Human leukocyte antigen (HLA)-A*0201-positive patients with metastatic melanoma who received tremelimumab monthly at the maximum tolerated dose in a phase I trial 31 were enrolled in an open-label expansion cohort between June 2004 and April 2005, and consented to donate repeated peripheral blood samples while receiving tremelimumab (10 mg/kg monthly) intravenously for up to 24 cycles. This study was conducted according to the Declaration of Helsinki and its amendments and relevant International Conference on Harmonization Good Clinical Practice guidelines. The protocol and consent forms, and all modifications, were approved by the University of California Los Angeles and University of Texas MD Anderson institutional review boards (approval numbers 03-01-059 and IDO3-0090, respectively). All patients provided written informed consent prior to any study procedures. Blood samples (40 mL) were collected from a peripheral vein during the screening period, on day 1 before the first dose of antibody (both baseline samples), on the day of each new dose in subsequent cycles, and 1 and 2 weeks after each monthly dose during the first 4 monthly cycles of therapy. When possible, a final blood sample was collected at the time the patient went off study for whatever reason. All patients underwent baseline and follow-up eye exams every 2 to 4 months throughout study participation following our previously-published eye surveillance protocol 3233. The primary endpoint was the determination of immune response to the melanoma-associated antigen recognized by T cells (MART1)-derived epitope MART1_26–36 by MHC tetramer assay across multiple time points. This was defined as detecting a 99% RCV for this assay and antigen, which corresponds to a minimum 80% change from baseline in the percentage of circulating antigen-specific T cells 29. Clinical responses were assessed by Response Evaluation Criteria in Solid Tumors (RECIST) 34. In addition, a clinical benefit response was defined for patients who were felt by the study investigators to have derived unequivocal clinical benefit after tremelimumab despite not meeting the standard criteria for response following RECIST.

Patients with surgically incurable stage IIIc or IV melanoma were eligible if they met the following major eligibility criteria: HLA-A*0201 positive, baseline level of circulating MART1_26–36-specific T cells above the lower limit of detection (LLD) by tetramer assay, previously defined as 0.03% of CD8⁺ T cells 29, disease measurable by RECIST, Eastern Cooperative Oncology Group performance status ≤ 1, and having received at least 2 doses of tremelimumab with samples for immune monitoring collected weekly between both doses. Major exclusion criteria were history of chronic inflammatory or autoimmune disease and presence of active brain metastases.

Peripheral blood mononuclear cells (PBMC) were isolated by Ficoll-Hypaque (Amersham Pharmacia, Piscataway, NJ) and cryopreserved in liquid nitrogen in Roswell Park Memorial Institute medium (RPMI) (Gibco-BRL, Gaithersburg, MD) supplemented with 20% (all percentages represent v/v) heat-inactivated human AB serum (Omega Scientific, Tarzana, CA) and 10% dimethylsulfoxide (Sigma, St. Louis, MO). Cryopreserved PBMC aliquots were thawed and immediately diluted with RPMI complete media consisting of 10% human AB serum and 1% penicillin, streptomycin, and amphotericin (Omega Scientific). Cells were washed and subjected to enzymatic treatment with DNAse (0.002%, Sigma) for 1 hour at 37°C. Cells were washed again and used immediately or were rested overnight in RPMI complete media in a 6% CO₂ incubator.

The following peptide epitopes were used: 1) 2 negative control epitopes: an HLA-A*0201-binding nonrelevant peptide (referred to as Negative peptide from here on) 35, and the HLA-A*0201 immunodominant peptide alpha fetoprotein (AFP)_325–332 (GLSPNLNRFL) derived from the oncofetal antigen AFP 36; 2) 2 infectious disease epitopes as positive controls, cytomegalovirus (CMV)pp65_495–503 (NLVPMVATV) and Epstein-Barr virus (EBV) BMLF1_259–267 (GLCTLVAML); and 3) 3 HLA-A*0201 immunodominant peptides derived from tumor rejection antigens: MART1_26–35 (ELAGIGILTV), tyrosinase_368–376 (YMDGTMSQV) and gp100_209–217 (ITDQVPFSV). All HLA-A*0201 tetramers were purchased from Beckman Coulter Inc., San Diego, CA, as peptide preloaded reagents, and the assay was performed following the manufacturer's instructions with minor modifications as previously described 29.

Interferon gamma (IFN-γ) ELISPOT assays were also performed as previously described 29. Briefly, PBMC were thawed from different time points and treated with DNAse. HLA-A*0201-transfected K562 (K562/A*0201), provided by Drs. Wolfgang Herr and Cedrik M. Britten (Johannes Gutenberg University, Mainz, Germany), were pulsed with the same peptide epitopes described for the tetramer assay and used as antigen-presenting cells. Then 1 × 10⁵ PBMC were mixed with 1 × 10⁴ peptide-pulsed K562/A*0201 in X-Vivo 10 media (BioWhittaker, Walkersville, MD) supplemented with 10% heat-inactivated human AB serum and seeded directly into anti-IFN-γ antibody coated ELISPOT plates for 20 hours.

Cocultures of thawed PBMC and peptide pulsed K562/A*0201 cells were plated as described for ELISPOT assays but were placed in triplicate in flat-bottom 96-well plates. Twenty hours later, supernatants were collected and frozen after centrifugation at 468 g for 10 minutes. Thawed supernatants from different time points were analyzed following the manufacturer's instructions using a 17-plex assay (Bio-Plex Human Cytokine 17-Plex Panel, Bio-Rad Laboratories, Hercules, CA). The cytokines quantified were IL-2, IL-4, IL-6, IL-8, IL-10, granulocyte-macrophage colony-stimulating factor (GM-CSF), IFN-γ, tumor necrosis factor alpha (TNF-α), IL-1β, IL-5, IL-7, IL-12 (p70), IL-13, IL-17, granulocyte colony-stimulating factor (G-CSF), monocyte chemoattractant protein 1 (MCP-1/MCAF), and macrophage inflammatory protein 1 α (MIP-1α). Data were analyzed using Bio-Plex manager software with 5PL curve fitting.

Unspecific antibody binding to Fc receptors from thawed PBMC was blocked with 100% adult bovine serum (Omega Scientific). PMBC were then stained using a panel of fluorescein-labeled antibodies against the following T-cell surface antigens: FITC-conjugated-UCHL1 (anti-CD45RO), APC-CY7-conjugated-SK3 (anti-CD4), APC-conjugated-FN50 (anti-CD69) (all from BD Biosciences, San Jose, CA), Pacific Blue-conjugated-S4.1 (anti-CD3) (Invitrogen, Carlsbad, CA), and ECD-conjugated-Immu-357 (anti-HLA-DR) (Beckman Coulter). Cells were fixed with 0.5% paraformaldehyde. Immediately before flow cytometric analysis, 5 μL of 7-amino-actinomycin D (7-AAD) was added to gate out dead cells. The 5 color flow cytometry staining was acquired by a FACSAria using Fluorescence Minus One approach 37. Analysis was performed with FCS Express (DeNovo Software, Thornhill, Ontario, Canada) software.

PBMC were first labeled with the following surface antibodies: Pacific Blue-conjugated-S4.1 (anti-CD3), AlexaFluor467-conjugated-RPA-T4 (anti-CD4), APC-CY7-conjugated-M-A251 (anti-CD25). Intracellular staining for FoxP3 protein was performed following the manufacturer's instructions using the PE-conjugated-PCH101 anti-FoxP3 antibody (eBioscience, San Diego, CA). Flow cytometric analysis was performed as described above.

Total RNA was extracted from thawed PBMC using RNeasy mini kit (Qiagen, Valencia, CA). Human FoxP3 mRNA expression was quantified using the iScript One-Step Quantitative reverse-transcriptase PCR kit (Bio-Rad) in an Opticon 2 (MJ Research, Ramsey, MN), and sample concentrations were corrected with human 18S rRNA 38. Amplification was conducted in a total volume of 25 μL for 40 cycles of 15 seconds at 95°C, 30 seconds at 60°C and 30 seconds at 72°C, with 2 beginning steps: 10 minutes at 50°C (to convert RNA to cDNA) and 15 seconds at 95°C (to inactivate reverse transcriptase). Samples were run in triplicate. FoxP3 primers were forward 5'-CAA GTT CCA CAA CAT GCG AC-3'; and reverse, 5'-ATT GAC TGT CCG CTG CTT CT-3' 39. 18S rRNA, primers were forward 5'-GC-CGA-AGC-GTT-TAC-TTT-GA-3' and reverse 5'-TCC-ATT-ATT-CCT-AGC-TGC-GGT-ATC-3' 38.

To generate a single-cell suspension and analysis of tumor infiltrating lymphocytes (TIL), tumors obtained from an outpatient excisional biopsy were decapsulated, minced with sterile surgical blades and enzymatically digested for 1 to 2 hours with DNAse I (0.1 mg/mL, Sigma) and collagenase D (1 mg/mL, Boehringer Mannheim, Indianapolis, IN) in 100 mL of AIM-V^® media (Gibco-BRL). Cells were plated in flasks in RPMI culture media and allowed to adhere for 2 hours. At that time, nonadherent cells, enriched for TIL, were collected and cryopreserved for tetramer analysis.

The previously defined 99% RCV 29 was applied to detect statistically significant changes in values above the lower limit of detection for the tetramer and ELISPOT assays. Data from all assays were normalized to an arbitrary scale from 0 to -4 or +4 change from baseline for the generation of heat maps for unsupervised and supervised hierarchical clustering of results. The mean of the results of 2 predosing samples was given a value of 0, and postdosing changes could be either positive or negative. For tetramer and ELISPOT assay results, positive changes < 99% RCV were scored as 1, twice the RCV was scored as 2, 3 times the RCV as 3, and 4 was scored for any result beyond this point. Negative changes compared with baseline were scored with the corresponding negative values from -1 to -4. The percentage change from baseline for surface flow cytometry results and QPCR was calculated for FoxP3 mRNA as described by Maker et al 40. A value of 1 was assigned for a positive or negative percentage change between 0 and 50% from baseline, 2 for percentage changes between 51% and 100%, 3 for percentage changes between 101% and 150%, and 4 for changes beyond 151%. Samples were analyzed by pairwise average-linkage cluster analysis 41 using dChip software initially developed for genome-wide microarray expression data analysis. In this analysis, data vectors typically assigned as gene probe sets were substituted with the aforementioned transformed tetramer, ELISPOT, flow cytometry, and QPCR serial data points. Samples were assigned vector columns, and assay points were assigned vector rows. In addition, a paired t test was used to determine significant changes in pre- and postdosing levels of the surface markers HLA-DR and CD45RO.

Twenty-three HLA-A*0201-positive patients with surgically incurable stage IIIc or IV MART1-positive melanoma provided a baseline blood sample for screening analysis of circulating CD8⁺ T cells specific for MART1_26–35 epitope. Patients with a baseline value within the measurable range for the tetramer analysis (≥ 0.03%) were selected. The inclusion of this criterion was based on the goal of quantification of both positive and negative changes in the number of circulating MART1_26–35-specific T cells after treatment with tremelimumab. Patients with baseline levels of MART1_26–35-specific T cells below the assay LLD would not contribute to assessing a potential decrease in circulating T cells specific for this antigen upon treatment with CTLA4 blocking antibodies 29, a possibility that has not been excluded in prior studies 8914194042. Fifteen patients were enrolled and 12 received 2 or more monthly doses of tremelimumab; 2 patients experienced rapid tumor progression before the second scheduled dose, and 1 patient never received the first dose due to rapid deterioration of performance status. The 12 patients who received at least 2 doses of tremelimumab were considered evaluable for immune-monitoring assays, and their characteristics are summarized in Table 1.

None of these patients developed a grade 4 treatment-related toxicity; however, 5 patients experienced treatment-related grade 3 toxicities. Two (1105A and 1108) developed marked increase in transaminases after 11 and 5 doses, consistent with hepatitis (Table 1). Acute viral hepatitis was ruled out in both cases. In both cases this was a reason for study discontinuation. Although neither patient received corticosteroids, both fully recovered within 2 months. One of these patients underwent 5 daily therapeutic plasma exchange procedures with efficient clearance of detectable levels of circulating tremelimumab. Two patients (1106 and 1114) developed grade 3 diarrhea after 4 and 2 doses. One underwent colonoscopy and biopsy, which demonstrated histologically proven colitis with lymphomonocytic infiltrates predominantly by CD4⁺ T helper cells and CD68⁺ macrophages (Figure 1). A final patient (1109) developed a generalized rash, which was described as leukocytoclastic vasculitis at biopsy. Clinically significant grade 2 toxicities included 2 cases of panhypopituitarism suggestive of hypophysitis, patient 1108 after 5 doses and patient 1113 after 22 doses of tremelimumab. This later patient also developed anterior uveitis after 21 doses that rapidly improved following treatment with corticosteroid-containing eye drops.

One patient (1108) achieved a complete response of 2 nodal metastases that were proven to be melanoma at baseline by fine needle aspiration. This response is ongoing 39+ months later. One patient (1111) had a major partial response of bulky in-transit metastasis, and remains relapse-free at 34+ months (Figure 2A and 2B). A third patient (1113) did not qualify as a responder by RECIST criteria but was determined to have unequivocal evidence of clinical benefit. A [¹⁸F]fluorodeoxy-glucose (FDG) positron-emission tomography (PET)-positive 4-cm psoas muscle mass became PET negative after 9 doses. Upon surgical resection, pathological analysis showed a 90% regression of melanoma with no disease progression at 33+ months. This patient is listed as having a pathological partial response. Only 1 of 5 patients with grade 3 toxicity had an objective clinical response.

T cells from treated patients were screened to determine whether responses to tumor-specific antigens developed as a result of treatment with tremelimumab. A mean of 2 baseline and 7 follow-up (range, 5 to 9) blood draws per patient were tested. Samples were analyzed by flow cytometry for negative tetramer-specific MART1_26–35-specific CD8⁺ T cells. Percentages of negative tetramer-specific and MART1_26–35-specific CD8⁺ T cells at each time point indicated increases beyond the 99% RCV in 4 patients on more than 1 occasion (Figure 3). However, these were spikes of peripheral appearance of melanoma antigen-specific T cells at isolated time points, as opposed to persistent elevations in several blood draws taken after dosing. Patient 1111, who had an objective clinical response to therapy, underwent biopsy of a responding lesion. MHC tetramer analysis of nonadherent cells from this lesion demonstrated a 6-fold enrichment of gp100_209–217-specific CD8⁺ T cells among total CD8⁺ T cells compared with peripheral blood (Figure 2C and 2D).

To efficiently present all of the data describing positive or negative changes across 12 patients, multiple time points, and the different test epitopes (a total of 719 assay results), the data were converted to a -4 to +4 scale based on the 99% RCV (as described in the Materials and Methods and Table 2) and color coded in shades of blue for negative changes and red for positive changes. These data were compiled in a heat map similar to the ones used to analyze data from gene expression profiling (Figure 4). As described for MART1_26–35-specific T cells, no definite trend was observed for changes over time in the proportion of CD8⁺ T cells reacting to any melanoma-(gp100_209–217, tyrosinase_368–376) or infectious disease-(EBV BMLF1_259–267, CMVpp65_495–503) specific antigen in any patient.

The heat map in Figure 4 also presents all the data generated by ELISPOT assay enumerating cells producing IFN-γ after short ex vivo peptide stimulation. Results are color coded following the -4 to +4 scale range with negative and positive changes defined by the 99% RCV compared with the results from the 2 baseline samples (Table 2). The data summarize replicate results from a mean of 9 (range, 6 to 11) time points analyzed from the 12 study patients testing reactivity to the negative control and test antigens for a total of 1,694 sample results. As with the tetramer assay, there was no evidence of a consistent change over time in the assay results for any of the tested epitopes.

The possibility that IFN-γ may not be the most significant cytokine to define the functional status of antigen-specific T cells after CTLA4 blockade was explored. Therefore, we set up cocultures of 2 × 10⁵ PBMC and 2 × 10⁴ K562/A*0201 cells pulsed with either negative, MART1_26–35, tyrosinase_368–376, and gp100_209–217 peptides in parallel with the ELISPOT assays of patients 1108, a responder, and 1110, a nonresponder. Twenty hours later supernatants were collected and analyzed with a bead-based multiplex assay of 17 cytokines (background [PBMC plus K562/A*0201] was subtracted from all samples). The PBMC cytokine profile from patient 1110 was under the assay limit of detection except for baseline IL-8, MCP-1, and MIP-1β. In contrast, cell supernatant from patient 1108 showed that IL-6, IL-8, G-CSF, MCP-1, and MIP-1β increased after dose 1 (week 2) and dose 2 (week 3) only. IFN-γ, TNF-α, IL-7, and IL-13 demonstrated a small increase after dose 2 (week 3). The cytokines IL-2, IL-4, IL-5, IL-10, IL-17, GM-CSF, and IL-12(p70) were not detected in supernatants of this patient (data not shown). From this preliminary analysis, we concluded that multicytokine analysis in these 2 representative patients was unlikely to provide additional information beyond the information obtained by the standardized IFN-γ ELISPOT.

Based on the emerging results suggesting that there was no consistent change in the number of circulating CD8⁺ T cells specific for melanoma or infectious disease antigens analyzed by either biochemical (tetramer) or functional (ELISPOT and multicytokine) assays, we sought evidence of T-cell activation regardless of antigen specificity. Cell surface expression of the activation-induced T-cell marker HLA-DR (MHC class II molecule) and the T-cell memory marker CD45RO have been reported to be increased after administering CTLA4-blocking monoclonal antibodies 89141940. Ten patients had at least 1 baseline and 1 postdosing aliquot of PBMC available for this analysis, for a total of 81 analyzed samples. There was no overall significant increase in the expression of these 2 surface markers on either CD4⁺ and CD8⁺ T-cell subsets comparing pre- and postdosing samples, although HLA-DR and CD45RO were significantly increased on CD4⁺ cells after the first dose (paired t test P = .02 and .01, respectively), and CD45RO was significantly increased on CD8⁺ cells after cycles 1 and 2 (paired t test P = .009 and .04, respectively). Results were then converted to percentage change from baseline as described by Maker et al 40 and assigned a score from -4 to +4 (Table 2) to allow compiling of data in the heat map (Figure 4).

An alternative possibility was that CTLA4-blocking antibodies affected CTLA4-positive Treg. Other groups have explored this possibility with divergent results 141940. In 8 patients (including the 3 patients with clinical benefit), aliquots of at least 1 baseline and 1 follow-up PBMC sample were available for analysis using QPCR to quantitate FoxP3 mRNA transcripts. We failed to detect a statistically significant change (either positive or negative) between pre- and postdosing results. Results from the 56 total tested samples were converted to the same scale used for the flow cytometry data (Table 2) and compiled in the heat map (Figure 4).

There were not enough cells derived from the 40 mL peripheral blood draws to perform functional assays of Treg-mediated suppression. In a further attempt to test a possible effect of tremelimumab on circulating Treg, expression of the intracellular FoxP3 protein in cells with the CD3⁺CD4⁺CD25^high phenotype was analyzed by flow cytometry. This included at least 1 baseline and 1 postdosing representative sample taken from 2 responders (1108 and 1111) and 3 nonresponders (1103, 1105A, and 1110), for a total of 19 analyzed samples. Data were again converted to percent change from baseline, scored with values between -4 and +4, and compared with results from FoxP3 mRNA analysis at the same time points. There was good concordance between the results of both assays at 9 time points, whereas at the other 10 time points both assays diverged at values ≥ 2 or ≤ -2 (data not shown). These discrepancies underscore the difficulty in efficiently quantifying Treg using currently available approaches.

The data compiled in the heat map presented in Figure 4 contain all available results of postdosing values compared with the mean of the 2 baseline assay results (with a score of 0). Because the data were converted to a normalized scale of -4 to +4 based on what were determined to be biologically significant changes over baseline, a hierarchical unsupervised clustering analysis could be performed to process all these data in a collective but exploratory approach. This analysis segregated the study patients randomly, with no clustering of patients by clinical response (Figure 4). Unsupervised hierarchical clustering also did not separate patients with grade 2/3 treatment-related toxicities from patients without treatment-related toxicity. We then conducted more limited unsupervised clustering including the results of the tetramer, IFN-γ ELISPOT, or FoxP3 QPCR data alone. Again, patients clustered randomly. However, when we conducted the unsupervised clustering for HLA-DR and CD45RO markers, the 3 clinical responders segregated along with several nonresponders with evidence of inflammatory toxicity (1105A with grade 3 hepatitis, 1106 with grade 3 diarrhea, 1109 with grade 3 leukocytoclastic vasculitis). This segregation was statistically significant (P = .05; Figure 5A) and suggests that patients with an objective tumor response and/or toxicity after administration of tremelimumab have a more active immune response detectable in peripheral blood.

The data set was then examined to explore which assay results better segregated the 3 clinical responders from nonresponders (Figure 5B). Increase in CD45RO on CD4⁺ cells after 3 doses and increase in HLA-DR on CD8⁺ cells after 2 doses significantly differentiated responders from nonresponders (2-fold cutoff difference, P = .03). These data strengthen the correlation between nonantigen-specific markers of immune activation on T cells and antitumor activity.