Tumour-retained activated CCR7+ dendritic cells are heterogeneous and regulate local anti-tumour cytolytic activity

Abstract

Tumour dendritic cells (DCs) internalise antigen and upregulate CCR7, which directs their migration to tumour-draining lymph nodes (dLN). CCR7 expression is coupled to an activation programme enriched in regulatory molecule expression, including PD-L1. However, the spatio-temporal dynamics of CCR7⁺ DCs in anti-tumour immune responses remain unclear. Here, we use photoconvertible mice to precisely track DC migration. We report that CCR7⁺ DCs are the dominant DC population that migrate to the dLN, but a subset remains tumour-resident despite CCR7 expression. These tumour-retained CCR7⁺ DCs are phenotypically and transcriptionally distinct from their dLN counterparts and heterogeneous. Moreover, they progressively downregulate the expression of antigen presentation and pro-inflammatory transcripts with more prolonged tumour dwell-time. Tumour-residing CCR7⁺ DCs co-localise with PD-1⁺CD8⁺ T cells in human and murine solid tumours, and following anti-PD-L1 treatment, upregulate stimulatory molecules including OX40L, thereby augmenting anti-tumour cytolytic activity. Altogether, these data uncover previously unappreciated heterogeneity in CCR7⁺ DCs that may underpin a variable capacity to support intratumoural cytotoxic T cells.

Fig. 1. Landscape and temporal dynamics of DCs in murine tumours.

a Experiment design. Tumours from Kaede transgenic mice were harvested 5, 24, 48, or 72 h after photoconversion. b UMAP of myeloid cells from scRNA-seq of FACS-sorted CD45⁺TER119^- Kaede-green⁺/Kaede-red⁺ cells 48 h after photoconversion of subcutaneous MC38-Ova tumours. DC, dendritic cell; LC, Langerhans cell; Mac, macrophage; Mast, mast cell; pDC, plasmacytoid DC. c UMAP of DCs from (b) and canonical marker gene expression in DC clusters (d). e Kernel density embedding of DCs by Kaede fluorescence and treatment group. f, g Flow cytometry of Kaede fluorescence in DCs subsets/states from MC38-Ova tumours: f Representative plots 48 h post-photoconversion, and (g) Time-course and quantification. h Confocal microscopy of MC38 tumours 72 h post-photoconversion. Arrows point to tumour-residing CCR7⁺ DCs (Kaede-red⁺CCR7⁺MHC-II⁺, dendritic morphology). Scale bar, 40μm. i Tumour DCs ordered by pseudotime, rooted in the cluster with highest proportion of Kaede-green DCs, and proposed maturation trajectory. j RNA velocity trajectory in tumour DCs. k Expression of Axl. Paired two-sided student’s t-test was used, data are shown as means ± s.d., and points represent independent mice (g). The results shown in (g) are from one experiment (24 h/72 h n = 5; 48 h n = 4 animals), representative of three independent experiments; and (h) are representative of three independent experiments (n = 7 animals).

Fig. 2. CCR7⁺ DCs that migrate to the dLN become phenotypically and transcriptionally distinct from tumour-residing populations.

a–c Flow cytometry of draining lymph nodes (dLN) and contralateral non-draining lymph nodes (ndLN) of MC38-Ova tumours; DCs that originate from photo-flashed tumours (tumour emigrants) carry the Kaede-red fluorescent profile, which enables tracking to LNs. a Representative plots 48 h after tumour photoconversion, b Time-course and quantification of Kaede-red cells among CCR7⁺ DCs, and c composition of Kaede-red DCs in dLNs. d, e UMAP of myeloid cells from scRNA-seq of FACS-sorted CD45⁺ Kaede-red cells from tumour-dLNs (MC38-Ova) and CD45⁺ cells from control LNs. f Gene set enrichment analysis (GSEA) of Kaede-red CCR7⁺ DCs in dLNs versus tumours (isotype control-treated). Signed P-adj indicate log₁₀(Benjamini-Hochberg adjusted P value) with the direction of enrichment. g Selected gene expression in scRNA-seq of tumour CCR7⁺ DCs (isotype control-treated) and Kaede-red CCR7⁺ DC tumour emigrants in the dLN. h Flow cytometry of CD80 and CD86 on CCR7⁺ DCs from MC38-Ova tumours or dLNs 48 h after photoconversion, with representative histograms. i Integration of scRNA-seq of Kaede-red DCs from tumour-dLNs with tumour DCs and label transfer. j Proportion of DC clusters by tissue and Kaede fluorescence from i. Paired two-sided student’s t-test (b, c), two-sided Wilcoxon rank-sum test with Benjamini-Hochberg multiple-testing correction (g), or one-way analysis of variance (ANOVA) and Šidák’s multiple comparisons test (h) were used. Points represent independent mice (b, c, h). Data are shown as means ± s.d. (b, c, h), or box (median; box, 25^th and 75^th percentile; whiskers, 1.5*inter-quartile range) and violin plots (g). The results shown in a–c are from one experiment (24 h/72 h n = 5; 48 h n = 4 animals), representative of three independent experiments; and h from two independent experiments (n = 8 animals).

Fig. 3. Tumour-residing CCR7⁺ DCs undertake an “exhausted” state with duration in the tumour, attenuated by anti-PD-L1 treatment.

a Expression of MHC-II transcripts and Cd74 over pseudotime in tumour DCs. Local regression (loess) was fit to scaled expression values. b Gene signature scores for “GO:BP dendritic cell antigen processing and presentation (GO:0002468)” by CCR7⁺ DC cluster. c Scaled enrichment scores of scRNA-seq of all DCs for “GO:BP DC antigen processing and presentation” and CCR7⁺ DC signature genes, coloured by pseudotime. d GSEA of Ccr7_DC.3 versus Ccr7_DC.1. Only significant pathways (Benjamini-Hochberg-adjusted p values (P-adj) < 0.05) are shown. e Differential gene expression between Ccr7_DC.1 and Ccr7_DC.3. Significant differentially expressed genes (DEG, P-adj < 0.01) from “Hallmark inflammatory response” are highlighted in red. f Milo differential abundance analysis. Bee-swarm plot shows treatment-associated differences in overlapping cellular neighbourhoods (individual points). Differentially abundant neighbourhoods at FDR < 0.05 are coloured. ‘Mixed’ refers to neighbourhoods where cells do not predominantly (>70%) belong to a single cluster. g Differential gene expression between Ccr7_DC.2 and Ccr7_DC.3. DEGs (P-adj < 0.01, log₂Fold-change > 0.5) are coloured. h Flow cytometry of OX40L expression (OX40L^+/Hu-CD4 reporter mice) on CCR7⁺ DC from MC38-Ova tumours and their dLN. Two-sided Wilcoxon rank-sum test (b), two-sided Wald test with Benjamini-Hochberg multiple testing correction (e, g), or paired two-sided student’s t-test (h) were used. Data are shown as box (median; box, 25^th and 75^th percentile; whiskers, 1.5*inter-quartile range) and violin plots (b), or means ± s.d. (h). Points represent independent mice (h). The results shown in h are from one experiment (n = 5 animals), representative of two independent experiments.

Fig. 4. CCR7⁺ DCs interact with CD8⁺ T cells in human tumours.

a Pearson correlation between cell proportions, from deconvolution of 521 bulk transcriptomes of colorectal adenocarcinoma biopsies (TCGA), CD274, and a CCR7⁺ DC gene signature. Box highlights the high correlation between CCR7⁺ DC and CD8⁺ T cell transcripts. Only significant correlations (p < 0.05) shown. b Pearson correlation between CCR7⁺ DC signature genes and effector CD8⁺ T cell signature genes in colorectal adenocarcinoma (COAD), breast cancer (BRCA), skin cutaneous melanoma (SKCM) and lung adenocarcinoma (LUAD) from TCGA. Points represent individual patient samples. c CellPhoneDB cell-cell communication analysis between myeloid cells and effector CD8⁺ T cells in scRNA-seq of human CRC (n = 62 patients). Edge width scaled to standardised interaction scores. Only significant interactions (p < 0.05) shown. Mono, monocyte; Mac, macrophage; Mast, mast cell; pDC, plasmacytoid DC; CD8_Teff, effector CD8⁺ T cell. Spatial correlation (Pearson R-value) of CCR7⁺ DC and effector CD8⁺ T cell signature scores in spatial transcriptomics (10X Genomics Visium) of independent human CRC tumour sections (d, n = 2), a human melanoma section (e, n = 1), and independent human breast tumour sections (f, n = 2). g Expression of selected genes in myeloid-T cell doublets from sequencing of physically interacting cells (PICseq) of non-small cell lung cancer (NSCLC) tumours (n = 10 patients), grouped by the myeloid cell identity in each myeloid-T cell doublet. Two-sided Pearson correlation was used (a-b). Data are shown as linear regression with 95% confidence interval (b).

Fig. 5. anti-PD-L1 promotes activating CCR7⁺ DC-cytotoxic CD8⁺ T cell interactions in the TME.

a UMAP of scRNA-seq of CD8⁺ T cells, from FACS-sorted CD45⁺ tumour-infiltrating lymphocytes 48 h after photoconversion of subcutaneous MC38-Ova tumours. b Proportion of T_EX_CD8T cells by Kaede fluorescence and treatment group. c Number of significant DEGs in anti-PD-L1 versus isotype control-treated tumours by cluster. DEGs calculated using Wilcoxon rank-sum test with Benjamini-Hochberg multiple-testing correction. d CellPhoneDB ligand-receptor analysis between tumour DCs and CD8⁺ T_EX cells. Only significant interactions (p < 0.05) shown. e Representative confocal microscopy images of MC38 tumours, showing co-localisation of CCR7⁺MHC-II⁺ DCs and CD3⁺CD8⁺ 4-1BB⁺/Ki-67⁺ T cells (arrows). Scale bar, 20 μm. (f–k) In vitro and ex vivo cultures. f Phenotype of bone marrow-derived DCs (BMDC) following culture with UV-irradiated MC38-Ova tumour cells. g Experiment set-up of DC:OT-I co-culture. h Representative flow cytometry of CD44⁺ OT-I cell activation and proliferation following co-culture with MC38-Ova-experienced CCR7⁺ BMDC; +/– anti-PD-L1, OX40L-expressing (+, OX40L^fl/fl) or OX40L-deficient (–, CD11c^creOX40L^fl/fl) BMDCs. i Quantification of h and granzyme B (GzmB) expression. Values shown are relative to cultures with isotype control antibodies and OX40L-expressing DCs (baseline), to facilitate comparisons across experiments. j Flow cytometry of CCR7⁺ DCs from subcutaneous MC38-Ova tumours following 8 h culture ex vivo with antibodies or recombinant IFNγ (rIFNg). k Flow cytometry of OT-I cells following co-culture with MC38-Ova-experienced CCR7⁺ BMDC; CCR7⁺ BMDCs pre-treated with rIFNg (+) or PBS (–), OX40L-expressing (+) or OX40L-deficient (–) BMDCs. One-way analysis of variance (ANOVA) and Šidák’s multiple comparisons test was used, data are shown as means ± s.d. (f, i–k). The results shown in (e) are representative of two independent experiments (n = 5 animals); f from one experiment (n = 4 biological samples) representative of three independent experiments; h, i from two independent experiments (n = 7 biological samples); j from one experiment (n = 5 animals), representative of two independent experiments; and (k) from one experiment (n = 4 biological samples), representative of two independent experiments.

Fig. 6. Conserved CCR7⁺ DC heterogeneity and CD8⁺ T cell crosstalk in human cancer.

a UMAP of scRNA-seq of CCR7⁺ DCs from human CRC (GSE178341, n = 62 patients). b Selected gene expression in CCR7⁺ DCs from tumour (T), normal adjacent tissue (N) and tumour-dLN from independent scRNA-seq data of human CRC (syn26844071, n = 63 patients). c UMAP of myeloid cells from scRNA-seq of human metastatic urothelial carcinoma (mUC, n = 11 patients). d single-cell gene expression (scGEX) profiles associated with responders (complete responder, CR; partial responder, PR) or non-responders (stable disease, SD; progressive disease, PD) in the IMvigor 210 trial (atezolizumab) for treatment of mUC; Scissor integration of bulk RNA-seq of tumour samples from IMvigor210 (n = 208 patients) and scRNA-seq from c. e Differential gene expression in myeloid cells associated with responders (CR/PR) versus non-responders (SD/PD) from d. f Gene signature scores for Ccr7_DC.2 transcripts in scRNA-seq of CCR7⁺ DCs from human breast cancer following treatment with anti-PD-1 antibodies; T cell clonotype expanders (E, i.e. responders, n = 9 patients) versus non-expanders (NE, i.e. non-responders, n = 20 patients). g PICseq (myeloid-T cell doublets) of human NSCLC; expression of CCR7⁺ DC-T cell ligand-receptor pairs, grouped by myeloid cell identity in each myeloid-T cell doublet (n = 10 patients). h Gene signature scores for “GO:BP T cell mediated cytotoxicity (GO:0001913)” (HLA genes removed), in myeloid-CD8⁺ T cell PICs grouped by the myeloid cell identity. i, j Spatial correlation (Pearson R-value) of CCR7⁺ DC signature scores and selected CCR7⁺ DC-ligand receptors expressed by CD8⁺ T cells, in spatial transcriptomics (10X Genomics Visium) of independent human CRC tumour sections (n = 2). Two-sided Wilcoxon rank-sum test with Benjamini-Hochberg multiple-testing correction (b, e), or two-sided Wilcoxon rank-sum test (f, h) were used. Data are shown as box (median; box, 25^th and 75^th percentile; whiskers, 1.5*inter-quartile range) and violin plots (b, f, h).