Multicellular tumor spheroids of human uveal melanoma induce genes associated with anoikis resistance, lipogenesis, and SSXs

Biopsies and cell cultures

UM biopsies from patients undergoing enucleation of the eye were included in this study. After enucleation, the ophthalmic pathologist excised fresh tumor tissue for use in research before formalin fixation for routine histopathological examination. The UM of the three donors (D1, D2, and D3) was classified as mixed (D1) and epithelioid (D2 and D3) types with a routine histopathological examination. Retrospectively, donors D1, D2, and D3 all had confirmed liver metastases. A fourth supplementary donor was added to the study after data were obtained. The UM of this donor, D(S), was classified as epithelioid, and the donor tissue underwent the same culture conditions as the tissue from donors D1, D2, and D3.

A fraction of the tissue was snap-frozen and stored at −80 °C. The remaining sample was minced with scissors in collagenase I and IV (1 mg/ml), before being incubated for 1 h at 37 °C. After dissociation, the tissue was cultured adherently for 7 days in RPMI 1640 (Invitrogen, Carlsbad, CA), 10% fetal bovine serum (FBS), penicillin/streptomycin (100 U/ml, P4333), and amphotericin B (2.5 µg/ml, A2942) in addition to gentamycin (75 µg/ml; Sanofi-Aventis, Gentilly, France) to ensure the removal of fibroblasts [22]. After 7 days of adherent culturing, the cells were trypsinized using Trypsin-EDTA (0.25%, T4049) and pelleted into three fractions of 100,000 cells. The first fraction of the cells was collected for RNA analyses, the second fraction for further adherent growth, and the third for non-adherent growth as MCTS. The term MCTS is used for this non-adherent culture of tumor cells, in accordance with the nomenclature, and is considered aggregation and compaction of tumor cells [21]. The cell fraction for MCTS culture was plated at a density of 500–1,000 cells per well on Corning Costar ultra-low attachment, polystyrene, round-bottom 96-well plates (CLS7007) in melanoma stem cell medium (MSCM) (1) and (2): (1) 30% human embryonic stem cell medium (hESC); (78% KnockOut DMEM/F12 (Cat. no. 12660–012, Thermo Fisher Scientific Inc., Waltham, MA), 20% KnockOut serum replacer (Cat. no. 10828–028, Thermo Fisher Scientific Inc.), 1% MEM non-essential amino acids (Cat. no. 11140–050, Thermo Fisher Scientific Inc.), 4 ng/ml basic fibroblast growth factor (b-FGF; Cat. no. 13,256-029, Thermo Fisher Scientific Inc.), 1% GlutaMAX (35,050-061, Thermo Fisher Scientific Inc.), and 1.4‰ 2-mercaptoethanol (M7522) and (2) 70% mouse embryonic fibroblast (MEF) conditioned medium (AR005, R&D Systems/Bio-Techne, Minneapolis, MN) [23] with penicillin/streptomycin (100 U/ml) and amphotericin B (2.5 µg/ml). The cells were collected after 12 days of cell culture and further embedded in paraffin for immunohistochemistry (IHC) or pelleted and stored at −80 °C for RNA analyses.

RNA isolation

RNA from fresh frozen primary tumors (D1, D2, and D3) was isolated using the Qiagen RNeasy kit (Qiagen, Hilden, Germany). Briefly, the tissue was placed in a 4.5 ml cryotube, and 500 µl of QIAzol (Qiagen) was added before the sample was disrupted using Qiagen TissueRuptor (Qiagen), according to the manufacturer’s recommendations. The sample was centrifuged at 18 400 ×g for 10 min to remove insoluble material before being processed with the Qiagen RNeasy kit with DNase. Samples were purified using the Zymo PCR inhibitor removal kit (Zymo, Irvine, CA). RNA from the pelleted samples (adherent and cultured spheres from D1, D2, and D3) was isolated as described above, except the disruption step using the Qiagen TissueRuptor. RNA concentration and purity were determined using NanoDrop (Wilmington, DE) and Bioanalyzer (Agilent 2100, Agilent, Santa Clara, CA). All nine samples had RNA integrity number (RIN) values above 8 before being analyzed with microarray and PCR [24].

Immunohistochemistry

The growth media in the 96-well plates was diluted by gently adding Hanks’ Balanced Salt solution (Thermo Fisher Scientific Inc.). Then the MCTS were allowed to make sediment before the media was carefully removed. A mixture of human plasma and thrombin (Sigma–Aldrich) was used to clot the MCTS together before fixation in 4% paraformaldehyde (PFA) and embedment in paraffin. Then 3.5 μm sections were cut and stained [25]. K_i-67 staining was performed using the Envision + Dual Link HRP (K4065, Dako, Glostrup, Denmark) and AEC + Substrate chromogen ready-to-use (k3461, Dako). Briefly, the K4065 kit protocol was followed until the addition of 3,3′-diaminobenzidine (DAB). After polymer horseradish peroxidase (HRP), 3-amino-9-ethylcarbazole (AEC) chromogen from the kit k3461 was added, and the sections were washed and counterstained with hematoxylin according to the k3461 protocol. Negative controls without primary antibody were included for all stainings. The following primary antibodies and dilutions were used (rabbit:rb, mouse:ms): K_i-67 (rb, 1:200; Thermo Fisher Scientific Inc.), SSX4 (rb, 1;50; Acris), anti-melanoma (a-melanoma) [HMB45 + MART1 (DT101 + BC199) + tyrosinase (T311)] (ms, 1:50; ab733, Abcam, Cambridge, UK), Perilipin (rb, 1:100; Santa Cruz Biotechnology Inc., Dallas, TX), and ANGPTL4 (rb, 1:500; Abcam). The secondary antibodies had the fluorescent marker Alexa Fluor 488 (1:500; Invitrogen). Hoechst (1:500; Invitrogen) was used for nuclear staining. The sections were analyzed using a Zeiss Axio Observer.Z1 fluorescence microscope (Zeiss, Oberkochen, Germany). Sections were also stained with hematoxylin and eosin (H&E) for morphological examination.

Microarray

Microarray analysis was performed at the Genomics Core Facility, Oslo University Hospital and Helse Sør-Øst. HumanHT-12 v4 Expression BeadChip (Illumina, San Diego, CA) was used for the analysis. It targets more than 31,000 annotated genes with 47,000 probes mainly derived from the National Center for Biotechnology Information Reference Sequence (NCBI) RefSeq Release 38 (November 7, 2009). For each sample, 440 ng of total RNA was amplified and labeled using the Illumina TotalPrep-96 RNA Amplification Kit protocol. The quantity of labeled copy RNA (cRNA) was measured using the NanoDrop spectrophotometer (Wilmington, DE). The quality and size distribution of the labeled cRNA were assessed using the 2100 Bioanalyzer. This was done to be able to hybridize equal amounts of successfully labeled cRNA to the arrays. For each sample, 750 ng of biotin-labeled cRNA was hybridized to the Illumina HumanHT-12 v4 Expression BeadChip. J-Express and rank product (RP) analysis were used to further identify differently expressed genes with ≥2 fold up- or downregulation and q values ≤0.05 between the different groups. One thousand permutations (1,000*) were run for each RP analysis [26].

Quantitative reverse-transcription PCR

RNA concentration and purity were measured using NanoDrop. Reverse transcription (RT) was performed using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Abingdon, UK) with 50 ng total RNA per 20 μl RT reaction. Copy DNA (cDNA) was diluted to a volume of 50 µl (1 ng/µl) after cDNA synthesis. Quantitative PCR (qPCR) was performed using the StepOnePlus RT–PCR system (Applied Biosystems) and Taqman Gene Expression assays following the manufacturer’s protocols (Applied Biosystems). The TaqMan Gene Expression Assays used include ANGPTL4 (Hs01101127_m1) and 18S (Hs03003631_g1). The thermal cycling conditions were 95 °C for 10 min followed by 40 cycles of 95 °C for 15 s and 60 °C for 1 min. All samples were run in duplicate (each reaction: 2.5 μl/2.5 ng cDNA in a total volume of 12.5 μl). The data were analyzed using the 2^−ΔΔCt method to find the relative changes in gene expression as a fold change between the samples. The uncultured tumor sample was chosen as the calibrator and equaled one, while the other samples had fold changes related to the uncultured tumor calibrator sample. The 18S probe, primers, and assay (Hs03003631_g1) were used as a loading control to quantify the differences in cDNA input between the samples.

RNAscope in situ hybridization

RNA in situ hybridization was performed using the RNAscope® 2.5 High Definition (HD)- Red assay (Advanced Cell Diagnostics, Hayward, CA) according to the manufacturer’s instructions using the standard pretreatment protocol. Sections were mounted using Prolong Gold with 4',6-diamidine-2'-phenylindole dihydrochloride (DAPI). RNAscope permits direct visualization of RNA in formalin-fixed, paraffin-embedded (FFPE) tissue with single molecule sensitivity and single cell resolution [27]. RNAscope Probe-Hs-SSX4–01 (Cat. no. 468,641, Advanced Cell Diagnostics) was used. Hybridization signals were detected with chromogenic reactions using Fast Red. Fast Red produces red fluorescence in addition to the red reaction product, thus providing a greater level of sensitivity [28]. The RNA staining signal was identified as red punctate dots. Each sample was quality controlled for RNA integrity with a probe specific to peptidyl-prolyl cis-trans isomerase B (PPIB) mRNA. Negative control background staining was evaluated using a probe (Cat.no. 3100439, Advanced Cell Diagnostics, Newark, NJ) specific to the bacterial dihydrodipicolinate reductase (DapB) gene (Gene ID EF191515). The sections were analyzed with a Zeiss Axio Observer.Z1 fluorescence microscope.

Pathway and gene ontology analysis

Data from the microarray analysis were imported into Ingenuity Pathway Analysis (IPA) software in the search for biologic pathways and Gene Ontology to identify potential networks. Principal component analysis (PCA) and unsupervised hierarchical clustering were performed using the Partek Genomics Suite software (Partek, Inc., Chesterfield, MO).

Transmission electron microscopy

Primary tissue from uncultured tumor D1 and the donor D(S) cultured as MCTS were fixed at 4 °C overnight in glutaraldehyde (0.1 M). The tissue was washed four times in cacodylate buffer (0.2 M) before post-fixation in a mixture of 1% osmium tetroxide and cacodylate buffer (0.2 M) for 60 min. The tissue was further rinsed in cacodylate buffer (0.2 M) before being dehydrated through a graded series of ethanol up to 100%. The tissue was then immersed in propylene oxide for 2 ×5 min and a mixture of Epon and propylene oxide before embedment in Epon. Ultrathin sections (60–70 nm thick) were cut on a Leica Ultracut Ultramicrotome UCT (Leica, Wetzlar, Germany), stained with uranyl acetate and lead citrate, and examined using a Tecnai12 transmission electron microscope (Phillips, Amsterdam, the Netherlands).

Cultivation of uveal melanoma

The cells cultured as MCTS grew as large aggregations involving the majority of the cells in the well (Figure 1A and insets). The MCTS were mitotically active, as seen with the positive Ki67 staining with a score of 1%, 2%, and 4% for donors D1, D2, and D3, respectively (Figure 1C). The melanoma profile of the MCTS was verified by staining for a-melanoma, a marker that recognizes HMB-45, MART-1, and tyrosinase. More than 90% of the cells in the MCTS-derived paraffin sections stained positive for this marker (Figure 1D).

Genetic clustering is determined by the culture conditions

The gene expression profiles of the UMs (D1, D2, and D3), uncultured, cultured as MCTS, or cultured as adherent primary cells, were comprehensively analyzed with microarray analysis. PCA was performed on raw data from the microarray with a false discovery rate (FDR) of 10%. This type of analysis clusters the samples and represents them on a three-dimensional space based on the differential relative gene expression. The PCA plot shows that the clustering was mainly determined by the culture conditions (Figure 2A).

The relative gene expression of UMs (uncultured, cultured as MCTS, or cultured as adherent primary tumor cells) was further investigated by performing an unsupervised hierarchical clustering with an FDR of 10% presented as a heat map (Figure 2B). The heat map shows significant downregulation of the surface markers in the cultured cells compared to the uncultured primary tumor biopsy. These markers reflect the cellular heterogeneity of the primary tumor and the loss of macrophages (CD68), endothelial cells (von Willebrand factor), and T-cells (CD3D, CD8A, and CD2) in the cell cultures (Table 1). Additionally, there is a marked downregulation of human leukocyte antigen (HLA) expression in MCTS (Table 1) and in the adherent primary tumor cells (Appendix 1). This finding is in accordance with the work of van Essen et al. who showed downregulation of HLA expression upon loss of tumor-infiltrating leukocytes [29].

The genes found to be upregulated in the unsupervised hierarchical clustering (Figure 2B) were in concordance with many of the genes found in the RP analysis (Table 1 and supplementary data). The RP analysis (q≤0.05) resulted in 206 genes ≥2 fold upregulated and 373 genes ≥2 fold downregulated in MCTS versus uncultured tumor biopsies. Two hundred eighteen genes were found to be ≥2 fold upregulated, and 552 genes were ≥2 fold downregulated in adherent cell cultures versus the uncultured tumor biopsies. Sixty-four genes were found to be ≥2 fold upregulated, and 71 genes were ≥2 fold downregulated in adherent cell cultures versus the MCTS.

The genes from the RP analysis were further analyzed with Ingenuity IPA software. The differences in molecular and cellular functions between the various culture conditions are shown in Figure 3.

There was a noticeable increase in the cellular strain in the MCTS compared to the uncultured tumor biopsies, indicated by increased free radical scavenging, enhanced drug metabolism, and the increase in lipid metabolism in the MCTS versus adherent cells and uncultured tumor biopsies. Associated pathways and molecules in lipid metabolism in the MCTS versus uncultured tumor biopsies are shown in Figure 3. Alterations in the lipid metabolism include seven networks: synthesis of lipids, steroid metabolism, metabolism of cholesterol, metabolism of lipid membrane derivatives, synthesis of cholesterol, and conversion of lipid and fatty acid metabolism.

MCTS display a genetic profile indicating EMT and anoikis resistance

Anoikis is a form of apoptosis induced by loss or inappropriate cell adhesion [30]. The process of epithelial-to-mesenchymal-transition (EMT) is considered an important feature of anoikis [31]. Rank product data revealed 3.5-fold upregulation of snail family transcriptional repressor 2 (SNAI2; Gene ID: 6591, OMIM 602150) and 0.6-fold downregulation of cadherin 1 (CDH1; Gene ID: 999, OMIM 192090; E-cadherin) in MCTS (Table 1). Anoikis resistance is also supported by the upregulation of pyruvate dehydrogenase kinase 4 (PDK4; Gene ID: 5166, OMIM 602527), an enzyme that inactivates pyruvate dehydrogenase (PDH), which is required for the conversion of pyruvate to acetyl-CoA. PDK4 is upregulated in response to loss of adherence (LOA) and reduces reactive oxygen species (ROS) strain [32]. Noticeably, there was strong upregulation of angiopoietin like 4 (ANGPTL4; Gene ID: 51129, OMIM 605910) in the MCTS (Table 1, Figure 4). ANGPTL4 has recently been shown to be associated with an angiogenic phenotype of UM, and thus being involved in metastatic spread [33]. ANGPTL4 is thought to contribute to anoikis resistance by inducing conformational changes that enable resistance to inducers of apoptosis [34,35]. ANGPTL4 is further known to stimulate intracellular lipolysis, thus supplying substrate for fatty acid oxidation (FAO) [35]. Upregulation of FAO in MCTS is indicated by upregulation of enoyl-CoA hydratase 1 (ECH1; Gene ID: 1891, OMIM 600696), peroxisomal D3,D2-enoyl-CoA isomerase (PECI; Gene ID: 10455, OMIM 608024), and acyl-CoA thioesterase 1(ACOT1, Gene ID: 641371, OMIM 614313; Table 1). FAO has been proven to be an advantageous metabolic trait for cancer cells and is linked to anoikis resistance [36].

MCTS culture conditions induce a metabolic shift toward a lipogenic profile

Microarray results indicated a metabolic shift toward a lipogenic profile in the MCTS. A high content of lipid droplets (LDs) and stored-cholesterol ester is strongly associated with tumor aggressiveness [37,38]. As shown in Table 1, ACOT1 (lipid metabolism), fatty acid desaturase 1 (FADS1, Gene ID: 3992, OMIM 606148; biosynthesis of unsaturated fatty acids), sterol-C4-mehtyl oxidase-like (SC4MOL, Gene ID: 6307, OMIM 607545), 7-dehydrocholesterol reductase (DHCR7, Gene ID: 1717, OMIM 602858), lanosterol synthase (LSS, Gene ID: 4047, OMIM 600909; cholesterol biosynthesis), and oxysterol binding protein like 9 (OSBPL9, Gene ID: 114883, OMIM 606737; intracellular lipid receptor) all showed marked upregulation in the MCTS cultures compared to primary tumors. SC4MOL, LSS, and FAD1 were also found to be upregulated in the MCTS cultures compared to the adherent cultures. The microarray results also demonstrated increased lipid storage by upregulation of perilipin 2 (PLIN2, Gene ID: 123, OMIM 103195). PLIN2 belongs to the perilipin family, members of which coat intracellular lipid storage droplets [39]. The presence of PLIN2 was verified with IHC of donors D1, D2, and D3 (Figure 4). Morphological examination with transmission electron microscopy (TEM) revealed numerous lipid droplets in the supplementary donor D(S) cultured as MCTS (Figure 5). The TEM images also showed numerous mitochondria.

MCTS cultures increase the expression of cancer and testis antigens

The synovial sarcoma X breakpoint (SSX, Gene ID: 6759, OMIM 300326) gene family consists of nine highly homologous members (SSX1–9) [40]. SSX expression is confined to the testis, placenta, at low levels in the thyroid, and in a wide range of tumors (including synovial sarcoma), thus making them interesting targets for cancer therapy [41]. SSXs have been linked to EMT and anoikis resistance [42].

The microarray results showed an increase in the expression of SSX4 in MCTS versus primary tumors and adherent cultures (Table 1). This presence of SSX4 mRNA was verified with RNAscope, while the SSX4 protein was verified with IHC staining. The proportion of cells expressing SSX4 in primary tumors and MCTS was found (Figure 6). Noticeably, the SSX protein was minimally expressed in the tumor biopsies.