Gangliosides from Human Melanoma Tumors Impair Dendritic Cell Differentiation from Monocytes and Induce Their Apoptosis

Dendritic cells (DC) ² are bone marrow-derived professional APC characterized by their unique capacity to elicit specific immune responses from naive T cells (1). They are found as a trace population in most tissues and fluids, where they exist at different differentiation/maturation stages interconnected by defined pathways of circulation (2). Most knowledge about the biology of DC has emerged from the recent possibility to generate DC in vitro from either CD34⁺ hemopoietic progenitors cultured in the presence of GM-CSF and TNF-α (3) or peripheral blood monocytes cultured in the presence of GM-CSF and IL-4 (4, 5).

A growing body of evidence has demonstrated that DC play a crucial role in the induction of tumor-specific immune responses. DC are capable of recognizing, processing and presenting tumor Ags to T cells, in turn initiating a tumor Ag-specific immune response (6). The transfer of Ags from tumor cells to DC, or cross-priming, is thought to be mandatory for processing of tumor-associated Ags and presentation by MHC class I to T cells (7, 8). Numerous clinical observations suggest that DC infiltration of solid tumors correlates with a better prognosis (9, 10). Conversely, tumor progression is associated with increased immunosuppression involving impairment of the DC system. Both circulating and tumor-infiltrating DC are functionally defective in tumor-bearing hosts (11, 12, 13, 14), a process that has been mostly related to tumor-derived soluble factors. Gabrilovich et al. first (15) demonstrated that production of vascular endothelial growth factor by tumor cell lines inhibits the early stages of DC development from human CD34⁺ hemopoietic progenitors in vitro. Using a similar model, Ménétrier-Caux et al. (16) showed that IL-6 and M-CSF from renal carcinoma cell lines block DC differentiation into potent immunocompetent cells and promote their commitment toward monocytic lineage. Additionally, a decreased longevity of DC may contribute to tumor-induced immunosuppression. Indeed, a large variety of murine and human tumor cell lines produce as yet unidentified soluble factors that cause increased apoptosis in DC (17, 18, 19, 20).

Gangliosides are ubiquitous, membrane-associated, sialic acid-containing glycosphingolipids that have been implicated in a broad range of biological functions, including cell growth and differentiation (21). It has long been known that the ganglioside composition is altered in tumor cells. For instance, while normal melanocytes only express GM3, melanoma cells overexpress a variety of gangliosides, the majority being GM3 and GD3 (22, 23). Furthermore, a number of tumors, including metastatic melanoma, shed gangliosides into their microenvironment (24, 25), and much evidence suggests that soluble gangliosides contribute to the tumor-induced immunosuppression (26, 27, 28, 29, 30). In a recent study Shurin et al. (31) demonstrated that GD2 and GM3 gangliosides released by neuroblastoma cell lines inhibit the generation of DC from CD34⁺ progenitors, as assessed by reduced expression of CD80, CD83, and CD86 and allostimulatory function.

In the present study we analyzed the effects of GM3 and GD3 gangliosides purified from human melanoma tumors on the other source of DC, i.e., monocytes cultured in the presence of GM-CSF and IL-4. We demonstrate that both GM3 and GD3 inhibit the phenotypic and functional differentiation of monocyte-derived DC and induce their apoptosis.

Gangliosides were purified from human melanoma tumors as previously described (32). Briefly, GM3 and GD3 were isolated by HPLC on a 250-4 Si100 column (Merck, Darmstadt, Germany) with a Hitachi L-6200 apparatus (Hialeah, FL) using a ternary gradient of hexane-isopropanol-water (55/36/9 to 55/30/15, v/v/v) at a flow rate of 0.25 ml/min. Fractions of 0.5 ml were collected, and elution was monitored by TLC on HPTLC silica gel 60 plates (Merck) migrated in chloroform/methanol/0.2% aqueous calcium chloride (60/35/8, v/v/v). The plates were visualized by heating at 150°C after spraying with a resorcinol-HCl reagent. The dried gangliosides were taken up in PBS, pH 7.4, and autoclaved for 20 min at 120°C (32). The sterile solutions of gangliosides were kept at 4°C until utilization. The stability of gangliosides in culture medium was checked after recovery of gangliosides from the medium by the recently reported method (33).

The culture medium was RPMI 1640 supplemented with l-glutamine (Life Technologies, Grand Island, NY), 1% gentamicin (Sigma-Aldrich, St. Louis, MO), and 10% heat-inactivated FCS (Myoclone; Life Technologies), hereafter called complete medium. Recombinant human GM-CSF (sp. act., 2 × 10⁶ U/mg) and IL-4 (sp. act., 2 × 10⁷ U/mg) were generous gifts from Schering-Plough Research Institute (Kenilworth, NJ).

Mononuclear cells were obtained from the peripheral blood of healthy donors by centrifugation on Ficoll-Hypaque (Pharmacia Biotech, St. Quentin en Yvelines, France). Cells (4 × 10⁷) were layered on a Percoll gradient (50% in PBS/5% FCS) and centrifuged at 450 × g for 25 min at 4°C. The monocyte-enriched fraction (nearly 70%) was collected from the interface while lymphocytes were recovered from the cell pellet. Monocytes were depleted of T and B cells using hapten-conjugated anti-CD3, CD7, CD19, CD45RA, and CD56 mAbs and anti-hapten Ig coupled to magnetic microbeads according to the manufacturer’s instructions (Monocyte Isolation Kit, Miltenyi Biotec, Bergisch Gladbach, Germany). The technique routinely resulted in >80% purity, as assessed by flow cytometry.

Purified monocytes (10⁶ cells/ml) were cultured for 6 days in six-well tissue culture plates (Costar, Cambridge, MA) in complete medium supplemented with recombinant human GM-CSF (200 ng/ml) and IL-4 (33 ng/ml), in the presence or the absence of melanoma-derived gangliosides at different concentrations. On days 2 and 4 cells were fed with fresh medium and cytokines.

For the study of CD40 ligand (CD154)-mediated maturation, fibroblastic L cells transfected with CD154 (provided by Schering-Plough, Kenilworth, NJ), were irradiated at 80 Gy and added to the monocyte cultures in a proportion of 1/10. Cells were collected 40 h later.

Cells were incubated for 30 min at 4°C with affinity-purified mouse mAbs at the appropriate concentration or with irrelevant isotype-matched mouse Igs at the same concentration. Cells were washed and, for indirect staining, further incubated for 30 min at 4°C with FITC-conjugated F(ab′)₂ of goat anti-mouse Ig Ab. The following mAbs were used: anti-HLA-DR-FITC (B8.12.2, IgG1), anti-CD54-FITC (84H10, IgG1), anti-CD80-FITC (mAb 104, IgG1), anti-CD83 (HB15A, IgG1), and anti-CD40 (mAb 89, IgG1), all from Immunotech (Marseilles, France); anti-CD1a-FITC (BB-1) and anti-CD86 (IT2.2, IgG1) from BD PharMingen (San Diego, CA); and anti-CD14-FITC (TUK 4, IgG1) from DAKO (Glostrup, Denmark). For double-color fluorescence, cells were first stained with FITC-conjugated mAbs, followed by PE-conjugated anti-CD1a or anti-CD86 mAbs. Fluorescence analysis was performed on a FACScan using LYSYS II software (BD Biosciences, Pont de Claix, France).

Quantification of apoptotic DC was assessed by two methods. The early exposure of phosphatidylserine residues on the cell surface was measured using the annexin V-FITC kit, with dead cells identified by propidium iodide (Immunotech). For this assay, 5000 events were collected on a FACScan II cytometer and analyzed with CellQuest software. The results were confirmed by investigating the occurrence of condensed or fragmented nuclei after incubating the cells for 30 min at 37°C with bisbenzimide (10 μg/ml), also called Hoechst 33342 (Sigma-Aldrich). Stained cells were examined by standard fluorescence microscopy, and the percentage of cells with nuclear features of apoptosis cells was calculated based on a total number of at least 400 cells.

Allogeneic T cells were isolated from the lymphocyte pellet obtained after Percoll gradient by rosetting with SRBC as previously described (34). The T cell population contained >95% CD3⁺, as assessed by flow cytometry. Mixed lymphocyte reactions were conducted in round-bottom microtiter plates by adding 10⁵ allogeneic T cells to varying numbers of DC. Triplicate cultures were maintained for 5 days at 37°C in a 5% CO₂ humidified atmosphere. T cell proliferation was measured by pulsing the cells with 1 μCi of [³H]methylthymidine (25 Ci/mmol; Amersham Pharmacia Biotech, Les Ulis, France) for the final 18 h of culture. Cells were then harvested, and incorporated thymidine was quantitated in a direct beta counter (Matrix 96; Packard Instruments, Meriden, CT). Results were expressed as the mean counts per minute ± SD of triplicate cultures.

PGE₂ and IL-10 production was assessed after a 6-day culture in the presence or the absence of gangliosides. Alternatively, IL-10 and IL-12 p70 were measured after a further activation of DC for 40 h in the presence of CD154-transfected fibroblasts. Cell supernatants were harvested and kept at −80°C before use. PGE₂ and cytokine concentrations were measured using ELISA kits purchased from R&D Systems Europe (Abington, U.K.): IL-10 (sensitivity, 7.8–500 pg/ml); IL-12 p70 (sensitivity, 0–40 pg/ml), and PGE₂ (sensitivity, 39–500 pg/ml).

Statistical analysis was conducted using Student’s t test. Values of p < 0.05 were considered statistically significant.

As previously reported (32, 35), human melanoma tumors mainly express and shed GM3 (NeuAc-Gal-Glc-Cer) and GD3 (Cer-Glc-Gal-NeuAc-NeuAc) gangliosides. These components were purified from human melanoma tumors as previously described (32) and added separately to monocytes at the onset of culture. The amounts of gangliosides to be added to the culture medium were selected to obtain a final concentration within the range of ganglioside concentrations reported in the sera of melanoma patients (35), i.e., 20–50 μg/ml. As assessed by FACS analysis, human monocytes cultured for 6 days in the presence of GM-CSF and IL-4 developed into floating DC characterized by the acquisition of CD1a and the loss of CD14 Ags at the cell surface. As shown in Fig. 1,A, both GM3 and GD3 gangliosides induced phenotypic alterations of monocyte-derived cells. The effect was dose-dependent, with GM3 being more efficient than GD3 in this process (Fig. 1,A). While either ganglioside did not substantially alter HLA-DR Ag expression at the cell surface, they induced a dose-dependent down-regulation of CD1a, CD54, CD80, and CD40 expression (Fig. 1 A). CD14 and CD83 remained negative, while, unexpectedly, CD86 expression was substantially up-regulated.

When gangliosides were added on day 4 instead of day 0 of culture, the decrease in DC-specific markers that occurred on day 6 was less pronounced (not shown). For example, the decrease in CD1a expression was 25–35 vs 50–60% for GM3 and GD3 at 20 and 30 μg/ml, respectively. By contrast, a similar inhibition of expression of DC markers could be observed whether gangliosides were added on day 0, day 2, or even day 4 of culture, providing that phenotypic analysis was conducted at least 4 days later. For example, in two experiments, the decreases in mean fluorescence intensity for CD1a were 57 ± 14, 60 ± 6, and 40 ± 11% when GM3 was added on days 0, 2, or 4 and phenotypic analysis was conducted on days 4, 6, and 8 of culture, respectively. This demonstrates that the inhibitory effect of gangliosides is time dependent and can be induced at any step of DC differentiation.

According to a recent paper by Kiertscher et al. (19), tumor supernatants induced an early and transient maturation of monocyte-derived DC, as revealed by up-regulation of HLA-DR, CD83, and CD86 Ag expression on day 4 of culture. To investigate whether gangliosides could have similar effects, monocyte-derived cells were harvested after a 4-day culture. We found in four experiments that, compared with control cells, CD1a, CD54, CD80, and CD40 Ags were already down-regulated on both GM3- and GD3-treated cells after the 4-day culture (not shown). By contrast, the percentage of CD86⁺ cells was increased in GD3- and GM3-treated DC (from 32 ± 8 to 63 ± 14 and 59 ± 11% in the presence of GM3 at 20 μg/ml and GD3 at 30 μg/ml, respectively). Moreover, both GM3 and GD3 induced a slight and transient increase in the percentage of mature CD83⁺ cells (from 12 ± 8 to 35 ± 10 and 34 ± 12% with GM3 at 20 μg/ml and GD3 at 30 μg/ml, respectively).

Kinetic analysis of ganglioside-induced phenotypic alterations was then conducted by adding GM3 or GD3 on day 2 of culture and analyzing the cells every day thereafter. As shown in Fig. 1,B, addition of gangliosides induced a substantial, but transient, increase in the mean fluorescence intensity for HLA-DR Ag on day 3 of culture. At that time, the percentage of DR⁺/CD1a⁺ cells was not significantly altered compared with that in the control culture. However, whereas this percentage increased from day 3 to day 6 in control monocyte cultures, it remained stable and ultimately decreased in the presence of both gangliosides. Similar observations could be made for CD1a⁺/CD54⁺ and CD1a⁺/CD80⁺ cells (not shown). Kinetic experiments also revealed the increased percentage of both CD86⁺/CD1a⁺ and CD86⁺/CD1a⁻ cells as soon as 1 day after the addition of gangliosides to the culture medium (Fig. 1 C). Taken together, these results demonstrate that both GM3 and GD3 gangliosides induced early activation of monocyte-derived cells and ultimately impaired their differentiation.

In all the experiments we observed that ganglioside treatment decreased cell viability at the end of the culture. Thus, compared with control cells, the viable cell yield on day 6 averaged 51.1 ± 13.3% in the presence of GM3 at 20 μg/ml and 75 ± 15.9% in the presence of GD3 at 30 μg/ml (mean ± SD of seven experiments). We therefore hypothesized that this effect might be due to an induction of apoptosis. To check this possibility, cells were treated with gangliosides on day 2 of culture and were analyzed every day thereafter for the presence of apoptotic cells. As assessed by annexin V-FITC/propidium iodide staining (Fig. 2, A and B), both GM3 and GD3 gangliosides induced a significant increase in both apoptotic and necrotic cells. Interestingly, the ganglioside-induced apoptotic cells could only be detected 2 or 3 days after their addition to the cells and increased thereafter (Fig. 2,A). The results were confirmed by Hoechst staining on day 6, showing an increased number of cells with condensed or fragmented nuclei in the ganglioside-treated cell suspensions (Fig. 2,C). It should be noted that differentiated monocyte-derived DC, i.e., day 4 or day 6 DC, remained sensitive to the GM3- and GD3-induced apoptosis (Fig. 2 D).

Cells from control or ganglioside-treated monocyte cultures were then evaluated for their allostimulatory capacity. To this end, graded numbers of viable monocyte-derived cells were added to allogeneic T cells in an MLR assay. As shown in Fig. 3, the cell allostimulatory function was significantly reduced in the presence of either GM3 or GD3 gangliosides, in a dose-dependent manner. In agreement with the results of phenotypic analyses, GM3 was more efficient than GD3 in this process. A significant inhibitory effect of GM3 and GD3 gangliosides could be detected after a 4-day culture (not shown).

We assayed the production of several factors by monocyte-derived DC after a 6-day culture in the presence or the absence of either GM3 or GD3. The factors were measured in the cell supernatants by ELISA. The production of IL-10 by control DC was low (5–20 pg/ml), and it did not increase in the presence of either ganglioside (not shown). By contrast, GD3 and especially GM3 induced a significant increase in PGE₂ production (Fig. 4) by monocyte-derived DC.

To investigate whether ganglioside treatment could have an effect on subsequent DC maturation, monocytes were cultured for 6 days in the presence or the absence of GM3 or GD3. Cells were then extensively washed and enumerated, and the same number of viable DC was activated for 40 h with CD154-transfected fibroblasts in the absence of gangliosides. The production of IL-10 and IL-12 was then assessed in the cell supernatants by ELISA. As shown in Fig. 5, GM3 treatment did not substantially modify the production of both cytokines by CD154-activated DC. By contrast, DC differentiated in the presence of GD3 showed an impaired ability to produce IL-12 compared with control DC. This could not be related to altered cell viability, since, concomitantly, GD3-treated DC produced enhanced amounts of IL-10.

Although the underlying mechanisms are not fully understood, it has long been known that gangliosides shed from tumors, including human melanoma, inhibit antitumor cellular immune responses. In vitro, tumor-derived gangliosides inhibit mitogen- or allogeneic cell-induced T cell proliferation, a process that has been related to an impaired Ag presentation by human monocytes (28). Addition of gangliosides to the tumor cell inoculum enhances tumor formation in mice (26) and inhibits tumor-specific cytotoxic T cell responses (29).

Gangliosides were also shown to inhibit myelopoiesis and hemopoiesis in vitro at several stages of development by a mechanism involving modulation of the maturation of precursor cells (36, 37). Recent papers suggest that gangliosides may also regulate DC generation. Shurin et al. (31) first reported that addition of neuroblastoma-derived GD2 and GM3 gangliosides to murine and human CD34⁺ hemopoietic precursor cells results in a significant decrease in CD83⁺- and CD86⁺-expressing cells at the end of the culture. Furthermore, the capacity of the cells to generate allogeneic T cell reaction was severely impaired. Inhibition of DC differentiation from human monocytes was recently reported by Wölfl et al. (38) using commercially available bovine brain ganglioside mixture. In that study, however, the inhibition could only be related to GM2, whereas GM3 and GD3 had no effect. Our present data extend these results by showing that both GM3 and GD3 gangliosides purified from human melanoma tumors profoundly impair phenotypic and functional differentiation of peripheral human monocytes. Indeed, both additions of GM3 and GD3 gangliosides to the culture medium ultimately inhibited the acquisition of DC-specific markers such as CD1a, CD54 adhesion molecule, and CD80 and CD40 costimulatory molecules, resulting in cells with poor allostimulatory properties. In the presence of gangliosides, the monocyte-derived cells lost CD14 and did not adhere to plastic culture wells, demonstrating that they were not directed toward the macrophage pathway. The apparent discrepancies between our study and the previous ones can probably be accounted for by the differential effects of gangliosides according to their tissue origin, i.e., brain vs melanoma cells and normal vs malignant tissue. This may reflect some differences in the ganglioside structure that may lead to different biological activities. Indeed, it is well known that structural differences exist between gangliosides of melanoma cells and neural tissue, notably in the ceramide backbone (39).

As opposed to their effect on CD34-derived DC (31), we found here that gangliosides induced an early maturation of monocyte-derived DC, followed by cell apoptosis. Induction of DC maturation has been reported by Kiertscher et al. (19) using supernatants from tumor cell lines, including melanoma. Gangliosides may play a role in this process, since we found here that GM3and GD3 gangliosides induced an early and transient increase in HLA-DR, CD54, and CD83 Ag expression as well as a sustained up-regulation of CD86 at the cell surface. Furthermore, this early tumor-induced DC maturation was followed by cell apoptosis, as shown by Kiertscher et al. (19) as well as in the present study.

A new finding in the present study was that both GM3 and GD3 melanoma-derived gangliosides induced significant apoptosis of monocyte-derived cells, as assessed by annexin V-FITC binding and Hoechst staining. It has long been known that tumor-derived factors induce both murine and human DC apoptosis, but the underlying mechanisms are still unclear. They involve down-regulation of Bcl-2 and up-regulation of Bax (17), a process that can be abrogated by CD154 transduction of tumor cells (18). The nature of the apoptotic factor in the tumor cell line supernatants was not identified, and apoptosis could not be neutralized by Abs against IL-10, vascular endothelial growth factor, TGF-β, or PGE₂ (19). According to the present results, tumor-shed gangliosides might be good candidates for the induced apoptotic effect on monocyte-derived DC.

We found that apoptosis, as well as phenotypic alterations, could be induced by gangliosides at any step of DC differentiation, and that both effects increased along with the time of contact and concentration of the gangliosides. One can hypothesize that in vivo, higher concentrations of gangliosides can be found in the vicinity of the tumors, and we indeed observed a dramatic drop in the viable cell yield following exposure to increased ganglioside concentrations (not shown). When monocyte exposure to gangliosides was limited to the first 1 or even 2 days of culture, and the cells were washed, the percentage of typical CD1a⁺/DR⁺ DC on day 6 closely resembled that in controls. Furthermore, the number of apoptotic and necrotic cells did not exceed the control value (not shown). This might suggest that accumulation of as yet undefined metabolites accounted for the observed effects. The apoptotic signal seems to take several days to be actually triggered following incubation with gangliosides. This delay may explain the lack of apoptosis observed by Shurin et al. (31), who examined the apoptotic effect of GD2 on CD34⁺ cells after incubation with the ganglioside for only 48 h, and such an incubation time is likely to be too short to actually detect an increase in apoptosis.

A recent paper reported that supernatants from murine melanoma cell lines induced apoptosis in bone-marrow derived murine DC by increasing C16 and C24 ceramide endogenous levels. Indeed, blocking ceramide synthesis by l-cycloserine prevented DC from tumor-induced apoptosis (20). However, the authors did not identify the tumor-derived factor that could enhance DC endogenous ceramide levels. According to the present results and because ceramide is a constitutive component of ganglioside structure, it is tempting to speculate that gangliosides might be metabolized into ceramide, which, in turn, would induce DC apoptosis. In keeping with this, the formation of free sphingosine and ceramide from GM1 ganglioside has been reported by cerebellar granule cells in culture (40). Furthermore, in preliminary experiments we showed that ³H-labeled ceramide could be detected in monocyte-derived DC after the addition of a ganglioside containing tritiated fatty acid to the cell culture (not shown).

An interesting finding is that GD3, but not GM3, impairs not only DC differentiation, but also their subsequent maturation induced by CD154 in the absence of the ganglioside. Namely, the GD3-treated DC produced a reduced amount of bioactive IL-12 and an increased amount of IL-10 compared with the control cells. A similar pattern of cytokine production has been reported after pre-exposure of developing DC to PGE₂ (41) at concentrations as low as 10⁻⁹ M, and we show here that gangliosides induced an increased production of PGE₂ by DC. However, both GM3 and GD3 were efficient in this process, which makes it unlikely that the altered production of IL-12 and IL-10 by maturing DC could be merely mediated by PGE₂. Nevertheless, the results suggest that GD3 will affect the activity of professional APC not only on the site of encounter, i.e., in the tissues neighboring the tumor, but also once they have left the tissues and migrated into regional lymph nodes. There, since IL-12 is critical for the induction of Th1 responses (42), the priming by IL-12-deficient APC may result in the generation of Th2 cytokine-biased responses and, therefore, hampers an efficient antitumor response.

Some antiganglioside Abs that showed strong antitumor effects in vitro gave disappointing results in clinical trials (43). One possible explanation of the poor beneficial effects for the patients would be the presence of shed gangliosides that can inhibit any activation of the immune system. The amount of such tumor-derived gangliosides that are known to be bound to the serum lipoproteins (44) may also be high enough to complex the antiganglioside Abs injected in the patients, thereby precluding the direct targeting of tumor cells by these Abs.

In conclusion, this study demonstrates for the first time that melanoma-derived GM3 and GD3 gangliosides severely impair DC differentiation from human monocytes and induce their apoptosis. It provides new insights into understanding the mechanisms underlying the suppression of immune recognition and subsequently tumor escape. These results should be considered in the development of new strategies to elicit antitumor immune responses, especially those involving DC.