Introduction
Melanoma has long been acknowledged to be a susceptible target for immunotherapy. A high number of melanomas are infiltrated by T cells, and the strength, phenotype, and activation status of infiltrated cells are associated with survival [
1].
Adoptive cell transfer (ACT), i.e., the infusion of ex vivo expanded tumor-specific T cells, resulted in objective tumor responses in melanoma patients [
2]. Previous ACT trials have demonstrated that clinical success required lymphodepletion prior to T-cell infusion and concomitant administration of high-dose interleukin-2 (IL-2) to maintain cell proliferation and activation status [
3].
In allogeneic stem cell transplantation, interferon-alpha (IFNα) is routinely used to replace the far more toxic IL-2 to sustain the activation status of infused donor cells [
4]. In addition, IFNα displays the capacity to upregulate HLA class I expression on tumor cells [
5]. These molecules are critical for tumor cell recognition by T cells and often (partially) downregulated in melanoma [
6]. Furthermore, IFNα is described to drive DC maturation required for generation of tumor-specific effector- and central-memory CD8+ T cells [
7,
8].
Therefore, we studied the feasibility and potential of ACT in combination with IFNα for the treatment of metastatic melanoma. Tumor-reactive T cells were obtained from PBMC, as highly effective melanoma reactive CTL are readily induced from PBMC [
9,
10] following stimulation with autologous tumor cells established from tumor biopsies.
Patients and methods
Melanoma patients
Patients with histologically proven cutaneous melanoma with verified progressive metastatic disease stage IV or unresectable stage III refractory to treatment before the start of the study and with at least one measurable target lesion were eligible for this phase I/II study. The protocol was approved by the Medical Ethics Committee of the Leiden University Medical Center and conducted in accordance with the Declaration of Helsinki. All patients gave written informed consent. Patients with brain metastases who are neurologically unstable and/or require use of dexamethasone and patients who are on continued chemo- or radiotherapy until 4 weeks before the start of IFNα injections were excluded from the study. The primary objective of this study was to investigate the feasibility and safety of adoptive T-cell transfer in combination with IFNα. Secondary endpoints include the best overall response defined as disease control (CR, PR, and SD) and evaluation of immunological parameters as predictors for response.
Treatment protocol
Patients received T cells as three consecutive infusions of on average 259 (range 38–474) million cells with a 3-week interval in combination with low-dose IFNα (3 million IU Roferon, Roche, Woerden, The Netherlands, subcutaneously daily) starting at day 7 before the first T-cell infusion and for a total of 12 weeks. Cryopreserved T cells were thawed and diluted to a total volume of 300 ml in PBS/2% human serum albumin (Albuman, Sanquin, Amsterdam, the Netherlands) and a final concentration of less than 1% DMSO. Then, the cells were administered intravenously over a time period of 30–60 min. At various time points before, during and after treatment, the tumor response was assessed by physical examination and tumor imaging (CT and MRI) and evaluated according to the Response Evaluation Criteria in Solid Tumors 1.0 (RECIST) [
11] and according to the immune-related response criteria (irRC) [
12]. Toxicity was scored using the CTC scale version 3.0.
Establishment of GMP-grade primary melanoma lines
Autologous melanoma cell lines were initiated and cultured using authorized Standard Operating Procedures (SOPs). Tumor tissue obtained by surgery was directly processed by mechanical dissociation and cultured in tumor medium consisting of Dulbecco’s MEM (DMEM, Lonza, Veriers, Belgium) supplemented with 8% irradiated, heat-inactivated GMP-grade pre-tested FCS, penicillin (50 U/ml), streptomycin (50 μg/ml), and
l-glutamine (4 mM) (all from Lonza). A stable cell line could be established (usually within 4–12 weeks) for approximately 50% of the patients that was subsequently used as a stimulator line for the generation of tumor-reactive T-cell cultures. Before the initiation of T-cell cultures, the melanoma origin of the used autologous tumor cells was confirmed by RT–PCR [
13‐
20] and they were shown to express at least three of the eleven melanoma-associated antigens studied (data not shown).
Furthermore, genotyping of HLA alleles was performed in all melanoma cell lines and corresponding PBMC to confirm their origin and revealed in addition that none of the used melanoma cell lines showed complete or partial genetic loss of HLA alleles. In addition, the surface expression of HLA class I and class II molecules was assessed by flow cytometry using Alexa-648-conjugated W6/32 HLA class I A/B/C (Serotec, Dusseldorf, Germany) and FITC-conjugated Bu26 HLA-DR/DP/DQ-specific antibodies (Bio-connect, Huissen, the Netherlands). The samples were measured on a FACSCalibur and analyzed using CellQuest Software and showed that HLA class I surface expression was not downregulated on the cell lines used for the induction of T cells and that all IFNγ-stimulated melanoma cells, except Mel AB and Mel CT, displayed variable surface expression of HLA class II (data not shown).
Generation and release of T-cell batches for infusion
Tumor-reactive T cells were cultured using a mixed lymphocyte tumor cell culture (MLTC) as described [
21]. Briefly, PBMC were isolated from heparinized venous blood and cryopreserved until use. Approximately 50 million PBMC were thawed, washed, counted and resuspended (1–2 × 10e
6 cells per ml) in T-cell medium; Iscoves MDM (Lonza) supplemented with 8% heat-inactivated pooled human serum (Sanquin Bloodbank, Dordrecht, The Netherlands), penicillin (50 U/ml), streptomycin (50 μg/ml), and
l-glutamine (4 mM) (all GMP-grade from Lonza) and plated 1 ml per well in a 24-well culture plate. Next, autologous tumor cells were lethally irradiated (100 Gy) and resuspended (1 × 10e
5 cells per ml) in T-cell medium. One milliliter of this tumor cell suspension was added per well containing PBMC. Human recombinant IL-4 (5 ng/ml, Cellgenix, Sieversen, Germany and Gentaur, Brussels Belgium) was added at day 0. Starting from day 2, low-dose (150 IU/ml) IL-2 (Aldesleukin, Novartis, Arnhem, The Netherlands) was added. Culture medium containing IL-2 was refreshed every 2–3 days. The T cells were cultured for a total of 4 weeks and stimulated weekly with irradiated tumor cells as described above. After 4 weeks of culture, the T-cell batches were evaluated for sterility, phenotype, and tumor reactivity before they were released for infusion. Phenotypic analysis was performed using a FACSCalibur and CellQuest software (BD Pharmingen, San Jose, CA, USA). T cells were released when they consist of more than 80% of T cells (CD3+CD4+ and CD3+CD8+) or NK cells (CD3-CD56+) (All conjugated-mAb were from BD Pharmingen except CD8-APC from DAKO, Heverlee, Belgium). In addition, the frequency of Treg CD4+CD25+FoxP3+ T cells was evaluated in the T-cell batches using CD4-APC and CD25-FITC mAb (both from BD Pharmingen) in combination with the human Treg/FoxP3-PE staining kit (eBiosciences, Vienna, Austria). Tumor reactivity was assessed by a
51Cr-cytotoxicity assay [
22], and in order to be released for infusion, T cells were required to specifically lyse autologous tumor cells but not autologous EBV-LCL or PHA blasts (difference >10% at an E/T ratio of 30). Alternatively, tumor reactivity was assessed by IFNγ ELISA (Sanquin, Amsterdam, The Netherlands) where T-cell batches were required to secrete >200 pg/ml IFNγ in response to autologous tumor cells and at least twice the amount induced by the negative controls, i.e., autologous EBV-LCL or PHA blasts [
23].
Analysis of T-cell specificity, polyclonality, and cytokine secretion by multiparameter flow cytometry
To examine T-cell reactivity, 0.5–1 × 10e
6 T cells were seeded in a 24-well plate in T-cell medium and stimulated with 0.5–1 × 10e
6 autologous melanoma cells. After 1 h, Brefeldin-A (10 μg/ml) was added and stimulation was continued overnight. The next day T cells were harvested, and part of the cells were stained with a mix of antibodies to CD154-PECy5, CD137-APC, CD4-PECy7, CD8-APCCy7, IFNγ-FITC, and IL-2-PE (all from BD Pharmingen) and CD3-Pacific Blue (DAKO) see [
24]. Non-stimulated T cells or T cells stimulated with PHA (5 μg/ml) served as controls. Responses were considered positive when the percentage of tumor-stimulated CD154+ and/or CD137+ T cells was at least three times the medium control (see Fig 1. of the on-line available supplementary materials).
Polyclonality of the T-cell batches was evaluated by the anti-IFNγ-FITC and IL-2-PE Ab by either one of the 8 different mixes of FITC- and/or PE-conjugated anti-TCRVβ mAb (BETAmark, Beckman Coulter) as previously described [
25]. The different T-cell clones were operationally defined as the percentage of activated tumor-specific CD4+ or CD8+ T cells expressing the same TCRVβ-chain, within the population of tumor-specifically activated CD154+ and/or CD137+ T cells, respectively. All samples were measured on a calibrated LSRII and analyzed using DIVA software (BD). A TCRVβ was considered dominant (>10%), subdominant (3–10%), or minor (<3%) based on the percentage of tumor-specific cells using the same TCRVβ [
25].
In parallel, T cells were stimulated with autologous melanoma cell lines in the absence of Brefeldin-A, and supernatant was harvested after 24 h in order to analyze the cytokine secretion using the human Th1/Th2 cytometric bead array (BD Pharmingen). T cells stimulated with PHA (5 μg/ml), medium alone, and autologous EBV-LCL or PHA blasts were included as controls. Antigen-specific cytokine production was defined by a cytokine concentration above the cut-off value (IFNγ 50 pg/ml; other cytokines 10 pg/ml) and >2 × the concentration of the medium control [
26].
Statistics
The overall survival was analyzed using Kaplan–Meier and SPSS version 2.0. The Cox regression could not performed due to the low patient number, and therefore no confidence interval and hazard ratio could be provided. Post hoc statistical analyses were performed to determine potential correlates with clinical responsiveness using the non-parametric Mann–Whitney Test or Wilcoxon Signed Ranks Test. All p values are two-sided and are considered significant when they are less than 0.05.
Discussion
The results of this phase I/II study clearly demonstrate that adoptive transfer of PBMC-derived tumor-specific T cells in combination with low-dose INFα is feasible and safe and can induce a disease control (CR, PR, and SD) in 50% of the treated patients displaying progressive metastatic melanoma before treatment without any major adverse events related to the infusion of T cells. Analyses of several different immune-related parameters, which are discussed in more detail below, resulted in a number of hypotheses that could explain clinical reactivity or the lack thereof. All together, these data warrant the conduct of a controlled clinical trial to confirm or reject our current observations.
All responding patients experienced an IFNα -induced mild and transient leucopenia including both reduced lymphocyte and in particular neutrophil numbers. Interestingly, high peripheral neutrophil numbers are described to be an independent prognostic factor for worse outcome of IL-2-based therapy in melanoma patients [
28]. Here, we observe that the level of neutrophil reduction after IFNα-conditioning correlates with increased survival and that all clinically responding patients display a clear reduction in peripheral neutrophil numbers after IFNα treatment. Notably, there was no difference in neutrophil numbers between these groups before treatment that could explain a difference in survival. Since expansion of IL-10-secreting immunosuppressive neutrophils has recently been shown to correlate with disease progression in melanoma [
29], the observed IFNα-mediated reduction in neutrophil numbers may have resulted in lower IL-10 levels but unfortunately this could not be tested in our current patient group because we did not obtain serum of all patients prior to therapy. A recent study in melanoma patients reported the reduction in regulatory T cells (Treg) following high-dose IFNα treatment [
30]. Whether the leucopenia observed in our study after a much lower dose of IFNα treatment also reduces the frequency of Treg or creates space for infused T cells and/or induces homeostatic cytokines IL-7 and IL-15, as has been suggested as mechanism to explain the beneficial effect of other lymphodepleting-conditioning regimens [
2], remains to be elucidated. Although IFNα is a pleiotropic cytokine and its use may contribute to the eventual clinical effect via a variety of pathways, a direct inhibitory effect of IFNα on the tumor is unlikely since single-agent IFNα treatment is only effective in much higher doses than the low dose used in our study, which was proven to be not effective in phase I/II studies [
31,
32].
To unravel potential mechanisms underlying the clinical responsiveness of patients, we performed a detailed characterization of the infused T-cell batches and observed that the expansion rate of T cells during the production phase was higher for cells infused in patients with a clinical response. This cannot be attributed to a reduced frequency of tumor-specific T cells in the PBMC or lack of antigen expression on autologous cell lines used to obtain the T-cell batches, since tumor-specific polyclonal T-cell populations were obtained in all patients. Whether the reduced expansion rate of T cells infused to non-responding patients is due to expression and/or secretion of immunosuppressive factors by their autologous tumor cells is subject of our current investigations. Alternatively, the reduced proliferation in the non-responding patients may be due to an intrinsic lack of proliferative capacity of their PBMC. Indeed, the proliferative capacity of ex vivo expanded tumor-infiltrating lymphocytes has been associated with tumor regression and has been linked to telomere length in these T cells [
33]. The reduced proliferation of T cells may also be due to the selective expansion of Treg, since at least in some cases (patient HL and JS), relatively high frequencies of CD4+CD25+FoxP3+ T cells were detected in the T-cell batches. Indeed, the accumulation of Treg in peripheral blood and locally in the tumor environment and metastatic lymph nodes of melanoma patients has previously been associated with impaired T-cell responsiveness [
34] and as such supports this idea. However, functional studies are required to appreciate their suppressive role during the mixed tumor T-cell cultures in vitro.
The stimulation of patients’ PBMC with autologous tumor cells, expressing the full spectrum of relevant HLA molecules and tumor antigens, leads to expansion of polyclonal CD4+ as well as CD8+ T cells, which may contribute to the current success. Although a recent case report shows that treatment of a melanoma patient with clonal NY-ESO-1-specific CD4+ T cells can be successful [
35], we aim at infusion of polyclonal T cells, thus preventing the appearance of antigen-loss variants frequently found after transfer of single-antigen-specific T cells [
36,
37]. Since polyclonal tumor-reactive T cells are infused in all patients but the clinical response does not merely correlate to the percentage or ratio of specific CD8+ versus CD4+ T cells, we conclude that the presence of a broad repertoire of tumor-reactive T cells by itself is not sufficient to induce a clinical effect but relies on other factors including their functional activity. Our patient number is too small to reach significance, but our results are in line with previous observations [
38‐
40], indicating that infusion of T cells that predominantly produce Th1 cytokines results in a better clinical outcome compared to infusion of T cells that produce mainly Th2 cytokines or no Th1/Th2 cytokines at all and that the presence of both tumor-specific Th1 cells and CTL mediates an effective anti-tumor response [
41].
In summary, the adoptive transfer of polyclonal tumor-specific T cells obtained after stimulation of PBMC with autologous tumor cells in combination with low-dose IFNα can result in durable clinical responses in stage IV melanoma patients supporting the idea to explore IFNα as an alternative conditioning regimen and cytokine for ACT trials. Although the use of IFNα also comes with some side effects, including the beneficial leukopenia and controllable psychological symptoms, it is far less toxic than high-dose IL-2, commonly used to support transferred T cells. The clinical responses reported here are associated with IFNα-induced lympho- and neutropenia and the proliferative capacity as well as the Th1/Th2 cytokine profile of the T cells used for infusion, underscoring the importance of measuring parameters that are associated with clinical reactivity of the infused T-cell batches to fully appreciate their in vivo effectiveness. Eventual combination of this approach with other treatment options may even further enhance the clinical outcome. One possibility is the combined use of ACT/IFNα with ipilimumab or negative immunoregulatory human cell surface receptor PD-1 (programmed death-1). These immune activating antibodies may result in further in vivo expansion of tumor-specific T cells and improved clinical effect after ACT/IFNα.
Acknowledgments
The authors wish to thank Dr. Els Persijn-van Meerten for experienced evaluation of CT and MRI scans and Prof. Hein Putter for statistical evaluation of the data. We also want to thank Conny Hoogstraten, Hannah Wensink, Inonge van Twillert and Loes van Eijk for their technical assistance.