Introduction

Immunotherapeutic modalities of cancer treatment have been increasingly gaining ground in the past few years. Understanding the mechanisms regulating antitumor immune response led to the development of a new class of immunotherapeutic agents targeting molecular interactions blocking T cell activation, the so called immune checkpoint inhibitors [3]. The first such agent, ipilimumab, which blocks CTLA-4 (cytotoxic T lymphocyte-associated antigen 4) was added to the therapeutic arsenal of advanced melanoma in 2011 [4, 5]. It also paved the way for agents targeting other immune regulatory pathways, of which antibodies blocking PD-1 or its ligand, PD-L1, represent the most promising treatment modality in a widening spectrum of tumor types. In a significant part of responding patients, immune checkpoint agents induce durable remission, showing unprecedented clinical efficacy in this patient population with advanced stage cancers.

Nevertheless, generally only a smaller proportion of patients benefit from immune stimulating antibody therapy. In the case of ipilimumab, long-term survival was observed in approximately 20% of patients [6]. On the other hand, serious side effects, often immune-related, were reported to occur in a higher percentage of patients [7]. To improve the benefit/risk ratio of individual patients, it is of primary importance to search for biomarkers that could predict the likelihood of therapeutic effect.

Although no validated predictive biomarkers are available yet for routine clinical use of ipilimumab, several candidates have been suggested, mainly concerning peripheral blood immune cells or serum factors. The most consistent finding was the observation of association between baseline absolute lymphocyte count, or its rise during treatment, and response to therapy or survival [8,9,10,11]. In some studies, correlation of high absolute eosinophil counts, low neutrophil/lymphocyte ratio or absolute monocyte counts, as well as low number or therapy-induced decrease of circulating myeloid-derived suppressor cells, with patient survival or response to ipilimumab treatment was described [8, 10,11,12]. Correlations of baseline number or therapy-induced change in T cells expressing specific, proliferation- or activation-associated markers (e.g., Ki67, and ICOS) with outcome have also been observed [8, 10]. Moreover, the presence of antibodies to NY-ESO-1 (New York esophageal squamous cell carcinoma-1) was found to correlate with clinical benefit of ipilimumab treatment, and patients with simultaneous presence of anti-NY-ESO-1 CD8+ (but not CD4+) T cell response experienced a clinical response and a survival advantage more frequently [13]. Among serum factors, high levels of lactate dehydrogenase (LDH), inflammatory markers as C-reactive protein (CRP), or vascular endothelial factor (VEGF) were found associated with poor outcome in ipilimumab-treated patients [8, 9].

The role of tumor-infiltrating immune cells in predicting the efficacy of CTLA-4 inhibitors was less intensely studied. In melanoma patients, ipilimumab therapy enhanced the expression of many immune-related genes in metastatic tumor samples, and most genes showing higher pretreatment expression in responders were immune-related [14]. In the same trial, clinical activity correlated with baseline staining for FOXP3 and indoleamine-2,3-dioxygenase (IDO), as well as with an increase in TIL during treatment, while no association was observed for pretreatment TIL or other immune cell subsets [15]. Enhanced tumor infiltration by CD4+, CD8+, HLA-DR+/CD45RO+, and FOXP3+ cells was reported in melanoma patients receiving tremelimumab, but no correlation with the therapeutic effect was observed [16].

A novel and intriguing field in the study of predictive biomarkers of immune checkpoint inhibitors is the analysis of somatic mutations and the resulting neoantigens in tumors [17]. Mutational burden was associated with long-term clinical benefit in ipilimumab-treated melanoma patients. Moreover, a neoepitope signature was identified that predicted the response to CTLA-4 blockade [18].

Most of the candidate biomarkers listed above were investigated only in one or a few studies, and none of them have been validated and can be used in routine clinical practice. The objective of this study was to explore tumor-infiltrating immune cell types as potential biomarkers predicting response to treatment and survival in melanoma patients receiving ipilimumab therapy.

Materials and methods

Patients and samples

Archived paraffin blocks of surgical tissue samples were collected from patients with metastatic melanoma who received ipilimumab treatment from 2010 to 2014 at four centers in Hungary. In the study, we included only cases with available tumor samples excised within 1 year before ipilimumab therapy and the study population consisted of 30 patients (1–25 lesions per patient). 86 samples were selected for the investigations: 52 lymph node metastases, and 34 subcutaneous/cutaneous metastases. The primary site was cutaneous in 28 cases and unknown in 2 cases. Most patients (n = 18) received ipilimumab treatment in the Expanded Access Program and 3 patients after drug commercialization (3 mg/kg four times every 3 weeks), while 9 patients participated in the CA-184-169 trial (receiving 3 or 10 mg/kg ipilimumab doses). Most patients (n = 26) had at least one prior systemic treatment; all of them received chemotherapy, while eight also had radiotherapy, one received BRAF inhibitor, and two patients were treated with electrochemotherapy. Twenty-six patients received all four doses while the remaining four patients received three cycles of ipilimumab treatment (three because of progression and one because of adverse events). Response assessment was based on immune-related response criteria (irRC) [19]. For our analysis, patients were considered “responders” if the best overall response was complete or partial response, or stable disease lasting for at least 6 months. 13 of the 30 patients belonged to this group, including 3 patients showing complete response lasting more than 24 months (27, 49, 57 + months). Patient and sample characteristics are shown in Table 1.

Table 1 Patient and sample characteristics
Full size table

Immunohistochemical detection of immune cell types

Three-micrometer sections from formalin-fixed, paraffin-embedded tumor samples were used in the study. Immunohistochemistry was performed as described earlier [20,21,22], using monoclonal antibodies against CD8, CD20cy, CD45RO, CD68 (Dako, Glostrup, Denmark), CD16, CD137 (Santa Cruz Biotechnology, Dallas, TX), CD134 (BD Biosciences Eastern Europe, Heidelberg, Germany), FOXP3 (236A/E7; eBioscience, San Diego, CA), NKp46 (R&D Systems, Abingdon, UK), CD4, and PD-1 (Cell Marque, Rocklin, CA). For detecting staining, we used the SS™ One-Step Polymer-HRP IHC Detection System (BioGenex, Fremont, CA), 3-amino-9-ethylcarbazole (AEC; Vector Laboratories, Inc., Burlingame, CA) for visualization, and hematoxylin counterstaining.

Evaluation of the immune reactions

Counting of labeled cells was performed by light microscope equipped with an eyepiece graticule, independently by two researchers who were blinded to the clinical information, and the mean value of their separate counts was used for the analysis. The number of the labeled cells within the metastases was registered in at least 10 (median: 20, range: 10–80) randomly chosen fields per section, using the graticule of 10 × 10 squares designating an area of 0.0625 mm2 at 400 × magnification. These fields were blindly chosen from different, non-adjacent areas of the metastases, omitting necrotic areas; in the case of smaller metastases, the whole tumor area was evaluated. Inter-observer concordance in density values was excellent (correlation coefficients between 0.6814 and 0.9156 for the different immune cell markers). For patients with more than one sample available, the mean labeled cell densities of all metastases studied were also determined. Cutoff levels were set up for each marker, based on the median of the given variable in the whole patient cohort, with minor adjustments for better discriminating power in some cases (46, 215, 430, 8, 10, 70, 83, 53, 5, 450, and 450 cells/mm2 for CD4, CD8, CD45RO, CD134, CD137, FOXP3, PD-1, CD20, NKp46, CD16, and CD68, respectively) in lymph node metastases. In cutaneous/s.c. metastases, the cutoff values used were 18, 21, 220, 4, 3, 20, 5, 4, 4, 370, and 500 cells/mm2, while in the case of all samples evaluated together, they were 24, 155, 440, 8, 7, 60, 28, 34, 4, 440, and 400 cells/mm2, in the same order. The proportion of patients with a mean cell density higher than the cutoff level was determined.

Statistical analysis

We used the Mann–Whitney U test for the statistical evaluation of differences in cell densities between different patient groups, χ2 test for comparing the proportions of samples with high cell densities, and the Pearson test for analyzing correlation between the densities of the different cell types. The Kaplan–Meier method and Mantel-Cox test were applied for evaluating survival. Univariate and multivariate Cox regression analyses were also performed using mean immune cell densities and patients’ age as continuous, while patient gender, disease stage, ECOG status, number of organs involved, LDH level, and previous treatments (chemo-, and radiotherapy) as categorical variables. Differences were considered significant in the case of p values ≤ 0.05. Statistics were calculated using the BMDP Statistical Software Pack.

Results

We determined the intratumoral density of immune cells expressing the different markers (CD4, CD8, CD45RO, CD20, CD134, CD137, FOXP3, PD-1, NKp46, CD16, and CD68) (Fig. 1) in 52 lymph node and 34 cutaneous or subcutaneous metastases. In lymph node metastases, CD45RO+ T lymphocytes, CD16+ cells, and CD68+ macrophages were the most numerous, followed by CD8+ T lymphocytes, FOXP3+ cells, CD20+ B lymphocytes, CD4+, and PD-1+ cells, while cells expressing the CD134 and CD137 activation markers, as well as NKp46+ NK cells were present only in small amounts (Fig. 2). For all cell types except CD68+ macrophages and NKp46+ NK cells, density values were significantly lower in subcutaneous/cutaneous lesions compared to nodal metastases (Fig. 2). Because of the large difference in immune cell densities between the two locations, the prognostic and predictive associations of immune cell infiltration were also evaluated separately in the two groups of metastases.

Fig. 1
figure 1

Immunohistochemical labeling of CD4+, CD8+, CD45RO+, FOXP3+, CD134+, CD137+, PD-1+, CD20+, NKp46+, CD16+, and CD68+ cells in metastatic melanoma samples (AEC, red)

Full size image
Fig. 2
figure 2

Density of immune cell types in pretreatment lymph node (n = 52) and subcutaneous/cutaneous metastasis samples (n = 34) from ipilimumab-treated patients. Boxes show first and third quartiles, horizontal lines are the medians, and whiskers indicate minimum and maximum values. LN lymph node, cut/sc cutaneous/subcutaneous

Full size image

Intratumoral infiltration of the labeled cells was evaluated in relation to response to ipilimumab treatment and disease outcome. Patients were categorized in two groups according to clinical efficacy of ipilimumab, showing either complete or partial response, or stable disease for at least 6 months (“responders”) or none of the above (“non-responders”). In lymph node metastases, mean densities of CD4+, CD8+, FOXP3+, CD134+ lymphocytes, CD20+ B cells, and NKp46+ NK cells were significantly higher in the responder group compared to non-responders (Fig. 3). For each cell type, a cutoff value was introduced based on the median of the given variable in the whole patient group (see in Materials and methods), and the proportion of patients with a mean intratumoral cell density exceeding this value was calculated and analyzed according to the efficacy of ipilimumab treatment. In this comparison, the above mentioned lymphocyte markers as well as CD137 also showed higher prevalence in the responders than in non-responders (Table 2). On the other hand, in subcutaneous/cutaneous metastases significant difference between responders and non-responders was found only in the proportion of patients with high mean density of CD68+ macrophages and CD16+ cells (Table 2). When all samples were evaluated together, significant association with response to treatment was found in the case of NK cell density values (p = 0.0182; not shown) and for proportion of patients with high density of NK cells as well as that of FOXP3+ cells and CD68+ macrophages (Table 2).

Fig. 3
figure 3

Density of immune cell types in pretreatment lymph node metastasis samples from ipilimumab-treated patients responding (resp.; n = 7) or not responding to treatment (nonresp.; n = 12). Circles: mean density values of samples from individual patients; horizontal line: median

Full size image
Table 2 Relationship of treatment response with proportion of patients with significant mean density of immune cells infiltrating metastases
Full size table

The densities of most of the studied immune cell types strongly correlated with each other and they frequently showed coordinate presence. In lymph node metastases, high expression of at least 7 of the 11 markers studied was found in 6 of the 7 responders (86%), compared to only 3 of 12 non-responders (25%) (p = 0.0106).

Kaplan–Meier analysis of survival according to the mean immune cell density in lymph node metastases revealed that high densities were associated with significantly longer overall survival (OS) in the case of 7 of the 11 cell types studied (Fig. 4a). The potential prognostic effect of immune cell densities evaluated as continuous variables (together with disease stage, patients’ age and gender, ECOG status, number of organs involved, LDH level, and previous treatments) was also analyzed using Cox’s proportional hazards model. In univariate analysis, CD4+, CD8+, CD45RO+, FOXP3+, and CD16+ cell densities were found significantly associated with overall survival (p = 0.0290, p = 0.0093, p = 0.0180, p = 0.0083, and p = 0.0047, respectively), besides ECOG status (p = 0.0009) and LDH (p = 0.0227). Multivariate analysis including all immune cell density values as well as clinicopathologic parameters identified ECOG status (p = 0.001) and FOXP3+ cell density (p = 0.004) as significant independent predictors of survival. Similar associations were found when all samples were evaluated together, using either Kaplan–Meier analysis (not shown) or Cox regression demonstrating significantly longer survival in case of high number of cells expressing CD4, CD8, CD45RO, FOXP3, CD16, CD68, or CD20 (p = 0.0348, p = 0.0136, p = 0.0113, p = 0.0121, p = 0.0055, p = 0.0168, or p = 0.0372, respectively), as well as in cases with better ECOG status (p = 0.0026) and normal LDH level (p = 0.0006). In multivariate analysis, LDH (p = 0.001) and the amount of FOXP3+ cells (p = 0.016) proved as independent predictive factors. In the s.c./cutaneous location, on the other hand, the mean density of CD16+ and CD68+ cells showed correlation with OS both in Kaplan–Meier analysis (Fig. 4b) and Cox’s proportional hazards model (p = 0.0197 and p = 0.0175, respectively); in this group, only LDH level proved as independent predictor of survival (p = 0.002). Disease stage, patients’ age and gender, the number of organs involved, and previous chemo- or radiotherapy were not found as significant parameters in univariate or multivariate analyses in either group studied.

Fig. 4
figure 4

Kaplan–Meier curves of overall survival for melanoma patients subdivided according to mean density of immune cells in lymph node (a), or cutaneous/subcutaneous metastases (b)

Full size image

Discussion

The introduction of ipilimumab, the first representative of the new class of immunotherapy, immune checkpoint agents, has revolutionized the treatment of metastatic melanoma, and paved the way for other immunomodulatory antibodies as PD-1/PD-L1 inhibitors, expanding the range of targeted tumor types. However, even in the case of the most efficient immunotherapeutic modalities, the majority of patients does not derive clinical benefit but are still at risk of developing side effects. Furthermore, the therapeutic arsenal of several tumor types has recently been widened with the introduction of other, potentially effective treatment modalities, e.g., targeted therapies, making it even more important to identify clinically usable predictive markers which could help in choosing the optimal treatment strategy for a given patient.

In our study, the density of several immune cell types, such as CD4+, CD8+ and CD45RO+ T cells, CD20+ B cells, lymphocytes expressing the activation markers CD134, CD137 or PD-1, FOXP3+ regulatory T cells, NKp46+ NK cells, CD68+ macrophages, and cells expressing CD16 (Fcγ receptor III), was determined in pretreatment surgical tumor samples of melanoma patients receiving ipilimumab therapy. Density values of each cell type were evaluated with regard to response to treatment and the outcome of the disease.

In this study, we restricted sample collection to metastases operated within 1 year before ipilimumab therapy, in an attempt to minimize potential changes in immune microenvironment during time elapsed between surgery and ipilimumab treatment. Furthermore, we evaluated surgical samples instead of small biopsies, and more than one metastasis per patient when possible, to reduce the confounding effect of intrapatient heterogeneity. To our knowledge, our study is the first that attempted to correlate density values of a panel of infiltrating immune cell types with response to treatment with ipilimumab in an everyday clinical setting, outside of clinical trials.

We found associations of treatment response and survival with intratumoral density of several infiltrating immune cell types. In lymph node metastases, the amount of CD4+, CD8+ T lymphocytes, FOXP3+ lymphocytes, cells carrying the CD134 activation marker, as well as CD20+ B cells and NKp46+ NK cells was higher in responders; the most pronounced difference could be seen in the case of the FOXP3 marker. The densities of CD45RO+, PD-1+, CD16+, and CD68+ cells showed correlation with survival, but not with treatment response. Analyzing subcutaneous/cutaneous metastases, on the other hand, resulted in significant associations of clinical response and survival with CD16+ and CD68+ cell density values. The amount of some of these immune cell types as CD8+ T lymphocytes and CD20+ B cells was also found prognostic in an earlier study on melanoma metastases (mostly lymph node and skin or soft tissue), while no association with survival was observed in the case of FOXP3+ cells [23].

Our results demonstrating association of T cell density with the efficacy of ipilimumab therapy fit well with the recent hypothesis suggesting different sensitivity of the so called “T cell-inflamed” and “non-inflamed” tumors to the various immunotherapeutic approaches, with higher probability of T cell-rich tumors to benefit from immunotherapies based on blocking immune suppressive mechanisms [24]. In support of this hypothesis, correlations between favorable disease outcome after checkpoint blockade and high expression of genes related to immune activity have been found in different tumor types [14, 25, 26]. Moreover, CD8+ T cell density in pretreatment biopsies proved predictive of response of melanoma patients to pembrolizumab treatment [27]; however, no such associations between the amount of infiltrating T cell subsets and clinical activity were apparent in the case of other anti-PD-1/PD-L1 agents or CTLA-4 blocking antibodies [15, 16, 25, 28]. A possible explanation for the lack of predictive power of T cell density in some studies could be that they analyzed together metastases of different locations, from patients with different tumor types in some cases. A potentially important finding of our study is that in melanoma patients immune cell densities as well as their predictive impact were different for lymph node vs. cutaneous/subcutaneous metastases, suggesting that evaluating all metastases without distinction according to their locations may not be optimal for revealing all differences between responders and non-responders.

According to our results, the factor most strongly predicting treatment response was the density of FOXP3+ cells infiltrating lymph node metastases. Similarly, strong baseline staining for FOXP3 and IDO was found to correlate with benefit from ipilimumab treatment in another study on melanoma patients [15]. The upregulation of immunosuppressive mechanisms in the tumors could also be connected to an active immune microenvironment, probably acting as counter-regulatory mechanisms. Indeed, Spranger et al. demonstrated correlation of IDO, PD-L1, and FOXP3 expression with each other and with the amount of CD8+ T lymphocytes in melanoma metastases, and proved in murine models that the presence of these inhibitory mechanisms was driven by infiltrating CD8+ T cells [29]. Our study also showed strong correlation between the densities of FOXP3+ cells and CD8+ or CD45RO+ lymphocytes (p < 0.001). Hypothetically, in T cell-rich tumors simultaneously present suppressive mechanisms may be responsible for inhibiting the antitumor effect of T lymphocytes, and therapeutic strategies aiming at blocking these mechanisms, such as immune checkpoint inhibitors, could be expected to be beneficial in these cases.

Recent studies have raised the possibility of ADCC-mediated intratumoral regulatory T cell depletion by Fcγ receptor expressing monocytes/macrophages as a mechanism of action of anti-CTLA-4 antibody therapy [30,31,32,33]. In metastases of melanoma patients receiving ipilimumab therapy, the density of CD16+CD68+ cells was higher in pretreatment samples of responders compared to non-responders [33]. Our findings partly correspond to these results, although significant differences in the density of CD16+ and CD68+ cells according to clinical response was found only in the case of cutaneous/subcutaneous metastases, while correlation with survival rather than with clinical response was seen in the case of nodal metastases. Together with the association of FOXP3+ cell density with treatment response [the present study and 15], these results may support a potential role of regulatory T cell killing by CD16+ effector cells in the effect of ipilimumab. We also observed a higher amount of the other potential ADCC effector NK cells in responders, however, the density of these cells was very low, which may question their possible biological significance.

We recognize the inherent limitations of our study caused by its retrospective nature. Moreover, the availability of sufficient surgical samples limits the number of patients that could be included in immunohistochemistry testing. On the other hand, the strengths of our analysis are that its data are derived mainly from a ‘real world’ patient cohort and that it evaluated pretreatment immune parameters, enabling the identification of predictive biomarkers that could help in making treatment decisions.

In conclusion, our data suggest that infiltration by FOXP3+ cells, CD4+, CD8+, CD134+ T lymphocytes, CD20+ B cells, and NKp46+ NK cells in lymph node metastases, as well as the prevalence of CD16+ cells and CD68+ macrophages in cutaneous/subcutaneous ones could be considered as candidate predictive markers in melanoma patients receiving ipilimumab therapy. Studies on larger patient cohorts are required to prospectively validate these and other parameters of tumor immune contexture as potential biomarkers of clinical response to ipilimumab treatment of melanoma patients. Investigation of the predictive value of these biomarkers in the case of anti-PD-1 or anti-PD-L1 agents as well as of anti-CTLA-4/anti-PD-1 combination therapies is also warranted.