Molecular analysis of melanoma-induced sentinel lymph node immune dysfunction

Abstract

Introduction

Sentinel lymph nodes (SLNs) of melanoma patients show evidence of tumor-induced immune dysfunction. Our previous works have shown that IL-10 and IFNγ co-regulate indoleamine-2,3-dioxygenase (IDO)-expressing immunosuppressive dendritic cells (DCs) in melanoma SLNs. The goal of this study is to examine the relationship between melanoma SLN tumor burden and the degree of SLN immune dysfunction as a model to study tumor-induced immune dysfunction. We hypothesize that SLN tumor burden correlates with the degree of SLN immune dysfunction.

Methods

Patients undergoing SLN biopsy for clinical stages I and II melanomas were enrolled in the study under an IRB-approved protocol. During the SLN biopsy, non-hot and non-blue portion of the SLN was harvested, flash-frozen in liquid nitrogen, and mRNA was extracted. By using quantitative real-time PCR, gene expressions of cytokines (IL-4, IL-10, IFNγ, TGFβ, GM-CSF) and the surrogates of immunosuppressive regulatory and effector cells (IDO-expressing DCs and Foxp3-expressing T-regs, respectively) were measured and correlated against the SLN tumor burden (MART1) and against each other. The data were log transformed for normalization. Statistical test used Student’s t-test and stepwise multivariate regression analysis. Statistical significance was determined at P < 0.05.

Results

SLNs of 74 patients were analyzed in this analysis. Ten of seventy-four patients (13.5%) had tumor-positive SLNs. MART1 gene expression showed a significant difference between the SLN (+) and SLN (−) groups (P = 0.04). Among the various cytokines, multivariate analysis showed that only IFNγ gene expression correlated independently with MART1 gene expression (P < 0.0001, r = 0.91). Similar multivariate analyses show that IFNγ (P < 0.0001, r = 0.78), IL-10 (P = 0.0037, r = 0.60), and TGFβ (P < 0.0001, r = 0.95) gene expressions correlated independently with IDO gene expression. IFNγ (P < 0.0001, r = 0.87) and GM-CSF (P = 0.042, r = 0.76) gene expressions correlated independently with Foxp3 gene expression. MART1 gene expression showed independent correlation with IDO (P = 0.0002, r = 0.75) and Foxp3 (P = 0.0002, r = 0.75) gene expressions.

Conclusion

SLN tumor burden correlates with immunosuppressive IDO and Foxp3 expressions within the SLNs of melanoma patients. Our data are consistent with our theory that melanoma induces expressions of specific cytokines, which in turn, stimulate immune suppressors within the SLN. This study also supports our previous finding that IL-10 and IFNγ co-regulate IDO within the SLN. In our data, IFNγ is the sole cytokine that correlates with the SLN tumor burden and seems to play a central role in tumor-induced immunological changes in the SLN immune microenvironment.

Introduction

Primary melanomas rarely kill people. The standard treatment for clinically early-stage melanoma is complete resection of the primary lesion and the tumor-draining lymph node (LN) basins. It is a systemic disease and/or distant recurrence (metastatic disease) that cause mortality. However, the term “recurrence” is a misnomer, as it actually represents eventual clinical manifestation of occult disease (micrometastasis) already present at the time of initial diagnosis. More than 80% of melanoma patients who will eventually die of melanoma have occult distant metastases at the time of their initial diagnosis of apparently localized primary melanoma. Unfortunately, currently available radiographic imaging studies used in staging work-up, such as CT, PET, and MRI cannot detect micrometastases. Thus, we are currently unable to determine who has any occult distant disease, or when they will progress to clinically more advanced disease. Clinically, we use TNM staging system to risk-stratify recurrence and death. Tumor involvement of the regional LNs marks the beginning of stage III disease and is the single most important prognostic factor in clinically early-stage melanomas [1, 2]. Patients with regional LN metastasis have worse prognosis than those without, even if they have the same Breslow thickness of their primary melanoma [2]. Even among the patients who develop systemic disease, those with history of having regional LN metastasis have significantly worse prognosis than those without [3]. In addition, among the stage III patients, those with higher nodal tumor burden, as measured by the size of nodal metastasis and/or number of tumor-involved LNs, have significantly worse prognosis [1–6]. These findings suggest that regional nodal metastasis and nodal tumor growth from micrometastasis to macrometastasis are not just passive events dependent on time, rather, they are active events reflecting independent measures of overall aggressive potential. Therefore, identification of factors associated with, if not causative of, nodal tumor growth from a micrometastatic deposit into a macroscopic nodule is a crucial step in understanding pathophysiology of metastatic tumor growth and ultimately, clinical manifestation of once occult metastatic disease.

Significant evidence exists to support the idea that melanoma progression and growth at a systemic level is partly dependent upon melanoma-induced systemic immune dysfunction. Therapeutic strategies that can recover patients’ own anti-tumor immunity can significantly deter disease progression. An important step in designing a rational therapeutic strategy based on such concept requires an understanding of in vivo microenvironmental immune factors that influence progression of micrometastases into macrometastases. Such study, however, poses a unique challenge in that we do not know where these micrometastases are, or when they will progress. There is hope, however. Although it is currently not possible to study directly the micrometastases and their microenvironments within other bodily organs, lymphatic mapping and selective sentinel lymphadenectomy (LM/SL) enable us to study such parameters in regional LNs. Sentinel lymph node (SLN) is the first LN draining a tumor bed, thus the first site of regional lymphatic metastasis [7]. Evaluation of SLN provides a unique opportunity to understand the initial phase of tumor-immune interaction. Many studies suggest that early-stage melanoma patients are immunosuppressed loco-regionally, including the SLNs [8–12]. SLNs show notable histological evidence of immune dysfunction as evidence by a significant decrease in T-cell-rich paracortical areas and interdigitating dendritic cells (IDCs), which are involved in antigen processing and subsequent T-cell activation [9]. In addition, LNs containing metastatic melanoma also contain significantly higher frequency of immunosuppressive, Foxp3-expressing, CD4⁺CD25^high T-cells (T-regs). Furthermore, melanoma SLNs show presence of indoleamine-2, 3-dioxygenase (IDO) expressing DCs, which are involved in induction of tolerogenic T-cells [13]. IDO is a rate-limiting enzyme in tryptophan catabolism [14]. Catabolism of tryptophan by IDO-expressing DCs has shown to inhibit T-cell proliferation in vitro [15]. In addition, IDO-expressing DCs have been shown to reduce T-cell response in vivo [16]. In vitro suppressor activities of IDO-expressing DCs can be abrogated by 1-methyl-tryptophan (1-MT), a competitive inhibitor of IDO. Clinical importance of SLN immunosuppression is demonstrated by a significantly worse prognosis among the patients with low SLN IDC density or high SLN IDO expression [4, 17].

Numerous in vitro studies suggest that local cytokine milieu can influence monocyte differentiation as well as DC maturation. For example, a monocyte population of peripheral blood mononuclear cells (PBMCs) can be differentiated into DCs by incubation in GM-CSF/IL-4-enriched media. A study by Palucka et al. indicates that monocyte-derived CD1a⁺/CD14⁻ DCs can revert to CD1a⁻/CD14⁺ macrophage phenotype, and vice versa, depending upon the locally dominant cytokine environment [18]. This plasticity is maintained until the terminal stage of DC differentiation. Geissman et al. demonstrate that GM-CSF/IL-4-derived CD1a⁺ DCs grown in TGF-β1 and TNFα show a persistent immature phenotype but become functionally mature once exposed to CD40L [19]. These studies suggest that local cytokine milieu and immune-stimulatory signals can have a significant impact on maturation and subsequent function of immune regulatory cells [18]. Interestingly, TGFβ has been shown to be an independent inducer of IDO [20]. Since loco-regional treatment with rhGM-CSF have shown to increase the SLN IDC density and SLN tumor-specific CD8⁺ T-cell reactivity, reversing the SLN immune dysfunction may also represent a valid therapeutic target [11, 21]. Though the concept of tumor-associated immune dysfunction has been around for some time, the precise mechanism remains largely unknown. We, and others, have shown that SLN immune dysfunction is induced by melanoma, and is mediated by differential regulation of SLN cytokine microenvironment, such as INFγ, GM-CSF, and IL-10 [11, 22].

Previous studies have shown that tumor-induced regional nodal and systemic immune dysfunctions are mediated partly by cytokines, via defect in DC and T-cell-mediated immunity [11, 23]. We have also demonstrated that immunosuppressive SLN cytokine profile can be reversed by removal of neighboring melanoma, and that rhGM-CSF therapy can potentially restore SLN immune function. Such findings are also paralleled in systemic setting, as INFγ and GM-CSF have been shown to be effective in select cohort of advanced-stage melanoma patients [24–26]. In this study, we exploited a unique opportunity within SLNs to understand the immunological molecular mechanism of metastatic growth and its association with SLN tumor burden. We hypothesize that growth and subsequent progression of SLN metastases are associated with differential cytokine microenvironment within the SLN that influence SLN immune milieu.

Methods

Patient accrual and pre-operative evaluation

From May 2007 to December 2009, all consecutive patients seen at The Cancer Institute of New Jersey (CINJ) Melanoma and Soft Tissue Oncology Program undergoing SLN biopsy for clinical stages I and II cutaneous melanomas were evaluated for enrollment in the study. Pre-operatively, the patients underwent routine pre-operative evaluation and lymphoscintigraphy to identify the lymphatic basin at risk. This prospective study was approved by the University of Medicine and Dentistry of New Jersey (UMDNJ)/CINJ Institutional Review Board.

Tissue procurement and storage

All patients underwent pre-operative peri-lesional and intra-dermal injection of technetium-99 labeled sulfur colloid by the members of Robert Wood Johnson University Hospital Nuclear Medicine Department within 6 h of the operation. In addition, peri-lesional and intra-dermal injection of 1% Lymphazurin was injected by the operating surgeon just prior to the operation. The SLNs were identified by using both Neoprobe and visual inspection of the blue dye. Our experience has shown that the radiotracer and the blue dye used to identify the SLN concentrate within a specific geographic portion of the SLN near the afferent lymphatics, producing a relatively “hot and blue spot” within the node. We also noted that if SLN metastasis is present, it will be located at this “hot and blue spot”. Therefore, in order to avoid compromise in pathologic diagnosis, non-hot and non-blue portion of the SLNs (<30%) was harvested intra-operatively and flash-frozen in liquid nitrogen by the members of the Tissue Analytical Services (TAS) at the CINJ. The specimens were cataloged and stored by the TAS under a standard operating protocol (SOP) and stored in the −70°C freezer.

RNA extraction and quantitative real-time PCR (qRT-PCR)

RNA was extracted from the specimens by using Trizol reagent and purified by using RNeasy mini kit (Qiagen, Valencia, CA) according to the manufacture’s guidelines. After purification, the RNA concentration was determined by absorbance at 260 nm on a spectrophotometer. A concentration of 5.0 μg of high-quality total RNA was used in cDNA synthesis by using random primers and SuperScript II (Invitrogen, Carlsbad, CA). Quantitative real-time PCR was performed by using Taqman^® 2× universal PCR Master mix and No AmpEaseUNG (Roche) on Strategene MX4000. Gene expressions of the following genes were analyzed: GAPDH, IL-4, IL-10, TGFβ, IFNγ, GM-CSF, IDO, Foxp3, and MART1. The results were normalized to GAPDH expression. Forward (F) and reverse (R) primers, and probe (P) sequences for each gene are as follows:

GAPDH (F-ACAACTTTGGTATCGTGGAAGG; R-CAGTAGAGGCAGGGATGATGTTC; P-ACCCAGAAGACTGTGGATGG), IL-10 (F-CCAAGCTGAGAACCAAGACC; R-CATGGCTTTGTAGATGCCTTT;P-GGGAGAACCTGAAGACCCTC), IFNγ (F-GCCAGGACCCATATGTAAAAG; R-GGCTCTGCATTATTTTTCTGTC; P-TGCAGGTCATTCAGATGTAGCGGA), GM-CSF (F-ATGTGAATGCCATCCAGGAG; R-CGGGTCTGTAGGCAGGTC; P-TCAGAAATGTTTGACCTCCAGGAGCC), IDO (F-ACTGTGTCTTGGCAAACTGG; R-CTTTACTGCAGTCTCCATCACG; P-AGCCCCTGACTTATGAGAACATGGACG), Foxp3 (F-CATGATCAGCCTCACACCAC; R-ATTTGGGAAGGTGCAGAGC; P-GATCAACGTGGCCAGCCTGGAAT), MART1 (F-ATGCCAAGAGAAGATGCTCAC; R-AGCATGTCTCAGGTGTCTCG; P-CAAGGCTCTGTATCCATTTCGTC), TGFβ (F-TGGAAACCCACAACGAAATC; R-TTTAACTTGAGCCTCAGCAGAC; P-TTCAACACATCAGAGCTCCGAGAAGC), IL-4 (F-ACTGCACAGCAGTTCCACAG; R-CTCTGGTTGGCTTCCTTCAC; P-AGCTGATCCGATTCCTGAAA).

The thermocycler parameters were 95°C for 10 min for initial denaturation followed by 40 cycles of denaturation at 95°C for 30 s and annealing and extension for 30 s. The annealing temperatures for genes were IL-4: 60°C, GAPDH: 60°C, TGFβ: 60°C, MART1: 60°C, Foxp3: 60°C, IDO: 58°C, GM-CSF: 55°C, IFNγ: 57°C, and IL-10: 57°C.

Statistical analyses

The results of qRT-PCR were normalized by using log transformation. Univariate analyses to determine the correlation between the variable were done by using Pearson’s r. Two group comparisons were performed by using Chi-square analysis, Wilcoxon Rank Sum test, or Student’s t-test (after log transformation). Multi-group comparisons were done by using Kruskal–Wallis test. Multivariate analyses to identify the cytokines associated independently with MART1, IDO, or Foxp3 were done by using multiple regression analysis. Multivariate logistic regression analysis was done in a stepwise fashion to identify the molecular and clinicopathologic factors associated independently with SLN tumor positivity. Statistical significance was determined at P = 0.05. Medcalc v.7.0 software was used to perform the statistical analyses.

Results

MART1 gene expressions are significantly different between SLN (+) and SLN (−) groups

SLNs of 74 patients were analyzed in this analysis. The clinicopathologic features of these patients are listed in Table 1. Thirty-three patients were female and 41 patients were male. Mean age at diagnosis was 55 years (median = 54), and mean Breslow thickness of the group was 2.49 mm. Ten of 74 patients (13.5%) had tumor-positive SLNs. Given that ulceration and mitotic rates are now included in T-staging of melanoma patients, we included those variables in our analyses. Table 2 shows the comparison of prognostic features in the SLN (+) and SLN (−) groups. Among the clinicopathologic features (age, Breslow, ulceration, mitosis), only Breslow thickness was significantly different between the two groups. However, given that we only had 10 (13.5%) patients with SLN tumor positivity, higher sample number might have shown that other factors could have been different between the two groups. MART1 gene expression was used as a surrogate for molecular tumor burden within the SLN, and thus we hypothesized that SLN (+) groups will have higher gene expression levels of MART1. MART1 gene expressions showed a significant difference between the SLN (+) and SLN (−) groups, demonstrating approximately tenfold (1 log unit) difference (Table 2). Considering that tumor infiltrating lymphocytes (TIL) at the primary site can potentially be a marker of immune functions, we also examined the relationship between TIL (Absent, Non-brisk, Brisk) and MART1, cytokines, IDO, and Foxp3. Univariate analyses showed no significant differences between the three TIL groups.

Table 1.

Clinicopathologic features of the 74 patients

	n	Mean	SD	Yes	No	Unknown
Total	74
Male	41
Female	33
Age (years)		55.4	13.7
Breslow (mm)		2.49	2.8
Ulceration				18	53	3
Mitotic rate ≥1/mm2				47	19	8
SLN positive				10	64
TIL (primary melanoma)
Absent	13
Non-brisk	33
Brisk	15
Unknown	13
Site
Trunk	22
Extremities	42
Head and neck	10

SD Standard deviation; SLN Sentinel lymph node; TIL Tumor infiltrating lymphocytes

Table 2.

SLN status and prognostic features

	n	SLN (+)	SLN (−)	P
Age (median values in years)	74	56	53	0.734*
Breslow (median values in mm)	74	4.1	1.4	0.004*
Ulceration (% positive)	69	40.0%	22.0%	0.591**
Mitosis (% ≥1/mm2)	64	90.0%	70.4%	0.814**
MART1 gene expression in SLN (mean of log values)	73	−3.48	−4.48	0.043***

* Wilcoxon rank sum test

** Chi-square test

*** Student’s t-test

IFNγ gene expression is central in SLN tumor burden and immune regulatory gene expressions

In univariate analyses, all of the tested cytokine gene expressions (IL-4, IL-10, IFNγ, GM-CSF, and TGFβ) showed statistically significant correlation with (P < 0.001) with MART1, IDO, and Foxp3 gene expressions (Table 3). Despite statistical significance, IL-10 gene expression showed only low level correlations with MART1 (r = 0.53), IDO (r = 0.47), and Foxp3 (r = 0.40). Multivariate analyses including all 5 cytokines as predictor variables and MART1 as the responder variable showed that only IFNγ gene expression correlated independently with MART1 gene expression (Fig. 1; P < 0.0001). In evaluating the cytokines associated (predictor variables) with IDO-expressing DCs (responder variable), the multivariate analyses including all 5 cytokines show that IFNγ (Fig. 2a; P < 0.0001), IL-10 (Fig. 2b; P = 0.0037), and TGFβ (Fig. 2c; P < 0.0001) gene expressions correlated independently with IDO gene expression. Similarly, in evaluating the cytokines associated with Foxp3-expressing T-regs, IFNγ (Fig. 3a; P < 0.0001) and GM-CSF (Fig. 3b; P = 0.042) gene expressions correlated independently with Foxp3 gene expression. MART1 gene expression showed independent correlation with IDO (Fig. 4a; P = 0.0002) and Foxp3 (Fig. 4b; P = 0.0002) gene expressions, confirming the link between melanoma and immunosuppressive SLN microenvironment.

Table 3.

Univariate analysis of cytokine gene expression

Correlation coefficient, r
Cytokine	MART1	IDO	Foxp3
IL-4	0.66	0.83	0.69
IL-10	0.53	0.47	0.40
IFNγ	0.87	0.83	0.89
GM-CSF	0.73	0.70	0.73
TGFβ	0.64	0.91	0.66

P < 0.001 for all values

Fig. 1.

Bivariate analysis of the factors that are shown to be significant in the multivariate analysis. IFNγ versus MART1 in SLNs of melanoma patients (P < 0.0001)

Fig. 2.

Bivariate analysis of the factors that are shown to be significant in the multivariate analysis. a IDO versus IFNγ in SLN’s of melanoma patients (P < 0.0001). b IDO versus IL-10 in SLNs of melanoma patients (P < 0.005). c IDO versus TGFb in SLN’s of melanoma patients (P < 0.0001)

Fig. 3.

Bivariate analysis of the factors that are shown to be significant in the multivariate analysis. a Foxp3 versus IFNγ in SLNs of melanoma patients (P < 0.0001). b Foxp3 versus GM-CSF in SLNs of melanoma patients (P < 0.0001)

Fig. 4.

Bivariate analysis of the factors that are shown to be significant in the multivariate analysis. a MART1 versus IDO in SLNs of melanoma patients (P < 0.001). b MART1 versus FOXP3 in the SLNs of melanoma patients (P < 0.001)

Breslow thickness and Foxp3 gene expression predictive of SLN status

In a stepwise fashion, a multivariate logistic regression model was used to identify the factors that are independently predictive of SLN status. Included variables were Breslow thickness and IFNγ, Foxp3, and IDO gene expressions. Mitotic rate and ulceration were not included in this model because these factors failed to show significance in our univariate model (Table 2). Seventy-two patient samples were included in the final model. The results show that only Breslow thickness (P = 0.002) and Foxp3 gene expression (P = 0.02) were independently predictive of SLN positivity.

Discussion

SLN is a rather unique environment in which we can evaluate the immune microenvironment when tumors are microscopic. Thus, we can use SLNs as a model to study metastatic microenvironments at various stages of tumor growth. We, and others, have shown that SLN tumor growth from micrometastasis to macrometastasis can adversely impact clinical outcome [1, 2, 4–6, 27]. Melanoma SLNs (when compared to neighboring non-SLNs) show notable histological changes in immune effector and regulator profiles evidenced by a significant decrease in T-cell-rich paracortical areas, as well as profound decrease in area and density of IDCs, which are involved in antigen processing and subsequent T-cell activation [9]. Decreased IDC density within SLNs is associated with increased tumor recurrence and death, suggesting that functional changes in SLN immune microenvironment can predict clinical outcomes [4]. These studies also suggest a functional role of melanoma in inducing such changes in SLN immune microenvironment. Our previous work has shown that in presence of melanoma, SLNs show higher gene expressions of IL-10, IFNγ, and IDO [11].

Another study showed up-regulation of Foxp3-expressing CD4⁺CD25^high regulatory T-cells (T-regs) in the LNs with metastatic melanoma [10]. In their study of 12 AJCC stage III melanoma patients who underwent complete LN dissection, Viguire et al. have shown that within the LNs that contain metastatic melanoma, the mean frequency of Foxp3-expressing CD4⁺CD25^high T-cells among the CD4⁺ T-cells was 11.06% [10]. This represents approximately twofold increase when compared to non-tumor containing matching LNs (6.2%, P < 0.0006) and/or autologous PBMCs (4.9%, P < 0.0001). Foxp3-expressing CD4⁺CD25^high T-regs suppress in vitro proliferation, and IFNγ and IL-2 production of CD4⁺CD25⁻ T-cells, and well as proliferation of CD8⁺ T-cells [10]. Furthermore, upon allogeneic stimulation, these T-regs produced significantly higher IL-10 (and TGFβ to lesser extent) when the LNs contained large metastatic melanoma, in contrast to minimal IL-10 production in setting of small (minimal) metastatic melanoma deposit in the LNs. These in vitro suppressor activities of the T-regs were abrogated in the presence of anti-IL-10 antibody (Ab), demonstrating the importance of cytokine environment in the functional activity of these T-regs [10]. These findings suggest that it is the prevailing cytokine microenvironment that determines the activity of the effectors (i.e., CD4⁺CD25^high T-regs), and may ultimately determine the functional immune status of the SLNs.

We, and others, have shown that the immune microenvironmental changes in SLN are induced by melanoma and are mediated by differential regulation of SN cytokine microenvironment. For example, Leong et al. have shown, by using ELISPOT assay, that in absence of SLN metastasis, the SLNs showed significantly higher expression of IFNγ, IL-2, and GM-CSF when compared to the neighboring non-SLNs from the same nodal basin [22]. However, in the presence of SLN micrometastasis, the noted elevations of IFNγ, IL-2, and GM-CSF in the SNs were no longer observed, further supporting a functional link between SLN metastasis and SLN immune microenvironment. Furthermore, by using semi-quantitative PCR, we have demonstrated that SLNs from early-stage melanoma patients expressed significantly higher IL-10 than matching adjacent non-SLNs. In our subsequent study, we examined the role of melanoma in influencing the cytokine microenvironment within the SLNs [11]. The results show that in presence of neighboring melanoma (including SLN metastasis), the SLNs showed significantly higher expression of IL-10, IFNγ, and IDO when compared to adjacent non-SLNs. When neighboring melanoma was completely removed, the observed elevations of IL-10, IFNγ, and IDO gene expression levels in the SLNs were no longer observed, suggesting the reversibility of SLN cytokine profile once melanoma is removed, potentially reversing SLN immune dysfunction. In addition, as shown in our pilot study, when the patients were injected pre-operatively with rhGM-CSF at their tumor sites, we noted a reversal of histological changes in the SLNs as evidenced by restoration of T-cell area, and IDC area and density, suggesting that nodal cytokine manipulation can potentially restore SLN immune function. Such conclusion is further supported by the finding noted in the study by Vuylstke et al. in which they show that pre-operative peri-lesional injection of rhGM-CSF results in increase tumor-specific cytotoxic CD8⁺ T-cells in the melanoma SLNs [21].

The data from this current study provide a strong support for this line of thought. Based on the studies mentioned, we hypothesized that melanoma SLN immune dysfunction correlates with the SLN tumor burden in vivo. Thus, we examined gene expressions of several cytokines that are likely to be involved in immune regulators (IDO-expressing DCs) and effectors (Foxp3-expressing T-regs) within the SLNs of melanoma patients undergoing LM/SL. Average SLN is about 1 cm³ in volume, and ethically, we cannot use most of it for research, lest we can compromise patients’ staging and medical care. Given that we only have access to very small amount of tissues, we decided to study gene expressions using qRT-PCR. We used MART1 gene expression as a surrogate for molecular tumor burden. Thus, we demonstrate that MART1 gene expression is significantly higher in the tumor-positive SLNs (Table 1). We hypothesized that SLN tumor burden influences the immune function via differential regulation of cytokines.

Our data show that among the various cytokines tested, IFNγ seems to play a central role in melanoma-induced SLN immune regulation (Table 3, Figs. 1, 2a and 3a), correlating independently and significantly with MART1, IDO, and Foxp3. It might be that during the initial phase of tumor-immune interaction, a chronic low-level inflammatory response induced by melanoma and mediated with IFNγ is what eventually leads to immunosuppression within the SLN microenvironment. In addition, IL-10 and IFNγ gene expressions independently correlated with IDO gene expression (Fig. 2a, b) supporting our previous finding that IDO may be co-regulated by IL-10 and IFNγ in human melanoma SLNs. Our data also suggest co-regulatory role of TGFβ in IDO expression (Fig. 2c), and this will be further studied in our future works. The data also show that IFNγ and GM-CSF independently correlated with Foxp3 expression. Paradoxically, it might be the case that “pro-inflammatory” cytokines can actually induce immune dysfunction in these patients. This would suggest that currently available immunotherapies can benefit from pharmacologically blocking IDO and/or Foxp3. Using an animal model, Quezada et al. [28] have shown that GM-CSF-transduced tumor cell vaccine (Gvax) showed increase in CD4⁺T_eff/T_reg ratio within the tumor, and better efficacy when combined with CTLA-4 antibody. To complete the hypothesis, our data also show that MART1 gene expression correlates independently and significantly with IDO and Foxp3 gene expressions (Fig. 4a, b), thus supporting the idea that it is melanoma tumor burden that incites DC-mediated T-cell dysfunction [9, 11]. Such idea is also supported by the fact that Foxp3 independently predicts SLN tumor status, as shown in our multivariate logistic regression analysis.

Summary and conclusion

Direct melanoma-draining LNs show evidence of immune dysfunction. It has been shown previously that such immune dysfunction is induced by melanoma and mediated by immunosuppressive cytokine microenvironment. The results shown in this study strongly supports the theory that initial proinflammatory microenvironment induced by melanoma eventually leads to immunosuppressive cytokine microenvironment within the SLN, resulting in induction of immunosuppressive regulators (IDO-expressing DCs) and effectors (Foxp3-expressing T-regs), with subsequent tumor growth within SLNs. Our results also show that IFNγ seems to play a central role in this complex cascade and warrants further analysis of its role in initial phase of tumor-induced immune suppression. In addition, one can infer from our study the merits of combining existing immunotherapies with agents that block IDO and/or Foxp3 in future human immunotherapy trials.