POLE mutations improve the prognosis of endometrial cancer via regulating cellular metabolism through AMF/AMFR signal transduction

In this study, 530 endometrial cancer patients with whole exome sequencing data, 528 patients with clinical and survival information, and 526 patients with RNA expression data in the TCGA-UCEC project were downloaded from the TCGA database. In brief, the Mutation Annotation Format (MAF) data generated by the mutect2 algorithm from whole exome sequencing data, RNA sequencing raw counts (HTSeq-Counts) data, RNA sequencing FPKM data (HTSeq-FPKM), clinical and survival information were downloaded from the TCGA database through the TCGAbiolinks R/Bioconductor package [28].

The detailed mutational information was extracted from the MAF file by the Maftools R/Bioconductor package [29]. Next, the plotmafSummary function was used to plot the summary of the MAF file for displaying the number of variant type and variant classification. The oncoplot function was used to plot the OncoPlot of the top 10 mutated genes and POLE. The lollipopPlot of POLE was plotted through the lollipopPlot function [29].

We used the Kaplan–Meier method and compared it with the log-rank test to depict the survival curves. Cox proportional hazards regression analysis was used for univariate and multivariate analyses to explore the association of clinical characteristics, POLE mutational status, and patient prognosis.

The differentially expressed genes (DEGs) between the mutant and the wild type POLE endometrial cancer samples were analyzed by the edgeR R/Bioconductor package [30]. Gene ontology (GO) and the Kyoto Encyclopedia of Genes and Genomes (KEGG) were used for the enrichment analysis of the function and signaling pathway of the DEGs through the hypergeometric algorithm. The value of p < 0.05 was considered significantly enriched.

Gene Set Enrichment Analysis (GSEA) was performed through the GSEA Pre-ranked tool in the GSEA software (http://​www.​broadinstitute.​org/​gsea/​). The value of log₂(FC) calculated by the edgeR package was used as the ranking metric. DEGs and all the other genes were both analyzed by GSEA. We used the C5 collection, which contains gene sets annotated by GO terms, and the C2 canonical pathway sub-collection contains 1329 gene sets that are annotated pathways mainly from the KEGG, BIOCARTA and REACTOME databases in the analysis.

RNA sequencing FPKM data of 526 endometrial cancer samples in the TCGA database were used to assess the gene expression levels. Unsupervised hierarchical clustering was used to distinguish groups based on the expression pattern of target genes. Expression values of these genes in the 526 samples (where rows indicate the identity of the genes, columns indicate the identity of the patients) were clustered using hierarchical clustering with Euclidean distance and ward linkage by the dendextend R package [31] and heatmap.2 function in the gplots R package.

In this study, 530 endometrial cancer patients containing the whole exome sequencing data in the TCGA-UCEC project were included for subsequent analysis.

Somatic alterations on a multiscale were identified in the exome of each endometrial cancer case including single-nucleotide polymorphisms (SNPs), short insertions (INS), and short deletions (DEL). In all 530 exome sequencing data of endometrial cancer samples, 886,371 somatic mutations were identified based on consensus calls from the mutect2 mutation-detection algorithm. 73.21% of endometrial cancer patients (Number = 388) contained less than 500 somatic mutations, 16.98% (Number = 90) contained 500–2000 somatic mutations, 2.64% (Number = 14) contained 2000–5000 somatic mutations and 7.17% (Number = 38) contained more than 2000 somatic mutations (Fig. 1a). The survival curve of endometrial cancer patients based on the number of mutations (mutation number > 500 or < 500) showed that the endometrial cancer patients had an improved prognosis when they held more mutations (P = 0.00033) (Fig. 1b). The patients with POLE somatic mutations had a higher tumor mutational burden compared with patients without POLE mutations (P = 1.26e-76) (Fig. 1c). These results suggest that POLE mutations may be an important prognostic factor in patients with endometrial cancer. In these somatic mutations, the major type of variants were SNPs (Number = 821,967), which was mainly the missense mutation (Additional file 1: Table S1). Silent, intron, UTR, 5′ flank, 3′ flank, splice region, and other synonymous SNPs were filtered and only non-synonymous variants were selected for further analysis. In the non-synonymous variants, most variants were SNPs and the major mutation was missense mutation (Additional file 1: Figures S1A, B). The top 10 mutated genes in endometrial cancer were: PTEN (57%), PIK3CA (48%), TTN (44%), ARID1A (43%), TP53 (36%), MUC16 (30%), PIK3R1 (30%), KMT2D (27%), CTCF (24%), and CSMD3 (24%). In addition, 15% of endometrial cancer patients had POLE somatic mutations as well, where more than 90% of POLE mutations were missense mutations (Fig. 1d).

POLE was the 123rd most frequently mutated gene in endometrial cancer with a mutational frequency of 80 in the 530 tumor samples (15%) (Fig. 1b, Additional file 1: Table S2). We identified 144 mutations in the POLE gene including 140 SNPs, 3 INS, and 1 DEL. In detail, the 144 mutations included 131 missense mutations, 7 nonsense mutations, 3 Frame_Shift_Ins, 1 Frame_Shift_Del, and 2 Splice_Site mutations (Additional file 1: Figures S1C, D). Most SNPs in POLE were substitutions between guanine and adenine (G/A) (Additional file 1: Figure S1E). Except for the 2 Splice_Site mutations, 142 somatic mutations occurred in exon regions containing 39 exons and more than 33% mutations occurred in exon9 (17.6%) and exon13 (16.2%) (Fig. 1e). The remaining 66.2% mutations were relatively evenly distributed in the other 37 exons (all exons contained less than 5% mutations). Most of the POLE mutations were missense mutations and these mutations were equally distributed from the N-terminal to the C-terminal of the POLE protein (Fig. 1f).

A total of 528 endometrial cancer patients with clinical and survival information, including histological type, histologic grade, clinical stage, age at initial pathologic diagnosis, hypertension, history of neoadjuvant treatment, and diabetes in the TCGA-UCEC project, were used in this series. The detailed clinical characteristics are listed in Table 1. To explore the relationship between the POLE mutations and the clinical characteristics, we employed the chi-square test to establish the relationship between the POLE mutations and histological type, histologic grade, clinical stage, and age at initial pathologic diagnosis. The results showed that POLE mutational status was significantly associated with the histological type (P = 0.0006766) and age at the initial pathologic diagnosis (P = 4.39e-05). No significant relationship was found between the POLE mutational status and the histologic grade or clinical stage (P > 0.05) (Additional file 1: Table S3).

The correlation between patient survival and histologic grade, clinical stage and POLE mutations, respectively, was analyzed by the Kaplan–Meier method and log-rank test. The results showed all three features correlate with patient survival with a P value less than 0.05 (Figs. 2a–c). Next, we conducted a Cox proportional hazards regression analysis to further investigate the correlation between patient overall survival and the four clinical features, POLE mutations, respectively. The P values of all five features were less than 0.05 in the univariate analysis, whereas, in the multivariate analysis, clinical stage (P = 2.50E-08), histologic grade (P = 0.00378), and POLE (P = 0.00194) mutational status were shown to be independent prognostic factors in endometrial cancer patients (Table 2). The results showed that endometrial cancer patients with low histologic grade (G1), early clinical stage (I+ II), and POLE mutations had favorable prognosis (Figs. 2a–c). Compared with the clinical stage, which showed the smallest P value in the Cox proportional hazards regression analysis and Kaplan–Meier curve in all three independent prognostic factors, histologic grade and POLE mutations were good stratified factors because it did not intersect in their survival curves.

Previous results suggest that POLE mutational status was associated with tumor mutational burden in endometrial cancers [32] consistent with our finding (Fig. 1c). To a certain extent, it is also possible that the hypermutated status might affect the prognosis of endometrial cancer patients. In order to increase the accuracy of the impact from POLE mutations and minimize the interference of hypermutated status on patient prognosis, we excluded hypermutated phenotypes (more than 500 mutations per sample) and found that the total tumor mutational burden decreased significantly. The survival curve of endometrial cancer patients showed that the exclusion of patients with hypermutated phenotypes altered the effects of POLE mutations on patient prognosis, hence the POLE mutations were no longer significantly associated with improved prognosis (Additional file 1: Figure S2). The variation in the relationship between POLE mutations and prognosis may be due to the limited sample size of the POLE mutant after removing the hypermutated phenotypes (Number = 10). The effect of mutational burden could not be considered in this study due to the limited number of samples.

Endometrial cancer patients with POLE mutations had favorable prognoses, indicating that mutant POLE may play an important role in the development and progression of endometrial cancer. To further investigate the role of POLE mutations in endometrial cancer, we used transcriptome sequencing data from 526 patients to analyze the global gene expression profile and the differential expression between the two groups in 80 POLE mutant patients and 446 patients with wild type POLE. A total of 1587 differentially expressed genes (DEGs) were obtained according to the cut-off criteria (P < 0.05 and FC > 1.5) including 725 upregulated genes and 862 downregulated genes (Figs. 3a). To clarify the functional differences between these two groups, we conducted the gene set enrichment analysis (GSEA), which is a bioinformatics method that determines whether a set of terms shows statistically significant enrichment [33]. GSEA using Gene Ontology (GO) terms as the gene sets showed the enrichment in categories like glucose metabolic process, isomerase activity, exonuclease activity, and intramolecular transferase activity in endometrial cancer samples with POLE mutations (Figs. 3b–e). The T cell receptor signaling pathway and the production of molecular mediators of immune response were also significantly enriched in the GSEA results (Additional file 1: Figures S3A, B). GO analysis using the hypergeometric algorithm of 1587 DEGs was similar as those of the GSEA, which indicated that cytokine receptor activity, cytokine binding, cytokine receptor binding, chemokine receptor activity, and carbohydrate binding were enriched (Additional file 1: Figure S3C).

To further enlighten us in targeting the cellular metabolism for POLE mutations, canonical pathways mainly from the KEGG, BIOCARTA, and REACTOME databases as the gene sets were used for GSEA. The results showed that cellular glucose metabolism such as glycolysis and gluconeogenesis were significantly enriched in endometrial cancer samples with POLE mutations (Figs. 4a, b). These findings suggest that POLE mutations can regulate cellular glucose metabolism in endometrial cancer patients.

In order to further study the function of POLE mutations in cellular metabolism through glucose metabolism, we analyzed the expression levels of 20 genes in glucose metabolism and 18 genes in glycolysis in endometrial cancer samples between the POLE mutant and the wild type. Endometrial cancer samples were reclassified into two clusters based on the gene expression patterns of these two pathways according to hierarchical clustering. We also examined the mutation rate of POLE according to hierarchical clustering. In glucose metabolism, cluster I had a higher gene expression pattern and exhibited a higher rate of POLE mutations; yet cluster II had a lower expression pattern and lower rate of POLE mutations (Fig. 4c). Therefore, the elevation of glucose metabolism significantly correlated with a higher rate of POLE mutations. Similarly, two main groups were also observed in glycolysis. Cluster I had a higher gene expression pattern and exhibited a higher rate of POLE mutations, while cluster II had a lower expression pattern and a lower rate of POLE mutations (Fig. 4d). Therefore, the elevation of cellular glucose metabolism significantly correlated with a higher rate of POLE mutations. These results suggest that the favorable prognosis of POLE mutants may be partly due to the regulation of cell metabolism.

To a certain extent, the results of this study have demonstrated that POLE mutations may impact endometrial cancer patient prognosis by regulating cellular glucose metabolism. However, the key molecules involved in the regulation of glucose metabolism by the POLE mutations are still unclear. To further target the possible core genes, we analyzed the process of glucose metabolism in detail. Phosphoglucose isomerase (PGI) is a glycolytic enzyme involved in the gluconeogenesis–glycolysis pathways. It is an extracellular cytokine as well, under the name of autocrine motility factor (AMF). Therefore, AMF/PGI plays a dual role as a phosphoglucose isomerase that catalyzes the interconversion of glucose-6-phosphate and fructose-6-phosphatein glycometabolism when it is effective as a cytokine [34‐36]. Transcriptomic profiling analysis of 587 endometrial cancer and normal samples revealed differences in the expression of AMF/PGI between endometrial cancer patients and normal samples. In endometrial cancer patients, the expression of AMF/PGI (fold change = 2.51, P = 6.3e-31) increased significantly (Fig. 5a). This result suggests that AMF/PGI may play an important role in endometrial cancer. On the other hand, the expression of AMF/PGI has also been detected in POLE mutant and wild type endometrial cancer patients. In endometrial cancer patients with POLE mutations, the expression of AMF/PGI was slightly higher than that in patients with wild type POLE (FC = 1.2, P = 0.00077), (Fig. 5b). AMFR/gp78 is a cell surface receptor for AMF/PGI, which is also located in the endoplasmic reticulum where it functions as an E3 ubiquitin ligase. Stimulation of AMFR/gp78 by its ligand AMF/PGI alters cell adhesion, proliferation and apoptosis. AMFR/gp78 expression has been reported in numerous types of human tumors [35, 37]. In POLE mutants and wild type endometrial cancer patients, the expression pattern of AMFR/gp78 was similar as that of AMF/PGI (FC = 1.26, P = 0.00002) (Fig. 5c). To further confirm the correlation between POLE mutations and the expression of these two genes, we used the chi-square test to establish the relationship between the POLE mutations and the expression levels of these two genes. The results showed that POLE mutational status was significantly associated with the expression of AMF/PGI (P = 7.041e-06) and AMFR/gp78 (P = 0.00017).

To further explore the relationship between AMF/PGI or AMFR/gp78 and prognosis of endometrial cancer patients, we analyzed the correlation between the expression levels of these two genes and patient survival by the Kaplan–Meier method and log-rank test. The patient survival curves showed no significant correlation between the expression levels of AMF/PGI or AMFR/gp78, respectively, and the prognosis of endometrial cancer patients (Additional file 1: Figures S4A, B). One interesting result was that the comprehensive expression of POLE and AMFR/gp78 was significantly correlated with patient prognosis. Patient survival curves showed high expressions of AMFR/gp78 and low expression of POLE, which were associated with the favorable prognosis of endometrial cancer patients (Fig. 5d). These results confirm the relationship between POLE mutations, AMF/PGI, AMFR/gp78 expression, and patient prognosis, suggesting that POLE mutations may play a vital role in patient prognosis by affecting the expression and signal transduction of AMF/PGI–AMFR/gp78 as well as affecting cell metabolism.