Systematic Development and Characterization of Enzyme-Free, Borax-Crosslinked Microneedles for Glucose-Responsive Insulin Delivery and In Vivo Glycemic Mitigation

Abstract

Background: Conventional insulin injections cannot mimic physiological pancreatic function and often lead to dangerous hypoglycemic events that glucose-responsive systems aim to prevent. Glucose-responsive microneedles (MNs) offer a promising closed-loop alternative. We developed an enzyme-free, glucose-responsive MN patch composed of a PVA/Dextran hydrogel dynamically crosslinked with borax, and evaluated its performance, biosafety, and in vivo efficacy. Methods: MNs were fabricated from PVA/Dextran via micromolding and crosslinked with borax. The formulation was systematically optimized based on mechanical properties and glucose-responsive release kinetics. Physicochemical properties, biosafety (cytotoxicity, skin barrier recovery, boron leaching), and in vivo efficacy in a type 1 diabetic mouse model were evaluated in comparison to a subcutaneous (SC) insulin injection. Results: The optimized MNs showed robust mechanics (per-needle fracture force approximately 1.0 N) for reliable skin penetration. The system demonstrated clear glucose sensitivity, with a release flux ratio ≥1.5 between hyperglycemic (e.g., 400 mg·dL^-1) and normoglycemic (100 mg·dL^-1) conditions and exhibited excellent reversibility under alternating glucose levels. The patch was highly biocompatible, with >95% cell viability, the only transient skin barrier disruption that fully recovered within 24 h, and had low boron release from patches in vitro. In vivo, the optimized sI-MN patch demonstrated a sustained, glucose-responsive release profile, maintaining blood glucose in diabetic mice near 100 mg·dL^-1 for approximately 8 h. This pharmacokinetic profile contrasts markedly with the rapid hypoglycemic nadir and rebound hyperglycemia observed with a standard subcutaneous insulin bolus, highlighting the patch's potential for mitigating hypoglycemia. Conclusions: The enzyme-free PVA/Dextran/borax MN patch enables autonomous, glucose-responsive insulin delivery. It provides more stable and safer glycemic control than conventional injections by mitigating the risk of hypoglycemia. By mitigating the hypoglycemic risk associated with bolus injections, this systematically optimized platform represents a potential step toward a safer, patient-friendly diabetes therapy, though significant challenges in duration and dose scaling remain.

Keywords: PVA/Dextran hydrogel; borate–diol chemistry; diabetes mellitus; enzyme-free; glucose-responsive microneedles; glycemic stability; hypoglycaemia; insulin delivery; smart drug delivery; transdermal drug delivery.

Figure 1

Schematic of insulin-loaded PVA/dextran microneedle (sI-MN) fabrication with controlled post-dip borate activation. (Step 1) Polymer dope. A cold, sterile blend of PVA and dextran is prepared to total solids of 18–22% (w/w) with trehalose (2% w/v) and recombinant human insulin (5 mg·mL⁻¹). (Step 2) Mold filling. The dope is dispensed onto PDMS-negative molds and driven into the cavities under vacuum (−60 to −90 kPa; 2–3 cycles × 60–120 s) until fully filled and bubble-free; the array geometry is 14 × 14 (about 150 functional needles within the active area), with individual needles of 900 µm height and 400 µm base diameter. Molds are then cold-dried at 4 °C for 2 h. (Step 3) Backing layer application and Stage-2 drying. A backing layer (PVA 15% w/w + glycerol 0.5% w/w) is cast to cover the mold and dried at 25 °C for 3–4 h (RH 35–40%) to mass-constant. (Step 4) Demolding and trimming. Patches are released and cut to size (e.g., Ø 15 mm for product patches; scaled as needed for mouse dosing). (Step 5) Controlled borax post-dip—full-array immersion (single pass). The entire dry microneedle array is immersed once in chilled sodium tetraborate solution (0.5% w/v, pH 8.2–8.4, 4–8 °C) for 5–10 s to form dynamic borate crosslinks throughout the array, then incubated at 4–8 °C, 60–80% RH for 15–20 min. (Step 6) Cold, humid incubation and finishing. Gentle PBS mist-rinse (3–5 s), cold-drying at 4 °C for 2 h, QC check, and sealing with desiccant for storage at 2–8 °C. Mechanism: at low glucose, borate crosslinks stabilize the PVA/dextran network, retaining insulin; at high glucose, competitive glucose–borate binding loosens the network to increase insulin diffusion (reversible, enzyme-free). Symbol key: insulin (blue), glucose (yellow), borate crosslinks (green), and PVA/Dextran chain (blue wavy line).

Figure 2

Optimization of insulin-loaded PVA/Dextran microneedles (sI-MN). (a) Fracture force per needle (N·needle⁻¹) as a function of PVA/dextran ratio (75:25→95:5) at 20% (w/w) solids. For each ratio: n = 3 patches, 10 needles/patch→30 points; dots = individual needles (jittered), thin horizontal ticks within each cluster = patch means, thick dashed lines at 1.0 N (design target) and 0.8 N (minimum acceptable). Blue shading marks the target zone (≥1.0 N). (b) R_0–6h (400/100) = early-time flux ratio Flux(400 mg·dL⁻¹)/Flux(100 mg·dL⁻¹) computed from 0 to 6 h cumulative-release slopes (definition stated in Section 2.2.6). Bars show mean ± SD of the cumulative-release ratio at 0−6 h in glucose 400 vs. 100 mg·dL⁻¹ (PBS, 37 °C). Dashed line = 1.5 (pre-specified responsiveness criterion). (c) Swelling at 6 h (%) at 0/100/400 mg·dL⁻¹ glucose for each ratio (mean ± SD; n = 3), color-coded consistently (0 = gray, 100 = orange, 400 = blue). (d) Fracture force per needle versus polymer solids (18−22% w/w) at the selected ratio. Plot elements and acceptance lines as in the figure; (a) blue shading = target zone. (e) R_0−6h (400/100) versus solids (18−22%); mean ± SD with dashed line = 1.5. (f) Encapsulation efficiency (EE, %) versus insulin in dope (2.5−10 mg·mL⁻¹). Bars = mean ± SD; overlay line on the right axis = between-patch CV (%). Dashed line at 80% EE denotes the design goal (acceptance ≥ 70%); violet shading highlights the region combining high EE and low CV. (g) Insulin remaining after dip (%) following borax activation at the indicated dip times (5−15 s; single immersion). Dashed line at 95% corresponds to insulin loss ≤ 5%; green shading indicates compliance. (h) Boron leaching (µg·patch⁻¹·day⁻¹; 24 h in PBS, 37 °C) versus dip time. Dashed line at 10 µg·patch⁻¹·day⁻¹ is the safety limit; pink shading denotes non-compliant conditions (e.g., 15 s). (i) R_0–6h (400/100) versus dip time (5−15 s), mean ± SD, dashed line = 1.5. Common conditions and replication. Post-dip activation used 0.5% (w/v) sodium tetraborate (borax), pH 8.2−8.4, 4−8 °C, single pass; swelling and release were measured in PBS (pH 7.4) at 37 °C. Unless stated otherwise: n = 3 patches/condition; mechanics used 10 needles/patch; swelling/release used n = 3 independent replicates. Acceptance thresholds (dashed lines/shaded bands): fracture ≥ 1.0 N·needle⁻¹ (minimum 0.8 N), R_0−6h ≥ 1.5, EE ≥ 80% (acceptable ≥ 70%), boron leaching < 10 µg·patch⁻¹·24 h⁻¹. Abbreviations: R_0−6h, ratio of cumulative insulin released at 0−6 h (400/100 mg·dL⁻¹); EE, encapsulation efficiency; CV, coefficient of variation.

Figure 3

Device architecture, insertion proof, acute function, and barrier recovery. (a) Overview of the circular insulin-loaded PVA/dextran/borax microneedle (sI-MN) patch (14 × 14 active needles) with representative images: handheld photograph, SEM top view, and fluorescence micrograph of eosin-labeled arrays showing uniform cavity filling. (b) Higher-magnification SEM illustrating tip sharpness, base regularity, and consistent inter-needle spacing; the designed needle height is 900 ± 30 µm. (c) Acute functional delivery in STZ-diabetic mice: dose-matched sI-MN versus subcutaneous (SC) insulin, showing rapid glucose lowering within the early post-application window (mean ± SD; n = 8). (d) Ex vivo insertion on porcine skin using eosin-labeled needle tips; immediately after patch removal, optical imaging shows a fully populated grid of micro-pores mirroring the array layout (high site occupancy). (e) Surface imaging reveals discrete vertical entry channels without lateral smearing; the apparent penetration depth is approximately 500 µm, consistent with the designed needle height and remaining well below the global buckling regime inferred from mechanical tests. (f) Transepidermal water loss (TEWL) recovery kinetics normalized to each subject’s baseline: both sI-MN and blank PDB-MN show a transient rise at removal and return close to baseline by 6 h, whereas the baseline (no patch) curve remains flat (mean ± SD; n as indicated). *** indicates a statistically significant difference (p < 0.001). “ns” indicates not significant (p ≥ 0.05).

Figure 4

Glucose-responsive release and storage stability of sI-MN. (a) Swelling kinetics of the PVA/Dextran/borax network at 0, 100 and 400 mg·dL⁻¹ glucose (PBS, pH 7.4, 37 °C). Data are mean ± SD (n = 3 independent patches per condition); curves separate by 2 h with 400 > 100 > 0 mg·dL⁻¹. (b) Cumulative insulin released over 0–24 h under the same media (mean ± SD, n = 3); higher glucose accelerates release throughout the window. (c) Stepwise on–off response: the medium alternates every 30 min between 100 (tan bands) and 400 mg·dL⁻¹ (blue bands) at 37 °C. Peaks at 30, 90 and 150 min coincide with the 400 mg·dL⁻¹ phases; troughs at 60, 120 and 180 min occur during 100 mg·dL⁻¹ phases (mean ± SD, n = 3). (d) Acute glucose-lowering after storage at room temperature (RT): plasma glucose (PGL) at 0 and 60 min in STZ mice for SC native insulin and dose-matched sI-MN stored at RT for 1, 2 and 8 weeks (mean ± SD; n = 5/group). Stored patches retain robust 60 min PGL reduction comparable to fresh controls. (e) Insulin integrity during storage expressed as % of initial content by ELISA (mean ± SD, n = 3 batches): 2–8 °C (foil + desiccant) shows the least loss, RT/desiccated intermediate, and RT/ambient RH the largest decline over 12 weeks. (f) Glucose-responsiveness during storage: flux ratio R_0–6h = flux(400)/flux(100) computed from 0 to 6 h release windows (mean ± SD, n = 3). The horizontal dashed line marks the a priori acceptance threshold (R_0–6h ≥ 1.5×); refrigerated, desiccated patches remain above this criterion over 12 weeks, whereas RT/ambient RH approaches the threshold. All in vitro studies were performed in PBS (pH 7.4) at 37 °C unless indicated. “ns” indicates not significant.

Figure 5

In vivo efficacy and glucose-tolerance tests (GTTs) in STZ-diabetic mice. (a) Plasma glucose (PGL, mg·dL⁻¹) over 0–12 h for PBS, PDB-MN, dose-matched subcutaneous insulin (SC) and sI-MN (patch applied at t = 0). (b) Plasma insulin (ng·mL⁻¹) for the same groups over 0–12 h. (c) AUC_0–12h (glucose; mg·dL⁻¹·h), box-and-whisker summary of Figure (a). (d–f) IVGTT at 2 h post-treatment (0.7 g·kg⁻¹ dextrose, i.v.): (d) PGL time-course; (e) plasma insulin; (f) IVGTT AUC_0–120min (glucose; mg·dL⁻¹·min). (g–i) IPGTT at 4 h post-treatment (1.5 g·kg⁻¹ glucose, i.p.): (g) PGL time-course; (h) plasma insulin; (i) IPGTT AUC_0–120min (glucose; mg·dL⁻¹·min). sI-MN blunts glucose excursions comparably to dose-matched SC insulin, while PBS/PDB-MN show minimal effects. n = 8 mice per group; data are mean ± SD.

Figure 6

Feeding challenge and day–night performance of sI-MN. (a) Plasma glucose (PGL, mg·dL⁻¹) over 48 h without meals in sI-MN-treated mice. Microneedles were replaced at 0, 12, 24, and 36 h (blue arrows beneath the axis). Day/night periods are shown as white/gray bands. Sampling every 3 h; PGL is maintained near-euglycemia between replacements with a mild pre-replacement drift. (b) With-meal comparison (0–24 h): PGL for control (no MN) and sI-MN with daytime meals at 1, 5, and 11 h (red arrows) and MN on/off at 0 and 12 h (blue arrows). sI-MN markedly attenuates postprandial spikes and stabilizes PGL; both groups converge overnight during fasting. (c) Dual-axis readout in sI-MN (0–24 h): PGL (mg·dL⁻¹), left axis (blue), and plasma insulin (ng·mL⁻¹), right axis (red), measured in the same cohort; insulin rises with a short delay after each meal, consistent with glucose-triggered release. Microneedles were replaced at 0 and 12 h (blue arrows), and meals were provided at 1, 5, and 11 h (red arrows). Sample size: n = 6–8 mice; values are mean ± SD.