Tumor-Infiltrating Lymphocyte Therapy for Melanoma and Other Solid Tumors: Looking Back, Yet Moving Forward

Abstract

Lifileucel, the first solid tumor adoptive tumor infiltrating lymphocyte (TIL) therapy product to receive regulatory approval in advanced melanoma, represents a critical achievement in the pursuit of improving outcomes using cellular therapies in patients with solid tumors. This review traces the development of adoptive TIL therapy from the initial human studies in melanoma, through recent advances in studies of other solid tumors, and previews ongoing and future areas for preclinical and clinical advances to improve upon this novel therapeutic strategy.

INTRODUCTION

In 1987, the Rosenberg Laboratory at the National Cancer Institute (NCI) in Bethesda, Maryland reported that autologous, tumor-infiltrating lymphocytes (TILs) from six patients with melanoma could be cultured and replicated in vitro with recombinant interleukin-2 (IL-2) for use in adoptive immunotherapy [1]. Since this initial report nearly 40 years ago, dozens of clinical trials have manipulated TILs for the treatment of patients with solid tumors. The recent milestone US Food and Drug Administration (FDA) approval of lifileucel for treatment of melanomas resistant to checkpoint inhibition, which follows the basic pioneering work from the NCI, serves as an important proof of concept that adoptive TIL therapy can be a life-saving tool for patients with advanced solid tumors. Dozens of trials are currently using adoptive TIL therapy on an ongoing basis. This is a review of the early clinical development of TIL therapy in some detail, identifying clinical and translational lessons learned to date as well as opportunities for better translational understanding, and thus looks forward to various ways adoptive TIL therapy is being refined and reimagined for next-generation cellular therapy products in solid tumors.

EARLY MELANOMA CLINICAL TRIALS

The first large case series utilizing adoptive TILs in patients with melanoma treated 86 patients at the NCI [2]. This autologous TIL process did not screen for specific antitumor reactivity, although half of the TIL isolates showed at least modest lysis against autologous tumor. The process was somewhat lengthy, taking a median of approximately 5 weeks and >6 weeks in 40% of patients. Twenty-nine patients did not receive any conditioning chemotherapy, and 59 patients received 1 dose of cyclophosphamide 25 mg/kg. Patients received a high number of TILs by today’s standards (>80% received >100 billion cells) and the mixture was somewhat heterogeneous, with most having a majority of CD8⁺ cells but a minority having large CD4⁺ or natural killer (NK) cell populations. Following cell infusion, patients were treated with an IL-2 regimen that is largely repeated across all subsequent adoptive TIL trials (unless noted otherwise), with 720,000 IU/kg administered every 8 hours for a maximum of 15 doses. In this trial, a remarkable objective response rate (ORR) of 34% was demonstrated, with no significant difference between those who received low-dose cyclophosphamide and those who did not (35% vs. 31%). However, all but 4 of these objective responses were partial responses (PRs), and the median duration of partial response was only 4 months. Responses were not associated with cell counts of any type (total CD3, CD8, CD4) but they were associated with a quicker doubling time (2.3 vs. 3.3 days), shorter median time in culture (33 vs. 43 days), and higher rates of lysis of autologous melanoma cells (25% vs. 10%). Toxicities were significant but manageable and reversible, and consistent with the expected toxicities of IL-2 and cyclophosphamide. One patient died of sepsis in this initial cohort.

Another early attempt to improve the response rate and duration of adoptive TIL therapy in melanoma took the approach of treating 13 patients with HLA-A 02:01 alleles using cloned tumor-reactive CD8⁺ cells specific for a known immuno-reactive melanoma peptide derived from the differentiation antigen gp100 [3]. To generate the high numbers of CD8⁺ cells, they used 3 replication cycles with OKT-3 and IL-2 in vitro. After TIL administration, IL-2 720,000 IU/kg was used, but importantly, no NMA-LD regimen was utilized. Disappointingly, no responses were identified, and because the infused cell reactivity was known, they proved that no infused cell product persisted beyond 2 weeks. Together with the insights from the responses from unselected TIL on the prior case series and the discovery of CD4⁺CD25⁺Foxp3⁺ regulatory T cells, the authors determined that nonmyeloablative lymphodepletion (NMA-LD) regimens are instrumental in manipulating the host environment to allow for proper propagation of the cell product and durable antitumor effect. They also wrestled with the inherent trade-off of early adoptive TIL trials: is it better to spend more time culturing more specific cytotoxic TILs or spend less time culturing more polyclonal TILs?

The first trial utilizing the NMA-LD regimen used to this day in standard lifileucel therapy—cyclophosphamide 60 mg/kg ×2 days on days −7 and −6, followed by fludarabine 25 mg/m² ×5 days on days −5 to −1—was tested in a multicohort phase 1 study of only 15 patients with HLA-A 02:01 allele, of which only 12 received NMA-LD [4]. The doses of IL-2 were also modified from 0 to 12 doses of 720,000 IU/kg. This cell product was also highly selected, with CD8⁺ peripheral blood T cell clones or TILs specifically recognizing MART-1 or gp100 peptides replicated in 3 in vitro cycles. With NMA-LD and what was likely a very highly differentiated cell product lacking naïve memory cells, only transient shrinkage and mixed responses could be achieved. These cells also did not persist in the only 2 patients that could be studied, neither of whom had clinical benefit.

The now-familiar NMA-LD, TIL infusion, and IL-2 administration regimen was utilized to its first great success in 2002, when 6 of 13 patients with advanced melanoma, the majority of whom were refractory to both IL-2 and platinum chemotherapy, achieved objective responses to therapy [5]. This TIL population was polyclonal and almost all utilized only 1 replication cycle after initial expansion, unlike the highly selected CD8⁺ peripheral blood mononuclear cell (PBMC) efforts that required 3 replications. This TIL product did select clones to expand based on reactivity to either autologous tumors or known melanoma cell lines. A median of 78 billion cells were infused (range: 23–153 billion) and a median of 9 IL-2 doses was administered. Responses were partial but largely durable using this NMA-LD and IL-2 combination, with 2 ongoing past 1 year at initial publication. They identified persistent clones underlying this response using bulk T cell receptor (TCR) beta sequencing using antibody testing. Of 6 tested patients, they identified persistent TCR-beta skewing of 1 or 2 distinct clones in the 2 ongoing responders that were not present at baseline.

One underemphasized aspect of TIL therapy in this early era was its inefficiency. These initial trials did not report CONSORT diagrams. The number of harvests that did not lead to viable TIL products was not routinely reported, nor was the number of viable TIL products that could not be used owing to a patient’s clinical deterioration during weeks of culturing. A 2004 publication was one of the first to indirectly present these data; in a cohort of 62 harvested patients, only 8 had been treated [6]. A larger follow-up publication systematically analyzed all attempts to generate TIL in 402 patients at the NCI from 2002 to 2007 [7]. In this era, TIL were screened for reactivity against the patient’s own fresh tumor, their own tumor when created as a cell line, or an HLA-A matched melanoma cell line. Viable TILs were generated from 94% of patients. Crucially, they applied a uniform criterion for TIL reactivity by measuring the amount of interferon gamma (IFNγ) released by both relative increase (≥2× background) and absolute value (≥400 pg/mL from 1 culture or 200 pg/mL from 25% of cultures from 1 patient). Using these criteria, successful reactive TILs were generated in 67% of patients. Lung and lymph node tumors were more likely to generate tumor-reactive TILs; gastrointestinal tract (and rarely attempted bone, brain) were statistically less likely to generate them. Perhaps unsurprisingly, active TILs were more likely generated from larger tumors on average (ie, mean size, 4 vs. 3.3 cm). No single clinical feature or combination had sufficient sensitivity or specificity to apply uniformly to future patients. The process was difficult for patients and physicians alike; only 107 of 269 patients (40%) with active TIL were eventually treated on clinical trials. The most common reasons for not pursuing therapy were clinical deterioration and development of brain metastases, which still plague current efforts today.

For those highly selected patients, however, efficacy was robust and much better than other available systemic therapies available at the time, including high-dose IL-2. Subsequent efforts in melanoma in the late 2000s to early 2010s often focused on improving the speed and efficiency of polyclonal TIL generation.

Treating with “Younger” TIL in Melanoma

Although the characteristics of naïve effector memory T cell lymphocytes were not fully fleshed out, there was clear evidence that the longer cells remained in stimulatory culture, the less likely they were to persist in patients. Indirect metrics of T cell stemness—such as longer telomere length, shorter time in culture, and improved in vivo persistence—were useful for predicting response to adoptive TIL therapy in patients with advanced melanoma [8].

Both the NCI [9] and Israeli groups [10] utilized bulk TIL and rapid expansion protocols that eliminated the need for multiple “microcultures” that were then screened for reactivity. In doing so, the time in culture and time to generation of TIL products could be reduced by several weeks. The number of screen failures due to lack of reactivity could also be greatly reduced. The Israeli group specifically reported in an early summary of their selected TIL versus young TIL experience that screen failures were reduced from 21/33 (64%) to 3/11 (27%), respectively [10]. The objective responses were also more favorable in these small, unselected cohorts. A later report on the same unselected young TIL Israeli cohort outlined a failure rate of only 21% in the first 55 patients. Of 32 treated patients, the ORR was 47% [11]. The NCI group reported a pooled 56% ORR and a 22% complete response (CR) rate for 93 patients across 3 cohorts treated with this TIL approach alone or 1 of 2 total body irradiation doses [12]. Similar to the smaller preliminary report, longer telomeres, higher percentage of CD8⁺CD27⁺ cells infused, and increased persistence at 1 month were associated with objective response (albeit with significant overlap between responses and nonresponses for each metric) [12]. In a large meta-analysis published in 2019 of 411 patients with melanoma treated by multiple centers, the ORR was 41% (95% confidence interval [CI], 35%–48%) and the CR rate was 12% (95% CI, 7%–16%) [13]. The vast majority of CRs were durable, and the majority of patients achieving PRs eventually progressed.

As a result of the work just described, the “unselected” approach to rapidly expanding TIL served as the backbone of the methods used in melanoma trials in the United States that led to the routine clinical adoption of lifileucel [14] and the randomized trial in Europe showing superiority of second-line TIL over ipilimumab [15]. Both of these methods are reviewed in detail by Smithy and Betof Warner [15,16] in this issue.

TIL in Nonmelanoma Solid Tumors

Although the initial adoptive TIL experience in the 1980s enrolled patients of multiple histologies [16], including renal cell carcinoma (RCC), non-small-cell lung cancer (NSCLC), and breast carcinomas, the greater success in the high-TMB (tumor mutational burden) cutaneous melanoma cohorts led to slower progress in many other solid tumors. Nonetheless, seminal studies in the past 10 years have demonstrated preliminary efficacy of TIL therapy in solid tumors outside of cutaneous melanoma with lower TMBs. These approaches have varied in methodology according to the solid tumor, and their success has increased with modern selection techniques of immune-reactive TIL cells for expansion. For example, in NSCLC, a recent pilot trial harvested 20 patients and successfully expanded 18 tumor-reactive TIL products. Sixteen patients were treated, of which 2 durable CRs and 1 transient PR was achieved (ORR 15% by intent-to-treat) [17].

In 2014, a case report was published of a woman with advanced, treatment-refractory cholangiocarcinoma treated on a multi-tumor cohort of adoptive TIL therapy for gastrointestinal malignancies [18]. The authors identified an endogenous CD4⁺ lymphocyte that recognized a missense mutation in the protein ERBB2IP via HLA-DQ, which was 1 of only 26 mutations detected on whole exome sequencing. Tumor regression was observed after transfer of 42 billion TILs, of which 25% were reactive against this mutation across 3 related CD4⁺ clones, and 4 doses of IL-2. After subsequent tumor growth, the patient was re-treated and remained in clinical benefit after 6+ months at the time of publication, along with TIL persistence. This served as an important proof of concept that CD4⁺ cells could mediate a TIL response, and that TIL responses were not restricted to high-TMB tumors such as cutaneous melanoma when the right reactive clone could be isolated. In fact, multiple immune-reactive epitopes were clearly present in patients with GI malignancies, as the same group later identified [19]. It seems that, for lower-TMB tumors, enriching for neoepitope-specific immune cells was the means by which one could utilize adoptive TILs for clinical benefit.

A similar case report was published from the same group utilizing a neoepitope-specific screen for TILs in a 50-year-old patient with metastatic colon adenocarcinoma to the lung [20]. An HLAC*08:02 restricted clone recognizing the MAP kinase driver alteration KRAS G12D was identified, expanded, and comprised 75% of the 148 billion cells administered to the patient with preceding NMA-LD and subsequent IL-2. This patient achieved a 9-month PR and, upon resection of the secondary resistant oligometastasis, there was documented loss of the HLA-C*08:02.

One important study enrolled 9 patients with chemotherapy-resistant advanced cervical carcinoma into a clinical trial of selected TILs that were reactive to the viral oncogenes E6 or E7 by HPV 16 or 18 [21]. Three of 9 patients responded, including two durable responses. Surprisingly, when detailed analyses were performed, the expanding, persistent TILs mediating clinical benefit in the 2 patients with CR were in fact not reactive to the HPV viral antigen, but instead were reactive to neoepitopes or a cancer germline antigen [22]. Nonetheless, HPV reactivity of the TIL culture was associated with clinical benefit, and the study was ultimately amended to necessitate at least 3 HPV+ cultures with IFNγ response [23]. Five of 18 patients (28%) responded in the cervical cohort, and 2 of 11 (18%) responded in the noncervical cohort. Consistent with most TIL trials, the CRs were durable, ongoing for >4 years, but the PRs lasted only a few months.

Estrogen receptor positive (ER+) breast carcinomas are generally insensitive to checkpoint inhibition. Zacharakis et al. presented a case report of a patient with ER+, HER2− breast carcinoma achieving a durable CR from TIL enriched for reactivity against 2 specific neoepitopes [24]. Using modern single-cell sequencing techniques, they demonstrated that one neoepitope was clonal but the other was likely subclonal. Interestingly, after administration of a study regimen that included NMA-LD, IL-2, and pembrolizumab, two additional neoepitope specific clones not specified arose and persisted (bringing the total to 2 clonal and 2 subclonal targeted neoepitopes), likely triggered by either “epitope spreading” from successful targeting, the resetting of the immune repertoire from the NMA-LD regimen, and/or the disinhibition from concurrent pembrolizumab. This case report is reassuring that in this common solid tumor generally thought to be immunologically “cold,” clonally selected adoptive TIL therapy can lead to a durable CR. A more mature follow-up from this trial, however, underscores the challenges inherent in selective TIL procedures [25]. Of 42 harvested patients, only 13 met reactive criteria and, of those, only 6 underwent therapy. The initial CR was the sole durable responder; two additional PRs were of 6- and 10-month duration. A similar approach was utilized in the low-TMB subset of melanomas arising in the uveal tract, with 7 of 20 evaluable patients selected for TIL therapy achieving an objective response, with only 1 persisting past 1 year [26].

Together, it is clear that for higher-TMB tumors, such as melanoma and NSCLC, bulk unselected TILs may be sufficient to lead to CRs in a minority of patients. For lower-TMB epithelial cancers that represent the majority of solid tumors, clonal selection of reactive TILs is necessary to achieve responses. We can utilize clinical and translational biomarkers of response to TILs to identify methods to improve efficacy across cancer types.

LOOKING FORWARD

Improved Identification of Lymphocytes Underlying Durable Clinical Benefit

For the use of bulk unselected TILs in advanced melanoma, data generally bear out the notion that “more is more.” In the meta-analysis of 112 patients with available data, responders had higher total cell dose infused versus nonresponders, and patients with >50 billion infused cells had a much better median overall survival (OS) than those who received <50 billion cells (NR vs. 4 months, P < .001) [13]. It is worth noting, however, that in the analysis of the phase 2 C-144-01 trial that led to the FDA approval of lifileucel, no relationship between infused TIL numbers was found with response [27]. This trial may have been underpowered to detect this relationship, however, and more data are required to understand the relationship between lifileucel cell dose and outcomes.

With the revolution in single-cell analytics and bioinformatics in the past decade, a dizzying array of immune subsets have been identified. Our ability to parse which immune cell subsets mediate which aspects of human health and disease has improved as much in the past decade as it has in perhaps the prior 30 years. Most notably, it is now widely understood that not all CD8⁺ TILs are created equal. Many CD8⁺ lymphocytes that chronically recognize their target neoepitope can reside in a dysfunctional, exhausted state, mediated by transcriptional factors such as TOX and expressing markers such as CD39 and checkpoints such as PD-1, TIM-3, and LAG-3 [28,29]. However, a minority of CD8⁺ effector and memory T lymphocyte populations with more “stem-like” properties can mediate durable antitumor responses from immune checkpoint blockades (eg, those that are TCF1⁺, CD28⁺, TIM-3⁻, and/or CD39⁻) [30,31].

The NCI group compared complete responders and nonresponders to TIL therapy in melanoma to investigate markers of response utilizing cytometry by time of flight (CyTOF) analysis at single-cell resolution. Their analysis suggested that, although the majority of infused TILs are PD-1⁺ and CD39⁺, a subset of CD8⁺ cells that are low in CD39 and CD69 represent stem-like lymphocytes that persist after infusion and retain antitumor efficacy. Their presence was 2.5× higher in CRs than NRs and improved OS in patients with higher rather than lower levels of CD39⁻CD69⁻ T cells was noted. Given the relatively lower rates of response to adoptive TILs in melanoma that is PD-1 and/or CTLA-4 resistant, it is tempting to speculate that identification of epitope-specific yet stem-like effector memory T lymphocytes becomes more difficult when the melanoma is immune edited following PD-1 blockade. This is also supported by retrospective analyses, which describe lower CR/PR rates when patients with advanced melanoma have experienced progression to prior PD-1 and/or CTLA-4 blockade [32,33].

The knowledge that neoepitope-specific T cells are largely exhausted offers diagnostic and therapeutic avenues. In a recent single-cell transcriptomic analysis of uncultured, neoepitope-specific T cells in 10 metastatic tumors of various tumor types, neoepitope-specific CD4⁺ and CD8⁺ cells were preferentially expanded in a more exhausted phenotype than their counterparts in the peripheral blood. By using candidate signatures containing features such as PD-1⁺ and TOX⁺, the authors could improve the prospective identification of single-cell neoepitope-specific TILs to where >50% were tumor reactive [34]. Another group has shown that sorting by CD137⁺ status can enrich for proliferative, epitope-specific T cells in ovarian carcinomas and melanomas [35]. These insights will help improve the ability to isolate and expand relevant patient-specific TILs and also may hint at targets for therapeutic exploitation.

There has been a relative lack of detailed data cross-referencing cell surface markers, especially in the majority of solid tumors that do not respond as well to immune checkpoint blockade or bulk TIL therapy. Recently, a single-cell resolution analysis of endogenous T cells recognizing antigen in epithelial ovarian carcinomas compared the expression and function of cells expressing multiple markers and their relationship with each other: PD-1, the co-stimulatory receptor CD137 (4-1BB), the lymphocyte-retention mediating integrin CD103 that binds E-cadherin, and the ectonucleotidase CD39 [36]. This analysis suggested that, of the analyzed markers shared among exhausted T cells, the CD137 marker was most specific for T cells that could be responsive to immune-based therapies. In this line, the MD Anderson group has investigated adding co-stimulation with CD137/4-1BB agonism in pre-replication for bulk TILs [37]. When combined with OKT-3 for CD3 stimulation and IL-2 in cutaneous and uveal melanomas, they found a significantly shorter time to culture than with IL-2 pre-replication alone, more consistent successful expansion in low-TMB uveal melanomas, and found that TILs retain CD28⁺ expression and low levels of TIM-3 and LAG-3, suggesting that they may retain some stemness. Clinical outcomes data are not yet published, to our knowledge, but are eagerly awaited.

Taken together, the improvements in identification of lymphocyte subsets in solid tumors have improved insights into therapy. An ideal tumor should have a high percentage of lymphocytes (to increase the chances of neoantigen-specific, not only bystander lymphocytes). However, it should also retain stem-like niches where memory CD4⁺ and CD8⁺ cells can proliferate to cytolytic effector phenotypes. There are some efforts underway to epigenetically preserve the “stemness” of the lymphocytes in TIL culture, but their preclinical data are not yet peer reviewed. As one example, LYL845 is a TIL product currently in phase 1 trials for solid tumors (NCT05573035) based on preclinical data suggesting that a distinct media formulation with certain cytokines in TIL expansion and culture that reduces hypoxia/glycolysis can retain a more proliferative, less exhausted TIL phenotype in the presence of autologous tumor [38].

Genetically Engineering TILs

Early efforts to manipulate TILs were often limited by the tools available in the early 2000s for transduction. These efforts utilized retroviral vectors that proved difficult to transduce cells with 1 copy with precision. This often led to lower efficiency of transduction or multiple genes/cell and potential cytotoxic or off-target effects. Compared with unmanipulated, unselected TILs, this would also lengthen the amount of time required to keep these cells in culture, which likely led to higher levels of terminal differentiation and exhaustion.

The first notable attempt to engineer TILs for increased activity was published in 2008 at the NCI Surgery branch, where 13 patients received TILs transfused with IL-2 utilizing a retroviral vector [39]. They achieved endogenous IL-2 production in approximately 25% to 35% of cells at best. Of 12 evaluable patients in this protocol, only 2 transient partial responses were observed. Translational analysis demonstrated a significantly shorter telomere length in transduced IL-2 TILs compared with either nonresponding or responding unselected TILs, offering support to the notion that the genomic tools of this era were too inefficient to be considered a reliable means of improving upon the unselected TIL process of the era.

More recently, genetic editing approaches such as CRISPR and TALENs have made genetic alterations more efficient, on-target, and reliable [40]. Efforts are underway in clinical trials targeting specific targets of interest. For example, chimeric immune receptors with specificity for a surface molecule, folate receptor alpha, enriched in cancers such as NSCLC and ovarian carcinomas, linked with co-stimulatory CD28/40 signaling domains can be transduced into TILs to trigger proliferation preferentially at tumor sites [37]. A first-in-human trial is reportedly forthcoming. A phase 1/2 trial of IOV-4001 (NCT05361174) is underway in melanoma and NSCLC; it utilizes a TALENs-based knockout of PD-1 in otherwise unselected TILs in an attempt to improve persistence and reduce exhaustion of infused TILs. Although PD-1 + expression in TILs is often a sign of more differentiated, exhausted TILs, it is not known whether abrogating PD-1/PD-L1 interactions will prevent exhaustion in the small population of stem-like TILs. A recent phase 1 trial offers some indirect support for this approach, however. CRISPR-mediated PD-1 deletion was tested in patients with refractory NSCLC [41]. They enrolled 22 patients to treat 12 with PD-1 disrupted PBMCs. Even though no a priori selection was performed, they demonstrated safety and feasibility of this approach with largely modest side effects and 1 patient achieved a progression-free survival of over 1 year.

Others are directly targeting transforming growth factor beta receptor 2 (TGFBR2) preclinically for potential use in ovarian, breast, and colorectal adenocarcinoma [42]. This is based on an understanding that TGFBR2 mediates multiple immune-suppressive mechanisms, including inhibiting CD4⁺ mediated Type 2 immunity and encouraging exhaustion of CD8⁺ cells [43]. Other direct modifications of TILs are aimed at cytokine expression, which is discussed below. Lastly, while outside of the scope of this adoptive TILs article, other manipulations to T cells, such as multi-step activations in chimeric antigen receptor-T (CAR-T) cells and human leukocyte antigen (HLA)-specific T-cell receptor (TCR)-expressing T cells, are also showing promise in solid tumors.

Improving the Co-administration Medication Regimen

Improving on IL-2

In terms of the cytokine used both in culture and in the clinic, traditional recombinant IL-2 (aldesleukin) is a crucial yet far from an ideal molecule. Its nonspecific alpha-receptor binding stimulates inhibitory Tregs that can inhibit TIL function. Its short half-life requires high-dose, frequent boluses, and its renally cleared pharmacokinetics restrict patient selection. To date, other molecules given intravenously have unfortunately not borne fruit, but new ideas are being developed. For example, an upcoming pilot trial is combining ANV14 [44], an engineered IL-2 molecule that selectively binds to the Th1-stimulating beta gamma subunits of the receptor rather than the Treg-stimulating alpha chain (NCT05869539), and would omit the use of high-dose IL-2 administration.

A phase 1 study of recombinant human IL-15, which will not stimulate Tregs but has Th1 effects similar to those of IL-2, was recently completed. Unfortunately, it caused significant hypotension that limited its dose escalation, although it did increase NK cell expansion as a pharmacodynamic marker of some biologic effect [45]. An interesting new approach utilizing IL-15 stimulation is OBX-115, which is a genomically modified TIL being tested in patients with advanced melanoma and NSCLC (NCTNCT06060613). These TILs have an acetazolamide-responsive transcription element, which triggers the alternative splicing of a constitutively expressed but degraded intracellular IL-15 to a locally secreted, membrane-bound form of IL-15. This would in theory allow for more targeted administration of a proliferative signal for TILs with less toxicity than systemic administration. Preliminary phase 1 data in 12 patients with melanoma suggest activity; there were responses in 6 patients, with immature follow-up [46]. This represents a proof of principle that patients need not receive IL-2 to respond to TILs.

Other cytokines are being explored as well. Recently, 33 patients with advanced melanoma were treated with TIL-expressing IL-12. Although they demonstrated transient antitumor efficacy with TIL doses 10 to 100× lower than traditional NCI bulk TIL protocols, numerous high-grade, life-threatening IL-12-related adverse events were observed, and the responses were short term [47]. Another upcoming trial will express IL-7 at the tumor microenvironment level where it could be better tolerated (NCT06204991).

Revisiting the Lymphodepletion Regimen

As noted above, the initial NMA-LD regimen developed by the Rosenberg laboratory has generally been used, largely unmodified, across 20 years in trials for melanoma and many solid tumors. The one aspect that was compared prospectively was the role of total body irradiation (TBI), which was added by the NCI group to intensify lymphodepletion in melanoma [48]. Administering 12 Gray initially showed promise, with an ORR of 72% in a cohort of 25 patients [48]. However, a larger randomized trial of 101 patients comparing 12 Gray TBI versus no TBI demonstrated no benefit in either CR rate or OS, leading to its abandonment from further study [49].

This NCI NMA-LD regimen was revolutionary for its time, but we now know that many hematologic lymphodepletion regimens for successful CAR-T therapies [50] generally utilize much less cyclophosphamide. For a patient who weighs 80 kg and has a body surface area of 1.9 m², the cyclophosphamide dose for lifileucel is more than double that of an agent such as tisagenlecleucel (ie, ~4.2 g for 750 mg/m² × 3 vs. 9.6 g for 60 mg/kg × 2). Prospective trials of lower cyclophosphamide are sorely needed. At present, there is 1 active small, single-center trial of 500 mg/m² cyclophosphamide × 3 doses with a translational immunologic endpoint that will provide valuable prospective data (NCT06151847). Additional multicenter retrospective analyses of standard lifileucel therapy must be done across treating centers to investigate the efficacy and safety of lower doses of cyclophosphamide when clinically indicated (eg, elderly, more frail patients, brain metastases).

Fludarabine can and should also be fine-tuned. Recent efforts in hematologic malignancies undergoing allogeneic stem cell transplantation or the CAR-T cell therapy axicabtagene ciloleucel [41] have demonstrated that a moderate “optimal” fludarabine exposure is associated with survival. Their work suggests that, based on routinely available clinical factors such as renal function, exposure can be estimated and fine-tuned to improve future outcomes. Notably, a recent study of transduced TCR-T cells engineered to recognize HLA-restricted KRAS G12D epitopes demonstrated an ongoing objective response despite using tocilizumab instead of fludarabine [51].

It is high time to interrogate the NMA-LD regimen. Despite its relatively “low-tech” status when compared with the other techniques above, learning how to safely modify this part of solid tumor TIL therapy may yet prove to be our biggest and fastest single improvement in treating more patients with these potentially life-saving regimens.

CONCLUSION

The use of TILs for adoptive immunotherapy has exploded in recent years, led by the tremendous advances in the academic and private sectors. Lifileucel is the first of undoubtedly multiple cell-based therapies that will change the way solid tumors are treated. However, we have identified key unanswered questions and areas for improvement, represented graphically in Figure 1.

Figure 1.

Schematic of adoptive TIL process and suggested future areas of improvement and optimization.

In the vast majority of solid tumors, bulk TIL therapy approaches have proven largely disappointing despite the notable successes highlighted above. We are now past the era in which we can naively call some tumors immunogenic and others not. It is incumbent upon the field to study how to translate nearly universal in vivo evidence of endogenous immune recognition of solid tumors into improvements in cellular therapies. We suspect that bulk, unmodified TILs are just the beginning, much as ipilimumab was the beginning of the immune checkpoint blockade revolution. Most patients with cancer will need selected immune-reactive or genetically modified TILs for antitumor efficacy. Thankfully, our knowledge of immune populations’ underlying response is advancing quickly. We must acknowledge that for every additional in vitro modification, a patient in dire need of therapy is waiting. Since the largest source of failure in modern TIL trials remains the “screen-fail” patient that usually does not get reported in the article, we must balance carefully the benefit of longer turnaround times with the potential downstream delay. By focusing on novel cytokines and adjustments to the NMA-LD regimen, we can reduce toxicity and broaden the pool of patients eligible for therapy. Physicians at treating centers should also continue to work together to pool data that will improve patient selection, especially given the significant biologic and financial toxicity of this approach.