Patients ≥18 years with metastatic, cutaneous, stage IIIc/IV melanoma, with at least one resectable metastatic lesion of sufficient size (≥3 cm in diameter), were eligible. Inclusion criteria included a clinical performance status ≤1, absolute neutrophil count ≥1.5 x 10⁹/L, platelet count ≥100 x 10⁹/L, hemoglobin >5 mmol/L, serum Alanine (Amino)Transaminase/Aspartate Aaminotransferase (ALAT/ASAT) <3 times the upper d of normal, unless patients had liver metastases (<5 times ULN) and serum creatinine clearance of 50 mL/min or higher. In addition, patients needed to be seronegative for HIV, hepatitis B antigen, hepatitis C antibody and syphilis. Exclusion criteria did not allow participation of patients with autoimmune disease, requirement of corticosteroids, two or more CNS metastases, symptomatic CNS metastasis and inability to receive HD IL-2.

All patients provided informed consent before enrolment. This study was registered as 2010-021885-31.

Patients underwent surgical resection of a melanoma metastasis to grow TIL in vitro (see below). Prior to infusion of the TIL product, patients received a non-myeloablative, lymphocyte-depleting preparative regimen consisting of cyclophosphamide (60 mg/kg/day x 2 days intravenous) and fludarabine (25 mg/m²/day x 5 days intravenous). Intravenous adoptive transfer of TILs was followed by HD intravenous IL-2 (Proleukin) (600 000 IU/kg/dose every 8 hours up to tolerance or up to 15 doses). Supportive care consisted of blood or platelet transfusions, which were given until spontaneous hematopoietic recovery.

A first complete assessment of evaluable lesions was conducted 6 weeks after cell infusion and periodically afterwards, at 12 and 18 weeks, and every 3 months thereafter for 2 years, to obtain best objective response by Response Evaluation Criteria in Solid Tumors (RECIST).

The generation of ‘young’ TIL was based on established protocols.23 26 In brief, TIL isolation and expansion was started by generation of a single cell suspension by enzymatic digestion of the resected metastatic tumor material obtained by surgery, using collagenase type IV (Sigma Aldrich) and Pulmozyme (Roche). Resulting cell suspensions were cultured in the presence of 6000 IU/mL IL-2 (Proleukin, Novartis) for 2–4 weeks until a minimum of 1×10⁸ cells was achieved. At this point, TIL were either frozen (for logistic reasons) or directly expanded using an REP. In this process, 35×10⁶ CD3 positive T cells were cultured in 50% Roswell Park Memorial Institute (RPMI)/ 50% AIM-V medium in the presence of 3000 IU/mL IL-2, 30 ng/mL anti-CD3 (OKT-3, MACS GMP CD3 pure, Miltenyi) and irradiated autologous PBMCs, at a 200:1 feeder cell: TIL ratio. Cultures were initiated in 35 T175 culture flasks and continued in a WAVE bioreactor from day 6 onwards.26 At day 14, TIL were harvested and concentrated by centrifugation, followed by a wash step with PBS containing 5% human serum albumin. TIL were then suspended in a physiological saline solution, supplemented with 2.5% human serum albumin (Sanquin Blood Supply) and 300 IU/mL IL-2, as the final infusion formulation.

TIL manufacturing was performed in a class B cleanroom (GMP EU Annex 1) with strict air pressure hierarchy for combining aseptic conditions with containment. Handling of product with direct exposure to the environment was performed in a class A (GMP EU Annex 1) biohazard laminar down-flow cabinet with a class B background. Performance of the cleanroom facility was monitored by measurement of the levels of microorganisms and particles both at operating state and at rest state. In addition, aseptic handling was validated by periodic medium fills. Final quality control of TIL products consisted of cell number (between 5×10⁹ and 2×10¹¹), viability (>70%), sterility, and the absence of tumor cells above level of detection. Cell number and viability was determined by visual counting using a hemacytometer under a microscope, according to Ph.Eur. 2.7.29. Sterility was confirmed using the BacT/ALERT system (Biomerieux), which was validated according to Ph. Eur. 2.6.27. The absence of detectable residual tumor cells was determined in a cytospin by immunohistochemical staining for the tumor markers MART-1, gp-100 and S100. The sensitivity of the test was determined to be 0.2% of tumor cells within a TIL population.

A small fraction of TIL infusion products was cryopreserved for subsequent phenotypic characterization by flow cytometry. In brief, after thawing, cells were rested for 24 hours, and then stained using antibodies directed against: CD3 (clone SK7), CCR7 (clone 150503), CD28 (clone CD28.2), CD62L (clone SK11) (all BD Biosciences), CD4 (clone S3.5), CD8 (clone 3B5), CD45RA (clone MEM-56), CD27 (clone CLB-27/1) (all Invitrogen) and PD-1 (clone J105, eBioscience). Fixable Violet Dead Cell Stain Kit (Invitrogen) was used to exclude dead cells. Cells were analyzed on a LSR2 SORP flow cytometer (BD Biosciences) and data was processed using Flowjo_V10 software.

A total of 2×10⁵ TILs were cultured with 2×10⁵ cells from tumor digest or tumor cell lines, in the presence of 1 µL/mL Golgiplug (BD Biosciences). After 5 hours of incubation, cells were stained using anti-CD8 antibody (clone SKI, BD Biosciences) and live/dead marker (Fixable Violet Dead Cell Stain Kit, (Invitrogen)). Intracellular cytokine production was detected after fixation and permeabilization with the Cytofix/Cytoperm kit (BD Biosciences) and antibodies against interferon-γ (IFN-γ) (clone B27), IL-2 (clone Mq1-17H12), and tumor necrosis factor α (TNF-α) (clone Mab11), all from BD Biosciences. Cells were analyzed on an LSR2 SORP flow cytometer and data were processed using Flowjo_V10 software.

Exome sequencing of tumor and autologous healthy PBMCs was performed to identify tumor specific mutations. DNA libraries were generated using the Illumina TruSeq DNA library preparation kit. Resulting gDNA libraries were enriched for exonic sequences using the Sure Select Human All Exon 50 Mb Target Enrichment system (Agilent, Cat.no. 5190–2851). RNA was isolated using the RNeasy purification kit (QIAGEN). Poly(A)-selected RNA libraries were prepared using the TruSeq RNA library protocol (Illumina, San Diego, California, USA), and resulting libraries were sequenced on an Illumina HiSeq2000 using 75 bp paired-end reads. Reads were aligned (Tophat version 1.3.3 (7)); expression values were calculated as fragments per kilobase of transcript per million mapped reads (FPKM) using cufflinks (V.1.0.2 (8)). Mutations identified in genes with FPKM=0 were excluded from further analysis. Subsequently, all potential 9, 10 and 11mers were filtered using three parameters: (1) likelihood of proteasomal processing (NetChop score >0.49); (2) predicted binding affinity to patient specific HLA alleles (with the highest ranking 35 peptides included per 100 mutations for each allele, yielding a cut-off of approximately 1700 nM for HLA-A*0201) using NetMHCpan and (3) A similarity-to-self test predicting the degree of similarity between the mutated epitope and the wild type counterpart.27 Predicted peptides were synthesized (LUMC, Leiden, The Netherlands), and major histocompatibility complex (MHC) multimers containing these epitopes were produced by microscale parallel UV-induced peptide exchange reactions.28 29 Subsequently, TILs and PBMCs from patients were analyzed for reactivity against these predicted neoantigens by a multiplexed peptide MHC multimer staining strategy, as described previously.29

For patients 3 and 4, a previously described immortalized B cell platform was also utilized for the detection of neoantigen reactivity.30 31 In brief, peptides that covered the mutant amino acid flanked by nine amino acids on each side were synthesized for each mutation and were loaded onto autologous immortalized B cells. Subsequently, B cells were incubated for 48 hours with sorted CD8⁺ cells from the TIL product, and recognition of mutant peptides was assessed by Cytokine Bead Array (BD Science). Peptides were first tested in pools of five peptides, that were subsequently deconvoluted using individual peptides. Any hits obtained were subsequently confirmed by MHC multimer-based flow cytometry.

TIL characteristics were compared for responders and non-responders, using the two-tailed Mann-Whitney U test (SPSS V.22).

In the period between 2005–2011 and 2004–2014, 13 patients were enrolled in this study. The prolonged follow-up of this patient group renders analysis of potential patient benefit more informative. For all patients enrolled, TIL products were successfully generated. Three patients did not receive the final TIL product because of rapid disease progression. All patients had been heavily pretreated with chemotherapy, immunotherapy and/or targeted therapy (see table 1).

Out of the 10 treated patients, two achieved a CR, three achieved a partial response (PR), one patient had stable disease for a period of 2 months and four patients showed progressive disease (table 2). None of these patients had received anti-PD-1 therapy, since this was not available during enrollment of this study. Figure 1A shows a waterfall plot of best clinical response.

For three patients with an objective response to TIL therapy, the presence of neoantigen-specific T cell populations in the TIL product and in the peripheral blood compartment was analyzed and related to clinical course.

Patient 3 was a 46-year-old male with BRAF V600E mutation positive melanoma with in-transit metastases, iliac lymph node metastases, liver metastases and an enlarged left-sided adrenal gland. During prior treatment with vemurafenib, this patient experienced complete resolution of the liver metastases, but progression in lymph node metastases. Subsequently, he was treated with ipilimumab but showed progressive disease after four courses. At the time of study enrolment, this patient showed large in-transit metastases in his right upper leg, as well as iliac lymph node metastases. In addition, the patient had an enlarged adrenal gland, but no signs of liver, pulmonary or brain metastases. One of the in-transit metastases was resected for generation of the TIL product, and the patient received 19.6×10¹⁰ TIL, followed by four boluses of HD IL-2. Because of hypotension and low urinary output not responding to fluid challenges, the patient was on the intensive care unit (ICU) for 2 days requiring inotropic agents. He recovered from TIL treatment without any sequelae, and was in CR 12 weeks after TIL infusion. He remains disease free for more than 9 years after TIL therapy (see figure 1B).

Patient 4 was a female of 44 years of age with stage M1b BRAF V600E mutation positive melanoma with metastases in the left axilla and upper lobe of the left lung. Previously, this patient underwent axillary lymph node dissection and excision of the lung metastasis via video-assisted thoracic surgery. Five months later, she developed several subcutaneous metastases and lymphadenopathy in the left supraclavicular region. Treatment with dacarbazine showed progression after two courses, and treatment with ipilimumab showed progression after four courses. Subsequent treatment with vemurafenib led to a PR that lasted 4 months. This patient received 7×10¹⁰ TIL followed by two doses of IL-2. One month after TIL infusion, her disease was in partial remission. In the months thereafter, no further shrinkage of metastases was observed, and all remaining visible lesions (five soft tissue and one lymph node metastases) were at that point removed by surgery. Unfortunately, her disease recurred 4 months later, for which she again underwent resection. Another 6 months later, the patient had developed multiple brain metastases, and she died 16 months after TIL therapy.

Patient 8 was a 40-year-old male who was treated elsewhere for stage I and later stage III melanoma. He underwent inguinal lymph node dissection, with a recurrence in the same area 8 months later. These lesions were resected, and 5 months thereafter he developed retroperitoneal lymphadenopathy, for which he again underwent resection. Three months later, he also developed mediastinal lymph node metastases and was treated in an NY-ESO-1 protein vaccine trial. Six months later his disease had progressed. Following two courses of ipilimumab, he developed grade 3 colitis and treatment was halted. Another 6 months later, due to disease progression of peritoneal and mesenteric lymph node metastases, he was referred to our institute for TIL treatment. The patient received 14.7×10¹⁰ TIL, followed by one dose of IL-2. One month after TIL infusion, the mesenteric lymph node metastases showed a PR. Ten months after TIL infusion, the patient was in CR and currently has an ongoing CR for 7 years after therapy.

All toxicities observed were expected and in line with previous studies using the same treatment regimen. No treatment-related deaths were observed. Median time in hospital for all patients was 16 days (range 12–31). All patients underwent the scheduled lymphodepleting chemotherapy regimen. Anticipated hematological toxicities were seen in all of the treated patients. A grade 4 decrease in neutrophils was seen in all patients. Grade 3–4 decrease in platelets was seen in seven patients (70%) and grade 3–4 anemia was observed in three patients (30%). Patients received platelet and red blood cell transfusions as required based on standard protocols. Patients received on average 2.5 units of platelets (range 0–6) and 2.5 units of red blood cells (range 0–16). TIL infusion was mostly accompanied by low-grade side effects such as dyspnea, fever, nausea and chills. One patient experienced grade 3 dyspnea, possibly due to an allergic reaction, shortly after TIL infusion for which she was treated with budesonide inhalation and additional oxygen (4–5 L/min). Except for chemotherapy related grade 3–4 toxicities, no other grade 3–4 toxicities were observed during, or shortly after TIL infusion. All patients received HD bolus IL-2 (600 000 IU/kg) starting 4 hours after TIL transfusion. A median of 3.5 cycles of HD IL-2 was administered to the patients (range 1–9). None of the patient received all 15 cycles of HD IL-2. All patients experienced febrile neutropenia grade 3–4 during the HD IL-2 period. Other commonly seen toxicities during the HD IL-2 period were chills (eight patients), dyspnea (eight patients), tachycardia (six patients) and oliguria (four patients). All patients received Granulocyte colony-stimulating factor (G-CSF) injections once daily (5 µg/kg, with a maximum of 600 µg) starting 24 hours after TIL infusion, until neutrophil recovery. Patients received broad-spectrum antibiotics and antipyretic drugs (acetaminophen and occasionally indomethacin), according to local guidelines. Four patients (40%) had to be admitted to the ICU. Three patients due to oliguria and the need for inotropic agents, and one patient due to the appearance of a pneumothorax, probably caused by the insertion of a subclavian catheter, for which he required chest drain insertion. Most IL-2-related toxicities were transient and resolved within a few days after discontinuation of IL-2. Autoimmune adverse events were seen in one patient (patient 6). This patient possibly developed clinical symptoms reminiscent of Vogt-Koyanagi-Harada-like syndrome with an anterior uveitis and hearing loss several days after TIL infusion. She was treated with prednisone eye-drops with quick resolution of visual complaints. Treatment of patients with a lymphodepleting chemotherapy regimen, followed by the infusion of TIL and HD bolus IL-2 appeared safe, as no therapy-related deaths were seen. Furthermore, all grade 3–4 toxicities were readily managed on the internal medicine ward or ICU, and no long-term side effects were seen.

All TIL products complied with the preset specifications. The average number of cells in the infusion product was 12.8×10¹⁰ (range 1.7–19.6), with an average pre-REP culture time of 16.7 days (range 11–21). The viability of all samples was above 95%. All cryopreserved infusion products were phenotypically and functionally characterized on coincubation with autologous tumor digest (see table 2 for an extensive overview).

In most infusion products, the majority of cells consisted of CD8⁺ T cells, with a smaller fraction of CD4⁺ T cells. In all tested samples, in vitro reactivity of both CD4⁺ and CD8⁺ T cells against autologous tumor digest could be detected at varying levels (with IFN-γ levels between 0.3% and 19% of total CD8⁺ cells). In addition, tumor-induced production of other T cell effector cytokines, such as TNF-α and IL-2, was noted. The majority of CD8⁺ cells were single producers of IFN-γ, suggesting a more advanced differentiation state.32 However, a smaller fraction of T cells in all patients was shown to produce a combination of multiple effector cytokines in response to tumor cell encounter (see online supplementary figure 1). In previous trials using TIL treatment in melanoma patients, a correlation between TIL culture time, percentage of CD8⁺, percentage of CD8⁺/CD27⁺, number of IL-2 infusions and clinical benefit has been described.19 20 22 In the current cohort, no such correlations could be detected potentially due to the small sample size.

Having observed T cell reactivity against tumor cells in all evaluated TIL infusion products, we set out to determine the occurrence of T cell responses against patient-specific neoantigens, using a previously developed pipeline.7 30 In order to be able to relate expansion of neoantigen-specific T cells to tumor regression, the infusion products of the three patients who either developed a PR (patient 4) or CR on TIL therapy (patient 3 and 8) were analyzed. In all three patients, neoantigen-specific T cell reactivity could be detected in the TIL product (see table 3), but with a major variation in response breadth and magnitude. Specifically, in patient 3, CD8⁺ T cell reactivity against five different neoantigens was observed, with CD8⁺ T cells directed against neoantigen RBM12 S>L forming a major component (29% of total CD8⁺ cells) of the therapeutic cell product. While patient 8 also experienced prolonged and ongoing clinical benefit of TIL therapy, T cell reactivity in the cell product used for therapy was only detected for a single predicted neoantigen, and this response was of a low magnitude (0.17% of total CD8⁺ T cells). Notably, total tumor reactivity of this TIL product, as measured using functional assays, was substantially higher (15% of CD8⁺ T cells, see table 2), either indicating the existence of neoantigen-specific T cell reactivities that were missed by the pipeline used, or indicating that reactivities against non-mutant antigens may form a major component in this particular patient. Note that MHC multimer screens were performed for six out of six HLA-A alleles, five out of six HLA-B alleles and none of the HLA-C alleles expressed by these patients. In patient 4, three unique neoantigen-specific T cell populations could be detected in the infusion product with frequencies of 0.007, 1.14% and 3.35% of total CD8⁺ cells. The infusion product of this patient also contained neoantigen-specific CD4⁺ T cells, with a frequency of 3.8% of CD4⁺ T cells against a single neoantigen, as reported previously.30

Having identified neoantigen-specific T cell populations in the TIL products of all responding patients, we subsequently determined whether engraftment and expansion of the different neoantigen-reactive T cell pools could be observed (see figure 2 and table 4).

Analysis of pretreatment blood samples of the three different patients revealed the presence of only low levels of T cells against the identified neoantigens in peripheral blood before TIL infusion (Patient 3, RBM12 _S>L : 0.03% MHC multimer⁺ of CD8⁺, patient 4, TTC37 _A>V: 0.002% MHC multimer⁺ of CD8⁺, ENTPD4 _P>L: 0.6% MHC multimer⁺ of CD8⁺, patient 8, RAD51AP1-002 _S>F: 0.008% MHC multimer⁺ of CD8⁺). The other neoantigen-specific T cell populations present in TIL products were not detected in the pre-treatment blood compartment (detection level: 0.002% MHC multimer⁺ of CD8⁺ T cells and at least 10 events in flow cytometry plot).

To understand to what extent TIL therapy boosted the capacity of the peripheral blood compartment to recognize this set of neoantigens, peripheral blood samples were monitored at different time points post-therapy. Strikingly, for six out of the nine neoantigen-specific T cell reactivities analyzed, a substantial increase in frequency in post infusion blood samples could be observed. One of the nine responses, specific for the ENTPD4 _P>Lmutation identified in patient 4, remained stable in blood on therapy. The other two neoantigen-specific T cell reactivities, which were both found at very low levels in the infusion products (patient 4, MAB21L1_v>m and patient 3, LRP3 _T>S), could not be detected above background in peripheral blood samples before or after therapy, precluding any assessment of change in frequency. Table 4 provides an overview of the minimum fold expansion in individual neoantigen-specific T cell responses prior to therapy and post-therapy.

The neoantigen-specific T cell population directed against RBM12 _S>L, which had the highest frequency in the TIL infusion samples, also showed the highest expansion in the blood. At 6 months post-therapy, a 750-fold expansion compared with pre-TIL infusion was observed. T cells reactive against the RBM12 S>L epitope remained detectable up to 34 months postinfusion (the last time point analyzed), at a frequency of 2.2%. The observed expansion and contraction of this neoantigen-specific T cell response that coincided with clinical response provides indirect temporal evidence for a role of this T cell pool in cancer regression. The low level presence of these neoantigen-specific T cells almost 3 years after development of a complete clinical response may both reflect the formation of T cell memory, or the development of an equilibrium between a clinically undetectable reservoir of tumor cells and an active T cell response.

In patient 4 (PR), analysis of the kinetics of neoantigen-specific T cell responses post-TIL therapy demonstrates that T cell reactivities against different neoantigens in an individual patient can show a disparate behavior. In this patient, one neoantigen-specific CD8⁺ T cell response showed a substantial increase (17 fold) on therapy, whereas a second neoantigen-specific T cell response remained stable on TIL infusion (see figure 2B). Both responses were already detectable before infusion at low frequencies and could be detected at the latest time point (8 months post-therapy) for which material was available. Interestingly, the previously reported CD4 response in patient 4 remained detectable in blood during this 8 month follow-up as well.

The single neoantigen-specific CD8⁺ T cell response that was observed in the TIL product from patient 8 (CR) was detectable at a very low frequency in the peripheral blood compartment prior to TIL infusion (0.008% MHC multimer⁺ of total CD8⁺ T cells). Following TIL infusion, this T cell response increased over time and peaked at 4 months after infusion at a 25-fold higher frequency than pretreatment levels. In the subsequent years, this T cell response remained detectable in blood at 0.05%, revealing an expansion and contraction pattern similar to that observed in patient 3.

Collectively, these results demonstrate that for the majority of neoantigen-specific T cell responses (six out of nine) analyzed, TIL therapy leads to a sizeable increase in T cell reactivity in the peripheral blood compartment, with an observed increase in neoantigen-specific T cell numbers between 8-fold and 750-fold.

Phenotypic analysis of the neoantigen-specific T cells populations in the infusion product demonstrated low CCR7 expression and a mixture of CD45RA negative and positive cells, indicating a mix of effector and effector memory like T cells. Remarkably, this distribution of phenotypes did not change over time, toward a more central memory phenotype, when this neoantigen-specific population was followed in blood over several months postinfusion (online supplementary figure 2A,B and 3A,B, data presented for patient 4 and 8). In addition, we observed that PD-1 expression, as a marker for T cell activation, was higher on the neoantigen-specific T cell population as compared with total bulk CD8⁺ in peripheral blood, and its expression did not change over time when looking at the neoantigen-specific population in blood after infusion (online supplementary figures 2C and 3C, data presented for patient 4 and 8).