Bio-inspired suction cup microneedle loaded fusion membrane targeting nano-drug delivery system treat androgenetic alopecia

Abstract

Androgenetic alopecia (AGA) involves androgen-driven dermal papilla cell (DPC) dysfunction compounded by oxidative stress and inflammation. We engineered a leech-inspired suction-cup microneedle platform SC/MN@memPDA to deliver metformin precisely to the dermal papilla. The system integrates 300-µm short microneedles with suction-assisted negative pressure to increase insertion depth, mesoporous polydopamine nanoparticles (MPDA) for high-capacity, H₂O₂-responsive MET release, and a fusion membrane cloak derived from M2 marcophages and DPCs for inflammation targeting and dermal papilla niche homing. This system shows anti-inflammation and anti-oxidant stress damage effect both in vitro and in vivo, demonstrates sufficient delivery depth and precise delivery site, and also reveals a positive regulatory effect on hair growth. Collectively, SC/MN@memPDA provides minimally painful, follicle-targeted delivery and reactivating regenerative signaling effect, offering a promising therapeutic strategy for AGA.

Graphical Abstract

Graphic abstract of SC/MN@memPDA for AGA treatment. A bioinspired suction-cup microneedle system (SC/MN@memPDA) combinesmesoporous polydopamine nanoparticles with dual cell-membrane coatings for follicle-targeted delivery, alleviating oxidative stress andinflammation while reactivating regenerative signaling to promote hair follicle regeneration in an AGA-like model. TMB: 1,3,5-Trimethylbenzene; F127: Pluronic F127; MET: Metformin; DPCs: Dermal papilla cells; memPDA: Mesoporous polydopamine nanoparticlescoated with dual membranes (dermal papilla cell membrane and M2 macrophage membrane); SC: Suction cup; MN: Microneedle; HF: Hairfollicle

Supplementary Information

The online version contains supplementary material available at 10.1186/s12951-026-04391-5.

Introduction

Androgenetic alopecia (AGA) is the most common type of hair loss, which key pathological features include the gradual depletion of dermal papilla cells (DPCs) and progressive follicular miniaturization, ultimately resulting in hair shedding [1, 2]. The pathogenesis of AGA is strongly linked to androgen signaling. Dihydrotestosterone (DHT) binds to androgen receptors (ARs) on DPCs, leading to their progressive apoptosis, which is recognized as a key mechanism [3–5]. Therefore, precise intervention in the dermal papilla region is regarded as a critical strategy for improving therapeutic outcomes. Currently approved clinical treatments showed marked interindividual variability and adverse reactions, and they fail to achieve efficient delivery into the follicular niche [6–8]. Thus, innovative therapeutic approaches that integrate safety, efficacy, and DPC niche targeting are urgently needed [9].

Microneedles (MNs) are minimally invasive, safe, and efficient platforms for transdermal drug delivery, which transport drugs into specific skin depths and significantly improve local bioavailability [10]. However, actual insertion depth typically reaches only 50–90% of the needle length [11]. Needles longer than 600 μm are often required for reliable dermal penetration, but these stimulate nerve endings, cause pain, and reduce patient compliance [12]. Short MNs offer low- or non-painful administration and better patient acceptance. Yet, their penetration depth is limited by skin viscoelasticity, rebound effects, and critical insertion resistance. Enhancing insertion efficiency while maintaining comfort remains a major challenge [13]. Prior studies have reported that moderate sub-atmospheric pressure can induce local tissue deformation and transient barrier perturbation, including changes in apparent skin thickness and microstructure, which may enlarge aqueous pathways and concentrate interfacial stress; collectively, these effects have been suggested to facilitate microneedle–skin contact, insertion, retention, and prolong micropore opening [14, 15]. In addition, the pressure gradient between the vacuum chamber and subcutaneous tissue facilitates drug diffusion [16]. Inspiration from nature, leeches achieve firm adhesion through suction-cup structures, reducing stratum corneum resistance and enhancing penetration depth under mild negative pressure, while driving substance transport into deeper tissues [17, 18]. Thus, integrating short MNs with suction-cup structures may improve transdermal delivery efficiency while retaining minimal pain characteristics.

Recent evidence indicates that AGA progression involves persistent oxidative stress and inflammation in the follicular microenvironment [19–21]. These processes accelerate DPC dysfunction and further suppress follicular regeneration. Targeting oxidative stress and inflammation has therefore emerged as a promising therapeutic approach. Metformin (MET), in addition to its glucose-lowering effect, activates the AMPK pathway to improve cellular metabolism and has antioxidant, anti-inflammatory, and microcirculatory regulatory effects [22]. Recent clinical studies showed that low-dose oral metformin alleviates inflammation and promotes hair regrowth in cicatricial alopecia, suggesting therapeutic potential in hair disorders [23]. However, oral administration carries risks of systemic side effects. Accordingly, a delivery platform capable of precisely targeting the dermal papilla region is required to maximize the local therapeutic potential of MET.

Polydopamine (PDA) is highly biocompatible and biodegradable, and can efficiently load drugs and release them in response to environmental stimuli without additional modification [24, 25]. Compared to non-mesoporous type, mesoporous PDA (MPDA) has a larger surface area and pore volume, resulting in higher drug-loading capacity and controlled release through multiple molecular interactions. Its nanoscale size also facilitates follicular penetration and prolongs local retention [26]. Recently, membrane coating targeting system has emerged as an effective strategy to enhance nanoparticles delivery [27]. M2 macrophage membranes retain chemokine receptors and immunoregulatory molecules, allowing them homing to inflammatory sites [28, 29], while DPC membranes provide homotypic recognition, promoting carrier adhesion and retention at the dermal papilla niche [30]. A fusion-membrane-coated MPDA system may therefore achieve synergistic follicular niche targeting and serve as an effective and precise delivery platform for AGA therapy.

In this study, we selected MET as the therapeutic agent, designed a leech-inspired “suction-cup microneedle” (SC-MN) transdermal platform. This system uses 300 μm short MNs, where suction-assisted negative pressure lowers the insertion threshold and enables deeper penetration. The fusion-membrane-coated MPDA system loaded inside the SC-MN efficiently encapsulate and release MET in a sustained manner with both “inflammation targeting” and “dermal papilla homing” effect. This platform enables precise delivery of MET to the dermal papilla region, significantly increasing local drug delivery efficiency, and activating follicular regeneration pathways. Collectively, this strategy provides a safe, effective, and follicle-targeted therapeutic approach for AGA .

Results and discussion

Metformin selection and SC/MN@memPDA characterization

Firstly, a drug-agnostic screening was performed to identify the drug. Genes associated with “oxidative stress” and “inflammation” were retrieved from GeneCards [31]. Their intersection was ranked based on standardized relevance scores and geometric means (Fig. 1A–B). Subsequently, gene–drug interactions were analyzed using DGIdb and drugs were aggregated at the network level to prioritize candidates that regulate the “inflammation–oxidative stress” consensus network (Fig. 1C) [32]. Among the retrieved drugs, metformin (MET) exhibited the highest multi-gene coverage, indicating its potential to modulate multiple nodes within the inflammation–oxidative stress network. In contrast, the top 14 candidates were mainly broad-spectrum cytotoxins, strong immunosuppressants, biologics, or highly toxic kinase inhibitors, which are limited by narrow therapeutic windows, restricted administration routes, high cost, and poor compliance. Moreover, many of these agents displayed non-specific or bidirectional effects on oxidative stress and inflammation, which are unfavorable for long-term follicular homeostasis. In comparison, MET demonstrated significant hits across multiple mechanistic modules—including AMPK–autophagy, NRF2 antioxidant signaling, NF-κB inhibition, and PI3K–PTEN pathways—consistent with disease-related pathological processes (antioxidation, anti-inflammation, metabolic remodeling). Together with its proven safety, accessibility, and feasibility for oral or topical use, MET was prioritized as the candidate drug for this study.

Fig. 1.

Preparation and characterization of SC/MN@memPDA. (A) Venn diagram of protein-coding genes associated with “oxidative stress” and “inflammation” from GeneCards. (B) Gene relevance ranking based on normalized relevance scores and geometric means. (C) Multi-node drug ranking obtained from DGIdb gene–drug interaction analysis, aggregated at the drug level; (D) Schematic illustration of the fusion membrane constructed from DPC membrane and M2 macrophage membrane; (E) Schematic diagram of memPDA nanoparticle preparation; (F) Fluorescence imaging of DPC membranes (DiO, green) and M2 macrophage membranes (DiI, red). Scale bar: 20 μm; (G) Western blot analysis; (H) TEM images of MPDA nanoparticles before (left) and after dual-membrane coating (right). Scale bars: 200 nm; (I) EDS elemental mapping of C, N, O, and Zn in MPDA. Scale bars: 100 nm; (J) Comparison of drug loading (DL%) and encapsulation efficiency (EE%) between PDA and MPDA; (K) Schematic illustration of SC/MN@memPDA fabrication; (L) Stereomicroscope image of the microneedle array. Scale bar: 500 μm; (M) SEM images of the SC/MN@memPDA patch. Scale bar: 500 μm; (N) Confocal fluorescence images of SC/MN@memPDA showing colocalization of RhB-labeled MPDA with FITC-labeled HA microneedle matrix. Scale bar: 200 μm. Results were presented as the mean ± SD, one way ANOVA was performed for multiple comparison tests, while Student’s t-test was performed for two independent groups (ns, no significance, *P < 0.05, **P < 0.01, ***P < 0.001)

Next, mesoporous polydopamine nanoparticles coated with fusion cell membranes (memPDA) was prepared (Fig. 1D–E). Under alkaline conditions, F127 micelles served as soft templates for dopamine polymerization, and subsequent template removal produced ordered or disordered mesoporous structures. Addition of TMB allowed further tuning of micelle size, optimizing surface area and pore volume. MPDA was then loaded with MET and coordinated with Zn²⁺. DPC membranes endowed homotypic recognition and adhesion to the dermal papilla region [33, 34], while M2 membranes conferred inflammation targeting and immunomodulatory functions [35–38]. To enhance targeting and niche homing capabilities, we developed a fusion-membrane coating system (memPDA). Fluorescent tracing demonstrated successful fusion of M2 macrophages and DPC membranes (Fig. 1F). Western blot analysis showed that the M2 marker F4/80 and the DPC-specific protein NCAM were both retained, along with ATPase (membrane enzyme) and GAPDH (internal control), indicating preservation of key proteins in the fusion membrane (Fig. 1G).

Further, through fusion and extrusion, memPDA has been prepared. TEM images revealed uniform quasi-spherical particles with regular mesoporous structures (Fig. 1H, left). After extrusion, a uniform membrane cloak layer was observed around the nanoparticles (white arrows, Fig. 1H, right), in clear contrast to bare MPDA, confirming successful fusion membrane encapsulation. Energy-dispersive X-ray spectroscopy (EDS) mapping verified the homogeneous distribution of C, N, O, and Zn elements, which confirmed Zn²⁺ loading (Fig. 1I). Compared to non-mesoporous PDA, mesoporous structure exhibited a larger surface area and pore volume, with approximately two-fold higher drug-loading (DL%) and encapsulation efficiency (EE%) (Fig. 1J). These structural advantages significantly increased overall drug loading, reduced burst release, and enabled more sustained release, establishing MPDA as an efficient and controllable delivery platform [39–41]. In vitro release experiment further demonstrated that MET-loaded MPDA showed concentration-dependent accelerated release under H₂O₂ stimulation, indicating excellent oxidative stress responsiveness and controlled-release properties (Fig. S1A).

To improve transdermal delivery, we further designed a leech-inspired “suction-cup microneedle” (SC/MN@memPDA) system. The vacuum-fabricated SC patch was a 1 × 1 cm silicone sheet with eight evenly distributed suction cups (3 mm diameter each), aligned with HA-based MN arrays of the same diameter. Each MN measured 300 μm in height with a 150 μm basal diameter (Fig. 1K–M). Mechanical testing confirmed strong negative-pressure adhesion of the suction cups to glass surfaces (Fig. S1B). MPDA-loaded MNs exhibited a stable crosslinked network, improving tensile, compressive, and bending resistance. Mechanical testing demonstrated that SC/MN@memPDA penetration force far exceeded the skin threshold of 0.058 N/needle, ensuring reliable insertion (Fig. S1C) [13, 42]. Confocal imaging showed uniform overlap of RhB-MPDA and FITC-HA signals within the MN matrix (Fig. 1N), confirming successful and homogeneous incorporation of memPDA into the MNs.

To improve dosing transparency and reproducibility, the absolute MET content and the post-application backing residual were quantified by HPLC. Each MN subunit (3 mm diameter) contained 0.1744 ± 0.0208 mg MET (n = 6). After application, the residual MET recovered from the retrieved backing layer was 0.0362 ± 0.0059 mg per MN subunit (n = 6). Therefore, the estimated MET dose delivered per MN subunit during a single application was 0.1382 mg (calculated as total MN content minus backing residual). For in vivo treatment, one SC/MN patch (1 × 1 cm) carrying eight MN subunits (3 mm diameter each; total active MN contact area ~ 56.5 mm²) was applied per session, corresponding to an estimated delivered MET dose of 1.1056 mg per session. Patches were applied once daily for 21 days, yielding a cumulative estimated delivered MET dose of 23.2176 mg over the treatment course.

Collectively, these results confirm that SC/MN@memPDA was successfully prepared, with high drug-loading and efficient transdermal delivery capacity, as well as H₂O₂ responsive controlled release ability.

MN@memPDA possessed anti-oxidant and anti-inflammation activity

To further evaluate the antioxidant and anti-inflammatory properties of MN@memPDA, we established an AGA model in DPCs using DHT stimulation [43]. Excessive reactive oxygen species (ROS) are hallmarks of oxidative stress, leading to follicular dysfunction and contributing to AGA progression. ROS detection with the DCFH-DA probe showed a marked increase in intracellular ROS in the ROSUP group, whereas the control group maintained the lowest levels. Blank MNs had little effect on ROS, and blank memPDA partially reduced ROS. In contrast, free MET and MN@memPDA showed the strongest inhibitory effects, restoring ROS levels close to those in the control group (Fig. 2A–B). Flow cytometry further confirmed these findings, demonstrating that both MET and MN@memPDA significantly reduced intracellular ROS, with MN@memPDA showing the greatest effect (Fig. 2C). These results indicate that MET-loaded MPDA particles provide potent antioxidant activity to counteract DHT-induced oxidative stress in DPCs. Notably, previous studies have demonstrated that polydopamine itself possesses intrinsic ROS-scavenging capability, contributing to its antioxidative potential [44–46].

Fig. 2.

MN@memPDA Possessed Anti-oxidant and Anti-inflammation Activity in vitro. (A) Detection of intracellular ROS levels in DPCs by DCFH-DA probe. Scale bar: 200 μm; (B) Quantitative analysis of ROS levels under different treatments; (C) Flow cytometry assessment of intracellular ROS levels; (D) Detection of mitochondrial membrane potential (ΔΨm) in DPCs under different treatments. Scale bar: 50 μm; (E) Quantitative analysis of ΔΨm levels; (F) qPCR analysis of oxidative stress–related genes (Nrf2, HO-1, FTH1, NQO1, SOD2, and GCLM) in DPCs under different treatments; (G) Immunofluorescence staining of Ki67 in DPCs under different treatments. Scale bar: 50 μm; (H) Quantitative analysis of Ki67-positive cells showing proliferative activity. ROS, reactive oxygen species; Ki67, marker of proliferation Ki-67; Nrf2, nuclear factor erythroid 2–related factor 2; HO-1, heme oxygenase-1; FTH1, ferritin heavy chain 1; NQO1, NAD(P)H quinone dehydrogenase 1; SOD2, superoxide dismutase 2; GCLM, glutamate–cysteine ligase modifier subunit. Data were presented as the mean ± SD, one-way ANOVA was performed for multiple comparison tests, and Student’s t-test was performed for two independent groups (*P < 0.05, **P < 0.01, ***P < 0.001)

Mitochondria are critical targets of oxidative stress, therefore mitochondrial membrane potential (ΔΨm) was assessed to evaluate mitochondrial protection. High ΔΨm produces red fluorescence aggregates, while loss of ΔΨm results in green monomers. As the results showed, red fluorescence predominated in control group, indicating intact ΔΨm. Green fluorescence increased markedly in CCCP group, confirming severe ΔΨm loss. MNs showed a similar pattern while memPDA partially restored ΔΨm. Notably free MET and MN@memPDA significantly preserved ΔΨm stability, MN@memPDA restored ΔΨm to near-control levels (Fig. 2D–E). Additionally, the expression of oxidative stress-related genes was detected (Nrf2, HO-1, FTH1, NQO1, SOD2, GCLM). DHT stimulation markedly upregulated these genes, while MN@memPDA treatment suppressed this stress-induced upregulation (Fig. 2F), indicating that exogenous antioxidant protection reduced oxidative burden and alleviated abnormal gene expression [47]. Thus, these results showed significantly anti-oxidant effect of MN@memPDA.

Since AGA progression is also associated with inflammatory infiltration around hair follicles, we further evaluated the anti-inflammatory effects of MN@memPDA. Live/dead and Ki67 staining revealed a higher proportion of dead cells and reduced Ki67-positive cells compared with controls. MN@memPDA significantly alleviated cell death and restored proliferation, whereas MET treatment was less effective, and memPDA or MNs alone showed only limited improvement (Fig. 2G–H; Fig. S2A-B). qPCR analysis showed that DHT strongly upregulated TNF-α, IL-1β, and IL-6 expression, which was markedly suppressed by MN@memPDA to near-control levels (Fig. S2C).

Taken together, MN@memPDA effectively scavenged intracellular ROS, preserved mitochondrial function, and alleviated inflammatory responses in vitro. These findings are consistent with our earlier drug-screening results targeting oxidative stress and inflammatory pathways, providing functional validation of MET as a therapeutic agent (Fig. 1A-C).

SC/MN@memPDA mitigates DHT-induced suppression of hair growth

To evaluate the therapeutic effect of SC/MN@memPDA, we established a DHT-induced alopecia mouse model (Fig. 3A) [48]. Representative in vivo application photographs are provided in Fig. S5. As delivery efficiency increased with application time and reached a plateau (≥ 90%) at 120 s (Fig. S6), we standardized the in vivo protocol by applying a gentle thumb preload for ~ 10 s to ensure conformal contact, followed by maintaining the patch in place for a total of 2 min.

Fig. 3.

SC/MN@memPDA Mitigates DHT-Induced Suppression of Hair Growth. (A) Schematic illustration of experimental design; (B) Overall appearance of dorsal skin on DAY 0, DAY14 and DAY 21 of treatment; (C) Representative hair coverage maps of dorsal skin at Day 0 (pink) and Day 21 (black) in different treatment groups; (D) Quantitative analysis of hair coverage percentage at Day 21; (E) Representative fluorescence images of dorsal skin sections showing in vivo distribution of membrane-coated MPDA (memPDA) versus uncoated MPDA; (F) Quantitative analysis of fluorescence intensity in the dermal papilla region; (G) Representative H&E staining of dorsal skin sections showing hair follicle morphology across groups. Scale bar: 100 μm; (H) Quantitative analysis of dermal papilla (DP) size (left) and hair follicle (HF) length (right) (n = 6). H&E, hematoxylin and eosin. Data were presented as mean ± SDs. One-way ANOVA was used for multiple comparison tests, and Student’s t-test was used for two independent groups (*P < 0.05, **P < 0.01, ***P < 0.001)

As shown in Fig. 3B, DHT treatment delayed the transition into the anagen phase compared with controls, confirming successful model establishment. Topical application of memPDA alone showed modest pro-growth activity, likely due to partial skin penetration of the nanoparticles. Simple SC/MNs demonstrated a certain degree of therapeutic effect, which may be attributed to microinjury from microneedle puncture [49]. By comparison, MN@memPDA without suction cups showed limited efficacy, whereas SC/MN@PDA achieved the better therapeutic outcome. This result suggested that restricted transdermal efficiency of 300 μm length short needles. Based on prior sub-atmospheric skin studies, mild suction-induced conformal sealing and local tissue deformation may facilitate microneedle–skin contact and insertion/retention, thereby contributing to improved transdermal delivery (Fig. 3C–D).

Moreover, coating MPDA with the fusion membrane targeting system (SC/MN@memPDA) further improved hair regrowth compared with non-targeted carriers. Accordingly, to assess the targeting and homing ability of the system, uptake tracing experiment was performed. MPDA and memPDA were covalently labeled with Alexa Fluor 488 (AF488) and incubated with DPCs at equal fluorescence doses. Confocal imaging showed that only a few green particles were observed in the cytoplasm of the MPDA group, whereas the memPDA group displayed much denser fluorescence signals around the cytoplasm and perinuclear regions, indicating markedly enhanced uptake (Fig. S3A-B). To further visualize the in vivo trageting advantage, exogenously labeled nanoparticles were tracked. Fluorescence imaging revealed that fusion membrane-coated MPDA preferentially accumulated in the dermal papilla region (Fig. 3E–F). These findings are consistent with the preserved homotypic adhesion molecules in the DPC membrane, supporting targeted delivery to the dermal papilla niche. Consistently, using the same AF488-labeling approach, AF488-memPDA showed markedly higher cell-associated fluorescence in DPCs than in primary dermal fibroblasts or HaCaT keratinocytes under identical incubation and imaging conditions (Fig. S8), supporting preferential uptake by dermal papilla cells.

Further H&E staining was used to assess follicular morphology, with quantitative analysis of dermal papilla volume and follicle length. In the DHT group, follicles were degenerated and dermal papilla regions were atrophic, consistent with the telogen phase. In contrast, all treatment groups showed intact follicular structures with well-preserved dermal papilla morphology, and both follicle length and papilla area were significantly larger than those in the DHT group (Fig. 3G–H).

Overall, suction-cup negative pressure and fusion membrane targeting system acted synergistically, achieved dermal papilla niche specific localization, constituting an efficient transdermal delivery strategy for AGA therapy.

Membrane coating combined with suction cup synergism significantly enhances the skin penetration depth of microneedles

To further evaluate delivery depth of the system, we visualized the drug penetration depth by performing drug-tracing studies in serial sections of mouse skin. The system was prepared with a DiO-labeled membrane and Rhodamine B-labeled MPDA, and three groups were compared: MN@PDA, MN@memPDA, and SC/MN@memPDA (Fig. 4A). In the MN@PDA group, fluorescence was largely restricted to ~ 150 μm and rapidly attenuated with depth, indicating limited penetration efficiency. Incorporation of the fusion membrane system (MN@memPDA) enabled detectable but weaker fluorescence at 200 μm, suggesting that membrane modification partially improved penetration and facilitated transport across the stratum corneum. Significantly, SC/MN@memPDA achieved the greatest penetration, with strong fluorescence signals still observed at 300 μm (Fig. 4B).

Fig. 4.

Evaluation of transdermal penetration depth of short microneedles with cell membrane modification and suction-cup design. (A) Schematic illustration of experimental design; (B) Representative confocal fluorescence images of skin sections at different depths following administration of MN@PDA, MN@memPDA, or SC/MN@memPDA. Scale bars: 100 μm; (C) Quantitative analysis of fluorescence intensity across skin layers and corresponding AUC (Area Under the Curve) values for transdermal delivery efficiency. Data were presented as mean ± SDs. One-way ANOVA was used for multiple comparison tests, and Student’s t-test was used for two independent groups (*P < 0.05, **P < 0.01, ***P < 0.001)

Penetration efficiency was subsequently quantified by measuring mean fluorescence intensity across skin layers and calculating the overall transdermal area under the curve (AUC). MN@PDA exhibited the lowest AUC, reflecting superficial drug retention. MN@memPDA displayed a significant increase compared with MN@PDA, confirming that membrane modification enhanced overall drug delivery. SC/MN@memPDA yielded the highest AUC, surpassing MN@memPDA, which demonstrated that the suction-cup structure markedly improved penetration efficiency through negative pressure (Fig. 4C).

To better distinguish initial deposition from time-dependent redistribution, we additionally performed cryosectioning immediately after patch removal (T0). At T0, SC/MN@memPDA already exhibited a deeper initial deposition profile and higher AUC across 0–300 μm (Fig. S11), whereas the 72 h profile (Fig. 4) reflects the combined effects of initial entry, subsequent diffusion, and tissue retention.

In summary, membrane coating coordinated suction cup enhanced tissue affinity and translocation, resulting in deeper and more uniform drug distribution. This strategy overcame the penetration limitations of short MNs and offers a promising approach for painless and efficient transdermal drug delivery.

Suction-cup engagement generates a mild sub-atmospheric pressure that improves conformal sealing and interface fixation, which may facilitate microneedle insertion/retention by locally deforming the skin. While we did not directly quantify insertion force under suction in this study, we observed stable attachment during application (Fig. S5) and a transient, reversible barrier perturbation evidenced by TEWL dynamics (Fig. S7). Prior sub-atmospheric skin studies have further shown that local sub-atmospheric pressure can induce measurable skin thinning and appendage opening, providing mechanistic context for the suction-assisted delivery concept.

SC/MN@memPDA attenuates DHT-induced oxidative stress and inflammation which restore hair growth signals

We next examined the effects of SC/MN@memPDA on follicular proliferative activity and key signaling pathways in vivo using immunofluorescence and immunohistochemistry. As the central pathway in AGA, AR expression directly reflects androgen-mediated suppression of hair growth. Immunohistochemistry showed AR-positive signals as brown staining. Compared with control group, AR staining intensity and positive area were markedly increased in the DHT group, indicating abnormal AR pathway activation and heightened androgen sensitivity. In contrast, SC/MN@memPDA treatment reduced AR staining intensity and positive range to near-control levels (Fig. 5 A–B). These results suggest that SC/MN@memPDA suppresses aberrant AR signaling, thereby alleviating DHT-induced inhibition and creating a favorable environment for follicular regeneration.

Fig. 5.

Immunohistochemical, immunofluorescence, and molecular analyses of SC/MN@memPDA in DHT-induced AGA-like mice. (A) Representative IHC staining of AR in dorsal skin sections. Scale bars: 100 μm; (B) Quantification of AR-positive area; (C) Representative immunofluorescence staining of Ki67, TGF-β1, and β-catenin in hair follicles. Scale bars: 200 μm; (D) Quantification of Ki67, TGF-β1, and nuclear β-catenin signals; (E-F) Statistical analysis of antioxidant gene expression levels; (G) Relative mRNA expression of hair growth–related genes (DKK1, TGF-β1, Wnt5a, Wnt3a, Wnt10b, Lgr5, BMP2/4, GLI1/2). AR, androgen receptor; IHC, immunohistochemistry; DKK1, dickkopf-related protein 1; Wnt, wingless-related integration site; Lgr5, leucine-rich repeat-containing G-protein–coupled receptor 5; BMP, bone morphogenetic protein; GLI, glioma-associated oncogene. Data were presented as mean ± SDs. One-way ANOVA was used for multiple comparison tests, and Student’s t-test was used for two independent groups (*P < 0.05, **P < 0.01, ***P < 0.001)

Immunofluorescence was then used to assess follicular proliferation and related molecular changes (Fig. 5C–D). In the control group, Ki67-positive signals were distributed in the bulge, outer root sheath, and matrix nuclei, indicating active proliferation within anagen phase [50]. However, under DHT treatment, Ki67 signals were markedly reduced, consistent with retention of telogen. Intervention groups showed varying levels of recovery. memPDA yielded only slight improvement, MN@memPDA and SC/MN@PDA enhanced Ki67 expression, and SC/MN@memPDA exhibited the strongest effect, with fluorescence intensity approaching control levels. To further evaluate inflammatory status, TGF-β1 expression was examined. DHT markedly elevated TGF-β1, reflecting a catagen-promoting and pro-fibrotic/inflammatory milieu, whereas all treatments reduced TGF-β1, with SC/MN@memPDA showing the most pronounced effect [51]. Finally, β-catenin expression was analyzed as a marker of Wnt pathway activity. In the control group, β-catenin was clearly detectable, while nuclear β-catenin was strongly reduced in DHT group, indicating Wnt suppression by DHT [52]. After intervention, especially in the SC/MN@memPDA group, nuclear positivity was restored to near-control levels, demonstrating effective reactivation of Wnt/β-catenin signaling and promoting entry of anagen. Together, these findings indicate that SC/MN@memPDA antagonizes aberrant AR activation, alleviates perifollicular inflammation, and reactivates Wnt/β-catenin signaling to restore hair growth.

To systematically evaluate oxidative stress and inflammatory regulation in vivo, we measured the expression of key genes in skin tissues. DHT reduced expression of Nrf2, HO-1, FTH1, NQO1, SOD2, and GCLM, reflecting impaired antioxidant defenses [47]. SC/MN@memPDA significantly restored these genes to near-control levels, whereas SC/MN@PDA and MN@memPDA had moderate effects, and SC/MN or memPDA alone were less effective (Fig. 5E–F). In vitro, short-term DHT stimulation induced ROS accumulation and compensatory upregulation of the same genes, which were reversed by MN@memPDA treatment [53]. These contrasting patterns indicate that chronic DHT exposure depletes the Nrf2 axis, while SC/MN@memPDA lowers ROS burden and reactivates antioxidant defenses [19, 54–56]. Thus, the system effectively mitigates oxidative stress and enhances antioxidant capacity.

Analysis of inflammatory cytokines showed marked upregulation of TNF-α, IL-1β, and IL-6 in DHT-treated tissues. SC/MN@memPDA strongly reduced these cytokines, more effectively than SC/MN@PDA or MN@memPDA, and nearly normalized expression to control levels. SC/MN and memPDA showed weaker effects (Fig. S3C). These results are consistent with the in vitro inflammation model, confirming that SC/MN@memPDA alleviates tissue inflammatory burden.

DHT treatment also disrupted key growth-related pathways. As shown in Fig. 5G, expression of DKK1, TGF-β1, and non-canonical Wnt5a was upregulated, while Wnt3a, Wnt10b, Lgr5, BMP2/4, and GLI1/2 were downregulated [57–60]. These changes suggest that DHT activates inhibitory and non-canonical Wnt signaling while suppressing Wnt/β-catenin and Hedgehog pathways, leading to arrest and miniaturization of hair follicles while SC/MN@memPDA reversed these abnormalities. DKK1, TGF-β1, and Wnt5a decreased, while Wnt3a, Wnt10b, Lgr5, BMP2/4, and GLI1/2 were restored to near-control levels. These results indicate coordinated pathway regulation that reactivates hair follicle growth signals.

To clarify the respective contributions of Zn²⁺ and metformin (MET) to the observed anti-inflammatory/anti-oxidative effects, we added component-separated controls (Zn²⁺ only, MET only, and Zn²⁺+MET). In RAW264.7 cells stimulated with LPS, the Zn²⁺+MET combination showed the greatest reduction of IL-6 and TNF-α levels (Fig. S9A). In DPCs, Zn²⁺+MET also produced the lowest ROS signal measured by DCFH-DA flow cytometry and the highest Ki67⁺ fraction (Figure S9B–D). Consistently in vivo, compared with Zn²⁺ alone, Zn²⁺+MET yielded more pronounced improvement in hair regrowth and follicular histology, accompanied by more favorable AR-associated readouts (Fig. S10).

To benchmark MET against an established AGA therapeutic mechanism, we performed a DHT-stimulated 5AR-luc reporter assay using finasteride as a reference drug and included a finasteride + MET group. MET significantly attenuated the DHT-induced reporter activation, and the finasteride + MET combination further enhanced the inhibitory effect compared with either single agent (Fig. S12).

Altogether, these results demonstrate that DHT-induced AGA is characterized by suppressed antioxidant pathways, enhanced oxidative stress and inflammation, and downregulated regenerative signaling axes, and SC/MN@memPDA effectively reversed these changes, supporting its role in promoting follicular regeneration through microenvironmental modulation and reactivation of key pathways.

Transcriptomic analysis shows immune reprogramming and restoration of hair-growth programs after SC/MN@memPDA treatment

To further explore the effects of SC/MN@memPDA on the skin local microenvironment, transcriptome sequencing of mouse skin tissue was performed. RNA sequencing was performed on dorsal skin tissues collected at the same time point from two groups: the DHT-induced AGA model group (DHT) and the SC/MN@memPDA-treated group (n = 3 biological replicates per group; one replicate corresponds to tissue from one mouse). Differentially expressed genes (DEGs) were determined for the comparison SC/MN@memPDA versus DHT.

A total of 5094 differentially expressed genes were identified between the SC/MN@memPDA and DHT groups, including 2993 genes that were downregulated in the SC/MN@memPDA group (Fig. 6A). Hair keratin–related genes such as Krt71, Krt72, and Krt73 were markedly upregulated, indicating that treatment promoted hair shaft keratinization and trichohyalin body formation, which may stabilize follicular structure and support normal hair production [61, 62]. Similarly, elevated Crnn and Padi3 expression suggested enhanced keratin crosslinking and restoration of epidermal barrier function [63, 64]. In contrast, genes related to perifollicular signaling and metabolism, including Pla2r1, Rarres1, and Pctp, were significantly downregulated, suggesting reduced lipid metabolism abnormalities and attenuation of inflammation-related signaling (Fig. 6B) [21, 65].

Fig. 6.

Transcriptomic profiling and pathway analyses of SC/MN@memPDA treatment in DHT-induced AGA-like mice. (A) Heatmap showing DEGs between SC/MN@memPDA and DHT groups; (B) Volcano plot of DEGs; (C) Gene Ontology (GO) enrichment analysis of DEGs; (D) Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis; (E) KEGG pathway clustering; (F) Protein–protein interaction (PPI) network of DEGs; (G) Gene Set Enrichment Analysis (GSEA) of epidermal development and extracellular matrix–related terms; (H) GSEA plots of NF-κB and Wnt signaling. DEGs, differentially expressed genes; GO, Gene Ontology; KEGG, Kyoto Encyclopedia of Genes and Genomes; PPI, protein–protein interaction; GSEA, Gene Set Enrichment Analysis

Protein–protein interaction analysis highlighted Ctnnb1 (β-catenin) as the top hub, followed by Tnf, IL6, Stat3, Egfr, Egf, and IL1b (Fig. 6F). These hubs linked the inflammatory axis (TNF/IL-6/IL-1β–STAT3) with regenerative pathways (Wnt/β-catenin, EGF–EGFR) [66, 67]. The network also enriched genes involved in regeneration and macrophage polarization (Wnt5a, Wnt7a/7b, Lef1, Sox9, Pparg) as well as ECM and adhesion remodeling molecules (Col1a1/1a2/3a1/5a1/5a2, Cdh2, Cav1, Sparc), reflecting a coordinated interaction between inflammation, stroma, and regeneration.

Gene Ontology enrichment revealed significant association with immune and inflammatory processes, including chemotaxis, leukocyte migration and activation, and regulation of leukocyte activation and differentiation (Fig. 6C). In the SC/MN@memPDA group, expression of genes in these pathways was broadly reduced, most prominently in IL-6 production, cytokine response, and leukocyte activation/migration, indicating that the treatment suppressed pro-inflammatory cascades, reduced immune cell recruitment and activation, and promoted macrophage rebalancing from M1 to M2 phenotypes, thereby lowering tissue inflammatory burden [68, 69]. KEGG analysis showed differential genes enriched in MAPK, PI3K–Akt, cytokine–cytokine receptor interaction, as well as cell adhesion molecules, ECM–receptor interaction, and Rap1/Calcium signaling (Fig. 6D) [70, 71]. These findings suggest that SC/MN@memPDA optimized the perifollicular regenerative microenvironment by suppressing inflammatory responses, lowering ROS, and reshaping lipid metabolism. KEGG clustering further grouped the enriched pathways into modules of inflammatory receptors/complement–coagulation, antioxidation, arginine/lipid metabolism, Ca²⁺/cGMP second messenger signaling, and cell cycle (Fig. 6E). Directional analysis indicated that SC/MN@memPDA suppressed pro-inflammatory receptor and complement cascades, reduced ROS-associated pathways, and reprogrammed arginine and lipid metabolism from the iNOS–NO axis toward ARG1–polyamine/proline synthesis, while enhancing cGMP/oxytocin-like anti-inflammatory signals and regeneration-related proliferative pathways. These changes supported macrophage transition toward an M2 reparative phenotype and optimization of the perifollicular microenvironment.

Gene set enrichment analysis (GO–GSEA) based on pre-ranked genes between SC/MN@memPDA and DHT groups showed positive enrichment for skin and epidermal development, epithelial differentiation, keratinization, cornified envelope, and keratin filament, along with extracellular matrix and collagen-related terms (extracellular matrix, collagen-containing ECM, ECM structural constituent/binding) (Fig. 6G). Enrichment of transmembrane receptor protein tyrosine kinase activator activity further indicated enhancement of growth factor receptor–driven regenerative signaling. These enrichments were consistent with the observed upregulation of Krt71/72/73 and Crnn/Padi3 as well as ECM remodeling, suggesting that treatment shifted transcriptional programs from inflammatory responses toward epithelial maturation and matrix stabilization, thereby supporting follicular recovery and anagen entry. Notably, NF-κB signaling was negatively enriched in the SC/MN@memPDA group (NES = − 1.15, P < 0.05), indicating global suppression of M1-related NF-κB activity, whereas Wnt signaling was positively enriched (NES = + 1.36, P < 0.05), consistent with reactivation of the β-catenin regenerative axis (Fig. 6H) [52].

Taken together, SC/MN@memPDA broadly reduced pro-inflammatory and oxidative stress–related gene expression while enhancing pathways related to regeneration and repair. Suppression of M1 inflammation, correction of lipid metabolic disturbances, activation of Wnt/EGFR signaling, and promotion of ECM repair jointly contributed to follicular regeneration. These findings suggest that SC/MN@memPDA promotes hair regrowth through a dual effect of reducing inflammatory burden and restoring regenerative programs.

MN@memPDA attenuates DHT-induced inflammation, promotes a reparative M2 phenotype, and improves the tissue microenvironment

To validate the transcriptomic findings, we investigated macrophage polarization differences. Immunofluorescence staining revealed that, compared with the DHT group, MN@memPDA markedly increased the proportion of F4/80⁺Arg1⁺ M2 macrophages while reducing F4/80⁺iNOS⁺ M1 macrophages, resulting in a significantly higher M2/M1 ratio (Fig. 7A–B) [72, 73]. MET also improved polarization but was less effective than MN@memPDA, with a residual M1 bias. MN and memPDA alone showed little effect, and M1 macrophages remained abundant, indicating a predominantly pro-inflammatory state. qPCR analysis confirmed that DHT stimulation strongly upregulated pro-inflammatory cytokines TNF-α, IL-6, and IL-1β, whereas MN@memPDA treatment significantly reduced their expression to near-control levels (Fig. 7C). At the same time, IL4ra and IL-10 were downregulated in the DHT group but were restored after MN@memPDA treatment, showing greater recovery than with MET or other interventions (Fig. 7D). These results demonstrate that MN@memPDA suppressed pro-inflammatory mediators while enhancing anti-inflammatory gene expression, thereby promoting M2 polarization at the molecular level and creating a more favorable immune microenvironment for follicular regeneration.

Fig. 7.

Evaluation of macrophage polarization in vitro and in vivo after SC/MN@memPDA treatment. (A) Representative immunofluorescence staining of M1 (F4/80⁺iNOS⁺) and M2 (F4/80⁺Arg1⁺) macrophages in DHT-stimulated RAW264.7 cells. Scale bars: 50 μm; (B) Quantification of M1/M2 macrophage ratios; (C) Relative mRNA expression of pro-inflammatory cytokines (TNF-α, IL-6, IL-1β) in RAW264.7 cells; (D) Relative mRNA expression of anti-inflammatory markers (IL4ra, IL-10) in RAW264.7 cells; (E) Relative mRNA expression of M1 markers (Nos2, IL-12b) and M2 markers (Arg1, Mrc1, IL4ra, IL-10) in dorsal skin of DHT-induced AGA-like mice. TNF-α, tumor necrosis factor-α; IL, interleukin; Nos2, nitric oxide synthase 2; Arg1, arginase-1; Mrc1, mannose receptor C-type 1; IL4ra, interleukin-4 receptor alpha. Data were presented as mean ± SDs. One-way ANOVA was used for multiple comparison tests, and Student’s t-test was used for two independent groups (*P < 0.05, **P < 0.01, ***P < 0.001)

To assess macrophage polarization in vivo, we examined skin tissues from DHT-induced AGA mice by qPCR for M1 markers (Nos2, IL-12b) and M2 markers (Arg1, Mrc1, IL-10, IL4ra). As shown in Fig. 7E, DHT markedly increased Nos2 and IL-12b expression, indicating sustained activation of NF-κB and STAT1 pathways that drive M1 polarization [73]. Elevated Nos2 suggested excessive production of reactive nitrogen species, which exacerbated oxidative and nitrative stress and further impaired the follicular microenvironment. Increased IL-12b indicated activation of the Th1 axis, likely enhancing IFN-γ secretion and amplifying inflammatory responses that disrupt follicular stem cell function. In contrast, anti-inflammatory and reparative markers IL4ra, IL-10, Arg1, and Mrc1 were significantly reduced in the DHT group, suggesting suppression of STAT6 and IL-4/IL-10 signaling [74, 75]. Reduced Arg1 implied weakened polyamine metabolism and diminished tissue repair and ECM remodeling, while decreased Mrc1, a canonical M2 marker, confirmed reduced M2 polarization and impaired local immunoregulatory function. Following intervention, SC/MN@memPDA produced the most pronounced improvement. Pro-inflammatory factors were markedly suppressed, while anti-inflammatory and reparative factors were strongly restored, approaching control levels. These findings indicate that SC/MN@memPDA inhibited the NF-κB/STAT1–iNOS pro-inflammatory pathway while activating the IL-4/IL-10–STAT6–Arg1/Mrc1 anti-inflammatory axis. This shift from M1 to M2 polarization reduced inflammatory cascades and oxidative stress, restored tissue repair capacity and ECM homeostasis, and thereby alleviated inflammation and improved the follicular microenvironment [68, 69]. Such changes created favorable conditions for follicular stem cell activation and regeneration.

Limitations and translational considerations

This work demonstrates a proof-of-concept peri-follicular delivery platform that couples suction-assisted short dissolving microneedles with biomimetic membrane-coated nanocarriers, enabling depth-resolved mapping of intradermal distribution. Translation to human AGA should nevertheless be approached cautiously. The DHT-induced C57BL/6 model provides a controlled androgen-challenge setting but represents an accelerated intervention and cannot fully capture the chronic, scalp-specific microenvironment of human AGA; interspecies and site differences in hair-cycle dynamics, follicle size and density, niche architecture, and immune and fibrotic remodeling may affect translatability. Moreover, the human scalp imposes practical constraints, including a deeper peri-follicular dermis, distinct biomechanics, and dense hair shafts that may complicate patch-type application. Accordingly, device adaptation and validation in human-relevant scalp systems will be important prior to clinical translation.

Conclusion

In this study, we developed a bioinspired suction-cup microneedle delivery system (SC/MN@memPDA). This system enables efficient drug loading, controlled release, and targeting/homing to the dermal papilla niche, which possesses markedly anti-oxidative stress and anti-inflammation abilities, finally restores hair growth signals. Overall, SC/MN@memPDA represents a safe, effective, and niche-targeted transdermal delivery strategy that offers a promising therapeutic option for androgenetic alopecia.

Experimental section

Chemicals and Materials

Anhydrous isopropanol, dopamine hydrochloride, 1,3,5-Trimethylbenzene (TMB), ammonium hydroxide solution (25–28%), metformin hydrochloride (MET), hyaluronic acid (20–40 kDa) and 1 H,1 H,2 H,2 H-Perfluorodecyltrichlorosilane (FDTS) were purchased from Macklin. Ecoflex 0030 was purchased from Smooth-On, Inc. (PA, USA). Sucker-cup and microneedle master molds were purchased from Taizhou Microchip Pharmaceutical Technology Co., Ltd., China. Pluronic F127, Rhodamine B, bicinchoninic acid (BCA) protein assay kit, reactive oxygen species assay kit and enhanced mitochondrial membrane potential assay kit with JC-1 were obtained from Beyotime. ZnSO₄·7 H₂O were obtained from Sigma. OCT embedding medium, Dulbecco’s Modified Eagle Medium (DMEM), phosphate-buffered saline (PBS, pH 7.4), Tris-HCl buffer (pH = 8.5) and 1% penicillin-streptomycin-amphotericin B solution were purchased from servicebio. Fetal bovine serum (FBS) was purchased from Dib. All reagents were used without further purification. Sequencing service were provided by Personal Biotechnology Co., Ltd. Shanghai, China.

Animals: C57BL/6J mice (6–7 weeks) were obtained from the Experimental Animal Centre of Southern Medical University (Guangzhou, China). All experimental procedures in-volving animals were carried out with the approval of the Institutional Animal Care and Use Committee of Southern Medical University (Application No: IACUC-LAC-20250728-003).

Drug-agnostic gene–drug interaction pipeline

A drug-agnostic workflow was employed to screen potential candidate drugs. First, the GeneCards database was queried using the keywords “oxidative stress” and “inflammation,” with the search restricted to protein-coding genes. After unifying the results to HGNC gene symbols and removing duplicates, the intersection of the two sets was extracted to generate the “inflammation ∩ oxidative stress” candidate gene set. To obtain a comparable composite relevance measure, the GeneCards relevance scores of both gene lists were subjected to quantile normalization, and their geometric mean was calculated as the primary ranking metric (with the smaller value between the two used as a conservative tie-breaker). Based on this index, the intersection genes were ranked. The ranked intersection set was then submitted to the DGIdb database for gene–drug interaction retrieval, followed by normalization of drug names and aggregation at the drug level. Within the top 100 intersection genes, the number of target genes covered by each drug was counted, thereby identifying candidate drugs with higher coverage.

Isolation of primary dermal papilla cells and M2-polarized macrophages

Primary DPCs were isolated from dorsal skin of 6–7-week-old C57BL/6 mice as previously described. Briefly, dermis–epidermis separation was achieved with Dispase II (2.4–2.5 U/mL, Sigma-Aldrich) at 4 °C for 12 h, followed by microdissection of hair follicles and collection of dermal papilla (DP) clusters under a stereomicroscope. DP explants were seeded in collagen I-coated 24-well plates and cultured in DPC medium (DMEM-LG supplemented with 10% FBS and 1% penicillin/streptomycin) at 37 °C in 5% CO₂. Cells were passaged at 80–90% confluence, and passages 2–4 were used for experiments.

RAW264.7 macrophages were maintained in DMEM-LG supplemented with 10% FBS. For M2 polarization, 2 × 10⁵ cells per well were seeded in 6-well plates and stimulated with recombinant mouse IL-4 (20 ng/mL; HZ-1004, Proteintech, USA) for 48 h. Successful polarization was verified by morphological changes (spindle-like and stellate) and expression of M2-associated markers.

Preparation and characterization of hybrid cell-membrane vesicles (DPC/M2-CMVs)

Cell membranes were extracted from DPCs and IL-4–induced M2 macrophages at 80–90% confluence. After washing with ice-cold PBS, cells were lysed in hypotonic buffer (10 mM Tris-HCl, 1 mM EDTA, with protease inhibitors) and homogenized. The lysates were subjected to differential centrifugation (800 g, 10 min; 10,000 g, 20 min; 100,000 g, 60 min) to isolate plasma membranes, which were washed, re-pelleted, and resuspended in PBS. Protein concentrations were determined by BCA assay.

Membranes from both cell types were mixed at a 1:1 protein ratio, incubated on ice, and extruded sequentially through 400 nm and 200 nm polycarbonate membranes (10 passes each). Aggregates were removed by centrifugation at 16,000 g for 10 min, and hybrid vesicles (2 mg/mL) were stored at 4 °C.

Membrane fusion was visualized by confocal microscopy (LSM 980, Carl Zeiss, Germany) after fluorescent labeling (DIO for DPC membranes and DIL for M2 membranes). Western blot was performed to confirm the retention of membrane proteins (F4/80, NCAM, Na⁺/K⁺-ATPase, and GAPDH).

Fabrication, drug loading, and release study of MPDA

MPDA were synthesized by dissolving dopamine (150 mg) and F127 (500 mg) in deionized water (20 mL) and anhydrous ethanol (20 mL), followed by sonication until complete dissolution. TMB (200 µL) was then added, and the mixture was stirred magnetically for 15 min. Ammonia solution (1 mL) was added dropwise, and the reaction proceeded at room temperature for 6 h. The resulting particles were collected by centrifugation, washed three times with ethanol: acetone (2:1, v/v), and further washed with ethanol before storage in ethanol suspension.

For Zn²⁺ loading, 10 mg of MPDA was dispersed in ZnSO₄·7 H₂O solution (10 mM, Tris-HCl buffer, pH 8.5), sonicated for 5 min, and incubated for 12 h under light protection. After centrifugation and washing, Zn-loaded particles were stored at 4 °C and verified by TEM with elemental mapping (JEM-F200, JEOL, Japan). For drug loading, Zn-loaded MPDA was dispersed in PBS (2 mg/mL) and mixed with metformin solution (10 mg/mL) at a volume ratio of 1:1. The suspension was incubated for 12 h, centrifuged, and washed once with PBS. The supernatant was collected for drug quantification, while the final MET-loaded particles (MPDA@MET) were resuspended in PBS (5 mg/mL). Morphology and elemental distribution were characterized by TEM and EDS, and particle size was determined using a Zetasizer Nano ZS90 (Malvern Panalytical, UK).

For in vitro drug release, MPDA@MET particles were sealed in dialysis bags and incubated in saline or H₂O₂ solutions (0.5, 1, and 5 mM) at 37 °C with shaking (150 rpm). Samples were withdrawn at predetermined time points (0–72 h) and replaced with fresh medium. Released metformin was quantified by UV-Vis spectroscopy at 232 nm, and cumulative release was calculated using a standard curve.

Cloaking of MPDA@MET with DPC/M2-CMVs (memPDA@MET)

MPDA@MET were coated with DPC/M2-CMVs at a membrane protein: MPDA mass ratio of 1:1. The suspension was extruded sequentially through 400 nm and 200 nm polycarbonate membranes (10 passes each) using a mini-extruder. Excess uncoated membranes were removed by centrifugation at 16,000 g for 10 min at 4 °C, and the resulting membrane-coated nanoparticles (memPDA@MET) were resuspended in PBS. The morphology and membrane coating were examined by transmission electron microscopy (JEM-F200, JEOL, Japan).

Fabrication of the bioinspired suction cup (SC)

A resin suction-cup mold (Taizhou Microchip Pharmaceutical Technology Co., Ltd., China) was sequentially sonicated in isopropanol and deionized water for 5 min each, dried at 60 °C for 15 min, and treated with UV–ozone under ambient pressure for 10 min. The mold was then placed in a sealed vacuum desiccator together with a small dish containing 20 µL of FDTS, and vapor-phase deposition was carried out under ambient pressure for 2 h. After the reaction, the mold was transferred to a 90 °C oven and cured for 30 min to form a fluorinated self-assembled monolayer (SAM) on its surface. Ecoflex 0030 was subsequently poured into the SAM-treated mold, degassed under vacuum, cured at room temperature for 4 h, and then demolded.

Fabrication and characterization of SC/MN@memPDA

To prepare SC/MN@memPDA, 10% hyaluronic acid (HA) solution was placed in an ice bath, and MPDA@MET@DPCM suspension was added dropwise under stirring. The mixture was allowed to stand at room temperature for 20 min, followed by vacuum degassing at 0.1 MPa for 5 min until bubbles were completely removed. The solution was dispensed onto the mold surface, centrifuged at 5000 g for 5 min, and excess liquid on the surface was aspirated. Additional 10% HA solution was applied to fill the mold and form the basal backing layer, followed by centrifugation at 4000 g for 5 min to eliminate voids. The mold was dried overnight at 30 °C and demolded to obtain MN-DPCM@MPDA@MET. Subsequently, an HA solution was spin-coated onto the contact region of the suction-cup apex using a spin coater (WS-400-6NPP, Laurell Technologies, USA) at 500 rpm for 1 min. The pre-fabricated MN-DPCM@MPDA@MET array was transferred onto the thin HA film and dried overnight at room temperature.

For characterization, the morphology and distribution of microneedles were examined using a stereomicroscope (VHX-7000, Keyence, Japan) and scanning electron microscopy (MIRA LMS, TESCAN, Czech Republic). The suction performance of SC on dry and wet surfaces was assessed using a universal testing machine (CMT4503, SANS, China). The force–displacement curves of the microneedles were obtained with a mechanical tester (5944, Instron, USA). Fluorescently labeled microneedles were fabricated by incorporating FITC-labeled HA (HA-FITC) as the microneedle matrix and Rhodamine B (RhB)-stained MPDA nanoparticles (RhB-MPDA) as the drug model. The distribution of nanoparticles within the microneedle shafts was visualized using a confocal laser scanning microscope (Stellaris 5, Leica Microsystems, Germany).

To account for potential process losses during dispensing/handling and to define absolute dosing at the single MN subunit level, the total MET content of individual 3-mm MN subunits was quantified by HPLC after complete extraction (n = 6). After in vivo application, the retrieved backing layer corresponding to each MN subunit was collected and completely extracted for HPLC quantification of MET remaining on the backing (n = 6). The net MET amount released from a single MN subunit during one application was calculated as (total MET content in the MN subunit − MET remaining on the retrieved backing). One SC/MN patch (1 × 1 cm) comprises eight MN subunits (3 mm diameter each) and was used per session in vivo.

In vitro cellular oxidative stress assessment

The intracellular ROS level was assessed using the fluorescent probe DCFH-DA to evaluate the antioxidant activity of MN@memPDA. DPCs were seeded in 24-well plates and cultured overnight to allow adherence. Except for the blank control (CTRL), cells were treated with Rosup reagent for 30 min to induce oxidative stress. Subsequently, all groups except the positive oxidative-stress control (ROSUP) were incubated for 4 h with culture medium containing either blank microneedles (MN), blank membrane-coated particles (memPDA), free metformin (MET), or MN@memPDA (n = 3 wells per group). Cells were then incubated with DCFH-DA for 20 min in the dark and imaged using an inverted automated microscope (ECLIPSE Ti2-E, Nikon, Japan). Relative fluorescence intensity was quantified with ImageJ. For flow cytometry, treated cells were harvested, washed with PBS, incubated with DCFH-DA in the dark for 20 min, washed to remove excess probe, resuspended in PBS, and analyzed on a flow cytometer (CytoFLEX N31376, Beckman Coulter, USA).

DPCs were seeded in 24-well plates and grouped as described above; cells treated with carbonyl cyanide m-chlorophenyl hydrazone (CCCP) served as the mitochondrial depolarization positive control. Following the indicated treatments, medium was removed and cells were gently washed with PBS. JC-1 working solution (prepared per manufacturer’s instructions) was added and cells were incubated in the dark for 20 min. After discarding the staining solution, cells were rinsed 1–2 times with buffer and covered with detection buffer. Red (Ex ≈ 535 nm/Em ≈ 590 nm) and green (Ex ≈ 485–490 nm/Em ≈ 530 nm) fluorescence were recorded using a confocal laser scanning microscope (LSM 980, Carl Zeiss, Germany). ImageJ was used to quantify fluorescence and calculate the red/green ratio for each group. An increased red/green ratio relative to the oxidative-stress control indicated preserved mitochondrial membrane potential and reduced oxidative damage, reflecting antioxidant protection. All measurements were performed under light-protected conditions with consistent cell density and incubation times, and in three independent replicates.

qPCR assessment of oxidative stress and antioxidant effects in DPCs

DPCs were divided into three groups: blank control, DHT-treated, and drug intervention (DHT combined with the corresponding drug, consistent with the groupings used for ROS assays). Following treatment, total RNA was extracted using a commercial RNA extraction kit, and RNA concentration and purity were determined. Equal amounts of RNA were reverse-transcribed into cDNA according to the manufacturer’s instructions. Quantitative PCR was performed using a SYBR Green system to evaluate the expression of oxidative stress- and inflammation-related genes, including Nrf2, HO-1, FTH1, NQO1, SOD2, GCLM, IL-6, IL-1β, and TNF-α. GAPDH served as the internal control. Cycling parameters and primer concentrations were set according to the manufacturer’s recommendations. Relative gene expression was calculated using the 2^–ΔΔCt method, comparing the DHT group with the blank control and the drug intervention groups with the DHT group.

In vitro immunofluorescence staining of Ki67 in DPCs

DPCs were seeded onto sterile round coverslips placed in 24-well plates and cultured until reaching ~ 70–80% confluence. Cells were divided into six groups: control, DHT, blank MN, blank memPDA, free MET, and MN@memPDA. For the DHT and drug treatment groups, cells were pretreated with 1 µM DHT for 24 h before drug administration. After treatment, cells were washed with PBS, fixed with 4% paraformaldehyde at room temperature for 15 min, permeabilized with 0.1% Triton X-100 for 10 min, and blocked with 5% BSA for 1 h. Primary antibody against Ki67 (28074-1-AP, Proteintech, USA) was applied overnight at 4 °C, followed by PBS washing and incubation with fluorescent secondary antibody (1:500, 1 h at room temperature in the dark). Nuclei were counterstained with DAPI. Samples were mounted with antifade reagent, and images were acquired using a confocal laser scanning microscope (LSM 980, Carl Zeiss, Germany).

Live/dead cell staining of DPCs

After the indicated treatments, DPCs were washed with PBS and incubated with a working solution containing Calcein-AM (2 µM) and propidium iodide (PI, 4 µM) for 20 min in the dark. Following staining, cells were washed and covered with phenol red–free medium. Images were obtained using an inverted automated microscope (ECLIPSE Ti2-E, Nikon, Japan). ImageJ software was used for image analysis and quantification of live/dead cell ratios.

In vivo evaluation of SC/MN@memPDA in a DHT-induced hair loss model

Male C57BL/6J mice (6–7 weeks old) were weighed and anesthetized by intraperitoneal injection of sodium pentobarbital. On day 0, dorsal hair was removed, and skin color was examined to confirm telogen phase. Telogen-phase mice were randomly divided into six groups: control, DHT, memPDA, MN@memPDA, SC/MN@PDA, and SC/MN@memPDA. From day 1, DHT was administered intraperitoneally once daily for 21 consecutive days to establish an androgen-dependent hair loss model, and group-specific treatments were administered in parallel. During the intervention, gross dorsal images were captured to monitor hair regrowth, and hair coverage was quantified.

On day 21, mice were euthanized, and dorsal skin tissues were harvested. Samples from the midline dorsal region were fixed and paraffin-embedded for histological analysis, while tissues from the bilateral dorsal regions were snap-frozen in liquid nitrogen and stored at − 80 °C for subsequent qPCR and transcriptome sequencing. Hematoxylin and eosin (H&E) staining was performed to evaluate follicular morphology across different groups.

Component-separated evaluation of Zn²⁺ and MET

RAW264.7 macrophages were seeded in 96-well plates and allowed to adhere overnight. Cells were stimulated with LPS (1 µg/mL) for 24 h to induce inflammation. Subsequently, cells were treated with the following conditions: control, Zn²⁺ only, MET only, and Zn²⁺+MET (dose-equivalent to the amounts used in the Zn²⁺/MET co-loading formulation). Cell culture supernatants were collected and the levels of IL-6 and TNF-α were quantified using ELISA kits according to the manufacturer’s instructions. Data are presented as mean ± SD (n = 3 biological replicates).

Mouse DPCs were seeded and treated under the same oxidative-stress induction protocol as described in Sect.  7.8. Cells were then incubated with Zn²⁺ only, MET only, or Zn²⁺+MET (dose-equivalent) for 4 h. Intracellular ROS was assessed using DCFH-DA staining. After incubation with DCFH-DA for 20 min in the dark, cells were washed, resuspended in PBS, and analyzed by flow cytometry (CytoFLEX N31376, Beckman Coulter, USA). Data are presented as mean ± SD (n = 3 biological replicates).

DPCs were seeded on coverslips and pretreated with DHT (1 µM, 24 h), consistent with Sect.  7.10. Cells were then treated with Zn²⁺ only, MET only, or Zn²⁺+MET (dose-equivalent). Ki67 immunofluorescence staining was performed as described in Sect.  7.10. Quantification was conducted in ImageJ from multiple randomly selected fields per sample. Data are presented as mean ± SD (n = 3 biological replicates).

To clarify the individual and combined contributions of Zn²⁺ and MET in vivo, additional component-separated control groups were included in the DHT-induced hair-loss model described in Sect.  7.12: Zn²⁺ only and Zn²⁺+MET. A topical minoxidil-treated group was also included as a positive-control comparator, and was evaluated in parallel under the same modeling timeline and endpoint readouts. The dosing and application procedure (including patch format, application duration, and treatment schedule) were kept identical to the main therapeutic study to ensure comparability. At the study endpoint, dorsal skin tissues were harvested for H&E staining and AR immunohistochemistry following the same protocols described in the main study.

In vitro cellular uptake of labeled MPDA and memPDA

In vitro uptake of nanoparticles was evaluated in DPCs. MPDA and membrane-coated MPDA (memPDA) were fluorescently labeled with Alexa Fluor 488 NHS ester (50–100 µg/mL in 0.1 M NaHCO₃ buffer, pH 8.3) for 1 h at room temperature in the dark. Free dye was removed by ultrafiltration, and the labeled nanoparticles were resuspended in PBS. DPCs were seeded on collagen I-coated coverslips in 24-well plates (1 × 10⁵ cells/well) and cultured overnight at 37 °C in a humidified 5% CO₂ atmosphere. Cells were then incubated with equal fluorescence doses of AF488-MPDA or AF488-memPDA for 4 h, washed, fixed with 4% paraformaldehyde, and counterstained with DAPI. Samples were mounted with antifade reagent and imaged using a laser scanning confocal microscope (LSM 980, Carl Zeiss, Germany).

For cross-cell-type comparison, mouse DPCs, primary mouse dermal fibroblasts (FBs), and HaCaT keratinocytes were seeded under identical conditions and incubated with AF488-memPDA for 24 h (equal fluorescence dose across cell types). After incubation, cells were washed thoroughly with PBS to remove unbound particles, fixed with 4% paraformaldehyde, and counterstained with DAPI. Confocal images were acquired using the same acquisition settings across groups within each experiment. Quantification was performed in ImageJ by measuring the mean cell-associated AF488 fluorescence intensity per cell (a.u.) from multiple randomly selected fields (n = 3 independent experiments; ≥5 fields per group per experiment).

In vivo perifollicular accumulation of AF488-labeled nanoparticles delivered by SC/MN patches

In vivo perifollicular accumulation was assessed in C57BL/6J mice. MPDA and membrane-coated MPDA (memPDA) were fluorescently labeled with Alexa Fluor 488 NHS ester. Seven-week-old male mice were randomized, shaved, and depilated 24 h before treatment. Under isoflurane anesthesia, SC/MN patches loaded with either AF488-MPDA or AF488-memPDA were applied to the dorsal skin following predetermined suction–application parameters. After patch removal, residual formulation was gently wiped off with PBS. At 12 h post-administration, mice were sacrificed, and dorsal skin samples from the application sites were harvested, embedded in OCT, and cryosectioned at − 20 °C. Sections were fixed, counterstained with DAPI, mounted, and imaged using a confocal laser scanning microscope (LSM 980, Carl Zeiss, Germany) under identical laser power, gain, and exposure settings across groups.

In vivo skin penetration behavior of SC/MN@memPDA

To evaluate transdermal penetration, MPDA was labeled with Rhodamine B and DPC/M2-CMVs were labeled with DIO. Seven-week-old male C57BL/6J mice were anesthetized with sodium pentobarbital, shaved, and depilated on the dorsal skin, and then randomly divided into three groups: MN@PDA, MN@memPDA, and SC/MN@memPDA. After the corresponding interventions, microneedles or suction-cup microneedle patches were removed, and mice were sacrificed 72 h later. To distinguish initial entry from time-dependent redistribution, a separate cohort was sacrificed immediately after patch removal (T0) and the skin was processed identically for serial-depth cryosectioning and confocal imaging. Dorsal skin was excised, trimmed into appropriate sizes, embedded in OCT, and cryosectioned at 10 μm thickness. Sections at 0, 50, 100, 150, 200, 250, and 300 μm from the skin surface were examined using confocal microscopy (LSM 980, Carl Zeiss, Germany). Fluorescence distribution and intensity were analyzed to assess delivery depth among groups.

Transepidermal water loss (TEWL) measurement

TEWL was measured on the shaved dorsal skin using a tewameter (Tewameter^® TM 300; Courage + Khazaka electronic GmbH, Cologne, Germany). Mice were acclimated under controlled ambient conditions prior to measurement. TEWL was recorded at -10 min (baseline, 10 min before treatment), then the patch was applied according to the in vivo administration protocol. After patch removal, TEWL was measured at 0 min (immediately after removal) and at 10, 20, 40, 60, and 90 min post-treatment. For each animal and time point, the probe was placed perpendicularly on the same treated area with minimal pressure until the reading stabilized; the recorded value was used for analysis. Data are reported as mean ± SD (n = 3).

Immunofluorescence and immunohistochemical analysis of dorsal skin

Immunofluorescence staining was performed to evaluate follicular microenvironment–related signaling using antibodies against Ki67 (28074-1-AP, Proteintech, USA), TGF-β1 (81746-2-RR, Proteintech, USA), and β-catenin (51067-2-AP, Proteintech, USA). Immunohistochemical staining was carried out to detect androgen receptor (AR) expression (22089-1-AP, Proteintech, USA), in order to assess androgen signaling activity in mouse skin and hair follicles and its modulation by DHT stimulation or drug intervention.

RNA isolation and quantitative real-time PCR

Frozen dorsal skin tissues were homogenized using a cryogenic sample grinder (75 Hz, 30 s grinding, 20 s pause, six cycles; LUKYM-I, Luka, China). Tissue homogenates were centrifuged at 12,000 rpm for 10 min, and the supernatants were collected for RNA extraction and purification. RNA extraction, reverse transcription, and quantitative PCR using the SYBR Green system were performed as described above. Gene expression associated with oxidative stress, inflammation, and hair growth was assessed. Hair growth–related genes included DKK1, TGF-β1, Lgr5, BMP2, BMP4, Wnt3a, Wnt5a, Wnt10b, GLI1, and GLI2. GAPDH served as the internal reference, and relative expression was calculated using the 2^−ΔΔCt method.

Establishment of the 5AR-Luc reporter cell line

A stable 5α-reductase reporter cell line (5AR-Luc) was generated using a lentiviral construct encoding a fusion of steroid 5α-reductase type 2 (SRD5A2) and firefly luciferase (pLVX-CMV-SRD5A2-Luc). NIH/3T3 cells were transduced with the lentivirus under standard culture conditions. At 48 h post-transduction, the medium was replaced with fresh complete medium containing puromycin (2 µg/mL) for antibiotic selection. After selection, surviving cells were expanded and used for subsequent drug screening experiments in the DHT-stimulated 5AR-luciferase assay.

Transcriptome analysis of mouse skin tissue via RNA sequencing

Dorsal skin tissues were harvested at the end of the experimental treatments, rapidly snap-frozen, and submitted to a professional sequencing company for transcriptome analysis (RNA-seq). Whole-transcriptome profiling of skin tissues was performed to evaluate the effects of DHT stimulation and drug intervention on the follicular microenvironment, inflammatory responses, and associated signaling pathways, thereby providing a systematic understanding of the underlying mechanisms of drug action.

Evaluation of macrophage polarization in vitro and in vivo

RAW264.7 murine macrophages were cultured in DMEM supplemented with 10% fetal bovine serum (FBS) at 37 °C in a humidified atmosphere with 5% CO₂ until reaching 70–80% confluence. Cells were seeded onto coverslips in 24-well plates and divided into three groups: control, DHT-treated, and drug intervention. Following treatment, immunofluorescence staining was performed to evaluate macrophage polarization. iNOS was used as the M1 marker, Arg1 as the M2 marker, and F4/80 as the macrophage marker. After incubation with primary and fluorescently labeled secondary antibodies, cells were imaged using a confocal laser scanning microscope (LSM 980, Carl Zeiss, Germany) to assess polarization states and fluorescence intensity, thereby determining the anti-inflammatory effects of treatment. Parallel cell groups were collected for RNA extraction, reverse-transcribed into cDNA, and analyzed by SYBR Green qPCR for IL4ra, IL-10, TNF-α, IL-6, and IL-1β expression, with GAPDH as the internal control.

Dorsal skin tissues were collected from each treatment group of mice, and total RNA was extracted. RNA concentration and purity were measured, and equal amounts were reverse-transcribed into cDNA. SYBR Green qPCR was performed to quantify the expression of M1-associated genes (NOS2 and IL-12b) and M2-associated genes (IL4ra, IL-10, ARG1, and MRC1). GAPDH was used as the internal control, and qPCR cycling conditions were set according to the manufacturer’s protocol.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

X.W. performed validation, investigation, and data curation, and wrote the original draft. D.Z. conducted formal analysis. M.Z., Y.W., and Z.F. provided resources. H.W. and J.Z. developed software and created visualizations. L.L. contributed to the investigation. Z.H. (Ziwei Hu) performed data curation. X.L. developed the methodology. Z.H. (Zhiqi Hu) supervised the project and managed project administration. Y.G. (Yuyang Gan) contributed to conceptualization and methodology and reviewed and edited the manuscript. J.W. reviewed and edited the manuscript and acquired funding. All authors reviewed the manuscript.

Funding

Research reported in this publication was supported by the Science and Technology Projects in Guangzhou (Grant No.2023A04J2319), the Natural Science Foundation of China (Grant No. 82473550), the Guangdong Basic and Applied Basic Research Foundation (Grant No. 2023A1515012054).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Competing interests

The authors declare no competing interests.