A stock solution of E7386 dissolved in DMSO to 10 mM and E7386 powder (Eisai Co., Ltd.) diluted with 100 mM hydrochloric acid was used for in vitro and in vivo studies, respectively.

A total of 45 surgical CRC specimens collected after the pathological examination of patients with CRC at the National Cancer Center Hospital (Tokyo, Japan) were received between February 2016 and February 2018 and used to prepare organoids/fibroblast culture and implantation into mice for the establishment of PDX. Of the 45 cancer patients, 60% were males and 40% were females, and average ages were 62 years old, ranging from 36 years old to 88 years old. All experiments were performed following the relevant domestic guidelines and regulations in Japan, including Ethical Guidelines for Medical and Health Research Involving Human Subjects. The use of surgical specimens in this study was approved by the Ethics Committee of the National Cancer Center (approval no. 2015-108) and written informed consent was obtained from all patients. In the present study, three cases of typical gene mutational statuses [case 21: KRAS, APC and TP53; case 28: PI3K catalytic subunit (PIK3CA) and TP53; and case 32: APC and TP53] were used (9). Clinical and pathological information has been previously reported (9).

The organoids and fibroblasts used in the present study were established from surgical specimens. The culturing of organoids and fibroblasts was performed as previously described (9). Briefly, for co-culture, organoids and fibroblasts were cultured in inserts with porous membranes and carrier plates (pore size, 1 µm; cat. nos. 353104 and 353504; Corning, Inc.). In total, 1 day before co-culture, fibroblasts (1×10⁴ cells/well) were seeded into 24-well plates with RPMI-1640 (cat. no. 189-02025; Fujifilm Wako Pure chemical Corporation) containing 10% FBS (cat. no. SH30075; Hyclone; Cytiva). Organoids (1×10⁴ cells/insert) were seeded into polymerized Matrigel (cat. no. 354234; Corning, Inc.) in cell culture inserts, followed by setting the inserts in companion carrier plates, and a medium for organoid culture was added into the basal compartments. The next day, the organoids were covered with Matrigel, and the inserts containing the organoids were transferred to the 24-well plates containing the fibroblasts. During co-culturing, the organoid culture medium was applied to both organoids and fibroblasts. After E7386 incubation for 6 or 24 h, organoids and fibroblasts in each insert and well were collected separately for the analysis of gene expression.

Surgical specimens were cut 2 or 3-mm cubes or established CRC organoids were implanted into the subcutis of female NOD.Cg-Prkdc^scidIl2rg^tm1Sug/ShiJic (NOG) mice at 7 weeks of age (In-Vivo Science, Inc.) under isoflurane anesthesia (4% for initiation and 2% for maintenance), and their body weights were 20.8±0.8 g at 14–16 weeks of age (at the start of drug administration). The mice were housed in plastic cages (n≤5 per cage) with recycled paper bedding in a specific-pathogen-free environment maintained at 22±1°C and 55±10% relative humidity, with 12-h light/dark cycle and were provided free access to the CE2 standard chow diet (CLEA Japan, Inc.) and tap water. Mouse experiments were performed at the Animal Facility of the National Cancer Center Research Institute (Tokyo, Japan) according to institutional guidelines. Ethics approval was obtained from the National Cancer Center Animal Ethics Committee (approval no. T17-006).

For the in vitro studies, E7386 at concentrations of 0, 10, 30, 100 and 300 nM in culture media was used, before cell viability was measured using the RealTime-Glo™ MT Cell Viability Assay solution kit (cat. no. G9711; Promega Corporation). Organoids/CAF samples planned for RNA analysis were collected 6 and 24 h after treatment, whereas those planned for metabolome analysis were collected 24 h after the addition of E7386. Triplicate samples of organoids and corresponding CAFs from each case were used for the measurement of cell viability.

For the in vivo PDX mouse model, E7386 solution was administered at 0 and 50 mg/kg body weight by gavage twice a day for 14 days. The number of mice used in the experiments were five of case 21, five of case 28 and three of case 32 per the control and E7386-treated group each. The size of the implanted tissue was measured using a caliper. In total, 6 h after the last drug administration, all mice were sacrificed by cervical dislocation under 4% isoflurane anesthesia before subcutaneous PDX tissues were excised cut into pieces, frozen and stored at −80°C or fixed with 10% neutral buffer formalin.

Total RNA was isolated using the RNeasy Micro Kit (cat. no. 74104; Qiagen GmbH). Triplicate samples were prepared for the RNA cocktail used for DNA microarray analysis from E7386-0, 30, and 100 nM groups. The quality of RNA samples and DNA microarray analysis [SurePrint G3 Human GE Microarray GE 8×60 K Ver.3.0 (cat. no. G4851C; Agilent Technologies, Inc.)] were evaluated as previously described (9). For RT-qPCR analysis, aliquots of total RNA were subjected to a reverse transcription with random primers using the High-Capacity cDNA Reverse Transcription Kit (cat. no. 4368814; Applied Biosystems; Thermo Fisher Scientific, Inc.). qPCR in duplicate in three independent experiments was performed using the SsoAdvanced™ Universal SYBR-Green Supermix (cat. no. 172-5271; Bio-Rad Laboratories, Inc.) in the DNA Engine Opticon^® 2 system (MJ Research, Inc.; Bio-Rad Laboratories, Inc.). The primer sequences are shown in Table SI. Data were normalized to the housekeeping gene β-actin, calculated using the 2^−ΔΔCq method (11) and are presented as the mean ± standard deviation (1.0-fold was as control).

To clarify the biological meaning of the key molecules, gene information obtained in the DNA microarray analysis was loaded into Metascape 3.5 (https://metascape.org) for GO enrichment analysis (12). Terms with a P<0.01, ≥3 and an enrichment factor >1.5 were collected before being grouped into clusters based on their membership similarity (κ-score >0.3).

For hydrophilic metabolite analysis, capillary electrophoresis time-of-flight mass spectrometry (CE-TOFMS) was used for cationic and anionic metabolite analyses. To extract the metabolites, cultured organoids from E7386-0, 30, and 100 nM groups were collected after adding 1.0 ml methanol containing internal standards (25 µM each of methionine sulfone and D-camphor-10-sulfonic acid). The suspension (400 µl) was then mixed with Milli-Q water and chloroform at a volumetric ratio of 5:2:5 and centrifuged at 9,100 × g for 10 min at 20°C. Subsequently, the aqueous layer (400 µl) was centrifugally filtered through a 5 kDa cut-off filter (cat. no. UFC3LCCNB-HMT; Human Metabolome Technologies, Inc.) to remove proteins. The filtrate was centrifugally concentrated and dissolved in 25 µl Milli-Q water containing the reference compounds (200 µM each of 3-aminopyrrolidine and trimesic acid) immediately prior to CE-TOFMS analysis [Agilent 1100 isocratic HPLC pump, Agilent G1603ACE-MS adapter kit, and Agilent G1607A CE-electrospray ionization (ESI)-MS sprayer; Agilent Technologies, Inc.] (13,14). The quantified data were standardized with the internal standards before being normalized to the DNA concentrations of each sample.

PDX tissues fixed with formalin were routinely processed into paraffin-embedded (FFPE) sections and stained with hematoxylin and eosin for histopathological evaluation using a transmission light microscope (BX51; Olympus Corporation).

Immunohistochemistry was performed using the FFPE sections and primary antibodies against β-catenin (cat. no. 610154; clone 14; BD Biosciences), phosphoenolpyruvate carboxykinase 2 (PCK2; (cat. no. 14892-1-AP; ProteinTech Group, Inc.), α-smooth muscle actin (αSMA; cat. no. M0851; clone 1A4; Agilent Technologies, Inc.) and very low density lipoprotein receptor (VLDLR; cat. no. ab203271; Abcam). Antigen retrieval was performed by autoclaving at 121°C for 10 min in citric acid buffer (pH 6.0) or proteinase K (FUJIFILM Wako Pure Chemical Corporation) treatment. For signal detection, Histofine MAX PO solution (cat. no. 424131; Nichirei Corporation) was applied, visualized with 3,3′-diaminobenzidine followed by counterstaining with hematoxylin. Negative controls without primary antibody addition was set for each antigen. Quantitative imaging analysis of αSMA-positive areas was performed using the HALO Imaging Analysis Platform (001-WS-HALO; Indica Labs, Inc.).

For the immunoblotting of β-catenin and PCK2, protein samples (10 µg) extracted with RIPA Lysis kit (cat. no. WSE-7420; ATTO Corporation) from the PDX tissues were subjected to SDS-PAGE. Immunoblotting and detection of chemiluminescence signals (cat. no. WSE-7120; ATTO Corporation) were performed as previously reported (15). Quantitative analysis was performed using Science Lab. 2005, Multi Gauge Ver. 3.0 software (FUJIFILM Wako Pure Chemical Corporation). For the internal control, a primary antibody against β-actin (cat. no. A1978; clone AC-15; Sigma-Aldrich; Merck KGaA) was used.

Quantitative data from in vitro and in vivo experiments, including cell viability assay, hydrophilic metabolomics, RT-qPCR and immunoblotting analysis results were presented as the mean ± SD. Homogeneity of variance in data was checked by Bartlett's test. When the data were homogenous, differences among the control and treatment groups were analyzed by one-way analysis of variance (ANOVA) or two-way ANOVA followed by Tukey's test using Easy R (EZR ver. 4.1.2) (16). In the heterogenous cases, Kruskal-Wallis test followed by Bonferroni test was applied. Immunohistochemical grade data were tested using Fisher's exact test in EZR. P-values of <0.05 were considered significant.

In total, three types of organoids, each harboring the following gene mutational status, were prepared (Fig. 1): i) APC, KRAS and TP53 mutations (case 21); ii) APC and TP53 mutations (case 32); or iii) PIK3CA and TP53 mutations (case 28). After comparison with the baseline cell viability of organoids or CAF monoculture in the E7386 (0 nM) group, cell viability was found to vary among the cases dependent on their origin. Increased cell viability of organoids following co-culture with CAFs was observed in cases 21 (P<0.05) and 32 (P<0.01) but not in case 28, which was reported previously (9). By contrast, an inhibitory effect of E7386 at 100 nM on the organoids from cases 21 and 32 was observed, which was more potent in the co-cultured groups (P<0.01). However, these inhibitory effects were diminished in the single-cultured groups from case 21 at 100 nM E7386. In case 28, inhibitory effects of E7386 on the organoids were observed in the co-culture group at 300 nM (P<0.05). In terms of CAFs, potentiating effects of co-culturing with organoids on viability were observed in case 28 (P<0.01) but not in cases 21 and 32. However, the degree of this enhancement was smaller compared with that in the organoids. Inhibitory effects of E7386 on the CAFs could be found at 100 nM, which was more potent in the single-culture groups from all three cases (P<0.01). There were slight differences in the sensitivity of the CAFs to E7386 among the co-cultured groups.

The effects of E7386 on gene expression in the organoids and CAFs were next evaluated using DNA microarrays. In the organoids, 46 genes were found to be upregulated associated with E7386 concentration (correlation coefficient >0.7) 24 h after E7386 treatment (Table SII). GO term enrichment analysis of these upregulated genes revealed that genes associated with stimulation, such as glutathione-specific γ-glutamylcyclotransferase 1 (CHAC1) and growth factor receptor bound protein 10 (GRB10), in addition to those associated with trans-sulfuration and one-carbon metabolism, including phosphoserine aminotransferase 1 (PSAT1), cystathionine γ-lyase and phosphoglycerate dehydrogenase (PHGDH), were amongst the most significantly enriched genes (Fig. S1A). In addition, UL16-binding protein (ULBP)1, which enhances susceptibility to natural killer (NK)-cell-mediated lysis of cancer cells (17) and PCK2, which regulates metabolic adaptation and enables glucose-independent tumor growth (18), were included amongst the upregulated genes. A number of the upregulated genes associated with E7386 concentrations were also categorized as ≥ two increased genes with significance (P<0.05) at 100 nM E7386, but associations between the gene expression changes and inhibition of the viability of CRC organoids by E7386 were uncertain (Figs. S2A and S3A). Among the genes that were significantly decreased by two-fold at 100 nM E7386, the cancer-related genes were not enriched (data not shown). RT-qPCR analysis was subsequently performed for CHAC1, ULBP1 and PCK2 to verify the results from the DNA microarray analysis. Elevated expression (P<0.01) of the three genes was observed in the single-culture organoids and after they were co-cultured with CAFs in the three cases (Fig. 2A). As a transport of small molecule-related genes (Fig. S1A), VLDLR expression was also found to be upregulated in association with E7386 concentrations according to results from DNA microarrays, which was confirmed by RT-qPCR in both single-culture organoids and co-cultures consisting of organoids and CAFs from case 32 but cases 21 and 28 (data not shown). The expression of 17 genes were downregulated associated with E7386 concentrations according DNA microarrays in the organoids following E7386 treatment (Table SIII). The results from the GO term enrichment analysis are shown in Fig. S1B. The downregulated genes included GRPEL1, histone deacetylase 2 (HDAC2) and farnesyl diphosphate synthase (FDPS). For the expression of FDPS and HDAC2, RT-qPCR analysis was performed to verify the results from the DNA microarray. No clear changes in FDPS expression could be found, whereas a trend toward a reduction in HDAC2 expression was observed in the organoids co-cultured with CAFs from cases 32 and 28 (Fig. 2B). For HDAC and FDPS expression, suggestive relationships with the Wnt/β-catenin signaling pathway were introduced (19,20).

In the corresponding CAFs for each organoid, 52 upregulated and 21 downregulated genes related to E7386 concentrations (correlation coefficient >0.9 and <-0.9, respectively) were found 24 h after the addition of E7386 (Tables SIV and SV). After they were evaluated at correlation coefficients of >0.7 and <-0.7, 458 upregulated and 431 downregulated genes were found (data not shown). GO term enrichment analysis of the upregulated genes revealed that those associated with cellular amino acid metabolism, including asparagine synthetase (glutamine-hydrolyzing; ASNS), PSAT1, PHGDH and glycyl-tRNA synthetase 1 (GARS), were amongst the most significantly enriched genes (Fig. S1C). In addition, those associated with gluconeogenesis, such as PCK2, activating transcription factor 4 and sestrin 2, were upregulated in the E7386-treated CAFs. Various upregulated genes related to E7386 concentrations were also categorized as ≥ two increased genes with a significance of P<0.05 at 100 nM E7386, but associations between the gene expression changes and inhibition of the viability of CAFs by E7386 were uncertain (Figs. S2B and S3B). Among these genes associated with glucose and amino acid metabolism, ASNS, PSAT1, PHGDH and PCK2 expression were also increased in the organoids. The genes downregulated associated with E7386 concentrations included those associated with the resolution of sister chromatid cohesion, such as centromere protein O, spindle pole body component 25 and structural maintenance of chromosomes 1A (Fig. S1D). In total, ≥ two decreased genes with significance at 100 nM E7386 included those involved in mitotic cell cycle processes, such as cyclin E1 and MYBL2 and in DNA replication, such as ATPase family AAA domain-containing 5 and zinc finger GRF-type-containing 1 (Fig. S3B). Upregulation of PCK2 (P<0.01) expression was then confirmed by RT-qPCR in the single-culture CAFs and in the co-culture with organoids in each case (Fig. 3A). As a gene decreased by E7386, diaphanous related formin 1 (DIAPH1) was reported to be associated with the differentiation of hepatic stellate cells into tumor-promoting myofibroblasts (21). Therefore, to evaluate the alterations in the general activity of CAFs, the expression levels of ACTA2, which is a classical gene associated with an activated status of fibroblasts (8), were examined. They were found to be partially downregulated in cases 32 and 28 (Fig. 3B).

The quantified levels of metabolites produced by the pentose phosphate pathway, glycolysis, tricarboxylic acid (TCA) cycle, amino acid metabolism and their derivatives were obtained from the organoid samples in triplicate after treatment with E7386 for 24 h. The data are presented graphically on a large-scale metabolomic map (Fig. S4) and summarized in Fig. 4. The concentrations of glucose-6-phosphate (G6P) and glucose-1-phosphate (G1P) in the control (0 nM E7386) of case 21 were higher compared with those of cases 32 and 28 (P≤<0.05). In addition, G6P and G1P levels in case 21 were reduced by E7386 treatment (P≤<0.001), but those of cases 32 and 28 were unchanged, suggestive of E7386-induced glucose depletion in case 21. As a response to this effect, gluconeogenesis followed by the reduced levels of metabolites in the TCA cycle was observed in organoids from case 21. In terms of amino acids, the majority of essential amino acids, including histidine, isoleucine, leucine, methionine, phenylalanine, threonine, tryptophan and valine, in addition to a number of nonessential amino acids, including alanine and glutamate, were found to be at higher levels in the control conditions from case 21 compared with those from cases 32 and 28. Furthermore, those of case 21 were reduced by E7386 treatment (Fig. S4).

PDX using samples from case 21 was successfully established, whereas organoids from case 28, which were engrafted and passaged in the subcutis of NOG mice, were used as PDXs. For case 32, organoids implanted into the subcutis of NOG mice, followed by confirmation of their engraftment, were used. No obvious changes after the administration of E7386 could be found in the size of tumor tissues (Figs. S5 and S6) or histopathological characteristics of PDX from cases 21 and 28 (data not shown). Evaluation could not be performed in case 32. Immunohistochemistry staining for PCK2 revealed an increased tendency for cytoplasmic granular positivity in the carcinoma cells of case 21 in the E7386 group (Fig. 5A), which was confirmed by immunoblotting (Figs. S7A and S8A), but not in case 28 (Figs. S7D and S8A). No clear changes in PCK2 positivity could be observed between the E7386 and control groups from cases 28 and 32 (Table SVI). VLDLR staining could be detected on the apical surface of the carcinoma cells in the control and E7386 groups of cases 21 and 28. However, in case 32, they were strongly visible not only on the apical surface but also on the intercellular membranes in the E7386 group (P=0.014; Fig. 5B; Table SVI). In terms of β-catenin, nuclear positivity in the carcinoma cells of case 28 tended to decrease in the E7386 group compared with that in the control group (Fig. 3C) but not in cases 21 and 32 (Table SVI). According to β-catenin immunoblotting, no notable changes in expression could be observed between the E7386 and control groups in cases 21 and 28 (Figs. S7B, S7E and S8B). As a marker for the activation of CAFs, immunohistochemistry staining for αSMA was performed. αSMA expression in the CAFs surrounding the CRC foci in the E7386 group from cases 21 or 28 showed a tendency to decrease compared with that in the control group (Figs. 5D and S9; Table SVI). The visually judged data was not necessarily reflected to the quantitative data.