IL-15 overexpression promotes memory program and anti-tumor activity of CD64 CAR T cells in a preclinical AML model

Abstract

The prognosis of relapsed or refractory acute myeloid leukemia (r/r AML) patients remains poor due to lack of novel therapies. We previous demonstrated that chimeric antigen receptor (CAR) T cells targeting CD64 have the potential to treat AML with minimal toxicity to hematopoietic stem/progenitor cells. However, the efficacy was limited in AML mouse models. Interleukin-15 (IL-15), a cytokine that promotes T cell survival and proliferation, has been shown to enhance CAR T cell activity. Here, we engineer CD64 CAR T cells with overexpression of IL-15 and evaluate the function. IL-15-armed CAR T cells exhibit enhanced cytolytic activity against AML cells, improve expansion and persistence in vitro, and favor a memory phenotype while reducing exhaustion and apoptosis. In mouse model, IL-15-armed CAR T cells show robust expansion, prolong mouse survival, and no obvious toxicity. These findings suggest that IL-15-armed CD64 CAR T cells may be a promising strategy for r/r AML.

IL-15-armed CD64 CAR T cells exhibit enhanced expansion, persistence, and anti-leukemia efficacy. This strategy prolongs survival in mouse models without obvious toxicity, presenting a promising therapeutic approach for acute myeloid leukemia.

Introduction

Acute myeloid leukemia (AML) is the most common acute leukemia in adults, marked by the abnormal proliferation of myeloid blasts in the bone marrow or other tissues^1,2. Despite advancements in treatment like chemotherapies and hematopoietic stem cell transplantation (HSCT), the overall survival rate of relapsed or refractory AML (r/r AML) patients has not shown significant improvement². Typically, the 5-year survival rate of AML patients is about 10%³. Furthermore, the median age at diagnosis for AML is 68 years. It is noteworthy that 54% of patients are diagnosed at 65 or older, with approximately one-third of cases occurring in individuals aged 75 years and above⁴. The feasibility of HSCT is often compromised in the geriatric population due to the high prevalence of concurrent medical conditions^5–7.

Acute myelomonocytic (M₄) and acute monocytic (M₅) leukemia are distinct subtypes of AML characterized by the excessive growth of abnormal mononuclear lineage cells and aggressive tissue infiltration. The M₄/M₅ subtypes of AML make up ~1/3 of all AML cases⁸. The therapeutic efficacy for M4/M5 AML is generally considered suboptimal in the clinic⁹. As a promising immunotherapy, chimeric antigen receptor (CAR) T cell therapy has demonstrated promise in treating B-cell malignancies. However, extending the success of CAR T cells to treat AML still faces challenges such as antigen heterogeneity, immunosuppressive microenvironments and poor CAR T persistence. Several tumor antigens such as CD33, CD123, and CLL-1 have been explored for AML treatment. Nevertheless, clinical studies of these CAR T cells for AML have reported suboptimal clinical outcomes^10–12. While CD33 is a potential target for M₄/M₅ AML, its expression on normal hematopoietic stem cells poses a significant safety risk. This underscores the urgent need to discover M4/M5-specific targets that can minimize off-tumor toxicity¹³.

CD64 (FcγRI), a Fc receptor for IgG, is constitutively expressed on AML cells, especially M₄ and M₅, whereas it is restricted to macrophages, monocytes, neutrophils, and myeloid progenitors in healthy bone marrow^14–16. Our previous study has shown that CD64 antigen is a potential target for AML CAR T therapy. However, a rapid relapse of AML disease was observed, which may be associated with insufficient proliferation of CD64 CAR T cells¹⁷. Studies have shown that CAR T cell-derived cytokines can boost AML survival, leading to CAR T cell exhaustion¹⁸. It is believed that inhibitory receptors and suppressor cells may limit the efficacy of CAR T cells in AML¹⁹. Interleukin-15 (IL-15) is a potent cytokine that has been shown to activate T cells and promote their differentiation into memory phenotypes. Specifically, IL-15 supports the long-term survival of CD8⁺ memory T cells and augments anti-tumor activity in vivo²⁰. In addition, overexpression of IL-15 has also been demonstrated to enhance the efficacy of CD19, CLL-1 and GPC-3 CAR T cells^21–23. Notably, IL-15 has also shown potency in promoting CAR T cell expansion in an AML model. This phenomenon may be attributed to the combined effects of IL-15 promoting expansion and sustained enhancement. We therefore engineered a novel CD64 CAR T cells with overexpressing IL-15 and compared them with CD64 CAR T cells. We performed the phenotypic characterization of the constructed IL-15 overexpression CAR T (IL-15 CAR) cells and evaluated them for anti-tumor activity, pharmacokinetics, and efficacy against CD64-positive-specific tumor cells. We found that IL-15 significantly improved CD64 CAR T cell expansion and anti-tumor activity, allowing the IL-15 CAR T cell product, but not the CAR alone, to eradicate the disease. Moreover, we have evaluated transcriptional profiles of IL-15 CAR T cells to elucidate signaling pathways mediated by IL-15.

Results

CD64 CAR T cells lack sustained disease control and have limited anti-AML efficacy

Our previous study has shown that the CD64 antigen is a potential target for AML. CD64 CAR T cells demonstrated robust tumor eradication at a higher dose¹⁷. However, the response rate of clinical trials of CAR T cells for AML is still unsatisfactory, and clinical settings are usually confronted with high tumor burden or limited CAR T cell manufacturing yields²⁴. It is important to know whether CAR T cells will work at a low dose. Therefore, we reduced the number of CD64 CAR T cells from 3×10⁶/mouse (the setting of our previous report) to 1×10⁶/mouse, but the number of tumor cells remained unchanged (Fig. 1A). CD64 CAR T cells became undetectable at day 20 post-infusion (Fig. 1B). Bioluminescence imaging (BLI) monitoring demonstrated that CD64 CAR T cells only transiently suppressed leukemic burden (Fig. 1C). Despite improved median survival compared to the PCDH (as the empty vector control) T cell group, a low-dose of CD64 CAR T cells could not completely eradicate tumors in mice (Fig. 1C, D). Thus, although CD64 CAR T cells exhibited initial efficacy, their expansion capacity and persistence may not be sufficient to sustain effective disease control in vivo.

Fig. 1. CD64 CAR T cells lack sustained disease control and have limited anti-AML efficacy.

A Schematic of mouse model. B The number of CD64 CAR T cells in tail blood. The PCDH T group was used as a flow-cytometric gating control for CAR⁺ cells and was not used for quantitative proliferation analyses. C Bioluminescence imaging of mouse model before and after CAR T cell infusion. D Survival curve of NSG mice, n = 3. CAR T cells were generated from one healthy human donor. Data in (B) are shown as mean ± SEM. Survival curve was compared using the log-rank Mantel–Cox test.

IL-15 overexpressing CD64 CAR T cells have enhanced cytolytic activity against AML

Prior research confirms that exposure to cytokines can enhance both the persistence and killing capacity of CAR T cells^21–23. Notably, IL-15 has shown potency in promoting CAR T cell expansion in an AML model²². To investigate whether IL-15 overexpression could facilitate durable anti-leukemic function, we incorporated the gene encoding human IL-15 into our CD64 CAR vector, linked via a P2A peptide (Fig. 2A). The process of manufacturing CAR T cells and subsequent experiments is shown in Fig. 2B. There was no difference in transduction efficiency between the two vectors, whether overexpressing IL-15 or not (Fig. 2C). However, CAR⁺ T cells were purified by flow cytometry, facilitating subsequent characterization assays (Supplementary Fig. 1A). The secretion of IL-15 after stimulation with PMA/ionomycin mixture was confirmed by ELISA assay, but in the absence of stimulation, IL-15 secretion was minimal or undetectable (Fig. 2D). Intracellular cytokine staining (ICS) was performed to assess cytoplasmic IL-15 expression. Under resting conditions, no cytoplasmic IL-15 expression was observed in IL-15 CAR T cells (Supplementary Fig. 1B). However, following stimulation with PMA/ionomycin and the subsequent inhibition of secretion with GolgiStop, cytoplasmic IL-15 expression was slightly detected within the cytoplasm (Fig. 2E). These results indicated that production and secretion of IL-15 may require strong activation.

Fig. 2. IL-15 overexpressing CD64 CAR T cells have enhanced anti-AML functionality.

A Schematic of CAR constructs. B Schematic of CAR T manufacturing procedure, detection and subsequent experiments. C Flow cytometric plots showing the percentage of CAR⁺. D The levels of IL-15 in supernatant of CAR T cell stimulated with PMA/ionomycin mixture, as determined by ELISA. E Representative histograms showing intracellular IL-15 expression in CAR T cells following stimulation with PMA/ionomycin mixture. F Cytolytic activity of CAR T cells against U937, THP-1 and MOLM-13 cells in vitro, detected by flow-based killing. G The levels of IL-2, IL-10, IFN-γ, and TNF-α in culture supernatant of CAR and U937, THP-1 or MOLM-13 cocultures after 24 h, as detected by LEGENDplex™. Data are from three independent biological replicates (n = 3) and each dot represents a different donor in (D, F, G). For all bar plots, data are shown as mean ± SEM. One-way ANOVA with Tukey’s multiple comparisons was used to assess significance in (D, F, G). ns not significant.

To determine whether overexpression of IL-15 or supplementary exogenous IL-15 more strongly enhances the function of CAR T cells, we first titrated supplementary IL-15 and identified 10 ng/mL as the optimal concentration (Supplementary Fig. 1C, D), which corroborates data from prior investigations^25,26. We performed subsequent assays to evaluate the function of CAR T cells cultured with 100 U/mL IL-2 or 100 U/mL IL-2 combination of 10 ng/mL IL-15.

We assessed the short-term CAR T cell cytotoxicity against CD64-positive cells, including U937 and THP-1, as well as the partially CD64-expressing MOLM-13 cells, whose CD64 expression was confirmed in our previous study¹⁷. The cytotoxicity assays were performed using flow-based killing, and precise counting beads were used to quantify residual tumor cells. Our results showed that IL-15 CAR T cells possessed markedly superior cytotoxicity to CAR T cells against U937 and MOLM-13 at 1:3 and 1:9 E:T ratios, as well as against THP-1 at 1:9 E:T ratio. Supplementing with exogenous IL-15 (CAR T + IL-15) induced significantly higher cytotoxicity against U937 at 1:3 and 1:9 E:T ratios and against THP-1 at 1:9 E:T ratio (Fig. 2F). IL-15 CAR T cells tended to exhibit higher cytotoxicity than CAR T + IL-15, though the difference was small and not statistically significant. The varied responses of tumor cell lines to CAR T cells may reflect their intrinsic heterogeneity. Using LEGENDplex™, we profiled cytokine secretion from CAR T cells co-cultured with three cell lines. Relative to CAR T cells, IL-15 CAR T cells released substantially elevated quantities of IL-2, IL-10, and IFN-γ when stimulated with U937 cells, higher IFN-γ with THP-1 cells, and higher IL-10 with MOLM-13 cells. Compared to CAR T + IL-15 cells, IL-15 CAR T cells showed significantly higher IFN-γ secretion upon stimulation with U937 and THP-1 cells but not MOLM-13 cells. CD64 CAR T cells secreted higher levels of TNF-α than IL-15 CAR T cells in response to all three cell lines, and also surpassed CAR T + IL-15 cells in response to U937 and MOLM-13 cells (Fig. 2G). However, no difference in granzyme A and granzyme B levels was observed among the three CAR groups when co-cultured with U937 and MOLM-13 cells. When co-cultured with THP-1 cells, IL-15 CAR T and CAR T + IL-15 cells showed significantly lower levels of granzyme A and granzyme B than CAR T cells (Supplementary Fig. 1E). These variations likely reflect intrinsic differences among the target cell lines, such as antigen density.

IL-15 overexpressing CD64 CAR T cells expanded in vitro maintain the memory phenotype

To evaluate the sustained capacity and phenotypic characteristics of the CAR T cells, a repetitive antigen challenge assay was employed (Fig. 3A). Cells were cultured in medium enriched with either 100 U/mL IL-2 alone or a mixture of 100 U/mL IL-2 and 10 ng/mL IL-15, with cell enumeration conducted every 72 h. U937 cells were replenished to maintain a 1:3 E: T ratio. Flow cytometry was used to distinguish T cells and U937 cell surface markers and counting beads were used to quantify cell numbers. A notable shift in the CD4⁺/CD8⁺ balance was observed, with the IL-15 CAR T group displaying a significantly higher proportion of CD8⁺ T cells, while CAR T + IL-15 cells had no effect on this ratio (Fig. 3B). As predicted by the short-term cytotoxicity assay, the benefits of IL-15 CAR T cells became more apparent when we measured T cell proliferation in repeated killing assays against the U937 cells. IL-15 CAR T cells exhibited a significant dominant expansion following coculture with U937 cells, approximately doubling the expansion observed in CAR T cells (Fig. 3C). In addition, supplementation of exogenous IL-15 modestly enhanced the expansion capacity of CD64 CAR T cells compared to the untreated control group. However, it remained inferior to that of IL-15 CAR T cells. Without continuous exogenous IL-15 supplementation, this expansion superiority diminished (Fig. 3C). Central memory, but not effector memory, is associated with better T cell proliferative vigor and extended T cell persistence^27,28. We analyzed the immunophenotype of CAR T cells on day 7, following repeated stimulation with target cells. IL-15 CAR T cells exhibited higher levels of central memory (CM; CD62L⁺CD45RA^-) and a trend toward increased stem cell memory (SCM; CD62L⁺CD45RA⁺CD95⁺) than CAR T and CAR T + IL-15 cells (Fig. 3D–F).

Fig. 3. IL-15 overexpression promoted CAR T cell proliferation and increased the proportion of central memory.

A Schematic of antigen repeat stimulation model. Quantification of CAR T cells was performed by using precise counting beads at three-day intervals. U937 cells were replenished to maintain a 1:3 E: T ratio. B The CD4⁺/CD8⁺ ratio on day 7 after initial stimulation with U937 cells. C CAR T cell expansion using repeated U937 stimulation in vitro. For the CAR T groups, the CAR⁺ cell count was shown. For the PCDH control group, the total surviving T cell count was used directly. Statistics was performed using the data from day 12. D–F Representative flow cytometric plots (D), scatter plots (E) and stacked bar charts (F) showing phenotype subsets compositions of CAR T cells on day 7 after initial stimulation with U937 cells. Data are from three independent biological replicates (n = 3) and each dot represents a different donor in (B, C, E, F). For all bar plots, data are shown as mean ± SEM. One-way ANOVA with Tukey’s multiple comparisons was used to assess significance in (B, C, E). ns not significant.

IL-15 overexpressing reduces apoptosis and alleviates exhaustion

Flow cytometry was employed to profile surface markers associated with activation, apoptosis, and exhaustion, with data quantified as the percentage of CAR⁺ for CAR T cell groups or percentage of T cells for PCDH group. Although CAR T cells showed higher general activation than PCDH controls, specific markers CD69 and CD25 did not vary significantly among the three CAR T groups (Fig. 4A, B). Of note, both IL-15 CAR T cells and CAR T + IL-15 cells exhibited a lower proportion of Annexin V-positive apoptotic cells than CAR T cells (Fig. 4C).

Fig. 4. IL-15 overexpression reduces apoptosis and alleviates exhaustion.

A Representative flow cytometric plots, scatter plots showing the percentages of CD69. B Representative flow cytometric plots, scatter plots showing the percentages of CD25. C Representative flow cytometric plots, scatter plots showing the percentages of Annexin V. D Representative flow cytometric plots, scatter plots showing the percentages of TIM-3 and LAG-3. E Representative flow cytometric plots, scatter plots showing the percentages of IFN-γ and TNF-α. All data were collected on day 7 after initial stimulation with U937 cells. For all bar plots, data are shown as mean ± SEM. Data are from three independent biological replicates (n = 3) and each dot represents a different donor. One-way ANOVA with Tukey’s multiple comparisons was used to assess significance. ns not significant.

T cell exhaustion is characterized as a state of functional impairment driven by the upregulation of inhibitory receptors, such as PD-1, LAG-3, and TIM-3, which correlates with compromised antitumor potency^29,30. Following stimulation with U937 cells, the CD8⁺ IL-15 CAR T subset displayed a marked reduction in LAG-3 and TIM-3 expression relative to both the CAR T and CAR T + IL-15 cells, whereas no such variation was detected within the CD4⁺ population (Fig. 4D). However, PD-1 expression was comparable among the three CAR T cell groups (Supplementary Fig. 2). Furthermore, we observed that stimulation enhanced the expression of IFN-γ and TNF-α by IL-15 CAR T cells relative to the other two CAR T groups (Fig. 4E).

IL-15 overexpressing CD64 CAR T cells exhibit a highly proliferative and central memory transcriptomic signature and upregulate genes of PI3K-Akt signal pathway

To examine gene expression changes caused by IL-15 overexpression, RNA sequencing was conducted. CD4⁺ and CD8⁺ CAR T cells were sorted using FACS Aria II, followed by RNA-seq. Venn diagrams of expressed genes in sorted CD8⁺ and CD4⁺ T cells showed a large core transcriptome shared by all groups (Supplementary Fig. 3A). Through DESeq2, we discovered genes with changes in expression (|log2 (fold change)| ≥1.5 and P < 0.05). The results revealed that T cells engineered with the CD64 CAR, compared with the empty vector PCDH, produced extensive transcriptional remodeling (1034 genes upregulated, 1214 downregulated in CD8⁺ cells; 840 genes upregulated, 1253 downregulated in CD4⁺ cells), which was involved genes such as cytokine-cytokine receptor interaction, regulation of T cell activation/chemotaxis and PI3K/ protein kinase B (AKT) signaling (Supplementary Fig. 3B, C). IL-15 CAR T cells elicited a modest transcriptional reprogramming compared to CAR T cells (72 upregulated, 112 downregulated genes in CD8⁺ cells; 40 upregulated, 72 downregulated in CD4⁺ cells), enriching pathways linked to PI3K/AKT signaling, mitochondrial gene expression, glycogen/purine metabolism, leukocyte proliferation, and oxidative-stress responses (Fig. 5A and Supplementary Fig. 3B). These signatures indicate that IL-15 primarily enhances proliferative capacity, survival and metabolic fitness rather than broadly amplifying cytokine transcription. Exogenous IL-15 supplementation induced only minor changes (30 upregulated and 33 downregulated in CD8⁺ cells; 18 upregulated and 22 downregulated in CD4⁺ cells) compared to CAR T cells, suggesting that overexpression of IL-15 is the main driver of the observed reprogramming (Supplementary Fig. 3B, D). Importantly, IL-15 CAR T still showed significant differences (107 upregulated, 154 downregulated in CD8⁺ cells; 69 upregulated, 71 downregulated in CD4⁺ cells) compared to CAR T + IL-15 cells, suggesting that overexpression of IL-15 reprograms T cells in a manner that could not be reproduced by transient IL-15 cytokine exposure (Fig. 5B and Supplementary Fig. 3B).

Fig. 5. IL-15 overexpressing CD64 CAR T cells exhibit a highly proliferative and central memory transcriptomic signature.

A Volcano plots (left) representing differentially regulated genes in CD8⁺ IL-15 CAR T and CAR T cells. The Gene Ontology (GO) processes (right) enriched by Metascape. B Volcano plots representing differentially regulated genes in CD8⁺ IL-15 CAR T and CAR T + IL-15 cells. Blue dots (A, B) are those not meeting criteria: P value ≥ 0.05; log2 fold change >1.5 or <1.5. C, D GSEA showing the genes involved in T cell proliferation (C) and central memory (D). E Dot plots displaying the expression of IL15, CD28, TCF7, EOMES genes. Data are from three independent biological replicates (n = 3) and each dot represents a different donor. The ratio paired t test was used to assess significance. ns not significant.

Gene set enrichment analysis (GSEA) demonstrated significant upregulation of the gene set involved in T cell proliferation and central memory in IL-15 CAR T versus CAR T and CAR T + IL-15 cells (Fig. 5C, D). CD8⁺ IL-15 CAR T cells showed higher expression of proliferation genes (IL15, CD28) and memory genes (TCF7, EOMES) than CD8⁺ CAR T cells, as well as higher IL15 and CD28 levels relative to CAR T + IL-15 cells. Among CD4⁺ cells, IL-15 CAR T cells exhibited greater IL15 expression than the other two groups and higher TCF7 expression than CAR T + IL-15 cells (Fig. 5E). A non-significant upregulation in the expression of the memory-associated markers LEF1, SELL, and CCR7 was observed in both CD8⁺ and CD4⁺ IL-15 CAR T cells compared to CAR T cells and CAR T + IL-15 cells (Supplementary Fig. 3E). LAG-3 and TIM-3 protein expression on CD8⁺ IL-15 CAR T cells was lower than on CD8⁺ CAR + IL-15 cells or CD8⁺ CAR T cells, but this difference was not reflected at the transcriptional level in either CD4⁺ or CD8⁺ cells (Supplementary Fig. 3F). Notably, while cytoplasmic and soluble TNF-α protein differed among groups by ICS and LEGENDplex™, TNF mRNA was comparable, indicating that IL-15 exerts post-transcriptional control of TNF-α expression (Supplementary Fig. 3G).

IL-15 overexpressing CD64 CAR T cells demonstrated superior anti-tumor activity and improved persistence in AML mouse model

The results suggested that the benefits of IL-15 overexpression extend beyond exogenous IL-15 supplementation in vitro assays. Consequently, we proceeded to assess the in vivo potency of these IL-15 CAR T cells using an AML xenograft model. In the initial experiments, a dosage of 1 × 10⁶ CD64 CAR T cells per mouse resulted in an observation window that was too short to examine cellular kinetics. To extend the observation window for the pharmacodynamic analysis of both CAR T and IL-15 CAR T cells and BLI kinetics, we subsequently increased the dose to 1.5 × 10⁶ CAR T cells per mouse. A single dose of 3 × 10⁵ luciferase-expressing U937 cells (U937-Luc) was intravenously injected into the NSG mice via the tail vein at Day -5 and a single dose of 1.5 × 10⁶ CAR or PCDH T cells was infused on Day 0 (Fig. 6A). The results showed that IL-15 CAR T cells cleared AML cells earlier than CAR T cells and maintained a longer remission (Fig. 6B). Flow cytometry was used to measure CAR T cell levels in the tail blood (Supplementary Fig. 4). IL-15 CAR T cells showed enhanced proliferation on day 20 and sustained this persistence through day 30, significantly outperforming the CAR T group (Fig. 6C). Both CAR T cell products extended the survival of mice with tumors compared to the PCDH control group, but IL-15 CAR T cells had a more pronounced positive impact on survival (median survival in days: IL-15 CAR T vs CAR T: 70 vs 30; Fig. 6D). Furthermore, BLI monitoring showed that leukemia was effectively eradicated in the IL-15 CAR T cohort by day 15, whereas the CAR T and PCDH groups continued to display high tumor burden (Fig. 6E).

Fig. 6. IL-15 overexpressing CD64 CAR T cells demonstrated superior antitumor activity and no obvious toxicity in vivo.

A Schematic of mouse model. B Quantification of the tumor burden as indicated by average radiance (p/sec/cm²/sr), n = 6. C The number of CAR T cells in tail blood, n = 6. D Survival curve of NSG mice, n = 6. The survival data were obtained through direct observation and not through bioluminescence imaging beyond day 28. E Bioluminescence imaging of mouse model before and after CAR T cell infusion. F Cytokine levels in tail blood collected on day 10 (n = 6) and day 20 (n = 5 in CD64 CAR T group and n = 6 in IL-15 CAR T group) post-infusion. CAR T cells were generated from two independent human donors, with three mice per group receiving cells from each donor. Data from both donors were pooled for statistical analysis to assess the efficacy. G Pathological analysis of the heart, liver, spleen, lung, and kidney of the representative three mice at their terminal stage by using HE staining. Magnification, ×20. For all bar plots, data are shown as mean ± SEM. Survival curve was compared using the log-rank Mantel–Cox test in (D). Multiple two-sided unpaired t tests were used to assess significance in (F).

The serum from the mice was collected on days 10 and 20, and the levels of cytokines and cytotoxic proteins were analyzed. The results showed that IL-15 CAR T cells produced significantly more cytotoxic protein perforin than CAR T cells on day 10, while cytokine IFN-γ exhibited opposite results on day 20 (Fig. 6F). Perforin and granzyme A were also increased in IL-15 CAR T at day 20, although the difference was not statistically significant. IL-2, IL-6, IL-10, IL-17A, TNF-α, and granzyme B levels did not differ significantly. This result suggests that the enhanced in vivo cytotoxicity of IL-15 CAR T cells may not be due to cytokines but to enhanced release of cytotoxic proteins.

We found that the IL-15 CAR T group maintained good activity levels and showed no abnormalities in general condition or skin/fur condition until the week before death, indicating that tumor progression did not significantly affect them. However, during the final five to seven days of life, the mice rapidly developed classic late-stage symptoms (e.g., markedly reduced activity, hunched posture, and matted fur). Mice were monitored daily and euthanized upon reaching a moribund state. Immediately after euthanasia, the heart, liver, spleen, lung, and kidney were harvested for histopathological evaluation. Pathological evaluation showed that the heart, liver, spleen and kidney revealed no appreciable inflammatory cell infiltration in IL-15 CAR T cell group, while alveolar spaces were infiltrated by neutrophils (arrows) to a lesser extent in both CAR T and IL-15 CAR T cell groups (Fig. 6G). The results suggested that IL-15 CAR T did not cause severe inflammatory organ toxicity in mice.

Discussion

We have previously reported a novel AML-associated antigen marker, CD64, as a promising target for CAR therapy. However, a rapid disease relapse of AML was observed, which may be associated with insufficient expansion of CD64 CAR T cells¹⁷. CAR T cell-derived cytokines have been shown to boost AML cell survival, leading to CAR T cell exhaustion¹⁸. Concurrently, checkpoint pathways and suppressor in AML also limit efficacy of CAR T cell therapy¹⁹. IL-15 has been recognized as a critical cytokine for T homeostasis, survival, proliferation, and effector function, prompting us to investigate its therapeutic overexpression in CAR engineering²⁰. Here, we describe the functional evaluation of CD64 CAR T cells engineered to overexpress IL-15. Further understanding of the favorable function of IL-15 overexpression CAR T cells might benefit more r/r AML patients.

In our study, IL-15 CAR T cells only showed detectable levels of cytoplasmic IL-15 and soluble IL-15 after PMA/ionomycin stimulation. The results indicate that IL-15 production by IL-15 CAR T cells is tightly regulated and significantly induced upon strong activation, which is consistent with prior reports on the regulation of endogenous IL-15^31,32. This suggests that overexpressed IL-15 may remain physiologically regulated by post-transcriptional and secretory pathways similar to those of endogenous IL-15. Practically, such regulation may help limit off-target inflammation under resting conditions, while preserving strong functional augmentation upon antigen encounter.

Our data suggested that while IL-15 supplementation transiently enhances CAR T expansion and function, stable overexpression of IL-15 provides a more robust and durable advantage in vitro. However, we did not compare the function of these two groups in vivo. First, treatment with exogenous IL-15 presents substantial clinical challenges due to its short half-life, resulting in the need to administer high and frequent doses or give continuous infusions^33,34. This pharmacokinetic profile inevitably increases the risk of toxicity. A clinical study demonstrated that patients receiving recombinant human IL-15 therapy may experience grade 3-5 toxicities, such as cytokine release syndrome (CRS), alveolar hemorrhage and prolonged bone marrow hyperplasia³⁵. Second, our in vitro results suggested that the benefits of IL-15 overexpression extend beyond those of exogenous IL-15 supplementation, potentially achieved through the synergistic effects of sustained autocrine signaling and intrinsic CAR signaling. Third, RNA-seq also indicated that exogenous IL-15 supplementation induced only minor transcriptomic changes in CAR T cells, whereas overexpression of IL-15 establishes a transcriptional state not recapitulated by transient exposure.

We observed that IL-15 overexpression CAR T cells exhibited superior expansion after tumor stimulation compared to CAR T cells and CAR T cells with IL-15 supplementation. The elevated proportion of central memory subset in the IL-15 CAR T group suggested that IL-15 signaling enhances memory phenotype development, potentially contributing to improved longevity and functional capacity of these engineered immune cells. In addition, it has been proven by other studies that central memory T cells are better at maintaining immune responses^26,36,37. RNA-seq showed that genes involved in memory phenotype were upregulated in IL-15 CAR T cells, consistent with their sustained functional persistence. Our results also suggested that IL-15 activates PI3K-AKT pathway, upregulating pro-survival programs that preferentially sustain less-differentiated memory T cells. IL-15 also promotes mitochondrial biogenesis, thereby reinforcing persistence and self-renewal capacities characteristic of TCM/TSCM. This phenotype reflects a direct and expected consequence of IL-15 signaling that aligns closely with established IL-15 biology^26,38. IL-15 overexpression appears to activate Wnt/β-catenin signaling, inducing the downstream transcription factor TCF/LEF-mediated of pro-survival and proliferation effector programs^39–41.

A critical limitation of sustained T cell proliferation is the progressive differentiation into terminally exhausted subsets, which results the diminished persistence and attenuated anti-tumor activity. CD8⁺ IL-15 CAR T cells had lower levels of the inhibitory molecules TIM-3 and LAG-3 proteins, despite no significant reduction in transcriptional levels, suggesting that IL-15 may inhibit the expression of inhibitory molecules at the post-transcriptional level.

In our study, IL-15 overexpression selectively enriched CD8⁺ T cell populations, aligning with established roles of IL-15 in driving CD8⁺ expansion and effector differentiation²⁶. This indicates that IL-15 can enhance the effector function of CAR T cells without increasing or possibly even decreasing their inhibitory effect. Flow cytometry and RNA-seq showed that the CD8⁺ and CD4⁺ IL-15 CAR T cell populations were similar in some phenotypes (promoted proliferation, sustained memory phenotype and increased expression of some proliferation markers, etc.). Although CD4⁺ and CD8⁺ T cells have some inherent differences in their response to IL-15, these findings indicate a common regulatory pathway that enhances the anti-tumor effectiveness of IL-15 CAR T cells.

Previous studies have generally reported increased TNF-α production under IL-15 transgene overexpression^22,42. In this study, we detected lower levels of soluble TNF-α in IL-15 CAR T cells after co-culturing with three tumor cell lines (U937, THP-1 and MOLM-13) than in CAR T cells. The discrepancy between our study and others may be due to specific variables, such as differences in target antigens, tumor models, and CAR structure. However, TNF-α mRNA levels were similar in both groups. Intracellular TNF-α protein levels, as measured by intracellular cytokine staining with GolgiStop to block secretion, were slightly higher in IL-15 CAR T cells. The reason for the reduced TNF-α secretion in IL-15 CAR T cells compared to CAR T cells without IL-15 overexpression may involve altered secretion pathways, which warrant further investigation.

In our previous study, CD64 CAR T cells effectively cleared tumor cells at a higher dose (3 × 10⁶ cells/mouse). However, the response rate of clinical trials of CAR T cells for AML is still unsatisfactory, and clinical settings are usually confronted with high tumor burden or limited CAR T cell manufacturing yields²⁴. Therefore, it is important to know whether CD64 CAR T cells  retain efficacy at a low dose. In this study, reducing the dose to 1 × 10⁶ CAR T cells/mouse established a more challenging therapeutic model than in previous studies. To enable clear comparisons of CAR T cell kinetics, we adopted a higher dose (1.5 × 10⁶ cells/mouse) in our studies of IL-15 CAR T cell in vivo function. This was necessary because the preliminary low-dose regimen, which revealed the limited efficacy of CD64 CAR T cells, provided an insufficient observation window for detailed pharmacodynamic analyses. The optimized model successfully captured the superior expansion and persistence of IL-15 CAR T cells, directly linking these kinetics to their enhanced therapeutic outcome. The differences in dosage reflect the different purposes of the two sets of experiments in this study.

Notably, preclinical studies have prominently highlighted the potential risk of IL-15-mediated toxicity in mice, which may manifest as weight loss, splenomegaly, and significant inflammatory alterations in the lungs and other tissues²². Sánchez-Moreno et al. conducted a study in which they developed a novel approach to reduce toxicity by modifying CAR T cells to co-express IL-15 and IL-15Rα simultaneously, limiting systemic exposure to IL-15 while maintaining anti-tumor activity²¹. Nevertheless, we did not observe obvious toxicities associated with IL-15 overexpression. No significant increase in pro-inflammatory cytokine TNF-α was observed with IL-15 CAR T, and other cytokines also showed a modest response in vivo. Specific CAR constructs, AML models and immune microenvironments may be able to influence cytokine dynamics and potentially modulate the inflammatory response to reduce toxicity while maintaining anti-tumor efficacy. In addition, previous study has shown that CD4⁺ CAR T cells play a major role in eliciting CRS⁴³. The reduction in the proportion of CD4⁺ CAR T cells may also contribute to the reduction of potential CRS. Furthermore, effective controlled release of IL-15 may allow for a more targeted and less toxic therapeutic approach, although this hypothesis requires further validation in future studies.

Although our study showed favorable efficacy, several limitations must be considered. An important limitation is using a single AML xenograft model, which may affect the generalizability of our findings, as different AML subtypes may exhibit varying responses to IL-15-enhanced CAR T cell therapy. Furthermore, previous reports suggest that IL-15-mediated toxicity may be a concern that cannot be ignored, although we did not observe obvious toxicity in our study. The colony formation assay using the same CD64 CAR construct was performed in our previous study and no obvious toxicity to human hematopoietic stem/progenitor cells (HSPCs) was observed. Furthermore, CD64 expression is low in HSPCs as determined by flow cytometry¹⁷. Using appropriate mouse models in the future, such as NSG mice engrafted with human CD34⁺ HSPCs, could help assess in vivo toxicity more effectively. In addition, we did not perform receptor-level or genetic-level blockade of IL-15 receptor to investigate its functional role in IL-15 signaling, which is likely complex and warrants a comprehensive, systemic study. Future work should employ CRISPR-Cas9 knockout of IL-15Rα on CAR T cells to fully elucidate the influence of IL-15 signaling on CAR T cell function.

In conclusion, this study demonstrates that the overexpression of IL-15 in CD64 CAR T cells significantly enhances anti-tumor activity, expansion, promotes TCM/TSCM, modulates gene expression, providing a promising strategy of CAR T cell therapies for r/r AML. Cytokine modulation represents a promising strategy to improve CAR T cell therapies, and our findings provide valuable insights for future therapeutic development and translating IL-15 CD64 CAR T therapy to the clinic to improve outcomes for r/r AML patients.

Methods

Cell lines

The leukemia cell lines used in the study included the U937, the THP-1, and the MOLM-13 cell lines, all of which were obtained from the Cell Resource Center of the State Key Laboratory of Experimental Hematology (Institute of Hematology & Blood Disease Hospital) in the study. Human cell lines were validated using short tandem repeat (STR) profiling, and all cell cultures were confirmed to be mycoplasma-negative prior to experimentation. Tumor cell lines were stably transduced with green fluorescent protein and luciferase (GFP⁺ luc⁺) by lentivirus, and purified on FACSAria II Cell Sorter (BD Biosciences). Leukemia cell lines were cultured in RPMI 1640, while HEK293T cells were cultured in high-glucose DMEM; both medium were supplemented with 10% fetal bovine serum (FBS). All cultures were kept at 37°C in a humidified 5% CO₂ incubator.

CAR construct

The CAR construct was designed using the humanized CD64-specific single chain variable fragment (scFv, clone H-22) sequence and the CD8α hinge, CD8 transmembrane (TM), 4-1BB and an intracellular CD3ζ signaling domain. A P2A peptide sequence connected the CAR sequence and transgene IL-15. Sequences were synthesized by GENEWIZ and subcloned into PCDH lentiviral empty vector. All sequences were validated by Sanger sequencing.

Lentiviral vector production

All of the CAR expression plasmids and empty vector PCDH were constructed using a third-generation lentiviral vector. CAR expression plasmids were co-transfected with Rev, pMDL (packaging plasmid) and VSV-G (envelope plasmid) into HEK293T cells (80–90% confluent) using polyethyleneimine (PEI, Polysciences) as the transfection agent. The viral supernatant was collected at 48 h post-transfection and centrifuged at 1000× g for 15 min, then filtered through a 0.45 μm filter. Lentiviral stocks were stored at −80 °C until use.

T cell isolation and CAR T cell production

Human blood was collected from healthy volunteers. All ethical regulations relevant to human research participants were followed. T cells were isolated with human T cell enrichment cocktail (Stemcell Technologies). Following a 48h stimulation with ImmunoCult™ Human CD3/CD28 T Cell Activator (Stemcell Technologies), the cells were transduced with CAR or empty vector (PCDH) lentiviruses. The PCDH T cell group refers to a control group using the empty lentivirus vector. Cultures were maintained in Excell Bio medium supplemented with 100 IU/mL IL-2 (T&L Biotechnology), with or without 10 ng/mL IL-15 (Peprotech). CAR T cells were purified on FACSAria II Cell Sorter (BD Biosciences). All of the experiments in the manuscript were performed using purified CAR T cells within 14 days of initial activation.

Flow cytometry

CD64-specific protein (ACRO) was used to detect CAR-expressing. The cells were incubated with monoclonal antibodies for 30 min at 4 °C, protected from light. Subsequently, the samples were resuspended in 200 μL PBS containing 2% FBS. The concentrations of antibodies used were the concentration recommended by the manufacturer. Cells were quantified by using Precision Counting Beads (BioLegend). Cell viability was tested by DAPI (Solarbio) or 7-AAD (BD Bioscience). Flow cytometry was performed on Canto II, LSR II or LSR Fortessa (all from BD Biosciences) with subsequent data analyzed by FlowJo.

Intracellular cytokine staining

CAR T cells or PCDH T cells were unstimulated or stimulated with PMA/ionomycin mixture (250×; Multi Science) and GolgiStop (1200×; BD Biosciences) for 4.5 h, surface proteins were stained by incubation with fluorochrome-conjugated antibodies. Following viability staining with Fixable Viability Dye (eBioscience), cells were processed using the Foxp3/Transcription Factor Staining Buffer Set (eBioscience) to achieve fixation and permeabilization.

Enzyme-linked immunosorbent assay

The supernatants were collected from CAR cultures stimulated with PMA/ionomycin mixture (250×) (Multi Sciences) overnight or unstimulated wells. ELISA was performed on 100 μL samples. IL-15 Human ELISA Kit was obtained from Neobioscience. Values were obtained using Synergy H4 Hybrid Microplate Reader (BioTek) to determine optical density (OD) and measured against standards provided, following by the manufacturer’s instructions.

Exogenous IL-15 supplementation

To define a working concentration for exogenous IL-15, conventional CAR T cells generated as described were cultured under identical conditions with recombinant human IL-15 (Peprotech) at 0, 1, 5, 10, or 30 ng/mL. IL-15 was added at culture set-up and re-added with each medium refresh to maintain the indicated concentration. Proliferation was quantified as fold-expansion of viable CAR⁺ cells for CAR groups or total T cells for PCDH group from baseline at the indicated time points (day 6 used for statistics), and short-term cytotoxicity was assessed against tumor cells at predefined E:T ratios.

In vitro killing assay

The cytolysis of CAR T cells was evaluated by 16-h killing assay using E: T ratios of 1:1, 1:3, and 1:9, determined by Flow Cytometry. Briefly, 3 × 10⁴ tumor cells were co-cultured with CAR T or control PCDH T cells in triplicate within U-bottom 96-well plates. Target-only wells served as negative controls. Cell viability was assessed using DAPI (Solarbio). To enable accurate counts of residual target cells, Precision Count Beads (BioLegend) were mixed into each tube. Specific lysis was determined by comparing the remaining live cell counts in the experimental group with those in the control PCDH T group.

Repeat antigen stimulation expansion

The pre-determination of CAR expression and cell count was conducted before conducting the killing assay. CAR T cells (2 × 10⁴) were co-cultured with U937-GFP⁺ cells (6 × 10⁴) in 96-well flat-bottom plates (triplicate) at an E: T ratio of 1:3. Three days later, cells were counted with flow cytometry using Precision Counting beads (BioLegend). CAR T cells were rechallenged with fresh target cells to reset the initial E:T ratio. This process was repeated for a total of four rounds. In total, 100 U/mL IL-2 or 100 U/mL and 10 ng/mL IL-15 was added in the medium. Fold expansion was calculated as (viable CAR T cells on day 3)/ (2 × 10⁴) for CAR groups, and (viable T cells on day 3)/ (2 × 10⁴) for PCDH group as control. The cumulative expansion was calculated by multiplying the fold changes obtained from consecutive rounds.

RNA-sequence (RNA-seq)

CAR T or PCDH cells were co-cultured with repeat U937 cell stimulation for 7 days. CD4⁺ and CD8⁺ CAR T cells were sorted using FACS Aria II. RNA was extracted by TRIZOL reagent (Thermo). Samples were sent to Novogene Biotech Co., Ltd. (Beijing, China). Briefly, the RNA integrity was assessed using the RNA Nano 6000 Assay Kit on the Bioanalyzer 2100 system. The mRNA-Seq libraries were sequenced using an Illumina Novaseq platform and 150bp paired-ended reads were generated. Quality control was performed using in-house Perl scripts of Novogene Biotech. Index for the human reference genome (v. hg38) was constructed using Hisat2 (v2.0.5). Subsequently, StringTie (v1.3.3b) was used to assemble the mapped reads for each sample. Differential expression analyses were performed using the DESeq2 R software package, which included those with absolute log₂ (fold change) ≥1.5 and p value < 0.05. GSEA was performed using GSEA software (http://www.broadinstitute.org/gsea), utilizing GOBP_T_CELL_PROLIFERATION gene sets from the Broad Institute Molecular Signature Database. T cell central memory gene set was obtained from the previously described study⁴⁴. For GSEA, thresholds for significant pathways were determined as: FDR  <  0.25 and P value  <  0.05. The top Gene Ontology (GO) processes were enriched by Metascape web-based platform (https://metascape.org)⁴⁵.

Mouse xenograft models

NSG mice (NOD-Prkdc^scid Il2rg^tm1/Bcgen) purchased from Biocytogen (Beijing, China) were housed under a specific pathogen-free (SPF) environment (12 h/12 h, light/dark cycle; 18–23 °C; 40-60% humidity). We have complied with all relevant ethical regulations for animal use. Facility staff monitored and recorded health status daily. Female mice between five and eight weeks old were given a tail injection of 3 × 10⁵ U937-GFP⁺luc⁺ cells on day −5. On day 0, they were infused with 1–1.5 × 10⁶ CAR⁺ T cells or PCDH T cells via tail injection. Using the IVIS system, the tumor burden of U937-GFP⁺luc⁺ was measured by BLI, and the results were analyzed with Living Image software (PerkinElmer). Survival was monitored throughout the study.

Histological analysis

Mice were euthanized upon reaching a moribund state. Key organs, including heart, liver, spleen, lung, and kidney were harvested, washed with PBS, and fixed in 4% paraformaldehyde for a minimum of 24 h. Subsequently, tissues were embedded in paraffin. Hematoxylin and eosin (H&E) was used for histopathological evaluation.

Analysis of cytokines

Mouse serum or coculture supernatant were harvested and preserved at −80 °C. Cytokines were detected using the LEGENDplex™ Human CD8/NK Panel (BioLegend) following the manufacturer’s protocol. Data collection was performed with Canto II (BD Biosciences), and the analysis was conducted using the ‘LEGENDplex’ Data Analysis online software (https://legendplex.qognit.com).

Statistical and reproducibility

Data are presented as mean ± SEM based on three independent biological experiments. Specific statistical methods are detailed within the figure legends. Significance was assessed using one-way ANOVA, whereas RNA-seq data were analyzed via ratio paired t tests to address inter-donor variability. The log-rank (Mantel–Cox) test was used to compare survival curves. Statistical analysis was performed using SPSS or Prism, with a P value of less than 0.05 indicating statistical significance. Final graphs were produced with Prism and FlowJo, and figures were assembled using Adobe Illustrator.

Ethics statement

Ethical approval for harvesting blood samples from healthy donors was granted by the Ethics Committee at the State Key Laboratory of Experimental Hematology (Institute of Hematology and Blood Disease Hospital, Chinese Academy of Medical Sciences & Peking Union Medical College) under approval number NKRDP2021009-EC-2. Informed consent was obtained from every participant. All animal procedures followed the institutional guidelines of laboratory animals in the State Key Laboratory of Experimental Hematology and were approved by the Institutional Committee.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

L.S. and X.F. designed the experiments. L.S. performed the majority of the experiments, generated experimental data. L.S. and X.F. wrote the manuscript. C.L. assisted with in vivo experiments. T.L. and C.W. assisted with some experiments during the revision process. H.C., A.P. and X.F. provided support and supervised the study. All authors are in agreement on the final version of the manuscript.

Peer review

Peer review information

Communications Biology thanks the anonymous reviewers for their contribution to the peer review of this work. Primary Handling Editors: Isabela Pedroza-Pacheco and Mengtan Xing. A peer review file is available.

Data availability

The corresponding author will address reasonable requests for datasets associated with this study, with a processing time of approximately three months. All essential data underpinning our conclusions are available within this article. RNA-seq data is available in the Genome Sequence Archive (GSA) database (https://ngdc.cncb.ac.cn/gsa-human/) under accession number HRA015107. Source data are provided in Supplementary Data 1.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s42003-026-09528-8.