Although cancer stem cells (CSC) are thought to be responsible for tumor recurrence and resistance to chemotherapy, CSC-related research and drug development have been hampered by the limited supply of diverse, patient-derived CSC. Here, we present a functional polymer thin film (PTF) platform that promotes conversion of cancer cells to highly tumorigenic three-dimensional (3D) spheroids without the use of biochemical or genetic manipulations. Culturing various human cancer cells on the specific PTF, poly(2,4,6,8-tetravinyl-2,4,6,8-tetramethyl cyclotetrasiloxane) (pV4D4), gave rise to numerous multicellular tumor spheroids within 24 hours with high efficiency and reproducibility. Cancer cells in the resulting spheroids showed a significant increase in the expression of CSC-associated genes and acquired increased drug resistance compared with two-dimensional monolayer-cultured controls. These spheroids also exhibited enhanced xenograft tumor-forming ability and metastatic capacity in nude mice. By enabling the generation of tumorigenic spheroids from diverse cancer cells, the surface platform described here harbors the potential to contribute to CSC-related basic research and drug development.

Significance:

A new cell culture technology enables highly tumorigenic 3D spheroids to be easily generated from various cancer cell sources in the common laboratory.

Since the initial discovery of stem cell–like cancer cells in acute myeloid leukemia (1), increasing evidence has supported the presence of a minor population of cancer cells in bulk tumors that are mainly responsible for tumor recurrence and drug resistance (2). These cells, termed cancer stem cells (CSC) or tumor-initiating cells, share many of the common characteristics of normal stem cells (1), including self-renewal capacity (3), intrinsic drug resistance (4), and differentiation ability (3). Hence, CSCs have attracted considerable interest in the fields of cancer research and drug development (5). CSCs are usually isolated from patient-derived tumor tissues through standard cell-sorting processes based on characteristic surface markers (6) and cell isolation based on spheroid-forming ability (7). However, the supply of patient-derived CSCs is limited, creating a bottleneck for CSC research (1, 3, 8). As an alternative, attempts have been made to isolate CSCs from conventional cancer cell lines, but only a small subpopulation (<1%–2%) of cancer cells expresses CSC-related surface markers, making such approaches also impractical for obtaining sufficient amounts of CSCs (2). A recent report showed that soft fibrin gels with well-controlled stiffness can selectively promote the growth of tumorigenic cells from among a pool of cancer cells, providing a useful method for CSC isolation based on tumorigenic behavior rather than putative CSC surface markers (9).

There is currently substantial and growing interest in developing methods that facilitate the formation of cancer cell spheroids because these three-dimensional (3D) structures are thought to better mimic in vivo tumor environments than two-dimensional (2D) monolayer cultures (10). Such spheroids, which have been employed for drug screening and efficacy tests, have been generated using a variety of methods, including seeding cells on hydrophilic ultra-low-attachment (ULA) surfaces (11) or on agarose gels with a concave shape (U-bottom; ref. 12), or by inserting cells into holes of hanging-drop culture plates (13). However, the tumor spheroids generated from cancer cell lines using the conventional methods do not systematically result in CSC enrichment (7, 14). In this context, there is a need for a facile method or simple technique that enables generation of tumorigenic spheroids from conventional human cancer cells with high efficiency and reproducibility. A few recent reports have suggested the possibility of bidirectional conversion between nontumorigenic cancer cells and CSCs (15). In light of the possibility of such a bidirectional conversion, we hypothesized that it may be possible to transform cancer cells to tumorigenic CSC-like cells if appropriate stimuli (chemical or biological) are provided on a culture surface (16, 17). Here, we report a general platform that promotes the transformation of diverse cancer cells to tumorigenic CSC-like 3D spheroids by simply culturing the cells on a specific functionalized surface.

Formation of diverse polymer thin films on cell culture plates or cover glasses via the initiated chemical vapor deposition process

The process for preparing pV4D4 thin films is described as a typical example; other polymer thin films (PTF) are also deposited directly on tissue culture plates (TCP) using the same protocol, with slight modifications of deposition conditions, as needed (see Supplementary Methods). For vaporization of the monomer, V4D4 (99%; Gelest) and tert-butyl peroxide (TBPO, 98%; Aldrich) were heated to 70°C and 30°C, respectively. Vaporized V4D4 and TBPO were introduced into an initiated chemical vapor deposition (iCVD) chamber (Daeki Hi-Tech Co. Ltd.) at flow rates of 1.5 and 1 standard cm3/min (sccm), respectively. The substrate temperature was maintained at 40°C, the filament temperature was kept at 200°C, and the iCVD chamber pressure was set to 180 mTorr. The deposition rate of pV4D4 film was estimated to be 1.8 nm/min. The thickness of pV4D4 films was monitored in situ using a He-Ne laser (JDS Uniphase) interferometer system.

Human cancer cell lines

Human ovarian cancer cell lines (SKOV3 and OVCAR3), human breast cancer cell lines (MCF-7, T47D, and BT-474), human liver carcinoma cell lines (Hep3B and HepG2), human glioblastoma cell lines (U87MG and U251), human colon cancer cell lines (SW480, HT-29, HCT116, and Caco-2), human lung cancer cell lines (A549, NCI-H358, and NCI-H460), and human prostate cancer (22RV1), cervical cancer (HeLa), melanoma (A375), and gastric cancer (NCI-N87) cell lines were purchased from Korea Cell Line Bank. Cell lines were authenticated by standard short tandem repeat DNA typing methodology. All cancer cells were Mycoplasma-free, tested using an e-Myco Mycoplasma PCR Detection Kit (iNtRON Biotechnology).

Cell culture conditions

SKOV3, T47D, BT-474, SW480, HT29, 22RV1, A549, NCI-H358, NCI-N87, OVCAR3, NCI-H460, and HCT116 cell lines were cultured in RPMI-1640 medium (Gibco) supplemented with 10% (v/v) FBS (HyClone), 1% (v/v) penicillin/streptomycin (P/S; Gibco), and 25 mmol/L HEPES (Gibco). MCF-7, Hep3B, HeLa, U251, and A375 cell lines were cultured in DMEM (Gibco) supplemented with 10% (v/v) FBS (HyClone) and 1% (v/v) P/S (Gibco). HepG2, U87MG, and Caco-2 cell lines were cultured in Minimum Essential Media (MEM; Gibco) supplemented with 10% (v/v) FBS (HyClone) and 1% (v/v) P/S (Gibco). All cells were maintained at 37°C in a humidified 5% CO2 atmosphere. The cell lines were used within 10 to 12 passages after thawing.

Spheroid formation on pV4D4

Cancer cells (1 × 106) were seeded on pV4D4-coated plates and cultured in RPMI-1640, DMEM, or MEM, as appropriate, containing 10% (v/v) serum replacement (Gibco), 1% (v/v) P/S (Gibco), and l-glutamine at 37°C in a humidified 5% CO2 atmosphere. The specific cell culture medium used for each cancer cell line is described in detail in Supplementary Information. For optimum spheroid growth, culture medium was refreshed every 2 to 3 days.

Spheroid formation using conventional methods

Hanging-drop 96-well plates (3D Biomatrix), U-bottom 96-well plates (S-Bio), and ULA 6-well plates (Corning) were used for conventional spheroid formation. Cells were plated at the following densities: 1 × 104 cells/50 μL for hanging drop, 5 × 104 cells/250 μL for U-bottom, and 5 × 105 cells/2 mL for ULA. For optimum spheroid growth, culture medium was refreshed every 2 to 3 days.

Immunocytochemistry

SKOV3 spheroids were transferred from ULA and pV4D4 plates to a 1.5 mL tube and fixed by incubating with a 4% paraformaldehyde solution (Sigma) for 30 minutes at room temperature. Fixed spheroids were incubated with 0.25% (w/v) Triton X-100 (Sigma) in Dulbecco's phosphate-buffered saline (D-PBS) solution for 10 minutes at room temperature, washed with D-PBS, and incubated with blocking solution (3% BSA in D-PBS). For laminin staining, spheroids were incubated with rabbit anti-human laminin primary antibody (1:100, cat. no. 11575; Abcam) for 12 hours at 4°C. After washing with D-PBS, the resulting spheroids were stained with rhodamine red-X–conjugated anti-rabbit secondary antibody (1:500, cat. no. R6394; Invitrogen) for 1 hour at room temperature, and then with Hoechst 33342 for 10 minutes. For tenascin-C (TNC) staining, SKOV3 2D controls or SKOV3 spheroids were incubated with rabbit anti-human TNC primary antibody (20 μg/mL, cat. no. AB19011; Millipore) for 12 hours at 4°C. After washing with D-PBS, cells and spheroids were stained with FITC-conjugated anti-rabbit secondary antibody (1:500, cat. no. sc-2012; Santa Cruz Biotechnology) for 1 hour at room temperature, and then with Hoechst 33342 for 10 minutes. For β-catenin staining, SKOV3 2D controls and SKOV3-ssiCSCs were incubated with mouse anti-human β-catenin primary antibody (1:100, cat. no. 13–8400; Invitrogen) for 1 hour at room temperature. After washing with D-PBS, cells were stained with TRITC-conjugated anti-mouse secondary antibody (1:1,000, cat. no. ab6786; Abcam) for 1 hour at room temperature, and then with Hoechst 33342 for 10 minutes. All fluorescent images were visualized using a confocal laser-scanning microscope (LSM 780, Carl Zeiss).

Flow cytometry analysis

Monolayer-cultured 2D control cancer cells and the corresponding pV4D4-cultured ssiCSC spheroids or ULA-cultured SKOV3 spheroids were trypsinized and then individually dispersed in buffer (D-PBS containing 1% FBS). SKOV3, MCF-7, Hep3B, and SW480 cancer cells were stained with allophycocyanin-conjugated anti-CD133 (1:100; eBioScience), FITC-conjugated anti-CD44 (1:200; BD Biosciences), phycoerythrin-conjugated anti-CD90 (1:100, MACS; Miltenyi Biotec), and FITC-conjugated anti-CD133 (1:100; Miltenyi Biotec) primary antibodies, respectively, and analyzed using a flow cytometry system (BD Calibur and BD LSR Fortessa).

For side population assays, 2D control cancer cells and ssiCSCs were dissociated using trypsin and then stained with Hoechst 33342 (ThermoFisher Scientific) in DMEM containing 2% FBS and 10 mmol/L HEPES buffer for 90 minutes at 37°C. Cells were then washed with HBSS containing 2% FBS and analyzed using a flow cytometry system (BD LSR Fortessa). Flow cytometry data histograms and plots were analyzed using FlowJo software (Tree Star Inc.).

RNA extraction and mRNA sequencing

mRNA was extracted from SKOV3 spheroids cultured for 8 days on pV4D4-coated plates and 2D control SKOV3 cells using a Magnetic mRNA Isolation Kit (NEB) according to the manufacturer's protocol. A library was prepared from DNase-treated mRNA using a NEXTflex Rapid Directional mRNA-Seq Kit (BIOO), as described by the manufacturer. Each library was sequenced on a HiSeq2500 system using the single-end method (50-bp reads). The sequenced reads were aligned to the human genome (version: Hg19) using STAR aligner (v.2.4.0; ref. 18). The HOMER software algorithm (19) and DESeq R package were used to investigate differentially expressed genes (DEG). Heatmap and MA plots were visualized using the pheatmap function and plotMA function, respectively, of the R statistical programming language v.3.3.0. (http://www.r-project.org/).

Gene ontology analysis and gene set enrichment analysis

The Gene Ontology (GO) analysis for up- and downregulated genes in SKOV3-ssiCSCs was carried out using the ConsensusPathDB database (http://consensuspathdb.org/). The significance threshold was defined by a P value less than 0.01. The GO analysis was repeated on a published expression dataset (GSE62905). Gene set enrichment analysis (GSEA; ref. 20) was performed using GSEA software (version 3.0) with 1,000 phenotype permutations and default values for other parameters. Gene sets used in this study were selected from the MSigDB hallmark gene sets (http://software.broadinstitute.org/gsea/msigdb/collections.jsp).

Animals and xenograft tumor formation

Female BALB/c nude mice (6 weeks old) were obtained from Orient Bio Inc. and housed under pathogen-free conditions in the animal facility at Korea Advanced Institute of Science and Technology. Mice were assigned randomly to experimental groups. The experiments themselves were not randomized, and investigators were not blinded to allocation during experiments and outcome assessment unless each section specifically included blind assessments. All surgeries were performed under isoflurane anesthesia, and every effort was made to minimize suffering. All animal procedures were reviewed and approved (approval number: KA2014–21) by the Korea Advanced Institute of Science and Technology's Institutional Animal Care and Use Committee for compliance with ethical procedures and scientific care.

For preparation of the human ovarian cancer xenograft model, 2D-cultured control SKOV3 cells or SKOV3-ssiCSCs dissociated from the corresponding spheroids were mixed with 50% Matrigel (Corning) at different serial dilutions (106 to 102 cells) and then s.c. injected into 6-week-old female BALB/c nude mice. Tumor formation was monitored for up to 120 days, and the formation of tumors was recorded when tumor volumes reached approximately 50 mm3. For preparation of the human breast cancer xenograft model, different serially diluted numbers (107 to 102) of 2D control cells or ssiCSCs from MCF7-Luc cancer cells were subcutaneously inoculated into 6-week-old female BALB/c nude mice. β-Estradiol 17-valerate (2.5 μg; Sigma), dissolved in 50 μL of sesame oil (Sigma), was subcutaneously administered to BALB/c nude mice via the neck every 10 days. For the human glioma xenograft model, different serially diluted numbers (106 to 102) of 2D control U87MG cells, ULA-cultured U87MG spheroids, or pV4D4-cultured U87MG-ssiCSC cells were mixed with 50% Matrigel and s.c. injected into 6-week-old female BALB/c nude mice. Tumor formation from MCF7-Luc and U87MG cells was monitored up to 90 days, and the formation of tumors was recorded when tumor volumes reached approximately 50 mm3.

Statistical analysis

Data were expressed as means ± SD. Statistical analyses were conducted using unpaired Student t tests with GraphPad Prism software. A P value< 0.05 was considered statistically significant.

Data availability statement

RNA sequencing data are available at the National Center for Biotechnology Information Gene Expression Omnibus data repository with the accession code GSE106848.

A PTF enables formation of 3D tumor spheroids from diverse human cancer cells

To introduce various surface functionalities on cell culture plates, we constructed a library of PTFs on conventional TCPs from various monomers using an iCVD process (21) and examined the ability of each PTF to promote formation of tumor spheroids (Fig. 1A). The chemical structures of a subset of the tested PTFs are shown in Supplementary Fig. S1A–S1F. When cells of the SKOV3 human ovarian cancer cell line were cultured on various PTFs, only one PTF surface—poly(2,4,6,8-tetravinyl-2,4,6,8-tetramethyl cyclotetrasiloxane), designated pV4D4—supported the formation of numerous multicellular spheroids within 24 hours. In contrast, these same cells grown on other PTFs exhibited a well-adhered, spread-out morphology similar to that of cells grown on TCPs (Fig. 1B). Encouraged by these preliminary findings, we further examined whether the spheroid-promoting ability of pV4D4 could be extended to other cancer cell lines. As shown in Fig. 1C, most human cancer cell lines regardless of origin or type rapidly formed multicellular spheroids (∼50–300 μm in diameter) within 24 hours, with high efficiency and reproducibility. Although the morphology of each spheroid varied from “cluster-of-grapes” appearance to densely packed spheres (Supplementary Fig. S2), these observations are indicative of the versatility of the PTF platform. Unlike a conventional hydrophilic ULA surface (11), the pV4D4 PTF surface, as characterized by Fourier transform infrared spectroscopy and high-resolution X-ray photoelectron spectroscopy (Supplementary Fig. S3A–S3C and Supplementary Table S1), was relatively hydrophobic with a water contact angle of approximately 90° (Supplementary Fig. S4) and had a smooth surface with a roughness similar to that of conventional TCPs (Supplementary Fig. S5A). In addition, there was little difference in the surface hardness between a pV4D4 PTF (200 nm in thickness) and a conventional TCP (0.510 vs. 0.517 GPa; Supplementary Fig. S5B). Notably, variations in the thickness of pV4D4 PTFs, which ranged from 50 to 300 nm, had no effect on their spheroid-forming ability (Supplementary Fig. S6), suggesting that a certain surface functionality (chemical or biological stimulus) present on pV4D4 rather than a mechanical cue like surface stiffness might be responsible for spheroid formation.

Figure 1.

Formation and characterization of cancer cell spheroids generated on a pV4D4 thin film from diverse human cancer cell lines. A, Schematic illustration of tumorigenic spheroid formation on a specific PTF surface. Human cancer cell lines were seeded on various functional PTFs prepared by polymerization of the corresponding monomers on conventional cell culture dishes using an iCVD process. During culture on the PTF surface, cancer cells become CSC-like cells, giving rise to tumorigenic multicellular spheroids. B, Morphologies of SKOV3 cells (2 × 104 cells/cm2) cultured for 24 hours in RPMI-1640 medium containing 10% (v/v) serum replacement on a conventional TCP and on various functional PTFs. Only the pV4D4 surface supported spheroid formation. C, Morphologies of cancer cell spheroids generated from various human cancer cell lines (2 × 104 cells/cm2) after culturing for 24 hours on a pV4D4 PTF surface. Cancer cell lines (and their tissue of origin): SKOV3 (ovarian), MCF-7 (breast), T47D (breast), BT-474 (breast), Hep3B (liver), HepG2 (liver), U87MG (brain), SW480 (colon), HT-29 (colon), 22RV1 (prostate), HeLa (cervical), A549 (lung), NCI-H358 (lung), and NCI-N87 (gastric). D, Shapes and morphologies of SKOV3 spheroids prepared using three conventional spheroid-forming methods (hanging-drop, U-bottom, ULA surface) and a pV4D4 surface after culturing the cells for 24 hours. E, Representative images showing immunostaining for laminin expression in 8-day SKOV3 spheroids cultured on a ULA or pV4D4 surface. A primary rabbit anti-human laminin antibody and a rhodamine red-X–conjugated anti-rabbit secondary antibody were used for laminin staining. Hoechst 33342 was used for staining nuclei. F, Expression of ALDH1A1 mRNA in day-8 SKOV3 spheroids prepared by conventional methods (hanging-drop, U-bottom, ULA) and by culturing on a pV4D4 surface, as quantified by qRT-PCR. SKOV3 cells cultured as a monolayer on TCPs for 8 days were used as a control (n = 3 independent experiments; *, P < 0.05; #, P < 0.01; and **, P < 0.005). Scale bars, 100 μm (B and C) and 20 μm (D).

Figure 1.

Formation and characterization of cancer cell spheroids generated on a pV4D4 thin film from diverse human cancer cell lines. A, Schematic illustration of tumorigenic spheroid formation on a specific PTF surface. Human cancer cell lines were seeded on various functional PTFs prepared by polymerization of the corresponding monomers on conventional cell culture dishes using an iCVD process. During culture on the PTF surface, cancer cells become CSC-like cells, giving rise to tumorigenic multicellular spheroids. B, Morphologies of SKOV3 cells (2 × 104 cells/cm2) cultured for 24 hours in RPMI-1640 medium containing 10% (v/v) serum replacement on a conventional TCP and on various functional PTFs. Only the pV4D4 surface supported spheroid formation. C, Morphologies of cancer cell spheroids generated from various human cancer cell lines (2 × 104 cells/cm2) after culturing for 24 hours on a pV4D4 PTF surface. Cancer cell lines (and their tissue of origin): SKOV3 (ovarian), MCF-7 (breast), T47D (breast), BT-474 (breast), Hep3B (liver), HepG2 (liver), U87MG (brain), SW480 (colon), HT-29 (colon), 22RV1 (prostate), HeLa (cervical), A549 (lung), NCI-H358 (lung), and NCI-N87 (gastric). D, Shapes and morphologies of SKOV3 spheroids prepared using three conventional spheroid-forming methods (hanging-drop, U-bottom, ULA surface) and a pV4D4 surface after culturing the cells for 24 hours. E, Representative images showing immunostaining for laminin expression in 8-day SKOV3 spheroids cultured on a ULA or pV4D4 surface. A primary rabbit anti-human laminin antibody and a rhodamine red-X–conjugated anti-rabbit secondary antibody were used for laminin staining. Hoechst 33342 was used for staining nuclei. F, Expression of ALDH1A1 mRNA in day-8 SKOV3 spheroids prepared by conventional methods (hanging-drop, U-bottom, ULA) and by culturing on a pV4D4 surface, as quantified by qRT-PCR. SKOV3 cells cultured as a monolayer on TCPs for 8 days were used as a control (n = 3 independent experiments; *, P < 0.05; #, P < 0.01; and **, P < 0.005). Scale bars, 100 μm (B and C) and 20 μm (D).

Close modal

Interestingly, individual cancer cells initially adhered to the pV4D4 surface, but soon began to spontaneously interact with each other to form multicellular spheroids (Supplementary Videos S1 and S2). As shown in the selected representative snapshots for the spheroid-forming behaviors on pV4D4 surface at early time points (Supplementary Fig. S7), cancer cells on the surface are kept interacting with surface and move around while being attached until they meet and interact with neighboring cells to form mini-spheroids. These active cell-to-surface and subsequent cell-to-cell interactions on the pV4D4 surfaces were not observed on the ULA surface in which SKOV3 cells seemed to prefer direct cell-to-cell interactions to cell-to-surface interactions (Supplementary Video S3).

Tumor spheroids generated on pV4D4 PTFs are enriched for CSC-like cells

We next compared the features of pV4D4 PTF–cultured cancer cell spheroids at days 4 and 8 with those prepared using other conventional spheroid-forming methods. Whereas hanging-drop and U-bottom methods caused SKOV3 cancer cells to form single, large, aggregated spheroids, the ULA and pV4D4 surfaces produced multiple and much smaller spheroids; those cultured on pV4D4 were also more uniform and slightly smaller than those cultured on ULA (Fig. 1D). An immunocytochemical analysis revealed a stark difference between SKOV3 spheroids cultured for 8 days on a ULA surface versus those cultured on a pV4D4 surface, showing that considerable amounts of laminin, a major component of extracellular matrix (ECM; ref. 22), were present within pV4D4-cultured spheroids, whereas much less amount of the ECM protein was deposited only at the periphery of ULA-cultured spheroids (Fig. 1E). Considering that tumor cells in vivo are surrounded by and interact with densely packed ECMs in tumor microenvironment (23), the deposition of abundant ECM proteins inside ssiCSC-tumor spheroids is a unique structural feature mimicking tumor tissue in vivo. Importantly, ECM in the tumor microenvironment is suggested to play a pivotal role in the development of the drug resistance, self-renewal, and tumorigenic properties of tumor-resident CSCs (22, 24).

The abundant expression of ECM within spheroids only in cells cultured on pV4D4 prompted us to examine the expression of CSC-associated genes. qRT-PCR analyses revealed that, among spheroids formed from SKOV3 cells using different methods, only those grown on pV4D4 showed a dramatic increase in the expression of aldehyde dehydrogenase 1 family member A1 (ALDH1A1), a putative CSC marker (Fig. 1F; refs. 25–27). Aldefluor assay further revealed that SKOV3 spheroids cultured on pV4D4 for 8 days showed a considerable increase in the ALDH-positive cell population compared with 2D-cultured control cells (17.5% vs. 3.8%; Supplementary Fig. S8). Spheroids grown on the pV4D4 surface also showed a striking increase in the expression of the typical self-renewal genes, OCT3/4, SOX2, and NANOG, compared with 2D-cultured SKOV3 controls grown on TCPs (Supplementary Fig. S9). These results suggest that the cancer cells within spheroids acquired some of the essential features of “stemness” (28, 29). Wound-healing assays further showed that cancer cells dissociated from day-8 pV4D4-cultured SKOV3 spheroids were able to migrate and fill the lined gap much faster (Supplementary Fig. S10A), and Transwell-based invasion assays showed that these cells were also able to penetrate a gel matrix more efficiently (∼4-fold) than 2D-cultured control cells (Supplementary Fig. S10B), demonstrating significantly enhanced capacity for cell migration and invasion. In addition, when cancer cells dissociated from day-8 pV4D4-cultured SKOV3 spheroids were seeded on conventional TCPs under the tumor sphere-forming conditions, they started to form spheroids spontaneously, suggesting maintenance of the acquired CSC-like properties (Supplementary Fig. S11). Collectively, these findings suggest that the pV4D4 surface provides certain stimuli that activate and transform SKOV3 cancer cells, giving rise to tumor spheroids that are significantly enriched for CSC-like cells. We have designated these CSC-like cells, surface stimuli–induced cancer stem cells (ssiCSC).

ssiCSC spheroids exhibit robust drug resistance

To test the generalizability of the pV4D4 method, we prepared various ssiCSC spheroids from other cancer cell lines and assessed CSC-related characteristics. Four human cancer cell lines with different tissue origins—SKOV3, MCF-7 (human breast cancer), Hep3B (human liver cancer), and SW480 (human colon cancer)—were chosen for the generation and analysis of ssiCSCs. Previously reported specific surface markers were used to identify putative CSCs for each cell line: ALDH1A1 for SKOV3 (26), CD44 (cluster of differentiation 44) for MCF-7 (30, 31), CD90 for Hep3B (32), and LGR5 (leucine-rich repeat-containing G-protein-coupled receptor 5) for SW480 (33). CD133 was used as a common putative CSC marker for all cell lines (6, 9, 27). Expression of CSC marker genes in ssiCSC spheroids was assessed by qRT-PCR after culture on pV4D4 surface for 4 and 8 days, and compared with that of the corresponding 2D controls cultured on TCPs. Genes for each cell-type–specific CSC marker were significantly upregulated in the corresponding spheroids, and expression of the common marker CD133 was increased in all ssiCSC spheroids (Fig. 2A). Interestingly, gene expression levels increased with time in culture, suggesting the CSC-like characteristics become strengthened over time. RT-PCR analyses also showed far higher expression of various CSC-related genes in all ssiCSC spheroids compared with 2D-cultured control cancer cells (Supplementary Fig. S12). Next, we used flow cytometry to quantify the fraction of putative CSC marker–positive cancer cells in the resulting spheroids at day 8 in culture. Flow cytometry revealed roughly a 10-fold increase in the expression of cell-type–specific CSC-associated surface markers (expressed as gene counts) in SKOV3, Hep3B, and SW480 ssiCSC spheroids compared with the corresponding 2D-cultured controls, except for CD44 in MCF-7 cells, which was upregulated to a lesser degree (Fig. 2B). Furthermore, although both 2D-cultured control SKOV3 cells and ULA-cultured SKOV3-spheroids showed only 0.1% and 1.9% of CSC-like CD133+ cell population, respectively, pV4D4-cultured SKOV3-ssiCSC spheroids, as expected, resulted in a dramatic increase of approximately 28.6% in the cell population (Supplementary Fig. S13), indicating a distinct difference in the CSC-generation ability between pV4D4 and the conventional spheroid-forming surfaces.

Figure 2.

Characterization of the CSC-like properties of various ssiCSC spheroids. A, Expression of CSC-associated markers in SKOV3-, MCF-7-, Hep3B-, and SW480-ssiCSC spheroids cultured on a pV4D4 surface for 4 and 8 days, as quantified by qRT-PCR (n = 3 independent experiments). The expression of putative CSC–related markers was significantly increased for each type of ssiCSC compared with the corresponding 2D-cultured controls, and also gradually increased with time in culture. B, Flow cytometric analysis of the CSC-associated marker–positive cell fraction from day-8 ssiCSCs and their corresponding 2D-cultured controls. Putative CSC markers: CD133 for SKOV3, CD44 for MCF-7, CD90 for Hep3B, and CD133 for SW480. C, Representative flow cytometry plots for side-population (SP) discrimination assays using Hoechst 33342 in 2D control cells and ssiCSCs from day-8 spheroids of SKOV3, MCF-7, Hep3B, and SW480 cells. D, Drug resistance of ssiCSCs from day-8 SKOV3, MCF-7, Hep3B, and SW480 cell spheroids was assessed after treatment with different concentrations of Dox for 24 hours (n = 4–6 independent experiments; *, P < 0.05; #, P < 0.01; and **, P < 0.005 for ssiCSCs vs. 2D controls). E, Expression of drug-efflux ABC transporter–related genes in SKOV3-ssiCSCs and 2D controls from day-8 cultures, as quantified by qRT-PCR (n = 3 independent experiments; *, P < 0.05; #, P < 0.01; **, P < 0.005; n.s., not significant).

Figure 2.

Characterization of the CSC-like properties of various ssiCSC spheroids. A, Expression of CSC-associated markers in SKOV3-, MCF-7-, Hep3B-, and SW480-ssiCSC spheroids cultured on a pV4D4 surface for 4 and 8 days, as quantified by qRT-PCR (n = 3 independent experiments). The expression of putative CSC–related markers was significantly increased for each type of ssiCSC compared with the corresponding 2D-cultured controls, and also gradually increased with time in culture. B, Flow cytometric analysis of the CSC-associated marker–positive cell fraction from day-8 ssiCSCs and their corresponding 2D-cultured controls. Putative CSC markers: CD133 for SKOV3, CD44 for MCF-7, CD90 for Hep3B, and CD133 for SW480. C, Representative flow cytometry plots for side-population (SP) discrimination assays using Hoechst 33342 in 2D control cells and ssiCSCs from day-8 spheroids of SKOV3, MCF-7, Hep3B, and SW480 cells. D, Drug resistance of ssiCSCs from day-8 SKOV3, MCF-7, Hep3B, and SW480 cell spheroids was assessed after treatment with different concentrations of Dox for 24 hours (n = 4–6 independent experiments; *, P < 0.05; #, P < 0.01; and **, P < 0.005 for ssiCSCs vs. 2D controls). E, Expression of drug-efflux ABC transporter–related genes in SKOV3-ssiCSCs and 2D controls from day-8 cultures, as quantified by qRT-PCR (n = 3 independent experiments; *, P < 0.05; #, P < 0.01; **, P < 0.005; n.s., not significant).

Close modal

Another key feature of CSCs is that they possess intrinsic or acquired resistance against chemotherapeutics owing to their ability to extrude drugs (34, 35). We, therefore, assessed the drug-efflux capacity of individual cancer cells dissociated from day-8 ssiCSC spheroids using a Hoechst dye-based side-population assay. These assays, performed using flow cytometry, showed a significant increase in the fraction of drug-efflux–positive cells in all four types of ssiCSCs compared with 2D-cultured controls. Specifically, the efflux-positive fraction was increased from 0% to approximately 13.8% for SKOV3 cells, 0.59% to approximately 9.6% for MCF-7 cells, 0.58% to approximately 9.2% for Hep3B cells, and 0.1% to approximately 10% for SW480 cells (Fig. 2C). Next, we evaluated the drug resistance of ssiCSCs against doxorubicin (Dox), a widely used anticancer drug (36, 37). Each type of day-8 ssiCSC spheroids was dissociated into single cells, which were then cultured as 2D monolayers on a conventional TCP surface and treated with different concentrations of Dox for 24 hours. Cell viability measurements using a WST-1 assay showed that, compared with 2D controls, ssiCSCs were highly resistant to Dox, even at concentrations as high as 50 μmol/L (Fig. 2D). Remarkably, both SKOV3- and SW480-ssiCSCs were highly resistant to Dox, and the latter exhibited even higher cell viability than untreated control cancer cells. Moreover, SW480-ssiCSCs maintained their drug resistance even after two subcultures on a TCP surface (Supplementary Fig. S14), implying that the original cancer cells were transformed to CSC-like cells. In contrast, the dissociated SKOV3 cells from ULA-cultured spheroids were much more sensitive to Dox compared with SKOV3-ssiCSC spheroids, showing IC50 value of 0.79 and 12.23 μmol/L, respectively (Supplementary Fig. S15).

Drug-efflux ability is mediated by a family of ATP-binding cassette (ABC) proteins (35, 38, 39). Accordingly, we analyzed expression of a panel of major multidrug-resistant (MDR) genes—ABCB1, ABCB2, ABCB5, ABCC1, and ABCG2 (40)—in SKOV3-ssiCSCs by qRT-PCR. All five MDR-related genes were highly upregulated in ssiCSCs compared with 2D-cultured controls; these increases were particularly striking for ABCB1 and ABCB5 (Fig. 2E). This significant upregulation of MDR genes in ssiCSCs was well correlated with our side-population assay results (Fig. 2C) and tests of Dox resistance (Fig. 2D). Taken together, our molecular and functional analyses of four types of ssiCSC spheroids suggest that, upon exposure to certain stimuli present on pV4D4 surfaces, cancer cells can be transformed to CSC-like cells that exhibit strong expression of CSC-related genes and robust drug resistance.

SKOV3-ssiCSC spheroids express stem cell–related genes on a genome-wide scale

To examine whether ssiCSC spheroids have typical genetic signatures of CSCs, we performed high-throughput mRNA sequencing (RNA-seq) on 8-day SKOV3-ssiCSC spheroids and 2D-cultured SKOV3 controls on a genome-wide scale. Using an adjusted P value of 0.05, we identified 2,086 genes with at least 4-fold differential expression (Fig. 3A). Interestingly, a majority of significantly upregulated genes (1,547 of 2,086) were found to be cell–cell adhesion molecules and ECM-associated genes (Fig. 3B), an observation that may explain the abundant expression of laminin in SKOV3-ssiCSC spheroids (Fig. 1E). To analyze the expression pattern of ECM genes in more details, we generated a heatmap displaying normalized expression levels of the genes listed in Fig. 3B (n = 56; Fig. 3C). Among the ECM genes whose expression increased (i.e., collagen, integrin, matrix metalloproteinase, and adhesion molecules), genes encoding LAMA1 (laminin subunit alpha 1), a constituent of laminin, and Wnt3a, which promotes tumorigenicity and stemness through Wnt/β-catenin signaling (41), were also highly upregulated. We next examined the 10 most highly expressed genes encoding cell–cell adhesion molecules listed in Fig. 3B. Notably, CLDN2, which is not only a target gene of Wnt/β-catenin signaling, but also encodes a key factor in cancer cell migration and invasion (42, 43), showed a significant increase in expression level (Fig. 3D). Collectively, these findings suggest that ECM–cell and cell–cell interaction-associated signaling pathways, such as Wnt/β-catenin, may contribute to the de novo acquisition of tumorigenicity and stemness properties by cancer cells in ssiCSC spheroids.

Figure 3.

Genome-wide gene expression profiling in SKOV3-ssiCSCs. A, MA plot showing log2-fold change of gene expression between SKOV3-ssiCSCs and 2D control cells. Golden-yellow marks indicate genes exhibiting fold changes greater than 2, as determined by DESeq analysis (adjusted P value < 0.05). Transcripts with reads per kilobase per million mapped (RPKM) values less than 0.03 were removed. B, GO terms associated with biological processes of upregulated genes in SKOV3-ssiCSCs. The negative log10P value is plotted on the x axis. C, Heatmap of ECM genes listed in B. Expression levels were expressed as relative values (log2) normalized to control signals. Red, high expression; blue, low expression. D, Bar graph of the 10 genes exhibiting the greatest increase in expression among cell–cell adhesion molecules. Values represent log2-fold change relative to 2D-control cells; results are presented as mean of two biological replicate samples. E, GSEA of the transcriptome of SKOV3-ssiCSCs compared with that of 2D-controls cells. The enrichment score (ES; y axis) reflects the degree to which a gene set was upregulated in ssiCSCs. NES, normalized enrichment score; FDR, false discovery rate. F, GO terms associated with biological processes of downregulated genes in SKOV3-ssiCSCs. The negative log10P value is plotted on the x axis. G, A four-set Venn diagram showing the overlap between gene sets assigned to cell-cycle–related GO categories—mitotic cell cycle, mitotic nuclear division, organelle fission, and sister chromatid segregation—presented in F. H, Heatmap of intersecting genes (n = 49) in G. Expression levels were presented as log2-fold change values normalized to control cells.

Figure 3.

Genome-wide gene expression profiling in SKOV3-ssiCSCs. A, MA plot showing log2-fold change of gene expression between SKOV3-ssiCSCs and 2D control cells. Golden-yellow marks indicate genes exhibiting fold changes greater than 2, as determined by DESeq analysis (adjusted P value < 0.05). Transcripts with reads per kilobase per million mapped (RPKM) values less than 0.03 were removed. B, GO terms associated with biological processes of upregulated genes in SKOV3-ssiCSCs. The negative log10P value is plotted on the x axis. C, Heatmap of ECM genes listed in B. Expression levels were expressed as relative values (log2) normalized to control signals. Red, high expression; blue, low expression. D, Bar graph of the 10 genes exhibiting the greatest increase in expression among cell–cell adhesion molecules. Values represent log2-fold change relative to 2D-control cells; results are presented as mean of two biological replicate samples. E, GSEA of the transcriptome of SKOV3-ssiCSCs compared with that of 2D-controls cells. The enrichment score (ES; y axis) reflects the degree to which a gene set was upregulated in ssiCSCs. NES, normalized enrichment score; FDR, false discovery rate. F, GO terms associated with biological processes of downregulated genes in SKOV3-ssiCSCs. The negative log10P value is plotted on the x axis. G, A four-set Venn diagram showing the overlap between gene sets assigned to cell-cycle–related GO categories—mitotic cell cycle, mitotic nuclear division, organelle fission, and sister chromatid segregation—presented in F. H, Heatmap of intersecting genes (n = 49) in G. Expression levels were presented as log2-fold change values normalized to control cells.

Close modal

To assess representative features of gene expression that reflect actual biological behavior in SKOV3-ssiCSCs, we used GSEA (Fig. 3E). These ssiCSCs showed marked upregulation of gene sets related to inflammation, STAT3, and KRAS. Inflammation promotes tumor progression by facilitating angiogenesis, invasion, and metastasis (44–47). In addition, the STAT3 signaling pathway enforces the maintenance of CSCs and plays an important role in KRAS-induced tumorigenesis (48, 49). This analysis thus further supports the conclusion that SKOV3-ssiCSC spheroids possess the genetic traits of CSCs. Not surprisingly, we found that SKOV3-ssiCSCs exhibited considerable downregulation of mitotic cell-cycle–related genes (Fig. 3F). A GO analysis of downregulated genes (n = 539) revealed over-representation of multiple GO terms related to cell-cycle control, including mitotic cell cycle, mitotic nuclear division, organelle fission, and sister chromatid segregation. To investigate the semantic similarity of these GO terms, we compared transcripts assigned to each GO category (Fig. 3G). A Venn diagram identified 49 genes that were commonly associated with the four GO terms; a heatmap of these intersecting genes showed a substantial reduction in their expression (Fig. 3H). Intriguingly, cancer cells within ssiCSC spheroids seemed to be in a quiescent state, which is an intrinsic property of stem cells that contributes to drug resistance. Next, we performed RNA-seq for both ULA-cultured SKOV3 spheroids and CD133+ cell population isolated from 8-day pV4D4-cultured SKOV3-ssiCSC spheroids and compared the gene expression with that of 8-day SKOV3-ssiCSCs (Fig. 3). As shown in Supplementary Fig. S16A, with an adjusted P value of 0.05 and at least 4-fold differential expression, the ULA-cultured spheroids shared merely a part of genes with the pV4D4-cultured ssiCSC spheroids (134 of 1,547 genes in upregulated genes and 40 of 539 genes in downregulated genes relative to 2D control, respectively). In contrast, the RNA-seq data of the isolated CD133+ cancer cell population were substantially overlapped with those of the original ssiCSC spheroids before separation (583 of 1,547 genes in upregulated genes and 185 of 539 genes in downregulated genes, respectively). A heatmap displaying normalized expression levels of typical ECM genes (upregulation) and cell-cycle–related genes (downregulation) further revealed the distinct difference between ULA-cultured spheroids and the isolated CD133+ cancer cells (Supplementary Fig. S16B). This result suggests that the CD133+ CSC-like cells contribute largely to the overall gene expression pattern of ssiCSC spheroids that contain both CSC-like cells and nontransformed cancer cells. Furthermore, we identified five major classes of subpopulation cells in 8-day–cultured SKOV3-ssiCSC spheroids by performing single-cell RNA-seq (Supplementary Fig. S17A). The single-cell transcriptome data reflected distinct gene expression patterns between each class of cells, many of which are involved in cell cycle, hypoxia, and metabolic processes (Supplementary Fig. S17B). Taken together, these genome-wide gene expression analyses support the conclusion that ssiCSCs possess the general molecular signature of CSCs and pV4D4-based spheroid-forming method is distinctly different from the conventional spheroid-forming ULA surface.

ssiCSC spheroids are highly tumorigenic in vivo

We next examined the in vivo tumorigenic capacity of ssiCSCs. SKOV3-derived ssiCSC spheroids were dissociated into single cells, serially diluted (102 to 106) in Matrigel, and subcutaneously inoculated into BALB/c nude mice (Fig. 4A). Xenograft tumor formation from the spheroid-dissociated cells was monitored for 120 days and compared with that for 2D TCP-cultured SKOV3 controls (Table 1). We found that 2D controls formed no tumors (0/5 mice) at doses of 105 cells per mouse or less, and formed tumors at a frequency of only 50% (2/4) at a dose of 106 cells per mouse (Table 1). In contrast, much smaller doses of ssiCSC-derived cells were capable of forming tumors at much higher frequencies. Specifically, the resulting frequencies of tumor formation were approximately 60% (3/5) for 105 cells, approximately 80% (4/5) for 104 cells, and approximately 20% (1/5) for 103 cells (Table 1). Considering how difficult it is to get xenograft tumors from human ovarian cancer (SKOV3) cells to grow in athymic nude mice without using SCID mice, the demonstrated tumorigenicity of SKOV3-ssiCSCs in vivo is impressive.

Figure 4.

Tumor-forming and metastatic ability of human ovarian cancer cell–derived ssiCSCs. A, Illustration of the overall process of tumor formation and liver metastasis after subcutaneous inoculation of serially diluted (102 to 106 cells/mouse) 2D-cultured control or SKOV3-ssiCSC spheroid-derived cells into the dorsal area of BABL/c nude mice. Prior to inoculation, day-8 ssiCSCs were dissociated from the corresponding spheroids into single cells and mixed with Matrigel. B, Representative images of livers dissected from mice that received either 2D control SKOV3 cells (1 × 106 cells/mouse; n = 5) or SKOV3-ssiCSC spheroid-derived cells (1 × 105 cells/mouse; n = 5) when palpable tumors were detected (30 days). C, Representative hematoxylin and eosin–stained images of liver tissues shown in B. Yellow arrows in the ×4 magnification image indicate metastatic lesions in the liver, and the dotted line in the ×20 magnification image delineates the boundary between normal tissue and tumor metastasis. D, IHC analysis of metastatic tumors in the liver depicted in B showed intense staining for TNC (arrowheads) near the boundary of the tumor and the invasive front. Images were obtained at ×4 and ×20 magnification.

Figure 4.

Tumor-forming and metastatic ability of human ovarian cancer cell–derived ssiCSCs. A, Illustration of the overall process of tumor formation and liver metastasis after subcutaneous inoculation of serially diluted (102 to 106 cells/mouse) 2D-cultured control or SKOV3-ssiCSC spheroid-derived cells into the dorsal area of BABL/c nude mice. Prior to inoculation, day-8 ssiCSCs were dissociated from the corresponding spheroids into single cells and mixed with Matrigel. B, Representative images of livers dissected from mice that received either 2D control SKOV3 cells (1 × 106 cells/mouse; n = 5) or SKOV3-ssiCSC spheroid-derived cells (1 × 105 cells/mouse; n = 5) when palpable tumors were detected (30 days). C, Representative hematoxylin and eosin–stained images of liver tissues shown in B. Yellow arrows in the ×4 magnification image indicate metastatic lesions in the liver, and the dotted line in the ×20 magnification image delineates the boundary between normal tissue and tumor metastasis. D, IHC analysis of metastatic tumors in the liver depicted in B showed intense staining for TNC (arrowheads) near the boundary of the tumor and the invasive front. Images were obtained at ×4 and ×20 magnification.

Close modal
Table 1.

Tumor formation and metastasis of SKOV3 in BALB/c nude micea

Number of tumor formation (liver metastasis)/number of injected animalsb
SKOV31001,00010,000100,0001,000,000Tumor-initiating cell frequency (95% confidence interval)P value (vs. 2D control)
2D Control 0/5 (0/5) 0/5 (0/5) 0/5 (0/5) 0/5 (0/5) 2/4 (0/4) 1: 1.73 × 106 (1: 4.36 × 105–1: 6.87 × 106 
ssiCSC 0/5 (4/5) 1/5 (4/5) 4/5 (4/5) 3/5 (5/5)  1: 4.11 × 104 (1: 1.55 × 104–1: 1.09 × 1051.39 × 10−7 
Number of tumor formation (liver metastasis)/number of injected animalsb
SKOV31001,00010,000100,0001,000,000Tumor-initiating cell frequency (95% confidence interval)P value (vs. 2D control)
2D Control 0/5 (0/5) 0/5 (0/5) 0/5 (0/5) 0/5 (0/5) 2/4 (0/4) 1: 1.73 × 106 (1: 4.36 × 105–1: 6.87 × 106 
ssiCSC 0/5 (4/5) 1/5 (4/5) 4/5 (4/5) 3/5 (5/5)  1: 4.11 × 104 (1: 1.55 × 104–1: 1.09 × 1051.39 × 10−7 

aTumor formation and metastasis were monitored up to 120 days.

bAll cells were dissociated into single cells and counted with a hemocytometer before s.c. injection.

Notably, we found that livers from ssiCSC-inoculated mice were filled with metastatic nodules, a strikingly abnormal appearance compared with livers of mice inoculated with 2D SKOV3 controls, which appeared normal (Fig. 4B). Histologic analyses of abnormal livers showed numerous metastatic lesions throughout the tissue, with a clear delineation between normal and tumorous areas. As expected given their normal appearance, the livers of mice injected with 2D control cancer cells showed no evidence of metastasis (Fig. 4C). Notably, mice inoculated with as few as 102 SKOV3-ssiCSC–derived cells also showed a high frequency of liver metastasis (4/5 mice; Supplementary Fig. S18; Table 1), indicating the enormously enhanced metastatic ability and tumorigenicity of SKOV3-ssiCSCs. Moreover, an IHC examination of liver metastases for the expression of TNC, a major component of the cancer-specific ECM and an essential component of the metastatic niche (50), revealed significant localization of TNC around the tumor boundary region at the interface with normal tissue (Fig. 4D), suggesting that tumor nodules found in the liver resulted from the metastasis of subcutaneously inoculated SKOV3-ssiCSCs.

Next, we tested the tumorigenicity of other cancer cell line–derived ssiCSCs. ssiCSCs derived from luciferase-transfected MCF-7 (MCF7-Luc) cells and U87MG human glioblastoma cells showed significant increases in tumor-forming ability compared with their corresponding 2D-cultured control cells (Supplementary Tables S2 and S3). Whereas 2D-cultured MCF7-Luc cells formed no tumors, even at a dose of 106 cells per mouse, its corresponding ssiCSCs formed tumors at a dose of 105 cells per mouse with high frequency (4/5 mice; Supplementary Table S2). Similarly, tumors formed from U87MG-ssiCSCs at a frequency of approximately 60% (3/5), even at 104 cells per mouse; in contrast, no tumors formed from ULA surface–cultured U87MG spheroids (Supplementary Table S3), indicating a stark difference in tumorigenicity between ULA- and pV4D4-cultured spheroids. Collectively, these results suggest that the pV4D4-cultured ssiCSCs possess dramatically enhanced tumor-forming ability compared with the original cancer cells and are further applicable to the preparation of various human xenograft tumor models that are notoriously difficult to form in athymic nude mice.

Tumorigenicity of ssiCSC spheroids is associated with activation of Wnt/β-catenin signaling

To explore the cellular and molecular mechanisms responsible for the stem cell–like characteristics of ssiCSCs, we turned our attention to several key signaling pathways related to tumorigenicity and stemness of CSCs, namely Notch, Hedgehog, and Wnt/β-catenin (51). Given that our genome-wide gene expression study of SKOV3-ssiCSCs revealed activation of the Wnt/β-catenin signaling pathway (Fig. 3), we first examined the expression of Wnt target genes (n = 46). Figure 5A shows that the expression of 30 of 46 Wnt/β-catenin target genes increased more than 1.5-fold in SKOV3-ssiCSCs together with a marked reduction in the expression of Dickkopf-related protein 1 (DKK1), a key inhibitor of the Wnt signaling pathway. qRT-PCR analyses also confirmed a dramatic reduction in DKK1 mRNA expression in 1-, 4-, and 8-day–cultured SKOV3-ssiCSC spheroids (Fig. 5B), indicative of activation of Wnt/β-catenin signaling pathways at an early time point in spheroid formation. qRT-PCR further revealed that this reduction in DKK1 expression was directly associated with significant increases in the expression of the Wnt/β-catenin downstream target genes, AXIN2 (axis inhibition protein 2) and MMP2 (Fig. 5B). Moreover, although qRT-PCR showed no evidence of changes in the level of β-catenin mRNA in ssiCSC spheroids, Western blot analyses indicated a significant reduction in phosphorylated β-catenin protein (Fig. 5C). In addition, immunostaining revealed substantial translocation of β-catenin into the nucleus of ssiCSCs, confirming activation of Wnt/β-catenin signaling pathways; by contrast, 2D-cultured SKOV3 cells showed little nuclear localization of β-catenin (Fig. 5D). Next, we further examined the effect of DKK1 protein on spheroid-forming ability and acquisition of CSC-like signature. Although numerous small-sized spheroids were observed at 6 hours and larger spheroids were formed at 24 hours under the normal ssiCSC culture condition (no DKK1 addition), the presence of excess DKK1 protein supplied to the culture media hindered spheroid formation at the same time frames and rather, formed a strange ameba-like cell network (Supplementary Fig. S19A). Moreover, after culturing for additional 3 days, we carried out qRT-PCR to examine the expression of Wnt signaling–related genes. As shown in Supplementary Fig. S19B, the expression of Wnt3a, Wnt target genes AXIN2 and MMP-2, and a putative CSC marker ALDHA1 was significantly reduced. This result supports the proposed mechanism that DKK1-mediated activation of Wnt/β-catenin signaling pathways might be responsible for the conversion of cancer cells to tumorigenic CSC-like phenotypes by the pV4D4 surface.

Figure 5.

Activation of Wnt/β-catenin signaling pathways in SKOV3-ssiCSC spheroids. A, Heatmap of Wnt target genes (n = 46). Expression levels were expressed as relative values (log2) normalized to control signals. B, Expression of DKK1 in SKOV3-ssiCSCs (days 1, 4, and 8) and AXIN2, and MMP-2 mRNAs in SKOV3-ssiCSCs (days 4 and 8) was quantified by qRT-PCR (n = 3 independent experiments; *, P < 0.05; #, P < 0.01; **, P < 0.005). Primers used are listed in Supplementary Table S4. C, Western blot analysis of phosphorylated β-catenin and total β-catenin in 2D controls and SKOV3-ssiCSCs (days 4 and 8). GAPDH was used as an internal protein standard. Cells were incubated first with primary antibodies against phosphorylated β-catenin (rabbit), β-catenin (mouse), and GAPDH (rabbit) and then with horseradish peroxidase–conjugated IgG or horseradish peroxidase-conjugated anti-mouse IgG secondary antibodies to detect the amount of each protein. D, Confocal images showing localization of β-catenin in 2D controls and SKOV3-ssiCSCs, detected by immunocytochemistry. Mouse anti-human β-catenin primary antibodies and TRITC-conjugated anti-mouse secondary antibodies were used for β-catenin staining, and Hoechst 33342 was used to stain nuclei. Scale bar, 20 μm. E, Confocal images showing immunostaining for TNC expression in 2D control cells and SKOV3-ssiCSC spheroids. Primary rabbit anti-human TNC antibody and FITC-conjugated anti-rabbit secondary antibody were used for immunostaining, and Hoechst 33342 was used to stain nuclei. Scale bar, 100 μm.

Figure 5.

Activation of Wnt/β-catenin signaling pathways in SKOV3-ssiCSC spheroids. A, Heatmap of Wnt target genes (n = 46). Expression levels were expressed as relative values (log2) normalized to control signals. B, Expression of DKK1 in SKOV3-ssiCSCs (days 1, 4, and 8) and AXIN2, and MMP-2 mRNAs in SKOV3-ssiCSCs (days 4 and 8) was quantified by qRT-PCR (n = 3 independent experiments; *, P < 0.05; #, P < 0.01; **, P < 0.005). Primers used are listed in Supplementary Table S4. C, Western blot analysis of phosphorylated β-catenin and total β-catenin in 2D controls and SKOV3-ssiCSCs (days 4 and 8). GAPDH was used as an internal protein standard. Cells were incubated first with primary antibodies against phosphorylated β-catenin (rabbit), β-catenin (mouse), and GAPDH (rabbit) and then with horseradish peroxidase–conjugated IgG or horseradish peroxidase-conjugated anti-mouse IgG secondary antibodies to detect the amount of each protein. D, Confocal images showing localization of β-catenin in 2D controls and SKOV3-ssiCSCs, detected by immunocytochemistry. Mouse anti-human β-catenin primary antibodies and TRITC-conjugated anti-mouse secondary antibodies were used for β-catenin staining, and Hoechst 33342 was used to stain nuclei. Scale bar, 20 μm. E, Confocal images showing immunostaining for TNC expression in 2D control cells and SKOV3-ssiCSC spheroids. Primary rabbit anti-human TNC antibody and FITC-conjugated anti-rabbit secondary antibody were used for immunostaining, and Hoechst 33342 was used to stain nuclei. Scale bar, 100 μm.

Close modal

Next, we searched for upstream signals that might have caused the significant reduction in DKK1 in ssiCSC spheroids. Interestingly, it has been shown that TNC, which is abundantly present in liver metastases of SKOV3-ssiCSCs (Fig. 4D), activates Wnt/β-catenin signaling pathways by downregulating DKK1 (52). Thus, to explore a possible link between TNC and DKK1, we immunostained day-8 SKOV3-ssiCSC spheroids for TNC. As shown in Fig. 5E, TNC was abundantly present throughout the spheroids, suggesting possible downregulation of its target DKK1 and thus activation of Wnt/β-catenin signaling pathways. ssiCSCs obtained from MCF-7, Hep3B, and SW480 spheroids also showed considerable expression of TNC (Supplementary Fig. S20A), accompanied by a dramatic reduction in DKK1 gene expression (Supplementary Fig. S20B), suggesting that the same Wnt/β-catenin signaling pathways may be involved in the generation of other ssiCSCs. Collectively, these findings strongly suggest that TNC-DKK1–mediated activation of Wnt/β-catenin signaling pathways could be responsible for the conversion of cancer cells to tumorigenic CSC-like phenotypes by the pV4D4 surface. However, it remains unclear what specific stimuli (chemical or biological) present on the pV4D4 surface trigger the activation of Wnt/β-catenin signaling in cancer cells. Answering this important question will require further studies.

As tumor-repopulating cancer cells, CSCs have been of intense interest to oncologists, and considerable effort has been dedicated to understanding their properties and developing candidate drugs that target them. A robust and versatile platform method that enables facile production of CSC-like cells would go a long way toward expediting CSC-related research. In the present study, we successfully developed a platform method capable of generating tumorigenic CSC-like spheroids by simply culturing a variety of conventional cancer cells on a PTF surface (pV4D4). No additional biological growth factors, genetic transfections, or chemical treatments were involved in the transformation process. Nevertheless, the usefulness and broad applicability of ssiCSC spheroids depends on how closely they recapitulate the features of aggressive tumors in vivo or patient-derived CSC spheroids. As confirmed by IHC imaging, the key ECM proteins, laminin and TNC, were abundantly expressed throughout ssiCSC spheroids, but were not present on tumor spheroids produced by conventional methods (Figs. 1E and 5E). Given that interactions between cancer cells and the ECM are important for maintaining and promoting CSC characteristics (22), we speculate that the presence of ECM within ssiCSC spheroids may explain their enhanced tumorigenicity. In addition, ssiCSC spheroids were not only significantly enriched for tumorigenic CSC-like cells, they also contained a certain fraction of nontumorigenic cancer cells. Such unique features—the presence of ECM and enrichment of tumorigenic CSC-like cells—suggests that our ssiCSC spheroid may be a better in vitro mimic or model for highly aggressive and malignant tumors in vivo.

Another key consideration is whether human cancer cell line–derived ssiCSCs obtained as described here are phenotypically and genetically similar to patient tumor-derived CSCs. We were not able to directly compare the genetic traits of SKOV3-ssiCSCs (Fig. 3) with those of patient-derived CSCs owing to a lack of genome-wide gene expression and GO analyses of patients with ovarian cancer CSCs. Instead, a comparison of DEG-driven GO terms between SKOV3-ssiCSCs and patient-derived liver CSCs (53) provides convincing evidence that our cancer cell line–derived ssiCSCs share several important features of gene expression profiles, such as ECM-related genes, with the patient-derived CSCs (Supplementary Fig. S21A and S21B). Although such direct comparisons between two cancer cells with different origins may not be appropriate, the similarities in some key genetic traits of CSCs suggest that our cell line–derived ssiCSCs have potential for future use as a feasible model for cancer research and drug development.

In conclusion, the findings presented here clearly demonstrate that a pV4D4-based cell-culture platform enables the conversion of conventional cancer cells to highly tumorigenic CSC-like spheroids with high efficiency, reproducibility, and versatility. In vitro molecular and functional analyses showed that pV4D4-cultured cancer cell spheroids are substantially enriched for tumorigenic cells that show dramatically increased drug resistance against an anticancer drug compared with 2D-cultured controls. We also confirmed that the in vivo tumor-forming ability and metastatic propensity of pV4D4-cultured CSC-like spheroids is greatly enhanced; thus, this system could be used as a platform for preparation of hard-to-form human xenograft tumor models in nude mice. Taken together, our results suggest that by providing a facile method of generating CSC-like tumor spheroids from diverse cancer cells, the PTF platform described here will contribute to CSC-related basic research and drug development.

No potential conflicts of interest were disclosed.

Conception and design: M. Choi, S.J. Yu, Y. Choi, H.R. Lee, S.G. Im, S. Jon

Development of methodology: M. Choi, S.J. Yu, H.R. Lee, E. Lee, E. Lee, S. Kang, J. Baek, S.G. Im, S. Jon

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Choi, S.J. Yu, Y. Choi, H.R. Lee, E. Lee, T.G. Lee

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Choi, S.J. Yu, Y. Choi, E. Lee, J. Song, T.G. Lee, D. Lee, S.G. Im, S. Jon

Writing, review, and/or revision of the manuscript: M. Choi, S.J. Yu, Y. Choi, E. Lee, D. Lee, S.G. Im, S. Jon

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Choi, S.J. Yu, Y. Choi, Y. Lee, J. Song, J.Y. Kim

Study supervision: E. Lee, D. Lee, S.G. Im, S. Jon

Other [identify differences in the samples used in the concept setting by surface analysis (additional information not included in the article)]: J.G. Son

This work was supported by a grant from the Samsung Research Funding Center of Samsung Electronics (Project Number SRFC-MA1501-01).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Bonnet
D
,
Dick
JE
. 
Human acute myeloid leukemia is organized as a hierarchy that originates from a primitive hematopoietic cell
.
Nat Med
1997
;
3
:
730
37
.
2.
Hanahan
D
,
Weinberg Robert
A
. 
Hallmarks of cancer: the next generation
.
Cell
2011
;
144
:
646
74
.
3.
Hope
KJ
,
Jin
L
,
Dick
JE
. 
Acute myeloid leukemia originates from a hierarchy of leukemic stem cell classes that differ in self-renewal capacity
.
Nature Immunol
2004
;
5
:
738
43
.
4.
Dean
M
,
Fojo
T
,
Bates
S
. 
Tumour stem cells and drug resistance
.
Nat Rev Cancer
2005
;
5
:
275
84
.
5.
Gupta
PB
,
Onder
TT
,
Jiang
G
,
Tao
K
,
Kuperwasser
C
,
Weinberg
RA
, et al
Identification of selective inhibitors of cancer stem cells by high-throughput screening
.
Cell
2009
;
138
:
645
59
.
6.
Singh
SK
,
Clarke
ID
,
Terasaki
M
,
Bonn
VE
,
Hawkins
C
,
Squire
J
, et al
Identification of a cancer stem cell in human brain tumors
.
Cancer Res
2003
;
63
:
5821
8
.
7.
Ishiguro
T
,
Ohata
H
,
Sato
A
,
Yamawaki
K
,
Enomoto
T
,
Okamoto
K
. 
Tumor-derived spheroids: relevance to cancer stem cells and clinical applications
.
Cancer Sci
2017
;
108
:
283
9
.
8.
Kaiser
J
. 
The cancer stem cell gamble
.
Science
2015
;
347
:
226
.
9.
Liu
J
,
Tan
Y
,
Zhang
H
,
Zhang
Y
,
Xu
P
,
Chen
J
, et al
Soft fibrin gels promote selection and growth of tumorigenic cells
.
Nat Mater
2012
;
11
:
734
41
.
10.
Yamada
KM
,
Cukierman
E
. 
Modeling tissue morphogenesis and cancer in 3D
.
Cell
2007
;
130
:
601
10
.
11.
Thoma
CR
,
Zimmermann
M
,
Agarkova
I
,
Kelm
JM
,
Krek
W
. 
3D cell culture systems modeling tumor growth determinants in cancer target discovery
.
Adv Drug Deliv Rev
2014
;
69–70
:
29
41
.
12.
Rotem
A
,
Janzer
A
,
Izar
B
,
Ji
Z
,
Doench
JG
,
Garraway
LA
, et al
Alternative to the soft-agar assay that permits high-throughput drug and genetic screens for cellular transformation
.
Proc Natl Acad Sci U S A
2015
;
112
:
5708
13
.
13.
Timmins
NE
,
Nielsen
LK
. 
Generation of multicellular tumor spheroids by the hanging-drop method
.
In:
Hauser
H
,
Fussenegger
M
,
editors
.
Tissue engineering.
Totowa, NJ
:
Humana Press
; 
2007
. p.
141
51
.
14.
Calvet
CY
,
André
FM
,
Mir
LM
. 
The culture of cancer cell lines as tumorspheres does not systematically result in cancer stem cell enrichment
.
PLoS One
2014
;
9
:
e89644
.
15.
Chaffer
CL
,
Brueckmann
I
,
Scheel
C
,
Kaestli
AJ
,
Wiggins
PA
,
Rodrigues
LO
, et al
Normal and neoplastic nonstem cells can spontaneously convert to a stem-like state
.
Proc Natl Acad Sci U S A
2011
;
108
:
7950
55
.
16.
Crowder Spencer
W
,
Leonardo
V
,
Whittaker
T
,
Papathanasiou
P
,
Stevens Molly
M
. 
Material cues as potent regulators of epigenetics and stem cell function
.
Cell Stem Cell
2016
;
18
:
39
52
.
17.
Mei
Y
,
Saha
K
,
Bogatyrev
SR
,
Yang
J
,
Hook
AL
,
Kalcioglu
ZI
, et al
Combinatorial development of biomaterials for clonal growth of human pluripotent stem cells
.
Nat Mater
2010
;
9
:
768
78
.
18.
Dobin
A
,
Davis
CA
,
Schlesinger
F
,
Drenkow
J
,
Zaleski
C
,
Jha
S
, et al
STAR: ultrafast universal RNA-seq aligner
.
Bioinformatics
2013
;
29
:
15
21
.
19.
Heinz
S
,
Benner
C
,
Spann
N
,
Bertolino
E
,
Lin
YC
,
Laslo
P
, et al
Simple combinations of lineage-determining transcription factors prime cis-regulatory elements required for macrophage and B cell identities
.
Mol Cell
2010
;
38
:
576
89
.
20.
Subramanian
A
,
Tamayo
P
,
Mootha
VK
,
Mukherjee
S
,
Ebert
BL
,
Gillette
MA
, et al
Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles
.
Proc Natl Acad Sci U S A
2005
;
102
:
15545
50
.
21.
Alf
ME
,
Asatekin
A
,
Barr
MC
,
Baxamusa
SH
,
Chelawat
H
,
Ozaydin-Ince
G
, et al
Chemical vapor deposition of conformal, functional, and responsive polymer films
.
Adv Mater
2010
;
22
:
1993
7
.
22.
Lu
P
,
Weaver
VM
,
Werb
Z
. 
The extracellular matrix: a dynamic niche in cancer progression
.
J Cell Biol
2012
;
196
:
395
406
.
23.
Riegler
J
,
Labyed
Y
,
Rosenzweig
S
,
Javinal
V
,
Castiglioni
A
,
Dominguez
CX
, et al
Tumor elastography and its association with collagen and the tumor microenvironment
.
Clin Cancer Res
2018
;
24
:
4455
67
.
24.
Qin
Y
,
Rodin
S
,
Simonson
OE
,
Hollande
F
. 
Laminins and cancer stem cells: partners in crime?
Semin Cancer Biol
2016
;
45
:
3
12
.
25.
Ginestier
C
,
Hur
MH
,
Charafe-Jauffret
E
,
Monville
F
,
Dutcher
J
,
Brown
M
, et al
ALDH1 is a marker of normal and malignant human mammary stem cells and a predictor of poor clinical outcome
.
Cell Stem Cell
2007
;
1
:
555
67
.
26.
Landen
CN
,
Goodman
B
,
Katre
AA
,
Steg
AD
,
Nick
AM
,
Stone
RL
, et al
Targeting aldehyde dehydrogenase cancer stem cells in ovarian cancer
.
Mol Cancer Ther
2010
;
9
:
3186
99
.
27.
Visvader
JE
,
Lindeman
GJ
. 
Cancer stem cells in solid tumours: accumulating evidence and unresolved questions
.
Nat Rev Cancer
2008
;
8
:
755
68
.
28.
Chiou
S-H
,
Yu
C-C
,
Huang
C-Y
,
Lin
S-C
,
Liu
C-J
,
Tsai
T-H
, et al
Positive correlations of Oct-4 and Nanog in oral cancer stem-like cells and high-grade oral squamous cell carcinoma
.
Clin Cancer Res
2008
;
14
:
4085
.
29.
Wang
M-L
,
Chiou
S-H
,
Wu
C-W
. 
Targeting cancer stem cells: emerging role of Nanog transcription factor.
Onco Targets Ther
2013
;
6
:
1207
20
.
30.
Al-Hajj
M
,
Wicha
MS
,
Benito-Hernandez
A
,
Morrison
SJ
,
Clarke
MF
. 
Prospective identification of tumorigenic breast cancer cells
.
Proc Natl Acad Sci U S A
2003
;
100
:
3983
8
.
31.
Sun
Y
,
Kim
HS
,
Park
J
,
Li
M
,
Tian
L
,
Choi
Y
, et al
MRI of breast tumor initiating cells using the extra domain-B of fibronectin targeting nanoparticles
.
Theranostics
2014
;
4
:
845
57
.
32.
Yang
ZF
,
Ho
DW
,
Ng
MN
,
Lau
CK
,
Yu
WC
,
Ngai
P
, et al
Significance of CD90+ cancer stem cells in human liver cancer
.
Cancer Cell
2008
;
13
:
153
66
.
33.
Vermeulen
L
,
De Sousa
E
,
Melo
F
,
van der Heijden
M
,
Cameron
K
,
de Jong
JH
, et al
Wnt activity defines colon cancer stem cells and is regulated by the microenvironment
.
Nat Cell Biol
2010
;
12
:
468
76
.
34.
Zhou
S
,
Schuetz
JD
,
Bunting
KD
,
Colapietro
A-M
,
Sampath
J
,
Morris
JJ
, et al
The ABC transporter Bcrp1/ABCG2 is expressed in a wide variety of stem cells and is a molecular determinant of the side-population phenotype
.
Nat Med
2001
;
7
:
1028
34
.
35.
Gottesman
MM
,
Fojo
T
,
Bates
SE
. 
Multidrug resistance in cancer: role of ATP-dependent transporters
.
Nat Rev Cancer
2002
;
2
:
48
58
.
36.
Tacar
O
,
Sriamornsak
P
,
Dass
CR
. 
Doxorubicin: an update on anticancer molecular action, toxicity and novel drug delivery systems
.
J Pharm Pharmacol
2013
;
65
:
157
70
.
37.
Bagalkot
V
,
Farokhzad
OC
,
Langer
R
,
Jon
S
. 
An aptamer–doxorubicin physical conjugate as a novel targeted drug-delivery platform
.
Angew Chem Int Ed
2006
;
45
:
8149
52
.
38.
Chaudhary
PM
,
Roninson
IB
. 
Expression and activity of P-glycoprotein, a multidrug efflux pump, in human hematopoietic stem cells
.
Cell
1991
;
66
:
85
94
.
39.
Scotto
KW
. 
Transcriptional regulation of ABC drug transporters
.
Oncogene
2003
;
22
:
7496
11
.
40.
Dean
M
,
Allikmets
R
. 
Complete characterization of the human ABC gene family
.
J Bioenerg Biomembr
2001
;
33
:
475
79
.
41.
Ying
Q-L
,
Wray
J
,
Nichols
J
,
Batlle-Morera
L
,
Doble
B
,
Woodgett
J
, et al
The ground state of embryonic stem cell self-renewal
.
Nature
2008
;
453
:
519
23
.
42.
Dhawan
P
,
Ahmad
R
,
Chaturvedi
R
,
Smith
JJ
,
Midha
R
,
Mittal
MK
, et al
Claudin-2 expression increases tumorigenicity of colon cancer cells: role of epidermal growth factor receptor activation
.
Oncogene
2011
;
30
:
3234
47
.
43.
Mankertz
J
,
Hillenbrand
B
,
Tavalali
S
,
Huber
O
,
Fromm
M
,
Schulzke
J-D
. 
Functional crosstalk between Wnt signaling and Cdx-related transcriptional activation in the regulation of the claudin-2 promoter activity.
Biochem Biophys Res Commun
2004
;
314
:
1001
7
.
44.
DeNardo
DG
,
Andreu
P
,
Coussens
LM
. 
Interactions between lymphocytes and myeloid cells regulate pro- versus anti-tumor immunity
.
Cancer Metastasis Rev
2010
;
29
:
309
16
.
45.
Grivennikov
SI
,
Greten
FR
,
Karin
M
. 
Immunity, inflammation, and cancer
.
Cell
2010
;
140
:
883
99
.
46.
Qian
B-Z
,
Pollard
JW
. 
Macrophage diversity enhances tumor progression and metastasis
.
Cell
2010
;
141
:
39
51
.
47.
Karnoub
AE
,
Weinberg
RA
. 
Chemokine networks and breast cancer metastasis
.
Breast Dis
2007
;
26
:
75
85
.
48.
Corcoran
RB
,
Contino
G
,
Deshpande
V
,
Tzatsos
A
,
Conrad
C
,
Benes
CH
, et al
STAT3 plays a critical role in KRAS-induced pancreatic tumorigenesis
.
Cancer Res
2011
;
71
:
5020
.
49.
Marotta
LLC
,
Almendro
V
,
Marusyk
A
,
Shipitsin
M
,
Schemme
J
,
Walker
SR
, et al
The JAK2/STAT3 signaling pathway is required for growth of CD44(+)CD24(–) stem cell–like breast cancer cells in human tumors
.
J Clin Invest
2011
;
121
:
2723
35
.
50.
Oskarsson
T
,
Acharyya
S
,
Zhang
XH
,
Vanharanta
S
,
Tavazoie
SF
,
Morris
PG
, et al
Breast cancer cells produce tenascin C as a metastatic niche component to colonize the lungs
.
Nat Med
2011
;
17
:
867
74
.
51.
Takebe
N
,
Miele
L
,
Harris
PJ
,
Jeong
W
,
Bando
H
,
Kahn
M
, et al
Targeting Notch, Hedgehog, and Wnt pathways in cancer stem cells: clinical update
.
Nat Rev Clin Oncol
2015
;
12
:
445
64
.
52.
Saupe
F
,
Schwenzer
A
,
Jia
Y
,
Gasser
I
,
Spenlé
C
,
Langlois
B
, et al
Tenascin-C downregulates Wnt inhibitor dickkopf-1, promoting tumorigenesis in a neuroendocrine tumor model
.
Cell Rep
2013
;
5
:
482
92
.
53.
Tong
M
,
Fung
T-M
,
Luk Steve
T
,
Ng
K-Y
,
Lee Terence
K
,
Lin
C-H
, et al
ANXA3/JNK signaling promotes self-renewal and tumor growth, and its blockade provides a therapeutic target for hepatocellular carcinoma
.
Stem Cell Reports
5
:
45
59
.