Targeting inflammation with chimeric antigen receptor macrophages using a signal switch

Abstract

Chimeric antigen receptor (CAR) T-cell immunotherapy has shown great success in clinical cancer, bringing hope to apply CAR strategies to other clinical settings. Here we developed a CAR macrophage (CAR-M) that recognizes the major inflammatory molecule tumour necrosis factor (TNF) and activates an intracellular IL-4 signalling pathway, thereby programming engineered macrophages for an anti-inflammatory function. CAR-M therapy has exhibited efficacy in mouse models of both acute and chronic inflammatory diseases. In kidney ischaemia reperfusion injury (IRI), infused CAR-Ms switched to an anti-inflammatory phenotype in inflamed kidney and attenuated kidney IRI. The anti-inflammatory phenotype of infused CAR-Ms switched off during the recovery phase of kidney IRI, coinciding with the disappearance of TNF. In Adriamycin-induced nephropathy, a model of chronic inflammatory disease, infused CAR-Ms maintained an anti-inflammatory phenotype for several weeks in response to sustained high levels of TNF and improved kidney function and structure. CAR-Ms also effectively reduced tissue injury in another organ, the liver. Human anti-TNF CAR-Ms exhibit anti-inflammatory phenotype and function in response to TNF. The CAR-M design, using signal switching, holds promise for the treatment of a broad range of acute and chronic inflammatory diseases.

Fig. 1. Generation and characterization of CAR-Ms.

a, Schematic representation of the anti-TNF CAR construct design for macrophages. The CAR contains a Flag tag epitope, a scFv antibody specific for TNF, the hinge and transmembrane domains of mouse CD8α and a mouse IL-4 receptor intracellular domain. b, Representative flow cytometric analysis of transduction efficiency (Flag-tag expression) on macrophages 3 days after transduction with CAR (blue, CAR-M; red, UTD-M). c, Quantitative analysis of the percentage of Flag+ cells on CAR-M and UTD-M. Data shown are the mean ± s.e.m. (n = 6 per group). ***P < 0.001. d, Representative flow cytometric analysis of mouse M2 markers p-STAT6, p-AKT, CD206, EGR2 and PD-L2 on CAR-M in response to TNF or IL-4. ISO, isotype. e, Flow cytometric analysis of mean fluorescence intensity (MFI) of M2 markers. Data shown are the mean ± s.e.m. (n = 4–6 per group) and are representative of 3 independent experiments. *P < 0.05, **P < 0.01, ***P < 0.001. f, Gene expression principal component analysis clustering from UTD, UTD.TNF, M1, M2, CAR-M, CAR-M.TNF, CARΔ-M and CARΔ-M.TNF. n = 4 per group. g, Scaled expression heat map of genes families in important pathways (accessed using molecular signatures database (MSIGDB)). TNFR, tumor necrosis factor receptor. h, Volcano plot of differentially expressed genes in CAR-M.TNF versus CAR-M. Blue indicates P_adj < 0.05 and log₂(fold change) >1 or <−1. Red triangles indicate M1-associated genes, and green triangles indicate M2-associated genes (n = 4 per group). Statistical significance was calculated using the QLFtest for DGE data. i, Gene Ontology and KEGG pathway analysis results of upregulated and downregulated pathways in CAR-M.TNF versus CAR-M. j, UTD-M and CAR-M were cocultured with FITC-labelled dextran in the presence or absence of TNF for 45 min. The uptake of fluorescent beads (MFI) was determined by flow cytometry. Data shown are the mean ± s.e.m. (n = 6 per group) and are representative of 2 independent experiments. ***P < 0.001. k, The UTD-Ms and CAR-M.TNF were incubated with PKH26 Red-labelled apoptotic cells for 3 h. The percentage of F4/80+ macrophages that had incorporated the apoptotic neutrophils was determined by flow cytometry. Data shown are the mean ± s.e.m. (n = 5 per group) and are representative of 3 independent experiments. **P < 0.01. l, Inhibition of CD4 T-cell proliferation was examined with various dosages of UTD-Ms and CAR-M.TNF. c.p.m, counts per minute. Data shown are the mean ± s.e.m. (n = 5 per group) and are representative of 2 independent experiments. **P < 0.01, ***P < 0.001 versus UTD-M. m, M1 macrophages were cocultured with TNF-activated CAR-Ms for 24 h. The mRNA expression of iNOS, IL-1β and IL-6 of M1 was examined by qPCR. Data shown are the mean ± s.e.m. (n = 4–6 per group) and are representative of 3 independent experiments. **P < 0.01. Source data

Fig. 2. Evaluation of protective effect of CAR-M in acute kidney injury model.

a, C57BL/6 mice were treated with CD45.1⁺ UTD-M, CARΔ-M, CAR-M or M2 at 6 h after bilateral IRI. Mice were euthanized at 48 h after IRI. b, Representative PAS-stained sections of kidney outer medulla from IRI mice treated with UTD-M, CARΔ-M, CAR-M or M2. Each panel shows low (top, bar = 250 µm) and high (bottom; corresponding to the boxed area, bar = 50 µm) power view. c, Semiquantitative assessment of tubular injury. **P < 0.01, ***P < 0.001. d, Serum creatinine levels were assessed in these mice. **P < 0.01, ***P < 0.001. e, Number of Gr-1⁺ neutrophils was assessed by immunofluorescence staining in the outer medulla of the kidney. h.p.f., high-power field. ***P < 0.001. f,g, The mRNA levels of pro-inflammatory cytokines (f) and chemokines (g) in the kidneys. Data shown are the mean ± s.e.m. (n = 6–8 per group). NS, not significant; *P < 0.05; **P < 0.01; ***P < 0.001. h, The biodistribution of XenoLight DiR-labelled UTD-M or CAR-M in unilateral IRI model. The IRI mice were euthanized, and organs were explanted for ex vivo imaging 48 h after injection. n = 4–5 mice per treatment group. i, The mRNA level of TNF in the kidneys over the course of kidney IRI. Data shown are the mean ± s.e.m. (n = 4 per group). j,k, CD45.1⁺ UTD-Ms and CD45.1⁺ CAR-Ms were sorted from various organs at 48 h after kidney IRI. Representative flow cytometric analysis of M2 markers CD206 and EGR2 on transfused UTD-M and CAR-M in kidney with IRI (j). The M2-like phenotype of transfused macrophages from various organs (k) was assessed by flow cytometry (CD206 and EGR2) and qPCR (Arginase and IL-10). Data shown are the mean ± s.e.m. (n = 4 per group). ***P < 0.001. l, The M2-like phenotype of transfused macrophages over the course of IRI was assessed by flow cytometry and qPCR. Data shown are the mean ± s.e.m. (n = 4 per group). ***P < 0.001 versus Kidney-sham. Source data

Fig. 3. Failed renoprotection of CAR-Ms in TNF-deficient mice with IRI.

a, TNF-deficient mice were treated with CD45.1⁺ CARΔ-M or CAR-M at 6 h after bilateral IRI. Mice were euthanized at 48 h after IRI. b, Representative PAS-stained sections of kidney outer medulla from IRI mice treated with CARΔ-M or CAR-M. Each panel shows low (top, bar = 250 µm) and high (bottom; corresponding to the boxed area, bar = 50 µm) power view. c, Semiquantitative assessment of tubular injury. d, Serum creatinine levels were assessed in these mice. e, Number of Gr-1⁺ neutrophils was assessed by immunofluorescence staining in the outer medulla of the kidney. f,g, The mRNA levels of pro-inflammatory cytokines (f) and chemokines (g) in the kidneys. Data shown are the mean ± s.e.m. (n = 6–8 per group). h,i, CARΔ-Ms and CAR-Ms were sorted from kidneys at 48 h after kidney IRI. Representative flow cytometric analysis of M2 markers (CD206 and EGR2 (h)) and M1 markers (CD38 and CD86 (i)) on transfused CARΔ-Ms and CAR-Ms in kidney with IRI. j,k, The M2 (j) and M1-like (k) phenotype of transfused macrophages from Sham or IRI kidney was assessed by flow cytometry and qPCR. Data shown are the mean ± s.e.m. (n = 4 per group). ***P < 0.001. Source data

Fig. 4. Evaluation of protective effect of CAR-M in chronic kidney disease model.

a, BALB/c mice were treated with UTD-M, CARΔ-M, CAR-M or M2 at day 7 after ADR injection. Mice were euthanized at 28 days. b, Representative PAS-stained sections of kidney from AN mice treated with UTD-M, CARΔ-M, CAR-M or M2. Each panel shows low (top, bar = 250 µm) and high (bottom; corresponding to the boxed area, bar = 50 µm) power view. c, Quantitative assessment of glomerular sclerosis and tubular damage. **P < 0.01, ***P < 0.001. d, Quantitative analysis of the positive area of Gomori Trichrome staining (kidney fibrosis). **P < 0.01, ***P < 0.001. e, Proteinuria and creatinine clearance were assessed at day 28 after ADR injection. **P < 0.01, ***P < 0.001. f,g, mRNA levels of pro-inflammatory cytokines (f) and chemokines (g) in kidneys. Data shown are the mean ± s.e.m. (n = 8–13 per group). *P < 0.05, **P < 0.01, ***P < 0.001. h, mRNA level of TNF in the kidneys over the course of AN. Data shown are the mean ± s.e.m. (n = 4 per group). *P < 0.05, **P < 0.01. i, Immunofluorescence double staining of CD90.1 and CD206 in kidney sections of AN mice at day 28. The white arrows indicate CD90.1⁺CD206⁺ CAR-Ms. Scale bars, 50 µm. j,l, CD90.1⁺ CAR-Ms were sorted from normal kidney and AN kidney at days 14 and 28 after ADR injection. Representative flow cytometric analysis of M2 markers (CD206 and EGR2, j) and M1 markers (CD38 and CD86, l) on transfused CAR-Ms from normal kidney and AN kidney at day 14. k,m, The M2 (k) and M1-like (m) phenotypes of transfused CAR-Ms was assessed by flow cytometry and qPCR. Data shown are the mean ± s.e.m. (n = 4 per group). ***P < 0.001. Source data

Fig. 5. Evaluation of protective effect of CAR-M in acute liver injury model.

a, C57BL/6 mice were treated with CD45.1⁺ UTD-M, CARΔ-M, CAR-M or M2 at 6 h after liver IRI. Mice were euthanized at 48 h after IRI. b, Representative haematoxylin and eosin-stained sections of liver from IRI mice treated with UTD-M, CARΔ-M, CAR-M or M2. Each panel shows low (top, bar = 250 µm) and high (bottom; corresponding to the boxed area, bar = 50 µm) power view. c, Quantitative assessment of necrotic areas in livers. *P < 0.05, ***P < 0.001. d, Serum alanine aminotransferase and aspartate aminotransferase levels were assessed in these mice. ALT, alanine aminotransferase; AST, aspartate aminotransferase. *P < 0.05, ***P < 0.001. e, Number of Gr-1⁺ neutrophils was assessed by immunofluorescence staining in liver. *P < 0.05, ***P < 0.001. f, mRNA levels of pro-inflammatory cytokines/chemokines in the livers. Data shown are the mean ± s.e.m. (n = 6–8 per group). *P < 0.05, **P < 0.01, ***P < 0.001. g, Immunofluorescence double staining of CD45.1 and CD206 in liver sections at 48 h after liver IRI. The white arrows indicate CD45.1⁺CD206⁺ CAR-Ms. Scale bars, 50 µm. h,j, CD45.1⁺ UTD-Ms and CD45.1⁺ CAR-Ms were sorted from various organs of mice at 48 h after liver IRI. Representative flow cytometric analysis of M2 markers (CD206 and EGR2, h) and M1 markers (CD38 and CD86, j) on transfused UTD-M and CAR-Ms from liver with IRI. i, The M2-like phenotype of transfused macrophages was assessed by flow cytometry (CD206 and EGR2) and qPCR (Arginase and IL-10). ***P < 0.001. k, The M1-like phenotype of transfused macrophages was assessed by flow cytometry (CD38 and CD86) and qPCR (iNOS and IL-1β). Data shown are the mean ± s.e.m. (n = 4 per group). ***P < 0.001. Source data

Fig. 6. Generation and characterization of human CAR-Ms.

a, Schematic representation of the anti-TNF CAR construct design for human macrophages. The CAR contains a Flag tag epitope, a scFv antibody specific for TNF, the hinge and transmembrane domains of human CD8α, and a human IL-4 receptor intracellular domain. b, Representative flow cytometric analysis of transduction efficiency (Flag-tag expression) on macrophages 3 days after transduction with CAR (blue, CAR-M; red, UTD-M). c, Quantitative analysis of the percentage of Flag+ cells on CAR-M and UTD-M. Data shown are the mean ± s.e.m. (n = 5 per group). ***P < 0.001. d, Representative flow cytometric analysis of IL-4 signalling pathways p-STAT6, p-AKT, p-JAK3 and p-TYK2 on CAR-M in response to TNF or IL-4. e, Flow cytometric analysis of mean fluorescence intensity (MFI) of IL-4 signalling pathways. Data shown are the mean ± s.e.m. (n = 5 per group) and are representative of two independent experiments. *P < 0.05, **P < 0.01, ***P < 0.001. f, Representative flow cytometric analysis of M2 markers CD206 and CD200R on CAR-M in response to TNF or IL-4. g, Flow cytometric analysis of MFI of M2 markers. Data shown are the mean ± s.e.m. (n = 5 per group) and are representative of 2 independent experiments. *P < 0.05; **P < 0.01; ***P < 0.001. h, Production of IL-10 by CAR-Ms after stimulation with TNF or IL-4. Data shown are the mean ± s.e.m. (n = 5 per group) and are representative of 2 independent experiments. ***P < 0.001. i, Inhibition of CD4 T-cell proliferation was examined with various dosages of CAR-Ms and CAR-M.TNF. Data shown are the mean ± s.e.m. (n = 5 per group) and are representative of 2 independent experiments. **P < 0.01 versus CAR-M. j, The human CAR-Ms and CAR-M.TNF were incubated with PKH26 Red-labelled apoptotic neutrophils for 3 h. The percentage of macrophages that had incorporated the apoptotic neutrophils was determined by flow cytometry. Data shown are the mean ± s.e.m. (n = 5 per group) and are representative of 2 independent experiments. ***P < 0.001. Source data

Fig. 7. CAR-Ms as inflammatory disease immunotherapy.

Our engineered CAR structure integrates an anti-TNF scFv as the extracellular domain with the intracellular domain of IL-4 receptor alpha (IL-4Rα). This design enables the capture of TNF and converts its signal into an IL-4 signal, promoting an anti-inflammatory response on macrophages. When capturing TNF, CAR-Ms exhibit an M2-like phenotype including elevated p-STAT6 expression and increased M2 markers (CD206, Arginase 1). In vivo, CAR-Ms reduce inflammation and injury in acute and chronic kidney inflammatory diseases (renal IRI and AN models), demonstrating substantial therapeutic potential. Remarkably, the CAR-Ms switched to the M2 phenotype only at the site of inflammation and reverted once the inflammation subsided, ensuring targeted and timely anti-inflammatory responses. CAR-Ms also protect against liver IRI, suggesting potential applications for inflammatory diseases in multiple organs.

Extended Data Fig. 1. Lentivirus transduction led to a transient pro-inflammatory macrophage phenotype.

(A) Volcano plot of differentially expressed genes in CAR-M versus UTD. Blue indicates Padj < 0.05 and log2 fold change >1 or <−1. Red triangles indicate significant M1-associated genes and green triangles indicate significant M2-associated genes (n = 4 per group). Statistical significance was calculated using the QLFtest for DGE data. (B) Hierarchical clustering of differentially expressed genes from UTD or CAR-M (n = 4 per group), 48 hours post transduction. The heatmap shows log2 fold-change in gene expression relative to UTD. (C) GO and KEGG pathway analysis results of up- and down-regulated pathways in CAR-M, compared to UTD. (D) Mouse bone marrow derived macrophages were transduced with CAR at MOIs of 0, 100, 250, 500, or 1000 IFU. CAR expression on macrophages correlated with MOI. Data shown are the mean ± SEM (n = 4 per group). Correlation was determined via nonlinear regression and Pearson. (E) Flow cytometry analysis of mouse M1 markers CD38 and CD86 and M2 marker CD206 in response to transduction with increasing MOIs of CAR by FACS. Data shown are the mean ± SEM (n = 4 per group) and are representative of 2 independent experiments. (F) Surface expression of M1 markers CD38 and CD86 after transduction with equivalent MOIs of CARΔ and CAR. (G) Production of IL-1β and IL-6 after transduction with equivalent MOIs of CARΔ and CAR. Data shown are the mean ± SEM (n = 4 per group) and are representative of 2 independent experiments. (H and I) Transient expression of M1 markers (CD38, CD86, iNOS, IL-1β and IL-6) on CAR-Ms during 14 days of in vitro culture. Data shown are the mean ± SEM (n = 4 per group). NS: not significant, *P < 0.05, **P < 0.01, ***P < 0.001 vs UTD. MOI: multiplicity of infection. Source data

Extended Data Fig. 2. TNF-α did not induce M2 phenotype in macrophages transduced with a truncated CAR (CARΔ).

(A) Volcano plot of differentially expressed genes in CARΔ-M + TNF-α versus CARΔ-M. Blue indicates Padj < 0.05 and log2 fold change >1 or <−1. Red triangles indicate significant M1-associated genes and green triangles indicate significant M2-associated genes (n = 4 per group). Statistical significance was calculated using the QLFtest for DGE data. (B) Hierarchical clustering of differentially expressed genes from CARΔ-M + TNF-α versus CARΔ-M (n = 4 per group). The heatmap shows log2 fold-change in gene expression relative to CARΔ-M. (C) GO and KEGG pathway analysis results of up- and down-regulated pathways in CARΔ-M.TNF-α, compared to CARΔ-M. (D) Quantitative PCR analysis of TNF-α signalling pathway in CARΔ-M in response to TNF-α. (E and F) Flow cytometric analysis of mouse M1 markers (CD38 and CD86) and M2 markers (p-STAT6, p-AKT, CD206, EGR2 and PD-L2) on CARΔ-M in response to TNF-α. Data shown are the mean ± SEM (n = 4 per group) and are representative of 3 independent experiments. **P < 0.01, ***P < 0.001. Source data

Extended Data Fig. 3. Characterization of CAR-M.

(A and B) Representative flow cytometric analysis of IL-4 signalling pathway p-JAK3 and p-TYK2 on CAR-M or CARΔ-M in response to TNF-α or IL-4. Data shown are the mean ± SEM (n = 4-6 per group). *P < 0.05, ***P < 0.001. (C) Quantitative PCR analysis of TNF-α signalling pathway on CAR-M in response to TNF-α. Data shown are the mean ± SEM (n = 4 per group) and are representative of 2 independent experiments. **P < 0.01. (D and E) Volcano plot of differentially expressed genes in CAR-M.TNF-α versus UTD and M2 versus UTD. Blue indicates Padj < 0.05 and log2 fold change >1 or <−1. Red triangles indicate significant M1-associated genes and green triangles indicate significant M2-associated genes (n = 4 per group). Statistical significance was calculated using the QLFtest for DGE data. (F and G) GO and KEGG pathway analysis results of up- and down-regulated pathways in CAR-M.TNF-α versus UTD and M2 versus UTD. (H) Venn diagram show the overlapping differently expressed genes from CAR-M.TNF-α versus UTD and M2 versus UTD. (I) Volcano plot of differentially expressed genes in CAR-M.TNF-α versus CARΔ-M.TNF-α. (J) Hierarchical clustering of differentially expressed genes from CAR-M.TNF-α versus CARΔ-M.TNF-α (n = 4 per group). The heatmap shows log2 fold-change in gene expression relative to CARΔ-M.TNF-α. (K) GO and KEGG pathway analysis results of up- and down-regulated pathways in CAR-M.TNF-α, compared to CARΔ-M.TNF-α. Source data

Extended Data Fig. 4. CAR-Ms did not switch to a pro-inflammatory phenotype in inflamed kidney of IRI mice.

(A and B) CD45.1 + UTD-Ms and CD45.1 + CAR-Ms were sorted from various organs at 48 hours after kidney IRI. Representative flow cytometric analysis of M1 markers CD38 and CD86 on transfused UTD-M and CAR-M in kidney with IRI (A). The M1-like phenotype of transfused macrophages from various organs (B) was assessed by flow cytometry (CD38 and CD86) and quantitative PCR (iNOS and IL-1β). Data shown are the mean ± SEM (n = 4 per group). ***P < 0.001. (C) The M1-like phenotype of transfused macrophages over the course of IRI was assessed by flow cytometry and quantitative PCR. Data shown are the mean ± SEM (n = 4 per group). Source data

Extended Data Fig. 5. CAR-Ms reduced TEC apoptosis and promoted TEC proliferation.

UTD-Ms or TNF-α-treated CAR-Ms were cocultured with ischemic kidney TECs (IRI-TECs) for 1–3 days. TECs were exposed to serum-free K1 medium alone as the nonischemic control (Ctrl-TEC). (A) Representative FACS analysis of apoptosis in TECs after 1-day coculture. (B) Frequency of early apoptosis (Annexin V + 7AAD- cells) and late apoptosis (Annexin V + 7AAD+ cells) in TECs after 1-day coculture. Data shown are the mean ± SEM (n = 4 per group) and are representative of 2 independent experiments. ***P < 0.001. (C) Numbers of kidney TECs were counted at each time point after coculture. Data shown are the mean ± SEM (n = 6 per group) and are representative of 3 independent experiments. ***P < 0.001. (D and E) The UTD-Ms and TNF-α-treated CAR-Ms were incubated with PKH26 Red-labeled ischemic kidney TECs (D) or nonischemic normal TECs (E) for 4 h. The percentage of macrophages that had incorporated the TECs was determined by flow cytometry. Data shown are the mean ± SEM (n = 4 per group) and are representative of 2 independent experiments. NS: not significant, **P < 0.01. (F) Kidney tubular epithelial cell apoptosis was measured by terminal deoxynucleotidyl transferase–mediated dUTP nick end-labeling assay on the representative sections of the kidney. (G) Kidney tubular epithelial cell proliferation was assessed by immunofluorescence staining for Ki67 on representative sections of kidney. Bar=50 µm. Data shown are the mean ± SEM (n = 6-8 per group). ***P < 0.001. Source data

Extended Data Fig. 6. CAR-Ms suppressed CD8 T cell activation in AN mice and reduced cytotoxicity of CD8 T cells on TEC in vitro.

(A) CD8 T cells were isolated from kidneys of AN mice treated with UTD-M or CAR-M. Inflammatory cytokines and cytotoxic molecules from CD8 T cells were assessed by flow cytometry and quantitative PCR. Data shown are the mean ± SEM (n = 4-6 per group). **P < 0.01, ***P < 0.001. (B, C) Tubular injury in AN mice is simulated by incubating kidney TECs with ADR for 24 hours. The injured kidney TECs (ADR-TECs) were then cocultured for 24 hours with CD8 T cells isolated from kidneys of AN mice treated with UTD-M or CAR-M. (B) Representative FACS analysis of apoptosis in TECs after 1-day coculture. (C) Frequency of early apoptosis (Annexin V + 7AAD– cells) and late apoptosis (Annexin V + 7AAD+ cells) in TECs after 1-day coculture. Data shown are the mean ± SEM (n = 4-6 per group). ***P < 0.001. Source data

Extended Data Fig. 7. Tracking of CAR-M in vivo.

(A) Biodistribution of XenoLight DiR-labeled CAR-Ms after IV administration in normal, renal IRI and AN mouse (n = 4 per group, the first mouse is a normal mouse that did not receive IV injection of CAR-Ms). Fluorescent signal was tracked over the course of 60 days (dotted line represents background fluorescence). The experiment was performed twice with similar results. (B) The normal, IRI and AN mice were euthanized, and organs were explanted for ex vivo imaging at 1, 10, 20, 40 and 60 days post injection (n = 4 mice per group). Macrophage mainly accumulated in liver, spleen, lung, and also infiltrated into diseased kidney. Source data

Extended Data Fig. 8. TNF-α did not induce M2 phenotype in human macrophages transduced with a truncated CAR (CARΔ).

(A) CARΔ is a truncated CAR lacking the IL-4 receptor intracellular domain. (B) Quantitative PCR analysis of TNF-α signalling pathway on human CARΔ-M in response to TNF-α. (C and D) Flow cytometric analysis of M1 markers (CD80 and CD86) and M2 markers (p-STAT6, p-AKT, CD206 and CD200R) on human CARΔ-M in response to TNF-α. Data shown are the mean ± SEM (n = 4 per group) and are representative of 2 independent experiments. ***P < 0.001. Source data