Effects of FGF4/HIF-1α axis-mediated neuronal glycolysis on neuropathic pain

Abstract

The pathogenesis of neuropathic pain is complex, and effective treatment methods are lacking in clinical practice. Recent studies have shown that glucose metabolism reprogramming may be involved in the process of neuropathic pain, but its role and molecular regulatory mechanisms in neuropathic pain are still unclear. In this study, a rat model of chronic constriction injury of the sciatic nerve (CCI) was established, and the pain threshold was evaluated through behavioural analysis. Morphological staining, transmission electron microscopy, transcriptome sequencing, Western blotting, immunofluorescence staining, ELISA, and the whole-cell patch clamp technique were used to systematically observe neuropathological changes, identify differentially expressed genes and associated pathways, and measure the expression levels of glycolysis-related indicators and the key regulatory factor fibroblast growth factor 4 (FGF4). The results showed that the pain threshold of rats decreased and that the structure of sciatic nerve tissue was damaged after CCI. Transcriptome sequencing of the sciatic nerve showed a significant increase in the expression levels of glycolysis-related indicators. Subsequent experiments confirmed that FGF4 expression was downregulated in the sciatic nerve and spinal dorsal horn after CCI, whereas the expression of hypoxia-inducible factor-1α (HIF-1α) and its key downstream glycolytic enzymes was upregulated, accompanied by increased levels of lactic acid and proinflammatory cytokines and decreased ATP levels. The spinal dorsal horn exhibited both synaptic structural abnormalities and neuronal hyperexcitability. Inhibiting HIF-1α alleviated pain and suppressed glycolysis, whereas the overexpression of FGF4 specifically reversed the increase in HIF-1α expression, inhibited neuronal glycolysis, and reduced neuroinflammation and central sensitization, ultimately effectively relieving pain. This study reveals the core role of the FGF4/HIF-1α-mediated regulation of neuronal glycolysis in neuropathic pain, providing a new theoretical basis and experimental evidence for a deeper understanding of the metabolic mechanism of neuropathic pain and the development of targeted treatment strategies.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s10194-026-02362-7.

Introduction

Neuropathic pain is a chronic pain syndrome caused by damage to or disease of the somatosensory nervous system. Its clinical manifestations include spontaneous pain, hyperalgesia, and trigger-induced pain, which seriously affect patients’ quality of life and social function [1, 2]. Although approximately 7%–10% of the global population is affected by neuropathic pain, existing first-line drug treatment strategies have significant limitations, including poor efficacy and significant side effects [3]. Therefore, an in-depth analysis of the core mechanisms of neuropathic pain and the identification of safe and effective intervention targets are urgently needed.

Chronic constriction injury (CCI) of the sciatic nerve, as a classic method for inducing peripheral nerve injury-related pain, can stably recapitulate the behavioural and pathological characteristics of human neuropathic pain (such as neuroinflammation, glial cell activation, and peripheral/central sensitization), providing a reliable research platform for exploring pain mechanisms [4]. This study focused on CCI model rats and involved high-throughput transcriptome sequencing analysis of injured sciatic nerve tissue. In recent years, increasing evidence has suggested that after peripheral nerve injury, sensory neurons not only experience abnormal signal transduction, but also undergo profound metabolic reprogramming [5]. The sequencing results of this study also showed that in the context of neuropathic pain, the expression of glycolysis-related genes such as hypoxia-inducible factor-1α (HIF-1α), hexokinase 3 (HK3), lactate dehydrogenase A (LDHA), phosphoglycerate mutase 1 (PGAM1) and phosphoglycerate kinase 1 (PGK1) tended to significantly increase, and ‌Gene Ontology (GO) analysis revealed the significant enrichment of neurons and synapses, strongly suggesting that an abnormal increase in neuronal glycolysis may be among the key mechanisms of CCI-induced pain.

The abnormal activation of glycolysis, which is the core pathway of cellular energy metabolism, may be closely related to changes in neuronal overexcitation, the inflammatory response, and synaptic plasticity [3]. Studies have shown that the final product of glycolysis, lactic acid, not only acts as a signalling molecule to directly activate acid-sensitive ion channels on nociceptors but also promotes neuroinflammation and amplifies pain signals [5]. Studies have also indicated that the metabolic intermediates produced during glycolysis are precursors for the synthesis of amino acids, lipids, and nucleotides, providing a material basis for abnormal repair of damaged nerves and synaptic remodelling, which may maintain the chronicity of pain [6]. Although these findings suggest that increased neuronal glycolysis may be the core mechanism involved in pain, the global changes and precise upstream regulatory mechanisms of this process in neuropathic pain models are still unclear.

In this study, we focused on neuronal glycolysis and clarified the pathological mechanism of neuropathic pain by systematically detecting changes in the activity of key enzymes, the accumulation levels of metabolites such as lactic acid and ATP, and their dynamic associations with the development of pain behaviour. After confirming the comprehensive activation of glycolysis, we further focused on revealing its upstream regulatory network. We explored the core role of the fibroblast growth factor 4 (FGF4)/HIF-1α signalling pathway in driving glycolysis and pain behaviour through molecular interventions, such as intrathecal injection of FGF4-AAV to overexpress FGF4 and inhibition of HIF-1α expression using HIF-1α siRNA. The aim was to elucidate the key molecular bridge connecting nerve damage, metabolic reprogramming, and pain behaviour, providing solid experimental evidence and promising therapeutic targets for novel analgesic strategies.

Methods

Antibodies

Rabbit anti-FGF4, rabbit anti-HIF-1α, rabbit anti-hexokinase 2 (HK2), rabbit anti-LDHA, rabbit anti-PGK1, rabbit anti-GAPDH, rabbit anti-Tubulin, rabbit anti-S100 calcium-binding protein β (S100β), rabbit anti-brain-derived neurotrophic factor (BDNF) and rabbit anti-postsynaptic density 95 (PSD-95) antibodies were purchased from Affinity Biosciences (USA). A mouse anti-NeuN antibody was purchased from Abcam (UK). Mouse anti-Iba1 and mouse anti-GFAP antibodies were purchased from Servicebio (China). Rabbit anti-PGAM1, mouse anti-HK3 and mouse anti-neurofilament 200 (NF200) antibodies were purchased from Proteintech (China). HIF-1α siRNA (HIF-1α-siRNA) and the negative control siRNA (NC-siRNA) were obtained from Genepharma (China). Adeno-associated virus (AAV) 9 vectors for FGF4 overexpression (pAAV-CMV> Kozak-Rat Fgf4 CDS [NM_001389212.1]-3xFLAG-P2A-EGFP, FGF4-AAV) and its empty control (pAAV-CMV> Kozak-EGFP, NC-AAV) were designed and manufactured by Cyagen (China).

Animals

Adult male Sprague-Dawley rats (age 6–7 weeks, 220–250 g) were purchased from the Laboratory Animal Center of Wenzhou Medical University. We housed the rats in an environment at 22–24℃ with regular alternation of day and night and ensured that they were provided sufficient water and food every day. Our studies were approved by the Animal Research Committee of Wenzhou Medical University and performed in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals.

CCI

The rats were anaesthetized with 2% sodium pentobarbital (30 mg/kg, i.p.). The sciatic nerve was fully exposed by isolating the skin and muscle in the right thigh of each rat. Silk thread (4 − 0) was used to ligate the sciatic nerve 3 times, with a spacing of 1 mm. The degree of tightness of the ligature was dependent on the slight twitch of the calf muscle or toe when knotting. The sciatic nerve in the sham-operated rats was exposed for 2–3 min without ligature.

Experimental design

First, based on a random number table, 88 rats were randomly divided into two groups: the sham group (n = 44, S) and the rats underwent CCI surgery group (n = 44, M), to clarify the pain behavioral characteristics and histopathological basis caused by CCI. The main research direction and key regulatory genes involved were determined by transcriptome sequencing of the sciatic nerve.

Subsequently, to verify whether HIF-1α has a regulatory effect on neuronal glycolysis in CCI model rats, 54 rats were subsequently randomized into three groups using a random number table method: the M group (n = 18), the M + NC-siRNA group (n = 18), and the M + HIF-1a-siRNA group (n = 18). The sequences of the siRNAs were as follows (5′–3′): HIF-1a-siRNA, sense GAGCUUUGGAUCAAGUUAATT, antisense UUAACUUGAUCCAAAGCUCTT; and NC-siRNA, sense UUCUCCGAACGUGUCACGUTT, antisense ACGUGACACGUUCGGAGAATT. HIF-1a-siRNA and NC-siRNA were dissolved in diethyl pyrocarbonate in water (20 pmol/µl). One week before CCI, intrathecal catheterization was performed on each group of rats. Starting on the 8th day after CCI, the rats in the M + HIF-1a-siRNA group and M + NC-siRNA group were injected intrathecally with the HIF-1a siRNA (10 µl) or NC-siRNA (10 µl), respectively, every 24 h for 7 consecutive days. The rats in the M group were also injected intrathecally with equal volumes of vehicle at the same time.

Finally, we further validated the critical role of FGF4 in regulating HIF-1a-mediated neuronal glycolysis by intrathecal injection of FGF4 overexpressing adeno-associated virus (pAAV-CMV> Kozak-Rat Fgf4 CDS [NM_001389212.1]-3xFLAG-P2A-EGFP, FGF4-AAV). The serotype of the virus was type 9, the promoter was cytomegalovirus (CMV), and the transgene cassette was Rat Fgf4 CDS [NM_001389212.1]-3xFLAG. Meanwhile, pAAV-CMV> Kozak-EGFP (NC-AAV) served as a negative control. 87 rats were randomly divided into three groups using a random number table method: the M group (n = 29), the M + NC-AAV group (n = 29), and the M+FGF4-AAV group (n = 29). Five weeks before CCI, the rats in the M+FGF4-AAV group and M + NC-AAV group were intrathecally injected with FGF4-AAV (5.0 × 10¹² vg/mL, 20 µl) or NC-AAV (5.0 × 10¹² vg/mL, 20 µl), respectively. The rats in the M group were also injected intrathecally with equal volumes of vehicle at the same time. The injection process involves slowly inserting a 50 µL microinjector along the L4-L5 interspinous space after anesthesia in rats. When there was a slight sense of disappointment and a clear tail tremor response was seen, it was considered successful entry into the intrathecal space. Subsequently, AAV was slowly injected at a constant speed for no less than 2 min. After the injection was completed, leave the needle for 1 min and then slowly withdraw. Afterwards, observe their activities, diet, and other conditions daily.

Transcriptome sequencing

On the 15th day after CCI, the right sciatic nerves of rats in the S and M groups were collected, and total RNA was extracted from the tissues using TRIzol reagent. Afterwards, the samples were fragmented and reverse transcribed, a library was constructed, and genes were sequenced using the Illumina Novaseq 6000 platform. Use HISAT2 (2.1.0) software to align the filtered reads to the reference genome to obtain their localization information. The gene expression level was normalized by fragments per kilobase of transcript per million mapped fragments (FPKM). PCA analysis were performed using R (v 3.2.0) to evaluate the biological duplication of samples. DESeq2 (1.22.2) software was used to standardize the gene count for each sample, calculate the fold difference, and perform a negative binomial distribution test for significance. Differentially expressed protein-coding genes were subsequently identified based on significance and the fold change. The parameters for differentially expressed genes (DEGs) were a threshold of q value (FDR-adjusted p) < 0.05 and |log2Fold Change|>1. GO analysis and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis were subsequently conducted on the DEGs. Afterwards, a protein–protein interaction (PPI) network of the DEGs was constructed using the STRING online database. Import the original interaction list into Cytoscape 3.10.0 software and analyze the PPI network using four algorithms: Maximum Clique Centrality (MCC), Edge Percolated Component (EPC), Density of Maximum Neighborhood Component (DMNC) and Degree. Select the top 5 genes ranked by each algorithm and visualize the PPI network of the genes and adjacent genes selected by each algorithm. Subsequently, the jVenn tool was used to perform intersection analysis on the gene sets obtained from the four algorithms and draw a Venn diagram to screen out common core genes. Notably, due to the ID mapping mechanism of STRING that merges functionally redundant homologous genes, the interaction information for LOC102549173 was integrated into its representative homolog LOC684762 for network visualization. The transcriptome sequencing data has been submitted to Gene Expression Omnibus (GEO) with accession number GSE317107.

Mechanical withdrawal threshold (MWT)

The MWT was determined with an electronic von Frey anaesthesiometer (IITC Life Sciences, CA, USA) before CCI and 3, 5, 7, 10, 12 and 14 days after CCI. The rats were placed in the testing apparatus and allowed to adapt for 20 min. When the rats were quiet, the probe was gradually applied vertically and forcefully to the root of the third and fourth toes of the hind limbs on the operated side. When the rats had positive reactions, such as foot lifting, foot licking and foot withdrawing, the reading (g) displayed by the tester at this time was recorded. The measurement was repeated 5 times at intervals of 5 min, the maximum and minimum values were removed from the measured values, and then the average value was taken as the final MWT value.

Thermal withdrawal latency (TWL)

The TWL was measured using an A37370 plantar tester (Ugo Basilee, Milan, Italy) before CCI and 3, 5, 7, 10, 12 and 14 days after CCI. The rats were placed in the testing apparatus and allowed to adapt for 20 min. After they were quiet, they were irradiated with infrared light emitted by the instrument at the root of the third and fourth toes of the affected side of the hind limbs of the experimental rats. If a rat exhibited evasive leg-raising movements, the heat source was automatically shut off, and the reaction time (s) displayed on the instrument was recorded. This process was repeated 5 times with an interval of 5 min, the maximum and minimum values were removed from the measured values, and then the average value was taken as the final TWL value.

Nissl staining

On the 15th day after CCI, the right sciatic nerves of the rats were removed and fixed with 4% paraformaldehyde at room temperature for 48 h before being embedded in paraffin. The tissue was then sliced, deparaffinized and rehydrated, and incubated with Nissl staining solution at 37℃ for 5 min. After dehydration and transparency, the slices were sealed with neutral resin, and the staining results were observed under an optical microscope.

Quantitative real-time PCR (qRT-PCR)

On the 15th day after CCI, total RNA was extracted from tissue samples using TRIzol reagent according to the manufacturer’s instructions and reverse transcribed to synthesize cDNA. qRT-PCR detection was subsequently performed using SYBR Green PCR Mastermix. The specific primers and their sequences used were as follows (5′–3′): FGF4 sense TACCTGCTGGGCCTCAAAAG, antisense TCCGAAGATGCTCACCACAC; HK3 sense TGGGGCTCCTTCTATGACGA, antisense CTCTGAGCACCAGGAACCAG; PGK1 sense GGGTGTGAATCTGCCACAGA antisense AGCTCCTCCCAAGATAGCCA; PGAM1 sense ACGATCTTACGATGTCCCGC, antisense AGGCTCTCACAGGAGGGTAG; LDHA sense CAGTCCACAGTGCAAACTGC, antisense GAACTCCCAGCCTTTCTCCC; HK2 sense ATGGAGTGGGGAGCATTTGG, antisense GCCGCTGATCATCTTCTCGA; HIF-1α sense GTGACCGTGCCCCTACTATG, antisense CGTAACTGGTCAGCTGTGGT and the housekeeping gene beta-actin (ACTB), sense TGTCACCAACTGGGACGATA, antisense GGGGTGTTGAAGGTCTCAAA. Three technical replicates were set up for each sample in the qRT-PCR reaction, and the average Ct value of the three replicates was taken for subsequent relative quantitative calculations. Relative expression was analysed using the 2^-ΔΔCt method.

Whole-cell patch clamp recording

On the 15th day after CCI, L4-6 spinal cord segments were collected from each group of rats, and the spinal cord segments were sliced into 300 μm sections using a vibratome. After the sections were revived for 0.5 h, the samples were transferred to the recording slot. Afterwards, artificial cerebral spinal fluid (ACSF) was circulated and perfused using a peristaltic pump while a 95% O₂/5% CO₂ mixture was continuously introduced. The main components of ACSF were as follows (in mM): 11 glucose, 3 KCl, 126 NaCl, 2.5 CaCl₂, 1.3 MgCl₂, 1.25 NaH₂PO₄ and 26 NaHCO₃. Then, a glass electrode was drawn using a microelectrode puller. After the liquid was injected into the electrode, it was connected to a patch clamp system, and the impedance of the electrode tip was tested. Fast electrical compensation was performed after cell sealing and membrane rupture, and spontaneous excitatory postsynaptic currents (sEPSCs) were recorded in the substantia gelatinosa (SG) neurons using patch pipette electrodes with a resistance of 5–10 MΩ. The main components of the patch pipette solution used to record sEPSC include (in mM): 0.5 CaCl₂, 135 K-gluconate, 5 ATP-Mg, 5 KCl, 5 HEPES, 5 EGTA and 2 MgCl₂. sEPSCs were recorded at a holding potential of -70 mV. After the neuron state stabilized, we input current through the glass electrode in a stepwise manner in whole-cell recording mode. At this time, the cell membrane potential increased accordingly, which triggered the generation of one or several action potentials (APs). Finally, the experimental data were analysed and organized using Mini Analysis and Clampfit software. This study only analyzed cells that maintained stable access resistance (< 30 MΩ) throughout the entire recording process. The access resistance and input resistance were continuously monitored throughout the entire experiment. If these parameters changed by more than 15%, the data was excluded. 2–3 neurons were recorded from each rat. To conduct statistical comparisons, the electrophysiological parameters (such as sEPSC frequency) of all neurons from the same rat were averaged and used as the unique representative value for that rat. The final inter group comparison was based on the number of animals (n = 4).

Western blotting

On the 15th day after CCI, the sciatic nerves and L4-6 spinal cord segments of each group of rats were completely homogenized in lysis buffer. The supernatant was collected, and the protein concentration was measured. After the separation gel and concentration gel were prepared, the proteins (30 µg per lane) were added to the wells. The samples were electrophoresed at a voltage of 120 V. The samples were then transferred to PVDF membranes, which were then incubated with milk for 2 h. The membranes were incubated with primary (FGF4, 1:1000, DF8948; HIF-1α, 1:500, AF1009; HK3, 1:5000, 67803-1-Ig; HK2, 1:500, DF6176; LDHA, 1:500, DF6280; PGAM1, 1:2000, 16126-1-AP; PGK1, 1:1000, DF6722; GAPDH, 1:5000, AF7021; Tubulin, 1:5000, AF7011) and secondary (Peroxidase Conjugated Goat anti-Rabbit IgG (H + L), 1:5000, BYE003; Peroxidase Conjugated Goat anti-Mouse IgG (H + L), 1:5000, BYE004) antibodies and developed in an exposure machine. We used the ChemiDoc XRS chemiluminescent imaging system to visualize the images. AlphaEaseFC 4.0 software was used to analyse the membranes. All data were expressed as the ratio of the grayscale values of the target protein bands to the grayscale values of GAPDH or Tubulin bands.

Immunofluorescence staining

On the 15th day after CCI, the sciatic nerves and L4-6 spinal cord segments of each group of rats were frozen and embedded in OCT compound. The embedded tissue blocks were placed in a frozen microtome for continuous sectioning at a thickness of 5 μm. The tissue was blocked with an appropriate amount of antibody blocking solution (PBS solution containing 10% normal goat serum and 0.3% Triton X-100) added dropwise. After incubations with primary and secondary antibodies, the nuclei were stained with DAPI for 10 min. Afterwards, an antifluorescence quencher was added to the sections, and the sections were observed under a microscope after sealing.

Transmission electron microscopy

On the 15th day after CCI, the sciatic nerve and spinal dorsal horn tissues from each group of rats were cut into several small strips using a blade. The tissues were fixed with glutaraldehyde. They were then placed in osmic acid and uranium acetate and incubated for 5 min. After gradient dehydration in acetone, the tissues were embedded in embedding solution. Ultrastructural changes in myelin sheaths and synapses were observed in semithin and ultrathin sections under a transmission electron microscope. The spinal dorsal horn tissue used for synaptic quantitative analysis was derived from 4 rats per group. Three ultra-thin sections were prepared and observed from the tissue blocks of each rat. Three fields of view were randomly selected from each slice, and ultimately 9 clear synaptic structures were analyzed for each rat. During statistical analysis, the average values of all tested synapses in each rat were first calculated, and then these average values were used for inter group comparison.

Determination of the lactic acid content

On the 15th day after CCI, the right sciatic nerve or L4-6 spinal cord segments of each group of rats were removed, and tissue samples were accurately weighed. The corresponding volume of physiological saline was added at a weight ratio (g)/volume (ml) = 1:9. After the tissues were cut into pieces, grinding beads were added and the tissues were mechanically homogenized under low-temperature conditions. The tissues were fully homogenized and centrifuged at 3000 rpm for 10 min. The supernatant was measured according to the instructions of the lactate content determination kit.

Determination of the ATP content

On the 15th day after CCI, the right sciatic nerve and L4-6 spinal cord segments of each group of rats were removed, and the tissues were accurately weighed and cut into pieces in EP tubes. Nine volumes of ATP reagent were added to the extract, and the samples were homogenized in a low-temperature grinder to prepare a 10% tissue homogenate. After the samples were boiled in water for 2 min, they were cooled with running water and centrifuged at 10,000 g and 4℃ for 10 min, after which the supernatant was removed and placed on ice. The ATP content of the supernatant was subsequently measured according to the instructions of the ATP content colorimetric test kit.

HE staining

On the 15th day after CCI, the right sciatic nerves of the rats in each group were collected and embedded in paraffin. The samples were sectioned in series, and then the slices were placed in xylene for dewaxing. The slices were subsequently placed in 100%, 100%, 80% and pure water for rehydration. The slices were stained with haematoxylin. After being washed, the slices were placed in hydrochloric acid ethanol. After being washed, the slices were placed in an eosin solution for staining. The slices were washed with water, dehydrated with graded ethanol solutions, and then placed in xylene. Finally, the slices were sealed with neutral gum.

ELISA

ELISA kits were purchased from Shanghai Boyun Biotech Co., Ltd., and the levels of IL-1β, IL-6 and TNF-α in the serum of the rats were detected according to the manufacturer’s instructions.

Statistical analysis

SPSS 23.0 statistical software was used to analyse the data. Prior to parametric analyses, data distributions were assessed for normality using the Shapiro-Wilk test, and no significant deviations from normality were detected. Homogeneity of variance was tested using Levene test. The data are presented as the mean ± SD. The pain threshold was analysed using repeated measures two-way analysis of variance (ANOVA) with Bonferroni’s post-hoc tests. Statistical significance between two groups was analyzed using Student’s t-test, and statistical significance between other multiple groups was analyzed using one-way ANOVA followed by Tukey’s post-hoc tests (the data satisfied homogeneity of variances) or Dunnett’s post-hoc tests (the data did not satisfy homogeneity of variances). A value of p < 0.05 was considered significant.

Results

CCI rats exhibited reduced pain thresholds and structural damage to sciatic nerve tissue

The pain threshold test results (Fig. 1A, B) revealed that the MWT and TWL of the rats continued to decrease within 7 days after CCI. On the 14th day after CCI, compared with those of S group, the MWT and TWL of the M group decreased (p < 0.01).

Fig. 1.

CCI reduced the pain threshold of rats and caused damage to the tissue structure of the sciatic nerve. (A) Thermal withdrawal latency (TWL). n = 15. Data were analyzed by repeated measures two-way ANOVA (Group × Day) with Bonferroni’s post-hoc tests. (B) Mechanical withdrawal threshold (MWT). n = 15. Data were analyzed by repeated measures two-way ANOVA (Group × Day) with Bonferroni’s post-hoc tests. (C) Observation of the ultrastructural changes of sciatic nerve myelin sheath under transmission electron microscopy. Scale bars, 10 μm, 2 μm. (D) HE staining. Scale bars, 100 μm, 20 μm. (E) Nissl staining. Scale bars, 50 μm, 20 μm. (F) Immunohistochemistry of NF200 (red) and S100β (green). Scale bars, 50 μm, 10 μm. (G), (H) Quantitative analysis of mean fluorescence intensity of S100β and NF200. n = 4. Data were analyzed using Student’s t-test. Columns represent the mean ± SD. ^$$p < 0.01 vs. the S group

The ultrastructural analysis of the sciatic nerve myelin sheath by transmission electron microscopy (Fig. 1C) revealed that the myelin sheath of the sciatic nerve in the S group presented a regular layered distribution, with a uniform, dense, and orderly arrangement, and that the axons maintained a complete shape with clear and distinguishable contours. The structure of the myelin sheath layer of the sciatic nerve in the M group was unclear, with folded and wrinkled layers, irregular shapes, and significant degenerative changes.

The morphological structure of the sciatic nerve on the affected side of the rats in each group was evaluated by performing HE staining (Fig. 1D). Longitudinal sections of the nerve showed that the sciatic nerve fibres in the S group were complete, the nerve fibres were arranged compactly, and the tissue structure was complete. Compared with those in the S group, the sciatic nerve fibres in the M group were scattered, the nerve bundle spacing was large, and the nerve structure was severely damaged.

Nissl staining (Fig. 1E) showed that the Nissl bodies of the sciatic nerve were stained light blue and appeared as fine granules. The structure of Nissl bodies in the S group was complete and regular, with uniform staining and a large quantity, whereas in the M group, Nissl bodies were loosely arranged and deformed, with nuclei displaced, uneven staining, and a significantly reduced quantity.

To further clarify the influence of CCI on sciatic nerve axons and myelin sheaths in rats, we measured NF200 and S100β expression using immunofluorescence (Fig. 1F-H). Compared with that in the S group, the immunoreactivity of NF200 and S100β in the sciatic nerve in the M group rats was significantly reduced (p < 0.01).

The pathological mechanism of CCI-induced neuropathic pain mainly involves immune inflammation and abnormal remodelling of neuronal synapses

Systematically studying changes in gene expression at the whole-genome level is beneficial for exploring the molecular mechanisms underlying the occurrence and development of neuropathic pain. We performed transcriptome sequencing on the affected sciatic nerve, and principal component analysis (PCA) showed significant separation of the sciatic nerve transcriptomes along PC1 in the S and M groups of rats, indicating extensive transcriptome differences in the sciatic nerve due to CCI (Fig. 2A). The sequencing results revealed that according to the parameters q < 0.05 and |log2FoldChange|>1, 5279 DEGs were identified between the M and S groups, of which 2795 were upregulated and 2484 were downregulated (Fig. 2B). Afterwards, we conducted enrichment analyses of the DEGs. The GO analysis (Fig. 2C) showed that the main enriched biological process terms were immune response, inflammatory response, response to lipopolysaccharide and chemokine-mediated signalling pathway. The main enriched cellular components were synapses and neuronal cell bodies. The enriched molecular functions included chemokine activity, signalling receptor binding and carbohydrate binding. KEGG pathway enrichment analysis (Fig. 2D) revealed that the DEGs were mainly involved in pathways related to neuropathic pain, including neuroactive ligand-receptor interactions, chemokine signalling pathways and cytokine-cytokine receptor interactions. These results suggest that the pathological mechanism of neuropathic pain mainly involves immune inflammation and abnormal remodelling of neuronal synapses.

Fig. 2.

Differentially expressed genes and enrichment analysis of sciatic nerves between the S and M groups. (A) PCA of the sciatic nerve in the S and M groups. n = 4. (B) Screen according to parameters q < 0.05 and |log2FoldChange|>1, and create a bar chart of differentially expressed genes between the S and M groups. (C) The top 10 biological process, cellular component, and molecular function in the GO analysis of differentially expressed genes between the S and M groups. (D) The top 20 signaling pathways in KEGG enrichment analysis of differentially expressed genes between the S and M groups

CCI promoted glycolysis in the sciatic nerve and exacerbated the inflammatory response

The transcriptome sequencing results showed that compared with the S group, the expression of genes related to glycolysis, such as HIF-1α, PGAM1, PGK1, HK3 and LDHA was significantly increased in the injured sciatic nerve of the M group of rats (Fig. 3A-F). We detected the protein expression levels of HIF-1α, PGAM1, PGK1, HK2, HK3 and LDHA in the sciatic nerves of the rats in each group using Western blotting (Fig. 3G-M) and found that CCI caused an increase in the protein expression levels of HIF-1α, PGAM1, PGK1, HK3 and LDHA (p < 0.05), but had no significant effect on HK2 (p > 0.05).

Fig. 3.

CCI caused increased glycolysis of the sciatic nerve and aggravated systemic inflammatory response. (A) Volcano plot of differentially expressed genes between the S and M groups. (B)-(F) Differential analysis of HIF-1α, PGAM1, PGK1, HK3 and LDHA expression levels in transcriptome data. n = 4. (G)-(M) Protein expression levels of HIF-1α, PGAM1, PGK1, HK2, HK3 and LDHA in sciatic nerve and quantification data of the expression of HIF-1α/GAPDH, PGAM1/GAPDH, PGK1/GAPDH, HK2/GAPDH, HK3/Tubulin and LDHA/Tubulin in each group. n = 6. Data were analyzed using Student’s t-test. (N) Lactic acid content of sciatic nerve in each group of rats detected by lactate detection kit. n = 6. Data were analyzed using Student’s t-test. (O) The ATP level of sciatic nerve in each group of rats detected by ATP detection kit. n = 6. Data were analyzed using Student’s t-test. (P)-(R) The content of IL-1β, IL-6 and TNF-α in the rat serum detected by ELISA. n = 5. Data were analyzed using Student’s t-test. Columns represent the mean ± SD. ^$p < 0.05, ^$$p < 0.01 vs. the S group

The levels of lactic acid and ATP in the sciatic nerve were detected using lactic acid and ATP detection kits, respectively (Fig. 3N, O). Compared with those in the S group, the rats in the M group presented an increased in lactate content (p < 0.01) and a decreased in ATP content (p < 0.01).

Serum IL-1β, IL-6 and TNF-α levels in the rats from each group were measured using ELISA (Fig. 3P-R). Compared with those in the S group, the levels of IL-1β, IL-6 and TNF-α in the M group were increased (p < 0.01).

CCI caused abnormal synaptic remodelling and increased neuronal activity in the spinal dorsal horn of rats

The ultrastructures of the spinal dorsal horn synapses in each group of rats were detected by transmission electron microscopy (Fig. 4A-C). Compared with those in the S group, the synapses in the M group exhibited smaller synaptic gaps (p < 0.01) and more synaptic vesicles (p < 0.01).

Fig. 4.

CCI caused abnormal synaptic remodeling and increased neuronal excitability in rat spinal dorsal horn neurons. (A) The ultrastructure of synapses in spinal dorsal horn neurons of rats in each group. Scale bars, 200 nm. (B) Width of synaptic space. n = 4. Data were analyzed using Student’s t-test. (C) Statistics of the number of synaptic vesicles in each group of rats. n = 4. Data were analyzed using Student’s t-test. (D) Schematic diagram of SG area in rat spinal dorsal horn and the state of SG neurons during glass microelectrode clamping. (E) Examples of sEPSCs recorded in spinal dorsal horn neurons. Scale bars, 40 pA, 5 s. (F) Cumulative distributions of inter-event interval and summarized data showed the frequency of sEPSC. n = 4. Data were analyzed using Student’s t-test. (G) Cumulative distributions of amplitude and summarized data showed the amplitude of sEPSC. n = 4. Data were analyzed using Student’s t-test. (H) Example voltage traces evoked by 80 pA inward current injection in spinal dorsal horn neurons. Scale bars, 20 mV, 0.2 s. (I) Summarized AP data recorded from spinal dorsal horn neurons. 0 ~ 120 pA, 10 pA steps, 1 s duration. (J) Spontaneous firing rates. n = 4. Data were analyzed using Student’s t-test. Columns represent the mean ± SD. ^$$p < 0.01 vs. the S group

The frequency and amplitude of sEPSCs in spinal dorsal horn SG neurons were detected by the whole-cell patch clamp technique (Fig. 4D-G). Compared with those in the S group, the frequency and amplitude of sEPSCs in the M group were significantly increased (p < 0.01). The cumulative frequency distribution curve of the M group was significantly shifted to the left compared with that of the S group. The cumulative amplitude distribution curve of the M group was significantly shifted to the right compared with that of the S group. In addition, the detection of APs in the SG neurons of the spinal dorsal horn (Fig. 4H-J) showed that compared with that in the S group, the frequency of AP firing in the M group was significantly increased (p < 0.01).

CCI altered the levels of HIF-1α PGAM1, PGK1, HK2, LDHA, lactic acid and ATP in the spinal dorsal horn but did not significantly affect the expression of HK3

Western blotting (Fig. 5A-G) showed that compared with those in the S group, the expression of the HIF-1α, PGAM1, PGK1, HK2 and LDHA proteins increased in the M group (p < 0.05). A significant difference in the expression of the HK3 protein was not observed between the S and M groups (p > 0.05).

Fig. 5.

CCI caused changes in the expression levels of HIF-1α, PGAM1, PGK1, HK2, LDHA, lactic acid and ATP in the spinal dorsal horn, but had no significant effect on the expression level of HK3. (A)-(G) Protein expression levels of HIF-1α, PGAM1, PGK1, HK2, HK3 and LDHA in dorsal horn of spinal cord and quantification data of the expression of HIF-1α/GAPDH, PGAM1/GAPDH, PGK1/GAPDH, HK2/GAPDH, HK3/Tubulin and LDHA/Tubulin in each group. n = 6. Data were analyzed using Student’s t-test. (H) Lactic acid content of spinal dorsal horn in each group of rats detected by lactate detection kit. n = 6. Data were analyzed using Student’s t-test. (I) The ATP level of spinal dorsal horn in each group of rats detected by ATP detection kit. n = 6. Data were analyzed using Student’s t-test. (J)-(P) Immunohistochemistry of HIF-1α (green)/NeuN (red), PGAM1 (green)/NeuN (red), PGK1 (green)/NeuN (red), HK2 (green)/NeuN (red), LDHA (green)/NeuN (red), PSD95 (green)/NeuN (red), BDNF (green)/NeuN (red). Scale bars, 200 μm, 50 μm. (Q)-(W) Quantitative analysis of mean fluorescence intensity of HIF-1α, PGAM1, PGK1, HK2, LDHA, PSD95 and BDNF in neurons. n = 4. Data were analyzed using Student’s t-test. Columns represent the mean ± SD. ^$p < 0.05, ^$$p < 0.01 vs. the S group

The levels of lactic acid and ATP in the spinal dorsal horn were detected using lactic acid and ATP detection kits, respectively (Fig. 5H, I). Compared with those in the S group, the rats in the M group presented an increased lactate content (p < 0.01) and a decreased ATP content (p < 0.01).

Immunofluorescence staining detection of the spinal dorsal horn (Fig. 5J-W) showed that the immunoreactivity of HIF-1α, PGAM1, PGK1, HK2, LDHA, PSD95 and BDNF in neurons in the M group was higher than that in the S group (p < 0.01).

CCI promoted glycolysis in the spinal dorsal horn, whereas HIF-1a siRNA inhibited glycolysis in CCI rats, reducing inflammation and increasing the pain threshold

As shown in Fig. 6A-F and J-N, compared with the S group, the mRNA expression levels of HIF-1α, HK2, LDHA, PGAM1 and PGK1 were increased in the spinal dorsal horn of CCI model rats (p < 0.01) but HK3 levels were not significantly changed (p > 0.05). Compared with the M + NC-siRNA group, the mRNA expression level of HIF-1α in the spinal dorsal horn of rats in the M + HIF-1a-siRNA group decreased by about 65%, and the mRNA expression levels of HK2, LDHA, PGAM1 and PGK1 were also reduced (p < 0.01). There was no significant difference in the mRNA expression of HIF-1α, HK2, LDHA, PGAM1 and PGK1 between the M and M + NC-siRNA groups (p > 0.05).

Fig. 6.

CCI promoted glycolysis in the spinal dorsal horn, while the use of HIF-1α siRNA inhibited glycolysis in CCI rats, reducing inflammation and increasing pain threshold. (A)-(F) qRT-PCR showing the expression of HIF-1α, HK3, HK2, LDHA, PGAM1 and PGK1 mRNA in the spinal cord. n = 6. Data were analyzed using Student’s t-test. (G) Schematic diagram of experimental procedure. (H) Thermal withdrawal latency (TWL). n = 15. Data were analyzed by repeated measures two-way ANOVA (Group × Day) with Bonferroni’s post-hoc tests. (I) Mechanical withdrawal threshold (MWT). n = 15. Data were analyzed by repeated measures two-way ANOVA (Group × Day) with Bonferroni’s post-hoc tests. (J)-(N) qRT-PCR showing the expression of HIF-1α, HK2, LDHA, PGAM1 and PGK1 mRNA in the spinal cord. n = 6. Data were analyzed using one-way ANOVA followed by Tukey’s or Dunnett’s post-hoc tests. (O) Lactic acid content of spinal cord in each group of rats detected by lactate detection kit. n = 6. Data were analyzed using one-way ANOVA followed by Tukey’s or Dunnett’s post-hoc tests. (P) The ATP level of spinal cord in each group of rats detected by ATP detection kit. n = 6. Data were analyzed using one-way ANOVA followed by Tukey’s or Dunnett’s post-hoc tests. (Q)-(S) The content of IL-1β, IL-6 and TNF-α in the rat serum detected by ELISA. n = 5. Data were analyzed using one-way ANOVA followed by Tukey’s or Dunnett’s post-hoc tests. Columns represent the mean ± SD. ^$$p < 0.01 vs. the S group, ^##p < 0.01 vs. the M + NC-siRNA group

The pain threshold test results (Fig. 6H, I) showed that the MWT and TWL of rats in the M group and M + NC-siRNA group were similar (p > 0.05). Compared with those of the M + NC-siRNA group, the MWT and TWL of the M + HIF-1a-siRNA group increased (p < 0.01).

The levels of lactic acid and ATP in the spinal dorsal horn were detected using lactic acid and ATP detection kits, respectively (Fig. 6O, P). Compared with those in the M + NC-siRNA group, the lactic acid content decreased (p < 0.01) and the ATP content increased (p < 0.01) in the M + HIF-1α-siRNA group. There was no significant difference in the levels of lactic acid and ATP between the M and M + NC-siRNA groups (p > 0.05).

The serum IL-1β, IL-6 and TNF-α levels in the rats from each group were measured using ELISA (Fig. 6Q-S). Compared with those in the M group, the expression levels of IL-1β, IL-6 and TNF-α in the M + HIF-1α-siRNA group were decreased (p < 0.01). There was no significant difference in the expression levels of IL-1β, IL-6 and TNF-α between the M and M + NC-siRNA groups (p > 0.05).

FGF4 was suggested as a potential core regulator of neuronal glycolysis in CCI rats

To further explore the key molecular mechanisms of glycolysis regulation in CCI model rats, we constructed a PPI network diagram of glycolysis-related genes (HIF-1α, HK2, LDHA, PGAM1 and PGK1) and the top 20 DEGs ranked in ascending order of q value using the STRING database and visualized it using Cytoscape 3.10.0 software (Fig. 7A-D, Fig. S1A-H). The results showed that the top 5 hub genes identified by the Degree algorithm were Hif1a, LOC684762 (representing the functional cluster of LOC102549173), Fgf4, Snca and Hk2; the top 5 hub genes identified by the DMNC algorithm were Hk2, Ldha, Pgk1, Fgf4 and LOC684762; the top 5 hub genes identified by the EPC algorithm were Hif1a, LOC684762, Fgf4, Snca and Pgk1; and the top 5 hub genes identified by the MCC algorithm were Hif1a, LOC684762, Fgf4, Snca and Pgk1. Further intersection analysis of the hub gene sets from the four algorithms revealed that LOC684762 and Fgf4 were the common core genes identified by all four algorithms (Fig. S1 I). In addition, both LOC102549173 and FGF4 also have interactions with glycolysis related genes (HIF-1α, HK2, LDHA and PGK1).

Fig. 7.

FGF4 was suggested as a potential core regulator of neuronal glycolysis in CCI rats. (A) Heat map of differentially expressed genes. n = 4. (B) Ranking the top 20 differentially expressed genes in ascending order of q-value. (C) Construct a protein-protein interaction network of differentially expressed genes through STRING online analysis database. (D) Based on the global PPI network (C), topological analysis using Cytoscape identified FGF4 as a hub gene. For clarity, the predicted interaction lines between FGF4 and key glycolytic genes (HIF-1α, HK2, LDHA, PGK1), as well as between HIF-1α and key glycolytic genes (HK2, LDHA, PGK1), were highlighted in red. (E), (F) Protein expression levels of FGF4 in sciatic nerve and quantification data of the expression of FGF4/GAPDH in each group. n = 6. Data were analyzed using Student’s t-test. (G) qRT-PCR showing the expression of FGF4 mRNA in sciatic nerve. n = 6. Data were analyzed using Student’s t-test. (H) Immunohistochemistry of FGF4 (green)/NeuN (red), FGF4 (green)/GFAP (red) and FGF4 (green)/Iba1 (red) in the spinal dorsal horn. Scale bars, 20 μm. (I), (L) Protein expression levels of FGF4 in dorsal horn of spinal cord and quantification data of the expression of FGF4/GAPDH in each group. n = 6. Data were analyzed using Student’s t-test. (J) Immunohistochemistry of FGF4 (green) and NeuN (red) in the spinal dorsal horn. Scale bars, 200 μm, 50 μm. (K) Quantitative analysis of mean fluorescence intensity of FGF4 in neurons. n = 4. Data were analyzed using Student’s t-test. Columns represent the mean ± SD. ^$$p < 0.01 vs. the S group

To further verify the expression changes of FGF4 in CCI model rats, we measured FGF4 expression in the sciatic nerve using Western blotting (Fig. 7E, F) and qRT-PCR (Fig. 7G). Compared with those in the S group, the FGF4 protein and mRNA expression levels in the sciatic nerve of rats from the M group were reduced (p < 0.01).

We also measured FGF4 in the dorsal horn of the rat spinal cord using immunofluorescence staining (Fig. 7H). The detection results showed that FGF4 was colocalized with NeuN, but rarely colocalized with GFAP and Iba1, indicating that FGF4 was mainly expressed in neurons, but rarely expressed in astrocytes or microglia. Further testing revealed that CCI reduced the immunoreactivity of FGF4 in neurons (Fig. 7J, K). Western blotting (Fig. 7I, L) showed that CCI inhibited the expression of FGF4 (p < 0.01).

FGF4 overexpression increased the pain threshold of CCI rats and inhibited glycolysis in spinal dorsal horn neurons

To further clarify the role of FGF4 in CCI induced neuropathic pain, we constructed FGF4-AAV adeno-associated virus to overexpress FGF4 and administered it by intrathecal injection to explore the effect of FGF4 overexpression in the spinal cord on CCI-induced neuropathic pain (Fig. 8A-C). To clarify which type of cells (neurons, astrocytes, or microglia) the virus primarily infects in the spinal cord, we costained EGFP with NeuN, GFAP and Iba1 using immunofluorescence (Fig. 8D). The results showed that the virus mainly transfected into neuronal cells, with a very small number transfected into astrocytes or microglia (Fig. 8E).

Fig. 8.

Intrathecal injection of FGF4-AAV increased the pain threshold of CCI rats and inhibited the expression of HIF-1α in neurons. (A) Schematics of AAV constructs overexpressing FGF4 or control. EGFP, enhanced green fluorescence protein; CMV, cytomegalovirus promoter; ITR, inverted terminal repeats. (B) Schematic diagram of experimental procedure. (C) Schematic diagram of AAV intrathecal injection. (D) Colocalization of EGFP (green) and NeuN (red), EGFP (green) and GFAP (red), and EGFP (green) and Iba1 (red) in L4-6 spinal cord segments. scale bar, 200 μm, 20 μm. (E) FGF4⁺NeuN⁺, FGF4⁺GFAP⁺ and FGF4⁺Iba1⁺ percentage statistics. (F) Thermal withdrawal latency (TWL). n = 15. Data were analyzed by repeated measures two-way ANOVA (Group × Day) with Bonferroni’s post-hoc tests. (G) Mechanical withdrawal threshold (MWT). n = 15. Data were analyzed by repeated measures two-way ANOVA (Group × Day) with Bonferroni’s post-hoc tests. (H), (I) Immunohistochemistry of FGF4 (green)/NeuN (red) and HIF-1α (green)/NeuN (red). Scale bars, 200 μm, 50 μm. (J), (K) Quantitative analysis of mean fluorescence intensity of FGF4 and HIF-1α in neurons. n = 4. Data were analyzed using one-way ANOVA followed by Tukey’s or Dunnett’s post-hoc tests. Columns represent the mean ± SD. ^##p < 0.01 vs. the M + NC-AAV group

The pain threshold test results (Fig. 8F, G) showed that the MWT and TWL of rats in the M group and M + NC-AAV group were similar (p > 0.05). Compared with those of the M + NC-AAV group, the MWT and TWL of the M+FGF4-AAV group increased (p < 0.01).

We subsequently detected the expression levels of FGF4, HIF-1α, PGAM1, PGK1, HK2, LDHA, PSD95 and BDNF in the spinal dorsal horn neurons of each group of rats using immunofluorescence staining (Figs. 8H-K and 9). Compared with the M + NC-AAV group, the immunoreactivity of FGF4 in spinal dorsal horn neurons of the M+FGF4-AAV group rats increased, while the immunoreactivity of HIF-1α, PGAM1, PGK1, HK2, LDHA, PSD95 and BDNF decreased (p < 0.01). No significant differences in the immunoreactivity of FGF4, HIF-1α, PGAM1, PGK1, HK2, LDHA, PSD95 and BDNF in spinal dorsal horn neurons were observed between the M and M + NC-AAV groups (p > 0.05).

Fig. 9.

FGF4-AAV inhibited the expression of PGAM1, PGK1, HK2, LDHA, PSD95 and BDNF in spinal dorsal horn neurons of CCI rats. (A)-(F) Immunohistochemistry of PGAM1 (green)/NeuN (red), PGK1 (green)/NeuN (red), HK2 (green)/NeuN (red), LDHA (green)/NeuN (red), PSD95 (green)/NeuN (red) and BDNF (green)/NeuN (red). Scale bars, 200 μm, 50 μm. (G)-(L) Quantitative analysis of mean fluorescence intensity of PGAM1, PGK1, HK2, LDHA, PSD95 and BDNF in neurons. n = 4. Data were analyzed using one-way ANOVA followed by Tukey’s or Dunnett’s post-hoc tests. Columns represent the mean ± SD. ^##p < 0.01 vs. the M + NC-AAV group

Western blotting (Fig. 10A, B) showed that compared with that in the M + NC-AAV group, the protein expression of FGF4 in the M+FGF4-AAV group increased (p < 0.01), and the protein expression of HIF-1α, PGAM1, PGK1, HK2, and LDHA in the M+FGF4-AAV group decreased (p < 0.01). There was no significant difference in the expression of FGF4, HIF-1α, PGAM1, PGK1, HK2 and LDHA between the M and M + NC-AAV groups (p > 0.05).

Fig. 10.

FGF4-AAV promoted the expression of FGF4 in the spinal dorsal horn of CCI rats, inhibited the expression of HIF-1α, PGAM1, PGK1, HK2 and LDHA, reduced abnormal synaptic remodeling, and decreased neuronal excitability. (A), (B) Protein expression levels of FGF4, HIF-1a, PGAM1, PGK1, HK2 and LDHA in dorsal horn of spinal cord and quantification data of the expression of FGF4/GAPDH, HIF-1α/GAPDH, PGAM1/GAPDH, PGK1/GAPDH, HK2/GAPDH and LDHA/Tubulin in each group. n = 5. Data were analyzed using one-way ANOVA followed by Tukey’s or Dunnett’s post-hoc tests. (C) Lactic acid content of spinal cord in each group of rats detected by lactate detection kit. n = 6. Data were analyzed using one-way ANOVA followed by Tukey’s or Dunnett’s post-hoc tests. (D) The ATP level of spinal cord in each group of rats detected by ATP detection kit. n = 6. Data were analyzed using one-way ANOVA followed by Tukey’s or Dunnett’s post-hoc tests. (E)-(G) The content of IL-1β, IL-6 and TNF-α in the rat serum detected by ELISA. n = 5. Data were analyzed using one-way ANOVA followed by Tukey’s or Dunnett’s post-hoc tests. (H) The ultrastructure of synapses in spinal dorsal horn neurons of rats in each group. Scale bars, 200 nm. (I) Width of synaptic space. n = 4. Data were analyzed using one-way ANOVA followed by Tukey’s or Dunnett’s post-hoc tests. (J) Statistics of the number of synaptic vesicles in each group of rats. n = 4. Data were analyzed using one-way ANOVA followed by Tukey’s or Dunnett’s post-hoc tests. (K) Examples of sEPSCs recorded in spinal dorsal horn neurons. Scale bars, 40 pA, 5 s. (L) Cumulative distributions of inter-event interval and summarized data showed the frequency of sEPSC. n = 4. Data were analyzed using one-way ANOVA followed by Tukey’s or Dunnett’s post-hoc tests. (M) Cumulative distributions of amplitude and summarized data showed the amplitude of sEPSC. n = 4. Data were analyzed using one-way ANOVA followed by Tukey’s or Dunnett’s post-hoc tests. (N) Example voltage traces evoked by 80 pA inward current injection in spinal dorsal horn neurons. Scale bars, 20 mV, 0.2 s. (O) Summarized AP data recorded from spinal dorsal horn neurons. 0 ~ 120 pA, 10 pA steps, 1 s duration. (P) Spontaneous firing rates. n = 4. Data were analyzed using one-way ANOVA followed by Tukey’s or Dunnett’s post-hoc tests. Columns represent the mean ± SD. ^##p < 0.01 vs. the M + NC-AAV group

The levels of lactic acid and ATP in the spinal dorsal horn were detected using lactic acid and ATP detection kits, respectively (Fig. 10C, D). Compared with those in the M + NC-AAV group, the rats in the M+FGF4-AAV group had a decrease in lactic acid content (p < 0.01) and an increase in ATP content (p < 0.01). There was no significant difference in the levels of lactic acid and ATP between the M and M + NC-AAV groups (p > 0.05).

FGF4 overexpression reduced the neuroinflammatory response in CCI rats and decreased the abnormal excitation of neurons in the spinal dorsal horn

The serum IL-1β, IL-6 and TNF-α levels in the rats from each group were measured using ELISA (Fig. 10E-G). Compared with those in the M + NC-AAV group, the levels of IL-1β, IL-6 and TNF-α in the M+FGF4-AAV group were decreased (p < 0.01). There was no significant difference in the expression levels of IL-1β, IL-6 and TNF-α between the M and M + NC-AAV groups (p > 0.05).

The ultrastructures of the spinal dorsal horn synapses in each group of rats were detected using transmission electron microscopy (Fig. 10H-J). Compared with those in the M + NC-AAV group, the synapses in the M+FGF4-AAV group exhibited larger synaptic gaps (p < 0.01) and fewer synaptic vesicles (p < 0.01). There was no significant difference in synaptic gaps and synaptic vesicles between the M and M + NC-AAV groups (p > 0.05).

The frequency and amplitude of sEPSCs in spinal dorsal horn SG neurons were detected using the whole-cell patch clamp technique (Fig. 10K-M). Compared with those in the M + NC-AAV group, the frequency and amplitude of sEPSCs in the M+FGF4-AAV group were significantly decreased (p < 0.01). The cumulative frequency distribution curve of the M+FGF4-AAV group was significantly shifted to the right compared with that of the M + NC-AAV group. The cumulative amplitude distribution curve of the M+FGF4-AAV group was significantly shifted to the left compared with that of the M + NC-AAV group. There was no significant difference in frequency and amplitude of sEPSC between the M and M + NC-AAV groups (p > 0.05). In addition, the detection of APs in the SG neurons of the spinal dorsal horn (Fig. 10N-P) showed that compared with the M + NC-AAV group, the frequency of AP firing in the M+FGF4-AAV group was significantly decreased (p < 0.01). The frequency of AP firing was not significantly different between the M and M + NC-AAV groups (p > 0.05).

Discussion

In this study, we found that CCI induced neuropathic pain is closely related to the HIF-1α-mediated increase in neuronal glycolysis, which can exacerbate neuroinflammatory responses and increase neuronal excitability. In addition, further research revealed that HIF-1α-mediated neuronal glycolysis is mainly regulated by FGF4.

The method for constructing the CCI model is simple, and its induced spontaneous pain-like manifestations are similar to the neuropathic pain manifestations of clinical patients, with a long duration of symptoms, making it an ideal model for studying neuropathic pain [4]. In this study, we constructed a CCI rat model and confirmed its stable mechanical and thermal pain sensitivity, accompanied by significant tissue damage to the sciatic nerve, decreased Nissl bodies, and myelin sheath destruction, which are consistent with the classic pathological features of neuropathic pain. NF200 forms the neuronal cytoskeleton together with microtubules and microfilaments, mainly providing structural support for axons and regulating the axon diameter, affecting nerve conduction velocity [7]. S100β is a Schwann cell marker that plays a crucial role in myelin formation and maintenance [8, 9]. Immunofluorescence staining showed that CCI reduced the immunoreactivity of NF200 and S100β, indicating damage to neuronal axons and myelin sheaths. These results provide a reliable behavioural and pathological foundation for the subsequent exploration of the underlying mechanism.

Subsequently, we performed transcriptome sequencing of the affected sciatic nerve of rats, and bioinformatics analysis showed that the pathological mechanism of CCI-induced neuropathic pain is mainly attributed to immune inflammation and the abnormal remodelling of neuronal synapses, which is consistent with the current mainstream direction of pain research. But when we carefully examined the DEGs list, we found that the expression levels of a series of key glycolytic enzymes/regulatory factors (such as HIF-1α, HK3, LDHA, PGAM1, PGK1) were significantly and consistently upregulated. We also performed Western blotting to measured the levels of HIF-1α, PGAM1, PGK1, HK3 and LDHA, and observed an increase in the expression levels of these proteins after CCI. Although these genes, as part of the “metabolic process”, may be masked by larger and more significant “immune response” or “neuronal synapse” related gene groups in overall enrichment analysis, and have not become the top-level enrichment entries, the magnitude of their individual changes and biological correlations have attracted our high attention.

In recent years, the role of energy metabolism reprogramming in neurological diseases has received increasing attention. The pathophysiological significance of neuronal glycolysis is particularly prominent. The traditional view holds that mature neurons mainly rely on efficient oxidative phosphorylation to meet their energy needs [10]. However, under stress states such as nerve damage, ischaemia, or inflammation, neurons undergo rapid adaptive metabolic changes characterized by a significant increase in glycolysis [11]. We observed an increase in lactic acid levels and a decrease in ATP production in the damaged sciatic nerve, which are typical metabolic features of increased glycolysis [12]. Among them, lactic acid is no longer considered a simple metabolic end product. As an important signalling molecule, it can directly activate acid-sensitive ion channels on nociceptive neurons by acidifying the extracellular microenvironment, thereby exacerbating synaptic dysfunction and inducing abnormal neuronal discharges [13]. Lactic acid can also promote the activation of microglia and astrocytes, thereby promoting the release of proinflammatory cytokines such as IL-1β, IL-6 and TNF-α, resulting in a vicious cycle of “glycolysis-lactate-neuroinflammation” [14]. This finding is consistent with our detection of elevated levels of proinflammatory cytokines such as IL-1β, IL-6 and TNF-α in the serum of CCI rats. Therefore, we speculate that CCI induced neuropathic pain may be related to neuronal glycolysis reprogramming, which is the direct cause of subsequent deterioration of neuroinflammation and abnormal remodeling of neuronal synapses.

The spinal dorsal horn is the primary centre in the central nervous system that analyses and processes pain information, and it plays an extremely important role in the transmission of pain information from the periphery to the central nervous system [15]. We also analysed the dorsal horn of the rat spinal cord and found that CCI can induce abnormal synaptic remodelling, including a decreased synaptic cleft and increased number of synaptic vesicles. Synapses are important components of information transmission between neurons, and neurotransmitters released from the presynaptic membrane need to cross the synaptic cleft to act on the postsynaptic membrane. The synaptic cleft is rich in neurotransmitter-degrading enzymes, and the wider the synaptic cleft is, the longer the time required for neurotransmitter diffusion and the greater the degree of enzymatic hydrolysis [16]. Therefore, the narrowing of synaptic gaps and the increase in the number of synaptic vesicles may be the structural basis for the increase in abnormal signal transmission and increased pain sensitivity in CCI rats.

The sEPSC is a postsynaptic membrane excitatory current caused by the release of excitatory neurotransmitters from the presynaptic membrane. Upon the detection of excitatory neurotransmitters, the cell membrane potential increases, and an AP can be triggered when the membrane potential reaches a threshold. Therefore, it can be said that sEPSC reflects the excited state of neurons to a certain extent, while AP is a manifestation of neuronal excitation [17]. In this study, we detected SG neurons in the spinal dorsal horn layer II using the whole-cell patch clamp technique and found that CCI increased the frequency and amplitude of sEPSC and promoted the release of AP, indicating that CCI can drive neuronal excitability and induce central sensitization. In order to further clarify whether the increased excitability of spinal dorsal horn neurons in CCI rats is related to enhanced glycolysis, we also detected glycolysis-related indicators through Western blotting. The results showed that CCI promoted the expression of HIF-1α, LDHA, PGAM1 and PGK1 in the spinal dorsal horn, but had no significant effect on HK3. Hexokinases (HKs) are the key enzymes participating in the conversion of glucose to glucose 6-phosphate in glycolysis, consisting of five isoenzymes (HK1, HK2, HK3, HK4 and HKDC1) [18]. These five members of the mammalian HK family with similar structure have tissue-specific expression and are closely related to cell type, metabolic status and function [18, 19]. Studies have shown that compared to other isoenzymes such as HK3, HK2 is more widely expressed in various cell types [20], especially in cells that require high-speed glycolysis (such as cancer cells, proliferating cells) or cells under metabolic stress (such as neuronal cells, glial cells), where its expression undergoes corresponding changes [19, 21]. Studies also showed that HK2 is a downstream target gene of transcription factors such as HIF-1α, playing a central role in pathological metabolic reprogramming [22]. Therefore, we also detected the protein expression levels of HK2 in the sciatic nerve and spinal cord dorsal horn, and the results showed that CCI could cause an increase in the expression level of HK2 in the spinal dorsal horn, but there was no significant change in the expression level of HK2 in the sciatic nerve, suggesting that the specific molecular executors of metabolic reprogramming may exist tissue-specific in different anatomical sites of neuropathic pain. After CCI, the axons of neurons in the sciatic nerve were directly exposed to the local microenvironment of physical damage, barrier disruption, and inflammation [23]. Research has shown that the expression of HK3 can be strongly induced by inflammatory signals [24, 25]. Therefore, its upregulation may represent a specific metabolic stress state initiated by peripheral neuronal axons in response to this “damaging microenvironment”. In contrast, although spinal dorsal horn neurons are also in a neuroinflammatory environment [26], the core pressure they bear comes from sustained nociceptive input induced “excitatory stress”, and their relatively chronic inflammatory signals may not have reached the dominant threshold for driving HK3 expression. As a target gene of HIF-1α, a large number of studies have shown that the expression level of HK2 is significantly upregulated in response to hypoxia, excitatory stress, and other conditions in the central nervous system [21, 27], which is consistent with the increased expression level of HK2 in the spinal dorsal horn of the neuropathic pain model in this study. In the sciatic nerve, although HIF-1α was upregulated, its transcriptional activity may be guided by strong inflammatory signals to other targets (such as HK3, pro-inflammatory factors, etc.) rather than HK2, but its specific regulatory mechanism still needs further exploration. Therefore, the results of this study also indicated that in neuropathic pain, neuronal glycolysis reprogramming is not a homogeneous process, but rather exhibits specificity in program selection based on its anatomical location and dominant stress signals. Further detection of spinal dorsal horn by immunofluorescence showed colocalization of HIF-1α, HK2, LDHA, PGAM1 and PGK1 with neuronal marker NeuN. These results suggest that glycolysis may provide neurons with the immediate energy and biosynthetic precursors required for high-frequency discharge and increased synaptic transmission. Moreover, increased glycolysis is synchronized with the upregulation of the expression of synaptic plasticity-related proteins such as PSD95 and BDNF, further indicating that metabolic reprogramming may directly participate in the formation of pain memory and chronicity by affecting synaptic structure and function.

HIF-1α is a key transcription factor involved in the cellular responses to hypoxia and environmental stress. Under normoxic conditions, HIF-1α is usually rapidly degraded; in response to hypoxia or specific signalling stimuli, HIF-1α is stably translocated to the nucleus, initiating the transcription of a series of target genes [28, 29]. Studies have shown that HIF-1α can promote glycolysis by inducing the transcription and translation of glycolytic enzymes such as HK2 and pyruvate kinase M2 (PKM2) [30, 31] and can inhibit the activity of pyruvate dehydrogenase (PDH) by activating pyruvate dehydrogenase kinase isoenzyme 1 (PDK1), thereby inhibiting the conversion of pyruvate to acetyl-CoA and ultimately inhibiting oxidative phosphorylation [32].

In this study, we observed a significant upregulation of HIF-1α expression in the sciatic nerve and spinal dorsal horn of CCI rats. In order to establish the causal role of HIF-1α in pain rather than just association, we conducted a functional impairment experiment. Intrathecal injection of HIF-1α siRNA effectively reversed various pathological changes induced by CCI. At the molecular level, it inhibited the activation of the glycolytic pathway; at the metabolic level, it reduced lactic acid levels and increased ATP levels; at the inflammatory level, it reduced the systemic inflammatory response; and ultimately, at the behavioural level, the pain threshold of the rats was significantly increased. These results unequivocally demonstrate that HIF-1α is necessary for the occurrence and development of neuropathic pain and is an effective target for interventions designed to modulate this pathway.

Therefore, what factors trigger HIF-1α expression after nerve injury? In order to further clarify the upstream regulatory mechanism, we constructed a PPI network by combining glycolysis-related genes with the top 20 DEGs ranked in ascending order of q value in the transcriptome sequencing. PPI network analysis showed that LOC102549173 and FGF4 are both core genes and interact with glycolysis related genes (HIF-1α, HK2, LDHA and PGK1). However, there are essential differences in the functional attributes between LOC102549173 and FGF4. LOC102549173 has been annotated as a member of the histone H3 family. As a structural component of nucleosomes, it mainly participates in chromatin packaging and global gene transcription regulation, rather than directly participating in signal transduction or metabolic pathways. If LOC102549173 were chosen as the research target, directly intervening in histone H3 (a structural protein) would present technical difficulties and specificity issues—intervening in a core histone affects all genes that rely on this histone for packaging, which encompasses nearly the entire genome, and any observed phenotypic changes cannot be attributed to any specific downstream event. On the other hand, the expression changes of LOC102549173 in this study may reflect CCI induced global chromatin remodeling, and its function is indirectly related to the focus of this study on “neuronal glycolysis metabolism regulation”, making it difficult to establish a clear causal chain. In contrast, FGF4, as a secreted signaling molecule, can activate intracellular signaling pathways through cell membrane receptors. Its functional intervention methods are mature and specific, and previous studies have shown that it exerts a regulatory effect on glucose metabolism [33]. Therefore, FGF4 was selected as the core research object in this study to further explore its molecular mechanism of regulating neuronal glycolysis in neuropathic pain. In this study, we found that the protein and mRNA expression of FGF4 were consistently and significantly downregulated in the sciatic nerve and spinal dorsal horn of CCI rats. Therefore, we speculate that FGF4 may play a crucial role in neuropathic pain as an upstream regulator of HIF-1α-mediated glycolysis.

To further validate the core role of FGF4, we administered intrathecal injection of FGF-AAV to rats to overexpress FGF4. The results showed that overexpression of FGF4 not only effectively reversed pain behaviour, but also specifically inhibited the expression of HIF-1α and its downstream glycolysis-related markers in spinal neurons, reduced lactic acid levels, and alleviated systemic inflammatory responses. Moreover, FGF4 overexpression reversed the synaptic ultrastructural abnormalities and increased excitability of SG neurons caused by CCI. These results clearly indicate that FGF4 exerts a protective effect against pain.

Immunofluorescence staining results showed that FGF4 mainly colocalizes with the neuronal marker NeuN in the spinal dorsal horn, and its expression is downregulated after CCI, strongly suggesting that neurons are one of the main cell sources of FGF4. Studies have shown that FGF4 mainly binds with high affinity to FGFR1 and FGFR2 [34–36]. Therefore, we speculate that FGF4 may exert regulatory effects through autocrine or paracrine activation of FGFR1/FGFR2 on neurons. After FGFR activation, classic downstream pathways include the phosphatidylinositol-3-kinase (PI3K)/ protein kinase B (AKT) signaling axis [37, 38]. The PI3K/AKT pathway is one of the main pathways regulating protein synthesis, cellular metabolism, and HIF-1α stability [39]. Therefore, we speculate that CCI may regulate the PI3K/AKT signaling pathway through FGF4/FGFR, promoting the expression of HIF-1α, ultimately leading to the accumulation of HIF-1α protein and driving glucose metabolism reprogramming. The overexpression of FGF4 observed in this study can specifically reduce the protein level of HIF-1α in neurons, which is highly consistent with this hypothesis. But the specific regulatory mechanism still needs further research in the future.

In this study, enhanced glycolysis occurred in both the sciatic nerve and spinal dorsal horn of CCI model rats, revealing the systematic features of metabolic reprogramming in neuropathic pain. These two parts represent the peripheral initiating site and central regulatory hub of pain respectively, and their metabolic changes not only have their own unique pathophysiological significance, but also form a synergistic network through complex interactions. This metabolic reprogramming is not only a passive response to energy adaptation, but also a key mechanism actively involved in the generation, transmission, and maintenance of pain signals. In the future, we will also comprehensively and systematically reveal the regulatory mechanism of glucose metabolism in neuropathic pain through the detection of extracellular acidification rate (ECAR), oxygen consumption rate (OCR), the classical target genes of HIF-1α (such as GLUT1, PDK1 and PFKFB3) and their nuclear translocation.

This study also had some limitations. Firstly, although the use of FGF4-AAV supports FGF4 being located upstream of HIF-1α, it cannot be ruled out that FGF4 may act through a parallel pathway partially independent of HIF-1α. In future research, reactivating HIF-1α in the presence of FGF4 overexpression to rigorously validate the necessity of HIF-1α will be an important step in improving the causal relationship of this regulatory axis. Secondly, although we used validated negative control siRNA, the general limitations of siRNA technology still cannot completely rule out the possibility of undetected off target effects or mild innate immune responses triggered by the delivery process. In the future, using gene knockout animals or more specific pharmacological tools for validation will help further consolidate the conclusions. Thirdly, long-term and widespread central nervous system gene therapy (AAV) still faces issues such as immunogenicity, off target expression, and long-term safety. As an alternative strategy, developing FGF4 mimetic peptides or small molecule agonists that can penetrate the blood-brain barrier may be a better choice.

In summary, this study reveals for the first time the core role of the FGF4/HIF-1α axis in neuropathic pain by regulating neuronal glycolysis (Fig. 11). Our findings not only provide a new theoretical framework for understanding the metabolic mechanisms of pain but also, more importantly, establish FGF4 and key links in the neuronal glycolysis pathway as highly promising new therapeutic targets, providing solid experimental evidence and new ideas for developing metabolic interventions for refractory neuropathic pain.

Fig. 11.

Schematic diagram of the mechanism of FGF4/HIF-1α axis in regulating neuronal glycolysis in CCI induced neuropathic pain. CCI caused downregulation of FGF4 expression, thereby relieving its inhibitory effect on transcription factor HIF-1α. Activated HIF-1α upregulated the expression of a series of key glycolytic enzymes, driving neuronal glycolytic metabolism reprogramming. This process caused an increase in the metabolic product lactate, which not only promoted the release of pro-inflammatory cytokines in glial cells, but also led to abnormal synaptic restructuring and hyperexcitability of spinal dorsal horn neurons, ultimately resulting in pain

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

CCI

Chronic constriction injury of the sciatic nerve

the S group

The sham group

the M group

The rats underwent CCI surgery group

FGF4

Fibroblast growth factor 4

HIF-1α

Hypoxia-inducible factor-1α

HK3

Hexokinase 3

LDHA

Lactate dehydrogenase A

PGAM1

Phosphoglycerate mutase 1

PGK1

Phosphoglycerate kinase 1

HK2

Hexokinase 2

GO

Gene Ontology

S100β

S100 calcium-binding proteinβ

BDNF

Brain-derived neurotrophic factor

PSD-95

Postsynaptic density 95

NF200

Neurofilament 200

KEGG

Kyoto Encyclopedia of Genes and Genomes

PPI

Protein-protein interaction

sEPSCs

Spontaneous excitatory postsynaptic currents

SG

Substantia gelatinosa

AP

Action potentials

DEGs

Differentially expressed genes

Author contributions

Conceptualization, WQY, YXW and WXQ; Data curation, WQY and CYW; Formal analysis, XH, RSQ and LSS; Methodology, YXT, WQY, CYW, XH and LSS; Validation, WQY and CYW; Writing – original draft, WQY and YXW; Writing – review & editing, WQY and WXQ. All authors read and approved the final manuscript.

Funding

This study was supported by the High-Level Innovation Team of the “Ou-Yue Talent Program” funded by the Wenzhou Science and Technology Bureau (No. 2024R2003), the Summit Advancement Disciplines of Zhejiang Province (Wenzhou Medical University-Pharmaceutics), and the Yunnan Provincial Major Science and Technology Projects (No. 202402AA310058 and No. 202502AS100005).

Data availability

The transcriptome sequencing data has been submitted to GEO with accession number GSE317107. Other datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

This experiment was approved by the Animal Research Committee of Wenzhou Medical University and followed the National Institutes of Health’s Guide for the Care and Use of Laboratory Animals.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.