Downregulation of Immune Response- and External Stimulus-Related Genes in CRC Organoids, and Their Re-Expression by Co-Culturing with CAFs


 Organoids derived from epithelial tumors have recently been utilized as a preclinical model in basic and translational studies. This model is considered to reproduce the features of cell-cell contacted and differentiated original tumor cells, but not the tumor microenvironment. In this study, we established organoids and paired cancer-associated fibroblasts (CAFs) from surgical specimens of colorectal carcinomas (CRCs), and evaluated gene expression profiles in organoids with and without co-culture with CAFs to assess interactions between tumor cells and CAFs in tumor tissues. We found that the expression levels of several genes, which are highly expressed in original CRC tissues, were downregulated in organoids but re-expressed by co-culturing with CAFs. They comprised immune response- and external stimulus-related genes, e.g., REG family and dual oxidases (DUOXs), which are known to have malignant functions, e.g., cell-proliferation and/or reducing apoptosis of epithelia and drug resistance for anti-cancer drugs in tumors. In addition, the degree of re-production of REG1 and DUOX2 in the co-culture system varied depending on CAFs from each CRC case. In conclusion, the co-culture system of CRC organoids with paired CAFs was able to partially reproduce the tumor microenvironment.


Downregulation of Immune Response-and External
Stimulus-Related Genes in CRC Organoids, and Their Re-Expression by Co-Culturing with CAFs Introduction Two-dimensional (2D) cancer cell lines that can be easily and reproducibly maintained in vitro have been mainly used for cancer studies targeting the clari cation of molecular mechanisms of carcinogenesis or evaluation of anticancer drugs. As most cancer cell lines were cultured under over-nourished conditions, differing from the in vivo environment, and repeated passages, the characteristics of original tumor tissues, such as expression of stem cell markers and differentiation marker genes, and proliferation, invasion, and drug metabolic abilities are altered in cancer cell lines [1][2][3] . The low probability of success of clinical trials is partly due to the use of cancer cell lines in preclinical evaluations of drug candidates [3][4][5][6] .
To promote e cient drug discovery, organoid culture systems have been utilized recently 7,8 .
An organoid is a miniaturized and simpli ed organ produced in vitro using 3D culture systems that resembles a realistic micro-organ. They are derived from one or a few cells from a tissue, and demonstrate natural cell-cell communication and undergo self-organization in 3D culture 9,10 . As an ex vivo model, cancer tissue-derived organoids are considered suitable to analyze direct reactivities of carcinoma cells to growth factors/cytokines, miRNAs, or synthetic compounds; however, stroma cells, including broblasts, immune cells, and vascular cells, which interact with carcinoma cells in original cancer tissues, are absent in organoid systems. The interactions between cancer cells and cancer associated broblasts (CAFs) vary and are considered to be important for carcinogenesis [11][12][13][14][15][16][17] . Moreover, the epithelial mesenchymal transition (EMT), which enables tumor cells to acquire resistance to anticancer drugs, was reported to be facilitated by the presence of CAFs [18][19][20] . To date, there have been limited investigations using co-culture systems of 3D-cultured cancer cells with CAFs on several types of cancers, e.g., pancreatic ductal adenocarcinomas 21,22 , prostate adenocarcinomas 23 , and esophageal carcinomas 24 , and interactions between cancer cells and CAFs were demonstrated to affect cancer aggressiveness and resistance to anti-cancer drugs.
The purpose of the present study was to clarify if the co-culture system of CRC organoids with CAFs reproduces the microenvironment between tumor cells and CAFs observed in the original cancer tissues.
First, comprehensive gene expression analyses between original tumor tissues and organoids were performed to characterize the features of the organoid culture system. Second, to identify gene expression changes in organoids induced by co-culturing with CAFs, paired CAFs from each case were established, and gene expression pro les between organoids with or without associated CAFs were compared. The present comparative screening of gene expression pro les of clinical tissue-derived CRC organoids with or without corresponding patient tissue-derived CAFs is highly important because the major molecules involved in interactions between tumor cells and CAFs have not been comprehensively evaluated in natural cell-cell communication-constructed and self-organizing in vitro systems.

Results
Baseline characteristics of CRC Basic patient and pathological data in the 45 CRC cases used for establishment of organoids and broblasts are summarized in Supplementary Table S1. Of the 45 cancer patients, 60% were males and 40% were females. This male per female ratio is comparable with Japanese Cancer Statistics (Cancer Registry and Statistics. Cancer Information Service, National Cancer Center, Japan (Monitoring of Cancer Incidence in Japan (MCIJ)). Average ages were 62 years old, ranging from 36 years old to 88 years old. The proportions of stage I, stage II, stage III, and stage IV cancers were 7%, 47%, 31%, and 15%, respectively, and 85% were found in the left colon (D, S, Rs, Ra, and Rb) and 15% in the right colon (T, A, and C). The detailed baseline characteristics of donor patients are summarized in Supplementary Table   S1 and Supplementary Fig. S1.

Mutation analysis by NCC Oncopanel
Mutations in original CRC tissues and established organoids were analyzed by targeted sequencing using the NCC Oncopanel test. The prevalence rates of APC, TP53, and KRAS mutations in original tissues were 91.1%, 80.0%, and 35.6%, respectively (Fig. 2). The frequently mutated genes including APC, TP53, and KRAS were highly consistent between our cohort and TCGA (The Cancer Genome Atlas) database 25 . Most of the mutations found in the CRC tissues were maintained in their organoids with higher mutation allele frequencies. However, there are some exceptions, suggesting a re ection of intratumor heterogeneity. Loss of the minor mutations in organoids was considered to partly be attributed to clonal evolution of cancer cells during the organoid culture (Table 1).
Then, we performed Sanger sequencing to con rm that the established broblasts did not harbor the mutations found in cancer tissues. The TP53 mutations of the cases #21, #28, and #32 were detected in their organoids but not in their broblasts ( Supplementary Fig. S2). Other mutations such as those of DDR2 in #21, ESR1 in #28, and APC in #32, were also not detected in the corresponding broblasts (data not shown).

Gene expression analysis by DNA microarray
Gene expression analyses by DNA microarray were carried out using original CRC specimens and organoids. Data sets for original tumor tissues and organoids from 5 cases were successfully adjusted and validated for principal components analysis (PCA). The PC1-axis demonstrated a different spatial distribution between the original tumors and organoids. The two groups had signi cant differences in gene expression distribution, indicating two transcriptionally distant populations (Fig. 3a). There were two typical gene expression pro les of PC1 genes: One was the gene groups highly expressed in organoids and the other was gene groups whose expression in organoids was lower than that in the original tumors. In total, 586 probes exhibited a similar gene expression pro le with the latter group of PC1 genes The cell proliferation ability signi cantly varied among CRC organoid lines (Fig. 4a). CAFs were also established from the same CRC specimens in some cases. To address the effects of cell-cell interaction between CRC organoids and corresponding CAFs, we developed a novel co-culture method using a chamber system. By co-culturing organoids with CAFs, cell viability of organoids increased by 1.2 to 1.5fold compared with corresponding single-culture organoids in three out of four cases (Fig. 4a). This suggested that the tumor microenvironment constructed by cell-cell communication between tumor cells and CAFs plays a role in the cell proliferative/anti-apoptotic abilities of tumor cells.
Identi cation of upregulated genes by co-culturing with CAFs To identify genes whose expression levels were different between organoids co-cultured with and without CAFs, we performed DNA microarray analysis. Volcano plot analysis demonstrated that the expression levels of 73 genes with 87 probes in organoids were reduced to less than half, whereas those of 177 genes with 238 probes were increased by more than double by co-culturing with CAFs (P < 0.05; Fig. 4b, Supplementary Tables S3 and S4). These upregulated genes included REG1A, REG1B, REG3A, REG3G, DMBT1, DUOXA2, DUOX2, SOCS3, and LOC340340, with more than 10-fold upregulation. Hierarchical clustering analysis con rmed that these CAF-induced genes were expressed at higher levels also in the original CRC tissues but not in single-culture organoids (Fig. 4c). To validate the microarray data obtained from the co-culture of CRC organoids and CAFs, two candidates for CAF-induced gene groups, REG1A, REG3A, DUOX2, and DUOX2A, were quanti ed by quantitative RT-PCR ( Fig. 4d-g). The levels of REG1A increased by more than 121 (#21), 480 (#28), and 21-fold (#32) when co-cultured with CAFs. The levels of REG3A increased by more than 26 (#21), 55 (#28), and 11-fold (#32). The levels of DUOX2 increased by more than 29 (#21), 3 (#28), and 33-fold (#32). The levels of DUOXA2 increased by more than 10 (#21), 3 (#28), and 15-fold (#32). This analysis con rmed that the microarray data is reliable, and the degree of induction of each REG family and dual oxidase gene by CAFs varied among cases.
Other than REG family and dual oxidase genes, microarray analysis suggested that expression of CEACAM6 and CEACAM7, members of the carcinoembryonic antigen-related cell adhesion molecule family, and MUC1, which inhibits the anti-tumor immune response, was upregulated by more than 2-fold (Supplementary Table S3).Although many cancer-related genes were induced by the co-culture with CAFs, the degree of their induction by CAFs varied among cases.
GO term enrichment analysis for upregulated genes in organoids by co-culturing with CAFs revealed that innate immune response, including "cell wall disruption in other organisms", "response to bacterium", "complement and coagulation cascades", and "interferon-gamma-mediated signaling pathway", were signi cantly enriched ( Supplementary Fig. S3a). These induced pathways by CAFs were abundantly expressed in CRC tissues but downregulated in single-culture CRC organoids ( Supplementary Fig. S3b).
CAFs derived from different cases exhibit case-speci c ability for REG1A induction in organoids Organoids derived from different cases exhibited varying degrees of induction of REG family and dual oxidase genes by co-culturing with paired CAFs, e.g., organoids from cases #21 and 32 had high reactions for dual oxidase genes, and those from case #28 had high reactions for REG family genes ( Fig.  4d-g). Next, combinations of organoids and CAFs derived from different cases were examined because CAFs consist of numerous cell types and their characteristics vary among lines 26,27 . Organoids of cases #28 and #32 co-cultured with CAFs from three different cases and corresponding normal mucosa-derived broblasts (NFs) exhibited almost identical induction patterns by co-culture with CAFs from each case and corresponding NFs ( Fig. 5a and b), suggesting the case-speci c ability of CAFs for the induction of REG1A in organoids. Thus, re-expressed genes in CRC organoids by co-culturing with CAFs may depend on the characteristics of each organoid and CAF line.

Discussion
We established 11 organoids from 45 CRC cases, and simultaneously prepared paired CAFs and normal broblasts from several of them. Interactions between cancer cells and CAFs using colon or lung cancer cell line-originated organoids and patient-derived CAFs were previously reported 28,29 . Other reports suggested that the molecular characteristics of CAFs vary among cases 30,31 . The newly established CRC model of organoids co-cultured with paired CAFs from the same case in the present study may provide novel insights into cell-cell communication between these cell types.
Target sequencing analyses of original tumor tissues revealed that mutation patterns varied among patients, and the most frequently mutated genes were APC, TP53, and KRAS, as previously reported 25 . In addition, most mutations found in the original tumor tissues were conserved in organoids 32 . The mutation rates in CRC organoids were higher than those in CRC tissues, with several exceptions (Table 1).
This may be because CRC organoids consist of only epithelial tumor cells, whereas CRC tissues include epithelial tumor cells and stromal cells, which do not have mutations. Indeed, CAFs did not carry the mutations found in corresponding CRC organoids ( Supplementary Fig. S2). The fold changes in the mutation ratios (early passage/late passage -up to passage 20) were almost identical (0.84-1.13) ( Table   1), as previously reported in ovarian cancer organoids 33 . This suggests that organoids from CRC passaged up to 20 times can be applied as an ex vivo model in basic and translational studies for the evaluation of anti-cancer drugs. Furthermore, cancer stem cells comprising CRC organoids and their proliferative ability were maintained in our culture conditions.
Next, to examine whether CRC organoids can reproduce the gene expression of the original tumor tissues, gene expression pro les were compared between the original tumor tissues and established CRC organoids using DNA microarray (Fig. 3). As a result, CRC organoids lacked gene expression from mesenchymal cell populations and blood cells, whereas each CRC organoid had a variable expression pro le for intestinal stem cell marker genes, e.g.
LGR5 and ORFM4, as observed in CRC tissues. This suggests that gene expression in CRC organoids is mostly conserved, but they lack cell-cell communication with the tumor microenvironment, including broblasts and immune cells.
In this study, a co-culture system using the CRC organoids and paired CAFs and/or normal broblasts from each case was prepared, and gene expression pro les for both CRC organoids with or without co-culture with CAFs were compared. As a result, we identi ed 177 genes, including REG family and DUOX family genes, which were markedly upregulated by more than 2-fold by co-culturing with CAFs in all three cases analyzed. Some of these upregulated genes were reported to have oncogenic functions. For example, REG family genes are known to be upregulated in several carcinomas compared with normal tissue [34][35][36][37] . REG family genes have cell-proliferative and anti-apoptotic functions 38 . REG3A is also known to act as an extracellular matrix (ECM)-targeted scavenger of reactive oxygen species (ROS) in a dose-dependent manner and to prevent ROS-induced mitochondrial damage due to acetaminophen overdose 39,40 . ROS is a key regulator for EMT, and is produced by NOX gene family and DUOX family genes 41 . Therefore, REG3A and DUOX2 gene expression is considered to be important for EMT. DUOX family genes also have cell-proliferative functions 41,42 . Moreover, DUOX2 was reported to have anticancer drug resistant activities through EMT 43 . GO term enrichment analysis for the above mentioned 177 upregulated genes revealed that co-culturing of CRC organoids with CAFs induced innate immune responses, suggesting that some oncogenic signal cascades induced by cell-cell interaction between organoids and CAFs were mediated by the signal pathways related to immune responses.
Other than immune response-and external stimulus-related genes, CEACAM6, a member of the carcinoembryonic antigen (CEA) family 44 , was also upregulated by more than 2-fold in CRC organoids by co-culturing with CAFs. CEA is normally produced in gastrointestinal tissue during fetal development, but its production stops before birth. Consequently, CEA is usually present at low levels in the blood of healthy adults (approximately 2-4 ng/mL) 45 . However, the serum levels increase in some types of cancer, suggesting its utility as a diagnostic marker and/or a drug e cacy marker in clinical tests. CEACAM6 was highly upregulated in colon cancer tissues and may therefore be a suitable candidate for a diagnostic marker of colorectal cancer 46 . CEACAM6 loss increases mitochondrial basal and maximal respiratory capacity. It also affects several hallmarks of pancreatic ductal adenocarcinoma (PDA), including brotic reactions, immune regulation, and energy metabolism, and was recently investigated as a novel therapeutic target in PDA 47 . Thus, there is a subset of genes expressed in CRC tumor cells in the original tissues, but not in the CRC organoids without co-culture with CAFs. These silenced genes in CRC organoids were induced by the co-culture with CAFs (Fig. 4c).The degrees of upregulation of the genes by co-culturing with CAFs varied among the CRC cases. We investigated whether these effects on the gene expression changes depend on the ability of CAFs or their compatibility with CRC organoids. Organoids from two CRC cases exhibited almost identical induction patterns by co-culture with CAFs from three different cases and paired normal mucosa-derived broblasts (Fig. 5). Therefore, the degrees of upregulation of several genes, e.g., REG1A, depend on the ability of CAFs from CRC. In addition, it should be noted that normal broblasts also induced gene expression, suggesting that some factors, possibly growth factors/cytokines secreted by CRC organoids or miRNA delivered by extracellular vesicles (EVs) secreted from CRC organoids, can induce gene expression to a similar degree as CAFs. To address which factors are essential for the induction of genes in CRC organoids, detailed analyses of the co-culture system of CRC organoids are needed. The upregulated genes mentioned above were related to malignancy of tumor cells, and the malignancy of cancer may partly depend on the characteristics of CAFs and/or original normal broblasts. Indeed, cell viability of organoids increased by 1.2 to 1.5-fold compared with corresponding single-culture organoids in three of four cases by co-culturing with CAFs (Fig. 4a); however, those in one case (#28) were not altered. We compared the e ciency of induction of the REG1A gene in organoids by co-culture with CAFs from three cases and it was the lowest by #28 CAFs. In conclusion, 1) gene mutation patterns, fold changes in the mutation allele frequencies, and expression pro les for intestinal stem cell marker genes were nearly similarly maintained in CRC organoids as in the original tumor tissues, suggesting that CRC organoids can be applied as an ex vivo model in basic and translational studies. 2) Expression levels of several genes, which are highly expressed in original CRC tissues, were downregulated in organoids but re-expressed by co-culturing with CAFs. They comprised immune response-and external stimulus-related genes, e.g., REG family and dual oxidases (DUOXs). In addition, gene induction by co-culturing with CAFs varied depending on the CAFs from each CRC case, suggesting that the present co-culture system of CRC organoids with paired CAFs partially reproduces the tumor-microenvironment.

Materials And Methods
Tissue sampling A total of 45 surgical specimens of colorectal carcinomas (CRCs) and adjacent normal mucosa were obtained between February 2016 and February 2018 at National Cancer Center Hospital. All experimental methods were carried out in accordance with relevant guidelines and regulations. The use of patients' surgical specimens in this study was approved by the ethics committee of the National Cancer Center, Tokyo, Japan (2015-108), and written informed consent was obtained from all patients. After sampling for pathological evaluation, they were stored on ice, and each sample was dissected into approximately 2-5-mm cubes and used for culture of organoids and broblasts. Remaining fragments were simultaneously frozen in liquid nitrogen and stored at −80˚C for isolation of DNA and RNA, or xed with 10% buffered formalin for preparation of tissue sections for morphological identi cation of the organoidand broblast-originated tumor tissues. The fragments for RNA isolation were stored in RNAlater solution (ThermoFisher Scienti c, Tokyo, Japan) at 4°C overnight before storage.

Organoid culture
The protocol employed for organoid culture was a modi ed version of those previously reported 48,49 .
The fragments of CRC tissues were washed with cold HBSS(-), minced with scissors, and washed again. In a 12-well plate, 65 μl of Matrigel (Corning, Bedford, MA, USA) /well was polymerized for 15 min at 37°C and a 650-μl cell suspension /well x 3 was seeded and incubated in a 37°C humidi ed CO 2 incubator.
After 24 hours, the supernatants were removed and 85 μl/well of Matrigel was overlaid. After Matrigel polymerization, the basal culture media containing different factor combinations [A:250 ng/ml of Rspondin 1 (Peprotech) + 20 ng/ml of Wnt-3a (Peprotech) +25 μM SB202190, B: 25 μM SB202190, and C: none] was used to select the most e cient growth media during several passages. The organoids were dispersed by Accumax and passaged approximately once a week. Zell Shield was not used after several passages. When organoids with more than 10 passages (P10) survived after a freeze and thaw cycle, they were de ned as "successfully established".

Fibroblast culture
Two to three tissue fragments were washed three times with HBSS(-), placed in a 60-mm dish, and minced with scissors. Then, culture medium, RPMI-1640, containing L-Glutamine (FUJIFILM Wako) and 10% FBS, penicillin-streptomycin was added into the dish and cells were incubated at 37°C in a humidi ed 5% CO 2 incubator. After they reached 70% con uency, cells were passaged using TrypLE Express dissociation reagent (ThermoFisher Scienti c).
Co-culture of organoids with broblasts using a chamber system Organoids were cultured in the cell culture inserts with a porous membrane and broblasts were cultured in the carrier plates (Corning). The pore size of the insert was 1.0 μm to allow the free exchange of media but not cells to migrate through. One day before starting the co-culture, broblasts were dissociated into single cells using TrypLE Express and 1 x 10 4 cells were cultured in a 24-well plate with RPMI-1640 containing penicillin-streptomycin, Amphotericin B, and 10% FBS. For organoid culture, cell culture inserts were set on the 24-well companion plate and 20 μl of Matrigel was polymerized on the insert. Organoids were dissociated by Accumax and resuspended in optimized media for organoids described above. The cell suspension (1 x 10 4 cells / 200 μl) was seeded onto the Matrigel and 620 μl of media for organoids was added to the basal compartment. Fibroblasts and organoids were incubated at 37°C in a CO 2 incubator.
The next day, apical and basal media were removed from the plate containing organoids attached on the Matrigel, and organoids were covered with 20 μl of Matrigel and polymerized at 37°C for 15 min. The inserts containing organoids were transferred to the broblast-containing compartment plate after changing the medium from that for broblasts to 820 μl of optimized media for organoids. The co-culture plate was incubated at 37°C in a CO 2 incubator for 96 hours, and organoids and broblasts were collected separately for gene expression analysis.

Targeted sequencing analysis
Genomic DNA from 45 samples of CRC and organoids of cases #11, #21, #25, #28, #32, #33, and #44 were prepared using NucleoSpin Tissue kit (Takara Bio, Kusatsu, Japan) according to the manufacturer's protocol. Targeted sequencing analyses of those DNAs were performed using the NCC Oncopanel v4 test, which can analyze mutations of 114 genes and ampli cations and fusions of 12 genes 50 . Procedures for targeted sequencing and data analysis were previously described 50 .
TP53 mutations identi ed by the NCC Oncopanel test were con rmed by Sanger sequencing. The PCR products including mutations were ampli ed using speci c PCR primers. Primers used for A578G and G589A mutations of #21 and #32, respectively, were forward: 5'-GGAGGTCAAATAAGCAGCAGG-3' and reverse: 5'-GGCCTCTGATTCCTCACTGA-3'. Primers used for a 45_48TCAG deletion of #28 were forward: 5'-CCCAACCCTTGTCCTTACCA-3' and reverse: 5'-CAGTCAGATCCTAGCGTCGA-3'. The ampli ed PCR products were directly sequenced by Sanger sequencing using the following primers. The primer for #21 and #32 was 5'-ACAACCACCCTTAACCCCTC-3'. The primer for #28 was 5'-CCCAACCCTTGTCCTTACCA -3' Gene expression analysis using DNA microarray Total RNA was extracted from original tumor tissues and established organoids of cases #11, #25, #28, #32, and #44 using NucleoSpin RNA Plus (TaKaRa) according to the manufacturer's protocol. Total RNA was extracted from three replicates of co-cultured organoids and broblasts for #21, #28, and #32 using the RNeasy Micro kit (QIAGEN, Tokyo, Japan). The quality of the RNA samples was evaluated using an Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA) and highly quali ed RNA samples with a RIN score > 7.0 were selected for further DNA microarray analysis. Triplicate samples were used as an RNA cocktail. Cy3-labeled cRNA was hybridized on SurePrint G3 Human GE Microarray GE 8 x 60K Ver.3.0 (Agilent Technologies) according to the manufacturer's instructions. The Subio Platform, Subio Basic plug-in, and Subio Advanced plug-in (ver 1.24) (Subio, Kagoshima, Japan) were used for data analysis.
Cell viabilityThe viability of organoids was assessed after co-culture with broblasts for 96 hours using the chamber system. After aspirating culture medium, 100 μl of fresh culture medium was added. Then, Matrigel including organoids was scraped using a mini cell scraper, 100 μl of CellTiter-Glo 3D reagent was added (Promega, Tokyo, Japan), and organoids were disrupted by pipetting. Suspensions were transferred to 96-well assay plates and incubated at room temperature for 25 min. After incubation, the luminescence was measured using Synergy H1 (BioTek, Tokyo, Japan).

Gene Ontology (GO) enrichment analysis
To clarify the biological meaning of the key modules, the gene information was loaded into Metascape (http://metascape.org) for Gene Ontology (GO) enrichment analysis 52 . Terms with a P-value <0.01, a minimum count of 3, and an enrichment factor >1.5 were collected and grouped into clusters based on their membership similarities (Kappa scores >0.3).

Statistical analysis
The associations between clinical factors and the establishment of organoids were tested using Fisher's exact test in EZR 53 . For qRT-PCR and cell viability assay, results were presented as the mean ± s.d. Differences between groups were analyzed by the Student's-t-test or one-way ANOVA using EZR. Pvalues of < 0.05 were considered signi cant.