combination of both. Segmentation was performed with the generalist deep learning-based segmentation algorithm Cellpose v0.6 (34). Multiple segmentation runs with increasing cell size diameter parameter (30-120px) were required to capture all cells.

Supplementary Table S3: List of pathogenic or likely pathogenic mutations in LMS based on 1) ClinVar (20); 2) functional classification as deleterious by both SIFT (21) and PolyPhen (22); and 3) truncating nature of the mutation.

Supplementary Table S2: Sig3 scores and FGA values in LMS tumor samples. Sig3+ samples are predicted with a classifier trained in WGS LMS data. FGA values are calculated on CNVkit-based copy number profiles (see methods).

Supplementary Table S4: Molecular features of patient-derived LMS models used in these studies (three LMS cells lines and one LMS PDX), including clinical features of LMS tumors from Dana-Farber/Brigham and Women’s Hospital from which they were derived

Supplementary Table S1: Cancer-associated genes interrogated by Oncopanel. Samples in this study were analyzed with three versions of the assay (v1, v2, v3 and v3.1) – the gene content of each panel is indicated

Supplementary Figure S1: A. Fraction of genome altered (FGA) in non-LMS cancers compared to LMS from GENIE v9.0 (37). B. FGA in LMS and in other tumor types in the GENIE dataset, including high grade sarcomas.

Supplementary Figure S5: Biochemical effects of increasing doses of doxorubicin in LMS04 cells, starting at unconventionally low doses.

Supplementary Figure S11: A-G. Relative body weight change of mice in cohorts treated with low-dose doxorubicin or pegylated liposomal doxorubicin, peposertib, and the combination of both, in the drug efficacy studies illustrated in Figure 5 and Supplementary Figure S9. H. Experimental design and drug doses utilized

Identifying genetic dependencies in human colon cancer could help identify effective treatment strategies. Genome-wide CRISPR-Cas9 dropout screens have the potential to reveal genetic dependencies, some of which could be exploited as therapeutic targets using existing drugs. In this study, we comprehensively characterized genetic dependencies present in a colon cancer organoid avatar, and validated tumor-specific selectivity of select pharmacologic agents. We conducted a genome-wide CRISPR dropout screen to elucidate the genetic dependencies that interacted with select driver somatic mutations. We found distinct genetic dependencies that interacted with WNT, MAPK, PI3K, TP53, and mismatch repair pathways and validated targets that could be exploited as treatments for this specific subtype of colon cancer. These findings demonstrate the utility of functional genomic screening in the context of personalized medicine.

Monocytic acute myeloid leukemia (AML) responds poorly to current treatments, including venetoclax-based therapy. We conducted in vivo and in vitro CRISPR-Cas9 library screenings using a mouse monocytic AML model and identified SETDB1 and its binding partners (ATF7IP and TRIM33) as crucial tumor promoters in vivo. . The growth-inhibitory effect of Setdb1 depletion in vivo is dependent mainly on natural killer (NK) cell-mediated cytotoxicity. Mechanistically, SETDB1 depletion upregulates interferon-stimulated genes and NKG2D ligands through the demethylation of histone H3 Lys9 at the enhancer regions, thereby enhancing their immunogenicity to NK cells and intrinsic apoptosis. Importantly, these effects are not observed in non-monocytic leukemia cells. We also identified the expression of myeloid cell nuclear differentiation antigen (MNDA) and its murine counterpart Ifi203 as biomarkers to predict the sensitivity of AML to SETDB1 depletion. Our study highlights the critical and selective role of SETDB1 in AML with granulomonocytic differentiation and underscores its potential as a therapeutic target for current unmet needs.

CRISPR/Cas9 library screens have revealed numerous oncogenes and tumor suppressors in malignant tumors. However, most of the genetic screens have been performed using in vitro culture assays, which may miss critical genes that specifically regulate tumorigenesis in vivo. To identify the therapeutic targets that specifically regulate tumorigenesis in vivo, we performed in vivo CRISPR/Cas9 library screens using two mouse models of myeloid tumors driven by MLL-AF9 and ASXL1/SETBP1 mutations (MA9 and cSAM cells, respectively). We ranked all genes according to in vivo specific essentiality, and identified a sarco/endoplasmic reticulum (ER) calcium-ATPase Atp2a2 (also known as SERCA2) as an in vivo specific tumor suppressor. To validate the results of our screens, we examined the effect of individual depletion of Atp2a2 in MA9 and cSAM cells. Interestingly, Atp2a2 depletion in these myeloid tumor cells showed contrasting results between in vitro and in vivo. Consistent with the screening results, Atp2a2 depletion accelerated the development of myeloid tumors in vivo. In sharp contrast, Atp2a2 depletion inhibited the growth of MA9 and cSAM cells in vitro. We then examined the effect of Atp2a2 depletion in MA9 cells on responses to cytotoxic drugs, cytarabine and a p53 activating agent, DS-5272. Treatment of C57BL/6 mice bearing MA9 cells with cytarabine or DS-5272 induced an enrichment of Atp2a2-depleted cells in the bone marrow, whereas no enrichment of Atp2a2-depleted cells was observed in vitro, again indicating that Atp2a2 depletion confers resistance to cytotoxic therapies in MA9 cells specifically in vivo. These results suggest that Atp2a2 is an in vivo specific tumor suppressor in myeloid tumors. Since ATP2A2 is known to play a key role in pumping Ca2+ from the cytosol into the ER lumen, we next investigated the effect of Atp2a2 depletion on Ca2+ homeostasis in MA9 cells. Atp2a2 depletion induced a decrease in ER Ca2+, an increase in cytosolic Ca2+, and an aberrant activation of Store-Operated Calcium Entry (SOCE) in response to the reduction in ER Ca2+ stores. Consequently, Atp2a2-depleted MA9 cells became dependent on Stim1 and Stim2, which play key roles in the SOCE pathway, indicating the essential role of SOCE in maintaining Ca2+ homeostasis in Atp2a2-depleted cells. The reduced ER Ca2+ also resulted in increased ER stress and downregulation of major histocompatibility complex classⅠ(MHC-I) in Atp2a2-depleted MA9 cells, which may promote immune evasion of myeloid tumors. Indeed, the proliferative advantage of Atp2a2-depleted MA9 cells in vivo was attenuated in the immunodeficient NSG mice and by the antibody-based CD8+ T-cells depletion in C57BL/6 mice. Regarding the differential influence of Atp2a2 depletion on AML growth in vitro and in vivo, we found that Atp2a2 depletion strongly induced cell cycle arrest in MA9 cells in the cytokine-rich in vitro culture, whereas it had little effect on the cell cycle in vivo. Thus, the context-dependent role of ATP2A2 in leukemogenesis could be partially explained by the reduced immunogenicity of Atp2a2-depleted cells and its differential impact on cell cycle progression in vitro and in vivo. Metabolomic analysis revealed that Atp2a2 depletion induced pro-tumorigenic metabolic alterations, such as enhanced purine metabolism and activation of the pentose phosphate pathway, which may also contribute to their increased leukemogenicity in vivo. Finally, we investigated the association between reduced ER Ca2+ levels and prognosis in human AML using RNA-seq data from Beat AML [Nature Medicine 26, 1852-1858 (2020)]. Consistent with the results from mouse AML models, human AMLs with low SERCA (ATP2A1+ATP2A2+ATP2A3) and ER Ca2+ uptake (ATP2A1/2/3-ITPR1/2/3-RyR1/2/3) scores showed significantly worse prognosis than those with high SERCA and ER Ca2+ uptake scores (p = 0.0296 and 0.0009, respectively). Thus, human AMLs with low ER Ca2+ appear to be resistant to current standard treatments with cytotoxic drugs. In conclusion, we performed the in vivo CRISPR/Cas9 library screens and identified Atp2a2 as a novel in vivo specific tumor suppressor in myeloid tumors. Deletion of Atp2a2 alters Ca2+ signaling, increases ER stress and confers resistance to cytotoxic drugs in myeloid tumors. Our data also suggest that SOCE is a synthetic lethal target for myeloid tumors with low ATP2A2 activity.

Identifying genetic dependencies in human colon cancer could help identify effective treatment strategies. Genome-wide CRISPR-Cas9 dropout screens have the potential to reveal genetic dependencies, some of which could be exploited as therapeutic targets using existing drugs. In this study, we comprehensively characterized genetic dependencies present in a colon cancer organoid avatar, and validated tumor-specific selectivity of select pharmacologic agents. We conducted a genome-wide CRISPR dropout screen to elucidate the genetic dependencies that interacted with select driver somatic mutations. We found distinct genetic dependencies that interacted with WNT, MAPK, PI3K, TP53, and mismatch repair pathways and validated targets that could be exploited as treatments for this specific subtype of colon cancer. These findings demonstrate the utility of functional genomic screening in the context of personalized medicine.