Adverse genomic alterations and stemness features are induced by field cancerization in the microenvironment of hepatocellular carcinomas

Abstract

Hepatocellular Carcinoma (HCC) commonly develops in chronically damaged liver tissues. The resulting regenerative and inflammatory processes create an adverse milieu that promotes tumor-initiation and progression. A better understanding of the hepatic tumor-microenvironment interaction might infer profound therapeutic implications.Integrative whole genome and transcriptome analyses of different tumor regions, the invasive tumor border and tumor-surrounding liver (SL) were performed to identify associated molecular alterations and integrated with our existing HCC database. Expression levels and localization of established CSC markers were assessed in pre-neoplastic lesions and confirmed in two independent patient cohorts using qRT-PCR, immunohistochemistry and immunofluorescence.Our results indicate that genomic and transcriptomic profiles between SL and different tumor regions are quite distinct. Progressive increase in genetic alterations and activation of pathways related to proliferation as well as apoptosis were observed in the tumor tissue, while activation of stemness markers was present in cirrhotic SL and continuously decreased from pre-neoplastic lesions to HCC. Interestingly, the invasive tumor border was characterized by inflammatory and EMT-related gene sets as well as activation of pro-survival signaling. Consistently, integration of gene expression signatures with two independent HCC databases containing 300 HCCs revealed that border signatures are predictive of HCC patient survival.Prognostic significance of the permissive liver microenvironment might be a consequence of a pro-oncogenic field effect that is caused by chronic regenerative processes. Activation of key oncogenic features and immune-response signaling indicates that the cross-talk between tumor and microenvironment might be a promising therapeutic and/or preventive target.

Keywords: field effect; hepatocarcinogenesis; liver cancer; microenvironment; stemness features.

Figure 1. Transcriptomic and genomic profiles of the different regions

A. Venn diagram demonstrating the overlap between the different gene expression signatures. B. Unsupervised hierarchical cluster analysis of the different regions based on the corresponding significant genes (SL vs tumor (2630 genes): upper panel; SL vs border (590 genes): middle panel; border vs tumor (100 genes): lower panel) C. Gene set enrichment analysis (GSEA) for each of the regions in comparison to all other regions (rest). Normalized enrichment score (NES) reflects degree of overrepresentation for each group at the peak of the entire set. Statistical significance calculated by nominal P value of the ES by using an empirical phenotype-based permutation test. D. Graphical representation of the genetic alterations in each region determined by DNAcopy. Amplifications are depicted and red and losses are depicted in blue.

Figure 2. Activation of proliferation in tumor tissues

A. Gene set enrichment analysis (GSEA) of the tumor regions in comparison to all other regions (rest) indicate an activation of proliferative gene sets. Normalized enrichment score (NES) reflects degree of overrepresentation for each group at the peak of the entire set. Statistical significance calculated by nominal P value of the ES by using an empirical phenotype-based permutation test. B. Proliferation of cells determined by Ki67 staining. Dashed bars indicating the separation between SL and T regions. Right panel shows the graphical representation as number of positive cells estimated based on 10 randomly selected view fields (20x magnification). Statistical evaluation based on Friedman- test for multiple group comparisons followed by Dunns posthoc test. (n = 22; P-values: *≤ 0.05; **≤ 0.05; ***≤ 0.001). The data are presented as mean fold differences ± SD.

Figure 3. Activation of stemness markers in peritumoral tissues

A. Activation of stemness marker in the different regions was determined by confocal imaging. Representative images for AFP (red) and EpCAM (green) staining (upper panel) and CK19 (lower panel) containing all different regions (T = tumor, B = border, SL = surrounding liver) are displayed. Dashed bars indicating the separation between SL and T regions. White bar representing 100μm. Graphical representation and statistical evaluation (right graphs) for each marker based on h-score and Friedman- test for multiple group comparisons followed by Dunns posthoc test. (n = 15; P-values: *≤ 0.05; **≤ 0.05; ***≤ 0.001). The data are presented as mean fold differences ± SD. B. Activation of stemness marker EpCAM (green) by confocal microscopy. Images contains representation of all different regions (M = metastasis, B = border, SL = surrounding liver). White bar representing 200μm. Graphical representation and statistical evaluation (right graphs) for each marker based on h-score and Friedman- test for multiple group comparisons followed by Dunns posthoc test. (n = 5; P-values: *≤ 0.05; **≤ 0.05; ***≤ 0.001). The data are presented as mean fold differences ± SD.

Figure 4. Activation of inflammatory gene sets and immune cells in the invasive tumor margin

A. Gene set enrichment analysis (GSEA) of the surrounding liver regions in comparison to all other regions (rest) indicate an activation of gene sets involved in macrophage as well as T cell activation/function. Normalized enrichment score (NES) reflects degree of overrepresentation for each group at the peak of the entire set. Statistical significance calculated by nominal P value of the ES by using an empirical phenotype-based permutation test. B. Representative H&E stainings are shown in the upper graph. Immunohistochemistry of CD68 and CD3 demonstrating activated macrophages and T cells. Lower panel shows representative images of PD-1 staining reflecting impaired T cell function. Dashed bars indicating the separation between SL and T regions. White arrows indicating selected positive cells. Right panels show the corresponding graphical representations as number of positive cells estimated based on 10 randomly selected view fields (20x magnification). Statistical evaluation based on Friedman- test for multiple group comparisons followed by Dunns posthoc test. (n = 22; P-values: *≤ 0.05; **≤ 0.05; ***≤ 0.001).

Figure 5. Prognostic implications of the identified gene expression signatures

Integration of the different gene expression signatures with our previously published dataset of 53 HCC patients [19]. Unsupervised cluster analyses (left) and Kaplan-Meier plots of overall survival (right) for each of the identified signatures are shown: A. SL vs tumor (2630 genes): upper panel; B. SL vs border (590 genes): middle panel; C. border vs tumor (100 genes): lower panel).