Drug-selected population in melanoma A2058 cells as melanoma stem-like cells retained angiogenic features – the potential roles of heparan-sulfate binding ANGPTL4 protein

Selection and characterization of drug-selected subpopulation in melanoma cells

We cultured different melanoma cells under detachment-impaired dishes, which enabled suspension culture and cell sphere formation. As seen in Figure 1A, most of the suspended melanoma cells formed irregular aggregates except Hs695t cells. For melanoma A375 cells, suspended cells accumulated but cell boundary remained distinguishable. Melanoma A2058 cells could partially form larger cell spheres, which implied the presence of cancer stem-like cells. It was known that stem-like subpopulation of tumor cells would be drug-resistant [12–14], so that we tried to enrich them by drug selection.

We treated melanoma A2058 cells using either one of different therapeutic agents with partial response in clinical trials. Sorafenib is the multikinase inhibitor that had been used to inhibit tumor cell proliferation. Sorafenib has been evaluated as a single therapy agent as well in combination with various chemotherapeutical drugs in several clinical trials [15, 16]. Carmustine is one of alkylation agents to interfere DNA replication and RNA transcription. It had been included as one component in Dartmouth regimen (carmustine, cisplatin, dacarbazine, and tamoxifen) in melanoma therapies [17–19]. Upon treatment with 20 μM sorafenib for 3 days, the cell proliferation was reduced as observed in significantly low cell numbers. Treatment with 20 μM carmustine resulted in significant cell death as seen with detached cells. Further suspension culture of A2058 cells after sorafenib treatment didn’t exhibit the characteristic cell spheres, while those after carmustine treatment retained the formation of cell sphere (data not shown).

To enrich these drug-selected subpopulation of melanoma A2058 cells, we sequentially and repeatedly selected by several cycles of carmustine treatments. Repeated cycles of selection, restoration, and cell amplification were done for several months, and the concentrations of carmustine used for selection were gradually increased from 20 μM to 100 μM. The drug-selected cells were less elongated (Figure 1B) and the invasion ability were largely-reduced than parental cells (Figure 1C). It suggested that these drug-selected cells were low-invasive. As seen in Figure 1D, drug-selected cells proliferated slower comparing with parental melanoma cells. Slower proliferation and less invasive ability might suggest the drug-selected subpopulations were less-malignant in cell studies. Since we suspected the selected subpopulation was melanoma-stem like cells, further characterization at the ability of sphere formation, content of side population, and xenograft tumor formation were performed.

In vitro sphere formation was used to demonstrate the activity of cancer stem cells. As seen in Figure 2A, suspension culture of parental cells mostly generated cell aggregates but those of drug-selected cells produced cell spheres. We had developed one novel approach, “spherocrit assay”, to quantify the level of sphere formation [20]. The number and the size of cell spheres both reflected their sphere forming abilities. Rigid and compacted spheres were collected into microcapillary tubes. The length of sphere column reflected the sphere forming ability. As seen in Figure 2A, drug-selected cells had longer sphere columns than parental cells did. Comparing the sphere column length generated from drug-selected cells after selection by 5, 6, or 7 cycles, they all had similar sphere-forming ability and were higher than that of parental A2058 cells (Figure 2B). This indicated that the period of long-term and repeated selection for 5 months was sufficient to enrich melanoma stem-like cells.

Next, cell fractions of side population were analyzed by Hoechst33342-exclusion assay on flow cytometer (Figure 2C). The content of side population in drug-selected cells was 3.32 % comparing with 0.81% in parental melanoma cells, which reinforced the enrichment of cancer-stem like cells after repeated selections. In addition, critical feature of cancer stem cells was the ability to form tumors by small number of cells. We performed xenograft tumor formation experiments in nude mice by inoculation of 2 x 10³ cells subcutaneously. As shown in Figure 2D, 4 out of 5 mice developed subcutaneous tumors within 3 months, and drug-selected cells generated larger tumors than parental cells did. Those data indicated this drug-selected subpopulation of melanoma cells were potential melanoma stem-like cells, which had features of slower proliferation, less cell invasion, higher sphere-forming ability, more side population cells, and higher tumor malignancy in vivo.

Melanoma-stem marker expression and altered protein phosphorylation in drug-selected melanoma cells

We also analyzed the expression of known melanoma-stem markers and potential signaling pathways. As seen in Figure 3, expression levels of several melanoma stem cell markers (NGFR, NES, ABCB5, BMI1, CXCR6, CD13, CD90, CD133, CD44, ABCG2, EpCAM, and CD20 [21–25]) were analyzed in parental or drug-selected cells. Of them, mRNA expression levels of some markers were higher but other markers were lower in drug-selected subpopulations than in those of parental cells as determined by qPCR. Further evaluation by western blot analysis also confirmed the increased expression of CD13, CD24, Ki67, and Sox2, but decreased CD133, ALCAM, and c-myc in drug-selected subpopulation (Figure 3). This suggested the melanoma stem-like characteristics in drug-selected cells as CD13⁺CD24⁺.

In addition, we analyzed the pathway activations for different receptor tyrosine kinases using antibody array. As seen in Figure 4A, the levels of several kinase phosphorylation were different between parental and drug-selected cells according to the dot intensities shown on the blot membranes. Of them, the phosphorylation of Akt-Thr308, Akt-Ser473, Erk1/2, S6 kinase, and Src were less in drug-selected cells. Only the phosphorylation of fibroblast growth factor receptor 3 (FGFR3) was higher in drug-selected cells.

The expression patterns of phosphor-tyrosine proteins were also compared by western blot. As seen in Figure 4B, the significant decreases in peak heights for drug- selected cells were seen at those around 44 and 60 kDa (indcaited by arrows). The peak at 44kDa corresponded to the phosphor-ERK (42/44 kDa). The peak at 60 kDa corresponded to the phosphoAkt (60kDa) or phosphor-Src (60kDa). These were consistant with the results from antibody array experiments (Figure 4A). The decrease in phospho-S6kinase (70/85 kDa) and increase in phospho-FGFR3 (125~165 kDa) were not significantly observed.

Molecular profiling of drug-selected melanoma cells showed high expression of ANGPTL4

In order to evaluate the mechanism why drug-selected cells were more malignant but the cell proliferation/cell invasion ability was less malignant, the gene expression profiling in parental and drug-selected melanoma cells were done using cDNA IlluminaBbeAdarray. Within all the 47,323 probes, there were 2,948 targets (2,445 targets upregulated; 503 targets downregulated) got changed with statistical significance. The microarray results were deposited at Gene Expression Omnibus with accession number GSE135689. These targets were analyzed by DAVID (Database for Annotation, Visualization and Integrated Discovery) v6.7 for functional gene clustering in different gene ontologies (biological process, cellular component, and molecular function). The biological processes perturbed by selection of drug-resistance were associated with angiogenesis, cell migration, survival, cell adhesion, which account for tumor malignancy. Stimulus responding, inflammation, and morphogenesis might associate with the feature of cancer stemness. In addition, almost all the changed genes were located at extracellular or plasma membrane regions, which explained the aforementioned change in biological processes. As to the aspects of molecular functions, they implied the receptor dimerization, as well as the glycosaminoglycan-binding, and the following signaling events would be dominant in drug-selected cells.

By the way, in the results of microarray profiling, NGFR, CD13, CD90, EpCAM, and Ki67 were significantly increased in drug-selected cell, while those of CD133 was downregulated. These were consistent with the aforementioned qPCR analysis (Figure 3). Several ABC family proteins, such as ABCA8, ABCC3, ABCA1, ABCG1, and ABCA9, were upregulated in drug- selected cells as seen in microarray data. As to epithelial-mesenchymal transition proteins, the expression levels of vimentin, E-cadherin, and N-cadherin were not significantly changed.

In order to specify the potential targets, we performed GSEA (gene set enrichment analysis) and leading edge analysis in the microarray results. It showed 187 upregulated and 1 downregulated gene sets out of 188 oncogenic signature gene sets (C6 collection in MSigDB v4.0) with significantly enrichment at nominal p-value < 1% upon drug-resistance. The leading edge analysis revealed ANGPTL4 and IL8 gene were mostly potential associated with oncogenic features upon selection of drug-resistance (Figure 5A). The microarray results also indicated ~50-folded and ~500-folded increased in ANGPTL4 and IL8 expression, respectively. The qPCR analysis further confirmed the increased expression in ANGPTL4 and IL8 gene expression in drug- selected cells.

It was known the increased expression of IL-8 protein correlated with melanoma malignancy. IL-8 protein in melanoma cells was recognized as autocrine [26] and its expression level was correlated with metastatic potential [27, 28]. IL-8 would stimulate multiple cell signaling pathways, including PI3K-AKT/PKC/MEK-ERK pathways in melanoma cells [29]. However, the ERK phosphorylation in drug-selected cells was decreased. The ANGPTL4 proteins belong to a superfamily of angiogenic proteins involved in lipid metabolism, energy homoeostasis, inflammation, cell differentiation, angiogenesis, or tumorigenesis, [30]. Previous literature indicated the expression of ANGPTL4 led to reduced ERK phosphorylation which was mediated by C-terminal domain [31]. In our selected subpopulation with high ANGPTL4 expression, we did see reduced ERK phosphorylation comparing with that of parental melanoma cells (Figure 4A and 4B). These could explain why drug-selected cells had higher tumorigenic potential but lower ERK phosphorylation. The role of ANGPTL4 protein in drug-selected melanoma cells might be important.

Upregulation of ANGPTL4 in drug-selected cells enabled their tube formation in vitro through sulfated proteoglycan

The expression level of ANGPTL4 was validated in parental or drug-selected cells. As seen in Figure 5A, qPCR analysis showed 5-folded increase in mRNA level for drug-selected cells. Cell surface expression level was higher in drug-selected cells as examined by flow cytometry (Figure 5A). Western blot showed increase ANGPTL4 protein bands at 45 and 54 kDa (glycosylated form), which confirmed the increased expression of ANGPTL4 gene. Another results from ELISA assay (data not shown here) also showed the higher secretion of ANGPTL4 protein which was consistent with the previous finding.

Since ANGPTL4 was known as angiogenic protein, we examined the in vitro angiogenic activity of parental and drug-selected cells by tube formation assay. As shown in Figure 5B, no in vitro tube was formed by parental cells. However, significant tubes were generated by drug- selected cells. In order to confirm that in vitro tube-forming was associated with ANGPTL4 protein secreted by drug-selected cells, we pretreated the cells with anti-ANGPTL4 antibody. As seen in Figure 5C, the angiogenic activity was indeed inhibited by anti-ANGPTL4 antibody which implied the in vitro tube formation was contributed by secreted ANGPTL4. In addition to higher expression of ANGPTL4 protein contributing higher angiogenic ability, we suspected some membrane receptor might also be responsible and was upregulated in drug-selected cells.

ANGPTL4 was previously recognized as potential glycosaminoglycan-binding proteins with preference toward heparan sulfate [32, 33]. We analyzed the expression level of syndecan-1 (SDC1), syndecan-2, and syndecan-4, as well as the biosynthetic enzymes associated with the levels of N-sulfate and 2-O-sulfate groups, in parental and drug-selected cells. N-sulfate was introduced by the enzymes of N-deacetylase/N-sulfotransferase (NDST gene products), which contains 4 different members. Addition of 2-O-sulfate group was achieved by solely heparan sulfate 2-O-sulfotransferase 1 (HS2ST1 gene product). In melanoma A2058 cells, only the expression of NDST1 gene was detected (data not shown). We analyzed the expression levels of these genes by qPCR. As seen in Figure 5D, SDC1 expression was higher (~2.5-folded) but SDC2 expression was less (~0.3-folded) in drug-selected cells. There was no significant difference in SDC4 expression between parental and drug-selected cells. The HS2ST1 gene expression was similar between parental and stem-like melanoma cells, but NDST1 gene expression was elevated at stem-like melanoma cells comparing with parental melanoma cells.

Experiment using biosynthesis inhibitors further conformed the importance of heparan sulfate in angiogenic ability of drug-selected cells. Sodium chlorate was used to abolish sulfate groups on glycosaminoglycan chains; while xylopyranoside was used to inhibit biosynthesis of glycosaminoglycan chain. As seen in Figure 5E, both sodium chlorate or xylopyranoside suppressed the level of in vitro tube formation. These results suggested membrane proteoglycan SDC1 and N-deacetylase/N-sulfotransferase-1 (NDST1) might associate with melanoma stemness, and its heparan sulfate chain might be the acceptors for ANGPTL4.

ANGPTL4 accounted for partial features of drug-selected subpopulation

Since high expression of ANGPTL4 was observed in drug-selected cells, we further investigated the roles of ANGPTL4 in cellular activities of drug-selected cell population. We suppressed the ANGPTL4 expression by transfection of specific shRNA against ANGPTL4 (shANGPTL4). Upon shANGPTL4 transfection, the mRNA expression and surface expression levels were all decreased as checked by qPCR and flow cytometry, respectively (Figure 6A). Since ANGPTL4 protein was known as angiogenetic factor and its higher expression correlated with tube forming ability (Figure 5B), we examined the effect of shANGPTL4 transfection on tube-forming ability of drug-selected cells. As shown in Figure 6B, suppression of ANGPTL4 expression did abolish tube formation, which reinforced the role of upregulated ANGPTL4 in promoting tumor malignancy.

Upon the suppression of ANGPTL4 expression, cell proliferation was slightly reduced (see Figure 6C, p-value > 0.05). It suggested that ANGPTL4 protein didn’t associate with the feature of slow proliferation in drug-selected cells. In addition, the shANGPTL4-transfected cells were more sensitive to carmustine as seen by less percentage of cell survival (Figure 6D). The effects of shANGPTL4 transfection on different cell activities was also examined. Sphere forming ability was significantly reduced upon shANGPTL4 transfection (Figure 6E). As seen in Figure 6E, the transwell migration ability of drug-selected melanoma cells remained less invasive as seen in Figure 1C. These suggested that ANGPTL4 protein determined the cell activities of tube formation, sphere formation, and drug-sensitivity.

We also examined the expression of melanoma stem-cell markers after suppression of ANGPTL4 expression. As shown in Figure 7A, reduced ANGPTL4 expression led to higher expression of NGFR, CD13, c-myc, SOX2, and lower expression of CD90, CD24, and ALCAM.

Comparing the results of melanoma stem-cell marker expressions in Figure 3 and Figure 7A, we suggested that high expression of ANGPTL4 in melanoma cells might associate with the CD90⁺/CD24⁺ phenotype. The expression patterns of phosphor-tyrosine proteins upon suppression of ANGPTL4 expression was investigated. As seen in Figure 7B, levels of phosphor-proteins with the molecular sizes of 44 and 24 kDa were increased by the suppression of ANGPTL4 expression. This suggested that reduced level of phospho-ERK in drug-selected melanoma cells was mediated by higher expression of ANGPTL4, which is consistent with previous observation [31]. The level of phosphor-protein of 24 kDa was not significantly different in drug-selected melanoma cells comparing that in parental melanoma cells. In summary, we suggested the higher ANGPTL4 expression in drug-selected melanoma cell subpopulation account for melanoma-stemness features of in vitro tube formation ability, sphere formation, drug-resistance, and reduced ERK phosphorylation.

Cell culture and selection of drug-selected cell populations in melanoma cells

Human melanoma A2058, A375, and Hs695t cells were maintained in culture dish (Corning Incorporated Life Sciences, Tewksbury, MA, USA) supplemented with DMEM medium containing 10 % (v/v) fetal bovine serum, 100 units/mL penicillin, and 100 μg/mL streptomycin at 37°C under 100% humidity.

To enrich carmustine-resistant melanoma cell subpopulation, melanoma A2058 cells were plated at density of 2x10⁶ cells for initial selection. The enrichment was done by repetitive cycles of chemoagent treatment, removal of chemoagent, and cell restoration/proliferation, which would be completed in one month. One enrichment cycle was completed till the optimal growth rate achieved with no cell detachment/death. Initially, melanoma cells were treated by 20 μM carmustine for 3 days and the concentrations of chemoagents used to select drug-resistant cells in each cycle were then gradually increased up to 100 uM. Totally 5 to 7 enrichment cycles were done to obtain drug-selected subpopulation of melanoma cells for the following characterizations. In order to compare in parallel, the parental A2058 cells were cultured and maintained along with enrichment procedure of 5-7 months.

ANGPTL4-specific shRNA (clone ID: TRCN0000056723) and control shRNA (shLuc, clone ID: TRCN0000072249) were purchased from the National RNAi Core Facility located at Institute of Molecular Biology/Genomic Research Center, Academia Sinica, Taiwan. Cells were transfected using TurboFect transfection reagent (Thermo Fisher Scientific Inc., Pittsburgh, Pennsylvania, USA) according to the manufacturer’s instruction. The transfected cells were selected and enriched under growth medium containing 5 μg/mL puromycin after shRNA transfection.

Cell proliferation, transwell migration assay, and cell survival assay

For cell proliferation assay, melanoma cells were collected by trypsinization and were counted by hemocytometer. The 2 x 10⁴ cells were plated onto 35 mm culture dish. The cells were incubated and grew for different day intervals. The cells were then trypsinized and the cell numbers were counted by hemocytometry.

For transwell migration assay, 1x10⁵ cells (in DMEM medium with 2% (v/v) FBS) were applied onto the culture insert (8 um pore size; SPL life sciences, Co. Ltd., Pocheon-si, Gyeonggi-do, Korea). The bottom wells contained 600 uL culture medium with 10% (v/v) FBS, and were left un-agitated in the cell incubator until observation. The migrated cells were visualized by staining with 0.2% (w/v) crystal violet.

For cell survival assay, 1x10⁴ cells were plated into each well of 96-well plate and incubated overnight. Different concentrations of chemoagents were added and incubated for 1 days in culture incubators. The cell viability was determined by Alamar Blue assay (Thermo Fisher Scientific Inc., Pittsburgh, PA, USA) to evaluate the mitochondria activity.

Sphere formation and spherocrit assay

For sphere formation, 1 x 10⁵ melanoma cells were seeded in 10-cm sterile plastic dish (Alpha-plus, Inc., TaoYuan, Taiwan) supplemented with 10 mL culture medium. The cell sphere were formed and enlarged for 2 weeks with medium renewal every 2-3 days.

One novel “spherocrit” assay was used to quantify the contents of cell spheres [20]. Cell aggregates or cell spheres cultured for 2-weeks were collected and processed as followed. In general, the cell spheres (rigid-packed cells) or cell aggregates (loose-packed cells) were pipetted to disrupt loose-packed cell aggregates. The rigid-packed cells were precipitated by gravity for 3 min and cell suspension was removed. The cell aggregates or cell spheres were then filled with 1 mL PBS and pipetted again. The procedure of collection, pipetting, precipitation, and suspension-removal was repeated for 3-times. The loose-packed cell aggregates were disrupted, while cell spheres remained intact. The remained cell spheres or cell aggregates were finally collected into microcapillary tube (VWR international, Radnor, PA, USA) and the bottom was stopped by toothstick connected with silicon-tube. The vertical-precipitated spheres were photographed and the lengths of cell columns were recorded, and the level of sphere formation was quantified [20].

Side population analysis by flow cytometry

Parental or drug-selected melanoma cells were dissociated with trypsin/EDTA, and pelleted by centrifugation. The cells were resuspended at 1x10⁶ cells/mL in culture medium. Hoechst33342 dye was added to a final concentration of 5 μg/mL and cells were incubated at 37°C for 90 mins. Propidium iodide was then added to final concentration of 5 μg/mL to label dead cells. The cells were filtered through 80 μm mesh (BD bioscience, Inc., San Jose, CA, USA) to obtain a single cell suspension before analysis. Analysis was performed on a MoFlo XDP (Beckman Coulter Taiwan Inc., Taipei City, Taiwan). The Hoechst33342 dye was excited at 355 nm and its fluorescence was dual-wavelength analyzed with emission for Hoechst blue at 445 nm, and Hoechst red at 650 nm. Side population fractions were selected by the cells with low Hoechst blue and low Hoechst red intensities as confirmed by verapamil pretreatment. The percentage was determined by Kaluza software (Beckman Coulter Taiwan Inc., Taipei City, Taiwan).