Treatment of allergic eosinophilic asthma through engineered IL-5-anchored chimeric antigen receptor T cells

Severe eosinophilic asthma (SEA) is a therapy-resistant respiratory condition with poor clinical control. Treatment efficacy and patient compliance of current therapies remain unsatisfactory. Here, inspired by the remarkable success of chimeric antigen receptor-based cellular adoptive immunotherapies demonstrated for the treatment of a variety of malignant tumors, we engineered a cytokine-anchored chimeric antigen receptor T (CCAR-T) cell system using a chimeric IL-5-CD28-CD3ζ receptor to trigger T-cell-mediated killing of eosinophils that are elevated during severe asthma attacks. IL-5-anchored CCAR-T cells exhibited selective and effective killing capacity in vitro and restricted eosinophil differentiation with apparent protection against allergic airway inflammation in two mouse models of asthma. Notably, a single dose of IL-5-anchored CCAR-T cells resulted in persistent protection against asthma-related conditions over three months, significantly exceeding the typical therapeutic window of current mAb-based treatments in the clinics. This study presents a cell-based treatment strategy for SEA and could set the stage for a new era of precision therapies against a variety of intractable allergic diseases in the future.


Introduction
Over 339 million people suffer from asthma worldwide 1 and patients with severe eosinophilic asthma (SEA) are at high risk of mortality and low quality of life [2][3][4] . SEA, characterized by eosinophilic inflammation, is a major phenotype of refractory asthma with poor clinical control 4 . Eosinophils have a prominent role in SEA pathogenesis, causing airway epithelial damage and bronchial remodeling 5 . Eosinophilia is closely related to higher exacerbation frequency and worse control, leading to decreased lung function 6,7 . Thus, strategies capable of inactivating or depleting eosinophils offer attractive therapies for SEA 8,9 . Currently, asthma symptoms control mainly relies on the daily administration of the inhaled corticosteroids (ICS) combined with β2 receptor agonists 10 , which can cause intolerable adverse reactions, including osteoporosis and hypertension 11 . For SEA patients, biological agents targeting the interleukin-5 (IL-5)/IL-5 receptor α (IL-5Rα) axis interfere with the pathologic functions of eosinophils and show promising therapeutic effects 12 . The human IL-5Rα, showing specific binding for IL-5, is expressed on mature eosinophils, basophils, and their progenitors 13,14 . When exposed to allergens, epithelialderived cytokine IL-33 upregulates the IL-5Rα expression on eosinophil progenitors 15 . Then the concomitant expression of IL-5 and IL-5Rα mediates the growth and terminal differentiation of eosinophil progenitors, which contributes to the subsequent development of blood and tissue eosinophilia in eosinophilic asthma patients with type 2-high inflammation 16,17 . Anti-IL-5 monoclonal antibodies, such as mepolizumab and reslizumab, have been approved for second-line treatment of SEA in the clinics by decreasing the level of eosinophils in the blood/sputum through neutralization of IL-5, and inhibition of eosinophil differentiation and activation to eventually reduce the exacerbation frequency in asthma patients [18][19][20] . Yet, their effects on the reduction of airway eosinophils are restricted 21,22 . Benralizumab, an mAb against the IL-5Rα, can directly eliminate eosinophils through antibody-dependent cell-mediated cytotoxicity 13 . Although benralizumab showed better efficacy than mepolizumab or reslizumab in improving pulmonary functions of SEA patients 23 , population-level response rates remain low and the reduction of annual exacerbation rates remains limited 24,25 . Moreover, poor bioavailability of mAb drugs results in a need for repeated administration over long timespans, which severely compromises patient compliance.
Recently, cellular adoptive immunotherapies based on CD19-specific CAR-T cells have shown remarkable efficacy in treating B cell malignancies [26][27][28] . The chimeric antigen receptor (CAR) is a fusion protein composed of an extracellular target-specific scFv-moiety and an intracellular T-cell receptor domain typically consisting of CD28 and CD3ζ, allowing for antigen-specific activation of T-cell killing in a strict scFv-dependent manner 29,30 . Using this empiric design principle, similar CAR-T strategies have been extensively tested for the treatment of other cancers such as neuroblastoma 31,32 , hepatocellular carcinoma [33][34][35] , as well as other diseases in the fields of autoimmune diseases [36][37][38] , cardiovascular diseases 39 and senescence-associated pathologies 40 . However, neither of the latter approaches could repeat or emulate the treatment efficacy of the inaugural anti-CD19 system, which forms the basis of all clinically approved CAR-T products that have hitherto arrived onto the market 30,[41][42][43] . In fact, a typical adverse effect of classical CAR-T therapies is the development of severe clinical anaphylaxis 44,45 . A few reports indicate that xenogenetically-derived scFv domains could largely contribute to such immunogenicity-related reactions 43,[46][47][48] , which restricts the application of CAR-T cell therapy in allergic patients.
In this work, we developed an scFv-independent, cytokine-anchored chimeric antigen receptor (CCAR) configuration that uses IL-5 as the extracellular targetbinding domain. When engineered into primary T-cells, IL-5-anchored CCAR-T cells specifically target IL-5Rαexpressing eosinophils and eosinophil progenitor cells 49,50 . IL-5-anchored CCAR-T cells effectively restricted eosinophil differentiation with obvious protection against allergic airway inflammation in murine asthma models.
To characterize the binding affinity of hIL-5-anchored CCAR and anti-hIL-5Rα CAR to the hIL-5Rα target, we used the adhesion frequency assay as previously reported 53,54 (Fig. 1l). Statistical analysis revealed no significant differences between these two CAR-T variants in terms of in-situ binding kinetics (Fig. 1m), i.e., effective affinity (A c K a ) (Fig. 1n) and effective on-rate (A c K on ) (Fig. 1o). However, the off-rate (K off ) of the hIL-5-anchored CCAR was slightly lower than that of the anti-hIL-5Rα CAR (Fig. 1p).
We then transferred anti-mIL-5Rα CAR-T cells into the allergic airway inflammation model ( Supplementary Fig.  S4b). Anti-mIL-5Rα CAR-T cells, which showed specific elimination of mIL-5Rα + cells in vitro ( Supplementary  Fig. S4a), could not significantly reduce eosinophil levels in vivo ( Supplementary Fig. S4c, d). Meanwhile, we compared the mCCL11-anchored, mCCL24-anchored, and mIL-5-anchored CCAR-T cells in vivo (Supplementary Fig. S5a). Interestingly, neither the mCCL11-anchored nor the mCCL24-anchored CCAR-T cells reduced eosinophil levels, while mIL-5-anchored CCAR-T cells showed promising results ( Supplementary Fig. S5b, c). Thus, we selected the IL-5-anchored CCAR-T strategy for further studies. Fig. 1 Design, characterization, and comparison of eosinophil-targeting CCAR-T cells. a Design of the anti-hIL-5Rα CAR-T cells using the scFv derived from human IL-5Rα mAb as the antigen-binding domain. b Cytotoxic activity of anti-hIL-5Rα CAR-T cells as determined by a bioluminescence assay using luciferase-expressing hIL-5Rα + or hIL-5Rα -U2OS cells as target cells. Differences between the CAR-T cell-treated and control group were examined by the two-way ANOVA, **** P < 0.0001. c Design of the hIL-5-anchored CCAR-T cells using human IL-5 as the antigen-binding domain. d Cytotoxic activity of hIL-5-anchored CCAR-T cells. Two-way ANOVA, **** P < 0.0001. e Design of the hCCL11-anchored CCAR-T cells. f Cytotoxic activity of the CCL11-anchored CCAR-T cells against target cells. Two-way ANOVA, **** P < 0.0001. g Design of the hCCL24-anchored CCAR-T cells. h Cytotoxic activity of the CCL24-anchored CCAR-T cells to target cells. Two-way ANOVA, **** P < 0.0001. i Cytotoxic activity of anti-hIL-5Rα CAR-T cells or hIL-5-anchored CCAR-T cells from healthy human donors against hIL-5Rα + U2OS cells. Two-way ANOVA, **** P < 0.0001. UTD-T, un-transduced T cells. j, k The production of interferon-γ (IFN-γ) in the supernatant of anti-hIL-5Rα CAR-T cells (j) or hIL-5-anchored CCAR-T cells (k) from healthy human donors after coculture with target cells for 24 h was determined by ELISA kit. Two-way ANOVA, **** P < 0.0001. l Schematic diagram of the biomembrane force probe assay for the adhesion frequency assay. m Binding specificities (P a ) of anti-hIL-5Rα CAR/hIL-5-anchored CCAR and hIL-5Rα as measured by adhesion frequency with different contact duration. n In-situ effective affinity (A c k a ) of the interaction between anti-hIL-5Rα CAR-hIL-5Rα or hIL-5-anchored CCAR-hIL-5Rα and hIL-5Rα. N.D., not detected. o In-situ effective on-rate (A c k on ) of the anti-hIL-5Rα CAR-hIL-5Rα bond and the hIL-5-anchored CCAR-hIL-5Rα bond. p Average off-rate (k off ) of the anti-hIL-5Rα CAR-hIL-5Rα bond and the hIL-5-anchored CCAR-hIL-5Rα bond. Twotailed t-test, * P < 0.05.

Functional assessment of hIL-5-anchored CCAR-cells in NSG mice
Next, we tested the function of hIL-5-anchored CCAR-T cells in vivo. First, we detected the CD69 expression on Jurkat cells to assess the CCAR-induced T cell activation. The hIL-5-anchored CCAR-Jurkat cells, which can be activated by target cell in vitro ( Supplementary Fig. S6), were significantly activated by hIL-5Rα + target cells as well after intraperitoneal injection of both hIL-5Rα + U2OS cells and hIL-5anchored CCAR-Jurkat cells in NOD/ShiLtJGpt-Prkdc em26Cd52 Il2rg em26Cd22 /Gpt (NCG) mice (Fig. 2a). Then, we tested the in vivo efficacy of hIL-5-anchored CCAR-T cells against target cells through bioluminescence imaging. Compared to UTD-T cells, both murine primary hIL-5-anchored CCAR-T cells (Fig. 2b, c) and human primary CCAR-T cells (Fig. 2d, e) dramatically eliminated hIL-5Rα + U2OS cells.

Targeting specificity of IL-5-anchored CCAR-T cells
To assess the function of IL-5-anchored CCAR-T cells in the murine asthma models, we designed murine variants based on mIL-5-anchored CCAR-T cells or CCAR-Jurkat cells specific for murine IL-5Rα (mIL-5Rα). The mIL-5-anchored CCAR-Jurkat cells were activated by target cells expressing mIL-5Rα after 24 h of coculture (Fig. 3a), whereas hIL-5-anchored CCAR-Jurkat cells showed no response to mIL-5Rα + target cells, confirming target specificity of the CCAR (Fig. 3b). Next, we transduced primary murine T cells with the mIL-5-anchored CCAR retrovirus (mIL-5-anchored CCAR-T cells) for cytotoxic activity assay. Similarly, mIL-5-anchored CCAR-T cells showed significant cytotoxicity against mIL-5Rα + target cells (Fig. 3c). Further, we performed a cell apoptosis assay using mouse primary mIL-5Rα + and Flow cytometry analysis of CD69 expression on mIL-5-anchored CCAR-Jurkat cells after coculture with target cells (U2OS cells) for 24 h. One-way ANOVA, **** P < 0.0001. b Jurkat cells were transduced with hIL-5-CCAR comprising a human IL-5 linked to human CD28 costimulatory and CD3ζ signaling domains (h.IL-5-h.28z). Flow cytometry analysis of CD69 expression on hIL-5-anchored CCAR-Jurkat cells after coculture with target cells for 24 h. Dunn's Kruskal-Wallis test. c Primary T cells from BALB/c mice were transduced with mIL-5-CCAR comprising a mouse IL-5 linked to mouse CD28 costimulatory and CD3ζ signaling domains (m.IL-5-m.28z). Cytotoxic activity of mIL-5-anchored CCAR-T cells against mIL-5Rα + U2OS cells. Two-way ANOVA, **** P < 0.0001, * P < 0.05. d Flow cytometry analysis showing the proportion of primary mIL-5Rα + cells derived from bone marrow after treating with mIL-5-anchored CCAR-T cells or UTD-T cells at a CCAR-T to target ratio of 6:1 in vitro for 8 h. UTD-T, un-transduced T cells. Two-tailed t-test, **** P < 0.0001. e Normalized cell death of primary mIL-5Rα + cells or mIL-5Rαcells after treating with mIL-5-anchored CCAR-T cells or UTD-T cells. Differences between mIL-5-anchored CCAR-T cells-treated and UTD-T cells-treated group were examined by two-way ANOVA, **** P < 0.0001. f Timeline of Eos differentiation induction, mIL-5-anchored CCAR-T cell administration, and flow cytometry analysis of Eos. The mIL-5-anchored CCAR-T cells were administrated at a CCAR-T/Target ratio of 3:1. Eos, eosinophil. g Flow cytometry plots showing the proportion of BM-derived Eos after treating with mIL-5-anchored CCAR-T cells or not. BM, bone marrow. Two-tailed t-test, *** P < 0.001. h Histogram of the cell count of BM-derived Eos. Brown-Forsythe and Welch ANOVA, * P < 0.05. mIL-5Rα − cells as target cells. In contrast to the UTD-T cells, mIL-5-anchored CCAR-T cells remarkably reduced the proportion of mIL-5Rα + cells (Fig. 3d) and displayed specific cytolysis (Fig. 3e). As the differentiation of eosinophils plays a crucial role in airway eosinophilia during SEA 16 , mIL-5-anchored CCAR-T cells were also applied on a bone marrow-derived eosinophil (BMDE) differentiation assay (Fig. 3f). We observed that administration of mIL-5-anchored CCAR-T cells could block eosinophil differentiation both on the cell proportion (Fig. 3g) and cell count level (Fig. 3h). Together, these results suggest that the mIL-5-anchored CCAR-T cells are capable of selectively and effectively eliminating eosinophils.

Protective effect of CCAR-T cells against allergic eosinophilic inflammation
Next, we assessed the effect of CCAR-T cells in allergic airway inflammation mouse models [55][56][57] . In an acute asthmatic inflammation model, 3 × 10 6 mIL-5-anchored CCAR-T cells were intravenously injected into recipient mice one week before administration of the extract of house dust mite (HDM) (Fig. 4a), and eosinophils were analyzed by flow cytometry using cell surface staining ( Supplementary Fig. S7). Indeed, administration of mIL-5anchored CCAR-T cells strikingly reduced both the proportion and the absolute number of eosinophils in bronchoalveolar lavage fluid (BALF) that are typically elevated during HDM-stimulated conditions (Fig. 4b-d).
In addition, CCAR-T cells brought a significant decrease in eosinophil levels in lung tissue of HDM-treated mice (Fig. 4e), as well as in the peripheral blood (Fig. 4f, g) and bone marrow (Fig. 4h). Collectively, these data demonstrate that the mIL-5-anchored CCAR-T cells can efficiently target and eliminate eosinophils in the HDMinduced allergic airway inflammation model. We next investigated the impact of mIL-5-anchored CCAR-T cells on the level of airway inflammation. IL-5 is a type 2 cytokine that promotes differentiation and activation of eosinophils and is, therefore, an essential biomarker for asthma 58,59 . We discovered that mIL-5-anchored CCAR-T cells resulted in a significant decrease in IL-5 level in BALF (Fig. 4i). Administration of CCAR-T cells also reduced the total number of inflammatory cells in BALF (Fig. 4j), indicating the remission of inflammatory infiltration in the airway. We further assessed the inflammation level in hematoxylin-eosin (H&E) stained pulmonary sections by semi-quantification. The notable differences in inflammatory scores following CCAR-T cells administration confirmed that mIL-5-anchored CCAR-T cells protect against airway inflammation (Fig. 4k, l).

Long-term efficacy of CCAR-T cells
To evaluate the duration of efficacy of the current IL-5-anchored CCAR-T strategy, we tested the effect of mIL-5-anchored CCAR-T cells in the ovalbumin (OVA)containing aerosols inhalation induced airway inflammation model. In the one-month model (Fig. 5a), we observed mIL-5-anchored CCAR-T-dependent reduction in the eosinophil levels in BALF (Fig. 5b, c), lung tissue ( Supplementary Fig.  S8a), and peripheral blood ( Supplementary Fig. S8b), and decreased IL-5 levels in BALF (Fig. 5d) as well as alleviated inflammation scores in lungs (Fig. 5e, f).
Moreover, the IL-5-anchored CCAR-T cells maintained effective control of asthma-related conditions for up to three months (Fig. 5g-l; Supplementary Fig. S9), including the eosinophil levels in BALF (Fig. 5h, i), peripheral blood ( Supplementary Fig. S9a) and bone marrow (Supplementary Fig. S9b), IL-5 levels in BALF (Fig. 5j) as well as the inflammatory infiltration in the airway (Fig. 5k, l). These findings imply that the CCAR-T concept might set a new standard for long-term inflammation protection for asthmatic patients.
Additionally, no differences in the proportion of Th1 cells, Th2 cells, or Treg cells could be observed following CCAR-T transfer, excluding putative effects of CCAR-T cells on endogenous T cell responses (Supplementary Fig. S10). Also, no significant impact on systemic inflammatory biomarkers, serum IL-6 and IFN-γ, was observed ( Supplementary Fig. S11).

Discussion
Eosinophilic inflammation plays a prominent role in SEA. According to reports 60-62 , eosinophil depletion does not increase the risk of helminth infection or affect the vaccine responses. Although current management or addon therapies for SEA can control the symptoms to a certain extent, inflammation relief is short-lasting, and asthma exacerbation continues. To address these concerns, we developed a cellular adoptive immunotherapy using design principles adopted from chimeric antigen receptor-T cells 29,30 and applied it to allergic asthma. To avoid anaphylaxis reactions typically observed in scFvdependent CAR-T therapies for cancer, we employed a ligand-anchored CAR design that allows cytokines to trigger target-specific T-cell killing. From a cell engineering perspective, the design of scFv-independent CARs is also highly advantageous in terms of time-and resource-efficiency, as the costly large-scale screening for antibody moieties can be omitted.
In this study, we have engineered the IL-5-anchored CCAR-T cells and verified their killing capacity in vitro and in mice. We showed that the IL-5-anchored CCAR-T cells exhibited efficacious and persistent control of eosinophilic asthma conditions in both the HDM and OVAstimulated acute inflammatory asthma models. Furthermore, IL-5-anchored CCAR-T cells maintained a constant effect on eosinophil reduction, IL-5 reduction, and prevention of airway inflammation over three months, exceeding the typical active therapeutic window of single mAb-injections of 4 weeks 11 .
During traditional CAR-T treatment, the monoclonal antibodies-derived scFvs could induce immune responses due to their high immunogenicity 43 . In this case, anti-mIL-5Rα CAR carrying scFvs might elicit anti-CAR responses, especially in the hypersensitive immune environment of allergen-induced asthma models involved in this study, which might contribute to the treatment failure of anti-mIL-5Rα CAR-T cells in vivo.
When exposed to allergens, eosinophil progenitors rapidly differentiate into a large number of mature g Quantification of Eos proportion in PB. One-way ANOVA corrected with the Tukey method, * P < 0.05, ** P < 0.01. h Histogram of Eos proportion in BM. BM, bone marrow. Kruskal-Wallis test, * P < 0.05, ** P < 0.01. i The concentration of IL-5 in BALF was determined by CBA kit. CBA, Cytometric Bead Array. Two-tailed Mann-Whitney test, ** P < 0.01. j Cell count of BALF total cells by microscope. One-way ANOVA corrected with the Tukey method, ** P < 0.01. k Representative images of the pulmonary sections stained with H&E. Scale bars, 100 μm. l Inflammation scores of the H&E-stained sections determined by semi-quantification. One-way ANOVA corrected with the Tukey method, * P < 0.05, *** P < 0.001. eosinophils. IL-5Rα is highly expressed on the surface of both eosinophil progenitors and mature eosinophils, while CCR3 is mainly expressed on mature eosinophils 63 . This might explain why CCAR-T cells targeting CCR3 failed to reduce eosinophils in vivo.
As the chronicity and the need for long-term or even life-long therapy are severe challenges during the treatment of eosinophilic diseases 64 , IL-5-anchored CCAR-T cells therapy is expected to solve these problems. Thus, the cytokine-anchored CCAR-T strategy not only showed unprecedented medical potential in SEA therapy but might also kick off a new era of cellbased precision medicine for the treatment of other eosinophilic diseases, such as chronic rhinosinusitis, P < 0.01 by two-tailed Welch's t-test. c Cell count of Eos in BALF in OVA-induced asthma model. * P < 0.05 by two-tailed Mann-Whitney test. d The concentration of IL-5 cytokine in BALF was determined by CBA. * P < 0.05, ** P < 0.01 by two-tailed Mann-Whitney test. e Representative images of the pulmonary sections stained with H&E. Scale bars, 100 μm. f Inflammation scores of the H&E-stained sections determined by semi-quantification. * P < 0.05 by two-tailed Mann-Whitney test. g Timeline of CCAR-T cell administration and sample analysis in the allergic airway inflammation model. h Flow cytometry analysis of Eos proportion in BALF. ** P < 0.01 by two-tailed Mann-Whitney test. i Cell count of Eos in BALF. * P < 0.05 by two-tailed Mann-Whitney test. j The secretion of IL-5 cytokine in BALF was determined by CBA. CBA, Cytometric Bead Array. Two-tailed Mann-Whitney test, * P < 0.05. k Representative images of the pulmonary sections stained with H&E. Scale bars, 100 μm. l Inflammation scores of the H&E-stained sections determined by semi-quantification. * P < 0.05, ** P < 0.01 by two-tailed Welch's t-test. eosinophilic esophagitis, and even chronic eosinophilic leukemia 64 .

Isolation, expansion, and genetic modification of human T cells
Peripheral blood was obtained from the healthy donors. Blood sampling was performed following the required ethical procedures. Lymphocytes were isolated by density gradient centrifugation following the manual of the Human Lymphocyte Separation Medium (DAKEWE). Human T cells were purified using the human CD3 T cell isolation kit (BioLegend), stimulated with CD3/CD28 T cell Activator Dynabeads (Gibco) and cultured at 10 6 /mL in X-VIVO 15 Serum-free Hematopoietic Cell Medium (Lonza), supplemented with 5 ng/mL human IL-7 (PeproTech) and 5 ng/mL human IL-15 (Pepro-Tech) 40

Cytotoxic activity assay
The cytotoxicity of CCAR-T cells, CAR-T cells, or UTD-T cells was determined by the luciferase-based assay as described previously 40,65 . In detail, 1 × 10 4 target cells, stably expressing firefly luciferase through retrovirus infection, were cocultured with killing cells at the indicated T/target ratios in white 96-well plates (Costar) for indicated incubation time. Target cells alone were plated at the same cell density for determining the maximal luciferase expression (relative light units, RLU). The culture medium was discarded carefully and 15 μg D-luciferin (GoldBio) in 100 μL PBS was added to each well after coculture. Emitted light was detected by the luminescence plate reader (SynergyMx M5, Molecular Devices) and was converted into lysis (%) according to the previous report 40 to characterize the cytotoxicity.

Adhesion frequency assay
The preparation of the red blood cells (RBCs) and the experimental procedure of adhesion frequency assay have been described in detail previously 53,54 . Briefly, for the preparation of the hIL-5Rα-coated RBCs, the human IL-5Rα (hIL-5Rα) extracellular domain linked with Avi-Tag was expressed, purified, and biotinylated. The biotinylated hIL-5Rα was linked to streptavidin-coated RBCs (SA-RBCs) to produce hIL-5Rα-coated RBCs which were then used for the adhesion frequency assay. For the adhesion frequency assay, it was used for measuring the in-situ binding kinetics of the hIL-5Rα and the anti-hIL-5Rα CAR/CCAR. In brief, this assay utilized micromanipulation to precisely operate the contact and retraction between the hIL-5Rα-coated RBCs and the anti-hIL-5Rα CAR/CCAR Jurkat cells.
The binding frequency P a was acquired with definite contact area A c and a series of preset contact time t c through 50 contact-retraction cycles. And the in-situ effective binding affinity A c K a and the off-rate k off were then calculated by the probabilistic kinetic model: Where m r and m hIL-5Rα are respective CAR/CCAR and hIL-5Rα molecular densities, which are determined by standard calibration beads on flow cytometry. In-situ effective on-rate A c k on was then calculated by:

Xenograft model in NCG mice
For the in vivo CCAR-induced Jurkat cell activation assay, the NCG mice were injected intraperitoneally with 1 × 10 7 hIL-5Rα + U2OS cells and 1 × 10 7 hIL-5-anchored CCAR-Jurkat cells at 0 h. Mice were sacrificed at 3 h or 6 h, and the CCAR-Jurkat cells were harvested from intraperitoneal lavage fluids and analyzed by flow cytometry.
For the in vivo CCAR-induced T cell cytotoxicity assay, the NCG mice were injected intraperitoneally with 3 × 10 5 hIL-5Rα + U2OS cells expressing firefly luciferase and 1 × 10 6 murine primary hIL-5-anchored CCAR-T cells or human primary hIL-5-anchored CCAR-T cells at day 0. 24 h later, the bioluminescence imaging of the mice was performed 10 min after intraperitoneal injection of 100 μL D-luciferin (30 mg/mL, GoldBio) on an IVIS Spectrum imaging system (Caliper) and the average radiance of hIL-5Rα + U2OS cells was measured through the Living Image software (Caliper).
Mouse model of eosinophilic asthma OVA-induced asthma model BALB/c mice were sensitized with 200 μL of 80 μg OVA (Sigma-Aldrich) emulsified in the aluminum adjuvant (Thermo Scientific) through intraperitoneal injection on day 0 and day 14, and control mice were administered with 200 μL saline (NS). On days 25-27, sensitized mice were challenged with 1.5% OVA in saline through aerosol administration for 40 min every time by an ultrasonic atomizer (Devilbiss). 24 h after the final challenge, mice were sacrificed for analysis.

HDM-induced asthma model
BALB/c mice received HDM (100 μg, D. pteronyssinus) in 50 μL saline through airway drip on day 0, day 7, and day 14, as described previously 56,68 . Control mice received 50 μL saline (NS) in the same way. Then mice were sacrificed 72 h after the final airway drip for analysis.

Detection of the inflammatory factors
The concentration of IL-5 in BALF supernatants was measured by mouse IL-5 enhanced sensitivity cytometric bead array assay (Enhanced CBA, BD), serum IL-6 by mouse IL-6 Enhanced CBA (BD), serum IFN-γ by mouse IFN-γ Enhanced CBA (BD), serum IL-13 by mouse IL-13 Enhanced CBA (BD) and serum IL-4 by mouse IL-4 Enhanced CBA (BD) following the manufacturer's manual. The human IFN-γ was measured by the human IFN-γ ELISA kit (AbClonal).

Perivascular inflammation score
The pulmonary sections were embedded in paraffin and stained with hematoxylin-eosin (H&E) after fixation. The score of the perivascular inflammation was determined by the degree of inflammatory cell infiltration and was assessed as 0-3 on a subjective scale, as described previously 69,70 , with slight modification. Briefly, 0 means no or occasional inflammatory cells distributed in the perivascular space; 1 means 1 layer of inflammatory cells surrounded in the perivascular space; 2 for 2-5 layers of inflammatory cells; 3 for more than 5 layers of inflammatory cells.

Statistical analysis
Data are presented as mean ± SEM. Statistical analyses were performed using GraphPad Prism software 8.0.
Comparisons in each experiment were described in the figure legends. All representative data were replicated in at least three independent experiments.