To evaluate the association between tumor differentiation and glucose metabolism in the thyroid cancer, we used two cohorts, papillary thyroid cancer samples of TCGA and PDTC/ATC samples obtained by GEO (GSE76039). The clinical and pathological characteristics of PTC patients and PDTC/ATC patients are summarized in Table 1 and Table 2.
Firstly, GLUT and glycolysis enrichment scores were estimated by gene expression data of the two cohorts and associated with the mutation status, BRAFV600E. The enrichment scores of GLUT and glycolysis were significantly different in PTC according to BRAFV600E mutation status (Figure 1). BRAFV600E positive PTC have higher GLUT signature and lower glycolysis signature (BRAFV600E positive group 20.5±1.1 vs. BRAFV600E negative group 19.1±1.5, t = -12.093, p < 0.0001 for GLUT; BRAFV600E positive group -0.28±0.77 vs. BRAFV600E negative group 0.24±1.11, t = 6.057, p < 0.0001 for glycolysis) than BRAFV600E negative PTC (Figure 1).
To evaluate the tumor differentiation, we employed tumor differentiation score, TDS, calculated by sixteen genes related to thyroid functions. The TDS was compared with glucose metabolic profiles. In PTC group, the TDS was negatively correlated with GLUT signature, but positively correlated with glycolysis signature (r = -0.585, p < 0.0001 for GLUT; Figure 2A, r = 0.334, p < 0.0001 for glycolysis; Figure 2B). We divided PTC into two subgroups according to the BRAFV600E mutation status. PTC with BRAFV600E showed a negative correlation between TDS and GLUT (r = -0.570, p < 0.001), while it showed a positive correlation between TDS and glycolysis (r = 0.352, p < 0.001). PTC without BRAFV600E showed a trend of negative correlation between GLUT and TDS (r = -0.177, p = 0.065) and no significant correlation between glycolysis and TDS. These correlations were significantly stronger in BRAFV600E negative group than positive group (z = 5.22 for GLUT, z = -4.48 for glycolysis, and p < 0.001 for both GLUT and glycolysis; Figure 2C-F). In PDTC/ATC group, the relationship between the TDS and GLUT signature showed a significant negative correlation (r = -0.682, p < 0.0001; Figure 2G). The TDS also showed a significant negative correlation to the glycolysis signature (r = -0.384, p = 0.019; Figure 2H). We also evaluated whether different cell types of PTC were associated with glucose metabolic signatures. Classical cell type PTC have higher GLUT signature and lower glycolysis signature than follicular cell type PTC (Classical cell type 20.1±1.3 vs. Follicular cell type 18.4±1.2, t = 11.547, p < 0.001 for GLUT; Classical cell type -0.13±0.94 vs. Follicular cell type 0.52±1.12, t = -5.653, p < 0.001 for glycolysis; Additional file 1: Figure S1A, B). The negative correlation between TDS and GLUT and the positive correlation between TDS and glycolysis were found in both cell types, classical and follicular types. The strength of their correlations showed no significant difference (Classical cell type: r = -0.467, p < 0.001 for GLUT, r = 0.228, p < 0.001 for glycolysis; Follicular cell type: r = -0.431, p < 0.001 for GLUT, r = 0.338, p = 0.001 for glycolysis; Additional file 1: Figure S1C, D). Individual genes that constitute TDS and glucose metabolic pathway were represented by heatmaps (Figure 3).
We assessed the association between signatures of glucose metabolism and patients’ prognosis. Kaplan-Meier survival curves of both PTC and PDTC/ATC patients with low glycolysis signature showed significantly better survival than the other group (p = 0.045, p = 0.015, respectively; Figure 4A, B). The hazard ratios (H.R.) of the influence of the glycolysis signature on the recurrence of PTC are presented in Table 3. In PTC, the high glycolysis signature alone had a significant influence on the recurrence-free survival (H.R. = 1.497; C.I. = 1.034–2.167; p = 0.033). The adjustment for age and gender maintained its influence on the recurrence-free survival (H.R. = 1.497; C.I. = 1.038–2.160; p = 0.031). Additional adjustment for T-stage and N-stage still demonstrated worse prognosis on PTC patients with high glycolysis (H.R. = 1.915; C.I. = 1.223–2.999; p = 0.005). GLUT score showed no significant correlation with PTC patients’ prognosis (p = 0.85; Figure 4C). On the contrary, we found that PDTC/ATC patients with low GLUT signature have significantly longer disease free survival than the other group (p = 0.0063; Figure 4D).