For this study, seven independent sets of tissue and serum samples from melanoma patients were analyzed. Furthermore, an additional set comprising 237 cDNA samples was utilized for a gene expression analysis of the S100 super-family. The study was in accordance with the ethical standards and was approved by the local ethics committee of the University Medical Center Mannheim (Project number 2010-318 N-MA).

Immunofluorescence analysis was performed as described previously [16], and nuclear staining was done with H33342 (EMD). The following primary antibodies were used: anti-S100A8/A9 (sc-33714, SantaCruz) and anti-S100B (ab189418, Abcam).

DNA collection, RNA preparation, procession via SAGE™, and cDNA gene expression analysis was done as described earlier [17, 18]. Briefly, tissue samples of 100 melanoma metastases, 67 primary cutaneous melanomas, and 70 melanocytic nevi were either collected at the Departments of Dermatology at the Universities of Cologne, Bonn or Aachen. Tissue samples were flash-frozen in liquid nitrogen immediately after surgery. Total RNA was isolated as described earlier [19]. A PIQOR™ (Parallel Identification and Quantification of RNAs) microarray (Miltenyi Biotec GmbH, Bergisch Gladbach, Germany) was designed on the basis of SAGE™ analysis according to the procedures previously described [19]. Cy5-labeled RNA from tumor or nevus samples was hybridized against a Cy3-labeled common skin reference pool as described earlier [17, 19]. Hybridization, scanning and data analysis were performed according to the PIQOR™ protocol and in compliance with the MIAME (minimum information about a microarray experiment) standards [19–21].

The analyzes of protein expression of S100A8/A9 in the above mentioned three independent TMAs were conducted using formalin fixed and paraffin embedded tissue according to a previous report [7]. Briefly, tissue punch samples (0.6 mm diameter) were taken from tumor or nevus tissue, respectively, and collected in a single TMA block each. Immunohistochemical staining was performed as described previously [7] using antibodies specific against S100A8/A9 heterodimer (sc-33714, SantaCruz). Evaluation of the stained slides was done by two blinded experienced investigators. In order to distinguish chromogen DAB from melanin pigment parallel sections stained lightly with H&E were used upon evaluation. Staining intensity was analyzed based on pathological scoring as previously described [7].

Three independent tissue microarrays (TMA) were utilized for analyzes of S100A8/A9 expression in melanoma tissue. TMA 1 contained samples of benign melanocytic nevi, non-metastasizing primary melanomas and metastasizing primary melanomas. TMA 2 and TMA 3 contained samples of primary melanomas and melanoma metastases deriving from patients who have all been diagnosed with metastatic disease. These two TMAs were designed to compare long-term vs. short-term survivors. Long-term and short-term referred to the timespan between first occurrence of metastatic disease (at this time the metastatic tissue samples were obtained) and death. Less than 12 months was considered short-term survival, more than 30 months was considered long-term survival. The TMAs were scored by experienced dermato-histopathologists in a blinded fashion regarding to the outcome of the patients.

Collection of serum and documentation of clinical data were performed after patients’ informed consent with institutional review board approval. Blood draw was done using gel-coated serum tubes (Sarstedt). After centrifugation, serum was immediately stored at − 80 °C. Serum concentration of S100A8/A9 was measured in duplicate using commercially available sandwich ELISA kits (Bühlmann Laboratories AG, Switzerland). S100B and LDH levels had been determined routinely during regular follow-up.

S100A8/A9 serum levels were measured in two independent sets of serum samples from 114 stage III and stage IV melanoma patients treated between 1990 and 2009 at University Hospital Essen, Germany (training set), and from 240 stage III and stage IV melanoma patients treated between 2007 and 2010 at University Hospital Tübingen, Germany (independent validation set), respectively. All samples were collected immediately after first diagnosis of stage III or stage IV disease, respectively. None of the patients had been treated systemically within 4 weeks before the blood withdrawal. Serum samples were chosen from the biobanks of the two University Hospitals Tübingen, and Essen, respectively, based on the amount of serum available. 32 of the samples of the training set were excluded due to missing follow-up information. None of the 354 patients of both sets were treated with CTLA-4 or PD-1 antibodies, nor with BRAF or MEK inhibitors. All patients included in this study were followed-up and were staged systematically by the departments for Dermatology of the two University Hospitals Tubingen and Essen according to the German melanoma guidelines.

S100A8/A9 serum levels were analyzed in two independent sets of 27 and 44 patients, respectively, treated with the anti-PD-1 antibody pembrolizumab at University Hospital Mannheim, Germany (pembrolizumab set 1) and at University Hospital Tubingen, Germany (pembrolizumab set 2). The serum samples of the pembrolizumab treated patients were collected prospectively after written informed consent was obtained.

Samples were selected according to the following criteria: histologically confirmed cutaneous melanoma, complete documentation of medical history, course of the disease, and follow-up. Follow-up time started at the date of initiation of pembrolizumab treatment and ended at the date of last follow-up or death. Primary endpoint in pembrolizumab set 1 was overall survival (OS). Median OS was not reached in pembrolizumab set 2. Therefore, primary endpoint in this set was progression-free survival (PFS). Patients received at least one cycle of pembrolizumab over 30 min at a dose of 2 mg/kg body weight. Treatment was repeated every 3 weeks according to the protocol approved by the European medicines agency (EMA). Staging was performed every 3 months according to the structured staging guidelines of the University Hospitals Mannheim and Tubingen, Germany. Radiologic responses were assessed using contrast-enhanced CT/MRI/PET-CT at around week 12 after the first pembrolizumab infusion and clinical response was defined based on immune-related response criteria (irRC). The peripheral blood was taken up to 5 days before or on the day of the first infusion.

All cut-off values in this study were determined utilizing a previously described algorithm which selects the ideal cut-off point based on minimizing the P-value [22]. For the TMA analyzes the cut-off value for the number of S100A8/A9 expressing cells was 55%, for all serum marker analyzes the cut-off value was 5.5 mg/l. Comparisons of continuous factors were done with two-sided Mann-Whitney U test. Estimates of cumulative survival probabilities according to Kaplan-Meier were compared using two-sided log-rank test. Multivariate Cox proportional hazard analyses were used to evaluate the independent effects of S100A8/A9 on survival. Throughout the analyses, p-values less than .05 were considered statistically significant. All analyses were carried out using R.

Immunofluorescence analysis of S100A8/A9 expression in tissue sections of primary melanomas revealed an exclusive and abundant expression of S100A8/A9 in cells of the TME, mainly granulocytes, whereas S100B expression was restricted to melanoma cells (Fig. 1a and b).Fig1Fig. 1

S100A8/A9 is expressed exclusively and abundantly in cells of the tumor microenvironment. a Representative image of S100A8/A9 (red staining) and S100B (green staining) antibody co-labelled tissue sections of a metastasized primary melanoma using immunofluorescence. b Representative images of melanoma samples from the tissue microarrays (TMA) stained by immunohistochemistry with a specific S100A8/A9 antibody showing a high proportion of cells with strong S100A8/A9 expression, and showing weak S100A8/A9 staining intensity, respectively. Abbreviations: infl = inflammatory cells of the tumor microenvironment, tu = tumor

To assess whether the finding of high numbers of S100A8/A9 expressing cells in metastases would translate into elevated S100A8/A9 serum levels in patients with impaired survival, we measured serum S100A8/A9 concentration in two cohorts of stage III and stage IV patients and conducted survival analysis (Additional file 1: Table S2). Univariate survival analysis showed that patients with elevated S100A8/A9 > 5.5 mg/l presented with significantly impaired OS in both cohorts in univariate (Fig. 4) as well as in multivariate analysis (Table 1).Fig4Fig. 4

Elevated S100A8/A9 serum levels are associated with impaired overall survival in melanoma patients. Kaplan-Meier survival curves for overall survival stratified by normal (≤5.5 mg/l) or elevated (> 5.5 mg/l) S100A8/A9 serum levels. a Depicts the training set (n = 114), b the independent validation set (n = 240). Hazard ratios were calculated using univariate Cox regression. Numbers in brackets indicate 95% confidence intervals. P-values were calculated using two-sided Log-Rank test. Abbreviations: HR = hazard ratio, P = P-value

To determine the prognostic impact of S100A8/A9 in the setting of immune checkpoint inhibition with PD-1 antibodies, its serum levels were determined in two independent cohorts comprising 27 and 44 patients, respectively (Additional file 1: Table S5). Patients with high baseline S100A8/A9 > 5.5 mg/l showed significantly impaired survival compared to patients with low baseline S100A8/A9 in two independent cohorts of patients treated with pembrolizumab (cohort 1: HR 5.37 [1.44–20.08], P = .0051; cohort 2: HR 10.70 [3.52–32.55], P < .0001) (Fig. 5). Elevated S100A8/A9 also remained significant in multivariate analysis including LDH > 2.5x upper limit of normal (ULN) and AJCC M stage (Table 2).Fig5Fig. 5

Elevated baseline S100A8/A9 serum levels are associated with impaired survival in patients treated with pembrolizumab. Kaplan-Meier survival curves for overall survival and progression-free survival stratified by normal (≤5.5 mg/l) or elevated (> 5.5 mg/l) S100A8/A9 serum levels in patients treated with the PD-1 antibody pembrolizumab in (a) pembrolizumab set 1 (27 patients), and in (b) pembrolizumab set 2 (44 patients), respectively. Hazard ratios were calculated using univariate Cox regression. Numbers in brackets indicate 95% confidence intervals. P-values were calculated using two-sided Log-Rank test. Abbreviations: HR = hazard ratio, P = P-value