Combination Nivolumab, CD137 Agonism, and Adoptive Cell Therapy with Tumor-Infiltrating Lymphocytes for Patients with Metastatic Melanoma

Over the past 15 years, immunotherapy has become a first-line treatment option for metastatic melanoma, largely due to its relatively high immunogenicity. T cells with the ability to specifically recognize melanoma antigens are critical to this immune-mediated response. Adoptive cell therapy (ACT) with tumor-infiltrating lymphocytes (TIL) exploits an in situ population of enriched, tumor-specific T cells, removes them from a suppressive microenvironment, and induces their expansion ex vivo in favorable conditions to high numbers for reinfusion. TILs are able to traffic to tumors if administered after lymphodepleting chemotherapy and with IL2 resulting in objective response rates (ORR) of 30%–50% across various institutional series (1–11). Despite these impressive results, improvements in treatment outcomes are potentially achievable through the incorporation of in vitro and in vivo agents that could reduce TIL generation time, enhance tumor reactivity, and TIL persistence. Our previous work with the inclusion of 4-1BB agonism (anti-CD137 agonistic antibody) to IL2 during in vitro TIL generation has consistently demonstrated enhanced kinetics of proliferation and increased tumor-specific activity of TILs in both mouse and human preclinical models (3, 12). Therefore, 4-1BB agonism was included in the TIL generation for this trial.

Programmed cell death protein-1 (PD-1 or CD279) is an immune checkpoint receptor expressed by activated T cells. Extensive work has demonstrated that T cells expressing PD-1 include a subset of tumor-reactive lymphocytes (13–16). However, interaction between PD-1 and its ligands programmed cell death ligand-1 and -2 (PD-L1 and PD-L2) in peripheral tissues such as tumors and stromal cells leads to the suppression of these tumor-reactive T cells (17–19). In murine models, blockade of PD-1/PD-L1 interactions demonstrated enhanced T-cell proliferation, persistence, and antitumor activity (17–20). In a pivotal phase III clinical trial of anti–PD-1 antibody for patients with metastatic melanoma, there were statistically significant improvements in progression-free survival (HR, 0.58) and overall survival (OS) at one year (HR, 0.69) when compared with ipilimumab, leading to FDA approval of anti–PD-1 for the first-line treatment of metastatic melanoma (21). The replacement of exhausted clones within the tumor microenvironment with a renewed infiltrate of tumor-specific clones is essential to the therapeutic efficacy of PD-1 blockade (22).

The combination of anti–PD-1 and TILs has the potential to harness a potent pool of tumor-reactive T cells and overcome their inhibition in vivo. With this in mind, we combined the anti–PD-1 blocking antibody, nivolumab, with TIL therapy in anti–PD-1 and TIL treatment-naïve patients with metastatic melanoma. Each of two cohorts in this trial included anti–PD-1 dosing following ACT with TILs, while the second cohort also received anti–PD-1 prior to surgical tumor harvest and during the TIL generation period. We reasoned that dual therapy may be beneficial in terms of TIL persistence upon infusion, and that pretreatment with anti–PD-1 may increase T-cell infiltration as well as tumor specificity. In a prior trial at our institution, combination ipilimumab with TIL ACT maintained treatment efficacy similar to previously reported trials, while minimizing patient drop-outs during TIL generation (23). Our institution has also recently reported the successful combination of anti–PD-1 and TIL transfer in patients with anti–PD-1 refractory metastatic non–small cell lung cancer with an initial reduction in tumor burden for 85% of patients and a median change in target lesion size of -35.5% (24).

Herein, we report a phase I clinical trial in which nivolumab was administered in combination with TIL ACT. The safety and feasibility of this approach were the primary endpoints, with the expectation that these results would inform the recommended phase II schedule for combination treatment. We further hypothesized that utilizing nivolumab prior to surgical tumor harvest and during the interval required for TIL generation would increase T-cell trafficking to the tumor and potentially improve the quality of the TILs generated, while mitigating against tumor progression in the time interval between the harvest and infusion of the TILs. Initial TIL generation was supplemented with anti–4-1BB (CD137) agonistic antibody to improve the rate of successful cultures, growth kinetics, and CD8⁺ T-cell composition (3). The trial also assessed the potential for nivolumab to impact the persistence of TILs after transfer, a characteristic that is critical for achieving durable clinical responses.

Patients with unresectable stage III/IV metastatic melanoma, an ECOG status of 0–1, and no prior anti–PD-1 or TIL therapy were eligible for enrollment. Other prior treatment for metastatic melanoma was allowed if completed at least three weeks before enrollment. Patients were assessed to determine their ability to tolerate ACT plus lymphodepletion and IL2 and were required to have measurable residual disease after harvest of at least one lesion for TIL generation. In addition, patients were required to be 18 years or older with adequate organ function and a positive EBV antibody titer. Patients were excluded if they were pregnant or breast-feeding or had an active systemic infection, serologic evidence of syphilis, Hepatitis B or C, HTLV-I or II, or HIV. The trial was carried out in compliance with ethical guidelines laid out in the Declaration of Helsinki, the International Conference on Harmonized Tripartite Guidelines for Good Clinical Practice, and the US Common Rule. All eligible patients signed a written informed consent to the Institutional review board (IRB)-approved clinical trial protocol (NCT02652455).

Dose-limiting toxicity (DLT) was defined separately for each attributable element in the protocol regimen. DLT related to ACT was defined as any nonhematologic grade ≥4 adverse events which occurred upon or up to six weeks after ACT, consistent with our previous TIL trials (23, 25, 26). DLT related to nivolumab was defined as any grade ≥3 immune-related adverse events definitely attributable to nivolumab occurring within the same DLT window of up to six weeks after ACT, with the exception of rash that was required to be grade ≥4.

Surgically harvested tumor specimens were minced into 1–3 mm³ fragments for TIL generation under sterile conditions. Tumor fragments (36–48 per patient) were placed into 24-well plates with media containing 6000 IU/mL IL2 (aldesleukin, Prometheus Laboratories, Inc.), 10% human AB serum (Access Biologicals LLC), and 10  μg/mL anti–4-1BB monoclonal agonistic antibody (BMS-663513; Bristol Myers Squibb). TILs were expanded under these conditions for up to 5 weeks, keeping individual fragments separate. Each well was assessed for confluency 2–3 times per week, expanded to new wells at >80% confluency, and supplemented with the aforementioned media containing 1 μg/mL anti–4-1BB every 3–4 days.

Excess tumor material that was not used to generate TIL was physically and enzymatically digested to prepare a single-cell suspension of target tumor cells. After mincing, tumor material was gently agitated in media containing collagenase II, collagenase IV, hyaluronidase, and DNase (Thermo Fisher Scientific). The digested material was then filtered, red blood cells lysed, and viable cells counted and cryopreserved until utilization in functional assays.

Expanded TILs were cocultured with tumor cells to determine antitumor reactivity. Autologous tumor cells from the tumor digest and HLA-mismatched control tumor cells were utilized as targets. TIL and tumor cells were plated at a 1:1 ratio in 96-well plates (1×10⁵ cells/well each) overnight. Where indicated, MHC-I interaction was impeded via the HLA-ABC blocking antibody clone W6/32 (BioLegend, Inc.) at 10 μg/mL. Coculture supernatants were collected and assayed for IFNγ) release by ELISA (R&D Systems and BioLegend, Inc.). TILs that produced >200 pg/mL IFNγ and at least a two-fold increase over the HLA-mismatched control were designated as tumor-reactive.

TILs from tumor-reactive fragments were collected and enumerated. A maximum of 12 GREX 100M flasks (Wilson Wolf) were seeded with 5.0×10⁶ TILs each for a total of 6.0×10⁷ TILs at the initiation of the rapid expansion protocol (REP). REP media consisted of 200 mL AIM-V (Invitrogen) and 200 mL RPMI containing 10% human AB serum, 1% HEPES, and 0.1% 2-mercaptoethanol. TILs were cultured at a 1:200 ratio with irradiated feeder cells pooled from three healthy donors. REP media were supplemented with 30 ng/mL OKT3 (Ortho Biotech) and 3,000 IU/mL IL2. On day 4, an additional 200 mL of the above REP media were added to each flask and supplemented with 3,000 IU/mL IL2. On day 7, TILs were resuspended, counted, and expanded into a maximum of 36 GREX 100M flasks. Each flask was seeded with 1.0–7.5×10⁸ TILs and filled to 1 L total volume of AIM V and 3,000 IU/mL IL2 and cultured for an additional 7 days. No anti–4-1BB was included in the rapid expansion steps.

After the REP, TILs were harvested, washed, and concentrated to between 100 and 500 mL for infusion. Viable cells were quantified after staining with acridine orange and propidium iodine on the Cellometer Auto 2000 (Nexcelom Bioscience). Infusion product TILs were labeled with 7-AAD, CD45, CD3, CD4, and CD8 antibodies (see Supplementary Table S1) followed by acquisition on a BD FACSCanto (BD Biosciences) and analysis by FlowJo software (RRID:SCR_008520; TreeStar, Inc.). Prior to infusion, patients were lymphodepleted with cyclophosphamide (60 mg/kg/day) on day -7 and -6, followed by fludarabine (25 mg/m²) on days -5, -4, -3, -2, and -1. After sterility testing, the TIL product was infused intravenously by gravity drip into the patient at 300 mL/hour. For cohort 1, nivolumab (BMS-936558; Bristol Myers Squibb) was administered intravenously (3 mg/kg) beginning two weeks after the TIL transfer, continuing every 14 days for six months, then every 90 days for 18 months unless unacceptable toxicity or tumor progression supervened. For cohort 2, nivolumab was administered intravenously (3 mg/kg) every two weeks for two doses prior to TIL harvest then was continued every two weeks during TIL propagation. Nivolumab was stopped two weeks prior to the TIL transfer then resumed at the identical dosing schedule to cohort 1 after the TIL transfer.

Peripheral blood mononuclear cells (PBMC) were isolated by Ficoll-Hypaque (GE Healthcare Biosciences) gradient separation and cryopreserved from blood draws at indicated weekly time points. PBMC collection was continued in this manner every three months while patients remained on study. Apheresis was performed at the six-week time point and processed as above.

An aliquot of the infusion product was cryopreserved at the time the product was prepared for infusion. Upon recovery, aliquoted TILs were washed with PBS, counted, and stained for flow cytometric analysis. Viability staining was completed according to the manufacturer's protocol (see Table 1). Antibody surface staining occurred at 4°C for 30 minutes, protected from light, in PBS containing 5% FBS, 1 mmol/L EDTA, and 0.1% sodium azide. Where indicated, intracellular staining for FOXP3 was performed following the manufacturer's protocol (BD Biosciences #560098 or Thermo Fisher Scientific #552300). Cells were stained at concentrations between 0.25 and 1.0 ×10⁶ TILs/sample and fixed in 2% paraformaldehyde until acquisition on a BD LSRII or BD Celesta (BD Biosciences) instrument. FlowJo software was utilized for FACS analysis and gates were drawn according to fluorescence minus one (FMO) staining on a per patient basis. See Supplementary Table S1 for panel information.

Cryopreserved TILs and PBMC samples were thawed into prewarmed media and counted. Where indicated, CD4⁺ and CD8⁺ TIL populations were isolated by magnetic bead separation (Miltenyi Biotec) on MACS columns from bulk TIL products. DNA was extracted using the Qiagen DNeasy Blood and Tissue Kit (Qiagen) according to the kit's protocol. DNA was quantified via Nanodrop, and deep level TCRβ sequencing was performed by the Moffitt Cancer Center (MCC) Molecular Genomics Core using the ImmunoSEQ TCRβ Kit v3 (Adaptive Biotechnologies). Data were uploaded to the Adaptive Biotechnologies server for analysis on the Adaptive ImmunoSEQ Analyzer 3.0.

A patient-derived xenograft (PDX) tumor line was established in NSG mice from subcutaneously injected tumor digest from patient 7. PDX tumors were harvested, physically and enzymatically digested, and injected subcutaneously into NOD.Cg-Prkdcscid Il2rgtm1Sug Tg(CMV-IL2)4–2Jic/JicTac (hIL2 NOG; RRID:IMSR_TAC:13440) mice (Taconic Biosciences). When tumors reached approximately 20–30 mm², TILs were adoptively transferred via intravenous tail vein injection. Tumor area was monitored by calipers. All animal studies were approved by the Institutional Animal Care and Use Committee review board (R IS00006131 and R IS00010199).

Excess tumor tissue from resected specimens were collected under separate IRB-approved nontherapeutic research protocols (MCC50148, MCC50193, MCC50232, and MCC50326) after informed written consent and processed and cultured as above. These nontherapeutic studies were conducted under the ethical principles laid out by the Declaration of Helsinki and in accordance with the US Common Rule.

Up to six patients were accrued per cohort. The primary endpoints of the trial were to determine the safety and feasibility of combination nivolumab with ACT using TILs. Safety and feasibility were defined as the ability to successfully treat at least 67% of the patients in each cohort without reaching DLT. The secondary endpoints consisted of the objective response rate confirmed at 12 weeks after TIL transfer and progression-free and overall survival. Responses were determined by RECIST 1.1 criteria by comparing the baseline radiographic and clinical assessments to those at 12 weeks following ACT with TILs. All statistical analyses were completed using GraphPad Prism v9 (RRID:SCR_002798; GraphPad Software).

The data in this report are available within this manuscript and its supplement or available upon reasonable request from the corresponding author.

Eleven patients with unresectable stage III/IV melanoma were enrolled between March 2016 and May 2018 (Table 1) at a single institution to receive combination nivolumab and ACT with TILs expanded in the presence of anti–4-1BB agonism (NCT02652455). The first six patients were enrolled in cohort 1 (C1) and were to receive TIL infusion followed by nivolumab treatment beginning two weeks after ACT. Five subsequent patients entered cohort 2 (C2) and were planned for pretreatment with nivolumab prior to tumor harvest and during TIL generation, in addition to the C1 regimen of nivolumab after ACT (Fig. 1A). Of these 11 patients, five were females and six were males, with a median age of 52 (range 37–68 years). The majority had no prior therapy, while four had previous adjuvant therapy and two each had received either interferon, ipilimumab, and/or combination BRAF/MEK–targeted therapy (Table 1). All patients were included in the intent-to-treat analysis of response; however, three patients were excluded from correlative analyses due to lack of specimen availability. Patient 2 was not treated with TILs due to progression with new brain metastases prior to lymphodepletion. Of note, this patient also had liver metastases and markedly elevated lactate dehydrogenase (LDH). Patient 5 was removed from all subsequent correlative analyses for noncompliance due to a lack of follow-up prior to first post-TIL imaging assessment. Patient 9 elected to forego surgical tumor harvest for TIL propagation due to a response to nivolumab prior to harvest.

The primary endpoints of safety and feasibility were met, because at least four patients (≥67%) per cohort were successfully treated with ACT without reaching DLT. Patients received a median of eight doses of nivolumab and six doses of IL2. Nonhematologic adverse events were primarily grade 3 and were largely attributed to either the cyclophosphamide/fludarabine lymphodepleting regimen or the IL2 and TIL transfer (Supplementary Table S1). One patient presented with grade 4 elevated lipase which was attributed to the cyclophosphamide/fludarabine preconditioning and did not alter the course of treatment. Two grade 3 toxicities definitely attributable to nivolumab occurred in a single patient. These events comprised grade 3 colitis that occurred outside of the six-week DLT window, and grade 3 rash that did not reach the grade ≥4 severity to be deemed a DLT.

Patient responses at six weeks after TIL infusion were confirmed at 12 weeks and constituted the secondary endpoint for the trial. The ORR was 36% with four patients achieving a partial response. Median OS was 23 months (both cohorts 23 months), while PFS was 5 months (C1: 5 months, C2: 12.5 months; Fig. 1B). Each of the three responding patients who received TILs had responses lasting more than 18 months, with two responses ongoing (46+ and 71+ months) at the time of publication (Fig. 1C). In total, 69% of the monitored target lesions decreased in size from baseline measurements after ACT, including three target lesions from a nonresponder (Fig. 1D and E).

There were no manufacturing failures, as TILs were successfully expanded from all patient samples utilizing media containing IL2 and anti–4-1BB. Aggregating all specimen samples, 59% of plated tumor fragments successfully generated TILs (Fig. 2A). After initial “pre-REP” outgrowth, TILs underwent REP and produced an average of 1,435-fold expansion, with no observed difference between cohorts (Fig. 2B). An average of 8.5×10¹⁰ TILs were infused per patient (range: 5.3×10¹⁰–1.3×10¹¹; Fig. 2C). TIL generation with combination anti–4-1BB and IL2 was compared with IL2 alone utilizing a separate set of fragments from trial-derived excess surgical specimens not utilized for treatment. Similar efficiency was observed between these two groups, although the majority of paired samples (5 of 6) attained an increase in the fraction of fragments that yielded TILs in the presence of anti–4-1BB compared with IL2 alone (Fig. 2D).

TIL products were composed primarily of CD8⁺ T cells at infusion (Fig. 2E). Of the CD4⁺ TILs, only a small fraction was CD25⁺ CD127⁻ FOXP3⁺ T cells, which were found to be statistically significantly enriched in the TIL product of nonresponders, as the three responders all had lower values than did the five nonresponders (P = 0.0357; Fig. 2F). When stratified by response, further phenotypic evaluation of the infusion product indicated that both CD8⁺ and CD4⁺ TILs from responders displayed increased LAG3 (CD223) and TIGIT surface expression (Fig. 2G). Infused TILs were predominantly effector memory (T_EM) or terminal effector (T_E) T cells, and we observed that responders were enriched for central memory T cells (T_CM) when compared with nonresponders (Supplementary Fig. S1). Tumor reactivity was detected from at least one tested fragment from each patient, except for Patient 11 who still achieved a partial response (Fig. 2H). Collectively, a mean of 39% of all tested fragments produced tumor-reactive TILs (C1: 50% vs. C2: 32%). Furthermore, TILs produced IFNγ in response to HLA-matched tumors, indicating that T-cell clones recognizing shared tumor antigens may be present within the infusion product.

To further characterize the infused TILs, TCRβ sequencing was performed. TIL products were similar in clonality between cohorts and between responders and nonresponders (Fig. 3A). Persistence was relatively high across all patients and trended toward increased values in C2 (Fig. 3B) with a greater ratio of expanded to contracted clones than their C1 counterparts (Fig. 3C and D). For each of the three long-term responders who received ACT, additional sequencing was performed from PBMCs isolated throughout the duration of response. The top ten most prevalent clones accounted for between 35% (patient 11) and 85% (patient 10) of the infusion product in these patients. Over 1 year after infusion (day 438 or day 452), these TIL clones were still detectable at substantial levels measuring 3%–4% of the entire peripheral blood repertoire. When this analysis was expanded to include the top 25 clones from each sample, greater than 50% of the TCRβ sequences originated in the TIL infusion product for the two patients in C2. The C1 long-term responder (patient 1) maintained 25% of clones in these same parameters two years following infusion (Fig. 3E and F).

We further investigated patient 7 (C2) as a case study as the clinical course of this patient included multiple recurrent lesions following ACT (Fig. 4A). The initial tumor (iTumor) of the patient generated an initial TIL (iTIL) product from fragments at a relatively high rate (Fig. 2H), and seven of these fragments were selected for REP based on IFNγ production in response to autologous tumor. The iTIL infusion product was predominantly CD8⁺ T cells (85.8%; Fig. 4A). After ACT, the patient received five additional doses of nivolumab and achieved a status of stable disease at 12 weeks. Each of the three target lesions regressed by greater than 30% from baseline (Fig. 1D). Despite this, the patient ultimately progressed after 5 months, manifested by two breakthrough brain metastases. These recurrences (rTumor1 and rTumor2) were acquired by surgical extirpation and TILs (rTIL1 and rTIL2, respectively) were expanded for nontherapeutic research purposes (Fig. 4A).

After 6 weeks, the iTILs were persistent in the periphery (89.8% TCRβ overlap) and similar in composition to the infusion product. However, substantially less clonal overlap was observed between iTILs and TILs generated from either of the brain metastases (Fig. 4B). Further analysis confirmed that the majority of iTIL clones were deleted between the iTILs and rTIL1 or rTIL2 (Fig. 4C). Analysis of the top ten iTIL clones, which comprised greater than 40% of the infused TIL, confirmed that each of these clones were excluded from the infiltrate in both of the recurrent lesions (Fig. 4D), indicating a complete loss of the dominant TIL clones originating from the infusion product. To further support this, we analyzed the top 25 shared clones within each TIL population. The majority of clones within the rTIL1 (73%) and rTIL2 (94%) repertoire originated independently of the infused TIL product, while only 2% of clones were shared across all three surgical specimens (Fig. 4E).

Further ex vivo analysis demonstrated that while the iTILs were highly responsive to the patient's iTumor (Fig. 2H), iTIL reactivity towards rTumor1 was substantially impaired both in PDX murine models and the corresponding in vitro coculture (Fig. 4G). In contrast, rTIL1 efficiently responded to rTumor1 in these PDX murine models (Fig. 4F and G), despite inability to control metastatic outgrowth in the patient. The ability of the iTILs to recognize the iTumor coupled with an inferior response of the iTILs toward the recurrent lesions is consistent with the patient's clinical trajectory and lack of response to ACT. Altogether, these data support a shift in the clonal distribution of the T-cell repertoire between the initial harvested specimen and subsequent recurrences.

Patient 8, another member of C2, received iTILs in combination with nivolumab treatment but recurred with a brain lesion after ACT. The brain lesion was resected and analyzed after surgical extirpation (Fig. 5A). The iTILs from the patient were efficiently generated from 79% of fragments and 92% of the tested fragments contained tumor-reactive TILs (Fig. 2H). This predominantly tumor-reactive infusion product primarily consisted of CD8⁺ T cells that were highly persistent within the circulation of the patient (Fig. 5B). This patient had stable disease at the 12-week time point before ultimately progressing and requiring additional surgical intervention at 15 weeks (rTumor). Again, rTILs were expanded for nontherapeutic research purposes from the rTumor sample and found to be CD8 predominant. Strong overlap (60%) in the TCRβ sequences between iTILs and rTILs suggested persistence and infiltration of the infused product into the recurrent brain metastasis (Fig. 5B). While many individual clones were not maintained, the two most prevalent clones in the iTIL population demonstrated a robust presence in the rTIL population, while steadily declining in the periphery (Fig. 5C and D). Furthermore, 67% of the clonal repertoire present in the rTIL population were shared with the top 25 clones among the infused TIL product (Fig. 5E), indicating a high degree of clonal continuity after ACT.

To investigate the function of these TILs, in vitro cocultures with tumor were performed. The iTIL product demonstrated high IFNγ production in response to both the iTumor and rTumor, indicating shared antitumor recognition between lesions. Only 30% of rTIL fragments produced low levels of IFNγ in response to the iTumor, and just 10% of rTIL fragments released IFNγ in response to rTumor (Fig. 5F and G), suggesting an intrinsic loss of rTIL function despite the substantial clonal overlap. Flow cytometric profiling of co-stimulatory receptor expression indicated that rTIL expressed diminished levels of 4-1BB on both CD8⁺ and CD4⁺ T cells as well as reduced TIM3 and OX40 on CD4⁺ and CD8⁺ T cells, respectively. In totality, these data support substantial tumor infiltration and persistence of tumor-reactive iTILs; however, the rTILs were marked by an inferior costimulatory profile suggesting these TILs may be lacking critical intrinsic activation receptors.

With these observed deficiencies in TILs isolated from recurrent metastases following treatment in patients 7 and 8, we sought to further investigate TIL quality and strategies to improve TIL production in the context of disease progression after prior anti–PD-1 therapy. In Fig. 2D, we found that the addition of anti–4-1BB to TIL cultures from anti–PD-1 naïve patients on the clinical trial resulted in a trend toward improved TIL generation, consistent with our previous work (3). From a separate group of patients with metastatic melanoma outside of the clinical trial, we cultured tumor fragments in media containing IL2 with or without supplemental anti–4-1BB. Prior progression on anti–PD-1 therapy had a clear negative effect on ex vivo TIL expansion with IL2 alone. The addition of agonistic anti–4-1BB plus IL2 to TIL cultures from patients who had relapsed on prior PD-1 blockade significantly improved TIL generation (P < 0.05; Fig. 6A). When patients in this category were stratified, the addition of agonistic anti–4-1BB resulted in a statistically significantly greater frequency of fragment expansion for patients who had failed a single course of anti–PD-1, while only a trend in this direction was observed for failure of multiple courses of anti–PD-1 therapy (Fig. 6B).

With IL2 alone, only 27% of tumor fragments from specimens previously treated with anti–PD-1 successfully generated TIL. In paired melanoma specimens, the addition of agonistic anti–4-1BB plus IL2 increased TIL culture success rate on a per fragment basis and improved TIL fold expansion (Fig. 6C and D). Impressively, of the specimens where TILs were not generated with single-agent IL2, the addition of anti–4-1BB rescued TIL production for 64% of these tumors (Fig. 6E). These data suggest that the addition of anti–4-1BB may be critical for future clinical TIL production as patients are treated with checkpoint blockade as first-line therapy with consideration of TIL therapy at the time of progression after checkpoint blockade.

We completed a phase I clinical trial for patients with anti–PD-1 naïve metastatic melanoma investigating the combination of anti–PD-1 and ACT with TILs generated in the presence of agonistic anti–4-1BB in vitro. The primary endpoint of this study was met as the treatment was determined to be both safe and feasible. Adverse events were primarily related to the lymphodepletion regimen and IL2 administered immediately following ACT. This study enrolled 11 patients who achieved an ORR of 36% and observed a decrease in 69% of all target lesions (1, 23–27). Three of these patients achieved a partial response lasting longer than one year following combination ACT and nivolumab therapy, while one additional patient responded to nivolumab preceding TIL treatment that was ultimately not administered.

Current timelines of TIL production can include up to eight weeks between tumor harvest and TIL administration, a substantial gap during which patients suffering from already advanced disease are susceptible to further tumor progression and clinical deterioration preventing scheduled ACT. Anti–PD-1 monotherapy is effective at reducing tumor burden in patients with metastatic melanoma, which may render a patient's tumor burden more amenable to TIL therapy (28). On the basis of prior studies, we hypothesized that pretreatment with nivolumab would curb patient progression during TIL generation and allow more patients to complete treatment with ACT (23). To this end, all five patients enrolled in C2 received TIL transfer, except Patient 9 who achieved a response on nivolumab monotherapy and elected not to complete ACT, while one of six patients in C1 progressed prior to ACT. The degree to which anti–PD-1 treatment alone contributed to patient responses in this trial was not specifically investigated; however, anti–PD-1 pretreatment provided patients the potential opportunity to respond prior to ACT in C2, as with Patient 9. Furthermore, we observed a decreased tumor burden in two additional patients in C2 with anti–PD-1 pretreatment between baseline and ACT. Responses were evaluated at 12 weeks following ACT by intent-to-treat analysis. The response rate in this study is similar to previously reported response rates involving TILs, including PD-1 refractory patients (23, 25, 26, 29). While the relatively small number of patients is a limitation of this study, these individual cases support the notion that anti–PD-1 pretreatment alleviated patient attrition when combined with ACT, and the intervening use of lymphodepleting chemotherapy following anti–PD-1 pretreatment did not appear to have a deleterious effect on subsequent tumor responses.

PD-1 is expressed by T cells enriched for antigen-specific reactivity and serves to negatively regulate T-cell function upon ligand interaction (14–16). We hypothesized that combination PD-1 blockade with ACT might improve treatment efficacy upon TIL infusion by preventing ligand engagement during antitumor responses to sustain T-cell responses. TILs from responders expressed elevated LAG3 and TIGIT, negative regulators of T-cell function induced after activation, suggesting these may be additional indicators of antigen-specific T cells (30–32). These represent possible targets for additional therapeutic intervention, especially with the recent FDA approval of an anti-LAG3 antibody in combination with nivolumab (33). In addition, infused TILs had high overall persistence in the peripheral blood of patients across both cohorts at six weeks after transfer and responders were enriched for central memory CD8⁺ T cells. In long-term responders, these enriched TIL clones were detected over one year after infusion and clones present within the TIL product comprised more than 25% of the peripheral repertoire in each of these patients, consistent with previous TIL trials that have correlated persistence with patient response (1, 34). Combination checkpoint blockade/ACT in this study led to durable responses, characteristic of ACT responses in other reports (1, 23, 25, 26, 35). The overall TIL persistence indicated that even nonresponders may have an altered immune landscape and could potentially be poised to benefit from additional immunotherapeutic options.

We further hypothesized that enriching for tumor-reactive T-cell clones would improve the quality of TILs available for infusion. To address this, we supplemented our ex vivo TIL cultures with anti–4-1BB. Previous studies demonstrated that the addition of anti–4-1BB antibody during TIL generation improved TIL yield and skewed the TIL population toward tumor-reactive CD8⁺ T cells (3, 36, 37). In this trial, substantially more TILs were infused in comparison to previous TIL ACT trials at our center, likely due to improved growth kinetics with addition of anti–4-1BB antibody (23, 25, 26). A paired comparison of TIL generation with and without supplemental anti–4-1BB demonstrated an increase in productive TIL fragments for five of six patients, including rescue of TIL expansion in one patient who failed to produce TILs with single-agent IL2. In addition, all eight patient samples produced tumor-reactive TILs upon in vitro coculture, with the exception of Patient 11. Despite this, Patient 11 achieved a partial response, which is still ongoing at the time of publication, after ACT with predominantly CD4⁺ TILs. We cannot exclude that this response was driven by nivolumab; however, it is possible that antigen-specific TILs were present but not detected in our assays. For example, CD4⁺ TIL antitumor reactivity is restricted by the ability of the tumor target to efficiently express MHC class II. This limitation often precludes the ability to accurately assess CD4⁺ TIL functionality, representing an unmet need in the field of tumor immunology and an active area of research in our laboratory.

We also investigated two patients who progressed with brain metastases following ACT, to better understand the mechanisms underlying therapeutic resistance after TIL transfer. We were able to efficiently generate TILs from each resected brain metastasis, indicating T-cell infiltration occurs in brain metastasis relapses, and that subsequent ex vivo expansion of TILs from these lesions is feasible. In patient 7, low TCRβ overlap between the iTILs and rTILs suggested inefficient trafficking or persistence of the iTILs at those novel sites, or reduced subsequent ex vivo expansion during rTIL generation. This shift in repertoire coincided with TIL antitumor recognition that was limited to the respective tumor origin of the TILs. These data support the recent report from Yost and colleagues describing clonal replacement following anti–PD-1 therapy but also may indicate a potential antigenic shift between the original target lesion and subsequent recurrences (22). Because of limited quality of the recovered iTumor specimen, we were unable to accurately compare the neoantigen mutanome between specimens, as well as other potential contributing mechanisms of therapeutic resistance, such as HLA loss and immune receptor ligand expression.

In contrast, analysis of Patient 8 demonstrated substantial overlap in the clonal profile between iTILs and rTILs, suggesting effective iTIL trafficking and infiltration into the relapsed tumor site. Despite successful infiltration, iTILs failed to mediate an antitumor response at this lesion indicating active tumor resistance to T cell–mediated elimination. Additional in vitro analysis demonstrated minimal recognition of autologous tumor by rTILs coupled with a deficient costimulatory profile, a known mechanism of immunotherapeutic resistance (38–43). Given that additional resistance mechanisms are possible, future analyses will aim to compare the transcriptome of iTILs to TILs recovered from those at sites of progression as well as to determine evidence of immunoediting in ACT. These case studies present opportunities to inform future trial designs to investigate additional monitoring and interventional strategies to elucidate the mechanisms and context underlying the efficacy of and resistance to ACT.

As anti–PD-1 therapy has become a widespread frontline treatment, including its use in the adjuvant setting, there are fewer patients eligible for TIL therapy who are anti–PD-1 treatment naïve. We performed a larger analysis with a separate group of metastatic melanoma patients outside of the clinical trial to determine the effects of disease progression from prior anti–PD-1 on TIL expansion. We found that surgical specimens resected from patients who had previously failed anti–PD-1 therapy generated substantially fewer TILs compared with anti–PD-1 treatment-naïve specimens. Agonistic anti–4-1BB antibody plus IL2 proved to be an effective strategy to rescue TIL expansion ex vivo following disease progression on anti–PD-1, particularly in samples unable to produce TILs when utilizing single-agent IL2. Going forward, this study underscores the importance of developing improved methods for TIL generation, especially in the context of disease progression after prior immunotherapy, although it has been previously shown that PD-1 refractory patients can be successfully treated with TILs (29).

Overall, combination therapy of ACT with TILs expanded ex vivo using anti–4-1BB along with systemically administered anti–PD-1, was found to be a safe and feasible treatment for patients with metastatic melanoma. This study represents the first peer reviewed report to our knowledge of TIL infusion in combination with PD-1 blockade in the anti–PD-1–naïve setting. Forthcoming ACT trials for melanoma will predominantly enroll patients who have progressed after anti–PD-1, and thus may necessitate additional combination strategies to enhance TIL generation such as anti–4-1BB used in this report.

M.S. Hall reports grants from Swim Across America, Iovance Biotechnologies, American Cancer Society, and National Cancer Institute during the conduct of the study; in addition, M.S. Hall has a patent for expansion of TIL pending, licensed, and with royalties paid from Iovance Biotherapeutics and reports common stock holdings in AbbVie, Inc., Amgen, Inc., BioHaven Pharmaceuticals, and Bristol Myers Squibb. J.E. Mullinax reports grants, personal fees, and other support from Iovance Biotherapeutics, grants from SQZ Biotech and Intellia Biotherapeutics, and personal fees from Merit Medical outside the submitted work; in addition, J.E. Mullinax has a patent for 18MB056, 21MA030 pending, licensed, and with royalties paid from Iovance Biotherapeutics and reports research support unrelated to this research from NIH-NCI (K08CA252642), Ocala Royal Dames, and V Foundation. A.M. Hall reports grants from Swim Across America, Iovance Biotechnologies, and American Cancer Society during the conduct of the study; in addition, A.M. Hall has a patent for expansion of TIL pending, licensed, and with royalties paid from Iovance Biotherapeutics and reports common stock holdings in AbbVie, Inc., Amgen, Inc., BioHaven Pharmaceuticals, and Bristol Myers Squibb. M.S. Beatty reports other support from Iovance Biotherapeutics during the conduct of the study. J. Blauvelt reports grants from Swim Across America, Iovance Biotechnologies, and American Cancer Society during the conduct of the study. H. Branthoover reports grants from Swim Across America, Iovance Biotechnologies, and American Cancer Society during the conduct of the study. A.D. Richards reports other support from Iovance Biotherapeutics, Swim Across America, Bristol Myers Squibb, and Prometheus during the conduct of the study. J.K. Teer reports grants from NCI during the conduct of the study as well as grants from Turnstone Biologics outside the submitted work; in addition, J.K. Teer has a patent for negative information storage model issued. N.I. Khushalani reports grants from Iovance Biotherapeutics and other support from Bristol-Myers Squibb and Prometheus during the conduct of the study as well as grants, personal fees, and other support from Bristol-Myers Squibb, Regeneron, and Replimmune; grants and personal fees from Merck, Novartis, and HUYA Bioscience; grants from Celgene, GlaxoSmithKline, and Modulation Therapeutics; personal fees from Instil Bio, Castle Biosciences, Genzyme, and Iovance Therapeutics; personal fees and other support from Nektar; and other support from Incyte, AstraZeneca, NCCN (from Pfizer), Bellicum Pharmaceuticals, Amarin Corporation, and Asensus Surgical outside the submitted work. J.S. Weber reports personal fees from BMS during the conduct of the study as well as personal fees from Instil Bio outside the submitted work; in addition, J.S. Weber has a patent from Iovance issued. V.K. Sondak reports other support from Iovance and Bristol Myers Squibb during the conduct of the study as well as personal fees from Iovance, Bristol Myers Squibb, Merck, Alkermes, Eisai, Genesis Drug Discovery & Development, Novartis, Regeneron, and Ultimovacs and grants from Neogene Therapeutics, Skyline DX, and Turnstone outside the submitted work. S. Pilon-Thomas reports grants from Swim Across America, Iovance Biotechnologies, and American Cancer Society and nonfinancial support from Bristol Myers Squibb during the conduct of the study as well as grants from Intellia Therapeutics, Provectus Biopharmaceuticals, Turnstone Biologics, Celgene, and Dyve Biosciences and personal fees from Seagen Inc. and KSQ Therapeutics outside the submitted work; in addition, S. Pilon-Thomas has a patent for expansion of TIL pending, licensed, and with royalties paid from Iovance, and Moffitt has also licensed IP to Tuhura Biopharma that is not related to this research. S. Pilon-Thomas is listed as a co-inventor on a patent application with Provectus Biopharmaceuticals that is not related to this research. A.A. Sarnaik reports grants, personal fees, and nonfinancial support from Iovance Biotherapeutics, nonfinancial support from Bristol Myers Squibb and Prometheus, and grants from Swim Across America during the conduct of the study as well as personal fees from Guidepoint, Defined Health, Huron Consulting Group, KeyQuest Health, Istari, Rising Tide Foundation, Boxer Capital, Gerson Lehrman Group, Society for Immunotherapy of Cancer, Physicians' Educational Resource, Medscape, and Medstar Health outside the submitted work. In addition, A.A. Sarnaik has a patent for 61/973,002 issued, licensed, and with royalties paid from Iovance Biotherapeutics; a patent for 62/612,915 issued, licensed, and with royalties paid from Iovance Biotherapeutics; a patent for 14/974,357 issued; a patent for 16/157,174 issued; and a patent for 17/279,327 issued. No disclosures were reported by the other authors.

M.S. Hall: Conceptualization, formal analysis, funding acquisition, investigation, visualization, methodology, writing–original draft, writing–review and editing. J.E. Mullinax: Conceptualization, funding acquisition, methodology, writing–original draft, writing–review and editing. C.A. Cox: Conceptualization, resources, supervision, validation, methodology, writing–review and editing. A.M. Hall: Resources, project administration, writing–review and editing. M.S. Beatty: Conceptualization, formal analysis, validation, visualization. J. Blauvelt: Resources, investigation. P. Innamarato: Conceptualization, investigation, writing–review and editing. L. Nagle: Investigation. H. Branthoover: Investigation. D. Wiener: Investigation. B. Schachner: Investigation. A.J. Martinez: Formal analysis. A.D. Richards: Resources, validation, investigation, project administration. C.J. Rich: Resources, investigation, project administration. M. Colón Colón: Resources, investigation, project administration. M.J. Schell: Formal analysis, validation, methodology, writing–review and editing. J.K. Teer: Software, methodology, writing–review and editing. N.I. Khushalani: Conceptualization, supervision, methodology, writing–review and editing. J.S. Weber: Conceptualization, supervision, methodology, writing–review and editing. J.J. Mulé: Conceptualization, resources, supervision, funding acquisition, methodology, writing–review and editing. V.K. Sondak: Conceptualization, supervision, methodology, writing–review and editing. S. Pilon-Thomas: Conceptualization, resources, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. A.A. Sarnaik: Conceptualization, resources, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing.

This work was funded by Swim Across America, Iovance Biotechnologies, and the Dr. Miriam and Sheldon G. Adelson Medical Research Foundation. The anti–4-1BB antibody was generously provided by Bristol Myers Squibb, as well as additional support. S. Pilon-Thomas was supported by an American Cancer Society–Leo and Anne Albert Charitable Foundation Research Scholar Grant (RSG-16- 117–01-LIB). A.A. Sarnaik was supported by NCI-5K23CA178083. J.E. Mullinax was supported by NIH-NCI (K08CA252642). M.S. Hall was supported by NCI-F31CA250320. Research reported in this publication was supported by the National Cancer Institute of the National Institutes of Health under Award Number F31CA250320. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. This work was supported in part by the Cell Therapies Core, Tissue Core, Flow Cytometry Core, and Molecular Genomics Core Facilities at the Moffitt Cancer Center. This work was also supported in part by the Cancer Center Support Grant P30 CA076292 from the National Cancer Institute. The authors would like to thank Autumn Joerger, Patrick Verdugo, Shayna Smeltzer, Zachary Sannasardo, Brittany Bunch, PhD, and Nermin Gerges, PhD, for their assistance with this work.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.