The cancer cell line SKOV3 was obtained from ATCC (USA). Cells were cultured in DMEM (cat. no. 043-30085; Wako, Japan), supplemented with 10% fetal bovine serum (FBS; Invitrogen, Carlsbad, CA, USA) and 100 U/ml penicillin/100 µg/ml streptomycin (Wako), and sub-cultured via 0.25% trypsin/EDTA (Wako) detachment. The FBS lot number was 1706567. For the starvation culture, low-glucose DMEM without glutamine (cat. no. 044-33555; Wako) that was not supplemented with FBS was used. All cells were grown in a humidified atmosphere at 37°C with 5% CO₂.

Dissociated single cells (1×10⁵ cells/ml) were seeded into conventional adherent plates (Nalge Nunc International; Thermo Fisher Scientific, Waltham, MA, USA) or ultra-low attachment plates (Corning Inc., Acton, MA, USA) and cultured for 48 h.

Data analyses were performed by Subio, Inc., (Kagoshima, Japan) with Subio Platform.

GSE 85293 contained 16 samples (ovary, 4; omentum, 4; ascites, 4; normal, 4). Probe sets that were not changed (|beta-value|<0.3) compared to normal were excluded. This gave us 153,045 probe sets. After that, the omentum probe sets that were changed compared to the ovary samples (|beta-value|>0.3) were extracted.

GSE 2109 contained 2158 samples, including ovary (86 cases) and omentum (38 cases). The probe sets in which the ABS call value was A (Absent) in more than 90% of cases were excluded. This gave us 38877 probe sets. This set was called the QC1 list. Then, the probe sets in which the processed signal average ranged from −0.1 to 0.1 both in omentum and ovary samples were excluded from the QC1 list. This gave us 16649 probe sets. This set was called the QC2 list. From the QC2 list, the genes that differed significantly in the omentum and ovary were extracted. The criteria were i) |fold change| ≥1.5 and ii) P-value less than 0.1 according to the Mann-Whitney U-test.

IL-6 was purchased from Wako. The final concentration of IL-6 was 50 ng/ml.

The same amount of protein from whole cell lysates was subjected to SDS-polyacrylamide gel (Bio-Rad, Berkeley, CA, USA) electrophoresis and then electrotransferred onto polyvinylidene difluoride membranes (Bio-Rad). Blotting was performed with Trans-Blot Turbo (Bio-Rad) according to the manufacturer's instructions. From blocking to the secondary antibody reaction, iBind Western System (Invitrogen) was used according to the manufacturer's instructions. The secondary antibodies were detected using Immobilon Western Chemiluminescent HRP Substrate (EMD Millipore, Billerica, MA, USA) according to the manufacturer's instructions.

Anti-PGC1α antibody, anti-ACOX1 antibody, anti-CPT1A antibody, anti-CD44 antibody and anti-c-kit antibody were purchased from Abcam (cat. no. ab54481, ab184032, ab128568, ab51037 and ab32363, respectively). Anti-EHHADH antibody and anti-ACSS2 antibody were purchased from SantaCruz (cat. no. SC-393123 and SC-398559, respectively). Anti-p-HSL (Ser 660) antibody, anti-p-AMPK (Thr172) antibody and anti-S6 antibody were purchased from Cell Signaling Technology (cat. no. 4126, 2535 and 2217, respectively). Anti-LDHA antibody was purchased from Proteintech (cat. no. 19987-1-AP). Anti-α-tubulin antibody was purchased from Millipore (cat. no. CP06).

It has been noted that epigenetics are involved in the metabolic shift toward lipid metabolism (11). Thus, we speculated that epigenetic regulation was also related to metabolic reprogramming in ovarian cancer between the primary site (ovary) and the metastatic site (omentum). GSE 85293 was the dataset of DNA methylation status from sites including the ovary and omentum. GSE 2109 was the dataset of gene expressions (microarray or RNA-sequencing) from many sites, including the ovary and omentum. Using these datasets, we investigated and compared the methylation status and gene expression patterns, focusing on the PPAR signaling pathway, a representative lipid metabolism pathway from the Kyoto Encyclopedia of Genes and Genomes (KEGG). As shown in Table I, we could not find a correlation between DNA methylation and gene expression, although the investigation was limited because the datasets used were from different patients and studies.

We could not obtain evidence that epigenetics (at least DNA methylation) were involved in metabolic reprogramming in ovarian cancer. We speculated that a more direct phenomenon was related to the metabolic shift toward lipid metabolism; in other words, detachment from the primary site and suspension in ascites itself affected metabolism (through morphological changes, for instance). Indeed, we have already demonstrated that cancer cells cultured in suspension are more likely to show an activated TCA cycle than cells cultured in adherent plates (8). In the present study, we focused on lipid metabolism. As shown in Fig. 1A, cells cultured in suspension showed increased expression of PGC1α, a sensor for lipid metabolism, compared with cells cultured in adherent plates (12). Additionally, the expression of some types of enzymes related to fatty acid metabolism was also increased, although the level of increase depended on the enzyme. While the expressions of these enzymes were increased in cells cultured in adherent plates when they were exposed to starvation, it seemed that cells cultured in suspension showed less sensitivity to starvation. This could be interpreted to indicate that cells in suspension already utilize fatty acid metabolism and maintain a relative advantage against starvation. Conversely, we considered the possibility that the phosphorylation of AMPK, a sensor for glycolysis, was decreased in cells cultured in suspension compared with cells cultured in adherent plates, as shown in Fig. 1B (13). The expression of LDHA, the indicator of glycolysis, was thought to be unchanged between the different cell culture statuses. Taken with the result from Fig. 1A, this finding suggests that cells cultured in suspension showed a metabolic shift from glycolysis toward lipid metabolism. Although the phosphorylation of AMPK was increased in cells cultured in suspension and exposed to starvation, as shown in Fig. 1B, it is possible that these cells were preparing for impending nourishment.

At present, it is well-documented that CSCs are related to metastasis (14). To add to this knowledge, we investigated how CSC markers changed according to the cell status. We considered CD44 and c-kit as CSC markers for ovarian serous carcinoma, as previously described (14). As shown in Fig. 1C, cells cultured in suspension showed decreased expression of CD44 and c-kit compared with cells cultured in adherent plates. Interestingly, the response during starvation differed between the two culture environments. While the expression of CD44 was decreased during starvation in both cells cultured in suspension and those cultured in adherent plates, the expression of c-kit increased in cells cultured in suspension but decreased in cells cultured in adherent plates. This phenomenon was considered advantageous for metastasis overall, as discussed below.

Above, we describe an in vitro model that mimics the condition of being detached from the primary site and suspended in ascites. Then, we performed an experiment that mimicked the status of ‘cells suspended in ascites and approaching the omentum’. We focused on IL-6, which is thought to be secreted from the omentum or other mesenchymal cells and is known to have some impact on cancer (stem) cells (9). In other words, we investigated the behavior of cancer cells in the presence or absence of IL-6. As shown in Fig. 2A, the expression of CD44 was increased in cells cultured in suspension when exposed to IL-6, while the expression of c-kit was decreased. These results seem paradoxical; however, they could be interpreted to indicate that the cells expressed CD44 and c-kit in relatively high but balanced manner. Indeed, the phosphorylation of S6, the indicator of mTOR pathway activation, was increased in cells cultured in suspension when exposed to IL-6, as shown in Fig. 2C.