Multi-omics analysis and validation of autophagy-related diagnostic biomarker in osteoarthritis

Abstract

Background

Autophagy plays an essential regulatory role in the development of osteoarthritis (OA), its specific regulatory mechanism remains unclear. This work aims to study the critical genes related to autophagy in OA and explore their potential value.

Methods

Datasets were downloaded from the publicly available GEO database, and differentially expressed autophagy-related genes (ARGs) were obtained. LASSO regression analysis, SVM-RFE algorithm, and RF algorithm were used to screen signature genes. The diagnostic accuracy of autophagy-related signature genes (ARSGs) in OA was determined through receiver operating characteristic (ROC) univariate analysis. The accuracy of the results was validated with an external dataset. Signature genes were analyzed for immune-related functions. Single-gene cluster enrichment analysis (GSEA) and single-gene set variation analysis (GSVA) were employed to demonstrate the potential biological functions of signature genes. Competing endogenous RNA networks, including miRNAs and lncRNAs, and drug regulatory networks were constructed to search for potential therapeutic targets. The expression levels of key ARSGs were verified through qRT-PCR, Western blot, immunohistochemistry (IHC) analyses, and animal model construction.

Results

A total of 12 up-regulated genes and 37 down-regulated genes were identified and found to be associated with autophagy regulation and inflammatory pathways. Screening identified three candidate genes, namely, CAPN2, ITGA3, and ERBB2. The GSE48556 dataset, as well as qRT-PCR, Western blot, IHC analyses, and animal model further demonstrated that ERBB2 was a key signature gene associated with autophagy and positively correlated with OA severity. The ROC curve showed that ERBB2 had a high diagnostic value. Immunofunctional correlation analysis revealed that the infiltration of immune cells, such as macrophages, neutrophils, and NK cells, was highly abundant in the group with low ERBB2 expression. GSEA and GSVA also revealed that low ERBB2 expression was associated with the activation of cellular immune states. Meanwhile, 30 potential therapeutic drugs for OA and several lncRNAs and miRNAs that regulate ERBB2 were identified.

Conclusions

ERBB2, an autophagy signature gene that is closely related to autophagy regulation and immune response in OA cartilage, holds promise as a bioindicator for the early diagnosis and targeted therapy of OA.

Introduction

Osteoarthritis (OA) as a chronic degenerative joint disease seriously affects the quality of life of older adults, thus attracting extensive worldwide attention. OA is characterized by cartilage degeneration [1], subchondral bone remodeling [2], meniscal degeneration [3], inflammation and fibrosis of both infrapatellar fat pad and synovial membrane [4,5]. Substantial loss of cartilage extracellular matrix causes cartilage destruction and fibrous tissue growth, which are the main pathological hallmarks of OA [6]. Numerous treatment options for OA exist. They include nonsurgical and surgical treatments, which aim to relieve pain and improve joint mobility. Nonsurgical treatments are primarily medications, which include acetaminophen, nonsteroidal anti-inflammatory drugs, and opioids [7]. Tissue engineering techniques have also been applied to the treatment of OA [8]. When OA progresses to the end stage, the patient’s clinical prognosis can only be improved via joint replacement. However, the above treatments have substantial limitations and cannot change the pathological course of OA; such a change necessitates restoring articular cartilage regeneration or changing the degenerative process [9,10]. Therefore, studying the pathomechanisms involved in the development of OA and searching for key targets in early OA prevention and treatment are urgently needed.

More and more evidence indicates that the weakening of autophagy in the context of inflammation is closely related to the development of OA [11]. As a protein degradation system relying on lysosomes, autophagy maintains stable intracellular energy metabolism and normal cellular functions by eliminating abnormal proteins, damaged organelles, or macromolecules [12,13]. Autophagy is negatively correlated with age and OA severity [14]. The weakening of autophagy in OA leads to the aggregation of different molecular proteins, thereby causing cellular functional defects, denaturation, and apoptosis. Deretic et al. pointed out that the disturbance of autophagy balance induces the occurrence of inflammatory diseases, including OA [15]. The key autophagy regulatory factors play an crucial role in the regulation of autophagy. Relevant studies have found that autophagy-related regulatory factors, such as Beclin-1, ATG5, and LC3-II, are associated with the occurrence and development of OA and pointed out the possibility of their use as therapeutic targets [16]. A previous study found increased levels of LC3 and Beclin-1 transcription in chondrocytes during the initial phase of OA. However, with the progression of OA, the expression of these key genes regulating autophagy decreases, and autophagy weakens, leading to the damage of chondrocytes and the acceleration of joint degeneration [17]. The up-regulation of the mTOR gene associated with autophagy in OA was also found to inhibit the expression of important autophagy-regulating genes, such as ATG3, ATG5, ATG12, LC3B, Beclin-1, and Bcl2, resulting in the increased apoptosis of chondrocytes and accelerated progression of OA [18]. This situation suggests that regulating the key targets of autophagy is a promising direction for delaying the progression of OA. Critical autophagy genes can play an essential role in the development of OA by regulating immune functions, such as immune cell growth, antigen presentation, inflammatory cytokine secretion, and inflammatory pathway activation [19,20]. The differential expression of key autophagy genes might influence the extent and function of immune cell infiltration in OA. In addition, polymorphically expressed autophagy-related genes (ARGs) are implicated in the occurrence of various immune diseases, including rheumatoid arthritis and systemic lupus erythematosus [21]. However, only a few studies have fully elucidated the relationship between the expression of key autophagy genes and immune cell infiltration in OA. Exploring immune infiltration in OA and how it relates to autophagy will enrich our knowledge of OA’s pathogenesis.

Therefore, studying the potential autophagy-related molecular targets and mechanisms of action in OA is of great clinical importance. This study will explore the potential autophagy key genes through cartilage tissue. In this work, we used bioinformatics, machine learning, and experimentally validated methods to identify autophagy-related signature genes (ARSGs) in OA. First, we obtained the characteristic genes that were differentially expressed between OA and normal samples by applying the GSE10575 and GSE51588 datasets and verified their diagnostic value by using the external dataset GSE48556. Subsequently, we further validated the results acquired on the basis of the GEO dataset through in vitro experiments (qRT-PCR, Western blot analysis, and immunohistochemistry (IHC)), and animal models. We also conducted an immunocorrelation analysis. Our study will help explore the potential biological targets related to autophagy in OA, deepen the understanding of the pathogenesis of autophagy in OA, and provide some evidence that key autophagy genes are involved in the progression of OA.

Materials and methods

Data downloading and processing

Original transcriptome datasets (GSE10575, GSE48556, and GSE51588) based on normal and OA samples were downloaded from the Gene Expression Omnibus (GEO) database (GEO, https://www.ncbi.nlm.nih.gov/geo/). The ‘limma’ R package was used to correct data and conduct log2 processing on data with large values to avoid error biases between individual sample data. The GSE10575 and GSE51588 datasets were merged and batch-corrected with the ‘sva’ R package, and the merged dataset was used as the training group. GSE48556 was employed as a validation group. The training group included 16 cases of normal cartilage and 46 cases of OA cartilage. The validation group comprised 33 normal samples and 106 OA samples. PCA was performed on the training group data by using the ‘ggplot2’ R package to observe the effect of batch correction. ARGs were obtained from the database at http://www.autophagy.lu/. The ‘limma’ R package was used to acquire the training group’s expression matrix and autophagy-related differentially expressed genes (DEGs). Screening was performed with adjusted p < 0.05. The parameter for the number of gene displays is set to the first 50. The download and processing of the data approved by the Medical Ethics Committee of the First Affiliated Hospital of Guangxi Medical University (approval number: 2023-S185-01).

Biological enrichment analysis

Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses were used to identify which molecular functions and biological pathways of the DEGs linked to autophagy are involved. The ‘clusterProfiler’ R package was used to run the analysis with p < 0.05.

Screening of candidate signature genes

Three machine learning algorithms were employed to forecast the disease status of OA to identify the condition’s prognostic features. LASSO regression analysis selects variables through regularized analysis algorithms. The SVM-RFE algorithm selects and visualizes the most relevant signature genes by using nonlinear kernel. Random forest (RF) is an algorithm that utilizes multiple decision trees to flexibly train and predict variables, with the screening criterion being an importance score greater than 1 [22]. The ‘glmnet’, ‘e1071’, and ‘randomForest’ R packages were applied in the above three machine learning methods to screen for signature genes. DEGs were analyzed, and signature genes were identified by using the three algorithms. Finally, the intersection of the results of the three analytical methods was considered as the candidate signature genes.

Accuracy assessment of signature genes

The ‘limma’ R package was first used to draw violin and line plots of the obtained intersection signature genes to analyze the differences between the expression of signature genes in the normal and OA group samples and analyze the diagnostic value of the intersection signature genes. The receiver operating characteristic (ROC) curves of ARSGs in the training group were plotted to assess the accuracy of ARSGs for OA diagnosis. The high or low diagnosticity of ARSGs was determined by assessing the area under the curve (AUC). The threshold was 0.8. Accuracy was considered high if the threshold was more significant than 0.8 [23]. Next, the external dataset GSE48556 was used to verify whether the validation group’s signature genes differed.

Coexpression analysis

The training group samples were separated into high- and low-expression groups in accordance with the expression of signature genes, allowing the identification of genes that differed between the two groups. The ‘limma’ R package was used for the analysis. Filtration was done with|Log2FC| >1 and adjusted p < 0.05. Next, the ‘corrplot’ R package was applied to conduct a coexpression analysis of differentially expressed and signature genes to observe which genes and signature genes had positive or negative regulatory relationships.

Single-gene cluster enrichment and variation analyses of signature genes

The R package ‘clusterprofiler’ was used for the single-gene cluster enrichment analysis (GSEA) of signature genes to observe which functions or pathways were active in the groups with high and low expression levels of signature genes. The peak of the curve in the upper left indicates functions or pathways that are active in the group with the high expression of signature genes, and the peak of the curve in the lower right shows that the functions or pathways that are active in the group with the low expression of signature genes. The parameter settings for the GSEA analysis were: The absolute value of the normalized enrichment score (NES) is >1, the q-value of the false discovery rate (FDR) is <0.25, and the nominal p-value is <0.05. In addition, the potential biological functional changes of ARSGs were assessed by using the single-gene set variation analysis (GSVA) algorithm [24]. c2.cp.kegg.Hs.symbols.gmt and c5.go.Hs.symbols.gmt served as the reference gene set in the analysis. The parameters of GSVA analysis were set as p < 0.05 and |log FC| > 0.2.

Analysis of immune function and cellular infiltration

An immune cell-related function analysis was conducted by using the ‘ggplot2’ R package with p < 0.05 as the analytical cutoff, and the results were visualized [25] to assess which immune-related functions differed between the groups with high and low expression levels of signature genes.

Potential drug prediction and competing endogenous RNA network construction for signature genes

The Gene–Drug Interaction Database (DGIdb) (https://www.dgidb.org/) was applied to explore potential target drugs for signature genes [26]. Three online databases were used to search for the potential target markers of signature genes in OA: miRanda (http://www.mocrorna.org), miRDB (http://www.mirdb.org/), and TargetScan (http://www.targetscan.org/) [27]. The spongeScan database explores the interactions among signature genes, miRNAs, and lncRNA. The obtained regulatory network was analyzed and visualized by using Cytoscape_v3.9.1 software [28].

Culture and extraction of rat chondrocytes

A total of six 3–7-day-old SD rats were purchased from the Animal Experiment Center of Guangxi Medical University. Knee cartilage cells were extracted from 3–7-day-old SD rats killed through cervical dislocation and cultured in vitro under sterile conditions. In detail, the procedure involved cutting the cartilage tissue into tiny pieces and digesting them for 30 min at 37 °C by using trypsin (Solabio, China). The samples were then immersed in high-sugar DMEM (Gibco, USA) containing collagenase II (1 mg/mL; Gibco, USA) and incubated for 4–6 h at 37 °C. Next, the chondrocytes were filtered, centrifuged, and resuspended. Cultures were performed at 37 °C in DMEM (Gibco, USA) containing a mixture of penicillin (1%; Solarbio, China) and fetal bovine serum (10%; Gibco, USA). In vitro investigations were performed by using third-generation chondrocytes. Chondrocytes were treated with interleukin 1-β (IL1-β, Gibco, USA) for 48 h to establish an OA cell model. This experiment was approved by the Medical Ethics Committee of the First Affiliated Hospital of Guangxi Medical University (approval number: 2024-E417-01).

RNA extraction and qRT–PCR

RNA was extracted from cells in accordance with the operating instructions of the RNAeasy Plus animal RNA extraction kit (Beyotime, R0032, China). Its concentration and purity were detected with UV spectrophotometry. Subsequently, 1000 ng of total RNA in a total volume of 20 μl was reverse-transcribed into cDNA by using a reverse transcription tool (Takara, China). Real-time qPCR was normatively performed in accordance with the operating instructions of SYBR-Green (Thermo Fisher Scientific, USA) as follows: 95 °C for 10 min, 95 °C for 15 s, and 60 °C for 1 min for 40 cycles. The instrument used was the 7500 Real-Time PCR system (Applied Biosystems, USA). The 2^−ΔΔCT method was used for analysis. The GAPDH gene was used as the standard for normalization. The experiment was repeated three times for all samples. Primer sequences are listed in Table 1.

Table 1.

Primer information on relevant genes in the study.

Gene name	Forward primer (5′ to 3′)	Reverse primer (5′ to 3′)
GAPDH	ACCCACTCCTCCACCTTTGAC	TCCACCACCCTGTTGCTGTAG
ERBB2	GGGCTGGCTCCGATGTGTTTG	CCGCTGTAGAGGGCTGAGGTC

Western blot analysis

Chondrocyte proteins were extracted by using a configured protein lysis buffer containing a phosphatase inhibitor mixture (ComWin Biotech Co., Ltd.), protease inhibitor (MedChemExpress, Monmouth Junction, NJ), RIPA lysate (tissue/cells) (Solarbio, Beijing, China), and PMSF (Solarbio, Beijing, China). BCA protein assay (Beyotime, China) was utilized to ascertain protein concentration. Approximately 50 μg of the protein sample was isolated using 15% (v/v) SDS-PAGE (Epizyme, China). The protein sample was isolated, and the isolated protein was transferred to a polyvinylidene fluoride (PVDF) membrane (Millipore, Billerica, MA, USA). The membrane was blocked with a protein-free rapid blocking buffer (Epizyme, China) for 30 min. The membrane was immersed in the corresponding primary antibodies of GAPDH (Proteintech, 10494-1-AP, 1:5000) and ERBB2 (Proteintech, 18299-1-AP, 1:1000) at 4 °C and incubated in a shaker overnight. The PVDF membrane was washed with TBST (Solarbio, Beijing, China) and incubated with the secondary antibody for 1 h at room temperature. Protein blotting signals were visualized and recorded by using ProteinSimple’s Wes system. Data were normalized against GAPDH for the protein ERBB2.

Immunohistochemistry

In this study, three normal human cartilage samples and three OA cartilage samples were collected from the First Affiliated Hospital of Guangxi Medical University (Nanning, China) between June 2024 and September 2024 for immunohistochemical staining to detect ERBB2 expression. Written informed consent of each patient was obtained during the sample extraction process. The inclusion criteria for patients in the OA group were as follows: (1) Age > 18 years old. (2) Patients who meet the diagnosis of knee OA and require surgical treatment. (3) Patients undergoing knee replacement for the first time. The exclusion criteria for patients in the OA group were as follows: (1) Patients with rheumatoid arthritis. (2) Patients with traumatic arthritis. (3) Patients with mental illnesses. The OA group included one male and two female patients, with an average age of 64 years. The inclusion criteria for patients in the normal cartilage group were as follows: (1) Normal articular cartilage that fell off due to traumatic fractures. (2) Patients without OA or rheumatoid arthritis. Exclusion criteria: (1) Patients with OA or rheumatoid arthritis. (2) Patients with diabetes or gout. (3) Patients who have undergone surgical treatment in the joint area before. The patients in the normal cartilage group included two male and one female patients, with an average age of 36.7 years. The specimens were fixed in 4% paraformaldehyde (Biosharp, China), decalcified by using 10% ethylenediaminetetraacetic acid (MCE, USA), and paraffin-embedded. Specimens were sectioned and incubated for 10–15 min at room temperature with 3% hydrogen peroxide (Solarbio, China) to remove endogenous peroxidase activity. Next, they were blocked with 10% goat serum (Gibco, USA) at room temperature for 20 min and then incubated overnight with the primary antibody of ERBB2 (Proteintech, 18299-1-AP, 1:300). The specimens were then incubated with the secondary antibody on the following day. Subsequently, staining was performed by using hematoxylin and diaminobenzidine (ZSGB-BIO, China). Finally, the specimens were dehydrated, sealed with neutral resin, and observed and recorded under an inverted phase contrast microscope (Olympus, Japan). This experiment was approved by the Medical Ethics Committee of the First Affiliated Hospital of Guangxi Medical University (approval number: 2024-E417-01).

Construction of rat knee OA models

Twelve 6-week-old male SD rats weighing 220–250 g were purchased from the Animal Experiment Center of Guangxi Medical University. Rats were randomized into two groups, normal and OA. Rats in the OA group: The anterior cruciate ligament and medial meniscus of the left knee joint were removed [29]. The normal group underwent sham surgery in which only the left knee joint cavity was opened and sutured without destroying the intra-articular structures. All rats were euthanized, and left knee samples were collected 8 weeks after surgery. Hematoxylin–Eosin (H/E) staining (Solarbio, China), SO/FG staining (Solarbio, China), and Osteoarthritis Research Society International (OARSI) scoring were performed on the specimens to evaluate cartilage damage [30]. IHC was performed to assess the expression levels of the ERBB2 protein. The animal experiments in this study were performed in accordance with institutional guidelines and approved by the First Affiliated Hospital of Guangxi Medical University and the Guangxi Medical University Animal Experimentation Center (2023-S185-01).

Statistical analysis

Bioinformatics data were analyzed by using R v.4.3.1 software. Wilcoxon test was conducted to compare the differences between two groups. The relationship between various variables was investigated by performing Spearman correlation analysis. In vitro data were expressed in the form of mean ± standard deviation, and one-way analysis of variance and independent sample t-test were used for between-group analysis. The statistical significance level was set at p < 0.05.

Results

Identification of autophagy-related DEGs

We obtained a gene expression matrix after merging the datasets and removing batch effects. PCA analysis of the merged dataset eliminated the effects of batch effects and the data were randomly distributed (Figure 1A and ^B). We identified autophagy-related DEGs through the differential analysis of the gene expression matrix. We found 49 DEGs, of which 12 were up-regulated and 37 were down-regulated (Figure 1C and ^D).

Figure 1.

Analysis of OA autophagy-related DEGs. (A) PCA results of two datasets before batch correction. (B) PCA results of two datasets after batch correction. (C) Heat map of DEGs. (D) Volcano map of DEGs.

Bioenrichment analysis

We used R software to perform KEGG and GO enrichment analyses on the obtained DEGs to search for potential biological activities and functional pathways associated with autophagy. GO analysis revealed that autophagy-related DEGs were involved in the regulation of autophagy, regulation of macroautophagy, reaction to nutrient levels, reaction to low oxygen levels, reaction to autophagy, and disassembly of mitochondrion, mitochondrial and organelle outer membrane, ubiquitin-like protein ligase and heat shock protein binding, translation regulator, and MAP kinase activity (Figure 2A). KEGG showed that the PI3K-Akt, mTOR, NOD-like receptor, FoxO, ErbB, and AMPK signaling pathways and other pathways were significantly enriched (Figure 2B).

Figure 2.

Bioenrichment analysis of DEGs. (A) GO enrichment analysis. (B) KEGG enrichment analysis.

Screening of signature genes

We further screened the 49 autophagy-related DEGs obtained earlier. LASSO regression algorithm screened out nine signature genes (ERO1L, CAPN2, ITGA3, BNIP3, ERBB2, GOPC, NLRC4, TUSC1, and BNIP3L) (Figure 3A). SVM-RFE screened out 12 signature genes (CAPN2, BNIP3, GOPC, ERO1L, ERBB2, ITGA3, TUSC1, ITGB4, TSC2, SAR1A, BNIP3L, and GABARAPL1) (Figure 3B). RF identified eight signature genes (ITGA3, CXCR4, EIF4EBP1, MAPK1, ERBB2, CANX, CAPN2, and IRGM) (Figure 3C). By taking the intersection of the signature genes identified through the above three methods, we finally identified three signature genes (CAPN2, ITGA3, and ERBB2) (Figure 3D). These signature genes may be the key genes related to autophagy in OA.

Figure 3.

Machine learning screening of ARSGs. (A) Signature genes screened from DEGs using the LASSO regression algorithm. (B) Signature genes are filtered from DEGs using the SVM-RFE algorithm. (C) Signature genes screened from DEGs using the RF algorithm with the criterion of significance score greater than 1. (D) The Venn diagram of intersecting genes.

Analysis and validation of signature genes

We plotted violin and line plots of the CAPN2, ITGA3, and ERBB2 (Figure 4A and B). We discovered that the expression levels of these three signature genes were higher in the OA samples than in the healthy samples. We plotted the ROC curves of these three signature genes and evaluated their diagnostic value on the basis of AUC values. The AUC values of CAPN2, ITGA3, and ERBB2 were 0.887, 0.882, and 0.827, respectively (Figure 4C). This finding suggests that these three signature genes are highly accurate as diagnostic genes. Through the verification of the external validation dataset GSE48556, we revealed that only ERBB2 differed between OA and normal samples, and its expression was elevated in OA samples (Figure 4D).

Figure 4.

Analysis and validation of signature genes. (A) Violin plot of characteristic gene expression in normal and OA samples. (B) Line plot of signature gene expression in normal and OA samples. (C) ROC curves to validate the diagnostic accuracy of signature genes. (D) Expression of ERBB2 in the GSE48556 dataset.

Analysis of the significance of the differential expression of ERBB2

Based on the expression level of the ERBB2 gene in the samples of each training group, the median value of ERBB2 gene expression was calculated. Next, the samples in the training group were categorized into high- and low-expression groups based on the median value of ERBB2. A total of 117 genes with significant differences between the high- and low-expression groups were identified. Among the DEGs, 97 were up-regulated and 20 were down-regulated in the group with high ERBB2 expression. We plotted the heat and volcano maps of the top 20 most significant DEGs (Figure 5A and ^B). Next, we performed coexpression analysis on ERBB2 and the top 20 most significant DEGs. Our results showed that ERBB2 had a positive regulatory relationship with BEST2, CEACAM19, SH2D5, BMP10, DPEP3, PRM2, and RLBP1 and a negative regulatory relationship with C20orf197, SOX7, RTN1, SETD1B, GBP1, SRF, and GPR126 (Figure 5C).

Figure 5.

Analysis of ERBB2 expression. (A) Heat maps of the top 20 DEGs in high and low ERBB2 expression groups. (B) Volcano map of DEGs in ERBB2 high and low expression groups. (D) Heat map of the correlation between ERBB2 and DEGs, with positive correlation in red and negative correlation in blue.

Analysis of the immune-related functions of ERBB2

The occurrence of autophagy in OA is closely related to immune-related functions [19]. We performed immune-related functional analysis on samples with high and low expression levels of ERBB2. In the high-expression group of ERBB2, the abundance of immune cells, such as macrophages, neutrophils, NK cells, T tumor-infiltrating lymphocytes, and regulatory T cells, and molecular functions, such as cytolytic activity, human leukocyte antigen, inflammation promotion, MHC class I molecules, and cell coinhibition, were down-regulated (Figure 6). This finding suggests that the high expression of ERBB2 in the development of OA may affect immune-related functions, inhibiting autophagy in OA.

Figure 6.

Immune-related function analysis in samples of high and low ERBB2 expression groups (*p < 0.05, **p < 0.01).

Biological function and pathway analyses of ERBB2

GSEA revealed the biological functions and pathways regulated in the high- and low-ERBB2 expression groups. We found that the detection of stimulus involved in sensory perception, sensory perception of chemical stimulus, signaling receptor regulator activity, neuroactive ligand, and receptor interaction were enriched in the group with high ERBB2 expression (Figure 7A and ^B). Surprisingly, in the group with low ERBB2 expression, we found that immune-related functions and pathways were significantly enriched in the group with high ERBB2 expression. These pathways included four different signaling pathways (immune response regulating, chemokine, B cell receptor, and adipokine signaling pathways), antigen processing and presentation of exogenous peptide antigens, positive regulation of defense response, regulation of innate immune response, and FC gamma R-mediated phagocytosis (Figure 7C and ^D).

Figure 7.

GSEA and GSVA analysis of ERBB2. (A–B) GSEA analysis of ERBB2 in high expression group. (C–D) GSEA analysis of ERBB2 in low expression group. (E) E GSVA analysis of biological processes, cellular components, and molecular functions of ERBB2. (F) Pathway-related GSVA analysis of ERBB2.

We performed GSVA to further explore the potential mechanisms through which ERBB2 affects OA progression (Figure 7E and ^F). Our results showed that low transcriptional levels of ERBB2 were involved in the regulation of nitric oxide-mediated signal transduction, neuronal action potential propagation, positive regulation of protein localization to the nucleolus, Ccr3 chemokine receptor binding, histone h3k4 demethylation, positive regulation of voltage-gated calcium channel activity, galactose metabolism, and steroid hormone biosynthesis. High transcriptional levels of ERBB2 participated in the negative regulation of centrosome replication, immunoglobulin production involved in immunoglobulin-mediated immune responses, nuclear receptor coactivator activity, 5-phosphoribose-1-diphosphate metabolic process, peptide and peptide lysine acetyltransferase activity, positive regulation of DNA recombination, glycosylphosphatidylinositol anchor biosynthesis, autophagy regulation, Notch signaling pathway, and endocytosis. The above findings suggest that the low expression of ERBB2 may induce chondrocyte autophagy by promoting the activation of the immune pathway, thus delaying the progression of OA. They enrich our understanding of the pathogenesis of OA.

Construction of the competing endogenous RNA network of ERBB2 and potential drug prediction

We constructed a competing endogenous RNA (ceRNA) network encompassing the target genes ERBB2, miRNA, and lncRNA to discover their interactions. The interaction network consisted of ERBB2, 18 lncRNAs, and eight miRNAs. We found that hsa-miR-593-3p, hsa-miR-323b-5p, hsa-miR-331-3p, hsa-miR-450b-5p, hsa-miR-1224-3p, hsa-miR-1225-3p, hsa-miR-1202, and hsa-miR-193a-5p may competitively regulate ERBB2. Moreover, these 18 lncRNAs can bind to these eight miRNAs to regulate ERBB2 (Figure 8A). We used the DGIdb database to predict potential target agents for ERBB2. A total of 30 potential therapeutic agents, including ERTUMAXOMAB, TALAZOPARIB, EMODIN, and others, were identified, and the results were visualized by using Cytoscape (Figure 8B).

Figure 8.

ceRNA Regulatory network and potential drug action network. (A) ceRNA network regulation diagram of lncRNA-miRNA-ERBB2. (B) Potential drug-signature gene network diagram of OA, red for ERBB2, blue for drug.

Gene and protein expression levels

We used different concentrations (10 and 20 ng/mL) of IL1-β to induce the OA chondrocyte model. QRT-PCR and Western blot analyses revealed that the transcription and protein expression levels of ERBB2 gradually increased with increasing IL-1β concentration (Figure 9A and B) (p < 0.05). We performed IHC on human cartilage samples to compare the differences in ERBB2 expression between normal and OA cartilage samples. The results showed that the positive rate of ERBB2 was higher in the OA group than in the normal group. This result further confirmed our findings (Figure 9C).

Figure 9.

ERBB2 Transcription levels and protein expression levels between normal and OA samples. (A) Expression level of ERBB2 in qRT-PCR. (B) Protein expression level of ERBB2. (C) Immunohistochemical results of normal human cartilage and OA cartilage from patients (*p < 0.05; **p < 0.01; ***p < 0.001).

Results of animal experiments

We constructed an OA model of the rat knee joint and evaluated cartilage damage in the knee joints of the normal and OA groups by H/E and SO/FG staining. In the results of H/E staining, the cartilage surface of the knee joint in the OA group was rough and discontinuous with meniscus damage compared to the knee joint in the normal group (Figure 10A). In the SO/FG staining results, the knee joints in the OA group stained unevenly and discontinuously (Figure 10B). The cartilage damage score (OARSI score) was higher in the OA group compared to the normal group (Figure 10C). This suggests that the rat knee OA model we constructed was successful. Detection of ERBB2 levels in cartilage tissue by IHC showed that ERBB2 levels were significantly higher in the OA group (Figure 10D and ^E).

Figure 10.

Tissue staining and immunohistochemical analysis. (A) Hematoxylin-eosin staining of normal and OA rat knee joints. (B) so/FG staining of normal and OA rat knee joints. (C) OARSI score. (D-E) Protein expression level and quantitative analysis of ERBB2 in knee cartilage by IHC (*p < 0.05; **p < 0.01; ***p < 0.001).

Discussion

Although the search for autophagy-related biomarkers and therapeutic targets is the perennial direction of OA therapy, related research has been slow. In this study, we found that ARSGs were closely associated with the development of OA. This finding deepened our understanding of the relationship between autophagy and OA pathogenesis. In addition, the results of immune correlation analysis revealed the possibility of immunotherapy, suggesting that ARSGs have the potential to serve as new biomarkers.

With the explosive growth of the global population and expansion of the aging population, OA, the most common type of arthritis, now affects an increasing number of people [31]. Irreversible aging processes and abnormal stress loads exacerbate the development of OA [32]. Early detection and intervention are critical for decelerating OA progression and improving joint function. When OA progresses to the middle and late stages, it can be diagnosed through relevant clinical symptoms and imaging tests [33]. However, finding practical early diagnostic tools for OA in actual clinical work is difficult [34]. Therefore, we screened 49 ARGs that were differentially expressed between OA and normal samples by using the publicly available GEO dataset. Our GO results showed that 49 autophagy-related DEGs were involved in the regulation of autophagy, reaction to nutrient levels and low oxygen levels, mitochondrial and organelle outer membrane, ubiquitin-like protein ligase and heat shock protein binding, translation regulator, and MAP kinase activity. KEGG showed that these DEGs participated in the PI3K-Akt, mTOR, NOD-like receptor, FoxO, ErbB, and AMPK signaling pathways. This finding suggests that autophagy-related DEGs regulate the inflammatory development of OA in multiple ways.

We then screened the 49 autophagy-related DEGs by using three machine learning methods to identify potential biomarkers in OA. The results showed that CAPN2, ITGA3, and ERBB2 were possible key genes. We found that only ERBB2 passed the validation based on the external validation dataset and that its expression was elevated in the OA samples. ROC analysis confirmed the high accuracy of ERBB2 as a diagnostic gene. We verified the difference between ERBB2 expression in normal and OA cartilage samples by applying in vitro cell experiments, clinical samples, and animal model. The ERBB2 expression found through this analysis was consistent with that found via our previous bioinformatics analysis. Moreover, ERBB2 expression in chondrocytes progressively increased as inflammation degree increased. On the basis of the degree of ERBB2 expression, we divided the samples from the training group into the high- and low-expression groups. We identified 117 significant DEGs, which we then analyzed for coexpression. We found that high ERBB2 expression can inhibit the expression of C20orf197, SOX7, RTN1, and other genes. This finding enriches our understanding of the role of autophagy in the pathogenesis of OA. GSEA and GSVA analyses revealed that ERBB2 is closely associated with various signaling pathways, including chemokine, B cell receptor, and adipokine signaling pathways. ERBB2, a tyrosine receptor kinase, plays an essential regulatory role in the proliferation and differentiation of cartilage, osteoblasts, and osteoclasts, and its aberrant expression has been associated with various diseases and is involved in the initiation of the PI3K–Akt and MAPK pathways [35,36]. Wang et al. noted that ERBB2, which is upregulated in postmenopausal women, might serve as a crucial gene for the development of OA and noted that it could be a target for treatment [37]. Hallbeck et al. note that an aberrant growth pattern may be induced by increased ERBB2 expression in synovial joint tissues, leading to inflammatory synovial hyperplasia [38]. ERBB2 was also found to be a signature gene of atopic dermatitis. The down-regulation of ERBB2 expression was involved in the pathogenesis of atopic dermatitis by regulating chemokines (CCL26 and eotaxin-3) [39]. It was also found that the reduced production of adipokines inhibited the targeting effect of ERBB2 in esophageal squamous cell carcinoma [40]. Our research team verified for the first time that ERBB2, a key gene of autophagy, is differentially expressed between normal and OA samples. This result suggests that ERBB2 is a potential autophagy-related diagnostic biomarker in OA.

We additionally investigated the possible relationship between ERBB2 expression and infiltrating immune cell subpopulations to obtain further knowledge of the regulatory function of ARGs in OA cartilage and fresh perspectives of the pathogenesis of OA. Studies have shown that autophagy is associated with immune regulation; immune cell infiltration plays a vital role in the development of OA and that its role includes two aspects, namely, innate and adaptive immunity; and the loss of control of autophagic homeostasis increases the severity of OA [15,41]. Our study revealed a positive correlation between ERBB2 expression and OA inflammation degree. Compared with the group with high ERBB2 expression, that with low ERBB2 expression had higher immune cell infiltration and immune function activation. These cells and functions involved leukocyte antigens, MHC class I molecules, macrophages, neutrophils, NK cells, tumor infiltrating lymphocytes, regulatory T cells, cytolytic activity, and inflammation promotion. Similar to our study, a previous study found that in OA synovial and periarticular fat, with the activation of innate immune response, the infiltrated cells are mainly macrophages and inflammatory cells [42]. The above results may be explained as follows: in the early stage of OA, the low expression of ERBB2 promotes the activation of autophagy. Subsequently, autophagy maintains the stability of intracellular energy metabolism through interaction with the immune system, ensures the normal function of cells, and prevents the progression of OA. When OA progresses to the later stage in an environment with high ERBB2 expression, the infiltration of related immune cells reduces, and the destruction of autophagy homeostasis leads to the weakening of clearance, which ultimately aggravates the progression of OA. Similar results were found by GSEA and GSVA. The group with low ERBB2 expression exhibited enrichment in immune response regulatory, cytokine, and B cell receptor signaling pathways; antigen processing and presentation; regulation of innate immune response; and phagocytosis. Recognizing the relationship of ERBB2 with immune cell infiltration and immune-related functions from the perspective of immunity will help reveal the potential mechanism of ERBB2 in OA cartilage and provide a new therapeutic target for OA.

We acquired a relational network consisting of ERBB2, 18 lncRNAs, and eight miRNAs to understand how ERBB2 and lncRNAs competitively bind miRNAs. Wang et al. reported that abnormally expressed lncRNAs in OA cartilage are involved in the pathogenetic process of OA and pointed out their important value as therapeutic targets [43]. This finding suggests that miRNAs or lncRNAs that directly or indirectly bind to ERBB2 have the potential to become new targets for the treatment of OA. Previous studies have discovered that miR-422a is associated with postmenopausal osteoporosis and expressed in high amounts in osteoclasts in the postmenopausal low-bone density group [44]. TOB2, a transducer of ERBB2, is believed to participate in osteoporosis and has a clear negative correlation with miR-422a [45]. Up-regulated miR-375 promotes the proliferation and osteogenic differentiation of human periodontal ligament stem cells by targeting the downstream ERBB2 transducer [46]. In addition, elevated miR-543, by acting as a promoter, promotes the osteogenesis of human periodontal ligament stem cells by inhibiting the ERBB2 transducer [47]. Herein, we identified 30 potential therapeutic agents, including ertumaxomab, talazoparib, and emodin. This suggests that they may be new targets for OA therapy.

However, our study has some shortcomings. The first is the small sample size of public database sources. The detailed mechanism of action of ERBB2 in OA also needs further investigation. The results of this study can only support the correlation between ERBB2 and immune cells. Further studies are required to verify the causal relationship between these two factors at the molecular level. The therapeutic efficacy of relevant potential drugs and interrelationship between lncRNAs and miRNAs need additional clinical validation.

Conclusion

We found that ERBB2 is a potential ARSG of OA and has crucial value for the early diagnosis and treatment of OA. The interaction of ERBB2 with immune cell infiltration suggests that autophagy may regulate the development of OA by modulating immune responses. The potential targeted therapeutic agents and lncRNAs and miRNAs in the ceRNA network results provide new insights for OA treatment.

Funding Statement

This work was supported by the Guangxi Medical and Health Appropriate Technology Development and Extension and Application Project (grant/award number: GZSY22-62), Nanning Qingxiu District Science and Plan Project (grant/award number: 2020018)

Ethics statement

In the experimental section, this study was conducted in accordance with the ethical standards of the Declaration of Helsinki and its subsequent amendments. Written informed consent was obtained from all individuals involved in the study for any identifiable data or images involved in the text. The download and processing of the GEO data, the collection of human tissues, and cell experiments were approved by the Medical Ethics Committee of the First Affiliated Hospital of Guangxi Medical University (approval number: 2024-E417-01). The animal experiments in this study were performed in accordance with institutional guidelines and approved by the First Affiliated Hospital of Guangxi Medical University and the Guangxi Medical University Animal Experimentation Center (2023-S185-01).

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The original public dataset used in this paper is available for download from the Gene Expression Omnibus (GEO, https://www.ncbi.nlm.nih.gov/geo/) platform. Additional data used to support the results of this study are available from the corresponding author upon request.