Introduction

Diabetes mellitus is a metabolic disorder characterized by persistently high levels of blood glucose. Impaired glycemic regulation, the hallmark clinical and pathological feature of diabetes, could lead to development of cardiovascular disease, renal failure, or even vision loss. Continuous and highly accurate glucose monitoring is essential for effective diabetes management and prevention of related complications1,2,3,4. In traditional ways, self-monitoring of blood glucose (SMBG) methods such as finger-prick or venous blood sampling were widely used clinically to provide accurate and instant blood glucose level information5,6,7. However, the glucose data gathered from these SMBG device fail to capture dynamic glucose fluctuations or trends, especially during critical periods such as sleep or physical activity. Continuous glucose monitoring (CGM) systems emerged to meet the demand of glycemic control and informs therapeutic strategies for diabetic patients. Utilizing an electrochemical glucose-sensing electrode that implanted subcutaneously, the CGM systems enabled measurement of glucose fluctuations from the interstitial fluid (ISF) to offer 24-h tracking of glucose variations and improve the quality of diabetes management8,9,10.

The electrochemical sensors embedded in the CGM system could perform real-time and continuous measurement of glucose concentration in ISF, allowing constant, uninterrupted monitoring of glycemic levels that facilitated fluctuation detection, trend analysis, and timely alerts for hyperglycemia or hypoglycemia events10,11, yet challenges still remain in current development of CGM devices3,12. For example, the implanted electrode of the CGM systems is long and may cause pain, tissue trauma and inflammation that promote the release of reactive oxygen species (ROS), an electrochemically active interferent that may affect the detection accuracy of molecules such as glucose or uric acid9,13. Microneedle (MN) with length ranging from 500 μm to 800 μm could penetrate the stratum corneum and epidermis without touching nerve endings or blood vessels, which can reduce the tissue damage and pain, and lower the risk of infection significantly14,15,16. MN-based CGM devices have recently emerged as effective tools for both biosensing and transdermal drug delivery17,18,19,20. For example, Kusama et al. reported a porous microneedle platform based on solid polymer with ionic conductivity for simultaneous extraction, detection, and therapeutic delivery of glucose within ISF21. Tehrani et al. developed a fully integrated wearable MN array designed for continuous and real-time fluctuation track of multiple metabolic markers, such as lactate, glucose, and ethanol during daily activities22.

Though the device portability and integration were improved markedly due to the miniaturization of sensing electrodes, a compromise in detection sensitivity is inevitable resulting from the reduction of the active sensing area. The development of high-performance MN-based glucose sensors is essential for enhancing the functionality of CGM systems. Graphene possessed outstanding properties such as high charge carrier mobility23,24 and excellent charge transport capability25,26, which made it highly promising material the fabrication and application of flexible electronics and wearable device27,28,29. Vertical graphene (VG), a nanomaterial composed of sparsely distributed graphene-containing nanosheets grown perpendicularly on a substrate30, retained graphene’s intrinsic properties while providing distinct structural advantages over planar graphene, such as a larger specific surface area and abundant edge defects. These features facilitated electron transfer and enhanced signal amplification, making VG highly suitable for high-performance miniature electrochemical sensor31,32,33,34. For instance, Liu et al. developed a portable electrochemical sensing platform based on vertically aligned graphene electrodes modified with Au nanoparticles for real-time tau protein detection of blood samples from Alzheimer’s disease patients35. However, it is challenging for current VG-based microneedle sensors to balance mechanical durability and signal stability, as they are vulnerable to damage during skin penetration that compromises the ability to sustain consistent long-term in vivo performance.

In this work, a VG-coated core-shell structured microneedle glucose sensor (VCMGS) was developed for continuous transdermal monitoring of glucose level fluctuations. The VCMGS consisted of a hollow microneedle serving as the shell for skin penetration and a VG -modified microneedle electrode as the core for transdermal glucose detection. The hollow microneedle offered strong mechanical strength to penetrate the skin, which could reduce the damage to the sensing module and enhance the sensor stability effectively. The modification of VG facilitated enhancement of electrochemical properties and detection sensitivity of the MN electrode, enabling reliable in-situ, and real-time acquisition of physiological signals from ISF (Fig. 1a). The VCMGS was demonstrated to be sensitively detect glucose concentration and exhibit a well-defined linear response, excellent selectivity, temporal stability, and reliable signal transmission performance. The MN electrode could work in a single-MN configuration, and be integrated with a circuit to form a miniatured wearable system. Both the in vitro and in vivo experimental results showed that the VCMGS was capable of monitoring glucose concentrations in real time with satisfying accuracy. This work offered a promising strategy that would find application for glucose level measurement in subcutaneous tissue fluids, possessing potential applicability and benefits in aiding the diagnosis and treatment of diabetes as well as the tracking of other disease-related biomarkers.

Fig. 1: Schematic illustration and fabrication workflow of the VG -coated core-shell microneedle glucose sensor (VCMGS) for in vivo glucose fluctuations tracking in interstitial fluid.
Fig. 1: Schematic illustration and fabrication workflow of the VG -coated core-shell microneedle glucose sensor (VCMGS) for in vivo glucose fluctuations tracking in interstitial fluid.The alternative text for this image may have been generated using AI.
Full size image

a The VCMGS comprised a VG -coated microneedle electrode as the working electrode, encapsulated by a hollow microneedle that simultaneously acted as both the counter and reference electrode. The core–shell design provided enhanced mechanical strength, ensured stable and minimally skin penetration for reliable ISF glucose measurement with wireless real-time data transmission. b Flow chart showing the fabrication of the VCMGS. Hollow MN were fabricated by laser cutting and subsequently coated with gold to obtain Au-coated HMN. These Au-coated HMN were sequentially modified with PMMA and Ag/AgCl ink to construct the HMN Reference electrode. Separately, acupuncture needle was laser-cut to prepare the microneedle substrate. Au was electrochemically deposited onto the substrate, followed by VG growth via plasma-enhanced chemical vapor deposition to obtain the VG-coated microneedle. Subsequent magnetron sputtering of Pt and immobilization of glucose oxidase was conducted to prepare the VG-coated microneedle sensing electrode, which was then assembled with the HMN Reference electrode to complete the VCMGS fabrication.(RE/CE: reference electrode/counter electrode; WE: work electrode; MCU: microcontroller unit; BLE: bluetooth low energy; AFE: analog front end; ISF: interstitial fluid.)

Result and discussion

The VCMGS comprised a VG-coated microneedle sensing electrode as the working electrode, encapsulated by a hollow microneedle reference electrode that simultaneously functioned as both the counter and reference electrode (Fig. 1b). The hollow microneedle provided robust mechanical strength for effective and stable skin penetration, enabling contact with ISF and continuous glucose measurement via the VG-based MN sensing electrode. Both electrodes were prepared based on an stainless steel substrate, which inherently suffered from poor electrochemical stability under electrical stimulation. This instability could cause polarization reactions or abrupt signal spikes during usage, thereby compromising the detection accuracy of the electrode. To prevent electrode polarization and potential loss during subsequent functionalization, an Au layer was first electrochemically deposited onto the electrode surface.

For the fabrication of the VG-based MN sensing electrode, VG was then grown onto the surface of the Au-coated MN electrode. Due to its unique structural properties, which provide a high specific surface area and excellent biocompatibility, VG enabled highly sensitive in vivo biomolecular detection and promoted continuous glucose monitoring. Radio frequency plasma-enhanced chemical vapor deposition (RF-PECVD) was introduced to grow VG directly on the Au-coated MN electrode surface to obtain the VG-coated MN electrode. The VG-coated MN electrode exhibited increased surface area and superior electrical conductivity, which effectively facilitated interfacial electron transfer and thereby improved the overall sensing performance. Additionally, a Pt layer was deposited onto the VG-coated MN electrode to enhance the catalytic decomposition of H₂O₂ and improve the electrochemical performance of the sensor. Then, glucose oxidase (GOx) was immobilized on the Pt-coated MN electrode surface to produce the GOx-coated MN electrode. For glucose detection, GOx on the electrode surface oxidizes glucose to produce H₂O₂, which is then detected electrochemically by the Pt layer under an applied bias voltage, generating amperometric signals that can be recorded by the working electrode.

For the hollow microneedles reference electrode, 23 G hollow microneedle tubes were employed and laser-cut to obtain sharp hollow microneedle (HMN). The inner sidewall of the HMN was first electrochemically deposited with Au to enhance the electrode conductivity. Next, an isolation layer was constructed via polymethyl methacrylate (PMMA) coating on the inner sidewall of the HMN to prevent electrical short circuits between the two electrodes of the VCMGS during usage (Supplementary Information Fig. S1). After Au deposition on the outer surface, the non-sensing area of the HMN was insulated accordingly. Then, a uniform layer of Ag/AgCl ink was coated onto the Au-coated HMN to obtain the HMN Reference electrode. The as-fabricated VG-based MN sensing electrode and the HMN Reference electrode were then assembled to form the VCMGS for real-time and transdermal glucose monitoring in ISF. The core–shell structure enhanced both the mechanical strength and signal stability of the sensor, ensuring reliable electrode performance during skin penetration and accurate glucose detection from ISF. The integration of VG enhanced the electrode’s sensitivity and anti-interference capabilities, beneficial for continuous and dynamic in vivo glucose monitoring.

Next, the surface morphology of the VCMGS electrodes with different functional decorations was characterized via scanning electron microscopy (SEM) to evaluate the feasibility and stability of the fabrication strategy. A thin gold layer (~10 nm) was deposited onto the electrode surface via magnetron sputtering prior to SEM analysis to enhance the surface conductivity and imaging quality. The morphology of each successive functional layer modified on the electrode surface were showed on Fig. 2a–e, including blank microneedle, Au-coated MN electrode, VG-coated MN electrode, Pt-coated MN electrode, and GOx-coated MN electrode (working electrode). The surface morphology of the microneedle was uniform and smooth, with rare apparent protrusions. For Au-coated MN electrode, a uniform and dense gold layer was deposited onto the electrode surface and fully preserved the integrity of the MN tip, which ensured sufficient mechanical strength during transdermal insertion and allowed efficient penetration of the stratum corneum to access the ISF directly. For VG-coated MN electrode, the graphene nanosheets were aligned perpendicularly and exhibited a rough surface morphology with ~1 µm interlayer spacing. The enlarged specific surface area provided abundant active sites for subsequent nanomaterial deposition and enzymatic functionalization, which in turn facilitated the electrode sensitivity improvement. As for the Pt-coated MN electrode, the Pt nanoparticles were uniformly distributed across the VG layer, with localized clustering that increased the specific surface area and enhanced catalytic activity to improve the electrochemical performance of the electrode. After GOx modification, the enzyme layer was evenly coated on the electrode surface to form dot-like protrusions. The increased surface area via VG facilitated efficient access to glucose molecules and promoted signal transduction on the electrode interface.

Fig. 2: Surface morphology characterization of working and reference electrode of VCMGS with different functional layers.
Fig. 2: Surface morphology characterization of working and reference electrode of VCMGS with different functional layers.The alternative text for this image may have been generated using AI.
Full size image

Enlarged views were included. a−e The surface morphology of the working electrode, including: (a) microneedle; (b) Au-coated MN electrode; (c) VG-coated MN electrode; (d) Pt-coated MN electrode; (e) GOx-coated MN electrode (Working electrode). f−h The surface morphology of the reference electrode: (f) hollow MN; (g) Au-coated hollow MN; (h) HMN reference electrode

The surface of the hollow microneedle exhibited a smooth surface and well-defined geometry with an outer diameter of ~550 µm, an inner diameter of ~300 µm, and a bevel angle of ~30°, enabling stable skin penetration and efficient ISF extraction (Fig. 2f). After Au deposition, the surface of the Au-coated HMN remained relatively smooth (Fig. 2g). Upon further coating with Ag/AgCl ink, numerous flake-like microstructures could be observed on the surface of HMN reference electrode (Fig. 2h), providing a structural foundation for the electrode stability. These results confirmed that both the working and reference electrode of the VCMGS exhibited well-defined morphology, distinct functional layers, and uniform interfacial integration, demonstrating the high stability and feasibility of the proposed fabrication strategy. The high specific surface area provided by VG and Pt nanoparticles enhanced the electrochemical performance and signal output of the GOx-coated MN electrode. Additionally, the functional layer of GOx enabled efficient and selective recognition of glucose to achieve real-time, stable, and continuous monitoring of glucose from ISF.

The structural characterization of the VG-coated sensing electrode was further carried out using Raman spectroscopy and transmission electron microscopy (TEM). As shown in Fig. 3a, the presence of a prominent G band at 1580 cm⁻¹ and a 2D band at 2686 cm⁻¹ confirmed the existence of graphene. The TEM image (Fig. 3b) revealed few-layer graphene sheets with high transparency and morphological features typical of two-dimensional materials. Then, the electrochemical performance of the VG-coated sensing electrode was evaluated in vitro using a commercial Pt electrode as the counter electrode and an Ag/AgCl electrode as the reference electrode via an electrochemical workstation. The cyclic voltammetry (CV) curves of the VG-coated sensing electrode were obtained in phosphate-buffered saline (PBS), with a scan rate of 100 mV/s ranging from −0.2 to 0.6 V to evaluate the electrode stability. The performance of the carbon electrode was also evaluated as the control group. As shown in Fig. 3c and Supplementary Information Fig. S2, both the VG-coated MN electrode and carbon electrodes exhibited highly overlapping CV curves over 10 consecutive cycles, indicating their excellent signal reproducibility. When applied with a potential of ~0.5 V, the current deviation of the VG-coated MN electrode was ~3.7%, while that of the carbon electrode was ~5.3%.

Fig. 3: In vitro performance characterization of the VCMGS.
Fig. 3: In vitro performance characterization of the VCMGS.The alternative text for this image may have been generated using AI.
Full size image

a Representative Raman spectrum of VG. b Transmission electron microscopy image of VG. c Electrochemical stability characterization of the VG-coated MN electrode via CV scanning in PBS over 10 continuous cycles. d CV curves showing the electrochemical performance of VG-coated MN electrode (1) and carbon electrode (2) via K₃[Fe(CN)₆] solution ranging from −0.2 V to 0.6 V. e Graph showing the electrochemical impedance spectroscopy (EIS) of VG electrodes with fitting results via simulation software. The orange dots represented the experimental data while the light orange lines represented the fitted curves. The corresponding equivalent circuit model were showed in the inset, where R₁ represented the solution resistance and C₁ corresponded to the double-layer capacitance. f Frequency-dependent impedance spectra of electrodes with different functional layers. The enlarged impedance profiles of Au-coated MN electrode, VG-coated MN electrode and Pt-coated MN electrode were showed in the inset, respectively. g Comparison of electrochemical impedance values of electrodes with different functional layers measured at 1000 Hz. h Amperometric signals recorded by the VCMGS in glucose concentrations varying from 0 mM to 14 mM. Black arrows indicated the sequential addition of glucose solution. The inset showed the linear calibration plot correlating amperometric response with glucose concentration. i Comparison graph showing the detection sensitivities of VG-based MN electrode versus carbon electrode in glucose concentration measurement. j Selectivity characterization of the VG-based MN sensing electrode. The response of the sensor to glucose concentration fluctuations was examined in the presence of common interfering species including uric acid, cholesterol, and ascorbic acid. k Comparative performance analysis between VG-coated MN electrodes and carbon electrode, including operational stability, charge storage capacity, detection sensitivity, and electrochemical impedance. All parameters were normalized by utilizing the performance of carbon electrode as the baseline. l Fluorescence microscopy image showing the VG-based MN stained with Rhodamine B. m Fluorescence microscopy image showing the cross-section of pig skin stained with Rhodamine B after VCMGS penetration. n Optical microscopy image of the integrated VCMGS. o Stability characterization of the fabricated HMN reference electrode. p Biocompatibility evaluation of the VG electrode. Live cells were stained with Calcein-AM, nuclei with Hoechst, and dead cells with propidium iodide

Additionally, the electrochemical performance of these electrodes in K₃[Fe(CN)₆] at varying concentrations (5–25 mM) was accessed to investigate the enhancement of VG growth on electrode sensitivity. For the CV curves of the VG-coated MN electrode, distinct oxidation and reduction peaks were observed at ~0.35 V and 0.1 V, respectively (Fig. 3d and Supplementary Information Fig. S3). The detection sensitivities were 7.68 µA/mM (R² = 0.963) for the oxidation peak and 9.67 µA/mM (R² = 0.970) for the reduction peak. The reduction peak potential of the VG electrode was utilized as a reference and the current at this potential for the carbon electrode was regarded as the reduction peak current, to compare the electrochemical performance of VG electrodes and carbon electrodes. As a result, the CV curves of carbon electrode exhibited sensitivities of 4.77 µA/mM (R² = 0.993) and 3.15 µA/mM (R² = 0.951) for the oxidation and reduction peaks, respectively. These results suggested that the growth of VG could enhance the electrode sensitivity by ~60% owing to the enlarged specific area. This result is comparable to or better than previously reported examples of glucose sensing technologies in the literature, as shown in Supplementary Information Table 1.

Subsequently, the electrochemical properties of these electrodes were characterized using electrochemical impedance spectroscopy (EIS). The solution resistance and double-layer capacitance of the electrodes were characterized in 25 mM K₃[Fe(CN)₆] solution, over an impedance frequency range of 10–10⁶ Hz. The semicircular portion of the Nyquist plot corresponded to the double-layer capacitance at the electrode–electrolyte interface, while the linear part reflected the charge-transfer kinetics. The obtained EIS data were fitted using the Zviewer to establish an equivalent circuit (Fig. 3e, inserted), showing a solution resistance (Ru = R1) of 7.01 Ω and a double-layer capacitance (Cdl = C1) of 3.99 × 10⁻⁹ C. These results suggested that the VG-coated MN electrode possessed excellent conductivity and electrochemical performance that were suitable for biomedical application.

Furthermore, the electrochemical properties of different MN electrodes were characterized via EIS measurement in PBS solution, including MN electrode, Au-coated electrode, VG-coated MN electrode and Pt-coated electrode. The impedance-frequency response curves of various electrodes (Fig. 3f) revealed that at low frequencies (<100 Hz), the impedance was significantly reduced by VG and Pt modification from ~ 400 kΩ to ~ 50 kΩ, which was advantageous for minimizing signal loss during detection and improving the signal-to-noise ratio in data acquisition. The impedance values of various electrodes at different frequencies (1 kHz, 100 Hz, and 10 Hz) were quantitatively summarized (Fig. 3f and Supplementary Information Fig. S4), with particular emphasis on comparisons at 1 kHz since most bio-signals were concentrated around this frequency. At 1 kHz, the impedance of MN electrode was ~4900 Ω (Fig. 3g), while the VG-coated electrode and Pt-coated electrode showed a markedly reduced impedance at ~350 Ω. The electrode impedance was reduced for ~12-fold due to the modification of VG or Pt nanoparticles, which increased the electroactive surface area and the active sites density of the electrode, thus facilitating electron transfer and enhancing current responsiveness. At lower frequencies (such as 10 Hz and 100 Hz), the Pt nanoparticle modification resulted in a more pronounced decrease in impedance (Supplementary Information Figs. S5–S6), suggesting that the incorporation of Pt nanoparticles could effectively improve the electrocatalytic activity and sensing capability of the electrode under low-frequency conditions.

After the integration of the VCMGS, the sensitivity and selectivity of the device were evaluated in vitro using simulated solutions. Given that the typical blood glucose levels of healthy individual ranged from 3.6 to 7.5 mM, the glucose concentration of the testing solutions was gradually varied from 2 to 14 mM to comprehensively cover this physiological range (Fig. 3h). The VCMGS was connected to the electrochemical working station, where the VG-based sensing electrode served as the working electrode, the HMN reference electrode as the counter and reference electrode with a constant potential of 0.5 V (vs. reference electrode) applied between them. Glucose concentrations in the testing solution were incrementally increased from 2 mM to 14 mM in 2 mM steps and the amperometric signals were recorded at a sampling interval of 0.1 s. During detection, glucose was catalyzed by GOx immobilized on the electrode surface, producing gluconolactone and hydrogen peroxide. The generated hydrogen peroxide underwent electrocatalytic oxidation at the surface of Pt nanoparticles with applied bias potential, releasing electrons that formed amperometric signals recorded by the electrode. In this way, the glucose concentration was linearly correlated with the signals collected by the VCMGS. The sensitivity of carbon electrode was also accessed as the control group. As shown in Fig. 3h, i, the amperometric signals collected via the carbon electrode increased from ~0.08 μA to 1.24 μA with rising glucose concentration (inset), demonstrating a sensitivity of 0.09 μA/mM with excellent linearity (R² = 0.99). For the VCMGS, the sensitivity in glucose sensing was 0.21 μA/mM (R² = 0.95), showing a ~ 2-fold improvement compared to the carbon electrode (Supplementary Information Fig. S7), suggesting that the electrochemical properties of the sensing electrode were enhanced by the VG modification in glucose detection.

Considering the potential interference from various electroactive species in the in vivo environment, the specificity of the VCMGS for glucose sensing was evaluated (Fig. 3j) based on sensitivity test. Various electroactive interfering substances were sequentially added into PBS, and the corresponding variations in amperometric signals were recorded via the VCMGS. Chronoamperometry test was conducted using an electrochemical workstation under a constant applied potential of 0.5 V, with a sampling frequency of 0.1 s for continuous signal monitoring. For each analyte addition, the final 50 s of the collected data were utilized for analysis. As shown in Fig. 3j, the addition of glucose resulted in a pronounced signal increase of ~2.1 μA, where the addition of common interfering species, such as uric acid (UA), cholesterol (CHO), and ascorbic acid (AA) induced signal changes below 0.03 μA. Upon subsequent addition of glucose, the VCMGS produced a marked sensing current (~2.1 μA) again, showing high specificity toward glucose detection in complex biological environments. For instant comparison, the amperometric responses induced by each interferent were normalized by regarding the current change produced by glucose addition as 100%. The signal variations caused by UA, CHO, and AA did not exceed 11% (Supplementary Information Fig. S8), demonstrating that the VCMGS could detect glucose stably with high accuracy under complex conditions.

Next, the electrochemical performance of the VG-coated MN electrode was normalized to that of the carbon electrode as a baseline to intuitively investigate the enhancement provided by VG. Key electrode attributes such as operational stability, impedance, charge storage capacity, and glucose sensing sensitivity were included for comparative analysis of electrode properties (Fig. 3k). In terms of charge storage capacity, the CV curves of different electrodes showed ~100% increase in electron charge storage compared to the carbon electrode, indicating the superior electron transport capability of the VG-coated MN electrode. As for sensing stability, the current fluctuation amplitude of the VG-coated MN electrode was reduced by ~30% compared to that of carbon electrode, suggesting enhanced operational robustness of the VG-based electrode. Additionally, the VG-coated MN electrode demonstrated a 2-fold increase in glucose sensing sensitivity and an 80% reduction in impedance. These findings confirmed that VG modification could significantly enhance the electrochemical performance of the electrode, including operational reliability, electrochemical impedance, charge storage capacity and sensitivity during glucose measurement.

Subsequently, the penetration capability of the VCMGS was evaluated to assess its suitability for in vivo application. In vitro characterization was conducted using freshly excised porcine skin to simulate human skin. Rhodamine B solution was first prepared, and the microneedle tips were briefly immersed to achieve uniform fluorescent staining on the VCMGS surface. The stained device was then placed perpendicular to the skin surface, and pressure was applied vertically to ensure insertion into the skin. After maintaining contact for 100 s, the VCMGS was withdrawn, and the treated skin was characterized under optical and fluorescence microscopy. Fluorescence imaging revealed red fluorescence on the VCMGS surface, indicating that a portion of the dye had been retained within the skin following insertion (Fig. 3l). Subsequently, cross-sectional slices of porcine skin (~200 μm thick) were prepared by cutting along the insertion tracks left by the VCMGS using a scalpel (Fig. 3m).

Fluorescence observed in the tissue surrounding the insertion site, along with fluorescence micrographs of cross-sections showing well-defined dye deposition closely matching the morphology of the VCMGS tips, confirmed successful skin penetration by the core-shell electrode (Fig. 3n). The maximum penetration depth of the VCMGS was ~400 μm, which was shorter than the tip length (~800 μm), likely due to the elastic resistance of the skin impeding full insertion. Notably, the penetration depth exceeded the combined thickness of the human stratum corneum (10–15 μm) and epidermis (50–100 μm), thus confirming the effective transdermal capability of the VCMGS. In addition, the mechanical strength of the HMN reference electrode contributed to an increased upper limit of force application during insertion.

The stability of the HMN reference electrode was also investigated, as the applied potential may significantly affect sensing accuracy. The performance of the prepared HMN reference electrodes was assessed by connecting both the HMN reference electrode and the standard saturated calomel electrode to the electrochemical workstation, where open-circuit potential time measurement was conducted over time to monitor the potential drift between them. The HMN exhibited an intergroup potential drift of ~0.1 mV in 50 s, demonstrating excellent potential stability during usage (Fig. 3o and Supplementary Information Fig. S9). Further, fresh chicken thighs were utilized to test the mechanical properties of VGMCS, conducting quantitative analysis of the force required for skin insertion and electrode penetration rate (Supplementary Information Fig. S10). First, fresh chicken legs were fixed to a mechanical testing platform (HLA-01, Yueqing Handpi Instruments Co., Ltd., Zhejiang, China), and the VCMGS was vertically mounted on a load-bearing force recorder (HLA-02, Yueqing Handpi Instruments Co., Ltd.), connected to a computer for data acquisition. With a speed of ~0.1 mm/s, the VCMGS was driven to vertically penetrate the chicken leg, and the force applied on the VCMGS was measured. The experiment included 3 parallel samples. As a result, as the VCMGS gradually penetrated the tissue, the force exerted progressively increased, reaching a maximum of 0.11 ± 0.03 N at an insertion depth of ~800 μm. As the VCMGS continued to advance into the tissue, the force on the electrode gradually decreases to 0. Subsequently, GOx-coated MN electrode was integrated into the VCMGS puncture chicken thigh meat for 1 s before withdrawal. The surface morphology of the electrode was characterized via SEM before and after tissue penetration (Supplementary Information Fig. S11). The experiment included 8 parallel samples. As a result, all MN electrodes were able to penetrate the chicken thighs effectively, with significant changes in the surface morphology and no shedding or defects occurring. This could be attributed to the protective effect of the GOx layer decorated on the electrode surface, coupled with the hollow microneedle shell which isolated the sensing electrode from direct contact with muscle tissue during insertion, thereby minimizing mechanical damage. These results demonstrated that the fabricated VCMGS could effectively penetrate tissue with low stress while maintaining the structural integrity of the device’s modified coating, enabling stable subcutaneous glucose monitoring in vivo.

In addition, the biocompatibility of the VG electrode for potential clinical applications was evaluated via an in vitro cell viability assay (Fig. 3p). The 3T3 fibroblast cell line was cultured in standard well plates to simulate native tissue conditions, and the VG sensing electrode was immersed in the culture medium and immobilized within the well without direct contact with the cells. In the control group, cells were cultured under identical conditions without the electrode. After 24 h of incubation, cell viability was assessed using Calcein-AM, propidium iodide, and Hoechst staining. As a result, the VG electrode group showed a cell survival rate exceeding 99%, comparable to the control group, indicating that the VG sensing electrode possessed low cytotoxicity and excellent biocompatibility.

The in vivo performance of the VCMGS was then evaluated using rat models. The study involved three healthy Sprague–Dawley (SD) rats and three diabetic SD rats. Each rat was first anesthetized using isoflurane gas, and the dorsal hair was shaved to expose bare skin. The VCMGS was then applied to the exposed area, and the sensing electrodes were inserted perpendicularly to ensure penetration through the stratum corneum and contact with ISF in the subcutaneous tissue (Fig. 4a). The VCMGS was connected to an electrochemical workstation for data recording over a 300 s period, with measurements taken every 15 min over a span of 3 h. To ensure accuracy in continuous monitoring, the average current during the final 50 s of each measurement was used as the representative response value. Parallel experiments were conducted on six rats. Based on the standard glucose concentration–current calibration curve obtained in vitro, the recorded data were converted to glucose concentrations correspondingly. At each time point, blood samples were simultaneously collected from the caudal vein of each rat and measured using a commercial glucometer to obtain reference blood glucose levels. These reference values were then utilized to validate the detection accuracy of the VCMGS system after experiment. The experiment was conducted for over 3 h per day. This study was repeated over three consecutive days (Fig. 4b). Considering the complexity of the in vivo environment, the glucose concentrations converted from data recorded by VCMGS were subsequently calibrated against reference glucose values during in the vivo experiment.

Fig. 4: In vivo performance characterization of the VCMGS in diabetic rats.
Fig. 4: In vivo performance characterization of the VCMGS in diabetic rats.The alternative text for this image may have been generated using AI.
Full size image

a Schematic diagram showing the in vivo experiment of VCMGS in rat models. The VCMGS penetrated the dorsal skin of the rat and contact with the ISF, where a potential was applied to gather amperometric signals for real-time and continuous glucose monitoring. b Timeline of the 3 h-detection experiment in 3 days for diabetic and healthy rats. N = 3 rats per group. c, d Time-dependent glucose concentration profiles in ISF of diabetic rats recorded by the VCMGS, with corresponding Clarke error grid analysis. Red dots indicated experimental data, blue asterisks represented calibration points, dark green dots denoted reference values measured by commercial devices, and the light green zone marked the normal glycemic range in healthy individuals

The glucose fluctuations of diabetic rats recorded via the VCMGS were illustrated in Fig. 4c. The red dots represented glucose concentrations measured by the VCMGS in diabetic rats, while green dots denoted reference blood glucose levels obtained using a commercial glucometer. Blue star markers indicated reference glucose concentrations used for establishing or calibrating the standard curve. The light green shaded area corresponded to the physiological glucose range in healthy individuals. In diabetic rats, a reduction in glucose levels was observed at ~90 min. For Rat 1, the glucose concentration ranged from 96.1 mg/dL to 396.6 mg/dL over the 3 day testing period. On Day 3, a rapid rise in glucose concentration was observed in Rat 1. In contrast, Rat 2 exhibited more effective glycemic regulation, with glucose concentrations ranging from 61.2 mg/dL to 208.8 mg/dL. For Rat 3, a clear decrease in glucose concentration was observed at ~90 min on each of the three days, with glucose levels ranging from 79.2 mg/dL to 327.4 mg/dL. Furthermore, Clarke error grid analysis was conducted to evaluate the accuracy of the VCMGS for glucose detection in rat models, where the average measurement errors over the 3 day testing period were 31.3% for Rat 1, 12.8% for Rat 2, and 14.6% for Rat 3 (Fig. 4d). The Clarke error grid results indicated that 63% of all data points obtained from the three diabetic rats across the three consecutive days fell within Zones A and B, corresponding to clinically acceptable deviations (error < 20%).

The glucose fluctuations recorded by the VCMGS in healthy rats were shown in Fig. 5a, where the purple dots represented glucose concentrations measured by the VCMGS in healthy rats, while green dots indicated reference blood glucose concentrations obtained using a commercial glucometer. Red star markers denoted the data points for standard curve calibration, and the light green shaded region corresponded to the physiological glucose range in healthy human. In healthy rats, an increase in blood glucose concentration was observed at the 90-min time point. For Rat 4, glucose levels ranged from 28.6 mg/dL to 206.4 mg/dL over the 3 day experiment. On Day 1 and Day 2, a marked elevation in blood glucose levels was observed. However, on Day 3, the fluctuation in glucose concentration was weak. For Rat 5, blood glucose levels ranged from 81.1 mg/dL to 212.8 mg/dL, with a gradual decrease in glucose level on Day 3. In the case of Rat 6, the blood glucose levels increased every day, with recorded values ranging from 121.6 mg/dL to 244.7 mg/dL. Clarke error grid analysis (Fig. 5b) suggested that the VCMGS exhibited average measurement errors of 12.5%, 15.3%, and 8.2% for Rats 4, 5, and 6, respectively, over the 3 day experiment. A total of 85% of the data points collected from the healthy rats across three consecutive days fell within Zones A and B, indicating clinically acceptable deviations (error < 20%). To better visualize glucose fluctuations and measurement accuracy, a heatmap was employed to summarize the dynamic changes in glucose concentrations detected by the VCMGS in different rats over time. Additionally, Clarke error grid analyses were performed across different animal groups to evaluate the sensing accuracy of the VCMGS (Fig. 5c). As a result, detection errors in healthy rats were markedly lower than those in diabetic rats, likely due to smaller reference glucose values in the healthy group, which led to reduced absolute deviation. Collectively, these findings demonstrated the capability of the VCMGS to continuously and in situ track interstitial glucose variations in live animals with high temporal resolution and clinically acceptable accuracy.

Fig. 5: In vivo performance characterization of the VCMGS in healthy rats.
Fig. 5: In vivo performance characterization of the VCMGS in healthy rats.The alternative text for this image may have been generated using AI.
Full size image

a, b Time-dependent glucose concentration profiles in healthy rats measured by the VCMGS, along with corresponding Clarke error grid analysis. Statistical comparisons were performed between VCMGS measurements and reference values obtained using commercial devices. Purple dots represented collected data points; red asterisks indicated calibration points; dark green dots denoted reference measurements by commercial glucometers; and the light green area marked the physiological glycemic range in healthy humans. c Heatmap summarizing dynamic glucose variations recorded by the VCMGS in rat over time. Clarke error grid analysis was employed to assess glucose detection accuracy across different rat groups. Error distribution across grid zones was illustrated in the plot

Subsequently, an embedded circuit was designed and fabricated to facilitate the in vivo subcutaneous glucose monitoring capability of the VCMGS. The structural and functional details of each module within the VCMGS system were depicted in Fig. 6a–c and Supplementary Information Figs. S12–S15. The VCMGS, battery, and printed circuit board (PCB) were integrated within a transparent 3D-printed enclosure to form the VCMGS system for in vivo glucose monitoring (overall dimensions: 40 × 24 × 13 mm; total weight: ~25.4 g). By integrating an analog front-end (AFE) chip with electrochemical sensing modules and a portable power supply, the VCMGS system was miniaturized, enabling in-situ, long-term subcutaneous metabolite monitoring in vivo. The VCMGS system comprised the VCMGS, embedded circuit and a smartphone-based software application. To minimize the overall footprint of the embedded circuit, component layout was carefully implemented on both the top and bottom layers of the PCB. The top layer of the PCB consisted of the main control chip, a Bluetooth Low Energy (BLE) module, and the power management unit. The bottom layer incorporated an AFE module for signal condition. Electrical interconnection between the VCMGS and the PCB was established through circuit vias. The signals collected by the VCMGS were filtered, amplified and digitalized by the AFE, which were then wirelessly transmitted to the host device via BLE communication (Fig. 6d).

Fig. 6: PCB design and functional validation of the VCMGS system.
Fig. 6: PCB design and functional validation of the VCMGS system.The alternative text for this image may have been generated using AI.
Full size image

a−c Structural schematic and optical images of the VCMGS system. The system comprised the VCMGS, an embedded circuit and a smartphone software interface. The hardware included four key modules: a microcontroller unit (MCU), Bluetooth Low Energy (BLE) module, analog front end (AFE), and the VCMGS. d Logic block diagram of the PCB circuit. e Performance evaluation of the PCB sensing circuit module, including the input-output current relationship for the glucose sensing module. f, g Comparative analysis of the detection performance between the PCB circuit and a commercial electrochemical workstation, along with corresponding normalized statistical results. h Customized application interface designed for integration with the PCB. i Photograph showing the application of the VCMGS on the forelimb of a live Bama miniature pig for continuous signal acquisition. j Time-dependent glucose concentration in subcutaneous tissue of the pig monitored by the VCMGS system. Green dots represented reference blood sampling values, and asterisks denoted calibration points. k Clarke error grid depicting the accuracy of subcutaneous glucose detection using the VCMGS system in vivo. Abbreviations: PWR, power management module; UART, Universal Asynchronous Receiver/Transmitter; SPI, Serial Peripheral Interface

Reliable signal transmission was essential for sensing systems, and thus the consistency between signal input and output on the PCB was evaluated. A closed-loop test circuit was established using precision resistors, across which a constant voltage was applied via the built-in potentiostat in the AFE module to generate a theoretical input current. Simultaneously, the system-recorded digital output was recorded as the measured signals. By sequentially replacing the precision resistors, corresponding input and output current values were recorded, and a linear regression analysis was performed to assess the experimental correlation (Fig. 6e). As the reference current increased from 0 to 3.3 μA, the output signals varied from 0 to 3.3 μA accordingly, exhibiting a high coefficient of determination (R² = 0.99994), which confirmed the system’s high accuracy in amperometric signal acquisition.

Furthermore, the performance of the VG-based MN sensing electrode was validated by connecting it to both the PCB and the electrochemical workstation (ECW) to assess the sensing capability of the PCB (Fig. 6f). As the glucose concentration increased from 2 mM to 16 mM, the current collected by the PCB ranged from 0.37 μA to 3.18 μA accordingly with the sensitivity of 0.726 μA/mM (R² = 0.962), while that recorded by the ECW varied from 0.36 μA to 3.19 μA with sensitivity of 7.59 μA/mM (R² = 0.990). To enable immediate comparison, signals acquired from both the PCB and the ECW were normalized based on the signal variation obtained between the lowest and highest glucose concentrations, serving as the reference baseline (Fig. 6g). The data obtained from the PCB was closely matched the trend measured with the ECW, indicating strong consistency between the two platforms. A customized smartphone application was also developed to interface with the PCB, enabling real-time data visualization on the host device (Fig. 6h). These results indicated a tight connection between the PCB and the VCMGS for stable signal acquisition and signal output and confirmed that this connection could be utilized for communication with a smartphone and perform continuous metabolite sensing.

Then, in vivo experiments were conducted in a healthy Bama miniature pig to assess the VCMGS’s ability for long-term subcutaneous glucose monitoring on large animals. Prior to the experiment, the pig was anesthetized using inhalational agents. The fully integrated VCMGS system was inserted through the skin of the forelimb and secured in place with a strap to ensure stable penetration through the stratum corneum and contact with the subcutaneous ISF (Fig. 6i). The VCMGS system was employed to continuously track glucose concentrations in the subcutaneous ISF of the pig’s forelimb for a duration of 1 h. At ~20 min into the experiment, the animal regained consciousness and exhibited vigorous limb movement. Considering the potential interference of motion artifacts on sensor performance, data acquisition was temporarily paused during this period. Once the animal was anesthetized again (at ~35 min into the experiment), signal recording was resumed. Concurrently, venous blood was collected from the region behind the pig’s ear within a range of ~5–8 min (typically 6 min), and a commercial glucometer was utilized to obtain reference blood glucose values for VCMGS accuracy evaluation. Considering the complex in vivo environment and potential interference from background signals, calibration was performed by aligning the sensor-derived glucose concentrations with the reference values measured by the commercial device. A two-point calibration method was employed using the glucose values at the initial and final time points of each data collection cycle as reference standards (Fig. 6j). During the experiment, the pig’s glucose levels fluctuated between 74 mg/dL and 90 mg/dL, with an overall stable trend. Notably, larger signal fluctuations were observed during the second measurement phase following animal awakening. Clarke error grid analysis (Fig. 6k) revealed a mean absolute relative error of ~15% ± 13%, with all data points falling within Zone A (error < 20%), indicating the high accuracy of the VCMGS system. These results confirmed that the developed VCMGS was conducive to real-time, in situ, and continuous tracking of subcutaneous glucose level in large animals, demonstrating its potential for long-term in vivo glucose monitoring applications. Owing to the combination of embedded circuits, the system could effectively support the operation and function of the VCMGS for real-time subcutaneous monitoring of bioanalytes.

The fabrication process developed in this work held the potential for clinical translation in several aspects. First, the VG growth process utilizes a low-pressure argon plasma atmosphere to generate radio-frequency ICP high-density, high-energy carbon radicals. These radicals knock out carbon atoms from graphite paper and are driven by the plasma onto the substrate surface30,36. This simple, safe, and rapid method was suitable for sustainable and large-scale VG growth. Consequently, the significant simplification of both the growth process and hardware equipment also facilitate future large-scale, repeatable, and low-cost growth. Furthermore, the miniaturization of the system’s overall structure minimizes wearer burden. By adopting the wearing style of fitness trackers, the device ensures long-term comfort during daily activities. Furthermore, replacing the rigid PCB with a flexible one enhances the system’s overall flexibility, significantly improving wearability and comfort. Exploring the use of more breathable, hypoallergenic medical-grade silicone or similar polymer materials for the casing and strap will improve skin breathability and reduce the risk of allergies associated with prolonged wear, further enhancing its suitability for long-term monitoring scenarios.

Conclusion

In this work, we developed a VG-coated core-shell microneedle glucose sensor (VCMGS) to track subcutaneous glucose detection for long-term in vivo glucose fluctuations tracking. The VCMGS device featured a hollow microneedle as an outer shell for skin penetration and a VG-modified microneedle electrode core for sensitive glucose detection. The core-shell structure ensured mechanical robustness to protect the sensing components and enhance sensor stability. The enhanced electrochemical performance and detection sensitivity due to VG modification allowed for reliable, real-time, in situ measurement of glucose levels from interstitial fluid, showing well linearity, high selectivity, stability over time, and consistent signal transmission. Supported by the embedded circuit and the customized smartphone interface, the VCMGS achieved accurate, real-time, and continuous glucose monitoring in both in vitro and in vivo experiments. The VCMGS presented a promising platform with potential for subcutaneous glucose measurement, supporting applications such as diabetes diagnosis and treatment, as well as the tracking of other biomarkers related to various diseases.

Materials and methods

Fabrication of VG-based MN sensing electrode (working electrode)

Stainless steel acupuncture needles (SUS316, diameter: 200 μm; Jiuzi Ridge Medical Equipment Co., Ltd., China) were used as the substrate material for fabricating the VG-based MN sensing electrode. The needles were cut into 15 mm segments using a laser marking system under the following conditions: laser wavelength of 1.06 μm, engraving speed of 600 mm/s, laser power of 20 W, and 1200 passes. The cut microneedle (MN) electrodes were immersed in an acidic detergent solution and subjected to ultrasonication for 1 h to remove surface oxide layers. The electrodes were then rinsed with anhydrous ethanol to remove residual detergent and dried in an oven at 75 °C for 1 h. Au was electrochemically deposited onto the outer surface of the cut stainless steel acupuncture needles (SUS MN) using a CHI760E electrochemical workstation (Chenhua Instruments, Shanghai, China). A gold foil electrode served as the counter electrode, an Ag/AgCl electrode as the reference electrode, and the microneedle as the working electrode. Electrodeposition was conducted in a 0.1 M gold sulfite solution for 450 s. The Au-coated MN electrodes were subsequently rinsed with anhydrous ethanol to remove residual plating solution and dried in an oven at 80 °C for 1 h to obtain the Au-modified MN. VG was grown on the surface of Au-coated MN electrode using radio frequency plasma-enhanced chemical vapor deposition (RF-PECVD). Prior to deposition, the Au-coated MN electrode was rinsed with 75% ethanol, dried, and placed into the reaction chamber. The chamber was evacuated and the substrate temperature was elevated to 1000 °C. Methan was introduced as the carbon source, with a total gas flow rate of 15 sccm, a chamber pressure of 12 Pa, and an applied RF power of 1200 W. The VG was grown for 15 min, after which the substrate was gradually cooled to room temperature to complete the deposition process. After growth, the substrate was gradually cooled to room temperature to complete the VG deposition. This surface modification process could also be alternatively performed by Shenzhen Yick Xin Technology Development Co., Ltd. Subsequently, Pt was electrochemically deposited onto the Au-coated MN electrode decorated with VG surface to enhance responsiveness toward hydrogen peroxide. Using a CHI760E electrochemical workstation, a Pt foil electrode (Shenzhen Yick Xin Technology Development Co., Ltd.) served as the counter electrode, the VG-coated MN electrode was used as the working electrode, and an Ag/AgCl electrode was employed as the reference electrode. Electrodeposition was carried out in a Pt sulfite solution for 800 s. The resulting electrode was gently rinsed with anhydrous ethanol to remove residual plating solution and air-dried under a fume hood for 1 h. The final structure was a solid microneedle electrode functionalized with Au/VG/Pt (Pt-coated MN electrode). Lastly, an outer glucose oxidase(GOx) enzymatic sensing layer was applied to enhance the responsiveness of the VG-based MN sensing electrode to glucose concentration variations. GOx solution was prepared by dissolving 200 mg of GOx (Sigma-Aldrich) in 4 mL of phosphate-buffered saline (PBS, pH 7.0, Sigma-Aldrich) to achieve a final concentration of 50 mg/mL. Separately, a bovine serum albumin (BSA) solution (80 mg/mL) was obtained by dissolving 400 mg of BSA (Sigma-Aldrich) in 5 mL of PBS. A 2.5% (v/v) glutaraldehyde (GA) solution was prepared by dissolving 250 μL of GA (Sigma-Aldrich) in 10 mL of deionized water. The GOx, BSA, and GA solutions were mixed at a volume ratio of 1:3:1 and gently agitated to obtain the enzyme immobilization solution. The Pt-coated MN electrode was subsequently dip-coated in the solution three times with uniform withdrawal, followed by overnight drying at room temperature to obtain VG-based MN sensing electrode.

Fabrication of Hollow Microneedle Reference electrodes (HMN Reference electrode)

Stainless steel syringe needles were employed as the substrate material for fabricating hollow microneedle (HMN) electrodes. Specifically, 23 G SUS304 stainless steel syringe tubes (Outer diameter ≈ 0.64 mm) were cut into 10 mm long microneedles using a YLP-F series fiber laser marking machine (Han’s Laser Technology Industry Group Co., Ltd., China). The laser cutting parameters were as follows: wavelength of 1.06 μm, engraving speed of 1000 mm/s, laser power of 25 W, and 2000 passes. The unmodified stainless steel syringe needles(hollow MNs) were immersed in an acidic detergent solution (LiGe Technology Co., Ltd., China) and ultrasonicated for 2 h to remove surface oxide layers. They were then rinsed with anhydrous ethanol to eliminate residual detergent and dried in an oven at 80 °C for 1 h. To enhance the electrochemical performance of the HMN, Au was subsequently electrodeposited using a CHI760E electrochemical workstation. A gold foil electrode served as the counter electrode, an Ag/AgCl electrode as the reference electrode, and the hollow microneedle as the working electrode. Electrodeposition was performed in a 0.1 M gold sulfite solution for 600 s. The Au-modified hollow MNs were rinsed with anhydrous ethanol to remove any residual plating solution and dried in an oven at 80 °C for 1 h to obtain the final Au-coated hollow MNs. To improve the stability of the Au-coated hollow MNs in sensing environments, the inner sidewall was insulated using polymethyl methacrylate (PMMA). A 5% (w/w) PMMA solution in ethyl acetate was slowly injected into the lumen of each MN using a syringe. The MNs were then dried in an oven at 80 °C for 1 h. Subsequently, the sensing region on the outer surface of the Au-coated hollow MN was protected by tightly wrapping it with transparent adhesive tape to form a shielding layer.

To insulate both the inner and outer surfaces of the Au-coated hollow MN, a layer of parylene was deposited. The MN samples were placed in the deposition chamber of a parylene coater (LACHIenterprise LH300, Biotime Nano Technology Co., Ltd., Suzhou, China), and 0.5 g of dichloro-[2,2]-paracyclophane dimer powder was loaded into the vaporization chamber. The chamber was evacuated, and the system was programmed as follows: the temperature was gradually increased from 25 °C to 150 °C to vaporize the dimer, which was then cracked into reactive 2,5-dichloro-p-xylylene monomers via a high-temperature quartz furnace at 650 °C. These monomers subsequently polymerized upon deposition on the MN surface as the system cooled to room temperature, forming a conformal insulating parylene layer. Next, the parylene layer on the outer surface of the Au-coated hollow MN was selectively removed to expose the sensing region by laser ablation using a laser marking system (wavelength: 1.06 μm; engraving speed: 10 mm/s; power: 10 W; 400 passes). Finally, Ag/AgCl conductive ink was uniformly coated onto the exposed sensing area and dried in an oven at 90 °C for 1 h, yielding the completed hollow MN electrodes (HMNs).

Integration of Vertical Graphene-Coated Core-Shell Microneedle Glucose Sensor (VCMGS)

After separately fabricating the working and reference electrode, they were assembled to construct a core–shell VCMGS. Specifically, the tail end of the SMN was inserted into the outlet of the HMN. Ethyl cyanoacrylate adhesive resin was then applied to the cut edge of the HMN to encapsulate the interface. The HMN was adjusted to align its opening with that of the SMN, ensuring precise coaxial alignment. Subsequently, ultraviolet light was used to cure the adhesive, thereby fixing the relative position of the SMN and HMN and completing the assembly of the core–shell structured VCMGS.

Morphological characterization of electrodes

The surface morphology of the VG-based MN sensing electrode and HMN Reference electrode after each functionalization step was characterized using a field-emission scanning electron microscope (FESEM, SUPRA60, Zeiss). Prior to imaging, a thin gold layer (~10 nm) was sputter-coated onto the electrode surfaces to enhance conductivity and minimize surface charging effects. In addition, the microstructure of the VG-coated MN electrode surface was further investigated using transmission electron microscopy (TEM, Titan3 G260-300). Raman spectra of the VG were obtained using an inVia confocal Raman microscope (Renishaw) to evaluate its structural characteristics.

In vitro electrochemical characterization of electrodes

The electrochemical performance of the fabricated electrodes was evaluated using a CHI760E electrochemical workstation in a conventional three-electrode setup. For electrochemical impedance spectroscopy, a Pt electrode was used as the counter electrode, an Ag/AgCl electrode as the reference electrode, and the fabricated MN electrode as the working electrode. Measurements were performed in a 25 mM potassium ferricyanide solution (K₃[Fe(CN)₆]) over a frequency range from 10 Hz to 10⁶ Hz.

For cyclic voltammetry (CV) measurements, a commercial Pt electrode was used as the counter electrode, an Ag/AgCl electrode as the reference electrode, and the fabricated MN electrode as the working electrode. The CV tests were conducted in a 25 mM K₃[Fe(CN)₆] with a potential range from –0.2 V to 0.6 V at a scan rate of 100 mV/s.

For chronoamperometri measurements, the HMN was used as both the counter and reference electrode, while the VG-based MN sensing electrode served as the working electrode. The measurements were conducted in PBS to evaluate the sensitivity of the electrode toward glucose detection. During the experiment, glucose concentrations were incrementally increased to 0, 2, 4, 6, 8, 10, 12, and 14 mM. A constant potential of 0.5 V was applied, and current was recorded at a sampling rate of 100 Hz for a duration of 300 s.

Selectivity characterization of VG-based MN sensing electrode

Building upon the sensitivity evaluation, the selectivity of the VG-based MN sensing electrode was assessed by immersing the electrodes in 10 mL of PBS. Interfering species were sequentially introduced at 300 s intervals in the following order: glucose (2 mM), uric acid (0.5 mM), cholesterol (5 mM), and ascorbic acid (0.1 mM). Chronoamperometric measurements were continuously performed to record the current response during each stage. For each addition, the last 50 s of the stabilized current signal were recorded and used for data analysis. The current change induced by glucose addition was used as the reference value. Current variations before and after the introduction of each interfering species were normalized against this baseline to visually illustrate the VG-based MN sensing electrode’s anti-interference capability.

Fluorescence labeling of microneedles

Rhodamine B was dissolved in PBS to prepare a staining solution at a concentration of 2 mg/mL. The tips of the VCMGS were dipped into a small volume of the staining solution to ensure uniform fluorescence labeling of the MN surface. The stained MNs were then vertically inserted into freshly excised porcine skin by applying downward pressure to ensure perpendicular penetration. After maintaining insertion for 100 s, the MNs were withdrawn and both the electrodes and treated skin were examined under optical and fluorescence microscopy. To visualize the insertion track, the porcine skin was sectioned along the insertion path using a razor blade to obtain cross-sectional slices, which were subsequently imaged under both optical and fluorescence microscopes.

Stability evaluation of HMNs

The stability of the fabricated HMN was assessed by measuring the potential difference between the HMN and a standard saturated calomel electrode (SCE) in PBS using the open circuit potential track in method. The HMN and SCE were connected to the working and reference electrode terminals, respectively, of the electrochemical workstation. The potential difference was recorded with a sampling interval of 0.1 s for a total duration of 50 s.

Biocompatibility evaluation of VG-coated MN electrodes

A 50 μL suspension of 3T3 fibroblast cells was seeded into each well and incubated for 2 h to allow cell attachment. The cells were cultured under controlled conditions at 37 °C in 5% CO₂ using Dulbecco’s Modified Eagle Medium (DMEM, 1×, Gibco) containing 4.5 g/L D-glucose, L-glutamine, and 110 mg/L sodium pyruvate. A square VG-coated MN electrode (1 cm × 1 cm) was gently placed on top of the cell-seeded well, immersed in the culture medium without exerting pressure on the cells. For the control group, cells were cultured without any added materials. After 24 h of incubation, the electrodes were removed and a live/dead cell assay was performed. Live cells were stained with Calcein-AM (green fluorescence), nuclei were stained with Hoechst dye (blue fluorescence), and dead cells were stained with propidium iodide (red fluorescence). Fluorescence images were captured using an inverted fluorescence microscope (MShot Photoelectric Technology Co., Ltd.), and the numbers of live and dead cells were quantified using ImageJ software based on the fluorescent signals.

Ethics declarations

All in vivo procedures were approved by the Institutional Animal Care and Use Committee (IACUC) of Sun Yat-sen University, in accordance with the national guidelines for the care and use of laboratory animals, animal welfare, and ethical standards (Approval No.: SYSU-IACUC-2023-001891). In vivo glucose monitoring experiments were conducted with technical support from personnel at Guangzhou Boyao Biotechnology Co., Ltd.

The experimental protocol for in vivo studies involving pigs was approved by the Institutional Animal Care and Use Committees (IACUC) of Sun Yat-sen University, Guangzhou Huateng Biotechnology Co., Ltd., Guangzhou Boaowei Biotechnology Co., Ltd. (Approval Nos.: IAEC-K-240925-01, IAEC-K-240925-02, IAEC-K-240925-03), and Guangzhou Bojin Biotechnology Co., Ltd. (Approval No.: BG-AMS-240915-CH01-Y). All animal handling procedures complied with the guidelines for the care and use of laboratory animals issued by the IACUCs of Sun Yat-sen University (Approval Nos.: SYSU-IACUC–2021–000022, SYSU-IACUC-2021-000020) and Guangdong Pharmaceutical University (Approval No.: GDPU-LAECVF-2022-008).

All experimental animals were obtained from the Laboratory Animal Center of Sun Yat-sen University (Guangzhou, Guangdong, China). Both healthy and diabetic models used male Sprague-Dawley (SD) rats under specific-pathogen-free (SPF) conditions.

In vivo glucose monitoring in rats using VCMGS

To evaluate the in vivo sensing performance of the VCMGS, glucose concentrations in the interstitial fluid (ISF) of healthy and diabetic rats were monitored. Three healthy and three diabetic male SD rats were used in this study. Following quarantine, all animals were acclimated for 7 days in the experimental environment. To induce Type 1 diabetes, three rats received a single intraperitoneal injection of streptozotocin at a dosage of 65 mg/kg. These rats were provided with standard chow and 10% sucrose water on the first day post-injection, followed by normal water thereafter. Diabetic model induction was considered successful when blood glucose levels remained stably above 11 mM, after which in vivo experiments were conducted. Considering inherent measurement errors in the meter, we selected a relatively stable measurement point early in the experiment (typically the second point) as the initial calibration reference. Since VCMGS signals may drift during experiments, we recalibrated using the reference blood glucose value at the midpoint (typically the seventh point). A calibration curve was established using these two points to correct the glucose concentrations recorded by VCMGS.

In vivo performance evaluation of the VCMGS was conducted on three healthy and three Type 1 diabetic SD rats. Prior to testing, rats were deeply anesthetized using 2% isoflurane (RWD Life Science Co., Ltd., China). The dorsal area was depilated using hair removal cream and a razor to fully expose the skin. The VCMGS were applied to the exposed skin area, allowing the sensor to penetrate the stratum corneum and access the ISF. A constant voltage of 0.5 V was applied between the work and counter electrodes, and the current signal from the VG-based MN sensing electrode was recorded for 100 s. Electrochemical detection was performed every 15 min. Simultaneously, blood samples were collected from the caudal vein and glucose levels were measured using a commercial glucometer to serve as reference data. Each experiment consisted of 12 detection cycles over a 3-h period. Each rat underwent one round of testing every 24 h, and the experiment was conducted over a period of 3 consecutive days.

For electrochemical measurements acquired by the VCMGS, the final 20 s of each 100-s current recording were extracted and averaged to obtain a representative steady-state current response. These values were then converted to glucose concentrations using the current-to-concentration calibration curve established from in vitro experiments. Given the complexity of the in vivo environment and potential background signal fluctuations, the initially converted glucose concentrations were further calibrated against the reference glucose readings obtained using a commercial glucometer. A two-point calibration method was employed, utilizing data from the second and seventh time points of each experimental session to adjust the sensor-derived concentrations. The resulting values were compared with the glucometer readings to assess the accuracy and reliability of the core–shell structured VCMGS for continuous in vivo glucose monitoring.

Design of the core–shell VCMGS circuit system

The embedded hardware system for the core–shell microneedle glucose sensor was based on the microcontroller unit (MCU, STM32F103C8T6), which interfaced with peripheral components including a power management module, analog signal conditioning module, and Bluetooth communication module. AD5941 analog front-end (AFE) chip (Analog Devices, China) was employed for signal acquisition and processing. The MCU was programmed to configure the AD5941 registers, enabling control over the potentiostat, transimpedance amplifier (TIA), and analog-to-digital converter (ADC), thereby executing the complete electrochemical measurement workflow from signal detection and conditioning to digitization. The acquired raw current data were transmitted wirelessly via a Bluetooth module (RF-BM-4044B4) to a custom-developed smartphone application. The app, designed using the Android Studio development platform, received and decoded the transmitted data and displayed the calibrated glucose concentrations in real time based on pre-established calibration parameters.

In vitro validation of circuit functionality

The signal acquisition accuracy of the hardware circuit was evaluated in vitro using a fixed resistor to simulate the equivalent impedance between electrodes in solution. A constant voltage was applied across the resistor using a potentiostat, and the resulting current (calculated based on the applied voltage and resistance value) was considered the theoretical input signal. The corresponding digital output from the monitoring system was treated as the actual measured value. The theoretical input current was gradually increased from 0 to 3.5 μA by adjusting the resistance. For each input level, the output data acquired by the system were recorded. A linear regression analysis was performed to compare the theoretical and measured values, thereby assessing the accuracy of the signal acquisition capability of the hardware system.

In vitro performance evaluation of the hardware circuit

The performance of the developed hardware circuit was evaluated by connecting the working electrode to both the custom-built system and a commercial electrochemical workstation. Glucose solutions of varying concentrations (0, 2, 4, 6, 8, 10, 12, 14, and 16 mM) were tested in a simulated physiological medium. For each concentration, current was recorded over a 300 s period, and the mean value of the final 20 s was used as the representative signal. Furthermore, the current responses at each glucose concentration were normalized using the current difference between the lowest (2 mM) and highest (16 mM) concentrations as the reference interval. This normalization enabled a more intuitive comparison of the sensing performance between the custom-developed hardware circuit and the commercial electrochemical workstation across different glucose levels.

Continuous glucose monitoring in live panamanian pigs using VCMGS

In vivo experiments were conducted using a healthy female Panamanian pig weighing ~35 kg (provided by Guangzhou Boaowei Biotechnology Co., Ltd., China) to evaluate the long-term subcutaneous glucose monitoring capability of the VCMGS system in a large animal model. An intramuscular injection of general anesthesia (Su Mian Xin II, Biofiven, Guangzhou, China) was administered via the cervical region. The animal was then secured with a respiratory mask and maintained under gas anesthesia using 2% isoflurane (Shuo Heng Biotechnology Co., Ltd., China). The front leg of the pig was shaved to expose the region for sensor placement. The VCMGS were inserted into the skin of the right forelimb to access subcutaneous ISF, and the system was secured in place using a smartwatch-style wrist strap. Continuous glucose monitoring was then carried out for 1 h. At 24 min into the experiment, the pig regained consciousness and exhibited vigorous limb movements. To minimize motion-induced artifacts, data collection was temporarily halted during this active period and resumed once the animal was re-anesthetized (approximately at the 36 min mark). The current signals collected by the VCMGS were processed by the AFE, converted into glucose concentrations based on a pre-established calibration curve, and digitized. The resulting data were wirelessly transmitted via the Bluetooth module.