Identification of susceptibility modules and genes through WGCNA and ceRNA network analysis in neuropathic pain-induced anxiodepression

Abstract

Neuropathic pain (NP) is frequently comorbid with anxiety and depression, yet the underlying molecular mechanisms in the brain remain poorly understood, hindering the development of targeted therapies. This study aimed to identify key transcriptional networks and regulatory pathways in the anterior cingulate cortex (ACC) associated with NP-induced anxiodepression. We analyzed transcriptomic data (GSE92718) from the ACC of a mouse model of chronic NP. By comparing differentially expressed genes at a time point manifesting anxiodepressive-like behavior (8-week post-injury) against those with pain alone (2-week), we constructed a weighted gene co-expression network (WGCNA). A key module (blue module) significantly correlated with the anxiodepressive phenotype was enriched for synaptic signaling (glutamatergic/GABAergic), neuroplasticity, and key pathways like MAPK and Ras. Within this module, we identified 7 pivotal lncRNAs and 5 hub mRNAs (Flt1, Slc38a2, Bmpr1b, Pdgfra, Gng2) via integrated lncRNA-mRNA-pathway and protein-protein interaction network analyses. Furthermore, we established a competing endogenous RNA (ceRNA) network, revealing a core regulatory axis comprising 3 hub lncRNAs, 5 hub mRNAs, and 40 miRNAs. The aberrant expression of the five hub mRNAs in the ACC was specifically validated in mice with anxiodepressive phenotypes using RT-PCR. Our findings unveil a critical ceRNA network and implicate dysregulated synaptic genes in the ACC as key drivers of NP-induced anxiodepression, providing novel insights into its molecular basis and highlighting potential diagnostic biomarkers and therapeutic targets.

Introduction

Neuropathic pain (NP) is a serious chronic disease caused by injuries or specific diseases that affect the somatosensory nervous system [1]. The incidence of NP is estimated to be 7% to 8% worldwide, and it has become a major public health problem with global aging [2]. NP patients are prone to several complications, such as anxiety, depression, cognitive impairment and sleep disorders, this fact is important to highlight [3, 4]. More importantly, the current prevalence of mood disorders among patients with NP stands at a notable 29.7%. Moreover, the cumulative incidence throughout their lifetime is alarmingly high, reaching up to 47% [3]. Managing patients with multiple conditions is much more complex than treating those with just NP or depression. Mental disorders remain among the top ten leading causes of the world’s disease burden [3]. Although an increasing number of studies have improved the understanding of NP-induced anxiodepression [5, 6], identifying an appropriate therapeutic method remains a challenge for physicians.

NP-induced anxiodepression arises from dysregulation within a broader brain network encompassing regions such as the anterior cingulate cortex (ACC), amygdala, insula, and prefrontal cortex [7]. Specifically, abnormal hyperactivity in the amygdala may amplify negative affect, while functional alterations in the prefrontal cortex can disrupt the cognitive evaluation of pain [8]. These regions do not act in isolation but form interconnected functional circuits, collectively contributing to the complex and bidirectional nature of pain-depression comorbidity [9]. Within this network, the ACC serves as a central hub for integrating the affective dimension of pain and modulating depressive symptoms [10]. Preclinical research indicated that injecting ibotenic acid into ACC lesions can effectively avert the anxiodepressive-like behaviors induced by NP without affecting allodynia [11]. With the progress of next-generation sequencing, transcriptome analysis has been used to explore brain pathology [12, 13]. Chronic nerve injury can cause extensive changes in gene expression in the ACC, some of which contribute to functional deficits [11, 14]. Although previous studies on expression profiling have identified several genes contributing to the depressive-like phenotype [15–17], there is still a lack of comprehensive systematic analysis of the intrinsic changes in the ACC in NP-induced anxiodepression. Notably, further exploration of the representative genes with internal connections is needed.

Exploring pain related biomarkers may be crucial for NP-induced anxiodepression [18]. Fortunately, weighted gene correlation network analysis (WGCNA) is an effective method of identifying the molecular pathways that are coregulated in a functionally coherent pattern [19]. WGCNA divides genes into diverse modules based on the similarity of coexpressed genes. To date, emerging research has constructed a co-expression gene network, and hub genes have been elucidated by the WGCNA algorithm in some types of depression [20, 21]. The application of the WGCNA algorithm to identify NP-induced anxiodepression-related genes and ceRNAs represents a novel approach. Traditional analysis relies on sequence complementarity between lncRNAs and mRNAs and overlook co-expression relationships between genes. In contrast, the WGCNA algorithm circumvents this limitation. Deng et al used WGCNA to identify key biological processes, signaling pathways and potential diagnostic biomarkers mediating the development of postpartum depression (PPD) [17]. However, the gene-related molecular features and phenotype-centric gene networks involved in NP-induced anxiodepression remain unclear.

The original analysis of the GSE92718 dataset by Barthas et al (2017) provided a seminal insight by identifying and functionally validating the critical role of MKP-1 in the ACC for NP-induced depression [14]. However, this focused approach, while did not leverage the dataset to explore the broader, systematically co-regulated gene networks and the potential involvement of non-coding RNA-mediated regulatory axes that might strengthen the anxiodepressive phenotype. To address this gap and uncover a more comprehensive molecular landscape, we employed WGCNA to identify the modules with highly correlated expression patterns that are collectively associated with the trait of interest—specifically, the transition from pain-alone (2-week) to pain-with-anxiodepression (8-week) states. Furthermore, we integrated this with competing endogenous RNA (ceRNA) network analysis to elucidate the potential post-transcriptional regulatory interplay between lncRNAs, miRNAs, and mRNAs within the key modules. Thus, our study represents a distinct and complementary systems biology re-analysis of the GSE92718 data, aiming to move beyond individual gene discovery towards identifying coordinated network-level dysregulations.

Materials and methods

Gene expression dataset

The microarray data under access number GSE92718 were downloaded from the National Center of Biotechnology Information (NCBI) Gene Expression Omnibus (GEO; www.ncbi.nlm.nih.gov/). In the GSE92718 dataset, samples were collected from the ACC of mice from which the sciatic nerve cuffing model was used to induce chronic NP and anxiodepression. Gene expression was investigated after 2 and 8 weeks in sham- or cuff-operated mice. The 8-week time point represented nociceptive model displaying an anxiodepressive-like phenotype. The 2-week time point corresponded to the nociceptive model without anxiodepressive phenotype. Thus, 24 samples in GSE92718 were processed, including 12 samples (2-week and 8-week) from the sham group and 12 samples (2-week and 8-week) from the cuff group.

Animal preparation and NP model

Male C57BL/6 J mice (6–8 weeks old; weighing 20–30 g) were obtained from the Laboratory Animal Center of Zhengzhou University and housed in the animal facility of the School of Basic Medicine, Zhengzhou University. The animals were maintained under a natural light–dark cycle at a controlled temperature of 23 ± 2 °C. Efforts were made to minimize animal suffering and optimize the number of animals used. Sample sizes were determined based on a power analysis (α = 0.05, power = 0.8) to ensure adequate statistical power. All experimental procedures were conducted in accordance with the National Institutes of Health (NIH) Guidelines for the Care and Use of Laboratory Animals and were approved by the Research Ethics Committee of the First Affiliated Hospital of Zhengzhou University (Approval No. 2023-KY-1032). Mice were randomly assigned to experimental groups, and all subsequent experiments and analyses were performed blind to treatment conditions.

An animal model of NP was generated by sciatic nerve cuffing surgery as previously described [22]. Briefly, adult male C57BL/6 J mice were anesthetized with 1.5%–2% isoflurane/oxygen. The NP model was induced by placing a polyethylene cuff around the right common sciatic nerve of the animal. The sham group underwent the same operation without cuff implantation.

Behavioral testing

Mechanical allodynia in mice was evaluated using the paw withdrawal threshold (PWT), measured via the von Frey test as previously described by Chaplan et al with the up-down method [23]. Before testing, mice were acclimated for 30 min daily over three consecutive days in a Plexiglas chamber (Xinruan Information Technology Co., Ltd., Shanghai, China) fitted with a metal mesh floor. After habituation, the plantar surface of each hind paw was stimulated perpendicularly with a series of von Frey filaments (Aesthesio, USA) of logarithmically increasing stiffness (0.04, 0.07, 0.2, 0.4, 0.6, 1, 1.4, and 2 g). Each filament was applied with enough force to bend for 5 s. A rapid withdrawal or lifting of the paw was recorded as a positive response. Following a positive response, the next thinner filament was applied; otherwise, a thicker filament was used in the subsequent trial. Trials were spaced 5 min apart.

Thermal hypersensitivity was assessed using the Hargreaves test, with paw withdrawal latency (PWL) serving as the outcome measure [24]. Mice were individually placed in test compartments (XR1880, Xinruan Information Technology Co., Ltd., Shanghai, China) on a temperature-regulated glass platform set at 30 °C. A radiant heat source was directed through an aperture onto the lateral plantar surface of each hind paw. The time interval from the onset of thermal stimulation to paw withdrawal was recorded as the PWL. Each paw was tested five times with 5‑min intervals between trials, and the results were averaged. A cut-off latency of 15 s was applied to avoid tissue injury from prolonged heat exposure.

To further assess anxiety-related behavior, the elevated plus maze (EPM) test was performed. The EPM apparatus (XR-XG201, Xinruan Information Technology Co., Ltd., Shanghai, China) comprised four arms (40 cm × 10 cm) elevated 50 cm above the floor, with two opposing open arms and two opposing enclosed arms fitted with 20 cm high black walls. Each mouse was placed on the central platform facing an open arm and allowed to freely explore the maze for 5 min. Time spent in the open arms was recorded and quantified using VisuTrack software, as described previously [25].

We referred to previous studies and used the (NSF) to evaluate anxiodepressive-like behavior in mice [26]. Prior to the experiment, mice were acclimated to the testing room for 30 min daily over three consecutive days. To ensure a state of hunger and increase food motivation, animals were fasted for 24 h before testing with free access to water. During the formal test, food was placed in the center of an open field (40 cm × 40 cm × 30 cm). Each mouse was positioned in one corner of the arena, facing away from the food, and the time from placement to the initiation of feeding was recorded as the feeding latency. Before testing each animal, the arena was cleaned with 70% ethanol to remove residual odors, and fresh food was provided.

The forced swimming test (FST) was also used to evaluate anxiodepressive-like behavior in mice [27]. Prior to the formal experiment, mice were acclimated to the testing room for 30 min daily across three consecutive days. During the formal test, a transparent cylindrical glass container (30 cm height, 11 cm diameter) was filled with water to a depth of 10 cm, maintained at 23–25 °C. Mice were gently placed into the water, and the duration of immobility was recorded. Immobility was defined as the cessation of active struggle, during which the animal remained floating with only minor limb movements necessary to keep its head above the water surface. The total test duration was 6 min; the first 2 min served as an adaptation period, and immobility time was measured over the final 4 min. At the end of the test, mice were promptly removed from the water, dried with a towel, and placed on a heating pad to maintain body temperature and prevent hypothermic stress.

Identification of differentially expressed genes

The microarray data from the GSE92718 dataset were annotated with mRNAs and miRNAs but not with lncRNAs. First, we annotated the lncRNA data according to the files of the GPL6887 Illumina Mousewg-6 v2.0 expression bead chip. The microarray data were analyzed using the Limma package in R to identify differentially expressed genes (DEGs). A linear model was fitted to the expression data. After significance analysis and FDR analysis, the DEGs were selected according to the following criteria: FDR < 0.05 and absolute fold change (FC) > 1.1.

Weighted Gene Correlation Network Analysis

WGCNA was used to identify modules containing genes with similar expression patterns via the package “WGCNA” in R software [19]. Cluster analysis was used to eliminate outliers based on DEGs. Then, a suitable soft-threshold power β value was determined to balance the relationship scale independence and mean connectivity. Next, a topological overlap matrix (TOM) was established based on the β value. Cluster dendrograms and eigengene adjacency heatmaps were generated to show gene clustering and module relationships. To determine the most important module for further study, the eigengene for each module was calculated. Eigengenes were used to estimate the module-trait relationships under anxiodepression conditions.

Gene ontology and KEGG pathway enrichment analysis

Gene Ontology analysis was used to analyze the main functions of the DEGs in terms of the Gene Ontology database [28]. The Kyoto Encyclopedia of Genes and Genomes (KEGG) was applied to analyze the potential regulatory pathways of DEGs involved in the WGCNA modules [29]. A Bonferroni-corrected P value of less than 0.05 was selected as the cut-off criterion for significant enrichment.

lncRNA-mRNA network and lncRNA-mRNA-pathway network construction

The genes with high connectivity in the module of interest were identified, and the top 100 genes (including lncRNAs and mRNAs) were selected. Genes with high connectivity are usually regulatory factors, while genes with low connectivity are usually downstream of the regulatory network. After extracting the hub genes, the weights of the co-expression relationships between genes were calculated. Then, we selected the lncRNA-mRNA and mRNA-mRNA co-expression regulatory relationships and constructed a lncRNA-mRNA network. According to the regulatory relationships of the lncRNA-mRNA network and pathways in which mRNAs participate, a lncRNA-mRNA-pathway network was constructed.

Protein-protein interactions network analysis

The PPIs between mRNAs were constructed by the Search Tool for the Retrieval of Interacting Genes (STRING) database [30]. An interaction score of 0.2 was regarded as the cut-off criterion, and the PPI was visualized.

CeRNA network construction

To calculate the possible target lncRNAs and mRNAs of miRNAs, a few databases were used. First, miRNA-mRNA pairs and miRNA-lncRNA pairs were predicted by searching for differentially expressed miRNAs (DEmiRNAs) in the miRanda, TargetScan, miRWalk and PITA databases [31]. Next, considering the predicted miRNA-mRNA/miRNA-lncRNA pairs and expression patterns of genes, miRNA-mRNA/miRNA-lncRNA pairs with negative correlations were ultimately identified in the blue module. Finally, ceRNA networks were constructed by integrating the miRNA-lncRNA-mRNA relationships via Cytoscape 3.4.0 software.

Quantification of genes by real-time PCR

After anesthetization with 5% chloral hydrate, the mouse brain tissue was quickly removed. Then, the ACC was quickly frozen in liquid nitrogen and stored at -80°C. Total RNA was procured from the samples using RNA Simple Total RNA Extraction Kit (TIANGEN, DP419). For reverse transcription and cDNA generation, Revert Aid First Strand cDNA Synthesis Kit (Thermo Fisher Scientific, USA) was used. To determine the mRNA level, we used CFX96 Real-Time PCR Detection System thermocycler (Eppendorf, Germany) and Maxima SYBR Green/ROX qPCR Master Mix (2X) (Thermo Scientific, USA). Specific primer pairs (5′-3′) for analysis of target and reference genes were selected by the software Primer Blast and produced by Thermo Fisher Scientific (USA). The sequences of the relevant target genes are shown in Table 1. The normalized relative quantity of cDNA target genes was determined using the 2 − ΔΔCT method.

Table 1.

Sequences of primers used.

Gene	Primer	Sequence
Flt1	Forward	GAGATAGGACTGCTGAACTGCGAAG
Flt1	Reverse	CGGCGGGCGTATTTGGACATC
Slc38a2	Forward	CGTTGGCGTTGGCATTCAATAGC
Slc38a2	Reverse	GAAGAACAGCGGGATGGCAGAC
Bmpr1b	Forward	ACTCAAGGCAAGCCAGCAATCG
Bmpr1b	Reverse	TTGGGTGGGATGTCAACCTCATTTG
Pdgfra	Forward	TGAGATCGAAGGCAGGCACATTTAC
Pdgfra	Reverse	GCGGCAAGGTATGATGGCAGAG
Gng2	Forward	AACACCGCCAGCATAGCACAAG
Gng2	Reverse	GGGGTCAGCAGAGGGTCTTCC
Gapdh	Forward	TCGGTGTGAACGGATTTGGC
Gapdh	Reverse	TCCCATTCTCGGCCTTGACT

Statistical analysis

Statistical analysis was performed with R software 3.4.0 and GraphPad Prism 8.0 (GraphPad Software, CA, USA). Data are presented as mean ± standard error of the mean (SEM). Data normality was assessed using the Shapiro-Wilk test, and homogeneity of variance was confirmed by Levene’s test. For PWT and PWL, the data were analyzed using two-way analysis of variance (ANOVA) with Bonferroni post hoc correction. For EPM, NSF and FST, the data were analyzed using two-tailed unpaired Student’s t-tests. For RT-PCR, the data among different groups were analyzed using one-way ANOVA followed by Tukey’s post hoc test. P < 0.05 was considered statistically significant.

Results

Gene preprocessing and screening of DEGs related to NP-induced anxiodepression

A flowchart of the research objects was presented in Fig. 1. Twenty-four sample transcriptional profiles obtained from the GSE database were preprocessed and normalized. First, we identified DEGs in the 2-week cuff group compared with the 2-week sham group that displayed nociceptive hypersensitivity without detectable anxiodepressive-like behaviors, yielding 68 lncRNAs (Fig. 2A) and 774 mRNAs (Fig. 2B). Next, we identified DEGs in the 8-week cuff group compared with the 8-week sham group, which represented both nociceptive and anxiodepressive-like phenotypes, and identified 77 lncRNAs (Fig. 2C) and 1101 mRNAs (Fig. 2D). To identify the DEGs associated with anxiodepressive-like phenotypes rather than those associated with nociceptive phenotypes, 1058 DEGs (70 lncRNAs and 988 mRNAs) were selected among the 8-week DEGs after removing duplicated genes from the 2-week DEGs (Fig. 2E, F).

Fig. 1. Schematic diagram of this study.

ACC anterior cingulate cortex, DEGs differentially expressed genes, DE differentially expressed, WGCNA weighted gene correlation network analysis, GO gene ontology, KEGG Kyoto encyclopedia of genes and genomes, PPI protein-protein interaction, RT-PCR real time polymerase chain reaction.

Fig. 2. Differential gene expression analysis.

A, B Heatmaps demonstrating the expression of DElncRNAs and DEmRNAs between 2-week groups following sciatic nerve cuffing surgery. C, D 8-week groups following sciatic nerve cuffing surgery. E, F 8-week DEGs after removing duplicated genes from 2-week DEGs.

Potential interconnections between DEGs revealed by WGCNA

To further evaluate the interconnection between DEGs, WGCNA was applied to identify modules containing genes with similar expression patterns. First, cluster analysis was used to eliminate outliers based on the differential gene expression of the samples. There were no outliers detected in the samples (Fig. 3). A total of 70 lncRNAs and 988 mRNAs were gained to establish the co-expression network. The network topology with soft-threshold powers from 1 to 20 was analyzed, and β values of 14 were confirmed in the lncRNA/mRNA co-expression network (Fig. 4A). In the cluster dendrogram, genes with similar expression patterns were assigned to modules represented with specific colors (Fig. 4B). There was not much similarity between modules according to the calculated relevance between modules from the eigengene adjacency heatmap (Fig. 4C). A total of 5 modules were ultimately obtained. The gray module (24 mRNAs) showed that some genes were unfit for grouping into any other modules. Finally, the module-trait relationships were further explored (Fig. 4D). The turquoise module and yellow module were positively associated with anxiodepressive conditions, while the blue module and brown module were negatively related to clinical traits. By considering the number of DEGs in respective modules and correlations between modules and traits, the turquoise, blue and brown modules were selected for further research.

Fig. 3. Sample clustering to detect outliers.

There was no outlier detected in the samples.

Fig. 4. WGCNA identifying the modules of interest.

A Analysis of soft-threshold power in the WGCNA. Left: Scale independence analysis for various soft-thresholding powers. Right: Mean connectivity analysis for various soft-thresholding powers. B Cluster dendrogram based on dissimilarity measurements of genes. The branches represent modules of highly interconnected genes. C Hierarchical clustering of modules and heatmap of eigengene adjacencies. Red corresponds to a positive correlation with high adjacency, while blue corresponds to a negative correlation with low adjacency. D Matrix of module-trait relationships. Each row corresponds to a module eigengene, and each column corresponds to a trait. Each module includes a corresponding correlation value R (upper number) and p value (lower number).

Identification of the most relevant module through functional enrichment analysis

Furthermore, module-specific GO enrichment and module-specific KEGG pathway enrichment were analyzed. Notably, the turquoise module was likely related to signal transduction, metabolic pathways and cell differentiation (Fig. 5A, B), while the brown module was strongly correlated with the apoptotic pathway (Fig. 5C, D). Interestingly, the blue module was significantly related to nervous system pathology, such as glutamatergic synapses, GABAergic synapses, pathways of neurodegeneration, metabolic pathways and protein transport (Fig. 5E, F). These results indicate that DEGs in the blue module may be prominently connected with synaptic signaling transmission among neurons in the ACC when NP mice exhibit anxiodepressive phenotype. Therefore, the blue module was used as the module of greatest interest for further exploration of hub genes related to neuropathy of anxiodepression-like behavior.

Fig. 5. Analysis of module-specifi c GO enrichment and module-specifi c KEGG pathway enrichment.

GO enrichment analysis of module specificity and KEGG pathway enrichment analysis of the A, B turquoise module, C, D brown module and E, F blue module.

Identifying blue-module hub lncRNAs and mRNAs in blue module

Then, the lncRNA-mRNA-pathway networks in the blue module were constructed. In the entire blue module, we identified 7 core lncRNAs that are critical to the NP association: NONMMUT046097.2, NONMMUT034211.2, NONMMUT033516.2, NONMMUT032934.2, ENSMUST00000195192.1, XR_875301.1 and NONMMUT022131.2 (Fig. 6A, C). Among the 7 identified lncRNAs, NONMMUT046097.2 may be the most important upstream regulatory factor for managing the greatest number of mRNAs in this network. Next, we established a lncRNA-mRNA-pathway network to calculate the top hub mRNAs in the co-expression regulatory relationship (Fig. 6B, C). Subsequently, a PPI network was constructed in search of the internal links between proteins encoded by mRNAs in the blue module (Fig. 7A). According to the regulatory power of mRNAs in the lncRNA-mRNA-pathway network and PPI network, the hub mRNAs were identified as Flt1, Slc38a2, Bmpr1b, Pdgfra and Gng2. The Flt1 binding 3 lncRNAs and Pdgfra regulated by NONMMUT046097.2 were closely associated with signal transduction pathways involved in anxiodepression, such as the Rap1 signaling pathway, MAPK signaling pathway, calcium signaling pathway, Ras signaling pathway and PI3K-Akt signaling pathway. Notably, Slc38a2, which is associated with 4 lncRNAs, and Gng2, which is regulated by NONMMUT032934.2, were significantly associated with synaptic transmission and synaptic plasticity, including glutamatergic synapses, GABAergic synapses and cholinergic synapses. Bmpr1b was strongly related to axon guidance, which was regulated by 3 lncRNAs.

Fig. 6. The lncRNA-mRNA-pathway network analysis of the most enriched (blue) module.

A LncRNA-mRNA network analysis of the blue module. B LncRNA-mRNA-pathway network analysis of the blue module. C The red color represents upregulated genes, while the green color represents downregulated genes. The triangles indicate lncRNAs, while the circles indicate mRNAs.

Fig. 7. PPI analysis and ceRNA network in blue module.

A The mRNAs involved in the blue module were used to construct a PPI network. B The ceRNA network was constructed for the blue module. Triangles indicate lncRNAs, squares indicate miRNAs, and circles indicate mRNAs. The blue color represents downregulated genes, while the red color represents upregulated genes.

Construction of ceRNA network for blue WGCNA module

Extensive evidence has revealed that the ceRNA network plays a crucial role in elucidating interactions between various types of RNA. Briefly, lncRNAs can share miRNA response elements to influence miRNAs matching target mRNAs by regulating gene expression. Considering that the WGCNA modules of interest mainly included lncRNAs and mRNAs with positive correlations, miRNAs might be negatively correlated with lncRNAs and mRNAs. To identify a core regulatory network with the top hub genes, we constructed a ceRNA network. There were 5 top hub mRNAs (Flt1, Slc38a2, Bmpr1b, Pdgfra and Gng2), 3 top hub lncRNAs (NONMMUT046097.2, XR_875301.1, NONMMUT034211.2) and 40 miRNAs in the ceRNA network of the blue module (Fig. 7B). Among the three key lncRNAs, NONMMUT046097.2 interacts with 32 miRNAs/mRNAs, whereas XR_875301.1 interacts with only 7 genes, and NONMMUT034211.2 interacts with 3 genes. Notably, the top hub lncRNAs and mRNAs were associated with the signal transduction pathways of anorexia and synaptic plasticity. Therefore, the ceRNA network in the blue module may play a key role in NP-induced anxiodepression-like behavior.

Validation of molecular changes in the ACC of NP mice

We assessed mechanical and thermal hypersensitivity in cuff-model mice at 2-, 4-, 6-, and 8-weeks post-surgery, and evaluated anxiety- and depression-like behaviors at week 8 (Fig. 8A). Behavioral pain tests revealed that compared with the sham group, the ipsilateral PWT and PWL were significantly reduced in the cuff group at all time points measured (2, 4, 6, and 8 weeks) (Fig. 8B, C). In the EPM, the 8-week cuff group exhibited a significant reduction in both the number of entries and time spent in the open arms (Fig. 8D–G), indicating anxiety-like behavior. Furthermore, results from the NSF test and FST suggested that mice in the 8-week cuff group displayed significant anxiety- and depression-like behaviors (Fig. 8H, I). Finally, the expression of the hub genes in the ACC of NP mice was determined by RT-PCR. Compared with those in the 8-week sham group, the expression levels of Flt1 (Fig. 8J), Slc38a2 (Fig. 8K), Bmpr1b (Fig. 8L) and Pdgfra (Fig. 8M) were significantly increased in the 8-week cuff group, whereas the expression of Gng2 (Fig. 8N, O) was dramatically decreased in the 8-week cuff group. However, the expression of these genes did not change significantly following 2 weeks of sciatic nerve cuff surgery compared with that in the 2-week sham group (Fig. 8J–O). These results were in accordance with the results of the DEG analysis.

Fig. 8. Behavioral changes and expression levels of the hub mRNAs in mice.

ATime schedule for behavioral experiments on mice. B The ipsilateral PWT at each time point (0, 2, 4, 6, and 8 weeks) (n = 8). C The ipsilateral PWL at each time point (0, 2, 4, 6, and 8 weeks) (n = 8). D Trajectory of 8-week sham group in EPM experiment. E Trajectory of 8-week cuff group in EPM experiment. F The number of entries in the open arms of 8-week sham group and cuff group (n = 8). G The time spent in the open arms of 8-week sham group and cuff group (n = 8). H The latency of feeds in 8-week sham group and cuff group (n = 8). I The immobility time of 8-week sham group and cuff group (n = 8). J Flt1 mRNA levels in the ACC (n = 4). K Pdgfra mRNA levels in the ACC (n = 4). L Bmpr1b mRNA levels in the ACC (n = 4). M Slc38a2 mRNA levels in the ACC (n = 4). N Gng2 mRNA levels in the ACC (n = 4). O Gng2 mRNA levels in the ACC. (O) Legends of two groups. *P < 0.05, **P < 0.01, ***P < 0.001 vs. the sham group. ACC anterior cingulate cortex.

Discussion

In the present study, we constructed a gene co-expression network and ceRNA network according to DEGs in the ACC of mice with NP-induced anxiodepressive-like phenotypes. First, transcriptome data from the GSE92718 dataset were used to statistically identify the differentially expressed mRNAs and lncRNAs, which were selected among the 8-week DEGs by removing duplications from the 2-week DEGs. Next, WGCNA was performed to explore co-expression modules, and the most important module (blue module) related to NP-induced anxiodepression was identified. Then, we elucidated the lncRNA-mRNA pathway network within the blue module, which primarily comprised 7 fundamental lncRNAs. The 5 hub mRNAs (Flt1, Slc38a2, Bmpr1b, Pdgfra and Gng2) in both the lncRNA-mRNA-pathway network and the PPI network might function as core regulatory factors in the anxiodepression induced by NP. Then, a ceRNA network was constructed to identify a pivotal regulatory network integrating the miRNA-lncRNA-mRNA relationships with the top hub genes. Finally, the expression of the top hub mRNAs was verified by RT-PCR.

Our findings are particularly relevant for understanding human chronic pain conditions that exhibit a high prevalence of comorbid affective disorders. Centralized pain disorders such as fibromyalgia are characterized by augmented central nervous system processing of pain and a high overlap with anxiety and depression, sharing potential mechanisms of central sensitization and glial dysregulation with our model. Similarly, chronic post-traumatic neuropathic pain is also frequently comorbid with post-traumatic stress disorder and major depression, representing another patient population where the ACC-centered molecular pathways identified here might be critically involved. Thus, while based on a murine model, our study provides molecular insights that could inform future research into these specific, debilitating human pain-affect comorbidities.

The DEGs extracted from the microarray data could be useful biomarkers for identifying psychiatric phenotypes [15, 16, 32]. However, mounting evidence has suggested that individual genes are partially involved in the pathogenesis of psychiatric disorders [33–35]. Importantly, extensive previous research has shown that complex mental phenotypes are closely related to an interconnection between genes and intricate co-expression networks [17, 20, 21]. WGCNA is a relatively innovative and comprehensive algorithm based on gene correlation, and recently, it has been used to detect modules, construct gene networks, and identify hub genes [19]. The WGCNA algorithm revealed 7 genes, including HNRNPA2B1, IL10, and RAD51, which may have potential as biomarkers for PPD diagnosis [17]. Another study identified key pathways involved in major depressive disorder (MDD), which are significantly associated with autophagy and cellular immune function [20]. It is well known that abnormal synaptic transmission mediates the development and maintenance of psychiatric disorders [36, 37], but few studies have predicted synaptic-related biomarkers of anxiety and depression via WGCNA. Interestingly, we found that there was a highly correlated module (blue module) that was clearly interrelated with psychiatric pathology. According to the comprehensive analysis, DEGs in the blue module were mainly enriched in synaptic function, the MAPK signaling pathway, the calcium signaling pathway, the Ras signaling pathway and metabolic pathways.

Disordered gene expression regulating synaptic function leads to changes in synaptic plasticity, which is the basis of psychiatric disorders [37]. In our study, we suggested that glutamatergic synapses and GABAergic synapses in the blue module were strongly related to NP-induced anxiodepression. Maintaining a proper balance of excitatory and inhibitory synapses is vital for the normal function of neuronal circuits [38]. He S et al reported that abnormally increased glutamatergic projections mediate NP-induced depressive behaviors in a rodent model and that inhibiting glutamatergic neurons in the nucleus of the solitary tract (NTS) relieves depression but not hyperalgesia [39]. In addition, both preclinical and clinical studies have shown that GABAergic synapses are disrupted in individuals with depressive disorder. When administered to rodents, GABA antagonists can induce depressive-like behaviors, while GABA agonists have a relieving effect [40]. Interestingly, the number of inhibitory GABA_A receptors in the brain may not decrease in a timely manner following pregnancy, which may be one of the most critical causes of PPD [41]. However, Narayan GA and colleagues provided convincing evidence that the glutamate to GABA ratio is not meaningfully correlated with depression severity or changes in depression severity after treatment [42]. Thus, the intrinsic connection between glutamatergic synapses and GABAergic synapses in depressive disorders still needs further exploration. Long-lasting adaptive changes in signaling pathways are involved in the remodeling of neuronal and synaptic plasticity, which is critical for depression-like behavior [43]. Undoubtedly, the MAPK signaling pathway plays a critical role in various depression-like behaviors [44]. MAPKs are mainly expressed in postmitotic neurons, where they respond to synaptic input changes and regulate neuronal activity and synaptic plasticity [45]. Moreover, inhibiting MAPK signaling activation in microglia could alleviate downstream inflammatory cytokine levels as a therapeutic strategy for PPD treatment [46]. The best-understood pathway upstream of the MAPK pathway is Ras, an evolutionarily conserved signaling cascade [47, 48]. More importantly, the Ras-MAPK signaling pathway modulates synaptic functions and is closely related to the pathogenesis of depression [48]. Wang T et al suggested that chronic diterpene ginkgolide therapy provides an antidepressant effect through the Ras-MAPK signaling pathway. Therefore, given the role of the Ras-MAPK signaling pathway in depression-like behaviors, it is predicted that Ras-MAPK signaling will be a prospective target for the clinical treatment of NP-induced anxiodepression. Recent findings also suggest that the calcium signaling pathway is the direct origin of synaptic plasticity [49, 50]. Dendritic spines allow calcium influx through calcium signaling pathways, ultimately causing long-term plasticity [49]. Ca_v1.2 expression and L-type calcium current amplitude are increased in an animal model of depression induced by chronic restraint stress (CRS), whereas Ca_v1.2^+/− mice have been reported to exhibit an antidepressant-like phenotype [50]. Taken together, the DEGs identified in the blue module may be most closely related to NP-induced anxiety and depression.

Within the entirety of the blue module, 7 pivotal lncRNAs have been characterized as indispensable factors in the association between lncRNAs and NP-induced anxiodepression, namely, NONMMUT046097.2, NONMMUT034211.2, NONMMUT033516.2, NONMMUT032934.2, ENSMUST00000195192.1, XR_875301.1, and NONMMUT022131.2. The lncRNA-mRNA-pathway regulatory network incorporating these 7 lncRNAs and 70 mRNAs was constructed. Among the 7 identified lncRNAs, NONMMUT046097.2 is thought to play a crucial role, as it governs the greatest number of mRNAs within the ceRNA network. It is noteworthy that none of the 7 identified lncRNAs have been previously explored in the context of anxiety or depression. Such a discovery may serve as an essential catalyst for future investigations in the pathobiology of NP-induced anxiodepression. What’s more, we found that the target genes (e.g., Flt1, Slc38a2) were also regulated by miRNAs. In the field of pain research, it has been demonstrated that the expression of miR-30b-5p is diminished in the spinal cord of rats with chronic constrictive injury. While overexpression of miR-30b-5p can reduce neuroinflammation in the spinal cord, thereby alleviating neuropathic pain in rats [51]. Meanwhile, Tan et al found that the absence of miR-34 mediated neuroinflammation in spinal cord and was involved in the inflammatory pain induced by complete Freund’s adjuvant in rats [52]. Additionally, it has been suggested that miR-207 can alleviate neuroinflammation and improve depression symptoms in mice with chronic stress by inhibiting the activation of astrocytes [53].

Then PPI and ceRNA network analysis were employed to further narrow the attention to specific genes in the blue module. Five mRNAs (Flt1, Slc38a2, Bmpr1b, Pdgfra and Gng2) and 3 lncRNAs (NONMMUT034211.2, NONMMUT046-97.2 and XR_875301.1) were considered hub nodes due to their greater connections with other nodes. In addition, the 5 hub mRNAs were validated by RT-PCR. Finally, aberrant expression of Flt1, Slc38a2, Bmpr1b, Gng2 and Pdgfra in the ACC was successfully verified in the 8-week cuff group but not in the 2-week cuff group. Flt1 is a gene encoding vascular endothelial growth factor receptor 1 (VEGFR1), which mediates the neuroprotective function of VEGF under pathological conditions [54]. Nunes F et al suggested that the Flt1 polymorphism rs7993418 is related to decreased depression severity by performing a genotype analysis of patients with depression [55]. The mechanism of the protective role of FLT1 in depression has not been well elucidated. Some investigators have considered that the antidepressant effects of FLT1 may be due to its ability to regulate synaptic plasticity and neuronal or glial protective factors [54, 56]. In addition, angiogenesis, inflammation and pain may be linked through pathophysiologic effects of VEGF in various diseases [57]. Similarly, Flt1 is also involved in neuroinflammation [58]. Higher level of FLT1 in cerebrospinal fluid may increase the risk of Alzheimer disease [59]. SLC38A2 (also termed SNAT2) acts as a glutamine transporter for maintaining the balance of the glutamate-glutamine cycle [60]. An increase in SNAT2 could increase glutamine influx into neurons and subsequently contribute to alterations in synaptic plasticity. Interestingly, clinical research has shown that SNAT2 expression is increased in the ACC of depressed patients who have committed suicide compared to that in depressed patients who have died from nonsuicide-related reasons [61]. Thus, the SNAT2-mediated glutamate-glutamine cycle may be a key pathway in NP-induced anxiety and depression. BMPR1b is a receptor subunit that binds to bone morphogenetic proteins (BMPs), which are members of the TGFβ superfamily [62]. Previous research has demonstrated that BMPR1a plays a decisive role in increasing the numbers of both oligodendrocytes and interneurons [63]. BMPR1b KO mice exhibit attenuated glial scarring in the chronic stages following spinal cord injury [64]. Moreover, it has recently been suggested that BMPR1b similarly protects against neuronal loss in neurodegenerative diseases [62]. Although the mechanism of BMPR1b in depression has still not been elucidated, increased BMPR1b may play a protective role in NP-induced anxiodepressive-like phenotypes according to previous studies. GNG2 is a G protein γ2 subunit that plays significant roles in cellular responses to external signals [65]. Pdgfra is a gene encoding platelet-derived growth factor receptor α (PDGFRα), and its activation can promote cell growth, division, and proliferation [66]. Activated PDGFRα can reduce the proliferation of glial cells after brain injury, accompanied by the inhibition of axonal neogenesis. Our lncRNA-mRNA pathway network analysis revealed that Gng2 was predominantly involved in both the glutamatergic and GABAergic synapse pathways, which was similar to the findings of previous studies of predicting biomarkers involved in MDD by machine learning models or transcriptomics and sequencing analysis [67, 68]. Additionally, PDGFRα, as well as SLC38A2, may be involved in cancer immunity [69]. The upregulation of Pdgfra expression can promote cancer-related NP via paracrine inflammatory factors [70]. Taken together, these findings indicate that NP-induced anxiodepression is strongly associated with several genes regulating synaptic plasticity and glial proliferation. However, further studies are needed to determine the underlying mechanisms involved.

Beyond elucidating pathogenesis, our findings point to tangible therapeutic possibilities. The identification of Flt1, Slc38a2, Bmpr1b, Pdgfra, and Gng2 as hub genes not only underscores their central role in NP-induced anxiodepression but also nominates them as compelling candidates for therapeutic intervention. Specifically, Flt1 (VEGFR1) represents a dual target for modulating neuroinflammation and aberrant angiogenesis, processes implicated in both pain and depression [71]. Pdgfra emerges as a critical node, with evidence showing that its inhibition in the hippocampus can concurrently alleviate pain and depression-like behaviors by suppressing the JAK2/STAT3 pathway [72]. Bmpr1b, a key receptor in the BMP signaling pathway, is implicated in nociceptive sensitization and has been proposed as a common pathway for antidepressant actions [73]. Gng2 may influence comorbid symptoms by regulating GABAergic signaling and inflammatory responses [65]. Although direct evidence for Slc38a2 in pain-depression comorbidity is limited, its role in glutamatergic transmission positions it within the crucial glutamate-glutamine cycle, a pathway frequently disrupted in affective disorders [74].

These potential targets intersect with known pathological mechanisms, suggesting opportunities for repurposing existing strategies or developing novel ones. For instance, anti-VEGF therapies targeting the FLT1 pathway, or modulating its associated neuroinflammatory pathways with existing drugs such as the antidepressant fluoxetine, or JAK/STAT inhibitors aimed at downstream of PDGFRα, warrant investigation for their dual efficacy [75]. Furthermore, the constructed ceRNA network, particularly the pivotal lncRNAs such as NONMMUT046097.2, opens a new avenue for RNA-targeted therapeutics to modulate the entire regulatory axis. Moving forward, validating these targets in vivo and exploring multidisciplinary approaches—combining neuromodulation with pathway-specific pharmacological agents—will be essential steps toward translating these molecular insights into effective treatments for patients suffering from NP and its debilitating emotional comorbidities.

While our study provides a comprehensive molecular landscape of NP-induced anxiodepression in a mouse model, we acknowledge that a direct comparison with human data is currently lacking. This limitation is primarily due to the scarcity of publicly available transcriptomic datasets from the ACC of patients specifically diagnosed with pain-induced depression. Nevertheless, the key modules and hub genes identified here present a set of high-priority candidates for future validation in relevant human. Such translational studies will be crucial to determine the conservation of these mechanisms and to assess their potential as diagnostic biomarkers or therapeutic targets in the clinical setting.

Conclusions

In summary, via bioinformatic approaches, we identified hub genes and ceRNA networks interrelated with NP-induced anxiodepression phenotypes. Five mRNAs (Flt1, Slc38a2, Bmpr1b, Pdgfra and Gng2) and 3 lncRNAs (NONMMUT034211.2, NONMMUT034211.2 and XR_875301.1) were considered to be key factors. Finally, a core ceRNA axis was obtained to integrate the miRNA-lncRNA-mRNA relationship in the anxiety and depression induced by NP. The present results provide potential targets and a deeper understanding of NP-induced anxiodepression. However, additional studies are required to fully clarify the molecular mechanisms involved in the development of NP-induced anxiodepression.

Author contributions

YH: Conceptualization, Formal analysis, Funding acquisition, Writing - original draft. YX: Conceptualization, Formal analysis, Visualization, Writing - original draft. FX: Conceptualization, Formal analysis, Visualization, Writing - original draft. XS: Formal analysis, Conceptualization, Methodology. FZ: Formal analysis, Visualization, Data curation. MX: Formal analysis, Visualization, Data curation. YL: Formal analysis, Visualization, Data curation. WZ: Methodology, Visualization, Data curation. XW: Conceptualization, Formal analysis, Funding acquisition, Writing - original draft. JY: Funding acquisition, Conceptualization, Supervision, Writing - review & editing.

Data availability

Data will be made available on request.

Competing interests

The authors declare no competing interests.