DAB2+ macrophages support FAP+ fibroblasts in shaping tumor barrier and inducing poor clinical outcomes in liver cancer

Background: Cancer-associated fibroblasts (CAFs) are the key components of the immune barrier in liver cancer. Therefore, gaining a deeper understanding of the heterogeneity and intercellular communication of CAFs holds utmost importance in boosting immunotherapy effectiveness and improving clinical outcomes. Methods: A comprehensive analysis by combing single-cell, bulk, and spatial transcriptome profiling with multiplexed immunofluorescence was conducted to unravel the complexities of CAFs in liver cancer. Results: Through an integrated approach involving 235 liver cancer scRNA-seq samples encompassing over 1.2 million cells, we found that CAFs were particularly increased in hepatocellular carcinoma (HCC) and intrahepatic cholangiocarcinoma (ICC). FAP+ fibroblasts were identified as the dominant subtype of CAFs, and which were mainly involved in extracellular matrix organization and angiogenesis. These CAFs were enriched in the tumor boundary of HCC, but diffusely scattered within ICC. The DAB2+ and SPP1+ tumor-associated macrophages (TAMs) reinforce the function of FAP+ CAFs through signals such as TGF-β, PDGF, and ADM. Notably, the interaction between DAB2+ TAMs and FAP+ CAFs promoted the formation of immune barrier and correlated with poorer patient survival, non-response to immunotherapy in HCC. High FAP and DAB2 immunohistochemical scores predicted shorter survival and higher serum AFP concentration in a local clinical cohort of 90 HCC patients. Furthermore, this communication pattern might be applicable to other solid malignancies as well. Conclusions: The interaction between DAB2+ TAMs and FAP+ CAFs appears crucial in shaping the immune barrier. Strategies aimed at disrupting this communication or inhibiting the functions of FAP+ CAFs could potentially enhance immunotherapy effectiveness and improve clinical outcomes.


Figure S1 .
Figure S1.Fibroblast subtype identification.(A) Dot plot showing marker gene expression of major cell types.(B) UMAP shows the sample type and individual sample distribution of all cells.(C) Stacking plot shows the proportion of major cell types in peripheral blood and five tissue samples.(D) UMAP shows the sample type and individual sample distribution of fibroblasts.(E) Feature plot shows the marker gene expression of selected fibroblast subtypes.(F) Heatmap showing sample preference of fibroblast subtypes, where OR > 1.5 was considered significantly enriched for that cell in that type of sample, and OR < 0.5 was considered significantly not enriched.(G) Volcano plot showing the difference in the proportion of major cell types in ICC (n = 31) versus AL (n = 14).(H) Volcano plot comparing the relative abundance of fibroblast subtypes in ICC versus HCC.(I) Heatmap showing the Ucell enrichment scores of key biological entries in the Fb_03_FAP subtype in different tissue types.(J) Bar plot shows the proportion of FAP + CAF from TCGA samples after deconvolution by CIBERSORTx.Paired point plot shows the high proportion of FAP + CAF in the paired tumor samples from the single-cell discovery cohort.*, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

Figure S2 .
Figure S2.Functional analysis of fibroblast subtypes.(A) Dot plot showing selected biological terms or pathways significantly enriched for each fibroblast subtype.(B) Heatmap shows the enrichment scores of relevant metabolic pathways for fibroblast subtypes calculated by R package scMetabolism; bar plot shows the overall metabolic score of each fibroblast subtype, with the vertical coordinate being the value after scaling of metabolic score.(C) Heatmap showing the enrichment of 50 key cancer hallmarks in different fibroblast subtypes.

Figure S3 .
Figure S3.Comparison of Scissor cell-related genes and fibroblast subtypes.(A) Volcano plot showing genes differentially expressed between Scissor + and Scissor -cells.Genes with |logFC| > 2, adj.P < 0.05 were considered significantly different.(B) Dot plot showing the expression preference of genes highly expressed in Scissor + or Scissor - cells in HCC and ICC.

Figure S4 .
Figure S4.Fluorescent expression of FAP in tumor and paracancer.Immunofluorescence images show that FAP is more abundantly expressed in tumor samples compared to paracancer.Viewed by SlideViewer with field of view sizes of 2000 μm and 400 μm.

Figure S5 .
Figure S5.Correlation and co-localization analysis of fibroblasts with macrophages.(A) Heatmap showing Spearman's correlation of the proportion of fibroblasts infiltrating with other major cell types in five independent bulk transcriptome cohorts based on CIBERSORTx deconvolution imputation.Scatterplots show the correlation between the proportion of fibroblasts infiltrated with macrophages.(B) Heatmap showing the correlation of spatial localization between cells based on R package Cards imputed in HCC and ICC ST slides, high correlation indicates high spatial localization between cells.(C) Cell score ST plot showing the presence of spatial co-localization of imputed fibroblasts and macrophages.

Figure S6 .
Figure S6.Identification of key TAMs.(A) UMAP shows the distribution of sample type and individual sample in macrophage subtypes.(B) Dot plot shows the expression of

Figure S7 .
Figure S7.Gene expression and spot annotation of ST boundary slides.(A) Spatial feature plot showing spatial expression of FAP + CAF and selected macrophage subtype marker genes.(B) Unbiased clustering of ST spots in HCC3 and HCC4 slides and cell type annotation for each cluster.Dot plots showing the expression of specific marker genes for each cluster in HCC3 and HCC4 slides.(C) Dot plot showing the results of GO enrichment analysis of the co-localized regions of FAP + CAF and TAM.

Figure S9 .
Figure S9.FAP + CAF recruit macrophages and promotes M2 polarization.(A) Heatmap showing the prior interaction potential of ligand and receptor from TAM to FAP + CAF.The dots represent genes significantly associated with survival of TCGA-LIHC patients (Cox P < 0.05), red represents better prognosis (HR > 1) and blue color represents worser prognosis (HR < 1).(B) Spatial dot plot showing the spatial expression of PDGFB in DAB2 + TAM and corresponding receptor genes in FAP + CAF in HCC1_L and HCC2_L slides.(C) Circos plot shows the weights of signal sent by FAP + CAF to other cell types in HCC or ICC; heatmap shows the weights of signaling exchange between all cell types.(D) Signal enrichment analysis based on cellular communication presents the strength of efferent signaling enrichment pathways of different cell types in HCC or ICC.(E) Heatmap shows the prior interaction potential of ligand and receptor from FAP + CAF to TAM; dot plot shows the expression of ligand and receptor genes in fibroblast subtypes and macrophage subtypes.(F) Dot plot showing the pathways to which FAP + CAF ligand genes are significantly enriched.

Figure S10 .
Figure S10.Cell communication of FAP + CAF in ICC.(A) Black heatmap shows significant top ligands and receptors between FAP + CAF and tumor or endothelial cells; red

Figure S11 .
Figure S11.Drugs predicted to block TAM-FAP + CAF interaction.(A) Dot plot showing the targeted drugs with the high correlation with the LRscore based on TCGA-LIHC samples.A higher positive correlation means that the drug is more sensitive for patients with a low LRscore, and a negative correlation means that the drug is more sensitive for patients with a high LRscore.(B and C) Bar plot showing the small molecule drugs predicted by the R package sc2MeNetDrug to block TAM-FAP + CAF communication, where a high negative enrichment score represents a higher likelihood that the drug will work; the network plot shows clustering based on drug structure.

Figure S12 .
Figure S12.Single-cell and bulk pan-cancer analysis of FAP + CAF and DAB2 + TAM.(A and B) UMAP show the identification of FAP + CAF and DAB2 + TAM and associated gene expression from the integrated pan-cancer single-cell cohort.(C) Bar plot showing the proportion of FAP + CAF and DAB2 + TAM infiltration based on gene FAP and DAB2 expression grouping.(D) KM curves showing that high FAP