Supplementary table 6: Stress-focused drug screens AUC differences in SW-480 cells Area Under the Curve (AUC) from each compound of the stress-focused drug screen in the presence or absence of LB-100. Compounds are ranked by the difference in the AUC between LB-100-treated and untreated samples.

Figure S1: LB-100 engages stress-related, inflammatory response, and mitogenic signaling transcriptional programs in CRC cells (A) Dose-response assays show the effect of LB-100 in 7 CRC models. Cell viability was estimated by resazurin fluorescence after 5 days in the presence of the drug or DMSO control. The normalized values are plotted. (B) The heat map shows all “Hallmarks” and “KEGG” molecular signatures significantly enriched by LB-100 (4 µM) in both HT-29 and SW-480 cells in at least one of the addressed time points. Asterisks indicate significance level (* p-value <0.05, ** p-value <0.01, *** p-value <0.001)

Figure S10: Acquired resistance to the combination of LB-100 and Adavosertib suppress anchorage-independent and tumor growth in CRC models (A) Endpoint proliferation of HT-29 and SW-480 parental and resistant cells growing attached or in anchorage-independent conditions. Cells were plated in the same density on regular or cell-repellent culture plates in the absence of drugs and grown for 5-6 days. Proliferation was addressed by CellTiter-Glo 3D® and is expressed as fold change over T0. (B) Growth curves of the individual tumors from SW-480 parental and resistant cells measured 3 times per week. The dashed line indicates the 1500 mm3 ethical sacrifice.

Figure S6: LB-100 and adavosertib induce histological response in orthotopic CRC PDXs (A) Representative Hematoxylin & Eosin (H&E) stainings at endpoint from PDOX2 and PDOX3 treated as indicated. Original magnification middle images: 15x, scale bar 1000 µm; right images: 200x. S indicates stroma, and the arrows point to the tumor-cell component. (B) Mice body weight variation of the 3 CRC PDXs across the experiments.

Figure S8: Acquired resistance to the combination of LB-100 and adavosertib suppressed malignant traits in CRC models (A) IncuCyte-based proliferation assays from HT-29 and SW-480 parental and resistant cells in the absence or presence of the combination (LB-100 4 µM + adavosertib 400 nM). (B) Chromosome counting and representative chromosome spreads from HT-29 and SW-480 parental and resistant cells. Nocodazole was added for 3h to block cells in mitosis. Cells were harvested by mitotic shake-off for spreading. Over 40 (HT-29 and HT-29-R) or 50 (SW-480 and SW-480-R) spreads were counted per cell line. Asterisks indicate significance level (**** p-value <0.0001) by two-tailed Mann-Whitney test. (C) Heatmaps showing the marker genes of each cluster from the scRNAseq analyses of HT-29 and SW-480 cells

Supplementary table 5: Stress-focused drug screens AUC differences in HT-29 cells Area Under the Curve (AUC) from each compound of the stress-focused drug screen in the presence or absence of LB-100. Compounds are ranked by the difference in the AUC between LB-100-treated and untreated samples.

Supplementary table 4: The composition of the stress-focused drug library Compounds comprising the stress-focused drug library with their respective targets.

Figure S1: LB-100 engages stress-related, inflammatory response, and mitogenic signaling transcriptional programs in CRC cells (A) Dose-response assays show the effect of LB-100 in 7 CRC models. Cell viability was estimated by resazurin fluorescence after 5 days in the presence of the drug or DMSO control. The normalized values are plotted. (B) The heat map shows all “Hallmarks” and “KEGG” molecular signatures significantly enriched by LB-100 (4 µM) in both HT-29 and SW-480 cells in at least one of the addressed time points. Asterisks indicate significance level (* p-value <0.05, ** p-value <0.01, *** p-value <0.001)

Figure S5: Combined toxicity of LB-100 and Adavosertib in PDAC and CCA models (A) and (B) Dose-response assays show the effect of LB-100 or Adavosertib in 4 PDAC and 4 CCA models, respectively. Cell viability was estimated by resazurin fluorescence after 5 days in the presence of the drug or DMSO control. The normalized values are plotted. (C) and (D) Long-term viability assays show 4 PDAC and 4 CCA models, respectively, treated with LB-100 or Adavosertib at the indicated concentrations. Treatments were refreshed every 2-3 days, and the cells were grown for 10-14 days before fixing, staining, and imaging. (E) and (F) IncuCyte-based proliferation assays from 4 PDAC and 4 CCA models, respectively, in the absence or presence of LB-100, Adavosertib, or the combination at the indicated concentrations.

Figure S3: High N-Myc levels sensitize neuroblastoma cells to LB-100, and PP2A knockdown sensitizes CRC cells to WEE1 inhibition (A) Western blots comparing N-Myc levels in isogenic neuroblastoma models. GAPDH was used as a loading control. (B) long-term viability assays compares the toxicity of LB-100 across these neuroblastoma models. LB-100 was refreshed every 2-3 days, and the cells were grown for 10 days before fixing, staining, and imaging. (C) Western blots show the knockdown of PPP2R1A in HT-29 and SW-480 cells. GAPDH was used as a loading control. (D) Dose-response assays show the effect of Adavosertib after PPP2R1A knockdown compared to a control shRNA in HT-29 and SW-480 cells. Cell viability was estimated by resazurin fluorescence after 5 days in the presence of the drug or DMSO control. The normalized values are plotted. (E) The heat map shows all “Hallmarks” and “KEGG” molecular signatures significantly enriched by LB-100 (4 µM), calyculin A (5 nM), or okadaic acid (10 nM) in HT-29 and SW-480 cells after 8 hours. Asterisks indicate significance level (* p-value <0.05, ** p-value <0.01, *** p-value <0.001).

Figure S9: Single-cell RNAseq identify transcriptional signatures downregulated in CRC cells after acquired resistance to the combination of LB-100 and adavosertib UMAP representations of HT-29 (A) and SW-480 (B) cells colored by the activity scores for the indicated pathways. UMAPs colored by sample of origin from both cell lines are present of the left for reference. The boxen plots show the pathway activity scores for parental (red) and resistant (blue) cells.

Figure S4: Combined toxicity of LB-100 and Adavosertib in CRC models (A) Dose-response assays show the effect of Adavosertib in 7 CRC models. Cell viability was estimated by resazurin fluorescence after 5 days in the presence of the drug or DMSO control. The normalized values are plotted. (B) Long-term viability assays show 7 CRC models treated with LB-100 or Adavosertib at the indicated concentrations. Treatments were refreshed every 2-3 days, and the cells were grown for 10-14 days before fixing, staining, and imaging. (C) IncuCyte-based proliferation assays from 7 CRC models in the absence or presence of LB-100, Adavosertib, or the combination at the indicated concentrations. (D) Dose-response assays show the toxicity of LB-100 and Adavosertib in BJ and HaCaT cells compared to the average across the 7 CRC cell lines listed in the supplementary table S1. Cell viability was estimated by resazurin fluorescence after 5 days in the presence of the drug or DMSO control. The normalized values are plotted. (E) Synergy matrices and scores for the combination of LB-100 and Adavosertib in BJ and HaCaT cells. Cells were treated with 5 concentrations of LB-100 (1, 2, 3, 4, and 5 µM) or Adavosertib (100, 200, 300, 400, and 500 nM) and all respective permutations for 4 days. The percentage of cell viability for each condition was estimated by resazurin fluorescence and normalized to DMSO controls. Synergyfinder.org web tool was used to calculate the ZIP synergy scores and generate the plots.

Supplementary table 3: Full list of genes whose knockout attenuated LB-100 toxicity in SW-480 cells in the CRISPR-KO screen FDR smaller or equal to 0.25 and log2 fold change greater or equal to 1 in treated/untreated comparison were criteria for hit selection.

Supplementary table 7: Full list of genes whose knockout was selectively toxic in the presence of LB-100 in SW-480 cells in the CRISPR-KO screen FDR smaller or equal to 0.25 and log2 fold change smaller or equal to -1 in treated/untreated comparison were criteria for hit selection.

Figure S7: Normal tissues from the orthotopic CRC PDXs are not affected by LB-100, adavosertib, or the combination. Representative Hematoxylin & Eosin (H&E) stainings of the heart, liver, lung, and spleen from the PDOX1 PDXs treated as indicated. Original magnifications are indicated.

Supplementary table 2: Full list of genes whose overexpression was selectively toxic in the presence of LB-100 in HT-29 cells in the CRISPRa screen FDR smaller or equal to 0.25 and log2 fold change smaller or equal to -1 in treated/untreated comparison were criteria for hit selection.

Cancer homeostasis depends on a balance between activated oncogenic pathways driving tumorigenesis and engagement of stress response programs that counteract the inherent toxicity of such aberrant signaling. Although inhibition of oncogenic signaling pathways has been explored extensively, there is increasing evidence that overactivation of the same pathways can also disrupt cancer homeostasis and cause lethality. We show here that inhibition of protein phosphatase 2A (PP2A) hyperactivates multiple oncogenic pathways and engages stress responses in colon cancer cells. Genetic and compound screens identify combined inhibition of PP2A and WEE1 as synergistic in multiple cancer models by collapsing DNA replication and triggering premature mitosis followed by cell death. This combination also suppressed the growth of patient-derived tumors in vivo. Remarkably, acquired resistance to this drug combination suppressed the ability of colon cancer cells to form tumors in vivo. Our data suggest that paradoxical activation of oncogenic signaling can result in tumor-suppressive resistance. Significance: A therapy consisting of deliberate hyperactivation of oncogenic signaling combined with perturbation of the stress responses that result from this is very effective in animal models of colon cancer. Resistance to this therapy is associated with loss of oncogenic signaling and reduced oncogenic capacity, indicative of tumor-suppressive drug resistance.

Inactivation of the DNA mismatch repair (MMR) system, due to (epi)genetic alterations of MMR genes, increases the frequency of mutations across the genome, creating a phenotype known as microsatellite instability (MSI). Cancers with this phenotype have been associated with a better prognosis for some time, but only since recently it has been recognised as a predictive biomarker of response to immunotherapy. Because MSI tumours accumulate more insertions and/or deletions in coding regions of the genome containing microsatellites, there is an increase in neoantigens resulting from reading frame shifts, which promotes immunogenicity. To investigate if additional genes exist that can cause an MSI phenotype, we developed a fluorescence-based sensor to identify genes whose inactivation increases the rate of frameshift mutations on microsatellite sequences in cancer cells. Using genome-scale CRISPR/Cas9 screens, we identified MED12 as a potential new regulator of microsatellite instability. Consistent with this, we found that MED12 mutant colon cancers that lack mutations in the known MMR genes are more likely to be of the MSI phenotype.

Targeted inhibition of aberrant signaling is an important treatment strategy in cancer, but responses are often short-lived. Multi-drug combinations have the potential to mitigate this, but to avoid toxicity such combinations must be selective and given at low dosages. Here, we present a pipeline to identify promising multi-drug combinations. We perturbed an isogenic PI3K mutant and wild-type cell line pair with a limited set of drugs and recorded their signaling state and cell viability. We then reconstructed their signaling networks and mapped the signaling response to changes in cell viability. The resulting models, which allowed us to predict the effect of unseen combinations, indicated that no combination selectively reduces the viability of the PI3K mutant cells. However, we were able to validate 25 of the 30 combinations that we predicted to be anti-selective. Our pipeline enables efficient prioritization of multi-drug combinations from the enormous search space of possible combinations.