More than 90% of human genes are transcriptional genes, but most of them do not encode proteins and are transcribed into non-coding RNAs (ncRNAs) [1]. NcRNAs are classified into two main types: housekeeping ncRNAs and regulatory ncRNAs. Regulatory ncRNAs include long ncRNAs (lncRNAs $>$200 nucleotides, nts), microRNAs (miRNAs, 21–23 nts), PIWI-Interacting RNA (piRNA, 25–33 nts) and circular RNAs (circRNAs) [2]. Regulatory ncRNAs interact with each other in a complex network to regulate both their abundance and gene expression [3].

Alzheimer’s disease (AD) is considered to be a complex heterogeneous disease. Its exact pathogenesis remains unknown, and has no effective treatment method [4]. Thus, there is a need for further research on underlying disease processes to better understand AD and aid the development of feasible treatment options. Considerable evidence shows that ncRNAs, particularly lncRNAs, miRNAs, and circRNAs, are a major class of regulatory molecules involved in the pathophysiology associated with AD [5].

LncRNAs are ncRNAs with a length of more than 200 nts and have been reported as important for cell regulation. Cytoplasmic lncRNA act as miRNA sponges or miRNA precursors to alter the expression and function of miRNA. They also recognize target messenger RNA (mRNA) and interact with cellular translation mechanisms. In SH-SY5Y cells, the upregulation of nuclear enriched abundant transcript 1 (lncRNA NEAT1) has been related to the downregulation of miR-27a-3p and upregulation of amyloid (A$\beta$), $\beta$-amyloid-precursor-protein-cleaver-enzyme 1 (BACE1) protein, amyloid precursor protein (APP), and tau protein [6]. Studies suggest that LncRNA Ribonuclease P RNA component H1 (Rpph1) participate in the regulation of the onset and progression of AD by competing with miR-330-5p and miR326-3p [7].

MiRNAs are small molecules with transcripts of 21–23 nts. Changes in the levels of several miRNAs and their targeted mRNA have been reported to affect the development and progression of the disease such as APP production, Tau aggregation, A$\beta$ production, protein misfolding, neuroinflammation, and oxidative damage [8]. For example, compared with controls, the expression of miR-485-3p was significantly increased in the brain tissues of AD patients [9], and MiR-485-3p antisense oligonucleotide (ASO) reduced the expression level of A$\beta$ via phagocytosis of A$\beta$ mediated by CD36 [9]. Additionally, it is reported that miR-146a overexpression reduces cognitive impairment, secretion of proinflammatory cytokines, A$\beta$ plaque accumulation and protects the neurons of APP/PS1 transgenic mice from damage [10].

CircRNAs are non-coding RNA molecules that do not have a 5${}^{\prime }$ end cap and a 3${}^{\prime }$ end poly (A) tail that form a ring structure by a covalent bond. CircRNAs have been reported to play a significant role in various biological functions as a miRNA sponge and transcription regulator. For example, circRNA_0003611 is augmented in AD patients. It was reported that silenced circRNA_0003611 might reduce neuronal damage triggered by A$\beta$ via adjusting the miR-383-5p/kinesin family member 1B (KIF1B) axis [11]. CircRNA Cwc27 is a neuronal-rich circRNA substantially expressed in the brain and significantly increases in AD patients and mice [12]. CircRNA Cwc27 deficiency significantly ameliorates AD-related pathological features and increases cognitive function [12]. CircRNA Cwc27 was shown to directly bind to purine-rich element-binding protein A (Pur-$\alpha$) and repress Pur-$\alpha$ recruitment to AD gene promoter clusters [13].

Competitive endogenous RNA (ceRNA) includes mRNA of coding-protein lncRNA and circRNA. Numerous studies have emphasized that these ceRNA molecules regulate miRNA target gene expression via competitively binding to the same miRNA by miRNA responsive elements (MREs). Recently, it was reported that the ceRNA network plays a critical role in many diseases, such as cardiovascular, neuroimmune and neurodegenerative disease [14, 15, 16]. Thus, the ceRNA network offers new approaches to understanding AD. Though some ceRNAs have been found related to AD progression, there is currently little discussion of the role of the ceRNA regulatory network in AD. In this study, data were obtained from the gene expression omnibus (GEO) database. Differentially expressed circRNAs (DEcircRNAs), mRNAs (DEmRNAs), lncRNAs (DElncRNAs), and miRNAs (DEmiRNAs) were screened from sequencing data. The targeted miRNAs and their related circRNAs and lncRNAs were forecast, an integrated ceRNA network was constructed, and their potential value for AD diagnosis was explored. The key pathways and regulatory mechanisms of central genes were analyzed and discussed using bioinformatics tools. MiRNAs, targeted mRNAs, related lncRNAs, and circRNAs involved in the ceRNA network may be possible biomarkers and treatment targets for AD. The workflow is shown in Fig. 1.

Fig. 1.

Study design flowchart. DEmRNAs, differentially expressed mRNAs; miRNAs, microRNAs; circRNAs, circular RNAs; lncRNAs, long ncRNAs; PPI, protein-protein interaction; ceRNA, competitive endogenous RNA.

Various ceRNAs are abnormally expressed in AD [23, 24, 25] and mounting evidence has shown the relationships among dysregulated lncRNAs, circRNAs, miRNAs, and mRNAs and the construction of AD-related ceRNA networks [26]. The ceRNA network is beneficial for a deeper understanding of the AD mechanism. In this study, transcriptome data of AD patients from the GEO database was used that contained mRNAs, circRNA, lncRNAs, and miRNAs. A total of 1381 DEmRNAs, 283 DEcircRNA, 484 DElncRNAs, and 372 DEmiRNAs were identified. A circRNA/lncRNA-miRNA-mRNA ceRNA network with biologic functions in AD was constructed following bioinformatic analysis. It is noteworthy that the ceRNA hypothesis indicates that there is a negative correlation between the ceRNA and miRNA expression levels but a positive correlation with mRNA expression. Therefore, the predicted relationships were integrated with the corresponding expression data to obtain more reliable results.

Through functional enrichment analysis, the upregulated DEmRNAs were mainly associated with regulation of response to biotic stimulus, regulation of defense response and regulation of innate immune response. Additionally, the downregulated DEmRNAs were mainly involved in translation, peptide biosynthetic process and peptide metabolic process. Importantly, ten hub genes were identified (RPL11, RPL34, RPL21, RPL22, RPL6, RPL32, RPL24, RPL35, RPL31 and RPL35A) using the CytoHubba tool. These ten hub genes were found to be downregulated in AD samples. The results of enrichment analysis showed that hub genes were mainly involved in ribosomes. These hub genes are members of a ribosomal protein family, which is important for the end and correctness of translation in the ribosome [27]. Ribosomal proteins were reported to be significant parts of ribosomes and participate in regulating bioprocesses [28]. Recently, it was reported that ribosomal proteins are tightly related to the progress of AD [29, 30, 31].

Previous investigation has shown that the low expression level of RPL11 is related to ribosome pathway dysfunction during AD development, which is the same as the conclusion of this analysis. There was evidence to suggest that RPL11 was correlated with brain maturation [32]. RPL11 contributed to the cortical neuron function damaged by ribosomal stress in neurodegenerative diseases [33]. Thus, it was speculated here that RPL11 may influence AD progression via the ribosome pathway. RPL34, a member of the ribosomal protein L34E family, is a highly conserved protein and has been demonstrated to participate in the occurrence and development of human malignancies. No evidence showed a role for RPL34 in AD development. The RPL21 gene was identified to be involved in Parkinson’s disease [34]. RPL22, a ribosomal protein, is induced by Cre recombinase, which primarily participates in the formation of the 60S subunit [35]. Recent research has demonstrated that RPL22 plays a significant role in normal T cell development to arrest new protein synthesis by regulating endoplasmic reticulum stress signal transduction [36] and deactivation or deficiency of the single allele of RPL22 has been found [37]. However, its association with neurodegenerative diseases such as AD has not been evaluated. RPL6 is a gene encoding a 60S subribosome, which plays an important role in oxidative phosphorylation and neuronal signal transduction in neurodegenerative diseases such as Parkinson’s disease [38]. RPL6 has been reported to be downregulated in Parkinson’s disease and induce 5${}^{\prime }$-adenosine monophosphate (AMP)-activated protein kinase signaling activation [39]. Similarly, down-regulation of RPL6 was observed in Parkinson’s disease peripheral blood mononuclear cell samples [40]. However, its function in AD is unclear. Knockdown of RPL32 significantly represses hepatocellular carcinoma proliferation [41]. RPL24 is another translation factor related to tumorigenesis. RPL24 reduction was found to be associated with over-expression of p53, suggesting that RPL24 in tumorigenesis is p53 dependent [42]. RPL35 encodes a protein of the 60s ribosomal subunit, which is part of the ribosomal protein L29 family. The L29 family proteins have been found to be usually related to the activation of oncogenes and tumor suppressor genes. However, little study of RPL35 function on neurodegenerative disease development has been found [43, 44]. RPL31, a 60S large ribosomal subunit protein, was demonstrated to be correlated with the cell cycle and regulate cancer development and progression via the p53 pathway [45]. Prior literature has shown that RPL35A was the first to be involved in Diamond-Blackfan anemia, which was found to act as a biomarker in tumor angiogenesis [46]. RPL35A has also been reported to be a significant down-regulator in an APP/PS1 mouse model of AD [47]. However, little is known about the relationship between RPL35A and AD development.

Three upregulated miRNAs (hsa-miR-3158-3p, hsa-miR-4435, and hsa-miR-3605-3p) were screened from AD samples in this ceRNA network. Among them, previous research has reported the upregulation of hsa-miR-4435 in AD samples [48]. This study was consistent with what is reported here. Additionally, the relationship of hsa-miR-3158-3p and hsa-miR-3605-3p with the pathogenesis of AD has not previously been reported. Similarly, five downregulated lncRNAs (TRHDE-AS1, SNHG10, OIP5-AS, LINC00926, and LINC00662) were found in AD samples. Three of them (SNHG10, OIP5-AS, and LINC00662) have been found to be correlated to the progression of AD by LncRNADisease v2.0. LINC00662 is a newly discovered long noncoding RNA. Previous study has shown that LINC00662 knockdown lowers the permeability of the blood brain barrier in the AD microenvironment due to enhanced tight junction-related protein expression levels [49]. Moreover, the ceRNA network constructed here predicted 26 differentially expressed circRNAs (see Table 4) that have been identified to be downregulated in AD. Among them, circRNA_0051620 was found to facilitate the occurrence and development of gastric cancer via acting as a miR-338-3p sponge and inhibiting ADAM17 [50]. However, there is little research on other circRNAs.

Notably, this study identified the clinical significance of several lncRNA-miRNA-mRNA and circRNA-miRNA-mRNA axes that included RPL35A and RPL35. LncRNA SNHG10 and OIP5-AS1 might be helpful for the development of AD via acting as a hsa-miR-3158-3p sponge and inhibiting the target mRNA RPL35A expression level. LncRNA LINC00662 might help the development of AD via acting as a sponge of hsa-miR-3605-3p and inhibiting the target mRNA RPL35A expression level. LncRNA OIP5-AS1 and LINC00662 might be helpful in the development of AD via acting as a sponge of hsa-miR-4435 and regulating target mRNA RPL35 expression level. Six differentially expressed circRNAs (hsa_circRNA_0090682, hsa_circRNA_0100723, hsa_circRNA_0076336, hsa_circRNA_0100174, hsa_circRNA_0001959, and hsa_circRNA_0104352) might regulate the development of AD via acting as a sponge of hsa-miR-3158-3p and inhibiting the target mRNA RPL35A expression. 11 circRNAs with differential expression (hsa_circRNA_0101760, hsa_circRNA_0002172, hsa_circRNA_0101758, hsa_circRNA_0103108, hsa_circRNA_0020990, hsa_circRNA_0051620, hsa_circRNA_0101472, hsa_circRNA_0002640, hsa_circRNA_0003057, hsa_circRNA_0102589, and hsa_circRNA_0104911) might regulate the development of AD via acting as a sponge of hsa-miR-4435 and inhibiting the expression level of target mRNA RPL35A and RPL35. Two circRNAs with differential expression (hsa_circRNA_0073593 and hsa_circRNA_0082335) might regulate the development of AD by sponging hsa-let-7d-3p and inhibiting the target mRNA RPL35A and RPL35 expression level. Ten differentially expressed circRNAs (hsa_circRNA_0090682, hsa_circRNA_0100946, hsa_circRNA_0104911, hsa_circRNA_0001276, hsa_circRNA_0100021, hsa_circRNA_0001399, hsa_circRNA_0026074, hsa_circRNA_0066331, hsa_circRNA_0076336, and hsa_circRNA_0100723) were found to be correlated to AD development via acting as a sponge of hsa-miR-330-5p and inhibiting the expression level of the target mRNA RPL35A. Hsa_circRNA_0025332 was found to be related to AD progression via acting as a sponge of hsa-miR-3605-3p and inhibiting the target mRNA RPL35A expression. Additionally, this analysis using the AlzData database showed that the target mRNA RPL35 and RPL35A were significantly associated with the clinicopathological features of AD. This study could provide new insights for understanding the interaction of ceRNA regulation and the pathogenesis of AD.

1381 differentially expressed overlapping mRNAs (711 downregulated and 670 upregulated) were identified between AD samples and control samples. 372 differentially expressed miRNAs (32 downregulated and 340 upregulated miRNAs) were obtained from AD samples compared with control samples. A total of 283 differentially expressed circRNAs (upregulation of 78 circRNAs and downregulation of 205 circRNAs) were screened. 275 upregulated lncRNAs and 209 downregulated lncRNAs were found between AD samples and control samples. Based on the ceRNA hypothesis, networks were constructed for lncRNA-miRNA-mRNA and circRNA-miRNA-mRNA, which include five lncRNAs (TRHDE-AS1, SNHG10, OIP5-AS, LINC00926 and LINC00662), 26 circRNAs, five miRNAs (hsa-miR-3158-3p, hsa-miR-4435, hsa-let-7d-3p, hsa-miR-330-5p and hsa-miR-3605-3p), and ten mRNAs (RPL11, RPL34, RPL21, RPL22, RPL6, RPL32, RPL24, RPL35, RPL31 and RPL35A). Based on the important role of RPL35 and RPL35A in AD reported in the AlzData database, the significance of the ceRNA network that includes RPL35A and RPL35 was discovered. These ceRNA networks are believed to be involved in the development of AD.

All the data supporting the results of this study are included in the manuscript and the Supplementary Documents.

LS, YW and HW designed the research study. LS, YZ and YW performed the research. LS and YZ conducted experiments. LS and YW analyzed the data. All authors contributed to editorial changes in the manuscript. All authors read and approved the final manuscript. All authors have participated sufficiently in the work and agreed to be accountable for all aspects of the work.

Not applicable.

Not applicable.

This work was supported by the project of department of health of Hebei Province (20200196), excellent youth fund for basic scientific research projects of Hebei North University (JYT2021005), youth fund for basic scientific research projects of Hebei North University (JYT2022008) and the Project of Hebei North University (H2022405030), China.

The authors declare no conflict of interest.