Introduction
The incidence of kidney cancer rises annually, up to 2.2% of the overall tumor incidence and 1.8% of the overall tumor mortality in 2020 [
1]. Renal cell carcinoma (RCC) presents 90% of kidney cancers and the most common type is clear cell renal cell carcinoma (ccRCC), accounting for approximately 80% of all RCC patients [
2,
3]. ccRCC is highly aggressive and does not respond to traditional radiotherapy and chemotherapy, so the fatality of ccRCC increases year by year. Hence, we must deepen our understanding of ccRCC, explore mechanisms of ccRCC metastases, identify reliable biomarkers to accurately diagnose tumors in the early stage, and develop more efficient measures to treat ccRCC patients.
Recently, dysregulated metabolism of cancer has drawn the attention of researchers. ccRCC is considered a metabolic disease since the oncogenic mutations are engaged with comprehensive metabolic pathways, so-called metabolic reprogramming [
4]. Such metabolism reprogramming enables tumor cells to generate adequate cellular fundamental components, like DNA, membrane structures, and molecules that module tumor energetics. Otto Heinrich Warburg found that most cancer cells yielded energy predominantly through an inefficient process, aerobic glycolysis [
5]. Notably, this phenomenon is more conspicuous in ccRCC than in ordinary tissues [
6]. Abnormal lipid metabolism is also outstanding in ccRCC, as substantial lipids accumulate in ccRCC cells and the levels of cholesterol esters and cholesterol, as well as triglycerides within ccRCC cells, are remarkably higher than those within normal tissues [
7]. Part of the reason is that the mutations of
SCD1,
FASN, and
ACC lead to the abnormal generation of acetyl-CoA and fatty acid, which then gradually aggregate and form lipid droplets in the cytoplasm [
8]. As an amino acid, Tryptophan is heavily utilized in ccRCC. The increased utilization results in immunosuppression, which promotes tumor growth and lowers the sensitivity to interferon-α-based immunotherapies [
9]. Taken together, the reprogramming of metabolic pathways in ccRCC is essential for cellular components, energy production, and avoiding immune surveillance.
Metastases are the major barrier to promising clinical outcomes, as more than 90% of cancer-associated death is the result of metastasis diseases. Unfortunately, up to 30% of ccRCC patients were diagnosed with metastases [
10]. Hence, the restraining and curing of metastases are the basis of improving cancer outcomes. The success of tumor metastases needs tumor cells invading through, or collaborating with stroma, regulating the tissue microenvironment, escaping immune surveillance, and developing resistance to anti-cancer therapies [
11,
12]. So far, tumor metastases are too complex to be fully elucidated and the integration of various expertise across scales like oncology, genetics, bioinformatics, and mathematics, is required to understand the phenomenon systemically and holistically.
It has been reported that metabolic reprogramming can help cancer cells adjust to varying tumor microenvironments during the invasion-metastasis cascade, and sustain metabolic flexibility and plasticity [
13,
14]. Moreover, metabolism-associated genes (MAGs) exert indispensable functions in cancer development and progression [
15]. Therefore, here in the present study, we seek whether the dysregulated metabolism promotes ccRCC metastases, analyze the interaction between MAGs and the outcomes of ccRCC, and explore the underlying mechanisms. By integrally analyzing GSE105261 and the cancer genome atlas kidney renal clear cell carcinoma (TCGA -KIRC) cohorts with 2131 comprehensive MAGs from kyoto encyclopedia of genes and genomes (KEGG) database [
16] and molecular signatures database (MSigDb), the metabolism-associated prognostic signature (MAPS) was constructed and further validated in two independent cohorts, E-MTAB-1980 and GSE22541. The MAPS was positively correlated with clinicopathologic parameters, and could accurately and independently forecast the outcomes of ccRCC patients. Furthermore, functional enrichment analysis showed the MAPS was accompanied by pathways related to dysregulated metabolism, tumor metastases, and immune responses. Besides, it was demonstrated that the MAPS also showed a strong capacity to estimate the sensitivity of target therapies and immunotherapies in ccRCC.
In summary, we built a powerful metabolism-associated prognostic signature that was valuable in forecasting patient prognoses, guiding the therapeutic selection, and offering potential pathways for exploring mechanisms related to ccRCC metastases standing on the point of metabolism reprogramming.
Discussion
As we mentioned before, up to 30% of ccRCC patients were diagnosed with metastases [
10]. Metastases are the most lethal aspect of cancer since they are difficult to be cured through surgical resection, conventional chemotherapy, and radiation therapy, with the fact that above 90% of all tumor-associated deaths are caused by metastasis diseases [
12]. Also, for patients with ccRCC, due to distant metastases and local recurrence, the outcomes are not satisfactory all the time [
41]. Meanwhile, ccRCC is considered a metabolic disease [
4], since multiple metabolic pathways are remarkably reprogrammed in ccRCC [
42]. Increasing evidence indicated that MAGs exerted essential functions in tumor development and progression [
15]. According to "the Warburg effect", glucose metabolism in tumors switches to aerobic glycolysis resulting in the accumulation of lactic acid [
43], glutamine [
44], and the low-PH tumor microenvironment (TME) [
45], which are responsible for promoting the invasion and migration of tumor cells. Additionally, lipid metabolism is also conspicuous in tumors like ccRCC with accumulating evidence showing that the phenotype of ccRCC cells is similar to that of adipotytes [
46], and the lipid metabolic reprogramming is tightly related to the emergence of metastases. For example, it was found that peroxisome proliferator-activated receptor gamma coactivator 1-alpha (PGC1A), as a lipid metabolic regulator activated by melatonin, promoted tumor slimming, a lipid browning process reducing catabolic state, and inhibited ccRCC metastases through uncoupling protein 1(UCP1)-dependent manner [
47]. Amino acids, like proline, serine, asparagine, and glycine, also affect the emergency of tumor metastases [
48‐
51]. Taken together, we can infer that the metabolic reprogramming in tumors closely affects the occurrence of metastases, while the discussion about this research hotspot in ccRCC is rare. In this article, we create a novel signature, according to the comprehensive MAGs which are also relevant to ccRCC metastases, to help forecast the outcomes of ccRCC patients, assist in choosing treatment regimens, and most importantly provide ideas for unveiling the mechanisms underlying the phenomenon that metabolic reprogramming affects tumor metastases, under the help of high-throughput methods and bioinformatics methodology.
First of all, we downloaded 2131 comprehensive MAGs from the KEGG database [
16] and MSigDb. Then through WGCNA, univariate Cox regression, differential expression analysis, LASSO regression, and multivariate Cox regression based on GSE105261 and TCGA-KIRC cohorts, we constructed a 12-gene-metabolism-associated prognosis signature, termed the MAPS by our team. The genes from the MAPS were closely associated with the ccRCC metastasis trait and affected the outcomes of ccRCC patients, and all of them were differentially expressed in ccRCC tumors. Using the median risk score as a cut-off, patients were assigned to low-risk and high-risk subgroups. The high-risk subgroup displayed extraordinarily worse survival. The ROC analyses demonstrated that the MAPS performed satisfactory correctness in forecasting 1-,3-and 5-year OS and the occurrence of progress. Additionally, both multivariate Cox regression and stratified survival analyses regarding clinicopathological parameters implied that the MAPS acted as an independent biomarker. Moreover, E-MTAB-1980 and GSE22541 were utilized to verify the practicability and universality of the MAPS. Taken together, the MAPS was a convincing biomarker for distinguishing the severity and predicting OS in ccRCC patients.
As the genes in the MAPS were screened out via the most metastasis trait-associated module in WGCNA, we further delved into the relationship between the MAPS and metastasis trait. Firstly, we found that the expression levels of these genes from the MAPS alter transcriptionally between primary ccRCC and metastatic samples and may trigger or accelerate the ccRCC metastasies. Moreover, in high-risk patients, the ratio of stage M1 to stage M0 was greater than that in low-risk patients and patients with metastases owned higher risk scores. Besides, The ROC curve illustrated that the AUC for forecasting the emergence of remote metastases was 0.709 in the TCGA-KIRC cohort and 0.762 in the E-MTAB-1980 cohort, respectively. These results implied the close connection between the MAPS and ccRCC metastases and validated that the MAPS had a reliable ability to forecast the emergence of ccRCC metastases. Next, the results from functional enrichment analysis exhibited that the MAPS was closely related to the metastasis trait. On the one hand, the enriched metabolism pathways in GO and GSEA, like fatty acid, arachidonic acid metabolism, cholesterol metabolism, and oxidative phosphorylation metabolism, implied that differentially expressed genes among high- and low-risk subgroups of the MAPS mainly focused on metabolic processes. On the other hand, GO analysis indicated that the MAPS were closely accompanied by the alteration of the metastasis process. Furthermore, GSEA also demonstrated that angiogenesis, Adherens junction, and EMT were enriched in the high-risk subgroup. Targeting the angiogenesis pathway is an attractive approach for cancer therapy, and it is worth mentioning that among tumors, RCC is more sensitive to VEGF inhibitors [
52]. Thus, we analyzed the sensitivity of common chemotherapeutic drugs and targeted therapeutics using the pRRophetic algorithm. It was illustrated that the high-risk patients showed more sensitivity to rapamycin, sorafenib, sunitinib, and pazopanib compared with the low-risk patients, while no difference in the sensitivity to temsirolimus and axitinib was observed among high- and low-risk subgroups. Furthermore, we identified that the mutation rate of
SETD2 in patients with metastatic ccRCC is higher than that without metastatic ccRCC and meanwhile
SETD2 mutated more commonly in high-risk patients. The enrichment of common oncogenic signaling pathways also demonstrated the fraction of samples affected by the WNT pathway was higher in the high-risk patients than that in the low-risk patients. These results reinforced that metastasis was an important trait for the high-risk subgroup from the viewpoint of somatic mutation.
Subsequently, we discuss how dysregulated metabolism affects ccRCC metastases. GSEA of the MAPS demonstrated that several key pathways, like Hedgehog signaling (Hh), Hippo signaling, and the Renin-angiotensin system, were enriched. These three pathways are simultaneously implicated in dysregulated metabolism and tumor metastasis. Aberrant Hh, Hippo, and Renin-angiotensin signaling could lead to a series of diseases associated with abnormal lipid metabolism [
53,
54], while it can also advance tumor progression and metastases [
55,
56]. In the following research, we tested the differences in the TIME among the high- and low-risk subgroups and found that both immune-related cells and molecules were boosted in the high-risk subgroup. Besides, the high-risk subgroup comprised more immunosuppressive cells like Tregs and macrophages, while anti-tumor Th17 cells were significantly downregulated in the high-risk subgroup. The chemokines for the recruitment, differentiation, and activation of macrophages and Tregs in TIME [
11] were elevated in high-risk subgroups, which was consistent with the result that macrophages and Tregs were enriched in high-risk subgroups. Tumor-associated macrophages (TAMs) constitute the main tumor-infiltrating immune cells in TIME and can promote tumor cell invasion and metastases [
57]. Besides, TAMs are metabolically active [
58], and metabolic alterations, like glucose metabolism, lipid metabolism, and glutamine metabolism can determine the functions of TAMs in cancer progression [
59]. Interestingly, apart from directly aiming at tumor cells, suppressing Hh signaling could also reconfigure the TIME to be active [
55]. Tumor-derived sonic hedgehog (SHH), the hg signaling ligand, acts at TAM to drive M2 polarization which inhibits CD8
+ T cell recruitment to the TIME and promotes tumor progression [
60]. Tregs act as pro-tumor cells in TIME. A higher frequency of Treg cells reflects poorer outcomes in various types of cancer [
22]. Taken together, the high infiltration of macrophages and Tregs provides an explanation for the poor outcomes of the high-risk subgroup and brings us thoughts that whether the immune cells are involved in the metabolism-mediated metastases of ccRCC tumor cells, which needs further validation. Besides, checkpoint inhibitors (ICIs) have been effective strategies for ccRCC metastases [
61]. We identified that the high-risk subgroup owned higher immune response scores and benefited more from immunotherapy. Thus, the MAPS can serve as an indicator to assist doctors to make clinical decisions in choosing the proper drugs.
Numerous prognostic signatures have been brought up, some of which were tightly related to metabolic abnormalities. For example, Bian et al. have constructed a cuproptosis-related prognostic signature [
62]. Their signature was based on the ten known cuproptosis-related genes, while our MAPS was based on 2131 comprehensive MAGs. Notably, the cuproptosis-related gene signature showed poorer performance in predicting the outcomes of ccRCC patients compared with the MAPS [
62]. Furthermore, the role of cuproptosis in ccRCC has not been reported and whether it was associated with ccRCC metastases was unknown. M.Alchahin et al. constructed a specific metastatic signature associated with poor prognosis according to the single-cell RNA-seq (scRNA-seq) studies [
63]. In contrast, our MAPS was based on the TCGA-KIRC RNA-seq data and their corresponding clinical information. Although both M.Alchahin et al. and this study launched signatures from the perspective of tumor metastases, our MAPS focused on the most notable feature of ccRCC, metabolism reprogramming, and tried to explore the latent mechanism in which dysregulated metabolism controlled ccRCC metastases.
In general, this study offers a novel comprehension of the invasion and metastases of ccRCC from the view of metabolism reprogramming. A12-gene MAPS was successfully constructed with the capacity to independently and accurately forecast the outcomes of ccRCC patients. Evermore, the MAPS owned prominent biological functions and clinical value, such as predicting the emergence of metastases, recognizing ccRCC patients with poor outcomes, and assisting clinical decisions.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.