SEARCH

SEARCH BY CITATION

Keywords:

  • Melanoma;
  • Stem cell;
  • Cell line;
  • Gene expression profiling

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. SUMMARY
  7. Acknowledgements
  8. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  9. REFERENCES
  10. Supporting Information

The ability of cell lines to accurately represent cancer is a major concern in preclinical research. Culture of glioma cells as neurospheres in stem cell media (SCM) has been shown to better represent the genotype and phenotype of primary glioblastoma in comparison to serum cell lines. Despite the use of neurosphere-like models of many malignancies, there has been no robust analysis of whether other cancers benefit from a more representative phenotype and genotype when cultured in SCM. We analyzed the growth properties, transcriptional profile, and genotype of melanoma cells grown de novo in SCM, as while melanocytes share a common precursor with neural cells, melanoma frequently demonstrates divergent behavior in cancer stem cell assays. SCM culture of melanoma cells induced a neural lineage gene expression profile that was not representative of matched patient tissue samples and which could be induced in serum cell lines by switching them into SCM. There was no enrichment for expression of putative melanoma stem cell markers, but the SCM expression profile did overlap significantly with that of SCM cultures of glioma, suggesting that the observed phenotype is media-specific rather than melanoma-specific. Xenografts derived from either culture condition provided the best representation of melanoma in situ. Finally, SCM culture of melanoma did not prevent ongoing acquisition of DNA copy number abnormalities. In conclusion, SCM culture of melanoma does not provide a better representation of the phenotype or genotype of metastatic melanoma, and the resulting neural bias could potentially confound therapeutic target identification. STEM CELLS 2012; 30:336–343.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. SUMMARY
  7. Acknowledgements
  8. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  9. REFERENCES
  10. Supporting Information

Brain tumor stem cells were first identified using a serum-free suspension culture method featuring supplementation with basic fibroblast growth factor (bFGF) and epidermal growth factor (EGF) [1] originally used to isolate adult neural stem cells as “neurospheres” [2]. Neurosphere culture has been shown to better preserve the genotype and phenotype of patient glioblastoma samples in comparison with adherent serum cultures, which acquired additional mutations and displayed a differentiated phenotype poorly representative of primary glioblastoma [3].

Culture using serum-free EGF/bFGF-supplemented media has become a common method to enrich for stem-like cells from many tumor types [4–7]. The original neural-specific media is now used as a general stem cell media (SCM), with little apparent consideration as to whether or not the factors that support neural tissue growth might have different effects on other cell lineages. Sphere culture has been used in studies examining stem cell characteristics and melanoma heterogeneity [8–10]. While they present no conclusive evidence that SCM enriches for tumorigenic melanoma cells, there is interest in the potential for SCM culture to provide a more representative model for melanoma [11] and other cancers. We cultured melanoma cells de novo in a consensus SCM and a standard serum-containing media and evaluated cellular function, gene expression profiles, and DNA copy number variation in reference to the original patient samples.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. SUMMARY
  7. Acknowledgements
  8. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  9. REFERENCES
  10. Supporting Information

Cell Culture

Tissue donors consented for tissue collection and protocols were approved by the Austin Health Human Research Ethics Committee, Melbourne, Australia. Table 1 contains clinical characteristics of samples. Tissue was mechanically dissociated and digested in Royal Park Memorial Institute (RPMI) 1640 media (Invitrogen, Mulgrave, Australia, www.invitrogen.com) using collagenase-IV and DNase. SCM consists of Dulbecco's modified Eagle's medium/F12, 1xB-27 minus Vit-A, 2 mM Glutamax (all Invitrogen), 50 ng/ml FGF and EGF (Biovision, Mountain View, CA, www.biovision.com), and antibiotics. RF10 contains RPMI 1640 (Invitrogen), 10% fetal calf serum (CSL, Melbourne, Australia, www.csl.com.au), and antibiotics. Nonadherent growth conditions used Corning (Lowell, MA, www.corning.com). Ultra-Low Attachment surfaces. Thrombin/plasma cell clots were formalin-fixed and paraffin-embedded for sectioning. Proliferation assays were performed as previously described [12] using 2,000 cells per well.

Table 1. Tissue sample/cell line characteristics
inline image

Nonobese Diabetic/Severe Combined Immunodeficiency (NOD/SCID) Xenograft Assay

A total of 8 × 105 cells in phosphate-buffered saline were injected subcutaneously into the flanks of NOD/SCID mice. Protocols were approved by the Austin Health Animal Ethics Committee, Melbourne, Australia.

Microarray Analysis

Nucleic acids were extracted using the AllPrep Mini Kit (Qiagen, Doncaster, Australia, www.qiagen.com) from cell pellets and snap-frozen tumor pieces. RNA was analyzed using Affymetrix HG-U133 Plus 2.0 arrays at Memorial Sloan-Kettering Cancer Center Genomics Core Laboratory (MSKCC-GCL). Raw data were imported into Partek Genomics Suite (PGS; Partek, St. Louis, MO) after frozen robust multiarray average normalization [13]. Differential expression was determined by two-way analysis of variance with a twofold change cut-off and false discovery rate (FDR) of 5% (gene lists in Supporting Information File 1). Gene set enrichment analysis (GSEA) [14] was performed with gene set permutation and 5% FDR (gene set lists in Supporting Information File 2). Gene sets from the MSigDB database (v3.0, categories: C2(all) and C5(gene ontology [GO] BP)) and additional publications were included [3, 15–17]. DNA was analyzed on Illumina Human610-Quad genotyping arrays at MSKCC-GCL and imported into PGS. Segmentation algorithm settings were: minimum markers, 15; p value, .0001; expected range, 0.5; amplification signal-to-noise ratio, 0.4; deletion signal-to-noise ratio, 0.8.

Quantitative Real-Time Polymerase Chain Reaction

cDNA was generated using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Mulgrave, Australia, www.appliedbiosystems.com.au). Quantitative polymerase chain reaction (qPCR) was performed using the QuantiFast SYBR Green PCR Kit (Qiagen). Primer sequences are available in Table 2.

Table 2. Quantitative polymerase chain reaction primer sequences
inline image

Immunohistochemistry

Immunohistochemistry (IHC) used the Dako Envision+ system for: SOX2 (Clone D6D9, 600 ng/ml; Cell Signaling Technology, Danvers, MA, www.cellsignal.com) and MelanA (Clone A103, 2.67 μg/ml; Santa Cruz Biotechnology, Santa Cruz, CA, www.scbt.com). Citrate buffer (pH 6.0) was used for antigen retrieval, with 3-amino-9-ethylcarbazole as chromogen. Slides were scanned on a ScanScope XT and exported from ImageScope v10 (Aperio, Vista, CA, www.aperio.com).

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. SUMMARY
  7. Acknowledgements
  8. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  9. REFERENCES
  10. Supporting Information

Melanoma Cells Cultured in SCM and RF10 Display Similar Growth Characteristics

RF10 lines grew as adherent monolayers, while those cultured in SCM grew as a mixture of adherent and floating cells, although occasionally solely adherent or solely as free-floating spheres. In vitro (Fig. 1A) and xenograft growth rates (Fig. 1B) demonstrated no consistent growth advantage for cells from either media.

thumbnail image

Figure 1. In vitro and in vivo characterization of RF10 and SCM melanoma cell lines. (A): Proliferation of paired cell lines grown in RF10 and SCM. Cell numbers were quantified every second day by measuring the absorbance of reduced MTS reagent secreted into the culture media. The assay was performed three times using independent cultures with triplicate measurements taken for each replicate. Data are represented as mean ± SEM. (B): Growth curves of xenograft tumors. A total of 8 × 105 cells were injected subcutaneously in to NOD/SCID mice. One mouse injected with LM-MEL-69 RF10 did not exhibit visible tumor growth during the experimental period; however, a small tumor was excised upon autopsy. Despite extensive variation in growth rates, all cell lines yielded a 100% tumor take rate. Cell dose was not titrated, and a relatively high number of cells were injected, as we aimed to characterize the tumors and not determine tumor initiating cell frequencies. All tumors remained localized subcutaneously without visible local invasion or infiltration by host cells upon histological examination. (C): Immunohistochemical staining for MelanA on sections from cell lines, xenograft tumors, and parental metastatic melanoma tissue samples. Scale bars = 100 μm; insets are a ×2 magnification of a portion of the full image. Abbreviations: MTS, 3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium; NOD/SCID, nonobese diabetic/severe combined immunodeficiency; SCM, stem cell media; Xeno., xenograft.

Download figure to PowerPoint

Expression Levels of Melanocyte Lineage Molecules Are Independent of Culture Conditions

MelanA staining patterns by IHC did not vary consistently between culture conditions but were distinct between cells from different patients (Fig. 1C). Growth as xenografts established expression levels more similar to the original tumor samples. These findings are representative of those for tyrosinase and microphthalmia-associated transcription factor (data not shown).

SCM Induces Expression of Neural Lineage Genes but Not Putative Melanoma Stem Cell Markers

Gene expression profiling of SCM cultures found overexpression of embryonic and neural lineage stem cell regulators such as SOX2 and FOXD3 and other neural lineage genes. GSEA identified association with: SCM cultures of melanoma [16], neural lineage GO categories, melanoma cells expressing high levels of differentiation markers [17], genes downregulated in melanoma metastasis [18], and glioma neurosphere profiling [3]. The similarity in expression profile of melanoma and glioma cells grown in SCM suggests that our SCM-induced profile is media-specific rather than melanoma-specific. SCM culture did not enrich for markers of stem-like melanoma cells [9, 19, 20] or other stem cell markers by microarray or qPCR (Fig. 2A). Switching cells between RF10 and SCM did not result in appreciable cell death and reversibly induced or diminished the expression of SCM-associated genes (Fig. 2B), which does not support the idea that SCM selects for an existing subpopulation of stem-like melanoma cells from patient samples. It suggests that most melanoma cells have the potential to adopt different phenotypes, which supports a recent report that melanoma cells appear to possess nonhierarchically organized, phenotypic plasticity allowing for reversible changes in surface antigen heterogeneity from purified populations in vivo [21]. Comparing cell lines grown in SCM as either monolayers or free-floating spheres yielded very few differentially expressed genes (Supporting Information File 1, Supporting Information Fig. 1).

thumbnail image

Figure 2. Gene expression in SCM and RF10 melanoma cell lines, xenografts, and patient tissue samples. (A): Quantitative polymerase chain reaction (PCR) analysis of stem cell markers in RF10 and SCM cell lines and xenografts. Data are presented as fold change calculated by dividing SCM expression levels by RF10 expression levels, each normalized to β-actin. No amplification of ABCB5 was observed for LM-MEL-82 SCM so a fold change was not calculated. The expression of ABCB5 in LM-MEL-69 SCM was 127-fold lower than LM-MEL-69 RF10. (B): Quantitative PCR analysis of SCM profile genes in cells switched between SCM and RF10. Cell lines were changed into the opposite media and passaged at least two times over several weeks before RNA extraction. Fold change is in reference to gene expression measurements taken prior to media switching. The expression of OLIG2 was downregulated 707-fold when LM-MEL-69 SCM was switched in to RF10 media. (C): Hierarchical clustering of SCM and RF10 melanoma cell lines, xenografts, and matched patient tissue samples with genes differentially expressed between SCM and RF10 cultures. Hierarchical clustering was performed using Pearson's dissimilarity, and average linkage on genes differentially expressed between SCM and RF10 cultures for each cell line pair individually (Supporting Information File 1). Green and red represent genes with expression above and below media, respectively. Replicate arrays are from independent cultures or separate regions of tissue samples. Microarray data have been deposited at ArrayExpress (http://www.ebi.ac.uk/arrayexpress) with accession number E-MEXP-3290. Abbreviation: SCM, stem cell media.

Download figure to PowerPoint

Genes Associated with Melanoma Metastasis and Poor Outcome Are Overexpressed in RF10 Cultures

RF10 cultures overexpressed ABCG2, and transforming growth factor β pathway components like TGFB2 and inhibitor of DNA binding (ID) family members. GSEA identified a gene set predictive of metastasis development in melanoma [22] and an invasive phenotype gene set derived from undifferentiated melanoma cells [17]. Additionally, gene sets related to DNA damage, repair, and proliferation were observed, and both are associated with adverse outcome in melanoma [23, 24]. In contrast to the SCM profile, there was no significant association between the RF10 profile and that of adherent glioma cell lines [3].

SCM Induces an Expression Profile Dissimilar to Patient Tumors

Hierarchical clustering was undertaken using genes differentially expressed between SCM and RF10 cultures (Fig. 2C). SCM cultures did not cluster with matched patient samples and demonstrated overexpression of genes that were not significantly expressed in RF10 cultures, xenografts, or tumor samples. The embryonic and neural stem cell transcription factor SOX2 was chosen as a representative marker of the SCM profile for IHC (Fig. 3). SOX2 is reported to be expressed in a subset of melanomas [25, 26]; however, we observed strong protein expression in all SCM cultures, including one derived from a SOX2 negative tumor.

thumbnail image

Figure 3. Immunohistochemistry for SOX2 in patient biopsies, SCM and RF10 cell lines, and xenografts. Scale bar = 100 μm; insets are a ×2 magnification of a portion of the full image. Bright red staining indicates SOX2 positivity. Dark brown cytoplasmic coloring is melanin. Tissue sections related to LM-MEL-69 and LM-MEL-82 showed rare SOX2 expressing cells, while the SCM cell lines and derived xenografts had a very high proportion of SOX2 positive nuclei. The RF10 cell lines derived from LM-MEL-69 and LM-MEL-82 showed infrequent cytoplasmic SOX2 staining; however, the RF10 xenografts displayed a SOX2 expression pattern more similar to the tissue samples. The tissue sample, RF10 cell line, and all xenografts for LM-MEL-62 were negative for SOX2 expression, while the SCM cell line displayed frequent SOX2 expression and nuclear localization. Abbreviations: SCM, stem cell media; Xeno., xenograft.

Download figure to PowerPoint

Xenograft Expression Profiles Are More Representative of Patient Melanomas

Comparing SCM and RF10 xenografts as a group yielded few differentially expressed genes, and the prominent SCM-induced neural expression profile was no longer evident. Hierarchical clustering revealed a clear association between xenografts and melanoma tissue samples, while the in vitro cultures comprised distinct clusters (Fig. 4), again fitting with reports that melanoma cells display plasticity in recapitulating patient tumor phenotypes in xenograft models regardless of the phenotype of the injected cells [21].

thumbnail image

Figure 4. Hierarchical clustering of SCM and RF10 melanoma cell lines, xenografts, and patient tissue samples with genes differentially expressed between in vitro and in vivo samples. Cell lines are identified by colored boxes to the left of the heat map. Hierarchical clustering was performed using Pearson's correlation and average linkage. Green and red areas of the heat map represent genes with expression above and below median, respectively. Replicate arrays were derived from independent samples. This clustering used genes differentially expressed between all cell line samples and all xenograft tumors (Supporting Information File 1), producing two main groupings. One contains all in vitro cultures from LM-MEL-59, LM-MEL-69, and LM-MEL-82. The other consists of xenografts and patient tissue samples, with a subgroup containing all material from LM-MEL-62 in which the same segregation of in vitro and in vivo samples was observed. The separation of all LM-MEL-62 material likely reflects intrinsic differences between this tumor and the others analyzed. The strong effect of patient origin can also be observed in samples related to LM-MEL-69 and LM-MEL-82, with the patient sample clustering adjacent to the matched xenografts as opposed to the larger tissue sample cluster. A panel of additional metastatic melanoma tissue samples was included, comprising tumors resected from lymph nodes, brain, skin, and liver. Abbreviation: SCM, stem cell media.

Download figure to PowerPoint

Copy Number Aberrations Accumulate in All Growth Conditions

Hierarchical clustering of DNA copy number data did not segregate cell lines from either condition with the matched patient tissue samples (Fig. 5A), suggesting no preferential preservation of copy number profile in either culture condition. Cell line pairs cultured for several months and xenografts showed a clear increase in copy number changes with prolonged growth period (Fig. 5B). No difference between culture conditions was observed when the genotyping data was analyzed for loss of heterozygosity (Supporting Information Fig. 2).

thumbnail image

Figure 5. DNA CN profiles of SCM and RF10 melanoma cell lines, xenografts, and matched patient tissue samples. (A): Hierarchical clustering and heat map of CN data. Amplifications are shown in green, deletions in red. Clustering was performed using Euclidian distance and average linkage. CN values were generated by comparing probe intensities from melanoma-derived tissues to intensities from matched patient blood. (B): Number of CN aberrations in patient samples, cell lines, and xenografts. Regions of significant CN change were defined using a segmentation algorithm. Genotyping array data have been deposited at ArrayExpress (http://www.ebi.ac.uk/arrayexpress) with accession number E-MTAB-741. Abbreviations: CN, copy number; SCM, stem cell media.

Download figure to PowerPoint

SUMMARY

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. SUMMARY
  7. Acknowledgements
  8. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  9. REFERENCES
  10. Supporting Information

Our results do not support SCM cultures of melanoma as a more stable or representative model system, as they consistently expressed a transcriptome that was less similar to patient tissue samples than RF10 cultures, and all samples accumulated copy number changes. We suggest that plasticity of melanoma cells allows for the homogeneous induction of a neural lineage expression profile in SCM cultures, which could seriously confound therapeutic target selection. The commonly used SCM culture method was developed specifically to support the growth of neural stem cells, and results similar to the ones in this study might potentially be observed with cancers from other lineages.

Acknowledgements

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. SUMMARY
  7. Acknowledgements
  8. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  9. REFERENCES
  10. Supporting Information

This research was supported by the Melanoma Research Alliance and Cancer Council Victoria. J.C. and I.D.D. are supported by Australian National Health and Medical Research Council Practitioner Fellowships. We apologize to colleagues whose work we could not cite or discuss in detail due to space constraints.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. SUMMARY
  7. Acknowledgements
  8. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  9. REFERENCES
  10. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. SUMMARY
  7. Acknowledgements
  8. DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
  9. REFERENCES
  10. Supporting Information

Additional Supporting Information may be found in the online version of this article.

FilenameFormatSizeDescription
STEM_786_sm_supplFigure1.pdf115KSupplementary Figure 1: Quantitative PCR analysis of SCM profile genes in SCM monolayer and sphere cultures Data are presented as fold change calculated by dividing sphere expression levels by monolayer expression levels, each normalized to beta-actin.
STEM_786_sm_supplFigure2.pdf152KSupplementary Figure 2: Loss of heterozygosity in SCM and RF10 melanoma cell lines, xenografts, and matched patient tissue samples Gray regions represent LOH, determined by comparing SNPs called heterozygous in normal tissue (blood) controls to the corresponding cell line and tumor tissue.
STEM_786_sm_supplTable1.xls335KSupplementary Table 1: Differentially expressed gene lists
STEM_786_sm_supplTable2.xls58KSupplementary Table 2: Gene Set Enrichment Analysis for SCM and RF10 melanoma cell lines.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.