Human gastric cancer AGS (CRL-1739) and NCI-N87 (CRL-5822) cell lines were purchased from the American Type Culture Collection. The AGS cell line was derived from a gastric adenocarcinoma that was located in the stomach (47), whereas the NCI-N87 cell line was derived from a liver metastasis of a gastric carcinoma (48). Both cell lines were cultured in RPMI-1640 medium (cat. no. R8758; Sigma-Aldrich; Merck KGaA) supplemented with 10% fetal bovine serum (FBS; cat. no. F2442; Sigma-Aldrich; Merck KGaA) and 100 U/ml penicillin-streptomycin (cat. no. 15140122; Gibco; Thermo Fischer Scientific, Inc.) at 37°C in a humidified incubator containing 5% CO₂.

AGS (1.5×10⁴ cells/well) and NCI-N87 (3×10⁴ cells/well) cells were seeded in 96-well plates and incubated at 37°C overnight. After which, cells were exposed to different concentrations of cisplatin (Pfizer, Inc.), doxorubicin (cat. no. 44583; Sigma-Aldrich; Merck KGaA), ATRA (cat. no. R2625; Sigma-Aldrich; Merck KGaA) or fenretinide (cat. no. H7779; Sigma-Aldrich; Merck KGaA) for 48 h at 37°C. As all the drugs evaluated were light-sensitive, drugs and treated cells were protected from light. Doxorubicin and the retinoids, ATRA and fenretinide, were dissolved in dimethyl sulfoxide (DMSO) according to the manufacturer's instructions. Cisplatin solution was directly dissolved in cell culture media.

After the treatments, the media was removed and the cells were incubated with MTT (final concentration, 0.5 mg/ml) solution dissolved in RPMI-1640 medium for 2 h at 37°C. The media was then removed and 95% ethanol was added to the wells to dissolve the formazan crystals. Cells incubated in media with only the vehicle (DMSO<0.05%) were used as a control and were used to define 100% viability. Absorbance values were determined at a wavelength of 570 nm using a microplate reader (BioTek Cytation 3 Imaging Multi-Mode Reader; Biotek; Agilent Technologies, Inc.).

AGS (1.5×10⁴ cells/well) and NCI-N87 (3×10⁴ cells/well) cells were seeded in 96-well plates and incubated at 37°C overnight. Fenretinide and cisplatin were added simultaneously for 48 h and the experiment was conducted with non-fixed ratio combinations. Two doses of cisplatin were selected, corresponding to values below the calculated IC₅₀. By contrast, three doses of fenretinide were chosen, one dose at the IC₅₀ value, one below it and one above it. This strategy aimed to evaluate whether the combination with fenretinide could reduce the concentration of cisplatin required, thereby mitigating the adverse effects associated with cisplatin.

Combined drug action was studied through median effect analysis according to the Chou-Talalay method (49). The combination index (CI) for non-constant combination and drug reduction index (DRI) were calculated using the CompuSyn software (version 1.0; ComboSyn, Inc.). The DRI provided an estimation of the extent to which a dose of an agent in combination can be decreased to achieve effect levels similar to a single drug. The CI was defined as follows: CI<1, synergistic effect; CI>1, antagonistic effect; and CI=1, additive effect.

Intracellular generation of ROS was measured using 2′,7′-dichlorofluorescein diacetate (DCFDA; cat. no. D6883; Sigma-Aldrich; Merck KGaA). DCFDA is a non-fluorescent probe that is internalized by the cell and when oxidized by ROS, is converted into the highly fluorescent 2′,7′-dichlorofluorescein. AGS (1.5×10⁴ cells/well) cells were seeded in black 96-well plates and left to adhere at 37°C for 5 h, before treatment with fenretinide (3 µM) for 20 h at 37°C. As a positive control of ROS production, cells were also treated with the oxidizing reagent tert-butyl hydroperoxide solution (TBHP; 1 mM; cat. no. B2633; Sigma-Aldrich; Merck KGaA) at 37°C for 2 h. Before the conclusion of TBHP incubation, the cells were incubated with DCFDA (60 µM) at 37°C for 30 min. Finally, the wells were rinsed twice with phosphate-buffered saline (PBS) and the intensity of the fluorescence was assessed using a BioTek Cytation 3 Imaging Multi-Mode Reader (Biotek; Agilent Technologies, Inc.), using 485 nm as the excitation wavelength and 528 nm as the emission wavelength. The basal production of ROS was determined in cells treated with medium alone; this value was subtracted from the fluorescence detected when cells were incubated with fenretinide.

AGS (8×10³ cells/well) and NCI-N87 (1.6×10⁴ cells/well) cells were seeded in 96-well plates and incubated at 37°C overnight. Cells were then treated with the antioxidant quercetin (80 µM) for 1 h before the addition of fenretinide (20–40 µM) or TBHP (0.6 µM, AGS; 4.8 µM, NCI-N87) as a positive control of ROS production for 4 h at 37°C. For quercetin, a concentration of 80 µM was selected as it was the highest dose that did not show toxicity in either cell lines. In the case of fenretinide, concentrations >IC₅₀ value that were reported from the MTT assay were used as the experiments were performed at 4 h instead of 48 h.

As various antioxidants interfere with the MTT assay, the sulforhodamine B assay was instead used to determine cell viability. Briefly, cells were fixed in 10% trichloroacetic acid for 1 h at 4°C and stained with 0.04% sulforhodamine B solution (cat. no. S1402; Sigma-Aldrich; Merck KGaA) for 30 min at room temperature. After which, the wells were rinsed three times, eachfor 30 sec, with 1% acetic acid at room temperature to remove the excess dye. Protein-bound dye was dissolved in 10 mM Tris base solution and the absorbance was determined at 530 nm using a microplate reader (BioTek Cytation 3 Imaging Multi-Mode Reader; Biotek; Agilent Technologies, Inc.).

AGS (8×10³ cells/well) and NCI-N87 (1.6×10⁴ cells/well) cells were seeded in black 96-well plates and incubated at 37°C overnight. After which, cells were treated with a moderately toxic dose of fenretinide (10 µM), which corresponded to the IC₂₅ value or avasimibe (20–40 µM) for 48 h. A fenretinide concentration <IC₅₀ value was selected as it allowed detection of a toxic effect while ensuring that most of the visual fields contained sufficient cells for meaningful quantification. This approach was particularly important as visual fields were randomly selected. Avasimibe (cat. no. PZ0190; Sigma-Aldrich; Merck KGaA), an inhibitor of acyl-CoA:cholesterol O-acyltransferases (ACATs), reduces LD synthesis and was therefore used as a control.

Cells were fixed with 4% paraformaldehyde for 10 min at room temperature and incubated in a 0.5 µg/ml solution of Nile red (cat. no. 19123; Sigma-Aldrich; Merck KGaA) in PBS for 10 min at 37°C. Images were captured using a fluorescence microscope with a red fluorescence protein filter (RFP; 531 and 593 nm) with a 20X objective under the same conditions of LED exposure, including integration time and gain, using a BioTek Cytation 3 Imaging Multi-Mode Reader (Biotek; Agilent Technologies, Inc.). For each treatment, images were captured from five independent wells. The Gen5 Image software (version 3.09; Agilent Technologies, Inc.) was used to quantify the fluorescence intensity. All images from each cell line were included in the experiment and an automatic setting was applied for preprocessing for RFP and bright-field channels. The cellular analysis tool was used to select the areas containing cells in the bright-field images, with threshold values for AGS and NCI-N87 of 500 and 4,000, respectively. The object size selection was set between 15–500 µm for both cell lines. The total signal fluorescence object sum intensity [Tsf (RFP 531,593)], total signal fluorescence object sum area [Tsf (bright field)] and the relative object sum intensity/object sum area were calculated for all images. The mean value and standard deviation were determined and statistical analyses were performed.

AGS (1×10⁶ cells/well) cells were seeded on plastic 24-well plates and incubated at 37°C overnight. Wounds were scraped in the middle of the well with a sterile pipette tip. To remove floating cells, wells were carefully rinsed twice with PBS. After which, cells were incubated with medium or fenretinide at a nontoxic concentration (6 µM) for 24 h. Each condition was performed in duplicate in media containing 5% FBS. AGS cell viability decreases considerably when FBS is completely depleted (50). It has been established that, when necessary, the medium can be supplemented with 2–5% FBS to maintain cell health during the assay (51). Since a low concentration of FBS was used in both the control and treated groups, cell proliferation was monitored through the entire wound-healing assay using the sulforhodamine B assay (data not shown). Light microscopy pictures were captured using a microplate reader (BioTek Cytation 3 Imaging Multi-Mode Reader; Biotek Instruments, Inc.) at 0 and 24 h time points. The percentage migration was calculated as the mean of three random measurements in each picture and expressed as a percentage of the control (time 0 h for each treatment) using the ImageJ software (version 1.49; National Institutes of Health).

For all the viability assays, dose-response curves were generated and the IC₅₀ values were calculated using the Slide Write Plus software (version 6.10; Advanced Graphics Software, Inc.). All quantitative experiments were independently repeated three times and data were expressed as the mean ± standard deviation. To evaluate the significance of the results between two groups, the Mann-Whitney test or independent t-test was applied. For comparisons of ≥3 groups, one-way ANOVA followed by Bonferroni's post-hoc test was applied. All tests were performed using GraphPad Prism (version 5; Dotmatics) or SPSS (version 22; IBM Corp.). P<0.05 was considered to indicate a statistically significant difference.

The present study evaluated whether the AGS and NCI-N87 cell lines had a different response to cisplatin and doxorubicin, which are both frequently used to treat gastric cancer. The primary gastric cancer cell line AGS was sensitive to both cytotoxic agents, however, doxorubicin was more efficient, as the IC₅₀ value was 0.39 µM, compared with cisplatin that had an IC₅₀ value of ~20 µM (Table I). The metastatic cell line NCI-N87 demonstrated the opposite effect; cells exhibited a notably low response to doxorubicin, with doses as high as 90 µM being poorly toxic, since cell viability remained close to 80%. By contrast, both cell lines were affected in a similar manner by cisplatin treatment at concentrations <20 µM (Fig. S1). However, at concentrations >20 µM, AGS cells presented lower viability compared with the NCI-N87 metastatic cell line.

The impact of the synthetic retinoid, fenretinide, was assessed using an MTT assay and its effects compared with those of its natural analog ATRA were recorded. ATRA was used to compare whether fenretinide-treated cells showed the same response observed with doxorubicin or cisplatin. Since ATRA has been shown to be ineffectively delivered in solid tumors (52), it was hypothesized in the present study that for fenretinide to be promising as a therapeutic drug in gastric cancer, the IC₅₀ values should be significantly lower than those for ATRA.

The results of the present study showed not only that NCI-N87 cells were less sensitive to ATRA compared with AGS cells, but also that both cell lines had lower IC₅₀ values when treated with fenretinide compared with ATRA. In fact, both cell lines had a similar response when treated with fenretinide, and the IC₅₀ values were in a similar range as those obtained when using cisplatin (Table I; Fig. S1).

As MTT assays are primarily based on the activity of mitochondrial enzymes and cell metabolic health (53,54), a sulforhodamine assay was also performed to allow the determination of cell density based on the measurement of cellular protein content (55). The results obtained using the MTT and sulforhodamine assays were consistent with each other (Fig. S1C-F).

It has been suggested that retinoids could be used as co-adjuvants in combination with chemotherapy (56), therefore it was evaluated whether fenretinide increased the toxic effect of cisplatin on gastric cancer cells. Cisplatin was selected as it was the only chemotherapy drug that showed a toxic effect on both cell lines in the present study. Additionally, the current first-line treatment for advanced or metastatic unresectable gastric cancer is a dual chemotherapy regimen based on a platinum compound and fluoropyrimidine, which suggests that anthracyclines may not be needed for optimal results (57).

Following combined treatment with fenretinide and cisplatin for 48 h, CompuSyn software was used to calculate the CI and DRI for cisplatin. The combination of three non-fixed toxic doses of fenretinide (similar to IC₅₀) with two different doses of cisplatin was evaluated (Table II; Fig. 1). Based on the CI value, it was suggested that in both cell lines, there may be an additive or partial synergistic interaction between both drugs when a high fraction of cells is affected (fraction affected >0.7). Notably, the additive effect was stronger in the NCI-N87 metastatic cell line; combination index values were closer to 0.9 and the drug reduction index values for cisplatin were at least twice those observed in AGS cells. However, when a low fraction of cells were affected, it was possible to observe CI values that suggested antagonism. However, for anticancer agents, synergy is more relevant to therapy when a high fraction of cells are affected (49).

Cancer cells exhibit increased ROS stress compared with normal cells, therefore targeting ROS levels may be useful as a therapeutic strategy (58). It was assessed whether fenretinide could induce an alteration in redox homeostasis. Following preincubation of both cell lines with the antioxidant quercetin and fenretinide, protection from cell death was only demonstrated in AGS cells (Fig. 2A and B). Nonetheless, the results of the present study suggested that other mechanisms besides ROS induction may also be implicated, as cell viability did not recover to levels similar to those in the control conditions. By contrast, for NCI-N87 metastatic cells, ROS production did not appear to be relevant for this effect as fenretinide did not enhance cell viability following TBHP exposure. The increase in ROS stress induced by TBHP was confirmed by the fact that the antioxidant quercetin was able to restore cell viability.

Notably, the TBHP concentration used to produce minimal toxic effect was increased ~8-fold compared with AGS cells.

To confirm the results observed in AGS cells, a DCFDA probe was used to detect intracellular ROS levels after treatment with fenretinide (Fig. 2C). After 20 h, a significant elevation in ROS was detected. However, this effect was smaller than the effect of TBHP, which was used as a positive control as it is a well-established ROS inducer (59).

Autophagy is implicated in fenretinide induced cell death (38). Xu and Fan (60) reported that LD content can be modified by autophagy. To assess this in the present study, AGS and NCI-N87 cells were stained with Nile red, a lipophilic agent that stains intracellular LDs. As a control, cells were incubated with avasimibe, an ACAT inhibitor that reduces LD formation (61).

In AGS cells, fenretinide significantly reduced Nile red fluorescence, which suggested a decrease in the number of LDs. By contrast, NCI-N87 cells treated with fenretinide showed no change in Nile red fluorescence in comparison with cells without treatment. Of note, when the NCI-N87 metastatic cell line was treated with avasimibe, LD content was significantly increased compared with the control cells (Fig. 3). A previous study reported that NCI-N87 cells regulate intracellular cholesterol differently from AGS cells. Furthermore, it was reported that NCI-N87 cells were more resistant to the effect of avasimibe when cell viability was assessed, suggesting the presence of a compensatory mechanism for cholesterol esterification in addition to ACAT-1 (50). This may partially explain why when treated with an inhibitor of LD formation, the opposite effect was observed. Since avasimibe changed LD content in these cells and fenretinide did not, these results suggest that fenretinide is not affecting the autophagy mechanisms that may break down LDs.

Previous studies have demonstrated that, under basal conditions, NCI-N87 cells require a prolonged period, usually 48 h, to exhibit a migration effect in the wound healing assay (62–64). In comparison with other gastric cancer cell lines, including AGS cells, NCI-N87 cell line demonstrate a more differentiated epithelial phenotype characterized by high expression levels of E-cadherin and zonula occludens proteins (65). This phenotype facilitates the formation of highly cohesive monolayers, thereby hindering their migration. As NCI-N87 cells have a limited capacity to migrate in the wound healing assay, only the effect of fenretinide in the migration of AGS cells was evaluated. After 24 h incubation with a nontoxic concentration of fenretinide, a significant reduction in the capability of AGS cells to migrate and close the wound was observed (Fig. 4).