Large‐scale transcriptome profiles reveal robust 20‐signatures metabolic prediction models and novel role of G6PC in clear cell renal cell carcinoma

Abstract Clear cell renal cell carcinoma (ccRCC) is the most common and highly malignant pathological type of kidney cancer. We sought to establish a metabolic signature to improve post‐operative risk stratification and identify novel targets in the prediction models for ccRCC patients. A total of 58 metabolic differential expressed genes (MDEGs) were identified with significant prognostic value. LASSO regression analysis constructed 20‐mRNA signatures models, metabolic prediction models (MPMs), in ccRCC patients from two cohorts. Risk score of MPMs significantly predicts prognosis for ccRCC patients in TCGA (P < 0.001, HR = 3.131, AUC = 0.768) and CPTAC cohorts (P = 0.046, HR = 2.893, AUC = 0.777). In addition, G6PC, a hub gene in PPI network of MPMs, shows significantly prognostic value in 718 ccRCC patients from multiply cohorts. Next, G6Pase was detected high expressed in normal kidney tissues than ccRCC tissues. It suggested that low G6Pase expression significantly correlated with poor prognosis (P < 0.0001, HR = 0.316) and aggressive progression (P < 0.0001, HR = 0.414) in 322 ccRCC patients from FUSCC cohort. Meanwhile, promoter methylation level of G6PC was significantly higher in ccRCC samples with aggressive progression status. G6PC significantly participates in abnormal immune infiltration of ccRCC microenvironment, showing significantly negative association with check‐point immune signatures, dendritic cells, Th1 cells, etc. In conclusion, this study first provided the opportunity to comprehensively elucidate the prognostic MDEGs landscape, established novel prognostic model MPMs using large‐scale ccRCC transcriptome data and identified G6PC as potential prognostic target in 1,040 ccRCC patients from multiply cohorts. These finding could assist in managing risk assessment and shed valuable insights into treatment strategies of ccRCC.


| INTRODUC TI ON
Renal cell carcinoma is one of the most common malignant tumours of the urogenital system, accounting for about 5% of all new cases of adult males and 3% of new cases of females. 1 According to statistics in the United States, there are about 73 820 new cases of kidney cancer and 14 770 deaths in 2019. 2 Clear cell renal cell carcinoma (ccRCC) is the most common and highly malignant pathological type of kidney cancer (accounting for 70%-85%). About 25%-30% of ccRCC patients have metastases at first diagnosis, and the 5-year survival rate of metastatic ccRCC is only 32%. In addition, even ccRCC patients who are initially effective in treatment will have disease progression after a period of time, at which time most patients will lack subsequent effective treatment.
In recent years, new immunotherapy represented by PD-1/PD-L1, CTLA4 inhibitors has rapidly emerged in the field of ccRCC treatment and has shown encouraging results for patients with advanced refractory disease. 3 In 2020, ASCO GU published the 5-year follow-up results of the CheckMate 025 study, showing the 5-year survival rate of monoclonal antibody second-line treatment is as high as 26%, which demonstrates the advantages of immunotherapy's survival benefits and new chapter of treatment strategies for high-risk ccRCC patients. [4][5][6] Immune check-point inhibitors combined with TKI play a variety of roles from the perspective of inducing anti-tumour immune normalization, inhibiting the development of advanced ccRCC and regulating tumour microenvironment (TME). Its success largely depends on deep understanding of tumour cells and TME interaction. 7,8 With the deepening of research, more evidence shows that not only the efficacy of immunotherapy depends on the activation of the tumour immune microenvironment, but also the efficacy of traditional treatments such as targeted therapy also depends on the strength of individual anti-tumour immune response. [9][10][11] Thus, it is of great significance exploring the underlying mechanism of TMEdriven tumorigenesis and development, improving the efficiency of various existing treatments and discovering novel precise targets for ccRCC therapies.
In TME, tumour cells and immune cells reprogram their metabolic patterns to adapt to the microenvironment of hypoxia, acidity and low nutrition. 12 For example, tumour cells show enhanced aerobic glycolysis (Warburg effect) but reduce oxidative phosphorylation, which has a great effect on T cell-mediated anti-tumour immune response and tumour-infiltrating myeloid cell activity; macrophages tend to be M2-type polarization, showing up-regulated fatty acid synthesis and β-oxidation. 13 The activation of tumour cell pro-cancer signals not only affects its own malignant biological behaviour, but also promotes the development of tumours. 14 It can also deflect the functional phenotype of tumour-infiltrating immune cells by changing the metabolic secretion profile and TME of tumour cells and induce the formation of tumour immune escape. 14,15 Therefore, the metabolic reprogramming of tumour cells and immune cells is crucial for understanding the game process of tumour cells' evil behaviour, tumour immune response and tumour immune escape and provides new directions for regulating tumour immunity.
The methods of constructing clinical prediction models based on retrospective data help researches realize the prediction of clinical outcomes with several prognostic factors, which is more likely to change our clinical practice and has strong clinical guidance value. 16 In 2018, RCClnc4 has been demonstrated to have precise prognostic significance in early ccRCC. This study aimed to first establish and validate an effective prognostic metabolic prediction models enrolled large-scale transcriptome metabolic genes for ccRCC patients. We suggested that the metabolic prediction models (MPMs) classifier could facilitate risk management and treatment strategies for ccRCC patients and identify novel targets in the MPMs co-network.

| Identification of metabolic differentially expressed genes
Forty-one metabolic pathways were selected according to KEGG pathways atlas. The 911 metabolic genes were utilized for identification of significant metabolic differentially expressed genes (MDEGs) using Limma R package (Version 3.6.3) with FDR < 0.05 and |logFC|>0.5. The intersective metabolic genes between TCGA and CPTAC cohorts were selected for further analyses.

| Cox regression analysis and receiver operating characteristic curve construction
All ccRCC patients from TCGA and CPTAC cohorts with complete transcriptome information and relevant clinical pathologic parameters were included for subsequent analysis. Univariate and multivariable Cox regression analyses were used to evaluate the independent prognostic value of the metabolic clusters using forest plot with the survival R package. The receiver operating characteristic curve (ROC) was constructed for traditional clinical pathologic parameter and the risk score of MPMs in both TCGA and CPTAC cohorts using survival ROC R package. The area under the curve (AUC) was utilized to assess the predictive value of these prognostic signatures. In addition, Nomogram was developed on the basis of all the independent prognostic factors in TCGA cohort.

| Tumour microenvironment purity assessment
ESTIMATE algorithm was utilized to evaluate total and immune scores using estimate package (http://r-forge.rproj ect.org; repos = rforge, dependencies = TRUE) in patients from TCGA cohort. Associated between tumour microenvironment (TME) purity and risk score of MPMs or G6PC expression was assessed using Pearson's r test.

| Differential G6PC mRNA expression and survival analysis
Protein-protein interaction network of 20 signatures in MPMs was constructed using Search Tool for the Retrieval of Interacting Genes (STRING; http://strin g-db.org, version 10.0) online database.
Differential expressed G6PC level was evaluated between ccRCC and normal samples from TCGA, CPTAC and RECA-EU cohort using Student's t test. Survival analysis of G6PC predicting prognosis ability was performed with GEPIA (http://gepia.cance r-pku.cn/detail.php###) in patients from TCGA cohort with cut-off value set as median. Kaplan-Meier method with 95% confidence intervals (95% CI) and log-rank test were used in survival analysis in CPTAC, RECA-EU and FUSCC. Best cut-off values were set using X-tile software. All patients at risk or patients numbers in different risk groups were shown in all survival plots.

| Glucose-6-phosphatase (G6Pase) expression in ccRCC and normal samples
G6Pase protein expression, coded by G6PC gene, was detected in ccRCC and normal samples from the human protein atlas (https:// www.prote inatl as.org/) and immunohistochemistry (IHC) data, including staining quantity, intensity, location and patients' data, were available online. Formalin-fixed, paraffin-embedded ccRCC tissues and human renal tissues were stained for anti-G6Pase using ab243319 (Abcam, USA) at 1/3000 dilution in FUSCC cohort and then independently evaluated by two experienced pathologists. The overall IHC score ranging from 0 to 12 was measured based on the multiply of the staining intensity and extent score, as previously described. 17 Low G6Pase expression group scores from 0 to 2, and high G6Pase group scores from 3 to 12.

| Immune cell infiltrations of G6PC in ccRCC
TIMER (Tumor IMmune Estimation Resource, https://cistr ome.shiny apps.io/timer /) is a web server for evaluating systematic various immune cells infiltration and clinical implications. In this study, correlation between infiltration of immune cells and G6PC copy number variation and expression levels were performed.

| Statistical analysis
All analyses were performed in the R (Version 3.6.0) and RStudio (Version 1.2.1335) and GraphPad Prism 7. Unless otherwise stated, results were considered statistically significant when P-value < 0.05. Twosided and p-values less than 0.05 were taken as significant in all tests.  Figure 1D) or CPTAC cohort ( Figure 1E). High-risk group was marked in red, and low-risk group was marked in blue.

| Differential G6Pase expression predicts outcomes in FUSCC cohort
A total of 322 ccRCC patients from FUSCC cohort were enrolled.
G6Pase was detected high expressed in normal kidney tissues (specifically in tubules cells rather than glomeruli cells), while not detected in ccRCC tissues from the Human Protein atlas ( Figure 6A). Meanwhile, significantly elevated G6Pase expression was found in normal tissues compared with ccRCC tissues from FUSCC cohort ( Figure 6B).
Clinicopathological characteristics in relation to G6Pase expression status were shown in 322 ccRCC patients from FUSCC cohort (Table S1)

| Most co-expressed genes and promoter methylation levels of G6PC in ccRCC
Top 50 co-expression genes with G6PC were extracted and shown in heat map in ccRCC ( Figure 7A

| Potential role of G6PC in ccRCC immune microenvironment
As the main regulator of glucose production, G6PC is highly expressed in liver and kidney tissues.  Figure 8C). Moreover, GSEA indicated that G6PC significantly involved in several signal pathways, including bile acid metabolism, fatty acid metabolism, epithelial mesenchymal transition and E2F targets in ccRCC ( Figure 8D-G). A total of 100 up-and down-regulated genes associated with differential G6PC expression were then visualized in ccRCC ( Figure S3).
Spearman's correlation and estimated statistical significance between G6PC expression and related genes and markers of immune cells were displayed in ccRCC patients using TIMER ( Table 1). F I G U R E 4 GSEA indicated significantly altered KEGG pathways based on differential risk score of MPMs in ccRCC patients with available transcriptomics data from TCGA and CPTAC cohorts. A, Top 5 significantly altered KEGG pathways in high-or low-risk ccRCC patients in TCGA cohort. B, Top 5 significantly altered significant KEGG pathways in high-or low-risk ccRCC patients of CPTAC cohort. C, Tumour environment purity was measured using ESTIMATE algorithm and showed a significant relationship with risk score of MPMs in ccRCC patients from TCGA cohort (r 2 = 0.2373, P < 0.0001). D, Immune purity in ccRCC environment significantly correlated with risk score of MPMs (r 2 = 0.3007,

| D ISCUSS I ON
The control of energy metabolism in human is a complex and cautious process, and metabolic disorders may lead to the occurrence and development of a variety of diseases. 18 For instance, abnormal lipid metabolism may reduce growth and impair fertility, while disorders in glucose metabolism can lead to diabetes and hypertension. 19,20 Importantly, the relationship between metabolic reprogramming and tumorigenesis has been paid more and more attention in recent years. Steven L. Gonias et al claimed that activation of lipid metabolism promotes tumour cell survival and tumour progression in pancreatic cancer. 21 Some studies found that abnormal glucose metabolism plays an key role in tumorigenesis. 22,23 Metabolic changes promote the proliferation of tumours microenvironment and also help us better understand the alterations of characteristics phenotypes and immune microenvironment of cancers. 24 For example, the activation of PI3K/AKT/mTOR and other carcinogenic pathways is related to the changes of bioenergy pathways such as glycolysis, fatty acid and glutamine metabolism, which provides a new target for tumour therapy. [25][26][27] ccRCC is one of the most common types of renal cell carcinoma in the world, and it is associated with poor prognosis because of its high metastasis and recurrence rate. 1,28 Metabolic reprogramming in ccRCC is most often associated with mutations in VHL, which occur in about 90% of cases. 29 In VHL mutant diseases, activation of metabolic pathways mediated by HIF leads to the activation of pathways contrary to the effects of hypoxia in normoxic environments. 30 Previous studies found that ccRCC produces energy mainly through the accumulation of lactic acid, 31,32 which is also called Warburg effect or aerobic glycolysis. HIF-1α, as the obvious driving force behind the Warburg effect in ccRCC, increases the expression of GLUT-1, thus promoting intracellular glucose uptake. 33,34 Interestingly, complex components in tumour microenvironment could exhibit metabolic stress on immune cells infiltrations, which can lead to immunosuppressive and tumour immune evasion. 24 The increased expression of GLUT-1 in ccRCC is associated with a decrease in the number of infiltrated CD8 + T cells, suggesting that glucose metabolism may suppress the immune system through another mechanism in renal cell carcinoma. 35 Normally, elevated glycolysis increases tumours immunity, immune check-point factors (PD-L1) expression levels on tumour cells, and thus imposed a favourable immunotherapy response in cancers. 36 Thus, the relationship between metabolic reprogramming and ccRCC microenvironment was worthy of further exploration.
G6PC (Glucose-6-Phosphatase Catalytic Subunit) is a protein coding gene, and it is closely associated with glycogen storage disease 37 and hypoglycaemia. 38   were extracted and shown in heat map in ccRCC. C, Promoter methylation levels of G6PC were significantly lower in primary ccRCC samples than normal samples (P < 0.0001). D, Promoter methylation levels of G6PC significantly climbed with elevated individual cancer stage and were the highest in samples with stage 4. E, Promoter methylation levels of G6PC significantly climbed with elevated individual tumour grade and were the highest in samples with grade 4. F, Promoter methylation levels of G6PC were significantly higher in ccRCC samples with nodal metastasis compared with pN0 patients (P < 0.05). ccRCC, clear cell renal cell carcinoma; pN0, pathological negative nodal metastasis status F I G U R E 8 Potential role of G6PC in pan-cancers and ccRCC microenvironment. A, As the main regulator of glucose production in the liver, high G6PC active expression is found in liver and kidney tissues. G6PC expression is significantly higher in normal tissue compared with renal cell carcinoma (KIRC, KIRP, KICH), while significantly lower in normal samples compared with hepatocellular carcinoma and cholangiocarcinoma. B, Copy number alteration of G6PC significantly correlated with environmental immune cells infiltration level. Elevated arm-level deletion of G6PC leads to inferior B cell, CD8 + cells, CD4 + cells, macrophage, neutrophil, dendritic cells infiltration compared with normal samples (P < 0.05). C, G6PC significantly participates in abnormal immune infiltration of ccRCC cells and microenvironment, showing significantly negative association with check-point immune signatures, dendritic cells, Th1 cells, MHC class I, cytolytic activity, inflammation promotion, HLA, APC co-inhibition and co-stimulation activities (cor.<−0.7). D-G, GSEA indicated that G6PC significantly involved in several signal pathways, including bile acid metabolism, fatty acid metabolism, epithelial mesenchymal transition and E2F targets in ccRCC. ccRCC, clear cell renal cell carcinoma; KIRC, kidney renal clear cell carcinoma; KIRP, kidney renal papillary cell carcinoma; KICH, kidney Chromophobe; GSEA, gene set enrichment analysis

| CON CLUS ION
In conclusion, this study first provided the opportunity to com- transcriptome data and identified G6PC as potential prognostic targets in 1040 ccRCC patients from multiply cohorts. These finding could assist in managing risk assessment and shed valuable insights into treatment strategies of ccRCC.

ACK N OWLED G EM ENTS
We thank the TCGA, CPTAC and ICGC database for survival data and gene expression profiles of ccRCC.

CO N FLI C T O F I NTE R E S T S
The authors declare no competing interests.

AUTH O R S' CO NTR I B UTI O N S
The work was performed in co-operation with all authors. YDW,

PATI E NT CO N S E NT FO R PU B LI C ATI O N
Not applicable.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data sets analysed in this study were obtained from online open-access databases or corresponding authors upon reasonable request.