Human malignant melanoma harbours a large fraction of highly clonogenic cells that do not express markers associated with cancer stem cells

Authors


Lina Prasmickaite, e-mail: linap@rr-research.no

Dear Sir,

Tumour initiation in various cancers has recently been attributed to the presence of rare cells with stem cell properties, so-called cancer stem cells (CSC). It has been reported that malignant melanoma also harbours small subpopulations of cells that express certain markers (e.g. CD20, CD133 and ABC-transporters) linked to tumour-initiating cells and that show exclusive abilities to form clones/tumours (Fang et al., 2005; Monzani et al., 2007; Schatton et al., 2008). However, a recent study by Quintana et al. (2008) revealed that a large fraction (∼27%) of melanoma cells had tumour-initiating abilities in vivo, independent of the expression of potential markers. The high frequency of tumourigenic cells together with the current lack of markers to distinguish such cells from non-tumourigenic counterparts raise doubts as to whether the CSC model applies to malignant melanoma.

In the present study we characterized clonogenic/tumourigenic properties and expression of stem cell-related markers in tumour cells isolated from four melanoma models that we have established from two patients with aggressive melanoma (designated Melmet 1 and Melmet 5). These models included early-passage in vitro cell cultures, traditional monolayers (MON) and non-adherent spheroids (SPH) grown in specialized stem cell medium, and the corresponding xenografts growing in mice. Here we show that a large fraction (20–60%) of single, randomly chosen melanoma cells were highly clonogenic, i.e. they grew anchorage-independently in soft agar (Figure 1A), formed non-adherent spheroids (Figure 1B–D) and could self-renew generating daughter spheroids, thereby displaying properties associated with tumourigenic potential. Individual cells demonstrated substantial heterogeneity in clonal growth rate and pattern (Figure 1D), which was also maintained in daughter spheroids, indicating that different melanoma cells could sustain heterogeneity with respect to colony formation.

Figure 1.

 Clonogenic potential and expression of stem cell-related genes in SPH- versus MON-derived melanoma cells. (A) Anchorage-independent growth in soft-agar. (B,C) Spheroid formation from single melanoma cells isolated from in vitro cultures (B) and corresponding in vivo xenografts (C) and cultured as one cell/well in stem cell medium in 96-well plates. Data indicate average % ± SD (n = 2). (D) Representative pictures (taken at day 17) illustrating the growth pattern from a single Melmet cell: a ‘compact’ large spheroid (left); a spheroid with disseminating cells, indicated by arrows (middle); a small ‘scattered’ colony (right). Bar, 200 μm. (E) Real-time PCR-based analysis of gene expression. Values in MON cells are normalized as 1, and the expression in the corresponding SPH cells is presented as relative increase (above 1) or decrease (below 1). Error bars indicate SEM from two different samples analyzed two or three times (OCT 4 data from a single experiment performed in duplicate). *P < 0.05.

Non-adherent spheroid formation has been considered as a method for enrichment for tumour-initiating/stem cells and as a measure of ‘stemcellness’ in various cancers (Singh et al., 2003; Zucchi et al., 2007), including melanoma (Fang et al., 2005). Here we show that although the clonogenic abilities of the SPH-derived cells compared to the MON cells seemed to be slightly higher, in general they were not remarkably different, and only three of six cases revealed statistically significant (P < 0.05) differences (Figure 1A–C). In addition, the stem cell-associated genes (GLI 1, BMI 1, OCT 4, SNAIL, SLUG and TWIST) were not consistently enriched in SPH cells compared to MON (Figure 1E), arguing against enhanced ‘stemcellness’ in spheroids, at least with respect to these genes or the associated pathways. However, spheroid cells (originating from two of three investigated patients) demonstrated enhanced tumour formation following subcutaneous (s.c.) injection in nude mice (Table 1), which is in agreement with an earlier study by Fang et al. (2005). Although comparison of the metastatic abilities displayed following systemic injection in vivo revealed no differences between SPH- and MON-derived cells, multiple metastases were induced efficiently in all cases (Figure S1). These data suggest that formation of spheroids is a rather weak criterion for the presence of subpopulations with exclusive clonogenic/tumourigenic potential and/or stem-like properties in malignant melanoma.

Table 1.   Limited dilution analysis of tumourigenicity of melanoma cells derived from MON and SPH cultures
Melmet cultures250 × 103 cells50 × 103 cells5 × 103 to 10 × 103 cells2 × 103 cells
Tumour take (%)Lag time (days)Tumour take (%)Lag time (days)Tumour take (%)Lag time (days)Tumour take (%)Lag time (days)
  1. Decreasing cell numbers of a single-cell suspension were injected s.c. into nude mice as indicated in the table, and the formation of palpable tumours was followed for 5 months. Tumour take indicates the number of tumours/number of injections. Lag time to first palpability is presented as mean ± SE from three (Melmet 1) and two (Melmet 5) independent experiments. N/A, not analyzed; n = 1 indicates a single case of tumour take.

  2. aP = 0.014; bP = 0.3.

  3. cTumour-initiating cell frequency is presented with the 95% confidence interval.

Melmet 1
 MON15/16 (94)22 ± 15/12 (42)48 ± 8a4/12 (33)63 ± 13b1/10 (10)67 (n = 1)
Tumour-initiating cell frequencyc: 1/71 300 (1/42 000 to 1/121 000)
 SPH14/14 (100)24 ± 612/12 (100)29 ± 2a10/12 (83)49 ± 5b7/12 (58)55 ± 3
Tumour-initiating cell frequency: 1/4000 (1/2200 to 1/7100)
Melmet 5
 MON18/20 (90)27 ± 31/12 (8)18 (n = 1)0/12 (0)N/AN/A
Tumour-initiating cell frequency: 1/150 000 (1/93 200 to 1/240 600)
 SPH5/6 (83)29 ± 45/6 (83)36 ± 35/12 (42)67 ± 8 N/AN/A
Tumour-initiating cell frequency: 1/50 000 (1/23 000 to 1/105 000)

Assessment of markers associated with tumour-initiating or stem cells, namely CD133, CD20, ABCG2, p75, cMET and aldehyde dehydrogenase (ALDH), revealed heterogeneous expression both in vitro (Figure 2A) and in vivo (Table S1). The majority of the investigated markers were identified only in tiny cell subpopulations, which does not agree with the large fraction of clonogenic cells observed in Figure 1(A–C) and which suggests a lack of correlation between these markers and clonogenicity. In contrast, a relatively large fraction of cells were positive for the markers cMET and ALDH. However, comparison of the clonogenic ability of single cells from subpopulations positive for cMET, ALDH and p75 versus subpopulations negative for these markers revealed no differences (Figure 2B). Only the p75+ cells showed a tendency to reduced clone formation compared to p75, which would be in line with a recent study by Held et al. (2010); however, the difference observed in our study was not statistically significant (P > 0.05). This indicates that the investigated markers do not discriminate melanoma cells with clonogenic abilities from cells lacking this property. There was also no difference in the in vivo tumourigenic potential of ALDH+ and ALDH cells; both subpopulations induced tumours in mice following s.c. injection of 200–10 000 cells (Prasmickaite et al., submitted), which corresponds to the in vitro clonogenicity data shown in Figure 2(B).

Figure 2.

 Lack of correlation between expressions of the assessed markers associated with tumour-initiating/stem cells and clonogenic potential in melanoma cells. (A) Flow cytometry-based identification of marker+ subpopulations. Data indicates average % marker+ cells ± SEM from at least three experiments. (B) Clonogenicity (spheroid formation) of marker+/high and marker–/low cells isolated from Melmet xenografts. Single cells from the sorted subpopulations were cultured as described in Figure 1(B, C). Error bars indicate SD (n = 2 for p75 and cMET and n = 7 for ALDH). *P < 0.05.

Summarizing, we have shown that in malignant melanoma, a large fraction of cells from phenotypically different subpopulations and independent of the expression of several assessed markers, exhibit clonogenic/tumourigenic properties. Although the current study cannot prove or disprove the findings of Quintana et al. (2008) regarding the high frequency of tumour-initiating melanoma cells that could not be distinguished by the examined markers, it does indicate a lack of markers for discriminating clonogenic from non-clonogenic melanoma cells. We anticipate that if such markers are identified in malignant melanoma in the future, they will most likely mark a large fraction of melanoma cells.

Ancillary