Prospective Study of the Evolution of Blood Lymphoid Immune Parameters during Dacarbazine Chemotherapy in Metastatic and Locally Advanced Melanoma Patients

Background The importance of immune responses in the control of melanoma growth is well known. However, the implication of these antitumor immune responses in the efficacy of dacarbazine, a cytotoxic drug classically used in the treatment of melanoma, remains poorly understood in humans. Methods In this prospective observational study, we performed an immunomonitoring of eleven metastatic or locally advanced patients treated with dacarbazine as a first line of treatment. We assessed by flow cytometry lymphoid populations and their activation state; we also isolated NK cells to perform in vitro cytotoxicity tests. Results We found that chemotherapy induces lymphopenia and that a significantly higher numbers of naïve CD4+ T cells and lower proportion of Treg before chemotherapy are associated with disease control after dacarbazine treatment. Interestingly, NK cell cytotoxicity against dacarbazine-pretreated melanoma cells is only observed in NK cells from patients who achieved disease control. Conclusion Together, our data pinpoint that some immune factors could help to predict the response of melanoma patients to dacarbazine. Future larger scale studies are warranted to test their validity as prediction markers.


Introduction
In the last few years, there have been many considerable changes in the treatment of metastatic melanoma, with the development of BRAF and MEK tyrosine kinase inhibitors and monoclonal antibody immunotherapies, such as anti CTLA-4 and anti PD-1, which have proved efficacy both in term of clinical response and overall survival. However, in metastatic melanoma patients who presented side effects, no indications or failure of targeted therapies or immunotherapy, ''classical'' cytotoxic chemotherapy is still used, namely dacarbazine (DTIC).
It has been shown that some cytotoxic drugs could affect the immune system and the antitumor immune response. The first mechanism is directly related to the cytotoxic property of these agents on cancer cells. The seminal discovery was that cancer death induced by some chemotherapies could prime CD8 + T cell antitumor immune response. This phenomenon is an essential contributor to the antitumor effect of some major anticancer drugs such as anthracyclines and oxaliplatin both in mice and humans [1][2][3][4]. The second phenomenon involves the capacity of some anticancer agents to selectively kill or affect the biology of some immune cells. Anticancer drugs can eliminate immunosuppressive cells and enhance antitumor immune responses [5][6][7] or mitigate cytotoxic antitumor immunity by inducing some immunosuppressive mechanisms [8]. In a recent work, using a mouse melanoma model B16F10, we identified DTIC immunological effect. While DTIC did not directly affect immune cells in this mouse model, we observed that DTIC triggered the expression of NKG2D ligands on tumor cells that led to activation of natural killer (NK) cells, interferon (IFN)c secretion, then activation of cytotoxic T cells [9,10]. We also observed that in vitro DTIC treatment enhanced NKG2D ligand expression on human melanoma cell lines. We thus formulated the hypothesis that for some patients, DTIC may also enhance NK cell toxicity toward melanoma cells and that this could be related to the clinical response to DTIC.
To address this question, we performed immunomonitoring of lymphoid subpopulations of patients with metastatic melanoma before and after a first cycle of DTIC treatment.

Patients
To monitor immune markers that could potentially be implicated in melanoma as prognosis and/or prediction for treatment response, we immunomonitored patients from the university Hospital of Dijon (France), bearing unresectable or metastatic melanoma. All patients gave their written informed consent and were treated with the cytotoxic chemotherapy DTIC at 1 g/m 2 every four weeks. The response rate was evaluated one month after the third cycle of treatment by CT scan, according to RECIST criteria 1.1. This trial received approbation of the ethics committee of the Hospital of Dijon under the number 2010/55 eudract n: 2010-02358-34 and was conducted following the Declaration of Helsinki's protocol. Patient whose death was related to cancer are counted as ''dead'', the others are censored after their last sign-of-life.

Blood sample preparation
Blood samples were obtained from patients before the first and second cycle of chemotherapy. Blood was collected in vacutainer tubes, with EDTA for determination of total leukocytes, lymphocytes and granulocytes concentration or with citrate for immunomonitoring. Citrated samples were diluted 1:1 with RPMI1640 (Lonza) and centrifugated on a cushion of lymphocyte separation medium (Eurobio). Peripheral blood mononuclear cells (PBMC) were collected and washed once with phosphate buffered saline (PBS) 5% bovine serumalbumin (BSA). 5% of the pellet was taken for further polymerase chain reaction (PCR) analysis, 10% for flow cytometry analysis and the remaining cells underwent NK cell separation with an NK cell isolation kit (Miltenyi Biotech), followed by separation on AutoMACS (Miltenyi).

Cytotoxicity assay
The MelC human melanoma cell line was a kind gift from Dr Anne Caignard. MelC cells, previously treated with PBS or DTIC (0.4 g/L) were incubated overnight either alone or with freshly purified NK cells from patients. Tumor cells were harvested and death was assayed using DAPI (49,69-diamidino-2-phénylindole) staining (Sigma Aldrich) by flow cytometry. Tumor cells were identified by their characteristic size and granularity, and cells positives for DAPI staining were considered dead.

PCR analysis
Total RNA from PBMC was extracted using TRIZOL (Invitrogen) as recommended by the manufacturer. The RNAs were rehydrated and measured with a NanoDrop 1000 (Thermo-Scientific). Reverse transcription was performed using the kit "M-MLV Reverse Transcriptase" (Invitrogen) as recommended by the manufacturer using a Biometra thermocycler. The quantitative PCR reaction was performed using the kit ''SYBR Green PCR Master Mix'' (Applied Biosystems) according to manufacturer's recommendations, on an Applied 7500 device (Applied Biosystems). Relative mRNA levels were determined using the difference in cycle threshold (DCt) method. Values were expressed relative to ACTB. We detected ACTB (with primers GATCATTGCTCC-TCCTGAGC and TGCGCAAGTTAGGTTTTGTC), EOMES (CAGCACCACCTCTACGAACA and CGCCACCAAACTGA-GTGAT), TBX21 (AACATCCTGTAGTGGCTGGTG and C-CACCTGTTGTGGTCCAAGT) and RORC (AAGCAGGAG-CAATGGAAGTG and GCAATCTCATCCTCGGAAAA).

Flow cytometry
Blood counts (leukocytes, granulocytes, lymphocytes and monocyte counts) were performed on the Xn2000 (Sysmex) device, which is the device used for clinical blood count in Dijon's hospital routine care. Fluorescent data were produced using a LSRII cytometer (BD Pharmingen). PBMC were stained in PBS/ 5%BSA with antibodies indicated in Table 1. All antibodies were purchased from Biolegend (Ozyme) and used according to the manufacturer's protocol. Analyses were done with FlowJo software (Tristar). Events were considered positive if their fluoresence was higher than matched-isotype control staining peak fluorescence. The gating strategy is summarized in Fig. S1 & S2.

Statistical analysis
We used Mann-Whitney's test for nonparametric means comparison. Student's t test was used when variance were tested comparable by Fisher-Snedecor's test. Paired tests were used when applicable. Linear correlations were assessed using the Pearson correlation coefficient. All tests were performed and graphed with GraphPad Prism 6 software. All p values were two-tailed. A p value inferior to 0.05 was considered statistically significant for all experiments and is indicated with *.

Patients
Eleven patients were included in this prospective observational trial. Patients' clinical characteristics are described in Table 2. In our cohort, one patient exhibited partial response, two showed disease stabilization and eight had progressive disease following DTIC treatment. Patients that showed partial response or disease stabilization after DTIC were grouped as ''disease control''. So, in our cohort, there were 3 (27%) patients with disease control and 8 (73%) with progressive disease, which is consistent with the published response rates to DTIC in the literature [11,12]. Survival analysis (Fig. S3) underscores that patients with progressive disease have a shorter survival than those with disease control (median survival 7 vs 19.5 months, respectively, Mantel-Cox'test p = 0.0379).

Blood count
A total blood count was performed before the first and second cycle of chemotherapy. One injection of DTIC induced no change in the total number of leucocytes (Fig. 1A). When looking at leukocyte types, we observed a drop in total lymphocyte count (Fig. 1B, p = 0.0439), whereas no significant modifications were observed for monocyte (Fig. 1C) or granulocyte counts (Fig. 1D). Neither lymphocyte nor monocyte counts before or after chemotherapy were associated with disease control (not shown). Surprisingly, patients with disease control showed a tendency of higher granulocyte counts after treatment compared to the progressive disease group (Fig. 1E).

T cells
We did not observe differences in total number of CD4 + or CD8 + T cells before and after chemotherapy ( Fig. 2A), although a tendency could be observed for both, accounting for the diminution of total lymphocyte count (Fig. 1B). When looking at memory and naïve subpopulations, we observed that chemotherapy did not significantly affect the number of naïve CD4 + T cells ( Fig. 2B) or naïve CD8 + T cells (not shown). However, high numbers of naïve (CD45RA + ) CD4 + T cells before chemotherapy are seen in patients with controlled disease (Fig. 2C). Neither the numbers of CD4 + or CD8 + T cells after one cycle of chemotherapy, nor naïve CD8 + T cell number are correlated with clinical outcome (Not shown).

T cell polarization
The polarization state of T cells was determined using the monitoring of master controller transcription factors using RT-qPCR. There was a strong correlation between expression of EOMES, coding for eomesodermin, before and after treatment and survival (Fig. 3A), thus suggesting that CD8 + or NK cells antitumor cytotoxic responses would be associated with longer survival. We did not find any correlation between the expression of TBX21, a gene coding for T-bet, before and after chemotherapy and overall survival (Fig. 3B). Conversely, after chemotherapy, the Th17 transcription factor, RORC, was associated with longer survival (Fig. 3C).

Regulatory T cells
We evaluated the proportion of FOXP3 positive cells amongst the CD3 + CD4 + double-positive lymphocyte cells. There was a tendency for the proportion of FOXP3 expression to positively correlate with tumor stage (Fig. 4A). We noted that chemotherapy did not significantly affect FOXP3 + cell numbers (Fig. 4B). However, we observed that, before chemotherapy, patients with disease control had fewer FOXP3 + than those with progressive disease (Fig. 4C).

NK
Treatment had no significant effect on total number of either CD56 hi or CD56 dim CD16 + NK cell subsets (Fig. 5A). We also did not detect significant modifications of the activation state of NK cell subsets after chemotherapy using CD69 labeling (Fig. S4). Furthermore, we did not see any association between CD56 dim CD16 + NK cell activation and response. However, we observed a higher activation of CD56 hi NK cells in patients with disease control after chemotherapy (Fig. 5B) thus suggesting that DTIC may help activation of CD56 hi NK cells in these patients.
We previously demonstrated that, in a mouse model, DTIC induces the expression of NK cell activating ligands on the surface of melanoma cells. This causes a better detection and lysis of DTIC-treated targets by NK cells [10]. To validate this observation in a human context, we tested in this cohort the cytotoxic properties of purified NK cells from patients before chemotherapy, against a human melanoma cell line ex vivo pretreated with DTIC at a dosage that triggers expression of NKG2D ligand [10]. We observed that NK purified from patients with progressive disease are unable to lyse either DTIC-treated targets or PBS-treated targets [13] (Fig. 5C, left panel), with a cytotoxicity -if any-comparable to the background mortality, i.e. the death of target cells cultured without NK cells, that is 6+/21% (IC95[3.7-8.2]) and 5+/21% (IC95[2.1-8.2]) for PBS-and DTIC-treated MelC cell, respectively. In contrast, NK cells from patients with disease control seem to have better capacity to kill pretreated DTIC target cells (Fig. 5C, right panel).

Discussion
To our knowledge, this study is the first that performs a systematic study of the effect of alkylating chemotherapy on all blood lymphoid cell subsets in melanoma patients. We observed that chemotherapy modestly but significantly affects the total number of lymphocytes, which is reduced after chemotherapy. This observation could be linked to the toxicity of this anticancer drug. However chemotherapy did not affect the number of CD8 + or CD4 + T cells or Treg or NK cells thus underscoring that such treatment is not a major immunosuppressive treatment.
High expression of EOMES gene, encoding eomesodermin is also correlated with longer survival. Eomesodermin has been described as a master regulator of cytotoxic CD8 + T cells [14], and can also be expressed by NK cells in mice [15]. This suggest that the intrinsic presence of a cytotoxic cellular response could be important for DTIC efficacy, thus corroborating our previous mouse findings that DTIC antitumor effects are dependent on CD8 T cells and NK cells [10].
After treatment, high expression of RORC, coding for RORct, a transcription factor responsible for CD4 + Th17 differentiation, correlates with better survival. Th17 cell induction has already been described with various melanoma treatments, such as vaccination [16], anti-PD-1 [17], anti-CTLA4 [18], as well as other anticancer drugs such as cyclophosphamide [19]. Furthermore, it has been described that Th17 cells can drive antigendependent tumor shrinkage [20,21]. We hypothesize that Th17 cell induction after treatment could be a surrogate marker of efficacy. This event could be related to the association of high granulocyte counts in responder patients after treatment, because Th17 cells are well known to be important in the activation and recruitment of neutrophils.
CD4 + T cell subsets seem to be associated with clinical outcome. In particular, the presence of higher number of naïve CD4 + T cells at baseline is associated with better response to DTIC. In patients  with low numbers of naïve CD4 + T cells, we suspect that there is an increase in the number of memory and exhausted lymphocytes. Indeed, in chronic inflammatory situations such as cancer, the presence of chronic antigen stimulation leads to an exhausted or senescent memory phenotype of CD4 + T cells. These cells are known to dampen anti-tumor immune response efficacy [22]. Such cells are known to accumulate in peripheral blood of cancer patients [23]. It is possible that the high number of exhausted lymphocytes could impair chemotherapy associated antitumor immune response, thus rendering such drug less effective. In addition, our study corroborates the previous finding that higher Treg rate is associated with advanced clinical stage [24]. Moreover, higher Treg number at baseline is associated with absence of response to DTIC. Such data suggest that high tumor burden and/or immunosuppressive state could impair antitumor efficacy of DTIC probably by inhibiting drug related antitumor immune response.
The major finding of this study is the relationship between NK cell activity and response to DTIC. We analyzed the two subsets of peripheral blood NK cells, that are described as functionnaly different [25]: CD56 hi NK cells can secrete high amounts of cytokines [26], and CD56 l uCD16 + are perceived as cytotoxic [27]. Our results show that, while DTIC did not affect the total number of NK cells, CD56 hi NK cells from patients with disease control have a higher activation status than the patients with progressive disease. We previously showed that DTIC can enhance NKG2D ligand expression by melanoma cells [10]. In a mouse model, we showed that this causes IFNc secretion by NK cells, followed by major histocompatibility complex expression by tumor cells, hence better recognition by T cells. We selected here a melanoma cell line in which DTIC ex vivo treatment induced up regulation of NKG2D ligands. We and others already described how these cells are killed by NK cells thanks to NKG2D-recognition [9,10,28]. Using this setting, we could see here that NK cells from responder patients seem to have a stronger capacity to kill DTIC-pretreated   cancer cells than NK cells from patients with progressive disease. This is coherent with the published literature, as Anne Caignard's group described how NK cell phenotypes and functions vary according to tumor stage and treatment [28,29]. Indeed, they showed that NK cell receptor NKP46 expression by NK cells seems diminished in patient bearing metastatic melanoma, and that chemotherapy could increase expression of NKP46 by NK cells, but IFNc secretion and degranulation are decreased after chemotherapy in ex-vivo assays, due to up-regulation of inhibiting factor NKG2A [28,29]. They also showed no variation of NKG2D expression by NK cells after chemotherapy or according to tumor stage. Our data support that alkylating chemotherapy such as DTIC is only effective in patients with functional NK cells and support that either some patients have intrinsic more proficient NK cells or that the tumor drives an immunosuppressive state, which blunts NK cell function. The second hypothesis is more probable because we described that pretreatment of melanoma cell line with DTIC enhanced the cytotoxicity of NK cells from healthy volunteers [10] and that high number of Treg before chemotherapy is associated with absence of response to DTIC.
Taken together, our study underscores some immune properties of DTIC and gives some preliminary data to isolate predictive factors of response. We believe that the two most promising factors are the importance of naïve CD4 T cell number and NK cell activity to predict response to chemotherapy. First, our results show that high Treg number and low number of naïve CD4 + T cells are associated with disease progression after chemotherapy. These patients are presumably those with exhausted memory cells and may greatly benefit from immune-based treatments, such as anti-CTLA4 or anti-PD1 antibody, because these two treatments aim to reverse immune inhibition and exhaustion. Furthermore, patients having highly cytotoxic NK cells could be better candidates to DTIC-based treatment, as they may benefit the most from its immune effects, enhancing NK-and CD8 + T cell based cytotoxicity. Larger studies are warranted to validate these results. Figure S1 Flow cytometry strategy. We present here the gating strategy for flow fluorocytometry analysis. We present two examples of NK cell CD69 labelling. (EPS)  Figure S4 Activation of NK cells is not related to chemotherapy. We compared NK cell activation before and after chemotherapy. We used CD69 labelling by flow cytometry to assess activation of CD56 hi (A) and CD56 dim CD16 + NK cells (B). (EPS)