Dual epitope anti-LILRB4 synthetic T-cell receptor and antigen receptor (STAR)-T-cell therapy for relapsed/refractory acute myeloid leukemia

Lv, Meng; Zheng, Hongli; Pei, Xuying; Lei, Lei; Chen, Wenbo; Bao, Yutong; Ding, Yiyang; Xu, Zhengli; Hu, Lijuan; Wan, Zhe; Zhao, Xueqiang; Zhou, Zhixiao; Lin, Xin; Huang, Xiaojun; Wang, Jiasheng; Zhao, Xiangyu

doi:10.1038/s41392-026-02765-7

Download PDF

Article
Open access
Published: 01 June 2026

Dual epitope anti-LILRB4 synthetic T-cell receptor and antigen receptor (STAR)-T-cell therapy for relapsed/refractory acute myeloid leukemia

Meng Lv ORCID: orcid.org/0000-0001-6625-9523¹^na1,
Hongli Zheng²^na1,
Xuying Pei¹^na1,
Lei Lei²^na1,
Wenbo Chen¹,
Yutong Bao¹,
Yiyang Ding¹,
Zhengli Xu¹,
Lijuan Hu¹,
Zhe Wan²,
Xueqiang Zhao²,
Zhixiao Zhou³,
Xin Lin ORCID: orcid.org/0000-0003-0956-3654³,
Xiaojun Huang ORCID: orcid.org/0000-0002-2145-6643^1,4,
Jiasheng Wang⁵ &
…
Xiangyu Zhao¹

Signal Transduction and Targeted Therapy volume 11, Article number: 203 (2026) Cite this article

Subjects

Abstract

Acute myeloid leukemia (AML) is a rapidly progressive malignancy with poor prognosis. To date, chimeric antigen receptor (CAR) T-cell therapy in AML has been limited by the lack of antigens with high specificity for AML cells. Here, we developed a synthetic T-cell receptor and antigen receptor-T (STAR-T) cell therapy targeting LILRB4, an immunosuppressive receptor highly expressed on monocytic AML blasts but not on hematopoietic stem cells. Two nanobodies with the highest affinity for LILRB4 were identified through phage display library screening and used to construct nanobody-based dual epitope anti-LILRB4 STAR-T cells, which exhibit more potent tumor inhibition in vitro and in vivo than single epitope anti-LILRB4 STAR-T cells or dual epitope anti-LILRB4 CAR-T cells do. We subsequently conducted the first human clinical trial (NCT05548088) involving 9 patients with LILRB4-positive relapsed and refractory (R/R) AML. Six patients completed the safety and efficacy evaluation (median follow-up, 10.7 months). No immune effector cell-associated neurotoxicity (ICANS) or grade ≥3 cytokine release syndrome (CRS) was observed. Three patients died due to laboratory-confirmed infections. The best overall response rate (ORR) was 50.0% (3/6) in the efficacy assessable set and 33.3% in the full analysis set. The number of LILRB4-positive STAR-T cells significantly increased, and the number of LILRB4-positive target cells decreased. Single-cell RNA sequencing revealed that monocyte-mediated suppression of autologous T-cell function may be a primary mechanism underlying the failure of STAR-T therapy in nonresponders. In conclusion, this first-in-human trial demonstrates the therapeutic potential of targeting LILRB4 with STAR-T-cell therapy in AML and warrants further investigation.

Introduction

Acute myeloid leukemia (AML) is an aggressive hematologic malignancy characterized by profound cellular heterogeneity, intrinsic drug resistance, and a highly immunosuppressive bone marrow microenvironment,^1,2,3 Despite advances in induction, consolidation, and allogeneic hematopoietic stem cell transplantation (allo-HSCT), disease relapse remains a major clinical challenge, particularly in monocytic AML.⁴ Although novel agents and targeted therapies have improved outcomes in selected subgroups, the overall prognosis of relapsed/refractory (R/R) AML remains poor.^5,6 With the development of chimeric antigen receptor (CAR)-T-cell therapy, remarkable success has been achieved in the treatment of B-cell hematologic malignancies.^7,8 However, translating this success to AML has been challenging, largely due to the lack of tumor-specific antigens. Most AML-associated targets, such as CD33 and CD123, are also expressed on normal hematopoietic stem and progenitor cells, raising concerns about on-target off-tumor toxicity and prolonged myelosuppression.⁹ Therefore, the identification of lineage-restricted, leukemia-selective targets remains a critical unmet need for the development of safe and effective adoptive T-cell therapies in AML.

Leukocyte immunoglobulin-like receptor B4 (LILRB4), also known as ILT3, has emerged as a promising candidate target in monocytic AML. LILRB4 is a transmembrane inhibitory receptor predominantly expressed on monocytic lineage cells, including macrophages, dendritic cells (DCs), and monocyte.^10,11 Notably, its expression is significantly upregulated in leukemic blasts of M4/M5 AML compared with normal counterparts. Importantly, LILRB4 has been detected on leukemia stem cells (LSCs)¹² but is largely absent from normal hematopoietic stem cells (HSCs),¹³ highlighting its potential as a selective therapeutic target. In addition, cells within the immunosuppressive tumor microenvironment, particularly myeloid-derived suppressor cells (MDSCs) and tumor-associated monocytes, exhibit high levels of LILRB4 expression.¹⁴ Functionally, LILRB4 contributes to leukemia progression through multiple mechanisms. AML cells with elevated LILRB4 expression have been shown to enhance tissue infiltration and promote immune evasion by suppressing T-cell activation and proliferation. Mechanistically, LILRB4 signaling can modulate downstream pathways that facilitate tumor cell migration and reshape the immune microenvironment toward an immunosuppressive state. Clinically, high LILRB4 expression is associated with adverse prognosis and inferior survival outcomes in AML patients.^14,15,16 Collectively, these findings underscore the dual role of LILRB4 in leukemic cell intrinsic aggressiveness and extrinsic immune suppression, supporting its development as a promising and rational therapeutic target for monocytic AML.¹⁷

Synthetic T-cell receptors (TCRs) and antigen receptors (STARs) are TCR-based chimeric antigen receptors that integrate key structural and functional features of both native TCRs and conventional CARs.¹⁸ STAR-T cells can specifically recognize tumor-associated antigens and initiate fully preserved TCR signaling cascades, leading to more physiological and coordinated T-cell activation. Compared with traditional CAR-T cells, STAR-T cells exhibit lower tonic signaling, enhanced antigen sensitivity, and improved signal fidelity, which collectively reduce exhaustion and enhance persistence. In addition, their modular design makes them particularly suitable for constructing dual-targeting or logic-gated systems to overcome tumor antigen heterogeneity. Preclinical studies have demonstrated that STAR-T cells can effectively eradicate both solid tumors and hematologic malignancies, with superior antigen sensitivity, robust in vivo expansion, and enhanced tissue infiltration capacity.^19,20 Moreover, emerging clinical evidence from a phase I trial in relapsed/refractory B-cell acute lymphoblastic leukemia indicates that STAR-T cells can induce rapid and profound tumor clearance with a favorable safety profile, achieving potent antitumor activity without severe treatment-related toxicities.¹⁹

Here, we immunized an alpaca with the recombinant extracellular domain of LILRB4 to generate a robust antigen-specific immune response. Subsequently, a high-diversity phage display library was constructed using the variable domains of heavy-chain-only antibodies (VHHs) derived from peripheral B cells of the immunized alpaca. Through iterative rounds of biopanning and affinity screening, we successfully isolated a panel of high-affinity VHH nanobodies specifically targeting distinct epitopes of LILRB4. These nanobodies were then engineered into STAR-T/CAR-T cell constructs, enabling the development of both single-epitope and dual-epitope targeting strategies. We systematically evaluated the therapeutic efficacy of these engineered T cells in vitro and in multiple in vivo xenograft models. Notably, compared with single-epitope STAR-T cells, dual-epitope anti-LILRB4 STAR-T cells (designated DE STAR-T cells) demonstrated significantly enhanced cytotoxicity, improved antigen recognition breadth, and superior effector function, including increased cytokine production and proliferative capacity. Importantly, DE STAR-T cells showed more durable tumor control and reduced antigen escape in vivo. Building upon these preclinical findings, we conducted a first-in-human phase I clinical trial of LILRB4-targeted STAR-T cells in patients with LILRB4-positive R/R AML (NCT05548088). In parallel, we performed comprehensive single-cell transcriptomic profiling of patient samples to investigate mechanisms underlying treatment response and failure. Our proof-of-concept results indicate that LILRB4-directed STAR-T-cell therapy represents a promising and potentially transformative therapeutic strategy for heavily pretreated R/R AML patients, including those who relapse after allo-HSCT, a population with otherwise limited treatment options.

Results

Design and preclinical evaluation of nanobody-based dual epitope anti-LILRB4-STAR T cells

LILRB4 is expressed in AML tumor cells. It mediates T-cell suppression and tumor cell infiltration via the immunoreceptor tyrosine-based inhibitory motif and serves as a potential specific marker for monocytic leukemia.¹² To develop a novel T-cell therapy product, we screened an immunized phage display library and identified two LILRB4-specific nanobodies (Supplementary Fig. 1). Subsequently, we established anti-LILRB4 STAR-T cells that targeted a single epitope (NLB4-STAR, NLB14-STAR) or dual epitope (NLB4/NLB14, DE-STAR) of LILRB4 to evaluate the potential ability of these cells for AML in vitro and in vivo (Fig. 1a). On the basis of the LILRB4 expression level in different target cells (Supplementary Fig. 2a), we selected MV4-11 cells with high LILRB4 expression and KASUMI-1 cells with low LILRB4 expression as target cells to assess the cytotoxicity of the transduced STAR-T cells. Compared with mock T cells, NLB4-STAR, NLB14-STAR and DE STAR-T cells exhibited significant cytotoxicity against MV4-11 (Fig. 1b, c) but no effect on LILRB4-negative target cells (Supplementary Fig. 2b). Notably, compared with NLB4-STAR or NLB14-STAR-T cells, DE STAR-T cells showed superior cytotoxicity to target cells, especially for low effector-to-target (E/T) ratios. Additionally, DE STAR-T expressed higher interferon-γ(IFN-γ) and interleukin-2(IL-2) expression level upon activation by target cells, compared to NLB4- or NLB14-STAR T cells (Fig. 1d).

Previous reports have shown that apolipoprotein E (APOE) and fibronectin (FN), as ligands of LILRB4, can induce T-cell suppression and facilitate tissue infiltration of AML cells.^21,22 APOE and FN can bind to LILRB4, which may competitively inhibit the binding of anti-LILRB4 STAR-T cells to LILRB4, consequently impairing the cytotoxic activity of anti-LILRB4 STAR-T cells. Before the STAR-T cells were cocultured with target cells in medium containing APOE or FN, we performed surface plasmon resonance (SPR) to confirm the ability of APOE and FN to bind to LILRB4. The SPR results indicated that APOE and FN could bind to LILRB4 (Supplementary Fig. 2c). However, in a target cell coculture system with APOE or FN, we observed no effect of APOE- or FN-containing medium on the effector functions of DE LILRB4 STAR-T cells (Fig. 1e, f). These findings suggest that DE STAR-T cells can bypass the interaction between APOE/FN and LILRB4 and directly eliminate AML cells. Moreover, MDSCs are known to exhibit immunosuppressive functions in AML therapy, and previous studies have shown that LILRB4 is highly expressed in MDSCs.^10,23 Our findings also revealed that DE STAR-T cells could effectively eliminate MDSCs, similar to their activity against targeted AML cells (Fig. 1g). Moreover, we detected LILRB4 expression in various cell populations among different cells in bone marrow and CD34-positive hematopoietic stem cells (CD34+ HSCs) from different healthy donors, and flow cytometry results revealed that compared with lymphocytes and myeloid cells, monocytes in bone marrow exhibited relatively high LILRB4 expression (Fig. 1h). Moreover, a small population of CD34+ HSCs expressed LILRB4, and DE STAR-T cells did not induce obvious cytotoxicity to CD34⁺ HSCs (Fig. 1i, j and Supplementary Fig. 2d) and exhibited significant activation (Fig. 1k and Supplementary Fig. 2e) after coculture for 24 h. These results suggested that DE STAR-T cells target only cells that express LILRB4 and do not damage LILRB4-negative CD34⁺ HSCs.

Next, we evaluated the antitumor efficacy of NLB4-STAR, NLB14-STAR and DE STAR-T cells in xenograft tumor models. In the MV4-11-luc model, mice were inoculated with 1 × 10⁶ MV4-11-luc cells, followed by the infusion of 4 × 10⁶ transduced STAR-T cells (Fig. 1l). Although neither NLB4-STAR, NLB14-STAR nor DE STAR-T cells completely eradicated MV4-11-luc cells (Fig. 1m, n), DE STAR-T cells demonstrated superior tumor control compared with NLB4- or NLB14-STAR-T cells. By day 7 postinfusion, DE STAR-T cells expanded significantly more than NLB14 STAR-T cells did, and the numbers of NLB4- and DE STAR-T cells were comparable at this time point. The number of DE STAR-T cells remained elevated for 1 week, whereas the numbers of NLB4 and NLB14 STAR-T cells decreased after day 7 (Fig. 1o).

Moreover, in the OCL-AML3-luc xenograft model, compared with NLB4-STAR and NLB14-STAR-T cells, DE STAR-T cells effectively suppressed tumor growth, resulting in a higher survival rate (Supplementary Fig. 2f–j). These results suggest that DE-STAR-T cells exhibit potent cytotoxicity against tumor cells while minimizing the effect of on-target, off-tumor toxicity to HSCs.

Compared with DE CAR-T cells, DE STAR-T cells mediate superior T-cell functions under low-LILRB4-expressing target cell stimulation

The results of the comparison between single-epitope-targeting and dual-epitope-targeting STAR-T cells suggested that DE STAR-T cells exhibited stronger cytotoxicity to conquer tumor cells. However, whether the antitumor efficacy of DE STAR-T cells surpasses that of DE-CAR-T cells remains unknown. Furthermore, given that compared with STAR-T cells, CAR-T cells exhibit lower sensitivity to antigens at low expression levels,^18,20 confirming whether the expression level of LILRB4 in patient-derived tumor cells is crucial for the use of LILRB4 in cell therapy.

We therefore analyzed tumor cells from various AML patients and quantified the expression levels of CLL1, CD123 and LILRB4. Our data revealed significantly lower expression of LILRB4 on AML tumor cells than on CD123 or CLL1 cells did (Fig. 1p), suggesting that CAR-T cells may exhibit low cytotoxicity against tumor cells with low LILRB4 expression.

To confirm the ability of DE-CAR-T cells and DE STAR-T cells to generate a low-abundance LILRB4 cell line, we engineered four DE-BBz-CAR constructs featuring varying nanobody arrangements and linkers (G₄S or M218) (DE-CAR1 to DE-CAR4, Fig. 1q and Supplementary Fig. 3a). These constructs were transduced into Jurkat cells or primary T cells via lentivirus for comparison with DE-STAR. Flow cytometry confirmed the robust surface display of all the DE-CAR constructs on T cells (Supplementary Fig. 3b), while their tonic signaling activity remained significantly greater than that of DE-STAR cells (Supplementary Fig. 3c), which is consistent with the findings of previous studies.^18,19

Since the expression level of LILRB4 in AML patient-derived tumor cells is lower than that of CD123 and CLL1, we chose the KASUMI-1 cell line (which has low LILRB4 expression) as the target cell line and compared the functional and mechanistic differences between DE STAR-T cells and DE-CAR-T cells. After 24 h of coculture at different E:T ratios, compared with DE-CAR-T cells, DE-STAR-T cells demonstrated stronger tumor cell killing ability (Fig. 1r) and cytokine secretion (Fig. 1s). Under low-antigen stimulation, DE STAR-T cells expressed prominent effector markers (CD25, PD-1, and LAG-3 expression; Supplementary Fig. 3d, e) and rapidly differentiated from high-CD62L-expressing cells (T central memory/T stem cell memory/T naive) into T effector cells with potent cytotoxic function (Supplementary Fig. 3i). After 7 days of coculture, compared with DE-CAR-T cells, DE-STAR-T cells exhibited better tumor cell elimination (Supplementary Fig. 3g). However, after being cocultured with KASUMI-1 cells for 7 days, DE STAR-T cells and DE-CAR-T cells exhibited comparable expression levels of PD-1 and LAG-3 (Supplementary Fig. 3f) and ultimately differentiated into T effector cells (Supplementary Fig. 3h, i).

To elucidate the differences in signaling pathways between DE-CAR-T cells and DE-STAR cells under low-antigen stimulation,^24,25 we performed western blotting to analyze TCR signaling and noncanonical NFκB signaling after coculture with KASUMI-1 cells at various time points. Western blotting revealed that TCR signaling was strongly activated in DE STAR-T cells, with rapid phosphorylation of CD3ζ, PLCγ, and LAT within 5 min of stimulation, followed by elevated p-AKT and p-ERK1/2 levels, significantly exceeding the activated responses in DE-CAR-T cells. With respect to the noncanonical NF-κB (ncNF-κB) signaling pathway, while DE-CAR-T cells exhibited increased baseline nuclear translocation of RelB and p52 due to tonic signaling, DE STAR-T cells exhibited substantial RelB/p52 nuclear accumulation upon antigen stimulation for 6 h, thereby increasing ncNF-κB signaling (Fig. 1t). These functional and mechanistic analyses demonstrate that compared with DE-CAR-T cells, DE-STAR-T cells achieve stronger activation and cytotoxicity even under low-antigen stimulation.

Given these in vitro results, we attempted to establish a KASUMI-1 cell xenograft tumor model to compare the antitumor efficacy of DE-CAR-T cells and DE-STAR-T cells in vivo. However, KASUMI-1 cells failed to expand in immunodeficient mice after inoculation (1 × 10⁶ to 3 × 10⁶ cells/mouse; data not shown) during a 2-week observation period. We therefore utilized MV4-11 cells (which express high levels of LILRB4) to establish an animal model for evaluating the comparative antitumor effects of DE-CAR-T cells and DE-STAR-T cells. Mice were inoculated with 5 × 10⁵ MV4-11-luc cells, followed by the infusion of 3 × 10⁶ transduced T cells (Fig. 1u). Under a low tumor burden model, DE-CAR-T cells and DE-STAR-T cells exhibited strong tumor control in the initiation stage (Fig. 1v). However, after infusion for 13 days, tumor relapse occurred in some mice in the DE-CAR-T-cell groups, and most of mice in the DE-STAR-T-cell group maintained tumor-free conditions (Fig. 1v). With respect to the expansion of transduced T cells in the different groups, DE STAR-T cells and DE-CAR-T cells exhibited proliferation in the early phase. However, after that, the number of DE STAR-T cells and DE-CAR-T cells also declined rapidly (Fig. 1w). These results demonstrate that DE STAR-T cells with high sensitivity to low-expression antigens can be effectively activated by tumor cells with low LILRB4 expression levels and exert significant antitumor effects. These findings suggest that DE STAR-T cells have superior potential for clinical application in the treatment of patients with LILRB4-positive AML.

Patient characteristics

Between May 2023 and June 2024, 12 patients with LILRB4-positive relapsed or refractory acute myeloid leukemia (R/R AML), all of whom were classified as FAB M4 or M5, were enrolled according to the predefined study design (Fig. 2a), and on the basis of the preclinical data, we chose DE STAR-T cells as the cell therapeutic product. One patient withdrew consent prior to apheresis, and two additional patients died from infections caused by carbapenem-resistant Klebsiella pneumoniae and Rhizomucor pusillus prior to cell infusion. Consequently, nine patients received cell infusions and were included in the analysis (Fig. 2b).

The patients’ characteristics are summarized in Table 1. The median age of the cohort was 36 years (range: 25–60). Eight patients had de novo AML, while one had AML secondary to relapsed chronic myelomonocytic leukemia (CMML). On the basis of the European Leukemia Net (ELN) 2022 classification, four patients had adverse genetic profiles. The median disease duration was 2 years (range: 0.5–3.82). Two patients had refractory disease, and seven (77.8%) patients relapsed following prior allo-HSCT. Three patients experienced HLA loss, and one patient experienced relapse after a second HSCT.

Table 1 Characteristics of the patients at baseline

Full size table

On average, patients had received three prior lines of therapy (range: 1–5), with HSCT counted as one line. Prior to lymphodepletion, the median bone marrow blast percentage was 28.4% (range: 6.0–81.5%). The median percentage of LILRB4-positive blasts was 74.6% (range, 44–94.4%). The screening procedures did not detect any cases of extramedullary disease.

Treatment

For the clinical trial, STAR-T cells were generated from either autologous sources (n = 4) or haploidentical donors postallo-HSCT (n = 5). The manufacturing of STAR-T products was successful for all patients who underwent cell collection. The manufactured products exhibited a median T-cell purity of 99.9% (range: 99.4–100%) and a median transduction efficiency of 54.75% (range: 32.4–75.5%). Notably, no residual AML cells were detected in the final STAR-T products (Supplementary Table 1). All patients received lymphodepletion conditioning with fludarabine (30 mg/m²) and cyclophosphamide (300 mg/m²) prior to STAR-T-cell infusion. The median vein-to-vein time (defined as the interval from apheresis to infusion) was 22 days (range: 18–186 days). Two patients underwent bridging chemotherapy before lymphodepletion, and all patients had active disease at the time of lymphodepletion. Nine patients received the planned STAR-T infusion, with doses ranging from 1 × 10⁶ cells/kg (n = 1, P1), 3 × 10⁶ cells/kg (n = 3, P2–P4), 6 × 10⁶ cells/kg (n = 4, P5–P8), to 1 × 10⁷ cells/kg (n = 1, P9).

Safety assessment

All adverse events (AEs) are summarized in Table 2. All nine patients who received STAR-T-cell infusion experienced grade ≥3 AEs. Pancytopenia was the most prevalent and affected 88.9% (8/9) of the patients. Moreover, 100% of the patients developed grade ≥ 3 cytopenia, including lymphopenia (n = 9), neutropenia (n = 9), thrombocytopenia (n = 8), and anemia (n = 8). Among the six patients who completed the safety assessments, one had an improvement in cytopenia to grade ≤2, and the other five patients still had unresolved grade ≥3 cytopenia after day 28 postinfusion. Other grade ≥3 AEs included elevated γ-GT (n = 1), elevated AST (n = 3), and electrolyte disturbances (n = 3). Three patients died from grade 5 infections in a state of pancytopenia before day 28 postinfusion. Patient 4 (3 × 10⁶ cells/kg group) died on day 8 from septic shock, possibly because of infection with metallo-β-lactamase-producing NDM⁺ Escherichia coli. Patient 5 (6 × 10⁶ cells/kg group) died on day 9 from infectious diffuse alveolar hemorrhage associated with parvovirus B19 and ESBL-positive Escherichia coli. Patient 7 (6 × 10⁶ cells/kg group) died on day 12 from severe pneumonia caused by carbapenem-resistant Klebsiella aerogenes and methicillin-resistant Staphylococcus aureus-positive human Staphylococcus. After careful multidisciplinary review, all of these patients died from infections with concordant microbiological evidence and clinical presentation occurring in the setting of profound neutropenia (ANC < 100 cells/µL). Therefore, the SRC did not classify these deaths as DLTs. Of the remaining six patients who completed the DLT assessment, none experienced DLTs, grade ≥3 cytokine release syndrome (CRS), or immune effector cell-associated neurotoxicity syndrome (ICANS). However, CRS occurred in 83.3% (5/6) of the patients. One patient had grade 1 CRS, and four patients had grade 2 CRS. CRS occurred within 24 h of infusion and persisted for a median of 3 days (range: 1–10 days). For the management of CRS, tocilizumab was administered to three patients, and corticosteroids were used for one patient (Supplemtary Table 2). Ultimately, all CRS-related symptoms resolved completely.

Table 2 Adverse events

Full size table

Clinical outcomes

The median follow-up duration for the six patients was 10.7 months (range: 1.47–16.9 months). Among the six patients who completed the primary endpoint evaluation, the best overall response rate (ORR) post-STAR-T infusion was 50.0% (3/6, Fig. 3a), corresponding to an ORR of 33.3% in the full analysis set. Specifically, one patient achieved complete remission (CR, 1/6, P6), albeit with positive minimal residual disease (MRD) (FCM, 2.65%), another patient reached a morphologic leukemia-free state (MLFS, 1/6, P8) but had positive MRD (FCM, 0.04%), and one patient achieved partial remission (PR, 1/6, P9) with positive MRD (FCM, 0.07%).

By day 28, only Patient 6 maintained a response, while Patients 8 and 9 lost their response. Patient 9 received a second infusion (1 × 10⁷ cells/kg) following relymphodepletion and subsequently achieved a PR.

Among the three patients who did not achieve a response by day 28, Patient 1 achieved MRD-negative CR after salvage chemotherapy and donor lymphocyte infusion (DLI). Patient 2 died of progressive disease within 3 months, while Patient 3 did not respond to salvage chemotherapy and had persistent active disease.

In this small cohort, clinical responses were observed only in patients receiving high doses of STAR-T cells. However, it is important to note that all patients in these higher-dose cohorts received allogeneic healthy donor-derived products, whereas the low-dose cohorts received autologous products, complicating the interpretation of a dose–response relationship. Notably, although no significant differences were observed in the STAR-T-cell manufacturing parameters between autologous and allogeneic sources, allogeneic STAR-T cells constituted the majority of the high-dose group. These findings suggest that compared with autologous T cells from leukemia patients, healthy T cells from donors may possess superior functional characteristics.

Among the four patients who were alive at the last follow-up, two patients (P6 and P9) underwent a second or third allo-HSCT after achieving CR or PR with STAR-T-cell therapy, respectively. Both patients have remained in MRD-negative CR for more than 16 months. Patient 1 maintained an MRD-negative CR for 19 months after receiving a chemotherapy and DLI. Patient 3 survived with active disease for 12 months at the last follow-up. The estimated 1-year overall survival (OS) was 44.4% (95% CI: 21.4–92.3%, Fig. 3b).

STAR-T expansion, changes in LILRB4-positive cells and cytokine secretion

Expansion of LILRB4 STAR-T cells was observed in patients postinfusion. STAR gene copies in peripheral blood (PB) peaked on day 7 (range: day 4–day 19). The median peak copy number was 9021 copies/μg DNA (range: 275–204,000). The percentage of LILRB4-positive cells in the PB was negatively correlated with the number of STAR-T cells (Fig. 4a and Supplementary Fig. 4), especially around the peak of STAR-T expansion. In bone marrow (BM), the expansion of LILRB4-positive STAR-T cells was also high, and the proportion of LILRB4-positive cells in the BM was also negatively correlated with the expansion of STAR-T cells (Fig. 4b and Supplementary Fig. 4).

Following LILRB4 STAR-T therapy, the percentage of patients with LILRB4-positive AML blasts tended to decrease, particularly in patients P1, P2 and P9. However, high numbers of LILRB4-negative AML blasts persisted, especially in patients P1, P2, and P3 (Fig. 4c). We also analyzed the peak expansion of the LILRB4-expressing STAR-T cells in the responding and nonresponding patients, although this difference did not reach statistical significance (Fig. 4d).

On the basis of dynamic monitoring of plasma cytokine levels after LILRB4 STAR-T-cell infusion, most patients showed a consistent pattern of transient cytokine release with considerable variation in magnitude. The expression of key cytokines (IL-6, IL-8, IL-10, and IFN-γ) typically peaked between days 7 and 14, which coincided with the peak of STAR-T-cell expansion, and returned to baseline by day 28. Patient P9 demonstrated the most synchronized and pronounced cytokine elevation closely mirroring the STAR-T-cell expansion kinetics, whereas patient P2 showed atypical cytokine profiles, potentially reflecting differences in the underlying disease state or individual immunological factors. (Supplementary Fig. 5 and Supplementary Fig. 6).

As LILRB4 is also expressed in monocytes,²⁶ to explore immune cell changes, we also detected monocyte and T lymphocyte levels after STAR-T infusion. We observed that CD14⁺ monocytes exhibited a characteristic biphasic pattern following STAR-T infusion, with transient depletion during the early phase (D0–D7) followed by gradual recovery in the later phase (D14–D28) (Supplementary Fig. 7). Additionally, at the time of infusion, patients exhibited a high CD4/CD8 ratio, which gradually normalized within 28 days post-STAR-T infusion (Supplementary Fig. 8).

Single-cell analysis reveals suppressed autologous T-cell function in nonresponder patients

To further explore the mechanisms underlying the lack of response to STAR-T therapy in nonresponder (NR) patients, we first examined the abundance of bone marrow e-MDSCs (Lin⁻ HLA-DR⁻ CD33⁺) and M-MDSCs (CD11b⁺ CD14⁺ CD15⁻) in two NR patients (P1 and P2) and two complete responder (CR) patients (P6 and P8). The results revealed that after treatment, the abundance of e-MDSCs did not significantly change across all the samples, whereas the abundance of M-MDSCs was universally downregulated, suggesting that STAR-T therapy may remodel the bone marrow microenvironment (Fig. 5a).

To elucidate the complex dynamic changes within the bone marrow microenvironment, we performed single-cell RNA sequencing on bone marrow mononuclear cells collected from the four patients before and after treatment. After quality control, 71,363 high-quality single cells were retained. On the basis of marker gene expression, the cells were annotated into 11 major cell populations: CD8 T cells, naive T cells, CD34 HSC-like cells, NK cells, B cells, monocytes, CLPs, erythroblasts, mast cells, plasma cells, and megakaryocytes (Fig. 5b). T-cell populations predominated in all patients before treatment, indicating that immune reconstitution had been achieved posttransplantation (Fig. 5c).

To further investigate the role of M-MDSCs in STAR-T-cell therapy, monocyte populations were reclustered into five populations: M-MDSCs (CD11b⁺ CD14⁺ CD15⁻), Mono1 (CD11b⁺ HLA-DR^low CD33⁻), Mono2 (CD11b⁺ HLA-DR^high CD33⁺), Mono3 (CD11b⁺ HLA-DR^low CD33⁺ CD15⁺), and Mono4 (CD11b⁻ HLA-DR^high CD33⁻) (Fig. 5d). Compared with other monocyte subsets, M-MDSCs exhibited significantly increased reactive oxygen species (ROS) activity (Fig. 5e). LILRB4 expression was predominantly observed in monocytes and CD34 HSC-like cells, which is consistent with previous findings (Fig. 5f). Within monocyte populations, LILRB4 was expressed mainly in Mono4, Mono3, and Mono2. Posttreatment, all patients had downregulated expression of LILRB4 in monocyte subsets, indicating that STAR-T therapy effectively alleviated the disease burden. Crucially, compared with those in the CR group, the ROS levels and T-cell inhibitory capacity in the NR group were significantly greater (Fig. 5g). These results suggest that autologous T cells in NR patients may be functionally impaired before therapy, potentially limiting the efficacy of autologous STAR-T cells.

To elucidate this phenomenon, we compared T-cell functional differences between the NR and CR groups before and after treatment. We found that prior to treatment, CD8+ T cells from patients who achieved CR significantly upregulated the expression of activation markers such as KLRK1 and GZMK, whereas CD8+ T cells from NR patients predominantly upregulated the expression of inhibitory markers such as MDM2 and BAX, indicating impaired T-cell function in the NR group (Fig. 5h). Functional enrichment analysis further revealed that T cells in the CR group exhibited increased activity related to cell killing and leukocyte-mediated cytotoxicity, whereas NR T cells were enriched in pathways such as signal transduction by p53 class mediators, which are potentially associated with the upregulation of MDM2 and BAX expression (Fig. 5i). T-cell dysfunction appears to be a major factor contributing to the failure of STAR-T therapy in NR patients.

To explore the potential mechanisms underlying pretreatment differences in T-cell function, we constructed an intracellular cell‒cell communication network in bone marrow. In pretreatment, potential malignant-like subclusters (Mono2, Mono4, and CD34 HSC-like) acted as dominant signal senders, whereas CD8 T cells and NK cells served as signal receivers, suggesting that monocyte subsets may be key regulators of CD8 T-cell function (Fig. 5j). Specifically, M-MDSCs, Mono2, and Mono4 suppressed T-cell activity through pathways including LGALS9 and MIF, with significantly stronger effects observed in the NR group (Fig. 5k). These findings suggest that T-cell dysfunction induced by the suppressive effects of monocytes may be a primary driver underlying the failure of STAR-T therapy in the NR group.

Discussion

AML is a hematologic malignancy with a dismal 5-year survival rate. Although CAR-T-cell therapy has achieved remarkable success in treating B-cell hematological malignancies, its application to AML remains a significant challenge.^27,28,29 The heterogeneity of AML cells, on-target, off-tumor effects in the immunosuppressive tumor microenvironment, and the lack of appropriate targets pose major obstacles to the efficacy of CAR-T-cell therapy in AML.^1,9 Most CAR-T-cell therapies for AML primarily target CD123, CD33, or CLL1.^27,30,31 However, these targets can lead to severe side effects in patients during treatment.³² Previous research has shown that LILRB4 is specifically overexpressed in the M4/M5 subtypes of AML cells and LSCs but is absent in HSCs and most other normal tissues, except for macrophages, DCs, and monocytes.¹⁶ These findings suggest that LILRB4 may be a promising target for treating M4/M5 subtypes of AML. Therefore, we selected LILRB4 as the target for developing anti-LILRB4 STAR-T cells to conquer AML in vitro and in vivo, after which we carried out a clinical trial to evaluate the therapeutic response of anti-LILRB4 STAR-T cells to the M4/M5 subtype of AML.

Unlike dual-target CAR constructs, which are difficult to engineer for potent effector functions,³³ the dual-target STAR construct simplifies the construct modification. Two single-chain variable fragment (scFv) sequences are directly fused to the constant regions of TCRα and TCRβ, yielding a dual-target STAR construct that inherently exhibits high antigen sensitivity and tissue infiltration capacity.^20,34 To reduce the potential for antigen escape owing to gene mutation and improve antigen binding ability, we mimicked the design of CARVYKTI and constructed a dual epitope-targeting construct by fusing two nanobodies targeting distinct epitopes of LILRB4 (DE-STAR).³⁵ DE STAR-T cells demonstrated potent elimination of various AML cell types in vitro and in vivo. For the clinical trial, we enrolled LILRB4-positive monocytic AML patients with a history of extensive prior therapy and relapse after allo-HSCT. Given the high drug resistance of M4/M5 subtypes,³⁶ STAR-T-cell infusion achieved an ORR of 50%. Notably, the high-dose group (≥6 × 10⁶ cells/kg) achieved a preliminary ORR in the assessable efficacy set, highlighting broad therapeutic potential, particularly in five patients with posttransplantation relapse. Among the three responders, relapse was not associated with loss of LILRB4 expression. Patient 9, who received a second STAR-T infusion, achieved a PR, demonstrating treatment feasibility even after initial response loss. Importantly, salvage chemotherapy, DLI, or subsequent allo-HSCT induced MRD-negative complete remission in patients who relapsed after LILRB4-STAR-T therapy. These observations support the biological relevance of LILRB4 as a therapeutic target in AML and suggest that prior exposure to LILRB4-STAR-T therapy, even without achieving durable remission, may be compatible with and not preclude the effectiveness of subsequent immune-based interventions such as DLI or repeat transplantation.^14,37

All treated patients developed grade ≥3 pancytopenia and severe infections, while one patient experienced recovery of cytopenia to grade ≤2 within 28 days post-infusion. Persistent cytopenia at the termination of the study in five patients highlights the vulnerability of this heavily pretreated R/R AML population.^29,38 Infection control remains a key consideration for future LILRB4-directed CAR-T studies in AML. To mitigate infectious risk, stricter patient selection should be applied, including the exclusion of patients with uncontrolled infections, together with systematic preinfusion screening for multidrug-resistant organisms and fungal colonization. Early differentiation between CRS and infection relies on rapid next-generation sequencing–based pathogen detection integrated with inflammatory biomarkers and clinical course. Antimicrobial strategies should be further individualized on the basis of colonization status and local resistance patterns to optimize safety outcomes.

The high incidence of prolonged cytopenia likely reflects a combination of baseline marrow failure, prior intensive therapies, and early post-CAR-T-cell inflammatory stress rather than direct LILRB4-specific off-tumor toxicity. In terms of therapeutic efficacy, the expansion of LILRB4-expressing STAR-T cells was associated with a significant reduction in the number of LILRB4-positive AML blasts, and the effects of LILRB4-expressing STAR-T cells on CD14⁺ monocytes and MDSC findings were consistent with those of previous studies.³⁹ Notably, in responsive or relapsed patients, we observed that LILRB4 STAR-T therapy may increase sensitivity to residual leukemia, potentially improving the efficacy of subsequent bridging chemotherapy and transplantation. Three patients achieved sustained remission after salvage therapy, although the underlying mechanisms warrant further investigation.

There are several limitations in this study that should be considered. First, these findings must be interpreted in the context of the very limited sample size, with only nine treated patients and six evaluated for efficacy. As such, this study is not powered to assess therapeutic efficacy or survival benefit, and any observed disease responses should be considered exploratory and hypothesis-generating rather than definitive evidence of clinical activity. Second, although three patients achieved transient disease remission, all experienced relapse, precluding a comprehensive therapy to achieve long-term efficacy.

To address these limitations, several optimization strategies warrant further exploration. First, combination approaches incorporating venetoclax may increase LILRB4 expression on leukemic blasts and potentially increase tumor sensitivity to STAR-T-mediated cytotoxicity. Second, the development of donor-derived or universal STAR-T products may improve T-cell quality, persistence, and antileukemic potency, thereby enhancing clinical efficacy. These strategies may help overcome current biological and manufacturing constraints and improve the therapeutic potential of LILRB4-directed STAR-T therapy in AML.

Collectively, these data support primarily the feasibility of LILRB4 STAR-T-cell manufacturing and administration while underscoring the critical need for improved patient selection, optimized timing, and infection-focused supportive care strategies in future studies. Larger, multicenter trials will be needed to more reliably define the safety profile, hematologic recovery kinetics, and potential clinical benefit of this approach.

Materials and methods

Cell lines and primary cells

Lenti-X-293T cells for lentivirus packaging were purchased from Takara Biomedical Technology and cultured in DMEM supplemented with 10% fetal bovine serum (FBS). LILR subtype-expressing 293 T cells were constructed by infecting 293T cells with lentiviriruses expressing different LILR subtypes and cultured in DMEM supplemented with 10% FBS. PBMCs for MDSC induction were purchased from Hemacare. THP-1, OCL-AML3, KASUMI-1, and MV4-11 cells were cultured in RPMI-1640 medium supplemented with 10% FBS. THP-1-luciferase GFP (THP-1-luc), THP-1-luciferase GFP-LILRB4 knockout (THP-1^LILRB4-KO-Luc), OCL-AML3-luciferase GFP (OCL-AML3-luc), KASUMI-1-luciferase GFP (KASUMI-1-luc) and MV4-11-luciferase GFP (MV4-11- luc) cells were established and stored in our laboratory and cultured in RPMI-1640 medium supplemented with 10% FBS. All the cells were cultured at 37 °C and 5% CO2 and regularly tested for Mycoplasma using PCR, and the results were consistently negative.

Isolation and screening of anti-LILRB4 nanobodies

Anti-LILRB4 nanobody colonies were isolated from alpaca VHH phage screening libraries. Briefly, alpaca was immunized with purified protein from the LILRB4 extracellular domain. Afterward, the nanobody sequences were amplified from the memory B cells of immunized alpaca to construct a phage screening display library. After 3–4 rounds of panning, the phage-enriched anti-LILRB4 nanobodies were infected into the host bacteria of TG1. Single colonies coated with the LILRB4 protein were then picked and identified by performing a phage ELISA. Positive clones were sequenced to obtain anti-LILRB4 nanobody sequences.

For anti-LILRB4 nanobody screening, the nanobodies identified from phage ELISA screening were fused to the N-terminus of the STAR structural β chain and subsequently inserted into a third-generation lentiviral vector. Following lentiviral packaging of LILRB4 STAR, primary human T cells were infected, and LILRB4 nanobodies were screened on the basis of their killing specificity and potency against LILRB4-positive and LILRB4-negative target cells. For anti-LILRB4 nanobody binding kinetics and epitope screening, the nanobodies were fused to human IgG1 Fc, and binding kinetics and epitope screening were performed by SPR. The binding specificity of the LILRB4 protein was also verified by flow cytometry in 293T cell lines expressing LILR subtype proteins (Supplementary Fig. 1i). Finally, in conjunction with binding kinetics and competitive assays, anti-LILRB4 nanobodies were selected because of their high specificity, high affinity, strong killing activity and ability to bind different epitopes (Supplementary Fig. 1a).

Enzyme-linked immunosorbent assay

Phage ELISA was conducted by coincubation of the phage display library with precoated LILRB4 antigen on an ELISA plate. Briefly, the unbound phages were washed with 0.1% PBST, and then the specifically bound phages were eluted with Gly-HCL and neutralized with Tris base. The neutralized phages were negatively eluted with an ELISA plate coated with 2% BSA. The titer of the phage pool after suction was detected. The culture was expanded according to the elution titer, after which the “positive elute-washing-elution-negative elute-amplification” cycle was carried out; the amount of antigen coated was reduced (from 200 ng/well to 100 ng/well), and the number of washes was increased (from 10 times to 20 times) to enrich the sequences that could bind to LILRB4.

In the epitope mapping ELISA, phage-displayed NLB4 or NLB14 nanobodies were coincubated with precoated LILRB4 antigen, and after the plate was cleaned to remove unbound phage, 10 pmol FN protein, ApoE protein or purified NLB4 or NLB14 nanobodies were added to the assay wells. Finally, the competitive binding of anti-LILRB4 nanobodies with FN or ApoE to the LILRB4 antigen was detected with an anti-phage antibody conjugated with HRP.

To detect cytokine release after the T cells were cocultured with the target cells, the culture medium was collected after 24 h, and cytokine release was measured using the following ELISA kits: human IL-2, human IFN-γ and human TNF-α. All samples were measured in triplicate according to the manufacturer’s protocol.

Antibody binding kinetics

The binding kinetics of anti-LILRB4 VHH-FC were determined on a Biacore 8 K system. Mouse anti-human IgG (Fc) (25 μg/mL) was immobilized on the assay chip at 10 μL/min for 360 s. Afterward, the chip was sealed with 1 M ethanolamine at 10 μL/min for 420 s. Anti-LILRB4 VHH was fused to human IgG1 Fc, and anti-LILRB4 VHH-Fcs were diluted to 5 μg/mL and injected into the capture channel at a flow rate of 10 μL/min, approximately 200 RU. The diluted LILRB4 protein was injected into the experimental channel at a flow rate of 30 μL/min for association and dissociation measurements.

Membrane proteome array

A membrane proteome array (MPA, Integral Molecular) was used to assess the off-target binding of anti-LILRB4 nanobodies. NLB4 and NLB14 nanobodies were fused to human IgG1 Fc (NLB4-FC, NLB14-FC). MPA was conducted by highly sensitive flow cytometry of more than 5220 kinds of human membrane proteins to detect and screen the potential cross-reactivity of nontarget proteins.

Competitive binding assay

A competitive binding assay was performed on a Biacore 8 K system. The LILRB4 protein was immobilized. NLB4-Fc or NLB14-Fc was diluted to 800 nM and injected into the capture channel at 30 μL/min for 180 s. Then, NLB14-Fc and NLB4-Fc were added into the corresponding channel at 30 μL/min for another 180 s for association measurements.

Lentivirus production

The lentivirus package used 3rd-Transfer vector mix with the packing plasmids pRSV-Rev and pMDLg/pRRE and the envelope plasmid pMD2. G plasmids at an 8:2:4:2 ratio and then transfected into Lenti-X-293T cells using polyethyleneimine. The virus-containing culture medium supernatant was collected at 48 h post-transfection. The virus was mixed with PEG8000 overnight at 4 °C and centrifuged for 30 min. All the viruses were stored at −80 °C for later use.

STAR and CAR construction and manufacturing

The STAR construct expresses the sequence of the LILRB4 nanobodies with optimized mouse TCR α and β chain constant regions. With respect to the monovalent LILRB4-STAR, the incorporation of the coding sequence of the self-cleaving 2A peptide between the TCR α and Vhh-TCR β coding sequences allows translational skipping at the 2A site to yield 2 polypeptides from a single mRNA molecule. For the dual epitope anti-LILRB4-STAR, two nanobody sequences were fused to TCR α and TCR β. For the DE-CAR, DE-BBz-CAR constructs featuring varying nanobody arrangements and linkers (G₄S or M218) (DE-CAR1 to DE-CAR4). All the constructs contained red fluorescent protein (RFP) as a reporter. A schematic of the single- or dual-epitope anti-LILRB4-STAR and CAR constructs is shown in Fig. 1a, q. The preparation and quality of single- or dual-epitope anti-LILRB4-STAR T cells are described in the Supplemental Methods.

Quantitative PCR

Genomic DNA (gDNA) was extracted from cells in 1 ml of whole blood using a QIAamp DNA Blood Midi kit. The gDNA was used as the template to detect the integrated STAR-T lentiviral vector using the TB Green Premix ExTaq (Tli RnaseH Plus) Kit (Takara Biotechnology, Shiga, Japan, RR420A). Real-time PCR amplification was carried out using an Applied Biosystems 7500 Real-Time PCR System (Life Technologies, MA, USA). A primer pair targeting the WPRE region was used to determine the STAR copy number, and a primer pair targeting the cellular RPP30 gene was used as an internal control. The primer pairs were experimentally validated using the following criteria: (i) a single gene-specific product was produced; (ii) the amplification efficiency ranged between 90% and 110%; and (iii) the cycle threshold (Ct) value of the no-template DNA control was more than 40.

Cytotoxicity assays

The cytotoxicity of T cells transduced with different LILRB4 STAR and DE-CAR constructs was determined by a Bio-Lite Luciferase Assay System. For the effector/target ratio assay, the transduced T cells and tumor cells (MV4-11-luc, OCL-AML3-luc, and KASUMI-1-luc) were cocultured at different effector/target ratios in flat-bottom 24-well plates with 4 × 10⁵ target cells in a total volume of 1 mL. Twenty-four hours later, changes in the morphology of tumor cells or CD34+ HSCs were observed by microscopy, or 150 μL of mixed T-cell target cell suspension was transferred to white-bottom 96-well plates, after which the firefly luciferase substrate was added, and after 10 min of shaking at 300 rpm/min in the dark, the luciferase activity was monitored by a Biotek Synergy H1 microplate reader. The percent cell viability was determined as 100×(experimental fluorescence intensity)/(negative control fluorescence intensity).

Flow cytometry

The following fluorophore-conjugated antibodies were used for cell surface molecule staining in PBSF: APC anti-mouse-TCR β chain, PerCP/Cyanine5.5 anti-mouse-CD45.2, PE/Cyanine7 anti-human CD3, APC anti-Huamn CD3, BV421 anti-Huamn CD45, FITC anti-human CD8, APC-cy7 anti-human CD8, APC anti-human CD4, APC anti-human CD85k (ILT3), FITC anti-human CD14, PE anti-human CD33, Brilliant Violet 421TM anti-human CD11b, Percp-cy5.5 anti-human PD-1, APC anti-human LAG-3, BV421 anti-human CD45RO, Percp-cy5.5 anti-human CD62L, APC anti-human CD69, APC anti-human CD25, Fixable Viability Dye-eFluor780 and Fixable Viability Dye-eFluor506. Flow cytometry data were acquired by a BD Fortessa 4 laser and analyzed by using FlowJo software.

MDSC induction and cytotoxicity assay

For MDSC induction, after PBMCs from healthy donors were thawed, CD14-positive monocytes were isolated using a monocyte isolation kit (stem cell, 19359). The monocytes were then transferred to RPMI-1640 medium supplemented with IL-6 (10 ng/ml), GM-CSF (10 ng/ml), and PGE2 (1 μg/ml) for 7 days. The induction of MDSCs was validated by flow cytometry analysis of CD33 and CD11b expression.

For MDSC cytotoxicity detection, MDSCs were labeled with CellTrace^TM Far Red dye (Invitrogen, C34564), followed by in vitro coculture of MDSCs and control target MV4-11 cells with LILRB4 STAR-T cells and MOCK-T cells, respectively, at an effector-to-target ratio of 5:1 for 24 h. After cell culture, PI staining was performed, followed by FACS detection to measure the target cell killing efficiency.

Animal assays

Six- to eight-week-old female nonobese diabetic/ShiLtJGpt-Prkdc^em26/Il2rg^em26/Gpt (NCG) mice were purchased from GemPharmatech Co., Ltd., and kept under specific-pathogen-free conditions at the Animal Facility of Peking University Health Science Center. All mouse experiments were conducted in accordance with Institutional Animal Care and Use Committee (IACUC)-approved protocols. For intravenous tumor models with high tumor burden, mice were injected intravenously with 1 × 10⁶ GFP-positive MV4-11-luc tumor cells or 2 × 10⁶ GFP-positive OCL-AML3-luc tumor cells in a 200 μl volume on day 0. To establish intravenous tumor models with a low tumor burden, mice were intravenously injected with 1 × 10⁶ GFP-positive MV4-11-luc tumor cells in a 200 μl volume on day 0. After 5 or 7 days, transduced T cells or mock-transduced T cells (containing RFP reporter protein) were intravenously infused in a 200 μl volume (cell numbers varied for different experiments; noted in the figures or figure legends). The mice were treated with analgesics and monitored for health conditions in accordance with the IACUC. Tumor progression was determined by bioluminescence emission on Lumina II (CALIPER) after intraperitoneal D-luciferin (YEASEN) injection. Additionally, mouse peripheral blood samples were collected through the orbital sinus, and different organs were isolated at appropriate end points. No randomization or blinding methods were used.

T-cell persistence detection

For detection of the persistence of transduced T cells, the T-cell infection rate and surface display were determined before infusion. After intravenous injection, mouse peripheral blood samples were collected through the orbital sinus at each time point, red blood cell lysis buffer was used to remove red blood cells, and then PBMCs were stained with PE/Cyanine7-conjugated anti-human CD3, BV421-conjugated anti-human CD45, and APC-conjugated anti-mouse-TCR β chain antibodies following standard protocols. Transduced T cells in PBMCs were acquired using a BD Fortessa 4 laser and analyzed by using FlowJo software.

Detection of LILRB4 in STAR-T cells before infusion

LILRB4 STAR expression levels were determined using flow cytometry after manufacturing. A PE-conjugated anti-mouse-TCR β chain antibody was used to detect the expression of the LILRB4 gene STAR. The following antibodies were used for lymphocytes and T-cell markers: PerCP/Cyanine5.5 anti-human CD45 antibody, APC anti-human CD3 antibody, and PE/Cyanine7 anti-human CD8 antibody.

LILRB4 expression in bone marrow cells and CD34+ HSCs

Bone marrow cells were obtained from patients, and CD34+ HSCs were isolated from healthy donors by a CD34 Microbead Kit to assess LILRB4 expression using flow cytometry. The following antibodies were used: BV421 anti-human CD45, PE/Cyanine7 anti-human CD3, FITC anti-human CD14, and APC anti-human CD85k (ILT3). Flow cytometry data were acquired by a BD Fortessa 4 laser and analyzed by using FlowJo software.

LILRB4 STAR-T-cell expansion and persistence after infusion

The pharmacokinetics of the LILRB4-expressing STAR-T cells after infusion were detected by flow cytometry and quantitative real-time PCR (qPCR). For flow cytometry, an anti-mouse-TCR β chain antibody was used to detect LILRB4 STAR, and an anti-human CD3 antibody and an anti-human CD8 antibody were used for T cells. For quantitative real-time PCR, genomic DNA was extracted from patients’ peripheral blood and bone marrow samples, and the sequences of the probe and primers targeting LILRB4 VHH were as follows: forward primer-CAGCAGTGATGGCTCTACCT, reverse primer: TCCTCAGGTTTCAGGCTGTT, and probe-TGCCCTTCACGCTATCGGCGT.

Cytokine detection after LILRB4 STAR-T-cell infusion

Plasma concentrations of cytokines and chemokines were measured by Bio-Plex Pro Human Cytokine 27-Plex Assays (Bio-Rad, Hercules, CA), according to the manufacturer’s recommendations, using the Bio-Plex 200 system (Bio-Rad).

Trial design

This investigator-initiated clinical trial aimed to assess the safety and efficacy of LILRB4 STAR-T cells in the treatment of relapsed/refractory acute myeloid leukemia. The clinical protocol has been registered on ClinicalTrials.gov (NCT05548088), and the study obtained approval from the Ethics Committee of Peking University People’s Hospital (Ethics Approval Number: 2022PHD007). The principal aim of this study was to evaluate the safety of LILRB4 STAR-T-cell therapy. The secondary aim was to conduct a clinical efficacy evaluation. This study obtained approval from the Institutional Ethics Committee and was carried out in accordance with the Declaration of Helsinki. Each participating patient provided signed informed consent. Eligible patients were aged between 18 and 70 years and were diagnosed with relapsed/refractory acute myeloid leukemia in which the bone marrow sample was LILRB4 positive. The Eastern Cooperative Oncology Group (ECOG) score ranged from 0 to 2. Patients were not expected to have central nervous system leukemia (CNSL) or solid organ transplantation or hematopoietic stem cell transplantation (HSCT) within 6 months prior to screening.

For clinical trials, patient apheresis products were collected, and PBMCs were activated and transduced with a lentiviral vector containing the dual epitope anti-LILRB4 STAR and subsequently expanded. All patients received conditioning regimens of intravenous fludarabine (25–30 mg/m²/d) and cyclophosphamide (250–300 mg/m²/d) for three consecutive days, followed by infusion of the LILRB4 STAR-T cells at escalating doses of 1 × 10⁶, 3 × 10⁶, 6 × 10⁶, and 1 × 10⁷ cells/kg. The details of the study design and eligibility requirements are provided in the Supplemental Methods (full clinical trial protocol).

Clinical assessments

The dose-limiting toxicity (DLT) of the subjects was observed after the LILRB4 STAR-T-cell infusion, and AEs and serious AEs were recorded, with a focus on CRS and immune cell-associated neurotoxicity (ICANS). The CRS and ICANS were graded according to the ASTCT consensus guidelines. Individual organ toxicity was graded in accordance with the National Cancer Institute Common Terminology Criteria for Adverse Events (CTCAE) Version 5.0. The antitumor response was initially assessed in all patients at 2 weeks and 4 weeks post-infusion, followed by monthly visits.

The efficacy evaluation included complete morphological remission (CR), complete remission with incomplete recovery of blood cell count (CRi), complete remission with partial recovery of peripheral blood cells (CRh), complete remission with negative minimal residual disease (CR (MRD^-)), a morphological leukemia-free state (MLFS), and partial remission (PR). The CMML patients had complete remission (CR), bone marrow complete response (BMCR), partial remission (PR), and hematological improvement (HI). Further details regarding the study procedures are provided in the Supplemental Methods (full clinical trial protocol).

Statistical analysis

All the statistical analyses were performed using SPSS software. No statistical methods were used to predetermine sample size. Statistical comparisons between multiple groups were performed by one-way ANOVA or two-way ANOVA. Survival data were analyzed using the log-rank test, and survival curves were generated using the Kaplan‒Meier method. P values < 0.05 were considered to indicate statistical significance. The statistical test used for each figure is described in the corresponding figure legend.

10x Genomics single-cell transcriptome sequencing

The protoplast suspension of bone marrow mononuclear cells was loaded into Chromium microfluidic chips with 30 v3 chemistry and barcoded with a 10X Chromium Controller (10X Genomics). RNA from the barcoded cells was subsequently reverse transcribed, and sequencing libraries were constructed with reagents from a Chromium Single-Cell 30 v3 reagent kit (10X Genomics) according to the manufacturer’s instructions. Sequencing was performed with an Illumina instrument (NovaSeq 6000) according to the manufacturer’s instructions (Illumina).

1.
Single-cell transcriptome analysis

Single-cell sequencing reads were aligned to the human reference genome (Hg38) using Cellranger v7.1.0 with default parameters. Genes expressed in at least three cells were retained in the results. Cells with fewer than 200 genes and with mitochondrial gene expression greater than 10% were excluded. Additionally, doublets were identified and removed using DoubletFinder v2.0.6.⁴⁰
2.
Bone marrow single-cell data dimensionality reduction and annotation

The cleaned single-cell data were subjected to the standard Seurat v5.3.0⁴¹ analysis pipeline for data dimensionality reduction and clustering of single-cell data. First, we merged the samples and used the normalizeData function to normalize the original count matrix and the FindVariableFeatures function to identify the top 2000 highly variable genes. The ScaleData function was used to perform z score transformation on the data so that the RunPCA function could reduce the dimensionality of the data. Since the cells came from different samples, batch effects needed to be removed. Here, we used the Harmony v1.2.3⁴² package to remove batch effects and then used the RunUMAP and FindNeighbors functions to perform dimensionality reduction, visualization, and unsupervised clustering of cells, respectively. FindMarkers was used to identify genes that were differentially expressed between cell clusters. Finally, on the basis of known cell markers, we annotated the cell types of all the clusters. The marker genes of 11 types of mouse bone marrow cells,⁴³ namely, CD8 T cells, naive T cells, CD34 HSC-like cells, NK cells, B cells, monocytes, CLPs, erythroblasts, mast cells, plasma cells and megakaryocytes, were obtained from CellTaxonomy.

In addition, gene sets related to reactive oxygen species metabolic processes and the negative regulation of alpha–beta T-cell activation were obtained from MSigDB (https://www.gsea-msigdb.org/gsea/msigdb). The AddModuleScore function was applied to calculate gene set scores, which represent the ROS activity and T-cell inhibitory capacity at the single-cell level, respectively.
3.
Cell–cell communication network analysis

Cells with fewer than 200 genes and with mitochondrial gene expression greater than 10% were excluded. Additionally, doublets were identified and removed using DoubletFinder v2.0.6.Cell–cell communication networks were constructed usingCellChat v2.1.2.⁴⁴ The netAnalysis_signalingRole_scatter functionwas used to identify incoming and outgoing signaling patterns ofeach cell population, and netVisual_bubble was applied to comparedifferences in cell–cell communication networks between groups.All functions were executed using default parameters

Data availability

The raw sequence data reported in this study have been deposited in the National Genomics Data Center (GSA-Human: HRA017004) and are publicly available at https://ngdc.cncb.ac.cn/gsa-human.

References

Isidori, A. et al. The role of the immunosuppressive microenvironment in acute myeloid leukemia development and treatment. Expert Rev. Hematol. 7, 807–818 (2014).
Article CAS PubMed Google Scholar
Molica, M., Breccia, M., Foa, R., Jabbour, E. & Kadia, T. M. Maintenance therapy in AML: the past, the present and the future. Am. J. Hematol. 94, 1254–1265 (2019).
Article PubMed Google Scholar
Vetrie, D., Helgason, G. V. & Copland, M. The leukaemia stem cell: similarities, differences and clinical prospects in CML and AML. Nat. Rev. Cancer 20, 158–173 (2020).
Article CAS PubMed Google Scholar
Cerella, C. et al. ATP1A1/BCL2L1 predicts the response of myelomonocytic and monocytic acute myeloid leukemia to cardiac glycosides. Leukemia 38, 67–81 (2024).
Article CAS PubMed Google Scholar
Dohner, H., Weisdorf, D. J. & Bloomfield, C. D. Acute myeloid leukemia. N. Engl. J. Med. 373, 1136–1152 (2015).
Article PubMed Google Scholar
Bose, P., Vachhani, P. & Cortes, J. E. Treatment of relapsed/refractory acute myeloid leukemia. Curr. Treat. Options Oncol. 18, 17 (2017).
Article PubMed Google Scholar
Maude, S. L. et al. Tisagenlecleucel in children and young adults with B-cell lymphoblastic leukemia. N. Engl. J. Med. 378, 439–448 (2018).
Article CAS PubMed PubMed Central Google Scholar
Park, J. H. et al. Long-term follow-up of CD19 CAR therapy in acute lymphoblastic leukemia. N. Engl. J. Med. 378, 449–459 (2018).
Article CAS PubMed PubMed Central Google Scholar
Zuo, S. et al. C-JUN overexpressing CAR-T cells in acute myeloid leukemia: preclinical characterization and phase I trial. Nat. Commun. 15, 6155 (2024).
Article CAS PubMed PubMed Central Google Scholar
Yang, T. et al. LILRB4, an immune checkpoint on myeloid cells. Blood Sci. 4, 49–56 (2022).
Article PubMed PubMed Central Google Scholar
Churchill, H. R. O. et al. Leukocyte immunoglobulin-like receptor B1 and B4 (LILRB1 and LILRB4): Highly sensitive and specific markers of acute myeloid leukemia with monocytic differentiation. Cytom. B Clin. Cytom. 100, 476–487 (2021).
Article CAS Google Scholar
Dobrowolska, H. et al. Expression of immune inhibitory receptor ILT3 in acute myeloid leukemia with monocytic differentiation. Cytom. B Clin. Cytom. 84, 21–29 (2013).
Article Google Scholar
John, S. et al. A novel anti-LILRB4 CAR-T cell for the treatment of monocytic AML. Mol. Ther. 26, 2487–2495 (2018).
Article CAS PubMed PubMed Central Google Scholar
Deng, M. et al. LILRB4 signalling in leukaemia cells mediates T cell suppression and tumour infiltration. Nature 562, 605–609 (2018).
Article CAS PubMed PubMed Central Google Scholar
Waanders, E. et al. Mutational landscape and patterns of clonal evolution in relapsed pediatric acute lymphoblastic leukemia. Blood Cancer Discov. 1, 96–111 (2020).
Article CAS PubMed Google Scholar
Liu, J. et al. LILRB4, from the immune system to the disease target. Am. J. Transl. Res. 12, 3149–3166 (2020).
CAS PubMed PubMed Central Google Scholar
Aribi, A. et al. A phase 1 study of IO-202, an anti-LILRB4 antibody, in chronic myelomonocytic leukemia and acute myeloid leukemia. Blood Neoplasia 2, 100126 (2025).
Article PubMed PubMed Central Google Scholar
Liu, Y. et al. Chimeric STAR receptors using TCR machinery mediate robust responses against solid tumors. Sci. Transl. Med. 13, https://doi.org/10.1126/scitranslmed.abb5191 (2021).
Wang, J. et al. A novel adoptive synthetic TCR and antigen receptor (STAR) T-cell therapy for B-Cell acute lymphoblastic leukemia. Am. J. Hematol. 97, 992–1004 (2022).
Article CAS PubMed Google Scholar
Huang, D. et al. TCR-mimicking STAR conveys superior sensitivity over CAR in targeting tumors with low-density neoantigens. Cell Rep. 43, 114949 (2024).
Article CAS PubMed Google Scholar
Gui, X. et al. Disrupting LILRB4/APOE interaction by an efficacious humanized antibody reverses T-cell suppression and blocks AML development. Cancer Immunol. Res. 7, 1244–1257 (2019).
Article CAS PubMed PubMed Central Google Scholar
Paavola, K. J. et al. The fibronectin-ILT3 interaction functions as a stromal checkpoint that suppresses myeloid cells. Cancer Immunol. Res. 9, 1283–1297 (2021).
Article CAS PubMed PubMed Central Google Scholar
Gong, L. et al. LILRB4 represents a promising target for immunotherapy by dual targeting tumor cells and myeloid-derived suppressive cells in multiple myeloma. Haematologica 109, 3650–3669 (2024).
CAS PubMed PubMed Central Google Scholar
Sommer, C. et al. Allogeneic FLT3 CAR T cells with an off-switch exhibit potent activity against AML and can be depleted to expedite bone marrow recovery. Mol. Ther. 28, 2237–2251 (2020).
Article CAS PubMed PubMed Central Google Scholar
Liu, Y. et al. Integrated scFv identification and CAR T cell generation for AML targeting in vivo. Int J. Cancer 158, 994–1007 (2026).
Article CAS PubMed Google Scholar
Wang, H. et al. LILRB4 on multiple myeloma cells promotes bone lesion by p-SHP2/NF-kappaB/RELT signal pathway. J. Exp. Clin. Cancer Res. 43, 183 (2024).
Article PubMed PubMed Central Google Scholar
Daver, N., Alotaibi, A. S., Bucklein, V. & Subklewe, M. T-cell-based immunotherapy of acute myeloid leukemia: current concepts and future developments. Leukemia 35, 1843–1863 (2021).
Article CAS PubMed PubMed Central Google Scholar
Saito, S. & Nakazawa, Y. CAR-T cell therapy in AML: recent progress and future perspectives. Int J. Hematol. 120, 455–466 (2024).
Article PubMed Google Scholar
Haubner, S., Subklewe, M. & Sadelain, M. Honing CAR T cells to tackle acute myeloid leukemia. Blood 145, 1113–1125 (2025).
Article CAS PubMed Google Scholar
Kim, M. Y. et al. Genetic inactivation of CD33 in hematopoietic stem cells to enable CAR T cell immunotherapy for acute myeloid leukemia. Cell 173, 1439–1453.e1419 (2018).
Article CAS PubMed PubMed Central Google Scholar
Sugita, M. et al. Allogeneic TCRalphabeta deficient CAR T-cells targeting CD123 in acute myeloid leukemia. Nat. Commun. 13, 2227 (2022).
Article CAS PubMed PubMed Central Google Scholar
Jin, X. et al. First-in-human phase I study of CLL-1 CAR-T cells in adults with relapsed/refractory acute myeloid leukemia. J. Hematol. Oncol. 15, 88 (2022).
Article CAS PubMed PubMed Central Google Scholar
Tong, C. et al. Optimized tandem CD19/CD20 CAR-engineered T cells in refractory/relapsed B-cell lymphoma. Blood 136, 1632–1644 (2020).
PubMed PubMed Central Google Scholar
Yu, L. et al. Converting TCR-based chimeric antigen receptor STAR into dual-specific targeting receptor for cancer immunotherapy. Mol. Ther. https://doi.org/10.1016/j.ymthe.2025.02.001 (2025).
Berdeja, J. G. et al. Ciltacabtagene autoleucel, a B-cell maturation antigen-directed chimeric antigen receptor T-cell therapy in patients with relapsed or refractory multiple myeloma (CARTITUDE-1): a phase 1b/2 open-label study. Lancet 398, 314–324 (2021).
Article CAS PubMed Google Scholar
Rubino, V. et al. IL-21/IL-21R signaling renders acute myeloid leukemia stem cells more susceptible to cytarabine treatment and CAR T cell therapy. Cell Rep. Med. 5, 101826 (2024).
Article CAS PubMed PubMed Central Google Scholar
Li, Z. et al. LILRB4 ITIMs mediate the T cell suppression and infiltration of acute myeloid leukemia cells. Cell Mol. Immunol. 17, 272–282 (2020).
Article PubMed Google Scholar
Bhagwat, A. S. et al. Cytokine-mediated CAR T therapy resistance in AML. Nat. Med. 30, 3697–3708 (2024).
Article CAS PubMed PubMed Central Google Scholar
Sharma, N., Atolagbe, O. T., Ge, Z. & Allison, J. P. LILRB4 suppresses immunity in solid tumors and is a potential target for immunotherapy. J. Exp. Med. 218, https://doi.org/10.1084/jem.20201811 (2021).
McGinnis, C. S., Murrow, L. M. & Gartner, Z. J. DoubletFinder: doublet detection in single-cell RNA sequencing data using artificial nearest neighbors. Cell Syst. 8, 329–337 e324 (2019).
Article CAS PubMed PubMed Central Google Scholar
Hao, Y. et al. Integrated analysis of multimodal single-cell data. Cell 184, 3573–3587.e3529 (2021).
Article CAS PubMed PubMed Central Google Scholar
Korsunsky, I. et al. Fast, sensitive and accurate integration of single-cell data with Harmony. Nat. Methods 16, 1289–1296 (2019).
Article CAS PubMed PubMed Central Google Scholar
Jiang, S. et al. Cell Taxonomy: a curated repository of cell types with multifaceted characterization. Nucleic Acids Res. 51, D853–D860 (2023).
Article CAS PubMed PubMed Central Google Scholar
Jin, S., Plikus, M. V. & Nie, Q. CellChat for systematic analysis of cell-cell communication from single-cell transcriptomics. Nat. Protoc. 20, 180–219 (2025).
Article CAS PubMed Google Scholar

Download references

Acknowledgements

This work was supported by the following: Noncommunicable Chronic Diseases-National Science and Technology Major Project (No. 2023ZD0501200), National Natural Science Foundation of China (Nos. 82350105, 82293664, 82341212, 31930039, and 82270228), Annual Fund from Tsinghua-Peking Center for Life Science, Science, Technology & Innovation Project of Xiongan New Area (No. 2023XACX0004), Beijing Outstanding Young Scientists Project (JWZQ20240101001), Beijing Natural Science Foundation (Z240019), Beijing Municipal Science & Technology Commission, Administrative Commission of Zhongguancun Science Park (No. Z241100009024043), and Peking University People’s Hospital (RZ2024-01), Peking University Medicine plus X Pilot Program-Key Technologies R&D Project (No. BMU2024YXXLHGG009).

Author information

These authors contributed equally: Meng Lv, Hongli Zheng, Xuying Pei, Lei Lei

Authors and Affiliations

Peking University People’s Hospital, Peking University Institute of Hematology, National Clinical Research Center for Hematologic Disease, Beijing Key Laboratory of Cell and Gene Therapy for Hematologic Malignancies, Peking University, Beijing, China
Meng Lv, Xuying Pei, Wenbo Chen, Yutong Bao, Yiyang Ding, Zhengli Xu, Lijuan Hu, Xiaojun Huang & Xiangyu Zhao
Bristar Immunotech Limited, Beijing, China
Hongli Zheng, Lei Lei, Zhe Wan & Xueqiang Zhao
School of Basic Medical Sciences, Tsinghua University School of Medicine, Beijing, China
Zhixiao Zhou & Xin Lin
Peking-Tsinghua Center for Life Sciences, Academy for Advanced Interdisciplinary Studies, Peking University, Beijing, China
Xiaojun Huang
Changping Laboratory, Beijing, China
Jiasheng Wang

Authors

Meng Lv
View author publications
Search author on:PubMed Google Scholar
Hongli Zheng
View author publications
Search author on:PubMed Google Scholar
Xuying Pei
View author publications
Search author on:PubMed Google Scholar
Lei Lei
View author publications
Search author on:PubMed Google Scholar
Wenbo Chen
View author publications
Search author on:PubMed Google Scholar
Yutong Bao
View author publications
Search author on:PubMed Google Scholar
Yiyang Ding
View author publications
Search author on:PubMed Google Scholar
Zhengli Xu
View author publications
Search author on:PubMed Google Scholar
Lijuan Hu
View author publications
Search author on:PubMed Google Scholar
Zhe Wan
View author publications
Search author on:PubMed Google Scholar
Xueqiang Zhao
View author publications
Search author on:PubMed Google Scholar
Zhixiao Zhou
View author publications
Search author on:PubMed Google Scholar
Xin Lin
View author publications
Search author on:PubMed Google Scholar
Xiaojun Huang
View author publications
Search author on:PubMed Google Scholar
Jiasheng Wang
View author publications
Search author on:PubMed Google Scholar
Xiangyu Zhao
View author publications
Search author on:PubMed Google Scholar

Contributions

Meng Lv: Writing—original draft, Validation, Methodology, Investigation, Formal analysis, Data curation. Hongli Zheng: Methodology, Investigation, Clinical trial design, Formal analysis, Data curation. Xuying Pei: Investigation, Formal analysis, Data curation. Lei Lei: Investigation, Data curation. Wen-Bo Chen: Investigation, Formal analysis, Data curation. Yutong Bao: Investigation, Data curation. Yiyang Ding: Investigation, Formal analysis, Data curation. Zhengli Xu: Formal analysis, Data curation. Lijuan Hu: Formal analysis, Data curation. Zhe Wan: STAR-T-cell preparation, Quality control. Xueqiang Zhao: Validation, Project administration, Conceptualization. Zhixiao Zhou: Investigation, Formal analysis, Data curation. Xin Lin: Validation, Supervision, Project administration, Funding acquisition. Xiaojun Huang: Validation, Supervision, Project administration, Funding acquisition. Jiasheng Wang: Writing—review and editing, Supervision, Project administration. Xiangyu Zhao: Writing—review and editing, Supervision, Project administration, Funding acquisition, Conceptualization. All the authors have read and approved the article.

Corresponding authors

Correspondence to Xin Lin, Xiaojun Huang, Jiasheng Wang or Xiangyu Zhao.

Ethics declarations

Competing interests

L.L., X.Q.Z., and X.L. have patents related to LILRB4 STARs. L.L. and H.L.Z. are employees of Bristar Immunotech Limited Co., Ltd. The remaining authors declare that they have no other competing interests.

Ethical approval

The animal studies were reviewed and approved by the Institutional Animal Care and Use Committee of Peking University Health Science Center (CIBR-IACUC-042), and the clinical trial was approved by the Ethics Committee of Peking University People’s Hospital (Ethics Approval Number: 2022PHD007).

Supplementary information

Supplement Materials (download DOCX )

Full clinical trial protocol-NCT05548088 (download DOCX )

Raw data (download PPTX )

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Cite this article

Lv, M., Zheng, H., Pei, X. et al. Dual epitope anti-LILRB4 synthetic T-cell receptor and antigen receptor (STAR)-T-cell therapy for relapsed/refractory acute myeloid leukemia. Sig Transduct Target Ther 11, 203 (2026). https://doi.org/10.1038/s41392-026-02765-7

Download citation

Received: 28 September 2025
Revised: 19 February 2026
Accepted: 27 February 2026
Published: 01 June 2026
Version of record: 01 June 2026
DOI: https://doi.org/10.1038/s41392-026-02765-7