GPX3 supports ovarian cancer tumor progression in vivo and promotes expression of GDF15

Abstract

Objective: We previously reported that high expression of the extracellular glutathione peroxidase GPX3 is associated with poor patient outcome in ovarian serous adenocarcinomas, and that GPX3 protects ovarian cancer cells from oxidative stress in culture. Here we tested if GPX3 is necessary for tumor establishment in vivo and to identify novel downstream mediators of GPX3's pro-tumorigenic function.

Methods: GPX3 was knocked-down in ID8 ovarian cancer cells by shRNA to test the role of GPX3 in tumor establishment using a syngeneic IP xenograft model. RNA sequencing analysis was carried out in OVCAR3 cells following shRNA-mediated GPX3 knock-down to identify GPX3-dependent gene expression signatures.

Results: GPX3 knock-down abrogated clonogenicity and intraperitoneal tumor development in vivo, and the effects were dependent on the level of GPX3 knock-down. RNA sequencing showed that loss of GPX3 leads to decreased gene expression patterns related to pro-tumorigenic signaling pathways. Validation studies identified GDF15 as strongly dependent on GPX3. GDF15, a member of the TGF-β growth factor family, has known oncogenic and immune modulatory activities. Similarly, GPX3 expression positively correlated with pro-tumor immune cell signatures, including regulatory T-cell and macrophage infiltration, and displayed significant correlation with PD-L1 expression.

Conclusions: We show for the first time that tumor produced GPX3 is necessary for ovarian cancer growth in vivo and that it regulates expression of GDF15. The immune profile associated with GPX3 expression in serous ovarian tumors suggests that GPX3 may be an alternate marker of ovarian tumors susceptible to immune check-point inhibitors.

Keywords: Antioxidant enzymes; GDF15; GPX3; Metabolism; Ovarian cancer.

Figure 1.. Knock-down of GPX3 decreases tumor burden in a syngeneic mouse ovarian cancer tumor model.

A. GPX3 was knocked down using 2 different shRNAs in mouse ID8 tumor cells. GPX3 expression assessed using sqRT-PCR (n=3 One-way ANOVA p=0.0013; Dunnett’s post-test). B. GPX3 decreases clonogenicity of ID8 cells (n=3 biological replicates. Superplot shows underlying technical replicates, One-way ANOVA p=0.0607; Dunnett’s post-test). C. ID8 tumor cells transduced with pLK0.1 control, shGPX3-41 or shGPX3-39 shRNAs were IP injected into immunocompetent C57BL/6J mice. GPX3 knock-down by shRNA-39 results in significant delay in the onset of ascites, which was used as endpoint for survival studies. Kaplan Meier curves show probability of ascites-free survival (pLK0.1 control n=12, shGPX3-41 n= 8, shGPX3-39 n=6; Log-rank Mantel-Cox test, P value = 0.0443). D. Quantification of total ascites volume collected at necropsy (Kruskal-Wallis test). E. Intraperitoneal tumor growth of ID8 tumor cells transfected with pLK0.1 control, shGPX3-41 or shGPX3-39 shRNAs in C57BL/6J mice was monitored using luminescence imaging F. Quantification of individual mouse tumor luminescence signals over time (pLK0.1 control n=12, shGPX3-41 n= 8, shGPX3-39 n=6; Total Flux p/s). G. Median tumor burden as assessed by luminescence on Day 54 (pLK0.1 control n=12, shGPX3-41 n= 8, shGPX3-39 n=6; Total Flux p/s; Kruskal-Wallis test). H. Tumor luminescence of isolated omenta at necropsy (pLK0.1 control n=12, shGPX3-41 n= 8, shGPX3-39 n=6; Kruskal-Wallis test). I. Tumor burden was detected on the diaphragm (i), peritoneal wall (ii) and omentum (iii). Representative images of mice at necropsy endpoint day 82 from pLK0.1control and shGPX3-41 groups are shown. A representative abdomen from the 5/6 mice that did not develop tumors, and the one of 6 mice in the shGPX3-39 group that developed tumors (lower panel) are shown from endpoint day 196.

Figure 2.. RNA sequencing demonstrates downregulation of pro-tumorigenic signaling pathways in response to GPX3 knock-down in OVCAR3 cells.

A. Volcano plot of differentially expressed genes (DEGs) following GPX3 knock-down using two different shRNAs (shRNA-76 & shRNA-78; n=3) relative to scramble shRNA control transduced OVCAR3 cells. B. Gene set enrichment analysis of RNA sequencing data following GPX3 knock-down (KEGG) demonstrates enrichment of tumor associated metabolic pathways in control cells relative to GPX3-knock-down cells. C. Small Molecule Query (L1000CDS²) was carried out to identify gene expression signatures that mimic or reverse gene expression changes following GPX3 knock-down. D. Transcription factor enrichment analysis (ENCODE & ChEA) on RNAsequencing data following GPX3 knock-down. E. Heat map of top differentially expressed protein coding genes (DEG) following GPX3 knock-down (z-scores; log 2-fold change >1, p adjust <0.05).

Figure 3.. Validation of GPX3-dependent GDF15 gene expression.

A. Stable shRNA mediated knock-down of GPX3 in OVCAR3 leads to decreased GDF15 mRNA expression (n=3, One Way ANOVA, Tukey’s post-test). B. shRNA-mediated knock-down of GPX3 in OVCAR4 cells results in decreased GDF15 mRNA expression (n=3, One Way ANOVA, Tukey’s post-test). C. Transient siRNA-mediated GPX3 knock-down in OVCA433 cells leads to decreased GDF15 mRNA expression (n=3, t-test). D. GDF15 mRNA expression in tumors from ID8 cells (from Figure 1), assessed by semi-quantitative real time RT-PCR (Kruskal-Wallis test). E. shRNA mediated knock-down of GPX3 in OVCAR3 cells decreases intracellular and F. Extracellular GDF15 protein expression (n=3, One Way ANOVA, Tukey’s post-test).

Figure 4.. GPX3 expression is associated with immune cell infiltration and PD-L1 expression.

A. Immune cell gene expression signatures established by Tamborero et al. (16) were correlated to GPX3 expression (z-score) in TCGA serous ovarian cancer specimens. GPX3 expression has significant positive correlation to macrophages, CD8+ T-cells and T-regs, and significant negative correlation with T-helper cells (Pearson r). B. Cibersort analysis of GPX3 expression to immune cell signatures in TCGA serous ovarian cancer specimens (Timer2.0, CIBERSORT-ABS, Pearson r, *P<0.05; **P<0.01; P<0.001). C. Expression of GPX3 mRNA in TCGA serous ovarian cancer specimens positively correlates with FOXP3 and PD-L1 expression (z-scores, n=300 specimens from TCGA with RNAseq data). D. Expression of GPX3 mRNA in ID8 tumors (from Figure 1) positively correlates with FOXP3 and PD-L1 expression.