肾癌组装体中的WNT–CAF–T细胞联系

Fig. 1 | Generation of patient-specific RCC assembloids that recapitulate therapeutic responses to ICIs. a,b, Representative immunofluorescence (a) and haematoxylin and eosin (H&E)-stained (b) images from a time-course analysis of patient-derived RCC assembloid formation in comparison to corresponding parental tumours from six patients (P-1 to P-6). Scale bars, 100 μm. c, Heatmap showing pairwise similarity between RCC assembloids (A) and their corresponding parental tumours (T) based on somatic mutation profiles. d, Proportion of shared and unique mutations in each patient-derived pair. e, Uniform manifold approximation and projection (UMAP) plots showing major cell populations identified by scRNA-seq in RCC assembloids and corresponding parental tumours from six patients. f, Dot plots showing the expression of selected marker genes for each cell type. g, Cell-type composition of RCC assembloids and matched parental tumours across six patient pairs. h, Pairwise comparison of cell-type proportions between RCC assembloids and matched parental tumours for each patient pair. i, Cell-type-specific transcriptomic similarity between RCC assembloids and their matched parental tumours. Detailed methods for these analyses are provided in the Methods. j, Representative immunofluorescence images of RCC assembloids treated with isotype control IgG or anti-PD-1. Scale bars, 100 µm. k,l, Quantification of cleaved caspase-3+ cells (k) and the proportion of cleaved caspase-3+ cells (l) in the indicated cell types following IgG or anti-PD-1 treatment. Three biologically independent assembloids per group were analysed, with three sections analysed per assembloid (n = 9). m, Summary of the clinical responses of patients with RCC to ICI therapy and corresponding responses in assembloids to anti-PD-1. Statistics: paired Wilcoxon signed-rank test with Benjamini–Hochberg false discovery rate (FDR) correction (g) or two-way analysis of variance (ANOVA) followed by Sidak’s multiple-comparison test (k,l); data show the mean ± s.e.m. (k,l). HVGs, highly variable genes; R, responder; NR, non-responder; NED, no evidence of disease; PD, progressive disease; PR, partial response; SD, stable disease.

Fig. 2 | Non-T cell-autonomous maintenance of CD8+ T cell stemness, mediated through a tumour–CAF–CD8+ T cell crosstalk, is required for anti-PD-1 responsiveness. a,b, Top, flow cytometry analyses (left) and quantification (right) of stem-like T cells in RCC assembloids (a) and parental tumours (b) from responders and non-responders. Bottom, expression profiles of exhaustion and effector markers in terminally differentiated (TD) and stem-like (Stem) T cell subsets from P-2. For a, three biologically independent assembloids were analysed per patient (n = 3); for b, three biologically independent samples were analysed per patient (n = 3). c, Flow cytometry analyses of stem-like T cells in four mix-and-match RCC assembloids (A–D) (Extended Data Fig. 5b). Blue, responder; red, non-responder. Three biologically independent assembloids were analysed per group (A, n = 48; B, n = 24; C, n = 24; D, n = 12). d, Flow cytometry analyses of the proportion of stem-like T cells in combinatorial RCC assembloids from responders. Three biologically independent assembloids were analysed per group (n = 3). e, Anti-PD-1 responses in combinatorial RCC assembloids from responders. Three biologically independent assembloids were analysed per group, with three sections analysed per assembloid (n = 9). Scale bars, 100 µm. f–i, Top, schematic of the experiment. Bottom, flow cytometry plots and bar graphs show the proportion of stem-like T  cells. Three biologically independent assembloids were analysed per group (n = 3). f, RCC assembloids lacking tumour cells and CAFs were cultured for 7 days in the presence of tumour-CM and CAF-CM. g, RCC assembloids lacking CAFs were cultured for 7 days in the presence of CAF-CM. h, RCC assembloids lacking tumour cells were cultured for 7 days in the presence of tumour-CM. i, CAFs and TII cells were cultured in a Transwell system, with CAFs in the bottom chamber treated with tumour-CM and TII cells cultured in the top Transwell insert. Statistics: two-tailed nested t-test (a,b), one-way ANOVA followed by Tukey’s multiple-comparison test (c,d) or two-way ANOVA followed by Sidak’s multiple-comparison test (e–i); data show the mean ± s.e.m. (a–i).

Fig. 3 | Identification of a WNT–activin A–T cell stemness axis in tumour– CAF–CD8+ T cell interactions. a, Relative expression of candidate genes in CAFs isolated from tumour tissue samples of four responders. Three technical replicates were analysed per group (n = 3). b, Relative expression of FGF2 in control and FGF2 KD CAFs and INHBA in control and INHBA KD CAFs from responders. Three biologically independent samples were analysed per group (n = 3). c, Proportion of stem-like T cells in RCC assembloids reconstituted with control CAFs, FGF2 KD CAFs or INHBA KD CAFs from responders. Three biologically independent assembloids were analysed per group (n = 3). d, Anti- PD-1 responses in RCC assembloids reconstituted with control or INHBA KD CAFs from responders. Three biologically independent assembloids were analysed per group, with three sections analysed per assembloid (n = 9). Scale bars, 100 μm. e, Predicted tumour–CAF signalling networks based on transcriptomic profiles of tumour cells and CAFs. Networks involving tumour cell–CAF interactions are shown. f, Heatmap showing expression of candidate genes in tumour cells from responders as determined by RNA-seq. g, Relative expression of candidate genes in CA9+ and CA9− cells isolated from tumour tissue samples of responders. Three technical replicates were analysed per group (n = 3). h, Relative expression of SEMA3B, TNF and WNT2B in control and candidate-gene KD tumour cells from responders. Three biologically independent samples were analysed per group (n = 3). i, Relative expression of INHBA in CAFs reconstituted with candidate-gene KD tumour cells from responders. Three biologically independent assembloids were analysed per group (n = 3). j, Proportion of stem-like T cells in RCC assembloids reconstituted with control tumour cells, SEMA3B KD tumour cells, TNF KD tumour cells or WNT2B KD tumour cells from responders. Three biologically independent assembloids were analysed per group (n = 3). k, Anti-PD-1 responses in RCC assembloids reconstituted with control or WNT2B KD tumour cells from responders. Three biologically independent assembloids were analysed per group, with three sections analysed per assembloid (n = 9). Scale bars, 100 μm. Statistics: two-way ANOVA followed by Sidak’s multiple-comparison test (b,d,h,k) or one-way ANOVA followed by Tukey’s multiple-comparison test (c,i,j); data show the mean ± s.e.m. (a–d,g–k).

Fig. 4 | Identification of tumoural WNT-unresponsive CAFs that confer resistance to anti-PD-1 therapy. a, Heatmap showing transcriptomic profiles of CA9+ and CA9− cells from responders and non-responders. b, Relative expression of WNT2B in CA9+ and CA9− cells from responders and non-responders. Three biologically independent assembloids were analysed per group (n = 3). c, Heatmap showing transcriptomic profiles of CAFs from responders and non-responders. d, Relative expression of INHBA in CAFs from responders and non-responders. Three biologically independent assembloids were analysed per group (n = 3). e, Relative expression of WNT target genes in CAFs from responders and non-responders. Three biologically independent assembloids were analysed per group (n = 3). f, Differential pathway analysis of CAFs between responders and non-responders, highlighting WNT-related pathways. g, Heatmap of WNT pathway-associated DEGs in CAFs from responders and non-responders. h, Relative expression of key WNT pathway regulator genes in CAFs isolated from tumour tissues of responders and non-responders. Three technical replicates per group were analysed (n = 3). i, Relative expression of LRP6, INHBA and WNT target genes in control and LRP6-expressing CAFs from non-responders. Three biologically independent assembloids were analysed per group (n = 3). j, Proportion of stem-like T cells in RCC assembloids reconstituted with control CAFs or LRP6-expressing CAFs from non-responders. Three biologically independent assembloids were analysed per group (n = 3). k, Anti-PD-1 responses in RCC assembloids reconstituted with control or LRP6-expressing CAFs from non-responders. Three biologically independent assembloids were analysed per group, with three sections analysed per assembloid (n = 9). Scale bars, 100 µm. l, Experimental design for evaluating anti-PD-1 responses in humanized mice and ALI organoids derived from patients with RCC (responders and non-responders). m, Anti-PD-1 responses in humanized mice with responder (P-1) tumours (left) or non-responder (P-2) tumours (right). Four mice were analysed per group (n = 4). n, Anti-PD-1 responses in ALI organoids from a responder (P-5) (left) and a non-responder (P-3) (right). Three biologically independent organoids were analysed per group, with three sections analysed per organoid (n = 9). Scale bars, 100 µm. Statistics: two-tailed nested t-test (b,d,e,h), one-sided Fisher’s exact test, with FDR correction for multiple comparisons (f), two-way ANOVA followed by Sidak’s multiple-comparison test (i,k,m,n) or unpaired two-tailed t-test (j); data show the mean ± s.e.m. (b,d,e,h–k,m,n).

Fig. 5 | Single-cell and spatial analyses of patient tumours reveal a WNT– activin A–T cell stemness axis and WNT-unresponsive CAF subsets associated with anti-PD-1 responses. a, Schematic for analysing pretreatment tumour tissue samples from responders and non-responders. b, UMAP plots showing major cell populations identified in patients with RCC (left; n = 6) and split into responders (n = 4) and non-responders (n = 2) (right). c, UMAP plot (left) and violin plot (right) showing WNT2B expression levels in tumour cells from responders (n = 4) and non-responders (n = 2). d, Dot plot showing expression of canonical marker genes for CAF subtypes. e, Pearson’s correlation coefficient (PCC) of average CAF gene expression profiles between responders and non- responders. f, Differential pathway analysis of CAFs from responders and non- responders, based on DEGs upregulated in responders, highlighting enrichment of WNT-related pathways. g, UMAP plots (left) and bar plot (right) of CAFs from responders and non-responders. h, Top, violin plots showing INHBA and LRP6 expression in CAFs stratified by INHBA expression status from patients with RCC (n = 6). Bottom, violin plots of INHBA and LRP6 expression in CAFs from responders (n = 4) and non-responders (n = 2). i, Bar plot showing the proportion of stem-like CD8+ T cells among total CD8+ T cells (top) and violin plot showing stem-like CD8+ T cell signature scores in CD8+ T cells (bottom) in responders (n = 4) and non-responders (n = 2). j, Representative immunofluorescence images of tumour tissue samples from responders and non-responders. Scale bars, 100 μm. k, Quantification of tumour-adjacent CAFs (COL1A1+ cells located within 50 µm of KRT8+ tumour cells), the proportion of LRP6+ CAFs among tumour-adjacent CAFs, tumour-infiltrating CD8+ T cells (CD8+ cells located within 50 µm of KRT8+ tumour cells) and the proportion of TCF1+ CD8+ T cells among total tumour-infiltrating CD8+ T  cells. All measurements were calculated per unit area (mm2). Three regions per section were analysed, with three sections analysed per patient (n = 9). Statistics: two-tailed Welch’s t-test (c,i (bottom)), one-sided Fisher’s exact test, with FDR correction for multiple comparisons (f), two-sided Wilcoxon rank-sum test (h), Fisher’s exact test (i, top) or two-tailed nested t-test (k). For box plots (c,h,i), centre lines indicate medians, box limits indicate the interquartile range and whiskers extend to the most extreme data points within 1.5× the interquartile range. Data show the mean ± s.e.m. (k).