Targeting Myc-driven stress vulnerability in mutant KRAS colorectal cancer

Mutant KRAS is a key driver in colorectal cancer (CRC) and promotes Myc translation and Myc-dependent stress adaptation and proliferation. Here, we report that the combination of two FDA-approved drugs Bortezomib and Everolimus (RAD001) (BR) is highly efficacious against mutant KRAS CRC cells. Mechanistically, the combination, not single agent, rapidly depletes Myc protein, not mRNA, and leads to GCN2- and p-eIF2α-dependent cell death through the activation of extrinsic and intrinsic apoptotic pathways. Cell death is selectively induced in mutant KRAS CRC cells with elevated basal Myc and p-eIF2α and is characterized by CHOP induction and transcriptional signatures in proteotoxicity, oxidative stress, metabolic inhibition, and immune activation. BR-induced p-GCN2/p-eIF2α elevation and cell death are strongly attenuated by MYC knockdown and enhanced by MYC overexpression. The BR combination is efficacious against mutant KRAS patient derived organoids (PDO) and xenografts (PDX) by inducing p-eIF2α/CHOP and cell death. Interestingly, an elevated four-gene (DDIT3, GADD45B, CRYBA4 and HSPA1L) stress signature is linked to shortened overall survival in CRC patients. These data support that Myc-dependent stress adaptation drives the progression of mutant KRAS CRC and serves as a therapeutic vulnerability, which can be targeted using dual translational inhibitors. Supplementary Information The online version contains supplementary material available at 10.1186/s43556-022-00070-7.


Introduction
Colorectal cancer (CRC) is the third most common cancer worldwide, with an estimated 1.2 million new cases and over 600,000 death annually [1]. Genetic alternations in oncogenes and tumor suppressors cooperate with epigenetic alterations to drive colorectal carcinogenesis [2]. KRAS is the most frequently mutated oncogene in human cancer. A comprehensive up to date analysis revealed that ~ 19% of cancer patients harbor RAS mutations and over 85% are in KRAS, equivalent to ~ 3.4 million new cases per year worldwide [3]. Mutational activation of KRAS is an early event in CRC development and occurs in about 40-50% of cases, with hot spots in codons 12, Ruan et al. Molecular Biomedicine (2022) 3:10 largely through the dissociation of Eukaryotic Translation Initiation Factor 4E (eIF4E) from its inhibitor 4EBPs upon their phosphorylation by MAPK/ERK and PI3K/ mTOR/AKT signaling [11]. Elevated protein synthesis and demand on quality control in cancer cells lead to increased phosphorylation of Eukaryotic Translation Initiation Factor 2 alpha (eIF2α) (S51, p-eIF2α), the core regulator of the "integrated stress response" (ISR) [12].
There are four known eIF2α kinases in mammals which are activated by distinct and overlapping stresses. Among them, general control nonderepressible 2 (GCN2) is the most ancient and activated by amino acid starvation. A modest increase in p-eIF2α inhibits cap-dependent translation, while facilitating the translation of stress-related transcription factors such as Activating Transcription Factor 4 (ATF4) and C/EBP Homologous Protein (CHOP, encoded by DDIT3) to promote adaptation and survival through widespread changes in transcription, translation, metabolism, and myriad effectors [12,13]. Failure to adapt leads to prolonged p-eIF2α elevation and CHOP induction, and subsequent cell death through apoptotic mediators such as death receptor 5 (DR5), p53 Upregulated Modulator of Apoptosis (PUMA), and Phorbol-12-myristate-13-acetate-induced protein 1 (also called NOXA) [12,13]. Myc is a master regulator of oncogenic growth and metabolism [14][15][16][17], in part through elevated p-eIF2α and p-eIF4E/4EBP1 that promotes stress adaptation and survival of cancer cells [18][19][20]. This Myc-translation axis was therefore suggested as an exploitable vulnerability particularly in metabolically active tumors with mutated or amplified KRAS [16,21]. However, targeting a single Myc target [14,15], metabolic pathway, or a single step in mRNA translation [7,8], yielded little clinical efficacy in most solid tumors. For example, inhibitors against mRNA Cap binding, eIF4A helicase [7,8], or p-eIF2α inducing agents such as proteasome (Bortezomib and Ixazomib) [12,22] and HSP90 inhibitors [23,24], lack potency or selectivity. On the other hand, many kinase inhibitors target translation indirectly by blocking the phosphorylation of eIF4E or 4EBP [7,8]. These include allosteric mTOR inhibitors Everolimus (i.e., RAD001) and Temsirolimus, and EGFR antibodies or inhibitors. However, mutant KRAS and BRAF represents a major resistance mechanism to single agent in CRC preclinical models or patients [25][26][27][28] due to complex feedback activation of survival pathways [4,29]. Dual inhibition of ERK/MAPK and PI3K/mTOR signaling generally produces unacceptable normal tissue toxicity [7,8].
We reasoned that Myc-driven stress adaptation is a critical survival mechanism in mutant KRAS CRCs, and Myc translation might be an useful therapeutic target [30]. In this study, we focused on FDA-approved agents and discovered that the combination of Bortezomib and Everolimus synergistically and selectively kills mutant KRAS CRC cells at concentrations where single agents had little or no toxicity. The efficacy of this combination is validated using patient derived organoids (PDOs) and xenografts (PDXs). Mechanistically, we showed that mutant KRAS-dependent vulnerability is mediated through elevated Myc, and its ablation leads to GCN2/p-eIF2α -dependent cell death with profound trancriptional signatures of proteotoxcity, oxidative stress and metabolic suppression. Our study provides a potential way to improve the treatment of mutant KRAS CRCs by targeting the deregulated Myc-ISR axis.

Bortezomib and Everolimus synergistically kill mutant KRAS CRC cells
To identify potential combination targeting mutant KRAS in CRC, we first performed a targeted drug screen using 11 translational inhibitors and HCT116 cells. These included pathway agents indirectly targeting p-4EBP1 or p-eIF4E, and ones targeting the cap binding complex (eIF4A, eIF4F assembly), or the 43S preinitiation complex (eIF2α). As expected, these agents showed massive differences in IC50 as single agent (over 20,000-fold, from low nM to sub mM). Cap analog 4Ei-1 was the least potent (at 500 µM or higher), while the p-eIF2α inducer bortezomib and eIF4A inhibitor were among the most potent (10-20 nM, Table S1). Direct or indirect kinase inhibitors had low to modest toxicity with IC50s ranging between 20-100 µM. The combination of Bortezomib (B) and Everolimus (R) (BR, hereafter) showed strong synergy in suppressing cell growth (Fig. 1a). Both are FDA-approved drugs and chosen for further study.
We validated the synergy of BR using three additional KRAS mutant CRC cells (DLD1, SW480 and LS180) and viability assay with combination index calculated (CI < 0.5) (Fig. 1a-b). The BR combination (B 5 nM and R 10 µM) strongly suppressed cell growth at 48 h (Fig. 1c), and long-term clonogenicity on day 14 with 24 h exposure. In contrast, either agent alone at the same doses showed little or no toxicity (Fig. 1d-e). The combination potently induced apoptosis as evidenced by increased Annexin V positive cells at 48 h ( Fig. 1f and Fig. S1a) and cleaved caspase-3 at 24 h in all four lines (Fig. 1g). Compared to either agent, BR induced much higher levels of phosphorylation of eIF2α (S51) and CHOP, and a strong reduction in Myc protein in all four lines (Fig. 1g). These KRAS mutated lines are all WT for BRAF but otherwise vary in mutational status of APC, CTNNB1 (β-catenin), PIK3CA, PTEN, and TP53 ( Fig. 1g and Table S2). We further confirmed induction of prolonged ISR with elevated ATF4, GRP78/BIP and spliced XBP1s in HCT 116 cells , Everolimus (R), or their combination (BR, 5 nM and 10 µM or specified ranges). a Cell growth at 48 h was assessed using MTS assay. b Calculated combination index (CI). The CI presents synergism (CI < 1), additive effect (CI = 1), and antagonism (CI > 1). Red area presents synergism, and green area presents antagonism. Fa represents fraction affected. c Attached cells were visualized by crystal violet at 48 h. (D) Colony formation assay. Cells were treated for 24 h and replated in drug free medium for 14 days before crystal violet staining. e The relative number of colonies as in (d) was normalized to untreated group (100%). f Apoptosis was quantitated by Annexin V + cells using flow cytometry. g Indicated proteins at 24 h was detected using western blotting. ACTIN is used as loading control. The status of major drivers is shown below. e, f, values are mean + s.d. (n = 3). ***p < 0.001 (Student's t-test, two tailed). B, R vs. BR by BR, not by single agent (Fig. S1b). BR and single agent reduced 4E-BP1 phosphorylation (S65/70) similarly, but had limited effect on 4E-BP1 (T37/46) or S6 phosphorylation (S235/236) (Fig. S1b). These results demonstrated that BR potently induces Myc ablation and stress-associated cell death in mutant KRAS cells.

The BR combination induces p-eIF2α-dependent killing of mutant KRAS CRC cells
To further explore the mechanisms underlying BRinduced cell killing, we conducted RNA-Seq analysis on HCT116 cells. Differentially expressed genes (DEGs) included 1231 upregulated and 1285 downregulated by two-fold and more at 24 h (p ≤ 0.005) (Fig. 2a). Gene ontology (GO) analysis indicated that upregulated genes are highly enriched in pathways for proteostress and unfolded protein response (UPR), including multiple HSP family members and classical ISR markers CHOP and Growth Arrest and DNA Damage-Inducible Protein 45B and 34 (GADD45B and GADD 34) ( Fig. 2a-b). Other highly upregulated pathways included extrinsic apoptosis, oxidative stress, and, surprisingly, immunity (Fig. 2b). The top 10 enriched pathways in downregulated genes were related to metabolism and cell cycle (Fig. S2a). Gene set enrichment analysis (GSEA) confirmed positive enrichment in misfolded proteins and leukocyte activation among others ( Fig. S2b and data deposit).
qRT-PCR analysis confirmed the induction of the proximal ISR regulators CHOP(DDIT3), GADD45B, GADD34, and ATF4, as well as apoptotic effectors DR5, PUMA, NOXA, and BIM by the combination, but not single agents in HCT 116 cells (Fig. 2c). The induction of apoptotic proteins was also confirmed by western blotting with cleaved casapase-8 and -9 (Fig. 2d). Using isogenic HCT 116 cells deficient in one of these four apoptotic effectors [31,32], we found that BR-induced cell killing is abrogated by DR5 knockout (KO) or PUMA KO, but minimally affected by NOXA KO or BIM KO (Fig. 2e-f ).
Prolonged p-eIF2α elevation and CHOP induction leads to cell death [12]. p-eIF2α can be inhibited by the overexpression of a phosphorylation defective and dominant negative mutant allele (serine 51 to alanine, herein referred to as S51A) [10,28]. The expression of eIF2AS51A significantly blocked BR-induced cell loss, with marked reduction in p-eIF2α, CHOP, DR5, and cleaved caspase-3 ( Fig. 2g-h), and expression of ISR targets CHOP, GADD45B, GADD34, ATF4, and apoptotic effectors (DR5, PUMA, NOXA and BIM) (Fig. S2c). Inhibition of apoptosis was confirmed by nuclear fragmentation assay and flow cytometry (Fig. S2d-f ). These data demonstrate that sustained p-eIF2α leads to the killing of mutant KRAS CRC cells through the activation of intrinsic and extrinsic apoptotic pathways upon BR treatment.

The BR combination triggers Myc-dependent activation of GCN2
The strong metabolic suppression by BR observed at 24 h ( Fig. S2a) prompted us to examine the role of GCN2, a classical metabolic stress sensor and eIF2α kinase [13]. Elevation of p-eIF2α was observed at 2-4 h in HCT 116, DLD1, SW480, and LS180 cells (Fig. 3a), in parallel with elevated p-GCN2 in 3 of 4 lines. Despite decreased Myc protein (Fig. 1g), little or no change in MYC mRNA was detected at 24 h in any line (Fig. S3a). Interestingly, loss of Myc and 4EBP1 (S65/70, not T37/46) was rapid and near complete at one hour, when p-GCN2, p-eIF2α and ATF4 just began to rise, continuing to 4 and 24 h in HCT 116 cells (Fig. 3b,  Fig. S3b). We detected no change in total eIF4E or p-eIF4E (S209) and decreased p-PERK, the UPR sensor (Fig. 3b). The levels of CHOP and death effectors such as DR5 and cleaved caspase-3 began to rise only at or after 12 h (Fig.  S3b). Consistent with Myc loss, significant enrichment of Myc down-regulated targets was evident at 24 h (Fig. 3c).
We further examined the role of GCN2 and Myc in BR-induced cell death. GCN2 siRNA markedly reduced BR-induced growth inhibition and apoptosis ( Fig. 3d-e, Fig, S3c), as well as the levels of p-eIF2α, ATF4, CHOP, caspase-3 cleavage, or ISR targets, with a minor effect on Myc loss ( Fig. 3f-g). MYC siRNA strongly reduced BR-induced cell death, p-GCN2/p-eIF2α/CHOP and caspase-3 cleavage (Fig. 3h-i). These results suggest that acute Myc loss in mutant KRAS CRC cells impairs stress adaptation and leads to GCN2/p-eIF2αdependent metabolic crisis and cell death.

The BR combination promotes mutant KRAS-Myc selective proteotoxicity
We further determined if elevated Myc in mutant KRAS CRC cells [30] is the target of BR. Compared to mutant KRAS CRC cells (n = 4), WT KRAS CRC cells (n = 4) showed higher IC50, reduced apoptosis, and lower induction of stress (p-eIF2α, CHOP, GADD45B), DR5, and cleaved caspase-3 ( Fig. 4a-b, Fig. S4a-b). To minimize the influence of genetic background, we took advantage of isogenic mutant and WT KRAS CRC cell pairs previously established [33]. We confirmed higher BR sensitivity in mutant KRAS isogenic cells (HCT116, DLD1 and SW48) by growth suppression and apoptosis induction ( Fig. 4c-d). qRT-PCR indicated induction of some ISR regulators and apoptotic effector genes (8) by BR in WT KRAS isogenic cells, albeit at much reduced levels compared to those in mutant KRAS counterparts (Fig. 4e). Myc protein, not MYC mRNA, was effectively ablated by BR within all lines ( Fig. 4e-f ). Isogenic WT KRAS cells showed lower basal Myc and p-eIF2α, and lacked the induction in CHOP, apoptotic targets, or cleaved caspase-3 at 24 h (Fig. 4f ).
We then used RNA-seq to compare global transcriptional response of isogenic WT and mutant KRAS HCT 116 cells to BR treatment. Consistent with the lack of acute stress or death, WT KRAS cells displayed a drastically reduced global transcriptomic response (Fig. S4c). BR-treated WT KRAS cells showed profound decrease of pathways in proteostress, intrinsic and extrinsic apoptosis, and HIF1α and immunity (-Log10 (pvalue) from 15-5) (Fig. 4g, Fig. S4d), and increase of pathways in development, cell cycle, metabolism, and chromosome segregation (Fig. S4e). Interestingly, top BR differential genes appeared to be regulated at opposite directions in this isogenic pair ( Fig. S4f-g). Despite little or no change in MYC mRNA, GSEA showed negative enrichment of Wnt-Myc in BR-treated mutant KRAS cells (Fig. 4g-h). Overexpression of MYC in WT KRAS cells increased p-GCN2/p-eIF2α/CHOP, and DR5 and cleaved caspase-3 at 24 h and cell death at 48 h upon BR treatment, with ( Fig. 4i-j). Together with results from MYC siRNA, these data demonstrate a fundamental role of Myc in regulating metabolic and transcriptomic response in mutant KRAS CRC cells.

The BR combination is effective against mutant KRAS CRC PDOs
Patient derived organoids (PDOs) retain the architecture, genotype, and phenotype of the patient's primary tumor and provide a rapid in vitro model for drug testing [34]. We then tested the efficacy of BR using mutant KRAS CRC PDOs. These PDOs had two distinct morphologies as cysts or clusters [35], and did not differ significantly in growth (Fig. S5a). PDOs were significantly more sensitive to the BR combination, compared to single agent ( Fig. 5a-b). The BR treatment induced marked apoptosis (cleaved-caspase-3) in the center of PDO, along with highly elevated p-eIF2α and expression of ISR effectors at 24 h (Fig. 5c-e, Fig. S5b).

MSS CRC PDXs
The above data suggest that Myc-driven stress adaptation might be a selective target in aggressive CRCs. Patient derived xenograft (PDX) models preserved tumor histology and heterogeneity [37], and were used to test the efficacy of the BR combination. We selected two PDXs (mutant KRAS and MSS). PDX1 (KRASG13D) and PDO1 were originated from the same tumor. PDX2 (KRASG12D) is therapy resistant and from a metastatic lesion of deceased patient after multiple lines of chemoand targeted therapies (FOLFIRI, FOLFOX, 5-FU, bevacizumab, onalespib, indimitecan, selumetinib and MK-2206). PDX1 and PDX2 were highly responsive to the BR treatment ( Fig. 6a-b, Fig. S6a), showing massive loss of proliferation (Ki67 index) and reduction in cellularity upon BR treatment (Fig. S6b-c). The BR treatment was well tolerated and induced a slight and transient weight loss compared to the control group, which recovered by day 9 (Fig. S6d).
To monitor drug-induced acute response in vivo, we first analyzed several ISR and cell death markers in PDXs on day 4, which is 24 h after the second BR treatment. BR group showed significant increases in p-eIF2α, CHOP and cleaved caspase-3 by staining and western blotting (Fig. 6c-e). qRT-PCR confirmed the induction of ISR effectors (Fig. 6f, Fig. S6e), and a similar pattern between PDO1 and PDX1. We used RNA-Seq to assess BR-induced global transcriptomic changes in PDX1, and identified 469 upregulated and 522 downregulated genes (filtered for human reads,twofold or more, p ≤ 0.005). Upregulated genes were highly enriched in proteo-and oxidative stress, apoptosis, and immunity ( Fig. 6g-h), while downregulated genes were highly enriched in metabolism such as oxidative phosphorylation and nucleotide and ATP biosynthesis (Fig. S6f ). Upregulated genes shared by HCT 116 and PDX1 were highly enriched in proteostress and extrinsic apoptotic pathway (Fig. S6gh). Like cell line studies, BR treatments significantly enriched Myc down-regulated targets (Fig. 6i), with little effect on MYC mRNA (Fig. 6g). These data collectively support that the BR combination targets Myc-dependent stress adaptation in mutant KRAS CRCs to promote ISRdependent metabolic crisis and cell death (Fig. 6j).

Discussion
Targeting mutant KRAS is a major clinical challenge and the "holy grail" in cancer therapy. Mutant KRAS is linked to poor prognosis in CRC, and promotes resistance to EGFR antibodies [4] and anti-PD-1 [38]. Resistance is in part mediated through feedback activition of adaptive responses to avoid drug or immune-mediated cell killing [4,39]. Here, we report a highly effective combination of two FDA-approved drugs, Bortezomib and Everolimus, against mutant KRAS CRC cells, PDOs, and PDXs. Mechanistically, the BR combination, but not single agents, ablates high Myc levels in mutant KRAS CRC, leading to unresolvable proteostress and cell death with a transcrptomic signature charaterized by proteotoxicity, oxidative stress, metabolic suppression, and immune activation (Fig. 6j). The vulerability to this drug combination is seletive to the mutant KRAS-Myc axis as demonstrated using isogenic and nonisogenic CRC cells. Our study therefore provides a potentially new therapeutic strategy to target Myc in mutant KRAS CRCs using FDAapproved drugs.
KRAS is a membrane-bound GTPase that cycles between GTP-bound active and GDP-bound inactive forms. Most oncognic mutations affect this on-off switch and lock the protein in the active form to drive cell proliferation, apoptosis resistance, and metastasis [4]. KRAS mutations are biochemcially distinct and appear to influence CRC patient outcomes [5], making it challening to develope allele spcific RAS inhibitors. Despite some encouraging data on G12C and G13C inhibitors in lung cancer [5,6], G12D and G12V mutations are the most prevalent KRAS mutations in CRC [3] associated with worse overall survival [40]. Indirectly targeting mutant KRAS also has not had much succes [4,29]. Our data suggest elevated Myc and protesostress as a druggable vunerability across mutant KRAS CRC cell lines, PDXs and PDOs using BR combination. Elevated proteostress is also reported in cancers with amplified KRAS [21] or MSI-high [41]. KRAS mutations are prevalent in pancreatic cancer (90%), lung cancer (20-30%), and endometrial cancer (18%) [3]. It is tempting to sepculate that this combination might be effective in other epithelial cancers with mutant KRAS.
Myc is a master regulator of oncogenic growth through extensive transcriptional and translational networks, and cooperates with a variety of cofactors [16,42,43]. Myc promotes stress adaptation and survival of cancer cells by increased autophagy [44], preservation of bioenergetics [45] through transient inhibition of RNA pol IImediated transcription [46] or protein synthesis through GCN2/p-eIF2α-dependent negative feedback [28]. Myc translation is highly regulated and enhanced by mutant KRAS [30,47]. However, this Myc-translation feed-forward loop is notoriously difficult to break [14,15]. Our data support that Myc ablation is likely required to disrupt these protective mechanisms in mutant KRAS cells 'addicted" to Myc (Fig. 6j). Consistent with this model, BR, but not single agents, rapidly ablates Myc protein, not mRNA, leading to sustained induction of p-GCN/p-eIF2α/CHOP. MYC siRNA decreased sensitivity while MYC over-expression increased sensitivity through the p-GCN/p-eIF2α/CHOP axis in mutant and WT KRAS isogenic cells, respectively. The combination is therefore necessary to push mutant KRAS CRC cells out of Myc-dependent adaptive "Goldilocks Zone" [8] and into metabolic crisis and cell death. However, the role of other eIF2α kinases in cell death cannot be ruled out due to their well-documented crosstalk and shared downstream targets [7,12,13,48]. CRCs are heterogeous and can be classified into several molecular groups based on gene expression [49]. The majority of CRC are MSI-L and MSS, associated with a high Wnt/Myc signature, and do not respond to immune checkpoint inhibitors such as anti-PD-1 [50,51]. Overwhelming evidence supports the cooperation of mutant KRAS and Myc in metabolic reprogramming and therapeutic resistance through the tumor microenvironment [18][19][20]52]. Here, we found that the 4-gene ISR signature (DDIT3, GADD45B, CRYBA4, and HSPA1L) is strongly associated with poor prognosis in CRC patients, but also with sensitivity to BR in vitro and in vivo. Our model (Fig. 6j) helps explain this paradox. As mutant KRAS cells become dependent on elevated Myc and ISR (p-GCN2/p-eIF2α) for metabolic adaptation and immune suppression, acute Myc ablation breaks this state by inducing even higher and sustained ISR and cell death (Fig. 6j). Mutant KRAS-Myc strongly increases the range of transcriptional response and level of p-eIF2α to BR treatment, but it remains likely that additional BR targets are involved given the large number of Myc cofactors [16,42,43]. Elevated p-eIF2α is a suggested marker of immunogenic cell death (ICD) [53,54]. The use of syngeneic models can help better understand drug-induced synthetic lethality for diffucult targets [55] through the tumor microenvironment (TME). The challenge however remains to develop mechanism-based clinical biomarkers distinct from activated oncogenes. Our data support that PDO and PDX might be useful in this regard, as those from the same patient showed similar stress-related pathway gene expression changes.
In summary, our study demonstrates a critical role of Myc-mediated stress adaptation in the surival of mutant KRAS CRC. The potent efficacy of the combination of FDA-approved mTOR and proteasome inhibitors is mediated through Myc ablation and induction of p-GCN2/p-eIF2α-dependent cell killing. With multiple FDA-approved agents in each class, the prevelance of KRAS mutations, and mechanistic biomarkers, it would be interesting to design clinical trials to evaluate potential benefit of BR or similar combinations in CRC patients.

Approval and protocols
All methods were performed in accordance with the relevant guidelines and regulations. The protocols for the use of recombinant DNA and animals included IBC201700136 and IACUC# 17071072. The protocol of establishing PDO/PDX includes REN11110076/ IRB0411047.

Cell culture, treatment, and transfection
The human colorectal cancer cell lines, including HCT116, DLD1, SW480, LS180, LIM1215, SW48, DIFI and RKO were obtained from the American Type Culture Collection (Manassas, VA, USA) ATCC. Isogenic KRAS pairs HCT116 (WT, G13D), DLD1 (WT, G13D) and SW48 (WT/G12V) cell lines were obtained from Bert Vogelstein at Johns Hopkins University [33]. PUMA KO [31], DR5 KO, NOXA KO, BIM KO [32,56] HCT 116 cells were generated in the lab. Information on major drivers or isogenic cell lines are found in Supplementary Materials (Table S1). Cell lines were regularly monitored for absence of Mycoplasma, approximately every 6 months. Any cell line is used for less than 2 months (10 or fewer passages) in culture upon thawing from LN tank. All cell lines were cultured in McCoy's 5A modified medium Cell viability was performed using the MTT assay kit (Promega, Madison, WI, #G3580) as described [30]. In brief, cells were seeded in 96-well plates at a density of 1 × 10 4 cells/well and treated with different agents for 48 h before the analysis. Combination index(CI) and fraction affected(Fa) values were calculated using Compusin software (https:// www. combo syn. com).
Crystal violet staining. Following various treatment, attached cells or clones were stained and with crystal violet (Sigma, St. Louis, MO, Cat# C0775) (3.7% Paraformaldehyde, 0.05% crystal violet in distilled water and filtered at 0.45 um before use) (2). For colony formation assays, equal numbers of cells were subjected to various treatments and plated into 12-well plates at different dilutions. Colonies were visualized by crystal violet staining 14 days after plating. Each assay was conducted in triplicate and repeated three times.

Apoptosis assays
Cell death and apoptosis was analyzed by nuclear staining with cells harvested from 12-well plates and Hoechst 33,258 (Invitrogen, Cat# 40,045), and Annexin V/propidium iodide (PI) followed by flow cytometry as described [57]. Experiments were repeated on two or more occasions (different days) with similar results. Adherent and floating cells were harvested, stained with Hoechst 33,258 (Invitrogen), and analyzed for apoptosis by nuclear staining and counting cells with condensed and fragmented nuclei. At least 300 cells were analyzed for each treatment. Annexin V/propidium iodide (PI) staining was performed. Flow cytometry plots and quantitation were based on the analysis of 20,000 cells for each condition. Results from one representative experiment are shown with fraction (%) of indicated population.

Western blotting and quantitative Reverse transcription PCR (qRT-PCR)
Western blotting was performed as previously described [26,30]. Details on antibodies are found in the Supplementary Materials (Table S4). Total RNA was isolated from cells or tissues using Mini RNA Isolation II Kit (Zymo Research, Orange, CA, Cat# R1054) according to the manufacturer's protocol. One μg of total RNA was used to generate cDNA using SuperScript III reverse transcriptase (Invitrogen, Carlsbad, CA, Cat# 18,064-014). Details on primers are found in the Supplementary Materials (Table S5). For organoids, pooling of 8 or more wells for each condition is necessary to prepare enough lysates or total RNA. For tumors, lysates were pooled from 3 randomly chosen tumors in each group. cDNA was synthesized from RNA pooled from 3 randomly chosen tumors in each group.

Patient derived CRC organoids (PDOs)
PDOs were established using surgically resected deidentified CRC tissues from the Biospecimen Core (PBC) at University of Pittsburgh with tissue collection under informed consent and usage approved by the Institutional Review Board at the University of Pittsburgh. CRC organoids were cultured in Matrigel (Corning) incubated with advanced DMEM/F12 (Invitrogen) medium with supplements including 50% (v/v) L-WRN-conditioned medium containing Wnt3a, R-spondin, and Noggin prepared as described [30,58]. More details on medium, passage and treatment are found in supplemental materials.

Patient derived CRC xenografts (PDX)
Animal experiments were approved by the University of Pittsburgh Institutional Animal Care and Use Committee. All methods were performed in accordance with the relevant guidelines and regulations. Patient-derived xenograft (PDX) tumors were established and propagated in 5-6-week-old female NOD.Cg-Prkdcscid Il2rgtm1Wjl/ Saju (NSG) mice (Jackson Laboratory, Bar Harbor, ME) as described (4, 10) using samples collected with IRB approval and obtained from the NCI.
Tumor-bearing mice were randomized into untreated and treated groups. Mice were treated with BR Injection every other day. B was given by intraperitoneal (IP) injection at 0.5 mg/kg, R was given by oral gavage (OG) at 10 mg/kg. Tumor growth was monitored by calipers, and tumor volumes were calculated according to the formula 1/2 × length × width 2 . Ethical endpoint was defined as a time point when a tumor reached 2 cm or more in any dimension. Tumor tissues were analyzed for histology, staining, protein, and mRNA expression. Selected tumors were pooled to prepare protein or RNA which was then used for western blotting, RT-PCR and RNA-seq. More Details on PDX establishment and treatment and analysis are found in supplemental materials.

RNA sequencing (RNA-Seq)
Total RNA was prepared from cells and tissues using TRIzol RNA Isolation Reagents (Sigma, St. Louis, MO, Cat# 15,596,026) according to manufacturer's instructions. Library construction, RNA sequencing (RNA-seq), and data analysis were performed by Novogene using the IIIumina HiSeq platform. Sample quality was assessed by HTSeq v0.6.1 to the count the read numbers mapped of each gene. FPKM (fragments per kilobase of transcript per million mapped reads) of each gene was calculated based on the length of a gene and read counts mapped to this gene. For PDX samples, Flow B was used to filter and map human reads with Fragments Per Kilobase Per Million (FPKM) calculated based on all mapped reads. For samples without biological replicates, readcount was adjusted by TMM, then differential expression significant analysis was performed by using the edgeR package, while the significant criterion are both p value < 0.005 and |log2(Fold Change)|> 1 (i.e., twofold).

Bioinformatics
More details on differential expression analysis and visualization such as Volcano plots, Gene Ontology (GO), Gene Set Enrichment Analysis (GSEA) and Venn diagram are found in supplemental materials.

Data deposit
Analyzed RNA-seq data on differential gene expression, including raw readcount and normalized abundance of all called genes in paired samples (DEG, no cutoff ), and differentially expressed genes (DEG_all, including both ups and downs, with indicated cutoff ), additional gene list (s) and analyses(GSEA) are deposited at DRYAD. This dataset has been assigned a unique identifier or DOI (https:// doi. org/ 10. 5061/ dryad. sf7m0 cg6h) for free public access upon the publication of manuscript.

Statistical analysis
Statistical analyses were carried out using GraphPad Prism software (VIII, GraphPad Software, Inc., La Jolla, CA). Comparisons between two groups were made by two-tailed, unpaired t test. Differences were considered significant if the probability of the difference occurring by chance was less than 5 in 100 (p < 0.05). The means ± one standard deviation (s.d.) were displayed in the figures. Sample size was determined using a combination of published work and power calculations.