Integrating the dysregulated inflammasome-based molecular functionome in the malignant transformation of endometriosis-associated ovarian carcinoma

Abstract

The coexistence of endometriosis (ES) with ovarian clear cell carcinoma (CCC) or endometrioid carcinoma (EC) suggested that malignant transformation of ES leads to endometriosis associated ovarian carcinoma (EAOC). However, there is still lack of an integrating data analysis of the accumulated experimental data to provide the evidence supporting the hypothesis of EAOC transformation. Herein we used a function-based analytic model with the publicly available microarray datasets to investigate the expression profiling between ES, CCC, and EC. We analyzed the functional regularity pattern of the three type of samples and hierarchically clustered the gene sets to identify key mechanisms regulating the malignant transformation of EAOC. We identified a list of 18 genes (NLRP3, AIM2, PYCARD, NAIP, Caspase-4, Caspase-7, Caspase-8, TLR1, TLR7, TOLLIP, NFKBIA, TNF, TNFAIP3, INFGR2, P2RX7, IL-1B, IL1RL1, IL-18) closely related to inflammasome complex, indicating an important role of inflammation/immunity in EAOC transformation. We next explore the association between these target genes and patient survival using Gene Expression Omnibus (GEO), and found significant correlation between the expression levels of the target genes and the progression-free survival. Interestingly, high expression levels of AIM2 and NLRP3, initiating proteins of inflammasomes, were significantly correlated with poor progression-free survival. Immunohistochemistry staining confirmed a correlation between high AIM2 and high Ki-67 in clinical EAOC samples, supporting its role in disease progression. Collectively, we established a bioinformatic platform of gene-set integrative molecular functionome to dissect the pathogenic pathways of EAOC, and demonstrated a key role of dysregulated inflammasome in modulating the malignant transformation of EAOC.

Keywords: endometriosis; gene expression microarray; gene-set integrative analysis; inflammasome; ovarian carcinoma.

Figure 1. Work flow of the two-stage strategy to discover gene signatures for EAOC

(A) Workflow of the gene set regularity model. The gene set regularity (GSR) index was computed by converting the gene expression ordering of gene elements in a gene set through the Gene Ontology (GO) term or canonical pathway databases. The informativeness of the GSR index was assessed by the accuracy of recognition, classification, and prediction by machine learning using binary or multiclass classifications. Functionome analyses were carried out to investigate the pathogenesis of endometriosis (ES), clear cell carcinoma (CCC), endometrioid ca (EC) and endometriosis-associated ovarian carcinoma (EAOC) by statistical methods, hierarchical clustering, and exploratory factor analysis. (B) Heatmaps and dendrogram for the three diseases. The dendrogram (left side of the heatmap) showed the relationship of the three diseases. When displayed on the heatmap, each of the three diseases computed through either the GO term gene sets showed a distinct pattern. However, the patterns were more similar between CCC and EC.

Figure 2. DNA microarray gene expression data mining of deregulated functions involving in the malignant transformation of EAOC

(A) Venn diagram of the deregulated GO term elements from exploratory factor analysis for the three diseases. The figure showed the results of the three diseases with the total factor elements from each of the disease. Their relationship was displayed on the Venn diagram to show the gene set numbers of all possible logical relations among the three diseases. The 9 commonly deregulated GO terms among ES, CCC and EC were listed on the right side of the figure. (B) The nine commonly deregulated GO terms among the ES, CCC, and EC, including ‘inflammasome complex’ was shown.

Figure 3. GO tree analysis

The GO tree of deregulated functions of CCC establish with the significant GO terms involving in the inflammation and immune system. After mapping to the GO tree, the similar or related GO terms were clustered together and shown the parent-child relationship. The table listed the immune or inflammation-related GO terms, the GOIDs and their p values in the GO trees.

Figure 4. Inflammasome complex correlate with survival outcome in EAOC patients

Kaplan–Meier plotter survival curves showed significant difference of EAOC survival with different expression level of inflammasome complex (NLRP3, AIM2, PYCARD, NAIP). HR = 5.14, 95% CI 1.47 to 17.92, p-value = 0.044; HR = 5.71, 95% CI 1.31 to 24.85, p-value = 0.086; HR = 2.62, 95% CI 1.03 to 6.65, p-value = 0.035; HR = 2.81, 95% CI 1.11 to 7.14, p-value = 0.023, respectively.

Figure 5. The survival of EAOC patients are correlate with inflammasome-related genes

Kaplan–Meier plotter survival curves showed significant difference of EAOC survival with different expression level of inflammasome-related genes (TNF, FOXO3, TLR7, NFKBIA). HR = 6.08, 95% CI 1.4 to 26.49, p-value = 0.061; HR = 3.15, 95% CI 1.24 to 8.02, p-value = 0.011; HR = 3.96, 95% CI 1.14 to 13.73, p-value = 0.019; HR = 3.13, 95% CI 1.03 to 9.53, p-value = 0.034, respectively.

Figure 6. Interaction analysis of identified genes

(A) The identified potential involving genes were subjected to a protein-protein interaction (PPI) analysis by establishing an interactive network from the STRING database (https://string-db.org). As members of inflammasome complex and inflammasome-related genes, their proteins showed intensive interactions. The average node degree is 3.56, and the PPI enrichment p-value is 3.33x10^-15, significantly more interactions than expected. (B) The p values of each gene in the three diseases were showed in the chart. The progressive changes of p values from ES to CCC and EC demonstrated that the NLRP3, AIM2, PYCARD, NAIP, TLR7, NFKBIA, TNF, FOXO3 would be the potential markers of prognosis in EAOC.

Figure 7. Immunohistochemistrical analysis of clinical samples from patients with ES, EC, and CCC

(A) Clinical samples from patients with ES (n = 13), EC (n = 15), and CCC (n = 15) were immunostained with anti-AIM2 antibody. (B) The expression levels of AIM2 in all clinical samples were quantified and presented in the chart. The mean values of AIM2 expression in EC and CCC were higher than that in ES. (C-E) Samples were stained with Ki-67 and AIM2. The case numbers of ES, EC, and CCC with high and low expression levels of Ki-67 and AIM2 were calculated and displayed in the chart. The percentages of each combination were also calculated. The AIM2 levels was positively correlated with Ki-67 levels.

Figure 8. Working model of the inflammasome in endometriosis associated ovarian cancer

This model presents the microenvironment in endometrioma of the ovary. Retrograded menstruation accumulated in ovary provoked DAMPs and caused chronic inflammation. Inflammasome related genes (NLRP3, AIM2, PYCARD, NAIP, TNF, FOXO3, TLR7, NFKBIA) were activated subsequently. Activated caspase can lead to cell pyroptosis with the consequence of the release of inflammatory cytokines. Finally, inflammatory cytokines induced oncogene over-expression then produced EAOC carcinogenesis.