Tumor-infiltrating lymphocytes appear to be a predictor of survival in many cancers, including cutaneous melanoma. We applied automated multispectral imaging to determine whether density and distribution of T cells within primary cutaneous melanoma tissue correlate with survival of metastatic melanoma patients after dendritic cell (DC) vaccination. CD3+ T cell infiltration in primary tumors from 77 metastatic melanoma patients was quantified using the ratio of intratumoral versus peritumoral T-cell densities (I/P ratio). Patients with longer survival after DC vaccination had stronger T-cell infiltration than patients with shorter survival in a discovery cohort of 19 patients (P = 0.000026) and a validation cohort of 39 patients (P = 0.000016). I/P ratio was the strongest predictor of survival in a multivariate analysis including M substage and serum lactate dehydrogenase level. To evaluate I/P ratio as a predictive biomarker, we analyzed 19 chemotherapy-treated patients. Longer survival times of DC-vaccinated compared with chemotherapy-treated patients was observed for high (P = 0.000566), but not low (P = 0.154) I/P ratios. In conclusion, T-cell infiltration into primary melanoma is a strong predictor of survival after DC vaccination in metastatic melanoma patients who, on average, started this therapy several years after primary tumor resection. The infiltration remains predictive even after adjustment for late-stage prognostic markers. Our findings suggest that the I/P ratio is a potential predictive biomarker for treatment selection. Cancer Res; 76(12); 3496–506. ©2016 AACR.

The incidence and, to a lesser extent, the mortality of cutaneous melanoma have rapidly increased over the past decades in many countries (1). Until recently, metastatic melanoma patients have had a 5-year survival rate of less than 20% (2) and systemic therapy showed only minimal effect (3). With the currently approved immune checkpoint inhibitors against CTLA-4 and PD-1, 10% to 40% of all metastatic melanoma patients have long-term benefit and survival rates seem to improve (4). As checkpoint inhibitors are expensive, often induce toxicity, and the majority of patients have no long-term benefit, there is a strong need for biomarkers that can predict, ideally at an early stage of disease, whether a patient will or will not benefit from therapy.

Currently, in melanoma, only a few biomarkers that accurately predict response to treatment have been found, for example, BRAF mutation status for vemurafenib. In contrast, multiple prognostic factors of patient outcome at initial diagnosis are well established, including histologic characteristics such as Breslow thickness, mitotic rate, and ulceration status. In the advanced metastatic setting, tumor substage (M1a/b/c) and serum lactate dehydrogenase (LDH) levels have significant prognostic value (2). These prognostic markers only lead to a rough estimation of survival and do not predict response to treatment. However, more recently, it has been suggested that LDH might also have a predictive value; elevated serum LDH levels correlate with dramatically lower response rates to anti-CTLA-4 antibodies (5), but the response rate in patients with normal serum LDH levels remains limited.

With the lack of simple and early biomarkers to be found in the blood, the focus has shifted toward the tumor microenvironment (6). As many primary tumors are heavily infiltrated by a complex repertoire of immune cells, which is considered to be indicative of a spontaneous host immune response to tumor antigens (7), it is hypothesized that the T-cell landscape in primary tumors has prognostic value and might also be of predictive significance. Most studies so far have been performed in colorectal cancer (8) and other solid tumors (9, 10), and have mainly focused on distant metastatic lesions. These studies have shown that infiltration of the tumor by T cells, especially CD3+ and CD8+ T cells, is associated with good prognosis (8). In primary cutaneous melanoma Clark and colleagues were among the first to show the importance of tumor-infiltrating lymphocytes (TIL) as an independent predictor of patient survival (11). In recent years, several studies have confirmed this finding (12–14), while others found no significant correlation between TILs and clinical outcome (15–17). Adverse effects of TILs are also documented, for example, the presence of large numbers of regulatory T cells or myeloid-derived suppressor cells clearly correlated with poor survival after immunotherapy (18).

Two factors could contribute to these partly conflicting results. First, in most cases, immune cell infiltration was assessed manually and qualitatively using various scoring systems. To assess T-cell infiltration in an objective and reproducible manner, we exploited a new quantitative digital imaging technique to analyze the density and distribution of T cells within primary tumors of metastasized melanoma patients. Second, the T-cell infiltration could only be relevant for a subset of patients or certain treatments. This might explain that for dendritic cell (DC) vaccination, and other immunotherapies, only a subset of patients appears to benefit (19–23). We previously demonstrated that skin-infiltrating T cells isolated from delayed-type hypersensitivity (DTH) biopsies strongly correlate with increased survival after DC vaccination (24–26). Therefore, we expected that if strong T-cell infiltration in the primary tumor does indeed have a positive effect on survival, this should be augmented with immunotherapy and therefore be more easily detectable in these patients. To test this hypothesis, we analyzed T-cell infiltration in two cohorts of patients that were treated with DC-based immunotherapy and compared the results with those in patients treated with standard chemotherapy.

Patients and samples

We retrospectively analyzed the primary cutaneous tumors of 77 metastatic melanoma patients. An initial discovery cohort consisted of 19 patients who were enrolled in prospective DC vaccination studies at our center. Subjects were selected on the basis of the availability of the primary tumor at our center and known BRAF/NRAS mutation status. Furthermore, we selected patients with an overall survival of less than 1 year (n = 11, short survivors) or more than 1.5 years (n = 8, long survivors).

The validation cohort included 39 additional patients who received DC vaccinations, of which the primary tumor was available at our center, independent of survival time and mutation status. From one patient of whom two primary melanomas were available, we included the primary tumor with the highest Breslow thickness in our analysis (both nonulcerated). Three patients had an initial incision biopsy of their primary melanoma, followed by a complete excision of the same tumor; for these patients, we included both specimens in our analysis.

In addition, we analyzed 19 patients receiving standard treatment, at that time comprising of chemotherapy (dacarbazine 850 to 1,000 mg/m2 i.v. at three weekly intervals), between April 2000 and April 2009. Subjects were selected on the basis of the availability of the primary tumor at our center. Patients that received prior or subsequent immunotherapy were excluded. Ethical approval was obtained from the appropriate Medical Ethical Review Board.

DC vaccination

All DC-vaccinated patients were enrolled in our DC vaccination studies between June 1999 and January 2011 and had metastatic disease upon inclusion. All studies were approved by the appropriate Medical Ethical Review Board and written informed consent was obtained from all patients. All patients were vaccinated with autologous DCs, obtained from leukapheresis, and loaded with tumor-associated antigens of gp100 and tyrosinase according to a schedule of three biweekly vaccinations with a maximum of three cycles. Variations in protocols included the type of DCs, route of administration, and method of antigen loading. For the exact details regarding the vaccination protocols, we refered to the following individual studies (20, 21, 27–31).

IHC

Slides of 4-μm thickness were cut from formalin-fixed, paraffin-embedded (FFPE) primary melanoma tissue blocks. The slides were deparaffinized, after which antigen retrieval using 10 mmol/L citrate buffer (pH 6.0; Skytek) for 10 minutes at 96°C was performed. After this pretreatment, the slides were placed in an Autostainer 480 (Thermo Scientific). In the stainer, the endogenous peroxidase was blocked using 3% hydrogen peroxidase in methanol (EMD Millipore) followed by primary antibody incubation (anti-CD3, Thermo Scientific, dilution: 1/40) for 60 minutes at room temperature. Next, incubation with Brightvision poly-HRP-anti Ms/Rb/Rt IgG (Immunologic BV; dilution: 1/2) was performed for 30 minutes at room temperature followed by a visualization step with the Vector Nova Red Substrate Kit (Vector Laboratories) for 7 minutes at room temperature. Between each step, samples were rinsed using PBS (Klinipath; dilution: 1/10). After visualization, the slides were manually counterstained with hematoxylin and enclosed with Quick-D mounting medium (Klinipath).

Tissue imaging and quantitative digital analysis

Whole tissue slides were imaged using Vectra Intelligent Slide Analysis System (Version 2.0.8, PerkinElmer Inc.). This imaging technology combines imaging and spectroscopy to collect entire spectra at every location of the image plane. Images of single stained tissues for each reagent and unstained tissue were used to build spectral libraries of the single dyes and of the melanin pigmentation, respectively, by using the Nuance Multispectral Imaging System (Version 3.0.2, PerkinElmer Inc.). The reagents used were hematoxylin for nuclear staining and nova red (Vector Laboratories) for CD3+ T cells. These spectral libraries were used to unmix the original multispectral images obtained with the Vectra imaging system, in particular, to separate the spectrum of melanin, which can interfere with the nova red chromogen signal, to obtain an accurate and specific quantification of the nova red–positive signal. A selection of 10 to 15 representative original multispectral images was used to train the inForm Advanced Image Analysis Software (Version 2.0.2, PerkinElmer Inc.) for quantitative image analysis (tissue segmentation, cell segmentation, and positivity score) as described previously (Supplementary Fig. S1; refs. 32, 33). All the settings applied to the training images were saved within an algorithm allowing batch analysis of multiple original multispectral images of the same tumor. A separate algorithm was generated per single patient applying the same settings for cell segmentation and positivity score but with different training for tissue segmentation due to the numerous morphologic differences between and within tumors, which is very common in melanoma. Vectra Review (Version 2.0.8, PerkinElmer Inc.) was used to select the areas for analysis; this consisted of the entire tumor parenchyma and the tumor margin (stroma surrounding the tumor with a thickness of 500–700 μm and stromal areas within the tumor not containing tumor cells). Thus, intratumoral CD3+ T cells were quantified in the tumor parenchyma and the peritumoral CD3+ T cells were quantified in the stroma, at the margin of the tumor.

To visualize the spatial distribution of stroma cells, melanoma cells, and T cells, each pixel representing a location less than 50 μm apart from the nearest cell nucleus according to the type of that cell was attributed with a certain color (dark, tumor; light, tumor margin). To appreciate the T-cell distribution patterns at low resolution, we clustered the T-cell positions using hierarchical mean linkage clustering, with a distance threshold of 150 μm (i.e., clusters whose centers were more than 150 μm apart were not joined).

Statistical analyses

Pairwise comparisons between groups were made using the Mann–Whitney U test. Correlations between intra/peritumoral CD3+ T-cell density ratios (I/P ratios) and survival times were measured using the linear (Pearson) correlation coefficient on log-transformed data. When dichotomizing I/P ratios, we used a cutoff of 1, that is, an I/P ratio was considered “high” if T-cell density was higher intratumorally than peritumorally and “low” otherwise. Overall survival was calculated from the date of first chemotherapy (chemotherapy cohort) or the date of leukapheresis (DC vaccination cohorts) to death of any cause. Time to M stage was calculated from the time of initial diagnosis to the time of reaching M stage. LDH levels were dichotomized in the analysis (≤/> the upper limit of normal) and gender was dummy-coded as male = 0, female = 1. Differences between Kaplan–Meier estimates of survival times were assessed using the log-rank test. AUC analyses used above-median survival time as the outcome variable, and multivariate AUC analysis was performed by fitting a logistic regression model and ranking the patients by their fitted score.

To compare survival times of chemotherapy-treated and DC-vaccinated patients, we matched each patient of the chemotherapy cohort to two patients from the DC vaccination cohorts who had the same LDH status (high/low) and the same status with respect to M substage 1c. These matching criteria were chosen based on the results of our multivariate analysis, which indicated that those were the most relevant prognostic criteria in our cohorts, whereas other common matching factors like gender and age appeared to play no role. Matching was performed using the R package “optmatch” (34). All statistical analyses were performed using the R platform for statistical computing (R Core Team, 2015; https://www.R-project.org).

Quantifying CD3+ T-cell infiltration in the primary tumor

To measure T-cell infiltration within primary tumors in a quantitative, automatic, and objective manner, whole-slide imaging covering the entire tumor cross-section was performed on three cohorts of patients (77 in total). Depending on the size of the primary tumor, 20 to 750 fields per patient were analyzed. Figure 1A shows typical examples of T-cell localization within the primary tumor of a patient where most T cells appear confined to the peritumoral area (margin; left), and a patient where T cells strongly infiltrate into the intratumoral area (tumor; right). A computer-generated overview (“landscape”) of this contrast can be seen for the entire tissue area analyzed in Fig. 1B, with a pronounced difference in T-cell density and location. As a simple metric to quantify the strength of T-cell infiltration, we computed the ratio of intratumoral over peritumoral CD3+ T-cell densities (I/P ratio), which is low in weakly infiltrated tumors and high in strongly infiltrated tumors. Landscapes of all primary tumors studied are shown in Supplementary Table S1. In the majority of tumors (45 of 77), we found either high (above-median; n = 21) or low (n = 24) T-cell densities in both the tumor and the margin, whereas 32 tumors showed contrasting T-cell densities, that is, low in the tumor and high in the margin (n = 16) or the other way round (n = 16). We studied the I/P ratio in primary melanoma tumors of metastatic melanoma patients with this new computational imaging system to evaluate a potential relationship with overall survival.

Figure 1.

Mapping out T-cell landscapes near primary tumors. A, typical example (×20) of CD3+ T-cell distribution in the primary tumor with weak (left) and strong intratumoral T-cell infiltration (right). B, tissue classification and T-cell localization in two of the analyzed slides. The I/P ratio is the result of (intratumoral T cells/mm2)/(peritumoral T cells/mm2) and therefore is a ratio of T-cell densities rather than T-cell counts. For illustrative purposes, individual T-cell positions less than 150 μm apart were clustered.

Figure 1.

Mapping out T-cell landscapes near primary tumors. A, typical example (×20) of CD3+ T-cell distribution in the primary tumor with weak (left) and strong intratumoral T-cell infiltration (right). B, tissue classification and T-cell localization in two of the analyzed slides. The I/P ratio is the result of (intratumoral T cells/mm2)/(peritumoral T cells/mm2) and therefore is a ratio of T-cell densities rather than T-cell counts. For illustrative purposes, individual T-cell positions less than 150 μm apart were clustered.

Close modal

Long survival after DC vaccination correlates with strong T-cell infiltration of the primary tumor

At first, we were interested to investigate patients that survived long upon therapy, and therefore selected a discovery cohort of 11 patients who survived <1 year (“short survivors”) and 8 patients who survived >1.5 years (“long survivors”) after DC-based immunotherapy. Baseline patient characteristics are shown in Table 1. On the basis of this discovery cohort, we aimed to determine whether strong T-cell infiltration is associated with long survival, and whether I/P ratio is an appropriate measure on which to base this prognosis.

Table 1.

Baseline characteristics in all cohorts

Discovery cohortValidation cohortChemotherapy
Number of patients 19 39 19 
Gender 
 Male 12 (63.2%) 25 (64.1%) 11 (57.9%) 
 Female 7 (36.8%) 14 (35.9%) 8 (42.1%) 
Age, years 
 Median (range) 57 (20–76) 51 (19–73) 57 (30–77) 
Breslow thickness, mm 
 Median (range) 3.3 (0.8–13.0) 2.1 (0.7–12.0) 2.4 (0.4–10.0) 
M stage at inclusion 
 M1a 5 (26.3%) 8 (20.5%)a 2 (10.5%) 
 M1b 4 (21.1%) 10 (25.6%) 3 (15.8%) 
 M1c 10 (52.6%) 21 (53.9%) 14 (73.7%) 
LDH at inclusion 
 Normal 13 (68.4%) 24 (60.5%) 11 (57.9%) 
 Elevated 6 (31.6%) 15 (39.5%) 8 (42.1%) 
Time to M stage, months 
 Mean (range) 31.0 (1.1–74.7) 39.5 (0.0–137.0) 48.6 (7.4–179.9) 
Overall survival, months 
 Mean (range) 22.4 (2.9–130.0) 12.5 (1.1–63.9) 6.6 (1.2–27.4) 
Discovery cohortValidation cohortChemotherapy
Number of patients 19 39 19 
Gender 
 Male 12 (63.2%) 25 (64.1%) 11 (57.9%) 
 Female 7 (36.8%) 14 (35.9%) 8 (42.1%) 
Age, years 
 Median (range) 57 (20–76) 51 (19–73) 57 (30–77) 
Breslow thickness, mm 
 Median (range) 3.3 (0.8–13.0) 2.1 (0.7–12.0) 2.4 (0.4–10.0) 
M stage at inclusion 
 M1a 5 (26.3%) 8 (20.5%)a 2 (10.5%) 
 M1b 4 (21.1%) 10 (25.6%) 3 (15.8%) 
 M1c 10 (52.6%) 21 (53.9%) 14 (73.7%) 
LDH at inclusion 
 Normal 13 (68.4%) 24 (60.5%) 11 (57.9%) 
 Elevated 6 (31.6%) 15 (39.5%) 8 (42.1%) 
Time to M stage, months 
 Mean (range) 31.0 (1.1–74.7) 39.5 (0.0–137.0) 48.6 (7.4–179.9) 
Overall survival, months 
 Mean (range) 22.4 (2.9–130.0) 12.5 (1.1–63.9) 6.6 (1.2–27.4) 

aIncludes one irresectable stage III melanoma patient.

Indeed, we found clear differences between short and long survivors when analyzing the T-cell distributions (Fig. 2A). The median I/P ratio was 0.16 in short survivors versus 4.48 in long survivors (P = 0.000026; Fig. 2B); there was a near-perfect concordance between high ratios (>1) and long survival. Importantly, there were no clear differences in tumor area and in margin area between short versus long survivors (Fig. 2C), which could have biased this analysis. Analyzing intratumoral and peritumoral T-cell densities separately (Fig. 2D) shows that high intratumoral T-cell densities appear beneficial, whereas high peritumoral densities appear detrimental. However, these individual differences are smaller than the differences in I/P ratio. Taken together, these data clearly show that I/P ratio is a diagnostically more relevant measure of T-cell infiltration than the individual T-cell counts and has a strong association with survival in this discovery cohort. Thus, we focused on I/P ratio as a measure of T-cell infiltration in our further analysis.

Figure 2.

Strong CD3+ T-cell infiltration of primary tumors tissue correlates with long survival after DC vaccination. A, relative tumor size, size of peritumoral area, and CD3+ count for short and long survivors in the discovery cohort. Overall survival is shown above each tumor in months. Each large bullet point represents 1,000 T cells within the tumor (white) or in the peritumoral area (black). B, differences in intra/peritumoral CD3+ T-cell density ratio (I/P ratio) between long survivors (n = 8) and short survivors (n = 11). C, size of the tumor area (tumor) and tumor margin (margin) in the analyzed tissue samples for short and long survivors. D, peritumoral (margin) and intratumoral (tumor) CD3+ T-cell densities (CD3+ T cells/mm2) are shown separately for short and long survivors.

Figure 2.

Strong CD3+ T-cell infiltration of primary tumors tissue correlates with long survival after DC vaccination. A, relative tumor size, size of peritumoral area, and CD3+ count for short and long survivors in the discovery cohort. Overall survival is shown above each tumor in months. Each large bullet point represents 1,000 T cells within the tumor (white) or in the peritumoral area (black). B, differences in intra/peritumoral CD3+ T-cell density ratio (I/P ratio) between long survivors (n = 8) and short survivors (n = 11). C, size of the tumor area (tumor) and tumor margin (margin) in the analyzed tissue samples for short and long survivors. D, peritumoral (margin) and intratumoral (tumor) CD3+ T-cell densities (CD3+ T cells/mm2) are shown separately for short and long survivors.

Close modal

Validation of T-cell infiltration as a survival predictor for metastatic melanoma patients who received DC vaccination

To validate our findings of the discovery cohort, we analyzed the primary tumors of a second and larger cohort of 39 metastatic melanoma patients that received DC-based immunotherapy (Table 1; Fig. 3A). Importantly, this cohort was not selected on the basis of patient survival. In line with the discovery cohort, there were no clear differences between the size of the tumor or the margin of short versus long survivors (data not shown). The differences in intratumoral and peritumoral T-cell densities between short and long survivors (Fig. 3B) are similar to those in the discovery cohort. Figure 3C shows a clear correlation (r = 0.58, P = 0.00012) between the I/P ratio and survival in the validation cohort. Stratification of this cohort by M substage at start of therapy (Fig. 3D and E) shows that T-cell infiltration is more strongly predictive for patients in substage M1a/M1b (r = 0.78, P = 0.00015) than for substage M1c (r = 0.31, P = 0.17). Survival was furthermore evaluated with Kaplan–Meier estimates comparing patients with low (<1, n = 33) and high (>1, n = 6) I/P ratios (Fig. 3F). The survival curves for these two groups were distinct, with strong T-cell infiltration (high I/P ratio) being linked to longer survival than weak T-cell infiltration (low I/P ratio; P = 0.0011). A ROC analysis (Fig. 3G) discriminating short survivors (bottom 50% of survival) and long survivors (top 50%) resulted in an AUC of 0.88. Thus, using three different statistical methods we consistently found an association between strong T-cell infiltration, as determined by the I/P ratio, and long survival after DC vaccination.

Figure 3.

Validation of the I/P ratio as a predictor of survival after DC vaccination. A, relative tumor size, size of peritumoral area, and CD3+ count of all patients in the validation cohort. Overall survival is shown above each tumor in months. Each large bullet point represents 1,000 T cells within the tumor (white) or in the peritumoral area (black). B, peritumoral (margin) and intratumoral (tumor) CD3+ T-cell densities (CD3+ T cells/mm2) shown separately for short and long survivors. C–E, correlation between I/P ratio and overall survival are shown on a log–log scale for all patients (C), patients with substage up to M1b (D), and patients with substage M1c (E). F, survival of patients with low (<1, n = 33) and high (>1, n = 6) I/P ratios (P = 0.0011, log-rank test). G, AUC analysis for discriminating short survivors (bottom 50% survival times) and long survivors (top 50%) using the I/P ratio (AUC of 0.88). H, comparison with two other relevant prognostic markers. A combined logistic regression model incorporating the ratio and LDH achieves an AUC of 0.95 (M substage alone: 0.68; LDH alone: 0.72), which is no longer improved by also taking the M substage into account.

Figure 3.

Validation of the I/P ratio as a predictor of survival after DC vaccination. A, relative tumor size, size of peritumoral area, and CD3+ count of all patients in the validation cohort. Overall survival is shown above each tumor in months. Each large bullet point represents 1,000 T cells within the tumor (white) or in the peritumoral area (black). B, peritumoral (margin) and intratumoral (tumor) CD3+ T-cell densities (CD3+ T cells/mm2) shown separately for short and long survivors. C–E, correlation between I/P ratio and overall survival are shown on a log–log scale for all patients (C), patients with substage up to M1b (D), and patients with substage M1c (E). F, survival of patients with low (<1, n = 33) and high (>1, n = 6) I/P ratios (P = 0.0011, log-rank test). G, AUC analysis for discriminating short survivors (bottom 50% survival times) and long survivors (top 50%) using the I/P ratio (AUC of 0.88). H, comparison with two other relevant prognostic markers. A combined logistic regression model incorporating the ratio and LDH achieves an AUC of 0.95 (M substage alone: 0.68; LDH alone: 0.72), which is no longer improved by also taking the M substage into account.

Close modal

T-cell infiltration remains predictive after combination with established late-stage prognostic markers

Given the long interval between excision of the primary melanoma and the development of distant metastasis and onset of treatment (mean time to M stage >3 years; Table 1), it was important to establish whether I/P ratio would still provide added value compared with the known prognostic markers in the metastatic setting. ROC analyses for predicting above-median survival (Fig. 3H) showed superior results of the I/P ratio (AUC = 0.88) compared with LDH (AUC = 0.72) and M substage (AUC = 0.68). Combining the I/P ratio with LDH in a logistic regression model, we even obtained an AUC value of 0.95. In a larger multivariate analysis (Table 2), the I/P ratio was the strongest predictor (P = 0.001; log OR, 0.4; 95% CI, 0.19–0.61) for survival in the validation cohort. LDH level remained an independent, but weaker predictor (P = 0.013; log OR, −0.37; 95% CI, −0.64 to −0.1), whereas M substage at start of DC-based immunotherapy, gender, Breslow thickness, age, and time to M stage did not show significant prognostic value. These multivariate analyses provide compelling evidence that strong T-cell infiltration in the primary melanoma is an independent predictor of survival in metastatic melanoma patients receiving DC-based immunotherapy, despite the fact that several years typically pass between resection of the primary tumor and the onset of treatment.

Table 2.

Multivariate analysis

Log OR (95% CI)P
Intercept 0.82 (0.05–1.6) 0.044 
I/P ratio (log100.4 (0.19–0.61) 0.001 
M stage at inclusion 0 (−0.18–0.19) 0.97 
LDH −0.37 (−0.64 to −0.1) 0.013 
Gender 0 (−0.3–0.31) 0.98 
Months to M stage 0 (0–0.01) 0.54 
Breslow thickness 0 (−0.06–0.06) 0.96 
Age at inclusion 0 (−0.01–0.01) 0.83 
Log OR (95% CI)P
Intercept 0.82 (0.05–1.6) 0.044 
I/P ratio (log100.4 (0.19–0.61) 0.001 
M stage at inclusion 0 (−0.18–0.19) 0.97 
LDH −0.37 (−0.64 to −0.1) 0.013 
Gender 0 (−0.3–0.31) 0.98 
Months to M stage 0 (0–0.01) 0.54 
Breslow thickness 0 (−0.06–0.06) 0.96 
Age at inclusion 0 (−0.01–0.01) 0.83 

NOTE: Logistic regression model showing that the difference in ratio in the validation cohort is not explained by including M substage, LDH, and other possible confounders as covariates.

T-cell infiltration is a potential predictive biomarker for DC vaccination

The I/P ratio in primary tumors would be tremendously useful as a predictive biomarker for (adjuvant) treatment selection due to its availability at an early stage of disease. Establishing a biomarker requires analysis of data from a randomized controlled trial, which is not available yet for DC vaccinations worldwide; so far, most DC studies are small stage I/II studies varying from 10 to 40 patients. Therefore, as an approximation, we compared our DC-vaccinated cohorts to a cohort of patients receiving standard treatment (chemotherapy) in the same time-frame, matched on known prognostic factors to minimize confounding. While this approach cannot provide definitive proof of biomarker status, it did allow us to probe our hypothesis that patients require strong intratumoral T-cell infiltration to benefit from immunotherapy.

We analyzed 19 primary melanomas of patients that had received chemotherapy (Table 1; Fig. 4A). Survival was evaluated with Kaplan–Meier estimates comparing patients with low (<1, n = 6) and high (>1, n = 13) I/P ratios (Fig. 4B). Although the survival curves suggest that strong T-cell infiltration may correspond to longer survival in chemotherapy-treated patients, the difference is not significant (P = 0.089). There was also no appreciable direct correlation between survival and I/P ratio (r = 0.12, P = 0.62; Fig. 4C), thus, the I/P ratio is not a strong predictor of survival in chemotherapy-treated patients. This is in line with our hypothesis that the effect of T-cell infiltration in the primary tumor needs to be augmented by immunotherapy to improve survival. To further support this hypothesis, we matched each chemotherapy patient to two DC vaccination patients based on LDH level and M substage (Fig. 4D). DC-vaccinated patients with low I/P ratios showed no significant difference in survival after the different treatments (Fig. 4E), indicating no survival benefit of DC-based immunotherapy when T-cell infiltration is weak. On the other hand, patients with high I/P ratios clearly survived longer after DC-based immunotherapy than patients with high I/P ratios treated with chemotherapy (Fig. 4F), and therefore seem to benefit from DC-based immunotherapy.

Figure 4.

Strong intratumoral T-cell infiltration is associated with a larger difference between DC vaccination and chemotherapy in a matched cohort. A, relative tumor size, size of peritumoral area, and CD3+ count of all patients in the chemotherapy cohort. Overall survival is shown above each tumor in months. Each large bullet point represents 1,000 T cells within the tumor (white) or in the peritumoral area (black). B, survival of patients treated with chemotherapy with low (<1, n = 6) and high (>1, n = 13) I/P ratios (P = 0.089). C, correlation between I/P ratio and survival in months are shown on a log–log scale for all patients. D, analysis of a combined cohort in which each of the 19 chemotherapy patients was matched to two patients from the DC vaccination cohorts (giving n = 57) that were matched on LDH and M1c substage. E, longer survival upon immunotherapy is not apparent in patients with low I/P ratio (<1; n = 36). F, yet, there is a significant difference for patients with high I/P ratio (>1; n = 21).

Figure 4.

Strong intratumoral T-cell infiltration is associated with a larger difference between DC vaccination and chemotherapy in a matched cohort. A, relative tumor size, size of peritumoral area, and CD3+ count of all patients in the chemotherapy cohort. Overall survival is shown above each tumor in months. Each large bullet point represents 1,000 T cells within the tumor (white) or in the peritumoral area (black). B, survival of patients treated with chemotherapy with low (<1, n = 6) and high (>1, n = 13) I/P ratios (P = 0.089). C, correlation between I/P ratio and survival in months are shown on a log–log scale for all patients. D, analysis of a combined cohort in which each of the 19 chemotherapy patients was matched to two patients from the DC vaccination cohorts (giving n = 57) that were matched on LDH and M1c substage. E, longer survival upon immunotherapy is not apparent in patients with low I/P ratio (<1; n = 36). F, yet, there is a significant difference for patients with high I/P ratio (>1; n = 21).

Close modal

In summary, our matched analysis indicates that the I/P ratio could indeed be a predictive marker for DC-based immunotherapy: patients with high I/P ratios appear to benefit from receiving DC-based immunotherapy, whereas patients with low I/P ratios do not. It will be important to test this result in future randomized trials with DC vaccination and other immunotherapies.

The rapidly evolving field of melanoma management and novel treatments mandates development of new biomarkers, especially given the high costs associated with current treatments and their substantial toxicity. While for targeted therapies, effective genetic tools are available to select patients who are most likely to benefit from treatment, that is, BRAF mutation status for vemurafenib, this is much less so for immunotherapy. Established factors that are known to correlate with survival after immunotherapy are mostly derived from metastases. When looking at immunotherapy in general, several studies have shown specific gene expression profiles, identified on pretreatment biopsies of melanoma metastases that correlate with outcome after immunotherapy (35). Besides gene profiling, multiple studies have focused on finding a predictive biomarker in the tumor microenvironment. For instance, the expression of FoxP3 and indoleamine 2,3-dioxygenase (IDO) in pretreatment biopsies of melanoma metastases have been linked to a positive association with clinical activity of anti-CTLA-4 antibody (36). Also, several studies investigated biomarkers for PD-1 pathway–blocking agents, in particular, the expression of the PD-1 ligand (PD-L1). Unfortunately, the many different PD-L1 clones and differences in assessment of the PD-L1 immunohistochemical staining complicate interpretation of the different studies (37). Nonetheless, in melanoma, there seems to be an association between PD-L1 expression on tumor cells and response to an anti-PD1-antibody (38). A recent study has focused both on the role of PD-L1 expression as well as CD8+ T cells in melanoma metastases and has shown that the density of pre-existing CD8+ T cells in both the tumoral area and the invasive tumor margin has a predictive value for the treatment outcome of patients receiving anti-PD-1 therapy (39). This study highlights the importance of pretreatment T-cell infiltration in the tumor area as it correlates with clinical outcome.

A disadvantage of most of these potential biomarkers is that tumor material from metastases can be difficult to access and only becomes available when patients reach systemic disease, and are therefore not suitable for decision-making at earlier stages of disease in which adjuvant immunotherapy might be beneficial. Furthermore, biomarkers observed in metastases might not be suitable for analysis of the primary tumor; for instance, the positive association between PD-L1 expression in metastases and melanoma-specific survival could not be confirmed in primary melanoma (40). Finally, markers like skin-infiltrating T cells isolated from DTH biopsies, although useful for monitoring response to therapy, cannot be used for pretherapeutic decision-making.

In this study, we investigated whether the T-cell landscape within the primary tumor is relevant for survival of 77 patients with metastatic melanoma. Given partly contradicting results in the literature, our approach aimed to exploit a novel observer-independent methodology, and focused on patients who received immunotherapy as we expected the relevance of immune infiltration to be higher in this setting. Our results show that strong CD3+ T-cell infiltration into the primary tumor is an excellent early predictor of longer survival in metastatic melanoma patients receiving DC-based immunotherapy. This was revealed in a discovery cohort and confirmed in a validation cohort, which showed a clear correlation between intratumoral CD3+ T-cell infiltration and survival. Multivariate analyses showed that the I/P ratio improves upon and can be combined with LDH levels, measured at the time treatment decisions must be made. This led to the most accurate survival prediction in our validation cohort (AUC = 0.95), which also outperformed the conventional classification based on the TNM staging system.

T-cell recognition of tumor cells relies upon presentation of tumor-specific antigens on MHC molecules. Therefore, a potential mechanism explaining the relevance of T-cell infiltration might be that weak T-cell infiltration into the primary tumor reflects poor immunogenicity of the tumor and thus also of potential future metastases. Weak immunogenic tumors might also be more difficult to treat with immunotherapy. If so, then intratumoral T-cell infiltration should also be an appropriate marker for selecting patients for treatment with immunotherapy. Our analysis of the matched cohort provides support for this hypothesis. When data from randomized trials with DC vaccination become available, it will be possible to confirm this predictive biomarker status.

Consistently, with earlier studies, we found high intratumoral T-cell densities to be beneficial. However, in the cohorts studied here, a high T-cell density in the margin appears to be detrimental. In theory, a high marginal T-cell density could reflect an efficient homing rate of T cells from the circulation towards the tumor site. But it could also reflect a poor infiltration rate of T cells from the margin into the tumor. By measuring the I/P ratio, we focus on the infiltration rate, which predicts survival better than either of the individual densities. This might indicate that the infiltration rate is of particular importance in primary tumors, which may not be the case in metastases.

A major advantage of analyses of T-cell infiltration in primary tumors is that it can be assessed at initial diagnosis, which could select patients at high risk for metastatic disease and a high I/P ratio for adjuvant immunotherapy. This is indeed of interest, as we have recently shown that DC vaccination more frequently induced functional tumor-specific immune responses in stage III melanoma patients, compared with patients with metastatic disease, in which their presence correlated with clinical outcome (25, 41). Also, other studies have shown a benefit in terms of recurrence-free survival of treatment with the anti-CTLA-4 antibody in the adjuvant setting (42). Anti-CTLA-4 has recently been approved by the FDA for stage III melanoma patients; yet, it is still questionable whether anti-CTLA-4 is to be recommended as an adjuvant treatment because of its high toxicity, unknown late toxicities, and awaited data on overall survival. Identifying the right patients for adjuvant treatment (i.e., high risk patients who do not develop metastasis due to adjuvant treatment) would optimize outcome, lower toxicity, and reduce costs. However, as we have not analyzed T-cell landscapes in stage III melanoma patients who did not develop distant metastases, further research is needed to determine the value of T-cell infiltration at this stage of disease.

Often, the immune system is not capable of efficiently controlling tumor growth despite proper infiltration of immune cells in the tumor. This inefficiency may be caused by several mechanisms, including decreased function of tumor antigen-specific T cells by immunosuppressive strategies caused by tumor cells. Therefore, we believe it will be important to reveal the phenotype and function of the CD3+ T cells. On the basis of previous studies examining the tumor microenvironment (8, 12, 43), we presume that the majority of the intratumoral cells are CD8+ T cells, as cytotoxic T cells are considered to be the most important effector cells in the tumor microenvironment. However, given that T cells in the margin appear detrimental, the peritumoral T cells could have a distinct, perhaps regulatory, phenotype. Further research should focus on unraveling the phenotype of the CD3+ T cells and the underlying mechanisms that give rise to the strong predictive value of the I/P ratio.

So far, most groups studying the correlation between TILs and patient survival classify the presence of TILs as absent, non-brisk and brisk (11, 15, 16). Some characterize TILs only as present or absent (44), while few use a more elaborate method that takes both the distribution and the density of TILs into account (12). Our method also takes distribution and density into account and does this in an automated and quantitative manner using the entire tumor slide. This way we are able to quantify all CD3+ T cells and assess their exact location. Computing I/P ratio requires sufficient availability of both peritumoral and intratumoral tissue. In this study, we aimed for including peritumoral tissue within 500–700 μm around the tumor, which was possible in most cases. Thus, we believe that this method is practically applicable, provides highly accurate predictive information, and may lead to more reproducible results than manual qualitative assessment.

In summary, we have shown that T-cell infiltration into the tumor is a highly accurate predictor of survival in metastatic melanoma patients receiving DC vaccination: on average, patients with a high I/P ratio appear to benefit from DC-based immunotherapy, whereas patients with a low I/P ratio do not. The predictive power of the I/P ratio does not wane when taking into account established late-stage prognostic markers, which is remarkable given that DC vaccination was only administered at metastatic stage, years after resection of the primary tumor. This finding underscores the importance of the tumor microenvironment throughout the entire course of disease. From a clinical perspective, assessing the T-cell landscape at initial melanoma diagnosis is attractive for two reasons: primary tumors are more easily accessible than metastases, and the marker is available early on for diagnostic decision-making and, potentially, (adjuvant) treatment selection. As all immunotherapies are based on the same principle, this tool may be generally applicable for all types of immunotherapy. Further studies are warranted to confirm that strong intratumoral T-cell infiltration is indeed a predictive biomarker and to investigate the prognostic and predictive potentials of the T-cell landscape for other types of immunotherapy in melanoma. Finally, this approach may be useful for the investigation of other tumor types, especially for other immunogenic tumors.

No potential conflicts of interest were disclosed.

Conception and design: A. Vasaturo, K.F. Bol, C.J.A. Punt, J.H.J.M. van Krieken, J. Textor, I.J.M. de Vries, C.G. Figdor

Development of methodology: A. Vasaturo, C.J.A. Punt, J.H.J.M. van Krieken, J. Textor, I.J.M. de Vries, C.G. Figdor

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. Vasaturo, A. Halilovic, K.F. Bol, D.I. Verweij, P.J.T.A. Groenen, I.J.M. de Vries

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Vasaturo, A. Halilovic, K.F. Bol, D.I. Verweij, W.A.M. Blokx, C.J.A. Punt, P.J.T.A. Groenen, J.H.J.M. van Krieken, J. Textor, I.J.M. de Vries, C.G. Figdor

Writing, review, and/or revision of the manuscript: A. Vasaturo, A. Halilovic, K.F. Bol, D.I. Verweij, W.A.M. Blokx, C.J.A. Punt, J.H.J.M. van Krieken, J. Textor, I.J.M. de Vries, C.G. Figdor

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. Vasaturo, A. Halilovic, K.F. Bol, D.I. Verweij, P.J.T.A. Groenen

Study supervision: C.J.A. Punt, I.J.M. de Vries, C.G. Figdor

The authors thank the technicians involved in the DC vaccination studies: Nicole Scharenborg, Annemiek de Boer, Mandy van de Rakt, Michel Olde Nordkamp, Jeanette Pots, Tom van Oorschot, and Tjitske Duiveman-de Boer and the technicians of the IHC Laboratory, under the supervision of Monique Link. The authors also thank Peter van Zwam, Jeroen van de Laak, and Harm Westdorp for their efforts in the initial phase of this research.

This work was supported by KWO grant KUN2009-4402 from the Dutch Cancer Society. I.J.M. de Vries received a NWO Vici grant (918.14.655). C.G. Figdor received the NWO Spinoza award and ERC Advanced grant PATHFINDER (269019). J. Textor was supported by a NWO Earth and Life Sciences grant (823.02.014).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Erdmann
F
,
Lortet-Tieulent
J
,
Schuz
J
,
Zeeb
H
,
Greinert
R
,
Breitbart
EW
, et al
International trends in the incidence of malignant melanoma 1953-2008–are recent generations at higher or lower risk?
Int J Cancer
2013
;
132
:
385
400
.
2.
Balch
CM
,
Gershenwald
JE
,
Soong
SJ
,
Thompson
JF
,
Atkins
MB
,
Byrd
DR
, et al
Final version of 2009 AJCC melanoma staging and classification
.
J Clin Oncol
2009
;
27
:
6199
206
.
3.
Palathinkal
DM
,
Sharma
TR
,
Koon
HB
,
Bordeaux
JS
. 
Current systemic therapies for melanoma
.
Dermatol Surg
2014
;
40
:
948
63
.
4.
Sharma
P
,
Allison
JP
. 
Immune checkpoint targeting in cancer therapy: toward combination strategies with curative potential
.
Cell
2015
;
161
:
205
14
.
5.
Kelderman
S
,
Heemskerk
B
,
van Tinteren
H
,
van den Brom
RR
,
Hospers
GA
,
van den Eertwegh
AJ
, et al
Lactate dehydrogenase as a selection criterion for ipilimumab treatment in metastatic melanoma
.
Cancer Immunol Immunother
2014
;
63
:
449
58
.
6.
Fridman
WH
,
Pages
F
,
Sautes-Fridman
C
,
Galon
J
. 
The immune contexture in human tumours: impact on clinical outcome
.
Nat Rev Cancer
2012
;
12
:
298
306
.
7.
Schreiber
RD
,
Old
LJ
,
Smyth
MJ
. 
Cancer immunoediting: integrating immunity's roles in cancer suppression and promotion
.
Science
2011
;
331
:
1565
70
.
8.
Galon
J
,
Costes
A
,
Sanchez-Cabo
F
,
Kirilovsky
A
,
Mlecnik
B
,
Lagorce-Pages
C
, et al
Type, density, and location of immune cells within human colorectal tumors predict clinical outcome
.
Science
2006
;
313
:
1960
4
.
9.
Zhang
L
,
Conejo-Garcia
JR
,
Katsaros
D
,
Gimotty
PA
,
Massobrio
M
,
Regnani
G
, et al
Intratumoral T cells, recurrence, and survival in epithelial ovarian cancer
.
N Engl J Med
2003
;
348
:
203
13
.
10.
Liu
H
,
Zhang
T
,
Ye
J
,
Li
H
,
Huang
J
,
Li
X
, et al
Tumor-infiltrating lymphocytes predict response to chemotherapy in patients with advance non-small cell lung cancer
.
Cancer Immunol Immunother
2012
;
61
:
1849
56
.
11.
Clark
WH
 Jr
,
Elder
DE
,
Guerry
Dt
,
Braitman
LE
,
Trock
BJ
,
Schultz
D
, et al
Model predicting survival in stage I melanoma based on tumor progression
.
J Natl Cancer Inst
1989
;
81
:
1893
904
.
12.
Azimi
F
,
Scolyer
RA
,
Rumcheva
P
,
Moncrieff
M
,
Murali
R
,
McCarthy
SW
, et al
Tumor-infiltrating lymphocyte grade is an independent predictor of sentinel lymph node status and survival in patients with cutaneous melanoma
.
J Clin Oncol
2012
;
30
:
2678
83
.
13.
Clemente
CG
,
Mihm
MC
 Jr
,
Bufalino
R
,
Zurrida
S
,
Collini
P
,
Cascinelli
N
. 
Prognostic value of tumor infiltrating lymphocytes in the vertical growth phase of primary cutaneous melanoma
.
Cancer
1996
;
77
:
1303
10
.
14.
Bogunovic
D
,
O'Neill
DW
,
Belitskaya-Levy
I
,
Vacic
V
,
Yu
YL
,
Adams
S
, et al
Immune profile and mitotic index of metastatic melanoma lesions enhance clinical staging in predicting patient survival
.
Proc Natl Acad Sci U S A
2009
;
106
:
20429
34
.
15.
Taylor
RC
,
Patel
A
,
Panageas
KS
,
Busam
KJ
,
Brady
MS
. 
Tumor-infiltrating lymphocytes predict sentinel lymph node positivity in patients with cutaneous melanoma
.
J Clin Oncol
2007
;
25
:
869
75
.
16.
Mandala
M
,
Imberti
GL
,
Piazzalunga
D
,
Belfiglio
M
,
Labianca
R
,
Barberis
M
, et al
Clinical and histopathological risk factors to predict sentinel lymph node positivity, disease-free and overall survival in clinical stages I-II AJCC skin melanoma: outcome analysis from a single-institution prospectively collected database
.
Eur J Cancer
2009
;
45
:
2537
45
.
17.
Burton
AL
,
Roach
BA
,
Mays
MP
,
Chen
AF
,
Ginter
BA
,
Vierling
AM
, et al
Prognostic significance of tumor infiltrating lymphocytes in melanoma
.
Am Surg
2011
;
77
:
188
92
.
18.
Karagiannis
P
,
Fittall
M
,
Karagiannis
SN
. 
Evaluating biomarkers in melanoma
.
Front Oncol
2014
;
4
:
383
.
19.
Lesterhuis
WJ
,
Aarntzen
EH
,
De Vries
IJ
,
Schuurhuis
DH
,
Figdor
CG
,
Adema
GJ
, et al
Dendritic cell vaccines in melanoma: from promise to proof?
Crit Rev Oncol Hematol
2008
;
66
:
118
34
.
20.
Jacobs
JF
,
Punt
CJ
,
Lesterhuis
WJ
,
Sutmuller
RP
,
Brouwer
HM
,
Scharenborg
NM
, et al
Dendritic cell vaccination in combination with anti-CD25 monoclonal antibody treatment: a phase I/II study in metastatic melanoma patients
.
Clin Cancer Res
2010
;
16
:
5067
78
.
21.
Tel
J
,
Aarntzen
EH
,
Baba
T
,
Schreibelt
G
,
Schulte
BM
,
Benitez-Ribas
D
, et al
Natural human plasmacytoid dendritic cells induce antigen-specific T-cell responses in melanoma patients
.
Cancer Res
2013
;
73
:
1063
75
.
22.
Wilgenhof
S
,
Van Nuffel
AM
,
Benteyn
D
,
Corthals
J
,
Aerts
C
,
Heirman
C
, et al
A phase IB study on intravenous synthetic mRNA electroporated dendritic cell immunotherapy in pretreated advanced melanoma patients
.
Ann Oncol
2013
;
24
:
2686
93
.
23.
Anguille
S
,
Smits
EL
,
Lion
E
,
van Tendeloo
VF
,
Berneman
ZN
. 
Clinical use of dendritic cells for cancer therapy
.
Lancet Oncol
2014
;
15
:
e257
67
.
24.
de Vries
IJ
,
Bernsen
MR
,
Lesterhuis
WJ
,
Scharenborg
NM
,
Strijk
SP
,
Gerritsen
MJ
, et al
Immunomonitoring tumor-specific T cells in delayed-type hypersensitivity skin biopsies after dendritic cell vaccination correlates with clinical outcome
.
J Clin Oncol
2005
;
23
:
5779
87
.
25.
Aarntzen
EH
,
Bol
K
,
Schreibelt
G
,
Jacobs
JF
,
Lesterhuis
WJ
,
van Rossum
MM
, et al
Skin-test infiltrating lymphocytes early predict clinical outcome of dendritic cell based vaccination in metastatic melanoma
.
Cancer Res
2012
;
72
:
6102
10
.
26.
Wimmers
F
,
Aarntzen
EH
,
Schreibelt
G
,
Jacobs
JF
,
Ja Punt
C
,
Figdor
CG
, et al
Early predictive value of multifunctional skin-infiltrating lymphocytes in anticancer immunotherapy
.
Oncoimmunology
2014
;
3
:
e27219
.
27.
Lesterhuis
WJ
,
de Vries
IJ
,
Schreibelt
G
,
Lambeck
AJ
,
Aarntzen
EH
,
Jacobs
JF
, et al
Route of administration modulates the induction of dendritic cell vaccine-induced antigen-specific T cells in advanced melanoma patients
.
Clin Cancer Res
2011
;
17
:
5725
35
.
28.
Lesterhuis
WJ
,
Schreibelt
G
,
Scharenborg
NM
,
Brouwer
HM
,
Gerritsen
MJ
,
Croockewit
S
, et al
Wild-type and modified gp100 peptide-pulsed dendritic cell vaccination of advanced melanoma patients can lead to long-term clinical responses independent of the peptide used
.
Cancer Immunol Immunother
2011
;
60
:
249
60
.
29.
Aarntzen
EH
,
Schreibelt
G
,
Bol
K
,
Lesterhuis
WJ
,
Croockewit
AJ
,
de Wilt
JH
, et al
Vaccination with mRNA-electroporated dendritic cells induces robust tumor antigen-specific CD4+ and CD8+ T cells responses in stage III and IV melanoma patients
.
Clin Cancer Res
2012
;
18
:
5460
70
.
30.
Aarntzen
EH
,
De Vries
IJ
,
Lesterhuis
WJ
,
Schuurhuis
D
,
Jacobs
JF
,
Bol
K
, et al
Targeting CD4(+) T-helper cells improves the induction of antitumor responses in dendritic cell-based vaccination
.
Cancer Res
2013
;
73
:
19
29
.
31.
Bol
KF
,
Figdor
CG
,
Aarntzen
EH
,
Welzen
ME
,
van Rossum
MM
,
Blokx
WA
, et al
Intranodal vaccination with mRNA-optimized dendritic cells in metastatic melanoma patients
.
Oncoimmunology
2015
;
4
:
e1019197
.
32.
Mansfield
JR
. 
Cellular context in epigenetics: quantitative multicolor imaging and automated per-cell analysis of miRNAs and their putative targets
.
Methods
2010
;
52
:
271
80
.
33.
Stack
EC
,
Wang
C
,
Roman
KA
,
Hoyt
CC
. 
Multiplexed immunohistochemistry, imaging, and quantitation: a review, with an assessment of Tyramide signal amplification, multispectral imaging and multiplex analysis
.
Methods
2014
;
70
:
46
58
.
34.
Hansen
BB
,
Klopfer
SO
. 
Optimal full matching and related designs via network flows
.
J Comput Graph Stat
2006
;
15
:
609
27
.
35.
Gajewski
TF
,
Louahed
J
,
Brichard
VG
. 
Gene signature in melanoma associated with clinical activity: a potential clue to unlock cancer immunotherapy
.
Cancer J
2010
;
16
:
399
403
.
36.
Hamid
O
,
Schmidt
H
,
Nissan
A
,
Ridolfi
L
,
Aamdal
S
,
Hansson
J
, et al
A prospective phase II trial exploring the association between tumor microenvironment biomarkers and clinical activity of ipilimumab in advanced melanoma
.
J Transl Med
2011
;
9
:
204
.
37.
Mahoney
KM
,
Atkins
MB
. 
Prognostic and predictive markers for the new immunotherapies
.
Oncology
2014
;
28
Suppl 3
:
39
48
.
38.
Taube
JM
,
Klein
A
,
Brahmer
JR
,
Xu
H
,
Pan
X
,
Kim
JH
, et al
Association of PD-1, PD-1 ligands, and other features of the tumor immune microenvironment with response to anti-PD-1 therapy
.
Clin Cancer Res
2014
;
20
:
5064
74
.
39.
Tumeh
PC
,
Harview
CL
,
Yearley
JH
,
Shintaku
IP
,
Taylor
EJ
,
Robert
L
, et al
PD-1 blockade induces responses by inhibiting adaptive immune resistance
.
Nature
2014
;
515
:
568
71
.
40.
Madore
J
,
Vilain
RE
,
Menzies
AM
,
Kakavand
H
,
Wilmott
JS
,
Hyman
J
, et al
PD-L1 expression in melanoma shows marked heterogeneity within and between patients: implications for anti-PD-1/PD-L1 clinical trials
.
Pigment Cell Melanoma Res
2015
;
28
:
245
53
.
41.
Bol
KF
,
Aarntzen
EH
,
Hout
FE
,
Schreibelt
G
,
Creemers
JH
,
Lesterhuis
WJ
, et al
Favorable overall survival in stage III melanoma patients after adjuvant dendritic cell vaccination
.
Oncoimmunology
2015
;
5
:
e1057673
.
42.
Eggermont
AM
,
Chiarion-Sileni
V
,
Grob
JJ
,
Dummer
R
,
Wolchok
JD
,
Schmidt
H
, et al
Adjuvant ipilimumab versus placebo after complete resection of high-risk stage III melanoma (EORTC 18071): a randomised, double-blind, phase 3 trial
.
Lancet Oncol
2015
;
16
:
522
30
.
43.
Tjin
EP
,
Krebbers
G
,
Meijlink
KJ
,
van de Kasteele
W
,
Rosenberg
EH
,
Sanders
J
, et al
Immune-escape markers in relation to clinical outcome of advanced melanoma patients following immunotherapy
.
Cancer Immunol Res
2014
;
2
:
538
46
.
44.
de Moll
EH
,
Fu
Y
,
Qian
Y
,
Perkins
SH
,
Wieder
S
,
Gnjatic
S
, et al
Immune biomarkers are more accurate in prediction of survival in ulcerated than in non-ulcerated primary melanomas
.
Cancer Immunol Immunother
2015
;
64
:
1193
203
.