Selective Accumulation of Mature DC-Lamp⁺ Dendritic Cells in Tumor Sites Is Associated with Efficient T-Cell-Mediated Antitumor Response and Control of Metastatic Dissemination in Melanoma

There is now accumulating evidence that many human cancers elicit antitumor immune responses. Indeed, isolation of specific CTLs has allowed cloning and characterization of genes that encode human tumor-associated antigens in different cancer types (1, 2). Melanoma is considered one of the most immunogenic tumor models because numerous specific and shared tumor antigens have been identified, and tumor-specific CTL responses contributing to spontaneous tumor immunosurveillance have been reported (3). However, few patients receiving the peptide-based vaccines developed to date have shown significant specific immune responses in correlation with clinical regression (4).

Mature dendritic cells (DCs) are the most potent antigen-presenting cells (5) capable of priming tumor-specific T cells, and their use in cancer immunotherapy appears to be a promising way to elicit and expand efficient antitumor immune responses (6, 7). However, little is known about the role of DCs in the natural immunosurveillance of cancer. It is generally postulated that spontaneous immunity against tumor antigens is elicited by DCs maturing in response to various inflammatory signals, but the sites and conditions in which DCs capture and/or encounter tumor antigens remain largely unknown. Indeed, most studies have addressed this question by use of nonspecific antibodies (8), whereas other reports describe the localization and phenotypes of DCs in tumor samples without clinical correlates (9).

Melanoma has become a paradigm for studying the interactions between tumor and host immune cells for the following reasons. First, the primary lesion occurs in the skin, the main residency compartment for Langerhans cells, which have high potential to migrate to skin-draining lymph nodes (LNs; Refs. 10, 11). Second, sentinel LNs (SLNs) are privileged sites of both T-cell priming and initial metastases, the extent of SLN involvement representing an independent predictive factor for survival (12, 13). Third, cutaneous and LN metastases appear to be the tumor sites most sensitive to vaccine-based immunotherapies (6, 14).

Using a panel of antibodies recognizing DC subsets and maturation markers, we addressed the clinical relevance of DC infiltration on regressing tumor lesions and SLNs in melanoma patients. Here we demonstrate, in a case report involving spontaneous regression of metastatic melanoma, dramatic infiltration of DC-Lamp⁺ DCs in close contact with effector memory Melan-A/Mart-1-specific CTLs in situ. In addition, in a series of 19 metastatic SLNs, we also show that the extent of DC-Lamp⁺ DC infiltration is significantly associated with a high incidence of tumor-free status after lymphadenectomy and a favorable clinical outcome. Altogether, our results strongly suggest that mature DCs play a key role in the local expansion and differentiation of protective tumor-specific immune responses at the first sites of melanoma metastasis.

All materials collected in this study were obtained after receipt of informed signed consent, as approved by the Institutional Review Board at the Gustave Roussy Institute and by the Legal Ethical Committee.

The patient was a 28-year-old male with an atypical nevus syndrome who in December 2000 developed a right axillar lymphadenopathy corresponding to a melanoma metastasis. The patient reported in the previous months a pigmented nodular lesion in the right arm that had regressed spontaneously. This residual lesion was removed, and a lymphadenectomy in the right axillar basin was performed. Histologically, the cutaneous lesion displayed features typical of regression: absence of associated melanoma cells, a flat epidermis and a fibrotic dermis containing ectasic vessels, heterogeneous melanin pigmentation, and a dense lymphocytic infiltrate. The lymphadenectomy revealed 8 metastatic LNs from a total of 24. After a 10-month disease-free period, a right subclavicular mass developed that corresponded, at surgical examination, to a LN 2 cm in diameter. Histologically, this LN was free of tumor cells and appeared almost completely necrotic with extensive peripheral fibrosis. LN samples were kept frozen for further analyses, and the primary cutaneous site was paraffin-embedded only. This patient remains disease-free 34 months after diagnosis without any adjuvant therapy.

All patients underwent surgical SLN resection at the time of or a few weeks after excision of the primary cutaneous melanoma, according to the recommended procedures (12). For this study, we selected a consecutive series of 19 patients with a melanoma-positive SLNs followed in our institution between June 1999 and March 2002. A few weeks after the SLN procedure, all underwent a regional completion lymphadenectomy to detect metastatic extensions in non-SLNs. The materials used consisted of paraffin-embedded sections from recorded paired SLN and non-SLN samples.

HLA-A*0201/Melan-A/Mart-1_26–35 A27L analog (ELAGIGILTV) and control HLA-A*0201/HIVgag_77–85 (SLYNTVATL) were synthesized as described previously (15). Total peripheral blood mononuclear cells or immunopurified CD8⁺ T cells, obtained after magnetic cell sorting with a MiniMACS device (CD8⁺ T Cell Isolation Kit; Miltenyi Biotec, Bergisch Gladbach, Germany), were stained with the PE-labeled tetramers (5 nm) at room temperature for 45 min in PBS containing 2% BSA, washed once, stained for cell surface differentiation antigens with monoclonal antibodies (mAbs) for 30 min at 4°C, washed again, and analyzed by flow cytometry. Conjugated anti-CD3, -CD8, and -CD45RA mAbs were purchased from BD Biosciences (Le Pont de Claix, France), and anti-TCRVβ2 and -TCRVβ3 were purchased from Beckman Coulter (Villepinte, France). Samples were also incubated with anti-CCR7 rat IgG2a mAb 3D12 (kindly provided by M. Lipp, Berlin, Germany) for 30 min at 4°C, washed, and additionally stained with an antigen-presenting-cell-labeled goat antirat IgM+IgG (Southern Biotechnology, Birmingham, AL). Fluorescent isotype-matched mAbs were used as controls. Direct ex vivo analysis was performed after acquisition of >10⁶ events on a FACSCalibur (BD Biosciences) with Cellquest software.

CD3⁺CD8⁺HLA-A*0201/Melan-A/Mart-1_26–35⁺ cells were sorted by flow cytometry with FACS Vantage (BD Biosciences) and cultured in V-bottomed 96-well plates either in bulk or after direct cloning in limiting dilution on feeder cells consisting in irradiated allogeneic peripheral blood mononuclear cells, EBV-transformed LAZ509 B cells, and MEL-FON tumor cells, a HLA-A2⁺ melanoma tumor cell line expressing the Melan-A/Mart-1 antigen (15). Growing clonal colonies were expanded at 37°C and 5% CO₂ in RPMI 1640 (Life Technologies, Inc., Paisley, United Kingdom) supplemented with 10% human AB serum (Institut Jacques Boy, Reims, France), 50 IU/ml recombinant human interleukin-2 (Proleukin; Chiron, Rattingen, Germany), and 3% T-cell growth factor, a supernatant from phytohemagglutinin/phorbol 12-myristate 13-acetate-allostimulated T cells. T-cell clones were characterized for immunophenotype by use of HLA-A*0201/Melan-A/Mart-1_26–35 tetramers and mAbs directed against CD3 and CD8 and by TCR-Vβ repertoire analysis with a panel of 21 anti-BV mAbs (IOTest β Mark; Beckman Coulter Immunotech, Marseille, France), and for cytotoxic activity in a standard 4-h ⁵¹Cr-release assay against MEL-FON tumor cells and Transporter associated with Antigen Processing (TAP)-deficient EBV-transformed T2 cells loaded with 10 μg/ml Melan-A/Mart-1_26–35 synthetic peptide. IFN-γ secretion in response to the same targets was also assessed by use of the high-affinity capture matrix assay according to the manufacturer’s instructions (IFN-γ secretion assay; Miltenyi Biotec).

Tissue samples were snap-frozen immediately after collection and cryopreserved in liquid nitrogen until use. Serial multiple 10-μm cryostat tissue sections were used for in situ tetramer staining. To block nonspecific binding sites, tissue sections were first recovered with 200 μl of 1× PBS containing 0.5% BSA and 10% FCS for 15 min at room temperature. After elimination of the buffer, PE-labeled HLA-A*0201/Melan-A/Mart-1_26–35 or peptide-irrelevant tetramer complexes were incubated at 2.6 μg/ml for 45 min at room temperature in a dark humidified atmosphere. The sections were washed with PBS and incubated with direct fluorescence or biotin-conjugated mAbs (anti-CD3, -CD8, -TCRVβ2, and -TCRVβ3) for an additional 30 min at 4°C. For triple staining, the biotin-coupled antibody was detected with streptavidin-Cy5.5 conjugate. After washing, sections were fixed with PBS-buffered 1% formaldehyde for 15 min at room temperature, mounted in antifading medium, and covered with a coverslip. Slide examination and micrography were performed with a fluorescence microscope (DMRXA; Leica, Weitzlar, Germany), and/or slides were evaluated with a Leica SP1 confocal microscopy system equipped with an Ar/Kr laser for the 488 and 568 nm lines and a HeNe633 laser (Leica Laser Technik; Mannheim, Germany) to excite the Cy5.5 label. Simultaneous acquisition of signals for the FITC-, PE-, and Cy5.5-labeled antibodies was performed at low illuminating laser power and with spectrophotometer-based separation of the fluorescence emissions. The fields of interest were selected on the basis of CD8- or TCRVβ2-FITC-labeled infiltrates to avoid bias in scoring of tetramer-PE staining before confocal image acquisition.

Sections 4-μm thick were prepared on glass slides from paraffin-embedded formaldehyde-fixed material and were deparaffinized and rehydrated. For all antibodies except for anti-CD123, -CCR7, and -CD68, slides were pretreated in sodium citrate at 68°C (pH 6) for 20 min. After a blocking step (Blocking agent, Ultra-tech HRP kit; Beckman-Coulter Immunotech), slides were incubated with the primary antibodies at the following dilutions: 1:1 for anti-CD1a and anti-CD68 (Immunotech); 1:20 for anti-MIP3α/CCL20 (Schering-Plough, Dardilly, France); 1:50 for anti-CD3, anti-CD4 (Dako, Trappes, France), and anti-MIP3β/CCL21 (R&D Systems, Abingdon, United Kingdom); 1:60 for anti-p55 (Dako); 1:100 for anti-CD8 (Dako) and TIA-1 (Beckman Coulter); 1:200 for anti-Langerin (Schering-Plough) and anti-CCR7 (BD PharMingen); 1:300 for anti-DC Lamp and anti-CD123 (Schering-Plough); and 1:500 for anti-CCR6 and anti-6CKine/CCL19 (R&D Systems) for 1 h at room temperature, followed by incubation with secondary antibodies and detection (Ultra-tech HRP kit). The slides were examined by two independent observers. Semiquantitative evaluation of DC-Lamp⁺ DC infiltration was made as follow: (a) the slide was examined at intermediate magnification (×100) to select the three areas containing the maximum number of DC-Lamp⁺ DCs; (b) the number of DC-Lamp⁺ cells per mm² was then counted at high magnification (×400) in 3-mm² sections per hot spot; (c) the final count used for DC-Lamp quantitative evaluation was the average number of the nine countings made on the slide. Statistical analyses comparing the groups according to DC-Lamp⁺ DC infiltration used Fisher’s exact method, and statistically significant differences were given at 95% confidence.

We report here the case of a young HLA-A*0201-positive patient presenting with a completely regressive primary cutaneous melanoma associated with regressing regional metastatic LNs. Postulating that this favorable clinical outcome might have resulted from an efficient antitumor immune response, we thus focused this study on the peripheral CTL response directed against the Melan-A/Mart-1 tumor antigen strongly expressed at the site of the first metastatic LNs, as assessed by reverse transcription-PCR (data not shown). As shown in Fig. 1,A, direct ex vivo staining of peripheral CD3⁺CD8⁺ T lymphocytes with HLA-A*0201/Mart-1_26–35 tetramers revealed 0.15% specific T lymphocytes, which produced four tetramer⁺ T-cell clones. Two expressed TCR-Vβ2⁺ with specificities for BV2J2.6 and BV2J2.7 gene segments, one expressed Vβ3⁺ (BV3J1.5), and one expressed Vβ13⁺ (BV13; Fig. 1,B). All of these CTL clones recognized not only TAP-deficient T2 cells pulsed with Melan-A/Mart-1_26–35 peptide, but also the naturally processed Melan-A/Mart-1 antigen presented by the HLA-A2⁺ MEL-FON melanoma cell line, as assessed by cytotoxicity (Fig. 1,B) and IFN-γ secretion assays (data not shown). Interestingly, half-maximal lysis of peptide-pulsed T2 cells was observed with peptide concentrations ranging from 5 × 10⁻⁹ to 5 × 10⁻¹⁰ m, demonstrating the high T-cell receptor (TCR) avidity of these clones (data not shown), which were all brightly stained with the tetramers (Fig. 1 A). Such high-avidity TCRs have been reported to be expressed by melanoma-reactive CTLs, whereas CD8⁺ T cells with impaired proliferative and effectors functions mostly display low-avidity TCRs (16).

Naive central and effector memory T lymphocytes can now be discriminated to delineate the stages of lymphocyte differentiation after antigen encounter and stimulation in vivo (17). To characterize the differentiation status of the HLA-A*0201/Mart-1_26–35 tetramer⁺ T lymphocytes found in this patient, we analyzed peripheral blood mononuclear cell-derived CD8⁺ T lymphocytes by four-color immunophenotyping using anti-CD45RA, -CCR7, and -TCRVβ mAbs. As shown in Fig. 1,C, we confirmed the predominance of naive (CD45RA^brightCCR7⁺) over memory (CD45RA^−/dimCCR7⁻) anti-Melan-A/Mart-1 T cells (18, 19). However, although the majority of tetramer⁺Vβ3⁺CD8⁺ T cells appeared to be naive, most te-tramer⁺Vβ2⁺ cells displayed a memory phenotype (Fig. 1 C). We thus addressed the relevance of such a melanoma-reactive memory T-cell response with high-avidity TCRVβ2⁺ cells in the spontaneously regressing lesions.

The lymphocytic infiltrate of the metastatic LNs contained a majority of CD3⁺CD8⁺ lymphocytes displaying a dominant Vβ2 specificity (data not shown). Importantly, a high proportion of lymphocytes were stained with the TIA-1 mAb, which recognizes perforin/granzyme-containing granules that are preferentially associated with memory/effector cytotoxic functions (Fig. 2,A). To determine the Melan-A/Mart-1 specificity of tumor-infiltrating CTLs in the first metastatic LN, we performed a double staining with PE-labeled HLA-A*0201/Mart-1_26–35 tetramer and FITC-labeled anti-CD8 or anti-Vβ2 mAbs. We observed significant numbers of tetramer⁺CD8⁺ T lymphocytes (e.g., >10 cells at high magnification) in the tumor areas (Fig. 2, B and C), approximately half of which were Vβ2⁺ cells (Fig. 2,E). However, a more precise quantification of specific T cells by flow cytometry analysis could not be performed because fresh LN fragments were not available at the time of lymphadenectomy. The specificity of tetramer staining was assessed with a peptide-irrelevant PE-labeled HLA-A*0201/HIVgag_77–85 tetramer on serial sections of the same LN (Fig. 2,D) and with the relevant PE-labeled HLA-A*0201/Mart-1_26–35 tetramer on a tumor-free alternative LN from the same axillary area (Fig. 2,F). Confocal microscopy revealed in double-positive cells an overlay of FITC and PE labeling concordant with the colocalization of TCR signals obtained with tetramers and anti-TCR mAbs (Fig. 2 E). Of note, tetramer staining was frequently polarized toward the T-cell membrane directly in contact with adjacent tumor cells with occasional intracellular vesicles, suggesting capping and/or recycling of TCR-containing vesicles. These features may also correspond to dense accumulation in lipid rafts and engagement of TCRs, an important process for T-cell activation; lipid raft integrity is also required for efficient MHC class I tetramer binding (20).

Although direct visualization by use of fluorescent MHC class I/peptide complexes of antigen-specific T cells on fresh tissue sections has already been reported (21, 22), the clinical application of this technique to human tumors remains limited by the very low frequency of infiltrating CTLs that are specific for a given epitope. Here we demonstrate, however, that such in situ tetramer staining can be applied successfully on informative sites, such as a metastatic LN draining a spontaneously regressive primary melanoma. The close contact of Melan-A/Mart-1-specific T lymphocytes with tumor cells suggests their active role in the efficient natural immunosurveillance ongoing in this patient.

The availability of a panel of mAbs directed against DC surface molecules enabled staining of paraffin-embedded tissue sections and allowed us to characterize the infiltrating DCs in the different samples collected from this patient. We first examined two Langerhans cell markers, CD1a and the lectin Langerin/CD207, which are both expressed by immature Langerhans cells. As for the noninvaded adjacent skin, numerous intradermal CD1a⁺ and Langerin⁺ cells were detected in the epidermis of the primary regressing cutaneous lesion (data not shown), but very few were observed in the draining metastatic LNs. Such findings contrasted with the detection of significant numbers of CD1a⁺ and Langerin⁺ cells in the subcapsular areas of normal LNs collected from other donors (data not shown). In addition, in contrast to what is also seen in normal LNs, we found no CD123/IL-3Rα⁺ plasmacytoid DCs in the LNs of our patient (not shown).

We also examined the expression of the actin-bundling protein p55/fascin and the lysosomal protein DC-Lamp, both of which are associated with a mature DC phenotype. In the metastatic LN (Fig. 3,A) as well as in the residual lymphocytic regions of the necrotic LN (not shown), we observed dense infiltration of p55⁺/DC-Lamp⁺ double-positive cells within the T-cell-rich regions and peritumoral areas, which contrasted with the scattered and mostly subcapsular localization in a nonmetastatic LN (Fig. 3,B). These cells exhibited the typical morphology of “mature” DCs and were surrounded by T lymphocytes forming clusters (Fig. 3, C, E, and G) that have been described as characteristic of ongoing immune reactions and have been observed at the periphery of human breast tumor cell clusters (9). DC-Lamp⁺ (not shown) or p55⁺ DCs (Fig. 3,G) frequently also accumulated around neomicrovessels within tumor areas. We made similar observations in the primary skin site, where p55⁺ DCs were often surrounded by lymphocytes in the regression area but not in the normal adjacent dermis (not shown), suggesting local recruitment and/or differentiation of DC precursors from the peripheral blood. Interestingly, in a metastatic LN (Fig. 3,C), the cellular localization of DC-Lamp was diffuse, including staining of peripheral dendrites, in contrast to that observed in a nonmetastatic LN (Fig. 3 D), in which the DC-Lamp staining appeared mostly concentrated in a “dot-like” juxtanuclear region of the cells. Considering that DC-Lamp has previously been reported to accumulate in juxtanuclear late lysosomal compartments during maturation (23), these findings suggest that an efficient immune response is more likely associated with a maturing DC phenotype rather than with a fully mature differentiation stage. Although the function of DC-Lamp has not been elucidated, our results indicate that its subcellular localization and/or its abundance on DC membranes in close apposition with tumor cells could be beneficial for tumor antigen capture and presentation or is associated with the presence of highly active antigen-presenting cells.

The scenario involving DCs, T lymphocytes, nonspecific effector cells such as macrophages, and tumor cells orchestrated to elicit an effective immune response in situ remained elusive. Nonetheless, we frequently observed veiled dendritic p55⁺ or DC-Lamp⁺ projections tightly apposed to tumor cells (Fig. 3, E and F), reminiscent of data from experimental in vitro systems in which DCs ingested tumor cell bodies and processed tumor-associated antigens to elicit specific immune responses (10). Because these DCs also interacted with several surrounding lymphocytes, one can postulate that such interactions may trigger in situ activation, expansion, and effector functions of antigen-specific T cells, culminating, as in this patient, in tumor regression.

Considering their critical role in DC chemoattraction, we assessed the expression of the CC chemokine receptors (CCR) CCR6 and CCR7 and their respective ligands, MIP-3α/LARC (CCL20), 6CKine/SLC (CCL19), and MIP-3β/ELC (CCL21). In the metastatic LN, we found that CCR6 was expressed predominantly within tumor areas by cells exhibiting a dendritic morphology (Fig. 4,A). Accordingly and as described previously (9), tumor cells and macrophages strongly expressed the CCR6 ligand CCL20 and clearly colocalized with CCR6⁺ DCs (Fig. 4,B). CCR7 was expressed by a few cells in tumor areas, i.e., tumor cells and lymphocytes, but not by infiltrating DCs (Fig. 4,C). In contrast, large dendrite-shaped CCR7^bright cells were detected occasionally in the T-cell-rich areas without tumor cell infiltration (Fig. 4 D), and the CCR7 ligands CCL19 and CCL21 were slightly positive on DCs and stromal and endothelial cells of microvessels (not shown). During their migration from blood to peripheral tissues, CCR6⁺ DC precursors appear to be attracted by a CCL20 gradient that is established by local inflammation (5, 10, 24). Peripheral immature DCs may then undergo a maturation process characterized by several modifications, including the reciprocal loss of CCR6 and acquisition of CCR7, with resulting responsiveness to CCL19/CCL21 (24). However, whether DC chemoattraction to LNs and DC maturation are concomitant processes remains controversial (25, 26). As reported in psoriasis (27), rheumatoid arthritis (28), and in some reports, cancer (9, 29), our findings of dominant CCR6/CCL20 expression suggest that the tumor-invaded LN should be considered as an inflammatory site rather than a typical lymphoid organ. Therefore, in this metastatic LN, maturing DCs might indeed display significant heterogeneity regarding expression of CCR receptors and could even transiently coexpress CCR6 and CCR7 in an intermediate stage of maturation, as has been reported recently (30). Altogether, these data point to a potential role of MIP3α/CCL20 in the local recruitment of DCs in metastatic LNs.

The most striking observation associated with spontaneous tumor regression in this case report was the dramatic accumulation of DC-Lamp⁺ mature DCs in metastatic LNs. However, because this finding was based on one informative patient, we assessed the presence of mature DCs in a series of 19 additional melanoma-positive SLNs. Because SLNs are identified as the first site of melanoma metastases, we postulated that the extent of mature DC infiltration may dictate the induction of efficient local antitumor immune responses. The following parameters were thus evaluated by two independent observers: (a) semiquantitative evaluation of the absolute number of DCs counted at sites of highest SLN infiltration, which led to the identification of two groups of patients: SLN/DC-Lamp^low (+ or ++, i.e., <100 or 100–300 DC-Lamp⁺/mm², respectively; n = 11) and SLN/DC-Lamp^high (+++, i.e., >300 DC-Lamp⁺/mm²; n = 8); (b) the aspect of DC-Lamp staining, i.e., dendritic or dot-like, that could correspond to a maturing or fully mature phenotype, respectively; (c) tumor infiltration of DC-Lamp⁺ cells and evidence for clusters of DC-Lamp⁺ cells with tumor cells and lymphocytes (in 10 of the 19 melanoma-positive SLNs in which tumor cells could be visualized). The results are summarized in Table 1. In the SLN/DC-Lamp^low group, DC-Lamp⁺ cells displayed a predominant dot-like lysosomal staining, whereas in the SLN/DC-Lamp^high group, DC-lamp⁺ cells displayed a predominant dendritic staining (P = 0.019) and an increased frequency of tumor infiltration and DC clustering, as mentioned in the case report. Furthermore, there was a striking inverse correlation between the density of DC-Lamp infiltrates and the occurrence of at least one melanoma-positive non-SLN in the subsequent regional completion lymphadenectomy performed a few weeks after the SLN resection. Seven of 11 patients in the SLN/DC-Lamp^low group had at least one metastatic non-SLN, in contrast to none of 8 patients in the SLN/DC-Lamp^high group (P = 0.013). We will attempt to validate this correlation between the presence of mature DCs in the initial metastatic lesion and the absence of secondary metastatic spread in a comprehensive multicenter study aimed at evaluating the prognostic value of DC-Lamp and other dendritic cell markers in disease-free and overall survival of a large cohort of patients with primary cutaneous melanoma.

Our data strongly suggest that mature DC-Lamp-expressing DCs play a pivotal role in melanoma immunosurveillance at the initial site of metastasis and in eliciting the generation of an effective antitumor response with local T-cell activation/expansion. These observations may also have implications regarding the identification of new prognostic groups of melanoma patients with high infiltration of DC-Lamp⁺ DCs in SLNs, in whom immunotherapy in early adjuvant settings could be successful.

We gratefully acknowledge Christophe Le Boullaire and Valérie Vélasco for excellent technical assistance, Patrick Bruneval for constant support in confocal microscopy, and Yann Lecluse and Pascal Batard for flow cytometry cell-sorting experiments.