Novel and efficient method for culturing patient-derived gastric cancer stem cells

2.1 Human samples

Tumor samples were collected from GC patients who underwent primary resections at the Kyoto University Hospital (KUHP, Kyoto, Japan) and Medical Research Institute Kitano Hospital (Osaka, Japan) from January 2016 to November 2022. Their diagnosis was confirmed through histopathological examinations by board-certified diagnostic pathologists.

2.2 L-WRN conditioned medium

The L-WRN cells expressing mouse Wnt3a, R-spondin 3, and Noggin were obtained from Dr. Thaddeus S. Stappenbeck (Cleveland Clinic). Conditioned medium (CM) from L-WRN cells was prepared according to a previous protocol.²² Quality control testing of L-WRN CM was conducted according to the validation procedures and guidelines reported previously.²⁶ A commercial L-WRN CM was purchased from Sigma-Aldrich.

2.3 Spheroid culture of human gastric cancer and normal gastric epithelial cells

Immediately after surgical resection, the excised stomach by operation was opened longitudinally, wrapped in gauze moistened with saline to prevent drying, and kept at room temperature. Sample specimens were collected within 1 h after the resection operation. From each stomach, one to four tumor pieces (100–1000 mm³ each) and one to two pieces of normal mucosa (500–2000 mm³) were collected in separate 15-mL conical tubes containing 5–10 mL ice-cold washing medium (Table S1). Sample tubes were kept on ice during transportation to the laboratory, and the isolation of epithelial cells and preparation of stem cell culture were performed within 6 h after sample collection (i.e., 7 h after the resection operation) according to a step-by-step protocol.²² Specifically, the specimen pieces were minced in a 60-mm Petri dish, digested with 1–3 mL collagenase solution (Table S1) at 37°C for 40–60 min and dissociated by pipetting. Epithelial cell clusters were filtered through a 100-μm cell strainer (Corning), collected in a 1.5-mL tube, and resuspended in Matrigel (Corning) or collagen type-I matrix (Cellmatrix, Nitta Gelatin). The cell-matrix mixture was placed at the center of each well of the 12-well cell-culture plate (30 μL/well; TPP). After polymerization of matrix materials at 37°C, GC and normal gastric epithelial (NGE) cells were cultured with the cancer medium and eL-WRN medium (epidermal growth factor [EGF]-containing 50% L-WRN CM), respectively (Table S1). The medium was changed every other day. To passage, we collected Matrigel-embedded spheroids and treated them with 2.5 g/L trypsin solution (Nacalai Tesque) at 37°C for 2–5 min. Collagen type-I–embedded spheroids were treated with collagenase solution at 37°C for 30 min, followed by trypsinization. Spheroids were dissociated into small cell aggregates by pipetting, and they were resuspended in Matrigel or collagen type-I. Dilution (based on the volume of matrix materials) was adjusted to one to six times depending on the growth rate and spheroid density. It should be noted that too much trypsinization and pipetting caused poor cell survival when spheroids grew poorly in early passages. The spheroid culture was considered successful when spheroids were expanded to 12 wells of a 12-well cell-culture plate.

2.4 Growth monitoring in spheroid culture using a cell imager

To monitor cell growth, we resuspended trypsinized spheroids in Matrigel or collagen type-I at a density of approximately 150 cell aggregates/μL. Subsequently, 3 μL cell-matrix mixture was distributed in each well of the 96-well cell-culture plate (TPP). After polymerization of matrix materials, cells were cultured in 100 μL of media. High-resolution cell images were obtained using a cell imager (Cell³iMager duos, SCREEN) every 3–4 days (Figure S1A). The area of each spheroid in each well was outlined using image processing software (Figure S1B). The volume of each spheroid was estimated using the following formula: spheroid volume (μm³) = 4/3 × {[spheroid area (μm²)]³/π}^1/2. The cell growth rate for each well was estimated as the proportion of total spheroid volume to that on initial measurement, and the growth effect index (GEI) was defined as the relative growth rate of an experimental group to that of its control group. At least three independent experiments were performed for each analysis.

2.5 Mutational analysis

The exonic regions of 409 cancer-related genes in GC-SC spheroids were sequenced using the Ion AmpliSeq Comprehensive Cancer Panel (Thermo Fisher), and the sequence alignment to the reference genome (hg19) and variant calling were performed at Macrogen Japan. We omitted the analyses of the primary tumors because we and others had shown homogeneity of driver-gene mutations in cancer and their stability during ex vivo culture.^{14, 27, 28} Detection of cancer-specific mutations was performed as we described previously with modifications.²⁷ Specifically, polymorphic alleles were removed from the called variants using the VCFtools program (V.0.1.13)²⁹ by referring to the GEM Japan Whole Genome Aggregation (GEM-J WGA) panel (https://togovar.biosciencedbc.jp/doc/datasets/gem_j_wga) or the profiles of NGE-SC spheroids from the same patients (when available). The selected variants were annotated using the ANNOVAR program,³⁰ and polymorphic alleles were removed again by referring to the Human Genetic Variation Database.^{31, 32} Subsequently, they were filtered to select nonsynonymous, frameshift, and splicing mutations with more than 20% frequency. Variant calls that appeared in more than two lines were eliminated as false-positive except for those identified in the COSMIC database. Other erroneous mutations were eliminated by surveying their coverage tracks on the Integrative Genomics Viewer software (V.2.12.3, Broad Institute).

2.6 Mutation detection from RNA sequencing (RNA-seq) data

To save time and cost, we took advantage of our transcriptome analysis data that we completed in most GC-SC spheroid lines. Namely, mutations in cancer-related genes were determined by deducing from the sequences of the RNA-seq data. Spheroid RNA samples were purified using the NucleoSpin RNA II kit (Takara Bio), and RNA-seq analysis was performed at Macrogen Japan. The sequence alignment to the reference genome (hg19) and variant calling were performed using the Subio Platform software (V.1.24.5853, Subio). Cancer-specific mutations in the exonic regions of expressed genes were detected with the same workflow as for the cancer panel.

Additional Materials and Methods can be found in Appendix S1.

3.1 Improvement of patient-derived gastric cancer stem-cell spheroid culture efficiency using a revised protocol

To culture GC-SC spheroids, we conducted two sets of experiments in which we collected tumor samples from 71 patients of the first cohort, followed by those from 33 patients of the second. To the first cohort samples, we applied our conventional method originally developed for CRC-SC spheroids (Table 1). Namely, we cultured tumor epithelial cells in a serum-containing cancer medium (Table S1) to propagate GC-SC spheroids.²¹ In contrast, NGE-SC spheroids were also established from normal mucosa of the same patients using the eL-WRN medium (Table S1) containing mouse Wnt3a, R-spondin 3, and Noggin.^{21, 22} The success rate for establishing GC-SC spheroids was 25% (18 of 71 cases; 95% CI, 15%–35%), whereas that for NGE-SC spheroids was 94% (67 of 71 cases; 95% CI, 89%–100%; Table 1; Table S2). To improve the low success rate, we revised our protocol in the following three points and tested its feasibility with fresh GC samples of the second patient cohort (Table 1). First and foremost, we scrutinized the sample collection maneuver from cancer tissues. One of the major reasons for our earlier failure in GC-SC spheroid establishment by our conventional method was likely the paucity of cancer stem cells in the sampled tumor pieces as estimated histopathologically in a retrospective manner (47% with 95% CI, 30%–64%; in 16 of the 34 failed cases; Figure 1A). Another minor cause was fungal contamination (9% with 95% CI, 2%–17%; in five of the 53 failed cases), particularly, of those samples from necrotic lesions that tended to accumulate fungi and/or hyphae (Figure 1B). Therefore, we collected more tumor pieces from wider and deeper areas, avoiding necrotic lesions to harvest cancer stem cells more reproducibly (Figure 1C,D). Importantly, the revised protocol reviewed by board-certified diagnostic pathologists of the collaborating hospitals did not affect pathological and molecular pathological assessment. Second, we embedded tumor epithelial pieces of each patient in both Matrigel and collagen type-I separately. This was because the different extracellular matrix (ECM) was preferred in some minority cases. Third, we added 5% L-WRN CM (containing Wnt ligands) to the cancer medium to help propagate Wnt-responsive GC-SCs, as the extent of dependence of GC-SC organoids on Wnt ligands has been variable.^{13, 33} Owing to these changes, we achieved a significantly higher success rate (88% with 95% CI, 77%–99%; 29 of 33 cases) as compared to that (25% with 95% CI, 15%–35%; 18 of 71 cases) with the first patient cohort (Table 1; Table S3). We failed in four of 33 cases because of heavy contamination with yeasts (two cases) or poor cell growth in early passages (two cases). Notably, five of 29 lines (17%) were established only when embedded in collagen type-I with a statistically significant difference (p = 0.008, Fisher's exact test), whereas three lines (10%) were only in Matrigel (Figure 2A). Regarding Wnt dependency, five GC-SC lines required L-WRN CM to maintain spheroid lines (Figure 2A). Our revised method also improved the culture efficiency in terms of the time needed for spheroid culture establishment, as the median time of the second cohort (21 days) was significantly shorter than that of the first cohort (33.5 days; Figure 2B).

Typically, GC cells formed spherical aggregates in either Matrigel or collagen type-I (Figure S2A), and they were highly proliferative in the cancer medium (Figure S2B). Their structures and expression of markers such as CDX2 and MUC2 recapitulated those in the epithelial components of their primary cancer tissues (Figure S2C). Consistent with a previous study,³³ culturing a Wnt-dependent spheroid line (HG22T) in the Wnt-free cancer medium accumulated signet-ring cell-like cells that were prominent in the primary tumor (Figure S2D). To assess the tumor-initiating activity in vivo, we injected GC-SC spheroids subcutaneously into immunodeficient mice, as we reported previously.³⁴ Three of the five GC-SC spheroid lines formed subcutaneous tumors in nude or NSG mice, and their epithelial structures were similar to those of the primary tumors (Figure S3A,B), indicating that most of our GC-SC spheroid lines contained abundant tumor-initiating cells. Genetic alterations of TP53 and APC were detected frequently in the first patient cohort (13 and five lines, respectively, of 18), whereas they were less frequent in the second cohort (10 and three lines, respectively, of 25), suggesting that the improved culture condition helped propagate niche factor-sensitive GC-SCs that did not carry these key driver mutations (Figure 2C; Tables S4 and S5). Based on the estimated amounts of mutational burden, we identified four hypermutated GC-SC spheroid lines in the first patient cohort (22%; four of 18 lines; Figure 2C; Figure S4A), which was confirmed for lack of mismatch repair proteins by immunohistochemistry (Figure S4B,C; Table S6).

Collectively, these results demonstrated that our revised method for GC-SC spheroids was more efficient than our previous one.

3.2 Collagen type-I stimulates the growth of some slow-growing gastric cancer stem-cell spheroids

A diffuse-type GC-SC spheroid line (HG18T) embedded in Matrigel grew very slowly in vitro compared with other lines in the first patient cohort. Diffuse-type GC cells often invade the stromal layer of gastric mucosa,⁴ suggesting that these cells have a higher affinity to collagen (e.g., collagen type-I) than Matrigel extracellular scaffold rich in laminin-1.³⁵ Therefore, we cultured HG18T and other spheroid lines separately in Matrigel and collagen type-I. Notably, HG18T spheroids preferentially proliferated in collagen type-I, whereas HG13T and HG15T in Matrigel. Other lines, HG6T, HG14T, and HG16T, showed little differences in growth between the two matrix materials without affecting the maintenance of spheroid lines because they more than quadrupled their cell volume in 6 days in either Matrigel or collagen type-I (Figure 3A,B). Thus, we decided to try both Matrigel and collagen type-I simultaneously but separately for primary culture of PD–GC-SCs, and empirically determine the matrix best suited for each GC-SC spheroid line.

3.3 Exogenous Wnt ligands stimulate the growth of some slow-growing gastric cancer stem-cell spheroids

Previous studies have shown that a subset of GC organoids is dependent on exogenous Wnt ligands such as Wnt and/or R-spondin for growth.^{13, 14} However, Wnt ligands cause predominant growth of NGE-SCs in primary culture, which necessitates another selection procedure to enrich GC-SCs.^{13, 14, 33} To resolve this problem, we hypothesized that a low concentration of L-WRN CM that contained Wnt ligands could stimulate the growth of Wnt-responsive GC-SC spheroids without affecting NGE-SCs. Before determining such a concentration of L-WRN CM, we titrated its activity to ensure the reproducibility of culture conditions. We determined mRNA expression levels of MKI67 (proliferation marker) and LGR5 (stem cell marker) in normal colonic epithelial SC spheroids cultured with eL-WRN media containing serially diluted L-WRN CM according to the previous guidelines for quality control testing.²⁶ As a result, we found that low concentrations of L-WRN CM (1%–10%) from two different sources (in-house and commercial media) stimulated MKI67 mRNA expression in a dose-dependent manner but failed to maintain LGR5 mRNA levels (Figure S5). Next, we conducted serial dilutions of in-house L-WRN CM with the cancer medium in the range of 0%–20% to titrate its effects on the growth of HG13T and HG18T, which showed the lowest growth rates among our GC-SC lines that we have established so far (Figure 3B). In both spheroid lines, 5%–10% of L-WRN CM supported the proliferation of GC-SC spheroids, whereas 5% CM of NGE-SC spheroids did not (Figure 4A,B; Figure S6A–C). Interestingly, 5% L-WRN CM stimulated the expression of the stem cell marker LGR5 in both HG13T and HG18T but not in NGE-SCs (Figure 4C). In contrast, L-WRN CM had smaller effects on the expression of the proliferation marker MKI67 in GC-SC lines than those in NGE-SCs (Figure 4C). These results suggested that supplementation with a low concentration (e.g., at 5%) of L-WRN CM should support self-renewal of Wnt-responsive GC-SCs without allowing that of NGE-SCs.