Main

Sentinel lymph node biopsy has become the standard of care for accurately staging primary cutaneous melanomas ≥1 mm in Breslow thickness in most major melanoma treatment centers worldwide.1 Patients with a positive sentinel lymph node that undergo completion lymphadenectomy still have variable 5-year survival rates ranging from 39–70% depending on factors such as the number of tumor-bearing nodes, sentinel lymph node tumor burden, presence or absence of primary tumor ulceration, and thickness of the primary melanoma.2 These patients may benefit from adjuvant therapies, which act to reduce the risk of relapse and dissemination to distant sites. Large randomized clinical trials have demonstrated a marginal benefit for adjuvant interferon, but associated significant toxicity has limited its role.3, 4, 5, 6

Recent clinical trials have demonstrated that patients with unresectable AJCC stage IIIC\IV metastatic melanoma treated with drugs inhibiting CTLA-4 on T cells (ipilimumab),7 or the mitogen-activated protein kinase (MAPK) pathway in BRAF mutant melanoma (vemurafenib, dabrafenib, cobimetinib, and trametinib) have significantly improved progression-free survival and overall survival.8, 9, 10, 11, 12 Similarly, improvements in clinical outcomes have been achieved following treatment with monoclonal antibodies (MAb) that target the programmed cell death receptor 1 (PD-1) (also known as CD279) immune checkpoint pathway.13, 14 Success in the metastatic setting with targeted therapies (BRAF and MEK inhibitors) and immunotherapies (CTLA-4, PD-1, and PD-L1 inhibitors) have prompted a number of ongoing phase III randomized clinical trials, as postoperative adjuvant therapies, assessing their efficacy in patients with high risk AJCC stage IIIB/C resectable metastatic melanoma (NCT01274338, NCT00636168, NCT01682083, NCT01667419, and NCT02362594). Whether they also have a role as adjuvant therapies in resected early stage (sentinel lymph node positive) metastatic melanoma is unknown.

PD-1 is expressed by a subset of activated T cells, as well as other immune cells15, 16 and the binding of PD-1 on T cells with its ligands programmed cell death ligand 1 (PD-L1, CD274)17, 18 and programmed cell death ligand 2 (PD-L2, CD273)19, 20 negatively regulates antigen receptor signaling and inhibits immune responses against PD-ligand expressing tumor cells and antigen-presenting cells. PD-L1 is expressed on subsets of T cells, B-cells, macrophages, and dendritic cells and may also be expressed by tumor cells thus facilitating a mechanism of escaping anti-tumor immunity. PD-L1 expression may be further upregulated by cytokines such as interferon gamma (IFN-γ) released by CD4+ helper T cells,21 which is believed to result in feedback suppression and evasion of immune responses against the tumor.22, 23 We have previously demonstrated that there is an influx of CD4+ and CD8+ lymphocytes into melanomas of patients treated with BRAF inhibitor alone24 and more recently with the combination of BRAF and MEK inhibitors.25 It has also recently been shown that a higher density of peritumoral CD8+ lymphocytes at baseline is predictive of response in patients treated with PD-1 inhibitors.26

It is currently unknown whether PD-1/PD-L1 inhibitors will improve survival when administered in an adjuvant setting, and whether known predictors of response such as tumor PD-L1 expression13 and peritumoral CD8+ lymphocytes,26 which have been identified in advanced stage metastatic melanoma patients treated with PD-1 inhibitors, will also predict responses in the adjuvant setting. In this study, we sought to (1) characterize the extent and subpopulations (cytotoxic, helper, and regulatory) of tumor-infiltrating and peritumoral lymphocytes and determine their prognostic significance, and (2) decipher whether the PD-1/PD-L1 pathway contributes to inhibition of the immune system, in sentinel lymph node-positive melanoma patients. The study also sought to inform whether there is a scientific rationale for the use of anti-PD-1 inhibitors in patients with sentinel lymph node positive melanoma and to identify subgroups of patients who are most likely to receive benefit from such treatments.

Materials and methods

Study Design and Patient Selection

This study was undertaken at Melanoma Institute Australia and the Royal Prince Alfred Hospital, Sydney, Australia with Human Ethics Committee approval (X11-0289, HREC/11/RPAH/444). One hundred and ten patients who underwent a sentinel lymph node biopsy procedure at Melanoma Institute Australia and its’ associated hospitals between 1993 and 2008 with pathologically identified metastatic melanoma in their sentinel lymph node were included in this study and, based on clinical follow up data, grouped into either ‘developed recurrence’ or ‘no recurrence’ groups matched for Breslow thickness, ulceration and maximum diameter of the metastasis in the sentinel lymph node. Of these, 60 sentinel lymph node biopsy samples were selected because they had tumor present in all the serial sections (7 × 4-μm thick sections), as assessed by H&E staining performed on the last slide. Patient demographics (age, gender), primary tumor characteristics (Breslow thickness, Clark level, ulceration, melanoma subtype), features of the sentinel lymph node biopsy (site of sentinel lymph node biopsy, number of positive sentinel lymph nodes, and maximum diameter of tumor), and results of the completion lymph node dissection were determined from review of clinical records in the Melanoma Institute Australia Melanoma Research Database. Recurrence-free survival and overall survival were calculated from date of the sentinel lymph node biopsy procedure to an event (either recurrence or death) or until last follow up (censored; Table 1).

Table 1 Clinical and pathological characteristics of the sentinel lymph node patients: (n=60)
Full size table

Immunohistochemistry

All immunohistochemistry staining was performed on 4-μm thick sections using an Autostainer Plus (Dako—Agilent Technologies) with appropriate positive and negative controls. Sections were baked for 60 min at 60 °C in a dehydration oven and heat-induced epitope retrieved in the PT link (Dako—Agilent technologies) using EnVision FLEX target retrieval solution for 20 min at 97 °C then cooled to room temperature in TBST Wash buffer for 5 min. Slides were incubated with the following antibodies at the following dilutions: CD3 (Cell Marque—MRQ39) 1:500, CD4 (Cell Marque—SP35) 1:100, CD8 (Cell Marque—SP16) 1:200, PD-1 (Cell Marque—MRQ-22/NAT105) 1:100, FOXP3 (Abcam—AB22510) 1:200 and PD-L1 (Merck—22C3) 1:1000. The Envision flex Mouse linker (K8022) was used to amplify the signal for PD-1 and PD-L1. Antibody detection utilized DAB chromagen for visualization according to the manufacturer’s instructions (Dako). Slides were then counterstained with hematoxylin.

Assessment of Tumor PD-L1 and Lymphocyte Subsets

Digital copies of all immunohistochemistry slides were made using the Nanozoomer 2.0 HT (Hammamatsu) whole slide scanner at × 20 magnification. For assessment of PD-L1 expression, the percentage of tumor cells demonstrating positive cytoplasmic membrane staining was determined and the intensity of staining was scored utilizing a semiquantitative scale (0–3+): no staining (0), weakly positive staining (1+), moderately positive staining (2+), and strongly positive staining (3+). Positive PD-L1 tumor staining was defined as ≥1% tumor cell expression. Digitally scanned CD3-stained tissue sections were used to generate semiquantitative cell counts within two distinct sentinel lymph node compartments; (1) The peritumoral region, defined as the junction between tumor and immune stroma along a 1.0 × 0.02 mm delimited zone, which whenever possible was selected to be located adjacent to the tumor periphery exhibiting the strongest level of PD-L1 expression; and (2) The intratumoral region, defined by a 0.2 mm2 bounded area containing an immune infiltrate considered to be most representative of the overall intratumoral lymphocytic density (Figure 1). The CD3-stained sections defined the intratumoral and peritumoral regions analyzed in serially cut sections stained for other T-cell subsets (CD4, CD8, PD-1, and FOXP3). The slides were reviewed and assessed by two observers (RV and HK).

Figure 1
figure 1

A schematic figure demonstrating a sentinel lymph node with metastatic melanoma. The number of immunoreactive lymphocytes for each of the immune markers was counted in the 1 × 0.02 mm peritumoral area (green line) at the interface between the tumor and the lymph node parenchyma. The number of immunoreactive lymphocytes for each of the immune markers was counted in the 0.2 mm2 intratumoral area (red box) within the tumor deposit. The example demonstrated is for PD-1+ lymphocytes.

Full size image

Statistical Analysis

Statistical analyses were conducted with ‘PASW Statistics 21’ SPSS, IBM. Bivariate correlations between the different immune markers and tumor PD-L1 expression were conducted using the Spearman’s rho test. Univariate survival analysis was carried out using the Kaplan–Meier method together with the log-rank (Mantel–Cox) test to calculate statistical significance. Cox regression analysis was used to generate univariate hazard ratios, 95% confidence intervals, and corresponding P values. Statistical significance was defined as a probability level <0.05.

Results

Patients and Biopsies

Sixty sentinel lymph node biopsy specimens were available for analysis. There were from 39 male and 21 female patients with a median age of 56 years (range=17–87). In the 60 patients, the primary melanoma was ulcerated in 52% (n=31) of cases, and the median Breslow thickness was 2.65 mm (range=0.85–6.5 mm). The breakdown of the primary melanoma subtypes were 34 nodular, 18 superficial spreading, 4 acral, 2 desmoplastic, 1 lentigo maligna, and 1 nevoid melanoma. The median number of positive sentinel lymph nodes was 1 (range=1–6) and the median maximum diameter of metastatic melanoma present in the sentinel lymph nodes was 2.76 mm (range=0.95–18 mm). There were 47 sentinel lymph node positive patients who subsequently elected to have a completion lymph node dissection, of whom 10 had metastatic melanoma present in one or more non-sentinel lymph nodes, with a median number of 1 positive non-sentinel node (range=1–12; Table 1).

Tumor PD-L1 and Immune Correlates

The membranous expression of tumor PD-L1 was highly heterogeneous with 26 (43%) samples showing expression in ≥1% tumor cell with a median percentage of tumor cells positive in positive cases of 4% (range=0–70% in all samples; Figure 2a,Table 2). The staining intensity of PD-L1 expression varied from negative (0) to strongly positive (3+). The immune markers CD3, CD4, CD8 and PD-1 showed membranous staining in lymphocytes, whereas FOXP3 was expressed within the nucleus of lymphocytes (Figures 2b–f, Table 2). All immune markers showed positive immune-reactivity, both at the interface between the tumor and the normal lymph node (peritumoral), and within the tumor deposit (intratumoral). Tumor PD-L1 expression significantly correlated with the presence of intratumoral CD3+ lymphocytes (R=0.329, P=0.011), peritumoral and intratumoral CD8+ lymphocytes (R=0.256, P=0.048; R=0.448, P=0.0003, respectively), peritumoral and intratumoral PD-1+ lymphocytes (R=0.311, P=0.016; R=0.425, P=0.001, respectively), and peritumoral and intratumoral FOXP3+ lymphocytes (R=0.351, P=0.006; R=0.395, P=0.002, respectively; Table 3). Using nonparametric analysis, patients with positive tumor PD-L1 expression had a larger tumor diameter in their sentinel lymph node biopsy (Mann–Whitney U—P=0.005). There were no significant associations between percentage and intensity of tumor PD-L1 or the immune cell markers with age, sex, number of positive nodes, Breslow thickness, or ulceration of the primary melanoma.

Figure 2
figure 2

Immunohistochemistry undertaken on serial sections from a sentinel lymph node biopsy containing metastatic melanoma. The expression of membranous tumor PD-L1 (a), CD3+ (b), CD4+ helper (c), CD8+ cytotoxic (d), FOXP3+ regulatory (e), and PD-1+ differentiated (f) lymphocytes at × 200 magnification is demonstrated. High-power inserts demonstrate intratumoral FOXP3+ and PD-1+ lymphocyte staining.

Full size image
Table 2 Median % tumor PD-L1 expression and intratumoral and peritumoral lymphocyte counts
Full size table
Table 3 Correlations between tumor PD-L1 and lymphocyte markers
Full size table

Recurrence-Free and Overall Survival

At the time of last follow up (October 2014) the median recurrence-free survival was 24 months (range=1–161 months) and 36 (60%) patients had experienced disease recurrence. The median overall survival at last follow up was 38 months (range=2–186 months). Twenty-four (40%) patients were still alive, of whom 23 were alive with no sign of recurrence and one was alive with disease status unknown. Cox proportional hazard ratio modeling was used to determine optimal cutoff for expression of the various markers for prediction of recurrence-free survival and overall survival. There was no correlation observed between PD-L1 expression and either overall survival or recurrence-free survival in this cohort of patients (Table 4). A positive correlation was observed between higher (>39.5) intratumoral CD3+ lymphocyte count and longer recurrence-free survival and overall survival (hazard ratio=0.36 (0.17–0.76), P=0.005; hazard ratio=0.29 (0.14–0.61), P=0.0005, respectively; Figures 3a and b, Table 4). There was also a positive correlation observed between higher (>24) intratumoral CD4+ lymphocyte count and longer recurrence-free survival and overall survival (hazard ratio=0.34 (0.15–0.77), P=0.007; hazard ratio=0.32 (0.14–0.74), P=0.005, respectively; Figures 3c and d,Table 4) and a positive correlation between the presence of higher (>29) intratumoral CD8+ lymphocyte count and longer recurrence-free survival and overall survival (hazard ratio=0.42 (0.21–0.85), P=0.013; hazard ratio=0.32 (0.19–0.78), P=0.006, respectively) (Figures 3e and f, Table 4). Interestingly, there was a negative correlation between the higher (>1.5) peritumoral PD-1+ lymphocyte count and longer recurrence-free survival and overall survival (hazard ratio=2.67 (1.17–6.13), P=0.016; hazard ratio=2.74 (1.14–6.76), P=0.019, respectively; Figures 3g and h, Table 4). There was no significant association between tumoral PD-L1 expression and recurrence-free survival or overall survival. A trend towards a negative correlation between higher (>7.5) peritumoral FOXP3+ lymphocyte count and longer recurrence-free survival and overall survival was observed, but this did not reach significance (hazard ratio=2.31 (0.87–6.15), P=0.085; hazard ratio=2.25 (0.85–5.97), P=0.096, respectively; Table 4).

Table 4 Correlation between immunohistochemistry and clinical outcome measures
Full size table
Figure 3
figure 3

Univariate Cox regression modeling showed that the recurrence-free survival (days) and overall survival (days) of metastatic melanoma-positive sentinel lymph node patients was positively associated with higher (>39.5, >24, >29, respectively) intratumoral (intra) CD3+ lymphocyte count (a and b), CD4+ lymphocyte count (c and d), and CD8+ lymphocyte count (e and f). There was also a negative association seen in the same patients with higher (>1.5) peritumoral (peri) PD-1+ lymphocyte count (g and h). Inserts demonstrate the optimal cutoff used for each of the markers to separate high and low lymphocyte counts (Significance taken at P<0.05).

Full size image

Discussion

Sentinel lymph node biopsy has become the standard of care for accurately staging melanoma patients with primary tumors >1 mm Breslow thickness and it provides important prognostic information.1, 27 Approximately 40% of patients with a positive sentinel lymph node will develop disease recurrence within 5 years and most of them will eventually die of melanoma.2 Adjuvant therapies have the potential to improve the outcome for these patients. Phase III clinical trial of adjuvant anti-CTLA-4 inhibitor at higher doses than administered in the metastatic setting has also proven to be effective in improving recurrence-free survival in patients at high risk of recurrence compared with placebo.28 Given the efficacy of anti-PD-1/PD-L1 antibodies in patients with advanced stage melanoma, these therapies may also provide benefit in sentinel lymph node positive patients. To the best of our knowledge this is the first study that has evaluated tumor PD-L1 expression, a known biomarker of response to anti-PD-1 inhibitors in advanced stage melanoma, and other immune markers in positive sentinel lymph nodes of melanoma patients. We sought to identify the phenotype of the tumor and immune cells at the tumor-immune stroma interface and within the tumor deposit in an attempt to provide a rationale for the use of anti-PD-1/PD-L1 checkpoint inhibitor therapy in the adjuvant setting for sentinel lymph node positive melanoma patients.

The challenge that faces clinicians who now have an assortment of treatments available for patients with metastatic melanoma is the identification of predictive biomarkers that will help to determine the greatest likelihood of responses to the already considerable array of available targeted and immune therapies. The membranous expression of tumor PD-L1 is heterogeneous both within any given melanoma and between various samples from the same patient.29, 30, 31 Despite this evidence, tumoral PD-L1 expression had been demonstrated to be predictive of response to anti-PD-1 inhibitors.32, 33 Nevertheless, recent studies have shown that some patients with ‘negative’ PD-L1 expression still have objective responses to anti-PD-1 inhibitors32 and anti-PD-L1 inhibitors;34 therefore, reportedly absent PD-L1 expression cannot be used to exclude patients from this form of therapy. In addition, patients with PD-L1 negative tumors treated with the PD-1 antibody nivolumab had an improved overall survival compared with both patients, classified as PD-L1 positive and negative, treated with chemotherapy.32 A recent study reported that peritumoral CD8+ lymphocytes were the most predictive of response to anti-PD-1 inhibitors.26 In our cohort of patients, the expression of tumor PD-L1 correlated highly with the presence of intratumoral lymphocyte subsets, including those that were PD-1 positive. These results support previously reported findings by our group and others,25, 26, 29 which suggest the presence of lymphocytes that secrete IFN-γ, as well as other co-contributing factors, drives the expression of tumor PD-L1 in an acquired adaptive immune evasion phenomenon. This model is also supported by our finding that patients with positive tumor PD-L1 have a higher median diameter of tumor in their sentinel lymph node compared to patients with tumors that were negative for PD-L1. This also suggests that the PD-1/PD-L1 axis may be contributing to disease progression in patients with larger sentinel lymph node metastases. Tumor PD-L1 can also be expressed in a constitutive manner, yet this only accounts for less than 5–10% of tumor PD-L1 positive cases25, 29 and these patients could potentially gain the greatest benefit from specific PD-L1 inhibitors, which show a significant association between tumor or immune cell PD-L1 expression and response to anti-PD-L1 inhibitors.35

The prognostic significance of tumor-infiltrating lymphocyte grade in primary cutaneous melanomas ≥0.75 mm thickness has been demonstrated previously and it has been shown that tumor-infiltrating lymphocytes can provide important prognostic information in a multivariate model with the presence of age, ulceration, tumor thickness and mitotic rate.36, 37 In one study, the presence of tumor-infiltrating lymphocytes in lymph node melanoma metastases predicted recurrence-free survival in multivariate analysis with significantly improved responses seen in patients treated with adjuvant interferon therapy.38 In our study we found that the presence of high intratumoral CD4+ helper T cells and CD8+ cytotoxic T cells in sentinel lymph node melanoma metastases predicted for recurrence-free survival and overall survival. To the best of our knowledge, prognostic associations of tumor-infiltrating lymphocytes and tumor-infiltrating lymphocyte subsets in sentinel lymph node metastases from melanoma patients have not previously been reported. Furthermore, the negative association between the presence of peritumoral PD-1+ differentiated lymphocytes and both recurrence-free survival and overall survival has not been previously reported in untreated sentinel lymph node positive melanoma patients and could underpin one of the mechanisms of immune resistance that melanoma cells employ to overcome eradication in the hostile microenvironment of the sentinel lymph node. The overexpression of FOXP3+ regulatory T cells in lymph nodes containing metastatic melanoma samples has been shown to inhibit the function of infiltrating CD4+ helper and CD8+ cytotoxic T cells through a cell-contact dependent mechanism,39 which supports the trend in our study of an increased hazard ratio >2 for recurrence-free survival and overall survival in patients with higher peritumoral FOXP3+ lymphocytes. It is possible that this association failed to reach significance due to the relatively low number of uncensored patients in this cohort (n=36), and multivariate analysis including tumor thickness, mitotic rate and ulceration of the primary tumor needs to be undertaken on a larger set of samples.

In conclusion, the frequent presence of tumoral PD-L1 expression in metastatic melanoma deposits in sentinel lymph nodes and the immune microenvironment establishes a scientific rationale for administering adjuvant immune checkpoint inhibitor therapy with either anti-PD1/PD-L1 inhibitor in AJCC stage IIIA sentinel lymph node positive melanoma patients with the potential for improving clinical outcomes. Furthermore, our data also demonstrated the prognostic significance of tumor-infiltrating lymphocytes and lymphocyte subsets in sentinel lymph node positive melanoma patients, which need to be taken into account when design, conduct and analysis of clinical trials assessing the efficacy of immune therapies targeted to tumor-infiltrating lymphocytes. It also has the potential for identifying those patients that will receive the greatest benefit from this treatment regimen.