CD57high Neuroblastoma Cells Have Aggressive Attributes Ex Situ and an Undifferentiated Phenotype in Patients

Background Neuroblastoma is thought to originate from neural crest-derived cells. CD57 defines migratory neural crest cells in normal development and is expressed in neuroblastoma. Methodology and Principal Findings We investigated the role of CD57 expression in neuroblastoma cells ex situ and in situ. Compared to CD57low U-NB1 neuroblastoma cells, CD57high cells developed tumors with decreased latency after orthotopic transplantation into adrenal glands of mice. In addition, CD57high U-NB1 and SK-N-BE(2)-C neuroblastoma cells were also more clonogenic, induced more spheres and were less lineage-restricted. CD57high cells attached better to endothelial cells and showed enhanced invasiveness. While invasion of U-NB1 cells was inhibited by blocking antibodies against CD57, neither invasion of SK-N-BE(2)-C cells nor adhesion of U-NB1 and SK-N-BE(2)-C cells was attenuated. After tail vein injection only CD57high cells generated liver metastases, while overall metastatic rate was not increased as compared to CD57low cells. In stroma-poor neuroblastoma of patients CD57high cells were associated with undifferentiated tumor cells across all stages and tended to be more frequent after chemotherapy. Conclusion Strong expression of CD57 correlates with aggressive attributes of U-NB1 and SK-N-BE(2)-C neuroblastoma cells and is linked with undifferentiated neuroblastoma cells in patients.


Introduction
High-risk neuroblastoma (NB) is an aggressive tumor of childhood whose prognosis has improved little for decades despite intensive multimodal therapy [1]. NB is an embryonic tumor thought to arise from immature derivatives of neural crest stem cells (NCSC) [2,3,4]. CD57, detected by the HNK-1 antibody [5], is a carbohydrate epitope expressed on adhesion molecules of NCSC and is associated with migration of these cells and their invasion and colonisation of tissues during development [6,7]. Thus, in NCSC, expression of CD57 is linked to a cellular behavior that also characterizes aggressive neuroblastoma. In line with this notion, strong expression of CD57 has been associated with histologically more immature NB cells [8]. However, experimental evidence investigating the role of strong CD57 expression in the aggressiveness of NB is lacking. We therefore set out to address this question using the short term NB culture U-NB1 and the NB cell line SK-N-BE(2)-C. We show that within these NB those cells selected for strong expression of CD57 have more aggressive attributes ex situ. In patients with stroma-poor NB strong expression of CD57 is associated with undifferentiated cells across disease stages and with residual tumor cells after chemotherapy.

U-NB1 cells strongly expressing CD57 show enhanced clonogenicity and induction of spheres
Clonogenicity correlates with the aggressiveness of tumors. The capacity of neuroectodermal tumor cells to initiate spheres under serum-free conditions is associated with their capacity to initiate tumors [9,10]. We therefore determined, whether the expression of CD57 in NB cells is associated with clonogenicity and induction of spheres. We first investigated NB cells without amplification of MYCN or deletion of 1p (U-NB1; Fig. 1A). These cells were established and maintained as spheres in serum-free medium, a procedure known to preserve the characteristics of the parental tumor [11]. 56% of U-NB1 cells expressed CD57, as determined by flow cytometry (Fig. 1B, upper left panel). CD57 expression was distributed in a bimodal fashion among U-NB1 cells, suggesting two subpopulations in regard to CD57 expression. This may be due to the fact that the U-NB1 culture, which has been propagated as spheres and with a low passage number, mirrors the heterogeneity of tumor cells found in patients' tumors. U-NB1 cells were fractionated by FACS into a CD57 high fraction (i.e., cells above the 80 th percentile of CD57 expression) and a CD57 low fraction (cells below the 20 th percentile) (Fig. 1B, upper panel). While the clonogenicity of U-NB1 cells and its fractions was very low, there was a significant increase in clonogenicity of CD57 high compared to CD57 low cells (Fig. 1B, lower panels). CD57 high cells also induced spheres at a 2.8-fold higher frequency than did CD57 low cells (Fig. 1C). Upon dissociation of CD57 high -derived spheres, cells fractionated again or left unfractionated induced new spheres with increased frequency upon recloning (p,0.001), suggesting that cells growing within spheres become adept at inducing new spheres. More important, CD57 high cells were always superior to CD57 low cells in inducing spheres.
Taken together, high expression of CD57 marks a population within U-NB1 cells with increased clonogenicity and an enhanced and sustained disposition to induce spheres. This suggests increased aggressiveness of CD57 high U-NB1 cells.

Clonogenicity and sphere induction of SK-N-BE(2)-C NB cells correspond to expression of CD57
We extended our study by investigating a MYCN-amplified NB cell line with deletion of 1p, SK-N-BE(2)-C. Flow-cytometric analysis showed that 70% of SK-N-BE(2)-C cells expressed CD57 in Gaussian distribution ( Fig. 2A, upper left panel). SK-N-BE(2)-C cells were sorted into CD57 low and CD57 high fractions ( Fig. 2A, upper middle and right panels). Clonogenicity of SK-N-BE(2)-C cells clearly segregated with expression of CD57, as clonogenic growth was confined to the CD57 high fraction, with very few colonies growing in the CD57 low fraction ( Fig. 2A, lower panels).
Expression of CD57 also correlated with the capacity of SK-N-BE(2)-C cells to initiate first, second and third generation spheres in serum-free medium (Fig. 2 B).
Compared to U-NB1 cells (Fig. 1) clonogenicity and, less so, sphere forming capacity of SK-N-BE(2)-C cells were higher. As in U-NB1 cells, strong expression of CD57 endows SK-N-BE(2)-C cells with increased clonogenicity and enhanced sphere induction, in vitro attributes of an aggressive cellular phenotype.
Increased frequency of peripherin 2 GFAP 2 cells in CD57 high U-NB1 and SK-N-BE(2)-C cells Less lineage-restricted and thus less mature tumor cells are usually more aggressive. We therefore analyzed, whether expression of CD57 is associated with a less lineage-restricted phenotype of U-NB1 and SK-N-BE(2)-C cells. U-NB1 cells have the morphology of so-called I-type NB cells (Fig. 1A, right panel), i.e., less lineage-restricted NB cells known to coexpress markers for both neuronal and non-neuronal lineages of the neural crest [12]. We therefore determined expression of peripherin, a neuronal marker, and glial fibrillary acidic protein (GFAP), a marker also expressed in glial cells, in U-NB1. Indeed, the majority of U-NB1 coexpressed peripherin and GFAP (Fig. 3A). We next investigated how expression of CD57 in U-NB1 cells correlates with the expression of peripherin and GFAP. Significantly more CD57 high cells were negative for both peripherin and GFAP compared to CD57 low cells (Fig. 3A). No significant differences were found in the percentage of cells double-or single-positive for peripherin and GFAP.
We showed that many SK-N-BE(2)-C cells, known I-type NB cells, also coexpress peripherin and GFAP. CD57 high SK-N-BE(2)-C cells tended to harbor more peripherin 2 GFAP 2 cells, although this trend was not statistically significant (Fig. 3B).
Taken together, the CD57 high fraction of U-NB1 contains significantly more double-negative cells than the CD57 low fraction, while SK-N-BE(2)-C cells show a trend in this direction. Assuming that double-negative cells are less lineage-restricted and thus less mature than double-positive cells this suggests increased aggressive potential of CD57 high cells.
CD57 high but not CD57 low U-NB1 and SK-N-BE(2)-C cells change lineage attributes upon forced differentiation in vitro Next we investigated the response of the fractions to forced in vitro differentiation towards attributes of peripheral sympathetic neurons and glial Schwann cells. Clones of CD57 low and CD57 high U-NB1 cells (Fig. 4A) and SK-N-BE(2)-C cells (Fig. 4B) were exposed to all-trans-retinoic acid (ATRA) and neuregulin for neural and glial differentiation, respectively. CD57 low cells were unresponsive to these agents, as expression of peripherin and neuregulin did not change. In contrast, CD57 high cells responded to the neuralizing ATRA by maintaining peripherin expression and decreasing expression of GFAP. The response of CD57 high cells to neuregulin was also pronounced, as expression of peripherin was markedly attenuated while GFAP expression was maintained. Thus, CD57 high U-NB1 and SK-N-BE(2)-C cells have an increased capacity to change lineage attributes upon forced neuronal and glial differentiation in vitro, which may suggest a less mature state of CD57 high cells.
Also of note, the number of double-negative clones significantly decreased upon forced differentiation of CD57 high U-NB1 and SK-N-BE(2)-C clones and tended to decease in CD57 low SK-N-BE(2)-C clones (Fig. 4 C). This is consistent with double-negative cells being more responsive to differentiation cues and thus supports the notion that double-negative cells are less lineagerestricted and hence less mature.
CD57 high U-NB1 cells initiate orthotopic tumors with shorter latency than CD57 low cells Tumor latency and tumorigenicity were investigated in an orthotopic model by transplanting FACS-separated CD57 high and CD57 low U-NB1 cells from one primary orthotopic tumor into the adrenal glands of mice (n = 8 for each group). CD57 high cells showed decreased latency of tumor formation compared to CD57 low cells (Fig. 5A, p,0.001), demonstrating increased aggressiveness of CD57 high U-NB1 cells. The incidence of tumor formation by CD57 high cells at the end of the experiment was not significantly different from CD57 low cells (8/8 vs. 5/8 mice, respectively, p = 0.24407 using Fisher's exact test). The tumors that developed consisted of densely packed small round blue cells typical of stroma-poor neuroblastic neuroblastoma (Fig. 5B).
CD57 high U-NB1 and SK-N-BE(2)-C cells have a higher propensity to invade and to adhere than CD57 low cells As CD57 expression defines migrating and colonizing neural crest cells during development we reasoned that CD57 may also be involved in invasion and adherence of neuroblastoma cells since these are derived from neural crest cells. To this end, we determined invasion of CD57 high and CD57 low U-NB1 and SK-N-BE(2)-C cells into matrigel and their adhesion to human umbilical vein endothelial cells (HUVEC). CD57 high U-NB1 and SK-N-BE(2)-C cells showed an increased capacity to attach to and foray into extracellular matrix (Fig. 6A) and to adhere to HUVEC (Fig. 6B). To determine, whether strong expression of CD57 U-NB1 cells were separated by flow cytometry into a CD57 high fraction (cells above the 80 th percentile of CD57 expression) and a CD57 low fraction (cells below the 20 th percentile). CD57 expression of unsorted and sorted cells is shown in histograms (upper panel). Solid lines correspond to CD57, dashed lines to isotype controls. For soft agar clonogenicity assays fractionated U-NB1 cells were seeded at 1000 cells per well into 24-well plates. 28 days after plating colonies were visualised by MTT staining. The numbers of colonies per well are depicted in the graph. The asterisk denotes p,0.05, using Student's t-test. Experiments were repeated three times, with similar results. (C) Enhanced induction of spheres by U-NB1 cells strongly expressing CD57. U-NB1 cells were separated by flow cytometry into CD57 high and CD57 low fractions. Fractionated as well as unsorted cells were plated at very low density in serum-free medium. The percentage of ''first generation'' spheres growing from single cells was determined. After dissociation of spheres this procedure was repeated twice (''second'' and ''third generation'' spheres, respectively). Results are shown as box plots (lower panel). Asterisks denote p,0.001, using Student's t-test. Experiments were performed twice, with similar results. doi:10.1371/journal.pone.0042025.g001 caused this behavior, invasion and adherence assays were also performed in the presence of blocking anti-CD57 antibodies. Proper function of the blocking antibody was confirmed by proving that it inhibited migration of a glioma cell line overexpressing CD57 but not of the parental cell line devoid of CD57 (Suppl. Fig. S1 A, B). Enhanced matrix invasion of CD57 high U-NB1 cells but not of CD57 high SK-N-BE(2)-C cells was inhibited by blocking CD57 (Fig. 6 A). Increased endothelial adhesion of CD57 high cells was not attenuated by blocking CD57 in both U-NB1 and SK-N-BE(2)-C cells (Fig. 6 B). To exclude that the antibody used for sorting prevented the blocking antibody from binding to CD57 or that internalization of antibody-bound CD57 decreased CD57 on the cell surface we performed sequential staining of CD57 with the sorting antibody and with a different anti-CD57 antibody. The sorting antibody was rapidly internalized and new CD57 swiftly appeared on the cell surface hence ruling out confounding effects of the sorting antibody on the blocking experiments (Suppl. Fig. S1 C).
Taken together, strong expression of CD57 in U-NB1 and SK-N-BE(2)-C cells is associated with attributes of aggressive cellular behavior but does not generally cause them.
Upon intravenous injection CD57 high U-NB1 cells are more proficient in forming liver metastases, as compared to CD57 low cells Having shown in vitro that some cellular mechanisms important for metastasis, such as adherence and migration, are preferentially active in CD57 high NB cells, we investigated metastasis of CD57 high U-NB1 cells. Orthotopically transplanted U-NB1 cells did not spontaneously metastasize. We therefore employed an experimental metastasis model. CD57 low and CD57 high cells isolated from orthotopic U-NB1 tumors expressing luciferase were injected into the tail vein of mice. A trend towards increased homing to lung and liver of CD57 high cells as compared to CD57 low cells was observed 30 minutes after injection (Fig. 7A). Few tumor cells were detected in spleen and bone marrow.
To investigate, whether increased homing of the CD57 high U-NB1 cells translated into growth of metastases, CD57 high and CD57 low cells were isolated from orthotopic U-NB1 tumors expressing luciferase and injected into the tail vein of mice. In vivo bioluminescence imaging revealed that CD57 high cells initiated metastases in 5 out of 8 mice injected, whereas 3 metastases developed in the 7 mice injected with CD57 low cells (Fig. 7B, left panel; p = 0.62 using Fisher's exact test). 4 out of 8 mice with CD57 high cells but none of the mice with CD57 low cells developed liver metastases (Fig. 7B, left and right panel; p = 0.049). Most CD57 high -initiated metastases were bigger than those of CD57 low cells (Fig. 7B). The luminescence signals corresponded to metastases detected by histology in liver and bone marrow (Fig. 7 C; and Suppl. Fig. S2). No microscopic metastases were found in organs negative for metastasis by bioluminescence imaging, confirming its specificity.
Taken together, these data show that strong expression of CD57 is associated with the ability of U-NB1 cells to form experimental liver metastases, without increasing overall metastatic rate.
In patients strong expression of CD57 is associated with undifferentiated NB cells across disease stages and with post-chemotherapy residual tumor cells To assess expression of CD57 in patient NB cells we investigated a panel of 16 NB that had been classified as Schwannian stroma-poor neuroblastoma according to the INPC classification (Table 1). Immunohistochemistry showed marked inter-and intratumoral heterogeneity of CD57 expression ( Fig. 8 A and Table 1). Using the strong expression of CD57 in natural killer cells as a positive internal control for CD57 high cells we determined that strong expression of CD57 was significantly associated with cytologically undifferentiated NB cells (Fig. 8, Table 1, p,0.001 using the Mann-Whitney test). In line, 9 out 11 (82%; 95% confidence interval 48-98%) undifferentiated and poorly differentiated NB contained CD57 high cells, compared to 1 out 5 (20%; 95% CI 1-72%) differentiating NB. CD57 high cells at diagnosis were more often found in non-metastatic stage 1-3 NB (5 out of 6 (83%), 95% CI 36-100%) compared to metastatic stage 4 NB (4 out of 10 (40%), 95% CI 12-74%).
Paired pre-and post-chemotherapy NB samples showed a trend to increased frequency of CD57 high cells in post-chemotherapy residual tumors in 7 out of 8 informative tumor pairs (Fig. 8B, C and Table 1; two-tailed p = 0.11 using the paired-sample Wilcoxon signed rank test). This is also illustrated by Fig. 8D that shows residual CD57 high cells in a tumor area rendered largely necrotic by chemotherapy.

Discussion
We show that CD57 high expressing neuroblastoma cells have aggressive attributes. Thus, we provide evidence that CD57 high cells from NB of different risk groups have enhanced sphereinducing capacity and clonogenicity, a less mature phenotype, an increased response to lineage-altering agents, decreased tumor latency, enhanced invasive and adhesive properties and a propensity to generate experimental liver metastases. In stroma-poor NB of patients strong expression of CD57 was found preferentially in undifferentiated cells across disease stages at diagnosis and in residual tumor cells after chemotherapy.
Increased clonogenicity and sphere-inducing capacity of CD57 high NB cells Anchorage-independent growth in soft agar and, to a lesser degree, sphere forming capacity was more pronounced in SK-N-BE(2)-C cells compared to U-NB1 cells, maybe in part due to MYCN being amplified and 1p being lost in SK-N-BE(2)-C cells [13]. Importantly, both anchorage-independent growth as well as sphere forming capacity in low-serum was more pronounced in the CD57 high fraction of U-NB1 and SK-N-BE(2)-C. There was a weak and sustained increase of sphere-inducing capacity of CD57 low cells upon recloning. This may be explained by adaption of CD57 low cells to the conditions of sphere culture unrelated to CD57 expression. Primary and secondary spheres from CD57 high cells could be fractionated into CD57 high and CD57 low cells. This implies that CD57 high cells must have generated CD57 low cells during sphere growth. The ability of NB cells to grow as suspended spheroids in low-serum conditions may reflect AKT-mediated anoikis resistance [10], enhanced growth factor independence or other, yet to be identified mechanisms. Intercellular adhesion mediated by homophilic interaction of CD57 [14] may add to the inclination of CD57 high cells to form spheres.
Anchorage-independent clonogenic growth is a defining in vitro property of aggressive cancer cells and the ability of NB cells to induce spheres correlates with their aggressiveness [10]. Thus, the propensity of CD57 high U-NB1 and SK-N-BE(2)-C cells for clonogenic and spheroid growth reflects an increased aggressiveness of the CD57 high cell population. The CD57 high fraction of both U-NB1 and SK-N-BE(2)-C cells harbored more cells neither expressing peripherin nor GFAP, markers for sympathetic and glial cells, respectively. The number of these double-negative clones decreased upon forced differentiation, consistent with these clones being of a less lineage-restricted phenotype. Thus, the CD57 high fraction of U-NB1 and SK-N-BE(2)-C cells contains less mature cells. In addition, CD57 high , but not CD57 low cells of both U-NB1 and SK-N-BE(2)-C could be forced to assume attributes of a predominant neuronal or glial phenotype. The increased capacity of these CD57 high NB cells to change lineage attributes provides additional support for the notion that this fraction contains less mature cells. Consistent with this notion, CD57 expression in SK-N-BE(2)-C cells has been shown to decrease upon differentiation to peripheral neuronal cells or glial cells [15]. The fact that CD57 marks neural crest stem cells, i.e., immature cells with increased capacity for differentiation (and sphere formation) lends further credibility to the association we observed in U-NB1 and SK-N-BE(2)-C cells between strong CD57 expression, less mature phenotype and increased capacity for changing lineage attributes (as well as enhanced sphere formation). The underlying molecular mechanisms for this association remain to be determined. As immaturity of cancer cells often correlates with enhanced aggressiveness, these findings may support the notion that U-NB1 and SK-N-BE(2)-C cells strongly expressing CD57 are more aggressive. Alternatively, the enhanced sensitivity of CD57 high NB cells to lineage cues may decrease these cells' aggressiveness should they encounter appropriate tumor-intrinsic or therapy-induced lineage cues.
Enhanced invasion and decreased tumor latency of CD57 high NB cells CD57 high U-NB1 and SK-N-BE(2)-C cells showed enhanced invasion into extracellular matrix in vitro. This may have contributed to the decreased tumor latency of CD57 high U-NB1 cells after orthotopic transplantation into adrenal glands of mice, the most common site of neuroblastoma. These data provide direct evidence of increased aggressiveness of CD57 high U-NB1 and SK-N-BE(2)-C neuroblastoma cells.
Increased adhesive, colonizing and metastasizing propensity of CD57 high NB cells Adherence of tumor cells to the endothelium of target organs, invasion of the subendothelial basal membrane and subsequent tumor cell growth are crucial late mechanisms in the metastatic cascade. We have shown that CD57 high U-NB1 and SK-N-BE(2)-C cells adhere better to endothelial cells and invade basal membrane matrix more extensively than CD57 low cells. While the matrigel trans-well chamber assay we employed cannot dissect the relative contributions of adhesion and migration to enhanced invasion of cells, the assay does show that invasion, the net result of adhesion and migration, is enhanced with CD57 high cells. After intravenous injection CD57 high U-NB1 cells were moderately more proficient in homing to and colonizing lung and liver.
induction of spheres. The propensity of single SK-N-BE(2)-C cells to generate spheres when plated at very low density in serum-free medium was compared between unsorted SK-N-BE(2)-C cells (upper panel, bar equals 30 mm), CD57 high cells and CD57 low cells. Results are calculated as the percentage of spheres generated from the cells seeded and are shown as box plots (lower panel). Asterisks denote p,0.001, using Student's t-test. Experiments were performed twice, with similar results. doi:10.1371/journal.pone.0042025.g002 Upon tail vein injection only CD57 high cells generated liver metastases, without an increase in overall metastatic rate. CD57 high U-NB1 also formed more metastases of large size.
In patient NB of all disease stages strong expression of CD57 is associated with undifferentiated NB cells Our experimental data were obtained from cell lines of patients with stroma-poor NB. We therefore investigated biopsies of stroma-poor NB. Poorly differentiated and differentiating NB were informative, as they contained both undifferentiated and differentiating tumor cells. In these tumors strong expression of CD57 was preferentially associated with cytologically undifferentiated NB cells across all disease stages. Accordingly, CD57 high cells were more often found in undifferentiated and poorly differentiated NB. This is consistent with the notion that CD57expressing undifferentiated NB cells mirror embryonic or early fetal sympathetic neuroblasts [8]. CD57 high cells at diagnosis were not increased in metastasizing disease, in line with a previous study that included ganglioneuromas and ganglioneuroblastomas [16]. . CD57 high but not CD57 low U-NB1 and SK-N-BE(2)-C cells are responsive to inducers of neuronal or glial lineage attributes. CD57 low and CD57 high U-NB1 cells (A) or CD57 low and CD57 high SK-N-BE(2)-C cells (B) were isolated by FACS. Cell fractions were plated in clonal density onto collagen-coated cover slips. Cells were treated over a ten day period with ATRA or with NRG1-ß1 for 14 days in medium supplemented with 2% FBS, followed by double-immunofluorescence staining for peripherin (green) and GFAP (red). Nuclei were stained with DAPI (blue). (C) Frequency of peripherin/GFAP double-negative clones upon forced differentiation as in (A) and (B). The asterisks denote p,0.05 using Student's ttest. Scale bars correspond to 50 mm. All experiments were repeated at least three times, with similar results. doi:10.1371/journal.pone.0042025.g004

CD57 high NB cells after chemotherapy
Following chemotherapy of patients the number of CD57 high NB cells within the residual tumor tended to increase. This may be due to selection of CD57 high NB cells, assuming that they are less chemosensitive than CD57 low cells. Alternatively, CD57 expression may be induced by chemotherapy, a presumption supported by the appearance of CD57 high cells in many NB that were devoid of CD57-positive cells prior to chemotherapy.
Possible mechanisms leading to CD57-associated characteristics of NB cells CD57 has been shown to mediate invasion and migration of neural crest stem cells [6,7] and has been associated with metastasis of melanoma, a neural crest stem cell-derived malignancy [17]. As CD57 defines migrating neural crest stem cells, from which NB is thought to arise, it appears plausible that some NB cells have co-opted from NCSC their CD57-associated propensity for migration and invasion. On a more mechanistic level CD57 is known to promote cell-cell interactions [14] and to mediate adherence of cells to laminin of the extracellular matrix via binding to integrins [18,19]. The latter mechanisms may have contributed to enhanced invasion of CD57 high U-NB1 cells.
That adherence of CD57 high NB cells to HUVEC and the invasion of CD57 high SK-N-BE(2)-C cells into matrix was not inhibited by anti-CD57 blocking antibodies in vitro seems to argue against a general role of CD57 in metastasis. However, these in vitro finding do not rule out that experimental metastasis in vivo could be inhibited by blocking antibodies. Regardless of the exact role of CD57 in experimental metastasis, CD57 expression did not correlate with metastatic disease in patients. Strong expression of CD57 may decrease shedding of NB cells from the primary tumor by increasing homophilic interactions between tumor cells. This may prevent any positive effect of CD57 on late metastasis suggested in the experimental metastasis assay from being operative in patients.
In conclusion, strong expression of CD57 is associated with aggressive attributes of NB cells ex situ and with undifferentiated and, to some degree, residual post-chemotherapy NB cells in patients.
C6 and G14 glioma cells were cultivated in Ham's F12 (PAA) supplemented with 15% horse serum (Biochrom), 2.5% heat -  . Only CD57 high cells generate liver metastases, while overall metastatic rate is not increased compared to CD57 low cells. (A) CD57 high U-NB1 cells injected intravenously home mildly better to lung and liver than CD57 low cells. U-NB1 cells from orthotopic tumors stably expressing the enhanced firefly luciferase gene were sorted into CD57 low and CD57 high fractions by flow cytometry. 5610 5 cells were injected into the tail vein of RAG2/2 cgc2/2 mice and the animals were sacrificed 30 min post injection. Organs were procured and lysed. Using a luciferase assay system luminescence was determined with a luminometer. Luminescence is expressed as relative light units. Because of different scales results are shown separately for lung and the other organs. Symbols represent the values of individual animals and bars the means. One representative of three independent experiments is shown. Student's t-test was used for statistical analysis. (B) CD57 high cells generate liver metastases in addition to bone marrow metastases. U-NB1 cells stably expressing firefly luciferase were sorted into CD57 low and CD57 high fractions by flow cytometry. 2610 6 cells were injected into the tail vein of NSG mice. Tumor development was assessed regularly beginning 21 days after tumor cell injection by luminescence imaging following luciferin injection. Mice are numbered for identification. The occurrence of liver metastases is depicted in the Kaplan-Meier plot, statistical analysis was performed using the log-rank test. (C) Histology confirms bone marrow and liver metastases. Mice were sacrificed 60-80 days post tumor cell injection and necropsy was performed. Organs were stained by H&E. Tail root of mouse CD57 high #2 and liver of mouse CD57 high #4 are shown. doi:10.1371/journal.pone.0042025.g007 inactivated FBS (Biochrom) and 2 mM L-glutamine at 37uC in an atmosphere of 5% CO 2 .
FACS cell -sorting CD57 high and CD57 low subpopulations of NB cells were isolated using phycoerythrin (PE)-conjugated mouse monoclonal CD57 antibody (Abcam, Cambridge, UK) and the FACSAria TM cellsorting system (BD Biosciences). 10 7 cells were incubated at 4uC in PBS (Biochrom) containing 5 ml CD57 antibody or unspecific PEconjugated IgM isotype control (BD Biosciences). Cells above the 80 th percentile of CD57 expression were sorted by flow cytometry into a CD57 high fraction, cells below the 20 th percentile of CD57 expression into a CD57 low fraction.
Sphere -forming assay U-NB1 and SK-N-BE(2)-C cells growing as spheres were sorted into CD57 high and CD57 low fractions as described above. Sorted cells were resuspended at a density of 1 viable cell/ml in serumfree DMEM/F12 containing 2% B27, 20 ng/ml rhEGF, 20 ng/ ml bFGF, 20 ng/ml LIF, 10 I.E./ml (5 mg/ml) heparin, 2 mM Lglutamine and penicillin/streptomycin. 1 ml of this suspension was added into each well of a 24-well plate with 14 mm cover slips at the bottom of the wells to avoid adherence. For the formation of second-generation spheres CD57 high neurospheres were dissociated and either left unfractionated or FACS-sorted into the two CD57 fractions and re-plated at clonal density. Third-generation spheres were generated in the same manner.
For immunocytological detection of CD57, cells were incubated at room temperature in PBS containing 5 ml PE-conjugated mouse monoclonal NK-1 CD57 antibody (Abcam, Cambridge, UK) or unspecific PE-conjugated IgM isotype control (BD Biosciences). Cells were washed, seeded on cover slips in 24-well plates and cultured over night. Subsequently, cells were incubated with HNK-1 anti-CD57 antibody (BD Biosciences) at RT for 1 h, followed by incubation with a secondary anti-mouse IgM-FITC antibody (BD Biosciences). Nuclei were stained with DAPI. Confocal microscopy was performed using the LSM 710 laser scanning microscope (Carl Zeiss, Jena, Germany).

Forced differentiation assay
NB cells from neurospheres were sorted into CD57 high and CD57 low fractions, plated either on poly-L-lysine-coated (SK-N-BE(2)-C) or on collagen type I-coated glass cover slips (U-NB1) at a concentration of 10 3 cells/ml and allowed to adhere overnight in growth media supplemented with 10% FBS. Medium was replaced with media containing 2% FBS and differentiationinducing cytokines. To differentiate, SK-N-BE(2)-C cells were treated with all-trans retinoic acid (ATRA; Sigma; 20 mM) or with human neuregulin1-b1 (NRG-1-b1; R&D Systems; 5 nM) for 10 days. For differentiation of U-NB1 cells 1 mM ATRA was used for 14 days. Differentiation medium was changed every 3 days. Differentiation was assessed by immunofluorescence cytology. For quantification of differentiation at least 144 colonies per group were analyzed at 4006 and 6006 magnification and scored as having colonies containing cells with red, green, red and green or unstained cytoplasm.
Blocking function of the HNK-1 anti-CD57 antibody C6 glioma cells do not express CD57 because they lack beta-1,3-glucuronyltransferase (GlcAT-P), a rate limiting enzyme of CD57 biosynthesis [20]. In contrast, C6 cells transfected with GlcAT-P cDNA (named G14 cells, [20]) express CD57. To verify the blocking capacity of the HNK-1 anti-CD57 antibody (BD) a 24-well transwell chamber system (8 mm pore size; BD Biosciences) was used. The bottom surface of the insert was coated with 10 mg/ ml mouse laminin (Invitrogen) over night at 4uC. C6 and G14 cells were labeled with DiI and incubated in serum-free alpha MEM medium (Gibco) with active or heat-inactivated blocking anti-CD57 antibody for 10 min at 37uC. 1.25610 5 cells were seeded in the top chamber. The bottom chamber was filled with serum-free alpha MEM medium. Cells were fixed 24 h post seeding with 3.7% paraformaldehyde. Membranes were cut out from the insert and mounted for detection of cells that migrated through the membrane. 24 low power visual fields were assessed using an inverse microscope (AX70; Olympus, Hamburg, Germany).

Matrigel invasion assay
BD biocoat tumor invasion system plates (BD Biosciences) containing a matrigel-coated PET membrane with 8 micron pore size were used. NB cells were sorted for the expression of CD57 by flow cytometry and stained with 10 mg/ml 1.19-dioctadecyl-3,3,39,39-tetramethyl-indocarbocyanine perchlorate (DiI, Molecular Probes, Göttingen, Germany). Cells were incubated with 30 mg/ml blocking anti-CD57 antibody (HNK-1, BD Pharmingen) or with heat-inactivated antibody as control for 10 min at 37uC. 2.5610 4 cells per well were seeded in RPMI 1640 medium without serum into the upper chamber. The lower chamber was filled with RPMI 1640 containing 10% FCS. Cells were fixed 24 h post seeding with 3.7% paraformaldehyde. Membranes were cut out from the insert and mounted for detection of cells that migrated through the membrane. 24 low power visual fields were assessed using an inverse microscope.

Adhesion assay on HUVEC
HUVEC were grown to complete monolayers in 24-well plates in Endothelial Cell Basal (EBM) Media (Lonza, Walkersville, MD) supplemented with EGM Single Quots (Lonza), 20 ng/ml recombinant human vascular endothelial growth factor (R&D Systems, Minneapolis, MN) and 20% FBS. NB cells were sorted for the expression of CD57 by flow cytometry and stained with 10 mg/ml 1.19-dioctadecyl-3,3,39,39-tetramethyl-indocarbocyanine perchlorate (DiI, Molecular Probes, Goettingen, Germany). Cells were incubated with 30 mg/ml blocking anti-CD57 antibody (HNK-1, BD Pharmingen) or with heat-inactivated antibody as control for 10 min at 37uC. 2.5610 4 NB cells were seeded per well on top of the HUVEC monolayer. After one hour under normal SK-N-BE(2)-C or U-NB1 culture conditions, respectively, cells were washed twice and fixed with 3.7% PFA. DiI-labeled cells per low magnification visual field were determined using an inverted microscope. 24 different visual fields were analyzed.
Orthotopic NB model [6][7][8] week old male RAG 2/2 /common c-chain 2/2 mice bred in the Animal Research Center of the University of Ulm were anesthetized with ketamin (Pfizer, Berlin, Germany; 25 mg/ml) and rompun (Bayer, Leverkusen, Germany; 20 mg/ml). After laparatomy 1610 4 viable U-NB1 cells suspended in 50 ml of a 1:3 mixture of BD Matrigel TM High Concentration (BD Biosciences) and DMEM/F12 medium without supplements were injected with a 27-gauge needle into the left adrenal gland. Tumors were resected and single cell suspensions were generated by mincing the tumor followed by chemical dissociation with Liberase Blendzyme 1 (Roche). Cells were sorted by FACS into CD57 high and CD57 low fractions and transplanted again orthotopically as described above. Mice were monitored twice a week for the presence of tumors by palpation and for morbidity.

Stable transduction of U-NB1 cells with firefly luciferase and detection of luciferase activity
U-NB1 cells were stably transduced with the retroviral expression vector pLib ELN (ProQuinase GmbH, Freiburg, Germany) carrying an enhanced luciferase and neomycin fusion gene and selected for luciferase-expressing neomycin-resistant cells. Luciferase activity was determined using the Luciferase Assay System (Promega Corporation, Madison, USA) and a luminometer (Berthold Technologies, Bad Wildbad, Germany) according to the manufacturers' recommendations.

Metastasis model
CD57 high and CD57 low fractions of U-NB1 cells expressing luciferase were generated by FACS sorting. 24 h post sorting 2610 6 cells suspended in PBS were injected into the tail vein of 7-10 week old NOD scid gamma (NSG) mice. Bioluminescence indicating tumor growth was monitored every 2 weeks starting 3 weeks after tumor cell injection. Mice were anesthetized with isofluran (Abbott, Wiesbaden-Delkenheim, Germany). Luciferin (Synchem, Felsberg/Altenburg, Germany; 100 mg/kg body weight) was injected intraperitoneally and images were taken using the Xenogen In vivo Imaging System IVIS 200 (Caliper Life Sciences, Hopkinton, USA) and Living Images software (Caliper Life Sciences).

Histology of metastases
Mice were sacrificed 60-80 days after tumor cell injection. Liver, lung, pelvis and tail of all mice and pancreas, fallopian tubes and soft tissue of selected mice were formalin-fixed for 2 h. Pelvises and tails were decalcified in 10% EDTA for 5-7 days. Tissues were embedded in paraffin, serially sectioned, stained with H&E and analysed for the presence of metastases using an AX70 PROVIS microscope (Olympus).

Immunohistochemistry of patient tumors
Immunohistochemistry was performed on paraffin-embedded tissue sections from the archived tissue files of the Institute of Pathology, University Medical Center Ulm, Germany. Staining was performed according to standard protocols using the EnVision kit (Dako, Carpinteria, CA). As primary CD57 antibody the clone NK1 (IgM) was used (Menrini, Newcastle, UK) at a dilution of 1:20 after pretreatment of the slides with pronase (1 mg/1 ml). Evaluation was carried out in a semiquantative fashion using CD57-positive natural killer cells that were always present in the tumors as positive internal controls.

Statistical analysis
Statistical analysis was performed using the PASW statistics software version 18.0 (IBM, Chicago, Illinois).

Ethics statement
Scientific use of the human tumor tissue employed in this study was approved by the ethics committees of the Ludwig-Maximilians-Universitä t (permit number 280-07) and of the Klinikum Augsburg (permit of 11.2.2005). Written consent of the patient's parents was obtained.
The animal studies were approved by the Regierungsprä sidium Tübingen (permit number 986). All animal experiments were carried out in strict accordance with institutional and state guidelines for animal welfare and all efforts were made to minimize suffering. Figure S1 HNK-1 anti-CD57 antibody blocks CD57 function. (A) G14 cells, but not parental C6 glioma cells, express CD57. Parental C6 glioma cells and G14 cells, i.e. C6 glioma cells overexpressing beta-1,3-glucuronyltransferase, a rate limiting enzyme of CD57 biosynthesis, were assayed for expression of CD57 by immunocytology. Nuclei are stained by DAPI. (B) HNK-1 anti-CD57 antibody blocks migration of C6 glioma cells overexpressing CD57. C6 and G14 cells were labeled with DiI and incubated with active or heat-inactivated blocking anti-CD57 antibody and added to laminin-coated transwell chambers. 24 hours post seeding cells that had invaded into and migrated through the membrane were fixed. (C) The NK-1 anti-CD57 antibody used for sorting is rapidly internalized and new CD57 epitopes are swiftly presented on the cell surface. SK-N-BE(2)-C cells were stained with NK-1 anti-CD57-PE (red) used for sorting and plated on cover slips in 24-well plates. 24 h post-staining, cells were incubated with unlabeled HNK-1 anti-CD57 followed by anti-mouse IgM FITC (green) and analyzed by confocal microscopy. (TIF) Figure S2 Histology confirms the presence of metastases detected by bioimaging. Mice were sacrificed 60-80 days post tumor cell injection. Livers were formalin-fixed, paraffinembedded and H&E-stained. Tail roots were formalin-fixed, decalcified, paraffin-embedded, H&E-stained and screened for metastases. The histology of mice identified in Fig. 7 is shown. (TIF)