Fig 2: Tumor‐infiltrating CD4+ TRMs correlate with patient prognosis and immunotherapy efficacy. (A) Representative IF staining images showing the localization of CD4 (orange), CD103 (red), and DAPI (nuclei, blue) in the tissue microarray. Scale bars, 25 µM. (B) Kaplan‒Meier overall survival (OS) curves stratified by the percentage of CD4+ TRMs in lung adenocarcinoma (LUAD, n = 90) (left) and lung squamous cell carcinoma (LUSC, n = 90) (right). High and low groups were defined using optimal cutoff values, and differences in OS were assessed using the log‐rank test. (C) Representative IF staining images showing the localization of CD4 (orange), CD103 (red), XCL1 (green), and DAPI (nuclei, blue) in the tissue microarray. Scale bars, 25 µm. (D) Kaplan‒Meier OS curves based on the percentage of XCL1+ CD4+ TRMs in LUAD (n = 90, left) and LUSC (n = 90, right). Optimal cutoff values were used to define high versus low groups, and statistical comparisons were performed using the log‐rank test. (E,F) Bar graphs showing the proportions of CD4+ TRMs among total CD4+ T cells across TNM stages in LUAD (E, n = 44, 26, and 20 for stages I, II, and III, respectively) and LUSC (F, n = 16, 27, and 18 for stages I, II, and III, respectively). (G,H) Bar graphs showing the frequency of XCL1+ CD4+ TRMs among CD4+ T cells across TNM stages in LUAD (G, n = 44, 26, and 20 for stages I, II, and III, respectively) and LUSC (H, n = 16, 27, and 18 for stages I, II, and III, respectively). (I,J) Bar graphs comparing the proportions of CD4+ TRMs (I) and XCL1+ CD4+ TRMs (J) among total CD4+ T cells from paired normal (left) and tumor tissues (right) from LUAD (n = 90) and LUSC (n = 90). (K,L) Kaplan‒Meier OS curves based on intratumoral CD4+ T‐cell abundance in LUAD (n = 90, left) and LUSC (n = 90, right). Optimal cutoff values were used to define high versus low groups, and statistical comparisons were performed using the log‐rank test. (M,N) Proportions of CD4+ TRMs among total CD4+ T cells (M) and the log2(fpkm+1) of JAML and XCL1 expression (N) in pre‐treatment, responsive, and non‐responsive groups. (O) Volcano plots presenting the DEGs between the responsive groups and non‐responsive groups. P value < 0.05, |LogFC| > 0.1. (P) Frequencies of CD4+ T cells among total T cells in pre‐treatment, responsive, and non‐responsive groups. (Q,R) Correlation analyses showing the relationship between CD4+ TRMs within CD4+ T cells (Q) or cDC1 within cDCs (R) and the pathological residual tumor rate following anti‐PD‐1 immunotherapy (n = 23). Spearman correlation coefficients (r) and P values are indicated. (S) Schematic model illustrating how tumor‐infiltrating CD4+ TRMs secrete XCL1 to recruit cDC1, which in turn promote CD8+ T‐cell‐mediated anti‐tumor responses; note that cDC1 recruitment is inhibited within the tumor microenvironment. This schematic was generated using BioRender (https://www.biorender.com). The data in (E–H, I,J, M,N, and P) are expressed as the mean ± SD. Statistical analyses were performed using unpaired Student's t‐tests (I, J) and one‐way ANOVA with Tukey's post hoc test (E–H, M, N, and P). ns, non‐significant; *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.

Fig 3: Functional Impairment of Tumor‐Infiltrating CD4+ TRMs. (A) Bulk RNA‐seq analysis revealing the heatmap of the mean log‐normalized (Z score‐standardized) expression of genes related to inhibition, tissue residency, costimulation, proliferation, migration, cytokines, Tfh‐associated genes, Treg‐associated genes, transcription factors and chemokines in the N‐TRM and T‐TRM. Differentially expressed genes (P value<0.05) are indicated by ‘+’ (upregulated) or ‘–’ (downregulated). (B) Volcano plot showing the results of the scRNA‐seq analysis of differentially expressed genes, plotted as the log fold change versus differences in the percentage of cells expressing each gene for TRMs in tumor versus normal tissues (adjusted P value<0.05, Wilcoxon rank‐sum test). (C–E) Gene set enrichment analysis (GSEA) of tumor versus normal CD4+ TRMs. (C) GO gene sets. (D) KEGG gene sets. (E) HALLMARK gene sets. Normalized enrichment scores (NESs) are shown. Red, gene sets enriched in tumor CD4+ TRMs; blue, gene sets enriched in normal CD4+ TRMs. (F) Bar plots showing the expression levels of tissue residency‐related genes (CD49a, CXCR6, Hobit, RUNX3, CD62L, S1PR1, CCR7, and ICAM2) between normal and tumor tissues using FCM analysis (n = 6 per group for all markers). (G) FCM analysis of XCL1 expression in CD4+ TRMs between normal (N‐TRM) and tumor (T‐TRM) tissues, with representative plots (left) and quantification (right) (n = 9). (H) Quantification of XCL1 expression (ng/mL) in N‐TRM and T‐TRM by ELISA (n = 9). (I) Multiplex immunofluorescence quantification of XCL1+ cells among CD4+ CD103+ T cells in paired normal and tumor tissues (n = 6). (J,K) Chemotaxis assay of cDC1 recruitment by CD4+ TRMs. (J) Representative FCM plots gated on CD45+ cells showing the cDC1 frequency in the N‐TRM group versus T‐TRM group. (K) Quantification of the cDC1 frequency among CD45+ cells in the lower chamber, comparing N‐TRM and T‐TRM (n = 10). (L) FCM analysis showing the percentages of cDC1 among cDCs between tumor and normal tissues (n = 16). (M) Multiplex immunofluorescence quantification of the spatial proximity between CD4+CD103+ T cells and cDC1 in normal and tumor tissues. Left, average distance from CD4+CD103+ T cells to the nearest cDC1; right, average distance from cDC1 to the nearest CD4+CD103+ T cell (n = 6). (N) Cell‒cell communication analysis of cDC1‐centric interactions in normal (left) and tumor (right) tissues. Bubble plots display outgoing signaling to cDC1 (x‐axis) and incoming signaling from cDC1 (y‐axis) across immune cell types, including CD4+ T cells, CD8+ T cells, NK/NKT cells, and other dendritic cell subsets. The bubble size reflects the interaction count, and the color denotes cell type classification. (O) Differential cDC1‐centered cell‒cell communication between tumor and normal tissues. Edges link cDC1 to partner immune subsets; edge width reflects the magnitude of change in aggregated interaction strength, and color indicates direction (red, higher in tumors; blue, higher in normal). (P) Dot‐heatmap depicting XCL1‐XCR1 and XCL2‐XCR1 signaling to cDC1 in tumor (T) and normal (N) tissues. Columns represent distinct ligand‐producing immune cell types (CD4+ T cells, CD8+ T cells, NK/NKT cells, and other dendritic cell subsets) engaging cDC1. (Q,R) Representative spatial transcriptomic analysis of CD4+ TRMs and cDC1 in normal (N) and tumor (T) tissues (Samples P25‐N and P25‐T). (Q) Spatial density maps showing the distribution of CD4+ TRMs and cDC1 across the tissue section. The color scale indicates local cell abundance. (R) Quantification of the mean nearest‐neighbor distances between CD4+ TRMs and cDC1 in both directions (n = 6). The data in (F–I, K–M, and R) are expressed as the mean ± SD. Statistical analyses were all performed using paired Student's t‐tests. ns, non‐significant; *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.

Fig 4: Tumor‐infiltrating CD4+ TRMs could recruit cDC1 toward TME. (A) KEGG pathway enrichment analysis comparing the TRMs and non‐TRM subsets. P value<0.05. (B) Bar plots depicting the log2(TPM+1) of XCL1 expression across CD4+ T‐cell subpopulations, as determined by Bulk RNA‐seq. (C) Representative FCM plots showing XCL1 staining in the T‐TRM, T‐SP, and T‐DN subsets. The fluorescence minus one (FMO) control is shown on the left. The percentages indicate XCL1+ cells within each subset. (D,E) Quantitative analyses of XCL1 expression in T‐TRM, TSP, and TDN cells by FCM (n = 8; D) and ELISA (n = 6; E). Data from each matched pair are connected by folded lines. (F,G) Gating strategy for the FCM analysis of the XCR1 expression in cDC1 and cDC2 in tumor tissues. Bar graph depicting the frequency of XCR1 expression in the cDC1 and cDC2 populations. n = 14. (H,I) Representative multiplex immunofluorescence images of normal lung tissue (H) and NSCLC tumor tissue (I) (Sample1) showing the colocalization of CD4 (cyan), CD103 (red), XCL1 (green), XCR1 (yellow), nuclei (DAPI, blue), and the merged images. Scale bars, 50 µM. (J,K) Multiplex immunofluorescence quantification. (J) Frequencies of XCL1+ cells among CD4+CD103+ and CD4+CD103− T cells. (K) Nearest‐neighbor distances (µM): left, CD4+ T cell to the nearest cDC1; right, cDC1 to the nearest CD4+ T cell; both comparisons of CD4+CD103− with CD4+CD103+ subsets. The paired lines denote matched samples. n = 6. (L,M) Chemotaxis assay of cDC1 recruitment by tumor‐infiltrating CD4+ T‐cell subsets (n = 7). (L) Representative FCM plots showing cDC1 among all cells in the lower chamber after migration toward T‐TRM, T‐SP, or T‐DN groups. (M) Quantification of the cDC1 frequency among live cells in the lower chamber, comparing the three CD4+ T‐cell subsets (n = 7). (N) FCM analysis revealing the Spearman correlation between cDC1 frequency and CD4+ TRM frequency in tumor tissues (n = 11). r, Spearman correlation coefficient; p, p value. (O) Representative spatial transcriptomic maps (Sample P11‐T) showing the density distribution of cDC1, CD4TRM, CD4TEM, CD4TCM, and CD4TN across the section. (P) Cell‒cell colocalization dot plot showing spatial associations between cDC1, NK/NKT cells, CD8+ T‐cell subsets, and CD4+ T‐cell subsets. (Q) Distribution of observed versus expected mean distances from CD4TRMs to cDC1s in tumor tissues, showing a significantly shorter observed distance (280 µm) compared with expected (1123 µM). (R) Cell‒cell mean distance matrix across all tumor samples. Distances (µm) were calculated from each source cell type (rows) to target cell types (columns). (S,T) Spearman correlation analysis of the scRNA‐seq data in tumor samples (n = 107). (S) cDC1 proportions among total cDCs versus CD4+ TRM proportions among CD4+ T cells. (T) cDC1 proportion among total cDCs versus average XCL1 expression in CD4+ TRMs. r, Spearman correlation coefficient; p, p value. (U) Cell‒cell interaction dot plot showing ligand‒receptor pairs between cDC1 and T‐cell subsets. Ligands are shown on the left, and source–target cell pairs are indicated on the top. Red circles denote significant interactions. (V) Representative FCM plots gated on CD45+ cells showing the cDC1 frequency in the lower chamber after chemotaxis with IgG control (left) or anti‐XCL1 blockade (right). (W) Quantification of the cDC1 proportion among CD45+ cells in the lower chamber for each condition (n = 6). The data in (B, D, G) are expressed as the mean ± SD. Statistical analyses were performed using paired Student's t tests (G, J, K, W) and one‐way ANOVA with Tukey's post hoc test (B‐E, M); ns, non‐significant; *p < 0.05; ** p < 0.01; *** p < 0.001; **** p < 0.0001.

Fig 5: JAML‐agonism in combination with PD‐1 blockade enhances CD4+ TRM‐mediated anti‐tumor immunity. (A) Experimental timeline: C57BL/6 mice were subcutaneously implanted with LLC cells and received i.p. injections of isotype IgG, anti‐PD‐1, anti‐PD‐1+recombinant CXADR, or anti‐PD‐1+JAML‐agonism on Days 7, 9, and 11 after tumor challenge. (B) FCM analysis of JAML expression in CD4+ TRMs. Left, representative plots (FMO, isotype, anti‐PD‐1) gated on CD4+ TRMs; right, quantification of JAML+ CD4+ TRMs (n = 10). (C,D) Multiplex immunofluorescence of LLC tumors across treatment groups (isotype, anti‐PD‐1, anti‐PD‐1+CXADR, and anti‐PD‐1+JAML‐agonism groups). (C) Representative images showing CD4 (red), CD103 (pink), XCL1 (green), nuclei (DAPI, blue) and merged images. Scale bar, 50 µm. (D) Quantification of XCL1+ CD4+ TRMs per mm2 field (n = 6). (E) FCM analysis of cDC1 within cDCs across treatment groups. Left, representative plots gated on cDCs showing the cDC1 frequency for isotype, anti‐PD‐1, anti‐PD‐1+CXADR, and anti‐PD‐1+JAML‐agonism groups; right, quantification of cDC1 among cDCs (n = 6). (F,G) Multiplex immunofluorescence of LLC tumors across treatment groups (isotype, anti‐PD‐1, anti‐PD‐1+CXADR, and anti‐PD‐1+JAML‐agonism). (F) Representative images showing CD8 (red), GZMB (yellow), PRF (green), nuclei (DAPI, blue), and merged images. Scale bar, 50 µM. (G) Quantification of GZMB+ CD8 T cells (left) and PRF+ CD8 T cells (right) per mm2 field (n = 6). (H) Macroscopic images of subcutaneous LLC tumors from mice treated with isotype, anti‐PD‐1, anti‐PD‐1+CXADR and anti‐PD‐1+ JAML‐agonism antibodies. (I) Tumor growth curves (left) and Kaplan‒Meier survival curves (right) for LLC‐bearing C57BL/6 mice among the four groups (n = 10). (J) Experimental timeline: C57BL/6 mice were orthotopically implanted with CMT‐167 cells and received i.p. injections of isotype IgG, anti‐PD‐1, anti‐PD‐1+recombinant CXADR, or anti‐PD‐1+JAML‐agonism antibodies on Days 7, 9, and 11 after tumor challenge. (K) In vivo IVIS Spectrum imaging of CMT‐167 orthotopic tumor‐bearing mice in the supine position at Day 50 post‐treatment with isotype IgG, anti‐PD‐1, anti‐PD‐1+CXADR, or anti‐PD‐1+JAML‐agonism antibodies. (L) Quantification of tumor bioluminescence by IVIS on Day 50, showing the average radiance (×106 p/s/cm2/sr) in the isotype, anti‐PD‐1, anti‐PD‐1+CXADR, and anti‐PD‐1+JAML‐agonism groups (n = 6). (M) Body weight and survival of CMT‐167 orthotopic tumor‐bearing mice. Left, Normalized body weight (%) over time in isotype, anti‐PD‐1, anti‐PD‐1+CXADR, and anti‐PD‐1+JAML‐agonism groups. Right, Kaplan‒Meier survival curves of the four treatment groups (n = 10). (N–P) FCM quantification of immune subsets in the CMT‐167 orthotopic model across treatment groups (isotype, anti‐PD‐1, anti‐PD‐1+CXADR, and anti‐PD‐1+JAML‐agonism groups). (N) Expression of XCL1 on CD4+ TRMs. (O) Percentage of cDC1 among cDCs. (P) Expression of GZMB (left) and perforin (right) on CD8+ T cells (n = 6). (Q) Experimental timeline: C57BL/6 mice were subcutaneously implanted with CMT‐167 cells and received i.p. injections of FTY720 (from Day 1, QOD), anti‐CD4 antibody (on Days 2–3, QD), adoptive transfer of WT or JAML‐KO CD4 TRMs (on days 8–10, QOD), and anti‐PD‐1 antibody (on Days 11, 13, and 15) after tumor challenge. Group assignment: All groups received FTY720 and anti‐PD‐1 antibodies. Group 0: control; Group 1: only WT CD4+ TRM transfer; Group 2: only JAML‐KO CD4+ TRM transfer; Group 3: anti‐CD4 plus WT CD4+ TRM transfer; Group 4: anti‐CD4 plus JAML‐KO CD4+ TRM transfer. (R) Macroscopic images of CMT‐167 tumors from mice in Groups 0–4, as defined in the experimental design. (S) Tumor growth curves (left) and Kaplan‒Meier survival analysis (right) of CMT167‐bearing mice across Group 0–4 (n = 10). (T) Representative multiplex immunofluorescence images of CMT‐167 tumors from Groups 0–4, showing CD4 (red), CD103 (yellow), XCL1 (green), XCR1 (white), nuclei (DAPI, blue), and merged images. Scale bars, 50 µm. (U) Average distance from CD4+CD103+ T cells to the nearest cDC1 (µM) in CMT‐167 tumors across Group 0–4. (V) Mean counts of cDC1 per mm2 field in CMT‐167 tumors across Group 0–4. Scale bar, 50 µM. The data in (B, D, E, G, L, N–P, and U,V) are expressed as the mean ±SD. Statistical analyses were performed using unpaired Student's t‐tests (B) and one‐way ANOVA with Tukey's post hoc test (D, E, G, L, N–P, and U,V). ns, non‐significant; *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.