Background
Cancer is regarded as a type of metabolic disease. Tumor cells can drive certain metabolic pathways to sustain their biological processes for growth and to adapt to complex tumor microenvironments (TMEs) [
1]. A well-established metabolic pathway that plays a prominent role in cancer progression is glycolysis, which is critical for supplying energy and producing metabolic end products, thus maintaining tumor cell survival [
2]. In addition to its functions in sustaining tumor growth, activation of glycolysis affects other phenotypic changes. For example, lactic acid produced by the glycolysis pathways induces the epithelial-to-mesenchymal transition (EMT) in lung cancer cells [
3]. Furthermore, a pan-cancer study reported that activated glycolysis is correlated with increasing tumor immunity [
4]. Hence, understanding the underlying relationship between glycolysis and cancer progression is a critical goal in cancer research.
Long noncoding (lnc)RNAs, which are longer than 200 nucleotides, can modulate gene expressions through various mechanisms. Several oncogenic signaling pathways, such as the cell cycle [
5], immune regulation [
6], and EMT mediation [
7], are linked to lncRNA regulation. Different types of lncRNAs have been revealed to promote glycolysis activation. In breast cancer, lncRNA-SNHG7, which is promoted by c-MYC regulation, can increase glycolysis through inhibiting miR-34a-5p expression [
8]. Another lncRNA, long intergenic noncoding RNA for IGF2BP2 stability, suppresses the degradation of IGF2BP2 by inhibiting the ubiquitination–autophagy pathway, leading to glycolysis upregulation in colorectal cancer [
9]. However, most studies have focused only on particular lncRNA candidates and their functions in promoting glycolysis. Therefore, we systematically investigated the association of lncRNAs with glycolysis in different cancer types and explored their related signaling pathways and clinical relevance.
In this study, we used pan-cancer data from The Cancer Genome Atlas (TCGA) to identify glycolysis-associated lncRNAs across 33 tumor types. We performed a consensus clustering analysis to classify these glycolysis-associated lncRNAs into distinct clusters. We then identified glycolysis-associated lncRNAs that exhibited key clinical effects in five cancer types. Finally, we explored the potential pathways and functions of glycolysis-correlated lncRNAs in association with oncogenic signaling such as EMT and immune regulation.
Discussion
Glycolysis, a metabolic pathway, supplies energy, promoting cancer malignancy. Although the evidence of a crosstalk between glycolysis and lncRNA regulation in cancer progression is mounting [
28,
29], a clinical link connecting lncRNAs and glycolysis is not fully understood. We conducted a pan-cancer scale analysis to identify glycolysis-associated lncRNAs in 33 cancer types. We determined that the prognoses of five cancer types, BLCA, PAAD, LGG, MESO, and UVM, were significantly correlated with glycolysis-associated lncRNA signatures. Furthermore, we comprehensively analyzed changes in gene mutations, molecular subtypes, TFs, oncogenic signal pathways, and immune cell infiltration in patients stratified by glycolysis-associated lncRNAs. Finally, we identified five lncRNAs, namely MIR4435-2HG, AC078846.1, AL157392.3, AP001273.1, and RAD51-AS1, which exhibited significant correlations with glycolysis across the five cancer types. In particular, MIR4435-2HG was suggested to play a critical role in connecting glycolysis, EMT, and immune cell infiltration in the cancers.
For the glycolysis evaluation of each cancer patient, we chose the ssGSEA-derived glycolysis score instead of the direct correlations of glycolysis-associated genes because of the large number of glycolysis-involved genes in this methodology (200 gene candidates). Directly correlating lncRNAs with these genes would generate hundreds of correlation results and p values, making evaluating the degree of association of each lncRNA with glycolysis activation difficult. Therefore, an ssGSEA algorithm was utilized to treat the 200 glycolysis-involved candidates as a gene set to perform a GSEA for individual patients. Through this method, we were able to compress the genes into a single score for each patient and capture the degree of glycolysis activation. We did not directly use the target genes of lncRNAs due to the complexity of the regulatory mechanisms of lncRNAs. The functions of lncRNAs in post-transcriptional regulation include the microRNA sponge, interaction with the chromatin modulator, and direct targeting of downstream RNA. In addition, only focusing on glycolysis-involved genes that belong to lncRNA direct targets would risk our overlooking other promising candidates. Furthermore, no bioinformatics tools that can accurately predict lncRNA target genes are yet available. Therefore, ssGSEA-derived glycolysis scoring may be a suitable tool for identifying glycolysis-associated lncRNAs.
Relationships between glycolysis and the immune resistance of cancers have recently been reported. Tumor cells utilize glucose and metabolically compete with T cells through impairing mammalian target of rapamycin (mTOR) activity and glycolytic activity in T cells, leading to the overriding of the capability of T cell-mediated cytotoxicity [
30]. In addition, highly activated glucose metabolism in cancer cells promotes lactate accumulation in the TME [
31]. This extracellularly accumulating lactate blocks lactate export from T cells, leading to the generation of dysfunctional aerobic glycolysis, a crucial mechanism for maintaining T cell effector function. In addition to immune regulation, metabolic reprogramming (including glycolysis) with highly invasive and drug resistance features was reported to possess the ability to transform cancer cell phenotypes toward EMT [
32]. The EMT process has also identified to be linked to immune evasion by cancer cells [
33,
34]. However, relationships among glycolysis, EMT, and immune regulation in cancers are still not fully understood. In our analyses, we determined that EMT and inflammatory responses, including immune-suppressive ligand expression, were enriched in cluster 3 cancer patients, as classified by glycolysis-associated lncRNAs. These findings suggest that the TME of cluster 3 cancer patients is surrounded with different types of immune cells, and the upregulated immune-suppressive ligands and EMT activation protect cancer cells from attack by immune cells. Therefore, the “hot” immune microenvironment in cluster 3 cancer patients implies that immune checkpoint blockade therapy might be suitable.
In our genomic mutation analysis, we identified specific gene mutations that were associated with glycolysis-associated lncRNA-stratified clusters in different cancers. Some of these genes have been reported to be involved in glucose metabolism. For example, in mice bearing RB1 null lung cancer [
35], upregulation of glucose transporter (GLUT) 1 and two rate-limiting enzymes in glycolysis, hexokinase-2 (HK2) and pyruvate kinase isozymes 2, was observed, suggesting that RB1 regulates glucose metabolism. In the present study, the cluster with high glycolytic activity mainly exhibited RB1 mutation in BLCA. Although loss of RB1 may serve as a poor prognosis predictor in BLCA [
36], the relationship between RB1 and glycolysis in BLCA remains unclear. Additionally, in PAAD and UVM, we identified KRAS and BAP1 mutations, respectively, as enriched in clusters with high glycolytic activity. In PAAD, KRAS mutation activates glycolytic signaling mainly through MEK activation and Myc-dependent transcription, resulting in the upregulation of GLUTs and rate-limiting enzymes of glycolysis, such as HK2, phosphofructokinase-1, and lactate dehydrogenase A (LDHA) [
37]. A multi
-omics analysis integrating transcriptome, metabolite, and genomic analysis revealed that UVM cells with BAP1 mutation maintain their energy demand through oxidative phosphorylation and glycolytic pathway [
38]. Additionally, the distinct metabolic features of mutant BAP1 and wild-type UVM further lead to different responses to metabolic inhibitors. These findings indicate that KRAS and BAP1 mutations respectively drive metabolic alterations in PAAD and UVM. In gliomas, the cluster with high glycolytic activity mainly belongs to IDH1 wild-type cancer patients, as per our findings. A previous study reported that glioma cells with IDH1 mutation produce 2-hydroxyglutarate, which promotes HIF-1α degradation, accompanied by the downregulation of glycolysis-related genes including
SLC2A1,
PDK1,
LDHA, and
SLC16A3 [
39]. By contrast, IDH1 wild-type gliomas exhibited an activated glycolysis pathway. Taken together, our lncRNA-stratified clusters not only exhibit different glycolytic activities but also distinct genomic mutations that have been reported to be involved in glycolysis signaling. However, whether these genomic mutations might be the drivers altering glycolysis-associated lncRNA profiles requires further investigation.
In our TF analysis, we determined that activated prominent TFs in highly glycolytic clusters across BLCA, LGG, PAAD, and UVM were mainly involved in TGF-β signaling and SMAD protein complex assembly. SMAD proteins are downstream signal transducers of the TGF-β signaling pathway, which functions as an immune-suppressive regulator in cancers. For instance, Smad3-mediated TGF-β signaling was reported to suppress the cytotoxic activity of NK cells by blocking the production of CD16-mediated IFN-gamma [
40]. Another SMAD protein, SMAD4, has dual roles in regulating NK immunity in a context-specific manner [
41]. In the initial phase of tumor formation, SMAD4 has a mainly positive effect in promoting the development and antitumor activity of NK cells. However, at the late stage of cancer development, NK cells are surrounded with TGF-β produced by tumor cells. In this condition, SMAD4 cooperates with p-SMAD2 and p-SMAD3 to suppress NK cell-mediated cytotoxicity. Our analyses revealed that the group with high glycolytic activity demonstrated a microenvironment with highly infiltrated NK cells. Furthermore, coordinated upregulation of SMAD2, SMAD3, and SMAD4 was also observed in highly glycolytic patients. Taken together, these results imply that highly glycolytic tumor cells might impede NK-mediated immunity through SMAD signaling, and such tumor cells might be vulnerable to SMAD4 inhibition. However, the relationship between glycolysis-related lncRNAs and SMAD complex regulatory mechanisms in tumor cells needs to be further validated.
Among the glycolysis-associated lncRNAs we identified, some lncRNAs have been reported to be directly involved in glycolysis signaling. For example, plasmacytoma variant translocation 1, which exhibited positive associations with glycolysis scores in six of thirty-three cancer types in our findings, was suggested to function as a microRNA sponge in suppressing miR-497 expression, leading to the promotion of HK2 upregulation and osteosarcoma progression [
42]. Nuclear factor (NF)-κB-interacting lncRNA (NKILA), which was positively correlated with glycolytic activity in the five cancer types, activates hypoxia-inducible factor 1α expression to promote the hypoxia-mediated Warburg effect on gliomas [
43]. Some lncRNA candidates in our findings might also be implicated in regulating glycolysis. For instance, cytoskeleton regulator RNA was reported to interact with Sam68, an RNA-binding protein that possesses an oncogenic function in cancers [
44]. Furthermore, Sam68 can upregulate the expression of pyruvate kinase isozymes M2 (PKM2), a key enzyme for glycolysis-dominated energy metabolism [
45], through facilitating the transport of PKM2 messenger (m)RNA from the nucleus to cytoplasm [
46]. MIR4435-2HG, an lncRNA positively correlated with glycolysis in 20 cancer types, enhances YAP expression for cancer progression [
47]. Furthermore, YAP is recognized as a TF and is responsible for upregulating the expression of glucose metabolism enzymes such as HK2 and 6-phosphofructo-2-kinase/fructose-2,6 biphosphatase 3 [
48]. However, their roles in the glycolysis process require further investigation.
Among the five lncRNAs that were consistently correlated with glycolysis scores across different cancers in our findings, MIR4435-2HG was identified as being highly associated with glycolysis and its correlated genes. By performing a multivariate linear regression adjustment, we also uncovered that immune and EMT-involved genes are positively correlated with MIR4435-HG expression. Several studies have demonstrated the oncogenic roles of MIR4435-2HG in cancer processes. MIR4435-2HG promotes gastric cancer cell migration and proliferation through Wnt/β-catenin signaling [
49]. The upregulation of MIR4435-2HG promotes oral squamous carcinoma cell proliferation through inducing TGF-β1 upregulation [
50]. Nevertheless, few studies have indicated links among MIR4435-2HG, glycolysis, and immune regulation in cancers. Our analyses suggest that MIR4435-2HG participates in interrelations among glycolysis, immune resistance, and EMT. The present study has limitations. The lack of public available RNA sequencing data limited us to validate the classification effectiveness of our genome classifier which is predominantly based on the glycolysis-associated lncRNAs or the gene candidates. More clinical data of individuals with other types of cancer or from large cohort studies are required for validation. Several lncRNAs in our findings have rarely been reported, especially those negatively correlated with glycolysis. These should be further investigated.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.