Electroacupuncture alleviates diabetic peripheral neuropathy: Regulates mitochondrial biogenesis and redistribution in peripheral nerve fibers

Abstract

BACKGROUND

Diabetic peripheral neuropathy (DPN) is a common chronic complication of diabetes, and current therapeutic options remain unsatisfactory. Mitochondrial dysfunction is considered a key contributor to DPN, and modulation of mitochondrial biogenesis may represent a promising therapeutic strategy. Electroacupuncture (EA) has shown beneficial effects in diabetic neuropathy, but its underlying mechanisms remain incompletely understood. We hypothesized that EA alleviates type 2 DPN (T2DPN) by enhancing silent information regulator 1 (SIRT1)/peroxisome proliferator-activated receptor-γ coactivator-1α (PGC-1α)-mediated mitochondrial biogenesis and mitochondrial redistribution, particularly in intraepidermal nerve fibers (IENFs).

AIM

To investigate whether EA alleviates T2DPN through SIRT1/PGC-1α-mediated mitochondrial biogenesis and redistribution.

METHODS

A rat model of T2DPN was established by high-fat diet feeding combined with streptozotocin injection (35 mg/kg). After successful model establishment, rats received EA stimulation at ST25 for 6 weeks. Metabolic parameters, behavioral tests, nerve conduction studies, immunofluorescence staining, western blotting, and three-dimensional imaging were used to evaluate mitochondrial content and distribution in IENFs and sciatic nerve tissue, as well as SIRT1 and PGC-1α expression.

RESULTS

Changes in body weight, blood glucose, behavioral responses, and nerve conduction confirmed the successful establishment of the T2DPN model. EA improved metabolic status, increased withdrawal thresholds, and ameliorated IENF loss, indicating protective effects on neural function and morphology. In T2DPN rats, SIRT1/PGC-1α-mediated mitochondrial biogenesis was reduced. EA treatment upregulated SIRT1 and PGC-1α expression in the sciatic nerve and increased mitochondrial content. Three-dimensional analysis showed more prominent mitochondrial redistribution in IENFs than in adjacent keratinocytes. In addition, SIRT1 inhibition attenuated the beneficial effects of EA on nerve function, while showing limited effects on systemic metabolic parameters.

CONCLUSION

EA protects against T2DPN and is associated with SIRT1/PGC-1α-mediated mitochondrial biogenesis and redistribution, with neural benefits partly dissociated from metabolic regulation.

Key Words: Acupuncture; Diabetic peripheral neuropathy; Type 2 diabetes; Mitochondria distribution; Mitochondria biogenesis

Core Tip: Current treatments for diabetic peripheral neuropathy (DPN) remain limited, and new therapeutic strategies are needed. This study suggests that electroacupuncture (EA) exerts protective effects in type 2 DPN rats and may be associated with silent information regulator 1 (SIRT1)/peroxisome proliferator-activated receptor-γ coactivator-1α-mediated mitochondrial biogenesis and mitochondrial redistribution, with more prominent changes observed in intraepidermal nerve fibers than in adjacent keratinocytes. Notably, inhibition of SIRT1 weakened the neural benefits of EA while having limited effects on systemic metabolic parameters, suggesting a partial dissociation between neural and metabolic effects.

INTRODUCTION

Type 2 diabetes mellitus (DM) (T2DM) is one of the main causes of morbidity and mortality worldwide. As the most prevalent complication of T2DM, diabetic peripheral neuropathy (DPN) usually presents as bilateral and symmetric damage with distal-to-proximal gradient of severity, of which the complexity is also reflected in the simultaneous emergence of pain hypersensitivity and numbness. China bears one of the heaviest diabetes burdens worldwide, and DPN is also highly prevalent in the Chinese population. A nationwide cross-sectional study in mainland China found that 57.2% of patients with T2DM had painful DPN, highlighting the substantial clinical burden of DPN in China[1,2]. Unfortunately, effective treatments for DPN that can be applied clinically remain limited at present.

The focus of research on DPN is shifting from specific dysregulated pathways to global nerve metabolism and energy support imbalance[3]. The paralgesia of DPN is blamed on the vulnerability of axons, which is closely related to the role of mitochondrial such as the function, integrity, movement, distribution, volume, number, biogenesis, etc.[4-6]. The defect of the antioxidant defense mechanism induced by hyperglycemia results in a higher requirement of mitochondria biogenesis to maintain functional homeostasis[7]. Emerging evidence suggests that mitochondria damage and defective biogenesis in peripheral nerves as the basis of the pathogenesis of DPN[8]. Together with the important role of axonal mitochondria in demyelinating lesions, these findings highlight the necessity of investigating mitochondrial distribution in diabetes and its regulation.

As a key transcriptional regulatory axis, the silent information regulator 1 (SIRT1)/peroxisome proliferator-activated receptor-γ coactivator-1α (PGC-1α) axis has been evidenced in numerous studies for the importance of mitochondria functional regulation, such as oxidative stress and bioenergy maintenance in DPN[9,10]. SIRT1 acts as a bioenergetic sensor and can respond sensitively to the energy change of cells. Through deacetylating targets like PGC-1α and regulating its activity, SIRT1 mediates mitochondrial biogenesis and energy metabolism adaptation in the nerve system, which thus supplements the increased energy gap in diabetes. Since the downregulation of SIRT1 was widely reported in T2DM and aggravates DPN’s progression, it was regarded as a potential therapeutic target for treatment[11,12].

Previous research pointed out that electroacupuncture (EA) has a positive regulatory effect on metabolic disorders, but whether it also affects related neural function remains unknown. The therapeutic effect of EA on neuropathy caused by various factors have aroused our interest, combined with its promotion of nerve regeneration inspires us to explore the underlying mechanisms especially DPN. EA has been reported to exert therapeutic effects on neuropathy and neuropathic pain via the regulation of mitochondria function and biogenesis[13-15]. Recent studies evidenced that PGC-1α expression positively rearranges mitochondrial axonal transport and distribution, which indicated the potential relationship between the EA positive effect and the mitochondrial redistribution, apart from the biogenesis[16]. In this study, we investigated whether EA treatment ameliorated peripheral neuropathy in high-fat diet (HFD)-streptozotocin (STZ)-induced T2DM rats, including glucose metabolic disorder, neuropathic pain, and nerve dysfunction. We further explored the potential mechanisms from the perspective of mitochondrial biogenesis and distribution, particularly in intraepidermal nerve fibers (IENFs), to better understand the neural mechanisms underlying its therapeutic effects.

MATERIALS AND METHODS

Animals

Male Sprague-Dawley rats, aged 8 weeks and weighing approximately 200-220 g, were purchased from the Model Animal Research Center of Nanjing Medical University (No. 202110A001) and housed in a temperature-controlled room under a 12 hours/12 hours light/dark cycle. A total of 40 rats were used in this study. The animals were randomly assigned to four groups, with 10 rats in each group: Control group, type 2 DPN (T2DPN) model group, EA group, and EA + selisistat (EX-527) group. The sample size was determined based on our preliminary experiments and previous studies using similar T2DPN rat models and EA interventions. Ten rats per group were used to meet the needs of behavioral, electrophysiological, and morphological analyses. Each rat was considered an independent biological replicate. All animal-related protocols were approved by the Institutional Animal Care and Use Committee of Nanjing University of Chinese Medicine, and all animals received humane care.

Preparation of T2DM rats with DPN

Animal models established by low-dose STZ intraperitoneal injection combined with a HFD were used in this study for DPN research, as they simulate the metabolic features and progression of human diabetes.

Type 2 diabetic rats with DPN were prepared using a HFD and low-dose STZ treatment as described previously[17,18]. Briefly, after 4 weeks of feeding with different diets, rats received an intraperitoneal injection of STZ (35 mg/kg, 0.1 M citric acid buffer, potential of hydrogen = 4.5), while nondiabetic rats received an equivalent volume of citrate buffer as controls. One week after STZ injection, rats with blood glucose levels > 16.6 mmol/L were defined as T2DM models. Five weeks after STZ injection, T2DM rats with peripheral neuropathy were identified through behavioral tests, including hind paw withdrawal threshold and latency. During the experiment, diabetic rats were fed a HFD containing 58% fat, while nondiabetic rats were fed a standard diet[19] (Supplementary Figure 1).

EA treatment

EA treatment was performed as described previously[20,21]. Briefly, rats were placed on a heating pad to maintain body temperature and anesthetized with isoflurane (2%-5%). EA treatment was conducted using pairs of unipolar stainless-steel needles at bilateral ST25 (Tianshu, located 5 mm lateral to the intersection between the upper two-thirds and lower one-third of the line connecting the xiphoid process and the upper border of the pubic symphysis). The electrical current was set at 2 mA with an alternating frequency of 2/15 Hz. EA treatment was performed for 20 minutes per day, 6 days per week, for 6 consecutive weeks throughout the experiment (Supplementary Figure 1).

EX-527 administration

EX-527, a selective SIRT1 inhibitor, was administered as described previously[22]. EX-527 was first dissolved in dimethyl sulfoxide and then diluted with physiological saline to a final concentration of 1 mg/mL. It was administered by intraperitoneal injection at a dose of 5 mg/kg every other day during the EA intervention period.

Behavioral tests

Behavioral tests were conducted 5 weeks after STZ injection and again after 6 weeks of EA treatment. All behavioral measurements were performed when the rats were awake and unrestrained. Behavioral assessments were conducted by an investigator blinded to group allocation.

Withdrawal threshold test: Rats were placed on a metal grid 30 minutes before the test for adaptation. A Von Frey filament was applied to the hind paw and raised at a constant speed with uniformly increasing force until paw withdrawal occurred. The corresponding force was recorded as the withdrawal threshold[23,24].

Withdrawal latency test: Rats were placed on a warm plate at 30 °C, with a movable heat source positioned under the plantar surface of the hind paw. The temperature was increased from 30 °C to 55 °C to avoid skin scalding. Paw withdrawal time was recorded to evaluate thermal sensitivity[25].

Metabolic phenotyping measurement

Weight, blood glucose, and glucose tolerance test measurement: Body weight and fasting blood glucose were measured weekly using established methods. For the glucose tolerance test (GTT), glucose was injected intraperitoneally at a dose of 2 g/kg body weight, and blood glucose levels were recorded using a glucometer (Roche Diagnostics, Mannheim, Germany) at 0 minute, 30 minutes, 60 minutes, 90 minutes, 120 minutes, and 150 minutes after injection.

Biochemical measurement: As described previously, serum insulin levels were measured using a corresponding assay kit according to the manufacturer’s protocol (Nanjing Jiancheng Bioengineering Institute, Nanjing, Jiangsu Province, China).

Nerve conduction measurement

Nerve conduction velocities (NCVs) of the sciatic nerve from the sciatic notch to the ankle were recorded using a PowerLab 8/35 system (ADInstruments, Australia). Briefly, rats were anesthetized, and body temperature was maintained at 34 °C using a heating lamp. The recording electrode was placed in the plantar interosseous muscle to record responses to electrical stimulation delivered at the sciatic notch and ankle, respectively, and the time difference was calculated. The distance between the sciatic notch and the ankle was then divided by the time difference between M waves to obtain the final motor NCV (MNCV), and by the time difference between H waves to obtain the final sensory NCV (SNCV).

Immunofluorescence and image analysis

Tissue preparation: For immunofluorescence staining of the skin, rats were euthanized with an overdose of isoflurane and transcardially perfused with 250 mL phosphate-buffered saline (PBS) and 250 mL 4% paraformaldehyde (PFA) at 4 °C. Plantar skin from the hind paws was separated and soaked in 4% PFA for 24 hours. After washing with PBS three times and immersion in 30% sucrose solution at 4 °C for 24 hours, skin tissues were embedded in optimal cutting temperature compound and frozen at -20 °C until further analysis.

For immunofluorescence staining of sciatic nerve fibers, rats were euthanized with an overdose of isoflurane, and the middle segment of the sciatic nerve was immediately dissected. The nerve segment was placed onto filter paper of a suitable size and fixed in 4% PFA at room temperature for 1 hour. After washing three times, the nerve segment was transferred to PBS, and the epineurium and perineurium were removed under a dissecting microscope. The nerve fibers were then separated into small bundles consisting of several nerve fibers using two acupuncture needles. A drop of PBS was placed on a slide, and a bundle of nerve fibers was transferred into the drop. The liquid was gently removed with filter paper, and the sample was stored at -20 °C until further analysis.

Immunofluorescence: All related antibodies are summarized in Supplementary Table 1. For analysis of related indicators in hind paw skin, floating immunofluorescence staining was performed based on published approaches[26]. Briefly, 40-μm-thick sections were cut using a cryotome and transferred to 96-well plates. Then, 150 μL of Image-IT FX Signal Enhancer, 1 × PBS, blocking solution containing 5% bovine serum albumin (BSA) and 0.3% TX-100, diluted primary antibodies, rinsing solution containing 1% BSA and 0.3% TX-100, diluted secondary antibodies, and rinsing solution were added to different wells, and the sections were sequentially transferred to the corresponding wells. After removal from the rinsing solution, the sections were placed on slides, and a drop of ProLong Gold antifade mounting reagent with 4’,6-diamidino-2-phenylindole (DAPI) was added to the sections. A glass coverslip was then placed over the sections, and the slides were kept in the dark until further analysis.

For analysis of related indicators in the sciatic nerve, slides with teased nerve fibers were washed three times with PBS after permeabilization and blocking. Diluted primary antibodies, 150 μL, were added to the slides and incubated at 4 °C overnight. After washing with PBS, 150 μL of diluted secondary antibodies was added to the slides and incubated at room temperature for 2 hours. ProLong Gold antifade mounting reagent with DAPI was then dropped onto the slides, and a glass coverslip was placed over the samples. The slides were kept in the dark until further analysis.

Confocal analysis: For analysis of IENFs, the epidermal-dermal junction was outlined by one investigator blinded to group allocation, and nerve fibers crossing the boundary were counted independently by a second blinded investigator. IENF density was calculated as the number of nerve fibers divided by the length of the epidermal-dermal junction, based on published research[27].

For quantification of epidermal nerve-specific or keratinocyte-specific mitochondria, a three-dimensional (3D) analysis technique was applied based on a published description[28]. Briefly, Alexa Fluor 488 and Alexa Fluor 594 were used to detect green and red fluorescence signals, respectively, and 3D z-series images were acquired at 1 μm z-step intervals. For each animal, 20-25 z-stack images were selected for 3D reconstruction and quantitative analysis. Imaris software (version 9.0) was used to visualize the 3D image sets and perform surface rendering. Nerve and keratinocyte surfaces were created independently according to their corresponding fluorescence signals to avoid cross-contamination. Mitochondrial signals were then isolated within each independently defined surface, and nerve-specific and keratinocyte-specific mitochondrial surfaces were generated separately. Surface rendering was performed using absolute fluorescence intensity thresholds, with the same thresholding criteria applied consistently across images within the same analysis. The generated mitochondrial surfaces were used to measure mitochondrial volume.

Transmission electron microscopy and image analysis: Sciatic nerve samples from different groups were dissected and fixed in 2.5% glutaraldehyde overnight, washed with 0.1 M phosphate buffer, and postfixed in 1% osmium tetroxide overnight at 4 °C. Specimens were then dehydrated in a graded ethanol series and embedded in 100% epoxy embedding medium. Thin sections were stained with uranyl acetate and lead citrate. Images were acquired using an electron microscope, and ImageJ was used to calculate the total number of nerve fibers in relatively independent fields. Transmission electron microscopy (TEM) image evaluation and quantitative analysis were performed by investigators blinded to group allocation.

Western blotting

Western blot analysis of sciatic nerve tissues was conducted as described previously[29]. The corresponding antibodies are summarized in Supplementary Table 1. Blots were quantified using ImageJ software (NIH, Bethesda, MD, United States).

Statistical analysis

All data are presented as the mean ± SEM. Unpaired Student’s t-test was used for two-group comparisons, and one-way analysis of variance was used for multiple-group comparisons. Statistical significance was defined as P < 0.05.

RESULTS

EA alleviates sensory hypersensitivity and improves nerve function in T2DPN rats

STZ-HFD rats exhibit typical features of T2DPN: To investigate glucose metabolism in rats after STZ-HFD induction, we examined related basic indicators. After STZ injection, rats in the T2DPN group displayed characteristic features of T2DM, including higher blood glucose levels and lower body weight compared with the control group (Figure 1A and B). Blood glucose levels in the T2DPN group remained high throughout the study (Figure 1A). Meanwhile, the body weight of rats in the T2DPN group increased more slowly than that in the control group, and this difference gradually increased, as body weight in the model group stopped increasing while that in the control group continued to increase steadily (Figure 1B).

Figure 1 Streptozotocin-high-fat diet-induced rats exhibit significant characteristics of type 2 diabetic peripheral neuropathy. A and B: Blood glucose (A) and body weight (B) of rats in the control and type 2 diabetic peripheral neuropathy (T2DPN) groups after streptozotocin (STZ) injection; C and D: Glucose tolerance test (C) and the related area under the curve (D) in the control and T2DPN groups; E-I: Serum levels of insulin (E), triglyceride (F), non-esterified fatty acid (G), low-density lipoprotein (H), and high-density lipoprotein (I) in the control and T2DPN groups; J: Withdrawal threshold; K: Withdrawal latency after STZ injection. Data are presented as mean ± SEM (n = 4). ^aP < 0.05. ^bP < 0.01. P indicated type 2 diabetic peripheral neuropathy group vs control group. STZ: Streptozotocin; T2DPN: Type 2 diabetic peripheral neuropathy; GTT: Glucose tolerance test; AUC: Area under the curve; Ctrl: Control; INS: Insulin; TG: Triglyceride; NEFA: Non-esterified fatty acid; LDL: Low-density lipoprotein; HDL: High-density lipoprotein.

Glucose tolerance is an important indicator for evaluating glucose sensitivity. Intraperitoneal glucose tolerance test (IPGTT) data showed that 30 minutes after glucose injection, blood glucose levels in the T2DPN group increased significantly and remained elevated for 120 minutes (Figure 1C and D). Together with the increase in serum insulin levels, these results suggested glucose metabolic disorder in the T2DPN group (Figure 1E).

Hyperlipidemia is another characteristic manifestation of T2DM. Detection of serum glucose and lipid metabolism indicators showed that, compared with the control group, rats in the T2DPN group had higher blood lipid levels, including triglyceride (Figure 1F), non-esterified fatty acid (Figure 1G), low-density lipoprotein (Figure 1H), and high-density lipoprotein (Figure 1I), suggesting that rats in the T2DPN group exhibited typical characteristics of T2DM.

Hyperalgesia or allodynia is a major behavioral manifestation of peripheral neuropathy. Mechanical and thermal pain responses were tested to assess behavior and determine whether STZ-HFD-induced rats developed peripheral neuropathy. As shown in Figure 1J and K, rats in the T2DPN group exhibited a decreased hind paw withdrawal threshold and decreased withdrawal latency after STZ-HFD induction. In summary, STZ-HFD-induced rats showed typical characteristics of T2DPN.

EA has a regulatory effect on peripheral neuropathy and metabolic disorders: Six-week ST25 EA treatment was performed 5 weeks after STZ injection in the EA group, and withdrawal threshold and latency were also measured. A gradual increase in mechanical withdrawal threshold and latency was observed during EA intervention (Figure 2A and B), suggesting a positive effect on hyperalgesia. To further explore the effects of EA, NCV levels were measured at the endpoint of the experiment. As shown in Figure 2C and D, EA increased MNCV and SNCV levels in T2DPN rats, indicating a regulatory effect on neurological function.

Figure 2 Electroacupuncture treatment has regulatory effect on peripheral neuropathy and metabolic disorders in streptozotocin-high-fat diet-induced rats. A-D: Behavioral and neurological function was assessed by withdrawal threshold (A), withdrawal latency (B), motor nerve conduction velocities (C), and sensory nerve conduction velocities (D) in different groups after electroacupuncture (EA) treatment; E-K: Basic metabolic indices included blood glucose (E), body weight (F), triglyceride (G), non-esterified fatty acid (H), intraperitoneal glucose tolerance test (I) and area under the curve (J), and insulin (K) in different groups after EA treatment. Data are presented as mean ± SEM (n = 4). ^cP < 0.05. ^dP < 0.01. P indicated electroacupuncture group vs type 2 diabetic peripheral neuropathy group. T2DPN: Type 2 diabetic peripheral neuropathy; EA: Electroacupuncture; MNCV: Motor nerve conduction velocities; SNCV: Sensory nerve conduction velocities; TG: Triglyceride; NEFA: Non-esterified fatty acid; IPGTT: Intraperitoneal glucose tolerance test; AUC: Area under the curve; INS: Insulin.

Compared to the T2DPN group, the level of blood glucose in the EA group decreased significantly after treatment while the average body weight increased slightly with no significance (Figure 2E and F). Data shows that EA has varyingly improved hyperlipidemia levels (Figure 2G and H). Regarding IPGTT and area under the curve, EA shortens the process of glucose returning to the baseline (Figure 2I and J), suggesting the tolerance and regulation of high blood glucose levels were improved after EA treatment. Further, explores the regulatory effect of EA on blood lipid levels. The insulin level was also reduced considerably after EA treatment (Figure 2K).

EA increases intraepidermal nerve fiber density and ameliorates sciatic nerve injury

Diabetic hyperalgesia is usually attributed to sensory nerve damage, which is reflected by a reduction in IENFs[30,31]. Immunostaining of hind paw skin showed a significant decrease in protein gene product 9.5-positive nerve fibers in the T2DPN group and a significant increase after EA treatment (Figure 3A and B), suggesting a protective effect on peripheral nerve fibers.

Figure 3 Electroacupuncture treatment increases intraepidermal nerve fiber density and ameliorates degeneration of sciatic nerve fibers in type 2 diabetic peripheral neuropathy rats. A: Representative images of protein gene product 9.5-positive intraepidermal nerve fibers (green) and 4’,6-diamidino-2-phenylindole staining (blue) in hind paw skin; B: Quantification of intraepidermal nerve fiber density in different groups. Data are presented as mean ± SEM (n = 3); C: Longitudinal hematoxylin-eosin-stained sections of the sciatic nerve in different groups; D: Representative transmission electron microscopy images of myelinated axons with pseudo-coloring of intact axons (blue mask) and degenerated axons (pink mask) (left), and magnified individual axons (right) in sciatic nerve sections from different groups; E: Representative images of unmyelinated axons (yellow mask) (left) and magnified Remak bundles (right) in sciatic nerve sections from different groups. Normal and damaged mitochondria are indicated by black arrows and arrowheads, respectively. Normal and increased lysosomes are indicated by blue arrows and arrowheads, respectively. Autophagosomes are indicated by magenta arrowheads; F: Related morphological analyses of axons per field; G: Percentage of abnormal fibers; H: Percentage of degenerated axons in different groups based on myelinated axons. Data are presented as mean ± SEM (n = 3). ^bP < 0.01 type 2 diabetic peripheral neuropathy group vs control group. ^cP < 0.05 electroacupuncture group vs type 2 diabetic peripheral neuropathy group. ^dP < 0.01 electroacupuncture group vs type 2 diabetic peripheral neuropathy group. T2DPN: Type 2 diabetic peripheral neuropathy; EA: Electroacupuncture; Ctrl: Control; IENF: Intraepidermal nerve fibers.

Hematoxylin-eosin (HE) staining of longitudinal sections of the sciatic nerve was used to assess the arrangement of nerve fibers. As shown in Figure 3C, nerve fiber arrangement in the T2DPN group was disordered and was improved to some extent after EA treatment. As axons usually degenerate in DPN, whether myelinated or unmyelinated, we further observed ultrastructural changes in nerve fibers by TEM using transverse sections of the sciatic nerve. As shown in Figure 3D-G, myelinated axons in the control group showed normal morphology, characterized by a clear lamellar structure and well-preserved axons (Figure 3D). In contrast, demyelinating degeneration of peripheral axons was observed in the T2DPN group, including vacuole-like deformation, separation of the lamellar structure, and shrunken axons. EA treatment improved the morphology of myelinated axons and ameliorated axonal atrophy. According to morphometric analysis of the myelin sheath, more than 50% of nerve fibers in the T2DPN group showed pathological changes, and approximately 20% of myelinated axons showed demyelination. After EA treatment, the percentages of abnormal fibers (Figure 3G) and degenerated axons (Figure 3H) were significantly decreased. No significant difference in the average number of axons per field was observed among the groups (Figure 3F).

Pathological changes also involved unmyelinated axons. In the T2DPN group, swollen and irregular axons, vacuolated mitochondria, and increased autophagic structures, including autophagosomes and lysosomes, were observed in Remak bundles. In contrast, normal morphology of unmyelinated axons was observed in the control group (Figure 3E).

EA upregulated the SIRT1/PGC-1α pathway and mitochondrial distribution in sciatic nerve fibers

The improvement in neural morphology and function mediated by EA treatment prompted us to explore the possible mechanisms. Dysfunction of mitochondrial biogenesis has been reported as a major mechanism and is increasingly regarded as a therapeutic target for peripheral neuropathy[32,33] since in DM, maintenance of energy supply requires the continuous production of new mitochondria. Previous studies have confirmed the important role of the SIRT1/PGC-1α axis in the regulation of mitochondrial function and biogenesis as well as in the improvement of DPN[34-36]. Western blot analysis of sciatic nerve tissue indicated that the expression levels of SIRT1, PGC-1α, and mitochondrial transcription factor A (TFAM) were decreased in T2DPN rats but increased after EA treatment (Figure 4), suggesting that EA treatment upregulated the SIRT1/PGC-1α pathway in T2DPN rats.

Figure 4 Electroacupuncture treatment upregulated the silent information regulator 1/peroxisome proliferator-activated receptor-γ coactivator-1α pathway in the sciatic nerve of type 2 diabetic peripheral neuropathy rats. A: Western blot analysis of silent information regulator 1 (SIRT1), peroxisome proliferator-activated receptor-γ coactivator-1α (PGC-1α), and mitochondrial transcription factor A (TFAM) in sciatic nerve tissue; B: Representative immunofluorescence images of pyruvate dehydrogenase (PDH)-stained mitochondria (red) in sciatic nerve fibers; C: Quantification of the mean fluorescence intensity of PGC-1α staining in sciatic nerve fibers; D: Representative immunofluorescence images of PGC-1α staining (green) in sciatic nerve fibers; E: Representative immunofluorescence images of PDH-stained mitochondria (red) at nodes of Ranvier marked by voltage-gated sodium channel 1.6 (NaV1.6) (green) in sciatic nerve fibers with axonal diameters of 5 μm and 2 μm. The white dotted frames indicate the positions of nodes of Ranvier, and the 2 × magnified images show mitochondria at the nodes of Ranvier; F-H: Quantitative analysis of SIRT1 (F), TFAM (G), and PGC-1α (H) protein expression in sciatic nerve tissue; I: Quantification of the mean fluorescence intensity of PDH-stained mitochondria in sciatic nerve fibers; J and K: Quantification of the relative mean fluorescence intensity of mitochondria at nodes of Ranvier in axons with diameters of 5 μm (J) and 2 μm (K). Glial fibrillary acidic protein was stained for glial cells (green in B and red in D), phalloidin was stained for axons (magenta in B, D, and E), and NaV1.6 was stained for nodes of Ranvier (green in E). Data are presented as mean ± SEM (n = 3). ^aP < 0.05 type 2 diabetic peripheral neuropathy group vs control group. ^bP < 0.01 type 2 diabetic peripheral neuropathy group vs control group. ^cP < 0.05 electroacupuncture group vs type 2 diabetic peripheral neuropathy group. ^dP < 0.01 electroacupuncture group vs type 2 diabetic peripheral neuropathy group. T2DPN: Type 2 diabetic peripheral neuropathy; EA: Electroacupuncture; Ctrl: Control; SIRT1: Silent information regulator 1; PGC-1α: Peroxisome proliferator-activated receptor-γ coactivator-1α; TFAM: Mitochondrial transcription factor A; Phall: Phalloidin; PDH: Pyruvate dehydrogenase; GFAP: Glial fibrillary acidic protein; NaV1.6: Voltage-gated sodium channel 1.6.

Mitochondria were stained and evaluated in sciatic nerve fibers. As shown in Figure 4, mitochondrial content was significantly decreased in the T2DPN group and was restored after EA treatment. Consistent with the Western blot analysis, immunofluorescence staining showed that EA treatment reversed the decreased PGC-1α expression in axons of sciatic nerve fibers (Figure 4). PGC-1α has been reported to positively affect axonal mitochondrial transport and distribution. The EA-mediated increase in PGC-1α expression prompted us to further investigate whether EA could regulate mitochondrial distribution in addition to mitochondrial biogenesis.

To further determine whether the decrease was also fiber structure-specific, we observed and compared mitochondria at the nodes of Ranvier. To identify mitochondria specific to the nodes of Ranvier, mitochondria, axons, and nodes of Ranvier in nerve fibers were co-labeled, and quantification was performed in sciatic nerve fibers with axonal diameters of 5 μm and 2 μm. As shown in Figure 4, mitochondrial content at the nodes of Ranvier in sciatic nerve fibers with different diameters was lower in the T2DPN group than in the control group. The increased mitochondrial content at the nodes of Ranvier after EA treatment further confirmed that EA may promote mitochondrial redistribution in axons.

EA exerts a greater effect on mitochondrial size distribution in IENFs than in keratinocytes

Mitochondria perform important cellular functions and play a central role in supporting peripheral nerve fibers. The high energy requirement and sensitivity of distal long axons depend on, and challenge, the flexible distribution and rapid transport of mitochondria[37,38], which prompted us to pay more attention to mitochondria in nerve endings. Here, mitochondria within IENFs were specifically stained, masked, and analyzed using 3D imaging and analysis techniques. As shown in Figure 5A-C, the proportion of smaller mitochondria (0.04-0.08 μm³) in IENFs was decreased in the T2DPN group, whereas the proportions of larger mitochondria (0.32-0.64 μm³, 0.64-1.28 μm³, and 1.28-2.56 μm³) were increased compared with the control group, indicating a significant shift of IENF mitochondria toward larger volumes. After EA treatment, this increasing trend in mitochondrial volume was ameliorated, suggesting an effect of EA treatment on mitochondrial size redistribution in IENFs.

Figure 5 Electroacupuncture has a greater effect on mitochondrial size distribution in intraepidermal nerve fibers than in keratinocytes. A: Representative immunofluorescence images showing protein gene product 9.5-positive intraepidermal nerve fibers (green), pyruvate dehydrogenase-stained mitochondria (red), and 4’,6-diamidino-2-phenylindole-stained nuclei (blue) in hind paw skin sections from rats in different groups. Micrographs of intraepidermal nerve fibers, keratinocytes, and their corresponding mitochondria outlined by white dotted and solid boxes are shown in the left and right panels, respectively. Nerve-specific mitochondrial surfaces (magenta) and keratinocyte-specific mitochondrial surfaces (turquoise) are shown in the left and right panels, respectively; B and C: Mitochondrial size distribution in intraepidermal nerve fibers was assessed by frequency histograms of the percentage distribution of nerve-specific mitochondrial volume, with fitted curves based on the mean value of each frequency (B), and heatmap of the percentage distribution of nerve-specific mitochondrial volume based on the control group data and plotted on a log₂ scale (C). A total of 473-509 mitochondria from 3 rats in each group were included; D and E: Mitochondrial size distribution in keratinocytes was assessed by frequency histograms of the percentage distribution of keratinocyte-specific mitochondrial volume, with fitted curves based on the mean value of each frequency (D), and heatmap of the percentage distribution of keratinocyte-specific mitochondrial volume based on the control group data and plotted on a log₂ scale (E). A total of 1486-1752 mitochondria from 3 rats in each group were included. Data are presented as mean ± SEM (n = 3). ^aP < 0.05 type 2 diabetic peripheral neuropathy group vs control group. ^bP < 0.01 type 2 diabetic peripheral neuropathy group vs control group. ^cP < 0.05 electroacupuncture group vs type 2 diabetic peripheral neuropathy group. ^dP < 0.01 electroacupuncture group vs type 2 diabetic peripheral neuropathy group. T2DPN: Type 2 diabetic peripheral neuropathy; EA: Electroacupuncture; Ctrl: Control; IENFs: Intraepidermal nerve fibers; Nuc: 4’,6-diamidino-2-phenylindole-stained nuclei; Mt: Mitochondrial.

The volume of mitochondria in keratinocytes adjacent to nerve fibers was also quantified to further investigate whether the changes in mitochondrial distribution and the effects of EA treatment were IENF-specific. As shown in Figure 5D and E, a similar mitochondrial size distribution was observed in the control group, whereas the proportion of small mitochondria (0.02-0.04 μm³) decreased slightly and that of large mitochondria (0.16-0.32 μm³) increased significantly in the T2DPN group, suggesting a possible shift of mitochondria toward larger volumes. EA treatment also affected mitochondrial size distribution in keratinocytes, although the shift appeared less marked than that observed in IENFs. The heatmap showing the proportions of different mitochondrial size ranges further demonstrated that, compared with mitochondria in keratinocytes, mitochondria specific to IENFs showed a more significant shift toward larger volumes, and EA treatment more effectively reversed this trend. In summary, in T2DPN rats, changes in mitochondrial size distribution and the corresponding effects of EA were more prominent in IENFs than in keratinocytes.

SIRT1 inhibition attenuates the effects of EA on nerve function but has limited effects on metabolic parameters

Previous studies have reported the important role of mitochondrial biogenesis regulation in the development of DPN, and SIRT1 is a key regulatory target. To further investigate whether the effects of EA were related to the regulation of SIRT1, EX-527 was administered by intraperitoneal injection during EA treatment to suppress SIRT1 expression (Supplementary Figure 1). As shown in Figure 6, compared with the EA group, rats in the EA + EX-527 group showed decreased withdrawal threshold and latency (Figure 6A and B) and reduced MNCV (Figure 6C), suggesting that SIRT1 inhibition attenuated the beneficial effects of EA on nerve function. Basic metabolic indices were also examined. Compared with the EA group, blood glucose and lipid levels in the EA + EX-527 group were slightly elevated, but without statistical significance (Figure 6D-F). In conclusion, inhibition of SIRT1 weakened the effects of EA on nerve function.

Figure 6 Selisistat injection attenuated the regulatory effects of electroacupuncture on nerve function and inhibited mitochondrial biogenesis in the sciatic nerve. A-C: Behavioral and neurological function was assessed by withdrawal threshold (A), withdrawal latency (B), and motor nerve conduction velocity (C) after electroacupuncture treatment and selisistat (EX-527) injection; D-F: Basic metabolic indices, including blood glucose (D), body weight (E), and total cholesterol (TC) (F), in different groups after electroacupuncture treatment and EX-527 injection; G: Representative Western blot images of silent information regulator 1 (SIRT1), peroxisome proliferator-activated receptor-γ coactivator-1α (PGC-1α), and mitochondrial transcription factor A (TFAM) in the sciatic nerve; H-J: Quantitative analysis of the relative protein levels of SIRT1 (H), PGC-1α (I), and TFAM (J) in the sciatic nerve. Data are presented as mean ± SEM (n = 3-5). ᵇP < 0.01, type 2 diabetic peripheral neuropathy group vs control group; ᵈP < 0.01, electroacupuncture group vs type 2 diabetic peripheral neuropathy group; ᵉP < 0.05, electroacupuncture plus selisistat group vs electroacupuncture group; ᶠP < 0.01, electroacupuncture plus selisistat group vs electroacupuncture group. EA: Electroacupuncture; EA + E: Electroacupuncture plus selisistat; MNCV: Motor nerve conduction velocity; T2DPN: Type 2 diabetic peripheral neuropathy; Ctrl: Control; TC: Total cholesterol; SIRT1: Silent information regulator 1; PGC-1α: Peroxisome proliferator-activated receptor-γ coactivator-1α; TFAM: Mitochondrial transcription factor A.

SIRT1 inhibition differentially affects mitochondrial changes in sciatic nerve fibers and IENFs

Analysis of sciatic nerve tissue showed that SIRT1 protein expression was significantly decreased after continuous EX-527 injection (Figure 6G), suggesting effective inhibition of SIRT1 expression. We also examined the levels of PGC-1α and TFAM, and the results showed that PGC-1α and TFAM protein expression levels were reduced compared with the EA group (Figure 6G-J). In conclusion, SIRT1 inhibition attenuated EA-induced mitochondrial biogenesis in the sciatic nerve.

Immunofluorescence staining was used to examine mitochondrial levels in sciatic nerve fibers. The results showed that, compared with the EA group, mitochondrial levels were also decreased to varying degrees after SIRT1 inhibition (Figure 7A-C). Finally, we examined whether SIRT1 inhibition affected the regulation of mitochondria by EA in IENFs (Figure 7D). Interestingly, compared with the EA group, SIRT1 inhibition significantly increased mitochondrial levels in IENFs (Figure 7E). Enlarged images suggested possible mitochondrial clustering in the EA + EX-527 group, which may partly account for the increased fluorescence intensity.

Figure 7 Selisistat injection decreased mitochondrial levels at nodes of Ranvier but increased mitochondrial levels in intraepidermal nerve fibers. A-C: Mitochondrial levels at nodes of Ranvier were assessed by representative immunofluorescence images of pyruvate dehydrogenase (PDH)-stained mitochondria (red) at nodes of Ranvier marked by voltage-gated sodium channel 1.6 (green) in sciatic nerve fibers with axonal diameters of 5 μm (A) and 2 μm (B), and quantification of relative mean fluorescence intensity (C). The white dotted frames indicate the positions of nodes of Ranvier; D and E: Mitochondrial levels in intraepidermal nerve fibers were assessed by representative immunofluorescence images of PDH-stained mitochondria (red) in intraepidermal nerve fibers marked by protein gene product 9.5 (green) (D), and quantification of relative mean fluorescence intensity (E). ^fP < 0.01 electroacupuncture plus selisistat group vs electroacupuncture group. EA: Electroacupuncture; EA + E: Electroacupuncture plus selisistat; PDH: Pyruvate dehydrogenase.

DISCUSSION

This study explored the potential molecular mechanisms underlying the protective effects of EA treatment against T2DM-induced neuropathy. Based on the beneficial effects of EA, we further investigated the possible mechanisms from the perspective of mitochondrial biogenesis in the sciatic nerve, which were further verified by SIRT1 inhibition.

EA has been widely used in the treatment of DM and has been reported to have advantages in T2DM and its complications[39,40]. Rigorous glucose control is effective in the prevention and treatment of diabetic neuropathy in T1DM, but its effects are limited in T2DM. Published studies suggest that this difference may be related to distinct underlying mechanisms, such as glucose metabolic disorder. ST25 is traditionally used for gastrointestinal and metabolic disorders. Recent studies have shown that EA at ST25 can modulate systemic metabolic function through specific neuroanatomical pathways[21], and the previous work also demonstrated a hypoglycemic effect of EA at ST25 through neural regulation of the pancreatic intrinsic nervous system[20]. Compared with acupoints more commonly used for DPN, such as ST36 and SP6, the present study focused on ST25 because of its potential role in neuro-metabolic regulation. In this study, low-dose STZ injection combined with a HFD was used to establish a model with characteristics similar to those of human T2DPN. EA improved blood glucose levels, GTT results, serum insulin levels, and blood lipid indicators, indicating positive regulatory effects on glucose metabolism and dyslipidemia.

After observing that EA improved nerve function and regulated hyperalgesia, we examined the structure of the sciatic nerve. HE staining and electron microscopy showed that EA improved the disordered arrangement and demyelination of sciatic nerve fibers (Figure 3). It has been reported that, under continuous stimulation by chronic hyperglycemia, disorders of multiple metabolic pathways promote oxidative damage and energy stress, which aggravate mitochondrial dysfunction and form a vicious cycle, eventually culminating in direct axonal injury[41-43]. The positive effects of EA on neural structure and function prompted further observation of neural mitochondria, and the results showed that EA improved the pathological manifestations of mitochondrial damage in nerve fibers (Figure 3).

A stable and sufficient mitochondrial level appears to be important for meeting the energy demands of nerves. When discussing axonal vulnerability, the role of mitochondria, including their function and localization, should not be ignored. Published studies have confirmed that upregulation of mitochondrial biogenesis may alleviate sciatic nerve axonal injury[44], which prompted this study to focus on the role of mitochondrial biogenesis in neuropathy.

SIRT1 is a nicotinamide adenine dinucleotide-dependent deacetylase that regulates multiple physiological processes, including energy homeostasis, cell survival, and metabolic sensitivity. Several studies have reported that upregulation of SIRT1 mediates the therapeutic effects of EA in metabolic disorders, neuropathy, and cognitive impairment[45,46]. We confirmed that EA upregulated SIRT1 levels in the sciatic nerve of rats with DPN, suggesting that SIRT1 may mediate the therapeutic effects of EA. In addition, the role of SIRT1 has been increasingly studied in neurodegeneration. Previous research has suggested that activation and upregulation of SIRT1 may improve sciatic nerve pathology and relieve neuropathic pain. Considering that SIRT1 can modulate inflammatory responses and oxidative stress through critical transcription factors, we speculate that the recovery of mitochondrial levels after EA may result from the comprehensive improvement of multiple cascade processes associated with increased SIRT1 levels.

PGC-1α is an important coactivator of many metabolic transcription factors and is closely associated with TFAM, which has been proven to be essential for mitochondrial DNA transcription[47,48]. Upregulation of PGC-1α protects against neurodegeneration and neuroinflammation caused by various factors, and its interaction with SIRT1 coordinately regulates many physiological processes, including mitochondrial biogenesis. Activation of the SIRT1/PGC-1α axis has potential neuroprotective effects in neuropathy, whereas impairment of the SIRT1/PGC-1α axis may aggravate neuropathy development by decreasing TFAM levels, suggesting a protective role of this axis in the prevention of peripheral neuropathy[49,50]. Based on the regulation of metabolism and the protection of nerve function after EA treatment, we examined and confirmed the increased expression of this axis in sciatic nerve tissue, suggesting that EA may alleviate DPN by upregulating the SIRT1/PGC-1α/TFAM pathway.

PGC-1α overexpression has recently been reported to promote the expression of key factors involved in mitochondrial transport, adaptor binding, and axonal maintenance[16], thereby improving axonal mitochondrial transport and distribution. This positive effect of PGC-1α may partly explain the effects of EA treatment on axonal mitochondrial distribution. In the present study, mitochondria and PGC-1α staining in sciatic nerve fibers showed consistent changes, especially in axons (Figure 4), suggesting that EA treatment may regulate mitochondrial distribution in axons along with its effects on the mitochondrial biogenesis axis.

The high energy demand of distal axons is usually met by local stationary mitochondria in internodes rather than by mitochondria newly supplied from the neuronal soma. The spatial heterogeneity of axonal energy demand is related to the physiological functions of specific structures. In fact, more mitochondria are allocated to regions of long axons with particularly high energy requirements, such as axonal branches and nodes of Ranvier. After observing that EA treatment restored the decreased mitochondrial distribution in axons, we further explored whether this regulation of mitochondrial distribution was also locally heterogeneous. In myelinated axons, nodes of Ranvier contain high-density ion pumps, which may correspond to locally elevated energy demand. Considering the special physiological functions of nodes of Ranvier, including saltatory conduction of action potentials, we next focused on mitochondria at the nodes of Ranvier. The increased mitochondrial content at the nodes of Ranvier in DPN rats after EA treatment indicates heterogeneous regulation of mitochondrial distribution, and this regulation appears to be consistent with the pattern of mitochondrial distribution under physiological conditions. These findings suggest that EA may help normalize the abnormal pattern of mitochondrial biogenesis and distribution in T2DPN.

After observing the possible effects of EA on mitochondrial distribution in the sciatic nerve, we further explored the regulation of mitochondria by EA in IENFs. Published studies have reported length-dependent differences in mitochondrial distribution in patients with DPN, indicating that mitochondrial dynamic disorder may aggravate the development of DPN[26]. Based on the regulation of mitochondrial biogenesis and distribution in the sciatic nerve, the possible effects of EA on mitochondrial distribution were further examined in IENFs. The larger mitochondrial signals observed in IENFs in the model group may provide evidence of mitochondrial swelling.

Several studies have reported that mitochondrial swelling is closely related to neuropathy and is regarded as a typical pathological feature[51,52]. Multiple studies have reported swollen mitochondria in DM models and suggested that this change may be related to increased mitochondrial membrane permeability caused by oxidative stress[53-56]. The smaller mitochondrial volume in the EA group compared with the model group may be due to upregulation of the SIRT1/PGC-1α pathway, which is considered to regulate oxidative stress.

In addition, the large mitochondrial signals observed in IENFs may reflect altered mitochondrial distribution or possible clustering, as some of these signals appeared irregular in surface morphology. Evidence has shown that some sensory terminals contain high levels of mitochondria to meet the possibly high energy demand associated with complex cellular processes[52,57]. To support the timely transmission of sensory signals, IENFs are likely to have a high energy demand and may experience an energy crisis under metabolic disorder. Mitochondrial dysfunction and increased mitochondrial volume caused by hyperglycemia may inhibit normal mitochondrial transport in IENFs, which may lead to mitochondrial aggregation and reduced IENF density. Many studies have reported the contribution of mitochondrial dysfunction to neuropathy and pointed out that recovery of mitochondrial transport may be a potential therapeutic strategy for neuropathy[32,58]. Mitochondrial redistribution in IENFs further indicated a possible underlying mechanism of the therapeutic effects of EA on DPN.

To further investigate whether the activation of SIRT1-mediated mitochondrial biogenesis by EA is a key pathway for DPN treatment, EX-527 was administered by intraperitoneal injection during EA treatment to suppress SIRT1 expression. We observed that EX-527 administration significantly attenuated EA-induced improvements in mechanical withdrawal threshold and MNCV (Figure 6A-C). This functional impairment aligns with the established role of SIRT1 as a master regulator of mitochondrial biogenesis and cellular stress resistance, suggesting that EA enhances nerve function, at least in part, by upregulating SIRT1-dependent pathways. Notably, EX-527 injection had no significant impact on basic metabolic parameters, including blood glucose and lipid levels (Figure 6D-F), indicating that the neuroprotective effects of EA are likely not directly related to its systemic metabolic regulation. This dissociation is particularly meaningful and interesting, as it underscores the specificity of SIRT1 in mediating the neural effects of EA rather than its potential metabolic modulation. This represents a critical distinction for targeting DPN, especially given that neuropathy secondary to T2DM often persists despite glycemic control.

Meanwhile, we observed that EX-527 injection reduced the EA-induced increase in SIRT1 expression in sciatic nerve tissue, accompanied by decreased levels of PGC-1α and TFAM. These findings further support the hypothesis that EA promotes mitochondrial biogenesis in the sciatic nerve by activating SIRT1. Immunolabeling of sciatic nerve fibers revealed a marked reduction in mitochondrial density following SIRT1 inhibition, particularly at the nodes of Ranvier, which are critical for saltatory conduction and highly dependent on mitochondrial energy supply. This spatial specificity may explain the reduced MNCV levels, as compromised mitochondrial function at the nodes of Ranvier could impair axonal conduction velocity.

In summary, this study demonstrated the positive effects of EA treatment on T2DPN, including the regulation of glucose metabolism and nerve function. The underlying mechanisms may be related to the upregulation of mitochondrial biogenesis and mitochondrial redistribution after activation of the SIRT1/PGC-1α pathway (Figure 8). Further research may focus on the exact role of SIRT1 in the therapeutic effects of EA on DPN, which may provide a more detailed and convincing explanation.

Figure 8 Electroacupuncture regulates axonal mitochondrial distribution via silent information regulator 1/peroxisome proliferator-activated receptor-γ coactivator-1α-mediated mitochondrial biogenesis to improve type 2 diabetic peripheral neuropathy. Under type 2 diabetes mellitus conditions characterized by hyperglycemia, hyperlipidemia, and insulin resistance, peripheral nerve axons undergo mitochondrial dysfunction and structural damage, thereby contributing to demyelination. Electroacupuncture at ST25 alleviates peripheral neuropathy by activating the silent information regulator 1/peroxisome proliferator-activated receptor-γ coactivator-1α pathway, which promotes mitochondrial biogenesis and improves mitochondrial distribution in peripheral nerve axons and nodes of Ranvier. EA: Electroacupuncture; SIRT1: Silent information regulator 1; PGC-1α: Peroxisome proliferator-activated receptor-γ coactivator-1α; T2DPN: Type 2 diabetic peripheral neuropathy; Nrf1: Nuclear respiratory factor 1; TFAM: Mitochondrial transcription factor A; Ac: Acetyl.

Several limitations should be acknowledged. The present study did not include a sham-EA group or a vehicle-control arm for EX-527, which should be considered when interpreting the specificity of the EA effect and the pharmacological inhibition results. In addition, TEM-based quantitative parameters, such as g-ratio and mitochondrial circularity, were not assessed, and the interpretation of mitochondrial aggregation was based mainly on morphological and fluorescence features. Further validation using mitochondrial dynamics-related markers is needed in future studies.

CONCLUSION

Our study suggested that the beneficial effects of EA on T2DPN rats might be associated with activation of the SIRT1/PGC-1α axis. This pathway may promote mitochondrial biogenesis in the sciatic nerve and regulate mitochondrial redistribution, especially in IENFs, thereby contributing to the improvement of nerve dysfunction and peripheral nerve injury.

ACKNOWLEDGEMENTS

We are grateful to our colleagues at Jiangsu Medical College, the Key Laboratory of Acupuncture and Medicine Research of Ministry and the Experiment Center for Science and Technology in Nanjing University of Chinese Medicine for their support in the preparation of this manuscript.