Cancer is characterized by unregulated cell development, which requires high energy levels to live in adverse environments such as low oxygen, dietary deficit, and chemotherapeutic drug intoxication43–45. Changing in metabolic pathways that control tumor energetics and biosynthesis, so-called metabolic reprogramming, is a prominent feature of ccRCC46. Inactivation of Von Hippel-Lindau (VHL) occurs in most sporadic ccRCC46,47. VHL increases proteasomal degradation of the transcription factor hypoxia-inducible factor (HIF) family, and its mutations activate pathways that reverse the effects of hypoxia in the normoxic environment, resulting in a condition of “pseudo-hypoxia”46–48. For example, glycolysis and lipid metabolism significantly influence ccRCC support, progression, and treatment choices through various biological processes9,11,46,49. The biosynthesis of glutathione (GSH) is equally important in ccRCC cells, as it is used to detoxify reactive oxygen species50. Consequently, as a center for several cellular signaling pathways, the adaptable and intricate metabolic network of ccRCC may offer novel targets for patient PET imaging, guided treatment, and enhanced prognosis51,52.
Previous bulk RNA-seq based transcriptome analyses of ccRCC multi-patient samples focused mostly on prognostic gene screening and clinical outcome prediction, with less consideration given to intercellular heterogeneity (STable 1). Meanwhile, scRNA-seq of ccRCC focuses on the research of cell origin53, identification and function of new cell groups54,55 and the heterogeneity of tumor cells24. However, it is challenging to correlate scRNA-seq analysis with prognosis because of the tiny sample size. Saout investigated the connection between certain ccRCC malignant cell subsets and patient prognosis via deconvolution approaches25. We employed a combination of scRNA-seq analysis and multi patient sample sequencing to study the association between particular cell subsets and prognosis in ccRCC and categorized ccRCC patients based on unique cell metabolism reprogramming, further choosing therapeutic targets.
To learn more about the metabolic heterogeneity of malignant cells in ccRCC samples, we initially performed scRNA-seq analysis. By combining functional analysis with the AUCell method, we discovered a rise in the abundance of a unique malignant cell subset dubbed META_active cells, which demonstrate greater metabolic activity than other cells in the microenvironment. Through the integration and analysis of transcriptome profiles of ccRCC samples from bulk RNA-seq and scRNA-seq, as well as patient-specific variations in cell composition and clinical outcomes from large cohorts, we have compiled the features of metabolism into a categorization system. According to the percentage of META_active malignant cells, samples are classified with high and low subgroups. The OS rate was worse in the high META_active group, suggesting that metabolic heterogeneity might play a significant role in determining the prognosis of ccRCC. As opposed to META_ inactive cells, META_ active cells communicate with other cells in the microenvironment more often and exhibit specificity in the PTN56 and THBS57 signaling pathways, which might be connected to the carcinogenesis and treatment resistance of ccRCC. Afterward, we integrated MRGs with DEGs in tumor samples to create a predictive model for assessing patients with ccRCC to find biomarkers that might predict variations in clinical outcomes. In the age of precision medicine, assessing the prognosis of patients with ccRCC and figuring out when to provide drugs is no longer possible by depending just on the Fuhrman grade or the AJCC TNM stage classification. Existing ccRCC prediction models frequently relied on separate cohort selection or did not include validation for more cohorts, resulting in poor model performance or overfitting. We gathered 10 popular machine learning methods and created 118 models to overcome these constraints. MRPS, the best model for ccRCC prediction, may be regarded as the combination of RSF and GBM after thorough analysis.
The 1, 3, and 5-year AUC values varied from 0.613 to 0.874 in the ROC curves, indicating that MRPS had a strong predictive effect on prognosis. It was also demonstrated that MRPS was an independent risk factor for the prognosis of ccRCC. The connection between MRPS and clinical characteristics was also examined. It appears that MRPS was not restricted to the predictive usefulness of OS, since the findings demonstrated a substantial correlation between risk scores and patients' distant metastases, tumor stage, and case grade. When it came to predicting the overall survival of patients with colorectal cancer, MRPS outperformed 51 published signatures and showed outstanding predictive ability across several datasets. Thus, MRPS could be useful in determining the prognosis for ccRCC and categorizing patients into high or low-risk invasive subgroups.
The high-risk subgroup's increased cell abundance of invasive Treg and TAM resulted in a more inhibitory immunophenotype. Multiple algorithms were used in this study to infer immune cell infiltration in ccRCC, and all of them concluded that Treg58 and TAM59 were upregulated in high-risk subgroups and that they were shaped by cell indication factors such as CD80/CD86 and chemokines IL-4, IL-10, and IL-35, thereby impeding anti-tumor immunity. Therefore, we examined the predictive value of MRPS in immunotherapy in this study. Low-risk samples’ OS was much longer than that of high-risk samples’ (P = 0.021). These findings imply that MRPS could offer a new perspective on ICI responses against ccRCC.
Potential therapeutic drugs for ccRCC patients were identified via established MRPS. Among them, clinical trials have been conducted on cancer patients to confirm the effects of sepantronium bromide60 and niclosamide61. Because of SGX-523’s unanticipated renal toxicity62, clinical development would no longer be conducted. Other medications' effectiveness and confounding have not yet been tested in clinical trials in cancer patients, and as technology advances, whether they can become the next generation of anti-tumor therapies in ccRCC patients must also be determined.
Finally, we looked at the possible physiological role that GGT6 members of the GGT family may have in the development and incidence of ccRCC. GGT6 knockdown enhanced the ability of ccRCC cells to proliferate and invade, indicating the viability of GGT6 as a ccRCC therapeutic target. We intend to further investigate its molecular mechanism in future studies.
Nevertheless, our study has several limitations. First, our findings are based on retrospective sequencing data and bioinformatics analysis; multicenter and prospective clinical studies are required to evaluate the robustness of MRPS. Second, more comprehensive experimental research is needed to understand the precise biological role of the genes in MRPS.
In conclusion, we created a stable and reliable prognostic signature of metabolic reprogramming including 17 genes based on a significant number of bioinformatics and machine learning programs. The model's predictive power for clinical characteristics, TIM and immunotherapy was also assessed. This MRPS model is a potential tool to guide clinical judgment and personalized therapy of ccRCC patients.