Abstract
The chemokine axis CCR6/CCL20 is involved in cancer progression in a variety of tumors. Here, we show that CCR6 is expressed by melanoma cells. The CCR6 ligand, CCL20, induces migration and proliferation in vitro, and enhances tumor growth and metastasis in vivo. Confocal analysis of melanoma tissues showed that CCR6 is expressed by tumor cells, whereas CCL20 is preferentially expressed by nontumoral cells in the stroma of certain tumors. Stromal CCL20, but not tumoral CCR6, predicted poor survival in a cohort of 40 primary melanoma patients. Tumor-associated macrophages (TAM), independently of their M1/M2 polarization profile, were identified as the main source of CCL20 in primary melanomas that developed metastasis. In addition to CCL20, TAMs expressed TNF and VEGF-A protumoral cytokines, suggesting that melanoma progression is supported by macrophages with a differential activation state. Our data highlight the synergistic interaction between melanoma tumor cells and prometastatic macrophages through a CCR6/CCL20 paracrine loop. Stromal levels of CCL20 in primary melanomas may be a clinically useful marker for assessing patient risk, making treatment decisions, and planning or analyzing clinical trials. Cancer Immunol Res; 6(3); 267–75. ©2018 AACR.
Introduction
Chemokines and their receptors are involved in multiple aspects of cancer progression. In addition to their chemotactic role guiding the metastatic process, chemokines can provide survival signals to cancer cells that promote tumor growth in otherwise nonsupportive microenvironments. Chemokines can also indirectly modulate tumor growth through their effect on nontumor cells in the stroma by inducing release of growth and angiogenic factors and by directing the trafficking of both antitumor and protumor leukocytes. Chemokines also participate in the response to cancer therapies and are targets for immunotherapy and chemokine-targeted therapy (1).
Cutaneous melanoma is an aggressive skin tumor in which chemokines and their receptors regulate progression (2). Originally described as autocrine factors for melanoma growth, the chemokines CXCL8 (IL8) and CXCL1-2 (GRO1-2), as well as their respective receptors CXCR1 and CXCR2, regulate melanoma tumor progression by affecting tumor growth, angiogenesis, and metastasis (3). CXCR4 expression renders melanoma cells responsive to CXCL12 and drives local invasion and metastasis (4, 5), and its expression in primary melanoma patients is associated with a higher metastatic risk and mortality rate (6). Expression of certain chemokine receptors by melanoma cells is associated with selective organ metastases with localization dependent on expression of particular ligands: CCR7 is associated with lymph node, which is rich in CCL19 and CCL21; CCR9 is associated with small intestine, which is rich in CCL25; and CCR10 is associated with metastasis in the skin, which is rich in CCL27 (7, 8).
Overexpression of CCR6 by cancer cells and aberrant signaling by its ligand CCL20 has been documented in many cancer types including colorectal, hepatocellular, pancreatic, ovarian, nasopharyngeal, and breast cancer (9–12). The purpose of this study was to characterize CCR6/CCL20 signaling and expression in human melanoma. We found that CCR6 is expressed by melanoma cells, whereas CCL20 is expressed by stromal cells in certain primary cutaneous melanomas. CCL20 stromal expression was associated with metastatic progression and patient survival. The stromal source of CCL20 in primary melanomas that developed metastasis was identified as protumoral TAMs. These cells coexpress TNF and VEGF, independently of their M1 or M2 profile.
Materials and Methods
Cells and antibodies
Human melanoma cells and associated fibroblasts were isolated from biopsied metastatic lesions, as previously described (13). TAMs and tumor-infiltrating lymphocytes (TIL) were purified by magnetic cell sorting using CD14 and CD3 microbeads (Miltenyi-Biotech) from human-homogenized melanomas (Miltenyi-Biotech), respectively. Monocytes were isolated with CD14 microbeads from healthy donors or melanoma patients' peripheral blood. For in vitro generation of macrophages and dendritic cells (DC), monocytes were incubated with GM-CSF (M1), M-CSF (M2), or GM-CSF/IL4 (iDC) for 7 days renewing cytokines every 3 days, as previously described (14). All cells, melanoma cell lines Mewo, BLM, A375, MV3, Skmel103, and Skmel147 (15); MEL-F melanocytes, Caco-2, NCI-H929, and Hek-293 (from ATCC) were grown in RPMI-1640 medium (Gibco) supplemented with 10% FCS (Harlan Sera-lab). Cells lines were mycoplasma-free and used before fifth passage from thawing. Cell lines were not reauthenticated during the last year, but checked for melanoma markers (HMB45, NG2, S100). Primary antibodies are specified in Supplementary Table S1.
In vitro measurements
For coculture assays, 5 × 104 melanoma cells were first cultured for 3 days in 12-well plates, and then 106 macrophages were added to each well. After 72 hours, coculture supernatants were collected for CCL20 assessment (DouSet-DY360; R&D Systems) and cells processed for quantitative real-time PCR (qPCR) analyses, separating macrophages from melanoma cells with CD14 microbeads. Oligonucleotides (Supplementary Table S1) were designed according to the Roche software for qPCR. Total RNA was extracted (RNeasy-kit; Qiagen) and retrotranscribed cDNA quantified using the Universal Human Probe Roche library (Roche-Diagnostics). Assays were made in triplicate and normalized to TBP and/or GAPDH expression (ΔΔCT method). Western blot and flow cytometry analyses were performed as previously described (14, 15).
Migration and proliferation assays
For chemotactic assays, cells were suspended in RPMI-1640 containing 0.1% FCS and 5 × 104 cells added on top of 8 μm pore transwell chambers (Costar). Then, cells were allowed to migrate for 4 hours at 37°C toward 0.1% FCS RPMI-1640 containing recombinant human chemokines (R&D Systems; PreproTech). Nonmigrated cells were removed from the upper part of the membranes with cotton swabs, and migrated cells attached to the bottom part were quantified. For wound-closure assays, 5 × 104 cells suspended in 1% FCS RPMI-1640 were added within wells of 1-mm insets (IBIDI) placed on 2 μg/mL collagen I (Nutragen) coated 24-well plastic plates and allowed to attach overnight. Then, insets were removed, plate wells were washed twice, and attached cells were allowed to migrate for 4 hours in 0.1% FCS RPMI-1640 ± rhCCL20 (R&D Systems). For quantification in both assays, cells were fixed with 4% formaldehyde, stained with propidium iodide (Sigma), and imaged at five random fields with a 10x objective with a fluorescence camera (sCMOS Orca-Flash-4.0/LT). Number of migrated cells and wound-closure distances were calculated using the FIJI software (National Institute Health, US). For proliferation assays, 105 cells were suspended in 1% BSA RPMI-1640 ± chemokines (R&D Systems), seeded on collagen-coated coverslips, and incubated for 24 hours at 37°C, adding bromodeuxyuridine (BrdUrd; Roche) 4 hours before fixation. Cells were then stained for BrdUrd and percentages of replicated nuclei estimated. This assay was used to study the effect of CCR6 downregulation in BLM cells using magnetofection (Biosciences) with siRNA (CKR-6/sc-35064; Santa Cruz Biotechnology) or blocked with anti-hCCR6 (clone 53103).
Melanoma xenograft models
NSG mice (The Jackson Laboratory) were maintained under specific pathogen-free conditions. At age 4 to 6 weeks, male mice were subcutaneously inoculated with 106 melanoma cells suspended in PBS ± 200 ng/mL rhCCL20 (R&D Systems; PreproTech). Mice were then subcutaneously injected with PBS ± 500 ng rhCCL20 every 3 days until tumors reached approximately 1 cm width, when they were resected, weighed, and frozen for further histologic analyses. For in vivo CCR6 blockade assays, mice were inoculated with 106 tumor cells (TCs) in PBS containing 10 μg/mL blocking anti-hCCR6 or IgG2b mouse antibodies, ±200 ng/mL rhCCL20 (added 30 minutes afterwards). Mice were then intravenously injected with 100 μg antibodies, ±500 ng rhCCL20 (supplied subcutaneously 2 hours later) every 3 days. In BLM tumor-bearing mice, spontaneous metastases were examined 10 days after resection of primary tumors, when lungs were extracted and fixed with Bouins fluid to quantify superficial metastases. These procedures were approved by the IiSGM animal care/use and Comunidad de Madrid committees (PROEX-14214).
Fluorescence confocal microscopy
Melanoma tissues were obtained from patients with primary and metastatic lesions who were undergoing surgical treatment, following the medical-ethics committee procedures of HGUGM. Fifty-seven cryopreserved melanoma samples were screened: 41 primary tumors and 16 metastases. Cryosections (5 μm) were fixed with cold acetone for 15 minutes, stained with different antibodies, and analyzed by confocal microscopy. For survival analyses, primary melanomas were triple-stained with CCR6 (clone 53103.111), CCL20 (ab9829), and Hmb45 antibodies. Then, several 20x fields (3–5 per sample) were analyzed at regions of interest (10–15 per field) identified along tumor (Hmb45+) and stromal zones, to measure tumor-CCR6 and stromal-CCL20 mean fluorescence intensities (MFIs), respectively. For in vivo TAM phenotyping, CD163-, CD163/CD11c-, CD11c/CD209-, or CD115/CD15-stained cells were segmented, and MFI of distinct markers was quantified at matched single cells. Similarly, for in vivo proliferation quantification, DAPI-stained nuclei were segmented and Ki67 MFI assessed. All quantifications were performed in a blinded way. Imaging was performed using the glycerol ACS APO 20x NA0.60 and oil ACS APO 40x NA1.30 immersion objectives of a confocal fluorescence microscope (SPE, Leica-Microsystems), and FIJI software was used for image quantification in all cases.
Prognostic significance analysis
To evaluate the prognostic significance of tumor-CCR6 and stromal-CCL20 expression in melanoma progression and survival, we used a cohort of 40 frozen human skin primary melanomas, obtained between the years 1999 and 2013, with known clinicopathologic features, including age, gender, tumor thickness (Breslow), ulceration, histologic subtype, location, American Joint Committee on Cancer stage, metastasis occurrence, and survival within the next 5 years to primary tumor resection. MFI values corresponding to two variables, stromal-CCL20 and tumor-CCR6, were classified into high or low levels using the median as cutoff points. Censured Kaplan–Meier curves were used to analyze the correlation with patient disease-free and overall survival; and the Cox-regression method (univariate and multivariate) was used to identify independent prognostic variables. The Mann–Whitney t test was used to evaluate the association with clinicopathologic features.
Statistical analyses
The Student t test, Mann–Whitney, Wilcoxon paired test, Spearman correlation, Cox-regression, and Log-rank analyses have been used in this study, as indicated. P < 0.05 was considered statistically significant.
Results
CCR6/CCL20 functional expression by melanoma cells
CCR6 expression by patient-derived melanoma cells and cell lines was investigated. Melanoma tumors are complex tissues infiltrated by leukocytes which may express CCR6. Therefore, for PCR and other analyses, leukocytes were removed from patient samples with the use of anti–CD45-coated magnetic beads. All cell lines studied and patient-derived TCs expressed CCR6 mRNA, compared with other cells known to express or not CCR6 (Fig. 1A). CCR6 protein expression was confirmed by Western blot, flow cytometry, and immunofluorescence. A specific 50-kDa band was detected in all melanocytic cells studied, with a variable but measurable CCR6 surface staining and a larger intracellular pool detected by flow cytometry of intact or permeabilized cells, respectively (Fig. 1B and C; Supplementary Fig. S1A–S1D).
Melanoma cells express CCR6 and respond to exogenous hCCL20. A, CCR6 mRNA expression (relative to TBP) in melanoma cells compared with other cell types (Hek-293 and NCI-H929 as negative controls; Caco-2 and iDC as positive controls). B and C, CCR6 protein expression in human melanocytes and melanoma cell lines as assessed by Western blot (B); flow cytometry (isotype, shaded; surface, black-line; total/permeabilized, gray line) and permeabilized cells immunofluorescence (C). D and E, Migratory response to exogenous chemokines (ng/mL), as assessed by wound-closure (D) and transwell (E) assays. F, Proliferative response to exogenous chemokines (ng/mL), as indicated. G and H, Proliferation assays achieved with BLM cells downregulated for CCR6 with siRNA (72 hours, G), or blocked with 10 μg/mL anti-hCCR6 antibody (H). CCR6 silencing was confirmed at both mRNA (relative to TBP/GAPDH average) and protein (normalized to GAPDH) levels (G). D to H, Data are given as the mean ± SD from two to three independent experiments performed in triplicate. Significant differences respective to untreated controls are shown (*, P < 0.05; t Student). Scale bar, as indicated.
Melanoma cells express CCR6 and respond to exogenous hCCL20. A, CCR6 mRNA expression (relative to TBP) in melanoma cells compared with other cell types (Hek-293 and NCI-H929 as negative controls; Caco-2 and iDC as positive controls). B and C, CCR6 protein expression in human melanocytes and melanoma cell lines as assessed by Western blot (B); flow cytometry (isotype, shaded; surface, black-line; total/permeabilized, gray line) and permeabilized cells immunofluorescence (C). D and E, Migratory response to exogenous chemokines (ng/mL), as assessed by wound-closure (D) and transwell (E) assays. F, Proliferative response to exogenous chemokines (ng/mL), as indicated. G and H, Proliferation assays achieved with BLM cells downregulated for CCR6 with siRNA (72 hours, G), or blocked with 10 μg/mL anti-hCCR6 antibody (H). CCR6 silencing was confirmed at both mRNA (relative to TBP/GAPDH average) and protein (normalized to GAPDH) levels (G). D to H, Data are given as the mean ± SD from two to three independent experiments performed in triplicate. Significant differences respective to untreated controls are shown (*, P < 0.05; t Student). Scale bar, as indicated.
To explore CCR6/CCL20 signaling, we performed cell proliferation and migration assays (Fig. 1D–F). CCL20 induced migration of BLM cells in wound-closure chemokinetic assays and in chemotactic assays, with a bell shape dose-response curve, whereas A375 cells responded to a broad range of CCL20 concentrations. CCL20 enhanced proliferation of BLM and A375 cells similarly to CXCL12 or CXCL8, which signal through other chemokine receptors, and showed no additional effects when the three chemokines were used together (CCL20, CXCL12, and CXCL8). To confirm specificity of CCR6/CCL20 signaling, we used either siRNA to silence CCR6 expression or mAb to CCR6 to block the receptor, showing abrogation of exogenous CCL20 effects but not of CXCL12 (Fig. 1G and H). These results indicated that melanoma cells express functional CCR6, which may contribute to proliferation and migration in response to exogenous CCL20.
CCL20 enhances tumor growth and metastasis in melanoma mouse models
In order to determine the effect of exogenous CCL20 on tumor growth, we developed xenografts with BLM, A375, and Skmel147 human melanoma cells, which were injected subcutaneously into NSG mice. Mice were then treated every 3 days with hCCL20 or vehicle; reaching concentrations >200 pg of hCCL20 per gram of tumor tissue after 0.5 hours of injection (Fig. 2A; Supplementary Fig. S1E–S1G). Exogenous hCCL20 increased tumor growth by 2-fold in BLM (P = 0.001) and A375 (P = 0.011) models. Significant differences were found with Skmel147 cells (P = 0.016). Both BLM and A375 tumors showed more proliferating Ki67+ cells in hCCL20-treated mice (Fig. 2B). We also analyzed vascularization and leukocyte infiltration in order to account for indirect host-mediated effects. No differences were found in stromal components between hCCL20- and vehicle-treated mice (Fig. 2C and D). CCR6 expression was limited to malignant human cells, whereas key host cells, such as F4/80+ TAMs, lacked CCR6 expression (Supplementary Fig. S1H). A blocking mAb to human CCR6 abrogated the effect on tumor growth of exogenous hCCL20 (Fig. 2E). In the absence of hCCL20 treatment, no differences were observed between xenografts treated with anti-CCR6 or isotype control Ab, precluding any role of endogenous CCL20. Next, to analyze the impact of hCCL20 treatment in the ability of BLM cells to metastasize to distant organs, we used a model of spontaneous metastasis. Lung weight and number of superficial metastases were significantly higher in mice treated with hCCL20 than those treated with vehicle (Fig. 2F). These results indicated that hCCL20 administration enhanced both local and metastatic tumor growth in xenograft models.
Exogenous hCCL20 induces melanoma tumor progression in vivo. A, Human melanoma xenografts (BLM, A375) developed in NSG mice subcutaneously injected with hCCL20 (+) or PBS (–) every 3 days (n = 5 mice/condition). Photographs of resected tumors and their weight values are shown. B, Percentages of proliferating cells (Ki67+ nuclei) relative to total cells (DAPI+ nuclei). Right, Ki67 (green) and DAPI (red) staining of A375 tumors from mice treated (+) or untreated (–) with hCCL20. Scale bar, 50 μm. C, Percentages of immune cells infiltrating tumors relative to total cells, including leukocytes (Cd45+), TAMs (F4/80+), DC (F4/80− Cd11c+), and neutrophils (PMN, F4/80− Ly6g+). D, Intratumoral Cd31+ vessel densities. E, In vivo blockade of hCCR6 during melanoma xenograft development in mice treated (+) or untreated (–) with exogenous hCCL20 (n = 4–5 mice/condition). Tumor weights are shown. F, Lung metastases 10 days after BLM primary tumors resection. Weight values, number of superficial metastases (“100”, when uncountable), and lungs fixed with Bouins solution are shown. P values are shown (Mann–Whitney test, n.s., nonsignificant).
Exogenous hCCL20 induces melanoma tumor progression in vivo. A, Human melanoma xenografts (BLM, A375) developed in NSG mice subcutaneously injected with hCCL20 (+) or PBS (–) every 3 days (n = 5 mice/condition). Photographs of resected tumors and their weight values are shown. B, Percentages of proliferating cells (Ki67+ nuclei) relative to total cells (DAPI+ nuclei). Right, Ki67 (green) and DAPI (red) staining of A375 tumors from mice treated (+) or untreated (–) with hCCL20. Scale bar, 50 μm. C, Percentages of immune cells infiltrating tumors relative to total cells, including leukocytes (Cd45+), TAMs (F4/80+), DC (F4/80− Cd11c+), and neutrophils (PMN, F4/80− Ly6g+). D, Intratumoral Cd31+ vessel densities. E, In vivo blockade of hCCR6 during melanoma xenograft development in mice treated (+) or untreated (–) with exogenous hCCL20 (n = 4–5 mice/condition). Tumor weights are shown. F, Lung metastases 10 days after BLM primary tumors resection. Weight values, number of superficial metastases (“100”, when uncountable), and lungs fixed with Bouins solution are shown. P values are shown (Mann–Whitney test, n.s., nonsignificant).
CCR6/CCL20 expression in primary cutaneous melanoma and metastasis
We next examined whether CCR6/CCL20 expression might have translational value by analyzing correlation with clinicopathologic factors and patient survival. We screened a collection of 40 primary cutaneous melanomas with a median patient follow-up of 67 months (range, 5–187 months) and 14 metastases (Supplementary Table S2). Frozen tissue sections were simultaneously triple-labeled for the melanosome protein Hmb45, CCR6, and CCL20 (Fig. 3A). The majority of Hmb45+ cells coexpressed CCR6 in all primary and metastatic lesions. By contrast, CCL20 was expressed in the stromal compartment. Therefore, to understand the prognostic value of CCR6/CCL20 expression in primary cutaneous melanoma, we quantified the MFI of CCR6 staining along Hmb45+ tumor nests, whereas CCL20 expression was evaluated at Hmb45− stromal bundles (Fig. 3B and C). MFI data of tumor-CCR6 and stromal-CCL20 expression were compared with clinicopathologic factors: higher expression of stromal-CCL20 correlated with unfavorable prognostic factors for cutaneous melanomas such as tumor thickness >2 mm (P = 0.023) and stage II–III (P = 0.002), whereas nonsignificant correlation with progression indicators was found for tumoral-CCR6 levels (Supplementary Table S3). Next, we used the Kaplan–Meier method to analyze whether tumor-CCR6 or stromal-CCL20 expression correlated with disease-free survival or overall survival of this group of 40 patients (Fig. 3D). We classified patients based on their respective tumor-CCR6 or stromal-CCL20 expression levels as “high” or “low,” using the median MFI value as cutoff point. High stromal-CCL20 staining correlated with shorter disease-free and overall survival (log-rank test, both P < 0.001), but no correlation was found between tumor-CCR6 staining and disease-free survival (P = 0.693). Finally, to determine whether stromal-CCL20 expression was an independent prognostic factor, we performed a multivariate regression analysis including gender, age, Breslow, and stage parameters (Table 1). This analysis showed that stromal-CCL20 expression level was an independent prognostic factor of disease-free survival [HR, 1.5; 95% confidence interval (CI), 1.25–1.76; P < 0.001] and overall survival (HR, 1.7; 95% CI, 1.2–2.5; P = 0.004) in this cohort. These findings highlighted the role of CCL20 expression by the stroma in predicting clinical behavior of primary cutaneous melanoma patients.
CCR6/CCL20 expression in human melanoma and correlation with patient survival. A, Human melanoma samples stained for Hmb45 (green), CCR6 (red), and CCL20 (white). Primary tumors with different stromal-CCL20 expression (“low” and “high”) are shown. “T,” tumor nest. Scale bar, 100 μm. B and C, Quantification of tumor-CCR6 (B) and stromal-CCL20 (C) expression. MFI ± SD (in arbitrary units, a.u.) are shown. Samples (40 primary tumors and 14 metastases) follow the same order in both histograms. D, Five-year Kaplan–Meier curves for tumor-CCR6 (disease-free survival) and stromal-CCL20 (disease-free and overall survival). The median values of the 40 primary melanomas were used to classify as “low” or “high” expressing samples (79 a.u. for tumor-CCR6; 72 a.u. for stromal-CCL20). P values are shown (log-rank).
CCR6/CCL20 expression in human melanoma and correlation with patient survival. A, Human melanoma samples stained for Hmb45 (green), CCR6 (red), and CCL20 (white). Primary tumors with different stromal-CCL20 expression (“low” and “high”) are shown. “T,” tumor nest. Scale bar, 100 μm. B and C, Quantification of tumor-CCR6 (B) and stromal-CCL20 (C) expression. MFI ± SD (in arbitrary units, a.u.) are shown. Samples (40 primary tumors and 14 metastases) follow the same order in both histograms. D, Five-year Kaplan–Meier curves for tumor-CCR6 (disease-free survival) and stromal-CCL20 (disease-free and overall survival). The median values of the 40 primary melanomas were used to classify as “low” or “high” expressing samples (79 a.u. for tumor-CCR6; 72 a.u. for stromal-CCL20). P values are shown (log-rank).
Univariate and multivariate Cox-regression analysis for disease-free survival
. | Univariate . | Multivariate . | |
---|---|---|---|
Disease-free survival (Cox regression) . | P . | HR (95% CI) . | P . |
Gender (F vs. M) | 0.070 | 5.307 (1.26–22.2) | 0.023 |
Age (years) | 0.919 | 1.022 (0.99–1.05) | 0.123 |
Location (H/L vs. T) | 0.547 | ||
Subtype (nod vs. others) | 0.246 | ||
Ulceration (yes vs. no) | 0.267 | ||
Breslow (mm) | <0.001 | 1.896 (1.33–2.69) | <0.001 |
Stage (I vs. II–III) | <0.001 | 1.433 (0.69–2.95) | 0.330 |
T-CCR6 (MFI, each 10 a.u.) | 0.490 | ||
S-CCL20 (MFI, each 10 a.u.) | <0.001 | 1.487 (1.25–1.76) | <0.001 |
. | Univariate . | Multivariate . | |
---|---|---|---|
Disease-free survival (Cox regression) . | P . | HR (95% CI) . | P . |
Gender (F vs. M) | 0.070 | 5.307 (1.26–22.2) | 0.023 |
Age (years) | 0.919 | 1.022 (0.99–1.05) | 0.123 |
Location (H/L vs. T) | 0.547 | ||
Subtype (nod vs. others) | 0.246 | ||
Ulceration (yes vs. no) | 0.267 | ||
Breslow (mm) | <0.001 | 1.896 (1.33–2.69) | <0.001 |
Stage (I vs. II–III) | <0.001 | 1.433 (0.69–2.95) | 0.330 |
T-CCR6 (MFI, each 10 a.u.) | 0.490 | ||
S-CCL20 (MFI, each 10 a.u.) | <0.001 | 1.487 (1.25–1.76) | <0.001 |
TAMs are the major stromal source of CCL20 in human melanoma tissues
To identify the CCL20 source in melanoma tumors, we isolated malignant cells and infiltrating leukocytes as macrophages and T cells (ref. 13; Fig. 4A; Supplementary Fig. S2A). To isolate TAMs, we used CD14 that was expressed by macrophages and correlated with CD163 (Supplementary Fig. S2B). TAMs showed higher expression of CCL20 mRNA than CD3+ TILs, malignant cells, cancer-associated fibroblasts (CAFs), or peripheral blood monocytes from the same patients. CCL20 protein secretion measured in the culture supernatant of TAMs, CAFs, and melanoma cells showed that TAMs were actively secreting large amounts of this chemokine (Fig. 4B). As expected, tissue supernatant from a mechanically disaggregated advanced primary melanoma also contained CCL20 (Fig. 4C). Therefore, we focused our interest in TAMs as the predominant source of CCL20 in melanoma tissues. However, neither proinflammatory (GM-CSF) M1 nor ant-inflammatory (M-CSF) M2 in vitro–differentiated macrophages secreted significant amounts of the chemokine (Fig. 4C), suggesting that the CCL20 high (CCL20hi) phenotype might be a distinctive TAM characteristic. Tissue supernatant from primary melanoma induced CCL20 secretion and mRNA upregulation by both M1 and M2 macrophages (Fig. 4C and D). Moreover, high CCL20 concentrations were secreted during coculture of M1 (∼100 fold) or M2 (∼50 fold) macrophages with BLM and A375 cells, but not with Skmel103 cells, compared with each corresponding cell type cultured alone (Fig. 4C). M1 macrophages separated from melanoma cells after coculture showed higher CCL20 mRNA expression than M2 macrophages or melanoma cells (Fig. 4D). These results indicated that primary tumor-derived factors or interactions with melanoma cells induced the CCL20hi secretory phenotype in both M1 and M2 macrophages.
TAMs are the major source of CCL20 in metastatic primary melanomas. A, Relative expression of CCL20 mRNA in cultured melanoma cell lines (n = 6, as in Fig. 1A) and CAF (n = 3 patients); and in freshly isolated TCs (n = 3), TILs CD3+ (n = 3), TAMs CD14+ (n = 5), and peripheral blood monocytes CD14+ from melanoma patients (PBMo, n = 3). B, CCL20 detected in supernatants of CAF (n = 3 patients), melanoma cell lines (n = 6, as in Fig. 1A), and isolated TAMs CD14+ (n = 3) after 24-hour incubation. C, Left, CCL20 secreted by M1 or M2 macrophages conditioned with an advanced primary melanoma tissue-supernatant (snt) for 48 hours (n = 3). Right, CCL20 detected in culture-supernatants of in vitro–differentiated macrophages (M1 or M2; n = 3) and TCs (Skmel103, A375, or BLM), alone or cocultured for 72 hours, as indicated. D, CCL20 mRNA expression in mutually conditioned, separated macrophages, and BLM cells (n = 3), or in tissue-supernatant conditioned macrophages. A–D, Data are given as the mean ± SD. E, Representative metastatic and nonmetastatic primary tumors stained for CD163 (red), CCL20 (green), and DAPI (blue). F, Dot plot showing CCL20 MFI values of single CD163+ macrophages from control skin, nevus, primary tumors (nonmetastatic and metastatic), and metastasis tissues. Mean ± standard error, and the percentage of TAMs expressing CCL20high (arbitrarily assigned to MFI > 75 a.u.), are shown for each sample. G, CCL20 MFI quantified in single-cell myeloid subpopulations of DC (red dots, CD11c+ CD163−) and TAMs (blue, CD11chi CD163+; black, CD11clow CD163+) from metastatic and nonmetastatic primary tumors (n = 3/group). H, DC (CD1c, CD141) and macrophage (CD115, CD14, CD163) markers MFI quantified in DC and TAM subsets (n = 5), as in G. I, CCL20 MFI quantified in single-cell myeloid subpopulations of PMN (green, CD15+ CD115−) and TAMs (black, CD15− CD115+) from 5 metastatic primary tumors. Illustrative images are shown, as indicated. Scale bars, 50 μm.
TAMs are the major source of CCL20 in metastatic primary melanomas. A, Relative expression of CCL20 mRNA in cultured melanoma cell lines (n = 6, as in Fig. 1A) and CAF (n = 3 patients); and in freshly isolated TCs (n = 3), TILs CD3+ (n = 3), TAMs CD14+ (n = 5), and peripheral blood monocytes CD14+ from melanoma patients (PBMo, n = 3). B, CCL20 detected in supernatants of CAF (n = 3 patients), melanoma cell lines (n = 6, as in Fig. 1A), and isolated TAMs CD14+ (n = 3) after 24-hour incubation. C, Left, CCL20 secreted by M1 or M2 macrophages conditioned with an advanced primary melanoma tissue-supernatant (snt) for 48 hours (n = 3). Right, CCL20 detected in culture-supernatants of in vitro–differentiated macrophages (M1 or M2; n = 3) and TCs (Skmel103, A375, or BLM), alone or cocultured for 72 hours, as indicated. D, CCL20 mRNA expression in mutually conditioned, separated macrophages, and BLM cells (n = 3), or in tissue-supernatant conditioned macrophages. A–D, Data are given as the mean ± SD. E, Representative metastatic and nonmetastatic primary tumors stained for CD163 (red), CCL20 (green), and DAPI (blue). F, Dot plot showing CCL20 MFI values of single CD163+ macrophages from control skin, nevus, primary tumors (nonmetastatic and metastatic), and metastasis tissues. Mean ± standard error, and the percentage of TAMs expressing CCL20high (arbitrarily assigned to MFI > 75 a.u.), are shown for each sample. G, CCL20 MFI quantified in single-cell myeloid subpopulations of DC (red dots, CD11c+ CD163−) and TAMs (blue, CD11chi CD163+; black, CD11clow CD163+) from metastatic and nonmetastatic primary tumors (n = 3/group). H, DC (CD1c, CD141) and macrophage (CD115, CD14, CD163) markers MFI quantified in DC and TAM subsets (n = 5), as in G. I, CCL20 MFI quantified in single-cell myeloid subpopulations of PMN (green, CD15+ CD115−) and TAMs (black, CD15− CD115+) from 5 metastatic primary tumors. Illustrative images are shown, as indicated. Scale bars, 50 μm.
Next, we analyzed macrophage CCL20 expression in tissues by multicolor confocal microscopy, including control skin, nevi, primary melanomas (grouped as nonmetastatic and metastatic primary tumors), and distant melanoma metastases (Fig. 4E and F). To quantify CCL20 at the single-cell level, we used CD163 to gate tissue macrophages because it is expressed by most macrophages (16). A subpopulation of TAMs (∼50%) showing CCL20hi (stated as MFI >75 a.u.) were detected in primary cutaneous melanomas from patients that developed metastasis during follow-up (referred from here as “metastatic primary melanomas”), whereas most TAMs from nonmetastatic primary tumors showed CCL20low (<75 a.u.), similar to skin and nevus macrophages. In secondary metastatic melanoma, the CCL20hi TAM subpopulation was observed in some samples but not in others. CCL20 MFI amounts correlated in single cells and stromal regions from primary tumors (R = 0.727), whereas no significant correlation was obtained between TAM density and stromal-CCL20 (Supplementary Fig. S2C and S2D). A minority of CCL20hi cells were CD163− and CD11c+ (Fig. 4G). We used DC markers (CD141 for DC1 subset and CD1c for DC2 subset) to characterize these cells as CD11c+CD1c+ classical DC2 (Fig. 4G and H; Supplementary Fig. S2E). In addition to being expressed by DC, CD11c was mostly expressed by a subset of CD14+ and CD115+ macrophages. These TAMs displayed lower CD163 expression and fell into two subsets: CD11chiCD163low and CD11clowCD163hi (Fig. 4H). CCL20hi expression was detected in all these myeloid subsets in primary tumors from patients that developed metastasis but not in tumors from nonmetastatic patients (Fig. 4G). In contrast, CCL20 was only barely detectable in polymorphonuclear cells (PMNs), identified as CD15+CD115− minor subset (Fig. 4I).
Protumoral TAMs secrete CCL20, regardless of polarization M1/M2 state
Because CD11c/CD163 expression identified two TAM subsets and CD11c expression in macrophages may indicate skewing toward M1 polarization (17–19), we analyzed in more detail M1/M2 polarization markers (17, 20, 21). Single-cell quantitative analysis of melanoma tissues showed a positive correlation between CD11c and pSTAT1 M1 markers and between CD163L1 and CD209 M2 markers; whereas M1 (CD11c and pSTAT1) and M2 (CD163L1 and CD209) markers correlated negatively between them (Supplementary Fig. S3A). The best negative correlation was obtained between CD11c and CD209 (R = –0.75; P < 0.001) allowing identification of the following subsets: CD11chiCD209low/– M1-like TAMs; CD11clow/–CD209hi M2-like TAMs; and a mixed M1/M2 population expressing both markers equally (Fig. 5A and B). Diverse TAM subsets displaying M1, M2, or M1/M2 mixed profiles were found in ten primary melanomas, with no significant differences in subset distribution with regard to subsequent nonmetastatic or metastatic patient evolution (Supplementary Fig. S3B). CCL20hi expression was detected independently of M1/M2 polarization state in most TAMs from the metastatic group (Fig. 5B), compared with low CCL20 expression by most TAMs from nonmetastatic melanomas (Figs. 4F and 5A). Altogether, these results suggested that CCL20 expression by TAMs in primary melanoma tumors may be the result of a metastatic activation state. TAM CCL20 expression might predict unfavorable prognosis of newly diagnosed patients better than the M1/M2 polarization phenotype.
CCL20 is expressed by TAMs displaying a protumoral profile, regardless of their polarization state. A and B, Nonmetastatic (A) and metastatic (B) primary tumors stained for CD11c (green), CD209 (red), CCL20 (white), and DAPI (blue). Dot plots of TAM subpopulations categorized as CD11chi CD209low (M1-like), CD11clow CD209hi (M2-like), and mixed phenotype (M1/M2), using 75 a.u. as arbitrary cutoff MFI point (n = 5/group; R, Spearman's coefficients). B, TAM CCL20 MFI values are shown for metastatic primary tissues (Wilcoxon paired test). C, Representative nonmetastatic and metastatic primary tumors stained for TNF (green), VEGF-A (green), CD163 (red), and DAPI (blue). D, Summary boxplot showing average MFI values of CD11c, CD209, CCL20, TNF, and VEGF-A expression in CD163+ TAMs, and VEGF-A in tumor nests, from metastatic and nonmetastatic primary melanomas (Mann–Whitney). Scale bars, 50 μm.
CCL20 is expressed by TAMs displaying a protumoral profile, regardless of their polarization state. A and B, Nonmetastatic (A) and metastatic (B) primary tumors stained for CD11c (green), CD209 (red), CCL20 (white), and DAPI (blue). Dot plots of TAM subpopulations categorized as CD11chi CD209low (M1-like), CD11clow CD209hi (M2-like), and mixed phenotype (M1/M2), using 75 a.u. as arbitrary cutoff MFI point (n = 5/group; R, Spearman's coefficients). B, TAM CCL20 MFI values are shown for metastatic primary tissues (Wilcoxon paired test). C, Representative nonmetastatic and metastatic primary tumors stained for TNF (green), VEGF-A (green), CD163 (red), and DAPI (blue). D, Summary boxplot showing average MFI values of CD11c, CD209, CCL20, TNF, and VEGF-A expression in CD163+ TAMs, and VEGF-A in tumor nests, from metastatic and nonmetastatic primary melanomas (Mann–Whitney). Scale bars, 50 μm.
Next, we analyzed the expression of known protumoral TAM markers, such as TNF and VEGF-A (22–24), to ask whether TAMs from primary lesions display a differential cytokine profile. As with CCL20, TAMs from metastatic primary melanomas and nonmetastatic tumors expressed different amounts of both TNF and VEGF-A (Fig. 5C and D; Supplementary Fig. S4A). TNF induced CCL20 secretion by macrophages (M2 > M1) suggesting an activating mechanism, but VEGF-A did not (Supplementary Fig. S4B). In addition to TAM, VEGF-A was also detected at tumor nests in certain samples, but not restricted to metastatic primary melanomas (Fig. 5D; Supplementary Fig. S4C). TAM expression of CCL20 correlated with TAM expression of TNF and VEGF-A in patient samples (R = 0.749 and R = 0.612, respectively; Supplementary Fig. S4D). Thus, the activated state of macrophages infiltrating primary melanomas may serve as a biomarker to identify patients with high risk of metastatic spreading and poor prognosis for survival.
Discussion
CCR6/CCL20 interactions promote tumor progression in several human cancers (9, 12). Here, we showed functional CCR6 expression by human melanoma cells, whereas CCL20 induced proliferation/migration in vitro and enhanced tumor growth/metastasis in vivo. We demonstrated an association in cutaneous melanoma patients between high stromal-CCL20 expression in the primary tumor and the subsequent development of metastasis during follow-up and poorer survival. Stromal-CCL20hi cells were identified as TAMs, which coexpressed TNF and VEGF-A in a signature of prometastatic TAM activation.
Tumors are complex tissues that contain both malignant and nonmalignant cells, which interact to influence the progression of the tumor (25). Analysis of the molecular phenotype of each cell type requires enough tissue to separate cells, but the small specimens available from primary skin melanomas preclude such analysis. Multicolor confocal microscopy coupled to single-cell quantitative image analysis provides tools to assess protein expression by specific cell types. We quantified CCR6/CCL20 expression in primary melanomas, showing that CCR6 was expressed in tumor areas and CCL20 in stromal areas, particularly by TAMs. We also purified specific cell types to confirm that CCR6 was expressed by fresh malignant cells, whereas CCL20 was secreted by TAMs. Our results blocking human CCR6 with mAb with or without addition of exogenous CCL20 proved paracrine instead of autocrine signaling in melanoma. We conclude that there is a synergistic interaction between malignant cells and other cells in the stroma through a CCR6/CCL20 paracrine signaling loop. By contrast, in most epithelial cancers such as colorectal, lung, and pancreatic adenocarcinomas or nasopharyngeal tumors, both CCR6 and CCL20 are expressed by malignant epithelial cells, suggesting autocrine selfstimulation (11, 26–28). Paracrine signaling between stromal-CCL20 and tumoral-CCR6 may be a characteristic of certain cancers, including melanoma.
Stromal-CCL20 expression was variable across patient samples. Long-term follow-up data showed a positive association between high stromal-CCL20 levels and incidence of subsequent metastatic spreading or melanoma-specific death. In our exploratory cohort of primary melanoma patients, stromal-CCL20 expression was associated with disease-free survival by multivariate analysis, together with gender and Breslow thickness, indicating that stromal-CCL20 expression could identify patients with higher risk of metastasis. Data from independent patient series will be necessary to confirm that stromal-CCL20 is a prognostic factor for cutaneous melanoma patients. By contrast, no association was found in our study between tumoral-CCR6 expression and metastatic progression.
CCR6 was originally discovered in dendritic cells (29). A proposed immunotherapy strategy is to attract DCs to cancer sites by engineering tumors to produce CCL20 or by intratumoral injection of CCL20 (30, 31). However, CCR6 expression by malignant cells in a wide variety of tumors, including our work in melanoma, cautions against choosing this chemokine for intratumoral immunotherapy. CCL20 is the main chemokine expressed by inflamed skin (32) and mucosal epithelium (33), and it may orchestrate a protumoral inflammatory microenvironment that could promote rather than hinder tumor progression. Indeed, CCL20 from different sources recruits Treg, Th17, and Th22 cells that maintain the sort of chronic inflammation and immunosuppression that could indirectly promote cancer progression (34–36).
Studies in cancer mouse models have revealed that TAMs are heterogeneous cells that adopt a spectrum of phenotypes between the extremes M1 and M2 polarization states, including mixed M1/M2 phenotypes (37, 38). Studies identifying distinct macrophage subpopulations in human tissues are scarce (39). In one study, CD11c expression was used to identify two subpopulations of human CD14+ macrophages isolated from decidua: a CD11chi subset with an inflammatory phenotype, and a CD11clo subset related with tissue remodeling and developmental functions (18). We used multiparametric single-cell microscopy to discriminate CD11c+ DC from macrophages that coexpress other markers as CD14, CD115, and CD163. This strategy identified a minor CCL20hi DC2 (CD1c+) subset as well as two subsets, CD11chi and CD11clo, of macrophages in human primary melanoma tissues. We performed a correlation analysis of the coexpression of several M1/M2 markers at the single-cell level in melanoma tissues. CD11c/CD209 showed the best negative correlation, indicating that these two markers were expressed by distinct TAM subsets. Thus, we quantified M1-like, M2-like, and mixed M1/M2 subsets in human primary melanoma. We found no statistically significant differences between nonmetastatic or metastatic patients, although larger patient series may be necessary to unravel this issue. Our results reveal that CCL20hi expression, irrespective of M1/M2 markers, is a prometastatic TAM phenotype that is associated with poor prognosis of patients with primary cutaneous melanoma. CCL20hi expression may be a unique characteristic of TAMs, as neither M1 nor M2 in vitro–differentiated human macrophages produced substantial amounts of CCL20. We found that both tumor “conditioned” M1 and M2 macrophages upregulated CCL20 expression in vitro, which is consistent with the detection of CCL20hi expression associated with both M1 and M2 TAMs in vivo.
We found that macrophages from metastatic skin melanomas coexpress CCL20, TNF, and VEGF-A, which together may define the inflammatory protumoral profile of TAMs. Consistent with our results, the tumor-promoting transcriptional profile of blood monocytes from renal cell carcinoma patients showed upregulation of, among other immune-related and protumoral genes, CCL20, TNF, and VEGF-A, as well as both upregulation and downregulation of M1 and M2 gene expression (40). Expression of these genes is activated by NF-κB (41–43), which is a central regulator of macrophage function in tumors and a hallmark of tumor-promoting chronic inflammation (44). Moreover, macrophage-derived TNF has been identified as a molecular inflammatory mechanism supporting survival of melanoma cells through upregulation of the melanoma survival factor MITF (22). Together, our data suggest that inflammatory macrophages expressing CCL20, TNF, and VEGF-A promote human melanoma progression, particularly in early metastasis. The protumoral signature of macrophages identified in this work may serve to identify patients with high risk of metastasis from primary melanoma and may guide early treatments.
Disclosure of Potential Conflicts of Interest
I. Márquez-Rodas is a consultant/advisory board member for Amgen, Bioncotech, Bristol-Myers Squibb, Merck Sharp & Dohme, Novartis, Pierre Fabre, and Roche. No potential conflicts of interest were disclosed by the other authors.
Authors' Contributions
Conception and design: R. Samaniego, A. Gutiérrez-González, P. Sánchez-Mateos
Development of methodology: R. Samaniego, A. Gutiérrez-González, A. Gutiérrez-Seijo, S. Sánchez-Gregorio, J. García-Giménez, E. Mercader, M. Relloso
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): E. Mercader, I. Márquez-Rodas, J.A. Avilés, P. Sánchez-Mateos
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): R. Samaniego, A. Gutiérrez-González, A. Gutiérrez-Seijo, I. Márquez-Rodas, M. Relloso, P. Sánchez-Mateos
Writing, review, and/or revision of the manuscript: R. Samaniego, E. Mercader, I. Márquez-Rodas, P. Sánchez-Mateos
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R. Samaniego, A. Gutiérrez-Seijo, P. Sánchez-Mateos
Study supervision: R. Samaniego, P. Sánchez-Mateos
Acknowledgments
This work was partially supported by the Ministry of Economy and Competitiveness ISCIII-FIS grants PI13/01454 (P. Sánchez-Mateos), PI17/01324 (P. Sánchez-Mateos and R. Samaniego), and PI16/00050 (M. Relloso), cofinanced by ERDF/FEDER Funds from the European Commission, “A way of making Europe.” A. Gutiérrez-Seijo is financed by the Comunidad de Madrid YEI-program. The authors would like to thank Julia Villarejo and Raul Campos-Fernandez for expert technical assistance, Jose María Bellon for help with statistical analyses, and Amaya Puig-Kröger, Francisco Sanchez-Madrid, and David Sancho for providing reagents/antibodies.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.