High endothelial venules (HEVs) in human melanoma lesions

Abstract

The presence of tumor-infiltrating lymphocytes (TILs) is a strong prognostic parameter for local dissemination and overall survival in melanoma. Lymphocyte migration from blood into peripheral tissues is mainly regulated by vascular endothelium. However, the blood vessels and mechanisms governing the recruitment of TILs in melanoma tumors remain poorly understood. Here, we show that high endothelial venules (HEVs), specialized blood vessels for lymphocyte extravasation into lymphoid tissues, are frequently found in melanoma tumors and are associated with high levels of lymphocyte infiltration. The analysis of 225 primary melanomas revealed that lymphocytes specifically infiltrated HEV-rich areas of melanoma tumors and that the density of MECA-79+ HEVs was variable among patients and strongly correlated with CD3+, CD8+ and CD20+ TIL densities. Inflammatory (CCL5, CXCL9, CXCL10 and CXCL11) and lymphoid (CCL21, CCL19 and CXCL13) chemokines as well as TH1 and naïve T-cell genes were overexpressed in melanoma samples with high densities of tumor HEVs. Mature dendritic cells (mDCs) were frequently found around tumor HEVs and densities of HEVs and DC-LAMP+ mDCs within tumor stroma were strongly correlated. DCs which maintain HEVs in lymph nodes, may thus also contribute to the regulation of HEVs in melanomas. Finally, we found significantly higher densities of tumor HEVs in melanomas with tumor regression, low Clark level of invasion and thin Breslow thickness (all p < 0.001). The strong association between tumor HEVs, TILs, mDCs and clinical parameters of melanoma, supports a critical role for HEVs in limiting malignant melanoma development through both naïve and effector T-lymphocyte recruitment and activation.

Introduction

The incidence of metastatic melanoma has increased over the past three decades and the death rate continues to rise faster than the rate of most other cancers. The World Health Organization estimates that there are 66,000 deaths from skin cancer annually worldwide, with approximately 80% due to melanoma. Although primary cutaneous melanoma is curable in its early stages with surgical resection, patients with metastatic melanoma have a median survival of less than 1 y.¹

Immunological factors have been shown to constitute a major predictor of good clinical outcome in solid cancers, including breast, colorectal and ovarian carcinomas.^2-6 The presence of tumor-infiltrating lymphocytes (TILs) in primary cutaneous melanomas is also a strong prognostic parameter for lymph node dissemination⁷ and overall survival.^8,9 Further implication of the immune system in melanoma come from a recent phase III study showing that monoclonal antibody Ipilimumab, targeting T-cell inhibitory receptor CTLA-4, improved overall survival in patients with metastatic melanoma.¹⁰ Even if immune-based therapies have led to progress in the treatment of metastatic melanoma,^11-13 tumor infiltration by activated lymphocytes is a current frontier to successful cancer therapy.¹⁴

Lymphocyte migration from blood circulation into peripheral tissues is regulated by vascular endothelium.¹⁵ However, blood vessels and mechanisms governing the recruitment of lymphocytes into melanoma remain poorly understood. Evidence has recently been provided that cancer development can alter blood vessel function and restrict lymphocyte infiltration into tumors.^16,17 In contrast, the characteristics of tumor blood vessels that facilitate large scale influx of lymphocytes within human melanoma lesions are presently unknown.

High endothelial venules (HEVs) are specialized blood vessels mediating lymphocyte extravasation in lymphoid organs and chronically inflamed tissues.^15,18-20 HEV endothelial cells express high levels of 6-sulfo sialyl Lewis X ligands, recognized by the HEV-specific antibody MECA-79, which mediate the initial capture and rolling of lymphocytes along the HEV vessel wall. Recently, we reported the frequent presence of HEVs in human solid tumors.²¹ In a retrospective cohort of 146 invasive breast cancer patients, we found that high densities of tumor HEVs were associated with T- and B-lymphocyte infiltration and significantly correlated with longer metastasis-free, disease-free and overall survival rates. However, the impact of tumor HEVs on lymphocyte infiltration and prognosis in melanoma remained to be determined.

In this study, we demonstrate that the density of tumor HEVs is variable among the 225 primary melanomas analyzed and inversely correlated with Breslow thickness and Clark level of invasion. We show that CD3⁺ and CD8⁺ T lymphocytes specifically infiltrate HEV-rich areas of melanomas and that lymphocyte infiltration in melanomas is correlated with the density of tumor HEVs. Together, these results suggest that tumor HEVs represent major gateways for lymphocyte infiltration into human melanoma lesions.

Results

Human melanomas contain variable numbers of HEV blood vessels

In order to analyze the presence of HEVs in melanoma, we performed immunohistochemistry with the HEV-specific antibody MECA-79 on tumor sections from a cohort of 225 patients operated for a primary melanoma between 2000 and 2012. Absolute numbers of MECA-79⁺ vessels present within the tumor area were quantified and the density of tumor HEVs (HEV/mm²) was calculated for each patient (Fig. 1A). MECA-79⁺ blood vessels were detected in almost two-thirds of the tumors analyzed (n = 152/225). In contrast, MECA-79⁺ vessels were never detected in normal skin samples distant from the tumor site (Fig. S1). The number of MECA-79⁺ HEVs detected within the tumor stroma was highly heterogeneous among the different melanomas ranging from 0 to 201 HEVs per tumor section with a median and a mean value of 6.5 and 24.7 HEVs per tumor section respectively (Fig. 1B). The density of HEVs within the tumor stroma was also highly variable among patients ranging from 0 to 27.2 HEVs/mm² with a median and a mean value of 0.81 and 3.1 HEVs/mm² respectively (Fig. 1B).

Figure 1. Human melanomas contain variable numbers of HEV blood vessels. (A) Representative picture from a digitized tumor slide stained with MECA-79 antibody showing numerous HEV blood vessels located at the invasive front of a melanoma lesion (dashed area). (B) Histograms showing the absolute number and the density of MECA-79⁺ HEVs in the tumor area quantified for 225 primary melanomas. Two representative pictures of melanomas with high and low numbers of MECA-79⁺ HEVs are shown. (C) Phenotypic characterization of tumor HEVs. Tumor HEVs express pan-endothelial cell markers CD31 and vWB, post-capillary venule-specific marker DARC and HEV-specific markers MECA-79, HECA-452, G72 and G152. Immunofluorescence staining of melanoma tumor sections was performed with the indicated antibodies. Counterstaining was performed with DAPI.

To further characterize the phenotype of tumor HEVs found in the stroma of human melanomas, we performed immunofluorescence staining against several blood vessel markers (CD31, vWB, DARC) and HEV markers (HECA452, G72, G152). These analyses revealed that the endothelial cells of melanoma tumor HEVs display a plump morphology, express pan-vascular endothelial cells markers CD31 and vWB, post-capillary veinule marker DARC and strongly react with HECA-452, G72 and G152 antibodies that recognize 6-sulfo sialyl Lewis X ligands for lymphocytes (Fig. 1C). Therefore, MECA-79⁺ HEVs found in human melanomas have the bona fide phenotype of HEV blood vessels.

The density of tumor HEVs correlates with clinical parameters of melanoma

We analyzed the clinical characteristics and prognostic factors of primary melanoma of our cohort according to the density of tumor HEVs (Table 1). We found no significant correlations between HEV densities and the age or sex of patients. In contrast, significant differences in the density of tumor HEVs were observed between different anatomic sites and melanoma histologic types (Table 1). The highest density of tumor HEVs was found in Dubreuilh melanoma followed by superficial spreading melanoma (SSM) whereas acral-lentiginous melanoma had the lowest HEV density (Fig. 2A). The density of HEVs was also significantly higher in melanomas showing signs of tumor regression (Fig. 2B). Breslow tumor thickness is the most powerful prognostic biomarker for staging primary cutaneous melanoma together with ulceration and lymph node involvement.²² Whereas no significant correlation was observed between HEV density and the presence of tumor ulceration or lymph node invasion (Table 1), the density of tumor HEVs was inversely correlated with Breslow thickness (Spearman r = -0.26 p < 0.001). Among patients with thin lesions (Breslow ≤ 1mm) the density of HEVs was significantly higher compared with patients with intermediate (Breslow 1.01–2mm) and thick lesions (Breslow > 2 mm) (Fig. 2C and Table 1). Finally, we found significant differences between the density of HEVs and Clark levels of invasion now recommended for staging thin melanomas when mitotic rate is not available.²² Patients with low Clark levels (I-II) had a significantly higher density of HEVs than patients with intermediate (III) and high Clark levels (IV-V) (Fig. 2D and Table 1). Altogether, these results showing that a high density of HEVs is associated with thin melanomas, low levels of invasion and signs of regression suggest that HEV density represents a favorable prognostic biomarker for malignant melanoma.

Table 1. Characteristics of patients with melanomas (n = 225) according to HEV density.

	N	%	Median	HEV/mm2 (Range)	P
Age
<60 ans	97	(43.1%)	0.7	(0.0:18.9)	p=0.32
>60 ans	128	(56.9%)	0.9	(0.0:27.3)

Sex
Male	115	(51.1%)	1.6	(0.0:18.1)	p=0.09
Female	110	(48.9%)	0.4	(0.0:27.3)
Missing	0

Anatomic site
Trunk	53	(26.2%)	2.4	(0.0:27.3)	p<0.001
Members	50	(24.8%)	0.0	(0.0:14.4)
Head	59	(29.2%)	3.7	(0.0:18.7)
Extremity	40	(19.8%)	0.0	(0.0:5.5)
Missing	23

Type
Dubreuilh	47	(21.6%)	3.9	(0.0:19.4)	p<0.001
SSM	100	(45.9%)	0.8	(0.0:27.3)
Acral-lentiginous	28	(12.8%)	0.0	(0.0:4.6)
Others	43	(18.7%)	0.5	(0.0:18.1)
Missing	7

Regression
No	171	(79.2%)
Yes	45	(20.8%)	0.6	(0.0:27.3)	p<0.01
Missing	9		3.9	(0.0:19.4)

Tumor ulceration
No	190	(88.4%)	1.0	(0.0:27.3)	p=0.41
Yes	25	(11.6%)	0.8	(0.0:13.0)
Missing	10

Breslow thickness
≤1mm	120	(57.1%)	2.6	(0.0:27.3)	p<0.001
1.01-2 mm	34	(16.2%)	1.4	(0.0:10.6)
>2 mm	56	(26.7%)	0.1	(0.0:8.7)
Missing	15

Clark level					p<0.001
I-II	80	(36.9%)	2.8	(0.0:27.3)
III	60	(27.6%)	1.3	(0.0:18.7)
IV-V	77	(35.5%)	0.1	(0.0:19.4)
Missing	8

Lymph node invasion
No	53	(69.7%)	0.8	(0.0:18.1)	p=0.11
Yes	23	(30.3%)	0.1	(0.0:5.1)
Missing	0

Figure 2. The density of tumor HEVs correlates with clinical characteristics and prognostic biomarkers of primary melanoma. The clinical characteristics and prognostic biomarkers of 225 primary melanomas were analyzed according to the density of HEVs. Significant differences in the density of tumor HEVs were found between patients classified according to melanoma histologic types (A), tumor regression (B), Breslow thickness (C) and Clark level of invasion (D). The line in the center of each box represents the median value of the distribution, and the upper and lower ends of the box are the upper and lower quartiles, respectively. Comparisons between groups were performed using the Kruskal-wallis test.

The density of tumor HEVs predicts lymphocyte infiltration in human melanomas

We observed that lymphoid infiltrates were always located around HEV-rich areas of melanomas suggesting an active role of these vessels in recruitment of TILs (Fig. 3A). Immunofluorescence staining revealed that immune cells around tumor HEVs were mainly CD3⁺ T lymphocytes (Fig. 3B and C) and to a lesser extent CD20⁺ B lymphocytes (Fig. 3D). A considerable fraction of CD3⁺ T lymphocytes were CD8⁺ cytotoxic T lymphocytes (Fig. 3E).

Figure 3. Phenotypic characterization of immune populations associated with HEV blood vessels in melanoma tumors. (A) Immune infiltrates are located around MECA-79⁺ HEV (red)-rich areas of melanoma lesions (dashed area). (B) MECA-79⁺ HEVs (green) are surrounded by numerous CD3⁺ T lymphocytes (red) that infiltrate melan-A⁺ tumors (blue). (C) CD3⁺ T cells (red) are seen attached to the luminal surface of MECA-79⁺ HEV endothelial cells (green) and extravasating through the vessel wall. (D) HEVs are present in tumor areas infiltrated mainly by CD3⁺ T cells (green) and CD20⁺ B cells (red). (E) A considerable fraction of CD3+ T cells (green) surrounding MECA-79⁺ HEVs (blue) are CD8⁺ T cytotoxic T cells (red).

To better define the link between HEVs and TILs in melanoma, we performed immunohistochemistry against CD3⁺ T cells, CD8⁺ cytotoxic T cells and CD20⁺ B cells on serial tumor sections (Fig. 4A). For each patient of our melanoma cohort, we quantified CD3⁺ T cells, CD8⁺ T cells and CD20⁺ B cells using semi-quantitative optical grading as previously described⁹ (grade 1: absent-low, grade 2: moderate, grade 3: high density of TILs). We observed that the density of tumor HEVs was highly correlated with the density of CD3⁺ (Spearman r = 0.71 p < 0.0001) and CD8⁺ (Spearman r = 0.67 p < 0.0001) T lymphocytes infiltrating melanomas (Table S1). In melanoma with high densities of CD3⁺ and CD8⁺ TILs (grade 3), the density of HEVs was significantly higher than in tumors with moderate (grade 2) or low (grade 1) densities of CD3⁺ or CD8⁺ TILs (Fig. 4B). In accordance with these results, most melanomas that did not contain HEVs (HEV^-) were poorly infiltrated by CD3⁺ or CD8⁺ T cells as compared with melanomas with HEVs (HEV⁺) (Fig. S2). The density of HEVs was also correlated with the density of CD20⁺ B lymphocytes infiltrating melanomas (Spearman r = 0.56 p < 0.0001; Fig. 4B), although this population represents a smaller fraction of TILs (Fig. 3D; Fig. S2). These results, together with the immunofluorescence staining data showing that HEV blood vessels are specifically located within lymphocyte-rich tumor areas, suggest that tumor HEVs are major gateways for lymphocyte infiltration into human melanomas.

Figure 4. The density of tumor HEVs predicts lymphocyte infiltration in melanoma. (A, B) Consecutive tumor sections from 225 melanomas were analyzed by immunohistochemistry with MECA-79, anti-CD3, anti-CD8 and anti-CD20 antibodies. The number of CD3⁺, CD8⁺ and CD20⁺ TILs was quantified by semi quantitative optical grading (grade 1, 2, 3 for low, moderate and high density of positive cells). (A) Representative pictures showing HEVs and the indicated immune populations in tumors with (upper panel) and without (lower panel) tumor HEVs. (B) Graphs showing the density of HEVs in melanoma tumors according to the density of CD3⁺, CD8⁺ and CD20⁺ TILs. p < 0.001; Kruskal-wallis test. (C) Graphs showing the densities of HEVs and CD3⁺ T cells (lower graph) in melanomas (n = 225) and invasive breast tumors (n = 146). *** p < 0.001; Mann Whitney test.

We previously described a similar association between the density of tumor HEVs and TILs in breast cancer using a cohort of 146 invasive ductal carcinomas.²¹ Comparison of HEVs densities between melanomas and breast tumors revealed that the mean HEV density was more than 10-fold higher in melanomas (3.13 ± 0.31) than in breast tumors (0.24 ± 0.05) (Fig. 4C). Similar to the HEV density, the mean density of CD3⁺ T cells was 10-fold higher in melanomas (3037.0 ± 746.1) than in breast tumors (366.2 ± 45.0) (Fig. 4C).

Tumor HEVs have no influence on Foxp3⁺ regulatory T-cell infiltration in melanoma tumors

Foxp3⁺ regulatory T cells (T_regs) suppress anti-tumor immune response and their presence has been associated with poor clinical outcome in many cancers.²³ We asked whether the presence of tumor HEVs could impact the density of Foxp3⁺ T-cell infiltrating melanomas. Immunofluorescence staining against Foxp3 on several melanoma tumor sections revealed that the presence of tumor infiltrating Foxp3⁺ CD3⁺ T cells in melanomas is independent of HEV presence (Fig. 5A). T_reg/T-cell ratio is a crucial parameter defining T_reg suppression in melanoma²⁴ so we evaluated the density of Foxp3⁺ T cells within lymphoid infiltrates surrounding melanoma lesions (Fig. 5B). We did not observe any correlation between HEVs and Foxp3+ cell densities (Fig. 5C), nor significant differences in Foxp3+ cells densities between samples with a low or a high density of HEVs (Fig. 5D). These results demonstrate that although tumor HEVs were correlated with the degree of T-lymphocyte infiltration in melanoma lesions, they had no influence on the densities of T_regs in these tumors.

Figure 5. The density of tumor HEVs has no influence on the density of tumor infiltrating Foxp3⁺ T cells. (A) Representative immunofluorescence staining of melanoma tumor sections with the indicated antibodies showing Foxp3⁺ T cells (red) within T cells infiltrates (green) in melanomas containing MECA-79⁺ HEVs (blue) or not. (B) Consecutive melanoma sections from representative tumors with a high and a low density of MECA-79⁺ HEVs (n = 30) were stained with anti-Foxp3 antibodies and the density of Foxp3⁺ cells (red arrow) within melanoma lymphoid stroma was calculated. (C) The density of Foxp3⁺ cells within peri-tumor lymphoid infiltrates is not correlated with the density of tumor HEVs. (D) The density of Foxp3⁺ cells within peri-tumor lymphoid infiltrates is similar in melanomas with a low and a high density of tumor HEVs. n.s p > 0.05 Mann Whitney test.

Melanoma tumor HEVs are associated with expression of specific chemokines, chemokine receptors and T_H1 genes

Chemokines orchestrate the trafficking and the correct positioning of immune populations in the different organs of the body. Given the strong association between HEVs and TILs in melanoma, we sought to determine chemokines associated with tumor HEVs in melanoma tumors by qRT-PCR analysis on mRNA extracted from 14 paraffin embedded formalin fixed melanoma samples containing HEVs (HEV+, n = 7) or not (HEV-, n = 7). We observed that transcripts coding for chemokines implicated in effector memory T-cell migration into peripheral tissues (CCL5, CXCL9, CXCL10 and CXCL11) were significantly upregulated in melanomas with HEVs (Fig. 6A). CXCR3, the receptor for chemokines CXCL9–11 was also overexpressed in HEV⁺ melanomas. Genes encoding chemokines mediating naïve T- and B-lymphocyte recruitment into secondary lymphoid organs (CCL19, CCL21 and CXCL13) were also significantly overexpressed in melanomas with HEVs. In contrast, chemokines CCL2 and CXCL12 which drive migration of monocytes were highly expressed in melanoma tissue samples independently of HEV presence (Fig. 6A). CCR7 and L-selectin (Lsel) which allow naïve lymphocyte migration through HEVs were also strongly overexpressed in HEV⁺ melanomas (Fig. 6A). These results suggest that HEVs may support naïve T-cell infiltration into melanomas. Immunofluorescence staining against naive lymphocyte marker CD45RA confirmed the presence of some CD45RA⁺ naïve T cells around HEVs (Fig. 6B). Nevertheless, most TILs located around HEVs in melanomas had a CD45RO⁺ memory phenotype (Fig. 4B). We analyzed expression of genes associated with T-helper (T_H) cell orientation in HEV ⁺ and HEV^- melanoma samples in order to better define effector T-cell responses associated with HEV presence in melanoma. We observed that genes related to T_H1 immune orientation (TBX21, IFNG) were significantly upregulated in melanomas with HEVs whereas expression of genes encoding cytokines associated with T_H2, T_H17 and Tregs (IL4, IL17A, IL-10 and TGFB1) was not significantly modified (Fig. 6C).

Figure 6. Expression of specific chemokines, chemokine receptors, T_H1 and naïve T cell genes in melanoma tumors containing HEVs. (A) Expression of genes related to lymphocyte migration was determined by qRT-PCR in 14 melanoma samples according to the presence of tumor HEVs (7 HEV⁺ vs 7 HEV^-). Relative mRNA expression levels were adjusted to the levels of the housekeeping gene YWHAZ, and are represented as mean relative expression (+SD) for HEV⁺ (red bars) and HEV^- (blue bars) tumors. (B) Representative immunofluorescence staining of melanoma tumor sections with the indicated antibodies showing numerous CD45RO⁺ lymphocytes and a few CD45RA⁺ naïve T cells around MECA-79⁺ HEVs. (C) Expression of genes related to T Helper orientation was determined by qRT-PCR in 14 melanoma samples according to the presence of tumor HEVs (7 HEV+ vs 7 HEV-). The levels are represented as mean relative expression (+SD) for HEV+ (red bars) and HEV- (blue bars) tumors. * p < 0.05, ** p < 0.01, *** p < 0.001; Mann Whitney test.

The density of tumor HEVs is correlated with the number of DC-LAMP⁺ dendritic cell clusters

We recently demonstrated that HEV blood vessels in mouse peripheral lymph nodes are maintained by dendritic cells (DCs).²⁵ To examine a potential link between the presence of DCs and tumor HEVs in melanoma tumors, we performed immunofluorescence and immunohistochemical staining of 30 melanoma tumor sections with antibodies against Fascin and DC-LAMP, two markers of mature DCs. We observed DCs clusters in close proximity to MECA-79⁺ HEVs in lymphoid infiltrates surrounding melanoma lesions (Fig. 7A and B). The amount of DC-LAMP⁺ infiltrating cells was evaluated through optical grading (grade 1, 2, 3 for low, intermediate and high number of DC-LAMP⁺ cells) and through the quantification of DC-LAMP⁺ DCs cluster within tumor areas. We found that the density of tumor HEVs was highly correlated with both DC-LAMP grade (Spearman r = 0.74 p < 0.001) and number of DC-LAMP⁺ cell clusters (Spearman r = 0.72 p < 0.001). The density of HEVs was significantly higher in melanomas with high numbers of DC-LAMP⁺ tumor infiltrating cells (Fig. 7C). Conversely, tumor sections with a high density of HEVs contained higher numbers of DC-LAMP⁺ clusters than tumors with a low density of HEVs (Fig. 7D). Altogether, these results indicate that the presence of tumor HEVs is correlated with the number of DC-LAMP⁺ DC clusters in the melanoma lesions.

Figure 7. The presence of tumor HEVs in melanoma lesions is correlated with the number of DC-LAMP⁺ DCs clusters. (A) Immunofluorescence staining showing Fascin⁺ DCs (red) around MECA-79+ HEVs (green) within melanoma lymphoid infiltrates. (B-D) Consecutive melanoma sections from tumors with a high and a low density of MECA-79⁺ HEVs (n = 30) were stained with antibodies directed against DC-LAMP and the number of DC-LAMP⁺ cells clusters was calculated. (B) Representative picture showing MECA-79⁺ HEVs in a tumor area infiltrated by DC-LAMP⁺ DCs. (C) The density of MECA-79⁺ HEVs is significantly higher in melanomas with a high amount of DC-LAMP⁺ infiltrating DCs (grades 2–3) as compared with melanomas with a low amount of DC-LAMP⁺ infiltrating DCs (grade 1). (D) The number of DC-LAMP⁺ DCs clusters is significantly higher in melanomas with a high density of tumor HEVs. *** p < 0.001; Mann Whitney test.

Discussion

Despite the critical role played by immunity in malignant melanoma prognosis²⁶ and therapeutic response,¹⁰ the mechanisms governing lymphocyte recruitment into melanoma tumors remain poorly defined. In this study, using a cohort of 225 primary melanomas, we provide strong evidence that HEV blood vessels represent major gateways for both T- (effector and naïve) and B-lymphocyte infiltration into melanoma lesions.

T_H1 adaptive immune responses are required for effective anti-tumor immunity^4,27 and T_H1 cytokines have been associated with regressing melanoma.^28,29 Our results demonstrating that tumor HEVs are associated with expression of genes (CXCR3, IFNG and TBX21) and chemokines (CXCL9–11) related to T_H1 effector T cells in melanoma samples suggest that HEV density could represent a crucial determinant of effector T_H1 lymphocyte infiltration into melanoma lesions. At the opposite, Foxp3+ regulatory T cells (T_regs) favor immune escape³⁰ and have been associated with adverse clinical outcome in a variety of malignancies.^31,32 We evaluated the density of T_regs within lymphoid infiltrates and found no statistical differences between melanomas with or without HEVs, suggesting that tumor HEVs do not stimulate preferential accumulation of T_regs within tumor stroma. In agreement with these results, expression levels of TGF-β and IL-10, two cytokines associated with immune suppression, were not statistically different between HEV⁺ and HEV^- melanomas.

We observed the presence of CD45RA⁺ naïve T lymphocytes around HEVs within tumor stroma and HEV⁺ melanomas were associated with overexpression of chemokines (CCL19, CCL21) and receptors (CCR7, L-selectin/CD62L) driving extravasation of naïve T lymphocytes into lymphoid organs. These results suggest that tumor HEVs, which express high levels of MECA-79⁺ sulphated L-selectin ligands, support naïve T-cell recruitment into primary melanoma lesions. Interestingly, de Chaisemartin et al. recently reported a similar association between HEVs, lymphoid chemokines and naïve T-cell infiltration in lung carcinomas.³³ Tumor HEVs may play an important role in anti-tumor immunity by facilitating the priming, activation, and differentiation of naïve T cells into effectors at the primary tumor site.^34-37 Priming of naïve T cells directly at the tumor site may bring several benefits: (1) the efficiency and the specificity of T-cell priming may be improved thanks to the higher antigenic load in situ than in the draining lymph node;³⁸ (2) a broader repertoire of tumor antigens may be presented to naïve T cells;³⁹ (3) newly generated effector T cells would not need additional migration steps to reach effector sites. Even if we can’t formally prove it, our study supports the possibility that naïve T-cell priming and differentiation may take place within human primary melanomas thanks to the presence of all the critical elements for effective T-cell responses: tumor HEVs and lymphoid chemokines for T-cell migration, mature DCs clusters around HEVs for T-cell stimulation.

Recent melanoma staging recommendations from American Joint Committee on Cancer (AJCC) indicate that Breslow tumor thickness remains the most powerful prognostic factor for localized melanoma overall survival.²² Our results revealed an inverse correlation between Breslow thickness and HEV density. Indeed T1 melanomas (Breslow ≤ 1mm) contained statistically more HEVs than T2 (Breslow 1.01–2 mm) and T3–4 (Breslow > 2 mm) melanomas. An inverse correlation was also observed between Clark level of invasion and HEV density demonstrating that HEV density is decreased when melanoma invade deeper through the dermis. Altogether, these results suggest that the density of tumor HEVs could represent an important prognostic biomarker in melanoma. Interestingly, although HEV density correlated with CD3⁺, CD8⁺ and CD20⁺ TILs (Table S1), the associations between HEV densities and melanoma prognostic biomarkers (Table 1) were stronger than those for TILs (Tables S2, S3, S4). The density of tumor HEVs, which correlates with the presence of mDCs and both effector and naïve T-cell infiltration, may thus represent a more powerful immune marker than different TILs sub-populations. Future studies will have to determine whether HEV density significantly impacts melanoma patient’s survival and represents a valuable independent marker for melanoma classification and medical care.

Melanoma is considered one of the most immunogenic solid tumors and the ideal candidate against which to develop various immunotherapeutic approaches.⁴⁰ However, most strategies have obtained limited results. The existence of immune regulatory mechanisms has been suggested to account for immunotherapy failure.⁴⁰ The inability of effector immune cells to reach tumor sites also represents a major limit to effective immunotherapy.^16,41 In this study, we provide strong evidence that tumor HEVs represent major gateways for infiltration of both effector T_H1 and naïve T lymphocytes into primary melanoma lesions. These results suggest that it may be possible to improve lymphocyte infiltration into melanomas by increasing the number of HEVs within tumor stroma. Our results revealed a strong correlation between the density of tumor HEVs and the number of DC-LAMP⁺ mDCs clusters within primary melanoma lesions. This observation suggests that mDCs within melanomas may not only function as antigen presenting cells but may also participate in the induction and/or maintenance of HEVs in the tumor microenvironment, similar to the role of DCs in the maintenance of HEVs from lymphoid organs.²⁵ Further characterization of the role of DCs in the regulation of HEV blood vessels may lead to novel therapeutic strategies for melanoma and other solid tumors.

Patients and Methods

Patients

Approval of the study was obtained from the scientific review board of the Institute Claudius Regaud (ICR, Toulouse, France). A cohort of 225 primary melanoma patients operated between 2000 and 2012 in the Institute Claudius Regaud and the CHU Toulouse Purpan were included in this study. The clinical characteristics of the patients are summarized in Table 1. The median age of the patients was 65 y and 51% were male. The most common site of melanoma was head (29%) followed by trunk (26%), members (25%) and extremity (20%). The most common melanoma histologic type was SSM (45.9%) followed by Dubreuilh (21.6%) and Acral-lentiginous (12.8%). The median Breslow depth was 0.9 mm. 36.9% of patients had Clark I-II levels, 27.6% Clark III and 35.5% Clark IV-V. Ulceration was present in 12% of patients and tumor regression was observed for 21%. Sentinel lymph node was evaluated for 76 patients among whom 30% had detectable metastasis.

Immunohistochemistry and immunofluorescence staining

Immunohistochemistry was performed on 5-μm consecutive sections from paraffin embedded tumor blocks using a Techmate Horizon slide processor (Dako, Trappes, france) as described previously.²¹ Details of the antibodies, fixatives and antigen retrieval methods used are provided in Table S5. Briefly, slides were incubated with primary antibodies for 1 h at room temperature, antigen-antibody complexes were detected using a peroxidase-conjugated polymer backbone coupled to secondary antibody system (EnVision, Dako) and 3,3′diaminobenzydine chromogen (Dako). For immunofluorescence detection, slides were incubated with fluorochrome-coupled secondary antibodies, diluted in PBS, BSA 1% for 1 h at room temperature and counterstained with DAPI.

Method for cell quantification

Tumor slides stained with MECA-79, anti-CD3, anti-CD8, anti-CD20, anti-Foxp3 and anti-DC-LAMP antibodies were scanned with a high-resolution scanner (NDP slide scanner, Hamamatsu and Panoramic 250 Flash, 3Dhistech). Absolute numbers of MECA-79⁺ vessels present within the tumor area (mm²) were quantified for each tumor slide and the densities of tumor HEVs (HEV/mm²) were calculated. The amount of CD3⁺, CD20⁺ and CD8⁺ infiltrating cells were evaluated by optical grading as described previously^8,9 (score 1, 2, 3 for absent-low, moderate, and high density of positive cells). Automatic cell counts of CD3⁺ cells were determined for 50 primary melanomas as described previously²¹ with image J software (NIH, Bethesda, USA) using a macro designed by R. Poincloux (IPBS, Toulouse) in order to validate optical grading (Fig. S3). The density of Foxp3⁺ cells was calculated by optical counting of positive cells on 5 representative fields within lymphoid infiltrates surrounding melanoma lesions (0.01 mm²; original magnification 100 x). We evaluated the relative amount of DC-LAMP⁺ infiltrating cells either by optical grading (grade 1, 2, 3 for low, intermediate and high number of tumor-infiltrating DC-LAMP+ cells) or by optical quantification of the number and density of DC-LAMP⁺ cells clusters within melanoma tumor stroma (Fig. S3). Quantification of vessels and immune scoring were performed by two independent observers (L.M. and S.L.).

Quantitative RT-PCR

An RNeasy FFPE isolation kit was used to isolate total RNA (Qiagen, Valencia, CA) from 14 paraffin-embedded melanoma tumor samples with or without tumor HEVs (detected by MECA-79 staining on adjacent paraffin embedded tumor sections). The integrity and the quantity of the RNA were evaluated using a bioanalyzer-2100 (Agilent technologies, Palo Alto, CA). cDNA was prepared by reverse-transcription using superscript VILO cDNA Synthesis Kit (Invitrogen, Paisley, UK). RT-PCR experiments were performed using Power SYBR Green mix with an ABI PRISM 7300HT (Applied Biosystems, Warrington, UK) according to manufacturer instructions. All reactions were done in duplicate and normalized to the expression of the house keeping gene YWHAZ. For each gene, relative expression was calculated by the Δ cycling threshold (CT) method as 2^{–(ΔCTsample)} with ΔCTsample = CT_gene – CT_YWHAZ.

Statistical analysis

Data were summarized by frequency and percentage for categorical variables and by median and range for continuous variables. Comparisons between groups were performed using the Kruskal-wallis test for continuous variables. Correlations between continuous variables were evaluated using Spearman's rank correlation test. All p-values reported were two-sided. For all statistical tests, differences were considered significant at the 5% level. Statistical analyses were performed using the STATA 11.0 (STATA Corp, College Station, TX) software.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Author Contributions

L.M. designed research, performed research, analyzed data and wrote the paper; S.L.G. performed research and analyzed data; T.F. analyzed data; L.L. designed research; N.M. designed research; P.R. designed research; I.G. designed research and analyzed data; J-P.G. designed research, analyzed data and wrote the paper.

Note

Supplemental materials can be found at: http://www.landesbioscience.com/journals/oncoimmunology/article/20492/