In this study, we used transcriptome data of thyroid cancers of two cohorts, PTC and PDTC/ATC to evaluate glucose metabolic profiles and tumor differentiation. In PTC, a trend of higher GLUT and lower glycolysis was found in tumors with BRAFV600E mutation and those with relatively poor differentiation. There is a controversy regarding the correlation of BRAFV600E, GLUT1, and tumor differentiation genes according to several previous reports (16-18). Our results are consistent with other results of higher GLUT1 in less differentiated thyroid cancer (16) and negative correlation between BRAFV600E mutation and tumor differentiation genes in PTC (17, 18). The results of TDS positively correlated with glycolysis and negatively correlated with GLUT were particularly found in PTC without BRAFV600E mutation. Moreover, the high glycolysis enrichment was significantly associated with poor clinical outcome, even it was associated with well differentiation in PTC. However, this paradoxical opposite direction of correlation was not found in PDTC/ATC cohorts, which showed both GLUT and glycolysis were negatively correlated with TDS and associated with poor clinical outcome.
We have found that thyroid cancers with poor differentiation showed higher GLUT expression (Figure 2). Cancer cells demand a higher amount of energy according to the progression, which is associated with enhanced aerobic glycolysis in the advanced cancer (19). The glucose demand of cancers cause overexpression of GLUT1 and/or GLUT3 to increase glucose influx (20). Our results were compatible with the previous studies since poorly differentiated thyroid cancers would need higher glucose uptake through GLUT expression (21). On the other hand, glycolysis signatures of thyroid cancers with poor differentiation were inconsistent between PTC and PDTC/ATC (Figure 2). In PTC, relatively well-differentiated tumors, glycolysis was positively associated with TDS. In PDTC/ATC, a negative correlation was shown between TDS and glycolysis. Considering PTC of TCGA data are relatively well-differentiated tumors compared with PDTC/ATC cohort, the association between the differentiation and glycolysis might have ‘U shape’ pattern; high glycolysis signatures were shown in ATC and some types of well-differentiated PTC. Moreover, increased glycolysis was associated with poor clinical outcome in spite of high TDS. It implies that the differentiation of thyroid cancer may not be the only factor that reflects the biological progression of thyroid cancer. Instead, in addition to differentiation, glucose metabolic profiles represented by glycolysis should be further considered to infer the progression of thyroid cancer, particularly for the specific subtypes of PTC, BRAFV600E negative and/or follicular variants.
Our results demonstrated that the signatures of GLUT and glycolysis can act as prognostic factors in predicting recurrence-free survival of thyroid cancer patients (Figure 4). The multivariate analysis revealed that the two variables, N-stage and the glycolysis signature, were significantly associated with the recurrence. As N-stage has been regarded as an important conventional biomarker related to prognosis which can be confirmed by surgical exploration, our results emphasized again the importance of the lymph node status in thyroid cancer as postoperative risk stratification. In terms of another prognostic marker in our results, glycolysis signature, it is notable that glucose metabolism profiles can be noninvasively estimated by FDG PET. GLUT score, calculated by GLUT1 and GLUT3, was reported as the major deterministic factor for the FDG uptake in various studies (10, 15, 21, 22), while a recent study showed a moderate correlation between FDG uptake and GLUT regarding a complex mechanism of glucose metabolism (23). Furthermore, glycolysis activity is also associated with FDG uptake in vivo (22, 24). According to the kinetic model of FDG, glycolysis activity is associated with FDG retention, which can be visualized by dual-time FDG PET (25, 26). In general, poorly differentiated thyroid cancers are known to have a worse outcome as compared to well differentiated thyroid cancers (27). However, a subset of well-differentiated carcinoma shows relatively poorly outcome in tumors with increased glycolysis with well-differentiated type. As GLUT and glycolytic activity were differently associated with TDS, noninvasive characterization using FDG PET and radioactive iodine imaging could play a role in risk stratification when considered with other prognostic factors as well as biological characterization of the tumor. As a future work, more specifically, dual-time FDG PET could be used for estimating glycolysis activity, which might identify a subtype of the tumor with enhanced glycolysis and well-differentiation.
Although, we analyzed glucose profiles of thyroid cancer in the two different cohorts with different differentiation, there are some limitations. Firstly, the PDTC/ATC sample size was small. Though we found the expected role of GLUT and glycolysis on the prognosis of PDTC/ATC patients, further studies with a larger group are needed. Another limitation was that the transcriptome data from two cohorts were not be combined and analyzed since those data were obtained from different resources, RNA sequencing and microarray. For PDTC/ATC patients, a few samples were acquired from recurred or metastatic tumors (recurred tumors for 6 samples and one sample from metastatic tumor), which might affect our analysis though recurred and metastatic tumors have similar glucose metabolic profiles with their primary tumors. Furthermore, noninvasive imaging biomarkers using iodine scan and FDG PET integrated with transcriptomic data could clarify our results of the association of differentiation and glucose metabolic profiles. We could expect the clinical application of our results as a future study, such as the estimation of glycolysis score based on gene expression profiles in fine-needle aspiration samples. In addition, according to our results, the expression of genes related to glycolysis may be examined by tissue samples in the clinic to stratify patients’ outcome, even though further studies focusing on clinical outcomes and clinical decision according to the glycolysis are needed.