Background
Colorectal cancer (CRC) is the third most common cancer and the fourth-leading cause of cancer death in the world [
1,
2]. Most CRCs originate from non-cancerous lesions by one or a combination of three different mechanisms: chromosomal instability, CpG island methylator phenotype, and microsatellite instability [
3]. Biomarkers of these cytogenetic alterations are of interest for diagnosis, prognostication, and anticancer drug development targeting CRC [
4]. Despite research efforts, genetic biomarkers currently have limited value as diagnostic or prognostic markers [
5].
Among the biomarkers, the solute carrier 2A (
SLC2A) gene family that encodes glucose transporter (GLUT) proteins has been widely investigated. GLUT proteins facilitate glucose influx into cancer cells which is necessary for cancer cell proliferation. Upregulation of
SLC2A genes is associated with poor prognosis in many cancers, including hepatocellular carcinoma, non–small cell lung cancer, and thyroid carcinoma [
6‐
8].
An association between overexpression of the subtypes of GLUT proteins and poor clinical outcomes has been reported in CRC [
9]. However, these studies were flawed by small sample sizes or subjective interpretation of immunohistochemical (IHC) staining. In this study, we analyzed the mRNA expression of all 14
SLC2A genes (corresponding to 14 GLUT proteins) and evaluated the associations with prognosis in CRC using data from the Cancer Genome Atlas (TCGA) database.
Discussion
Most cancer cells favor glycolytic energy metabolism over mitochondrial metabolism and oxidative phosphorylation chain for energy production, even in the presence of oxygen (Warburg effect). This explains why malignant cells overexpress GLUT family proteins, which is a plasma membrane transport system. Glucose can be translocated into the cell only via GLUT proteins. Expression and subcellular distribution of GLUT proteins are regulated by different signaling molecules and pathways such as
PI3K,
HIF,
p53,
Myc, and
AMPK LBk1 [
11,
12].
In the current study, we used the TCGA database to explore the clinical significance of SLC2 family genes in CRC. IHC staining of GLUT expression in cancer cells can be diverse which is a downside with IHC staining. This may be caused by incorrectly interpreted IHC stain results or different GLUT positive thresholds [
9]. Messenger RNA-gene expression analysis from TCGA data is superior to IHC and may best predict cancer prognosis in TCGA data [
13]. The current study is the first to report the expression of
SLC2A genes in CRC using the TCGA database.
In the current study, age and AJCC TNM stage were associated with survival in CRC patient. As expected, in this study, age > 65 was linked with worse prognosis for CRC patients. This may be due to aging itself, or due to co-morbidities such as cardiovascular disease or postoperative complications [
14]. Older patients are less likely to be treated with resectional surgery than younger patients and have poorer survival outcomes [
15]. We included age > 65 as a factor for multivariate analysis because age > 65 is useful for prognostication. We found that AJCC stage III and IV, which has nodal metastasis or organ metastasis, was associated with OS and DFS in the TCGA database. AJCC TNM stage was established based on OS, and lymph node involvement in itself is considered to have a strong influence on OS and DFS in CRC [
16]. In oral squamous cell carcinoma and papillary thyroid carcinoma, GLUT3 was known to be a prognostic marker for OS, and was associated with advanced cancer stage (AJCC TNM stage III and IV) which has nodal metastasis [
8,
17]. In current study, we found high expression of
SLC2A3 was a prognostic factor for predicting OS and DFS in CRC and was not associated with AJCC TNM stage. Similar to current findings, GLUT3 is a significant marker of poor prognosis in laryngeal carcinoma, with no significant differences in nodal or distant metastasis between the GLUT3 negative and GLUT3 positive groups [
18]. Since
SLC2A3 or GLUT3 genes are associated with prognosis in CRC as well as thyroid or laryngeal carcinoma, it would be worthwhile to investigate whether these cancers have similar GLUT-dependent metabolic pathways.
In the current study, we analyzed the mRNA expression values of all 14
SLC2A family genes and found that
SLC2A3 is independently associated with both OS and DFS after adjusting for age and AJCC TNM stage in CRC patients. Among the GLUT family, GLUT1 is associated with prognosis in several cancers, including CRC. GLUT3 has high affinity glucose uptake, similar to GLUT1 [
19]. GLUT3 is overexpressed in human carcinomas including CRC [
20]. Although GLUT1 and GLUT3 have many similarities, including expression pathways, the effects of GLUT3 on outcomes in several cancer varieties are not as well understood as GLUT1. This is the first report to study the effects of
SLC2A3 on CRC patients’ survival.
The following reasons may explain why the expression of
SLC2A3 affects CRC prognosis. GLUT3 is induced by hypoxia-inducible factor (HIF) formation in response to hypoxia in carcinomas [
11]. HIF-1 promotes tumor metastasis into distant and more oxygenated tissue through the transcriptional activation of oncogenic growth factors such as transforming growth factor beta3, epidermal growth factor, and others [
21]. Solid tumors with high hypoxia levels are more malignant, more likely to metastasize, and have a worse prognosis [
11]. GLUT3 is also induced by Akt involved in the Warburg effect. In cancer cells, Akt increases expression of GLUT1 and GLUT3 by causing degradation of p53.It may reflect the activity of hypoxia independent oncogenic pathways [
11,
19].
SLC2A1 expression was associated with poor DFS but not with OS. This corresponds well with a study which reported that GLUT-1 expression is associated with poor DFS but not with OS in rectal cancer patients [
9]. To further evaluate the association between the expression of
SLC2A1 and rectal cancer, we performed a subgroup analysis for rectal cancer patients using a univariate Cox regression analysis. Although we were unable to confirm our findings in the validation set due to a lack of primary site information, in the discovery set,
SLC2A1 expression in rectal cancer patients (
n = 72) was significantly associated with DFS (HR 1.57, 95% CI: 1.04–2.38;
P = 0.03). We also found no association between
SLC2A1 expression and OS (HR 1.54, 95% CI: 0.96–2.46;
P = 0.07) in rectal cancer patients. Our findings were consistent with the findings of the previous study.
In addition to
SLC2A family genes, the following 12 genes are known to be involved in glucose metabolism:
MTOR,
RICTOR,
HIF1A,
MYC,
PDK1,
PDK2,
PDK3,
PDK4,
PIK3R1,
PKM,
POU2F-1, and
RPTOR [
12]. We conducted an analysis to evaluate if the expression of the genes was associated with survival outcomes of CRC. We found no association between the other 12 genes’ mRNA expression and survival outcomes of CRC in both discovery and validation sets (Additional file
1: Table S1). Based on these findings, we concluded that among the glucose metabolism regulating genes, only the
SLC2A3 gene is significantly associated with the survival outcomes of CRC.
There are several genes known to be associated with the prognosis of CRC:
BRAF,
KRAS,
HIF,
TP53 and thymidylates synthase (TYMS) [
22,
23]. The
BRAF mutation generates an abnormality in the MEK/ERK signaling pathway in CRC [
24] and has been reported to be associated with poor prognosis by many CRC studies. However, reports vary on its association with survival [
25]. Mutation of
KRAS, a proto-oncogene, activates
RAS signaling pathways, but its association with CRC survival is not clear [
24,
26]. In the current study, we analyzed the association of
BRAF and
KRAS expression level with survival, and found neither to be associated with OS or DFS in the discovery set (Additional file
1: Table S2, Table S3 and Additional file
2: Figure S1). Likewise, in the recent study conducted by authors’ group, the mutational status of
BRAF or
KRAS was not associated the prognosis of CRC [
27]. For
HIF,
TP53 and
TYMS, there are studies which reported the expression of the genes are not related to the prognosis of OS of CRC [
28‐
30]. In this study, we also found that
HIF,
TP53 and
TYMS were not associated with OS or DFS (Additional file
1: Table S4). The discrepancies among the studies regarding the prognostic value of these genes may result from differences in patient cohorts, available co-variates, or statistical methods.
The short observation period of the discovery set patients (31.4 months median follow-up period) is a limitation of our study. Secondly, the relationship between the expression of
SLC2A mRNA and GLUT protein in CRC has not been confirmed. Although one study reported a close correlation between the expression of
SLC2A1/
SLC2A3 mRNA and that of GLUT1/GLUT3 proteins in thyroid carcinoma [
31], further studies are needed to investigate whether the expression of
SLC2A mRNA correlates with expression of GLUT protein mRNA in CRC. Another limitation of our results is that mRNA gene expression value is not a readily available parameter, especially in clinical settings, due to the high cost of storage and processing of fresh tissue. Its application may become wider in the near future when cost is reduced and stable mRNA expression can be obtained through formalin-fixed paraffin-embedded tissue samples.