The circRNA-mediated ceRNA molecular regulatory network in fatigue-type type 2 diabete

Abstract

Fatigue is a common but poorly understood issue in type 2 diabetes (T2DM) that affects quality of life. Although ceRNA networks regulate disease progression, their role in T2DM-related fatigue (F-T2DM) is unclear. This study developed a circRNA-mediated ceRNA network to uncover the molecular interactions causing fatigue in F-T2DM. The study included healthy control group (Control, n = 21), F-T2DM group (n = 21), and non-fatigue type 2 diabetes patients (NF-T2DM, n = 21). By combining high-throughput sequencing to screen differentially expressed circRNAs (F-T2DM vs Control: 1144; F-T2DM vs NF-T2DM: 1303) and mRNAs (F-T2DM vs Control: 912; F-T2DM vs NF-T2DM: 1190), it was found that hsa_circ_0078539 and hsa_circ_0026239 were significantly upregulated in F-T2DM compared to both Control and NF-T2DM groups, and their host genes were involved in cytoskeleton remodeling. The GO/KEGG enrichment analysis combined with weighted gene co-expression network (WGCNA) of F-T2DM compared with Control indicated that the core pathways of F-T2DM focused on actin cytoskeleton dynamic regulation, AMPK signaling pathway, tricarboxylic acid cycle, and oxidative stress response. In the enrichment analysis of F-T2DM and NF-T2DM, cytoskeleton dynamics regulation, AMPK signaling pathway, and tricarboxylic acid cycle were further enriched, and the specific activation of reactive oxygen metabolism balance and AGE-RAGE pathway was also observed. Further, through multi-database prediction and experimental verification, a F-T2DM-specific ceRNA network was constructed, and key regulatory axes hsa_circ_0044623/hsa-mir-129-5p/MYLK3, hsa_circ_0002622/hsa-mir-200b-3p/RAB21, and hsa_circ_0078539/hsa-mir-4695-3p/SLC7A14 were screened out. The ceRNA regulatory network in human and animal samples was confirmed using RT-qPCR. These axes drive the pathological process by regulating myocardial contractility efficiency, glucose transport, mitochondrial energy metabolism, and insulin signaling pathway. This study clarified the molecular regulatory mode of patients with fatigue type 2 diabetes from the perspective of ceRNA network, providing a new direction for the research on diabetes classification and diagnosis.

Background

Diabetes is a rapidly growing epidemic. According to the report of the International Diabetes Federation (IDF) in 2021, approximately 537 million adults worldwide are affected by diabetes. It is projected that by 2045, this number will reach 783 million [1]. In the population of type 2 diabetes, poor glycemic control often leads to a vicious cycle of comorbid symptoms such as depression and fatigue. These factors not only undermine the patients' self-management efficacy but also result in a significant decline in their quality of life [2]. In the face of this serious public health issue, it is particularly urgent and important to conduct in-depth research on the pathogenesis and treatment strategies of T2DM. Recent studies have revealed that T2DM patients with significant fatigue symptoms exhibit unique transcriptional characteristics in peripheral blood mononuclear cells, involving abnormal molecular regulatory networks of key pathways such as mitochondrial energy metabolism and inflammatory response [3]. With the breakthroughs in the field of non-coding RNA research, circular RNAs have gradually become a new focus in the study of disease mechanisms due to their special covalent closed-loop structure and tissue-specific expression patterns [4, 5]. This type of RNA can simultaneously bind to multiple members of miRNA families through the ceRNA mechanism, using miRNA response elements (MREs) as molecular sponges. This enables it to effectively relieve the post-transcriptional inhibitory effect of miRNAs on target gene mRNAs [6].

The ceRNA mechanism plays a significant role in key physiological and pathological processes such as actin and tissue repair. Taking sperm development as an example, circLIMA1 has been confirmed to be a core molecule regulating nuclear actin depolymerization. The circLIMA1 specifically adsorbs regulatory factors such as miR-7a-5p and miR-214-3p through its miRNA binding domain (MREs) by the fluorescence reporter system verification. This ceRNA network can relieve the inhibitory effect of miRNA on genes related to sperm nuclear actin depolymerization, resulting in sperm nuclear actin remodeling disorder [7]. In the field of chronic wound healing in diabetes, high-throughput sequencing revealed that the host genes of differentially expressed circRNAs were significantly enriched in the actin cytoskeleton regulation pathway, such as hsa_circRNA_104175 and other molecules. Through the ceRNA network constructed by adsorbing miR-223-5p, miR-514a-3p, etc. miRNAs, they may affect the actin dynamics related to vascular smooth muscle contraction and epidermal cell migration [8].

Insulin homeostasis is crucial for maintaining the balance between glucose and insulin. As an example of circRNA-mediated ceRNA network regulating insulin resistance, the research by Lin team [9] revealed the molecular pathway by which circRNF111 antagonizes insulin resistance through the ceRNA mechanism. At the functional level, gene knockdown of circRNF111 in vitro significantly inhibited the glucose uptake ability of adipocytes; and the in vivo gene silencing experiment mediated by recombinant adenovirus further confirmed that the downregulation of circRNF111 could exacerbate the glucose intolerance and insulin sensitivity impairment induced by high-fat diet in obese mice. Mechanism analysis indicated that circRNF111 sponges miR-143-3p through adsorption, releasing the post-transcriptional inhibition of miR-143-3p on insulin-like growth factor 2 receptor (IGF2R), thereby activating the PI3K/AKT signaling pathway to improve insulin resistance. Yan et al. [10] discovered that hsa_circ_0071336 was abnormally highly expressed in peripheral blood of T2DM patients, and through dual luciferase reporter gene and protein interaction verification, it was revealed that this circRNA could block the targeting inhibition of miR-93-5p on glucose transporter 4 (GLUT4) by competitive binding, thereby enhancing the glucose transport efficiency under insulin stimulation.

This study innovatively integrates multi-omics analysis strategies to systematically analyze the molecular characteristics and pathological regulatory networks of F-T2DM. Through Illumina high-throughput sequencing combined with bioinformatics analysis, the first circRNA-miRNA-mRNA regulatory network specific to F-T2DM was constructed. The experimental design covered the control group, F-T2DM group, and NF-T2DM group. Through differential expression profile analysis, weighted gene co-expression network (WGCNA), cross-validation, and gene expression correlation verification, the core regulatory axes closely related to syndromes were screened out. Finally, the results were verified through in vivo experiments and PCR experiments. This study focuses on the association patterns of circRNA and mRNA, not only providing new evidence for revealing the molecular basis of fatigue-type diabetes, but also attempting to establish a theoretical bridge between fatigue-type diabetes and ceRNA regulatory network, laying the foundation for developing specific diabetes and targeted intervention strategies.

Materials and methods

Patient and public involvement

The patients with F-T2DM in this study were from Hepingli Hospital and National Medical Hall of Beijing University of Traditional Chinese Medicine from December 2018 to September 2021, and the healthy physical examination population was recruited from the physical examination center of Hepingli Hospital and Water Conservancy Hospital. The study was approved by the Ethics Committee of Beijing University of Traditional Chinese Medicine (BUCM), with the ethical batch number of (2017BZHYLL0105). All operations were carried out in accordance with the Helsinki Declaration, and all participants signed written informed consent forms. Both the researchers and the subjects should abide by the principle of confidentiality.

F-T2DM) subject selection criteria: (a) all patients met the diagnostic criteria of the American Diabetes Association criteria (2018);(b) fasting blood glucose ≥ 7.0 mmol/L, glycosylated hemoglobin (HbA1C) ≥ 6.5%;(c) patients with significant symptoms of fatigue shall be diagnosed by three doctors of the rank of associate professor or above. and (d) patients with complete background information. Exclusion criteria: (a) patients with severe liver, kidney, lung, or systemic disease;(B) patients with malignant tumors;(c) patients with neurological disease;(d) patients with vascular disease, and immune disorders. After being enrolled, all three groups had their fasting venous blood collected for subsequent experiments.

Library construction and sequencing

After collecting peripheral blood from the subjects, RNA extraction, library construction, sequencing on the instrument and data quality control were carried out. The data processing and cleaning methods adopted in this study were consistent with those in our previous research [11]. High-throughput sequencing was performed using the Illumina paired-end 150 bp sequencing strategy. After quality control, the clean reads were compared and analyzed using the software packages Tophat2 (http://tophat.cbcb.umd.edu), Hisat2 (http://ccb.jhu.edu/software/hisat2), and STAR (http://code.google.com/p/rna-star). The final libraries obtained all RNAs except ribosomal RNA and small RNAs (such as microRNA, siRNA, etc.), including lncRNA, mRNA, and circRNA.

Analysis strategy for differentially expressed circRNAs and mRNAs

This study focuses on the detection of circRNAs and conducts dual cross-validation by integrating two complementary algorithms, find_circ and CIRI, to enhance the specificity of detecting circular cleavage sites. To eliminate technical bias and achieve cross-sample comparability, the study designs differentiated standardization processes for circRNAs and mRNAs respectively:

1. circRNA quantification: Based on junction reads count (readCount ≥ 2), the TPM (Transcripts Per Million) model is used for cross-sample standardization. The calculation formula is:

1

2. mRNA quantification: The number of reads aligned to the exon regions of genes is counted using HTSeq. It is standardized by the FPKM (Fragments Per Kilobase per Million mapped reads) model. The calculation formula is as follows:

2

Ultimately, in this study, DESeq2 was employed to conduct the differential expression analysis of circRNAs and mRNAs based on the negative binomial generalized linear model. The screening criteria for differentially expressed genes in this study were |log2 Fold Change|> 1 (i.e., fold change > 2 times), and P < 0.05. The WGCNA analysis and KEGG/GO enrichment analysis in this study were carried out according to the research methods previously reported [11]. We used the R package clusterProfiler for functional enrichment analysis and applied WGCNA with R package [12] to identify modules linked to phenotypic traits.

Construction of circRNA-miRNA-mRNA Regulatory Network

This study utilized datasets of differentially expressed circRNAs and mRNAs to analyze the ceRNA regulatory network through bioinformatics. First, miRNA binding sites on differential circRNAs were predicted using the ENCORI database, and targeted mRNAs of these miRNAs were identified using a combination of TargetScan, miRDB, and miRWalk databases. Only mRNAs predicted by all three databases were retained. These mRNAs were then intersected with differentially expressed mRNAs from transcriptome sequencing to identify co-regulated candidate genes. Perl scripts were employed to integrate circRNA-miRNA and miRNA-mRNA interactions, and a visual regulatory network was subsequently generated using Cytoscape. In accordance with the ceRNA competitive network principle, which posits a positive correlation between circRNA and mRNA expression, molecular pairs exhibiting a negative correlation between circRNA expression and their predicted target mRNA were excluded. This approach facilitated the construction of upregulated and downregulated ceRNA networks.

Verification of the expression correlation of the core regulatory axis

Based on the ceRNA regulatory network theoretical framework, this study screened key regulatory axes through the analysis of circRNA-mRNA co-expression networks. Using the Spearman rank correlation analysis method (Spearman |r|> 0.3, P < 0.05), circRNA-mRNA pairs with significant positive co-expression characteristics were strictly selected. By integrating miRNA target regulation databases and functional analysis, the core circRNA-miRNA-mRNA regulatory axes with pathophysiological significance were ultimately established, providing a theoretical basis for subsequent functional mechanism research.

Experimental animals

In this study, 7-week-old SPF-grade C57BL/6 J mice (n = 16, balanced for sex) were used, with ethical approval from the Experimental Animal Ethics Committee of Beijing University of Chinese Medicine. The mice were randomly assigned to a control group (n = 8) receiving standard feed and an experimental group (F-T2DM, n = 8) fed a high-fat, high-sugar diet (H10045 formula: 45% fat, 20% protein, 35% carbohydrates). Following a five-week regimen of a high-fat diet, the experimental group was subjected to an intraperitoneal injection of streptozotocin (STZ, 120 mg/kg) to induce insulin resistance, thereby establishing a model of type 2 diabetes. Throughout this phase, the high-fat diet was sustained. Insulin resistance was assessed via fasting plasma glucose (FPG) measurements. The high-fat dietary regimen was continued until the eighth week. Concurrently, fatigue levels were evaluated using grip strength and endurance tests. Post-experiment, mice were sacrificed under sterile conditions, and pancreatic, liver, and skeletal muscle tissues were dissected, weighed, and stored at -80 °C after rapid freezing in liquid nitrogen.

Hematoxylin–Eosin (HE) staining

The pancreatic islets and skeletal muscle tissues were excised from each mouse and subjected to fixation in formalin for a duration of 24 h. Following fixation, the tissues underwent gradient dehydration and were subsequently embedded in paraffin for sectioning. The resulting sections were stained with hematoxylin and eosin and mounted on slides using resin. Images were then captured to assess the pathological alterations.

Real-time quantitative PCR, RT-qPCR

Extraction of total RNA

Total RNA was extracted from blood samples preserved in Trizol using the following method: Blood samples were moved from -80 °C to pre-chilled RNase-free microcentrifuge tubes. For animal tissues, 50 mg was placed on ice, mixed with 1 mL Trizol, homogenized, and incubated for 5 min. Chloroform was added at a 5:1 ratio, followed by 15 s of vortexing and 3 min at room temperature. Phase separation occurred via centrifugation at 12,000 × g for 15 min at 4 °C. The upper aqueous phase was transferred to a new tube, and 0.5 mL of pre-chilled isopropanol was added. After vortexing for 10 s, samples were incubated at room temperature for 10 min to precipitate RNA. The RNA pellet was centrifuged at 12,000 × g for 10 min at 4 °C, and the supernatant was discarded. It was then washed twice with 1 mL of 75% ethanol (in DEPC-treated water) by resuspending and centrifuging at 7,500 × g for 5 min at 4 °C to eliminate salts. Finally, the pellet was air-dried for 5–10 min at room temperature and dissolved in DEPC-treated water according to the expected concentration.

Reverse transcription

Preparation of reverse transcription reaction system for circRNA and mRNA (reagent kit number AG11706): 5 × Evo M-MLVRT Master Mix * 2 ul, Total RNA * 500 ng, and RNase free water up to 10 ul. Program settings for circRNA and mRNA reverse transcription: 37 ℃ for 15 min, 85 ℃ for 5 s. The reaction products should be temporarily stored at 4 ℃ or on ice.

Quantitative PCR

A 96-well optical reaction plate was used for the detection system, with each well containing a 20 µl reaction volume. The system comprised 10 µl of 2 × SYBR Green Pro Taq HS Premix III, 1 µM gene primers (2 µl), 100 ng cDNA (2 µl), and 6 µl RNase-free water. Primers were designed by Hunan Aikor Biotechnology Co., Ltd. using a divergent strategy for circRNAs. Primer sequences are in Table 1. After adding samples, the plate was sealed and amplified on a fluorescence quantitative PCR instrument with the following settings: Stage 1 at 95 °C for 30 s; Stage 2 (40 cycles) with 95 °C for 5 s and 60 °C for 30 s.

Table 1.

Primer Sequences

Gene	Primers sequence (Human)	Primer Sequences (Mouse)
β-actin	F: TGGCACCCAGCACAATGAA R: CTAAGTCATAGTCCGCCTAGAAGCA	F: CATCCGTAAAGACCTCTATGCCAAC R: ATGGAGCCACCGATCCACA
hsa_circ_ 0044623	F: GCCCAACAGGCAAAAATGAAG R: GCATGGCCTCGTCTGATCC	F: GCCCAACAGGCAAAAATGAAG R: GCATGGCCTCGTCTGATCC
MYLK3	F: CTATGACGCCTTCGAGAGCA R: CCACATCCAGCTCAGTCAGG	F: GGTCTTGTTCACGAGGCAGA R: CCTTTAGCTTCTCCCGAGGC
hsa_circ_ 0002622	F: CCTGGAATCACGAAGCACAG R: ACTATGGTCCTTAAATCTGGTGC	F: CCTGGAATCACGAAGCACAG R: ACTATGGTCCTTAAATCTGGTGC
RAB21	F: TTCAGCCAAACAGAACAAAGGA R: GGTCTGGGCTTGAGGTTCAT	F: GACAAGCACATCACCACCCT R: GCGTGGAATCTCTCCTGACC
hsa_circ_ 0078539	F: ACCGTGGGATGCTCAAAGAT R: CAGCTCTGCATCCATGGTG	F: ACCGTGGGATGCTCAAAGAT R: CAGCTCTGCATCCATGGTG
SLC7A14	F: TTTGCGGTCTGGTGCTTTGT R: CTCTGTGGCGTAGGAGAAACC	F: CACATCCATCCCTTACGCCA R: ACTTCGCAGCATAGAAGCCA

F: Forward Prime; R: Reverse Primer

Result processing

After collecting Ct values at the experiment's end, data analysis used dual normalization. First, circRNAs and mRNAs were normalized with β-actin as the reference gene [13, 14]. Then, further calibration was done using the healthy control group. The calculation of relative target gene expression strictly followed the classic Livak 2-ΔΔCt algorithm: ΔCt = Ct(target)—Ct(β-actin), ΔΔCt = ΔCt(experimental group)—ΔCt(control group), and the final relative expression level = 2-ΔΔCt. Melting curve analysis verified reaction specificity.

Statistics and analysis

The study used SPSS 20.0 for statistical analysis. Continuous variables were shown as mean ± standard deviation. Normality was checked with the Shapiro–Wilk test. Normally distributed data were compared using t-tests or one-way ANOVA with Bonferroni correction for multiple comparisons; non-normal data were analyzed with non-parametric tests. A significance level of 0.05 was used, with P < 0.05 considered significant.

Results

Clinical characteristics of subjects

This study strictly included the control group (n = 21), the fatigue-type T2DM group (F-T2DM, n = 21), and the non-fatigue-type T2DM group (NF-T2DM, n = 21), totaling 63 subjects. The baseline characteristics of all subjects were well-matched: the age range of all three groups was 40–75 years old, the gender distribution was balanced, and there was no significant difference in BMI among the groups. As shown in Table 2. Compared with the Control group, the fasting plasma glucose (FPG) and glycosylated hemoglobin (HbA1c) levels in the F-T2DM group and NF-T2DM group were significantly increased, which met the diagnostic criteria for T2DM (FPG ≥ 7.0 mmol/L, HbA1c ≥ 6.5%). There were no significant differences in the observed indicators between the F-T2DM group and the NF-T2DM group.

Table 2.

Basic Information and Biochemical Indicators of the Subjects (x ± s, n = 21)

	Control	F-T2DM	NF-T2DM
Gender(M/F)	16/5	12/9	13/8
Age(year)	49.90 ± 11.32	55.90 ± 11.17	55.33 ± 8.95
FPG(mmol/L)	5.49 ± 0.54	10.07 ± 3.24***	11.64 ± 5.38***
HbA1c(%)	5.74 ± 0.27	8.51 ± 2.03***	8.84 ± 2.00***
BMI(kg/m2)	24.63 ± 1.74	25.06 ± 3.63	26.56 ± 3.31

Note: FPG: fasting plasma glucose; HbA1c: glycosylated hemoglobin; BMI: body mass index; Compared with the Control group, * indicates P < 0.05, ** indicates P < 0.01, and *** indicates P < 0.001

F-T2DM vs control

Screening of differentially expressed genes

Figure 1A displays the research design groupings. Based on the strict screening threshold (|log2 Fold Change|> 1, P < 0.05), this study identified 1144 significantly differentially expressed circRNAs between F-T2DM and Control. Among them, 315 were up-regulated and 829 were down-regulated, with the down-regulation accounting for 72.5%. The differential expression levels were visualized through volcano plots (Fig. 1B). At the mRNA level, 912 differentially expressed mRNAs were identified in F-T2DM vs Control (223 up-regulated, 689 down-regulated, with the down-regulation accounting for 75.5%), as shown in Fig. 1C. By ranking the differentially expressed circRNAs based on their expression levels, we found that the highly expressed circRNAs with significant up-regulation included hsa_circ_0007205, hsa_circ_0007908, hsa_circ_0022723, hsa_circ_0051987, and hsa_circ_0000970, while those with significant down-regulation included hsa_circ_0007643, hsa_circ_0072481, hsa_circ_0007034, hsa_circ_0001458, and hsa_circ_0011938.

Fig. 1.

Transcriptome analysis of circRNAs between the F-T2DM group and the control group (A) Schematic diagram of the research cohort design. It shows the recruitment situation of the normal control group (n = 21) and the F-T2DM group (n = 21). (B) Volcano plot of circRNA differential expression. Note: The horizontal axis: log2(F-T2DM / Control expression fold change); the vertical axis: -log10(P). Blue marks significantly down-regulated genes, orange marks significantly up-regulated genes, gray indicates non-significant genes, and dotted lines indicate the significance boundary. (C) Volcano plot of mRNA differential expression. (D) Enrichment analysis of circRNA host genes in KEGG pathways. The height of the bar chart represents the significance of pathway enrichment, while the line chart indicates the number of genes enriched on this pathway. (F) GO functional annotation. BP refers to biological process; CC refers to cellular component; MF refers to molecular function. The size of the dots indicates the number of enriched genes, and the color depth corresponds to the significance

Functional regulatory network analysis of host genes of differentially expressed circRNAs

The results of KEGG pathway enrichment analysis are shown in Fig. 1D. Compared with the Control group, the differentially expressed circRNAs' host genes in F-T2DM were significantly enriched in metabolic core pathways such as actin cytoskeleton regulation, endocrine resistance, AMPK signaling pathway, thermogenesis, metabolic pathways, insulin resistance, lysine degradation, and ABC transporters, suggesting that circRNAs may be involved in the pathological process of fatigue by influencing cell dynamic remodeling and energy metabolism imbalance. Further GO functional annotation (Fig. 1E) indicates that compared with the Control group, the differentially expressed circRNAs' host genes in the F-T2DM group mainly participated in biological processes such as actin cytoskeleton dynamic regulation, oxidative stress response, cAMP-dependent protein phosphorylation, etc. The encoded proteins are located in the inner mitochondrial membrane and actin cytoskeleton domains, and at the molecular function level, they are mainly characterized by ATP hydrolase activity, GTPase binding, protein kinase regulation, and cysteine-type endopeptidase activity.

Weighted gene co-expression network analysis (WGCNA) of mRNA expression matrix

Based on weighted gene co-expression network analysis, this study successfully constructed the gene co-expression network between F-T2DM and Control groups, and the clustering situation of genes is shown in Fig. 2A. Through the dynamic shuffling tree algorithm, 7 functional modules were screened out. Among them, the yellow-green module showed a significant positive correlation with the clinical features of F-T2DM (r = 0.45, P < 0.001, Fig. 2B). The KEGG pathway enrichment analysis of the core genes (2155 genes) of this module indicated that it mainly involved the tricarboxylic acid cycle, regulation of actin cytoskeleton, thermogenesis regulation, and peroxisome signaling pathways, suggesting that it may exacerbate the pathogenesis of the fatigue type diabetes by regulating energy metabolism homeostasis and cytoskeleton (Fig. 2C).

Fig. 2.

Weighted gene co-expression network analysis of mRNA (A) mRNA co-expression network module gene clustering tree. (B) Module and phenotype correlation heatmap. Each square's upper left corner represents its Pearson coefficient, and the lower right corner represents its statistical significance. (C) Core module KEGG pathway enrichment. The height of the column chart indicates the enrichment significance, and the line chart shows the number of pathway genes

F-T2DM vs NF-T2DM

Based on the transcriptome analysis of F-T2DM patients and healthy individuals in the previous stage, in order to precisely analyze the molecular characteristics of the fatigue feature, this study further established a comparative research model between F-T2DM and NF-T2DM. This focused research design can effectively eliminate the interference of diabetes-related basic metabolic disorders on the research results.

Screening of differentially expressed genes

Comparing circRNA and mRNA profiles between F-T2DM and NF-T2DM groups, 1,303 circRNAs were differentially expressed, with 183 upregulated and 1,120 downregulated. For mRNA, 1,190 genes were differentially expressed in the F-T2DM group, with 940 upregulated and 250 downregulated. See Fig. 3A-C for visualization.Based on the expression levels, we ranked the differentially expressed circRNAs. We found that the circRNAs with significantly elevated expression levels included hsa_circ_0000876, hsa_circ_0064306, hsa_circ_0078539, hsa_circ_0026239 and hsa_circ_0038891, while those with significantly decreased expression levels included hsa_circ_0007643, hsa_circ_0029920, hsa_circ_0060618, hsa_circ_0019696 and hsa_circ_0068724.

Fig. 3.

Differentially Expressed Genes between F-T2DM and NF-T2DM (A) Research cohort design schematic diagram. It shows the recruitment situation of the F-T2DM group (n = 21) and the NF-T2DM group (n = 21). (B) Volcano plot of circRNA differential expression. Blue markers indicate significantly down-regulated genes, orange markers indicate significantly up-regulated genes, green indicates non-significant genes, and dotted lines indicate the significance boundary. (C) mRNA differential expression volcano plot. (D) Enrichment analysis of circRNA host genes in KEGG pathways. The rich factor is the ratio of differentially expressed genes in a pathway to the total genes in the pathway's annotation database. The color gradient indicates statistical significance of pathway enrichment (using -log10 transformed P values), with warmer colors showing higher significance. Dot size correlates with the number of differential genes in the pathway. (E) GO functional annotation of circRNA host genes. Gene Percent (%) represents the proportion of differentially expressed genes in the pathway, with color indicating significance and column height reflecting the absolute number of differential genes

A comparative analysis of differentially expressed circRNAs between the F-T2DM and control groups, as well as the NF-T2DM group, revealed specific circRNAs that may serve as molecular markers for fatigue-type diabetes. Notably, the expression levels of two circRNAs, hsa_circ_0078539 (host gene:EZR) and hsa_circ_0026239 (LARP4), were markedly elevated in individuals with F-T2DM compared to both the Control and NF-T2DM groups. EZR is a protein implicated in cytoskeletal remodeling, while LARP4 acts as a mechanical binding partner of filamin A, an actin cross-linking protein [15, 16], suggesting a potential role in cytoskeletal remodeling in F-T2DM. Furthermore, hsa_circ_0007643 (YWHAE) was consistently downregulated in both groups.

KEGG enrichment analysis of host genes of differentially expressed circRNAs

Based on the KEGG pathway enrichment analysis (Fig. 3D), it was found that compared with NF-T2DM, the host genes of the differentially expressed circRNAs in F-T2DM showed significant enrichment characteristics in key metabolic regulatory pathways. The core pathway clusters can be summarized as follows: 1) Energy metabolism hubs (AMPK signaling pathway, citric acid cycle); 2) Cytoskeleton dynamics (actin regulatory pathway, actin molecules); 3) Oxidative stress response system (reactive oxygen metabolism, AGE-RAGE signaling pathway); 4) Protein homeostasis regulatory mechanism (ubiquitin–proteasome system, ATP-dependent chromatin remodeling); 5) Substance transport system (ABC transporters). This suggests that the pathological essence of fatigue-type diabetes may involve a vicious cycle triggered by energy metabolism disorders, imbalance of the oxidative-antioxidative system, and accumulation of advanced glycation end products.

GO functional annotation of host genes of differentially expressed circRNAs

Further GO functional annotation (Fig. 3E) indicates that, compared with NF-T2DM, the differentially expressed circRNAs of F-T2DM show systematic regulatory characteristics at the levels of biological process, subcellular localization and molecular function. At the biological process level, the host genes are mainly involved in cell dynamics regulation (regulation of actin cytoskeleton organization), metabolic homeostasis maintenance (involving citric acid cycle, gluconeogenesis regulation), stress response system (oxidative stress response, cAMP signal-mediated cellular adaptation), and mitochondrial system management (involving membrane organization and matrix homeostasis). Their encoded proteins are located in domains such as actin cytoskeleton, trans-Golgi network, cell leading edge, spliceosome complex and mitochondrial matrix, and molecular function analysis reveals that they are closely related to energy-metabolism-related enzyme activities (ATP-dependent activities, protein kinase catalytic activities), redox homeostasis regulators (Fe-S cluster binding activity, cysteine-type peptidase activity) and epigenetic regulatory elements (histone binding domains, chromatin remodeling activities). GO analysis shows that these genes cause energy metabolism disorders and cell morphological abnormalities through influencing cytoskeleton, mitochondrial and oxidative stress dysregulation.

F-T2DM vs NF-T2DM

Acquisition of candidate circRNAs in the circRNA-miRNA-mRNA network

After transcriptome sequencing, a total of 1144 significantly differentially expressed circRNAs were obtained between the F-T2DM group and the Control group. Among them, 544 circRNAs had the ability to bind to miRNAs, so these 544 were temporarily designated as candidate circRNAs.

Acquisition of candidate miRNAs in the circRNA-miRNA-mRNA network

Based on the circRNA dataset with differences, 544 candidate circRNAs were predicted to have miRNA binding sites through the ENCORI database (http://starbase.sysu.edu.cn/index.php). The results showed that a total of 2654 different miRNA binding sites were involved, and these 2654 sites were used to locate the candidate miRNAs.

Acquisition of candidate mRNAs in the circRNA-miRNA-mRNA network

Using a multi-database integration strategy (TargetScan, miRDB, miRWalk), we predicted mRNA targets for 2654 candidate miRNAs. Of these, 2231 miRNAs were found to have mRNA targets, resulting in 6997 unique mRNA targets. Transcriptome sequencing revealed 912 significantly differentially expressed mRNAs between the F-T2DM and Control groups. By intersecting these with the predicted targets, 71 candidate miRNAs were identified (Fig. 4).

Fig. 4.

Venn diagram of mRNA prediction and sequencing for F-T2DM vs Control

Construction of the up-regulated circRNA-miRNA-mRNA network

Among the 71 candidate mRNAs, 17 were up-regulated. These 17 mRNAs were selected as the mRNA for constructing the ceRNA network. Based on these mRNAs, the Perl script was used to infer the data of the ternary relationship of circRNA-miRNA-mRNA. The 17 mRNAs corresponded to 67 miRNAs, and the 67 miRNAs corresponded to 454 circRNAs. The visualization regulatory network was constructed using Cytoscape. According to the principle of the competitive ceRNA network, the expression levels of circRNAs and mRNAs were positively correlated. In the upregulated network, we preserved the circRNA-miRNA-mRNA relationship in accordance with the competing endogenous RNA (ceRNA) hypothesis, whereby both circRNA and mRNA exhibited upregulation. Conversely, in the downregulated network, we preserved the relationship in which both circRNA and mRNA were downregulated. Finally, the up-regulated circRNA-miRNA-mRNA network was successfully constructed. The network included 262 nodes (178 circRNAs/67 miRNAs/17 mRNAs) and 1000 interaction edges (Fig. 5).

Fig. 5.

Upregulated circRNA-miRNA-mRNA Regulatory Network in F-T2DM vs Control Note: Square-shaped nodes represent circRNAs, diamond-shaped nodes represent miRNAs, and circular nodes represent mRNAs; red and pink indicate upregulation, while blue indicates unknown

Construction of the down-regulated circRNA-miRNA-mRNA network

Among the 71 candidate mRNAs, 54 were down-regulated. These 54 mRNAs were selected as the mRNA for constructing the ceRNA network. Based on these mRNAs, the Perl script was used to infer the data of the ternary relationship between circRNAs, miRNAs and mRNAs. The 54 mRNAs corresponded to 200 miRNAs and 200 miRNAs corresponded to 530 circRNAs. The visualization regulatory network was constructed using Cytoscape. According to the principle of the competitive ceRNA network, the expression levels of circRNAs and mRNAs were positively correlated. We preserved the relationship wherein both circRNA and mRNA exhibited downregulation. Finally, the up-regulated circRNA-miRNA-mRNA network was successfully constructed. The network included 595 nodes (341 circRNAs/200 miRNAs/54 mRNAs) and 4201 interaction edges (Fig. 6).

Fig. 6.

Downregulated circRNA-miRNA-mRNA Regulatory Network in F-T2DM vs Control Note: Square-shaped nodes represent circRNAs, rhombic-shaped nodes represent miRNAs, and circular-shaped nodes represent mRNAs; green color indicates down-regulation, and blue color represents unknown

Verification of expression

This study analyzes differential gene expression profiles between F-T2DM and NF-T2DM patients. Using cross-validation, we confirmed the specificities of the up-regulated and down-regulated ceRNA networks. The results demonstrated a significant differential expression of the MYLK3 gene (FC = 0.455, P = 0.016) and RAB21 gene (FC = 0.639, P = 0.038). Notably, significant differences in the expression levels of MYLK3 and RAB21 were observed in both the comparisons between F-T2DM and Control, as well as F-T2DM and NF-T2DM. These findings suggest that MYLK3 and RAB21 may serve as critical regulatory nodes in the pathophysiology of F-T2DM.

To validate the core signaling axis in the ceRNA network, this study examined the gene expression correlation between circRNAs and mRNAs, with results in Table 3. For MYLK3, three candidate circRNAs were identified: hsa_circ_0004855, hsa_circ_0003706, and hsa_circ_0044623, based on criteria (Spearman |r|> 0.3, P < 0.05). Cross-validation of F-T2DM and NF-T2DM patients' gene profiles showed that only hsa_circ_0044623 was specific to diabetes (FC = 0.499, P = 0.013). According to the down-regulated ceRNA network constructed in this study, a down-regulated ceRNA axis was finally established: hsa_circ_0044623—hsa-mir-129-5p—MYLK3 (Table 4).

Table 3.

Verification of the Correlation between circRNA and mRNA Gene Expression

circRNA	mRNA	r_spearman	pval_spearman
hsa_circ_0004855	MYLK3	0.503	0.001
hsa_circ_0003706	MYLK3	0.375	0.014
hsa_circ_0044623	MYLK3	0.335	0.030
hsa_circ_0001504	MYLK3	-0.329	0.033
hsa_circ_0003496	MYLK3	-0.321	0.038
hsa_circ_0058443	RAB21	0.571	0.000
hsa_circ_0002622	RAB21	0.397	0.009
hsa_circ_0078539	SLC7A14	0.325	0.036
hsa_circ_0078539	KLHL30	0.395	0.010
hsa_circ_0026239	TM4SF1	0.375	0.014
hsa_circ_0026239	SLC7A14	0.142	0.370
hsa_circ_0007643	CKS2	0.131	0.407

Note: r_spearman: Spearman's rank correlation coefficient, reflecting the strength of association between the expression levels of circRNA and mRNA

Table 4.

Establishment Status of the Weakness-type Diabetes Mice Model (Mean ± SD)

Group	FPG(mmol/L)	Fasting weight(g)	Food intake (g)	Water intake (g)	Urine volume (mL)
Control	5.57 ± 0.64	26.52 ± 3.65	2.15 ± 0.16	3.19 ± 0.19	0.79 ± 0.06
F-T2DM	12.89 ± 1.31***	25.29 ± 3.48	2.81 ± 0.34***	5.70 ± 1.47***	2.11 ± 0.46***

Compared with the Control group, * P < 0.05, ** P < 0.01, *** P < 0.001

For the gene RAB21, after screening and cross-validation, the final result shows that hsa_circ_0002622 has specificity for diabetes (FC = 0.291, P = 0.000) and conforms to the ceRNA principle. Based on the ceRNA network constructed in this study, a ceRNA axis that is down-regulated was finally established: hsa_circ_0002622—hsa-mir-200b-3p—RAB21.

We conducted an in-depth analysis of circRNAs (hsa_circ_0078539, hsa_circ_0026239, and hsa_circ_0007643) that exhibited significant expression differences in both F-T2DM VS Control and F-T2DM VS NF-T2DM, and ranked highly in terms of FC values. Finally, we confirmed three up-regulated ceRNA axes: (1) hsa_circ_0078539—hsa-mir-4695-3p—SLC7A14; (2) hsa_circ_0078539—hsa-mir-7847-3p—KLHL30; (3) hsa_circ_0026239—hsa-mir-4310—TM4SF1.

In vivo experiments

Establishment of the model of mice with fatigue-type diabetes

We assessed mice in the Control and F-T2DM groups for fatigue symptoms. By the 4th month, the F-T2DM group showed significantly higher fasting blood glucose and increased 24-h food, water intake, and urine output, indicating T2DM symptoms. Although fasting body weight showed a decreasing trend, it wasn't statistically significant. The F-T2DM group also exhibited reduced grip strength and endurance, confirming the fatigue-type diabetes model's suitability for further PCR studies (Table 5).

Table 5.

Establishment Status of the Weakness-type Diabetes Mice Model (Mean ± SD) (Continued)

Group	Pull force (N)	Endurance (s)
Control	1.93 ± 0.26	300.00 ± 0.00
F-T2DM	1.64 ± 0.13***	166.82 ± 79.52***

Compared with the Control group, * P < 0.05, ** P < 0.01, *** P < 0.001

Observation Results of the General Conditions of Laboratory Animals

In the experiment, mice were systematically observed for behavior and physiology. Initially, all groups were healthy, displaying energy, responsiveness, shiny fur, and normal activity. Post-intervention, the F-T2DM group showed fatigue-related changes, including reduced activity, lethargy, slow movements, preference for quiet, increased body size, greasy fur, clustering, and hair loss. The Control group remained normal.

Experimental verification of the ceRNA regulatory axis

After RT-qPCR verification, in the F-T2DM group of human peripheral blood samples, the expression levels of MYLK3 and RAB21 genes, as well as the circRNAs that indirectly regulate these two genes—hsa_circ_0044623 and hsa_circ_0002622, were all coordinately downregulated (Fig. 7A, B). The expression of SLC7A14 gene and its indirectly regulated hsa_circ_0078539 also showed a synchronous significant upregulation trend (Fig. 7C). It is worth noting that the gene expression trends in mouse skeletal muscle tissue samples were highly consistent with those in human peripheral blood samples (Fig. 7D-F). Moreover, the RT-qPCR identification results were consistent with the sequencing results, which fully demonstrated that these circRNAs and mRNAs could form a complex regulatory network in the form of circRNA-miRNA-mRNA network, participating in the gene expression regulation process within the organism.

Fig. 7.

RT-qPCR Results Note: Figures A-C illustrate the RT-qPCR detection results of peripheral blood samples from humans. Figures D-F represent the RT-qPCR outcomes of skeletal muscle tissue samples from mice. * indicates P < 0.05, ** indicates P < 0.01, and *** indicates P < 0.001

HE staining of mouse pancreas and skeletal muscle

The histological analysis of pancreatic tissue in experimental animals, as illustrated in Fig. 8A, demonstrated that the F-T2DM group exhibited significant deviations from the Control group. Specifically, the F-T2DM group showed irregular islet morphology, pronounced lipid vacuolation, infiltration of inflammatory cells, and degeneration or necrosis within the pancreatic islets. Furthermore, the morphological assessment of skeletal muscle, depicted in Fig. 8B, revealed through HE staining that the F-T2DM group experienced muscle fiber bundle atrophy, reduced volume, hyperchromatic nuclei, and a disordered arrangement of muscle fibers.

Fig. 8.

Representative HE-stained tissue samples. Note: Representative HE-stained tissue samples from Control and F-T2DM groups. (A)Pancreas; (B) Skeletal muscle. Scale bars = 200 μm

Discussion

According to the complementary pairing rule of bases, miRNA inhibits or promotes the translation of mRNA by binding to the non-translated regions of target genes. RNA can influence the inhibitory effect of miRNA on target genes through competing for the same miRNA to bind together. Such RNA is called ceRNA [6]. Studies have shown that many circRNAs contain a large number of miRNA response elements and can act as efficient ceRNA by binding to miRNA and exerting miRNA sponge function, adsorbing miRNA and effectively inhibiting the binding of miRNA to the non-translated regions of target genes, thereby achieving the regulatory effect on target genes [17]. The ceRNA network composed of circRNA-miRNA-mRNA has attracted widespread attention in disease research. In recent years, more and more studies have found that the circRNA-miRNA-mRNA regulatory network plays an important role in the pathogenesis of T2DM.

Through omics analysis in this study, it was found that F-T2DM patients exhibited significant transcriptional heterogeneity. Compared with the Control group, both circRNAs and mRNAs in the F-T2DM group were mainly downregulated. This expression pattern suggests that circRNAs may positively regulate mRNA expression through the ceRNA mechanism, thereby influencing the progression of fatigue-type diabetes. However, compared with the NF-T2DM group, circRNAs in the F-T2DM group were mainly downregulated, while mRNAs showed an upward trend. This contradictory phenomenon implies that relying solely on the ceRNA mechanism may not be sufficient to explain the symptom of fatigue, and its pathological process may involve more complex regulatory levels—such as epigenetic modifications or abnormal activation of transcription factors. Future studies need to further analyze through multi-omics integration.

It is worth noting that through cross-group comparative analysis strategy screening, it was found that in the cross-analysis of differentially expressed circRNAs between the F-T2DM group and the Control group as well as the NF-T2DM group, there were circRNA molecules with significant common differential expression trends, suggesting that these molecules might serve as specific molecular markers for F-T2DM and play a key role in disease classification. There were two circRNAs that were consistently downregulated in the comparison between the two groups: hsa_circ_0007643 (host gene: YWHAE). There were also two circRNAs that were significantly upregulated in both groups: hsa_circ_0078539 (host gene: EZR) and hsa_circ_0026239 (host gene: LARP4). EZR is a protein related to cytoskeleton remodeling, and LARP4 is a mechanical binding partner of filamin A, a myosin cross-linking protein. Both are closely related to cell proliferation and migration, and participate in biological processes such as cell adhesion, amyloid cell migration, and cell junction assembly [15, 16, 18–21]. This suggests that they may affect the development of F-T2DM by regulating cytoskeleton remodeling and cell migration.

Compared with the NF-T2DM group, the KEGG enrichment analysis of the differentially expressed circRNAs in the F-T2DM group showed that they were significantly enriched in actin cytoskeleton regulation, AMPK signaling pathway, citric acid cycle (TCA cycle), AGE-RAGE signaling pathway, and oxidative stress-related pathways. The dynamic remodeling disorder of actin cytoskeleton not only affects the insulin secretion process of pancreatic β-cells [22], but also has a close relationship with fatigue and weakness [23], directly leading to typical symptoms of fatigue and weakness. In addition, the excessive aggregation of actin stress fibers in endothelial cells can disrupt the intercellular connections mediated by cadherin in vascular endothelial cells [24], inducing increased microvascular permeability and hemodynamic disorders. As the core regulatory hub of energy metabolism [25, 26], the inhibition of the AMPK pathway may lead to mitochondrial dysfunction [27, 28], resulting in insufficient ATP production and directly causing symptoms such as fatigue. It is worth noting that the activation of the AGE-RAGE signaling pathway suggests that the accumulation of advanced glycation end products (AGEs) may exacerbate oxidative stress damage and pathological damage of F-T2DM through endoplasmic reticulum-mediated inflammatory responses [29] and the excessive accumulation of ROS in mitochondria [30].

The GO functional annotation further reveals that compared with the NF-T2DM group, the differentially expressed circRNAs in the F-T2DM group coordinately regulate energy metabolism and cell morphology homeostasis through cytoskeleton, mitochondria and oxidative stress. Among them, the imbalance of phosphorylation of the actin cytoskeleton may lead to the obstruction of the directional transport of β-cell secretory granules, while mitochondria will induce ATP synthesis disorders. In addition, iron-sulfur clusters (Fe-S clusters) are key cofactors involved in electron transfer, enzyme activity regulation and redox balance in the biological system. Abnormal assembly of iron-sulfur cluster proteins leads to the insertion disorder of Fe/S clusters in complex I, resulting in mitochondrial energy metabolism disorders and oxidative stress, and subsequently affecting the insulin secretion function of pancreatic β-cells [31].

This study, by integrating high-throughput sequencing and bioinformatics analysis, for the first time constructed the unique ceRNA regulatory network of fatigue type 2 diabetes mellitus (F-T2DM), and revealed the molecular mechanism by which circRNAs regulate downstream mRNA expression through competitive adsorption of miRNAs. The study found that circRNAs that were significantly downregulated in F-T2DM patients, such as hsa_circ_0044623 and hsa_circ_0002622, respectively, relieved the inhibitory effects on target genes MYLK3 and RAB21 by binding to miRNAs (hsa-mir-129-5p and hsa-mir-200b-3p). In F-T2DM patients, the significantly upregulated hsa_circ_0078539 relieved the inhibitory effect on target gene SLC7A14 by binding to hsa-mir-4695-3p.

Among them, MYLK3 protein (myosin light chain kinase 3) is a key enzyme regulating myocardial contraction. Its down-regulation leads to the weakening of phosphorylation of myosin light chain (MLC), resulting in disordered sarcomere structure (such as misalignment of Z-lines and loose arrangement of myofilaments), directly reducing myocardial contraction efficiency [32–34]. This is manifested as decreased cardiac output and compensatory increase in heart rate, which leads to insufficient blood perfusion in various tissues and organs throughout the body, reduced supply of oxygen and nutrients, and insufficient energy supply to muscles and other tissues, thereby causing fatigue. At the same time, it is closely related to oxidative stress. Abnormal MYLK3 may activate the NF-κB pathway, promoting the release of inflammatory factors and myocardial oxidative damage [35], further aggravating the disorder of glucose metabolism.

The RAB21 protein can regulate glucose transport and mitochondrial energy metabolism. At the level of glucose transport, RAB21 maintains the cellular surface GLUT1 level by regulating the recycling of glucose transporter SLC2A1/GLUT1 mediated by the retromer complex. Its deficiency leads to the incorrect sorting of GLUT1 to lysosomes for degradation, resulting in reduced glucose uptake. This process directly affects cellular energy homeostasis, especially in insulin-sensitive tissues (such as muscle and liver), which may exacerbate insulin resistance and accelerate the progression of metabolic disorders in diabetes [36]. Additionally, in the regulation network of mitochondrial energy metabolism, RAB21 interacts with mitochondrial outer membrane protein MFN2 to direct internalized epidermal growth factor receptor (EGFR) to mitochondria. Through the mitochondrial-localized phosphatase PTPRJ, it dephosphorylates EGFR and inhibits its downstream proliferative signals. This process not only regulates growth signals but also may interfere with ATP generation by affecting mitochondrial dynamics (such as fusion/fission balance), leading to an insufficient cellular energy supply and directly aggravating fatigue symptoms [37]. Further studies have found that RAB21 deficiency can enhance lysosomal activity. Although it can supplement energy through accelerating substrate degradation in the short term, long-term excessive lysosomal activity may cause metabolic imbalance (such as abnormal lipid accumulation) and aggravate mitochondrial oxidative stress [36]. In summary, RAB21 jointly exacerbates the pathological process of fatigue-type diabetes through dual pathways of regulating glucose transport disorder and mitochondrial energy metabolism imbalance.

SLC7A14 protein, as a lysosomal γ-aminobutyric acid (GABA) transporter located on the lysosomal membrane, is upregulated under conditions of high-fat diet or insulin resistance. Its overexpression in the liver leads to an increase in GABA levels within lysosomes. GABA inhibits the activity of mTOR complex 2 and reduces the phosphorylation of AKT (protein kinase B) at Ser473, thereby weakening insulin signaling and ultimately causing decreased insulin sensitivity and glucose metabolism disorders in the liver. Experiments have confirmed that overexpression of SLC7A14 aggravates insulin resistance, while gene knockout or loss-of-function mutants significantly improve insulin sensitivity and glucose tolerance [38].

Through the F-T2DM mouse model induced by high-fat diet combined with STZ, this study successfully simulated the "three excesses and one deficiency" symptoms and fatigue characteristics, such as decreased grip strength endurance. RT-qPCR results showed that the expression trends of MYLK3, RAB21, SLC7A14, and the regulatory circRNA were consistent in the F-T2DM group, and were highly consistent with the results of human samples.

These research results indicate that the ceRNA network may be involved in the pathogenesis of fatigue-type diabetes by regulating myocardial contraction, abnormal glucose transport, mitochondrial energy supply disorder, actin cytoskeleton dynamic remodeling, and oxidative stress, among other processes. The molecular regulatory axis composed of key genes MYLK3 (associated with cardiovascular-related fatigue), RAB21 (associated with energy metabolism disorder), SLC7A14 (associated with insulin resistance), and their related circular RNAs may play a core regulatory role in the pathological process of F-T2DM. Furthermore, detecting the expression levels of MYLK3, RAB21, SLC7A14, and their related circular RNAs in blood or tissue samples can help distinguish between ordinary T2DM and the F-T2DM subtype; at the same time, developing agonists or inhibitors targeting these genes or their circular RNAs is expected to achieve intervention in the disease progression.

Limitations of this study

While this study has advanced our understanding of the link between ceRNA networks and F-T2DM, it has limitations. Firstly, this study predominantly utilizes bioinformatics prediction and expression correlation analysis to elucidate regulatory mechanisms, with a particular focus on microRNAs (miRNAs). These miRNAs are primarily predicted through database analyses and lack supporting clinical data. Consequently, the molecular interactions within the circRNA-miRNA-mRNA axis necessitate experimental validation. Additionally, the small sample size and single-center data source may reduce statistical power and introduce selection bias, limiting the findings' generalizability. Future research should involve multi-center collaborations and larger sample sizes to validate these findings and improve the reliability and applicability of the results.

Funding

This project was supported by the Major Project Cultivation Program of Xinjiang Medical University (XYD2024ZX06), the Ministry of Education's Key Laboratory open project for Research of High Incidence Diseases in Xinjiang., China (2024A05), the Key Research and Development Project of Xinjiang Uygur Autonomous Region, China (2024B03034-2), and Beijing Hospitals Authority Clinical medicine Development of special funding support (ZLRK202308).

Data availability

The datasets used and analyzed during the current study are available from GEO (GSE277813).

Declarations

Ethics approval and consent to participate

The study was approved by the Ethics Committee of Beijing University of Traditional Chinese Medicine (BUCM), with the ethical batch number of (2017BZHYLL0105). All procedures were performed in accordance with the Declaration of Helsinki, and all participants received written informed consent.

Consent for publication

Both written and verbal consent was obtained from the subjects of the study.

Competing interests

The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.