CD4 T细胞通过IL-3与髓系TNF损伤肿瘤血管

Fig. 1: CD4+ T cells control tumor growth and induce perivascular myeloid clusters. (A) (top) Experiment schematic; (bottom) Tumor size in B6 mice injected with 0.5×106 MC38-GP cells one day after injection of the indicated number of naïve Smarta TCR transgenic CD4+ T cells (Smarta, red lines) or of PBS as a control (Ctrl, blue line). Data are presented as mean ± the standard error of the mean (s.e.m.) from one experiment (n = 4 mice per group), representative of three independent experiments. (B) (top) Experiment schematic; (bottom) Tumor growth in Rag1−/− mice injected with 0.5×106 MC38-GP cells (right flank) and 0.5×106 parental MC38 cells (left flank) one day after injection of 104 Smarta cells (red line) or PBS (Ctrl, blue line). Data are presented as mean ± s.e.m. from one experiment (n = 4 mice per group), representative of two independent experiments. (C) (top) Experiment schematic; (bottom) Tumor growth in Rag1−/− mice injected with 0.5×106 MC38-GP cells (right flank) and 0.5×106 I-Ab–/– MC38-GP (C10, left flank) cells one day after injection of 104 Smarta cells (Smarta, n = 5 mice) or PBS (Ctrl, n = 4 mice). Data are presented as mean ± s.e.m. from one experiment, representative of two independent experiments. (D) (top) Experiment schematic; (bottom) C10 tumor growth in indicated mice receiving PBS (Ctrl) or 104 Smarta cells (Smarta) on day 4 after injection of C10 cells in each flank. Data are presented as mean ± s.e.m. from one experiment (n = 8 tumors from 4 mice per group; two tumors per mouse), representative of two independent experiments. (E) C10 tumors from Ctrl-treated (n = 2 mice) and Smarta-treated Rag1−/− mice (n = 4 mice) were processed on day10 p.t.i. using the 10x Genomics FFPE Visium Spatial technology. (left) Scaled expression of the classically activated macrophage gene signature (color scale at bottom, Table S2). (right) Violin plot of scaled expression of the signature in spots from Ctrl- and Smarta-treated tumors. Data shown are pooled from two independent experiments(see Methods). (F) C10 tumor growth in Rag1−/− (round symbols) or tKO (square symbols) mice receiving PBS (Ctrl, blue symbols) or 1×104 Smarta cells (Smarta, red symbols) on day 4 p.t.i. Ctrl: Rag1−/− and tKO (8 tumors from 4 mice for each genotype). Smarta: Rag1−/− (6 tumors from 3 mice), tKO (10 tumors from 5 mice). Data are presented as mean ± s.e.m. from one experiment, representative of four independent experiments. (G) Representative immunofluorescence staining (for CD45, CD31, CD3) of C10 tumors collected on day 10 p.t.i. in Rag1−/− mice injected with either PBS (Ctrl) or 104 Smarta cells (Smarta) on day 4 p.t.i. Dashed boxes (left) delineate regions of interest (ROI) magnified on the right. (H) Percentage of the tumor section occupied by peri-vascular clusters in experiments shown in (G). Each dot represents a tumor. Data are presented as mean ± s.e.m. from one experiment (n = 7 Ctrl tumors from 4 mice and n = 10 Smarta tumors from 5 mice), representative of at least three independent experiments. (I) High-magnification d10 p.t.i. immunofluorescence analysis, as described in (G), of C10 tumors growing in Rag1−/− mice or Rag1−/− Ccr2−/− Cx3cr1−/− (tKO) mice injected with Smarta cells on day 4 p.t.i. (J) Quantification of peri-vascular cluster areas in experiments shown in (I), as described in (H). Each dot represents a tumor. Data are presented as mean ± s.e.m. from one experiment (n = 4–8 tumors each group from 2–4 mice), representative of two independent experiments. (K) Spatial distribution of inflammatory (CD64+ CD86hi MHC-II+, cyan) and regulatory (CD64+ CD206+, magenta) macrophages in C10 tumors on d10 p.t.i. in Smarta-treated Rag1−/− mice. Peri-vascular cluster areas are shown as deep orange regions and are superimposed by macrophages colored as cyan (inflammatory) or magenta (regulatory). Each dot represents one cell. Two ROI defined on the left are shown at higher magnification on the right. (L) Spatial attraction analysis of inflammatory (top) and regulatory (bottom) macrophages to peri-vascular clusters, plotting the distribution of distances of macrophages to the boundaries of peri-vascular structures as shown in (K). Line ①, observed distribution; line ②, randomized distribution. Negative distances are assigned to cells within cluster boundaries. Data shown are representative of three independent experiments. (A–D, F) P values are calculated with a two-way analysis of variance (ANOVA). (H) P value calculated with a two-tailed, unpaired Student’s t test. (J) P values are calculated with one-way ANOVA. ns P > 0.05, *P < 0.05, **p <0.01, ***p <0.001, ****P < 0.0001.

Fig. 2: Tumor transcriptomic profiling. (A-C) scRNAseq analysis of CD11b+ NK1.1− CD45+ myeloid cells isolated from Ctrl- or Smarta-treated C10 tumors in Rag1−/− mice on day 10 p.t.i., sorted as in fig. S6A. Data are pooled from two captures derived from two independent experiments (see Methods). (A) UMAP plot showing cells from Ctrl- and Smarta-treated tumors. Each dot represents a cell and is color-coded by either condition (top) or SingleR-annotated cell type (bottom, as defined in fig. S6C). (B) UMAP plots showing cells from Ctrl- and Smarta- treated tumors, color-coded by groups defined by unsupervised clustering. (C) Row z-scored expression of selected genes of L10 signature among mono-mac groups; color scale on the right. (D-J) Xenium analysis of Ctrl- and Smarta-treated C10 tumors in Rag1−/− mice on day 10 p.t.i. Data shown are pooled from two independent experiments, with a total of n = 2 mice per group (see Methods). (D) Row z-scored expression (color scale on the right) of indicated genes among cell groups defined by unsupervised clustering of single-cell assigned gene expression, performed on the whole data set (4 tumors from 4 mice). Group identity is shown in fig. S7B; “others” refers to pooled cells from all groups not individually analyzed in the heatmap. (E) Frequency of cells in each group within the indicated sample, relative to the total cellularity in each condition. (F) Pearson correlation (color scale on the right) between Mono-Macrophage groups fold change vectors across CD11b+ NK1.1− CD45+ myeloid scRNAseq and Xenium datasets, computed as in Ref. (33). (G) Row z-scored expression (color scale on the right) of group 13 signature genes among mono-mac groups (4, 5, 8, 13, 20) in Ctrl- and Smarta-treated tumors. (H, I) Percentage of cells from each group (y-axis) among cells within the indicated distance (x-axis) of Smarta cells (H) in all Smarta-treated tumors, or of endothelial cells (I) in Ctrl- or Smarta-treated tumors. Numbers and colors refer to group identity (fig. S7B). (J) Left panels show the spatial distribution of group 6, 17, 13 and 26 cells in one representative Ctrl- (top) and Smarta- (bottom) treated C10 tumor, with each dot representing a cell. The right panels provide a high-magnification view of the dashed box in left panels, showing inferred cell segmentation. Colors refer to group identity. (K) Survival probability of colon adenocarcinoma patients stratified by levels of the L10 gene signature score, normalized by FCGR1A (encoding CD64). Hazard ratio (HR) and P value of Cox regression fit are shown. Data is from the TCGA database. (L) Gene set enrichment analysis (GSEA) of the L10 signature (as defined in methods, Table S2) on genes ranked by log2 fold change between long-term and short-term survivors among a cohort of PPP2R1A-mutant, platinum-resistant ovarian cancer patients treated with combined anti–CTLA-4 and anti–PD-1/PD-L1 immunotherapy. NES (normalized enrichment score). P (adjusted p-value).

Fig. 3: Non-lymphocyte-derived TNF mediates CD4+ T cell-dependent tumor control. (A-D) scRNA-seq analysis of Smarta TIL in Rag1−/− mice on day 10 p.t.i., sorted as in fig. S9B. Data are pooled from two captures derived from two independent experiments (see Methods). (A) UMAP plots show TIL color-coded by groups defined by high-resolution clustering (left plot) or by scaled expression of indicated genes (right 5 plots). (B) Violin plots show single-cell expression of Tbx21 mRNA within Smarta TIL groups. (C) Growth of C10 tumors in indicated mice receiving PBS (Ctrl) or 104 Smarta cells of the indicated genotype. Data are presented as mean ± s.e.m. from one experiment (n = 8 tumors from 4 mice per group; two tumors per mouse), representative of two independent experiments. (D) Violin plots show scaled expression of indicated genes within Smarta TIL. (E) Schematic of the experiments shown in (F) and (H). (F) Growth of C10 tumors in Rag1−/− mice receiving (i) PBS (Ctrl, empty circles), or Ifng−/−Smarta cells (filled circles) and (ii) indicated antibody (red symbols) or a control IgG (purple symbols). Data are presented as mean ± s.e.m. from one experiment (n ≥ 5 mice per group), representative of at least two independent experiments for each treatment. (G) Growth of C10 tumors in Rag1−/− mice that, on day 4 p.t.i., received PBS (Ctrl, n ≥ 4 mice), 104 Ifng−/− Smarta cells processed for Crispr-Cas9 deletion of Il3 (gIl3, n = 10 mice), Tnf (gTnf, n = 7 mice) or Rosa26 (gRosa, as a control, n ≥ 6 mice). Data are presented as mean ± s.e.m.; data are pooled from two independent experiments for gIl3 and are from one experiment, representative of five independent experiments for gTnf. Nucleofection slightly but reproducibly affected tumor control by Smarta cells (as seen in an independent experiment in fig. S10B). (H) Growth of PyMT-GP tumors in Rag1−/− mice in the same experimental conditions as in (G). Data are presented as mean ± s.e.m. from one experiment (n ≥ 4 mice per group), representative of two independent experiments. (C, F-H) P values are calculated with a two-way analysis of variance (ANOVA). ns P > 0.05, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Fig. 4: IL3-producing CD4+ T cells induce TNF-producing peri-vascular macrophages to mediate tumor control. (A) Representative immunofluorescence staining (for TNF, CD45, CD64, and CD86) of C10 tumors collected on day 10 p.t.i. in Rag1−/− mice injected with either PBS (Ctrl) or 104 Smarta cells (Smarta) on day 4 p.t.i. A computationally generated image of TNF expression (TNF surface, second column) was generated from the acquired TNF staining signal (left column) and superimposed on CD45, CD64, and CD86 staining signals (right three columns). (B) Quantification of relative TNF cytokine abundance in Ctrl- and Smarta-treated tumors from experiments shown in (A). Each dot represents one tumor from one mouse. Data are presented as mean ± s.e.m. from one experiment (n = 4 mice per group), representative of two independent experiments. (C) Imaris spatial attraction analysis of TNF+ cells relative to peri-vascular cluster in one Smarta-treated tumor from the experiment shown in (B). Line ①, observed distribution of TNF+ cells. Line ②, randomized distribution. Data shown are representative of two independent experiments. (D) Representative immunofluorescence staining shows TNF production by CD64+ CD86+ macrophage-enriched peri-vascular clusters in Smarta-treated tumor collected on day 10 p.t.i. in Rag1−/− mice. (E) The percentage of indicated cells among TNF+ cells in Smarta-treated tumors from experiments shown in (D). Each dot represents one tumor from one mouse. Data are presented as mean ± s.e.m. from one experiment (n = 4 mice per group), representative of two independent experiments. (F) Average expression of Tnf (y-axis) in cells within the indicated distance (x-axis) from Il3+ (red line) or Il3− (cyan) Trbv13–1+ group 26 cells, in the tumor bed from Xenium- analyzed Smarta-treated tumors. (G-L) Analyses of C10 tumors collected on d10 p.t.i. from Rag1−/− mice injected on d4 p.t.i with 104 Ifng−/− Smarta cells that had been processed for Crispr-Cas9 deletion of Il3 (gIl3, n = 4 mice) or Rosa26 (gRosa, as a control, n = 5 mice). Data are from one experiment, representative of two independent experiments. (G) Representative immunofluorescence staining for TNF and CD45. An image of TNF expression (TNF surface, second column) was computationally generated from the acquired TNF staining signal (first column) and superimposed on CD45 staining signals (fourth column). (H) TNF staining mean fluorescence intensity (MFI) was measured on individual CD64+ TAMs within the peri-vascular clusters. Each dot represents one cell. Bars indicate median. (I) Percentage of TNFhi CD64+ TAMs among total CD64+ TAMs within peri-vascular clusters. TNFhi CD64+ TAMs are defined by TNF MFI greater than 5 in (H). Each dot represents one tumor from one mouse. Error bars indicate mean ± s.e.m. (J) TNF MFI of Smarta T cells. Each dot represents one cell. Bars indicate median. (K) Number of Smarta T cells per mm2. Each dot represents one tumor from one mouse. Error bars indicate mean ± s.e.m. (L) Percentage of the tumor section occupied by peri-vascular clusters. Each dot represents one tumor from one mouse. Error bars indicate mean ± s.e.m. (M) (left) Experiment schematic. C10 (middle) and PyMT-GP (right) tumor growth in mixed bone marrow chimeras made by reconstituting irradiated Rag1−/− hosts with a donor cell mix comprising tKO and either Rag1−/− (purple) or Rag1−/− Tnf−/− (red) precursors, and receiving either PBS (Ctrl, open blue circle) or 1×104 Ifng−/− Smarta cells (solid circles) on day 4 p.t.i. Data are presented as mean ± s.e.m. from one representative experiment of three independent experiments (C10, n = 5–7 mice per condition per experiment; one tumor per mouse), or from one experiment (PyMT-GP, n = 6–8 tumors from 3–4 mice per condition, two tumors per mouse). (B, H-L): P values are calculated with two-tailed, unpaired Student’s t-tests. (M) P values are calculated with a two-way analysis of variance (ANOVA). ns P > 0.05, *P < 0.05, **P < 0.01, ****P < 0.0001.

Fig. 6: CD4+ T cell-induced, TNF-dependent vascular damage. (A) (top) Experiment schematic; (bottom) Representative IBEX TER-119, CC3 and CD31 immunofluorescence staining of C10 tumors from Smarta-injected Rag1−/− mice treated with either IgG or anti-TNF antibody. Dashed boxes (left images) define ROI shown on the right at higher magnification. (B) Quantification of erythrocyte extravasation, as defined in Fig. 5G, within indicated tumors. Each dot represents one tumor. Data are presented as mean ± s.e.m. from one experiment (Ctrl: n = 4 tumors from 2 mice; Smarta+IgG and Smarta+anti-TNF: n = 6 tumors from 3 mice per group), representative of two independent experiments. (C) Blood vessel width of indicated tumors, quantified as in Fig. 5E. Bars indicate mean; data are from one experiment (Ctrl: n = 4 tumors from 2 mice; Smarta+IgG and Smarta+anti-TNF: n = 6 tumors from 3 mice per group), representative of two independent experiments. (D) Percentage of CC3+ area within indicated tumors, expressed relative to the whole tumor area in the analyzed section. Each dot represents one tumor. Data are presented as mean ± s.e.m. from one experiment (Ctrl: n = 4 tumors from 2 mice; Smarta+IgG and Smarta+anti-TNF: n = 6 tumors from 3 mice per group), representative of two independent experiments. (E) Representative IBEX immunofluorescence images of C10 tumors from Smarta-injected Rag1−/− mice treated with either IgG or anti-TNF antibody, as in (A). Dashed boxes (left) define ROI shown magnified on the right. In the top (IgG) panel, ROI ① is a region with NG2-unbounded vessels, ROI ② a region with NG2-bounded vessels. Data shown are representative of two independent experiments. (F) Spatial attraction analysis of indicated cells to NG2-unbounded blood vessel, in Smarta cells and IgG treated tumor section shown in (E), computed across the whole tumor section using Imaris 10.0, and showing observed and randomized cell distributions as indicated. Data shown are representative of two independent experiments. (G) Left panel: representative area magnified from panel E (top, plain box) showing apoptotic endothelial cell and pericytes within IgG-treated C10 tumors that had received Smarta cells. Right panels: percentage (i) of apoptotic (CC3+) endothelial cells (left) or pericytes (right), expressed relative to the total number of each cell, and (ii) of NG2- unbounded vessel area (CD31+ NG2−), expressed relative to total blood vessel area (CD31+) (far right). Percentages were computed on the whole tumor section from indicated mice, in experiments as described in (E). Each dot represents one tumor. Data are presented as mean ± s.e.m. from one experiment (Ctrl and Smarta+IgG: n = 4 tumors from 2 mice per group; Smarta+anti-TNF: n = 6 tumors from 3 mice), representative of two independent experiments. (B-D, G) P values are calculated with a one-way analysis of variance (ANOVA). ns P > 0.05, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Fig. 7: CD4+ T cell-induced TNF causes tumor cell death and targets non-immune stromal cells. (A) Representative immunofluorescence staining for tomato lectin, CD31, TER-119, and cleaved caspase-3 (CC3) in C10 tumors harvested on day 10 p.t.i. from Rag1−/− mice injected with either PBS (Ctrl) or 1x 104 Ifng−/− Smarta cells (Smarta) on day 4 p.t.i. Mice received Tomato-lectin perfusion 30 min before tumor harvest. (B) Quantification of vascular integrity and erythrocyte (RBC) extravasation shown in (A). Shown are the percentage of lectin+ CD31+ endothelial cells relative to total CD31+ cells (top left) and the area of lectin− CD31+ regions relative to the total tumor area (bottom left) in each condition; the area percentage of lectin− CD31+ regions (top right) and erythrocyte (RBC) extravasation (bottom right), as defined in Fig. 5G, within CC3+ and CC3− tumor regions in Ifng–/– Smarta-treated C10 tumors. Each dot represents one tumor. Data are presented as mean ± s.e.m. from one experiment (n = 6 tumors from 3 mice per group), representative of two independent experiments. (C) Representative immunofluorescence images of Smarta-treated C10 tumors collected from Rag1−/− mice on day 10 p.t.i., and stained for TER119 (red blood cells, RBC), CD31 (endothelium), cleaved caspase 3 (CC3), HIF1A, and Thy1.1 (marking C10 tumor cells). Dashed boxes define ROI in each section, shown on the right at higher magnification. Data shown are representative of two independent experiments. (D–E) Overlap between HIF1A+ and CC3+ areas in the full tumor section shown in (C). (D) shows the percentage of HIF1A+ areas overlapping with CC3+ areas, and (E) shows the reverse. Each dot represents one tumor from one mouse. Data are presented as mean ± s.e.m.; pooled from two independent experiments, with a total of n = 6 mice. (F) Spatial attraction analysis of HIF1A+ (left) and CC3+ cells (right) to extra-vascular RBC, computed across the whole tumor section using Imaris 10.0. Data shown are representative of two independent experiments. (G-K) Analyses on C10 tumors growing in bone marrow chimeras generated by injecting Rag1−/− precursors into Rag1−/− or Rag1−/− Tnfrsf1a−/− irradiated hosts, and that had. Received either PBS (Ctrl) or 1×104 Ifng−/− Smarta cells on day 4 p.t.i. (G) Experiment schematic. (H) Tumor growth in indicated conditions. Data are presented as mean ± s.e.m. from one experiment (n = 4–6 mice per group), representative of two independent experiments. (I) Representative TER-119, CC3, and CD31 IBEX staining. Dashed boxes (left images) define ROI shown on the right at higher magnification. Data shown are representative of two independent experiments. (J) Quantification of RBC extravasation, as defined in Fig. 5G, within indicated tumors. Each dot represents one tumor from one mouse. Data are presented as mean ± s.e.m. from one experiment (n = 2–5 mice per group), representative of two independent experiments. (K) Blood vessel width of indicated tumors, quantified as in Fig. 5E. Bars indicate mean; data are from one experiment (n = 2–5 mice per group), representative of two independent experiments. (B) P values are calculated with a two-tailed, unpaired Student’s t test. (H) P values are calculated with a two-way analysis of variance (ANOVA). (B, J, K) P values are calculated with a one-way analysis of variance (ANOVA). ns P > 0.05, *P < 0.05, **P < 0.01, ****P < 0.0001.

Fig. 8: TNF-dependent, IL3-dependent CD4+ T cell tumor control is seen with diverse T cell populations, tumor types, and antigens (A) Tumor size in indicated mice injected with 0.5 × 106 C10 tumor cells one day after injection of 0.5 × 106 wt Smarta cells (middle), Ifng–/– Smarta cells (right), or of PBS as a control (Ctrl, left). Data are presented as mean ± s.e.m.; pooled from three independent experiments, with a total of n = 6–16 mice per group. (B) MC38 tumors were collected on day 15 p.t.i. from Rag1–/– mice that on day 4 p.t.i had received either PBS (Ctrl) or 5 × 106 naive Ifng–/– CD4+ T cells. Left panel: Representative IBEX staining of tumors for TER-119, cleaved caspase-3 （CC3） and CD31. Right panel: Quantification of RBC extravasation, as defined in Fig. 5G, in indicated tumors. Each dot represents one tumor. Data are presented as mean ± s.e.m. from one experiment (n = 6 tumors from 3 mice per group), representative of two independent experiments. (C) (top) Experiment schematic; (bottom) Growth of B16 tumors in Rag1–/– mice receiving (i) PBS (Ctrl, empty circles), or 1×104 Trp1 TCR–transgenic CD4 cells (Trp1 Tg CD4, filled circles), and (ii) IgG (black symbols), αIFNγ plus αTNF antibodies (red symbols) or αIFNγ antibody alone (purple symbols). Data are presented as mean ± s.e.m. from one experiment (n = 4–5 mice per group), representative of two independent experiments; (D) Left panel: Representative IBEX staining for TER-119, CC3, and CD31 in B16 tumors collected on day 12 p.t.i. from Rag1–/– mice treated as in C. Right panel: Quantification of erythrocyte (RBC) extravasation, as defined in Fig. 5G, in the indicated tumors. Each dot represents one tumor from one mouse. Data are presented as mean ± s.e.m.; pooled from two independent experiments, with a total of n = 3 mice (Ctrl), n = 7 mice (Trp1 Tg CD4 + αIFNγ), and n = 8 mice (Trp1 Tg CD4 + αIFNγ + αTNF). (E) (top) Experiment schematic; (bottom) Growth of B16 tumors in Rag1–/– mice that, on day 4 p.t.i., received PBS (Ctrl, n = 5 mice), 5×105 Trp1 transduced Ifng–/– CD4 T cells processed for Crispr-Cas9 deletion of Il3 (gIl3, n = 4 mice), or Rosa26 (gRosa, as a control, n = 4 mice). Data are presented as mean ± s.e.m. from one experiment, representative of two independent experiments. (F) H&E staining (far left) and spatial distributions of group Mono_Mac_1, vascular cell groups (Endo_1, Endo_2, Endo_3, and pericytes; left), group tumor_2 (Hypoxia associated cells), and CD4 T cells [defined as (CD40LG | CD4) > 0 and FOXP3 = 0 within group Lymphocytes; right] in a representative human ovarian tumor (patient P1). Each dot represents a single cell. Far-right panel shows high-magnification view of the boxed region in the preceding panel. Cells are defined by computationally-inferred segmentation and colored as indicated, based on cell group identity defined in fig. S16AB. (G) Volcano plot [x-axis: average log2 fold change; y-axis: −log10–transformed adjusted P value (false discovery rate, FDR)] showing differentially expressed genes between Mono_Mac_1 cells located near CD4 T cells (distance < 10 μm) and those located farther from CD4 T cells (distance ≥ 10 μm) in all three patients. Selected genes are annotated. Red, blue, and gray symbols indicate genes preferentially expressed in Mono_Mac_1 cells near CD4+ T cells, far from CD4+ T cells, or showing no significant difference, respectively. P values were calculated using the Wilcoxon rank-sum test implemented in the Seurat FindMarkers function. (H) Left panels show the spatial distribution of group 9 (Hypoxia associated tumor cells, purple), group 26 (Smarta cells, red), group 13 (mono-mac cells, green), and vascular cell groups (groups 6 and 17, shown together as gray) in one representative Ctrl- (top) and Smarta- (bottom) treated mouse C10 tumor, with each dot representing a cell. The right panels provide a high-magnification view of the areas boxed in left panels, showing inferred cell segmentation. Group identities are defined in fig. S7B, and sections are the same as shown in Fig. 2J. (A, C, E) P values are calculated with a two-way analysis of variance (ANOVA). (B) P values are calculated with a two-tailed, unpaired Student’s t test. (D) P values are calculated with a one-way analysis of variance (ANOVA). ns P > 0.05, *P < 0.05, ****P < 0.0001.