Identification of three genes lactic acid metabolism and transporter related as prognostic models of ccRCC patient.
Given that lactic acid metabolism is disordered in the TME, we first compared whether there are significant differences in the expression of lactic acid metabolism and transporter related genes in clear renal cell carcinoma tissues and normal adjacent kidney tissues. Among all the twenty-one lactic acid metabolism and transporter related genes, twenty genes have significant expression differences. Compared with normal tissues, nine genes are up-regulated and eleven genes are down-regulated in cancer tissues (Figure 1A), which is consistent with lactic acid metabolism disorder[12]. In order to make the prognostic model more robust, we used a small sample of E-MTAB-1980 cohort as a training set to construct a prognostic model, and a total of three genes (PNKD, SLC16A8, SLC5A8) were screened to construct the model (Figure 1B), their respective weights can be found in Table 2. Subsequently, we used the aforementioned formula to calculate the risk score of each patient in the E-MTAB-1980 cohort, and divided them into a high-risk group and a low-risk group (high score means high risk) based on the median score for survival analysis. There is a significant difference in OS (hazard ratio = 4.117, 95% CI: 1.810 - 9.362, p < 0.0001) between the two groups of patients, and patients in the high-risk group have a lower survival rate (Figure 1C). Homoplastically, in the TCGA cohort, it can be observed that the OS (hazard ratio = 1.909, 95% CI: 1.414 - 2.579 p < 0.0001) of ccRCC patients has similar result (Figure 1D). Finally, we integrated the selected genes and the clinical information of the patients into the heamap for analysis, and the correlation between the risk score and gene expression and clinical pathological information can be clearly observed (Figure 1E).
Table 2: Selected genes and associated weights in the prognosis model.
Genes
|
Weights
|
PNKD
|
0.43603432
|
SLC16A8
|
-0.05299190
|
SLC5A8
|
-0.07168776
|
Risk score is strongly correlated with the patient’s clinicopathological information.
We want to know whether there is a difference in risk scores in different subgroups of clinicopathology information. In the TCGA cohort, we found that the risk score gradually increased as the Grade increased (Figure 2A). Similarly, the patient's risk score will also gradually increase with the increase of Stage (Figure 2B). The T stage of renal cell carcinoma mainly represents the size of the tumor, and the risk score also increases as the tumor increases (Figure 2C). In addition, the most important thing is that when the risk score increases, the risk of tumor metastasis also gradually increases (Figure 2D, E). Taken together, the higher the lactic acid metabolism risk score, the more likely the tumor is to develop into a serious clinical pathological state. It can be explained that our prognostic model is strongly related to clinicopathological information.
The constructed prognostic model can be used as a prognostic factor.
Considering that the risk score can significantly distinguish the survival rate and clinicopathological information of ccRCC patients, we want to know whether the risk score can be used as a prognostic factor for ccRCC patients. Therefore, we performed univariate Cox regression analysis in two independent cohorts, and the hazard ratio of the risk score in the E-MTAB-1980 cohort was 6.406, p <0.0001. Simultaneously, we also got analogous conclusion in the TCGA cohort (Table.3). In summary, we can conclude that the risk score of lactic acid metabolism and transport can be used as a prognostic factor for ccRCC patients.
Table 3: Univariate Cox regression analysis in the TCGA cohort and E-MTAB-1980 cohort.
Cohort
|
TCGA cohort
|
E-MTAB-1980 cohort
|
HR (95% CI)
|
p-Value
|
HR (95% CI)
|
p-Value
|
Age
|
1.030(1.016-1.043)
|
<0.001
|
1.044(1.002-1.087)
|
0.040
|
Gender
|
0.952(0.693-1.306)
|
0.759
|
2.265(0.673-7.626)
|
0.187
|
Grade
|
2.333(1.896-2.871)
|
<0.001
|
2.982(1.671-5.320)
|
<0.001
|
Stage
|
1.907(1.666-2.182)
|
<0.001
|
2.289(1.654-3.167)
|
<0.001
|
Risk score
|
3.323(2.072-5.327)
|
<0.001
|
6.406(2.571-15.960)
|
<0.001
|
Risk score can accurately predict the survival time of ccRCC patients.
In order to explore whether our prognostic model can predict the survival time of patients, we performed ROC curve analysis on patients in two independent cohorts. In the E-MTAB-1980 cohort, patients' 1-, 3- and 5-year AUC reached 0.71, 0.72, and 0.77, respectively (Figure 3A). Simultaneously, In the TCGA cohort, the 1-, 3- and 5-year AUC of the patients were 0.65, 0.65, and 0.63, respectively (Figure 3B). Therefore, the prognostic model we established can be used as a reference factor for predicting survival of ccRCC patients.
More Treg cells in the tumor microenvironment of patients with high score.
Lactic acid metabolism can affect the distribution and phenotype of immune cells[13]. In order to explore whether there are significant statistical differences in the proportion of immune cells in patients with different risk groups, we used the CIBERSORT algorithm for analysis. In the TME of ccRCC patients, we found that the top three immune cells with the largest proportion are M2 type macrophages, CD4+ T memory cells, and CD8+ T cells (Figure 4A). So as to show the proportion of immune cells in each patient more intuitively, we visualized the proportion of immune cells in each patient (Figure 4B). To further study the relationship between our risk model and immune cell infiltration, we analyzed the differences in immune infiltration of ccRCC patients between high-risk and low-risk groups (Figure 4C). We found that the tumors of patients in the high-risk group had more Treg cells and macrophages, and fewer dendritic cells and CD8+ cells than those in the low-risk group. This may be related to the worse the prognosis of ccRCC patients with higher CD8+ T cell infiltration[14]. In addition, it may also be related to the ability of lactic acid to supply Treg cell metabolism and proliferation[15]. In short, for the first time, we established a prognostic model based on lactic acid metabolism and transporter related genes to analyze the infiltration of immune cells in the TME.
WGCNA reveals that high risk score is related to cell cycle.
Finally, in order to explore the potential mechanism of our model for predicting prognosis, we performed a weighted gene co-expression network analysis in the TCGA cohort based on the previously identified differentially expressed genes. We analyzed 2936 differentially expressed genes in the TCGA cohort through WGCNA. According to the suggestion of pickSoftThreshold, the soft threshold power of the β value is set to 14 (Figure 5A). Subsequently, all genes related to ccRCC are hierarchically clustered into 10 gene modules (Figure 5B). The correlation analysis shows that the green model (MEgreen) has the highest correlation with the risk score (Figure 5C). Then we performed a functional enrichment analysis of the genes in the green module, and we found that the most relevant signal pathway for risk score is the cell cycle (Figure 5D). Ultimately, we performed protein-protein interaction analysis (Figure 5E) on the genes in the first three signal pathways of Figure D.