Frequencies of circulating MDSC correlate with clinical outcome of melanoma patients treated with ipilimumab

Introduction

Primary melanomas in early stages are curable by surgery. After metastasis formation, however, patients have a median survival of <1 year. Chemotherapy may induce occasional tumor responses [1] but does not improve overall survival. Until 2010, long-term responses in metastatic melanoma were achieved only after adoptive transfer of autologous T cells [2], a therapy applicable to a selected minority of patients with good performance status, or after high dose IL-2 treatment [3].

About 50 % of melanomas express BRAF-activating mutations [4], which are V600E and V600K substitutions in 95 % of cases [5]. Treatment with BRAF inhibitors, MEK inhibitor, or a combination of both prolonged progression-free survival [5–7]. Despite these great advances, patients treated with BRAF or MEK inhibitors experience disease progression after a mean of about 6 months [7]. Recently, the fully humanized anti-CTLA-4 antibody ipilimumab was reported to increase overall survival of stage III/IV metastatic melanoma patients [8, 9]. The treatment received FDA approval for non-resectable and metastatic melanoma in March 2011 [10]. CTLA-4 is expressed on activated T cells, serving as “immune checkpoint” molecule which prevents overwhelming cytotoxicity and consequent autoimmunity and tissue damage. Upon activation, T cells express CTLA-4 which binds to its ligands CD80 and CD86 with higher affinity than CD28. Several mechanisms of action have been proposed explaining the strong effect of CTLA-4 blockade [11].

It is well accepted that tumor-reactive T cells in vivo upregulate CTLA-4. In situ, they show an exhausted, less functional phenotype [12]. Treatment with the CTLA-4-blocking antibody ipilimumab increased T cell activation [13] and their proliferation [14] in stage III/IV melanoma patients. Higher T cell activation correlates with tumor regression or disease stabilization in about 11–13 % of patients [8, 15]. However, autoimmune responses are also promoted, as 10–15 % of patients develop severe grade 3–4 immune-related adverse events [8]. Interestingly, these autoimmune manifestations correlate with clinical responses [15].

Tumor lesions may increase in size with the onset of inflammation during ipilimumab treatment, followed by size reductions frequently several months after the completion of ipilimumab dosing [16]. Traditional tumor response criteria are therefore not fully appreciating the clinically relevant effects of anti-CTLA-4 treatment. The new immune-related response criteria take into account that clinical responses can develop late, possibly only after initial tumor mass increase due to inflammation [17]. These criteria have already proven to be highly relevant for oncoimmunological-patient assessments [18], despite that they are not yet fully established for phase III studies. According to these criteria, the first tumor assessment is performed 12 weeks after therapy start.

Identification of biomarkers that are associated with clinical responses to ipilimumab may help to identify patients likely to benefit from the treatment. In bladder cancer, increased frequencies of CD4⁺ ICOS^high T cells correlated with a better clinical outcome [19]. For melanoma, this correlation remains to be confirmed as results are controversial [14, 20]. Absolute lymphocyte counts after 2 infusions of ipilimumab [18] and the presence of NY-ESO-1-specific antibodies and of specific CD8⁺ T cells [21] correlated positively with clinical outcome in melanoma. Furthermore, high levels of Ki67⁺ EOMES⁺ CD8⁺ T cells were associated with improved relapse-free survival in melanoma patients [14].

While most studies focused on T cell and antibody responses in patients treated with ipilimumab [13, 14, 20–24], we were particularly interested in studying myeloid cells. In humans, several myeloid-derived suppressor cells (MDSC) subpopulations have been described termed monocytic and granulocytic MDSC, respectively [25–28]. Different malignancies favor the accumulation of different MDSC phenotypes in patients [29]. In melanoma patients, circulating CD14⁺ HLA-DR⁻ MDSC were described to be enriched as compared to healthy donors (HD) [30–32]. However, MDSC also vary between different studies of the same cancer type, e.g., human renal cell carcinoma [33–35]. MDSC in humans are widely accepted to have low or absent HLA-DR expression [25–28], whereas some experimental mouse models showed that MHC-II expression on MDSC may be important for suppression of CD4⁺ T cells [36].

In our study of patients treated with CTLA-4-blocking antibody, we hypothesized that improved T cell activity with stronger proliferation [14] and enhanced cytokine secretion [23] might be paralleled with alterations in the myeloid cell compartment. For example, IFN-γ production may play a role because it contributes to MDSC recruitment [37]. Besides the known T cell intrinsic mechanisms of CTLA-4, several T cell extrinsic pathways have been described, for instance by competition for co-stimulation via CD80/CD86 (B7), or B7-mediated outside-in-signaling altering myeloid cells, particularly antigen-presenting cells, in their capacity to activate or tolerize immune responses [38]. To obtain more information about mechanisms regulating immune responses in cancer patients, we analyzed MDSC in peripheral blood of advanced melanoma patients receiving the standard regimen of 4 3-weekly infusions of ipilimumab (3 mg/kg). Our results show significant correlations between frequencies of monocytic MDSC and clinical responses in patients treated with ipilimumab.

Materials and methods

Results

Discussion

In the present study, we set out to perform a detailed flow cytometry analysis of MDSC in patients with cutaneous melanoma, with a special focus on potential effects of anti-CTLA-4 therapy consisting of 4 3-weekly cycles of ipilimumab (Yervoy^®) at 3 mg/kg. In order to avoid in vitro artifacts due to freezing–thawing procedures and delayed blood processing, we analyzed fresh blood samples directly after withdrawal. We optimized the analysis for monocytic MDSC by using PBMC prepared with conventional density gradient centrifugation, as MDSC are purified together with mononuclear cells. This procedure is an alternative to whole blood stainings, where a higher background due to less pure cell fractions is a common obstacle.

In concordance with the literature, we found that monocytic lin⁻ CD14⁺ HLA-DR⁻ MDSC were significantly enriched in PBMC from melanoma patients compared to HD [30–32]. This difference occurred only as a trend when comparing HD and untreated patients, likely because fewer samples were available from untreated patients. As some of the vemurafenib-treated patients show higher frequencies of lin⁻ CD14⁺ HLA-DR⁻ cells, we cannot exclude a possible relationship between this treatment and high MDSC frequencies. We did not have enough patients and samples to follow this trend, but, most importantly, vemurafenib patients do not account for the significant difference in frequencies of lin⁻ CD14⁺ HLA-DR⁻ cells in melanoma patients and HD, as this difference remains significant when vemurafenib-treated patients were excluded from the analysis. Confirming our results, Mandruzzato et al. [44] identified IL-4R-alpha expression on monocytic cells enriched in cancer patients compared to healthy controls, but as opposed to our data no HLA-DR expression was assessed.

We observed a trend pointing toward enrichment of monocytic MDSC in patients with tumor burden, compared to tumor-free patients. In line with this, for patients with non-active disease, reduced frequencies of CD14⁺ HLA-DR⁻ MDSC have been described for melanoma patients and of CD14⁻ CD33⁺ CD11b⁺ MDSC for patients with breast cancer [30, 45].

So far, very few is known about MDSC frequencies in melanoma patients treated with ipilimumab. It is important to study how MDSC behave in the context of an environment with increased T cell activity, in order to estimate drug-induced changes in the myeloid cell compartment of patients. For example, T cells from ipilimumab-treated patients secrete higher amounts of IFN-γ [23] which is known to recruit MDSC [37]. Therefore, one could expect that ipilimumab-treated patients accumulate higher MDSC frequencies. In contrast, our data show no significant differences in MDSC frequencies over time in ipilimumab-treated patients, indicating that these cells are tightly regulated. In contrast, the patient-to-patient variability was high, even among patients with the same tumor staging.

When testing the potential of ipilimumab as neoadjuvant in stage III b-c patients, a drop of MDSC frequencies 6 weeks after treatment induction was reported [46]. Furthermore, Tarhini et al. have analyzed several MDSC subsets in melanoma patients receiving the anti-CTLA-4 antibody tremelimumab together with IFN-α. In contrast to our study, a decrease in frequency of monocytic MDSC was seen at days 29 and 85 [47] when cryopreserved samples were used. We analyzed fresh PBMC, immediately after blood withdrawal from patients and healthy controls, because our results with cryopreserved cells were inconsistent (manuscript in preparation). Besides technical differences, there are further reasons why the studies are not directly comparable. For example, different ipilimumab dosing was used in the neoadjuvant study, and anti-CTLA-4 therapy was conducted in combination with IFN-α in the tremelimumab study.

Most probably, T effector and T regulatory cells are the most direct targets of anti-CTLA-4 therapy. However, CTLA-4 expression has been observed in small quantities on the surface and more pronounced intracellular in monocytes [48, 49], suggesting that ipilimumab could eventually act on monocytes or MDSC directly. Comin-Anduix et al. reported changes in the phosphorylation status of p38, ERK1/2, Stat-1, Stat-3 and Stat-5 in monocytes of patients treated with the anti-CTLA-4 antibody tremelimumab. Furthermore, monocytes of clinical responders and NR to tremelimumab differed in phosphorylation levels of Akt and Stat-6 [49]. In contrast to these observations, our study did not reveal CTLA-4 expression on monocytic cells and also no changes of MDSC frequencies were induced by ipilimumab treatment.

We did not find any correlation between serum LDH levels at baseline and accumulation of MDSC in blood of melanoma patients. If MDSC recruitment is altered in patients with higher serum LDH levels, reflecting strongly progressive disease, this might take place only in the tumor microenvironment, remaining undetectable in the blood.

Patients with bulky metastatic disease (M1c) showed a trend toward higher frequencies of monocytic MDSC than patients with metastases limited to skin and/or lymph nodes (M1a). For the more immature MDSC population of lin⁻ CD33⁺ CD11b⁺ cells, this correlation has been described already in a patient cohort comprising of different malignancies [50]. Interestingly, in our study, patients with stage IV M1a metastatic disease showed MDSC levels comparable to HD, indicating that MDSC increased only later, i.e., with tumor spread to visceral sites, beyond skin and lymph nodes. This might involve tumor secreted VEGF, known to play a role in MDSC recruitment [51]. Possibly, higher amounts of MDSC might facilitate metastasis formation, e.g., by creating an immunosuppressive environment.

When distinguishing clinical responders and non-responders within the different metastatic stages, we identified a subgroup of M1b and M1c patients. These patients had higher frequencies of monocytic MDSC and were mostly clinical non-responders. The subsequent analysis of clinical responders and non-responders addressed this effect.

Clinical benefit of ipilimumab treatment can take months to develop. Initially, tumor lesions might even increase in size due to local recruitment of immune cells and inflammation [16]. Therefore, the tumor response evaluation by standard RECIST criteria at week 12 after treatment may not reflect actual anti-tumor activity. As recommended recently, we applied the immune-modified RECIST criteria [17] and evaluated the tumor response by CT scans performed at week 12, 16 and/or 24 after treatment. Our immune monitoring of patients treated with ipilimumab focused on identifying biomarkers associated with clinical response, with the aim of finding predictive markers. Our data show for the first time that MDSC frequency in PBMC of melanoma patients with ipilimumab treatment correlates significantly with the patient’s clinical response. This is also observed as a trend in patients with highly aggressive bulky metastatic disease (M1c), whose treatment options were very limited before the introduction of ipilimumab.

In contrast to our study, the frequency of the more immature lin⁻ HLA-DR⁻ CD33⁺ CD11b⁺ MDSC was reported to decrease stronger in patients responding to tremelimumab in combination with IFN-α [47]. For the reasons mentioned above (different patient treatments, different laboratory techniques), the results of the studies cannot be compared directly.

It has been shown previously that patients with high baseline expression levels of immune-related genes and an immune-active tumor microenvironment in tumor tissue are more likely to respond favorably to ipilimumab [52]. Our data on peripheral blood are in line with this observation, as they show that patients with low frequencies of MDSC, implying less immune suppression and better reactivity, have improved clinical outcome.

Recently, a correlation between the frequency of CD14⁺ HLA-DR⁻ cells and overall survival could be detected in renal cell carcinoma patients [53]. Furthermore, frequencies of lin⁻ HLA-DR⁻ CD33⁺ CD11b⁺ MDSC correlated with overall survival in patients with pancreatic and esophagogastric cancer [54]. In contrast, our data did not reveal significant correlations between MDSC frequencies and overall survival. To our knowledge, we provide the first evidence that MDSC frequencies in PBMC correlate with clinical response to treatment with ipilimumab. Although our study was retrospective and involving a relatively small number of patients, it suggests that MDSC frequencies correlate with outcome of anti-CTLA-4 treatment. Prospective clinical trials assessing MDSC subpopulations and frequencies as potential biomarkers of response to CTLA-4 blockade are therefore warranted to validate our observations.

Electronic supplementary material

Below is the link to the electronic supplementary material.