Residual ANTXR1+ myofibroblasts after chemotherapy inhibit anti-tumor immunity via YAP1 signaling pathway

Abstract

Although cancer-associated fibroblast (CAF) heterogeneity is well-established, the impact of chemotherapy on CAF populations remains poorly understood. Here we address this question in high-grade serous ovarian cancer (HGSOC), in which we previously identified 4 CAF populations. While the global content in stroma increases in HGSOC after chemotherapy, the proportion of FAP⁺ CAF (also called CAF-S1) decreases. Still, maintenance of high residual CAF-S1 content after chemotherapy is associated with reduced CD8⁺ T lymphocyte density and poor patient prognosis, emphasizing the importance of CAF-S1 reduction upon treatment. Single cell analysis, spatial transcriptomics and immunohistochemistry reveal that the content in the ECM-producing ANTXR1⁺ CAF-S1 cluster (ECM-myCAF) is the most affected by chemotherapy. Moreover, functional assays demonstrate that ECM-myCAF isolated from HGSOC reduce CD8⁺ T-cell cytotoxicity through a Yes Associated Protein 1 (YAP1)-dependent mechanism. Thus, efficient inhibition after treatment of YAP1-signaling pathway in the ECM-myCAF cluster could enhance CD8⁺ T-cell cytotoxicity. Altogether, these data pave the way for therapy targeting YAP1 in ECM-myCAF in HGSOC.

Fig. 1. Decrease in CAF-S1 and CAF-S4 content after chemotherapy in HGSOC.

A Hematoxylin eosin saffron (HES) staining in paired (before and after chemotherapy) HGSOC (Retrospective Curie 1 cohort). Scale bars, 50 μm and 25 μm (insert). B EPCAM IHC staining. Scale bars, 50 μm and 25 μm (insert). C Percentage (%) of epithelium (from HES) in HGSOC before and after chemotherapy (N = 35 patients, n = 70 matched samples). Data are shown using paired (Left, two-sided paired Wilcoxon test) and unpaired (Right, two-sided Mann-Whitney test) statistical analyses. D Intensity of EPCAM staining in epithelial cells, ranging from 0 to 4 (N = 35 patients). Paired (Left, two-sided paired Wilcoxon test) and unpaired (Right, two-sided Mann-Whitney test) analyses. E Percentage of the fibroblastic stroma (from HES) in HGSOC (N = 35 patients). Paired (Left, two-sided paired Wilcoxon test) and unpaired (Right, two-sided Mann-Whitney test) analyses. F IHC staining of FAP, CD29, SMA and FSP1 CAF markers on serial sections of paired HGSOC. Scale bars, 50 μm. G CAF marker H-scores in HGSOC (N = 35 patients). Two-sided paired Wilcoxon test. H Decision tree algorithm defining CAF identity (See Methods). I Repartition of CAF populations enrichment in HGSOC based on the decision tree shown in (H) (N = 35 patients). Two-sided Chi-square test. J Flow cytometry plots showing CD29, FAP, SMA and FSP1 protein levels in viable fibroblasts from HGSOC samples collected before (Left, treatment-naïve) and after chemotherapy (Right) (See also Supplementary Fig.S1 showing the gating strategy). K Left, Quantifications of EPCAM⁺ epithelial cells among viable cells (N = 20 HGSOC tumor samples). Two-sided unpaired t-test. Right, same as in Left for CAF among viable cells (N = 8 treatment naïve, N = 12 after chemotherapy). Two-sided Mann-Whitney test. L Left, % of CAF populations in HGSOC tumor samples (N = 8 treatment naïve, N = 12 after chemotherapy). Two-sided Chi-square test. Right, Bar plot showing the % of FAP^High (myCAF) and FAP^Low-Med (iCAF) among CAF-S1 (N = 8 treatment naïve, N = 12 after chemotherapy). Two-sided Mann-Whitney test. Data are presented as mean ± SEM.

Fig. 2. Inverse correlation between CAF-S1 and CD8⁺ T cell content upon chemotherapy.

A IHC of CD3⁺, CD8⁺ and FOXP3⁺ TILs before and after chemotherapy. Epithelial (E) and stromal (S) compartments, and stromal immune cell percentages are indicated. Scale bars, 50 μm and 25 μm (Inset). B Number of CD3⁺, CD8⁺ and FOXP3⁺ TILs per mm² of total HGSOC sections before and after chemotherapy (N = 35 patients, n = 70 matched samples). Two-sided Wilcoxon paired test. C Number of CD3⁺ and CD8⁺ TILs per mm² in HGSOC stromal and epithelial compartments, respectively (N = 35 patients). Two-sided Wilcoxon paired test. D Same as (B) reported to the stromal or epithelial content in each sample (N = 35). Two-sided Wilcoxon paired test. E Correlation matrix between variations (after/before chemotherapy) of FAP, SMA, FSP1 and CD29 H-scores and the number of CD3⁺, CD8⁺ and FOXP3⁺ TILs per mm² (N = 35). Variations are assessed by a delta score (Δ) (See Methods for calculation). Positive (red) and negative (blue) correlations with p-values < 0.1 (Spearman test after Benjamini & Hochberg correction for multiple testing) (N = 35). Square sizes are proportional to P-values and color intensities to the correlation coefficients (P-value is specified in square). F Correlation plots with linear regression lines comparing the variations upon chemotherapy in the content of CAF-S1 (assessed by ΔH-score FAP) and the number of CD3⁺ (Left), CD8⁺ (Middle) and FOXP3⁺ (Right) TILs per mm² of total section. Each dot represents one tumor (N = 35). Two-sided Spearman correlation test. G CD4 and CD8 protein levels in HGSOC samples (Left, treatment-naïve and Right, after chemotherapy). H % of CD3⁺ TILs among CD45⁺ cells (Left) and % CD8⁺ TILs among CD3⁺ T cells (Right) assessed by flow cytometry (N = 8 treatment naïve; N = 8 after chemotherapy). Two-sided unpaired Student t-test. I, J Correlations plots with linear regression lines between % of CD8⁺ TILs and % of FAP^High CAF-S1 (I) and FAP^Low-Med CAF-S1 (J) assessed by flow cytometry (N = 16 patients). Two-sided Spearman correlation test. Data are presented as mean ± SEM.

Fig. 3. The proportion of the ANTXR1⁺ ECM-myCAF cluster is the most reduced by chemotherapy in HGSOC.

A ANTXR1 staining in paired HGSOC. Scale bars, 50 μm. B ANTXR1 H-scores (N = 35 patients). Two-sided paired Wilcoxon test. C Flow cytometry plots showing FAP and ANTXR1 proteins to distinguish ANTXR1⁺ FAP⁺ myCAF and ANTXR1⁻ FAP⁺ iCAF. D Percentage of ANTXR1⁺ myCAF and ANTXR1⁻ iCAF among CAF-S1 population in HGSOC (N = 7 treatment-naïve, N = 9 after chemotherapy). Two-sided two-way ANOVA test. E–G ANTXR1 expression from scRNAseq (Left) and % of ANTXR1⁺ CAF-S1 among total CAF-S1 (Right) in HGSOC Prospective Curie cohort 2 (N = 12, 5 treatment-naïve and 7 after chemotherapy) (E), HGSOC Turku cohort (N = 11 paired before/after treatment) (F) and breast cancer (BC) cohort (N = 42, 31 treatment-naïve and 11 after chemotherapy) (G). Two-sided Wilcoxon test (E–G, Left) and two-sided Fischer’s Exact test (E–G, Right). H Left, UMAP of 5 618 CAF-S1 from HGSOC Curie cohort colored by treatment status (Up) or by CAF-S1 clusters predicted using label transfer (Bottom). Right, % of CAF-S1 clusters among CAF-S1. Two-sided Fisher’s exact test. I Same as in (H) for 5 658 CAF-S1 from HGSOC Turku cohort. J Percentage of CD8+ TILs in the retrospective SCANDARE Curie 2 cohort (N = 70 samples: 45 treatment-naïve, 25 after chemotherapy). Two-sided Mann-Whitney test. K Same as (J) % of ECM-myCAF (Left) and Detox-iCAF (Right). L, M Same as (J, K) according to response to chemotherapy for ECM-myCAF (L, Left), Detox-iCAF (L, Right) and CD8+ TILs (M). Two-sided Mann-Whitney test. N ANTXR1 expression in HGSOC Visium sections at time of diagnosis and in residual disease. Scale bars, 1 mm. O ANTXR1 expression (Left) and mean proportion of deconvoluted ANTXR1⁺ (black) and ANTXR1^- (grey) CAF-S1 among CAF-S1 (Right) in stromal spots of Visium sections (N = 4 Treatment naïve, n = 13 438 spots and N = 6 After chemotherapy, n = 24 205 spots). Two-sided Wilcoxon test. P Percentage of each deconvoluted CAF-S1 cluster in the 10 Visium sections. Two-sided Wilcoxon test. Data are presented as mean ± SEM. In boxplot the center line, box limits and whiskers indicate the median, upper and lower quartiles and 1.5x interquartile rage.

Fig. 4. Down-regulation of YAP1/TEAD-dependent pathway in ECM-myCAF after chemotherapy in HGSOC.

A The 50 most variables TFs before and after chemotherapy in the CAF-S1 HGSOC Curie scRNAseq dataset. B Expression of TEAD-target genes in CAF-S1 (Left, n = 1968 Treatment-naïve, n = 3 650 After chemotherapy) and in ECM-myCAF (Right, n = 1099 Treatment-naïve, n = 107 After chemotherapy). N = 12 HGSOC patients. Two-sided Mann Whitney test. C Expression score of TEAD-target genes (Left) and abundance of deconvoluted ECM-myCAF per spot (Right) in Visium sections. Scale bars, 1 mm. n = 13 438 Treatment naïve spots; 24,205 after chemotherapy spots). D Expression score of TEAD-target genes in stromal spots (n = 2410 spots). N = 4 Treatment-naïve; 6 after chemotherapy sections. Two-sided Wilcoxon test. E YAP1 staining in paired HGSOC. Scale bars, 50 μm. F YAP1 H-scores in stroma (Left) and epithelium (Right) (N = 35 patients). Two-sided paired Wilcoxon test. G Correlation plots with linear regression lines between H-scores of FAP and YAP1 in the stroma before (Left) and after chemotherapy (Right). Each dot represents one tumor (N = 35 patients). Two-sided Spearman correlation test. H Same as in (G) for ANTXR1 and YAP1 in stroma. I Representative views of nuclear YAP1 staining in paired HGSOC. Scale bars, 50 μm. J Left, % of CAF with nuclear YAP1 staining among total YAP1⁺ CAF, Right, % of cancer cells with nuclear YAP1 staining before and after chemotherapy (N = 35 patients). Two-sided paired Wilcoxon test. K Correlations with linear regression lines between FAP H-scores and % CAF with nuclear YAP1. Each dot represents one tumor (N = 35 patients). Two-sided Spearman correlation test. L Same as in (K) for H-scores of ANTXR1. M Merged staining of serial IHC for ANTXR1 and YAP1 before and after treatment. Scale bars, 50 μm. N YAP1 mean intensity in ANTXR1 + CAF in paired patients (n = 83,101 stromal cells analyzed). Two-sided Mann-Whitney test. O % of YAP1⁺ and YAP1^- ANTXR1 + CAF in paired patients (n = 83 101 stromal cells analyzed). Two-sided Fisher’s exact test. In boxplot the center line, box limits and whiskers indicate the median, upper and lower quartiles and 1.5x interquartile rage.

Fig. 5. ECM-myCAF segregate away from cytotoxic CD8 + T lymphocytes.

A Abundance of deconvoluted ECM-myCAF (Right) and differentiated CD8⁺ TILs (Left) per spot in Visium sections. The bottom right section shows distinct pathological responses after treatment, one non-responding with high proportion of residual ECM-myCAF (Left part) and one responding with a reduced ECM-myCAF content (Right part). B Non-negative matrix factorization of the deconvolution output on 4 Treatment-naïve; 6 after chemotherapy Visium sections. Colors and sizes of circles indicate the scaled cell type abundance. C Spatial distribution and score intensity of ECM-myCAF- and CD8⁺ T cell-enriched factors (Factor 6 and 7) on a HGSOC sample after treatment. D Distribution of the closest distances between CD8-enriched spots and ECM-myCAF-enriched or depleted spots (n = 10) in a radius of 1 mm. E Co-staining of Pan-cytokeratin, ANTXR1 and CD8 in a HGSOC section after treatment. Scale bar = 1 mm. F Cell segmentation of the section shown in (E). Colors represent clustering results identifying cancer cells (blue), CD8⁺ TILs (red), ANTXR1⁺ (green) and ANTXR1^- CAF (black). Scale bars = 1 mm. G Same as in (F) area considered for co-occurrence analysis shown in (H). H CD8⁺ TILs co-occurrence probability ratio (See Methods) with ANTXR1⁺ and ANTXR1^- CAF in (G). I ANTXR1 (brown) and CD8 (red) IHC co-staining in paired HGSOC sections. Scale bars, 50 μm. J Left, HGSOC residual sample segmentation. Scale bar = 50 μm. Right, CD8⁺ TILs co-occurrence probability ratio with ANTXR1⁺ and ANTXR1^- CAF in the Left section. Correlation plot with linear regression line between H-score of YAP1 in the stroma (K) or in the epithelium (L) and the number of total CD8⁺ TILs in HGSOC after chemotherapy. Each dot represents one tumor (N = 35). Two-sided Spearman correlation test. M ANTXR1 (brown) and CD8 (red) IHC co-staining with YAP1 stained on serial paired HGSOC sections (N = 2 paired HGSOC). Scale bars, 50 μm. N Left, Co-occurrence probability ratio between CD8⁺ T lymphocytes and residual YAP1⁺ ANTXR1⁺ CAF in HGSOC after treatment. Right, Neighbors enrichment score computed on the spatial connectivity graph between CD8⁺ T lymphocytes, YAP1⁺ ANTXR1⁺ CAF, and other CAF.

Fig. 6. ECM-myCAF dampen CD8⁺ T lymphocyte cytotoxicity through a YAP-1 dependent mechanism.

A Up, Representative flow cytometry plots showing CD8 and PD-1 protein levels in control condition (CD8⁺ T cells alone) (−) or co-cultured with Detox-iCAF or ECM-myCAF primary fibroblasts transfected with an untargeted siRNA (siCtrl) or with two different siRNA targeting YAP1 (siYAP1(1), siYAP1(2)). The population of interest (CD8⁺ PD-1⁺) is represented in red and the isotype control in black. Bottom, Bar plots showing the % of PD-1⁺ T cells among CD8⁺ T lymphocytes alone or in presence of Detox-iCAF or ECM-myCAF (Left) and transfected with siCtrl or siYAP1 (Middle and Right). Data are mean ± SEM (n = 5 independent experiments). P-values from paired Wilcoxon test. B−D Same as (A) for Granzyme B⁺ (B), Perforin⁺ (C) and IFN-γ⁺ (D) CD8⁺ T lymphocytes. P-values from paired Student t-test. E Left, Representative flow cytometry plots showing CAOV3 cell death after 24 h of incubation with CD8⁺ T lymphocytes pre-incubated with Detox-iCAF or ECM-myCAF primary fibroblasts transfected with an untargeted siRNA (siCtrl) or with two different siRNA targeting YAP1 (siYAP1(1), siYAP1(2)). Right, Bar plots showing the % of cancer cell death after incubation with CD8 + T lymphocytes. Data are mean ± SEM (n = 5 independent experiments). P-values from paired Student t-test. F Bar plots showing the % of migration of CD8 + T lymphocytes after 24 h of transwell co-culture with Detox-iCAF or ECM-myCAF primary fibroblasts transfected with an untargeted siRNA (siCtrl) or with two different siRNA targeting YAP1 (siYAP1(1), siYAP1(2)). Data are mean ± SEM (n = 8 independent experiments). P-values from unpaired Student t-test.

Fig. 7. Summary: impact of chemotherapy on CAF-S1 and CD8 T cell density in HGSOC.

HGSOC TME is heterogeneous and composed of different CAF subsets, including the FAP⁺ CAF-S1 and FAP^- CAF-S4 myofibroblastic populations. Chemotherapy globally increases the stromal content, but reduces the proportions of both CAF-S1 and CAF-S4 myofibroblasts. The combination of several high-throughput technologies shows that, among the CAF-S1 population, the content in the ANTXR1⁺ ECM-myCAF cluster decreases the most following chemotherapy, while the proportion of the ANTXR1^- detox-iCAF cluster increases. Concomitantly, chemotherapy increases CD8⁺ T lymphocyte density. Indeed, ECM-myCAF reduce CD8⁺ T cell cytotoxicity through the YAP1 signaling pathway, thereby explaining that the more the content in ECM-myCAF decrases after chemotherapy, the more CD8⁺ T cell infiltrate the tumor upon treatment. Figure 7 was created with Biorender.com.