Positioning and reversible suppression of CCR7+ dendritic cells in perivascular tumor niches shape cancer immunity

Abstract

Tumor-resident CCR7⁺ dendritic cells (DCs) are key determinants of antitumor T cell responses. Here, we examined the localization of CCR7⁺ DCs within tumors and the impact of this positioning on antitumor immunity. Spatial, single-cell, and intravital analyses of human cancers and mouse models reveal that CCR7⁺ DCs form perivascular clusters. Fibroblasts surrounding venous blood vessels produced CCL19, guiding CCR7⁺ DCs into perivascular niches. Regulatory T (Treg) cells frequently contact perivascular CCR7⁺ DCs, suppressing CD40 expression and CD4⁺ and CD8⁺ T cell activation. Treg cell depletion restored CD40 expression by CCR7⁺ DCs, enhanced immunostimulatory programs, and improved T cell-dependent tumor control. Anti-PD-1 not only increased perivascular CCR7⁺ DC clustering and IL-12 production but also strengthened Treg-DC interactions through a CCL22-dependent mechanism, limiting therapeutic efficacy. CCR7⁺ DCs expressed both co-stimulatory and co-inhibitory molecules, which may underlie their capacity for antitumor activation and concurrent vulnerability to suppression. Modulating the mechanisms that form and restrain CCR7⁺ DC perivascular immune hubs may improve cancer immunotherapy.

Keywords: T cells; cancer immunity; cancer immunity cycle; chemokines; dendritic cells; immunotherapy; regulatory.

Figure 1.. CCL19 anchors CCR7⁺ DCs to perivascular sites

(A) Representative images and quantification of CCR7⁺ DCs (panCK⁻HLA-DR⁺LAMP3⁺, yellow) near BVs (CD31⁺PDPN⁻, magenta), or LVs (CD31⁺PDPN⁺, cyan) in human tumors (HNSCC, NSCLC, and EC). Scale bar represents 20 μm. Whole-tumor sections were analyzed for EC and NSCLC. Numbers of fields of view (FOVs) analyzed per HNSCC sample are as follows: HNSCC1–04 n = 7; HNSCC1–06 n = 16; HNSCC1–07 n = 11; HNSCC2–01 n = 126; HNSCC2–06 n = 455; HNSCC2–09 n = 180; HNSCC2–11 n = 122; HNSCC2–12 n = 79; HNSCC2–15 n = 205; HNSCC2–26 n = 293; HNSCC2–35 n = 175. One bar = one patient (Table S1). (B) Representative images and quantification of CCR7⁺ DCs (FSCN1⁺; yellow) located near BVs (CD31⁺LYVE-1⁻; magenta) or LVs (CD31⁺LYVE-1⁺; cyan) in mouse tumors (MC38, B16F10, and D4M3.A-OVA). Scale bar represents 10 μm. Whole-tumor sections were analyzed. One bar = one mouse. (C) Frequencies of BV-, LV- and non-vessel-associated CCR7⁺ DCs in mouse MC38 tumors 3 days post anti-CD40 or anti-PD-1 treatment. Whole-tumor sections were analyzed. One bar = one mouse. (D) (Left) Representative images of CCR7⁺ DCs (FSCN1⁺; yellow) located near BVs (CD31⁺LYVE-1⁻; magenta) in MC38 tumors inoculated in Ccr7^ko/wt and Ccr7^ko/ko mice, 3 days post anti-PD-1 treatment. (Right) Distribution of the area of CCR7⁺ DC surfaces in clusters relative to their distance to closest BVs and plotted as percentage of total CCR7⁺ DC cluster area. CCR7⁺ DC surfaces from clusters associated with LVs and those not in clusters were excluded from the analysis. Scale bar represents 20 μm. Whole-tumor sections were analyzed. One dot = average value of all clusters in each genotype (Ccr7^ko/ko n = 5 mice, 56 clusters; Ccr7^wt/ko n = 6 mice, 28 clusters; and Ccr7^wt/wt n = 3 mice, 19 clusters). Two-way ANOVA with multiple comparisons, mean with SEM; ****p < 0.0001 for comparison at 10 and 20 μm from closest BVs. (E) (Left) Representative images of CCR7⁺ DCs (FSCN1⁺; yellow) located near BVs (CD31⁺LYVE-1⁻; magenta) and Ccl19 (Ccl19-eYFP⁺Tomato⁺; white) in Ccl19-ieYFP reporter mice (left image) or CCL21 (white, right image) in MC38 tumors. (Right) Frequencies of perivascular CCR7⁺ DC clusters associated with Ccl19-covered BVs or within CCL21⁺ areas of the tumors among total perivascular CCR7⁺ DC clusters. Scale bar represents 20 μm. Whole-tumor sections were analyzed. One dot = one mouse. Unpaired t test, mean with SEM; ***p < 0.001. (F) (Left) Representative images of CCR7⁺ DCs (FSCN1⁺; yellow) located near BVs (CD31⁺LYVE-1⁻; magenta) in MC38 tumors inoculated in Ccl19^wt/wt and Ccl19^ko/ko mice, 2 days post anti-PD-1treatment. (Right) Quantification of BV- or LV-associated CCR7⁺ DC clusters in MC38 tumors from Ccl19^wt/wt and Ccl19^ko/ko mice. Scale bar represents 20 μm. Whole-tumor sections were analyzed. One dot = one mouse, whiskers represent min to max. Unpaired t test; *p < 0.05. (G) Heatmap depicts log₂-transformed averaged expression of Ccl19 in indicated immune and non-immune populations in the TME of multiple mouse tumor models (breast,^, lung [and GSE201247], and pancreatic^,). (H) (Left) Synthetic images of CCR7⁺ DCs (yellow), blood endothelial cells (BECs; magenta), lymphatic endothelial cells (LECs; cyan), and CCL19⁺ fibroblasts (green) in one representative NSCLC patient analyzed by spatial transcriptomics. (Right) Box plots depict the enrichment scores of CCL19⁺ fibroblasts within the neighborhood of BV-associated CCR7⁺ DCs, in four human NSCLC. Data are shown for both permuted (median enrichment scores from 1,000 permutations) and observed datasets. Scale bar represents 20 μm. Whole-tumor sections were analyzed. One dot = one sample. Paired t test, whiskers represent mean to max; *p < 0.05. (I) Heatmap depicts log₂-transformed averaged expression of CCL19 in indicated immune and non-immune populations in the TME of multiple human cancer types (HNSCC, n = 40, n = 18 patients; CRC, n = 23, n = 64 patients; ESCC, n = 58 patients; NSCLC, n = 32, n = 7 patients; BRCA, n = 29 patients; and PRCA, n = 18 patients). A cross indicates that the cellular population was not detected. See also Figures S1–S4.

Figure 2.. Tregs populate perivascular CCR7⁺ DC niches in human and mouse tumors

(A) Heatmaps depict the enrichment of immune and non-immune cell types in the immediate neighborhood of CCR7⁺ DCs in NSCLC spatial transcriptomic data (n = 4). (B) (Left) Representative FOV displaying CCR7⁺ DCs (HLA-DR⁺LAMP3⁺; yellow) located near BVs (CD31⁺PDPN⁻; magenta) and Tregs (CD4⁺FOXP3⁺; white) in one HNSCC sample using high-plex whole-tissue imaging. Scale bar represents 20 μm. (Right) Box plots display the frequencies of BV-associated, LV-associated, and non-vessel-associated CCR7⁺ DCs close (<5 μm) to Tregs among all tumor CCR7⁺ DCs with nearby Tregs. Wilcoxon test, whiskers represent min to max; *p < 0.05. (C) Correlations between CCR7⁺ DCs and Tregs within CD45⁺ cells, as determined by scRNA-seq in multiple human cancer types. Spearman rank correlation; significant correlations are shown with a fitted red line. (D) (Left) Scheme outlining the analyses of CCR7⁺ DCs and Tregs in NSCLC samples. Patients with numerous (>5) CCR7⁺ DC clusters (n = 12) were selected for downstream analyses. (Right) Frequency of CCR7⁺ DCs (CD11c⁺LAMP3⁺) with at least one nearby (<50 μm) Treg (CD4⁺FOXP3⁺) in each individual patient. Numbers of FOVs analyzed per sample are as follows: NR01, n = 126; NR06, n = 455; NR09, n = 180; NR12, n = 79; NR26, n = 293; R11, n = 122; R15, n = 205; R35, n = 175; R37, n = 459; R45, n = 276. (E) (Left) Scheme outlining the analysis of tumor biopsies from HNSCC patients before immunotherapy (pre-IO). Patients were divided into non-responders (NR, n = 5) and responders (R, n = 5) based on the assessment of clinical response at 6 months. (Right) CCR7⁺ DC shortest distance to Tregs, T_CONV, and CD8⁺ T cells in NR versus R tumors. Data are shown for all CCR7⁺ DCs compiled (NR tumors, n = 1,457 cells; R tumors, n = 1,324 cells). Unpaired t test, whiskers represent min to max; ****p < 0.0001. Numbers of FOVs analyzed per sample as in (D). (F) (Left) Scheme outlining the analyses of CCR7⁺ DC-CD8⁺ T cell niches. (Right) Frequencies of CCR7⁺ DC-CD8⁺ T cell niches with or without Tregs in their proximity (<100 μm). Two-way ANOVA with multiple comparisons, whiskers represent min to max; *p < 0.05. Numbers of FOVs analyzed per sample as in (D). (G) Representative FOV displaying CCR7⁺ DCs (FSCN1⁺ cells; FSCN1 in yellow) located near BVs (CD31⁺LYVE-1⁻ cells; CD31 in magenta) and Tregs (FOXP3⁺ cells; FOXP3 in white) in untreated MC38 tumors. Scale bar represents 50 μm. (H) Correlations between the numbers of CCR7⁺ DCs and Tregs per mg of tumor tissue, as determined by fluorescence-activated cell sorting (FACS) analyses of MC38 and D4M3. A tumors. Spearman rank correlation; significant correlations are shown with a fitted red line. (I) Box plots show the frequencies of tumor CCR7⁺ DCs close (<5 μm) to Tregs that are associated to BVs or LVs in MC38 tumors (n = 7). Whole-tumor sections were analyzed. Paired t test, whiskers represent min to max; ****p < 0.0001. See also Figures S5 and S6.

Figure 3.. Tregs reversibly suppress antitumor functions of CCR7⁺ DCs at perivascular niches

(A) (Left) Scheme outlining the experimental setup for bulk RNA-seq analyses of tumor-derived CCR7⁺ DCs. (Right) GO pathway enrichment analyses performed on differentially expressed genes (DEGs) in CCR7⁺ DCs in MC38 tumors (n = 4) from Treg-depleted (FoxP3-DTR) compared with Treg-sufficient (WT) mice. Bar plot indicates the −log₁₀ raw binomial p-values of the top 10 most enriched pathways in CCR7⁺ DCs. (B) (Left) Experimental setup for ex vivo stimulation of OT-I CD8⁺ T cells with tumor CCR7⁺ DCs. (Right) Percentage of OT-I CD8⁺ T cells that proliferated after 5-day culture with OVA_257–264 peptides-loaded CCR7⁺ DCs isolated from WT or Treg-depleted tumors. As a control, CCR7⁺ DCs without OVA_257–264 peptides were used. Two-way ANOVA with multiple comparisons, whiskers represent min to max; **p < 0.01. (C) (Left) Relative gene expression levels analyzed by bulk RNA-seq. Each dot represents one mouse (n = 4), whiskers represent mean to max. Unpaired t test with multiple comparisons; *p < 0.05. (Right) Representative histogram of CD40 protein expression and relative mean fluorescence intensity (MFI) measured by FACS and expressed both as normalized values and absolute MFI. Each dot represents one mouse (n = 18), whiskers represent min to max. Unpaired t test; **p < 0.01. (D) Analyses of cDCs in tumor-draining lymph nodes. Absolute cell counts (left, n = 10) and MFI of CD40 expression (right, n = 18) measured by FACS in migratory cDCs (CCR7⁺CD8α⁻) from WT or Treg-depleted mice. Whiskers represent mean to max. (E) (Left) Experimental setup for ex vivo analyses of tumor CCR7⁺ DCs isolated from anti-PD-1-treated mice that received or not αCD25^NIB mAbs. (Right) CD40 protein expression measured by FACS and expressed both as normalized values and absolute MFI. Each dot represents one mouse (n = 4 WT and n = 6 FoxP3-DTR), whiskers represent min to max. Unpaired t test; **p < 0.01. (F) (Left) Overall survival analyses of MC38 tumor-bearing mice treated, or not treated, with αPD-1 and αCD25^NIB mAbs, and in which CD4⁺ or CD8⁺ cells were depleted or not (n = 8 or 9 mice/group). Log-rank Mantel-Cox test; *p < 0.05, ***p < 0.001, and ***p < 0.0001. (Right) Percentage of tumor-free mice on day 60 in the indicated treatment groups. (G) (Left) Experimental setup for ex vivo stimulation of OT-I CD8⁺ T cells with tumor CCR7⁺ DCs as in (B). The DCs were obtained from mice receiving anti-PD-1 immunotherapy and that were treated or not with αCD25^NIB mAbs. (Right) Percentage of OT-I CD8⁺ T cells that proliferated after 5-day culture with OVA_257–264 peptide-loaded CCR7⁺ DCs. Each dot represents one mouse (n = 8 and n = 7), whiskers represent min to max. Two-way ANOVA with multiple comparisons; *p < 0.05. (H) (Left) Scheme outlining bone marrow chimeras with inducible Cd40-deficiency in cDCs and the treatment schedule. (Right) Growth curves of MC38 tumors inoculated in zDC^iDTR:Cd40^WT and zDC^iDTR:Cd40^KO bone marrow chimeras treated with αPD-1, αCD25^NIB, or αPD-1 + αCD25NIB combination (n = 8–10 mice/group). Mean with SEM. Two-way ANOVA with multiple comparisons; *p < 0.05 and ****p < 0.0001. See also Figures S7 and S8.

Figure 4.. Immunotherapy enhances perivascular tumor CCR7⁺ DC interactions with neighboring Tregs via Ccl22

(A) (Left) Outline of in vivo two-photon time-lapse microscopy in MC38-H2b-Cerulean tumors. (Middle) Representative FOVs displaying the migratory behavior of Tregs (FoxP3-mRFP, magenta) in untreated (top) or anti-PD-1-treated (bottom) mice in perivascular regions containing Il12-eYFP⁺ DCs (yellow). Scale bar represents 20 μm. Treg migratory tracks are shown in white. (Right) Observed contact duration between DCs and Tregs in each experimental group (n = 4). Each dot represents one DC, whiskers represent mean to max. Unpaired t test; ***p < 0.001. (B) As in (A), but in Il12-p40-eYFP; FoxP3-mRFP; Ccl22^−/− mice (n = 5). (C) Change in contact duration between DCs and Tregs after anti-PD-1 treatment, comparing Ccl22^WT and Ccl22^−/− mice. Each dot represents one mouse, bar represents mean with SEM. Unpaired t test; *p < 0.05. (D) Quantification of the cumulative interactions of CCR7⁺ DCs with Tregs over time of acquisition, comparing Ccl22^WT and Ccl22^−/− mice. Shaded areas indicate confidence intervals. (E) Change in tumor volume in WT and Ccl22-deficient mice left untreated (left), or upon anti-PD-1 therapy (right). Unpaired t test, whiskers represent min to max; *p < 0.05. See also Figure S9.

Figure 5.. Individual CCR7⁺ DCs co-express stimulatory and regulatory molecules

(A) (Left) FACS plots display the gating strategy for the analyses of IL-12-eYFP⁺CCR7⁺ DCs. (Right) Graphs show MFI values of indicated stimulatory and inhibitory markers in cDC1s, cDC2s, CCR7⁺ DCs, and IL12-eYFP⁺CCR7⁺ DCs. Expression of Ccl22 transcripts was analyzed by PrimeFlow. Each dot represents one mouse (n = 5), bar represents mean with SEM. Ordinary one-way ANOVA test was performed; *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001. (B) Heatmap displays the expression of genes (with stimulatory or inhibitory function) at single-cell level in individual DCs in each state in mouse tumor models. (C) As in (B), but in HNSCC tumors. See also Figure S10.