Identification of metabolic dysregulation and biomarkers for clear cell renal cell carcinoma

Dear Editor,

Clear cell renal cell carcinoma (ccRCC) is the most common RCC histology and a metabolic disease characterized by reprogramming of energetic metabolism, enabling cancer cells to proliferate rapidly and survive.¹ Understanding these metabolic changes may provide opportunities for discovering biomarkers, improving tumour detection and developing new therapeutic strategies.

This study aimed to identify and validate metabolites as biomarkers to improve clinical diagnosis, facilitate risk stratification and predict prognosis in ccRCC. We used targeted metabolomics to screen for metabolic biomarkers and compared the differential metabolites between our data and other six published studies^2-6 (Table S1). Detailed information on metabolite measurement and methods is provided in the Supporting Information.

In our cohort, 90 tissues (30 from early patients, 30 from advanced patients and 30 from normal tissues) were collected from 60 ccRCC patients (Table S2). Principal component analysis (PCA) and partial least squares discriminant analysis (PLS-DA) analyses revealed a clear separation between the two groups (Figure S1A,B). The trend of quality control samples and permutated R2, and Q2 values indicated good analytical reproducibility and low risk of overfitting (Figure S1D,E). The differential analysis demonstrated that 163 metabolites (67 metabolites up-regulated and 96 metabolites down-regulated) from 23 distinct classes were significantly altered in tumour tissues (Figure 1A–F and Table S3). Then, we divided patients into early and advanced cohorts to perform differential analyses. The PCA and PLS-DA analyses showed distinct differences (Figure S1B,C). 79 metabolites were significantly different between early tumour and paired normal tissues (39 metabolites up-regulated and 40 metabolites down-regulated) (Figure 1G and Table S3), and 64 metabolites significantly distinguished advanced tumour from paired normal tissues (25 metabolites up-regulated and 39 metabolites down-regulated) (Figure 1H and Table S3). The pathway enrichment analysis of differential metabolites revealed eight intersected KEGG pathways. Two metabolic processes were substantially enriched in the early-stage cohort, while three processes were enriched in the advanced-stage cohort (Figure 1I,J and Table S4).

To understand the differences in metabolites between early and advanced stages, we used the unsupervised clustering analysis. Among early patients, two early clusters (A and B) were identified, and patients in early cluster A had the worst prognosis. Similarly, in the advanced cohort, we identified three advanced clusters (A, B and C) and found that patients in the advanced cluster B had the worst prognosis and patients in the advanced cluster C had a better prognosis (Figure 2A–D and Figure S1F–L). Then, we found elevated tryptophan and creatine in early-cluster B. Similarly, in advanced-cluster C, high creatine was found (Figure 2E,F). Considering valuable biomarkers are rarely utilized in ccRCC patients, we screened metabolites using random forest analysis. Based on mean decrease accuracy, we selected the top 10 metabolites in the total, early, and advanced cohorts (Figure 2G–I). A receiver operating characteristic (ROC) showed the top 10 metabolites had high predictive accuracy (area under the curve [AUC] > 0.9) in distinguishing patients (Figure 2J–L and Table S5). Then, we combined all 10 metabolites and found the combined model also had high predictive accuracy (AUC > 0.9) (Figure 2M–O). These results showed that these 10 metabolites could be valuable parameters in the prediction of patients.

To further validate the predictive capability, we compared external validation datasets from six independent studies with our data and took the intersection. After excluding two studies without stage information, four validation studies were utilized.^2-5 All intersected metabolites had better predictive performance in validation datasets (Figure 2P–W). Although only N-Phenylacetylglutamine was found in an advanced cohort from test cluster 2, it exhibited good predictive ability (Figure S1J).

Next, we evaluated the role of metabolites in predicting prognosis using external data. We set our data as the training cohort to construct the risk model, and other studies with full clinical information were used as validation cohorts. We first divided our data into total RCC cohort (n = 60), early (n = 30) and advanced (n = 30) RCC cohort. Using lasso analysis, we constructed risk models and found the high-risk group presented worse outcomes in the total and early RCC cohort (Figure 3A–D). However, there was no significant difference between the two groups in the advanced RCC cohort (Figure 3E). Additionally, the ROC curve demonstrated a strong capacity of the risk model in predicting OS (AUC = 1) in total and early RCC cohorts (Figure 3F, G). These results suggest that the risk model may effectively evaluate the survival of RCC patients. Then, we validated the risk model using the Hakimi et al. dataset. The K-M curve indicated that the high-risk groups had worse prognosis in our total and early RCC cohort (Figure 3H,J). However, due to the low sample number, we found although there is a trend between high and low-risk groups, the difference did not reach statistical significance (Figure 3I,K). Therefore, further validation is still necessary.

Finally, we compared all differential metabolites between our data and others, to determine whether metabolite alterations were consistent across populations. After comparing metabolomics data in Hakimi et al.,² Hu et al.,⁴ Popławski et al.⁷ and Li et al.³ studies, we found the content of most metabolites was consistent with our results. However, creatine, citrulline and inosine content were apparently different (Figure 4A–C and Figure S1M–Y). We further mined their effect on ccRCC. We found creatine was abnormally low in tumour samples (Figure 4D). Then, we explored IC₅₀ of creatine and used CCK8 and plate cloning experiments to determine its effect on proliferation, which indicated that creatine (5.28 mM) markedly inhibited RCC cell proliferation (Figure 4E–H). Wound healing assay, Transwell migration and invasion assay demonstrated that creatine significantly inhibited ccRCC cell migration and invasion (Figure 4I–K). Referring to previous research, citrulline and inosine were supplemented at 16.67 mM⁸ and 0.19 mg/ml,⁹ separately. Citrulline suppressed ccRCC cell proliferation, migration and invasion, while inosine promoted RCC cell progression (Figure 4L-Q). We speculated variations in metabolite concentration may be partially attributed to differences in study samples and contexts.

Accurate prognostication of oncological outcomes is crucial for ccRCC. Although many gene-based models exist, metabolite-based signatures are rare.¹⁰ We identified key metabolites and constructed a metabolites-based risk model, which provides predictive and prognostic information independent of clinicopathologic factors. We also evaluated three metabolites, offering insights into the molecular mechanisms and novel therapeutic targets. In short, our study discovered novel biomarkers for diagnosing ccRCC and predicting prognosis at different stages.

Bin Zheng., Kan Liu., Qing Ouyang. and Xiubin Li. designed research. Bin Zheng., Kan Liu. and, Qing Ouyang. performed research. ShengPan Wu., Li Wang., Tongyu Jia., Shouqing Cao. and Ji Feng contributed reagents and resources. Bin Zheng. and Xiubin Li. wrote the paper in discussion with other co-authors. Xin Ma. and Xu Zhang. conceived and directed the study.

The authors declare no conflict of interest.

This work was supported by the National Natural Science Foundation of China (81802804, 82403839), Sino-German Mobility program (M0735), PLA General Hospital Youth Independent Innovation Science Fund Growth Project (22QNCZ029), PLA General Hospital-Third Medical Center Discipline Innovation and Development Special Fund Project (2024BJ-04) and Fostering Fund of Chinese PLA General Hospital for National Excellent Young Scholar Science Fund (2020-YQPY-006).

All studies were approved by the Ethics Committee of the third medical centre of PLA General Hospital.