Uveal Melanoma Cell Lines Contain Stem-Like Cells That Self-Renew, Produce Differentiated Progeny, and Survive Chemotherapy

Abstract

Purpose.: Uveal melanoma (UM) cells in high-metastatic risk tumors have an undifferentiated molecular signature indicative of a more primitive cellular phenotype. Given mounting evidence for the existence of cancer stem cells (CSC), the authors investigated whether UM cell lines retain a population of self-renewing tumorigenic cells.

Methods.: Single-cell cloning and spheroid culture studies were used to study the presence of a CSC-like population in two cell lines derived from the primary tumor (Mel270) and metastatic liver lesion (Omm2.5) of the same patient.

Results.: Mel270 and Omm2.5 cells exhibited distinct clonal morphologies in adherent culture akin to holoclones, meroclones, and paraclones. Holoclones were large colonies of tightly packed small cells, which could be serially passaged (> 10 generations) to produce colonies of all three types; paraclones were small colonies of flattened cells that could be passaged for only one or two generations to produce further paraclones. Mel270 and Omm2.5 cells surviving cisplatin treatment produced significantly more holoclones than untreated cells (P < 0.05), suggesting enrichment for this CSC-like subpopulation. Mel270 and Omm2.5 cells also formed melanomaspheres (MS) when grown at clonal density in nonadherent culture. MS possessed self-renewal capacity to generate further MS and when replated to adherent culture yielded colonies of all three types. Mel270 and Omm2.5 holoclones and MS also demonstrated antigenic heterogeneity expressing markers associated with both a primitive migratory neural crest phenotype, and a more differentiated phenotype.

Conclusions.: These data suggest the presence in UM cell lines, of a CSC-like subpopulation with enhanced self-renewal and proliferative capabilities that could more appropriately model therapeutic efficacy.

Despite successful ocular treatment, almost 50% of uveal melanoma (UM) patients develop metastatic disease, usually involving the liver. We can reliably identify which patients will develop metastatic disease, according to: (1) the clinical stage of the ocular tumor (i.e., tumor dimensions and extent); (2) the histologic grade (i.e., cytomorphology and mitotic rate); and (3) chromosomal abnormalities and gene expression profiling (i.e., chromosome 3 loss and/or chromosome 8q gain; class 2 molecular signature). ^{1 –3} However, while we can identify individuals with a poor prognosis, chemotherapy rarely prolongs life and long-term survival is rare unless the hepatic metastases are resectable. ^{4 –6} There is a real clinical need, therefore, to identify biological disease modifiers and thus effective therapeutic targets, which when inhibited might suppress one or more steps in the metastatic cascade, thereby preventing disease progression. To achieve this, we need a better understanding of the cellular processes involved in UM development and progression. Recently, functional analysis of the genes expressed in class 1 and class 2 UMs revealed that class 1 tumors expressed genes associated with a more differentiated neural crest and/or melanocyte phenotype. ⁷ In contrast, the class 2 tumors expressed high levels of genes found in a primitive neural and/or ectodermal cell population. In a review by the same group, there is reference to unpublished data that class 2 tumors demonstrate more stem cell-like properties and are able to undergo multi-lineage differentiation. ⁸ This is consistent with the observation in many tumor types, including cutaneous melanoma, that a malignant subpopulation of cells exists, which initiate and maintain tumor growth—the so called “cancer stem cells” (CSC). 

The concept of CSCs has gained momentum over the last decade and many excellent reviews ^{9 –13} emphasize the important implications of such a model for understanding tumor biology and cancer treatment. In particular, it is thought that cancer treatments, such as chemotherapy, target rapidly-dividing tumor cells but any CSCs are not destroyed because of enhanced survival mechanisms, so that the tumor can then regrow. This may be especially relevant for UM because these tumors show high levels of chemo- and radioresistance. It has also been suggested that although many tumor cells may enter the bloodstream it is only the CSC that is able to produce metastases at distant sites. ^{14 –16} Thus, to develop effective ways to suppress one or more steps in the processes of tumor progression and metastatic spread, it is essential that we are able to characterize CSCs and the signaling pathways controlling them. Studies of this kind in UM, however, are hampered by problems associated with the acquisition and long-term culture of primary human UM cells, and have been slow because it has been hard to unequivocally identify CSC's. Recent studies using cancer cell lines from a range of tumor types, including prostate, breast, and head and neck, have demonstrated retention of a subpopulation with stem cell-like characteristics. ^{17 –20} Thus, the present study examines whether a cellular hierarchy exists in UM cell lines, such that they retain a population of self-renewing tumorigenic stem-like cells. Demonstration of such a cell population would allow the subsequent use of the cell lines to study their regulation and behavior and responsiveness to therapeutics. 

Materials and Methods

Cell Culture

The primary human UM cell line Mel270 and a cell line derived from a liver metastasis of the same patient, Omm2.5, originally derived by Tim Murray at the Bascom Palmer Eye Institute, ²¹ were kindly provided by Martine Jager (Leiden University, Leiden, The Netherlands). Both cell lines were routinely maintained in medium (RPMI medium; PAA Laboratories, Somerset, UK) supplemented with 10% fetal calf serum (FCS; PAA Laboratories), 2 mM l-glutamine (Sigma, Dorset, UK), and antibiotics. For all studies, UM cell lines (70% confluent) were harvested in EDTA with trypsin (0.25%) and resuspended in cloning medium to generate a single-cell suspension. Cloning medium (1:1 αMEM/Quantum 3.2.1 [PAA Laboratories]) with 10% FCS, 2mM l-glutamine, and antibiotics. 

Single-Cell Cloning

UM cell lines were harvested and diluted to a density of 100 cells per mL. Subsequently, 10 μL was dispensed into each well of a 96-well culture plate (approximately 1 cell per well) and 100 μL of cloning medium added. Four hours after plating, each well was checked with an inverted microscope (Nikon TS100; Jencons, Leicestershire, UK). Wells containing only a single cell were marked; wells with no cells or with more than one cell were excluded. Wells were checked daily and a further 100 μL of cloning medium added after seven days. After 10 to 14 days in culture, the colonies were counted, examined, and classified as holoclones, meroclones, or paraclones. 

In each experiment, the following were also conducted: individual holoclones were trypsinized and transferred to separate wells of a six-well plate for further culture and analysis. Up to 10 wells containing paraclones were also trypsinized, but the cells were then pooled and plated to a single well of a six-well plate. Clones that formed in the six-well plates underwent further passage and replating at a density of 200 cells per cm². Before each passage and replating, the numbers of holoclones, meroclones, and paraclones per well was determined; and 100 μL reagent (TRIzol; Invitrogen, Paisley, UK) was added to each of the remaining wells containing single holoclones. The wells were then pooled for RNA extraction and reverse transcriptase-polymerase chain reaction (RT-PCR) analyses, as described below. 

Nonadherent Sphere Assay

Cells were seeded in 20 mL cloning medium, at a density of 1000 cells per mL into three 75-cm ² flasks per cell line coated with poly 2-hydroxyethyl methacrylate (poly-HEMA; Sigma) to prevent cell attachment. Fresh medium (10 mL) was added to the flasks every 7 days. After 14 days, melanomaspheres (MS) of at least 50 μm in diameter (size was determined using an eyepiece graticule with crossed scales) were counted. To propagate MS, spheres were collected by centrifugation (300g) on Day 14, and then enzymatically dissociated by a short incubation in EDTA with trypsin, followed by mechanical dissociation with a 200 μL pipette tip to obtain a single-cell suspension. After cell counting, cells were either reseeded at a density of 1000 cells per mL into poly 2-hydroxyethyl methacrylate-coated 75 cm² flasks containing 20 mL cloning or basal medium as before ² ; lysed with reagent (TRIzol; Invitrogen) and RT-PCR analyses performed as described below; or diluted as described above for single-cell cloning and replated to each well of a 96-well plate for analysis of colony formation in adherent culture. 

Colony Heterogeneity

Adherent Culture.

UM cell lines were plated to 8-well chamber slides at a clonal density of 200 cells per cm², and examined by phase contrast microscopy as colonies developed. Colonies in each well were then fixed for 10 minutes in 2% formalin in phosphate buffered saline (PBS) before washing in PBS with 0.2% Triton X-100 and blocking with 10% normal goat serum and 1% bovine serum albumin in PBS. The antibodies used for immunofluorescence staining were mouse monoclonal antibodies against HMB45 (DAKO, Ely, Cambridgeshire, UK; 1:200), MelanA (DAKO; 1:100), vimentin (DAKO; 1:200), MITF (Novocastra, Newcastle, UK; 1:100), Pax3 (R&D Systems, Abingdon, UK; 1:50), Sox10 (R&D Systems; 1:50), rabbit monoclonal antibodies against Slug (SNAI2, Cell Signaling, New England Biolabs, Hertfordshire, UK; 1:400), Notch1 (Cell Signaling, 1:200) and a rabbit polyclonal antibody against CD117 (c-kit; DAKO; 1:200). After incubation for either 30 minutes at room temperature (RT; HMB45, MITF, MelanA, vimentin, CD117) or overnight at 4 °C (Pax3, Sox10, Slug, and Notch1), wells were washed in PBS and either goat anti-mouse Alexa Fluor 488 (Molecular Probes, Invitrogen Ltd, Paisley, UK) or goat anti-rabbit Alexa Fluor 555 (Molecular Probes) added for 1 hour at RT. Controls for specificity included: omission of primary antibodies for the rabbit polyclonal, rabbit IgG, and the use of isotype controls for the mouse monoclonals. Slides were mounted (Vectashield containing DAPI; Vector Laboratories, Peterborough, UK) for visualization of nuclear morphology, viewed on a fluorescence microscope (Leica DMRB; Leica Microsystems UK Ltd, Buckinghamshire, UK) and photographed (Zeiss AxioCam; Carl Zeiss Ltd, Hertfordshire, UK). 

Nonadherent Sphere Culture.

After 14 days, MS were gently pelleted by centrifugation (1200 rpm for 3 minutes) and fixed in 3.7% formalin in PBS for 10 minutes at RT. MS were then washed in PBS and again gently pelleted. Warmed agar (30 to 100 μL) was added to the tubes and the colonies pelleted in the agar by centrifugation at high speed. The solidified agar plug was then transferred to a tissue cassette. Normal tissue dehydration and clearing (to remove dehydrants) was performed before paraffin embedding and cutting of 4-μm sections for immunohistochemistry. In brief, sections were dewaxed in xylene, rehydrated, and endogenous peroxidase was blocked with H₂O₂ in methanol. Heat-induced epitope retrieval (EDTA pH 7.0) was performed using a pressure cooker, and the slides stained using an autostainer (DakoCytomation; DAKO) according to the following protocol. After washing with Tris-buffered saline (TBS), the slides were incubated for 40 minutes with the following antibodies: mouse monoclonal antibodies against HMB45 (DAKO; 1:200), MelanA (DAKO; 1:100), vimentin (DAKO; 1:200), MITF (Novocastra; 1:100), and Ki67 (DAKO; 1:200); a rabbit polyclonal antibody against CD117 (c-kit; DAKO; 1:200). Bound antibody was detected using a horseradish peroxidase system (Dako ADVANCE) according to manufacturer's instructions. Peroxidase was visualized with diaminobenzidine or a substrate kit (Vector AEC; Vector Laboratories) before counterstaining with Mayer's hematoxylin and mounting with a resin-based or aqueous mountant, respectively. 

RNA Extraction and Reverse Transcriptase Polymerase Chain Reaction

Total RNA was prepared (TRIzol) according to manufacturer's instructions. In all cases, RNA was subsequently incubated with DNase I (Turbo DNA Free kit; Ambion Ltd, Cambridgeshire, UK) to digest genomic DNA. Up to 1 μg of RNA was incubated in a 20-μL reaction containing 100 U reverse transcriptase (MMLV; Retroscript, Ambion) at 45 °C for 1 hour. RT-PCR was conducted to determine whether the cells expressed markers associated with a primitive migratory neural crest phenotype; PAX3, SOX10, SNAI2, and NOTCH1. Hot start PCR was performed in 25-μL reactions containing 50 ng cDNA, 20 nM of forward and reverse primers and 1 U DNA polymerase (SuperTaq; HT Biotechnology, Cambridge, UK), using a PCR reaction mix (Retroscript; Ambion). −cDNA and RNA lacking the reverse transcription step were included as controls; glyceraldehyde 3-phosphate dehydrogenase (GAPDH) was used as a housekeeping gene. Primers and RT-PCR conditions are given in Table 1. 

Table 1.

 

RT-PCR Primers and Conditions

	Primers	RT-PCR Conditions
SOX10 Forward	5′-AAGGTCAAGAAGGAGCAGCA-3′	32 cycles of 94°C, 30-second denaturation; 56°C, 30-second annealing; and 72°C, 30-second extension
SOX10 Reverse	5′-ACTTTCGTTCAGCAGCCTC-3′
SNAI2 Forward	5′-CTTTTTCTTGCCCTCACTGC-3′
SNAI2 Reverse	5′-AGCAGCCAGATTCCTCATGT-3′
PAX3 Forward	5′-TGTTTTTATCAACGGCAGG-3′	40 cycles of 94°C, 30-second denaturation; 56°C, 30-second annealing; and 72°C, 30-second extension
PAX3 Reverse	5′-CCGCGTCCTTGAGTAATTT-3′
NOTCH1 Forward	5′-AGAGAGCTCCTGCTTCAACG-3′
NOTCH1 Reverse	5′-CACACCAGTGCACAAGGTTC-3′
GAPDH Forward	5′-GAGTCAACGGATTTGGTCGT-3′	32 cycles of 94°C, 30-second denaturation; 52°C, 30-second annealing; and 72°C, 30-second extension
GAPDH Reverse	5′-GACAAGCTTCCCGTTCTCAG-3′

Analysis of Cell Type Surviving Chemotherapy

Mel270 and Omm2.5 cells plated to six-well plates (12,500 cells per cm²) were treated with the concentration of cisplatin previously shown to reduce cell number by 50% (IC₅₀). or medium alone for 24 hours, after which time cells remaining in the wells were harvested and counted. Two hundred cells per cm² were then plated to separate wells of a six-well plate in 2 mL cloning medium. An additional 1 mL of cloning medium was added every 7 days and the number of holoclones, meroclones, and paraclones formed were counted after fixation and staining with 70% Gentian violet in ethanol after 20 days in culture. 

Results

Clonogenic Capacity of Uveal Melanoma Cell Lines

In culture, each cell line contained cells with differing morphologies; Mel270 cells were spindle in shape while Omm2.5 cells showed a mixed morphology of spindle, flat polygonal, and dendritic forms. When plated at clonal density in adherent culture, Mel270 and Omm2.5 cells formed colonies after 10 to 14 days. Examination of the colonies formed indicated the presence of three morphologic types (Fig. 1A), as has been reported for other cancer cell lines ^{18 –20} and as originally described by Barrandon and Green. ²² The colony types were either: large, consisting of tightly packed small ovoid or spindle shaped cells—akin to holoclones; medium to large, consisting of loosely-packed small ovoid, spindle, and flat polygonal forms—akin to meroclones; and small, consisting of large flat polygonal cells and some dendritic cells—akin to paraclones. The parental Mel270 cell line yielded 39.0% holoclones, 44.8% meroclones, and 16.2% paraclones (Fig. 1B) and the parental Omm2.5 cell line produced 29.6% holoclones, 23.4% meroclones, and 47.0% paraclones (Fig. 1B). Further analysis of the proliferative potential of the cell line-derived holoclones and paraclones demonstrated that cells isolated from individual holoclones and replated at low density, gave rise to all three colony types (Fig. 1C). Continued passage and plating at low density of holoclone-derived cultures produced all three colony types in similar proportions during the course of the study (> 10 generations), demonstrating an ability to sustain long-term expansion in culture. In contrast, cells isolated from paraclones and replated at low density gave rise to further paraclones only (Fig. 1C). By passage two for the Mel270 cell line and passage three for the Omm2.5 cell line, cells isolated from these cultures failed to produce any additional colonies indicating that the proliferative capacity of these cells had been exhausted. 

Figure 1.

 

(A) Phase contrast images of colony morphology for Mel270 and Omm2.5 cells in adherent culture. Single Mel270 and Omm2.5 cells were plated to each well of a 96-well plate in cloning medium. Three main colony types were observed and classified as follows: holoclones (Holo), large colonies of tightly packed small ovoid or spindle shaped cells; meroclones (Mero), medium to large colonies of loosely packed small ovoid, spindle, and flat polygonal cells; and paraclones (Para), small colonies of large flat polygonal cells and some dendritic cells. The frequency of each type of cell clone formed (B) by the parental Mel270 and Omm2.5 cell lines as determined in three independent 96-well plate experiments (mean ± SD) (C) after replating of a single holoclone and pooled paraclones.

Melanomaspheres Contain Cells That Undergo Self Renewal

Nonadherent sphere culture assays have increasingly been used to examine the self-renewal capability of putative CSCs. ^{23 –27} Mel270 and Omm2.5 cells formed floating clusters of cells, MS, within two weeks of plating to nonadherent plates. 

MS formed by the Mel270 cell line consisted of loosely-packed cells that showed a significantly higher degree of pigmentation with increasing passage, than the more tightly packed spheroids formed by Omm2.5 cells. At passage three, this was consistent with a higher number of cells expressing HMB45 in Mel270 MS than those formed by Omm2.5 cells (Fig. 2B). An increased MS forming efficiency was noted at passage three for both cell lines, which then remained stable at passage four (Fig. 2A). To examine the relationship between MS and holoclones, Mel270 and Omm2.5 cells were isolated from passage three MS and clonally replated to adherent culture. Mel270 and Omm2.5 sphere-derived clones showed morphologic characteristics consistent with all three colony types, holoclones, meroclones, and paraclones, in proportions that were not statistically significantly different from those of the parental cell lines (Fig. 2C). 

Figure 2.

 

Melanomasphere (MS) formation in nonadherent culture. (A) Mel270 (dark gray bars) and Omm2.5 (light gray bars) cells were plated a low density to tissue culture flasks coated to prevent cell attachment. The number of floating spheres was determined after 14 days as a percentage of the number of cells plated. Cells were isolated from spheres at each passage and replated at the same density. (B) Antigenic heterogeneity of Mel270 and Omm2.5 MS isolated at passage three and fixed and embedded in agar. Sections through the MS were examined for the expression of MITF, MelanA, HMB45, Ki67, c-kit, and vimentin (Vim). Antigenic positivity is indicated by brownish-red staining and the sections were counterstained with hematoxylin (blue). (C) Frequency of cell clones formed after low density replating to adherent culture, of cells isolated from individual MS at passage three.

Holoclones and Melanomaspheres Display Antigenic Heterogeneity and Express Markers Associated with Both a Primitive Migratory Neural Crest Phenotype, and a More Differentiated Phenotype

In adherent culture, holoclones and paraclones derived from the Mel270 and Omm2.5 cell lines showed differing patterns of staining with antibodies against markers of committed melanocytes (Melan A and HMB45), the neural crest-derived melanocyte lineage (c-kit and MITF), and the intermediate filament marker vimentin. Typically, Mel270 and Omm2.5 holoclones and MS showed nuclear MITF staining in the majority of cells (Figs. 2B and 3). Heterogenous staining for MelanA and vimentin was noted in holoclones formed by both Mel270 and Omm2.5 cells (Fig. 3), however, both antigens were expressed in almost 100% of MS cells (Fig. 2B). Ki67 was also present in the majority of Omm2.5 MS cells in contrast to Mel270 MS, which showed positivity in only approximately 10% to 20% of cells (Fig. 2B). This was consistent with a much higher expression of HMB45 in Mel270 holoclones and MS (Figs. 2B and 3), and the ability of these cells to produce pigment in suspension culture. HMB45 was rarely detected in Omm2.5 MS cells. c-Kit positivity was detected in almost all Mel270 MS cells but was not present in Omm2.5 MS cells (Fig. 2B). We have previously observed Omm2.5 cells in adherent culture to be c-kit negative (data not shown). 

Figure 3.

 

Antigenic heterogeneity of Mel270 and Omm2.5 holoclones and paraclones. The expression of MITF, MelanA, vimentin (Vim), and HMB45 was determined by immunofluorescence after the growth of clones in adherent culture. HMB45 was not detected in Omm2.5 holoclones or paraclones. Cell nuclei were stained with DAPI (blue); positive immunofluorescence for each antigen was detected with Alexa Fluor 488 (green).

Mel270 and Omm2.5 clonogenic units were also examined for the expression of four markers associated with a more primitive migratory neural crest phenotype. ²⁸ As shown in Figure 4, both Mel270 and Omm2.5 holoclone-, paraclone-, and MS-derived cells expressed PAX3, SOX10, and SNAI2, at the mRNA level as determined by standard RT-PCR. NOTCH1 mRNA was detected in both Mel270 and Omm2.5 holoclones and MS but showed very low expression levels in Omm2.5 paraclones and was absent from Mel270 paraclones (Fig. 4). At the protein level, Pax3 and Sox10 were expressed in the majority of Mel270 and Omm2.5 holoclone and paraclone cell nuclei, although Pax3 expression appeared to be much weaker in the Omm2.5 clones (Fig. 5). Slug (SNAI2) and active Notch1 protein were observed in the majority of Mel270 and Omm2.5 holoclone and Omm2.5 paraclone cell nuclei, however, they were not detected in Mel270 paraclone cells (Fig. 5). 

Figure 4.

 

Expression of markers associated with a primitive migratory neural crest phenotype in Mel270 and Omm2.5 holoclones, paraclones, and MS. The expression of SOX10, SNAI2, Notch1 and PAX3 mRNA was determined by RT-PCR using RNA isolated from Mel270 and Omm2.5 holoclones, paraclones, or MS. Lanes: M, marker; 1, Mel270 holoclone; 2, Mel270 MS passage two; 3, Mel270 MS passage three; 4, Omm2.5 holoclone; 5, Omm2.5 MS passage three; 6, −RT sample; 7, −cDNA; 8, Omm2.5 holoclone; 9, Omm2.5 paraclone; 10, Mel270 holoclone; and 11, Mel270 paraclone.

Figure 5.

 

Antigenic heterogeneity of Mel270 and Omm2.5 holoclones and paraclones. The expression of Pax3, Sox10, Slug, and Notch1 was determined by immunofluorescence after the growth of clones in adherent culture. Slug and Notch1 were not detected in Mel270 paraclones. Cell nuclei were stained with DAPI (blue; left panels for each cell line); positive Pax3 and Sox10 immunofluorescence was detected with Alexa Fluor 488 (green; right panels for each cell line). Positive Slug and Notch1 immunofluorescence was detected with Alexa Fluor 555 (red; right panels for each cell line).

The Mel270 Cell Line Contains a Clonogenic Subpopulation of Cells That Are Able to Survive Chemotherapy

To examine whether the Mel270 and Omm2.5 cell lines possess a CSC-like chemoresistant phenotype, we determined the clonogenic capacity of cells surviving treatment with an IC₅₀ dose of cisplatin. Mel270 and Omm2.5 cells were seeded to 96-well plates at a density of 12,500 cells per cm² in cloning medium. After 24 hours, medium was removed and either fresh cloning medium alone or containing cisplatin (0–5.0 μg/mL) (Sigma) was added to the wells (n = 5 per treatment) for an additional 24 hours. The medium was removed and fresh cloning medium alone added to all wells. This was designated Day “0.” Plates (96-well) were then fixed at various times between 0 and 10 days in 10% trichloroacetic acid (Sigma) for 1 hour at 4 °C. Cell number was measured using 0.4% sulforhodamine-B (SRB; Sigma) in 1% acetic acid as a colorimetric end point. ²⁹ Readings were made at 570 nm on a microplate reader and the data analyzed (Excel, 2007; Microsoft, Berkshire, UK) to determine the IC₅₀ value for cisplatin. We hypothesized that by treating the UM cells with a drug such as cisplatin, which targets dividing cells, the rapidly proliferating transit-amplifying cells would be killed and slowly dividing CSC-like and differentiated cells would remain. Table 2 shows that the number of holoclones formed by both Mel270 and Omm2.5 cells selected by their ability to survive cisplatin treatment was significantly increased compared with untreated cells (P < 0.05), while the number of Omm2.5 meroclones and paraclones were significantly reduced. Although the numbers of Mel270 meroclones and paraclones were reduced in the cisplatin-treated cultures compared with untreated cells, this was not statistically significant. 

Table 2.

 

Phenotypic Analysis of the Frequency of Clone Types Formed after Treatment of Mel270 and Omm2.5 Cells with Medium Alone or an IC₅₀ Dose of Cisplatin

	Total Colonies Formed (%)
	Mel270		Omm2.5
	Control	Cisplatin (IC50 0.3 μg/mL)	Control	Cisplatin (IC50 0.5 μg/mL)
Holoclone	33.3 ± 5.3	44.5 ± 3.9*	28.6 ± 2.8	56.7 ± 9.4*
Meroclone	40.3 ± 5.5	31.9 ± 7.5	27.7 ± 1.2	12.7 ± 3.8*
Paraclone	26.3 ± 7.1	23.6 ± 3.7	43.7 ± 2.7	30.6 ± 5.6*

 Data show the mean ± SD from triplicate wells.

*   P < 0.05 compared with untreated control cells.

Discussion

This study presents novel evidence for the existence of a tumor cell hierarchy in two UM cell lines: one derived from a primary UM and one from a metastatic liver lesion of the same patient. Through in vitro clonal analyses, serial passaging, replating assays, immunophenotyping, and RT-PCR, we also provide evidence that the UM cell lines examined contain self-renewing cancer cells as well as cells with differing proliferative capacity and ability to undergo differentiation. 

Normal adult tissues contain a heterogeneous cell population with respect to morphology, function, and gene and protein expression patterns. This cellular heterogeneity is believed to reflect the hierarchical organization of the tissue, with rare self-renewing stem cells able to generate a transient proliferating cell population that undergoes terminal differentiation. These properties are essential for repair and homeostasis in many tissues. ^30,31 Even human tissues traditionally considered stable, undergoing only minimal or slow turnover throughout adult life, such as the eye, are also today known to contain specific stem cell populations. ^32,33 Indeed, multipotent neural crest-derived stem cells have been isolated from the normal adult mouse cornea. ³⁴ The similarity between normal tissue homeostasis and cancer has led to the CSC hypothesis, in which the tumor is maintained by a small subset of cells with unique functional properties, the CSC population. Over the last decade evidence for the CSC model has been reported in leukemia ³⁵ and many solid tumors. ^{9,36 –39} Recent studies of cutaneous melanoma demonstrated the presence of cells within the tumor with stem cell-like characteristics, based on the expression of biomarkers (CD133, CD166, and nestin) or ABC-transporter family members. ^{40 –44} However, data using putative CSC markers to isolate cell subpopulations have been inconsistent and progress has been slow, hampered by various factors including tissue procurement limitations; the ability for only short-term cultures; and the inability to unequivocally identify CSCs. In vitro systems that retain stem cell characteristics would greatly facilitate our ability to understand the molecular mechanisms regulating asymmetrical cell division, a key feature and potential therapeutic target of CSCs, as well as designing models that more appropriately determine therapeutic efficacy. In support of this, in vitro clonogenicity has been shown to correlate well with in vivo tumor initiating abilities. ^17,19  

We observed in the Mel270 and Omm2.5 cell lines the presence of a subpopulation of cells capable of indefinite self-renewal and with increased resistance to chemotherapeutic insult. This subpopulation produced holoclones and MS that contained cells at varying stages of maturation, including proliferating and differentiating progeny reminiscent of stem-like and transit-amplifying cell populations with differentiation capabilities. The proportion of holoclone forming cells was relatively high for both Mel270 (39 ± 9.5%) and Omm2.5 (29.6 ± 5.4%) cells, however, this is consistent with the recent report by Quintana et al. ⁴⁵ who demonstrated a significantly higher frequency of tumor initiating cells (approximately 25%) derived from cutaneous melanoma patients when compared with other tumor types. Variation in the percentage of holoclones formed by cell lines has previously been reported, for example in prostate cell lines this ranged from virtually none in the PC3 cell line to 20% to 40% in the DU145 cell line. ²⁰ This has been attributed to differences in plating density during routine subculture, which can result in selection pressures to repopulate in a similar manner to those occurring in vivo. These studies highlight the importance of maintaining consistent culture conditions when performing these types of analyses. The ability to produce differentiated cells, as determined by HMB45 positivity and the presence of brown pigmentation varied between the two cell lines with both properties being largely absent in Omm2.5 holoclones and MS. This may be explained by the fact that Omm2.5 cells are c-kit negative, a factor, which in normal melanoblasts is essential for MITF induction of tyrosinase expression and the production of mature pigmented melanocytes. ⁴⁶ The therapeutic importance of cancer cells with stem cell-like properties relates to their ability to evade cell death in response to therapeutic insult, resulting ultimately in tumor recurrence and metastatic spread. In this study we have shown that Mel270 and Omm2.5 cells growing in culture demonstrate an increase in the proportion of holoclones formed, together with reduced meroclone and paraclone formation after treatment with cisplatin, compared with untreated cells. Previous studies have suggested a number of mechanisms contributing to apoptotic resistance including increased activation of Notch ⁴⁷ or high levels of survivin. ⁴⁸ In addition, a recent study demonstrated that the apoptotic resistance of a stem cell-like population of head and neck carcinoma cells was associated with a tendency to dwell longer in G2. ⁴⁹ It should be mentioned that while the extensive self renewal demonstrated over > 10 passages for holoclones and four passages for MS is an indicator of stem-like cell activity, it does not provide a measure of stem cell frequency. Further studies are necessary to identify markers that could be used to isolate these cells for more detailed characterization. To begin to examine regulatory pathways that may control CSC-like activity and which could provide candidate molecular markers of this cellular population, we hypothesized that uveal melanoma cells harness neural crest developmental pathways. Thus, we examined the expression of several factors reported to regulate both the migration and lineage specification of neural crest stem cells, from which melanocytes are derived, and the maintenance of a melanocyte stem cell population in skin. ^{28,50 –53} Indeed, the expression of PAX3, SOX10, SNAI2, and NOTCH1 mRNA in holoclones and MS and Pax3, Sox10, Slug (SNAI2), and active Notch1 protein in holoclones of both cell lines lends further support to this possibility. Interestingly, only active Notch1 and Slug protein were found to be differentially expressed between the holoclone and paraclone cells in the Mel270, but not Omm2.5, cell line. It is well established that Pax3 and Sox10 play a key role in the regulation of MITF expression during melanocyte development and differentiation ^52,54 and in particular, Sox10 can inhibit differentiation and maintain neural crest stem cell potential. Notch1 was recently described in cutaneous melanoma as a key factor contributing to melanoma development by allowing melanoma cells to survive hypoxic stress and actively proliferate. ⁵⁵ In addition, one could speculate that these factors contribute to the migratory potential of the cells and their ability to colonize at distant sites. Slug, in particular, has been shown to regulate epithelial to mesenchymal transition as well as migratory behavior in melanoma and breast tumors. ^56,57 Association of Slug expression with resistance to chemotherapeutic agents has also been described in mesotheliomas. ⁵⁸ It was interesting to note that both holoclones and paraclones formed by the Omm2.5 cell line expressed all neural crest lineage markers at the protein level while for the Mel270 cell line, expression of Slug and active Notch1 were absent in paraclones. Whether this relates to other differences observed for the Omm2.5 cell line compared with Mel270 cells, for example c-kit negativity, high Ki67 index and low and/or absent HMB45 expression is unclear. Further functional studies are now necessary to uncover the role of these markers in subpopulations of uveal melanoma cells. 

In conclusion, the results presented demonstrate that by using the described clonogenic assays, malignant UM cell lines can provide a model system with which to investigate the effectiveness of particular therapeutic strategies against malignant stem-like cells as well as to understand the regulation of their behavior, numbers, and distribution. 

Footnotes

 Supported by the Eye Tumor Research Fund, Charitable Funds, and Royal Liverpool University Hospital NHS Trust.

Footnotes

 Disclosure: H. Kalirai, None; B.E. Damato, None; S.E. Coupland, None