TREM2hi resident macrophages protect the septic heart by maintaining cardiomyocyte homeostasis

Abstract

Sepsis-induced cardiomyopathy (SICM) is common in septic patients with a high mortality and is characterized by an abnormal immune response. Owing to cellular heterogeneity, understanding the roles of immune cell subsets in SICM has been challenging. Here we identify a unique subpopulation of cardiac-resident macrophages termed CD163⁺RETNLA⁺ (Mac1), which undergoes self-renewal during sepsis and can be targeted to prevent SICM. By combining single-cell RNA sequencing with fate mapping in a mouse model of sepsis, we demonstrate that the Mac1 subpopulation has distinct transcriptomic signatures enriched in endocytosis and displays high expression of TREM2 (TREM2^hi). TREM2^hi Mac1 cells actively scavenge cardiomyocyte-ejected dysfunctional mitochondria. Trem2 deficiency in macrophages impairs the self-renewal capability of the Mac1 subpopulation and consequently results in defective elimination of damaged mitochondria, excessive inflammatory response in cardiac tissue, exacerbated cardiac dysfunction and decreased survival. Notably, intrapericardial administration of TREM2^hi Mac1 cells prevents SICM. Our findings suggest that the modulation of TREM2^hi Mac1 cells could serve as a therapeutic strategy for SICM.

Fig. 1. Single-cell atlas of cardiac immune cells during sepsis progression.

a, Schematic diagram showing scRNA-seq pipeline of murine cardiac immune cells during sepsis progression. SS, steady state. b, Uniform Manifold Approximation and Projection (UMAP) plots of 29,537 cardiac immune cells allocated into 15 clusters from the hearts of 14 WT mice at SS, 3, 7 and 21 d after CLP. Cardiac cells from 3–4 mice were mixed as one sample. c, UMAP plots of cardiac immune cells colored by cell type. d, Heat map showing selected cluster marker genes (top, color-coded by cluster and condition) with exemplar genes and cell-type annotation labeled. e, UMAP plots of cardiac immune cells from WT mice in different time points of sepsis, annotated by cell type as in c. Each time point is contributed by either three (SS or 21 d) or four mice (3 or 7 d). f, Bar plots showing the distribution of immune cell types from WT mice at different time points of sepsis (Extended Data Figs. 1 and 2). Source data

Fig. 2. A distinct cardiac macrophage subset is associated with the recovery of SICM.

a, UMAP plots of 24,058 monocyte-macrophages allocated to the eight clusters from Fig. 1c. b, Heat map showing top 20 cluster marker genes (top, color-coded by cluster and condition) with exemplar genes and cluster annotations labeled. c, UMAP plots of the monocyte-macrophage compartment from WT mice at SS, 3, 7 and 21 d after CLP, annotated by cluster as in a. d, Bar plots showing cluster distribution within the monocyte-macrophage compartment from WT mice after CLP. e, Representative contour plots showing flow cytometric analysis of cardiac CD45⁺CD11b⁺F4/80⁺CD163⁺RETNLA⁺ macrophages from WT mice after CLP. f, Quantification of CD163⁺RETNLA⁺ macrophages per mg of cardiac tissue. Bars show mean ± s.e.m. Two-sided P values were determined by one-way analysis of variance (ANOVA) with Tukey’s multiple comparison test. g,h, Correlation of cardiac CD163⁺RETNLA⁺ macrophage numbers with EF and cTnI levels. Data are presented as two-tailed Spearman’s rank correlation. Dashed lines represent 95% confidence intervals. Each symbol represents one animal (SS, n = 8 mice; CLP 3 d, n = 8 mice; CLP 7 d, n = 9 mice; CLP 21 d, n = 10 mice) (f,g). Results represent four independent experiments in this figure (Extended Data Fig. 3). Source data

Fig. 3. Developmental trajectory and lineage tracing track the fate of Mac1 subset in SICM.

a, The Monocle prediction of the monocyte-macrophage developmental trajectory with Seurat’s cluster information in Fig. 2a mapped alongside. b, The Monocle prediction of monocyte-macrophage developmental trajectory with each Seurat-based cluster shown separately. c, Heat map of top 50 DEGs along with the pseudotime. The relative position of individual subsets across pseudotime is illustrated below. d, Representative contour plots showing the expression of tdTomato (CX3CR1) on gated CD45⁺CD11b⁺F4/80⁺CD163⁺RETNLA⁺ macrophages, CD45⁺CD11b⁺F4/80⁺ (non-CD163⁺RETNLA⁺) macrophages and CD45⁺CD11b⁺F4/80⁻LY6C⁺ monocytes. Graph showing the percentage of tdTomato⁺ cells isolated from hearts of WT mice at each time point along sepsis. NS, not significant. e, Representative contour plots showing the expression of tdTomato (CX3CR1) and CD163 on gated CD45⁺CD11b⁺F4/80⁺RETNLA⁺ macrophages and graph showing absolute numbers of CD45⁺CD11b⁺F4/80⁺CD163⁺RETNLA⁺tdTomato⁺ macrophages isolated from hearts of WT mice at each time point following sepsis. f, Representative histograms showing the expression of Ki67 in CD45⁺CD11b⁺F4/80⁺CD163⁺RETNLA⁺ macrophages and graph showing quantification of mean fluorescence intensity (MFI) of Ki67 at SS, 3 and 7 d after CLP. Dots represent individual subjects (d–f); SS, n = 5 mice; CLP 3 d, n = 6 mice; CLP 7 d, n = 7 mice. Data are shown as mean ± s.e.m.; two-sided P values were determined by one-way ANOVA, followed by Games-Howell’s (d) or Sidak’s (d–f) multiple comparison test; results represent three independent experiments (Extended Data Fig. 4). Source data

Fig. 4. TREM2 is highly expressed on Mac1 subset and essential for Mac1 cells self-renewal in SICM.

a, Dot plots showing the top 20 biological processes for the upregulated DEGs of Mac1 analyzed by GO. Red arrows indicate endocytosis-related pathways. b, UMAP plots and violin plots showing the expression of Trem2 in the monocyte-macrophage compartment. c, Representative immunofluorescence images of CD163 (green), TREM2 (red) and nuclei (blue) in cardiac tissue from WT mice at SS (n = 5) and 7 d after CLP (n = 5). The dotted lines indicate CD163⁺TREM2⁺ macrophages. Scale bars, 20 μm. d, Representative histograms of flow cytometric analysis of TREM2 expression in cardiac CD45⁺CD11b⁺F4/80⁺CD163⁺RETNLA⁺ and CD45⁺CD11b⁺F4/80⁺(non-CD163⁺RETNLA⁺) macrophages from WT mice at SS and 7 d after CLP. e, UMAP plots showing cell clustering results corresponding to Fig. 2a for a total of 13,405 monocytes-macrophages in WT and Trem2 knockout (KO) hearts at 7 d after CLP. f, Cluster distribution within the monocyte-macrophage subsets in WT and Trem2-KO hearts at 7 d after CLP. g, Percentages and absolute numbers of CD163⁺RETNLA⁺ macrophages analyzed by flow cytometry from hearts of WT (n = 6) and Trem2-KO (n = 6) mice at 7 d after CLP. h, Representative contour plots showing the expression of Ki67 on gated CD45⁺CD11b⁺F4/80⁺CD163⁺RETNLA⁺ macrophages and graph showing percentages of Ki67-expressing cells in WT (n = 5–6) and Trem2-KO (n = 6–7) hearts at SS and 3 d after CLP. Every symbol represents a mouse (g,h). Data are presented as mean ± s.e.m.; two-sided P values were determined by unpaired t-test (g) and one-way ANOVA with Sidak’s multiple comparisons test (h). Results represent three independent experiments (c,d,g,h). See also Extended Data Fig. 5. Source data

Fig. 5. TREM2 promotes the uptake of cardiomyocyte-derived mitochondria by Mac1 in septic heart.

a, Immunofluorescence images and 3D reconstruction showing presence of Tom20⁺ material in cardiomyocyte-derived exophers from hearts of septic Card^RED mice. b, Presence of Tom20⁺ material in cardiomyocyte-derived exophers from hearts of Card^RED mice at SS and 3 d after CLP. Graph showing exopher numbers per field of view (FOV); n = 6 per group. c, TEM image of a mononuclear cell taking up mitochondria by extended pseudopods in septic hearts (CLP 7 d). d, Images showing cardiomyocyte-derived exophers (red) containing mitochondria (Tom20, white) present in TREM2⁺ macrophages (green) from hearts of Card^RED mice (n = 4). e, Cardiomyocyte-derived mitochondria (mt-Dendra2, green) present in TREM2⁺ macrophages (red) from hearts of MitoCard mice (n = 4). f, Cardiomyocyte-derived mitochondria (mt-Dendra2, green) localized in lysosomes (LAMP1, white) of TREM2⁺ macrophages (red) in hearts of MitoCard mice (n = 3). g, TREM2⁺ macrophages (green) took up cardiomyocyte-derived mitochondria (mt-Keima-458, cyan) and some mitochondria in an acidic environment (mt-Keima-561, red) from hearts of AAV9-Tnnt2-mt-Keima infected mice. h, CD163⁺ macrophages (green) engulfed cardiomyocyte-derived mitochondria in hearts of WT and Trem2^−/− mice infected with AAV9-Tnnt2-mt-Keima, respectively. Graph showing percentages of mt-Keima⁺ mitochondria in CD163⁺ macrophages (n = 5 per group). For each animal, we randomly selected five visualization areas and five Mac1 cells were analyzed. The ratio of the area of mt-Keima mitochondria in a single Mac1 cell to the area of the same Mac1 cell was calculated. The ratio in the KO group was normalized to the ratio of the control group. i, The incorporation of cardiomyocyte-derived tdTomato protein in CD45⁺CD11b⁺F4/80⁺CD163⁺RETNLA⁺ macrophages and CD45⁺CD11b⁺F4/80⁻LY6C⁺ monocytes from WT → Card^RED or KO → Card^RED chimeras 7 d after CLP. Graph showing quantification of MFI of tdTomato (n = 6–7 per group). j, Presence of Tom20⁺ material (cyan) in cardiomyocyte-derived exophers (red) from hearts of WT → Card^RED and KO → Card^RED chimeras 7 d after CLP. Graph showing the exopher numbers per FOV (n = 6 per group). Scale bars are indicated in the images. Bars show as mean ± s.e.m. (b,h,j) and median with interquartile range (i). Two-sided P values were determined by unpaired t-test (b,h,j) and Mann–Whitney U-test (i). Results represent four (b), three (d–f) and two (g–j) independent experiments (Extended Data Figs. 6 and 7). Source data

Fig. 6. Trem2 deficiency exacerbates the cardiac dysfunction following sepsis.

a, Survival analysis of WT (n = 18) and Trem2^−/− (n = 30) mice after CLP performance was monitored for 7 d. b,c, WT and Trem2^−/− mice were subjected to CLP and cardiac function was examined at SS and 3, 7 and 21 d after CLP. Graph showing EF % (b) measured by echocardiography (n = 7–8 mice for each group) and the levels of cTnI (c) in the serum (n = 5–8 mice for each group). d, Representative M-mode echocardiography images and EF % of WT → WT chimeras (Trem2^+/+Ch) and Trem2^−/−→WT chimeras (Trem2^−/−Ch) 7 d after CLP. e, Representative continuous-wave Doppler echocardiography images and E/A ratio of Trem2^+/+Ch and Trem2^−/−Ch chimeras 7 d after CLP. f, Graph showing CO measured by echocardiography. g,h, Graphs showing the levels of cTnI (g) and LDH (h) in the serum of Trem2^+/+Ch and Trem2^−/−Ch chimeras 7 d after CLP (d–h, n = 11 mice for each group). i, Graphs showing protein levels of ANP and BNP in the serum of Trem2^+/+Ch and Trem2^−/−Ch chimeras 7 d after CLP. j, Graphs showing protein levels of tumor necrosis factor (TNF)-α, IL-1β, IL-6 and CCL2 in the heart tissues of Trem2^+/+Ch and Trem2^−/−Ch chimeras 7 d after CLP (i,j, n = 10 mice for each group). k, Representative TEM images (left) and mitochondria injury score (right) of Trem2^+/+Ch and Trem2^−/−Ch chimeras 7 d after CLP. Scale bars, 5 μm. Bar graph showing mitochondria injury score per FOV. Each symbol represents a FOV of TEM images. Each group has four mice and five FOVs were randomly selected for assay from each animal. l,m, Graphs showing the levels of ATP (l) in heart tissue lysates and the levels of serum lactate (m) of Trem2^+/+Ch and Trem2^−/−Ch chimeras 7 da after CLP (l,m, n = 11 mice for each group). Each symbol represents one animal (b–j,l,m). Bars show as mean ± s.e.m. (b–f,h–l) and median with interquartile range (g,m). Two-sided P values were determined by Kaplan–Meier log-rank test (a), two-way ANOVA with Sidak’s multiple comparisons test (b,c), unpaired t-test (d–f,h–k) and Mann–Whitney U-test (g,m). Results represent at least four (b,c) and three (d–m) independent experiments (Extended Data Fig. 8). Source data

Fig. 7. Intrapericardial administration of TREM2^hi Mac1 cells protects SICM.

a, Schematic illustration of TREM2^hi Mac1 cells transplantation in WT mice. TREM2^hi Mac1 (CD45⁺CD11b⁺F4/80⁺CD163⁺RETNLA⁺) and non-Mac1 (CD45⁺CD11b⁺ F4/80⁺ and non-CD163⁺RETNLA⁺) cells isolated from WT mice were mixed with MG and transplanted into the pericardial cavity of WT mice (2 × 10⁵ cells per animal) immediately after CLP. Control mice were injected with MG only. Mice were killed and analyzed 3 d after the transplantation. b–e, Representative M-mode echocardiography images (b) and graphs showing EF % (c), FS % (d) and CO (e) measured by echocardiography (b–e, +MG, n = 5 mice; +non-Mac1, n = 9 mice; +Mac1, n = 10 mice). f,g, Graphs showing levels of cTnI and LDH in the serum. h, Graphs showing mRNA levels of Anp and Bnp in the heart tissues. i, Graphs showing protein levels of ANP and BNP in the serum. j, Graphs showing mRNA levels of Tnfα, Il1β, Ccl2 and Il6 in heart tissues. k, Graphs showing protein levels of TNF-α, IL-1β, IL-6 and CCL2 in heart tissues. l, Representative TEM images (left) and mitochondria injury score (right). Scale bars, 5 μm. Each symbol represents a FOV. Each group has four mice and five FOV were randomly selected for assay from each animal. m, Graph showing the levels of ATP in heart tissue lysates. n, Graph showing the levels of serum lactate. (f–k,m,n, +MG, n = 5 mice; +non-Mac1, n = 9 mice; +Mac1, n = 9 mice). Each symbol represents one animal in c–k, n, m. Bars show as mean ± s.e.m. Two-sided P values were determined by one-way ANOVA followed by Games-Howell’s (c–h,j,n,m) or Tukey’s (i,k) multiple comparisons test. Results represent four independent experiments (b–n). See also Extended Data Figs. 9 and 10. Source data

Extended Data Fig. 1. The dynamic changes of cardiac function, injury biomarkers and inflammatory cytokines in SICM, related to Fig. 1.

a-d, Representative echocardiography images (a) and EF% (b), FS% (c), and CO (d) measured by echocardiography in Wild-type (WT) mice at steady state (SS), 1, 3, 7, and 21 days after CLP (SS, n = 6 mice; CLP 3 d, n = 6 mice; CLP 7 d, n = 5 mice; CLP 21 d, n = 6 mice). e,f, Graphs showing the levels of cTnI (e) and LDH (f) in the serum. g,h, Graphs showing mRNA levels of Anp (g), Bnp (h) in the heart tissues; n = 5-6 (e,f,g,h) mice for each treatment. i, Heat map showing the scaled expression of 14 selected cytokines (row) across 22 heart tissue lysate samples (column), colored by different conditions. In b-h, each symbol represents one animal and data are presented as mean ± SEM. Two-sided P values were determined by one-way ANOVA followed by Games-Howell’s (b-d,f) or Tukey’s (g) multiple comparisons test, Kruskal–Wallis with Dunn’s multiple comparison test (e,h). Exact P values are shown. Results represent four independent experiments (b-h). Source data

Extended Data Fig. 2. Single-cell characterization and flow cytometric analysis of the cardiac immune cells during sepsis progression, related to Fig. 1.

a, Violin plots of average unique molecular identifier (UMI) (left) and gene (right) numbers of scRNA-seq at different stages of sepsis. Each dot represents a cell. b, UMAP plots showing the expression of selected marker genes for the defined cell types in Fig. 1c. c, Violin plots showing the expression of selected marker genes in all defined cell types corresponding to Fig. 1c. d, Gating strategy of 10-color flow cytometry for identifying various types of immune cells in murine hearts. e, Graph showing percentages of cardiac immune cells during sepsis progression (n = 4 - 5 mice for each treatment). Each symbol represents one animal in e. Data are presented as mean ± SEM. Two-sided P values were determined by one-way ANOVA followed by Tukey’s multiple comparisons test, n.s., not significant. In e, experiments were performed four times. Source data

Extended Data Fig. 3. Identification of Mac1 subset in the septic heart, related to Fig. 2.

a, UMAP plots showing the expression of selected marker genes for the monocyte-macrophage subpopulations. b, Gating strategy to identified Mac1 cells (CD45⁺CD11b⁺F4/80⁺CD163⁺RETNLA⁺). c, Graph showing percentages of CD163⁺RETNLA⁺ cardiac macrophages during sepsis progression (SS, n = 8 mice; CLP 3 d, n = 8 mice; CLP 7 d, n = 9 mice; CLP 21 d, n = 10 mice). Each symbol represents one animal. Bars show mean ± SEM. Two-sided P values were determined by one-way ANOVA with Games-Howell’s multiple comparisons test. d, Representative immunofluorescence images showing the presence of CD163 (green), CD68 (red), and nuclei (blue) in cardiac tissues during sepsis progression. n = 6 mice per group. Scale bars, 20 μm. Exact P values are shown. Results represent four independent experiments (c,d). SS, steady state. Source data

Extended Data Fig. 4. Developmental trajectory and lineage tracing track the fate of Mac1 subset in SICM, related to Fig. 3.

a, Monocle prediction of monocyte-macrophage compartment developmental trajectory with pseudotime mapped along. b, Cluster-defining gene expression plotted along the pseudotime, with Seurat’s cluster mapped along. c, Schematic illustration of lineage tracking Mac1 cells in Cx3cr1^CreERT2:Rosa26^tdTom mice. d, Representative contour plots showing the expression of tdTomato (CX3CR1) on gated CD115⁺CD11b⁺ blood monocytes, CD45⁺CD11b⁺F4/80⁺ cardiac macrophages, CD45⁺CD11b⁺F4/80⁺CD163⁺RETNLA⁺ macrophages and graph showing percentages of tdTomato-labeled cells in those cells (0 week, n = 4 mice; 1 week, n = 5 mice; 3 weeks, n = 5 mice; 5 weeks, n = 2 mice; 6 weeks, n = 3 mice). Bars are mean ± SEM. e, Representative contour plots showing the expression of Ki67 on gated CD45⁺CD11b⁺F4/80⁺CD163⁺RETNLA⁺ macrophages and graph showing percentages of Ki67 expressing in CD45⁺CD11b⁺F4/80⁺CD163⁺RETNLA⁺ macrophages at steady state (SS), 3 days and 7 days after CLP. Each symbol represents one animal. Data are presented as mean ± SEM. Two-sided P values were determined by one-way ANOVA with Games-Howell’s multiple comparisons test. Exact P values are shown. Results represent five (d) and two (e) independent experiments. Source data

Extended Data Fig. 5. TREM2 is highly expressed on Mac1 subset and essential for Mac1 remodeling during SICM, related to Fig. 4.

a, Volcano plots showing the gene expression fold change (x-axis, log2 scale) and p-value significance (y-axis, -log10 scale) in Mac1 subset of steady state (SS) versus Mac2 subset (left), Mac3 subset (middle) or Mac4 subset (right) of 3 days after CLP. Selected significant genes were labeled. Two-sided P values were determined by the Mann-Whitney U test. b, Violin plots showing the expression of Trem2 in all defined immune cell types corresponding to Fig. 1c. c, Graph showing the quantification of mean fluorescence intensity (MFI) of TREM2 expression in CD45⁺CD11b⁺F4/80⁺CD163⁺RETNLA⁺ and CD45⁺CD11b⁺F4/80⁺(non-CD163⁺RETNLA⁺) cardiac macrophages at steady state and 7 days post-CLP (n = 8 mice for each group). Each symbol represents one animal. Data are presented as mean ± SEM. Two-sided P values were determined by Kruskal–Wallis with Dunn’s multiple comparison test. d, UMAP plots showing cell clustering results of the monocyte-macrophage compartment from hearts of WT littermate controls and Trem2-KO mice at steady state and 3 days after CLP. Monocyte-macrophages under each condition are downsampled to 1000 cells. The subpopulations obtained by Seurat were identical to those previously identified in Fig. 2a according to their marker genes. e, Bar plots showing cluster distribution of the monocyte-macrophage subsets in WT littermate controls and Trem2-KO mice at steady state and 3 days after CLP. f, Graph showing percentages of cardiac immune cells in WT controls and Trem2-KO mice at 0, 3 and 7 days post-CLP (n = 4-5 mice for each treatment). Every symbol represents a mouse. Data are presented as mean ± SEM, Two-sided P values were determined by Mann-Whitney U test, n.s., not significant. Results represent three independent experiments (c,f). Source data

Extended Data Fig. 6. Sepsis induces extracellular damaged mitochondria accumulation and TREM2^hi Mac1 cells eliminate cardiomyocyte-derived mitochondria during sepsis, related to Fig. 5.

a, TEM images showing vesicles in hearts of WT mice 3 days post-CLP. b, TEM images showing cardiac interstitial mitochondria vesicles (red arrowheads) of WT mice at steady state (SS) or 3 days post-CLP. Bar graph showing free mitochondria numbers per field of view (FOV). Each symbol represents a FOV. Five fields of view are randomly selected for assay from each animal (n = 3). c, Schematic illustration of the gating strategy for collecting vesicles from hearts of mice. d, Flow cytometric analysis of purified vesicles stained with the mitochondria-selective probes MitoTracker Green and MitoNIR (n = 3 per group). e, Mitochondrial membrane potential in cardiac vesicles and cultured fibroblasts as assessed by MFI levels of MitoNIR in basal conditions and in the presence of depolarizing (FCCP) or hyperpolarizing (Oligomycin) agents (n = 3-4 per group). f, Pseudo-colored representative TEM images of two mononuclear cells (grey) taking up vesicles containing mitochondria (brown) from cardiomyocytes (green) of WT mice 7 days post-CLP. g, The incorporation of cardiomyocyte-derived tdTomato protein in CD45⁺CD11b⁺F4/80⁺CD163⁺RETNLA⁺ macrophages, CD45⁺CD11b⁺F4/80⁺ (non-CD163⁺RETNLA⁺) macrophages and CD45⁺CD11b⁺F4/80^-LY6C⁺ monocytes from αMHC^Cre/+ (control) and αMHC^Cre/+:Rosa26^TdTom (Card^RED) mice at SS and 7 days post-CLP (control mice, n = 4; Card^RED SS, n = 5; Card^RED CLP 7d, n = 8). h, The incorporation of cardiomyocyte-derived mt-Dendra2 protein in CD45⁺CD11b⁺F4/80⁺CD163⁺RETNLA⁺ macrophages, CD45⁺CD11b⁺F4/80⁺(non-CD163⁺RETNLA⁺) macrophages and CD45⁺CD11b⁺F4/80^-LY6C⁺ monocytes from αMHC^Cre/+ (control) and αMHC^Cre/+:mtD2^Flox/Flox (MitoCard) mice at SS and 7 days post-CLP (control mice, n = 4; MitoCard SS, n = 6; MitoCard CLP 7d, n = 6). i, Proportions of TUNEL⁺ and TUNEL⁺cTnI⁺ cells in hearts of WT mice at SS and 3 days post-CLP (n = 6 mice per group). All scale bars are indicated in the images. Data are presented as mean ± SEM (b,e,g,h,i). Two-sided P values were determined by Mann-Whitney U test (b), or by Kruskal–Wallis with Dunn’s multiple comparison test (e,g,h,i). Results represent three (a,b,d,g-i) and two (e,f) independent experiments. Source data

Extended Data Fig. 7. Trem2 deficiency impairs the uptake of cardiomyocyte-derived mitochondria by Mac1 subset in septic heart, related to Fig. 5.

a, Schematic illustration of the cardiomyocyte mitochondria labeled with AAV9-Tnnt2-mt-Keima virus. Right panel showing the mt-Keima fluorescence signals in hearts of WT mice at weeks 0, 3 and 6 after virus infection respectively. Keima-tagged mitochondria were indicated by different fluorescence in neutral (Keima 458 nm; green) and acidic (Keima 561 nm; red) environments. Scale bars, 20 μm. b, Scatter plots showing the average expression levels (log10(FPKM+1) scale) of total macrophage populations from WT littermate controls (y-axis) and Trem2^-/- mice (x-axis). c, Violin plots showing the selected endocytic gene expression in total macrophage from WT littermate controls and Trem2^-/- mice. d, Study protocol of bone marrow transplantation from WT or Trem2^-/- donors into Card^RED recipient mice. e,f, Representative TEM images showing cardiac interstitial mitochondria vesicles from WT and Trem2^-/- mice at steady state (f) and 7 days post-CLP (e), Scale bars, 5 μm. Bar graph showing the quantification of free mitochondria numbers per field of view (FOV). Each symbol represents a FOV (n = 3 mice for each treatment). Five fields of view are randomly selected for assay from each animal. Data are presented as mean ± SEM (e,f). Two-sided P values were determined by Mann-Whitney U test (e,f). Results represent two independent experiments (e,f). Source data

Extended Data Fig. 8. Trem2 deficiency exacerbates cardiac dysfunction in sepsis, related to Fig. 6.

a-i, WT and Trem2^-/- mice were subjected to CLP and cardiac function was examined at steady state (SS) and 3, 7 and 21 days post-CLP. Graph showing FS% (a); CO (b) measured by echocardiography; Graph showing the levels of LDH in the serum (c); Graphs showing mRNA levels of Anp (d), Bnp (e), Tnfα (f), Il-1β (g), Ccl2 (h) and Il-6 (i) in the heart tissue lysates (a-b, n = 7–8 mice; c, n = 6-8 mice; d-i, n = 6-7 mice for each treatment). j, Study protocol of bone marrow transplantation (BMT) from WT and Trem2^-/- donors (CD45.2 background) into WT recipients (CD45.1 background). k, Representative contour plots of CD45.1 and CD45.2 expressions on gated CD45⁺CD11b⁺F4/80⁺CD163⁺RETNLA⁺ Mac1 subset in recipient mice 1, 3, 5 and 8 weeks after BMT. l, Graph showing percentages of CD45.2⁺ cells in CD11b⁺F4/80⁺CD163⁺RETNLA⁺ Mac1 subset after BMT (n = 4-7 mice for each group).m, Representative contour plots of TREM2 expression on gated CD45⁺CD11b⁺F4/80⁺CD163⁺ macrophages in hearts of both WT→WT chimeras (Trem2^+/+Ch) and Trem2^-/-→WT chimeras (Trem2^-/-Ch) 8 weeks after BMT. Graph showing Trem2 mRNA levels in heart tissue lysates (n = 6 mice for each group). n, Graphs showing mRNA levels of Anp and Bnp in the heart tissue lysates 7 days post-CLP (n = 11 mice for each group). o, Graphs showing mRNA levels of Tnfα, Il1β, Ccl2 and Il6 in the heart tissue lysates 7 days post-CLP (n = 11 mice for each group). Each symbol represents one animal. All data are presented as mean ± SEM. Two-sided P values were determined by Mann-Whitney U test (a-i,m-o) or by Kruskal-Wallis with Dunn’s multiple comparison test (l). Results represent at least four (a-i,l) and three (m-o) independent experiments. Source data

Extended Data Fig. 9. Transplanted cardiac macrophages are detected by flow cytometry and immunofluorescence, related to Fig. 7.

a, TREM2^hi Mac1 cells (CD45⁺CD11b⁺F4/80⁺CD163⁺RETNLA⁺) and non-Mac1 (CD45⁺CD11b⁺F4/80⁺ and non-CD163⁺RETNLA⁺) cells were isolated from CD45.1 mice. Mac1 or non-Mac1 cells (2 × 10⁵ cells per mouse) were mixed with Matrigel (MG), and transplanted into the hearts of CD45.2 mice, respectively. Seven days after the transplantation, macrophages from hearts of recipients (CD45.2) were analyzed with flow cytometry assay. b,c, Graphs showing percentages of total Mac1 (b) and CD45.1⁺ Mac1 (c) cells within the CD45⁺ populations from hearts of recipients (CD45.2) at 1, 3, and 7 day(s) post the cell transplantation, respectively (n = 6 mice for each treatment). d,e, Graphs showing percentages of total non-Mac1 (d) and CD45.1⁺ non-Mac1 (e) cells within the CD45⁺ populations from hearts of recipients (CD45.2) at 1, 3, and 7 day(s) post the cell transplantation, respectively (n = 5-6 mice for each treatment). f,g, TREM2^hi Mac1 cells (CD45⁺CD11b⁺F4/80⁺CD163⁺RETNLA⁺) and non-Mac1 (CD45⁺CD11b⁺F4/80⁺ and non-CD163⁺RETNLA⁺) cells were obtained from the hearts of WT mice by FACS and stained by CMTMR. Then, both types of cells (2 × 10⁵ cells/mouse mixed with Matrigel) were respectively injected into the pericardial cavity of recipient mice immediately after CLP. f, Representative immunofluorescence images showing the transplanted cells in the epicardium (Epi), intermyocardium (Inter), and endocardium (Endo) of recipient mice at 3 days post the cell transplantation. CMTMR (red) signals indicating Mac1 or non-Mac1 cells; epicardial layer (E.L.); myocardium (Myo); inner layer (I.L.). Scale bar, 20 μm. g, Representative immunofluorescence images showing the transplanted CMTMR labeled Mac1 cells in MitoCard mice at 3 days post CLP. 3D reconstruction images of the dashed box area, illustrating that cardiomyocyte-derived mitochondria (mtDendra2⁺, green) presented in the transplanted cells (CMTMR⁺, red). Nuclei were shown in blue (DAPI). Scale bar, 10 μm. n = 6 mice per group (f,g). Bars show mean ± SEM, Two-sided P values were determined by one-way ANOVA with multiple comparisons (b-e). Results represent four (b-e) and two (f,g) independent experiments. Source data

Extended Data Fig. 10. Intrapericardial administration of WT Mac1 cells protects SICM in Trem2^−/− mice, related to Fig. 7.

a, Schematic illustration of Mac1 cells transplantation in Trem2^-/- mice. Mac1 (CD45⁺CD11b⁺F4/80⁺CD163⁺RETNLA⁺) cells isolated from WT and Trem2^-/- mice were mixed with Matrigel (MG), respectively, and then transplanted into the pericardial cavity of Trem2^-/- mice (2×10⁵ cells/animal) immediately after CLP. Control mice were injected MG only. Mice were analyzed 3 or 7 days after the transplantation. b-d, Graphs showing EF% (b), FS% (c) and CO (d) measured by echocardiography. e,f, Graphs showing the levels of cTnI (e) and LDH (f) in the serum. g,h, Graphs showing mRNA levels of Anp (g) and Bnp (h) in the heart tissue lysates. i,j, Graphs showing mRNA levels of Tnfα, Il1β, Ccl2 and Il6 in the heart tissue lysates at 3 (i) and 7 (j) days after CLP. k, Graph showing ATP content in heart tissue lysates. l, Graphs showing the levels of serum lactate. b-l, each symbol represents one animal (n = 6 mice for each treatment). All data are presented as mean ± SEM. Two-sided P values by Kruskal-Wallis with Dunn’s multiple comparison test, n.s., not significant. m, Schematic diagram showing TREM2^hi Mac1 phagocytoses cardiomyocyte-derived mitochondria protect the septic heart. scRNA-seq uncovers a cardiac macrophage subset that was highly expressing TREM2 and CD163; Sepsis stimulates the release of massively cardiomyocyte-derived mitochondria; TREM2^hi Mac1 cells scavenge defective mitochondria and protect the septic heart. Results represent two (b-l) independent experiments. Source data