Targeted neural stem cell-derived extracellular vesicles loaded with Sinomenine alleviate diabetic peripheral neuropathy via WNT5a/TRPV1 pathway modulation

Abstract

Background

Diabetic peripheral neuropathy (DPN) is one of the most prevalent and debilitating complications of diabetes, marked by chronic neuroinflammation, immune dysregulation, and progressive neuronal degeneration. Current treatments offer limited efficacy, largely focusing on symptomatic relief rather than addressing the underlying disease mechanisms. There is a critical need for disease-modifying therapies that target the molecular basis of DPN.

Results

In this study, we developed a novel targeted nanotherapeutic system—ZH-1c-EVs@SIN—composed of neural stem cell-derived extracellular vesicles (NSC-EVs) modified with the ZH-1c aptamer and loaded with the anti-inflammatory compound sinomenine (SIN). This system was specifically designed to target microglia and inhibit the WNT5a/TRPV1 signaling pathway. Transcriptomic profiling of microglia revealed key gene networks implicated in DPN pathology and responsive to SIN treatment. Functional assays demonstrated that ZH-1c-EVs@SIN facilitated a shift in microglial phenotype from pro-inflammatory M1 to anti-inflammatory M2, significantly reduced inflammatory cytokine expression, and restored levels of neuronal regulatory proteins. Nanoparticle tracking analysis and transmission electron microscopy confirmed optimal vesicle size and morphology, while fluorescence imaging showed efficient uptake by microglia. In vivo studies in a murine model of DPN revealed marked improvements in pain-related behavior and histopathological signs of nerve damage.

Conclusion

ZH-1c-EVs@SIN represents a promising therapeutic strategy for DPN, offering targeted immunomodulation and enhanced neural repair via regulation of the WNT5a/TRPV1 signaling axis. This nano-delivery platform introduces a novel and precise approach to intervening in diabetic neuropathy and may be applicable to other neuroinflammatory conditions.

Graphical abstract

Mechanism of ZH-1c-EVs@SIN Mediating the WNT5a/TRPV1 Pathway to Improve Immune-Inflammatory Homeostasis in the Treatment of DPN in Mice.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12951-025-03678-3.

Introduction

Diabetic peripheral neuropathy (DPN), impacting nearly half of individuals with diabetes, represents a prevalent and debilitating complication [1]. The main symptoms of DPN include pain, sensory abnormalities, numbness, and neurological dysfunction, all of which severely impair patients’ quality of life and functional independence [2–4]. The pathogenesis of DPN is complex, involving metabolic disorders, microvascular damage, oxidative stress, nerve injury, and immune-inflammatory responses [5]. Despite the availability of various drugs to relieve DPN-related pain and symptoms, effective treatments that promote nerve repair remain limited [6, 7]. With growing understanding of DPN pathophysiology, the role of immune-inflammatory homeostasis in disease progression has garnered increasing attention [8]. Accordingly, this study aims to explore innovative therapies that modulate immune-inflammatory responses to promote neural repair in DPN.

Sinomenine (SIN), a natural alkaloid with potent anti-inflammatory and neuroprotective properties, has demonstrated promise in the treatment of neurological disorders [9]. Studies show that SIN regulates inflammatory signaling, mitigates nervous system damage, and exerts therapeutic effects in various disease models [10–12]. Microglia, the primary immune cells of the central nervous system, exhibit dual functions in neurodegeneration: pro-inflammatory M1 microglia exacerbate neuronal damage, while anti-inflammatory M2 microglia facilitate repair [13]. Notably, SIN has been shown to suppress M1 microglial activation and promote the M2 phenotype, providing a strong theoretical basis for its application in DPN [14, 15]. However, SIN’s clinical use is hampered by poor bioavailability and limited targeting capacity [16–18]. Therefore, developing an efficient and targeted SIN delivery system is a critical focus of current research.

The WNT5a/TRPV1 signaling pathway plays a key regulatory role in neuroinflammation and neuropathic pain. WNT5a modulates microglial polarization and neural repair processes [19, 20], while TRPV1, a calcium ion channel, is involved in pain transmission and inflammatory responses [21]. Emerging evidence suggests a synergistic interaction between WNT5a and TRPV1 in regulating microglial activity, modulating inflammation, and promoting nerve regeneration [22, 23]. Targeting this signaling axis has shown promise for improving neural function and alleviating DPN-related pain [24, 25]. Therefore, elucidating the effects of SIN on the WNT5a/TRPV1 pathway holds significant translational and therapeutic value.

In recent years, extracellular vesicles (EVs) have emerged as innovative drug delivery systems due to their intrinsic ability to carry bioactive molecules—including proteins, RNAs, and lipids—with high biocompatibility and low immunogenicity. These properties enable EVs to cross biological barriers and deliver therapeutic agents to target tissues with precision [26–28]. In the context of neurological diseases, EVs have shown great potential for delivering neuroprotective agents, modulating the neural microenvironment, and promoting nerve regeneration [29, 30]. This study employs neural stem cell-derived extracellular vesicles (NSC-EVs) as carriers for SIN, capitalizing on their inherent neuroregenerative capabilities and microglial targeting properties. Compared to mesenchymal stem cell-derived EVs, NSC-EVs are enriched with neurotrophic factors (e.g., BDNF) and molecules associated with neural repair [29]. Therefore, NSC-EVs combined with SIN offer a dual advantage: efficient crossing of biological barriers and enhanced targeting of microglia in the damaged neural microenvironment. Furthermore, modification of NSC-EVs with aptamers can improve targeting specificity and enhance SIN delivery efficiency. This targeted approach facilitates the phenotypic transition of microglia from M1 to M2, ultimately promoting nerve repair and restoring immune-inflammatory balance.

The primary objective of this study is to construct a targeted delivery system—ZH-1c aptamer-modified NSC-derived EVs loaded with SIN (ZH-1c-EVs@SIN)—to modulate the WNT5a/TRPV1 signaling pathway in microglia, improve immune-inflammatory homeostasis, and promote neural repair in a murine DPN model. This approach not only provides new insights into the therapeutic potential of SIN but also underscores the clinical promise of EV-based drug delivery systems. By enhancing drug bioavailability and minimizing systemic side effects, EV-mediated targeted delivery offers a powerful strategy for the treatment of DPN and other neurological disorders. This research presents a novel therapeutic platform with implications for personalized medicine and precise neuroimmune modulation.

Results and discussion

Therapeutic effects of SIN on DPN mice

Previous studies have demonstrated that SIN has significant efficacy in preventing and alleviating neuroinflammation, as well as promoting the repair of nerve damage [31–33]. To evaluate its therapeutic potential in DPN, we first established a DPN mouse model (Fig. 1A). Compared with the control group, mice in the DPN group exhibited significantly increased body weight, reduced serum insulin levels, and elevated blood glucose levels. Treatment with SIN led to a significant increase in insulin levels, along with reduced body weight and lower blood glucose levels (Fig. 1B-D). Behaviorally, DPN mice demonstrated reduced paw withdrawal threshold (PWT) to mechanical stimuli and shortened paw withdrawal latency (PWL) to thermal stimuli—hallmarks of mechanical allodynia and thermal hyperalgesia, respectively. Notably, SIN treatment significantly increased both PWT and PWL (Fig. 1E-F).

Fig. 1.

Validation of SIN’s Therapeutic Effects on DPN Mice. (A) Schematic of the DPN mouse model; (B) Body weight measurements for each group of mice; (C) Serum insulin levels in each group; (D) Blood glucose levels in each group; (E) Mechanical allodynia test results; (F) Heat hyperalgesia test results; (G-H) Gait analysis results for each group of mice; (I) IF staining showing microglial activation in the lumbar spinal cord of mice. Data are presented as mean ± S.D., and comparisons among multiple groups were performed using one-way ANOVA, while data from animals at different time points were analyzed using two-way ANOVA. * indicates p < 0.05 between groups, ** indicates p < 0.01, *** indicates p < 0.001, ns indicates no significant difference between groups. N = 6 mice per group

Furthermore, we evaluated the impact of each treatment on motor function. As shown in Fig.s 1G-H, DPN mice exhibited a marked reduction in hind limb strength compared to the contralateral, unaffected limb. Gait analysis revealed a smaller paw contact area in DPN mice, likely reflecting pain-avoidant behavior and neuromuscular weakness. SIN-treated mice displayed improved hind limb function and a more normalized gait pattern. Quantitative analysis of blood flow and PGP9.5 (a marker of intraepidermal nerve fibers) in the footpad skin revealed a significant reduction in nerve fiber density in the DPN group (Fig. S1). In contrast, these symptoms were alleviated in the SIN-treated group. Additionally, we observed robust microglial activation in the spinal cord of DPN mice. In contrast, SIN-treated mice showed reduced microglial activation, comparable to levels observed in control animals (Fig. 1I).

These results confirm the successful establishment of the DPN mouse model and demonstrate the therapeutic effects of SIN on DPN mice.

SIN mediates the WNT5a/TRPV1 pathway to promote M2 polarization of microglia

Microglial polarization is increasingly recognized as a pivotal regulator of DPN pathogenesis [34–36]. To clarify its mechanistic role, we isolated microglia from lumbar spinal cords of Control, DPN, and SINtreated mice for highthroughput RNA sequencing. Comparison of Control versus DPN groups revealed 152 differentially expressed genes (DEGs) implicated in DPN progression (Fig. S2A). Comparison of DPN versus SIN groups identified 218 DEGs associated with SIN’s therapeutic action (Fig. S2B). Intersection analysis yielded 41 overlapping genes (Fig. S2C-D). Applying Least Absolute Shrinkage and Selection Operator (LASSO) regression, we pinpointed WNT5a and TRPV1 as key regulators (Fig. S2E-G). Notably, WNT5a is a potent endogenous agonist of TRPV1 [25], and both promote proinflammatory M1 microglial polarization [37, 38].

To probe causality, we overexpressed TRPV1 in microglia via lentiviral transfection; real-time quantitative polymerase chain reaction (RT-qPCR) and Western blot confirmed robust overexpression (Fig. S3A-C). Flow cytometry and immunofluorescence (IF) substantiated sequencing findings: relative to Control, highglucose (HG) microglia showed elevated WNT5a/TRPV1 and M1 markers with reduced M2 markers. SIN treatment reversed these changes, lowering WNT5a/TRPV1 levels, suppressing M1 polarization, and enhancing M2 polarization. Crucially, TRPV1 overexpression blunted SIN’s effects (Fig. 2A-D, Fig. S3D-E), indicating that SIN modulates polarization through the WNT5a/TRPV1 axis.

Fig. 2.

Regulation of Microglial Polarization by SIN. (A) IF staining showing M1 (CD86) and M2 (CD206) microglial markers in each group, scale bar = 25 μm; (B) Quantitative analysis of IF results; (C) The flow cytometry gating strategy was used to identify M1-type microglia (CD86) and M2-type microglia (CD206), and the proportions of CD86-positive and CD206-positive microglia in each group; (D) Statistical quantification of flow cytometry results; (E) Schematic representation of co-culture between microglia and neurons; (F) ELISA analysis of inflammatory cytokines in the supernatants from co-cultured cells; (G-H) TUNEL staining of neuronal apoptosis (H) and its statistical analysis (G), scale bar = 50 μm; (I-J) Flow cytometry analysis of neuronal apoptosis (I) and corresponding statistical results (J) in co-cultured cells. Data are presented as mean ± S.D., and comparisons among multiple groups were performed using one-way ANOVA.* indicates p < 0.05, ** indicates p < 0.01 between groups. All cell experiments were repeated three times

Given SIN’s reported neuroprotection [39], we examined whether its regulation of microglia indirectly benefits dorsal root ganglion (DRG) neurons. Microglia and DRG neurons were cocultured (Fig. 2E). Enzyme-linked immunosorbent assay (ELISA) quantified cytokines, while TdT-mediated dUTP Nick-End Labeling (TUNEL) staining and flow cytometry assessed neuronal apoptosis. HG conditions elevated proinflammatory cytokines (TNFα, IL-6) and neuronal apoptosis, and lowered antiinflammatory cytokines (TGFβ, IL-10). SIN + oeNC reduced proinflammatory cytokines and apoptosis while boosting TGFβ and IL-10 versus HG. Conversely, SIN + oeTRPV1 negated these antiinflammatory and neuroprotective gains (Fig. 2F-J).

These results indicate that SIN regulates microglial polarization through the WNT5a/TRPV1 pathway, exerting anti-inflammatory and neuroprotective effects.

SIN mediates the WNT5a/TRPV1 pathway to ameliorate DPN

To investigate the in vivo mechanism by which SIN mediates the WNT5a/TRPV1 pathway, we measured body weight, insulin levels, and blood glucose levels in the mice from different groups. Compared to the SIN + oe-NC group, the SIN + oe-TRPV1 group exhibited a notable increase in body weight and blood glucose levels, while insulin levels decreased (Fig. 3A-C). Additionally, the PWT and PWL values in the SIN + oe-TRPV1 group were distinctly higher than those in the SIN + oe-NC group, indicating diminished pain sensitivity (Fig. 3D-E).

Fig. 3.

Effects of SIN on the WNT5a/TRPV1 Pathway in DPN Mice. (A) Schematic representation of lentiviral injection in DPN mice; (B) Body weight measurements across groups of mice; (C) Serum insulin levels in each group; (D) Blood glucose levels in each group; (E) Mechanical allodynia test results; (F) Heat hyperalgesia test results; (G) Gait analysis results. Data are presented as mean ± S.D.; comparisons between two groups were performed using an independent samples t-test, and data from animals at different time points were analyzed using two-way ANOVA. * indicates p < 0.05 between groups. N = 6 mice per group

Next, we evaluated motor function across the groups. As shown in Fig. 3F, the SIN + oe-TRPV1 group displayed a larger paw contact area during standing, suggesting a notable improvement in limb weakness and neuropathic pain (Fig. 3G). Quantitative blood flow analysis and morphological data from PGP9.5 (a marker for intraepidermal nerve fibers) immunoreactivity in the footpad skin demonstrated a significant increase in nerve fiber density in the SIN + oe-TRPV1 group (Fig. S4A-B).

Western blot analysis revealed that TRPV1 overexpression evidently increased TRPV1 protein levels in spinal cord tissue, while WNT5a expression remained unchanged relative to the SIN + oe-NC group (Fig. S4C-D). To evaluate microglial polarization in vivo, immunofluorescence staining of spinal cord sections was performed. The DPN group exhibited elevated M1 (pro-inflammatory) microglia and reduced M2 (anti-inflammatory) microglia compared to controls. Conversely, the SIN + oe-TRPV1 group showed a significant increase in M2 microglial polarization and a reduction in M1 polarization (Fig. 4A-D). Primary DRG neurons were cultured from each group, and βIII-tubulin IF staining revealed that the length of neurites was evidently diminished in the DPN group compared to the Control group, accompanied by worsened neuroinflammation. However, in the SIN + oe-TRPV1 group, neurite length was distinctly longer than in the SIN + oe-NC group, and neuroinflammation in DRG neurons was markedly alleviated (Fig. 4E-H).

Fig. 4.

Effect of SIN on Microglial Polarization via the WNT5a/TRPV1 Pathway (A-B) IF staining showing M1 microglial marker expression (A) and M2 microglial marker expression (B), scale bar = 50 μm (original), scale bar = 10 μm (magnified); (C-D) Quantitative analysis of IF results; (E-F) IF staining showing the longest neurite length in primary DRG neurons (E) and its statistical analysis (F), β-III Tubulin is marked in red, 3 images were collected per mouse, scale bar = 25 μm; (G-H) IF analysis of iNOS expression in DRG neurons (G) and its statistical results (H), scale bar = 25 μm. Data are presented as mean ± S.D., and comparisons among multiple groups were performed using one-way ANOVA. * indicates p < 0.05 between groups, ** indicates p < 0.01. N = 6 mice per group

These results collectively confirm that SIN mediates the WNT5a/TRPV1 pathway to regulate microglial polarization, thereby improving DPN pathology.

Synthesis and characterization of ZH-1c-EVs@SIN

Based on the previous findings, we have demonstrated that SIN mediates the WNT5a/TRPV1 pathway to regulate microglial polarization and has therapeutic effects on DPN. To address the issue of significant activation of the WNT5a/TRPV1 pathway during DPN and the poor targeting ability of SIN, we aimed to develop a nanomaterial capable of targeting microglia in the lumbar spinal cord during DPN while also enhancing the therapeutic efficacy of SIN (Fig. 5A).

Fig. 5.

Synthesis and Characterization of ZH-1c-EVs@SIN. (A) Schematic diagram of ZH-1c-EVs@SIN synthesis; (B) Schematic representation of NSC-EVs isolation; (C) Size distribution of EVs (green) and EVs@SIN (red) as analyzed by NTA; (D) Loading efficiency of SIN into EVs@SIN; (E) Schematic representation of ZH-1c binding to microglia (left) and the structure of ZH-1c; (F) Agarose gel analysis demonstrating the conjugation of ZH-1c to EVs or EVs@SIN using EDC-NHS coupling; (G) Fluorescence microscopy showing the binding efficiency of ZH-1c to EVs in each group (scale bar: 1 μm); (H) Laser scanning microscopy showing the uptake of EVs by microglia, with red fluorescence representing DIR-labeled EVs and blue indicating DAPI-stained nuclei (scale bar: 25 μm). All cell experiments were performed at least three times. Data are expressed as mean ± S.D. Comparisons between two groups were performed using an independent samples t-test. * indicates p < 0.05 between groups

First, we isolated EVs from NSCs (Fig. 5B). Western blot analysis confirmed the presence of canonical EV markers (Alix, TSG101, CD81), validating the identity of the isolated vesicles (Fig. S5A). Transmission electron microscopy (TEM) and nanoparticle tracking analysis (nNTA) revealed the typical disc-shaped morphology and size distribution of EVs, ranging from 50 to 300 nm (Fig. S5B-C). Following SIN encapsulation into EVs, we performed dynamic light scattering (DLS) to assess size and surface charge. In addition, the zeta potential and size distribution of EVs, EVs@SIN, and ZH-1c-EVs were measured. The results showed that the mean size of the unmodified EVs was (110 ± 15) nm, with a size distribution ranging from 70 to 150 nm, which is characteristic of typical EVs. The zeta potential was (−35 ± 5) mV, indicating that the EV surface is negatively charged, which helps maintain stability in solution. After ultracentrifugation and NTA quantification, the EV concentration was (5.0 ± 1.0) × 10¹⁰ particles/mL. For EVs@SIN, the mean size increased to (125 ± 18) nm, which may be due to SIN loading altering the structure of the EVs or causing some aggregation. The zeta potential shifted to (−40 ± 6) mV, likely due to the chemical nature and distribution of SIN on the EV surface. The concentration decreased to (3.5 ± 0.8) × 10¹⁰ particles/mL, possibly because some EVs were lost during the loading process. For ZH-1c-EVs, the mean size further increased to (140 ± 20) nm as a result of aptamer modification, which may change the surface morphology and charge distribution. The zeta potential was (−42 ± 7) mV. After ultracentrifugation, the concentration was (3.0 ± 0.7) × 10¹⁰ particles/mL, indicating that aptamer modification and subsequent processing caused a certain degree of loss (Table S1).

While native EVs displayed a narrow size distribution, EVs@SIN showed increased heterogeneity and the presence of aggregates. The encapsulation efficiency of SIN was approximately 18%. Using high-performance liquid chromatography (HPLC), we quantified the actual SIN loading as 0.24 ± 0.03 µg per mg of EV protein (n = 3), corresponding to a suspension concentration of 0.73 µM. This data provides a direct reference for designing in vivo dosing (e.g., an injection of 100 µL EVs@SIN would deliver approximately 0.288 µg SIN), and combined with the 18% encapsulation efficiency, confirms that ZH-1c-EVs have a stable loading capacity for SIN (Fig. 5C-D).

ZH-1c has been demonstrated to be an aptamer molecule that targets the transmembrane protein CD64, which is highly expressed on the surface of activated microglia during neuroinflammation (Fig. 5E) [40]. To confirm the successful conjugation of the ZH-1c aptamer to EVs, agarose gel electrophoresis and fluorescence microscopy were performed. As shown in Fig. 5F, the mobility of the ZH-1c-EVs and ZH-1c-EVs@SIN was hindered due to the introduction of EVs, validating the attachment of the aptamer to the EV surface. Fluorescence microscopy in Fig. 5G further confirmed this observation.

To evaluate the targeting and uptake of EVs by microglia, DiR-labeled EVs were co-incubated with cultured microglia for 24 h. Laser scanning confocal microscopy revealed no detectable fluorescence in the PBS and SIN groups, while the remaining groups showed clear red fluorescence. Notably, the ZH-1c-EVs and ZH-1c-EVs@SIN groups exhibited the strongest fluorescence signals (Fig. 5H).

In summary, we successfully developed ZH-1c-EVs@SIN capable of precise SIN delivery.

ZH-1c-EVs@SIN promotes M2 polarization and inhibits M1 polarization of microglia

Next, we examined the effect of different EV treatments on microglial activation using IF staining. The IF results showed no significant differences in the activation levels of microglia across the groups, indicating that none of the EV treatments had cytotoxic effects or notably activated or inhibited microglial cells (Fig. S6A-C). However, when microglia were treated with the different EV formulations, the protein expression levels of WNT5a and TRPV1 were notably diminished in all groups except for the PBS, EVs, and ZH-1c-EVs groups. The ZH-1c-EVs@SIN group showed the most pronounced reduction in protein expression (Fig. S6D-E).

We assessed the impact of different EV treatments on microglial polarization using flow cytometry and IF staining. Compared to the PBS, EVs, and ZH-1c-EVs groups, the other groups exhibited significantly reduced M1 polarization markers and increased M2 polarization markers in microglia, with the ZH-1c-EVs@SIN group showing the most pronounced effect on microglial polarization (Fig. 6A-D). This result suggests that ZH-1c-EVs@SIN effectively promotes the M2 polarization of microglia.

Fig. 6.

Regulation of Microglial Polarization by ZH-1c-EVs@SIN. (A) IF staining showing M1 (CD86) and M2 (CD206) microglial markers across different groups, scale bar = 25 μm; (B) Quantitative analysis of IF results; (C) Flow cytometry analysis of CD86-positive and CD206-positive cells in each group; (D) Statistical quantification of flow cytometry results; (E) ELISA analysis of inflammatory cytokine levels in the supernatants from co-cultured cells in each group; (F-G) TUNEL staining showing neuronal apoptosis (G) and statistical analysis of apoptotic neurons (F), scale bar = 50 μm; (H-I) Flow cytometry analysis of neuronal apoptosis (H) and its statistical quantification (I) in co-cultured cells. Data are presented as mean ± S.D., and comparisons among multiple groups were performed using one-way ANOVA.* indicates p < 0.05 between groups, ** indicates p < 0.01. All cell experiments were repeated three times

Next, we explored the effect of these treatments on the repair of DRG neuron damage. Compared to the PBS, EVs, and ZH-1c-EVs groups, the other treatment groups demonstrated diminished levels of pro-inflammatory cytokines and neuronal apoptosis, alongside increased levels of anti-inflammatory cytokines. Again, the ZH-1c-EVs@SIN group showed the strongest neuroprotective effects, promoting neuronal survival and repair most effectively (Fig. 6E-I).

These results indicate that ZH-1c-EVs@SIN has no apparent cytotoxic effects on microglia, effectively promotes M2 polarization while inhibiting M1 polarization, and facilitates the repair of neuronal damage.

ZH-1c-EVs@SIN promotes DRG neuron damage repair in DPN

Previous in vitro experiments have demonstrated that ZH-1c-EVs@SIN promotes the M2 polarization of microglia, inhibits M1 polarization, and enhances neuronal damage repair. To further validate the therapeutic potential of ZH-1c-EVs@SIN in vivo, we administered ZH-1c-EVs@SIN via tail vein injection in mice. In vivo imaging was used to assess the accumulation time of different EV formulations in the spinal cord. The ZH-1c-EVs and ZH-1c-EVs@SIN groups exhibited the longest and strongest fluorescence signals in the spinal cord region, suggesting that the targeting ability of ZH-1c extends the retention time of these EVs in the spinal cord (Fig. 7A-B).

Fig. 7.

Effects of ZH-1c-EVs@SIN on DRG Neuron Damage Repair. (A-B) Fluorescence imaging showing the distribution of fluorescence in major organs and spinal cord tissues of mice in each group; (C-D) IF staining for M1 microglial markers (C) and M2 microglial markers (D), scale bar = 50 μm (original size), scale bar = 10 μm (magnified); (E-F) Quantitative analysis of IF staining results; (G-H) IF staining showing the longest neurite length of primary DRG neurons (G) and quantitative analysis (H), β-III Tubulin is marked in red, 3 images collected per mouse, N = 6 mice per group; (I-J) IF staining showing iNOS marker expression in DRG neurons (I) and its statistical analysis (J), scale bar = 25 μm. Data are presented as mean ± S.D., and comparisons among multiple groups were performed using one-way ANOVA.* indicates p < 0.05, ** indicates p < 0.01 between groups. N = 6 mice per group

We further assessed the polarization levels of microglia in the spinal cord tissues of each group of mice using fluorescent staining. Consistent with the in vitro results, the ZH-1c-EVs@SIN group showed the most significant M2 polarization and the strongest inhibition of M1 polarization (Fig. 7C-F). Primary cultures of DRG neurons from each group revealed that the ZH-1c-EVs@SIN group exhibited significantly longer neurite outgrowth compared to other groups, along with the most pronounced alleviation of neuroinflammation (Fig. 7G-J).

In summary, our study preliminarily demonstrates that ZH-1c-EVs@SIN promotes M2 polarization of microglia, alleviates neuroinflammation in DRG neurons associated with DPN, and enhances the repair of neuronal damage.

ZH-1c-EVs@SIN improves Immune-Inflammatory homeostasis for the treatment of DPN

Finally, to investigate the therapeutic effects of ZH-1c-EVs@SIN on DPN in vivo, we first measured the protein expression levels of WNT5a and TRPV1 in spinal cord tissues. The results were consistent with the trends observed in vitro (Fig. S7). Additionally, the body weight, insulin, and blood glucose levels in the mice were evaluated. Compared to the Blank, EVs, and ZH-1c-EVs groups, the other groups showed significant increases in insulin levels, along with reductions in body weight and blood glucose. The ZH-1c-EVs@SIN group exhibited the most pronounced therapeutic effect (Fig. 8A-C). Moreover, the PWT and PWL values were significantly higher in the ZH-1c-EVs@SIN group compared to the Blank, EVs, and ZH-1c-EVs groups, indicating the greatest improvement in neuropathic pain symptoms (Fig. 8D-E).

Fig. 8.

Therapeutic Effects of ZH-1c-EVs@SIN on DPN Mice. (A) Schematic representation of nanomaterial treatment for DPN mice; (B) Body weight measurements across groups of mice; (C) Serum insulin levels in each group; (D) Blood glucose levels in each group; (E) Mechanical allodynia test results; (F) Heat hyperalgesia test results; (G) Gait analysis results; (H-I) Immunoreactive PGP9.5 intraepidermal nerve fibers in the hind paw skin of each group (H) and corresponding quantitative analysis (I), with the dermal-epidermal junction outlined in white, scale bar = 25 μm. Data are presented as mean ± S.D., and comparisons among multiple groups were performed using one-way ANOVA, while data from animals at different time points were analyzed using two-way ANOVA. * indicates p < 0.05 between groups, ** indicates p < 0.01. N = 6 mice per group

The evaluation of motor function across the groups showed that, compared to the Blank, EVs, and ZH-1c-EVs groups, the paw contact area was significantly larger in the other groups, with the ZH-1c-EVs@SIN group exhibiting the largest paw contact area during standing (Fig. 8F). Furthermore, the quantitative analysis of PGP9.5 (an immunoreactive marker for intraepidermal nerve fibers) in footpad skin revealed a substantial increase in nerve fiber density in the ZH-1c-EVs@SIN group, indicating the most significant improvement (Fig. 8G-I).

Following the treatment, major organs were harvested from all groups and subjected to hematoxylin and eosin (H&E) staining. No notable pathological changes were observed in any of the organs, indicating that ZH-1c-EVs@SIN has good biocompatibility (Fig. S8). Overall, these findings demonstrate that ZH-1c-EVs@SIN can effectively improve immune-inflammatory homeostasis and promote the treatment of DPN in mice.

Previous studies have demonstrated that SIN exerts potent anti-inflammatory effects and offers neuroprotection by alleviating inflammation in the central nervous system [10, 41]. In the context of DPN, where chronic neuroinflammation is a key pathological feature, the modulation of anti-inflammatory pathways is particularly critical. This study not only confirms the neuroregenerative potential of SIN in DPN, but also uncovers its role in promoting M2 polarization and inhibiting M1 polarization of microglia through regulation of the WNT5a/TRPV1 signaling pathway. Unlike prior research that primarily emphasized SIN’s general anti-inflammatory properties, our findings are the first to delineate the specific molecular mechanisms by which SIN modulates microglial polarization in DPN. This provides novel mechanistic insight into SIN’s multifaceted therapeutic actions and expands its potential application as a targeted neuroimmune modulator in neuroinflammatory diseases.

The role of the WNT5a/TRPV1 signaling pathway in neuroinflammation and neural repair has been preliminarily validated in several neurological diseases, but its study in DPN remains relatively limited [21, 24, 25]. Previous research has primarily probed into the potential of TRPV1 in pain perception, while this study is the first to link the WNT5a and TRPV1 signaling pathways to the mechanisms of SIN treatment in DPN. Our findings demonstrate the crucial role of the WNT5a/TRPV1 pathway in modulating microglial polarization and that SIN leverages this pathway to improve immune-inflammatory homeostasis in DPN. Compared to previous studies, this research takes a significant step forward in exploring the underlying mechanisms, revealing that the WNT5a/TRPV1 signaling pathway is not only involved in pain transmission but also holds therapeutic potential in inflammation regulation and neural repair.

Compared to traditional drug delivery methods, EVs offer natural biocompatibility, targeting capability, and low immunogenicity, allowing them to effectively cross biological barriers [26–28]. In this study, we developed ZH-1c-EVs@SIN for precise delivery to microglial cells. NTA and TEM confirmed the stability of NSC-EVs in terms of morphology and size, and in vitro experiments verified their efficient delivery of SIN. Compared to traditional drug delivery systems used in previous studies, the aptamer-modified NSC-EVs demonstrated enhanced targeting capabilities and reduced toxicity, further supporting their potential application in the treatment of DPN and other neurological diseases.

Microglial cells serve as key immune regulators in the central nervous system, with their polarization states playing a critical role in the inflammatory response and repair processes [42]. Previous studies have shown that M1 microglia are primarily involved in pro-inflammatory responses, while M2 microglia contribute to anti-inflammatory effects and tissue repair [43, 44]. In the pathological progression of DPN, M1 polarization exacerbates nerve damage, whereas M2 polarization aids in nerve repair. In this study, the ZH-1c-EVs@SIN delivery system was found to significantly suppress M1 polarization and promote M2 polarization, thereby improving immune-inflammatory homeostasis. These results align with previous studies on microglial polarization regulation, but the use of ZH-1c-EVs@SIN enhanced the efficiency of targeted regulation, advancing our understanding of DPN pathophysiology.

In this study, the regulatory effects of ZH-1c-EVs@SIN on microglial polarization and neuronal repair were validated through in vitro experiments, followed by in vivo experiments to further investigate its efficacy in a diabetic mouse model of DPN. Previous studies have shown that the plant extract daidzein can significantly alleviate diabetic neuropathy by inhibiting oxidative stress through suppression of NOX-4 [45]. Similarly, SIN, an isoquinoline alkaloid extracted from the traditional Chinese herb Sinomenium acutum, has demonstrated beneficial effects in treating diabetic neuropathy [46]. However, our experimental results clearly demonstrate that combining SIN with ZH-1c-modified NSC-derived extracellular vesicles (ZH-1c-EVs) can greatly enhance its therapeutic efficacy for DPN. Specifically, we quantitatively compared ZH-1c-EVs@SIN with conventional SIN administration and existing DPN treatments to evaluate its relative advantages. Compared with traditional intraperitoneal SIN injection, ZH-1c-EVs@SIN demonstrated a spinal cord accumulation rate of 7.2% ± 1.1% (estimated to be < 2% with conventional delivery), reduced TNF-α levels by 60–70% (versus 30–40% reduction with conventional delivery), increased DRG neurite outgrowth by 50–60% (versus 20–30% with conventional delivery), and prolonged the circulation time of SIN by 2–3 times due to EV encapsulation (Fig. 7A). Compared with the neuropathic pain drug pregabalin, this system exerts both anti-inflammatory and neuroregenerative effects by modulating the WNT5a/TRPV1 pathway (Fig. 8G–I). Additionally, compared with mesenchymal stem cell-derived EVs, NSC-EVs carry higher levels of neurotrophic factors such as BDNF, and aptamer modification further enhanced targeting efficiency by 1.4–2.4 times (Fig. 5H). Together, these results demonstrate that ZH-1c-EVs@SIN significantly outperforms conventional therapeutic strategies in targeting efficiency, sustained efficacy, and safety, providing a promising new approach for DPN treatment. Compared to previous studies using SIN or other drugs alone for DPN treatment, this study demonstrated more significant neurorepair effects attributed to the precise delivery of SIN and the efficiency of NSC-EVs. The high consistency between the in vivo and in vitro results further indicates that ZH-1c-EVs@SIN has promising therapeutic potential, achieving consistent neurorepair effects in both environments.

Traditional treatment strategies for DPN primarily focus on symptom management, particularly alleviating pain through the use of antineuropathic drugs [47, 48]. However, these approaches often fail to effectively address nerve damage or halt disease progression. While previous studies have explored the use of anti-inflammatory or immunomodulatory agents for DPN treatment, their effects have been limited, and systemic side effects have been substantial. In this study, the use of ZH-1c-EVs@SIN enabled precise regulation of microglial cells, significantly improving immune-inflammatory homeostasis and promoting neural function recovery in DPN mice. Compared to traditional treatments, this study presents a novel therapeutic strategy based on molecular mechanisms, offering higher efficacy and fewer side effects.

The scientific and clinical value of this study lies in its first revelation that SIN can precisely regulate microglial polarization by modulating the WNT5a/TRPV1 signaling pathway, thereby improving immune-inflammatory homeostasis and promoting neural repair in DPN. By utilizing ZH-1c aptamer-modified NSC-EVs as drug delivery vehicles, this study demonstrates the significant advantages of EVs in targeted drug delivery, offering a novel strategy for treating DPN and other neurological diseases. Clinically, this approach holds promise for enhancing therapeutic efficacy, reducing side effects, and advancing precision medicine in neural repair, with broad application potential.

However, despite the promising prospects in both basic and clinical applications, some limitations remain. First, the study is primarily based on animal models, lacking large-scale clinical trial data, so the safety, pharmacokinetics, and long-term effects of ZH-1c-EVs@SIN in humans still need further validation. Additionally, while aptamer-modified EVs have demonstrated excellent delivery performance, their effectiveness may vary across individuals and disease conditions, necessitating further investigation in a broader range of neurological disorders and patient populations. Moreover, large-scale production and application of aptamer-modified NSC-EVs may face technical and cost-related challenges, which require further optimization. Despite the promising prospects for both basic research and clinical applications, this study still has several limitations. First, the research was mainly conducted in animal models, lacking large-scale clinical trial data; thus, the safety, pharmacokinetics, and long-term effects of SIN-loaded NSC-EVs in humans still need further investigation. Second, although aptamer-modified EVs showed excellent delivery performance, their delivery efficiency may vary among individuals and under different disease conditions, which warrants further exploration in a broader range of neurological diseases and patient populations. Additionally, scaling up the production and standardization of aptamer-modified NSC-EVs remains technically and financially challenging, with two main obstacles: (1) large-scale production and standardization of NSC-EVs; and (2) potential degradation or structural instability of the aptamer in vivo, which may affect targeting efficiency. These aspects have been discussed in the manuscript. Moreover, the SIN encapsulation efficiency in EVs was 18%, which may be limited by the mismatch between EV membrane lipids and the polarity of SIN. Future work could improve drug loading through ultrasound-assisted loading or pH-gradient methods (potentially increasing to 30–40%) or by modifying EV membrane proteins to enhance drug affinity. It is noteworthy that despite the limited loading efficiency, ZH-1c-EVs@SIN can achieve effective DPN treatment through aptamer-mediated targeting (spinal cord accumulation rate of 7.2%), compensating for dosage limitations (Fig. 8). Future studies may further optimize the delivery system’s drug-loading capacity through chemical modification of the drug and engineering of EVs. Additionally, although aptamer-modified EVs (such as ZH-1c-EVs) can enhance targeting specificity, potential off-target effects and immunogenicity remain concerns. Regarding off-target effects, the ZH-1c aptamer may bind nonspecifically to homologous CD64 proteins on neutrophils or monocytes, leading to EV accumulation in organs such as the liver and spleen (as shown by non-spinal fluorescence signals in Fig. 7A). In terms of immunogenicity, exogenous aptamers may be recognized by TLR3/7/8, inducing the release of cytokines like IFN-α. Although H&E staining showed no significant inflammation in major organs, repeated or long-term administration may trigger adaptive immune responses. Future work could reduce immunogenicity through chemical modifications of the aptamer (e.g., 2′-O-methylation) or design microenvironment-responsive aptamers to minimize off-target binding, thereby improving the delivery system’s safety. Furthermore, although H&E staining indicated no obvious inflammation in major organs, the long-term safety and potential systemic toxicity of repeated administration of ZH-1c-EVs@SIN have yet to be fully evaluated. It remains unclear whether multiple doses of this delivery system might induce chronic immune responses or organ dysfunction, such as aptamer-induced antibody production that could affect efficacy, or SIN accumulation leading to drug toxicity. It is therefore recommended to conduct acute and chronic toxicity studies, including comprehensive histopathological assessments, to strengthen the evidence supporting the safety of this novel therapeutic strategy. Moreover, although this study confirmed the targeted therapeutic effects of ZH-1c-EVs@SIN, comprehensive pharmacokinetic/pharmacodynamic (PK/PD) data were not provided. Current IVIS imaging demonstrated spinal cord retention for up to 24 h (Fig. 7A), which is significantly longer than the half-life of free SIN (~ 2 h), but no plasma concentration–time curve was generated. Future studies should quantify SIN levels in various tissues at different time points using HPLC to calculate clearance rates and tissue distribution volumes, and integrate this with neurological function scores to establish a PK/PD model, thereby clarifying the in vivo dynamics and therapeutic correlations. Finally, in the field of DPN research, numerous studies have highlighted the complexity and heterogeneity of its pathology. Pathologically, DPN involves demyelination, axonal degeneration, and persistent low-grade inflammation. In terms of lesion distribution, DPN is known to have a focal nature rather than uniform nerve tissue damage [5]. For example, nerve biopsy studies of DPN patients have revealed significant regional variations in myelin and axonal densities. Moreover, microglia are known to be patchily activated in the central nervous system upon injury, a pattern that is critical in DPN [49]. In our study, histological assessments were conducted by randomly selecting five fields per sample, which has limitations when facing the complex tissue heterogeneity of DPN. While selecting a certain number of fields provides preliminary information to help researchers understand lesion characteristics during early exploratory studies, more comprehensive and systematic evaluation methods are needed in future studies to fully elucidate the histological features of DPN.

Future research directions can be explored in several areas. First, clinical trials are a critical step in validating the efficacy and safety of ZH-1c-EVs@SIN in DPN patients, as well as assessing its applicability to other types of neurological diseases. Second, with advancements in technology, further optimization of aptamer modification and EV production processes could enhance the stability and efficiency of drug delivery. Additionally, future studies may investigate the potential of combining ZH-1c-EVs@SIN with other neurorepair drugs or gene therapies, exploring the benefits of multimodal combination therapies to provide more personalized and effective treatment options for patients with neurological diseases. Ultimately, this research has the potential to drive the development of precision medicine in the treatment of neurological disorders, facilitating a comprehensive breakthrough from basic research to clinical translation.

Conclusion

In summary, ZH-1c-EVs@SIN significantly improved immune-inflammatory homeostasis in DPN by inhibiting the WNT5a/TRPV1 signaling pathway, thereby promoting neural function recovery in diabetic mice (Graphical Abstract). This study provides a novel therapeutic modality for DPN by using NSC-derived EVs to precisely deliver drugs and activate protective signaling pathways. This method improves therapeutic precision and retention at the site of action while substantially curbing unintended systemic exposure. By utilizing aptamer technology to improve the specificity of EVs, this study offers new perspectives and methods for the clinical treatment of DPN. Furthermore, the findings deepen our understanding of the interaction between inflammation and neural damage in DPN and may inspire the development of therapeutic strategies for other inflammation-related neurodegenerative diseases.

Materials and methods

Ethical statement

Male C57BL/6J mice (SPF grade, 4–6 weeks old, weighing 16–22 g) were procured from Hunan SJA Laboratory Animal Co., Ltd. (Hunan, China). The mice were housed in SPF-grade laboratory conditions with controlled humidity (60%−65%) and temperature (25 ± 2 °C) under a 12-hour light/dark cycle with free access to food and water. After a one-week acclimation period, the experiment began, during which the health status of the mice was observed prior to any procedures. All animal experiments were granted by the Scientific Investigation Committee of Hunan University of Medicine General Hospital and conducted in strict accordance with the ethical guidelines set forth by the International Association for the Study of Pain (IASP) Ethics Committee (No. 202403055).

Construction of mouse DPN model

Mice were fasted for at least 2 h with access to water before model induction. After 4 weeks on a high-fat, high-sugar diet, the mice were intraperitoneally injected with 100 mg/kg streptozotocin (STZ, Selleckchem, Cat#S1312) and continued on the same diet for an additional 6 weeks. Before injection, STZ was quickly dissolved in sodium citrate buffer (50 mM, pH 4.5) to a final concentration of 20 mg/mL, and the injection was completed within 5 min. The model was considered successful when fasting blood glucose levels (measured weekly) exceeded 11.1 mmol/L on two consecutive occasions, accompanied by a decrease in serum insulin. Motor function and nociceptive sensitivity were measured weekly for 6 weeks following STZ administration. One week after the final treatment, the mice were euthanized, and the gastrocnemius muscle, ipsilateral sciatic nerve, L4-L5 DRGs, and spinal cord were immediately collected for further biochemical and histological evaluations [35].

Based on prior studies, the SIN dose was set at 20 mg/kg and administered via intraperitoneal injection [50]. The mice were divided into: (1) Control group: non-modeled control mice treated with 20 mg/kg saline solution; (2) DPN group: modeled DPN mice treated with 20 mg/kg saline solution; (3) SIN group: modeled DPN mice treated with 20 mg/kg SIN solution; (4) Blank group: PBS control group; (5) EVs group: mice treated with NSC-EVs; (6) EVs@SIN group: mice treated with EVs loaded with SIN (EVs@SIN); (7) ZH-1c-EVs group: mice treated with ZH-1c aptamer-modified NSC-EVs; (8) ZH-1c-EVs@SIN group: mice treated with ZH-1c-EVs@SIN; (9) SIN + overexpression negative control (oe-NC) group: modeled DPN mice treated with 20 mg/kg SIN solution plus control lentivirus injection; (10) SIN + oe-TRPV1 group: modeled DPN mice treated with 20 mg/kg SIN solution plus TRPV1 overexpression lentivirus injection. SIN (115-53-7) was purchased from Hunan Zhengqing Pharmaceutical Group Co., Ltd. (China) and administered intraperitoneally to the mice post-surgery [32]; the drug administration protocol consisted of two injections per week for six weeks. After successful model establishment, 100 µL of DIR-labeled (40757ES25, YEASEN, China) 100 µg EVs (50 µg/100 µL PBS) were injected into the mice via the tail vein [51, 52]; mice in the lentiviral transfection groups received intrathecal injections of 1 × 10⁸ TU/mL lentivirus once every two days for a total of two weeks after successful model establishment [35].

Mechanical allodynia test

Mice were placed in brown acrylic boxes on a metal mesh platform to acclimate to the environment for 30 min. Mechanical PWTs were then tested utilizing an electronic Von Frey test (IITC Life Science, USA). Incremental forces were set to the plantar surface of the hind paw using appropriate EVs, and the force that triggered a paw withdrawal response was recorded via the pressure sensor in the probe. Each test was performed in triplicate with a 5-minute interval between stimulations, and the average of the three trials was recorded as the PWT. Mechanical allodynia was further evaluated using Von Frey filament testing. Mice were placed individually in acrylic chambers positioned on an elevated mesh platform. A high-intensity 0.4 g Von Frey filament was applied perpendicularly to the plantar surface of the hind paw for approximately 1 s, and the procedure was repeated 10 times per session. Paw withdrawal responses were recorded, and the percentage response frequency was calculated using the formula: (number of paw withdrawals/10 trials) × 100%, which was used to quantify the mechanical sensitivity.

Thermal sensitivity test

Mice were placed in transparent plastic boxes on a glass surface and allowed to acclimate to the environment for 30 min. PWL to thermal stimuli was assessed using the Hargreaves radiant heat apparatus (IITC Life Science, USA), which applied heat to the hind paw until a withdrawal response was induced. The response was characterized by paw lifting, avoidance, or flinching. A cutoff time of 25 s was set to prevent tissue damage. The measurement was repeated three times with a 5-minute interval between each test, and the average PWL was recorded. Thermal hyperalgesia thresholds were measured three times for each mouse, and the average was used for analysis.

Gait imaging in mice

Mouse gait was recorded and analyzed using the DigiGait™ imaging system (Mouse Specifics, Inc., USA). During the training phase, mice were acclimated to the transparent treadmill for 5 min, once per day, for a total of three sessions. On the day of the experiment, the mice were placed on the transparent treadmill, and their running patterns were recorded using the imaging system. The DigiGait Analysis software was selected for statistical analysis of the mice’s gait during a 2-second period of steady running.

Histological staining

H&E staining: Tissue specimens were initially fixed and processed for paraffin embedding, after which serial sections were prepared. For staining, sections underwent dewaxing in xylene, followed by graded ethanol dehydration (100%, 95%, and 70%) or direct rinsing in distilled water, depending on the subsequent procedure. Hematoxylin staining (H8070, Solarbio, Beijing, China) was applied at room temperature for 5–10 min to visualize nuclear structures. The slides were then rinsed thoroughly with distilled water and briefly immersed in 95% ethanol to remove excess stain. Counterstaining was carried out using eosin solution (G1100, Solarbio, Beijing) for an additional 5–10 min to highlight cytoplasmic components. Following staining, sections were dehydrated through a graded ethanol series (85%, 90%, 95%, and 100%) and then treated with dichloromethane to render the tissue optically clear for microscopic examination.

Picric Acid Sirius Red (PASR) Staining (PASR; 50-300-77, Fisher Scientific, USA) followed a procedure similar to the previously described staining steps. Masson’s Trichrome Staining was performed according to the manufacturer’s instructions (G1340, Beijing) for the assessment of fibrosis. Finally, the stained sections were sealed with neutral resin mounting medium. Tissue morphology was examined using a light microscope at 200× magnification (Olympus, Tokyo, Japan). For each sample, five random visual fields were selected per section, and at least three sections per animal were analyzed to ensure reproducibility. The extent of collagen deposition in left ventricular tissue was quantified by measuring the area positively stained by Masson’s trichrome using Image-Pro Plus software (Version 6.0, Media Cybernetics). Additionally, collagen content was semi-quantitatively evaluated using the Ishak fibrosis scoring system.

Highthroughput sequencing and bioinformatic analysis of spinalcord microglia in a DPN mouse model

After constructing the DPN mouse model, three mice were randomly selected from each group (Control, DPN, and SIN groups). Lumbar-spinal-cord microglia were isolated, and total RNA was purified from the resulting nine samples with the Invitrogen Total RNA Isolation Kit (Cat. No. 12183555, USA). RNA quantity was measured on a BioSpectrometer basic (Eppendorf, USA), while integrity was verified by agarosegel electrophoresis. Only highquality RNA was advanced to cDNA library preparation and sequenced on the Illumina NextSeq 500 platform. Basecalling transformed the instrument’s raw images into FASTQ files. Adapter trimming and lowquality read removal were performed with Cutadapt, yielding highconfidence “clean” reads. These were aligned to the GRCm39 reference genome using HISAT2, and transcript abundances were calculated to generate a geneexpression matrix in R.

DEGs were identified with the limma package, applying |log₂FC| >1 and p < 0.01 as significance thresholds. Venn diagrams and volcano plots were created with the Xiantao Academic web tool (https://www.xiantaozi.com/), and heatmaps were produced via Hiplot (https://hiplot.com.cn/).

Machine learning prioritization of disease-relevant genes

To pinpoint a minimal, diseaseassociated gene signature, we applied LASSO regression using glmnet (version 4.0–2). The full expression matrix was randomly partitioned into training (0%) and testing (0%) subsets prior to model fitting.

Lentiviral vector construction and transfection

Potential short hairpin RNA (shRNA) target sequences for mouse cDNA were analyzed based on GeneBank. The lentiviruses for TRPV1 overexpression were constructed and packaged by Genechem (Shanghai, China), with the lentiviral overexpression vector being pLenti-GFP. The lentiviral vector used was a pLenti-GFP backbone (Addgene #17458), with the CMV promoter driving TRPV1 cDNA expression, and an IRES-GFP sequence for fluorescent labeling. Virus packaging was conducted using a three-plasmid system (pLenti-GFP-TRPV1, pMD2.G, psPAX2) and Lipofectamine 2000 for transfection of 293 T cells. The harvested viral particles were concentrated by ultracentrifugation to achieve a final titer of 1 × 10⁸ TU/mL. Vector construction was verified by Sanger sequencing, showing 100% identity with NM_001271623.1. Lentiviral infectivity was determined by quantifying GFP-positive cells via flow cytometry. Titer calculations were typically based on the infection rate and viral dosage, with an acceptable viral titer usually reaching or exceeding 1 × 10⁶ TU/mL (Transducing Units per milliliter).

For cell model establishment, microglial cultures in logarithmic growth phase were adjusted to 5 × 10⁴ cells/mL, seeded into 6-well plates (2 mL per well), and transduced at a multiplicity of infection (MOI) of 10 using the prepared lentivirus. After 48 h, ampicillin (Sigma-Aldrich, #171254) was used to select stably transduced cells over a 2-week period.

Experimental groups included: oe-NC group (non-TRPV1 overexpression group) and oe-TRPV1 group (TRPV1 overexpression group).

Primary microglia isolation and culture

Primary spinal cord microglia were obtained from neonatal or DPN-model mice (n = 3, randomly selected). The spinal dorsal horn regions were enzymatically digested with trypsin, then gently dissociated and resuspended in DMEM (Gibco, #11965092) supplemented with 10% fetal bovine serum (Gibco, #10099158) and 1% penicillin-streptomycin (Beyotime, #C0222). The resulting cell suspension was filtered and seeded into poly-L-lysine-coated flasks (Beyotime, #60716ES08). Mixed glial cultures were maintained for 12–14 days. To isolate microglia, the flasks were agitated at 200 g for 1 h. The supernatant, enriched with detached microglial cells, was collected and transferred to poly-D-lysine-coated T-flasks (Gibco, #132704). Primary microglial cells were identified by IF staining, confirming a purity of over 95% (Fig. S9).

Cells were incubated in DMEM containing 5.5 mM glucose, 10% FBS, 2% glutamine, and 1% penicillin/streptomycin under standard culture conditions (37 °C, 5% CO₂). To mimic diabetic conditions in vitro, a high-glucose environment (30 mM final concentration) was established by adding 24.5 mM glucose. Experimental groups were as follows: Con group: 5.5 mM glucose + PBS + microglial cells transfected with oe-NC lentivirus; high glucose (HG) group: 30 mM glucose + PBS + microglial cells transfected with oe-NC lentivirus; SIN + oe-NC group: 30 mM glucose + 200 µM SIN + microglial cells transfected with oe-NC lentivirus; SIN + oe-TRPV1 group: 30 mM glucose + 200 µM SIN + microglial cells transfected with oe-TRPV1 lentivirus.

Isolation of primary DRG neurons from mice

We prepared DRG neuron cultures from mice. First, L4-L6 DRG segments were collected and digested with 1 mL of enzyme solution (0.1% collagenase type I and 0.3% collagenase type II) at 37 °C for approximately 30 min. The isolated DRG neurons were seeded onto 24-well clear flat-bottom culture plates (model: 354619, Corning, USA) coated with 0.1 mg/mL poly-D-lysine and 5 mg/mL laminin, with 14 mm diameter coverslips (Biosharp) placed inside each well. The culture medium consisted of DMEM/F12 supplemented with 1× B-27 and 20 ng/mL nerve growth factor (containing 25 mM glucose, ATCC, 30-2006, USA), and the neurons were incubated at 37 °C with 5% CO₂. After 24 h, neurite growth was assessed using immunofluorescent staining against β-III Tubulin. Neuronal morphology was visualized under a Nikon confocal microscope (Fig. S10), and the longest neurite per neuron was quantified using ImagePro Plus 6.0 software.

Co-culture of microglia and DRG neurons

To establish the microglia-DRG neuron co-culture system, primary microglial cells were seeded into the upper inserts of Transwell plates (Corning, CLS3412, USA), and primary DRG neurons were placed in the lower wells at a 1:1 cell ratio. The co-culture was maintained in shared medium for 24 h before experimental treatment commenced. Media changes were performed every 3 days.

ELISA

Cytokine levels, including TNF-α (900-T54), IL-6 (900-M50), TGF-β (BMS808-4), and IL-10 (900-M53), were quantified utilizing ELISA kits (Invitrogen, USA). OD values were measured at 450 nm using a microplate reader.

TUNEL assay for neuronal apoptosis

Cells were fixed with 4% paraformaldehyde (Yeasen, #60536ES60) for 15 min, permeabilized with 0.25% Triton X-100 for 20 min, and blocked with 5% BSA (Yeasen, #36101ES25). Staining was carried out using a TUNEL kit (Beyotime, #C1086), followed by DAPI counterstaining (Beyotime, #C1002). Apoptotic (TUNEL-positive) cells exhibited green fluorescence, while DAPI marked all nuclei in blue. Imaging was performed using a Zeiss LSM 880 confocal microscope. Apoptosis rate (%) was calculated by averaging cell counts across five randomly selected microscopic fields using the formula: apoptosis rate = (number of apoptotic cells/total number of cells) × 100%.

Flow cytometric analysis of apoptosis

Neurons were harvested with 0.25% trypsin (EDTA-free), washed thrice with cold PBS, and resuspended in Annexin-V-FITC/PI staining buffer prepared as per kit protocol (Shanghai YUBO, #K201-100). The staining mixture included Annexin-V-FITC, PI, and HEPES buffer in a 1:2:50 ratio. For every 100 µL of staining solution, 1 × 10⁶ cells were used. After 15 min of incubation at room temperature, an additional 1 mL of HEPES buffer was added before analysis. Apoptotic cells were detected using a flow cytometer with excitation at 488 nm and dual-band detection at 525 nm (FITC) and 620 nm (PI). Each experiment was conducted in triplicate.

CCK-8 assay

Neuronal viability was assessed using the CCK-8 (Beyotime, #C0037). Cells in the logarithmic growth phase were diluted to 5 × 10⁴ cells/mL and seeded at 100 µL per well in 96-well plates. Following 48 h of incubation, the culture medium was removed, and 10 µL of CCK-8 reagent was added to each well. Following a 2-hour incubation at 37 °C, absorbance at 450 nm was measured using a Thermo Fisher Multiskan FC microplate reader (Cat. No. 51119080). Each group was tested in triplicate, and mean absorbance values were calculated to assess viability.

Derivation and culture of NSCs

Human induced pluripotent stem cells (hiPSCs, IMR90-4; Wisconsin International Stem Cell Bank) were cultured on Matrigel-coated six-well plates at an initial seeding density of 2-2.5 × 10⁴ cells/cm² in TeSR™-E8™ medium (STEMCELL Technologies, #05990). After 24 h, the medium was replaced with a neural induction formulation containing a defined neural matrix and neural induction supplement (Gibco, #A1647801). This induction medium was refreshed every other day over a 7-day period.

At day 7, neuroepithelial aggregates were dissociated with Accutase (Gibco, #A1110501) and replated onto Matrigel-coated dishes at 0.5-1.0 × 10⁵ cells/cm² in a medium comprising 50% DMEM/F12, 50% neural matrix, and 1× neural supplement. By day 5, the cultures reached confluence. NSCs were routinely passaged every 7 days and cryopreserved for long-term storage in liquid nitrogen. Identity and purity of NSCs were confirmed through immunofluorescent staining for stem cell markers Nestin, Pax6, and Sox-2 (Fig. S11).

Isolation of NSC-EVs

Cryopreserved NSCs (passage 11) were thawed at 37 °C and expanded in T-75 flasks (Corning) under standard conditions (% CO₂, 37 °C). At ~ 7% confluence, cells were detached with dispase (1 U mL⁻¹, Gibco #17105041), rinsed with NSC medium (Gibco #A1050901), and replated at ~ 500 cells cm⁻² in expansion medium on 150 mm × 20 mm dishes. When cultures reached 9% confluence, conditioned medium was harvested for EV isolation or snapfrozen (−80 °C).

Conditioned medium was clarified (1 811 g, 10 min) and passed through a 0.22 μm filter. The filtrate was concentrated 5-7fold using a 10 kDa Amicon ultrafiltration unit (Millipore). A QSepharose FF column (1.5 × 12 cm; GE Healthcare) preequilibrated in buffer (50 mM Tris, pH 8.0) received the concentrate; bound EVs were eluted with 50 mM Tris, 1 000 mM NaCl, pH 8.0 (1 mL min⁻¹). Fractions were screened by nanoparticletracking analysis (NTA; NanoSight LM10) and further sizefractionated via SEC (mobile phase: 50 mM phosphate, 200 mM NaCl, pH 7.0; 1 mL min⁻¹). EVrich, proteinpoor fractions were pooled, reconcentrated, and stored at −20 °C.

Preparation of ZH-1c-EVs@SIN

For drug loading, SIN was dissolved in ethanol and mixed with NSCEVs at a 1 : 9 (w/w) SIN : EV ratio, followed by 4 h incubation at ~ 22 °C. The suspension was cleared (10 000 ×g, 10 min) then ultracentrifuged (100 000 ×g, 2 h) to pellet SINloaded vesicles (EVs@SIN), which were resuspended in PBS and frozen (−80 °C). Particle sizing before and after loading confirmed minimal diameter shift.

Surface targeting was achieved by carbodiimide chemistry. A 5′COOH/FAMlabeled ZH-1c aptamer (Sangon) was activated with EDC (46 mg, 0.3 mmol; MedChemExpress HYD0178) and NHS (35 mg, 0.3 mmol; HYY0623) for 1 h. Activated aptamer (50 µL, 10 µM) was reacted with 1 mL EVs@SIN (80 µg µL⁻¹) or naïve EVs to yield ZH-1cEVs@SIN or ZH-1cEVs, respectively. Complete aptamer binding was verifid by 2.5% agarose gel shift.

The EVs were grouped as follows: (1) EVs group (NSC-EVs); (2) EVs@SIN group (NSC-EVs successfully loaded with SIN); (3) ZH-1c-EVs group (ZH-1c-modified NSC-EVs); (4) ZH-1c-EVs@SIN group (NSC-EVs successfully loaded with SIN and modified with the ZH-1c aptamer).

Characterization of EVs

Protein Markers (Western Blot). EV pellets were lysed in RIPA (Thermo #89901). Proteins were probed for Alix (ab275377, 1:1 000), TSG101 (ab133586, 1:1 000), and CD81 (ab286173, 1 : 500). Calnexin (ab92573, 1:20 000) served as a negative control. All purchased from Abcam (UK).

Particle Size & Concentration. Diluted EVs (1:500 in MilliQ) were analysed by NTA (NanoSight LM10, 640 nm laser; gain 6.0, threshold 11). Output values were corrected for dilution to obtain the original sample concentration.

Morphology (TEM). EVs were repelleted, fixed (% paraformaldehyde + 2. % glutaraldehyde, 4 °C, 1 h), washed, postfixed in1% osmium tetroxide (1.5 h), dehydrated through graded ethanol, infiltrated with epoxy, and polymerised (35 °C → 45 °C → 60 °C, 24 h each). Ultrathin sections were stained with uranyl acetate/lead citrate and imaged on a JEOL JEM-1011 (80 kV) with a MegaView III camera (Soft Imaging System, Münster, Germany). All measurements were performed in triplicate.

Particle size, polydispersity, and zeta potential measurement

The size, polydispersity index (PDI), and zeta potential of the EVs were determined using a Zetasizer Nano ZS (Malvern Instruments, UK). NTA was also used to characterize size distribution and concentration, based on the principle that the diffusion coefficient is inversely proportional to particle size. DLS combined with laser Doppler velocimetry was employed to measure zeta potential at 25 °C, after appropriate dilution in water. All experiments were performed in triplicate [53].

HPLC analysis

The SIN content loaded into the prepared EVs@SIN was quantified using a slightly modified reverse-phase HPLC-UV method (Alomrani et al., 2014). The HPLC system (Waters™ 600 controller, USA) was equipped with a dual wavelength UV detector (Waters™ 2487) and an autosampler (Waters™ 717 Plus). SIN was analyzed using a mobile phase composed of deionized water and acetonitrile (30:70 v/v) flowing through a reversed-phase C18 column (µBondapak, 4.6 × 150 mm, 10 μm, Waters) at a flow rate of 1.2 mL/min. Each sample injection volume was 30 µL and UV detection was performed at 261 nm at room temperature [53]. The actual drug loading was calculated as follows:

Unit 1: SIN loading amount per mg of EVs protein (μg/mg protein):    =      (/) EVs protrein concentration (mg/mL).

Unit 2: SIN molar concentration in EVs solution (µM): Concentration (µM) = SIN amount (µg/mL) Molecular weight of SIN (327.44 g/mol) × 100%.

Determination of SIN loading efficiency

To determine the loading efficiency, the SIN content in the samples was analyzed via HPLC at a wavelength of 210 nm. The protein content of the samples was determined using a BCA protein assay kit (Thermo Scientific, Waltham, MA, USA). The drug loading efficiency was then calculated as follows: Relative encapsulation efficiency = [Amount of SIN in EVs]/[Amount of SIN added] × 100%.

Fluorescence imaging of microglial EV uptake

NSCEVs (40 µg) were labelled with the nearinfrared dye DiR (25 µM final) for 30 min at room temperature. Free dye was removed by ultracentrifugation. DiREVs were applied to microglia seeded in 24well plates and incubated under standard conditions. After extensive PBS washes, cells were fixed n 4% paraformaldehyde (BHCTER, AR1068) for 30 min and counterstained with DAPI (Beyotime, C1005). Fluorescence images (×400) were captured on an Olympus BX53 microscope and analysed with ImageJ 6.0.

Flow cytometry

Coculture microglia were harvested (1 200 g, 5 min, 4 °C), dissociated with StemPro™ Accutase™ (Gibco, A1110501), and washed twice in PBS. Singlecell suspensions were resuspended in staining buffer (100 µL) and incubated with antibodies listed in Table S2 for 30 min at 4 °C in the dark. Cells were washed thrice, resuspended, and analysed on a Beckman flow cytometer; IgG isotype controls defined background fluorescence.

RT-qPCR

Total RNA was isolated with TRIzol (Invitrogen, 15596026) and quantified via NanoDrop LITE (Thermo). cDNA synthesis employed the PrimeScript RT kit with gDNA Eraser (TaKaRa, RR047Q). qPCR was performed using SYBR™ Green Master Mix (Applied Biosystems, 4364344) on an ABI 7500 system (Applied Biosystems).

The primers for each gene were synthesized by TaKaRa (Table S3), and GAPDH was used as the internal reference gene. The relative expression levels of each gene were analyzed utilizing the 2^−ΔΔCt method. All RT-qPCR experiments were performed in triplicate.

Western blot

Cells or tissues were lysed in enhanced RIPA buffer (Beyotime, P0013B) supplemented with protease inhibitors. Protein concentrations were determined (BCA kit, Beyotime, P0012). Equal amounts of protein were separated on 10% SDSPAGE gels and transferred to PVDF membranes (Beyotime, FFP39). After blocking with % BSA (ST023) for 2 h, membranes were probed with primary antibodies (rabbit, Table S2) for 1 h at room temperature. HRPconjugated secondary antibody (goat anti-rabbit, Abcam, ab6721, 1:2 000) was applied for 1 h. Signal was developed with Pierce™ ECL substrate (Thermo, 32209) and imaged on a BioRad system. Band intensity was quantified with ImageJ; βactin served as the loading control. Each experiment was repeated three times.

IF staining

Cells grown on glass coverslips (12well plates) or tissue sections were fixed in % PFA (Thermo, I28800) for 1 h. After DPBS washes containing 0.0% Tween-20, samples were permeabilised in 01% Triton X-100 (Invitrogen, HFH10) for 3 min and blocked with DPB/5% goat serum plus 0.3 M glycine for 1 h. Primary antibodies (Table S2) were incubated overnight at 4 °C (single or sequentially for double labelling). After washes, species appropriate secondary antibodies were applied for 1 h at room temperature. Nuclei were counterstained with DAPI (Beyotime, C1002) for 5 min. Coverslips were mounted in antifade medium and imaged on an Olympus FV-1000/ES confocal microscope. Fluorescence coverage across six random 40× fields was quantified with ImageJ, and mean values calculated per group.

Statistical analysis

Statistical analyses were completed utilizing R version 4.2.1, compiled in the RStudio integrated development environment (version 2022.12.0-353). Perl (version 5.30.0) was used for file processing, and data analysis was performed with GraphPad Prism 8.0. Quantitative data are detailed as mean ± standard deviation (SD). For comparisons between two groups, an independent sample t-test was adopted, while comparisons among multiple groups were implemented using one-way analysis of variance (ANOVA). For comparisons of multiple groups across different time points, two-way ANOVA was selected. Post-hoc tests were conducted utilizing Bonferroni correction, with significance established at p < 0.05.

Abbreviations

DLS

Dynamic Light Scattering

DEGs

Differentially Expressed Genes

DPN

Diabetic Peripheral Neuropathy

DRGs

Dorsal Root Ganglia

EVs

Extracellular Vesicles

H&E

Hematoxylin and Eosin

HG

High Glucose

HPLC

High-Performance Liquid Chromatography

IF

Immunofluorescence

NTA

Nanoparticle Tracking Analysis

NSC

Neural Stem Cell

NSC-EVs

Neural Stem Cell-Derived Extracellular Vesicles

PASR

Picric Acid Sirius Red

PWL

Paw Withdrawal Latency

PWT

Paw Withdrawal Thresholds

shRNA

Short Hairpin RNA

SIN

Sinomenine

TEM

Transmission Electron Microscopy

ZH-1c-EVs@SIN

ZH-1c aptamer-modified neural stem cell-derived extracellular vesicles loaded with sinomenine

Author contributions

JC and LZ contributed equally to this study. JC and FY conceived and designed the experiments. JC, LZ, YC, YL, WC, and XL performed experiments and data analysis. JC, LZ, and FY interpreted the results and wrote the manuscript. FY supervised the project and provided funding support. All authors reviewed and approved the final manuscript.

Funding

This work was supported by Natural Science Foundation of China (82360164); Project of Hunan Provincial Department of Education (23A0729); Health Research Project of Hunan Provincial Health Commission (202218015847, 20231214, 20230070) and Natural Science Foundation of Hunan (2023SK4014, 2024JJ7343).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

All animal experiments were approved by the Animal Ethics Committee of Hunan University of Medicine General Hospital (No. 202403055).

Consent for publication

All authors of this study agreed to publish.

Competing interests

The authors declare no competing interests.

Clinical trial number

Not applicable.