Mesenchymal stromal cells with chimaeric antigen receptors for enhanced immunosuppression

Abstract

Allogeneic mesenchymal stromal cells (MSCs) are a safe treatment option for many disorders of the immune system. However, clinical trials using MSCs have shown inconsistent therapeutic efficacy, mostly owing to MSCs providing insufficient immunosuppression in target tissues. Here we show that antigen-specific immunosuppression can be enhanced by genetically modifying MSCs with chimaeric antigen receptors (CARs), as we show for E-cadherin-targeted CAR-MSCs for the treatment of graft-versus-host disease in mice. CAR-MSCs led to superior T-cell suppression and localization to E-cadherin⁺ colonic cells, ameliorating the animals' symptoms and survival rates. On antigen-specific stimulation, CAR-MSCs upregulated the expression of immunosuppressive genes and receptors for T-cell inhibition as well as the production of immunosuppressive cytokines while maintaining their stem cell phenotype and safety profile in the animal models. CAR-MSCs may represent a widely applicable therapeutic technology for enhancing immunosuppression.

Extended Data Fig. 1 |. EcCAR-MSC effects on CART effector functions in in vivo tumor model.

a, Comparison of Ecad⁻NALM6 and Ecad⁺ NALM6 tumor flux measurements across control mice without CART19 cell infusion. Displaying mean ± SEM with statistics by 2-way ANOVA (n = 4–5 mice per group). b, c, Relative levels of luc⁺ Ecad⁻NALM6 (left) and luc⁺ Ecad⁺ NALM6 (right) as measured by luminescence following 24-hour in vitro coculture with UTD-MSCs or EcCAR-MSCs at varying MSC:NALM6 ratios. Data displaying mean ± s.d. with statistics by 2-way ANOVA (n = 2 replicates per group). d, Schema of NALM6 and JeKo-1 tumor models: NSG mice were engrafted with luciferase⁺ CD19⁺ Nalm6 or JeKo-1 cells (1 × 10⁶ i.v.) and treated with CART19 (1 × 10⁶ cells i.v.) and irradiated Ecad⁺ cell line. Mice were then randomized to receive UTD-MSCs or EcCAR-MSCs (1 × 10⁶ cells i.p.) and monitored biweekly for BLI and survival. Image created with BioRender.com. e, Tumor flux following CART19 infusion in Jeko-1 model, comparing EcCAR-MSCs, UTD-MSCs, or no MSC treatment. Data showing mean ± SEM with statistical analysis by 2-way ANOVA (n = 2–5 mice per group) f, Tumor flux measurements across MSC administration groups following CART19 infusion in NALM6 model. Displaying mean ± SEM with statistics by 2-way ANOVA (n = 3–4 mice per group, 2 independent experiments). g, Survival outcomes of EcCAR-MSCs compared to UTD-MSCs and no MSC control groups. Statistics by Kaplan-Meier survival analysis (n = 3–4 mice per group, 2 independent experiments). For all panels, ns=p ≥ 0.05 and significant p values are displayed.

Extended Data Fig. 2 |. RNAseq pathway analysis and cytokine secretion.

a, Differentially expressed genes in unstimulated EcCAR-MSCs vs. UTD-MSCs, Ecad-stimulated vs. unstimulated EcCAR-MSCs, Ecad−stimulated vs. unstimulated UTD-MSCs, and Ecad-stimulated EcCAR-MSCs vs. UTD-MSCs. Data displaying significantly unregulated and downregulated gene counts within comparisons with adj. p value < 0.01 and ± 1-log fold change. Transcriptional alterations induced by Ecad stimulation of CAR-MSCs included 2362 significant genes vs. EcCAR-MSC alone and 3032 significant genes vs. Ecad stimulated UTD-MSCs. Transcriptional alterations induced by CAR transduction included 606 significant genes. Transcriptional alterations induced by Ecad stimulation included 206 significant genes. b, Ingenuity Pathway Analysis (IPA) revealed upregulated canonical pathways in unstimulated EcCAR-MSCs vs. UTD-MSCs. Dashed line across x axis represents statistically significant enrichment for all pathways -log(p ≤ 0.05). (n = 3 MSC donors per group). c, Graphical Summaries generated through IPA machine learning algorithm illustrating most significant entities activated in unstimulated EcCAR-MSCs vs. UTD-MSCs and d, stimulated EcCAR-MSCs vs UTD-MSCs. Canonical pathways and activated molecules were used to predict meaningful functional impacts between datasets. e, Additional serum cytokine elevations found in peripheral blood from EcCAR-MSC-treated tumor xenograft mice as compared to UTD-MSC and control. Cytokines include macrophage-derived chemokine (MDC), growth related alpha protein (GRO), granulocyte macrophage colony-stimulating factor (GM-CSF), monocyte chemotactic protein 3 (MCP-3), and FMS-related tyrosine kinase 3 ligand (Flt-3L) in pg/mL. Data showing mean ± s.d. with statistical analysis determined by multiple t tests (n = 4–6 mice per group). For all panels, significant p values are displayed.

Extended Data Fig. 3 |. EcCAR-MSC safety profiles within in vivo canine models.

a, Schema for CAR-MSC manufacturing and safety analysis in healthy canine models: EcCARs with cross reactivity to human, mouse, and canine Ecad were lentivirally transduced into human MSCs and expanded in vitro for subsequent i.p. injection into healthy canine subjects. Subgroups were monitored for hematological and organ toxicity for 28 days. Image created with BioRender.com. b, Complete blood count levels displayed as a determinant of hematopoietic safety following administration of EcCAR-MSCs. This includes white blood cells, monocytes, lymphocytes, neutrophils, and platelets with short term (3 day) and long term (28 day) blood level monitoring following in vivo EcCAR-MSC injection as compared to baseline. Displaying mean ± s.d. of blood composition with statistics by 1-way ANOVA (n = 3 subjects per experimental group). c, Total protein, BUN, creatinine, albumin, and alkaline phosphatase levels depicted for safety confirmation with short term (3 day) and long term (28 day) monitoring following in vivo EcCAR-MSC injection. Displaying mean ± s.d. of blood composition with statistics by 1-way ANOVA (n = 3 subjects per experimental group). d, Bodyweight changes in healthy canines following administration of EcCAR-MSCS as compared to control. No significant differences in body weight changes were found between groups. Data showing mean ± s.d. of % weight change from baseline with statistical analysis performed by ordinary 1-way ANOVA, (n = 3 subjects per experimental group) e, Data displaying transverse colonic tissue sections of canines through H&E staining 28 days following administration at 20x and 40x magnification following treatment with human EcCAR-MSCs (left) or control (right). For all panels, ns=p ≥ 0.05.

Fig. 1 |. MSCs are transduced to express CAR and maintain stem-like features following transduction and stimulation.

a, Data showing CAR expression following lentiviral transduction of MSCs with increasing concentrations of protamine sulfate enhancer (50 μg ml⁻¹ and 100 μg ml⁻¹). Representative data of MSC donors (n = 6). CAR expression shown by goat anti-mouse antibody linked to allophycocyanin (GAM-APC). b, Ridge plot showing anti-Ecad-CAR (turquoise) and anti-CD19-CAR control (blue) detected on MSCs by flow cytometry. Displaying mean ± standard deviation (s.d.) representative data of MSC donors (n = 6). CAR expression shown by protein L (κ-chain) antibody linked to phycoerythrin (PE). c, CAR expression on EcCAR-MSC compared with UTD control 2 days and 11 days following transduction and ex vivo expansion. Displaying mean ± s.d. results from MSC donors (n = 3). Statistics by two-way ANOVA (n = 3 replicates per donor). d, Ridge plot showing EcCAR expression on MSCs across different primary biological MSC donors. Matched biological donor UTD and CAR-MSCs labelled across colours (grey for donor 1, pink for donor 2, and turquoise for donor 3). Data representing ≥10 independent experiments on six MSC donors. e, Stem phenotype of EcCAR-MSC versus UTD-MSC assessed by CD105, CD90, CD73, CD34, CD45 and CD14 surface markers using flow cytometry. Displaying mean ± s.d. results from MSC donors (n = 4). Statistics by two-way ANOVA (n = 3 replicates per donor). f, Upregulated (left) and downregulated (right) phenotypic enrichments of EcCAR-MSCs versus UTD-MSCs by CellMarker Augmented gene set enrichment analysis. Significant genes selected by ≥1 or ≤−1 log(fold change) and ≤0.05 adjusted P values (P_adj) via differential expression analysis on three MSC donors. Dashed line across x axes represent statistically significant enrichment for all pathways −log(P ≤ 0.05). Not significant (NS) is P ≥ 0.05, and significant P values are displayed.

Fig. 2 |. EcCAR-MSCs demonstrate superior antigen-specific suppression of primary T cells and signalling in vitro.

a, Ridge plots displaying establishment of Ecad⁺ NALM6 (turquoise) with matched Ecad⁻ NALM6 (blue) to be used as Ecad cell-based stimulation in in vitro suppression assays and in vivo tumour models. b, Absolute numbers of CD3⁺ T cells following co-culture with EcCAR-MSCs or UTD-MSCs with increasing concentrations of soluble Ecad (0 ng ml⁻¹, 250 ng ml⁻¹ and 1,000 ng ml⁻¹). Displaying mean ± s.d. results from MSC donors (n = 3). Statistics by two-way ANOVA (n = 3 replicates per donor). c, Stem phenotype of EcCAR-MSC with increasing (0–1,000 ng ml⁻¹) soluble Ecad stimulation by surface markers using flow cytometry. Displaying mean ± s.d. results from MSC donors (n = 6). Statistics by 2-way ANOVA. d–f, Absolute numbers of CD3⁺ T cells following co-culture with EcCAR-MSCs or UTD-MSCs derived from three biological MSC donors with or without Ecad⁺ or Ecad⁻ cell-based stimulation. Displaying mean ± s.d. across MSC donors (n = 3). Statistics by two-way ANOVA (n = 2 or 3 replicates per donor). g, Stem phenotype of EcCAR-MSCs or UTD-MSCs following co-culture alone or with Ecad⁺ or Ecad⁻ cell-based stimulation by surface markers using flow cytometry. Displaying mean ± s.d. results from MSC donors (n = 3). Statistics by two-way ANOVA. h, Western blot displaying phosphorylation (pTyr783) of downstream CAR signalling effector PLCγ in EcCAR-MSCs stimulated with Ecad⁺ K562, Ecad⁻ K562 or alone in culture at a 2:1 K562 to MSC ratio for 15 min. Fold change of normalized band intensities are mean ± s.e.m. with statistics by paired t-test (n = 3 MSC donors). Uncropped images can be found in Source Data. NS is P ≥ 0.05, and significant P values are displayed.

Fig. 3 |. EcCAR-MSCs induce superior immunosuppression in tumour and GvHD xenograft models.

a, Schema of Ecad⁺ tumour xenograft model: NSG mice were engrafted with luciferase (luc)⁺ Ecad⁺ or luc⁺ Ecad⁻ NALM6 cells (1 × 10⁶ cells i.v.) and treated with CART19 (1 × 10⁶ cells i.v.). Mice were then randomized to receive UTD-MSCs or EcCAR-MSCs (1 × 10⁶ cells i.p.) on days 0 and 24 and monitored for tumour burden by BLI. Image created with BioRender.com. IVIS, in vivo imaging system. b, Ecad⁺ (positive) NALM6 tumour flux measurements in photons per second (p s⁻¹) across mice receiving UTD-MSC or EcCAR-MSC doses (arrows) with CART19 infusion. Displaying median ± s.e.m. with statistics by two-way ANOVA (n = 4 or 5 mice per group, showing one representative of two independent experiments). c, Ecad⁻(negative) NALM6 tumour flux measurements in p s⁻¹ across mice receiving UTD-MSC or EcCAR-MSC doses (arrows) with CART19 infusion. Displaying median ± s.e.m. with statistics by two-way ANOVA (n = 4 or 5 mice per group, showing one representative of two independent experiments). d, Schema of human PBMC-induced GvHD xenograft model: human PBMCs were injected (25–30 × 10⁶ cells i.v.) into NSG mice along with EcCAR-MSCs or UTD-MSCs (1 × 10⁶ cells i.p.). MSCs were administered as treatment on days 0 and 21 of the experiment (arrows). Mice were monitored for weight loss, the development of clinical GvHD symptoms (scored on the basis of weight loss, diarrhoea, posture, activity, fur texture and skin integrity) and survival. Image created with BioRender.com. e, Graph depicting percent weight change from baseline in GvHD xenografts following treatment with UTD-MSCs, EcCAR-MSCs or no treatment doses (arrows). Displaying mean ± s.e.m. with statistics by two-way ANOVA (n = 5–10 mice per group, showing one representative of three independent experiments). f, Left: GvHD clinical scoring severity results following treatment with EcCAR-MSCs, UTD-MSCs or no MSCs. Right: Representative mouse pictures. Displaying mean ± s.e.m. with statistics by two-way ANOVA (n = 4 or 5 mice per group, showing one representative of three independent experiments). g, Survival outcomes following treatment with EcCAR-MSCs compared with UTD-MSCs and no MSC control. Statistics by Kaplan–Meier survival analysis (n = 5–10 mice per group, showing one representative of three independent experiments). NS is P ≥ 0.05, and significant P values are displayed.

Fig. 4 |. EcCAR-MSCs display antigen-specific activation and trafficking to Ecad⁺ colonic target tissue in acute GvHD xenograft models.

a, Schema of acute GvHD xenograft model to test luciferase (luc)⁺ GFP⁺ CAR-MSC scFv specificity and efficacy: NSG mice were first irradiated at 250 cGy to further prime an inflammatory environment. Human PBMCs were next injected (10–15 × 10⁶ cells i.v.) into mice along with luc⁺ GFP⁺ Ecad⁻ CAR-MSCs or luc⁺ GFP⁺ CD19-CAR-MSCs (1 × 10⁶ cells i.p.). Mice were monitored for long-term weight loss, the development of clinical GvHD and survival. Satellite mice were isolated 1 week following MSC injection for assessment of MSC localization in colonic target organs by bioluminescent imaging and immunofluorescent staining. Image created with BioRender.com. b, Graph depicting percent weight change from baseline in acute GvHD xenografts following treatment with Ecad-CAR-MSCs, CD19-CAR-MSCs or no treatment. Displaying mean ± s.e.m. with statistics by two-way ANOVA (n = 4 or 5 mice per group, three independent experiments). c, Survival outcomes following treatment with Ecad-CAR-MSCs compared with CD19-CAR-MSCs and no MSC control. Statistics by Kaplan–Meier survival analysis (n = 4 or 5 mice per group). d, Mouse peripheral blood assessment comparing absolute number of human CD3⁺ T cells between Ecad-CAR-MSC and CD19-CAR-MSC treatment groups by flow cytometry 2 weeks following MSC injection. Displaying mean ± s.d. with statistics by ordinary one-way ANOVA (n = 7–10 mice per group, showing two combined independent experiments). e, Left: Representative bioluminescent imaging across luc⁺ CD19-CAR-MSC and Ecad-CAR-MSC treated mice Right: Percent MSC flux to colon relative to total flux detected across all organs. Data showing mean ± s.d. with statistics by unpaired t-test (n = 4 or 5 mice per group, showing one representative of two independent experiments). f, Immunofluorescent-based quantification of EcCAR-MSC versus CD19-CAR-MSC localization to Ecad⁺ colon tissue. Data determined by the percent of MSC⁺ colonic crypts per focal image. Displaying mean ± s.d. with statistics by ordinary one-way ANOVA (n = 3 mice per group). g, Immunofluorescent imaging of mouse colonic tissue isolated from acute GvHD xenograft models 7 days following GFP⁺ MSC administration. Comparisons between CD19-CAR-MSC and Ecad-CAR-MSC (green) colocalization with Ecad⁺ (red) colonic regions. Cell nuclei were stained with DAPI (blue) with each colour channel displayed to make merged image. Images obtained at 40× magnification capturing 1× crop area with 20 μm scale bars. NS is P ≥ 0.05, and significant P values are displayed.

Fig. 5 |. Activation of antigen-specific immunosuppressive signalling pathways identified in EcCAR-MSCs.

a, Heat map depicting the distinct gene expression profile of stimulated EcCAR-MSC+Ecad samples highlighting the functional impact of CAR antigen-specific stimulation through hierarchical clustering with unstimulated EcCAR-MSC, stimulated UTD-MSC+Ecad and unstimulated UTD-MSC samples. Colour bar shows gene fold counts normalized across all samples (P_adj ≤ 0.05) (n = 3 MSC donors per condition). b, PCA depicting gene expression profiles across six samples clustering by derived MSC donor in stimulated versus unstimulated UTD-MSCs. c, PCA of stimulated versus unstimulated EcCAR-MSCs reveals unique clustering by stimulation group. PC, principal component. d, Graphical summary by IPA machine learning algorithm illustrating most significant entities activated in Ecad-stimulated versus unstimulated EcCAR-MSCs. Factors including upstream regulators, canonical pathways and biological functions were combined to predict meaningful functional impacts. Analysis revealed the apoptosis of leukocytes (centre) to be the most significantly enriched (P_adj ≤ 0.0001) functional pathway directly associated with all 17 activated molecules predicted in the dataset, including CD28 CAR signalling molecule (left). e, IPA reveals upregulated canonical pathways in stimulated versus unstimulated EcCAR-MSCs. Dashed line across x axis represents statistically significant enrichment for all pathways −log(P ≤ 0.05) (n = 3 MSC donors per group). f, Normalized gene count comparisons across unstimulated UTD-MSCs, stimulated UTD-MSCs, unstimulated EcCAR-MSCs and stimulated EcCAR-MSCs. CD28-linked transcription factor genes (NFkB1, JUN, RELB and IRF1) and downstream effector genes (TRAF1, TLR3 and FYN) were upregulated in stimulated EcCAR-MSC group only. Displaying mean ± s.d. Gene counts normalized across groups. Statistics by one-way ANOVA. Significant P values are displayed.

Fig. 6 |. Increased cytokine secretion, surface marker expression and subsequent T-cell modulation identified following EcCAR-MSC stimulation.

a, Heatmap displaying upregulated cytokines across Ecad⁺ versus Ecad⁻ cell line stimulation on T cells alone, UTD-MSCs and EcCAR-MSCs by multiplexed cytokine assay. Data show degree of cytokine fold change as measured by multiplexed assay normalized by smallest (0) and largest (100) values per cytokine (n = 2 technical replicates per assay). b, Inhibitory surface marker expression across UTD-MSCs and EcCAR-MSCs with Ecad⁺ cell line stimulation as compared with Ecad⁻ cell line stimulation. Displaying mean ± s.d. results with statistics by two-way ANOVA (n = 3 replicates per donor). c, Enriched serum cytokines (IL-10, TNFα, G-CSF, eotaxin and fibroblast growth factor 2 (FGF-2) by pg ml⁻¹) at 2 weeks identified only in EcCAR-MSC-treated mice from tumour xenografts. Displaying mean ± s.e.m. with statistics by ordinary one-way ANOVA (n = 4–6 mice per group). d, Prevention of GvHD-induced weight loss following single dose treatment of EcCAR-MSCs versus UTD-MSCs. Displaying mean ± s.e.m. of percentage weight change. Statistics by two-way ANOVA (n = 5 mice per group). e, Mouse peripheral blood assessment comparing absolute number of human CD3⁺ T cells between EcCAR-MSC and UTD-MSC treatment groups by flow cytometry 2 weeks following first MSC injection. Displaying mean ± s.d. with statistics by unpaired t-test (n = 5 mice per group). f, Human CD4⁺ and CD8⁺ T cells quantified in mouse peripheral blood 2 weeks following EcCAR-MSC treatment. Displaying mean ± s.d. cell counts. Statistics by two-way ANOVA (n = 5 mice per group). g, Alteration in human CD4⁺ to CD8⁺ T-cell proportions in mice treated with EcCAR-MSCs versus UTD-MSCs. Displaying mean ± s.d. of percentage cell subsets. Statistics by two-way ANOVA (n = 5 mice per group). h, Human CD4⁺CD25⁺CD127^dim Treg cell subsets 4 weeks following MSC treatment in GvHD xenograft mice. Displaying mean ± s.d. percantage of cells. Statistics by unpaired t-test (n = 4 mice per group). Significant P values are displayed.

Fig. 7 |. CD28ζ signalling domain within EcCAR-MSCs is required for optimal immunosuppressive efficacy.

a, Schema of EcCAR-MSC construct designs used to further investigate CAR-MSC mechanism of action. Constructs contained the full CD28ζ, CD28 alone, CD3ζ alone, and null (no signalling domain) EcCAR-MSCs with extracellular heavy and light chain, hinge, transmembrane (TM) and intracellular domains. b, Ridge plots displaying CAR expression on MSCs for CD28ζ, CD28, CD3ζ and null EcCAR-MSC constructs detected by flow cytometry. Displaying CAR percentage expression representing ≥5 independent experiments on MSC donors (n = 4). c, Degree of CD3⁺ T-cell suppression following Ecad⁺ cell line stimulation across each EcCAR-MSC signalling domain subtype or T cells alone in culture with Ecad⁺ and matched Ecad⁻ cell line. Displaying mean ± s.d. representative of four primary T-cell donors. Statistics by one-way ANOVA (n = 3 technical replicates per donor). d, Schema of acute GvHD xenograft model testing of CAR-MSC signalling domain efficacy: NSG mice were first irradiated at 250 cGy to further prime inflammatory environment. Human PBMCs were next injected (10–15 × 10⁶ cells i.v.) into mice along with one of the following MSC groups: CD28ζ, CD28, CD3ζ, null EcCAR-MSCs, UTD-MSCs (1 × 10⁶ cells i.p.) or ‘no MSC’ treatment dosed on days 0 and 25. Mice were monitored for long-term weight loss, the development of clinical GvHD symptoms and survival. Image created with BioRender.com. e, Graph depicting percentage weight change from baseline in acute GvHD xenografts following treatment with each EcCAR-MSC signalling domain subtypes or no treatment on days 0 and 25 (arrows). Displaying mean ± s.e.m. with statistics by two-way ANOVA (n = 5 or 6 mice per group, showing one representative of two independent experiments). f, GvHD clinical scoring (Supplementary Table 2) severity results depicting symptom progression following treatment with each EcCAR-MSC subtype. Displaying mean ± s.e.m. with statistics by two-way ANOVA (n = 5 or 6 mice per group, showing one representative of two independent experiments). g, Survival outcomes following treatment with each EcCAR-MSC subtype. Data depicting trends in enhanced survival, but without statistical significance, are included with prospective P value measurements. Statistics by Kaplan–Meier survival analysis (n = 5 or 6 mice per group, showing one representative of two independent experiments). NS is P ≥ 0.05, and significant P values are displayed.

Fig. 8 |. EcCAR-MSC safety and clearance profiles across tissues.

a, Schema for luciferase (luc)⁺ MSC persistence model: luc⁺ GFP⁺ UTD-MSCs or luc⁺ GFP⁺ CAR-MSCs were injected into human PBMC-primed NSG mice with or without CAR stimulation through co-administration of irradiated Ecad⁺ cells and followed by serial biweekly BLI. Image created with BioRender.com. b, Pictorial representation of luciferase⁺ MSC expansion and clearance in mice from 3 to 24 days following administration. c, Graph depicting flux of BLI as a measure of MSC expansion and clearance kinetics following administration of EcCAR-MSCs or UTD-MSCs with or without additional Ecad stimulation. Displaying mean BLI with statistics by two-way ANOVA (n = 4 or 5 mice per group). d, Absolute number of keratinocytes following 24 h co-culture with UTD-MSCs or EcCAR-MSCs. Displaying mean ± s.d. results with statistics by paired t-test (n = 3 replicates per donor). e, Ecad expression identified on keratinocytes following 24 h co-culture with UTD-MSCs or EcCAR-MSCs. Displaying mean per cent ± s.d. results with statistics by paired t-test (n = 3 replicates per donor). f, Heatmap shows upregulated cytokines across UTD-MSCS, EcCAR-MSCs and keratinocytes at baseline and following 24 h co-culture conditions. Data show cytokine levels measured by multiplexed assay normalized by smallest (0) and largest (100) values per cytokine (n = 2 technical replicates per assay). g, Absolute number of bronchial cells following 24 h co-culture with varying ratios of UTD-MSCs or EcCAR-MSCs. Displaying mean ± s.d. results with statistics by paired t-test (n = 3 technical replicates). h, Immunofluorescent-based quantification of EcCAR-MSC versus CD19-CAR-MSC localization to Ecad⁺ lung tissue. Data determined by the percentage of MSC⁺ regions per focal image. Displaying mean ± s.d. with statistics by ordinary one-way ANOVA (n = 3 mice per group). i, Immunofluorescent imaging of mouse bronchial (lung) tissue isolated from acute GvHD xenograft model 7 days following GFP⁺ MSC administration (Fig. 4a). Comparisons between CD19-CAR-MSC and Ecad-CAR-MSC (green) treated mice with interrogation of colocalization to Ecad⁺ (red) bronchial regions. Cell nuclei are stained with DAPI (blue) with each colour channel displayed to make merged image. Images obtained at 40× magnification capturing 1× crop area. Scale bars, 20 μm. Data showing mean ± s.d. with statistics by unpaired t-test (n = 4 or 5 mice per group, two independent experiments). NS of P ≥ 0.05.