An engineered three-dimensional gastric tumor culture model for evaluating the antitumor activity of immune cells in vitro

Monolayer tumor culture models have been used for evaluating the antitumor activity of immune cells in vitro. However, their value in this research is limited. We used human gastric cancer cells (BGC823) and collagen hydrogel as a matrix to establish an engineered three-dimensional (3-D) tumor culture model in vitro. Tumor cells grew in 3-D culture and formed spheroids in the collagen matrix. Evaluation of the antitumor activity of cytokine-induced killer (CIK) cells revealed that, compared with the 2-D cell culture models, CIK cells migrated towards the tumor cells and destroyed the spheroids and tumor cells in the engineered 3-D tumor culture model. The cytotoxicity of CIK cells against the tumor cells in the engineered 3-D tumor culture model was lower than that in 2-D tumor culture models at 12–36 h post-interaction, but there was no significant difference in the cytotoxicity at later time points. Further analysis indicated that dendritic cell-activated CIK cells had a significantly higher level of cytotoxicity against tumor cells, compared with CIK and anti-CEA/CD3-treated CIK cells, in the engineered 3-D tumor culture model. Our data suggest that the engineered 3-D gastric tumor culture model may better mimic the interaction of immune cells with tumor cells in vivo than the 2-D tumor culture models, and may be used for evaluating the antitumor activity of immune cells in vitro.


Introduction
Traditionally, the biological behavior of solid tumors and their responses to immune cells in vitro are examined using stabilized tumor cell lines grown in a two-dimensional (2-D) monolayer culture (1). Although this strategy has provided valuable information with regard to the cytotoxicity of immune cells to tumor cells, it is not sufficient to mimic the antitumor activity of immune cells in vivo. Solid tumor cells grown in three-dimensions (3-D) in vivo support each other and also interact with the extracellular matrix (ECM). The behavior of a single tumor cell is regulated by its interactions with its immediate neighbors and the ECM (2). Similarly, the antitumor activity of immune cells may be affected by the 3-D spatial array and ECM (3,4). In this aspect, the in vitro 2-D tumor cell culture model has limited value in the evaluation of the clinical efficacy of the cytotoxicity of immune cells in vivo. Therefore, it is necessary to establish an in vitro tumor model that accurately reflects the in vivo immune cell-tumor cell interactions. Such in vitro models may be used for successful screening of tumor-reactive immune cells prior to testing their in vivo antitumor immunity in animal models, as well as for the treatment of patients.
Tissue engineering practices have suggested that the 3-D culture of cell models in vitro is better than the planar cultures of cell lines, particularly in mimicking the in vivo cell-cell and cell-matrix interactions (5). Furthermore, it has been reported that the cellular microenvironment, cellcell and cell-matrix interactions within 3-D tissues greatly influence cellular functions, including the adhesion, motility, invasiveness and metastasis of cancers (6). Thus, 3-D culture may be an important strategy for the evaluation of the behavior of tumor cells and the efficacy of immunotherapeutic intervention. There are various methods available for the establishment of 3-D tumor cultures in vitro, including the spinner flask method (7), gyratory rotation system (8) and liquid-overlay technique (9). While the established 3-D tumor models have been successfully used for the evaluation of the effects of drugs (10-12), they are not suitable for the screening of immune cells. Our previous study used a tissue engineering approach and type I collagen hydrogel as a matrix to successfully create hepatic tissue in vitro (13). However, whether the engineered 3-D tumor tissues cultured with a similar approach may be used for evaluating the efficacy of antitumor immune cells has not been explored. Given that immune and tumor cell interactions in the engineered 3-D tumor model resemble those in vivo, we hypothesized that the evaluation of the antitumor activity of immune cells in an engineered 3-D tumor model may be better than that in a 2-D tumor culture model.
In the present study, we first employed tissue engineering technology to establish an in vitro 3-D gastric tumor model. Furthermore, we compared the growth of tumor cells and the cytotoxicity of cytokine-induced killer (CIK) cells in 2-D culture and the engineered 3-D tumor culture models. In addition, we compared the cytotoxicity of CIK cells, dendritic cell-activated CIK (DC-CIK) cells and anti-CEA/CD3 bi-specific singlechain antibody-activated CIK cells (CIK-CEA/CD3-bscAb) in the engineered 3-D model. Our results indicate that this engineered 3-D gastric tumor model may be a simple and useful approach to screen antitumor immune cells in vitro.

Materials and methods
Cell culture. The human undifferentiated stomach adenocarcinoma cell line BGC823 was purchased from the Chinese Academy of Medical Sciences (Beijing, China). Tumor cells were cultured in Dulbecco's modified Eagle's medium (DMEM; Sigma, St. Louis, MO, USA) containing 10% fetal calf serum (FCS; Hyclone, Logan, UT, USA) as monolayer growth in 25-cm 2 plastic tissue culture flasks at 37˚C in 5% CO 2 and 95% humidity in an incubator (HF212UV; Heal Force, Shanghai, China).
Establishment of different types of gastric tumor cell culture models in vitro. BGC823 cells (2x10 5 ) were mixed with 500 µl collagen type I (2 mg/ml) prepared from Sprague Dawley rat tails (14) in 2X DMEM (500 µl) and adjusted to pH 7.2 using 0.1 M NaOH on ice. Subsequently, the mixture was added to 96-well plates (1x10 4 cells/50 µl/well) and incubated at 37˚C in 5% CO 2 for 10-15 min, followed by adding 50 µl complete culture media into each well for the establishment of the 3-D engineered model of gastric tumors in vitro.
Traditional gastric tumor culture models were established and used as controls. Briefly, BGC823 cells (1x10 4 cells/well) were cultured in 96-well plates and used as a monolayer tumor culture model. The same numbers of BGC823 cells were cultured in 96-well plates that had been coated with collagen type I (30 µl/well) and used as the liquid-overlay tumor culture model.
Following culture for varying periods of time, the cells in the monolayer and liquid-overlay models were stained with hematoxylin and eosin (HE) and directly examined by the Lab-Tell ® II Chamber slide system (Nunc, Roskilde, Denmark) on an inverted phase contrast microscope (TE2000-U, Nikon, Tokyo, Japan). The cells in the engineered 3-D model were fixed and paraffin-embedded, followed by staining with HE and examinination under a microscope.
Immune cell isolation and culture. Heparinized peripheral blood samples (50 ml) were obtained from healthy donors and peripheral blood mononuclear cells (PBMCs) were isolated by density gradient Ficoll-Paque centrifugation (TBD, Tianjin, China). PBMCs at 1x10 7 cells/ml were incubated in Cellix 901 serum-free medium (SIMA, Beijing, China) in tissue culture flasks for 3 h and the non-adherent cells were harvested, followed by gentle washing. The adherent cells were exposed to fresh Cellix 901 medium containing rh-IL-4 (1,000 U/ml) and GM-CSF (1,000 U/ml) every other day for seven days. Then the DC cells were loaded with BGC823 tumor cell lysate (20 µg/ml) in the presence of 3 µl/ml Pseudomonas aeruginosa preparation (Weikexi, Haikou, China) and cultured for 48 h. The cells were harvested, washed and stained with fluo-rescent-conjugated antibodies against CD80, CD83, CD86 and HLA-DR (Becton Dickinson, San Diego, CA, USA), followed flow cytometry analysis on a FACScan (Becton Dickinson).

Generation of CIK and DC-CIK cells.
The harvested nonadherent cells were stimulated with 1,000 U/ml IFN-γ (SIMA) in Cellix 601 serum-free medium (SIMA) at 37˚C for 24 h and activated by anti-CD3 (20 µg/ml) and 1 µg/ml anti-CD28 (Mabworks Biotech, Beijing, China) in the presence of 1,000 U/ml IL-2 for 4-5 days to induce CIK cells. Subsequently, one portion of the CIK cells was mixed with mature DCs at a ratio of 100:1 and cultured for 1-2 days to generate DC-CIK cells. The remaining CIK cells were continually cultured for up to 13-14 days following stimulation. The phenotypes of CIK and DC-CIK cells were characterized by flow cytometry analysis using antibodies against CD3, CD4, CD8 and CD56 (Becton Dickinson).
Tumor cell transfection and CIK cell labeling. BGC823 cells (1x10 6 ) were transfected with 6 µg pEGFP-C1 plasmid (Clontech, Mountain View, CA, USA) using the X-fect transfection reagent (Clontech) according to the manufacturer's instructions. The CIK cells (2x10 7 cells) were labeled with 8 µl PKH26 (Sigma), according to the manufacturer's instructions. The efficiency of cell transfection and labeling was examined by flow cytometry analysis.

Antitumor activity of CIK cells in different tumor growth models
Morphology. Unmanipulated and pEGFP-C1-transfected BGC823 cells were cultured in 96-well plates for 24 h for the generation of different gastric tumor growth models. Subsequently, individual wells of tumor cells were reacted in triplicate with CIK and PKH26-labeled CIK cells (1x10 5 cells/well) at an effector:target cell (E:T) ratio of 10:1. The morphology of tumor cells was examined under an inverted phase contrast microscope, and a confocal laser scanning microscope (ACAS-208, Olympus, Tokyo, Japan) was used to observe the immune cell-tumor cell interactions and the antitumor activity of CIK cells.
Cytotoxic effects. The cytotoxicity of CIK cells against individual types of tumor cells was measured by lactate dehydrogenase (LDH) assay using the CytoTox 96 nonradioactive cytotoxicity assay kit (Promega, Madison, WI, USA), according to the manufacturer's instructions. Briefly, BGC823 cells (1x10 4 /well) were cultured for 24 h and then reacted in triplicate with 5x10 4 cells/well or 1x10 5 cells/well CIK cells (E:T ratios of 5-10:1). The BGC823 cells were used as the target cells and the CIK cells as the effector cells. The BGC823 and CIK cells alone were used as controls (Target spontaneous and Effector spontaneous), and BGC823 cells treated with lysis solution were used as maximum control (Target maximum). The supernatants of the cultured cells were harvested, and the levels of LDH in the supernatants were determined by reading at 492 nm on an ELISA analyzer (DNM-9602G, Perlong, Statistical analysis. Data are expressed as the mean ± SD. The difference between groups was analyzed by ANOVA and Student's t-test using the SPSS 17.0 statistical software (SPSS, Inc., Chicago, IL, USA). P<0.05 was considered to indicate a statistically significant result.

Results
Cancer cells form spheroids in collagen matrix. Compared with other tumor culture methods (Fig. 1A and B), human gastric cancer (BGC823) cells grew in 3-D culture and formed tumor spheroids in the engineered culture model (Fig. 1C). Phase contrast micrograph images showed that the numbers and sizes of tumor cells spheroids increased with time (Fig. 1D). Histological characterization revealed that the tumor cells formed numerous spheroids in the collagen matrix in the engineered culture model (Fig. 1G and H). However, there were no tumor cell spheroids in the monolayer culture ( Fig. 1E and F).

Characterization of immune cells. To study the interaction of immune cells with tumor cells in different culture models, we prepared human DCs and CIK and DC-CIK cells in vitro.
Flow cytometry analysis indicated that the CIK cells were >95% CD3 + and 70% CD8 + , but only ~30% CD3 + CD56 + cells. Similarly, the purity of DC-CIK cells was >90% and the majority of DC-CIK cells were CD3 + , CD4 + , CD8 + and CD56 + . The generated mature DCs had 90% purity and the majority of DCs had high levels of CD80, CD86, CD83 and HLA-DR expression.

CIK cells migrate into the matrix and kill cancer cells.
To visualize the interaction of CIK cells with tumor cells, we transfected tumor cells with pEGFP-C1 plasmids and found that 79.49% of the tumor cells were GFP-positive ( Fig. 2A). Similarly, we labeled the CIK cells with PKH26 and found that 99.85% of CIK cells exhibited a red color (Fig. 2B). After adding CIK cells to the different tumor culture models, we found that the CIK cells had direct contact with tumor cells in the control models within a short time (Fig. 3A and B). By contrast, the CIK cells first infiltrated into the collagen matrix (Fig. 3C) and then migrated towards and surrounded the tumor  (Fig. 3D). Finally, the CIK cells destroyed the spheroids (Fig. 3E) and killed the tumor cells (Fig. 3F). Confocal laser scanning and histological analyses indicated that the CIK cells migrated into the collagen matrix and destroyed the spheroids and tumor cells (Fig. 3G and H).

Cytotoxicity of CIK cells in different tumor culture models.
Quantitative analysis of the cytotoxicity of CIK cells in different tumor culture models revealed that following interaction of CIK cells with tumor cells for varying periods of time, the cytotoxicity of CIK cells against tumor cells in the engineered tumor culture model at 12 h post-interaction was low and gradually increased over time. The cytotoxicity of CIK cells against tumor cells in the engineered tumor culture models at 12, 24 and 36 h post-interaction was significantly lower than that in the control models regardless of the ratios of E:T in our experimental system (Fig. 4). However, there was no significant difference in the cytotoxicity against tumor cells among the tested models at 48 h post-interaction.

Discussion
The 2-D cell culture models have been widely used for studying the formation, function and pathology of tumors (15). It has been suggested that 2-D cell cultures do not accurately reflect the cellular environment in vivo due to the significant differences in the spatial array and cellular surroundings (16).
In the present study, we used the tissue engineering approach to establish a 3-D gastric tumor culture model in vitro. Our results showed that tumor cells grew in the collagen matrix and formed spheroids in the matrix. The engineered tumor 3-D culture model has a 3-D structure and should be better than the 2-D cell culture models for mimicking the tumor growth in vivo.
Engineering tissues have been traditionally developed for the repair of damaged tissues and the reconstruction of organs. The engineered scaffold materials to build the 3-D tumor models have been used for evaluating the activity of antitumor drugs in vitro (17)(18)(19). There are few studies using the engineered tumor models to evaluate the antitumor activity of immune cells in vitro. In the present study, we used collagen as the scaffold material to establish a tumor 3-D culture model. Collagen is a commonly used scaffold material in tissue engineering and is also the primary structural protein of ECM that maintains the stability of the basement membrane (20). Collagen has a porous structure that supports tumor cell growth to form a 3-D tumor-like structure. The porous structure of the scaffold provides a huge surface for cells to adhere to and to interact with the scaffold. The porous structure also allows the migration or entry of cultured cells into the scaffold and nutrients to diffuse into the scaffold to support the growth of the seeded cells (21). In addition, while collagen is liquid at a low temperature, it is polymerized to form a collagen gel after increasing the temperature and adjusting the pH to alkaline. Moreover, the collagen gel is translucent and can be easily paraffin-embedded, which allows the examination of the immune and tumor cell interactions by microscopy. Therefore, collagen has a number of advantages in establishing the 3-D tumor model for evaluating the antitumor activity of immune cells in vitro.
Immunotherapies have been widely used for the treatment of cancer patients in the clinic (22)(23)(24)(25). DCs are important antigen-presenting cells and T cells are critical for antitumor immunity. Currently, the antitumor activity of immune cells in vitro is primarily evaluated by the 2-D cell culture (26,27). In this experimental system, the immune cells are able to contact the tumor cells directly, leading to cytotoxicity against tumor cells. However, this model poorly reflects the cytotoxic process of immune cells against tumor cells in vivo, as tumor cells are surrounded by ECM in vivo. Indeed, immune cells have to migrate through the ECM barriers, such as the basement membrane (28). Collagen is the primary structural protein of ECM and may be considered as an artificial ECM (29). We found that tumor cells formed spheroids in the engineered 3-D tumor culture model and that the cytotoxic process of CIK cells against tumor cells in the 3-D tumor culture model was slower than that in the 2-D tumor cell culture models. We observed that CIK cells migrated in the collagen matrix, surrounded the tumor spheroids, invaded the tumor spheroids and finally killed the tumor cells. Therefore, the cytotoxic process of immune cells

A B
in the engineered 3-D tumor culture mode may be similar to that in vivo. Antigen-specific T cells have strong cytotoxicity against tumor cells. CEA is expressed by BGC823 tumor cells, and the CEA/CD3 bi-specific single-chain antibody inhibits the growth of CEA + tumor cells in vivo (30). We compared the cytotoxicity of CIK, DC-CIK and CIK-CEA/CD3-bscAb cells in the engineered 3-D tumor culture model. We found that DC-CIK cells displayed early and strong cytotoxicity against the tumor cells, although the cytotoxicity of these immune effector cells was comparable at 48 h post-interaction in the engineered 3-D tumor culture cell model. Therefore, the engineered 3-D tumor culture model may be used for evaluating the cytotoxicity of different types of immune cells in vitro.
In conclusion, we used collagen as the matrix to establish an engineered gastric tumor 3-D culture model, in which gastric tumor cells formed spheroids. Furthermore, our data indicated that the established engineered 3-D tumor culture model was used for effectively evaluating the cytotoxicity of different types of activated immune cells in vitro.