Fig 2: Systemic osteopontin blockade restores immune control, and tumor cells are the predominant source of SPP1.(A) Myeloid cells, macrophages, and myeloid-derived suppressor cells in tumors from the indicated treatment groups (VT: n = 11; anti-SPP1: n = 4; B20 LH: n = 9; B20 HH: n = 10). One-way ANOVA with Tukey’s post hoc test for multiple comparisons. (B) M1 and M2 macrophage fractions across the same treatment groups. (VT: n = 11; anti-SPP1: n = 4; B20 LH: n = 9; B20 HH: n = 10). One-way ANOVA with Tukey’s post hoc test for multiple comparisons. (C) Tumor growth and OS after anti–PD-L1, anti-SPP1, B20-4.1.1, or their combinations. Controls (n = 36); anti–PD-L1 (n = 23); anti-SPP1 (n = 5); anti-SPP1 + anti–PD-L1 (n = 6); HH (n = 20); LH (n = 10); B20-4.1.1+anti-SPP1 (n = 8); B20-4.1.1 + anti-SPP1 + anti–PD-L1 (n = 7). Two-way ANOVA followed by Tukey’s multiple-comparison test. (D) Phenotypic consequences of tumor cell versus myeloid SPP1 depletion, showing a predominant contribution of the tumor-epithelial compartment (n = 7–21 mice per group). Two-way ANOVA followed by Tukey’s multiple-comparison test. (E) Hypoxia development (in VT- or B20-4.1.1–treated tumors) is partially corrected in the epithelial KO model (n = 5–7) but not in the BM model (n = 9–19). One-way ANOVA with Tukey’s post hoc test for multiple comparisons. Data are presented as mean ± SEM. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

Fig 3: Single-cell analysis identifies SPP1-centered epithelial-to-myeloid communication in HH tumors.(A) UMAP representation of the major tumor and immune cell populations identified by scRNA-seq. (B) Relative distribution of nonimmune and immune cell populations across treatment groups. (C) Heatmap of the number of inferred cell-cell interactions in HH versus LH tumors. (D) Heatmap of interaction strength across cell populations in HH versus LH tumors. (E) Signaling flow analysis highlighting pathways enriched in HH or LH tumors. Blue or red coloring of pathways indicates a significant difference between HH and LH (Wilcoxon’s test, P < 0.05; gray, nonsignificant). (F) SPP1-mediated interaction strength across sender and receiver cell populations in vehicle-treated tumors and in LH or HH tumors after B20-4.1.1. (G) Transcriptional Spp1 levels in tumors treated with VT (n = 6), anti-CSF1R (n = 5), LH (n = 10) or HH (n = 9) B20-4.1.1, and anti-CSF1R+B20-4.1.1 (n = 5). Experiment was performed in triplicate. One-way ANOVA with Tukey’s post hoc test for multiple comparisons. Data are presented as mean ± SEM. **P < 0.01.

Fig 4: Time-course and clinical run-in analyses to support an osteopontin-centered adaptive program under antiangiogenic pressure.(A) Tumor SPP1 transcript levels during the first 6 weeks of anti-VEGF treatment (T0: n = 5; T1: VT n = 3, B20 n = 3; T3: VT n = 10, B20 n = 9; T6: VT n = 6, B20 LH n = 8, and B20 HH n = 11). (B) Macrophage abundance over the same time window (T3: VT n = 5, B20 n = 7; T6: VT n = 10, B20 LH n = 11 and B20 HH n = 15). (C) Transcriptional levels of VEGFA along the first 6 weeks of isotype or anti-VEGF treatment (T0: n = 5; T1: VT n = 3, B20 n = 3; T3: VT n = 10, B20 n = 9; T6: VT n = 3, B20 LH n = 4, and B20 HH n = 6). (D) Evolution of HIF1α staining during treatment (T0: n = 5; T1: VT n = 3, B20 n = 3; T3: VT n = 6, B20 n = 6; T6: VT n = 5, B20 LH n = 6, and B20 HH n = 6). (E) Design of the NCT01484080 clinical trial, including baseline and post–run-in sampling in the nintedanib-containing arm. (F) Individual changes in SPP1 expression and hypoxia score during the nintedanib run-in phase. (G) GSEA plots comparing pre– and post–run-in samples according to baseline SPP1 levels. (H) Baseline SPP1 levels according to pathological complete response category (Symmans 0/1 versus 2/3) in the nintedanib-containing arm (experimental arm) and in the paclitaxel-alone arm of clinical trial NCT01484080. Experimental arm: Symmans 0/1: N = 5; average SPP1 = 42.3; Symmans 2/3: N = 19; average SPP1: 119.9; P = 0.35. Standard arm: Symmans 0/1: N = 3; average SPP1 = 142.2; Symmans 2/3: N = 33; average SPP1 = 127.9; P = 0.74).

Fig 5: Extracellular osteopontin reprograms macrophages through CD44/integrin B3-linked signaling.(A) Transwell migration of macrophages in response to vehicle, recombinant SPP1, or VEGF-A. Scale bars: 200 μm. Right: quantitation chart. Experiment involved analyzing 10 images for each condition across 3 independent experiments (total 30 images per condition). One-way ANOVA with Tukey’s post hoc test for multiple comparisons. (B) Expression of M2-related transcripts in macrophages after exposure to recombinant SPP1 (n = 4). One-way ANOVA with Tukey’s post hoc test for multiple comparisons. (C) Representative images of immunoblots showing phosphorylation of STAT3 in response to stimulation (5 minutes) with SPP1 (5 μg/mL) or vehicle (VT) in RAW264.7 cells, in presence or absence of a CD44 inhibitor (anti-mouse/human CD44 IM7 antibody, 10 μg/mL; upper panel) or an ITGβ3 inhibitor (ITGB3-IN-1, 5 μM; lower panel) (n = 3). (D) Same as in C showing the effect of p65 phosphorylation in response to SPP1 (15 minutes) (n = 3). (E) PD-L1 regulation in RAW264.7 cells in the same conditions as C and D (n = 4). One-way ANOVA with Tukey’s post hoc test for multiple comparisons. Data are presented as mean ± SEM. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.