Low-intensity focused ultrasound-activated piezoelectric gel bandage for diabetic wound repair and neuropathic pain relief

Abstract

Nociceptive hypersensitivity and prolonged wound healing due to the interaction of peripheral neuropathy and local immune disorders are key issues that need to be addressed in diabetic foot ulcer (DFU) regenerative repair. Here, we present a low-intensity focused ultrasound (LIFU)-activated piezoelectric gel bandage for DFU wound repair and neuropathic pain relief. Acting as an artificial “skin”, this self-powered, functionalized bandage not only temporarily shields tissue from external environment, but also serves as a drug reservoir that can in situ release nitric oxide upon LIFU to promote macrophage polarization, generate piezoelectric current to desensitize TRPV1 nociceptor for on demand neuropathic pain relief and produce reactive oxygen species to eliminate pathogens in male rodents. More importantly, this regimen can promote CGRP neuropeptides release from sensory neurons to invigorate M2-like macrophages-mediated protective cutaneous immunity. This LIFU-activated immune-nociceptor modulation strategy is potentially applicable to other non-healing tissue regeneration.

Introduction

Diabetic foot ulcers (DFU) are one of the most devastating complications of advanced diabetes mellitus, with high disability and mortality rates, imposing heavy clinical and economic burdens on society¹. More than half of DFUs become infected, with about 20% of moderate to severe infections leading to amputation, and the mortality after amputation exceeds 70% within 5 years for all diabetic patients^2,3. Long-term poor glycemic control in diabetic patients not only creates a persistent local inflammatory microenvironment that leads to immune disorders and vasculopathy, but also causes peripheral neuropathy that results in abnormal foot sensation and protective function deficiency^4,5. These factors work together to make DFU highly susceptible to infection and prolonged healing when exposed to adverse external factors. Current treatment of DFU mainly focuses on systemic glycemic control, microvascular recanalization, infection prevention and local wound care, focusing solely on the healing effect but meeting with rather limited success^6,7,8. Thus, developing comprehensive therapeutic regimens to simultaneously remodel immune homeostasis and relieve peripheral nerve perception becomes indispensable for DFU management.

Diabetic wound microenvironment, as a local factor that participates in the pathogenesis and continues to influence the wound repair process with the support of systemic factors, consists mainly of fibroblasts, immune cells, endothelial cells and various cytokines^9,10. Of these, the immune components play a key role in coordinating tissue healing. During chronic inflammation in diabetes, macrophages are recruited to the trauma site and release cytokines to modulate the immune microenvironment¹¹. Notably, macrophages are highly plastic and can be categorized into pro-inflammatory state (M1-like macrophages) and anti-inflammatory state (M2-like macrophages) depending on their response to different stimuli and functional phenotypes. M1-like macrophages, activated by Th1 cytokines (e.g., γ-interferon, tumor necrosis factor-α) and exogenous danger signals, exhibit strong cytotoxicity and pro-inflammatory properties, whereas M2-like macrophages mainly exert anti-inflammatory and pro-cell proliferation effects¹². Besides, studies have shown that macrophages are involved in the regeneration process of damaged peripheral nerves. Modulation of macrophage polarization is expected to reverse the local inflammatory microenvironment and support peripheral nerve repair to promote diabetic wound recovery^13,14.

In addition to a dysregulated immune microenvironment, abnormal pain perception caused by peripheral neuropathy remains a non-neglectable problem affecting diabetic wound healing and even patients’ life quality¹⁵. Considering that the expression, transport and phosphorylation status of ion channels within sensory neurons directly affect functional abnormalities, targeting the regulation of specific ion channels may remodel diabetic peripheral nerve perception. Transient receptor potential vanilloid-1 (TRPV1), a receptor highly expressed in dorsal root ganglia (DRG) neurons, is closely involved in the pain process¹⁶. TRPV1 can be activated by deleterious chemical, mechanical and thermal insults, triggering calcium influx, and in turn, a cascade of events linked to neurogenic inflammation and nociception^17,18. Growing studies have validated the long-term analgesic efficacy of TRPV1 agonists or antagonists in diverse preclinical chronic pain models^19,20. However, currently available selective TRPV1 agonists, such as capsaicin (Cap), have limited clinical efficacy due to their local irritation, such as stinging, burning, and erythema. Therefore, enormous efforts have been devoted to developing efficacious TRPV1 agonists without the adverse effects associated with Cap application.

Local electrical stimulation has been proven to successfully stimulate receptor internalization, macrophage polarization and tissue regeneration; however, the application of implanted electrodes remains restrictive due to side effects such as invasiveness, infection, and electrode breakage. A self-powered therapeutic strategy based on physical principles thus offers an ideal solution^21,22,23. Barium titanate (BaTiO₃, BT), a representative piezoelectric material, converts acoustic mechanical energy into electrical energy, enabling ultrasound (US)-driven wireless stimulation; meanwhile, the piezoelectric catalytic process produces reactive oxygen species (ROS) to participate in the wound healing process²⁴. These materials exhibit micro-pressure-induced polarization, dynamic built-in electric fields, and excellent biocompatibility^25,26. Studies confirm that US-triggered piezoelectric BT systems significantly shorten the healing cycle of bacterially infected wounds and improve immune homeostasis^27,28. Critically, the ROS generated by piezoelectric catalysis needs to be properly regulated to coordinate its dual role of antibacterial and healing^29,30,31. Low-intensity focused ultrasound (LIFU), with its precise energy focusing capability, avoids tissue damage and discomfort caused by excessive energy (e.g., significant suppression of acute pain perception in the human dorsal anterior cingulate) while enhancing therapeutic efficiency³². Therefore, the synergistic interaction between LIFU and piezoelectric materials presents a promising strategy for achieving enhanced therapeutic outcomes.

Herein, we introduce a LIFU-activated piezoelectric gel (BT-dNO@Gel) bandage with immune homeostasis modulation and neuropathic pain relief functions for comprehensive repair of diabetic wounds. This self-powered system was composed of polyethyleneimine (PEI) and polyvinyl alcohol (PVA) cross-linked hydrogel encapsulating piezoelectric BT nanoparticles with surface-coated nitric oxide (NO) donor capable of sequentially releasing NO, generating ROS and outputting localized direct current in response to LIFU stimulation. First, the use of non-invasive LIFU as a switch for precise spatiotemporal control avoided local damage or discomfort caused by implanted electrodes or excessive incident energy. Second, the outer hydrogel layer possessed favorable biocompatibility, adhesion, and mechanical flexibility to conformably fill complex wound areas and effectively transmit US and electrical currents into deep regions. Finally, the subsequent on-demand released NO synergistically with free radicals and piezoelectric currents generated by BT mechanical deformation accelerates the wound healing process through combating pathogen infection, polarizing macrophages to M2-like phenotype, desensitizing TRPV1 nociceptor and releasing calcitonin gene-related peptide (CGRP). This LIFU-mediated immune-nociceptor modulation strategy not only presents a valuable approach for diabetic wound care, but also holds great potential for other chronic or acute wounds.

Results

Preparation and characterization of the BT-dNO@Gel bandage

The BT-dNO nanoparticles (NPs) with both piezoelectric and NO release properties were first prepared by stepwise surface modification processes and NO donor loading (Fig. 1a). Since bare BT NPs are prone to agglomeration, hydroxylation modification (BT-OH) and polydopamine surface coating (BT-OH-pDA) were employed to enhance their hydrophilicity, conductivity and biocompatibility. The typical peak at 1430 cm⁻¹ in the Fourier transform infrared (FTIR) spectra and the typical peak of C=C stretching of the benzene ring at 1565 cm⁻¹ in the Raman spectra confirmed the existence of OH and pDA, respectively (Supplementary Fig. 1a, b). Notably, since the surface modification processes may form a covering layer on the BT-OH surface through π-π stacking or hydrogen bonding interactions, this can interfere with the vibration freedom of free hydroxyl groups, leading to a weakened or broadened 1430 cm⁻¹ peak that merges into the baseline noise. X-ray photoelectron spectroscopy (XPS) and UV-vis spectra confirmed the successful loading of NO donor (N-diazeniumdiolates, NONOate) (Fig. 1b, c). Thermogravimetric analysis (TGA) demonstrated that the mass contents of NONOate was 3.62 wt%, which was in agreement with the loading content (3.95 ± 0.14%) determined by the absorbance of NONOate at 253 nm in UV-vis spectra (Fig. 1d). Transmission electron microscopy (TEM) images and energy dispersive spectroscopy (EDS) elemental mapping displayed that BT-dNO NPs were polygonal-like shape with a thin polymer layer on the surface and a uniform distribution of Ba, Ti, N, O and C elements (Fig. 1e and Supplementary Fig. 2a, b). The average hydrated particle size of BT-dNO was 80.49 ± 12.11 nm with the zeta potential of −17.9 ± 0.55 mV measured by dynamic light scattering (DLS) (Fig. 1f and Supplementary Fig. 3). Furthermore, the synthesized BT-dNO NPs possessed high colloidal stability, as observed by being well dispersed in various solutions within a week, which may be attributed to their negative charge and the anti-fouling effect of the deprotonated hydroxyl and pDA coatings (Fig. 1g).

Fig. 1: Preparation and characterization of BT-dNO@Gel bandage.

a A schematic illustration of the BT-dNO@Gel design. b Full XPS scanning spectra of the nanoparticles. c UV-vis absorbance spectra of the nanoparticles. d TGA curves of the nanoparticles. e Representative TEM image and elemental mapping of the BT-dNO. f The average hydrodynamic diameter of the BT, BT-OH, BT-OH-pDA and BT-dNO determined by DLS. g Time-dependent average hydrodynamic diameter of the BT-dNO in various physiological conditions (Inset: digital image of bare BT and BT-dNO dissolved in different media after 1 week at room temperature). Data were presented as mean ± SD (n = 3 independent experiments). h Digital images showing the formation of BT-dNO@Gel. i Cryo-SEM images of the BT-dNO@Gel. j Representative fluorescent images of BT-dNO@Gel, in which BT-dNO was labeled with Cy3 (red) and the blank hydrogel was labeled with FITC (green). All images were acquired from the same BT-dNO@Gel sample. k Amplitude oscillation strain test of the BT-dNO@Gel with alternated strain switching from 1% to 500% for four cycles (Inset: photographs of macroscopic autonomous healing ability of the BT-dNO@Gel). l Swelling rate of the BT-dNO@Gel. m Digital photos (right) and relative quantification (left) of hemostasis evaluation on rat tail hemorrhage model. Data were presented as mean ± SD (n = 3 rats per group). Statistical difference was calculated using two-tailed unpaired Student’s t-test. ****P < 0.0001. Representative images were from n = 3 independent experiments in (e, g–k, m). Source data are provided as a Source Data file. Elements in (a, h) were created using templates from BioRender. Li, X. (2026) https://BioRender.com/vfilyre.

To immobilize the BT-dNO NPs on wound surface for long-lasting efficacy, PEI/PVA gel with good acoustic permeability and adhesion property were designed as the bandage substrate (Fig. 1a). Briefly, BT-dNO NPs were added into phenylboronic-acid-functionalized branched PEI (PBA-PEI) to form the precursor solution, followed by mixed with PVA solution to experience a solution-gel transition under imine, boron ester and hydrogen bonding interactions without the need for additional stimuli (Fig. 1h). The characteristic UV-vis absorption peak of PEI-PBA at 230 nm as well as FTIR absorption peak of C=N bond at 1645 cm⁻¹ and B-O-C bond at 1371 cm⁻¹ confirmed the reaction of amino group with aldehyde group and the formation of boron ester bond (Supplementary Fig. 4a, b). Dynamic rheological tests showed the storage modulus (G′) was much higher than the loss modulus (G″) values within 30 min with a frequency range of 0.1–100% Hz, indicating the successful formation of BT-dNO@Gel network (Supplementary Fig. 5a, b). Biological cryo-scanning electron microscopy (Cryo-SEM) presented a typical porous micro-structure of BT-dNO@Gel, which may contribute to exudate absorption and oxygen diffusion (Fig. 1i). As demonstrated by confocal imaging, the red Cy3 signals from BT-dNO NPs displayed a uniform distribution pattern in the FITC-labeled gel, further indicating the satisfactory encapsulation (Fig. 1j).

Considering that gel bandages are vulnerable to deformity by mechanical external forces, the viscosity and self-healing capacity of BT-dNO@Gel were first tested. Both G′ and G″ values were relatively stable in the strain range from 0.1% to 100%, indicating that the BT-dNO@Gel were in the linear viscoelastic region with favorable viscosity (Supplementary Fig. 6a, b). Macroscopic healing tests using rhodamine B (red) and methylene blue (blue) staining exhibited that two separate gels in contact with each other were able to heal rapidly without any external interference, and the healed gel could be lifted by its own weight without cracks (Fig. 1k). Further continuous-step strain test yielded the same results. When a high dynamic strain (500%) exceeding the critical value was applied, the value of G′ was dropped from ~100 Pa to ~10 Pa and lower than that of G″, whereas once adopting a lower strain (1%), the values of G′ and G′′ can immediately recovered almost to the original values even after four alternating cycles, demonstrating the autonomous healing behavior of the BT-dNO@Gel (Fig. 1k). An ideal wound dressing should be able to absorb tissue exudates and establish a moist local microenvironment conducive to wound repair. As expected, the BT-dNO@Gel displayed a time-dependent swelling behavior in PBS at 37 °C with a swelling rate of about 90% in 1 h and reached an equilibrium swelling state in ~2 h, indicating that the prepared bandage is sufficient to effectively absorb wound exudate (Fig. 1l). The hemostatic potential of BT-dNO@Gel was further investigated. In both the rat tail hemorrhage model and liver hemorrhage model, BT-dNO@Gel treatment reduced rat blood loss to a great extent, which can be attributed to strong adhesion between gel bandage and tissues (Fig. 1m and Supplementary Fig. 7). In addition, BT-dNO@Gel degraded almost completely in physiological media within about 10 days (Supplementary Fig. 8). The above results indicated the successful fabrication of BT-dNO@Gel bandage with tight adhesion, high self-healing, strong swelling and excellent hemostatic capacity, which is well suited for extracorporeal wound dressing applications.

BT-dNO@Gel bandage functionality in response to LIFU

Under LIFU irradiation, the outer NONOate of the BT-dNO decomposes to release NO, and the inner BT core would produce an unbalanced charge on its surface and undergo a cascading redox reaction process, which leads to the generation of ROS and a localized electric field (Fig. 2a). The LIFU-responsive NO production profile of the BT-dNO@Gel was first evaluated under physiological conditions using the classical Griess assay. NONOate is an exogenous NO donor and free NONOate displayed a concentration-, LIFU intensity-, and irradiation time-dependent NO release features (Supplementary Fig. 9a–c). Subsequent grafting NONOate onto the surface of BT NPs also produced a suitable concentration of NO that is beneficial for wound healing, and the NO-responsive release property remained undiminished after further encapsulation into the PEI/PVA gel (Fig. 3b). The NO production yield was not only positively linear with the concentration of BT-dNO in the gel, but also strongly correlated with LIFU intensity or duration (Fig. 2c, d and Supplementary Fig. 10).

Fig. 2: NO release and piezoelectric catalytic performance of BT-dNO@Gel bandage.

a Schematic diagram of the piezocatalytic effect of BT-dNO@Gel under LIFU activation. b–d The NO release profiles under different formulations (b), different BT-dNO@Gel concentrations (c) and different LIFU irradiation durations (d). e XRD pattern of the bare BT and the BT-dNO. f Schematic diagram of the PFM test and the height sensor of BT-dNO. g Piezoresponsive phase and amplitude curves of BT-dNO. h The output voltage signal of BT-dNO and BT-dNO@Gel with or without LIFU irradiation. i The sonocurrent response of BT-dNO and BT-dNO@Gel in the LIFU intensity range of 0–1.5 W cm⁻². j The ROS generation capacity under different formulations determined by rhodamine B degradation assays. k, l ESR spectra of ·O₂ ⁻ (k) and ·OH (l) trapped by DMPO. Data were presented as mean ± SD (n = 3 independent experiments). Statistical significance was determined using a one-way ANOVA test followed by Tukey’s multiple comparison analysis. ns, not significant, *P < 0.05, **P < 0.01 and ****P < 0.0001. Source data are provided as a Source Data file. Elements in (a) were created using templates from BioRender. Li, X. (2026) https://BioRender.com/vfilyre.

Fig. 3: In vitro pro-repair, anti-inflammatory and antibacterial properties of BT-dNO@Gel bandage.

a, b Representative live/dead staining images (a) and quantitative analysis of live/dead ratio (b) of HUVECs after different treatments. c, d Representative migration images (c) and relative quantification (d) of HUVECs after different treatments for 24 h and 48 h (Scale bar: 200 µm). e–g Representative images (e) and quantitative analysis of total length (f) and mesh numbers (g) representing the tube formation ability of HUVECs after different treatments. h Flow cytometry analysis of macrophage polarization after different treatments. i Relative quantification of the ratio of M2- to M1-like macrophages. j SEM images of S. aureus treated by the representative formulations for 4 h at 37 °C. The arrows showed the membrane rupture with cytoplasm outflow. Representative images were from n = 3 independent experiments. k, l Confocal images (k) and relative quantification (l) of S. aureus stained with O11 after different treatments. m, n Representative photographs of S. aureus CFUs (m) and quantitative analysis of bacterial survival rates (n) after different treatments. o Representative S. aureus live/dead staining images after different treatments. Live bacteria stained with SYTO 9 (green) and dead bacteria stained with PI (red), respectively. p Protein leakage from S. aureus after different treatments. G1, Control; G2, LIFU; G3, Gel; G4, BT@Gel; G5, BT@Gel+LIFU; G6, BT-dNO@Gel+LIFU. Data were presented as mean ± SD (n = 3 biologically independent samples). Statistical significance was determined using a one-way ANOVA test followed by Tukey’s multiple comparison analysis. ns, not significant, *P < 0.05, **P < 0.01 and ****P < 0.0001. Source data are provided as a Source Data file.

Then, the piezoelectric catalytic performance of the BT-dNO@Gel was systematically estimated. The X-ray diffractometer (XRD) pattern of BT-dNO showed a peak splitting occurred around 2θ = 45°, resulting in the separated (002) and (200) reflections, verifying the preservation of tetragonality to generate piezoelectricity (Fig. 2e). A typical butterfly amplitude curve with a ramp voltage starting from −10 to +10 V observed in piezoforce microscopy (PFM) suggested a constantly varied strain induced by the applied external electric field (Fig. 2f, g and Supplementary Fig. 11a, b). Besides, on the phase chart, there was an ~180° phase switch in the local circuit of piezoelectric hysteresis, exhibiting the desirable piezoelectric efficacy of BT-dNO. The LIFU-response output voltage of BT-dNO@Gel was measured using a high-precision benchtop digital multimeter. The BT NPs hardly produced any voltage in the absence of LIFU (Fig. 2h). In contrast, the output voltage of either BT or BT-dNO NPs was markedly increased under LIFU stimulation at 1.5 W cm⁻², indicating that surface modification and the drug loading process did not affect the piezoelectric properties of BT NPs. More importantly, the voltage transmission capability of BT-dNO NPs was amplified to hundreds of millivolts after encapsulation in gel substrate, suggesting that higher electrical signal outputs could be obtained from BT-dNO@Gel at the same LIFU parameter. In line with the results of the piezoelectric voltage output, the real-time piezoelectric current generated by the gel bandage under LIFU stimulation showed that BT-dNO generated a ‌peak current density of 0.12 ± 0.04 μA cm⁻², especially after encapsulated in gel matrix, the current transmission capacity of BT-dNO nanoparticles was amplified about 10 times (Fig. 2i). The built-in electric field of BT-dNO NPs under LIFU stimulation enables piezoelectric separation of electrons and holes to catalyze the ROS generation. The rhodamine B degradation assay confirmed the ROS generation capacity of BT-dNO@Gel under LIFU activation, and the ROS yield was positively correlated with BT-dNO concentration, LIFU power and duration (Fig. 2j and Supplementary Fig. 12a–c). Electron spin resonance (ESR) was adopted to systematically identify and quantify the ROS types produced by BT-dNO@Gel upon LIFU using 5,5-dimethyl-1-pyrroline-N-oxide (DMPO) spin-trapping agent. Characteristic 1:2:2:1 and 1:1:1:1 quartet signals were observed in BT-dNO and BT-dNO@Gel groups under LIFU irradiation, confirming the generation of hydroxyl radicals (·OH) and superoxide anions (·O₂⁻), respectively (Fig. 2k, l). The relative contributions of each ROS were quantified by signal deconvolution, revealing ·OH as the predominant species. These results suggested that LIFU can be used as a switch to spatiotemporally control NO release, direct current generation and ROS production from BT-dNO@Gel bandage, thus facilitating subsequent biological applications.

In vitro pro-repair, anti-inflammatory and antibacterial performance of BT-dNO@Gel bandage

Prior to cellular experiments, the cytotoxicity of BT-dNO@Gel was first assessed. CCK-8 assay showed negligible effect of BT-dNO (less than 0.3 mg mL⁻¹) and PEI/PVA gel (less than 3 mg mL⁻¹) on HUVEC viability (Supplementary Fig. 13a, b). Meanwhile, the hemolysis test exhibited that the blood cell hemolysis rate of BT-dNO was less than 5% within the safe range (less than 0.3 mg mL⁻¹), suggesting its favorable hemocompatibility (Supplementary Fig. 14 and Supplementary Table 1). Consistent with this, live/dead staining and relative quantitative analysis revealed that either LIFU alone or BT-dNO@Gel combined with LIFU did not affect the cellular activity of HUVECs (Fig. 3a, b). Cell migration is essential for wound healing. The scratch assay showed that the combination of LIFU and BT@Gel or BT-dNO@Gel significantly promoted the migration of HUVECs compared with LIFU irradiation alone, and in particular, the 48 h migration rate of HUVECs reached 84.7% after BT-dNO@Gel+LIFU treatment (about 2.5 times higher than the control group) (Fig. 3c, d). Similar phenomena were observed in keratinocyte (HaCat cell line) and fibroblast (HSF cell line), confirming the high biocompatibility and robust pro-cell migration ability of BT-dNO@Gel+LIFU treatment regimen (Supplementary Fig. 15a–h). To elucidate the mechanism of the pro-healing effect, we first measured the NO release kinetics at the cellular level. Griess assay displayed that both BT-dNO and BT-dNO@Gel liberated NO in situ under LIFU activation. Under preset conditions, the maximum instantaneous concentration of NO released from BT-dNO@Gel was about 0.2 μM, which is considered the beneficial concentration range for wound healing (Supplementary Fig. 16a, b). It is noteworthy that a small amount of NO was still detectable after the removal of LIFU, which may be due to the slow spontaneous release of NO from the NONOate donors. Consistent with our conjecture, blocking NO signaling with NS-2028, a soluble guanylate cyclase (sGC) inhibitor, significantly decreased the cell migration rate of HUVECs, indicating the released NO was the main contributor of LIFU-activated BT-dNO@Gel bandage to promoting cell migration (Supplementary Fig. 17a–e). Moreover, tube formation results revealed that BT-dNO@Gel+LIFU treatment achieved greater total tube length and more tube meshes than BT@Gel+LIFU treatment (Fig. 3e–g). This satisfactory healing efficacy may be attributed to the released NO and piezoelectric current from BT-dNO@Gel under LIFU activation that stimulated the growth and migration of endothelial cells.

Given the pivotal role of macrophages in orchestrating the inflammatory phase of wound healing, we assessed the ability of our piezoelectric gel bandage to promote the anti-inflammatory M2-like macrophage polarization essential for the resolution of inflammation and tissue repair. Expressions of the inflammatory marker CD86 and the anti-inflammatory marker CD206 in macrophages (RAW264.7) were assessed using cellular immunofluorescence (IF) staining and flow cytometry. IF staining showed that BT-dNO@Gel+LIFU concurrently downregulated lipopolysaccharides (LPS) and interferon-γ (IFN-γ)-induced M1-like macrophages while upregulating M2-like macrophages (Supplementary Fig. 18a, b), suggesting that BT-dNO@Gel+LIFU could encourage macrophages’ repolarization from M1- to M2-like phenotype. Similar findings were verified in flow cytometry, where BT-dNO@Gel+LIFU treatment significantly increased the M2/M1 ratio by about 6-fold compared to the M1 control, underscoring the anti-inflammatory properties of the piezoelectric gel bandage (Fig. 3h, i).

Infection is one of the major obstacles to diabetic wound healing; dressings with antimicrobial activity would be helpful in impeding pathogen invasion and boosting wound recovery. The bactericidal effect of BT-dNO@Gel bandage was tested in Gram-positive bacteria represented by S. aureus and Gram-negative bacteria represented by E. coli. As shown by the SEM, bacteria in the control group exhibited an intact morphology with a smooth surface, while partial cytoplasmic efflux from bacterial membrane damage was seen in the Gel and BT@Gel groups (Fig. 3j). Notably, significant morphological disruption was observed in the BT@Gel+LIFU and BT-dNO@Gel+LIFU groups, with the bacterial membrane severely wrinkled and disrupted, causing inner contents leakage. This confirmed the antibacterial efficacy of LIFU-activated BT-dNO@Gel. Several studies have shown that ROS can imbalance the antioxidant system of bacteria, thereby inhibiting their growth and activity^33,34. The fluorescent probe O11 is a common tool to visualize bacterial ROS generation, and strong ROS green fluorescence was detected in BT@Gel and BT-dNO@Gel-treated S. aureus after LIFU irradiation (Fig. 3k, l). Then, the antimicrobial properties of BT-dNO@Gel against S. aureus were evaluated using the typical spread plate method. Compared to LIFU alone, which had no apparent effect on bacterial growth, the blank gel treatment inhibited bacterial viability to a certain extent (Fig. 3m, n). This is in line with previous studies that cationic polyamine PEI can interact with microbial membranes to cause membrane disruption³⁵. Both BT@Gel and BT-dNO@Gel treatment dramatically reduced S. aureus survival rate in the presence of LIFU, as manifested by substantial protein leakage from the bacteria, which was also confirmed by the bacterial live/dead staining (Fig. 3o, p and Supplementary Fig. 19). A similar phenomenon was observed in E. coli (Supplementary Fig. 20a–h). To confirm whether ROS at this bactericidal concentration would cause damage to normal cells, we assessed the production of ROS in HUVECs, which was consistent with the previously non-toxic results. Piezo-catalyzed ROS production within HUVECs was lower than that of the positive control (Rosup concentration < 50 μg mL⁻¹) and progressively decreased over time, which was within the safe threshold range of cell viability (Supplementary Fig. 21a–f). The LIFU-triggered ROS generation from BT-dNO@Gel indeed affected bacteria survival but not HUVECs, probably due to the fact that the respiratory chain of eukaryotic cells is located in intracellular mitochondria and their superior ROS-scavenging systems, which are more resistant to external ROS interference than bacteria^36,37,38. Overall, the above data indicated that LIFU-activated BT-dNO@Gel bandage possessed excellent pro-cell migration, angiogenic, macrophage polarization and bactericidal capacities in vitro.

Repairing efficacy of BT-dNO@Gel bandage on diabetic wound

Biosafety is a prerequisite for clinical application and translation; the local and systemic toxicity of BT-dNO@Gel plus LIFU was assessed before in vivo treatment. Healthy rats were subjected to PBS or BT-dNO@Gel plus LIFU treatment, respectively. H&E staining of local skin tissues collected on day 14 after treatment showed no significant difference between the two groups in terms of either cellular morphology and structure or inflammatory cell infiltration, suggesting the as-prepared bandage had no toxicity to normal skin (Supplementary Fig. 22a). Meanwhile, analysis of blood specimens from rats in the middle and late stages of treatment revealed that the blood cells and liver & kidney function indicators of treated rats were all within normal levels (Supplementary Fig. 22b). Consistently, H&E staining of major organs after treatment also revealed no abnormalities (Supplementary Fig. 22c). The above results indicated that the BT-dNO@Gel+LIFU treatment regimen had satisfactory biocompatibility with high translational potential.

Skin wounds can self-heal within a few days under normal physiological conditions. In diabetic wounds, however, the sustained inflammatory microenvironment caused by chronic hyperglycemia presents a significant challenge to normal physiological healing³⁹. Encouraged by the favorable efficacy in vitro, the repair effect of BT-dNO@Gel was next evaluated in a full-thickness skin wound model of streptozotocin (STZ)-induced diabetic rats (Fig. 4a). Successful modeling was defined as random blood glucose concentration >16.7 mmol L⁻¹ measured by blood sampled from the tail vein after 1 week. Two weeks after injection, a 2 cm full-thickness skin wound was created on the back of the rats using a sterile perforator. Diabetic rats with back wounds were randomly divided into 6 groups: (1) control, (2) LIFU, (3) Gel, (4) BT@Gel, (5) BT@Gel+LIFU and (6) BT-dNO@Gel+LIFU and treated every other day (dose of BT was 0.3 mg mL⁻¹, dose of BT-dNO was 0.3 mg mL⁻¹, LIFU parameters were 1.0 MHz, 1.5 W cm⁻², 50% duty cycle and 120 s duration). During the observation period, there were no significant differences in body weight of the rats in each group, and the blood glucose level of all diabetic rats remained consistently >16.7 mmol L⁻¹ (Supplementary Fig. 23a, b). Recognizing that the therapeutic efficacy of BT-dNO@Gel+LIFU treatment involves the regulation of NO release, whose overproduction would interfere with wound healing, we first investigated the kinetics of NO release within the wound skin tissue following treatment. BT-dNO@Gel bandage coated on diabetic wounds rapidly released NO under LIFU stimulation, reaching a peak concentration of about 0.25 μM, and then gradually decreased and maintained at a relatively low level, which is considered beneficial for wound healing (Supplementary Fig. 24a, b). It is noteworthy that a small amount of NO was still detectable after the removal of LIFU, which may be due to the slow spontaneous release of NO from the NONOate donors. Digital photos and quantitative analysis of the wound sites showed that all the gel-containing groups facilitated the healing process to some extent compared to the control group with significantly delayed wound healing, especially in the BT-dNO@Gel+LIFU treatment group, which may be attributed to well-adherent hydrogels providing a suitable matrix-like environment for tissue regeneration (Fig. 4b, c). Notably, the wound closure rate of BT@Gel with LIFU was significantly higher than that without LIFU, suggesting a pro-healing effect mediated by piezoelectric stimulation. Consistent with expectations, the BT-dNO@Gel+LIFU group achieved optimal wound repair, with a 93.15% wound closure rate compared to the control. Moreover, this treatment regimen exhibited an obviously faster healing rate than recombinant bovine basic fibroblast growth factor (rb-bFGF) dressing, which is commercially available for wound repair, in promoting diabetic wound healing (Supplementary Fig. 25a–d).

Fig. 4: In vivo wound healing efficacy on diabetic rat models.

a Schematic diagram of the in vivo study design on diabetic rat wound model. b Representative photographs of diabetic wounds and traces of wound closure from day 0 to day 14 in different groups (Scale bar: 5 mm). c Relative wound closure rates of rats after different interventions (n = 5 rats per group). d Representative images of H&E staining of rat skin tissue (Scale bar, 300 µm (top) and 80 μm (enlarged)). e Representative images of Masson staining and collagen I and collagen III IHC staining of rat skin tissue (Scale bar: 80 μm). f Representative IF staining images of angiogenic markers CD31 (green), α-SMA (red) and Ki67 (pink) in diabetic wounds of different groups on day 14 (Scale bar: 50 μm). White arrows indicate CD31 and α-SMA double-positive vessels. g, h Quantitative analysis of granulation tissue width (g) and epithelial thickness (h) from rat skin tissue H&E staining on day 14. i Quantitative analysis of collagen deposition from rat skin tissue after different interventions on day 14. j Quantitative analysis of CD31⁺ vessels per unit area. k Quantitative analysis of mature blood vessels with CD31⁺/α-SMA⁺ per unit area. l Quantitative analysis of CD31⁺Ki67⁺ cell rate in diabetic wounds of different groups. m, n Serum EGF (m) and VEGF (n) secretion levels in different groups. o, p Representative IF staining images (o) and quantitative analysis (p) of CD86 (red) and CD206 (green) in diabetic wounds of different groups on day 14 (Scale bar: 50 μm). q–s ELISA analysis of TNF-α (q), IL-10 (r) and IL-6 (s) from the wound tissue lysates extracted at day 14 after treatment. G1, Control; G2, LIFU; G3, Gel; G4, BT@Gel; G5, BT@Gel+LIFU; G6, BT-dNO@Gel+LIFU. Data were presented as mean ± SD (n = 3 rats per group). Statistical significance was determined using a one-way ANOVA test followed by Tukey’s multiple comparison analysis. ns, not significant, *P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001. Source data are provided as a Source Data file. Elements in (a) were created using templates from BioRender. Li, X. (2026) https://BioRender.com/vfilyre.

Consistent with accelerated macroscopic closure, H&E staining revealed significant histological improvements in BT-dNO@Gel+LIFU-treated wounds, as displayed by abundant neoplastic granulation tissue in the wound center with fewer inflammatory cells and increased epidermal viability compared to untreated diabetic controls (Fig. 4d, g, h). Moreover, more collagen deposition and highly mature and oriented collagen structures, which are essential for cell proliferation, differentiation and wound healing, were observed in the BT-dNO@Gel+LIFU group compared to the other groups, showing better extracellular matrix (ECM) remodeling (Fig. 4e, i and Supplementary Fig. 26). The co-analysis of K14-labeled keratinocytes, vimentin-positive fibroblasts, and CD31-marked endothelial cells provides critical insights into the ‌epidermal-dermal-vascular tri-lineage synergy during skin regeneration^40,41,42. Dual IF staining for K14 and Ki67 revealed a significantly expanded zone of proliferating basal keratinocytes (K14⁺Ki67⁺) at the wound edge after BT-dNO@Gel+LIFU treatment, directly supporting the accelerated re-epithelialization observed in H&E (Supplementary Fig. 27a, b). Similarly, dual IF staining with vimentin and Ki67 showed an obviously increased percentage of proliferating fibroblasts after BT-dNO@Gel+LIFU treatment, which is essential for ECM production (Supplementary Fig. 27c, d). In parallel, there was a noticeably higher density of CD31⁺α-SMA⁺ vessels in the BT-dNO@Gel+LIFU group, along with an increased proportion of proliferating endothelial cells (CD31⁺Ki67⁺), suggesting the active neovascularization and maturation (Fig. 4f, j–l). This phenomenon was further verified in enzyme-linked immunosorbent assay (ELISA), where the BT-dNO@Gel+LIFU group secreted the highest amount of epidermal growth factor (EGF) and vascular endothelial growth factor (VEGF) compared to the other groups (Fig. 4m, n). These results demonstrated that BT-dNO@Gel bandage combined with LIFU can significantly accelerate diabetic wound healing by promoting re-epithelialization, collagen deposition and amplified angiogenesis.

Inflammation is crucial in the wound repair process and reducing inflammation level will help the wound move into the proliferative phase⁴³. Macrophages, as essential immune cells in the regulation of inflammation during wound repair, are highly plastic and capable of phagocytizing pathogens to promote wound healing⁴⁴. To elucidate the therapeutic mechanism of LIFU-activated BT-dNO@Gel bandage, we assessed the macrophage status in the diabetic wound microenvironment. The in vivo M2/M1 macrophage ratio was confirmed by IF staining for CD206 and CD86. Compared to other groups, higher populations of M2-like macrophages and lower populations of M1-like macrophages were observed in BT@Gel+LIFU and BT-dNO@Gel+LIFU groups on both day 7 and day 14, which resulted in the greater M2/M1 ratio, indicating decreased pro-inflammatory and increased anti-inflammatory capacities (Supplementary Fig. 28a–c and Fig. 4o, p). Notably, BT-dNO@Gel+LIFU treatment exhibited the highest anti-inflammatory efficacy, which can be attributed to the synergistic action of released NO and the piezoelectric current triggered by LIFU. To gain mechanistic insights, we employed a mouse model as a well-established system for analyzing immune cell responses and profiling macrophage activation following the treatments. M2-like macrophages (CD11b⁺ F4/80⁺ CD206⁺) and M1-like macrophages (CD11b⁺ F4/80⁺ CD86⁺) were quantified by flow cytometry, and the results were in agreement with the IF trends (Supplementary Fig. 29). In addition, compared to the persistent inflammatory responses in the control, LIFU, Gel and BT@Gel groups, the expression of pro-inflammatory cytokines including tumor necrosis factor-α (TNF-α) and interleukin-6 (IL-6) at the wound site decreased dramatically after BT-dNO@Gel+LIFU treatment, while the expression of anti-inflammatory cytokine interleukin-10 (IL-10) increased substantially (Supplementary Fig. 28d–f and Fig. 4q–s). Taken together, these findings conclusively showed that BT-dNO@Gel+LIFU treatment could effectively regulate local immune-inflammation homeostasis to accelerate diabetic wound healing.

Antibacterial efficacy of BT-dNO@Gel bandage against infected diabetic wounds

Inspired by the commendable antibacterial effect of LIFU-activated BT-dNO@Gel bandage in vitro, we further investigated its efficacy against rat diabetic infected wounds. Diabetic rat model with a full-thickness wound was constructed as mentioned before. For the S. aureus infection wound model, a suspension of S. aureus (150 μL, 10⁸ CFU mL⁻¹) was first dropped onto the skin wound site and bacterial infection was confirmed 24 h later by standard swab culture and solid agar plate counting (Fig. 5a and Supplementary Fig. 30). The infected diabetic rats were randomly divided into 6 groups and treated every other day according to the aforementioned regimens. Digital photos of the infected wounds were recorded, and corresponding wound areas for each group were calculated. The findings revealed that groups containing gel all promoted wound healing to some extent, especially BT-dNO@Gel+LIFU treatment achieved a 90.16% wound healing rate on day 14 (Fig. 5b–d). Meanwhile, sample fluid from the wound edges of each group was collected on days 7 and 14 after different interventions. Despite the fact that obvious bacterial colonies were observed in all groups on day 7, it was evident that colony counts in BT@Gel+LIFU and BT-dNO@Gel+LIFU groups were slightly less than those in other groups, which could be attributed to the bactericidal effect of ROS generated by LIFU-activated piezoelectric catalytic reaction (Fig. 5e, f). Notably, the antibacterial effect of BT@Gel+LIFU and BT-dNO@Gel+LIFU groups was particularly excellent by day 14, with only a trace of bacterial growth. Recognizing that the efficacy of antibacterial treatment with piezoelectric gel bandage lies in the regulation of ROS generation, whose overproduction leads to oxidative stress. Therefore, we investigated the kinetics of ROS release in wound skin tissue after treatment. DHE staining showed that low levels of ROS fluorescence were detected in rat skin tissues after BT-dNO@Gel+LIFU treatment (Supplementary Fig. 31a–d). Over time, the ROS fluorescence was gradually absorbed and degraded, decreasing in concentration after exerting an effective antibacterial effect and thus progressing toward the direction of favorable wound healing. Consistent with the antibacterial results, wounds treated with BT-dNO@Gel+LIFU had narrower granulation tissue width and thicker regenerated epithelial tissue (Fig. 5g, i, j). Meanwhile, BT-dNO@Gel+LIFU-treated wounds had thicker collagen bundles with better alignment, orientation and collagen deposition density than other groups (Fig. 5h, k).

Fig. 5: Anti-bacterial restorative efficacy against S. aureus-infected diabetic wound.

a Schematic diagram of the study design on S. aureus-infected diabetic wound model. b Representative photographs of S. aureus-infected diabetic wounds from day 0 to day 14 in different groups (Scale bar, 5 mm). c Traces of wound closure from day 0 to day 14 in different groups. d Relative wound closure rates of rats after different interventions (n = 5 rats per group). e, f Bacterial colonies (e) and survival rates (f) of S. aureus-infected wounds on days 7 and 14 in each group (Scale bar, 10 mm). g Representative images of H&E staining of rat skin tissue (Scale bar, 300 µm (top) and 80 μm (enlarged)). h Representative images of Masson staining of rat skin tissue (Scale bar, 80 μm). i, j Quantitative analysis of granulation tissue width (i) and epithelial thickness (j) from rat skin tissue H&E staining. k Quantitative analysis of collagen deposition from rat skin tissue after different interventions. l Representative IF staining images of CD86 (red) and CD206 (green) in diabetic wounds of different groups on day 14 (Scale bar, 80 μm). m Quantitative analysis of M2/M1 macrophages in diabetic wounds of different groups on day 14. n–p ELISA analysis of TNF-α (n), IL-10 (o) and IL-6 (p) from the wound tissue lysates extracted at day 14 after treatment. G1, Control; G2, LIFU; G3, Gel; G4, BT@Gel; G5, BT@Gel+LIFU; G6, BT-dNO@Gel+LIFU. Data were presented as mean ± SD (n = 3 rats per group). Statistical significance was determined using a one-way ANOVA test followed by Tukey’s multiple comparison analysis. ns, not significant, *P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001. Source data are provided as a Source Data file. Elements in (a) were created using templates from BioRender. Li, X. (2026) https://BioRender.com/vfilyre.

Considering that a high bacterial burden would worsen inflammatory response and exacerbate immune disorders at the wound site, we wondered whether bacterial infections affect the regulation capacity of BT-dNO@Gel+LIFU on the local inflammatory microenvironment. IF staining showed that even in the infected diabetic wound model, BT-dNO@Gel+LIFU treatment could still promote macrophage polarization toward the M2-like phenotype, especially in the early stages of wound repair, as verified by the higher CD206 expression and lower CD86 expression at the wound site (Fig. 5l, m). Unsurprisingly, inflammatory indicators gradually recovered as the immune disorder resolved, with a decrease in pro-inflammatory cytokines (e.g., TNF-α and IL-6) and an increase in anti-inflammatory factors (e.g., IL-10) (Fig. 5n–p). Similar results were observed in E. coli-infected diabetic wound, where BT-dNO@Gel+LIFU treatment significantly promoted refractory wound regenerative repair by reducing bacterial load and modulating the inflammatory microenvironment (Supplementary Fig. 32a–p). Collectively, our results validated the antibacterial repair effectiveness of LIFU-activated BT-dNO@Gel in diabetic wounds infected with S. aureus and E. coli.

Antinociceptive efficacy of BT-dNO@Gel bandage under LIFU

Interactions between the local inflammatory microenvironment of diabetic wounds and resident immune cells can alter nociceptive messaging through releasing pro-inflammatory cytokines and chemokines⁴⁵. To test the antinociceptive effect of BT-dNO@Gel upon LIFU irradiation, a diabetic neuropathic pain (DNP) model was constructed (Fig. 6a). Naive rats of the same weight were used as negative controls without receiving any treatment. Von Frey filament testing and hot plate assay were applied to test mechanical and thermal pain behavior, respectively. Compared to naive rat, DNP rat exhibited significantly lower pain thresholds, as displayed by decreased mechanical paw withdrawal threshold (PWT) and thermal paw withdrawal latency (PWL) (Fig. 6b, c). The successfully modeled DNP rats were then randomly divided into six groups and received different treatments according to the aforementioned regimens. Digital photos of DFU and corresponding quantitative analysis of wound area revealed that the BT-dNO@Gel+LIFU treatment group achieved the best wound healing rate among other groups on day 14 (Supplementary Fig. 33a–c). In addition, compared with the PBS group, BT-dNO@Gel+LIFU therapy significantly increased the pain threshold in DNP rats, manifested by an increase in both 50% PWT and PWL (Fig. 6d, e). More importantly, the pain thresholds of the diabetic rats did not decrease during the observation period, suggesting that our LIFU-mediated immune-nociceptor modulation strategy had a durable analgesic effect. In contrast, the pain threshold of non-diabetic wound rats was recovered rapidly by day 14 regardless of treatment, which was consistent with their uncompromised healing capacity (Supplementary Fig. 34a–c).

Fig. 6: Neuropathic pain resolving effect of BT-dNO@Gel bandage under LIFU.

a Schematic diagram of the study design on rat DNP model. b, c Changes of 50% PWT (b) and thermal PWL (c) in normal and DNP rats within 28 days (n = 5 rats per group). d, e Changes of 50 % PWT (d) and thermal PWL (e) of rats during treatment (n = 3 rats per group). f Schematic diagram of Cap or LIFU-activated BT-dNO@Gel bandage-induced analgesia. g Fluorescence images and quantitative analysis of TRPV1 expression in DRG neurons under different treatments (n = 18 biologically independent samples) (Scale bar, 10 μm). h Ca²⁺ influx fluorescence images and quantitative analysis of DRG neurons under different treatments (n = 18 biologically independent samples) (Scale bar, 15 μm). i–k Representative IF images showing Na_v1.8 (red), TRPV1 (green) and DAPI (blue) expressions (i) and relative quantification of Nav1.8⁺TRPV1⁺/Nav1.8⁺ (j) and MFI of TRPV1 (k) in L4/L5 DRG on the ipsilateral side after intervention (n = 18 slices from 3 rats per group) (Scale bar, 200 μm). l–n Representative IF images showing GFAP (red), Iba-1 (green) and DAPI (blue) expressions (l) and relative quantification of GFAP (m) and Iba-1 (n) in L4/L5 spinal dorsal horn after intervention (n = 18 slices from 3 rats per group) (Scale bar, 100 μm and 50 μm (magnification)). G1, Control; G2, Cap; G3, BT-dNO@Gel+LIFU; G4, Cap (30 min); G5, BT-dNO@Gel+LIFU (30 min). Data were presented as mean ± SD. Statistical significance was determined using two-tailed unpaired Student’s t-test and one-way ANOVA test, followed by Tukey’s multiple comparison analysis. ns, not significant, *P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001. Source data are provided as a Source Data file. Elements in (a–c, f) were created using templates from BioRender. Li, X. (2026) https://BioRender.com/vfilyre.

Next, we wonder whether the BT-dNO@Gel plus LIFU exerts antinociceptive effects in vivo via targeting TRPV1, a critical pain nociceptor overexpressed in DRG sensory neurons. Cap, a selective TRPV1 channel agonist, was introduced as a positive control. High concentrations of Cap induce TRPV1 activation and Ca²⁺ influx, which causes receptor desensitization of TRPV1 ion channels^46,47. TRPV1 receptor desensitization then decreases TRPV1 response to inflammatory mediators or endogenous ligands, and in turn reduces TRPV1-dependent hyperalgesia (Fig. 6f). As expected, Cap (1 μM) treatment markedly promoted TRPV1 expression in DRG neurons, which was instead diminished with prolonged incubation time (Fig. 6g). Similar phenomenon and more pronounced TRPV1 inactivation were observed after BT-dNO@Gel plus LIFU treatment. Meanwhile, intracellular Ca²⁺ levels were measured in DRG neurons using calcium imaging techniques. Compared to the control group with no significant change in Ca²⁺ signal, both Cap and BT-dNO@Gel plus LIFU treatment induced a dramatic increase in intracellular Ca²⁺ fluorescence signal, which then diminished at about 30 min (Fig. 6h). The above results suggested that the antinociceptive efficacy of BT-dNO@Gel plus LIFU strategy may be due to impaired Ca²⁺ influx resulting from TRPV1 receptor desensitization.

To further validate the mechanism of BT-dNO@Gel plus LIFU-induced TRPV1 desensitization in vivo, L4/L5 DRG and spinal cord of rats were collected 24 h after intervention. Na_v1.8 is a voltage-gated sodium channel highly expressed in DRG nociceptive neurons‌. As a ‌functional and structural marker‌ of nociceptors, Na_v1.8 expression is tightly linked to neuronal excitability and pathological pain states, making it a robust complementary target for assessing sensory neuron activity⁴⁸. In the rat DNP model, the ‌co-upregulation of Na_v1.8 and TRPV1 expression‌ reflects neuronal response mechanisms to persistent noxious stimuli. After BT-dNO@Gel+LIFU intervention, the observed ‌reduction in activity of Na_v1.8⁺TRPV1⁺ neurons‌ demonstrates a significant correlation with the attenuation of mechanical allodynia, providing critical insights into their role in mediating pathological pain (Fig. 6i–k). Given that the glial region is involved in neuronal signaling, particularly the transmission of nociceptive messages, we further examined the activation of astrocytes (GFAP⁺) and microglia (Iba-1⁺) in the spinal dorsal horn glial region. In line with previous results, BT-dNO@Gel+LIFU treatment markedly inhibited the activation of microglia and astrocytes in DNP rats (Fig. 6l–n). The endogenous opioid pathway and α₂-adrenergic receptor pathway are also common potential pain targets^49,50. To ascertain whether the analgesic effect of BT-dNO@Gel+LIFU depends on these pathways, we examined the effects of the nonselective opioid receptor antagonist Naloxone and the α₂-adrenergic receptor antagonist Yohimbine on BT-dNO@Gel+LIFU-induced analgesia through the Von Frey filament testing and hot plate assay. As shown in Supplementary Fig. 35, intraperitoneal administration of Naloxone or Yohimbine did not attenuate the analgesic potency, suggesting that BT-dNO@Gel+LIFU exerts antinociceptive effects independently of the function of opioid receptors and α₂-adrenergic receptors. Collectively, the above results suggested that our piezoelectric gel bandage hinders Ca²⁺ influx by desensitizing TRPV1 nociceptor under LIFU activation, thereby attenuating nociceptive messaging-related neuronal excitability for effective neuropathic pain relief.

BT-dNO@Gel bandage orchestrated neuro-immune cross-talk

To comprehensively explore transcriptional dynamics after LIFU plus BT-dNO@Gel bandage treatment, we collected total RNA from skin wounds of well-established DNP rats and analyzed the transcriptome landscape. Principal component analysis (PCA) demonstrated clear separation among the naive, control and BT-dNO@Gel+LIFU treatment groups, confirming distinct global transcriptomic signatures (Fig. 7a). The Venn diagram showed that 565 differentially expressed genes (DEGs) were shared among the three groups (Fig. 7b). Hierarchical clustering of DEGs showed significant differences between the naive group and control group, highlighting the extensive transcriptional reprogramming induced by diabetes and injury (Supplementary Fig. 36a, b). Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses revealed that genes altered by diabetes and ulcers were significantly enriched not only in glucose and lipid metabolism pathways, but primarily in pathways related to inflammation and oxidative stress, ECM remodeling, and cell differentiation, migration and response to stimulus (Supplementary Fig. 36c, d). This profile reflects a molecular landscape skewed toward chronic inflammation, compromised tissue regeneration and hyperalgesia, characteristic of the non-healing diabetic wound. Crucially, comparison of BT-dNO@Gel+LIFU group and control group identified 2942 significant DEGs (1292 upregulated and 1650 downregulated), demonstrating the potent ability of piezoelectric gel bandage to modulate the wound transcriptome (Fig. 7c). KEGG and GO analyses revealed that BT-dNO@Gel+LIFU treatment significantly reversed diabetes-induced expression patterns dysregulation, as verified by shifts in key transcriptional pathways including inflammation-immune response regulation, ECM remodeling, cell surface receptor signaling pathway and cytokine-cytokine receptor interaction (Fig. 7d, e). Furthermore, targeted analysis displayed that genes encoding immune and inflammatory regulation, tissue repair, and pain regulation in the BT-dNO@Gel+LIFU group were shifted significantly toward the naive profile compared to the control group, which was in accordance with the alleviation of pain sensitivity symptoms and remodeling of immune homeostasis (Fig. 7f).

Fig. 7: Neuro-immune cross-talk driven by BT-dNO@Gel bandage under LIFU.

a PCA analysis presenting differentially expressed genes as identified by RNA-seq in the naive, control, and BT-dNO@Gel+LIFU groups. b Venn diagram portraying differential gene counts in the naive, control, and BT-dNO@Gel+LIFU groups. c Volcano plots of differentially expressed gene. d KEGG enrichment analysis of differentially expressed gene. e A chord diagram showing GO enrichment terms. f Heat map of differentially expressed genes. g Expression of Calca after different treatments on DRG neurons by RT-qPCR. h ELISA analysis of CGRP release in the supernatant of cultured DRG neurons stimulated for 24 h. i–l Representative IF images showing Na_v1.8 (red), CGRP (yellow), CD206 (green) and DAPI (blue) expressions (i) and relative quantification analysis of Na_v1.8 nerves (j), CGRP area % (k) and CD206⁺ cell per field (l) (Scale bar, 100 μm). m Flow cytometry analysis and relative quantification of macrophage polarization after different treatments (n = 3 biologically independent samples). G1, control; G2, BT-dNO@Gel+LIFU; G3, BT-dNO@Gel+LIFU + CGRP8-37. Data were presented as mean ± SD (n = 3 rats per group). Statistical significance was determined using a one-way ANOVA test followed by Tukey’s multiple comparison analysis. ns, not significant, *P < 0.05, **P < 0.01 and ****P < 0.0001. Source data are provided as a Source Data file.

The transcriptome atlas outcomes impelled us to further explore the mechanism of neuro-immune interplay in DFU wound regeneration. Recent studies revealed that the sensory nerves located within cutaneous tissue may signal dermal immune cells through the neuropeptide CGRP during the healing process⁵¹. Given the importance of peptidergic sensory neurons in mediating neuro-immune communication during tissue healing, we evaluated the expression of CGRP neuropeptide after treatment. RT-qPCR results showed that BT-dNO@Gel+LIFU treatment upregulated Calca expression by 3.4 times in DRG neurons, which was consistent with the ELISA assay that BT-dNO@Gel+LIFU treatment increased soluble CGRP release in DRGs (Fig. 7g, h). Further multiplex IF co-staining for CGRP with definitive neuronal markers Na_v1.8 and M2-like macrophage markers CD206 revealed that the CGRP signal occurred specifically in Na_v1.8 cutaneous nociceptors, and its spatial association with CD206⁺ macrophages suggested a neuro-immune axis driving wound repair (Fig. 7i). Quantitative analysis showed that compared with the control group, BT-dNO@Gel+LIFU treatment significantly promoted CGRP release and subsequent polarization of CD206⁺ macrophages, while the polarization process of macrophages toward anti-inflammatory phenotype could be blocked by ‌CGRP receptor antagonist (CGRP8-37), which did not affect the release of soluble CGRP in skin tissue (Fig. 7j–l and Supplementary Fig. 37). To directly validate the correlation between CGRP signaling and macrophage phenotype, we assessed the polarization of RAW264.7 macrophages in vitro, a widely used and well-established model for studying macrophage activation. Expressions of the inflammatory marker CD86 and the anti-inflammatory marker CD206 in macrophages were assessed using flow cytometry. The results showed that LPS and IFN-γ treatment significantly increased the proportion of M1-like macrophages, while IL-4 treatment increased the proportion of M2-like macrophages (Fig. 7m). Compared with the M1 positive control group, fewer CD86-labeled M1-like macrophages and more CD206-labeled M2-like macrophages were detected in the α-CGRP group. After blocking CGRP signaling, the proportion of CD206-marked macrophages decreased, which directly confirmed the role of CGRP in inducing macrophage phenotypic transformation and anti-inflammatory regulation. The above findings confirmed that the LIFU-mediated immune-nociceptor modulation strategy can orchestrate subcutaneous sensory nerve and M2-like macrophages interactions to achieve wound regenerative repair and analgesia by invigorating protective cutaneous immunity (Fig. 8).

Fig. 8: Schematic diagram of the LIFU-mediated immune-nociceptor modulation strategy for DFU repair and analgesia.

Acting as an artificial “skin,” this self-powered, functionalized bandage temporarily shields tissue from the external environment and also serves as a drug reservoir. Upon LIFU activation, the in situ liberated nitric oxide accelerated tissue repair via promoting immunomodulation, angiogenesis and endothelial cell migration. Concurrently, the embedded barium titanate undergoes deformation upon ultrasonic mechanical force, producing free radicals to fight bacterial infections and outputting direct current for neuroregulation, desensitizing nociceptive receptors and promoting the release of CGRP neuropeptides from sensory neurons to activate M2-like macrophage-mediated protective cutaneous immunity. This neuro-immune interaction significantly accelerated the diabetic wound healing process and elevated the pain threshold in diabetic neuropathic pain (DNP) rats. LIFU low-intensity focused ultrasound, ROS reactive oxygen species, NO nitric oxide, M1 M1-like macrophage, M2 M2-like macrophage, DNP diabetic neuropathic pain, DRG dorsal root ganglia. Elements were created using templates from BioRender. Li, X. (2026) https://BioRender.com/vfilyre.

Discussion

DFU is one of the most serious complications of diabetes and poses tremendous health burdens to patients worldwide. Approximately 50% to 60% of ulcers become infected, leading to a significant increase in amputation risk and mortality rate⁵². DFU is caused by multiple factors that ultimately lead to skin breakdown and prolonged healing, in which the interaction between local immune disorder and peripheral sensory neuropathy plays a critical role⁵³. Most current treatments focus solely on inflammatory microenvironmental modulation, neglecting the role of neurologic factors in the tissue repair process and obtaining modest efficacy. Thus, the development of effective integrated regimens with synergistic immunomodulatory and neuropathic pain relief for comprehensive care of DFU is highly desirable.

Here, we developed a self-powered piezoelectric gel bandage BT-dNO@Gel for complicated diabetic wound management (Fig. 8). Specifically, PEI/PVA hydrogel with tight adhesion, high self-healing, strong swelling and excellent hemostatic capacities was employed as an extracorporeal wound dressing to prevent pathogen invasion and facilitate exogenous ultrasonic transmission. To achieve spatiotemporal on-demand modulation at the wound site, we chose LIFU as a therapeutic trigger switch to optimize the penetration depth and minimize adverse impact on non-target tissue. NO, as an endogenous molecule, has diverse biological effects at different concentrations, including vasodilation, immunomodulation and cell proliferation^54,55,56. The in situ produced NO under LIFU activation at the wound site via NONOate decomposition overcomes the routine NO delivery challenge caused by its short half-life. Electrical stimulation is considered an effective method for ion channel modulation and nerve repair^21,22,57. The piezoelectric BT loaded in the bandage can convert the mechanical forces generated by LIFU into microcurrents and simultaneously generate ROS, which enables powerless, precise and safe electrical stimulation for neuromodulation and antibacterial infection.

Systematic experimental results indicated that our LIFU-mediated immune-nociceptor modulation strategy exerted potent efficacy in combating pathogen infection, polarizing and activating macrophage and desensitizing TRPV1 nociceptors, which ultimately accelerated diabetic infected wound healing and improved pain threshold in DNP rats. Further genetic analysis revealed that BT-dNO@Gel+LIFU treatment altered the pre-existing pro-inflammatory transcriptome landscape of DNP rat to an anti-inflammatory niche that facilitates wound repair by upregulating gene expression related to immune and inflammatory regulation, tissue repair, and pain regulation. Nociceptor-derived CGRP has been shown to induce an anti-inflammatory transcriptional programme in macrophages. Further mechanistic exploration revealed that LIFU-mediated immune-nociceptor modulation strategy resulted in increased CGRP release from sensory neurons, which regulated the release of anti-inflammatory cytokines by activating M2-like macrophages. The elevated anti-inflammatory cytokines, in turn, alleviated nociceptive signaling in sensory neurons, which may be used to explain the elevated rat pain threshold after combination treatment. It should be noted that although our findings highlight the role of BT-dNO@Gel+LIFU in promoting CGRP release from nociceptors to modulate macrophage polarization, this focus does not preclude the involvement of ancillary pathways (e.g., bidirectional communication with endothelial cells)⁵⁸. Furthermore, consistent findings from physiologically relevant rat diabetic wound models as well as mechanistic studies using well-established murine macrophage culture systems and immune cell analyses collectively strengthen the overall dataset and support the proposed CGRP-mediated pathway. Across all experimental conditions, the piezoelectric hydrogel bandage consistently improved the local immune microenvironment, promoted a shift toward anti-inflammatory macrophage phenotypes, and alleviated neuropathic pain, underscoring its therapeutic potential in diabetic wound repair.

There are several limitations and challenges to overcome in future research. First, our current data strongly support TRPV1 desensitization as a primary mechanism; however, consistent with the complex nature of nociceptive modulation, we acknowledge potential contributions from secondary mechanisms that our current experimental design may not have fully captured. Reliable alternative pathways worth further investigation, such as TRPA1-mediated calcium signaling cross-talk, endocannabinoid receptor modulation in DRG and descending serotonergic facilitation via 5-HT3 receptors. Second, while we employ the M1/M2-like macrophage terminology for clarity and consistency with the majority of published literature, it should be noted that macrophage polarization exists along a continuum in vivo, with numerous intermediate states beyond this dichotomous classification. Third, the parameters of LIFU may need to be re-evaluated for larger animal or human studies. Since larger animals or humans may have different anatomical and physiological bases compared to rats, including variations in skin, muscle or fat thickness, LIFU parameters may need to be adjusted to effectively trigger the piezoelectric efficacy. Fortunately, this should be easily achieved based on the development of ultrasound therapy, as it has been approved by the FDA for widespread clinical applications, including Parkinson’s disease, hypertension, tumor ablation, bone healing and cartilage regeneration, etc.^59,60,61. Finally, the piezoelectric gel bandage still faces the challenge of clinical translation. Due to insufficient preclinical research data and a lack of unified guidelines for biological safety evaluation, further exploration and verification are needed before clinical application.

In conclusion, our study presented a LIFU-activated immune-nociceptor modulation strategy based on self-powered piezoelectric gel bandage capable of remodeling immune homeostasis and desensitizing peripheral nerve perception for wound repair and neuropathic pain relief of DFU. Our findings provide a potential approach for the development of regenerative medicine, especially for diabetic patients with peripheral neuropathy. This LIFU-mediated neuro-immune regenerative axis modulation strategy, either as standalone treatments or in combination with existing therapies, holds great promise for addressing chronic wounds and other non-healing tissues.

Methods

Animals, cell lines and bacterial strains

All animal studies comply with policies of the National Ministry of Health and obtained approval from the Laboratory Animal Center of Shanghai Tenth People’s Hospital (SHDSYY-2023-4674). Male Sprague-Dawley (SD) rats (170–200 g) and male BALB/c mice (6–8 weeks) were ordered from Salccas Biotech Co., Ltd. All animal experiments were conducted exclusively with male animals. This approach was taken to eliminate the potential confounding effect of estrogen on wound healing and glycemic control in the diabetic model. Animals were housed in a specific pathogen-free environment under controlled conditions (20 ± 3 °C, about 40–70% humidity and 12 h light/dark cycle). Cell lines (HUVECs (CRL-4053), HSF (CRL‑4066), HaCat (CRL‑4048) and RAW264.7 (TIB-71)) and bacterial strains (S. aureus (6538) and E. coli (8739)) were obtained originally from American Type Culture Collection (ATCC). Cells were cultured in a humidified incubator containing 5% CO₂ at 37 °C using specific medium (Dulbecco’s Modified Eagle Medium (DMEM, Gibco) supplemented with 1% penicillin-streptomycin (Gibco) and 10% fetal bovine serum (FBS, Gibco)). Cells were routinely tested for mycoplasma contamination every 2 months. S. aureus and E. coli were suspended in Luria-Bertani medium (Solarbio) and cultured in oscillation incubator at 37 °C. Bacteria were harvested during the exponential growth phase and adjusted to 1 × 10⁸ CFU mL⁻¹ prior to use.

Materials

Barium titanate (BT, 99.9%), NONOate (98%), polyethyleneimine (PEI, 99%) and polyvinyl alcohol (PVA, 99%) were obtained from Rhawn. Dopamine (98%) and rhodamine B (≥99%) were purchased from Macklin. Phenylboronic acid (PBA, 98%) and 1-(3-dimethylaminopropyl)-3-ethylcarbodiimide hydrochloride (EDC, 99%) were obtained from Bidepharm. FITC (≥90%) and cyanine3 (Cy3, ≥95%) were purchased from BioFroxx GmbH. Methylene blue (≥70%), Capsaicin (Cap, ≥95%), Naloxone (≥97%) and Yohimbine (≥99%) were gained from Aladdin. Calcium ion indicator (Fluo-4 AM), Griess reagent kit and DCFH-DA were obtained from Beyotime. Streptozotocin (STZ, ≥98%) was purchased from Maokang Biotech Co., Ltd. Rat VEGF and EGF ELISA kits were acquired from Enzyme-linked Biotech Co., Ltd. Rat TNF-α, IL-10, IL-6 and CGRP ELISA kits were acquired from Multi Sciences Biotech Co., Ltd. α-CGRP (99.8%) and rat CGRP-(8-37) (99%) were purchased from MedChemExpress.

Synthesis and characterization of BT-dNO

In detail, BT (1 g) was refluxed in hydrogen peroxide (H₂O₂) aqueous solution (50 mL, 30 wt%) at 105 °C under stirring. The hydroxylated modified BT (BT-OH) was cooled to room temperature (RT), then collected after centrifugation (7200×g, 5 min), washing (three times with deionized water) and drying (60 °C in a vacuum oven for overnight). BT-OH (12.5 mg) was dissolved in Tris buffer at pH 8.5 (25 mL, 10 mM) and homogenized using an ultrasonicator. Fresh dopamine (2.5 mg) was rapidly added to the solution to initiate oxidative polymerization, with a following overnight agitation at RT. Polydopamine-coated BT-OH (BT-OH-pDA) was then collected by centrifugation (8900×g, 10 min). Finally, NONOate aqueous solution (4 mL, 1 mg mL⁻¹) was dropwise added to the BT-OH-pDA solution (8 mL, 1 mg mL⁻¹) to obtain BT-dNO through π–π attraction. The loading content of NONOate on BT-dNO was determined by UV-vis spectrum at 253 nm and by TGA (NETZSCH STA 449 F5). During TGA, the sample was heated from RT to 600 °C at a rate of 5 °C min⁻¹ in a nitrogen environment.

Preparation and characterization of BT-dNO@Gel bandage

PEI (1 g) was dissolved in methanol (5 mL), followed by adjustment of the pH of the PEI solution to 5.0–6.0 with HCl (1 M). PBA powder (1 g) was dissolved in DMF (5 mL). EDC (2 g) was then added to the PBA solution, followed by the prepared diluted PEI solution. The pH value was maintained within 5.0 to 6.0 throughout the 3 h reaction process. The product was dialyzed against deionized water for 3 days by changing deionized water three times daily. Finally, the PEI-PBA concentrated solution was collected and vacuum freeze-dried for reserve. BT-dNO solution (0.3 mL, 2 mg mL⁻¹) was thoroughly mixed with PEI-PBA solution (1 mL, 40 mg mL⁻¹), followed by the addition of PVA aqueous solution (1.3 mL, 40 mg mL⁻¹) to form a BT-dNO@Gel complex. The biological cryo-SEM (FEI Quanta 450) was used to examine the surface and section morphology of the BT-dNO@Gel bandage. The rheological properties were evaluated using a Thermo Haake Rheostress rheometer (RS6000, Thermo Scientific).

Characterization

Chemical structures were characterized by XPS (Thermo Escalab 250Xi) and Raman microscope (Renishaw inVia). UV-vis spectra and FT-IR spectra were recorded using UV-vis-NIR spectrophotometer (PE Lambda 950) and Nicolet FT-IR IS 10 spectrometer, respectively. TEM (FEI Tecnai G2 F30) was used to analyze sample morphology and elemental composition. Particle size as well as zeta potential were detected by a laser particle analyzer (Zetasizer Nano ZS90). XRD analysis was performed using a Rigaku D/MAX-2200 PC XRD diffractometer with Cu Kα radiation (λ = 1.5406 Å). The piezoelectric output was measured under an electrochemical analysis instrument (CHI660E). PFM measurement was conducted by an atomic force microscope (AFM, Bruker Dimension ICON) equipped with a ferroelectric test module. BT-dNO samples were dispersed onto a conductive Pt-coated silicon substrate. A Cr/Au-coated conductive probe was used in contact mode. During imaging, a 3 V AC excitation voltage was applied to the tip, with the drive frequency tuned to the contact-resonance range (250–400 kHz). Amplitude and phase images were recorded simultaneously for each scanned area.

Swelling and degradation assays

Hydrogel samples were freeze-dried and recorded for initial dry weights (W₀), then immersed in PBS solution (5 mL, 37 °C) simulating physiological conditions. The hydrogel was removed at the specified time point and gently blotted with filter paper to remove surface moisture. For the swelling test, the hydrogels were weighed as W_ts, and the swelling ratio was determined according to: swelling ratio (%) = (W_ts-W₀) / W₀ × 100%. For the degradation test, the hydrogel samples were lyophilized and weighed as W_td. The degradation percent of the hydrogel was computed by using the formula: degradation percent (%) = (W₀-W_td) / W_td × 100%.

Hemostatic assay

For rat tail amputation model, after anesthesia with isoflurane, the rats were secured to a corkboard. Approximately 50% of the tail was surgically removed using sterile scissors. Following tail transection, BT-dNO@Gel (500 μL) was applied to the wound. Blood loss was continuously monitored during hemostasis, while wounds without any treatment were used as controls. For liver hemorrhaging model, rats were anesthetized with isoflurane, and the liver was exposed via abdominal incision. After gently removing excess surface fluid, the mouse liver was positioned on a pre-weighed filter paper. An incision of 5 mm was made in the liver using micro scissors to induce bleeding, followed by immediate application of 500 μL BT-dNO@Gel. After complete hemostasis, bleeding time and the increase in filter paper weight were recorded and compared with the control group.

NO release and piezoelectrocatalytic properties

The NO release performance of BT-dNO was assessed using the Griess assay. The release profiles under various conditions were determined by recording characteristic absorbance (450 nm) using a UV-vis-NIR spectrophotometer and a microplate reader (Thermo Multiskan FC). NO level was determined based on a standard calibration curve generated using NaNO₂, with untreated deionized water serving as the control. Rhodamine B degradation and ESR were used to characterize ROS generation of the gel bandage, and DCFH-DA was used as an indicator of ROS at the cellular level. BT-dNO (5 mg) was dispersed in aqueous rhodamine B solution (5 mL, 10 mg L⁻¹) and stirred for 1 h to establish adsorption-desorption equilibrium. Following various treatments, the samples were centrifuged to collect transparent supernatants, and the residual rhodamine B was measured by UV-vis-NIR spectroscopy. Bruker A300-10/12 ESR spectrometer further detected different types of ROS. DMPO (Sigma) served as the spin-trapping reagent for ·OH and ·O₂⁻, and the resulting signals were recorded using an ESR spectrometer at ambient temperature.

Cytotoxicity evaluation

Cytotoxicity and cell proliferation assay were assessed using the CCK-8 assay (Dojindo). HUVECs, HSF and HaCat cells were pre-seeded into 96-well plates (5 × 10³ cells per well) and cultured overnight. BT-dNO solutions and gel leachates at varying concentrations were added and incubated for another 24 h. After adding CCK-8 reagent to each well, cell viability was assessed by measuring absorbance at 450 nm using a microplate reader. Moreover, cells were subjected to different treatments for 2 days, and their viability was gauged using a Calcein-AM/PI Double Staining Kit (Dojindo), followed by observation using confocal laser scanning microscopy (CarlZeiss LSM900). Fresh red blood cells (RBCs) were collected from healthy male SD rats and suspended in PBS for the hemolysis assay. The RBC suspension was incubated with BT-dNO solutions at different concentrations for 2 h at 37 °C. Hemolysis was quantified by measuring the supernatant absorbance at 570 nm and calculated using: Hemolysis (%) = (OD _sample − OD _negative) / (OD _positive − OD _negative) × 100%. The negative control and positive control were deionized water and PBS, respectively.

Scratch assay

HUVEC, HSF and HaCat cells were seeded into 6-well plates (5 × 10⁵ cells per well) and cultured until confluent. Use a sterile micropipette tip to create a scratch in the culture dish, then rinse to remove suspended cells and debris. Subsequently, the medium was replaced according to the groups and subjected to various treatments. Images were captured by a microscope at different culture times (0, 24, and 48 h). Migration rate was determined according to the following formula: cell migration (%) = (S₀ − S_t) / S₀ × 100%. S₀ and S_t denote the scratch area at 0 h and the predetermined time, respectively.

Tube formation assay

Matrigel was thawed overnight in an ice bath (4 °C). Diluted Matrigel (200 μL) was pre-added to 24-well plate to allow gelation. HUVECs (5 × 10⁴ cells per well) were then evenly seeded onto the polymerized Matrigel and incubated with different materials for 24 h. After incubation, the cells were stained with 0.5 µM Calcein-AM dye and observed under a fluorescence microscope. The junctions, total tube length and meshes were quantified using ImageJ software (version 2.1.0).

Macrophage polarization analysis

RAW264.7 (1 × 10⁵ cells per well) were seeded in confocal dishes and cultured until confluent. The M1-like phenotype was induced by co-incubation with LPS (100 ng mL⁻¹, ABclonal, RPM0001) and IFN-γ (20 ng mL⁻¹, ABclonal, RP01070), while the M2-like phenotype was induced by IL-4 (20 ng mL⁻¹, ABclonal, RP01161) stimuli. IF staining (Leica SP8 LSCM) and flow cytometry (BD LSRFortessa™ X-20) were performed to evaluate macrophage polarization status. Antibodies used in this assay included anti-CD86 (diluted at 1: 400, Abcam, ab119857, clone no. GL-1), anti-CD206 (diluted at 1: 500, Abcam, ab64693), Alexa Fluor® 488-conjugated secondary antibody (diluted at 1: 800, Abcam, ab150157), Alexa Fluor® 594-conjugated secondary antibody (diluted at 1: 800, Abcam, ab150080), APC anti-CD86 (0.2 μg per test, Biolegend, 159215, clone no. A17199A), PE anti-CD206 (0.2 μg per test, Biolegend, 141705, clone no. C068C2).

For flow cytometry analysis, the harvested tissue was first digested into a single-cell suspension using hyaluronidase/collagenase II. Antibodies used for staining included: APC-Cy7 anti-CD45 (Biolegend, 103115, clone no. 30-F11), APC anti-F4/80 (Biolegend, 157305, clone no. QA17A29), FITC anti-CD11b (Biolegend, 101205, clone no. M1/70), PE-Cy7 anti-CD86 (Biolegend, 105013, clone no. GL-1), PE anti-CD206 (Biolegend, 141705, clone no. C068C2). Each antibody was diluted to a final working concentration of 0.2–0.5 μg per test. Data were collected and analyzed using a flow cytometer (BD LSRFortessa™ X-20) and FlowJo software (10.6.2).

In vitro antibacterial assessment

SEM (Sigma 300) was used to observe the bacteria (S. aureus and E. coli) morphology after different treatments. The ROS level of bacteria was assessed using the bacterial ROS detection kit (BestBio) and imaged under an upright microscope (Leica DM6 B). Spread plate assay was adopted to evaluate the antibacterial effect. In brief, prepared bacterial suspensions (15 μL, 10⁷mL⁻¹) were added to the pre-treated 24-well cell plate and received different group treatments. Bacterial suspensions were harvested with sterile PBS (1 mL) and diluted approximately 800-fold. Bacterial survival was determined by enumerating CFUs after 18 h of incubation on agar plates. All experiments were conducted in triplicate, and survival rates were calculated as: survival rate (%) = (C / C₀) × 100%. C₀ and C denote colony numbers in control and treated groups, respectively. In addition, bacterial suspensions subjected to different treatments were passed through 0.22 μm filters, and the concentration of bacterial protein leakage was calculated by the BCA protein quantification kit (Beyotime). To evaluate bactericidal efficacy, each sample was tested using a viability/cytotoxicity assay kit (L6060, UElandy) and observed under an upright microscope (Leica DM6 B, Germany). The bacterial reduction rate (%) was calculated as: reduction rate (%) = (N / N₀) × 100%. N₀ and N denote the total and the dead bacterial count, respectively.

Study in diabetic wound model

For the establishment of the diabetic models, animals were fasted for more than 12 h prior to STZ administration. The rat diabetes model was induced by a single i.p. injection of STZ (50 mg kg⁻¹), while the mouse diabetes model was established through five consecutive i.p. injections of STZ (50 mg/kg). Blood glucose levels in all animals were measured twice a week using a glucometer (Accu-Chek®) for 4 weeks after STZ injection. Successful modeling of diabetic animals was defined as randomized glucose >16.7 mmol L⁻¹ for two consecutive readings. After modeling, the diabetic animals were anesthetized (1% pentobarbital sodium) and a back full-thickness skin wound (2 cm diameter for rats and 8 mm diameter for mice) was created using a round punch under aseptic conditions. The successfully modeled animals were randomly divided into six groups: Control (PBS), LIFU, Gel, BT@Gel, BT@Gel+LIFU and BT-dNO@Gel+LIFU (n = 5 rats per group). The dose of BT or BT-dNO was 0.3 mg mL⁻¹, LIFU parameters were 1.5 W cm⁻², 50% duty cycle, 1.0 MHz and 120 s duration, and freshly prepared gel bandage was replaced every other day to ensure sustained bioactive ion release. Wound images were acquired at predetermined time points to evaluate the healing process. The wound closure rate was calculated based on the following formula: wound closure (%) = (A₀ − A_t) / A₀ × 100%. A₀ and A_t denote the original wound area and the wound area at different time points, respectively.

Study in rat diabetic infected wound model

For diabetic rat with infected wound model, after experiencing the above STZ-induced diabetes and establishing the back skin defect, the S. aureus or E. coli suspension (150 μL, 1 × 10⁸ CFU mL⁻¹) was spread onto the wounds. Wound infection was confirmed 24 h after inoculation by routine swab culture followed by spread plate analysis on solid agar. The successfully modeled rats were then randomly allocated into groups based on the aforementioned grouping (n = 5 rats per group). Wound healing was monitored by photographing and measuring wound areas at predetermined time points, while bacterial infection was assessed by swab culture and spread plate methods on days 7 and 14.

Study in rat DNP model

For DNP model, the diabetic rats were first constructed according to the aforementioned method, and the blood glucose status and behavioral changes were monitored. One month after injection, rats with blood glucose levels >16.7 mmol L⁻¹ and 50% PWT decreased more than 50% from baseline indicated successful establishment of the DNP model. The DNP rats were anesthetized (1% pentobarbital sodium), and a right foot full-thickness skin wound (8 mm diameter) was created to establish the DFU model on day 0. The successfully modeled DNP combined with DFU rats were randomly divided into six groups to receive the different treatments described above (n = 5 rats per group). The wound healing process and behavioral changes were monitored during treatment.

Behavior tests

Mechanical nociception was determined using the Von Frey filaments by measuring paw withdrawal responses to stimulation on the plantar surface of rat hind paw. Prior to routine testing, each rat received a 30 min adaptation period on pain test mesh rack covered with Plexiglas chambers. Adaptation was confirmed when the rat entered a quiet state with reduced voluntary activities. Von Frey filaments of increasing force (1.4–26 g) were applied vertically to the central plantar surface of the hind paw for 6–8 s. Paw withdrawal or licking was considered a positive response, proceeding to the next lower-force filament, whereas failure to respond led to using the next higher-force filament. The testing sequence was continued until four measurements were collected after the first behavioral change, or until four successive negative responses or five successive positive responses occurred. The 50% PWT (denoted in grams, g) was determined by the up-down method of Dixon. Hot nociception was evaluated using hot plate test, specifically by timing the latency of rapid thumping of the hind paw, hind paw-licking or jumping. Rats were placed on the plate at RT for 30 min prior to formal testing. Subsequently, they were transferred to a hot plate heated to 55 °C for observation. Measurements were conducted three times with a 15 min interval between each session, and a 30-s cut-off time is set to avoid foot burns.

Isolation, cultivation and experimentation of DRG neurons

DRG neurons were isolated from rat embryos on day 18 of embryonic development and immediately plated into cold DMEM medium. Then DRGs were then washed with DMEM medium and digested with an enzyme mixture solution (DNase Ⅰ, trypsin and collagenase) at 37 °C for 20 min until cell detachment. Next, the cell suspension was filtered with a cell strainer (100 μm) and purified with Percoll. DRG neurons were resuspended in Neurobasal medium with 2% B-27 (Gibco) and cultured in an incubator.

For IF staining of TRPV1, DRGs were pre-seeded in confocal dishes (1 × 10⁵ cells per well) and subjected to different treatments. After PBS washing, the neurons were fixed with 4% polyformaldehyde (PFA) at RT (15 min), and incubated in QuickBlock™ Blocking Buffer (Beyotime, P0260) for 15 min. After triple washing by PBS, the specimens were incubated with anti-TRPV1 primary antibody (diluted at 1:200, ABclonal, A8564) overnight at 4 °C, and then incubated with Alexa Fluor® 488-conjugated secondary antibody (diluted at 1:800, Abcam, ab150077) for 1 h at RT. Before being observed under CLSM, the nuclei were stained with DAPI.

For calcium imaging, DRG neurons were inoculated into confocal dishes (1 × 10⁵ cells per well) and cultured until cell attachment. Fluo-4 AM (4 μM) was added to the serum-free DMEM medium and co-incubated with neurons in the dark at 37 °C for 30 min. Neurons were then washed and treated with Cap or BT-dNO@Gel in culture media. LIFU stimulation was carried out at 1.5 W cm⁻² for 30 s. The fluorescence images were captured using CLSM.

Histological analysis

To assess wound closure status, epidermal regeneration, immune infiltration and analgesia mechanism, the rats skin tissues, L4/L5 DRGs and spinal cord dorsal horn were harvested and fixed in 4% PFA for 20 min–1 h, followed by dehydration and paraffin embedding. Tissue sections of 5 μm thickness were cut using a microtome for histological analysis. Re-epithelialization and collagen deposition were evaluated by H&E staining, Masson’s trichrome staining and immunohistochemistry staining of collagen I (dilution at 1: 500, Abcam, ab270993, clone no. EPR24331-53) and collagen III (dilution at 1: 200, Abcam, ab7778). ROS levels within wound tissue were assessed using DHE dye (Beyotime, S0064S). For IF staining, tissue sections underwent a 90 min blocking step followed by overnight incubation with the primary antibody at 4 °C and a 2 h incubation with the secondary antibody at RT. Prior to imaging (Leica DM3000 and Zeiss Axio Imager Z1), cell nuclei were stained with DAPI. Primary antibodies involved in experiment include Ki67 (dilution at 1: 500, Abcam, ab15580), CD31 (dilution at 1: 100, Abcam, ab222783, clone no. EPR17260-263), α-SMA (dilution at 1: 500, Bioss, bsm-33187M, clone no. 3F9), K14 (dilution at 1: 100, Thermo Fisher Scientific, MA5-11599, clone no. LL002), vimentin (dilution at 1: 1000, ABclonal, A19607, clone no. ARC0086), CD68 (dilution at 1:100, Abcam, ab283654, clone no. EPR23917-164), Arg1 (dilution at 1: 200, CST, #93668, clone no. D4E3M), CD206 (dilution at 1: 500, Abcam, ab64693), CD86 (dilution at 1: 200, ABclonal, A16805), Na_v1.8 (dilution at 1: 2000, Abcam, ab307817, clone no. EPR25132-222), TRPV1 (dilution at 1: 200, ABclonal, A8564), GFAP (dilution at 1: 500, CST, #3670, clone no. GA5), Iba-1 (dilution at 1: 200, CST, #17198, clone no. E4O4W) and CGRP (dilution at 1: 200, CST, #14959, clone no. D5R8F).

Transcriptome sequencing

Total RNA was isolated from wound tissues using TRIzol reagent at the end of treatment. RNA quality and quantification were tested using Agilent 2100 Bioanalyzer and ND-2000 NanoDrop spectrophotometer. Library construction and sequencing were performed on an Illumina HiSeq 2000 system. Differential expression analysis among the naive, control and BT-dNO@Gel+LIFU groups was analyzed using DESeq2 (v1.4.5) with a significance threshold of Q value ≤ 0.05. KEGG and GO enrichment analysis of DEGs were conducted using Phyper based on the hypergeometric test, and statistical significance was corrected by Bonferroni adjustment with a Q value threshold of ≤0.05.

Real-time quantitative polymerase chain reaction (RT-qPCR)

Total RNA was extracted with the Eastep Total RNA Extraction Kit (Promega). Complementary DNA (cDNA) was synthesized from total RNA using the HiScript® II First Strand cDNA Synthesis Kit (Vazyme Biotech Co., Ltd.). RT-qPCR analysis was performed with SYBR Green PCR Master Mix (Biotechnology), and gene expression was normalized to Gapdh. Primers were ordered from Sangon Biotech. The primer sequences for Calca and Gapdh used were as follows⁵¹:

Rat_Calca forward primer: 5’-GAAATGCCACCTTTTGACAGTG-3’, Rat_Calca reverse primer: 5’-TGGATGCTCTCATCAGGACAG-3’, Rat_Gapdh forward primer: 5’-AGGTCGGTGTGAACGGATTTG-3’, Rat_Gapdh reverse primer: 5’-TGTAGACCATGTAGTTGAGGTCA-3’.

Statistical analysis

All experiments were performed with at least three biological replicates. Quantitative data were presented as mean ± SD. Statistical analyses were performed using GraphPad Prism software (version 8.2.1), IBM SPSS software (version 27.0.1.0), and OriginPro 2024 software. A two-tailed unpaired Student’s t-test was applied for comparisons between two groups, whereas one-way ANOVA with Tukey’s post hoc test was used for multiple-group comparisons. Statistical differences were denoted as ns, not significant, *P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.