A TIM-3–Fc decoy secreted by engineered T cells improves CD19 CAR T-cell therapy in B-cell acute lymphoblastic leukemia Free

B-cell acute lymphoblastic leukemia (B-ALL) is a highly aggressive hematological malignancy. Patients are risk stratified based on age, response to induction/consolidation therapy, and recurrent molecular/cytogenetic abnormalities.¹ Despite affecting individuals of all ages, B-ALL is the most common childhood cancer, accounting for ∼35% of pediatric tumors.² Unfortunately, the prognosis of B-ALL is drastically worse in patients with refractory or relapsed (R/R) disease.^3,4 

CD19-directed chimeric antigen receptor (CAR) T cells have transformed the treatment landscape for R/R B-ALL. However, despite impressive initial complete remission (CR) rates, >50% of patients experience relapse within 1 year after treatment.^5,6 Two types of relapse are observed: (1) CD19^– relapse, primarily driven by leukemia cell–intrinsic mechanisms, including lineage switching, reduced membrane expression of CD19 due to splice variants/truncating mutations, or phenotypic escape from CD34⁺CD22⁺CD19^– (pre)leukemic cells; and (2) CD19⁺ relapses, often associated with poor T-cell function or loss of CAR T-cell persistence.^7-11 Regardless of the mechanism, patients with B-ALL who relapse after CAR T-cell therapy face a dismal prognosis with limited therapeutic options.¹² 

The bone marrow (BM) is the primary site for leukemic cells and a common location for relapse. The interaction between leukemic and immune cells, specifically the expression of immune checkpoint (IC) receptors (ICRs) and their ligands, has been linked to leukemia progression and therapy resistance.^13-18 A key mechanism underlying relapse is the suppression of T cells through the activation of ICRs.¹⁹ Indeed, several studies have shown that TIM-3⁺, PD-1⁺TIM-3⁺, and CTLA-4⁺TIGIT⁺ T cells promote B-ALL progression and confer poorer survival.^20,21 Cancer cells often evade immune surveillance by upregulating IC ligands on various cell types within the tumor microenvironment (TME), leading to T-cell exhaustion.²² In fact, mesenchymal stromal cells (MSCs) are important components of the BM niche and have been implicated in B-ALL pathogenesis and drug resistance.^12,23-25 However, despite their importance in the development of more effective therapies, the expression of ICRs on T cells and their corresponding ligands on blast cells and MSCs within the TME throughout B-ALL progression remains poorly characterized. Additionally, CAR T cells also express ICRs and are susceptible to T-cell–intrinsic immunoregulatory exhaustion signals,^12,26 suggesting that upregulation of ICRs in CAR T cells may represent a resistance mechanism. Importantly, combinatorial approaches with ICR blockers, such as monoclonal antibodies and small molecules, have shown promise in clinical settings.^27-30 

Here, we performed a comprehensive analysis of ICR and ligand expression in leukemic blasts, CD4⁺ and CD8⁺ T cells, and MSCs from pediatric and adult B-ALL BM samples at diagnosis (Dx) and relapse. We observed a significant increase in TIM-3 expression in T cells and galectin-9 expression in blasts and MSCs as the disease progressed. Furthermore, we found that galectin-9 impairs the function and persistence of CD19-directed CAR (CAR19) T cells. Because the available murine and human functional anti–TIM-3 antibodies do not disrupt the interaction between TIM-3 and galectin-9,^31,32 we developed a TIM-3–Fc decoy receptor. By delivering this decoy either through primary T cells coadministered with CAR19 T cells or by CAR19 T cells themselves by incorporating it into a bicistronic all-in-one CAR19–TIM-3–Fc construct, we significantly enhanced the antileukemia efficacy and persistence of CAR19 T cells in in vivo models of xenografts derived from patients (PDXs) with B-ALL. Mechanistically, we observed an in vivo expansion of both transduced and bystander adaptive and innate effector and memory T cells after treatment with CAR19–TIM-3–Fc T cells. Our findings support the clinical evaluation of TIM-3–Fc decoy–armored CAR19 T cells for the treatment of R/R B-ALL.

All experimental studies using primary samples were approved by the Ethics Committee on Clinical Research of the Clinic Hospital of Barcelona (HCB/2017/0781). Primary cells were obtained after written informed consent, in accordance with the Declaration of Helsinki. Buffy coats were obtained from the Catalan Blood and Tissue Bank after institutional review board approval (HCB/2018/0030). Peripheral blood (PB) mononuclear cell processing and B-ALL cell lines used are detailed in supplemental Methods, available on the Blood website.

Nonobese diabetic Cg-Prkdc^scidIl2rg^tm1Wjl/SzJ (NSG) mice (The Jackson Laboratory, Bar Harbor, ME) were housed under pathogen-free conditions. For all in vivo experiments, sublethally irradiated (2 Gy) NSG mice received 1 × 10⁶ B-ALL PDX cells by IV infusion, followed 4 days later by an IV infusion of 1 × 10⁶ to 2 × 10⁶ T cells (untransduced, TIM-3–Fc, CAR19, coadministration of CAR19 and TIM-3–Fc, or bicistronic all-in-one CAR19–TIM-3–Fc T cells; n = 5-6 mice per group). B-ALL engraftment and T-cell expansion was monitored weekly in PB by fluorescence-activated cell sorter (FACS). Animals were euthanized when they had lost >20% of their weight or showed signs of disease. Plasma was collected by blood centrifugation (1500g, 10 minutes, 4°C) obtained by intracardiac puncture. All in vivo procedures were performed in accordance with the institutional animal care committee of the Barcelona Biomedical Research Park (DAAM7393/HRH170029).

We used the ComplexHeatmap package in R to display in heat maps the median of ICRs and their ligands in human BM samples. Other data are presented as mean ± standard error of the mean. All in vitro experiments were performed in triplicate using 3 different donors. Parametric tests were used for in vitro studies: Student t test to compare 2 groups and 1-way analysis of variance to compare ≥3 experimental groups. Nonparametric tests were used for in vivo studies: the Mann-Whitney U test to compare 2 groups and the Kruskal-Wallis test to compare ≥3 experimental groups. All P values are 2-tailed, and statistical significance was defined as P value ≤.05. All analyses were performed using Prism software v.8.0 (GraphPad Prism, San Diego, CA).

Detailed information about the immunophenotyping of primary samples and cell lines, generation of CAR19 and TIM-3–Fc decoy expression vectors, lentiviral production and T-cell transduction, cytotoxicity assays, soluble TIM-3 detection and galectin-9 interactions, production and purification of the TIM-3–Fc decoy, clinical-biological correlation analyses, CAR19-treated patients’ cohort, and spectral flow cytometry are provided in supplemental Methods.

ICs have been linked to immunosurveillance and immunotherapy failure in cancer, prompting the development and clinical use of IC blockers. However, little is known about the expression of ICRs and their ligands in B-ALL. We investigated the protein expression of several IC axes in BM samples at Dx and relapse from pediatric (Dx, n = 32; relapse, n = 25) and adult patients with B-ALL (Dx, n = 15; relapse, n = 13) and compared it with expression in BM from age-matched healthy donors (n = 21; Table 1). Specifically, we analyzed the ICRs BTLA, CTLA-4, LAG-3, PD-1, and TIM-3 in CD8⁺ and CD4⁺ T cells, whereas their ligands HVEM, CD80, CD86, MHC-II, PDL-1, PDL-2, galectin-9, and HMGB1 were examined in blasts (normal B cells in nonleukemic BM) and MSCs (Figure 1A; supplemental Figures 1-4).

Our analysis of ICR protein expression in CD4⁺ and CD8⁺ T cells revealed that only CTLA-4 and TIM-3 expression was significantly higher in patients with B-ALL at Dx than in control BM and persisted high at relapse (Figure 1B; supplemental Figure 4). Notably, TIM-3 expression on CD8⁺ T cells was remarkably high at both Dx and relapse, whereas all ICRs were significantly upregulated in CD4⁺ T cells at relapse (Figure 1B; supplemental Figure 4). We then examined the expression of IC ligands in normal B cells and MSCs from nonleukemic BM and in blasts and MSCs from the BM of patients with B-ALL. Our results indicated that galectin-9 and HMGB1 (TIM-3 ligands) were significantly more highly expressed in Dx B-ALL blasts than in normal B cells from nonleukemic BM, whereas CD86 (CTLA-4 ligand), galectin-9, and HMGB1 were the most significantly upregulated ligands at relapse (Figure 1B-C; supplemental Figure 4). Additionally, galectin-9 was the only IC ligand significantly upregulated in BM-MSCs from patients with B-ALL at both Dx and relapse (Figure 1B-C; supplemental Figure 4). No significant correlation was observed between the expression of ICRs and their ligands with patient age, gender, molecular/cytogenetic subtype, or allogeneic (allo) vs nonallotransplantation status (Table 1; data not shown). These findings highlight the TIM-3–galectin-9 axis as the most highly expressed IC axis in T cells, blasts, and MSCs throughout the progression of B-ALL.

To investigate the prognostic significance of galectin-9 in B-ALL, we analyzed a pediatric R/R B-ALL data set (n = 64 patients) from the TARGET Data Matrix Portal. Our analysis revealed that patients with higher galectin-9 levels had significantly poorer event-free survival (P = .002; Figure 1D). Moreover, we analyzed the expression of TIM-3 in CAR19 T cells from the infusion products of 37 R/R patients with B-ALL who had received CAR19 T cells (varnimcabtagene autoleucel) at our institution. We observed a poorer event-free survival in patients with CAR19⁺ T cells expressing high levels of TIM-3 (Figure 1E). Collectively, TIM-3–galectin-9 axis may represent a promising immunotherapeutic target for B-ALL.

To explore the role of the TIM-3–galectin-9 axis in CAR T-cell function against B-ALL, we generated CAR19 T cells according to standard protocols and cocultured them with 3 primary B-ALL samples at an effector-to-target (E:T) ratio of 1:4 (Figure 2A; supplemental Figure 5). FACS analysis of TIM-3 expression on CAR19 T cells and galectin-9 expression on B-ALL blasts revealed that TIM-3 expression on CAR19 T cells was significantly higher after exposure to B-ALL cells for 24 hours than in CAR19 T cells alone. Similarly, galectin-9 expression in CAR T-cell–resistant B-ALL cells was consistently much higher after coculture with CAR19 T cells (Figure 2B). To investigate the effect of galectin-9 on CAR19 T cells, we cultured CAR19 T cells with or without recombinant human galectin-9 (rhGalectin-9) for 48 hours. rhGalectin-9 treatment significantly increased apoptosis (Figure 2C) and decreased the proliferation of CAR19 T cells, as measured by Annexin-V and eFluor 670 retention assays, respectively (Figure 2D). Furthermore, CAR19 T cells exposed to rhGalectin-9 showed significantly reduced cytotoxic activity against SEM and NALM6 B-ALL cells at various E:T ratios (Figure 2E). To validate these findings, we generated B-ALL cell lines overexpressing galectin-9 (Figure 2F) and compared the cytotoxic effects of CAR19 T cells against galectin-9–overexpressing B-ALL cells and wild-type counterparts. Consistently, CAR19 T-cell killing was significantly lower in galectin-9–overexpressing target cells than in nontransduced cells (Figure 2G). These findings highlight the detrimental effect of galectin-9 on the homeostasis and function of CAR19 T cells and suggest that inhibiting its interaction with TIM-3 in T cells may enhance CAR19 T-cell efficacy in R/R B-ALL.

TIM-3 is known to have multiple ligands, and murine and human anti–TIM-3 antibodies with reported functional efficacy disrupt TIM-3 binding to PtdSer and CEACAM1 but not to galectin-9.^31,32 Given these observations, we developed a strategy to block the binding of any TIM-3 ligand (including galectin-9) expressed on B-ALL cells to the TIM-3 receptor expressed on CAR19 T cells. To achieve this, we engineered a TIM-3 decoy comprising the extracellular domain of TIM-3 fused to a silent Fc region to increase stability and in vivo half-life (Figure 3A; supplemental Figure 6A). This decoy protein is designed to capture TIM-3 ligands, preventing their interaction with the receptor. The TIM-3–Fc decoy was purified by chromatography and showed a glycosylated dimer structure and a dose-dependent interaction with immobilized rhGalectin-9 (supplemental Figure 6B-D). We then demonstrated that purified soluble TIM-3–Fc decoy effectively rescued rhGalectin-9–mediated inhibition of CAR19 T-cell cytotoxicity against SEM and NALM6 cells (supplemental Figure 6E). Lentiviruses were then generated to deliver the TIM-3–Fc decoy in primary T cells, and successful transduction of activated T cells was confirmed (tdTo⁺; Figure 3B). Transduced T cells exhibited higher levels of TIM-3 on the cell surface (Figure 3C), likely attributed to TIM-3–Fc binding to TIM-3 ligands (decoration). Importantly, no differences in proliferation were observed between untransduced and TIM-3–Fc–transduced T cells during in vitro expansion (Figure 3D).

We next tested the functionality of this strategy in vivo using a galectin-9–expressing B-ALL PDX model (60% galectin-9⁺ cells). NSG mice were injected IV with 1 × 10⁶ B-ALL cells and, 4 days later, mice received IV 2 × 10⁶ of the following T cells: untransduced T cells, TIM-3–Fc T cells (25% transduction), CAR19 T cells (25% transduction), and CAR19 T cells plus TIM-3–Fc T cells (25% transduction for each; Figure 3E). PB was monitored weekly by FACS for leukemia growth and T-cell persistence (supplemental Figure 7A). Analysis of blasts in PB revealed complete leukemia dissemination in the control groups (untransduced and TIM-3–Fc) at week 3, whereas animals treated with CAR19 T cells alone or in combination with TIM-3–Fc T cells achieved CR. To functionally assess CAR19 T-cell persistence, animals were rechallenged with the same PDX and monitored for leukemia relapse. Although no blasts were detected in PB 4 weeks after rechallenging, there was a significant increase in T cells in mice treated with CAR19 T cells + TIM-3–Fc T cells (Figure 3F). This was confirmed at the end point (week 7), when a consistent increase in both total T cells and green fluorescent protein (GFP⁺) CAR⁺ T cells was observed in PB, BM, and spleens of CAR19 T-cell + TIM-3–Fc T-cell–treated mice (Figure 3G).

To better determine whether the increased T-cell population observed in animals treated with CAR19 T cells + TIM-3–Fc T cells was responsible for the greater antileukemic effect, we conducted an in vivo study under stress conditions. We reduced the T-cell dose to half (1 × 10⁶) and lowered the CAR19 transduction efficiency to 17.5%, comparing the following groups: untransduced T cells, TIM-3–Fc T cells (32.5% transduction), CAR19 T cells (17.5% transduction), and CAR19 T cells + TIM-3–Fc T cells (17.5% and 32.5% transduction, respectively; Figure 3H). End point analysis revealed a significantly lower tumor burden in the PB and spleen of CAR19 T-cell + TIM-3–Fc T-cell–treated animals than in CAR19 T-cell–treated mice (Figure 3I). Analysis of T-cell persistence revealed that CAR19 T-cell + TIM-3–Fc T-cell–treated mice consistently exhibited the highest number of total and CAR⁺ T cells in PB, spleen, and BM compared with other groups (Figure 3J). Finally, to assess T-cell functionality, we sorted human T cells from mice treated with CAR19 T cells or CAR19 T cells + TIM-3–Fc T cells and performed an ex vivo cytotoxic assay with B-ALL cells at an E:T ratio of 1:1 for 24 hours (supplemental Figure 7B). Results showed that T cells from animals treated with CAR19 T cells + TIM-3–Fc T cells had a greater cytotoxic effect ex vivo, with lower numbers of blasts and higher numbers of T cells, than T cells from mice treated with CAR19 T cells (supplemental Figure 7C). Collectively, the secretion of a TIM-3–Fc decoy by primary T cells enhances the persistence and efficacy of CAR19 T cells in vivo and ex vivo.

We next sought to engineer TIM-3–Fc as a secreted decoy in CAR19 T cells. We designed a CAR19–TIM-3–Fc bicistronic construct (“all-in-one” CAR19–TIM-3–Fc), incorporating the CAR19 structure, an F2A domain, the previously characterized TIM-3–Fc structure, and a T2A peptide followed by GFP (Figure 4A). The transduction rates of the CAR19–TIM-3–Fc bicistronic construct into activated primary T cells were significantly lower than those of the single CAR19 construct (13.5% ± 2.2% vs 54.5% ± 3.6%; Figure 4B). We first confirmed that CAR19–TIM-3–Fc T cells secrete TIM-3–Fc (5.3 ± 0.4 ng/mL) in vitro (Figure 4C). The proliferation rates of untransduced cells, CAR19 T cells, and CAR19–TIM-3–Fc T cells were very comparable (Figure 4D). Additionally, we determined the CD4:CD8 ratio and its distribution among naïve, central memory, effector memory (EM), and terminally differentiated EM T-cell subsets in untransduced, CAR19, and CAR19–TIM-3–Fc T cells and found no differences (Figure 4E-F). Similarly, no overall differences were observed in the expression of the ICRs TIM-3, LAG-3, and PD-1 in transduced T cells between CAR19 and CAR19–TIM-3–Fc T cells (Figure 4G). Notably, a significantly higher percentage of both CD8⁺ and CD4⁺ T cells expressing the activation marker CD69 was found in transduced (GFP⁺) CAR19–TIM-3–Fc T cells than in CAR19 T cells (Figure 4H).

We next evaluated the cytotoxic function of CAR19–TIM-3–Fc T cells in vitro. Four independent primary B-ALL cells expressing galectin-9 were used for cytotoxicity assays at an E:T ratio of 1:8 for 48 hours (Figure 4I). CAR⁺ T cells were identified as 7AAD^–CD3⁺GFP⁺ and live target cells as 7AAD^–CD3^– (Figure 4J). Compared with CAR19 T cells alone, CAR19–TIM-3–Fc T cells demonstrated enhanced cytotoxicity, with significantly lower absolute numbers of live blasts and significantly higher numbers of GFP⁺ T cells and total T cells, as well as elevated interferon gamma release (Figure 4K), confirming a superior B-ALL cell killing and T-cell expansion of CAR19–TIM-3–Fc T cells in vitro.

We then assessed the in vivo efficacy of CAR19–TIM-3–Fc T cells using a B-ALL PDX stress model. Four days after PDX injection, 1.5 × 10⁶ of either CAR19 or CAR19–TIM-3–Fc T cells (12% transduction efficiency; 175 000 transduced cells) were administered (Figure 5A). No significant weight loss was observed between CAR19- and CAR19–TIM-3–Fc–treated mice during the follow-up period (Figure 5B). At the end point, soluble TIM-3–Fc was detected in plasma only in those animals treated with CAR19–TIM-3–Fc T cells (24.4 ± 4.4 ng/mL), confirming in vivo secretion of TIM-3–Fc by bicistronic CAR⁺ T cells (Figure 5C). Analysis of antileukemic effects revealed that the number of blasts in PB, spleen, and BM was significantly lower in animals treated with CAR19–TIM-3–Fc T cells than in animals treated with CAR19 T cells (Figure 5D). Furthermore, the total number of circulating GFP⁺ T cells was 4.5-fold higher in animals treated with CAR19–TIM-3–Fc T cells (Figure 5E), confirming an enhanced CAR19–TIM-3–Fc T-cell activity in vivo.

To further characterize the in vivo–expanded T-cell populations, spleens and BM were harvested from B-ALL xenografted mice 4 weeks after treatment with CAR19 T cells (n = 5) and CAR19–TIM-3–Fc T cells (n = 5). Single-cell suspensions from spleen and BM were analyzed by spectral flow cytometry using a 24-color T-cell module of the Euroflow-Immune monitoring antibody panel^33,34 (Figure 6A). The Uniform Manifold Approximation and Projection (UMAP) representation of 1 917 663 T cells, including both transduced (GFP⁺) and nontransduced/bystander (GFP^–) CAR19 and CAR19–TIM-3–Fc T cells, was analyzed from spleen and BM samples across all mice. Major T-cell subsets, innate T cells, and functional and maturation-associated T-cell subsets were classified and quantified (Figure 6B; supplemental Table 3). CAR19–TIM-3–Fc–treated mice exhibited a notable expansion of multiple transduced and bystander T-cell subpopulations in both spleen and BM compared with those treated with CAR19 alone, suggesting a greater polyclonal T-cell response elicited by CAR19–TIM-3–Fc (Figure 6C). We comprehensively characterized the distinct T-cell populations that underwent a preferential significant expansion in each organ (supplemental Table 3). In the spleens of animals treated with CAR19–TIM-3–Fc T cells, the GFP⁺ T-cell populations that showed the most pronounced expansion were effector γδ T cells, consistent with their established cytotoxic capacity and their ability to migrate to lymphoid tissues.^35,36 Within the GFP^– T-cell fraction in the spleen, there was a predominance of innate effector CD8⁺ and CD4⁺ mucosal-associated invariant T cells (MAITs), together with CD8⁺ and CD4⁺ memory T cells. In T cells obtained from the BM of mice treated with CAR19–TIM-3–Fc T cells, there was an overexpansion of GFP⁺ CD4^–CD8^– and CD8⁺ EM T cells, as well as GFP^–CD4^–CD8^–, GFP^–CD4⁺CD8⁺ memory T cells, and effector GFP^–CD4⁺ MAITs; all strong cytotoxic cells.^36-39 Finally, among all the T-cell populations significantly expanded in CAR19–TIM-3–Fc–treated mice, 4 GFP^– T-cell populations were shared between the spleen and BM (Figure 6C, illustrated in bold). These included cytotoxic memory CD4⁺CD8⁺ T cells and innate effector CD4⁺ MAITs, which may be attributable to the bystander effect of the secreted TIM-3–Fc decoy. This multimodal identification of transduced and bystander T cells expanded in vivo provides insights into the diversity of T-cell responses after adoptive immunotherapy with CAR19–TIM-3–Fc T cells.

CAR19 T cells have revolutionized the treatment of R/R B-ALL, but unfortunately >50% of patients relapse within a year.⁶ Several factors influence the effectiveness and durability of CAR T-cell therapy, including the design of the CAR constructs, the target of choice, antigen density, antibody affinity and avidity, tumor burden, TME, and other factors that confer immune resistance to tumor cells.^9,10,40 CAR T-cell persistence after infusion is crucial for long-term outcomes, because lower T-cell persistence correlates with worse overall survival in B-cell malignancies.⁴¹ A number of emerging strategies hold promise for improving CAR T-cell function and persistence. These include optimizing the design of CAR constructs, increasing the proportion of memory T cells, altering the TME through cytokine signaling, and overcoming IC inhibition with IC blockers.^{9,10,40,42,43} Here, we extensively characterized the protein expression of several ICRs and their ligands during B-ALL progression. Specifically, we investigated ICR expression on T cells and IC ligand expression on B-ALL blasts and MSCs, which may play an important role in the course of the disease.^13,14,44 

The TIM-3–galectin-9 axis was the only IC pathway consistently upregulated in all 3 cell types examined in our B-ALL cohort, as well as in CAR19 T-cell/B-ALL cocultures. The TIM-3–galectin-9 axis has been studied in autoimmune diseases, infections, and various cancer types,^45-51 but its role in cancer remains unclear. Some studies have linked galectin-9 expression to poor outcomes in urinary tract tumors and non–small cell lung cancer, whereas others have associated it with better outcomes in breast, melanoma, and colon cancers.⁵² The impact of the TIM-3–galectin-9 axis in the context of CAR T cells remains unexplored. Our data demonstrate that galectin-9 clearly impairs CAR19 T-cell cytotoxicity by increasing CAR T-cell apoptosis and suppressing T-cell proliferation. Consistent with these findings, some studies have shown that anti–galectin-9 monoclonal antibodies (mAbs) inhibit tumor growth and extend survival in mouse models of lung cancer, acute myeloid leukemia, and T-cell acute lymphoblastic leukemia (T-ALL), and other studies have shown that these antibodies also protect T cells from galectin-9–induced cell death and enhance T-cell–mediated cytotoxicity of tumor cells.^53-55 

We have developed an innovative therapeutic strategy based on an engineered TIM-3–Fc decoy that can be delivered either by primary T cells coadministered with CAR19 T cells or by CAR19 T cells themselves through a bicistronic all-in-one CAR19–TIM-3–Fc construct. Our findings suggest that the soluble TIM-3 decoy enhances CAR19 T-cell persistence through both autocrine and paracrine mechanisms. Importantly, the enhanced T-cell expansion amplified the cytotoxic effect of the CAR19 construct, particularly the “all in one” CAR19–TIM-3–Fc construct, which outperforms CAR19 T cells alone, demonstrating therapeutic benefits both in vitro and in vivo. Because CAR19 T cells target CD19⁺ blasts, the TIM-3–Fc decoy secreted locally in the site of disease by CAR19–TIM-3–Fc T cells is likely to enhance the antileukemic effect compared with the TIM-3 decoy secreted by other non-CAR T cells.

Decoys occur naturally in humans and are very useful for targeting receptors/ligands with redundant functions. Indeed, receptor-ligand promiscuity is a hallmark of the chemokine/cytokine network but is also evident for ICRs, such as TIM-3, which can bind multiple ligands at the same or distinct sites within the extracellular domain.^56-58 To our knowledge, TIM-3 decoys have not yet been described. Although several anti–TIM-3 mAbs have been used clinically, their therapeutic efficacy is limited, and significant antitumor effects have been demonstrated only when coadministered with anti–PD-1 or anti–PD-L1 mAbs.^59-61 In contrast to the TIM-3 decoy, available anti–TIM-3 antibodies bind to a specific site on the TIM-3 receptor, disrupting its binding to PtdSer and CEACAM1 but not to galectin-9, thereby compromising the inhibition of the TIM-3 inhibitory pathway.^31,32 In addition to mAbs, intrinsic IC silencing in CAR T cells using CRISPR/Cas9 has shown promise^62-66; however, a recent study showed that PD-1 silencing can impair CAR T-cell function.⁶⁷ Current data indicate that ICR ablation or IC blockers have yielded more promising results in the context of CD28-based than 4-1BB–based CAR T cells.^63-66,68 Additionally, 2 independent studies using TIM-3 knockout mouse models demonstrated impaired T-cell responses, underscoring the critical role of TIM-3 in optimal T-cell function.^69,70 Accordingly, using an IC blocker while preserving the intrinsic function of the ICR seems the most promising approach.

Spectral flow cytometry studies enabled us to comprehensively identify in vivo overexpanded T-cell populations in B-ALL–xenografted mice treated with CAR19–TIM-3–Fc T cells. We observed overexpansion of specific functional and maturation-associated subpopulations of both adaptive and innate transduced and bystander T cells in mice treated with CAR19–TIM-3–Fc T cells. Notably, innate effector and memory T cells, including γδ T cells, MAITs, and double-negative T cells, were overexpanded in CAR19–TIM-3–Fc–treated mice. Given the potential exhaustion and proliferation of these innate CAR19 T cells due to their potent cytotoxic activity,^38,71-75 the local secretion of the TIM-3–Fc decoy may enhance their persistence. Furthermore, recent studies have highlighted the prognostic significance of preexisting rare CAR⁺ T-cell populations, such as MAITs and γδ T cells, in R/R patients with B-ALL treated with CAR19 T cells.^76,77 A comparable cellular composition is anticipated in patients because apheresis products are collected before lymphodepletion for CAR T-cell expansion and transduction during manufacturing.

In conclusion, our innovative approach using CAR19–TIM-3–Fc T cells, which continuously secrete a TIM-3 decoy in vivo, improves CAR19 T-cell persistence and efficacy in preclinical B-ALL PDX models. Indeed, this strategy of engineering T cells to secrete the TIM-3–Fc decoy could potentially be applied to other malignancies involving the TIM-3–galectin-9 axis or to other CAR T-cell products based on different costimulation domains, which have shown less persistence than 4-1BB–based CAR T cells.^78,79 Future studies exploring conditional TIM-3–Fc secretion may offer improved system control and safety.

Samples included in this study were provided by the collection of samples of the Josep Carreras Leukemia Research Institute (C002922) and were processed following standard operating procedures with the appropriate approval of the Ethics and Scientific Committees.

Core support of research in the P.M. laboratory is provided by CERCA/Generalitat de Catalunya and Fundació Josep Carreras-Obra Social la Caixa; additional support is provided by the European Research Council (grants ERC-PoC-957466 IT4B-TALL and ERC-PoC-101100665 BiTE-CAR); Horizon 2020 (grant 101057250-CANCERNA), the Ministerio de Economía, Industria y Competitividad (MINECO; grants PID2019-108160RB-I00 and PID2022-142966OB-I00/MCIN/AEI/10.13039/501100011033 and Feder Funds), MINECO/European Union NextGenerationEU (grants CPP2021-008508 and CPP2022-009759); the Spanish Association Against Cancer (AECC; grants PRYGN234975MENE, and PRYGN211192BUEN), the Health Institute Carlos III (ISCIII; grant PI20/00822), the Uno Entre Cien Mil Foundation, Dutch Josep Carreras Leukemia Stifftung (DJCLS; grant 02 R/2023), MINECO (grant PLE2021-007518), and the ISCIII-RICORS within the Next Generation EU program (plan de Recuperación, Transformación y Resilencia). Research in the L.Á.-V. laboratory is funded by the Spanish Ministry of Science and Innovation MCIN/AEI/10.13039/ 501100011033 (grants PID2023-148429OB-I00, PID2020-117323RB-I00, PDC2021-121711-100, CPP2022-009762, and CPP2022-009765); the ISCIII/FEDER (grants DTS20/00089 and PMPTA22/00167), the AECC (grants PROYE19084ALVA and PRYGN234844ALVA); the CRIS Cancer Foundation, Spain (grants FCRIS-2021-0090 and FCRIS-2023-0070); the Fundación “La Caixa” (grant HR21-00761 project IL7R_LungCan); the Comunidad de Madrid (grant P2022/BMD-7225 Next Generation CART MAD), and the Fundación d'Estudis i Recerca Oncológica (BBASELGAFERO2024.01). A.F. was supported by a Juan de la Cierva postdoctoral fellowship (FJC2021-046789-I); N.T. was supported by a PhD fellowship (FPU19/00039) from the Spanish Ministry of Science and Innovation; and C.P. was supported by a Contratos predoctorales de formación en investigación en salud fellowship (FI21/00161) from ISCIII. L.R.-P. is supported by a predoctoral fellowship from the Immunology Chair, Universidad Francisco de Vitoria/Merck. S.R.Z. was supported by a Marie Sklodowska Curie Fellowship (GA 795833).

Contribution: A.F., S.R.Z., L.Á.-V., C.B., and P.M. conceptualized and designed the study; A.F., R.L.-G., S.R.Z., L.R.-P., A.M.M., M.V., M.G.-M., N.F.-F., H.R.-H., N.T., C.P., T.V.-H., A.M., and A.P.-P. contributed to sample acquisition of data; A.F., R.L.-G., S.R.Z., A.O., L.Á.-V., C.B., and P.M. analyzed and interpreted data; E.G., J.-M.R., J.R., M.C., M.R.-O., E.A., P.B., J.L.F., M.J., E.A.G.-N., F.L., R.W.S., S.Q., P.V., V.O.-M., N.M.-C., and J.D. contributed to sample preparation; A.F., C.B., and P.M. wrote the manuscript; L.Á.-V., C.B., and P.M. are the guarantors; and all authors contributed to review and editing.

Conflict-of-interest disclosure: L.Á.-V. is a cofounder of Ledartis, a spin-off company focused on unrelated interests, as well as of STAb Therapeutics, a spin-off company from the Research Institute Hospital 12 de Octubre (imas12). P.M. is a cofounder of OneChain Immunotherapeutics, a spin-off company from the Josep Carreras Leukemia Research Institute. The remaining authors declare no competing financial interests.

Correspondence: Pablo Menéndez, Josep Carreras Leukemia Research Institute, Carrer Casanova 143, 08036 Barcelona, Spain; email: pmenendez@carrerasresearch.org; Clara Bueno, Josep Carreras Leukemia Research Institute, Carrer Casanova 143, 08036 Barcelona, Spain; email: cbueno@carrerasresearch.org; and Luis Álvarez-Vallina, H12O-CNIO Cancer Immunotherapy Clinical Research Unit, Spanish National Cancer Research Center, c/Melchor Fernandez Almagro, 3, 28029 Madrid, Spain; email: lalvarezv@ext.cnio.es/ lav.imas12@h12o.es.