Interferon Stimulated Gene 15 Constitutively Produced by Melanoma Cells Induces E-Cadherin Expression on Human Dendritic Cells¹ Free

DCs³ are potent inducers of immunity, the properties of which can be strongly influenced by the nature of the microenvironment they infiltrate (1). Experimental data indicate that the migratory behavior of mature DCs can be profoundly different, even when their apparent antigen-presenting capacity is similar (2). In particular, homing of mature DCs to secondary lymphoid organs, following the induction of specific chemokine receptors (3, 4), allows priming of naïve T cells. On the other hand, induction of chemokines specifically attracting antigen-specific T cells (5, 6) will enhance effector functions in the periphery and recruitment of memory lymphocytes (7). Derangements of these migratory patterns could crucially influence the outcome of immune responses, including those that are tumor specific (8, 9).

The presence of DCs in malignant tissues has been extensively studied, with conflicting results. For some types of cancers, such as oral (10), ovarian (11), colorectal cancers (12) and renal carcinoma (13), DC infiltration has been associated with prolonged patients survival and reduced metastatic disease, specially if associated with T-cell infiltration (10, 12).

On the other hand, factors secreted by neoplastic cells can compromise the antigen-presenting cell functions of infiltrating DCs, thereby favoring tumor immune escape. In a consistent number of reports, discrete tumor-derived factors have been shown to prevent DC differentiation and maturation and hamper the induction of antitumor immunity (14, 15, 16, 17). Along with these observations, we have previously identified discrete melanoma cell lines inducing E-cadherin expression on monocyte-derived DCs in vitro, potentially impairing their migratory behavior (9).

In the present study, melanoma cell clones endowed with different functional capacities were subjected to gene profiling to identify differentially expressed genes. We provide evidence that defined soluble factors are responsible for the melanoma-induced modulation of DC phenotype.

Cells were cultured in RPMI 1640 supplemented with 1% Ciproxin (Bayer, Zürich, CH), 2 mm l-glutamine, 1% nonessential amino acids, 1% sodium pyruvate, 5 × 10⁻⁵ m 2-mercaptoethanol and 10% FCS (Life Technologies, Inc., Paisley, Great Britain), thereafter referred to as complete medium. LPS content in serum was tested by Limulus LAL assay, and only LPS-free batches were used. Human recombinant IL-4 was produced in our laboratory. Granulocyte/macrophage-colony stimulating factor was provided by Novartis (Basel, CH).

The original cell line Me67 was generated in our laboratory from a metastatic melanoma. Lines Me67.3, Me67.5, Me67.9, and Me67.10 were derived by culturing the parental cell line in limiting dilution at 0.3 cells/well in 96-well plates. All cell lines were grown in complete medium and were free from Mycoplasma infection, as monitored by specific reverse transcription-PCR. Cell lysates were prepared by three cycles of freeze and thawing.

Immature DCs were generated from human peripheral blood mononuclear cells according to published methods (18). Briefly, monocytes were purified by positive sorting using anti-CD14 conjugated microbeads (Milteny I., Bergisch-Gladbach, D). The sorted cells were cultured for 6–7 days in complete medium supplemented with 50 ng/ml granulocyte/macrophage-colony stimulating factor and 1000 units/ml IL-4.

Immature DCs were cocultured with either tumor cells at a 5:1 ratio, in the presence of tumor cell culture supernatant or their lysates (1:2 dilution). In some experiments, purified rabbit IgG anti-ISG15 (19), goat anti-IFNARI, mouse anti-CCR5 IgG1 antibodies, or their isotype matched controls (R&D Systems, Oxon, United Kingdom and BD Biosciences, Heidelberg, Germany), were added to the cultures at a 50 μg/ml final concentration. Cell cultures were analyzed, after 24-h incubation time, by FACS staining.

The DC phenotype was monitored by cell surface staining using FITC-conjugated mouse antibodies from BD Biosciences (Heidelberg, Germany) to human CD86 (clone IT2.2) and CD15 (clone MMA). The mouse antihuman E-cadherin (clone SHE78–7; R&D Systems) antibody was used in combination with a goat antimouse IgG_2a FITC-conjugated (Southern Biotechnology Associates, Birmingham, AL) antibody. Samples were analyzed on a FACSCalibur (Becton Dickinson, Mountain View, CA) using propidium iodide to exclude dead cells.

Cultured melanoma cells were harvested by scraping, and total cellular RNA was extracted (20). Ten μg from each sample were reverse transcribed, labeled, and processed by using a commercial kit (Affymetrix, Santa Clara, CA) according to the supplier’s instructions. Upon alkaline heat fragmentation, cDNA were hybridized to the arrays following standard procedures as supplied with the microchips (Affymetrix). Raw data were collected with a confocal laser scanner (Hewlett Packard, Palo Alto, CA), and pixel levels were analyzed using a commercial software (GeneChip v3.1; Affymetrix). Three repeats for each array were performed. Expression levels for each gene were calculated as normalized average difference of fluorescence intensity as compared with hybridization to mismatched oligonucleotides, expressed in arbitrary units. On average, >25% of the genes under investigation were positive in the cell lines tested. A threshold of 20 normalized average difference units was assigned to any gene with a calculated expression level <20, because mRNA levels in this low range could not be reliably assessed.

CXCL1, CCL5, IL-1β, IL-6, and IGF-II production was determined by quantitative ELISA assays (sensitivity, ≥10 pg/ml) using cell supernatants of confluent cultures. Antibody pairs and standards were provided by BD Biosciences or R&D Systems. ISG15 was detected by ELISA assay (19). Release of type I IFNs was quantified on HeLa cells as described elsewhere (21). Assays were performed on coded samples.

The original tumor specimen from which Me67 melanoma lines were derived, conserved as paraffin-embedded material, was retrieved and analyzed as follows. Serial sections were incubated overnight at 4°C with anti-ISG15 or isotype-matched control antibodies, followed by avidin-biotin-peroxidase complex (ABC; Vectastain, Burlingame, CA).

We have shown previously that melanoma cell lines can activate monocyte-derived DCs in vitro (9). To obtain an insight into the molecular mechanisms underlining these modifications, we first analyzed the effects of coculturing Me67 melanoma cells, their conditioned culture medium, or their lysates with human DCs. Induction of E-cadherin, as well as the increased expression of CD15 and CD86 on DCs was observed in the presence of tumor cells or their culture supernatant, whereas cell lysates were unable to induce these effects (Fig. 1 A). Thus, phenotypic DC modulation induced by Me67 was cell contact independent and relied on the production of active soluble factors present in the cell culture medium.

To identify the molecular entity(s) responsible for these effects, specific cell clones were derived from Me67 parental cell line. Supernatants from clones Me67.3 and Me67.9 induced E-cadherin expression, whereas clones Me67.5 and Me67.10 were devoid of any modulatory effect on DC phenotype (Fig. 1 B). These clonal cell populations, all expressing the same genetic background, were subjected to gene profiling to identify differentially expressed gene sequences.

Total cellular RNA extracted from Me67.3, Me67.9, Me67.5, and Me67.10 melanoma clones was processed for hybridization to oligonucleotide arrays containing probe sets from ∼7000 full-length human genes. A total of 52 expressed genes were selected based on a >3-fold changes between E-cadherin inducing (Me67.3 + Me67.9) and noninducing (Me67.5 + Me67.10) clones and grouped according to their putative physiological function (Table 1). Most of the genes identified were specific for membrane-associated, intracellular, or nuclear proteins and therefore excluded from further analysis.

A total of 11 genes encoding secreted proteins were found to be expressed at higher levels in clones Me67.3 and Me67.9 than clones Me67.5 and Me67.10 (Table 1). Among these, 5 were specific for proteins present in the extracellular matrix such as peptidoglycan, collagen, fibronectin, and mucin (Table 1). Because these proteins are involved in the formation of three-dimensional structures and poorly detectable in solution, they were excluded from further investigations.

The other genes were analyzed in detail. Two genes for chemotactic factors (GRO-α/CXCL1 and RANTES/CCL5) and 4 encoding cytokines (IL-1β, IL-6, IGF-II, and ISG15) were strongly expressed in clones Me67.3 and Me67.9 but barely detectable in clones Me67.5 and Me67.10 (Table 1).

When the corresponding conditioned melanoma cell culture media were assessed for proteins detection, the chemokine CXCL1 and the cytokine IL-1β were found to be produced to different extents by three of the melanoma lines (Fig. 2, A and C, respectively). No quantitative differences existed for IL-6 and IGF-II secretion among the different melanoma clones (Fig. 2, D and E). Considering these patterns, the involvement of these factors in the modulation of DC phenotype was unlikely. On the contrary, production of the chemokine CCL5 and the cytokine ISG15 was consistently different among the two groups of melanoma clones. More than 4000 pg/ml of CCL5 and >7000 pg/ml of ISG15 were measured in the supernatants of clones Me67.3 and Me67.9 but were undetectable in supernatants from clones Me67.5 and Me67.10 that were devoid of any DC-conditioning capacity.

Moreover, genes encoding IFN-inducible proteins (i.e., rig-g, the IFN-induced M_r 56,000 protein encoding sequence, ISG20, trip14, 27-sep, and 16-jun) were consistently expressed in clones Me67.3 and Me67.9 but poorly amplified in clones Me67.5 and Me67.10. Indeed, the two groups of clones differed for the capacity to secrete type I IFNs; clones Me67.9 and Me67.3 produced 64 and 128 IU/ml of type I IFNs, respectively, whereas Me67.5 and Me67.10 clone supernatants were completely negative (Fig. 2 G). Considering the immunomodulatory properties of type I IFNs, these data might be of relevance also in the experimental system described here.

Altogether, by combining the gene chip technology with quantitative protein detection assays, we were able to identify three soluble factors (i.e., RANTES, ISG15, and type I IFNs most likely involved in the DC phenotypic modulation induced by Me 67.3 and Me67.9 conditioned media.

To define to what extent RANTES, ISG15, and/or type I IFNs induced DC phenotypic modulation, the cellular in vitro assays were repeated in the presence of specific neutralizing antibodies. The following experiments crucially depended on the use of reagents unable to induce any modification of DC phenotype per se, e.g., devoid of endotoxin contamination. Commercially available anti-CCL5 antibodies were excluded from our tools because they were found to induce up-regulation of CD83 and CD86 on immature DCs. Anti-ISG15, anti-IFNARI, and anti-CCR5 antibodies were suitable for our assays.

Neutralization of ISG15 markedly suppressed the melanoma-conditioned modulation of DC phenotype (Table 2). The results of three different experiments, independently performed, confirmed that ISG15 played a crucial role in inducing expression of E-cadherin and strongly influenced up-regulation of CD15 induced by both Me67.3 and Me67.9 conditioned media (Table 2, Exp. I, II, and III). In contrast, the up-regulation of CD86 expression on melanoma-conditioned DCs was not reproducibly dependent on ISG15. Antibodies to IFN receptors did not inhibit E-cadherin induction and CD15 expression but slightly affected the up-regulation of CD86 induced by both Me67.3 and Me67.9 conditioned media (Table 2, Exp. IV). Finally, because anti-CCR5 antibodies did not interfere with the induction of E-cadherin induced by Me67.3 conditioned medium (Table 2, Exp. V), involvement of CCL5 in conditioned media DC phenotypic modulation can be unlikely.

In previous work, tumor-infiltrating CD15⁺ cells with a typical DC morphology were identified within malignant melanoma tissues. The potential in vivo relevance of our in vitro findings relied on the expression of ISG15 in original tumor specimens. Indeed, immunohistochemical staining of Me67 metastatic melanoma tissue on paraffin sections revealed that tumor cells were strongly positive for ISG15, in contrast to tumor-infiltrating lymphocytes that were completely negative (Fig. 3,A). In malignant cells, ISG15 protein was localized within the cytoplasm (Fig. 3,B). WM9 melanoma human cells grown in nude mice as xenografts were examined, in the same settings, as specific control. These cells were strongly positive in immunocytochemistry only upon treatment with IFN-β, a potent inducer of ISG15 (Fig. 3, C and D, respectively). Altogether, eight tumor specimens were analyzed in immunohistochemistry. Among these, four samples were found strongly positive, three showed a moderate positive staining, and one specimen was negative (data not shown). These findings suggest that constitutive expression of ISG15 by malignant cells might frequently occur in vivo, at least in melanomas.

In this study, we explored the capacity of melanoma cell-derived proteins to induce phenotypic modulation of human monocyte-derived DCs (9). We performed microchip analysis on cellular reagents, all from the same genetic background, combined with standard ELISA assays, and this narrowed our search from 52 genes most likely involved in the DC phenotypic modulation induced by melanoma-conditioned media down to three soluble products, ISG15, type I IFNs, and CCL5.

ISG15 cytokine is strongly induced by IFN-α/β stimulation in different types of cells, including epithelial tumor cell lines in vitro (19, 22), and requires a functional proteasome (23). Its expression in malignant tissues was, thus far, never investigated. ISG15 is synthesized as a M_r 17,000 precursor protein and processed to a mature M_r 15,000 product by cleavage of the COOH-terminal amino acidic tail (24). Mature ISG15 may be released as monomer or in the form of high molecular weight conjugates (25). At present, the nature and the biological functions of these conjugates are not fully understood (26), although the immunoregulatory properties of ISG15 have been demonstrated on T lymphocytes and natural killer cells. In particular, exposure to ISG15 in vitro induces IFN-γ production by T lymphocytes and proliferation of natural killer cells (27).

Our experimental data support the hypothesis that ISG15 is crucially involved in the modulation of DC phenotype induced by melanoma cell conditioned medium: (a) the ISG15 gene was expressed at higher magnitude in tumor cells that induced DC phenotypic modulation than in cells devoid of this capacity; (b) high amounts of ISG15 protein were detected in the corresponding conditioned media and, remarkably, in original tumor specimens; and (c) most importantly, de novo expression of E-cadherin on monocyte-derived DCs in vitro was not inducible in the presence of anti-ISG15 antibodies, and up-regulation of CD15 and CD86 were strongly hampered in these conditions. In contrast, neutralization of type I IFN receptors partially inhibited CD86 up-regulation, confirming previous results (6), but did not affect E-cadherin induction; anti-CCR5 antibodies did not have any effect. Finally, in vitro exposure to recombinant CCL5 (used in the range 10–1000 ng/ml) and IFN-α2a (from 10 to 1000 IU/ml) never resulted in E-cadherin expression on immature DCs (data not shown). Thus, the most straightforward explanation of these results is that E-cadherin induction required ISG15 expression, whereas CD86 up-regulation relied on the contribution of both ISG15 and tumor-derived type I IFNs.

Impaired DC mobility, eventually caused by E-cadherin expression, and sequestration of DCs into malignant tissues have been described previously as potential mechanisms of tumor immune escape not affecting antigen-presenting capacities (8, 9). These data could have important implications for cancer immunotherapy. As interventional strategy, neutralization of tumor-derived ISG15 might be of difficult application. On the contrary, targeting the migratory pattern of immunocompetent cells might represent a possibility of immune intervention. Regarding DC mobility, it has been shown that cytokines promoting mobilization of circulating DCs (i.e., Flt3 ligand) could increase infiltration of DCs in the peritumoral region and enhance the responses to recall antigen in cancer patients (28). However, the newly recruited DCs might also become the target of tumor-derived factors and be sequestered into the malignant tissue. In this context, the use of cytokines promoting recruitment and expansion of effector T cells could be suggested to counteract DC immobilization (6, 29). The aim will be the concomitant intratumoral infiltration of DC and T lymphocytes that has been related to more favorable prognosis in different types of malignancies (10, 12).

Thanks to M. Ferrantini, B. Hassel, and A. L. Haas for providing reagents and for helpful discussions and to E. Shultz and P. Zajac for technical help.