Background
Intervertebral disc degeneration (IDD), a complex multifactorial and age-dependent condition, is a critical contributor to low-back pain (LBP) [
1]. The increasing incidence of IDD not only affects the life quality of patients because of chronic pain and disability, but also causes a severe financial burden to the health care system [
2]. Moreover, current prevention, diagnosis, and treatment modalities for IDD are hampered due to its ambiguous pathogenic mechanisms. Previous studies have shown that IDD is characterized by the abnormal proliferation of nucleus pulposus (NP) cells, degradation of proteoglycans and collagen in the extracellular matrix (ECM), imbalance of homeostasis, and accelerated transition of NP cells from a healthy state to a catabolic and degenerative state [
3]. It is known that the cellular function of all life forms relies closely on metabolism [
4,
5]. Metabolic genes are believed to play essential roles in various cellular functions and many diseases such as hepatic fibrosis, central nervous system diseases, and hematologic malignancies [
6‐
8]. However, investigations concerning the role of metabolic networks in IDD remain sparse. Hence, in-depth comprehension of metabolic gene expression and the regulatory mechanisms of noncoding RNA may provide novel insights into the onset, development, and diagnosis of IDD.
Long noncoding RNAs (lncRNAs), a group of RNA molecules consisting of more than 200 nucleotides with no or feeble protein-coding ability [
3], have been of great interest because of their ability to regulate gene expression in various pathological and biological processes, such as cell proliferation and apoptosis [
9,
10]. MicroRNAs (miRNAs), consisting of approximately 18–22 nucleotides without coding function, mediate post-transcriptional regulation of target messenger RNA (mRNA) [
11]. Ample studies indicate that lncRNAs can act as natural microRNA (miRNA) sponges to bind miRNAs, acting as competing endogenous RNAs (ceRNAs) [
12‐
14]. The ceRNA crosstalk plays a critical role in modulating gene expression [
15]. Accumulating evidence has demonstrated the aberrant expression of several mRNAs, miRNAs, and lncRNAs in NP cells in IDD and explained the relationship between them with respect to autophagy, apoptosis, and cell cycle [
11,
16,
17]. Nevertheless, metabolic gene-associated ceRNA regulated IDD progression remains largely unstudied. Hence, further exploration of metabolism-related ceRNA crosstalk in NP cells will benefit our understanding of the metabolic gene regulatory network, which is of considerable significance for understanding IDD pathogenic mechanisms. Besides, it is vital to identify the key metabolic gene(s) and relevant key regulated noncoding RNAs in the IDD. In addition, we previously screened core RNAs of IDD based not only on differential expression (DE) but also using the machine learning technology, namely support vector machine (SVM) [
18].
Currently, the definitive diagnosis of IDD is sometimes difficult because the intervertebral disc has a particular anatomical structure. In addition, LBP usually has atypical clinical features [
19], and the definitive diagnosis is not often achieved using clinical imaging methods [
20]. It is even more difficult to correctly diagnose dorsal disc migrations [
21], and misdiagnosis often occurs [
22]. A transcriptomic signature is beneficial for clear and early diagnosis and timely treatment of patients with IDD [
23], especially those that are at a high risk of IDD. Currently, the metabolic gene signature is a collection of metabolism-related genes that can characterize the outcome events; it accounts for the top differentially regulated genes in many disorders but is relatively poorly explored in IDD [
24,
25]. In the current study, we aimed to explore the key metabolic gene(s) signature and its key ceRNA regulatory systems to provide insight into IDD pathogenesis. Combining database prediction and gene expression validation, we explored the metabolic abnormalities of NP cells in IDD and revealed their underlying metabolic, molecular signatures and regulation profiles based on microarray datasets.
Discussion
The metabolic homeostasis and the construction of the ECM of intervertebral discs mainly depend on its active NP cells, which form the inner core of the intervertebral disc. An imbalance of cellular functions because of metabolic imbalance, nutrient deprivation, dysregulated apoptosis, and some transcellular signaling interrupts homeostasis in NP cells, resulting in IDD [
31‐
33]. Furthermore, the metabolic signature represents a broad molecular signature of systemic metabolism and covers multiple cellular metabolic pathways [
34]. The metabolic signature and its relevant mechanism are critical for understanding the cellular mechanisms of various diseases. Accordingly, it would be reasonable to assume that NP cell metabolic disorder might be a pathogenic factor of IDD. However, the metabolic investigations and associated ceRNA regulation mechanism of NP cells in IDD remaine inconclusive. The ceRNAs regulate the expression of various genes and play a significant role in the pathogenesis of many diseases [
35]. Identifying ceRNA regulation with the metabolic signatures in NP cells would shed a novel light on IDD. Thus, based on the GEO dataset, we identified a novel gene metabolic signature and related ceRNA regulatory mechanisms to gain insight into the IDD pathogenic mechanism.
In the present study, 45 DEGs and 36 DElncRNAs were identified. MiR-338-3p was the only key DEmiRNA, which was upregulated in IDD. A vast ceRNA network was constructed to reveal the underlying molecular signatures of IDD. GO annotation revealed that the DEmRNAs/DEGs were mostly enriched in the PI3K-Akt signaling pathway. The PI3K-Akt signaling pathway participates in the synthesis of ECM, regulating apoptosis and cell proliferation in IDD. Furthermore, the GO annotation showed that many biological processes were categorized in response to oxidative stress of NP cells, which indicates that oxygen and oxygen-associated processes might play an essential role in IDD. The KEGG pathway enrichment analysis showed that the MAPK signaling pathway, IL-17 signaling pathway, apoptosis, and TNF signaling pathway might be related to IDD. Based on batch normalization, we found that miR-338-3p was the only DEmiRNA validated in two datasets and was correctly predicted for lncRNA-miRNA and mRNA-miRNA binding. However, there were 45 overlapping DEGs and 948 genes defined by the metabolic gene set. To further determine the key metabolic signature and critical lncRNA(s) of the 36 DElncRNAs in IDD, we used the SVM method. Thus, AC063977.6 was identified as the critical lncRNA with recursive feature elimination, based on the results of SVM and related expression regulation of DElncRNAs. Moreover, PFKFB2 is the key metabolic gene with recursive feature elimination, based on the results of SVM, metabolic gene set, and related expression regulation of DEmRNAs.
According to the online database prediction, AC063977.6 directly targets miR-338-3p, whereas miR-338-3p directly targets PFKFB2, suggesting that AC063977.6, miR-338-3p, and PFKFB2 might form an axis that modulates NP cell metabolism in IDD. Furthermore, the gene expression validation based on microarray data of GSE116726 and GSE19943 further enhanced the credibility of this axis. In conclusion, we found that AC063977.6, miR-338-3p, and PFKFB2 formed an axis in IDD to modulate the metabolism of NP cells. This study revealed that miR-338-3p was upregulated in degenerating NP cells, accompanied by downregulation of PFKFB2 expression. Moreover, AC063977.6, acting as a ceRNA sponging miR-338-3p, was significantly attenuated in IDD progression.
The central part of the DEGs is
PFKFB2, which is an isoform of the PFKFB family that mainly regulates glycolytic metabolism and catalyzes the fructose 2,6-bisphosphate (Fru-2,6-P) [
36]. PFKFB2 is a key bifunctional enzyme involved in glycolytic metabolism [
37]. Moreover, glycolysis significantly diminishes cellular oxidative stress [
38,
39]. Similarly, bioinformatics analysis with the GEO microarray datasets revealed that the DEGs were significantly enriched in the oxidative stress process in GO enrichment. Compared to normal cells, several cancer cells exhibit elevated expression levels of
PFKFB [
40]. However, in this study, we found that the expression level of
PFKFB2 in NP cells with IDD was lower than that in control NP cells, suggesting that reduced glycolysis may increase the level of cellular oxidative stress in NP cells; thus, it may be associated with IDD.
Accumulating evidence suggests that lncRNAs and miRNAs dysregulate essential pathological processes of IDD, such as apoptosis, angiogenesis, ECM degradation, and inflammatory responses [
41]. The miR-338-3p originates from the apoptosis-associated tyrosine kinase gene, and it is dysregulated in many tumors and plays distinct roles in different diseases [
42]. The upregulation of miR-338-3p could promote glioma cell invasion and metastasis of lung cancer, whereas the downregulation of miR-338-3p was related to poor outcomes in gastric cancer [
43‐
45]. In this study, miR-338-3p was remarkably upregulated in IDD, serving as an impetus with metabolic dysregulation by targeting
PFKFB2, thereby disturbing the metabolism of NP cells. IDD severely threatens the health of patients, and identification of genetic signature is beneficial for its early diagnosis and early treatment [
23]. Moreover, a strong negative correlation was observed between IDD and
PFKFB2 expression levels in both the training cohort and validation cohort, which indicates that low expression of
PFKFB2 is a key factor in IDD. The results suggest that
PFKFB2 is an player in IDD. According to our experience in clinical work and reported literature, misdiagnosis of IDD occasionally occurs, which often leads to incorrect treatment and even unnecessary surgery. Based on our results, the metabolic signature could be used in providing clues for IDD mechanism research, and may aid in surgery for differential diagnosis in clinical practice.
Limitations of this study
There are some limitations to this study. First, we focused only on the regulatory roles of target genes without further classification into specific subgroups according to their functions, which limited the excavation of those data. Second, further in vitro and in vivo experiments on the sophisticated regulating mechanism involved with AC063977.6 and miR-338 in IDD progression are urgently needed. Third, although a small sample size is a common problem in orthopedic research, it is undeniable that this study is based on a small sample size (a total of 46 samples were included). Sufficient in vitro and in vivo experiments are warranted to establish the functions and mechanisms of the lncRNAs and miRNAs involved in the pathogenesis of IDD and may further promote the translation of precise gene diagnosis in IDD.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.