Thirty three patient-derived HGSOC cell lines (from patients 1–23) and 5 primary cell cultures with predominant mesothelial cells (derived from patients 6, 7, 9, 11 and from an additional patient for #13088) (Figure 1A) were previously established and molecularly characterized (Kreuzinger et al., 2019). All information pertaining to these cell lines is available from the Expasy Cellosaurus cell line database (https://web.expasy.org/cellosaurus). RNA sequencing data were previously published (Kreuzinger et al., 2019).

Cells were cultivated at 37°C and 5% CO₂ in DMEM Medium supplemented with 10% Fetal Bovine Serum, 10,000 Units/ml Penicillin and 10,000 µg/ml Streptomycin (all from Gibco Life Technologies, CA, United States). Before splitting, cells were rinsed first with PBS and thereafter, trypsinized with 0.05% Trypsin-EDTA (1x) (Gibco Life Technologies) (splitting ratio 1:2–1:4). Cells were stored in liquid nitrogen in medium supplemented with 10% DMSO (Sigma-Aldrich, St. Louis, United States).

AllPrep DNA/RNA Mini Kit (Qiagen, Hilden, Germany) was used for DNA and RNA extractions according to the manufacturer’s instructions. The yield and purity were calculated by measuring the absorbance at 260/280 nm with Nanodrop 1,000 (Thermo Scientific, MA, United States). DNA was stored at 4°C and RNA at −80°C. For cDNA synthesis, the Omniscript Reverse Transcription Kit (Qiagen, Hilden, Germany) was used and the cDNA was stored at −80°C.

Gene expression of SLCO4A1 (ENSG00000101187) obtained via RNA sequencing was validated by RT-qPCR using the ABI prism 7900HT system (Applied Biosystems; Thermo Fisher Scientific, Inc.) and TaqMan Gene Expression assay for SLCO4A1 (Hs00249583_m1) (Thermo Fisher Scientific, Inc.). All reactions were performed in duplicate. A cell line with a higher expression of the target gene was used to generate standard curves. By comparing the CT value of a certain sample relative to the standard curve, a relative gene expression value for each sample was calculated and expressed as fold-change. Statistical significance (p < 0.05) between the expression of each individual SLCO in HGSOC and mesothelial cell cultures was calculated using the Mann-Whitney test (SPSS 27; IBM Corp.). p < 0.05 was considered to be statistically.

To determine the potential influence of gene mutations on gene expression, Sanger sequencing was performed on a particular region that may contain the g.323G>A (p.Val78Ile; Accession No. AB031051.1) mutation. Primer pairs (sense, 5′-AGA​CGT​GAG​CTT​GCT​AAC​CAG-3′; antisense, 5′-TGG​TAG​CTG​TGC​AGG​TCA​TAG-3′) were designed and PCR was performed using DNA samples under the following thermocycling conditions: 40 cycles of 30 s at 94°C, 60 s at 65°C and 80 s at 72°C. PCR products were purified using the GFX PCR DNA kit (Amersham Biosciences Corp.) and sequenced by Eurofins Genomics LLC. The CLC mainwb5 program was used for data evaluation.

For protein extraction, the cell pellet was resuspended in cold RIPA buffer (Sigma-Aldrich; Merck KGaA) with proteinase inhibitor (Sigma-Aldrich; Merck KGaA) and incubated on ice for 30 min. The lysates were centrifuged at 12,000 rpm for 20 min at 4°C, after which the supernatant was collected and stored at −80°C. For protein quantification, a Bradford Assay was used. A total of 20 µg protein was mixed with 5X Laemmli Sample Buffer containing 5% β-mercaptoethanol, and heated for 5 min at 95°C. Subsequently, the proteins were loaded on a Mini Protean stain-free Tris-Glycine extended gel (TGX, 8%–16%; Bio-Rad Laboratories, Inc.) and separated at 100 V for 60 min under denaturing conditions. Samples were then transferred onto nitrocellulose membranes at 25 V and 4°C using electrophoresis Power Supply 301 equipment (Amersham Pharmacia Biotech, Inc.). After 24 h, the membranes were blocked with 5% dried milk in TBS-0.1% Tween for 1 h at room temperature. The membranes were then incubated with the following primary antibodies for 60 min at room temperature: OATP4A1 (rabbit IgG; Clone H-145; 1:100; Santa Cruz Biotechnology, Inc.) and GAPDH (Rabbit IgG; Clone 14C10; Cell Signaling Technology, Inc.). After washing the membrane three times with TBS-Tween, the membrane was incubated with anti-rabbit horseradish peroxidase (HRP)-conjugated secondary antibody (Cell Signaling Technology, Inc.) at room temperature for 1 h. Target proteins were detected using the LuminataTM Crescendo/Classico Western HRP Substrate (Merck KGaA) for chemiluminescent detection with FluorChem Q (Alpha Innotech). The blots were evaluated according to their intensity and volume using open-source software Image Lab 6.0.

Formalin-fixed paraffin-embedded (FFPE) tumor tissues were obtained from the Department of Pathology, Medical University of Vienna. Informed consents were obtained from all patients prior to sample collection. All processes were approved by the local ethical committee (EK366/2003, EK260/2003). Immunohistochemistry was performed on 3 µm sections cut using the automatic microtome HM355S (Microm International GmbH, Walldorf, Germany). The sections were dried overnight and deposited at −20°C before being further processed. Thereafter FFPE sections were treated with xylol (Fisher Scientific, Loughborough, United Kingdom) and a series of descending ethanol concentrations. To permeabilize the tumor tissue, sections were incubated in 0.5% Triton X (Merck Millipore, Darmstadt, Germany) diluted in PBS for 15 min at RT. After washing in PBS and in distilled water 3 times for 2 min each, antigen retrieval was performed by boiling the slides in a citrate buffer (pH 6) (Merck Millipore) in a microwave oven at 850 W for 2.5 min, which was followed by 160 W for 10 min. They were slowly cooled down to RT for 1–2 h. The endogenous peroxidase was blocked by incubating the slides for 10 min in 3% H₂O₂ (Merck Millipore) diluted in methanol (Thermo Fisher Scientific). Afterwards, the samples were washed twice in distilled water, each time for 2 min. To block unspecific binding, the samples were covered with Dako Ultra V Block (Agilent Technologies, Vienna, Austria) for 7 min and then washed in PBS-Tween twice. The primary antibodies were diluted to the optimal concentration in Dako Antibody-Diluent (Agilent Technologies). OATP-E/SLCO4A1 antibody (Rabbit IgG, Clone H-145; at 1:100; Santa Cruz, United States) and Negative Control rabbit IgG (BioCare Medical, CA, United States) were used. The slides were then incubated with the antibody at 4°C overnight.

Then the slides were rinsed and washed twice with PBS-Tween, each time for 3 min. Dako LSAB2 System—HRP was applied in a 3-step procedure, by adding a biotinylated anti-rabbit and anti-mouse immunoglobulin for 30 min, streptavidin conjugated to horseradish peroxidase for 30 min and a substrate-DAB solution for about 5–20 min. The slides were rinsed with PBS-Tween twice, 3 min between each step. The reaction was observed under the microscope and stopped by washing with distilled water for 3 min.

To stain the nucleus, Hematoxylin Gill III (Merck Millipore, Darmstadt, Germany) was added to the sections for 1–2 s and the slides were immediately rinsed in water to stop the reaction. Slides were covered with Kaiser’s glycerol-gelatine (Merck Millipore) and dried at RT.

FFPE sections were treated with xylol and descending ethanol concentrations. For antigen retrieval, a citraconic anhydride buffer was used. PBS/Tween containing 5% goat serum was applied for blocking. First, staining was done with the rabbit anti-human antibody against OATP4A1 (HPA030669, Atlas Antibodies AB), and, thereafter, a mouse-monoclonal antibody against cytokeratin 19 (A53-B/A2, Abcam) as a marker for epithelial ovarian cancer cells (Menz et al., 2021) was applied. An Alexa Fluor 488-labelled anti-rabbit antibody and an anti-mouse Alex-Fluor 568-labelled antibody (Abcam, Cambridge, MA, United States) were added as secondary antibodies. For negative controls, the primary antibodies were replaced by non-immunogenic IgG. To stain the nuclei, the sections were incubated for 10 min with DAPI (0.2 µg/ml in PBS). After rinsing with distilled water, the sections were mounted with Fluoromount G and covered by cover slides.

For the identification of SLCO4A1 related pathways, the tumor cell line samples were categorized for high and low SLCO4A1 levels. Eight cell lines demonstrating the highest and eight cell lines with the lowest SLCO4A1 expression were selected for analysis. In cases where more than one cell line derived from the same patient appeared in the same group, the mean was calculated. GSEA 4.1.0 (https://www.gsea-msigdb.org/gsea/index.jsp) was used for the analyses. The data sets c2.cp.kegg.v7.1.symbols.gmt for KEGG and c2.cp.biocarta.v7.2.symbols.gmt for BIOCARTA were selected from the Molecular Signatures Database (MsigDB: https://www.gsea-msigdb.org/gsea/msigdb/index.jsp).

All statistical analyses were performed using IBM SPSS version 27, except the heatmap, which was generated using R version 3.5.2. The heatmap function in R 3.5.2 was used to hierarchically cluster log-transformed expression values with respect to cell lines and genes. Pearson’s correlation coefficient was used to quantify the correlation between pairs of variables in case of a linear association. Spearman’s correlation coefficient was used otherwise in case of a monotone association. p-values < 0.05 are assumed to indicate statistical significance.

We first compared the RNA levels of selected SLCOs in the 33 ovarian cancer cell lines and the five primary cell cultures containing mainly mesothelial cells (Kreuzinger et al., 2019) by RNA sequencing. SLCO2A1, 2B1, 3A1, and 4A1 were detected at considerable levels in the cell lines (Figure 1B). While levels of SLCO2A1, SLCO2B1 and that of the most abundantly expressed SLCO3A1 were higher in mesothelial cells than in ovarian cancer cells (p = 0.002 and p = 0.03, respectively), SLCO4A1 mRNA levels were higher in ovarian cancer cells (median of 304 read counts) as compared to mesothelial cells (median of 41 read counts) (Figure 1; p = 0.05).

The strong correlation between RNA sequencing and RT-qPCR data indicating the reliability of the sequencing data (Supplementary Figure S1; Pearson correlation coefficient: R = 0.86; p < 0.001).

Sanger sequencing revealed that the heterozygous mutation g.323G>A (p.Val78Ile), previously described in colon cancer and healthy control patients, was present in 2 out of 7 investigated cell lines. This SNP, which is located in the first intracellular loop of the protein distant to the binding site of drugs, was previously shown to have no effect either on the expression levels nor on the uptake of the OATP surrogate substrate sodium-fluorescein (Buxhofer-Ausch et al., 2020).

Comparison with clinical data of patients showed that SLCO4A1 mRNA level correlated neither with the origin of the tumor cells (either from tumor tissue, n = 8; or ascites; n = 23; or pleural fluid, n = 2) nor with the time of sample collection (at the primary diagnosis, n = 20 or during disease progression and chemotherapy, n = 13), (p = 0.85 and 0.98, respectively). For example, the cell line with the highest SLCO4A1 expression (2,502 read counts in cell line 14433) and the cell line with a SLCO4A1 level below the detection limit (cell line 17330) were both derived from tumors of patients at the time of the first diagnosis prior to platinum-based chemotherapy. Furthermore, cell line 13363 and 15233 were derived from patient 3 before and after chemotherapy, respectively and had similar low SLCO4A1 expression (5 and 20 read counts, respectively).

We also examined the SLCO4A1 encoded OATP4A1 in the ovarian cancer cell lines by Western blotting and showed OATP4A1 immunoreactive protein in ovarian cancer samples by IHC and double-IF staining. On the Western blot, the immunoreactive band at approximately 70 kDA corresponded to the predicted molecular mass of the transporter (Figure 2). An additional band at 62 kDA in the mesothelial cells 13914_2 may reflect the non-glycosylated protein. The semi-quantitative evaluation of the immunoreactive band shows a moderate correlation between the immunoreactive protein and relative SLCO4A1 expression levels (Pearson correlation coefficient = 0.52, p = 0.024; Supplementary Table S1; Supplementary Figure S1). By IHC staining, we showed that SLCO4A1/OATP4A1 in FFPE samples reflects the SLCO4A1 expression in the tumor cell lines derived from the tumor of the same patient (Figures 3A–D). The transporter was mainly visible on the membrane and in the cytoplasm of the tumor cells (Figures 3B–D), similar as the observation in other tumors including colon carcinoma (Kleberg et al., 2012; Buxhofer-Ausch et al., 2019). As expected, immunoreactive OATP4A1 was not detected in adipose tissues. Double staining with an antibody against CK19 (red color), which was previously shown to be highly present in the ovarian cancer cells (Svoboda et al., 2018; Menz et al., 2021), revealed that OATP4A1 (green color) is rarely detectable in cells together with CK-19 but rather in cells that do not express this marker (Figures 3E,F). This is in line with the findings that cells with high SLCO4A1 levels have an overall gene expression pattern more similar to that of mesothelial cells (see Figure 4).

To examine the possible relationship between SLCO4A1 and genes coding for enzymes and transporters and genes in the pathways for prostaglandin and lipid turnover, steroid hormone receptors, and their transcriptional regulators, a set of 77 genes was selected (Supplementary Table S2). Spearman correlation analysis of individual genes indicated that ABCC3, an efflux pump from the family of multidrug resistance-related transporters, has a strong and significant positive correlation to SLCO4A1 in the ovarian cancer cells (r = 0.75, p < 0.001). Correlation values for all other genes were lower (|r|=<0.6).

Unsupervised clustering of the 77 genes revealed two major groups of cell lines (Figure 4). Group 1 with lower levels of SLCO4A1 consists of 15 ovarian cancer cell lines from nine patients (median SLCO4A1 levels: 72 read count). Group 2 with significantly higher levels (p = 0.005) of SLCO4A1 (median: 412 read counts) includes the remaining 18 ovarian cancer cell lines from 14 patients and the 5 mesothelial cell lines, the latter with SLCO4A1 levels of 41 read counts (median). Within the group of mesothelial cells, the gene expression pattern is rather uniform and gives these cells a separate sub-cluster next to three ovarian cancer cell lines from group 2, which are adjacent to group 1 cell lines. The residual ovarian cancer cell lines belonging to group 2 are located at the right side of the diagram. The overall gene expression pattern of group 2 ovarian cancer cells shows a high association with the expression pattern of genes in the mesothelial cells, which are associated with cell metabolism and proliferation.

Importantly, cell lines derived from the same patient had the closest relationship in the gene expression pattern independent of the time of establishment (at the primary diagnosis or later during disease progression and exposure to chemotherapy) as visible from the column clusters. We also show that only minor changes in the overall gene expression pattern during tumor progression and chemotherapy were detectable in the ovarian cancer cell lines, which were derived from the same patients at the time of diagnosis or later after tumor progression and chemotherapy. It seems that some populations of ovarian cancer cells keep their phenotype during tumor progression. Nevertheless, cancer cell populations with other properties and an altered phenotype produced by clonal selection, which takes place in HGSOC (Lambrechts et al., 2016) might get lost during the isolation procedure.

In order to identify which pathways might be associated with the expression of SLCO4A1, eight ovarian cancer cell lines with the highest (median 1,420 read counts) and eight cancer cell lines with the lowest SLCO4A1 expression levels (median 18 read counts) (p = 0.001) were selected to perform the gene set enrichment analysis (GSEA). The database Kyoto Encyclopedia of Genes and Genomes (KEGG) and the BIOCARTA database were used for the analysis.

The resulting heat map is shown in Figure 5. Among the top 50 enriched genes, also ABCC3 coding for MRP3, an efflux pump for non-conjugated organic anions and glucuronide-conjugates of endogenous and exogenous compounds (Keppler, 2011) was present, suggesting that a close relation between the uptake and efflux by these transporters may exist. Interestingly, also the expression of SLCO4A1 Antisense RNA 1 (SLCO4A1-AS1) was related to high SLCO4A1 levels.

In the group of ovarian cancer cells with low SLCO4A1 levels, SLC25A12 coding for the SLC25A12 transporter was identified among the most significant enriched genes. SLC25A12 is responsible for the calcium-dependent exchange of cytoplasmic glutamate with mitochondrial aspartate (Palmieri et al., 2001) and may therefore contribute to disturbances in the mitochondrial function (Figure 5).

The most significant pathways associated with high and low SLCO4A1 expression are given in Table 1. The enrichment score (ES) reflects the degree to which the genes are overrepresented at the top or bottom of the entire ranked gene list. The values are normalized to the mean enrichment of random samples of the same size (NES) and range from 2.06 to 1.73. The nominal (NOM) p-value estimates the significance of the observed enrichment score for a single gene set and the False-Discovery rate (FDR), which gives the estimated probability that a gene set with a given enrichment score represents a false positive finding (normalized for gene set size). The latter values are normalized for gene set size.

From the 16 pathways identified in the BIOCARTA and the seven in the KEGG analysis associated with higher SLCO4A1 expression, the TALL-1, CD40, NF-KB and TNFR2 signaling pathways (BIOCHARTA) and the Nucleotide-binding, oligomerization domain (NOD)-like receptors (NLRs) and the adipocytokine signaling pathways (KEGG) were the most prominent ones (FDR < 0.125, p < 0.01, Table 1). These particular pathways are related to NF-kB signaling, suggesting a prominent role of the transporter in NF-kB related inflammatory processes.

Lower expression of SLCO4A1 was significantly associated with the mitochondrial electron transport chain (ETC) pathway, which is important for ATP production and the maintenance of reactive oxygen species (ROS) homoeostasis in cells. This pathway had been identified in many tumors, including in HGSOC and was indicated as an potential therapeutic target (Nayak et al., 2018) (BIOCARTA analysis, Table 1).