Prognostic and predictive role of a metabolic rate‐limiting enzyme signature in hepatocellular carcinoma

Abstract Objectives Abnormal expression of metabolic rate‐limiting enzymes drives the occurrence and progression of hepatocellular carcinoma (HCC). This study aimed to elucidate the comprehensive model of metabolic rate‐limiting enzymes associated with the prognosis of HCC. Materials and Methods HCC animal model and TCGA project were used to screen out differentially expressed metabolic rate‐limiting enzyme. Cox regression, least absolute shrinkage and selection operation (LASSO) and experimentally verification were performed to identify metabolic rate‐limiting enzyme signature. The area under the receiver operating characteristic curve (AUC) and prognostic nomogram were used to assess the efficacy of the signature in the three HCC cohorts (TCGA training cohort, internal cohort and an independent validation cohort). Results A classifier based on three rate‐limiting enzymes (RRM1, UCK2 and G6PD) was conducted and serves as independent prognostic factor. This effect was further confirmed in an independent cohort, which indicated that the AUC at year 5 was 0.715 (95% CI: 0.653‐0.777) for clinical risk score, whereas it was significantly increased to 0.852 (95% CI: 0.798‐0.906) when combination of the clinical with signature risk score. Moreover, a comprehensive nomogram including the signature and clinicopathological aspects resulted in significantly predict the individual outcomes. Conclusions Our results highlighted the prognostic value of rate‐limiting enzymes in HCC, which may be useful for accurate risk assessment in guiding clinical management and treatment decisions.


| INTRODUC TI ON
Hepatocellular carcinoma (HCC) is a highly aggressive solid malignancy accounting for 90% primary liver cancer and is the fourth leading cause of cancer death worldwide. 1 Unfortunately, HCC is always diagnosed at advanced stage with extreme hepatic dysfunction, resulting in poor prognosis of HCC patients. 2 Moreover, current clinical tools to predict prognosis are limited to a set of clinical and pathologic variables, such as the TNM stage, which mainly relies on anatomical information without biological characteristics.
Therefore, there is an urgent need to further elucidate the molecular mechanisms of HCC and identify specific prognostic and predictive biomarkers, which would be of great importance to improve the prognosis of HCC patients.
Metabolic reprogramming has been reported to be involved in tumorigenesis and development. 3,4 Metabolic dysregulation is associated with the various progressions of different cancers, including tumour growth, metastasis, angiogenesis and drug resistance. 5,6 Recently, several metabolic rate-limiting enzymes in glycolysis, gluconeogenesis, pentose phosphate pathway (PPP), and fatty acid oxidation (FAO) and tricarboxylic acid (TCA) cycles 7,8 have been identified as biomarkers and drug targets, such as phosphofructokinase platelet (PFKP), 9 fructose-1,6-bisphosphatase 1 (FBP1) 10 and UDP-glucose 6-dehydrogenase (UGDH). 11 Therefore, identification of the enzymes from metabolic rate-limiting enzyme database systematically may provide more prognostic biomarkers and therapeutic targets in HCC.
Several studies have kept tabs on the systematic analysis of the role of metabolism in glioblastoma, gastric cancer and endometrial cancer. 12,13 However, the comprehensive analysis of the interrelation between rate-limiting enzymes and clinical prognosis of HCC patients has not been reported yet. In this study, we found that metabolic pathways were more enriched in diethylnitrosamine (DEN) and CCL 4 induced HCC animal model. Meanwhile, by conjointly analysing the RNA sequencing (RNA-seq) data of animal model and TCGA data of HCC, we found a series of abnormal metabolic rate-limiting enzymes involved in HCC progression. Then, we established a signature based on RRM1, UCK2 and G6PD, which was also validated by an independent HCC cohort. Taken together, our study demonstrated abnormal metabolic pathways were involved in HCC development and provides a novel signature based on metabolic rate-limiting enzyme for predicting the clinical outcome of HCC patients.

| Mouse models for HCC
The C57BL/6 background mice were purchased from Nanjing Biomedical Research Institute of Nanjing University (Nanjing, Jiangsu, China) and maintained in SPF facilities. At 14 days, the mice were injected with the carcinogen DEN (25 mg/kg, Sigma-Aldrich), and flowing by intraperitoneal injections of 10% CCl 4 (5 mg/kg, once a week, Sigma-Aldrich) at age of 4 weeks. The ultrasonic inspection was performed at indicated time. The mice were sacrificed at 5 months finally. Part of the tumour tissues and normal tissues was fixed in 4%paraformaldehyde for H&E analysis, and the remaining tissues were placed in liquid nitrogen for RNA-seq, qRT-PCR and Western blotting.

| RNA-seq analysis
Total RNA was extracted from mouse liver tumour tissues (n = 3) and paired normal tissues (n = 3). The quality and quantity of the RNA were assessed by a NanoDrop TM ND-1000. Denaturing agarose gel electrophoresis was used to assess RNA integrity.
The mRNA extraction was performed using a NEB Next RPoly(A) mRNA Magnetic Isolation Module. RNA libraries were constructed using a KAPA Stranded RNA-Seq Library Prep Kit (Illumina).
The RNA sequencing service was provided by ShuPu (Shanghai, China) BIOTECHNOLOGY LLC. R packages were used to screen the differentially expressed genes between tumour and normal tissues, Gene Ontology (GO) and KEGG pathway analyses were performed to explore the biological functions using 'clusterprofiler' R package, and enrichment analysis was presented by R packages 'ggplot2' and 'GO plot'.

| Data collection
111 human rate-limiting metabolic enzymes were selected from the rate-limiting enzymes database according to previous study, 14 and the complete list of these genes encoding the enzymes is included in Table S1. score. Moreover, a comprehensive nomogram including the signature and clinicopathological aspects resulted in significantly predict the individual outcomes.

Conclusions:
Our results highlighted the prognostic value of rate-limiting enzymes in HCC, which may be useful for accurate risk assessment in guiding clinical management and treatment decisions.
All the human RNA expression data and corresponding clinical information of 374 HCC patients were downloaded from TCGA (PanCancer Atlas)(https://tcga-data.nci.nih.gov/tcga/), and differentially expressed metabolic rate-limiting enzymes were screened by R package 'limma' with a cut-off criterion of |logFC| > 1 and P < .05.
The clinical characteristics of enrolled patients from the three HCC cohorts were included in Table S2. The expression level of prognostic associated rate-limiting metabolic enzymes between cancerous and normal samples was displayed via package 'pheatmap' and 'ggplot' respectively.

| Human specimens
All HCC patients in our independent cohort were included from the Nanjing Drum Tower Hospital, the Affiliated Hospital of Nanjing

| RNA extraction and qRT-PCR analysis
Total RNAs were extracted from human and mouse HCC tissues and associated non-cancerous tissues using TRIzol reagent

| Western blotting
The Western blot protocol was performed as previously described. 15 The antibodies used were as follows: anti-RRM1 (Protein tech,

| Immunohistochemistry
The immunohistochemistry staining of formalin-fixed, paraffinembedded tissue sections was performed following the manufacturer's instructions. 16 Briefly, formalin-fixed, paraffin-embedded tissue sections were deparaffinized with xylene and decreasing concentrations of ethanol. After antigen retrieval and protein blocking steps, tissue sections were incubated with primary antibodies

| Clonogenic survival assay
Hep3B cells were transfected with indicated siRNAs of RRM1, UCK2 and G6PD for 48h; then, the cells were trypsinized and cultured in 12-well plates with 500 cells/well for 2 weeks respectively. For scoring colonies, the cells were fixed in 1ml methanol for 15 min and stained with Giemsa for 10 min. and the number of colonies was quantified, and each colony containing cells >50 were counted.

| Statistical analysis
Statistical analysis was performed with the R software (ver-sion3.6.1). We used chi-square test for categorical variables and chi-square or Fisher's exact test for contingency tables. Univariate Cox regression analysis was performed to identify prognostic associ- to perform the time-dependent ROC curve analysis, the 'rms' package to consolidate the risk score and clinical characteristics for nomogram construction. All statistical tests were two-sided, and P < .05 was considered to be significant.

| Abnormal expression of metabolic rate-limiting enzymes in HCC
The detailed flow chart of the study design is displayed in Figure 1A.
To investigate the functional pathways in hepatocellular carcinogenesis, we first constructed DEN and CCL 4 induced HCC mice model.

| Identification of clinical prognosis of the metabolic rate-limiting enzymes
Intriguingly, by conjointly analysing the differentially expressed  Figure 3E and Figure S4A). Simultaneously, univariate Cox regression analysis demonstrated that the protein levels of RRM1, UCK2 and G6PD were significantly correlated with OS in HCC patients ( Figure S4B). Furthermore, multivariate Cox regression analysis also revealed that RRM1, UCK2 and G6PD could act as an independent predictive marker for the prognosis of HCC ( Figure S4C). Meanwhile, Kaplan-Meier survival analyses showed that HCC patients with high levels of RRM1, UCK2 and G6PD had worse OS (P = 7.36e-05; P = 1.687e-02; P = 1.492e-03 respectively) ( Figure 3F-H). Taken together, these results suggest that the expression of RRM1, UCK2 and G6PD are up-regulated in HCC and all of them could act as an independent prognostic factor for HCC.

| Construction of a signature based on the metabolic rate-limiting enzymes
Based on the expression levels of RRM1, UCK2 and G6PD from entire TCGA data cohort, the following formula was derived to calculate prediction model risk score for each patient: Risk score = (0.006 × expression value of RRM1) + (0.075 × expression value of UCK2) + (0.005 × expression value of G6PD). Then, we randomly divided the 235 HCC patients from TCGA into the training cohort (n = 116) and the internal cohort (n = 119). With the risk score formula, patients in the training cohort were stratified into high-risk (n = 54) and low-risk (n = 62) subgroups according to mean risk score ( Figure 4A), and the expression levels of RRM1, UCK2 and G6PD were higher in the high-risk group ( Figure 4D). Furthermore, it showed that the patients in high-risk group had shorter OS than those in the low-risk group ( Figure 4G, J).
To further validate the efficiency of this signature, the internal cohort (n = 119) and the total HCC patients from TCGA data (n = 235) were used to assess ( Figure 4B and Figure S5A), as shown in Figure 4E and Figure S5B, the expression of these three genes was higher in the high-risk subgroup. Survival analysis of the two cohorts also showed that HCC patients in the high-risk group had worse prognosis than those in the low-risk group ( Figure 4H, K and Figure S5C, D), which further confirmed the results in the training cohort.
We then further assess the predictive ability of the signature in external independent validation cohort ( Figure 4C). Intriguingly, it showed that HCC patients with high-risk scores generally had higher protein levels of RRM1, UCK2 and G6PD than those with low scores ( Figure 4F), and HCC patients in the low-risk group had a better prognosis than those in the high-risk group ( Figure 4I, L).  in HCC patients ( Figure S6A,B), while multivariate Cox regression analysis showed only the signature had an independent prognostic ability ( Figure 5A,B). In the external independent validation cohort, univariate Cox regression analysis also showed that TNM stage, differentiation grade, tumour burden, lymph node metastasis and signature were substantially associated with survival in HCC patients ( Figure S6C), while multivariate Cox regression analysis showed that the signature and lymph node metastasis were indicated as independent predictive marker for the prognosis of HCC patients ( Figure 5C). Next, we compared our classifier with the existing clinicopathological features, it showed that, in the training cohort, TNM stage and tumour burden were significantly associated with risk score (Figure 5D), which were confirmed in the internal cohort ( Figure 5E). Consistently, in the independent validation cohort, more patients with advanced TNM stage, serous tumour burden and high grade were seen in high-risk group compared with low-risk group ( Figure 5F).

| The prognostic value of the signature in three cohorts of HCC
To further assess the accumulative effects of the metabolic rate-

| Development of a prognostic nomogram to predict the individual outcomes of HCC patients
Based on the signature and clinicopathological characteristics, we built a comprehensive prognostic nomogram to estimate overall survival probability for 5 years in HCC patients using the independent cohort. Two independent prognostic parameters (metabolic rate-limiting enzymes signature and lymph node metastasis) were integrated into the nomogram, which showed that higher total score was associated with shorter OS of HCC patients ( Figure S6D).
Furthermore, we found the predicted value was more consistent with the actual value, which were confirmed by the calibration curve of the nomogram for the survival probability at 3 or 5 years ( Figure S6E,F).

| RRM1, UCK2 and G6PD promote HCC proliferation in vitro
To further elucidate the function of these three rate-limiting enzymes in HCC cells, we first investigated RRM1, UCK2 and G6PD expression levels in HCC cell lines and normal hepatocytes by Western blot analysis. Our data suggested that RRM1 and UCK2 were preferentially expressed in most HCC cell lines, except for Huh7 cell, whereas high expression of G6PD was observed in Hep3B and Huh7 cells ( Figure 7A). Subsequently, we transfected Hep3B cell with two specific siRNAs of RRM1, UCK2 and G6PD, and the protein level of RRM1, UCK2 and G6PD was markedly reduced respectively ( Figure 7B,D,F). We also performed colony formation studies to elucidate the functional roles of the three in vitro, and the results showed that knockdown of RRM1, UCK2 and G6PD significantly suppressed the clonogenic ability of HCC cells (Figure 7C,E,G). Taken together, these data suggested that RRM1, UCK2 and G6PD may act as oncogene that promotes HCC proliferation.
F I G U R E 5 Multivariate Cox analysis in three HCC cohorts. (A-C) Multivariate Cox analysis was performed in the TCGA testing cohort, TCGA internal cohort and independent cohort. (D-F) Heat maps of the expression of RRM1, UCK2 and G6PD with corresponding clinicopathological characteristics in the TCGA testing cohort, TCGA internal cohort and independent cohort. **P < .01; ***P < .001 showed that high-risk score group was related to poorer prognosis in three independent cohorts. In addition, metabolic rate-limiting enzyme signature and clinical characteristics were identified as independent overall survival-related variables, which were incorporated into the nomogram, indicating that the nomogram can be regarded as a valid tool for clinical diagnosis and treatment of HCC patients. Furthermore, we utilized univariate and multivariate Cox regression analyses to compare our signature with the published signatures, and the results indicated that our signature similar with that of a published 10-gene metabolic signature (Weng_ signature) and had a relatively better predictive output than the 4-gene metabolic signature. These findings suggested that this signature had powerful capacity to predict the prognosis of HCC patients, which may be helpful to guide significance for the decision-making of clinical treatment.

| D ISCUSS I ON
In summary, for the first time, we identified and validated a novel signature based on three-metabolic rate-limiting enzymes, which was associated with the prognosis of HCC patients. Our study not only demonstrates abnormal metabolic pathways contribute to HCC progression, but more importantly provides a new metabolic ratelimiting enzyme prognostic signature for HCC, which was useful for accurate risk assessment in guiding clinical management and treatment decisions.