(A) Tumor development after subcutaneous injection into NSG mice of single melanoma cells derived from twelve patients with metastatic disease. Dots represent individual tumors and are plotted against tumor growth rate (left axis). Lines represent the mean growth rate of all tumors derived from each patient. Red stars represent the proportion of single cells derived from each patient melanoma that formed tumors (right axis). Scale bars = 1cm. (B) Linear regression analysis of the mean (±SD) growth rate of clonal tumors derived from the same twelve patients plotted against the empirically determined frequency of tumorigenic cells in their parent tumors. r² value shows the Pearson correlation co-efficient, indicating no significant correlation between tumor growth rate in NSG mice and tumorigenic cell frequency in NSG mice.

(A) Eight clonal tumors were established from single cells obtained from patient 481. 100 cell aliquots from each of these tumors were serially transplanted into secondary and tertiary NSG mice. All clonal tumors were successfully passaged twice after being established from single cells, suggesting that all tumorigenic cells had unlimited tumorigenic potential (scale bar=1cm). (B) Clonal tumors derived from melanomas from five other patients were passaged similarly. Every attempt was successful. (C-H) Tumor growth rates for patients 481 (C), 501 (D), 526 (E), 528 (F), 530 (G) and 534 (H). Each group of three bars indicates the growth rates of a clonal tumor (white) and its descendent first (grey) and second (black) generation tumors. During each passage, 2-6 secondary or tertiary injections were performed for each tumor line and nearly all such injections gave rise to tumors. The growth rates of these tumors are shown as mean±SD (*, p<0.05 by T-test indicates significantly different growth rates in secondary and tertiary tumors).

(A) Frequency of ABCB5⁺ cells versus frequency of tumorigenic cells in melanomas from nine patients. AJCC stage is the American Joint Committee on Cancer clinical stage of the patient at the time of melanoma removal. Frequency of ABCB5⁺ cells indicates the percentage of cells that stained above isotype control background by flow cytometry. The frequency of tumorigenic cells in each tumor was determined by either single cell or limiting dilution injections into NSG mice. (B-C) Separation by flow cytometry of ABCB5^- (blue) and ABCB5⁺ (red) melanoma cells from patients 491 (B) and 526 (C). Percentages indicate the frequency of cells that stained more strongly than isotype control (left). Reanalyses of sorted cells is shown to the right. Each plot shows viable, human HLA⁺ cells and excludes mouse hematopoietic (CD45, TER119) and endothelial (CD31) cells as described previously (Quintana et al., 2008). (D) Tumor formation after injection into NSG mice of unfractionated, ABCB5^-, and ABCB5⁺ cells isolated as in Figure 3B-C from 3 different patients (308: n=2 experiments, 491: n=1, 526: n=1). Almost every injection of 10 ABCB5^- cells or 10 ABCB5⁺ cells from these three melanomas formed a tumor.

(A) Frequency of CD271⁺ cells versus frequency of tumorigenic cells in melanomas obtained directly from 13 patients and from 9 xenografted tumors (≤ 2 passages). NGFR5 and C40-1457 anti-human CD271 antibodies were compared side-by-side and gave similar results (data not shown). (B) Linear regression analysis of the percentage of cells expressing CD271 in tumors from 15 patients, plotted against the frequency of tumorigenic cells in the same tumors. r² value (the Pearson correlation co-efficient) indicates no correlation. (C) Separation by flow cytometry of CD271^- (blue) and CD271⁺ (red) melanoma cells from patients 597, 600, 608, 610, 631, 641, 491, 526 and 534. Reanalysis of sorted cells is shown to the right. Each plot shows viable, human HLA⁺ cells and excludes human or mouse hematopoietic (CD45+, Glycophorin A or TER119+) and endothelial (CD31+) cells. (D) Tumor formation after injection into NSG mice of unfractionated, CD271^- and CD271⁺ cells purified as in Figure 4C directly from six patients or from three xenografted (≤ 2 passages) tumors. Both CD271^- and CD271⁺ cells readily formed tumors with similar efficiency when isolated from stage III or IV metastatic tumors (600, 608, 631 and 641). CD271^- cells were more tumorigenic in primary cutaneous tumors (597, p=0.001; 610, p=0.005), though the less tumorigenic CD271⁺ cells accounted for only 2-12% of cells in these tumors. See also Table S1.

(A-H) Spontaneous metastasis from a subcutaneous melanoma that arose from the injection of a single melanoma cell derived from a xenograft obtained from patient 481. Fifteen weeks after injection, a subcutaneous tumor was observed at the site of injection (B-D) that metastasized to lymph nodes (not shown), ovaries (E), pancreas (not shown), and liver (F). Immunostaining of the subcutaneous tumor (D), ovary (G) and liver (H) confirmed the presence of S100+ melanoma cells (in brown). (I-Q) Melanoma metastasis from subcutaneous tumors that arose from the transplantation of CD271^- or CD271⁺ cells obtained directly from patient 608 (I). Metastases developed in the kidneys (J, L, N, P) and lungs (K, M, O, Q) of NSG mice 23 to 32 weeks after transplantation, irrespective of whether CD271^- (J-M) or CD271⁺ (N-Q) cells were transplanted. Metastasis developed with similar efficiency from tumors derived from CD271^- and CD271⁺ cells (I). Sections of kidney (J,P) and lungs (K, Q) show infiltrated S100+ melanoma cells (in brown). Similar results were obtained when injecting unfractionated, CD271^- or CD271⁺ cells derived from xenografted tumors from patient 205, without Matrigel (R). Scale bars = 1 cm (C, E, F, L-O) or 100 μm (D, G, H, J, K, P, Q).

(A) Analysis by flow cytometry of the heterogeneous expression of MCAM, CD29, A2B5, CD151, E-Cadherin, CD44, HNK1, CD10, N-Cadherin, CD49d, CD54, L6, c-kit, CD49b, CD9 and CD49e. The expression of each marker was analyzed in xenografted tumors derived from at least 9 different patients (see Table S2 for details). Blue and red gates show the selection of marker^-/low and marker^+/high cells for transplantation studies (Figure 6B), based on isotype labeling (indicated with a vertical line in each plot). (B) Tumor formation after injection into NSG mice of 10 unfractionated, 10 marker^-/low, or 10 marker^+/high cells isolated by flow cytometry from xenografted tumors obtained from 3 to 5 different patients (except for CD10 and CD49e, which were tested in cells from 2 different patients). Tumors readily formed from every fraction of cells such that no marker distinguished tumorigenic from non-tumorigenic melanoma cells. See also Figure S2 and Table S2.

Expression of ABCB5 (A), CD166 (B), A2B5 (C), CD151 (D), CD54 (E), CD44 (F), CD9 (G), CD29 (H), N-Cadherin (I), CD271 (J), CD49e (K), CD49f (L), L1-CAM (M), E-Cadherin (N) and c-kit (O) in parent tumors (upper left) compared with expression in secondary tumors derived from marker^-/low and marker^+/high fractions (top right and bottom right, respectively). Bottom left panels show reanalyses of the sorted cell fractions used to generate secondary tumors. See also Figure S3. Every marker was tested in 2-4 separate melanomas, except for CD44, CD49f, E-Cadherin and c-kit, which were tested in one.