Fig 2: M1hot tumor-associatedmacrophages (TAMs) support tissue-resident memory T cells (TRM) cell maintenance by allowing uptake of fatty acids. (A) Frequency of CD103high, CD103intermediate and CD103low tumors among those categorized by chemokine (C-X-C motif) ligand 9 (CXCL9) expression in TAMs (M1hot, M1intermediate, M1cold; numbers of patients above each bar). (B) Heatmap showing expression of genes related to Tissue Resident Memory features (TRM phenotype) in CD8+ T cells isolated from the same tumors from patients with non-small cell lungcancer (NSCLC) were M1hot TAMs and M1cold TAMs were isolated and studied. (C) Gene set enrichment analysis (GSEA) of 246 gene sets in the transcriptome of CD8+ T cells from patients with NSCLC with M1hot TAMs versus those with M1cold TAMs, presented as the running enrichment score. Statistical significance by Kolmogorov-Smirnov test. (D) Expression of FABP3, FABP4 and FABP5 in non-tumor-associatedmacrophages (NTAMs), M1hot TAMs and M1cold TAMs (log2 normalized counts). Statistical significance by ordinary one-way analysis of variance. (E) Macrophages derived from blood (BDM) treated with interfern (INF)-γ/lipopolysaccharides (LPS) (M1hot-like BDM, red) or interleukin (IL)-4 (M1cold-like BDM, blue) cultured with bodipy lipid probe for 30 min (n=7). Lipid uptake measured by flow cytometry and expressed as mean fluorescence intensity. Wilcoxon test, two-tailed. *P≤0.05 (F) M1hot-like BDM (red) or M1cold-like BDM (blue) were co-cultured with CD3+ cells and bodipy lipid probe for 30 min (n=6). Uptake of lipid represented as Mean fluorescence intensities (MFI) in TRM (CD8+CD103+), CD8+ (CD8+CD103-) and CD3+ (CD8-CD103-) cells was determined by flow cytometry. Wilcoxon test, two-tailed. *P≤0.05. ns, non-significant for p value >0.05, *p≤0.05, **p≤0.01 and ****p≤0.00001. Related to online supplementary figure S6 and tables S5, S6.

Fig 3: Single-cell RNA-sequencing (RNA-seq) reveals a tumor-associatedmacrophages (TAM) subset with dual M1 and M2 features. (A) t-SNE visualization of ~9000 CD19-CD20-CD56-CD3-CD45+ single cell transcriptomes obtained from two tumors and two matched lung samples. Each symbol represents a cell; color indicates Seurat clustering of cells identifying eight clusters. (B) Left, distribution of clusters in tumor and lung. Right, pie chart representing the relative proportions of cells in each tissue-resident memory T cells (TRM) cluster. (C) Single-cell RNA-seq analysis of genes differentially expressed between purified populations of TAM and non-tumor-associatedmacrophages (NTAM) (figure 1B) overlaid over the average expression of TAM-enriched and NTAM-enriched clusters. Horizontal breaks separate genes commonly upregulated or downregulated. Rows are clustered with average linkage. (D) Single-cell RNA-seq analysis of M1 and M2 associated signatures (figure 2A, B) as per (figure 3C). (E) Seurat normalized expression of indicated transcripts (cluster colored as per figure 3A), overlaid across the t-SNE plot, with expression levels represented by the color scale. (F) Percentage of cells expressing a given transcript in highlighted cluster, the median of the values from the other clusters (1, 5, 6, excluding 7) is shown. (G) WGCNA of the cluster 3 specific transcripts in 40 bulk tumor transcriptomes, visualized in Gephi the nodes are sized according to the number of edges (connections). (H) Confocal images of a representative TAM versus NTAM sample of a patient with non-small-cell lung cancer (NSCLC), showing co-expression of chemokine (C-X-C motif) ligand 9 (CXCL9) with matrix metallopeptidase 12 (MMP12). Scale bar refers to 10 µm. Related to online supplementary figure S3 and tables S1, S4.

Fig 4: Impact of treatment with urelumab variants on liver pathology. Humanized mice received either 3 or 6µg/g of a human IgG4 isotype control, urelumab (αCD137) or deglycosylated urelumab (αCD137 PNGaseF) and liver pathology was studied twenty days after antibody injection. (A, B) Representative hematoxylin/eosin stained liver sections (A) of mice treated with 3 or 6µg/g of human IgG4 isotype control, urelumab and deglycosylated urelumab variants (scale bar 100µm) and quantification of the mean infiltrate area (B) of immune cell infiltrates (hIgG4: n=3, αCD137: n=6, αCD137 PNGaseF: n=4). (C) Shown is the percentage of mice with detectable immune cell infiltrates in the liver. (D, E) Immunofluorescent staining of liver sections (scale bar 100µm) (D) and quantification of the infiltration (E) of CD3+ T cells (upper panel) and the relative amount of CD3+CD8+ T cells within the CD3+ T cell population (lower panel) (hIgG4: n=9, αCD137: n=7, αCD137 PNGaseF: n=7, n represents the number of analysed images of two independent mice per group). Statistical testing was performed by using a Shapiro-Wilk and Kruskal Wallis test with a Dunn’s multiple comparison test. *p<0.05; **p<0.01.

Fig 5: Expression of key regulators of cellular metabolism in T cells of patients with MDD(A–C) mRNA expression levels of CPT1a (A), SLC2A (encoding GLUT1) (B), and TNF (C) in T cells of patients with MDD (n = 25) compared with matched controls (n = 25). Gene expression is shown as fold change relative to housekeeping genes (TBP and IPO8). Individual patient data and median are shown. Wilcoxon signed-rank test was used to compare groups.(D-E) The numbers of metabolome-wide significant correlations (of 3,511 detected features) are shown for CD3 and CD14 cellular mitochondrial readouts (BR, basal respiration; ATP, ATP production; SR, spare respiratory capacity) (D). Top known metabolites significantly correlated with CD3 CPT1a expression (E). CPT1a, carnitine palmitoyltransferase 1 A; GLUT1, glucose transporter 1; TNF, tumor necrosis factor; HC, healthy controls; MDD, major depressive disorder.