A combinatorial CAAR-T cell strategy to eliminate factor VIII inhibitors in preclinical models of hemophilia A

Summary

Chimeric autoantibody receptor T (CAAR-T) cell therapy extends CAR-T technology to target autoreactive immune cells, offering a novel approach for treating autoimmune disorders. In hemophilia A, the development of anti-factor VIII (FVIII) inhibitors poses a major clinical challenge. This study investigates the therapeutic potential of CAAR-T cells engineered with FVIII C1 or A2 domains to eliminate inhibitor-producing B cells. In vitro cytotoxicity assays confirmed that C1 and A2 CAAR-T cells exhibited comparable killing efficiency against their respective targets. In an F8KO hemophilia A mouse model, both CAAR-T cells significantly reduced circulating FVIII inhibitors, while their combination further suppressed inhibitor re-emergence after FVIII rechallenge. In a humanized NCG mouse model, combined CAAR-T therapy effectively inhibited in vivo target cell expansion. These findings demonstrate that CAAR-T therapy, particularly the combinatorial approach, holds promise for addressing FVIII inhibitor formation in hemophilia A, offering a targeted and effective treatment strategy.

Graphical abstract

Highlights

• •FVIII-targeted CAAR-T cells eliminate inhibitor-producing B cells in hemophilia A models
• •The FVIII C1 domain is a functional and immunologically relevant CAAR-T target
• •Combinatorial targeting of FVIII A2 and C1 domains enhances inhibitor clearance in vivo
• •FVIII CAAR-T cells show persistence and a favorable safety profile in mice

Hematology; molecular biology; immunology

Introduction

Hemophilia A (HemA) is an X-linked recessive bleeding disorder caused by mutations in the coagulation FVIII gene, resulting in deficient or dysfunctional FVIII protein. This disease primarily affects males and is clinically characterized by lifelong recurrent spontaneous bleeding into joints and muscles, severely impairing patients' quality of life and potentially causing disability.^1,2 The FVIII gene is located at the distal end of the long arm of the X chromosome (Xq28). The encoded FVIII protein comprises multiple domains arranged sequentially as A1-A2-B-A3-C1-C2, with the A and C domains playing critical roles in the protein’s functional stability and immunological properties.^3,4

Currently, replacement therapy using plasma-derived or recombinant FVIII (rFVIII) concentrates is the mainstay for preventing and controlling bleeding episodes in HemA patients. However, since reinfused rFVIII is recognized as a “neoantigen” by the patient’s immune system, approximately 20%–30% of severe HemA patients and 3%–13% of mild to moderate cases develop neutralizing antibodies (inhibitors) against FVIII during treatment.^1,3 These inhibitors neutralize FVIII’s coagulation function, significantly reducing therapeutic efficacy, sometimes rendering treatment ineffective, and imposing substantial economic and life-threatening burdens.

Each domain of FVIII has the potential to be recognized as an epitope by the immune system and contribute to inhibitor formation. Among them, the A2 and C1 domains are particularly immunogenic, frequently triggering B cell-mediated immune responses.^5,6 The A2 domain is essential for FVIII procoagulant function, as it participates in the intrinsic tenase complex and facilitates factor X activation; notably, its structural exposure and functional importance also make it a frequent target of inhibitor antibodies.^7,8 The C1 domain, recently recognized for its dual role in both immune regulation and immunogenicity, also binds von Willebrand factor (vWF) to limit FVIII uptake by antigen-presenting cells (APCs), thereby modulating immunogenic potential.^9,10,11,12 Therefore, the inhibitors targeting these domains are often abundant in HemA patients.^13,14

To address the challenges posed by FVIII inhibitors, several innovative strategies have been explored, including gene therapy, the engineering of low-immunogenic rFVIII variants, and immune tolerance induction.^15,16,17,18 Among these, CAR-T technology has emerged as a promising therapeutic platform for autoimmune diseases.^19,20 In particular, chimeric autoantibody receptor T (CAAR-T) cells, which display disease-relevant antigenic epitopes on their surface to specifically recognize and eliminate autoreactive immune cells, have shown encouraging results in preclinical and early clinical studies of pemphigus vulgaris, muscle-specific kinase antibody-positive myasthenia gravis, and autoimmune encephalitis.^21,22,23,24

Given that inhibitor formation in HemA can be conceptually regarded as a treatment-induced autoimmune process. Therefore, CAAR-T cell technology can be explored as a strategy to eliminate FVIII-specific B cells. Previous studies have developed CAAR-T and CAAR-Treg cells targeting the A2 and C2 domains of FVIII, both of which demonstrated effective clearance of reactive B cells.^25,26

In this study, we report for the first time the construction of a CAAR-T cell targeting the FVIII C1 domain. Using both in vitro and in vivo models, we demonstrated that C1-CAAR-T cells are effective in eliminating FVIII inhibitors, with A2-CAAR-T cells serving as a comparator. Given the clinical challenge that HemA patients often develop a heterogeneous population of FVIII inhibitors against multiple domains, we hypothesized that a combinatorial approach utilizing both A2- and C1-specific CAAR-T cells could enhance inhibitor clearance and suppress the generation of new inhibitors upon continued FVIII exposure. Our results in both FVIII gene knockout (F8KO) and NOD-SCID IL2Rγnull (NCG) mouse models confirmed this hypothesis, providing a preclinical rationale for the application of domain-targeted combinatorial CAAR-T cell therapy as a potential strategy to overcome FVIII inhibitor complications in HemA treatment.

Result

Design of CAAR-T cells to target FVIII inhibitors-producing cells

As shown in Figure 1A, FVIII needs to be cleaved into heavy and light chains before it can function properly in the blood clotting process. Within the active FVIII complex, the A2 domain on the heavy chain binds to activated factor IX to help drive coagulation. Meanwhile, the C1 domain on the light chain interacts with phospholipids on platelet surfaces and with vWF, helping to stabilize FVIII, regulate its clearance from the body, and influence how the immune system recognizes it. Recent studies highlight the critical role of the C1 domain in FVIII immunogenicity. Importantly, neutralizing antibodies that target the C1 domain make up a significant portion of FVIII inhibitors, making C1 an attractive target for therapeutic intervention.¹⁴

Figure 1.

Construction of CAAR-T cells targeting FVIII inhibitors and their functional characterization

(A) Schematic diagram of FVIII proteolytic fragments. The domain structure of the FVIII protein after physiological cleavage is shown, illustrating its division into a heavy chain (A1-A2) and light chain (A3-C1-C2).

(B) Design of chimeric autoantibody receptor constructs. Two CAAR constructs were generated using the A2 or C1 domains of FVIII as target antigens. Each CAAR consists of the FVIII-derived domain fused to CD28 transmembrane and co-stimulatory regions, CD3ζ signaling domain, and a P2A self-cleaving peptide linked to GFP for expression tracking.

(C) Flow cytometric analysis of CAAR expression on human T cells. Top, GFP reporter was used to assess transduction efficiency of A2-CAAR and C1-CAAR constructs, with both showing high GFP⁺ proportions compared to control T cells. Bottom, HA-tag detection was used to confirm surface expression of CAAR constructs; the HA-tag was inserted between the A2 or C1 extracellular domain and the CD28 transmembrane region within the CAAR plasmid.

(D) Characterization of engineered K562 target cells. BOIIB2- and RHD5-expressing K562 cells were sorted by flow cytometry based on strong mCherry fluorescence, achieving >97% purity in both lines.

(E) Cytotoxic activity of CAAR-T cells against antigen-expressing target cells. Cytotoxicity assays were performed using LDH release to evaluate the lytic activity of A2-CAAR T and C1-CAAR T cells at various effector-to-target (E:T) ratios (0.3:1, 1:1, and 3:1). A2-CAAR T cells were tested against BOIIB2 targets, and C1-CAAR T cells against RHD5 targets. UTD were used as controls. Results are shown as mean ± SEM (n = 4). Experiments were independently repeated 3 times. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001.

(F) IFN-γ secretion by CAAR-T cells upon target recognition. ELISA quantification of IFN-γ released by CAAR-T cells co-cultured with their respective antigen-expressing targets at an E:T ratio of 3:1. Both A2-CAAR T and C1-CAAR T cells showed significantly increased cytokine release compared to UTD controls. Results are shown as mean ± SEM (n = 4). Experiments were independently repeated 3 times. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001.

(G) Degranulation activity of CAAR-T cells assessed by CD107α expression. Flow cytometry analysis of CD107α surface mobilization in CAAR-T cells after 4 h co-culture with target cells at an E:T ratio of 3:1. Quantification and mean fluorescence intensity (MFI) demonstrates elevated degranulation in both A2-CAAR and C1-CAAR groups compared to UTD controls. Results are shown as mean ± SEM (n = 4). Experiments were independently repeated 2 times. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001.

Moreover, it has also been shown that A2 domain-redirected regulatory T cells can suppress immune responses against FVIII.²⁵ Therefore, we designed two types of CAAR-T cells, recognizing the A2 or C1 domain antibodies, to broadly eliminate FVIII inhibitors (Figure 1B). In our design, either the A2 or C1 domain serves to redirect the CAR, recognizing B cells that produce anti-FVIII antibodies. These domains were linked to the intracellular signaling regions of CD28 and CD3z via transmembrane domain, and tagged with HA for detection. In addition, a P2A-tandemed GFP reporter gene allowed to track the CAR-expressing T cells.

The CAAR constructs were cloned into lentiviral vectors and used to transform activated human T cells. Flow cytometry showed that the percentage of GFP was 60%, and exhibited correlation with the CAAR expression on cell surface detected with anti-HA staining. Both A2-CAAR and C1-CAAR T cells showed similar transduction efficiencies, with GFP offering feasible and sensitive indication (Figure 1C).

In vitro cytotoxic activity of CAAR-T cells against FVIII inhibitor-producing cells

To test whether the engineered CAAR-T cells could specifically recognize and kill FVIII inhibitor-producing cells, we construct target cell models by inserting single-chain variable fragment (scFv) derived from known anti-A2 (BOIIB2) and anti-C1 (RHD5) antibodies into K562 cells.^21,22 These engineered cells mimic FVIII inhibitor-producing B cells and were sorted to ensure the purity (Figure 1D).

Next, A2-CAAR and C1-CAAR T cells were co-cultured with their respective target cells. After 24 h, we assessed killing efficacy using the lactate dehydrogenase (LDH) release assay. The LDH was released from the dead cells, and more LDH suggested more cytotoxic activity. Compared to untransduced (UTD) T cells, both A2-and C1-CAAR T cells showed significantly enhanced killing against their target cells (Figure 1E). Therefore, both A2-CAAR and C1-CAAR T cells demonstrated specific and potent killing effects in a dose-dependent manner.

We further assessed CAAR-T cell functionality by measuring interferon gamma (IFN-γ) secretion and CD107a expression post-co-culture. ELISA revealed IFN-γ levels increased more than 4-fold in the CAAR-T group compared to the UTD T group. Moreover, flow cytometry showed approximately 50% higher CD107a expression in CAAR-T cells (Figures 1F and 1G), indicating robust immune activation and cytotoxic potential.

Taken together, these findings demonstrated that CAAR-T cells built using the FVIII A2 or C1 domains can specifically recognize and destroy B cells that produce FVIII inhibitors. These engineered cells exhibit both target specificity and immune activation, highlighting their promise as a new type of cell-based therapy to eliminate FVIII inhibitors in patients with HemA.

In vivo therapeutic efficacy of CAAR-T cells in the HemA mouse model

To investigate whether CAAR-T cells can suppress FVIII inhibitors in vivo, we employed a mouse model that mimics how patients with HemA develop these inhibitors. F8KO mice, lacking immune tolerance to FVIII due to genetic knockout of the FVIII gene, were subcutaneously immunized for 3 weeks with 10 IU of recombinant human FVIII mixed with aluminum adjuvant (Figure 2A). Post-immunization serum analysis confirmed inhibitor levels of 40–60 μg/μL by ELISA (Figure 2C), demonstrating successful model induction.

Figure 2.

In vivo evaluation of CAAR-T cells targeting FVIII inhibitors in a hemophilia A mouse model

(A) Schematic of the FVIII inhibitor mouse model and CAAR-T cell treatment timeline. FVIII-deficient mice were immunized subcutaneously with recombinant FVIII (rFVIII and 10 IU) once per week for 3 consecutive weeks (day −21, −15, and −8) to induce anti-FVIII inhibitor formation. On day 0, mice received adoptive transfer of CAAR-T cells. CD45.1⁺ donor mice were used as the T cell source for tracking transferred cells.

(B) Design of chimeric autoantibody receptor constructs. Two CAAR constructs were generated using the A2 or C1 domains of FVIII as target antigens. Each CAAR consists of the FVIII-derived domain fused to mice CD28 transmembrane and co-stimulatory regions, mice CD3ζ signaling domain, and a P2A self-cleaving peptide linked to GFP for expression tracking.

(C) Measurement of FVIII inhibitor titers in serum following CAAR-T therapy. Inhibitor titers were measured by ELISA over time. Compared to UTD, both A2-CAAR and C1-CAAR treatments significantly reduced FVIII inhibitor levels in serum starting from day 0 and continuing through day 70. The ELISA detection threshold was defined based on the IgG levels of unimmunized F8KO mice, all of which were undetectable. n = 5 per group; experiments were independently repeated 3 times. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

(D) Detection of CAAR-T cell persistence in the spleen. Immunohistochemical staining of splenocytes harvested on day 35 post-transfer revealed the presence of CD45.1⁺ donor CAAR-T cells in mice treated with A2-CAAR and C1-CAAR, but not in the UTD group (brown DAB signal). HA-tag staining indicates the presence of infused CAAR-T cells in the spleen. Scale bars, 50 μm.

(E) Schematic illustration of the FVIII re-challenge model. F8 knockout mice were first immunized subcutaneously with 10 IU rFVIII on day −21 to induce inhibitor formation, followed by infusion of CAAR-T cells on day 0. On day 14, a second dose of rFVIII (10 IU) was administered to assess the durability of immune suppression.

(F) Quantification of FVIII inhibitor levels over time. On day 14, mice received an FVIII re-challenge, and plasma inhibitor levels were measured on day 35. Mice treated with combined C1 + A2 CAAR-T cells showed significantly reduced inhibitor titers compared to the A2-CAAR T alone group, indicating enhanced suppression of anti-FVIII immunity. n = 5 per group; experiments were independently repeated 2 times. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

(G) CAART therapy reduces splenic cells that secrete FVIII inhibitors in hemophilic mice. After establishing the hemophilia mouse model, mice were randomized to receive an infusion of UTD or CAART. On day 21 post-infusion, splenocytes were collected and antigen-specific ELISpot assays (FVIII-coated plates) were performed to detect cells secreting FVIII inhibitors.

(H) Bethesda assay quantification of inhibitory anti-FVIII antibody titers in plasma 21 days post-infusion. Groups: UTD, A2-CAART, C1-CAART, and A2+C1-CAART. Data are shown as mean ± SD. Significance: ns, not significant; ∗p < 0.05; ∗∗∗p < 0.001.

(I) Representative flow cytometry plots of splenocytes collected on day 35 post-infusion. The percentages of CD45.1⁺CD44⁺ cells, indicating persisting donor-derived CAAR-T memory T cells in the spleen, are shown for each group.

To evaluate the in vivo efficacy in the mouse, we modified the CAAR design by substituting human CD28 and CD3ζ signaling domains with their mouse counterparts (Figure 2B). Mouse CAAR-T cells were generated from CD45.1 donor mouse, allowing distinction from CD45.2-expressing F8KO recipient mice. Notably, inhibitor levels in the mice declined significantly within 7 days post-infusion of A2-CAAR or C1-CAAR T cells, returning to baseline by day 70 (Figure 2C). To assess the long-term survival and immune memory potential of infused CAAR-T cells, spleens were collected on day 70 for analysis. Immunohistochemistry detected clusters of CD45.1⁺ cells and HA-Tag⁺ cells in the spleens of treated mice (Figure 2D), indicating CAAR-T cell persistence and localization to lymphoid tissues. At day 21 post-infusion, FVIII-specific ELISpot analysis of splenocytes showed that both A2-CAART and C1-CAART treatment markedly reduced the number of FVIII inhibitor-secreting cells compared with the UTD control, indicating effective in vivo clearance of pathogenic B cells (Figure 2G). Moreover, the combined A2 + C1 CAART treatment achieved the most pronounced reduction. Consistently, plasma Bethesda assays showed substantially lower titers of inhibitory anti-FVIII antibodies in all CAART-treated groups than in the UTD group, with the combination group exhibiting the greatest decline (Figure 2H). These results, combined with prior in vitro findings, confirm that both CAAR-T constructs effectively target and eliminate FVIII inhibitor-producing cells, and maintain for long-term in vivo.

Given that FVIII inhibitors often target multiple domains within regions A and C, we compared A2-CAAR T cells alone to a combination of A2- and C1-CAAR T cells (Figure 2E). Both approaches reduced inhibitor levels, but the combination therapy showed a more significant reduction, with serum inhibitor levels in the dual-therapy group being 2-fold lower than those in the A2-CAAR-only group by day 14 (Figure 2F). Notably, the combination group maintained low inhibitor levels even after rFVIII re-exposure, suggesting enhanced immune memory. At the study endpoint, flow cytometric analysis of splenocytes revealed a population of CD45.1⁺CD44⁺ cells in CAAR-T-treated mice (Figure 2I), consistent with a memory T cell phenotype. These findings suggest that CAAR-T cells persisted in vivo and may contribute to sustained protection against inhibitor recurrence.

Safety assessment of CAAR-T therapy

To evaluate the safety profile of CAAR-T therapy, histological analyses of the heart, liver, spleen, and kidneys from treated mice were performed using hematoxylin and eosin (H&E) staining. The results showed no evidence of tissue damage or pathological alterations in any group, including those receiving both A2- and C1-CAAR-T cells (Figure 3). In addition, longitudinal monitoring of body weight during the first 7 days post-infusion revealed no significant differences between CAAR-T-treated and control mice (Figure S1), further supporting that the treatment was well tolerated in vivo.

Figure 3.

Histopathological safety evaluation of CAAR-T cell therapy in major organs

H&E staining was performed on major organs harvested from mice treated with UTD, C1-CAART, and A2-CAART, or combination A2 + C1-CAART to assess off-target toxicity. Tissue sections from liver (A), spleen (B), kidney (C), and heart (D) were examined under light microscopy. Scale bars, 100 μm.

Evaluating efficacy of CAAR-T cells in a humanized mouse model

To advance the clinical application of CAAR-T therapy, we assessed human CAAR-T cells in humanized immunodeficient NCG mice, a preclinical model. We simulated a polyclonal FVIII inhibitor response by subcutaneously injecting a 1:1 mix of K562 cells expressing BOIIB2 or RHD5 antibodies on the cell surface, targeting the A2 and C1 domains of FVIII. Once tumors were established, mice received either UTD T cells (10 million cells/mouse) or a combination of A2 and C1 CAAR-T cells (5 million cells for each CAAR/mouse). The CAAR-T group showed significant tumor suppression compared to controls, with significant tumor regression from day 14 and near elimination by day 24. Final tumor size and weight were significantly lower in the treated group (Figures 4B and 4C).

Figure 4.

Evaluation of CAAR-T cell efficacy in a humanized xenograft tumor model

(A) Experimental schematic of humanized tumor model and CAAR-T treatment. NCG immunodeficient mice were subcutaneously inoculated with a 1:1 mixture of K562 cells expressing the human autoantigens BOIIB2 and RHD5. On days 3 and 7 post-inoculation, mice received intravenous administration of a combination of human A2-CAAR and C1-CAAR T cells. Tumor progression was monitored until the experimental endpoint on day 24.

(B) Tumor volume kinetics post-treatment. Tumor growth was significantly suppressed in mice receiving A2 + C1 CAAR-T cells compared to untreated controls. Volumetric measurements were taken at multiple time points, with significant differences observed by day 18 and onward. n = 5 per group; ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

(C) Tumor weight at endpoint. Final tumor weights were recorded on day 24. The A2 + C1 CAAR-T treatment group showed a dramatic reduction in tumor burden compared to the UTD group. n = 5 per group; ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

(D) Persistence of human T cells in host tissues. Flow cytometric analysis was performed at endpoint to detect residual human T cells (hCD3⁺GFP⁺) in spleen and bone marrow. Mice treated with mixed CAAR-T cells demonstrated significantly higher frequencies of hCD3⁺GFP⁺ cells compared to controls.

(E) Histological analysis of tumor tissues by H&E. Representative hematoxylin and eosin (H&E) staining images of tumor sections from the UTD and mix-CAAR-T groups at study endpoint. Scale bars, 100 μm.

On day 24, flow cytometry analysis of the spleen and bone marrow detected human CD3⁺GFP⁺ T cells in treated mice, confirming in vivo CAAR-T cell activation and persistence (Figure 4D). H&E staining of tumor sections 10 days post-CAAR-T cells infusion revealed robust lymphocyte infiltration, with fewer viable tumor cells compared to the UTD T cell group. In contrast, tumors in UTD-treated mice remained dense and largely intact, exhibiting minimal immune infiltration (Figure 4E). These findings confirm the potent and specific in vivo activity of human CAAR-T cells against FVIII inhibitor-producing cells in a humanized mouse model.

Discussion

B cells, as key components of the adaptive immune system, exhibit dual roles under various pathophysiological conditions. On one hand, they contribute to host defense by producing protective antibodies against pathogens. On the other hand, they can generate pathogenic autoantibodies or activate T cells via antigen presentation and costimulatory pathways, thereby playing a central pathogenic role in multiple autoimmune diseases,^27,28 including FVIII inhibitor formation in hemophilia, systemic lupus erythematosus,²⁹ autoimmune encephalitis,²² and myasthenia gravis.²³ Consequently, B cell-targeted depletion strategies have become a research focus in recent years for treating autoimmune disorders, with the successful application of CD19 CAR-T cell therapy providing a strong theoretical and technical foundation for B cell-related diseases. In this study, we report for the first time the construction of CAAR-T cells targeting the FVIII C1 domain, which were validated in an F8KO mouse model immunized with human rFVIII. The results demonstrated that this cell therapy effectively eliminated high-titer FVIII inhibitors, showing promising therapeutic potential and safety, thus offering a novel approach for inhibitor eradication.

The unique “spikes” structure of the C1 domain facilitates binding to phosphatidylserine on activated platelet surfaces, promoting assembly of the tenase complex and significantly enhancing coagulation efficiency.^30,31 Notably, the C1 domain plays a key part in FVIII uptake by APCs and antigen presentation, influencing FVIII clearance rates and immune recognition.^9,11 Previous studies have shown that anti-C1 antibodies accelerate FVIII clearance and interfere with multiple FVIII functions, and they are commonly found in inhibitor-positive patients and significantly affect FVIII half-life and immunogenicity. Among the FVIII domains, the C1 domain displays relatively higher sequence conservation compared with A2 and C2.¹⁰ Inhibitors targeting the C1 domain may possess higher pathogenic potential due to their ability to disrupt critical functional interactions of FVIII with vWF and phospholipid surfaces. Therefore, it is rational for selecting the C1 domain as a specific antigenic target in our CAAR-T cell design. By focusing on a structurally conserved and immunologically relevant domain, our approach aims to achieve efficient depletion of pathogenic inhibitors while minimizing unintended immune toxicity.

Several studies have demonstrated the efficacy of CAAR-T/Treg cells targeting the A2 and C2 domains—either individually or in combination—in both in vitro and in vivo models.^25,26 However, although the C1 domain plays a similarly critical role in the immunological recognition of FVIII, and anti-C1 antibodies are considered highly pathogenic, the therapeutic potential of C1 as a CAAR-T target has not yet been fully explored. In this study, we constructed a novel C1-CAART based on the domain’s pivotal role in immunogenic responses. We systematically evaluated its capacity to selectively eliminate FVIII inhibitor-producing B cells and to reduce inhibitor levels both in vitro and in vivo. Given that the C1 domain is relatively conserved, it was important to consider whether the C1-CAAR structure might interact with vWF in vivo, potentially influencing its cytotoxic activity. Our in vitro assays demonstrated that exposure of C1-CAAR-T cells to various concentrations of vWF did not lead to noticeable activation, suggesting that the C1 domain does not cross-react with vWF and that the observed therapeutic effects are unlikely to result from nonspecific activation. This finding further supports the safety profile of the C1-CAAR design (Figure S2).

The hemophilia patients receiving recombinant FVIII frequently develop antibodies against multiple antigenic epitopes. Therefore, single-target CAAR-T therapy may be insufficient to comprehensively eliminate all B cell clones producing inhibitors, limiting its broad and durable clinical application. To address this, we further developed and applied a dual-target CAAR-T cell approach simultaneously targeting the A2 and C1 domains. Notably, in combination therapy experiments, the dual-target CAAR-T cells not only exhibited enhanced efficacy in clearing FVIII inhibitors during the initial immunization phase but also effectively prevented inhibitor resurgence upon subsequent recombinant FVIII re-challenge. Importantly, in preclinical evaluations, the dual-target CAAR-T cells effectively eliminated tumor cells mimicking pathogenic B cells in the humanized mouse model and demonstrated a favorable safety profile. Importantly, such a strategy may more closely reflect the clinical reality, as HemA patients undergoing FVIII replacement therapy often develop polyclonal antibody responses targeting multiple domains of the FVIII molecule. Therefore, dual-target CAAR T cell therapy holds promise as a more comprehensive and clinically relevant approach for managing diverse and persistent FVIII inhibitors.

In summary, we developed C1-CAAR T and used A2-CAAR T as a control to validate that C1-CAART monotherapy and A2 + C1 dual-target therapy effectively eliminated inhibitor-producing B cells in mice and significantly reduced the levels of FVIII inhibitors in the serum. To further simulate clinical applications, we performed a rechallenge in mice following the infusion of CAART, mimicking the scenario in which patients require continuous rFVIII treatment. The combined therapy continued to demonstrate good efficacy in inhibiting inhibitor production, providing the evidence for the clinical translation of CAAR-T technology in treating immune complications of HemA.

Limitations of the study

This study has several limitations. The functional evaluation of the engineered CAAR-T cells was performed in vitro and in animal models. Although these experiments demonstrated effective clearance of FVIII inhibitors, they cannot fully recapitulate the human physiological environment or the complex immune status of patients with HemA. Therefore, further studies incorporating clinical samples from patients are required to validate the efficacy and translational potential of CAAR-T therapy.

Resource availability

Lead contact

Requests for further information and resources should be directed to and will be fulfilled by the lead contact, Dr. Yan Chen (cheny658@mail.sysu.edu.cn).

Materials availability

This study did not generate new unique reagents.

Data and code availability

All data in this paper will be shared by the lead contact upon request. This paper does not report original code.

Acknowledgments

This study was supported by the Shenzhen Science and Technology Innovation Commission (grant no. JCYJ20220530144207017), National Natural Science Foundation of China (82171825), Health Commission of Guangdong Province Program (0920220202), and Futian Healthcare Research Project (FTWS015).

Author contributions

Y.C. and X.H. designed and supervised the study; W.L., Q.Y., and S.L. performed the experiments and conducted data analysis; X.Z. investigated and wrote the draft; R.W. and L.C. were responsible for volunteer recruitment; Y.C. provided funding support for the study. All the authors read and approved the final version of the manuscript.

Declaration of interests

The authors declare no competing interests.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
anti-human CD3/CD28 antibodies	Gibco	Cat#11161D
APC anti-human CD107a (LAMP-1) Antibody	Biolegend	Cat#328619
APC anti-mouse CD45.1 Antibody	Biolegend	Cat#110713
Pacific Blue™ anti-human CD3	Biolegend	Cat#980004
HA Tag Recombinant Rabbit Monoclonal Antibody	Invitrogen	Cat#MA5-27915
Pacific Blue™ anti-mouse CD44 Antibody	Biolegend	Cat#156005
CD45.1 Recombinant monoclonal antibody	Proteintech	Cat#84325-2-RR
CD3 Monoclonal Antibody	Invitrogen	Cat#14-0038-82
Critical commercial assays
MojoSort™ Human CD3 T cell Isolation Kit	Biolegend	Cat#480022
MojoSort™ Mouse CD3 T cell Isolation Kit	Biolegend	Cat#480024
ImmunoCult™-XF T cell Expansion Medium	STEMCELL	Cat#10981
Cytotoxicity LDH Assay Kit-WST	DOJINDO	Cat#CK12
Human IFN-γ(Interferon Gamma) ELISA Kit	Elabscience	Cat#E-EL-H0108
ALL-IN-ONE Mouse ELISPOT Accessory Kit	DAKEWE	Cat#2230014
Chemicals, peptides, and recombinant proteins
BOIIB2	U.S. Patent Office	Patent No.7858089 B2
RHD5	U.S. Patent Office	Patent No.7829085 B2
Recombinant Human Coagulation Factor VIII (sinocelltech)	SinoCellTech	Drug Approval Number S20190004
Experimental models: Cell lines
HEK 293T	ATCC	N/A
K-562 cell line	ATCC	N/A
Experimental models: Organisms/strains
F8KO	Shanghai Modal Organisms	N/A
C57BL/6JGpt-Il2rgem1Cd390in1/Gpt	Jiangsu GemPharmatech Co., Ltd.	N/A
Software and algorithms
Graphpad Prism 8	Graphpad Prism 8	N/A
Flow jo version 10.8.1.	Flow jo version 10.8.1.	N/A
Other
6-well cell culture plates	NEST	Cat#714011
12-well cell culture plates	NEST	Cat#712002
96-well cell culture plates	NEST	Cat#713011
Multiscreen® 96 well Plate, hydrophilic PVDF membrane	MERCK	Cat#MSHVS4510

Experimental model and study participant details

Ethics declarations

Ethics approval and consent to participate. All animal experiments and human sample studies were reviewed and approved by the Ethics Committee of Sun Yat-sen University (SYSU-IACUC-2024-B1075) and the Ethics Committee of the Eighth Affiliated Hospital of Sun Yat-sen University. All procedures involving animals were conducted in accordance with institutional guidelines and national regulations for animal care. Human peripheral blood samples were obtained with informed consent.

Method details

Plasmid construction and cell culture

The human FVIII gene plasmid was purchased from Miaoling Bioscience. The second-generation CAR backbone was based on sequences previously established in our laboratory. The scFv sequences of anti-A2 (BOIIB2, Patent No.7858089 B2) and anti-C1 (RHD5, Patent No.7829085 B2) antibodies were obtained from published literature. The FVIII-CAAR fragment was cloned into the PHR-GFP vector, and the scFv fragments of BOIIB2 and RHD5 were constructed into the PHR-mCherry vector.

The K562 cell line was obtained from Suya Biotechnology and cultured in RPMI-1640 medium supplemented with 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin at 37°C with 5% CO₂. Human T cells were isolated from peripheral blood mononuclear cells (PBMCs) using a T cell isolation kit (Biolegend, 480133), followed by culture in STEMCELL ImmunoCult-XF T cell Expansion Medium (STEMCELL, 10981). T cells were activated with anti-human CD3/CD28 antibodies (Gibco, 11161D) for 48 h. Mouse T cells were isolated from splenocytes following the manufacturer’s protocol and activated in 12-well plates pre-coated with anti-mouse CD3/CD28 antibodies for 48 h.

CAAR T cell generation

Lentiviral particles were generated by co-transfecting 293T cells with pMD2.G, psPAX2, and the target plasmid at a ratio of 3:9:9. Viral supernatants were collected and concentrated using PEG8000. The concentrated virus was used to transduce pre-activated human or mouse T cells. After 48–72 h, transduction efficiency was assessed by flow cytometry based on GFP or mCherry expression.

In vitro cytotoxicity assay (LDH release)

CAAR-T cells and target cells were co-cultured in 96-well plates at various effector-to-target ratios for 18 h. Following incubation, the cytotoxicity was evaluated using a commercial LDH release assay kit (DOJINDO, CK12) according to the manufacturer’s protocol. Absorbance was measured at 490 nm.

IFN-γ secretion assay

To assess T cell activation, CAAR-T cells were co-cultured with target cells at E:T = 3:1 ratios for 12–24 h. Supernatants were collected, and interferon-gamma (IFN-γ) levels were quantified using ELISA according to the manufacturer’s instructions (Elabscience, E-EL-H0108).

Animal models and in vivo CAAR T therapy

Hemophilia A mouse model (F8KO)

FVIII gene knockout (F8KO) mice were obtained from Shanghai Modal Organisms. Female mice aged 6–8 weeks were subcutaneously immunized weekly with 10 IU of recombinant human FVIII protein (SinoCellTech, SCT800) mixed with aluminum adjuvant in a total volume of 200 μL. Blood was collected via retro-orbital bleeding 7 Days post-immunization. Plasma was separated by centrifugation and inhibitor levels were assessed via ELISA. After three rounds of immunization, mice received tail vein injections of either CAAR-T cells or untransduced control T cells.

Tumor xenograft model

NCG mice (6-8-week-old females) were purchased from Jiangsu GemPharmatech Co., Ltd. Mice were subcutaneously injected with 1 × 10⁶ target cells. On day 3 and day7 post-inoculation, 1 × 10⁷ CAAR-T cells were administered via tail vein injection. Tumor size was measured every 4 days using calipers, and tumor volume was calculated.

Immunohistochemistry

At the end of treatment, mice were sacrificed and major organs including heart, liver, spleen, and kidneys were harvested and fixed in 4% paraformaldehyde. Tissues were embedded in paraffin and sectioned. Slides underwent standard deparaffinization, antigen retrieval, and blocking. Primary antibodies were applied according to the manufacturer’s protocols, followed by DAB staining and hematoxylin counterstaining. Tissue sections were visualized under a light microscope to assess inflammation and cellular infiltration.

ELISA for FVIII inhibitor detection

96-well ELISA plates were coated overnight at 4°C with 2 IU of recombinant human FVIII protein per well in coating buffer. Mouse plasma was obtained by retro-orbital bleeding, incubated at 4°C for 1 h, centrifuged, and diluted for use as the primary antibody. HRP-conjugated secondary antibodies were used for detection. After color development and stopping the reaction, OD values were measured at 450 nm using a microplate reader.

Flow cytometry analysis

CD107α degranulation assay

To assess the cytotoxic function of CAAR-T cells, they were co-cultured with target cells in flat-bottom 96-well plates at an E:T ratio of 3:1. Anti-CD107α fluorescent antibody (Biolegend, 328619) was added at the beginning of co-culture. After 1 h, 5 μL of GolgiStop (BD, 555029) was added and incubation continued for another 3–4 h. Cells were harvested, washed, and analyzed via flow cytometry for degranulation marker expression.

In vivo detection of CAAR-T cell persistence

At study endpoints, spleens and bone marrow were collected. Spleens were mechanically dissociated and filtered through a cell strainer. Bone marrow cells were flushed with PBS and filtered similarly. Red blood cells were lysed using RBC lysis buffer. Cell suspensions were stained in FACS buffer.

Donor-derived CAAR-T cells were identified using specific markers

In F8KO mice, donor T cells were distinguished using anti-CD45.1 antibody (Biolegend, 110713) (on a CD45.2 host background); In NCG mice, human CAAR-T cells were identified using anti-human CD3 antibody (Biolegend, 980004).

Samples were analyzed using flow cytometry and data were processed using FlowJo software to assess cell persistence and distribution.

Quantification and statistical analysis

Statistical analysis and graphing were performed using GraphPad Prism 8.0 software. Continuous variables are presented as mean ± standard deviation (Mean ± SD). For the comparison of differences between two groups, data with a normal distribution were analyzed using a t test, and data that did not meet the normality assumption were analyzed using the Mann-Whitney U test. For multiple group comparisons, data with a normal distribution were analyzed using one-way analysis of variance (ANOVA), and data that did not meet the normality assumption were analyzed using the Kruskal-Wallis test. A p-value <0.05 was considered statistically significant, with NS indicating no significant difference; ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001. All tests were two-sided.

Published: February 18, 2026