Root-system-inspired core–shell microneedles enable spatiotemporal sequential therapy via ROS scavenging, angiogenesis, and capillary-driven lipid removal for enhanced fat graft survival

Abstract

Background

Autologous fat grafting is widely used in reconstructive and esthetic surgery, but its clinical outcomes are limited by early ischemia–hypoxia, oxidative stress, delayed vascularization, and subsequent lipid accumulation-induced inflammation. Current therapeutic strategies generally focus on single-stage regulation and lack temporal coordination with the dynamic pathological evolution of transplanted adipose tissue. Here, we developed a root system-inspired core–shell microneedle (MN) platform capable of sequentially regulating the graft microenvironment through early vascular promotion and later lipid removal.

Methods

Vascular endothelial growth factor-loaded modified silk fibroin methacryloyl/grooved poly(lactic-co-glycolic acid) core–shell MNs (VEGF@mSF/gPLGA–MNs) were fabricated by integrating a reactive oxygen species (ROS)-responsive mSF shell with a lipid-adsorbing gPLGA core. The physicochemical properties, ROS scavenging ability, VEGF release behavior, and lipid adsorption capacity of the MNs were systematically characterized. Their biological effects were evaluated using in vitro cellular assays and a mouse autologous fat grafting model. Graft survival, vascularization, inflammatory regulation, and adipose remodeling and the underlying molecular mechanisms were assessed by imaging, histological analyses, immunofluorescence staining, ribonucleic acid sequencing, and western blotting.

Results

The mSF shell rapidly responded to oxidative stress and degraded during the early stage after implantation, enabling localized VEGF release while alleviating ROS-induced cellular damage. Following shell degradation, the exposed gPLGA core facilitated directional lipid adsorption owing to its lipophilic properties and groove-mediated capillary transport. In vitro studies demonstrated improved endothelial migration, angiogenesis, ROS clearance, and mitochondrial protection. In a mouse fat graft model, VEGF@mSF/gPLGA–MNs significantly improved graft retention, reduced cystic degeneration and fibrosis, increased vascularization, and promoted adipose tissue remodeling. Transcriptomic and protein analyses revealed that lipid metabolism-related pathways, including PPAR signaling, were activated, and that inflammatory pathways, such as the NF-κB, TNF, IL-17, and toll-like receptor signaling pathways, were suppressed.

Conclusions

The root system-inspired VEGF@mSF/gPLGA–MN platform enables spatiotemporally sequential regulation of the fat graft microenvironment by coordinating early vascular reconstruction with subsequent lipid clearance. This strategy provides a promising approach for improving fat graft survival and may offer a generalizable paradigm for regenerative therapies involving dynamic pathological transitions.

Highlights

• A root-system-inspired core–shell microneedle is designed, featuring a reactive oxygen species-responsive shell and a grooved/porous lipophilic core.

• The system sequentially alleviates early oxidative stress and removes late necrotic lipids through capillary-driven adsorption, reshaping the graft microenvironment.

• It significantly enhances graft survival, reduces inflammation, and maintains adipocyte viability in vivo.

• The strategy offers a new paradigm for treating lipid-driven inflammatory diseases through local physical intervention.

Background

Adipose tissue reconstruction is critical for esthetic procedures, including facial augmentation and mammoplasty, as well as for repairing soft tissue defects caused by congenital anomalies, trauma or other surgical operations [1–3]. However, graft failure remains common, primarily because of insufficient blood supply to the transplanted tissue and poor integration with the host tissue induced by inflammatory responses [4–7]. Technical improvements, such as the use of vascularized fat flaps and minimally invasive micro-fat grafting protocols, have improved outcomes by promoting the initial stage of vascularization and reducing mechanical injury [8, 9]. However, these approaches address primarily the early phase of graft failure, and a critical downstream phase, governed by inflammatory pathology, is frequently overlooked. This downstream phase is initiated by the inevitable ischemia–reperfusion injury and oxidative stress that occur, which lead to extensive adipocyte necrosis [10–12]. The subsequent release of cytotoxic lipid droplets and damage-associated molecular patterns (DAMPs) triggers a robust inflammatory cascade [7, 11, 13]. A key problem arises when the phagocytic clearance of the excessive lipid debris overwhelms both resident and recruited macrophages, driving macrophages toward a proinflammatory state [13, 14]. This results in a self-perpetuating cycle of chronic inflammation, impaired formation of new blood vessels, and fibrosis, ultimately leading to graft resorption and clinical failure [15, 16]. This observation highlights the urgent need for active intervention strategies to disrupt this pathological cycle, thereby improving graft survival and long-term retention.

Adipocytes serve as the primary lipid-storage cells and endocrine regulators within adipose tissue [17]. Large amounts of their intracellular lipids are released into the transplant microenvironment because of ischemia/reperfusion injury-induced necrosis, a pathological process termed “lipid spillover”, which includes free fatty acids, triglycerides, and related species [18, 19]. Abnormal lipid spillover and deposition that induce tissue damage have been widely reported in multiple systemic disorders, such as atherosclerosis, non-alcoholic steatohepatitis, and severe acute pancreatitis [20–24]. This evidence highlights “lipid overload” as a common, cross-organ mechanism that triggers inflammation and cellular dysfunction [25–27]. Within the context of fat grafts, the same mechanism underlies a core pathological event: DAMPs released from necrotic cells combine with the spilled lipids to form a potent inflammatory signaling complex [12, 14]. Free fatty acids, in particular, serve as precursors of proinflammatory mediators such as prostaglandins, directly activating and amplifying the local inflammatory cascade. Consequently, the necrotic core evolves from a purely ischemic region into an inflammatory niche that continuously emits “lipid signals” [28, 29]. Excessive extracellular lipid droplets overwhelm the local clearance capacity, providing persistent stimulation that drives macrophage infiltration [30]. During phagocytosis, macrophages readily experience “phagocytic overload”, which drives their polarization toward a proinflammatory M1 phenotype and induces the secretion of large quantities of cytokines, including TNF-α and IL-1β [11]. This lipid-driven, self-amplifying inflammatory feedback loop severely disrupts the stable microenvironment required for angiogenesis and tissue repair, thereby serving as a key determinant of graft failure. Although recent research has largely focused on improving graft vascularization or increasing adipocyte viability (e.g. through stem cell-based strategies) [31, 32], the critical role of sustained inflammation driven by necrotic lipids has often been overlooked. Therefore, targeting post-necrotic lipid release and its metabolic consequences is essential for disrupting the vicious cycle of fat graft failure.

Recently, a variety of biomaterial-based strategies have been developed for lipid metabolism-related disorders, including nanoparticles that regulate intracellular lipid metabolism [33], nanocarriers that promote reverse lipid transport to achieve plaque regression [34], and agents that directly eliminate lipid droplets [35]. However, these approaches rely predominantly on enhancing the host’s intrinsic metabolic clearance mechanisms rather than physically removing lipids from the lesion site, resulting in limited efficiency and a delayed response. Moreover, interventions targeting early-stage key pathological events after fat grafting, namely, oxidative stress and insufficient vascularization, also face formulation-related limitations. Boronate ester-based materials (e.g. silk fibroin derivatives) have been developed to rapidly scavenge ROS [36]. Nevertheless, conventional formulations such as hydrogels often exhibit poor adhesion and limited retention at graft sites, hindering precise spatiotemporal regulation of the early local microenvironment.

As a minimally invasive and precisely targeted local delivery platform, microneedles (MNs) offer a novel strategy to overcome the aforementioned limitations [37]. They not only enable long-term retention and controlled release of therapeutics at the target site but also provide the physical potential to actively remove pathological substances owing to their microscale structure [38]. Previous studies have designed channels or grooves in MNs to load, deliver, or adsorb active components. Nevertheless, such work often overlooks the intrinsic physicochemical properties of the material itself, especially its affinity and adsorption capacity for specific targets [39, 40]. Poly (lactic-co-glycolic acid) exhibits excellent biocompatibility and controllable degradation, but its inherent lipophilicity gives it outstanding potential to specifically adsorb and remove lipid droplets, thereby providing an ideal material basis for constructing highly efficient lipid-clearing MNs [41]. Therefore, engineering an integrated MN system that combines these advantages—the capability to respond to early oxidative stress through smart materials and actively clear lipids with its lipophilic core—represents a highly promising strategy for achieving coordinated intervention and breaking the vicious cycle of fat graft failure.

Herein, we present a core–shell MN system engineered for spatiotemporally coordinated fat grafting therapy (Figure 1). Inspired by the efficient nutrient uptake system of plant roots, which relies on capillary forces and large specific surface areas, our design features a functional core capable of adsorbing lipids (Figure 1c). The shell is fabricated from a silk fibroin hydrogel chemically grafted with phenylboronic acid and cross-linked with poly (vinyl alcohol). This hydrogel acts as a ROS-responsive matrix for localized vascular endothelial growth factor (VEGF) delivery. This design enables dual functions in the early phase after grafting: rapid ROS scavenging to alleviate oxidative stress and controlled VEGF release to promote angiogenesis. The core is fabricated from poly(lactic-co-glycolic acid) (PLGA) and contains a porous tip region generated using thermally removable gelatin, while the entire shaft is patterned with microgrooves through mold fabrication. These structural features mimic the high surface area architecture of root systems, synergistically increasing the capillary-driven uptake and physical adsorption of necrotic lipid droplets, thereby facilitating their subsequent removal (Figure 1a and b). Through this integrated design, the MN system operates in a sequential and coordinated manner, much like the way roots simultaneously manage adsorption and environmental interactions (Figure S1). Initially, the ROS scavenging shell locally releases VEGF to promote early vascularization while mitigating oxidative damage. Subsequently, the porous and grooved PLGA core actively adsorbs and facilitates the removal of necrotic lipid droplets, thereby alleviating a major source of sustained inflammation. These actions collectively shift the immune microenvironment from a proinflammatory state to a pro-regenerative state, maintaining mitochondrial homeostasis in adipocytes by preserving membrane potential and ultrastructural integrity, and thereby promoting adipocyte survival (Figure 1d). In this study, we systematically characterized the material properties, validated the ROS-responsive VEGF release and efficient lipid adsorption capabilities in vitro, and evaluated the efficacy of the system in a murine fat grafting model. Our results demonstrate that this MN platform significantly increases graft survival, reduces inflammation, and maintains adipocyte viability. Therefore, this core–shell MN system establishes a new therapeutic paradigm for fat grafting and provides critical insights for managing other disorders driven by pathological lipid accumulation.

Figure 1.

Fabrication, animal administration and underlying mechanisms of VEGF-loaded core–shell MNs for improving the survival rate of transplanted fat tissue. (a) Preparation of the MN core and mSF hydrogel loaded with VEGF. (b) Construction of photopolymerized core–shell MNs. (c) Application of VEGF@mSF/gPLGA–MNs in animal models. (d) Mechanism through which VEGF@mSF/gPLGA–MNs increase fat graft survival: promoting angiogenesis, reducing oxidative stress, suppressing inflammatory pathways, inhibiting M1 macrophage polarization, promoting adipose browning, and reshaping the microenvironment to accelerate tissue remodeling. VEGF vascular endothelial growth factor, VEGF@mSF/gPLGA–MNs VEGF-loaded modified silk fibroin/grooved poly(lactic-co-glycolic acid)–microneedles

Methods

Materials

In this study, unless otherwise specified, all the chemicals and reagents were of analytical grade and were used as received from commercial vendors without further purification. Common cell culture reagents, namely, fetal bovine serum (FBS), high-glucose Dulbecco’s modified Eagle’s medium (DMEM), and phosphate-buffered saline (PBS), were purchased from Gibco (MA, USA), while penicillin–streptomycin solution (5000 U/ml) and trypsin were obtained from Solarbio Science & Technology Co., Ltd (Beijing, China). For adipogenic induction, 3-isobutyl-1-methylxanthine (IBMX), dexamethasone (DEX), and insulin were purchased from Sigma–Aldrich (MO, USA). Biomaterials, specifically silk fibroin methacryloyl (SilMA) and lithium phenyl (2,4,6-trimethylbenzoyl) phosphinate (LAP), were obtained from Engineering for Life (Suzhou, China).

Specialized assay kits and matrices were obtained as follows: the Cell Counting Kit-8 (CCK-8) was purchased from UElandy Biotechnology Co., Ltd (Suzhou, China); a suite of kits from Beyotime Biotechnology Co., Ltd (Shanghai, China), including the JC-1 mitochondrial membrane potential (MMP) assay kit, live/dead cell staining kit, mitochondrial deep red fluorescence staining kit, and ROS assay kit (with DCFH-DA); and Matrigel® basement membrane matrix was acquired from Corning Inc. (NY, USA). Additionally, the fluorescent contrast agents sodium fluorescein and rhodamine B (RB) were purchased from Aladdin Biochemical Technology Co., Ltd (Shanghai, China). Enzyme-linked immunosorbent assay (ELISA) kits for VEGF were purchased from ABclonal Technology Co., Ltd (Wuhan, China).

Reagents acquired from Shanghai Macklin Biochemical Technology Co., Ltd (Shanghai, China) included the polymer materials poly(vinyl alcohol) (PVA; MW 195000), gelatin (Gel; MW 50000), PLGA (MW 90000) and small molecules [N-(3-dimethylaminopropyl)-N′-ethylcarbodiimide hydrochloride (EDC), 2-mercaptoethanol (2-ME), N-hydroxysuccinimide (NHS), ethyl acetate (EA), butyl acetate (BA), 3-aminophenylboronic acid (APBA), and sodium hydroxide (NaOH)]. For immunodetection, antibodies against CD86 and CD206 for flow cytometry were purchased from 4A Biotech Co., Ltd (Suzhou, China). Antibodies against CD31, perilipin, uncoupling protein 1 (UCP-1), and PPARγ for immunofluorescence staining were obtained from Abcam plc (Cambridge, UK).

Custom-fabricated MN molds, the tip diameter and spacing of which were tailored to our experimental requirements, were acquired from Taizhou Microchip Co., Ltd (Taizhou, China).

Bioinformatic pipeline for lipid-associated disease transcriptomics

Transcriptomic datasets for lipid-associated diseases, including nonalcoholic steatohepatitis, atherosclerosis, and acute pancreatitis (GSE33814, GSE100927, and GSE194331, respectively), were retrieved from the Gene Expression Omnibus (GEO; http://www.ncbi.nlm.nih.gov/geo) [42–44]. In addition, the adipose tissue RNA-seq dataset GSE203599 was obtained from GEO and included control adipose tissue (n = 5) and fat graft samples harvested on Days 3 (n = 5) and 7 (n = 5) after transplantation [45]. Inflammation- and lipid-related gene sets were downloaded from the Molecular Signatures Database (https://www.gsea-msigdb.org/gsea/msigdb). The differential expression analysis in disease models were done in R by using the DESeq2 and limma packages, and plotted to show the major transcriptional alterations [46, 47]. The phenotypic variations in the biological processes and signaling pathways were evaluated by gene set variation analysis (GSVA) to measure the pathway activity scores related to the pathologies associated with lipids [48]. Single-sample gene set enrichment analysis (ssGSEA), which is within GSVA, was then used to calculate the macrophage infiltration enrichment score, the inflammation score and the lipid metabolism score in each sample. These scores allow for systematic correlation analyses among the three phenotypic dimensions [49].

Modification of silk fibroin and preparation of the VEGF-loaded modified silk fibroin hydrogel

Silk fibroin (SF) was first modified to incorporate functional groups with the desired properties. The carboxyl groups of APBA (0.006 g, 0.28 mmol) were activated in 3 ml of DMSO using EDC (0.0546 g, 0.28 mmol) and NHS (0.033 g, 0.28 mmol) for 2 h. The activated APBA solution was then added to a solution of SilMA (0.3 g in 40 ml of ultrapure water). The reaction mixture was stirred at room temperature for 48 h, after which the pH was adjusted to ~7.4 with 0.1 M NaOH. The resulting solution was dialyzed against ultrapure water for 3 days (14 000 Da MWCO; the water was refreshed every 4 h) and lyophilized, and the purified SilMA–APBA conjugate was stored at 4°C until further use.

To prepare the photoresponsive VEGF-loaded modified silk fibroin (VEGF@mSF) hydrogel, recombinant VEGF was first dissolved in PBS to a working concentration of 40 μg/ml. This VEGF solution was then mixed with 20% (w/v) PVA solution at a 1:1 (v/v) ratio and left to stand briefly at room temperature. Separately, a 20% (w/v) solution of phenylboronic acid-functionalized methacrylated silk fibroin (SilMA–APBA) was prepared, to which 0.1% (w/w) of the photoinitiator LAP was added. Equal volumes of the two precursor solutions were mixed and thoroughly vortexed to ensure homogeneity, thereby inducing rapid gelation to afford the photoresponsive VEGF@mSF hydrogel. After lyophilization, the samples were characterized by Fourier transform infrared (FTIR) spectroscopy, X-ray photoelectron spectroscopy (XPS), and solution-state proton nuclear magnetic resonance (¹H NMR) spectroscopy to determine their chemical structures and compositions.

Design of the PDMS Moulds

The MN height and groove length were selected on the basis of preliminary optimization experiments considering skin penetration requirements, fabrication feasibility, structural stability, and capillary transport performance.

Three custom-designed polydimethylsiloxane (PDMS) molds were fabricated to meet the experimental requirements with the following structural specifications. Mold A was made of MNs that had a conical tip of tapered edges, with the center-to-center distance between two adjacent MNs equal to 750 μm, the base diameter of the tip of 330 μm and the tip height of 750 μm. Mold B is modified on the basis of Mold A by putting four grooves on the lateral surface of each conical tip, where each groove is oriented from the base plate toward the MN tip. Each groove is 3 μm in width and 500 μm in length, and they do not reach to the MN apex to keep the mechanical strength of the tip. Mold C has the identical center-to-center distance between MNs as that in Mold A, but the tips are engineered to have an enlarged base diameter of 430 μm and height of 1000 μm. All three molds have a rectangular base plate 7.5 mm in length and in each plate there is an array of 81 MNs arranged in a 9 × 9 pattern, and for Mold A and Mold B there is liquid drainage grooves in the lateral sides of the base plates.

Preparation and characterization of the microneedle patch

The boiling point was modulated, and the volatilization of liquid during the MN semi-solidification was reduced by mixing EA and BA at a volume ratio 1:1 on the basis of Raoult law, where bubble was not formed. PLGA was dissolved in this binary solvent system and a 10% (w/v) solution was prepared. This was mixed with a 10% (w/v) GEL solution at a volume ratio 9:1 to afford a homogenous suspension; this was kept in appropriate conditions before use.

The pure PLGA solution and the PLGA–GEL mixed suspension were each diluted to a final concentration of 5% (w/v). Aliquots of the respective mixtures were dispensed into Mold A and Mold B for MN fabrication. Entrained air bubbles were eliminated by defoaming at room temperature by centrifugation at 3000 rpm for 15 min; this procedure was carried out 2-3 times to deaerate completely. In the PLGA–GEL system, the centrifugal force also promoted enrichment of the GEL-containing aqueous phase at the MN tips. The filled molds were left undisturbed at room temperature overnight, and additional PLGA solution was intermittently replenished to compensate for the solvent evaporation and volume loss. After demolding, the resulting MN arrays were briefly immersed in an 80°C water bath, by which the GEL within the tip regions of PLGA–GEL-derived MNs was selectively leached out via phase separation to generate porous architectures. Poly(lactic-co-glycolic acid) MNs (PLGA–MNs) and grooved PLGA MNs (gPLGA–MNs) were prepared successfully by using this protocol.

The prefabricated VEGF@mSF and mSF hydrogels were also cast into Mold C and were centrifuged at 3000 rpm for 5 min for uniform spreading and also for some preliminary deaeration. The PLGA–MNs and gPLGA–MNs were put into the filled molds, and then a second centrifugation step was performed in order to make sure that the MNs are in intimate contact with the hydrogel matrix and also to remove the excess liquid as well as the residual air bubbles. The molds were then irradiated with ultraviolet (UV) in order to have the photocrosslinking of the hydrogel, and the MN-hydrogel constructs were demolded to have the final MN assemblies.

Characterization of the physicochemical properties of the VEGF@mSF/gPLGA–MNs

For fluorescence visualization, RB-labeled MN cores and sodium fluorescein-labeled MN shells were fabricated following the procedure described above. All the samples were morphologically characterized by scanning electron microscopy (SEM; ZEISS Sigma 300) and imaged by confocal laser scanning microscopy (CLSM; Leica TCS SPB) to assess the spatial distribution of the labeled components. To evaluate the mechanical properties, compression tests were performed on the MN samples using an electronic universal testing machine (CMT6104; Meister Industry, Zhongshan, China) at a crosshead speed of 0.2 mm·min⁻¹. In the skin puncture experiment, the MNs were used to puncture the skin of normal C57 mice; photographs of the punctured sites were acquired, and the excised tissues were subsequently stained with hematoxylin and eosin (H&E) for histological evaluation.

VEGF release from the VEGF@mSF/gPLGA–MNs

In order to study the release behavior of VEGF from the MN patches, in vitro release assays were performed by immersing two MN arrays separately in PBS and 1 mM hydrogen peroxide (H₂O₂) at 37°C with constant stirring. After predetermined time intervals, 500 μL of the supernatant was taken from each well, and an equal volume of fresh PBS was added so as to have constant release volume. The concentration of VEGF in the collected supernatants was determined with an ELISA kit. H₂O₂ medium (1 mM) was used to mimic the in vivo microenvironment with hypoxia as well as high level of ROS. So, with this medium we can evaluate the release of VEGF from the MNs in the presence of oxidative stress conditions.

For the in vivo studies, the MN patches were put over the dorsal skin overlying the fat transplantation site in mice. At specific times in 7 days after the implantation, the transplanted adipose tissues were taken out and homogenized. The concentration of VEGF in tissue extract was quantified with an ELISA kit following the instructions of the manufacturer. We applied this approach for the quantification of the local retention and accumulation of VEGF in the grafted adipose tissue.

Computer simulations of drug release and oil adsorption

The simulations were done with COMSOL Multiphysics (COMSOL Inc., Burlington, MA, USA). The MN core was modeled as a conical geometry with tapered edges, a base diameter of 330 μm, a height of 750 μm and four longitudinal grooves (each of 3 μm wide and 500 μm long) on the lateral surface. An outer shell encapsulates a drug-coated layer. The shell base diameter and total height is increased to 430 and 1000 μm, respectively, forming a MN with a nested core–shell configuration.

Drug release simulations were conducted using the Transport of Diluted Species module to characterize VEGF diffusion from the shell into the surrounding adipose tissue under diffusion-dominated conditions without convective transport. The initial concentration of VEGF was set to 1 in the shell domain and 0 elsewhere. A no-flux boundary condition was applied to the top surface of the patch to prevent upward mass loss, while the outer boundary of the tissue domain was defined as an open boundary to approximate an infinite diffusion domain.

The equation governing drug transport is described by Fick’s second law:

where is the normalized drug concentration and is the diffusion coefficient, which is defined as spatially dependent across different domains.

Specifically, the diffusion coefficients were set as follows: and for both simulated microenvironmental conditions, while that of the surrounding tissue was . A transient solver was used to compute the spatiotemporal evolution of the concentration fields. The cumulative drug release fraction was defined as follows:

where represents the spatially averaged concentration within the coating domain at time .

To quantify depth-dependent VEGF distribution and delivery efficiency within tissue, three representative axial sampling regions beneath the patch were defined for concentration profile extraction: top (500 μm beneath the patch), middle (1000 μm, aligned with the MN tip), and bottom (1500 μm, at the tissue base).

Capillary uptake of lipids from the tissue into the MN structures was simulated using the Laminar Flow module coupled with the Level Set method, enabling resolution of multiphase nonlipid interfaces. The system was constructed under an axisymmetric assumption and consisted of three computational domains: a lipid reservoir, the solid MN structure, and the surrounding nonliquid phase.

Initially, the lipid reservoir was fully filled with oil, while the outer MN channel was occupied by air. The initial velocity field in all fluid domains was set to zero ().

The interface between the lipids and air was tracked using the Level Set function , defined as follows.

The interface evolution is governed by the following:

where is the interface thickness defined by the mesh size and is the reinitialization parameter set to the maximum characteristic velocity of the system.

Fluid motion is described by the incompressible Navier–Stokes equations:

where lipid properties were defined as and .

The surface tension forces at the interface are expressed as follows:

where is the surface tension coefficient, is the curvature, and is the Dirac delta function localized at the interface.

Wetting behavior at solid boundaries was incorporated through a contact angle-dependent boundary force, as follows.

To distinguish capillary performance between MN designs, distinct contact angles were assigned as follows: grooved MNs , smooth MNs , and neutral wetting surfaces .

Boundary conditions were defined as follows: the bottom reservoir was assigned hydrostatic pressure with a fixed lipid phase (), while the top boundary was set to atmospheric pressure (). Gravity coupled with surface tension drove spontaneous upwards lipid migration along the MN surface until force equilibrium was reached.

The temporal evolution of lipid transport was quantified by tracking the interface displacement, penetration height, and phase volume fraction within the MN-adjacent region, enabling direct comparison of wicking efficiency between grooved and smooth structures.

Biocompatibility of VEGF@mSF/gPLGA–MNs

3 T3-L1 preadipocytes were cultured in high-glucose DMEM supplemented with 10% FBS and 1% penicillin–streptomycin at 37°C with 5% CO₂, and the medium was changed every 2–3 days until the cells reached confluence. At 48 h after reaching confluence (Day 0), the cells were induced with induction medium A (DMEM supplemented with 10% FBS, 0.5 mmol·L⁻¹ IBMX, 1 μM DEX, and 10 μg·ml⁻¹ insulin) for 48 h, after which induction medium B (DMEM supplemented with 10% FBS and insulin) was added for 48 h of incubation, followed by maintenance medium (DMEM supplemented with 10% FBS), which was replaced every 2 days. On Day 8, adipogenesis was assessed by Oil Red O staining, which was performed by first fixing the cells with 4% paraformaldehyde (PFA), staining with 0.5% Oil Red O, and visualization under an inverted phase-contrast microscope. The lipids were eluted with isopropanol, and the optical density (OD) was measured at 510 nm.

The in vitro biocompatibility of the MN extracts was assessed using human umbilical vein endothelial cells (HUVECs) and differentiated 3 T3-L1 adipocytes. MN extracts were prepared by immersing one patch in 2 ml of DMEM at 37°C for 24 h with stirring, followed by filtration and sterilization. HUVECs or adipocytes were seeded at 2 × 10⁴ cells/well in 96-well plates and cultured in complete DMEM or MN-conditioned medium. Cell growth was evaluated by a CCK-8 assay, which included measurement of the OD at 1, 3, and 5 days, and cell viability was assessed by live/dead staining on these same days.

To evaluate in vivo biocompatibility, peripheral blood samples were collected from mice 1 week after autologous fat transplantation for routine hematological analysis, and heart, liver, spleen, lung, and kidney tissues were harvested for H&E staining to assess potential adverse effects on hematopoietic parameters and major organs.

Microneedle degradation in vivo

To investigate the in vivo degradation behavior of the gPLGA–MNs cores, MN patches were implanted into the autologous fat graft sites of the mice. The gPLGA–MNs cores were taken back from the graft tissues at defined time points (Days 0, 1, 2, 3, 4, 5, 6, and 7) after the implantation. The samples were washed with PBS to remove the biological tissue remaining in the graft and then dried at room temperature until they reached constant weight. The remaining mass of each sample was determined with an analytical balance, and the residual mass percentage was given with respect to the initial mass. The degradation profile of the gPLGA–MNs cores was determined on the basis of the daily changes of the residual mass.

The gPLGA–MNs cores retrieved on Day 7 were rinsed with PBS, dried in air, and sputter-coated with gold before SEM observation in order to study the structural stability of the MN cores after implantation. In particular, the changes in the surface morphology, groove structure and the whole architecture of the gPLGA–MNs cores after in vivo implantation were observed by SEM.

Directional lipid adsorption by the microneedles in vitro

Lipid fluid was prepared by mixing the oil fraction isolated from discarded human adipose tissue (obtained following liposuction) with Oil Red O staining solution at a volume ratio of 1:1. The oil fraction was collected by ultracentrifugation at 12 000 rpm for 15 min. The adipose-derived oil fraction contained triglycerides, free fatty acids, and other endogenous lipid components released from adipose tissue, thereby providing a lipid environment for evaluating lipid adsorption behavior that is more physiologically relevant than simplified dye-based systems. To evaluate directional lipid transport, the MN tip was positioned vertically and gradually lowered until it contacted the lipid fluid on a horizontal substrate. Upon contact, movement was stopped, and lipid uptake by PLGA–MN cores without grooves, grooved PLGA MN cores (gPLGA–MNs), and grooved modified silk fibroin MN cores (gmSF–MNs) was recorded using a digital camera. PLGA–MNs and gPLGA–MNs were used to evaluate the effect of groove architecture, whereas gmSF–MNs served as a groove-matched material control to investigate the contribution of PLGA-mediated lipophilicity to lipid adsorption and uptake. The lipid adsorption behaviors of the three groups were compared to assess the mechanisms underlying capillary-driven directional lipid uptake.

Lipid adsorption by the microneedles in grafted adipose tissue in vivo

To evaluate the lipid adsorption capacity of the MNs within grafted adipose tissue, MN patches were implanted into autologous fat grafts in mice and retrieved at predetermined time points (0, 0.5, 1, 2, 3, 4, 5, and 7 days post-surgery). After retrieval, the MNs were gently rinsed with PBS to remove loosely attached debris and air-dried.

The distribution of lipids on the MN surface was first qualitatively visualized using Oil Red O staining under optical microscopy, providing direct morphological evidence of lipid accumulation on both the grooved and non-grooved structures.

For quantitative analysis, the MN cores were isolated and immersed in 2 ml of n-hexane, followed by gentle shaking for 24 h to ensure complete extraction of the adsorbed lipids. The extracts were collected, and the lipid content was quantified by infrared spectroscopy on the basis of the C–H stretching adsorption band (2800–3000 cm⁻¹) using a triglyceride standard curve for calibration. The lipid mass was calculated from the measured concentration, and the corresponding volume was derived assuming a lipid density of approximately 0.9 g/ml. Blank MNs were processed in parallel as controls, and all measurements were performed in triplicate.

Immunomodulatory capacity of the grooved poly(lactic-co-glycolic acid)–microneedles

A lipid extract from discarded human adipose tissue was filtered, sterilized, deposited in a culture dish, and then subjected to adsorption by the gPLGA–MNs. DMEM containing serum was added to establish a lipid exudate microenvironment, into which RAW264.7 macrophages were seeded. After incubation, the cells were harvested, stained with anti-CD206 and anti-CD86 antibodies, washed to remove unbound antibodies, and analysed by flow cytometry to determine the M2/M1 polarization ratio.

Protective effect of microneedles against oxidative stress damage in adipocytes in vitro

To examine the protective effect of the MNs, we first evaluated adipocyte viability under varying H₂O₂ concentrations to identify a range that induces moderate oxidative stress without excessive cytotoxicity. An in vitro oxidative stress injury model was established using a CCK-8 assay and live/dead cell staining to determine the optimal H₂O₂ concentration for induction. Differentiated 3 T3-L1 adipocytes were then treated with MN extracts; survival was assessed by live/dead staining, and intracellular ROS levels were detected via DCFH-DA staining. The ROS clearance rate was determined by flow cytometry. Finally, the fluorescence intensity was quantified using ImageJ software, and the relative fluorescence intensity was normalized to that of the control group.

Antioxidant effect of the microneedles on grafted adipose tissue in vivo

To evaluate the oxidative stress microenvironment within grafted tissue, the level of ROS in the transplanted adipose tissue was quantitatively assessed at predetermined time points (0, 0.5, 1, 2, 3, 4, 5, and 7 days post-surgery). Briefly, fresh adipose tissue samples were harvested and immediately immersed in a 10 μM DCFH-DA probe solution (in serum-free DMEM) for 30 min at 37°C in the dark. After thorough washing with PBS to remove excess probe, the samples were homogenized, and the resulting supernatants were collected. The fluorescence intensity of 2′,7′-dichlorofluorescein (DCF), which is indicative of intracellular ROS levels, was measured using a microplate reader with excitation and emission wavelengths of 488 and 525 nm, respectively. The protein concentrations in the supernatants were determined using a BCA assay for normalization of the fluorescence intensities.

Mitochondrial function in adipocytes after microneedles-mediated protection against oxidative stress injury

To evaluate the effect of the MNs on mitochondrial function under oxidative stress, differentiated 3 T3-L1 adipocytes were pretreated with MN extracts and subsequently exposed to H₂O₂ to induce injury.

The MMP was measured by JC-1 staining (5 μg·ml⁻¹, 20 min, 37°C, dark), after which the cells were washed with PBS and imaged with a confocal laser scanning microscope (ex/em: 488/530 nm for green monomers and 561/590 nm for red aggregates). The MMP was evaluated by determining the red/green fluorescence intensity ratio.

Mitochondrial morphology and distribution were assessed by MitoTracker Red CMXRos staining (100 nM, 30 min, 37°C, dark) and confocal imaging (ex/em: 579/599 nm) and by transmission electron microscopy (TEM) after fixation with 2.5% glutaraldehyde, post-fixation with osmium tetroxide, dehydration, and embedding. Ultrathin sections (70 nm) were stained and examined at 80 kV to evaluate cristae integrity and morphology. Quantitative analysis of the JC-1 red/green fluorescence ratio was performed using ImageJ, and the data were used to evaluate the protective effect of MNs on mitochondrial integrity and function under oxidative stress.

Angiogenic properties of the microneedle patches in vitro

In the cell scratch assay, HUVECs were seeded in 6-well plates. Once the cells approached 90% confluence, a straight scratch was introduced in the center of each well with a sterile 1 ml pipette tip. The medium was then replaced with fresh complete DMEM containing MN extracts, and co-culture was performed for 12, 24, and 48 h. Cell migration and proliferation were imaged at 0, 12, 24, and 48 h. Closure of the scratch gap was quantified using ImageJ software.

In the in vitro tube formation assay, 100 μL of Matrigel was aliquoted into each well of a 24-well plate and allowed to polymerize at 37°C for 30 min. HUVECs were then seeded onto the gel at 2 × 10⁴ cells per well in serum-free DMEM containing MN extracts, with serum-free medium alone as a control. After 8 h of incubation at 37°C with 5% CO₂, capillary-like structures were visualized under a fluorescence microscope, and the number of branching points was quantified using ImageJ software.

In the Transwell migration assay, 100 μL of a HUVEC suspension (1 × 10⁵ cells) in DMEM supplemented with 10% FBS and 1% P/S was added to the upper chamber of a Transwell insert. The chambers were placed in the wells of a 24-well plate containing 500 μL of the respective MN extracts, with standard complete medium as the control. After being incubated at 37°C in 5% CO₂ for 24 h, the cells were fixed, stained with crystal violet solution, and photographed. The number of migrated cells was quantified by analysing the images with ImageJ software.

Effect of the microneedle patch on autologous fat graft survival in mice in vivo

All animal procedures followed the National Institutes of Health guidelines and were approved by Nanchang University. Eight-week-old male C57BL/6 mice were randomly divided into four groups (n = 20): CON (no patch), mSF–MNs, mSF/gPLGA–MNs, and VEGF@mSF/gPLGA–MNs. Under isoflurane anesthesia, the dorsum was shaved and depilated. Autologous adipose tissue was harvested from the inguinal region, minced, and washed with PBS. A standardized 120 mg graft was prepared and implanted into a single subcutaneous pocket created in the midline of the dorsum, after which the incision was closed with 6–0 polypropylene sutures. Postoperative analgesia was given as appropriate. Immediately after surgery, the MN patches specific to each group were applied to the graft site; the CON group did not receive a patch. The patches were removed after one week under brief anesthesia.

At 1, 4, and 12 weeks after surgery, six mice per group per time point were randomly selected and euthanized, and the grafted fat tissue with the overlying skin was excised. The samples were weighed and then fixed in 10% neutral buffered formalin for subsequent paraffin embedding. At 4 and 12 weeks, high-frequency ultrasound was performed before euthanasia to evaluate fat retention and vascularization. Sections were prepared for H&E staining and immunohistochemical analysis.

Histology and immunofluorescence staining

Fixed fat and skin tissues were processed for histological and immunofluorescence analyses. The tissues were first embedded in paraffin and cut into 5-μm sections for H&E staining and Masson’s trichrome staining.

For immunofluorescence staining, the sections were dewaxed, rehydrated through a graded ethanol series, and subjected to heat-induced antigen retrieval in citrate buffer (pH 6.0) for 5 min. Nonspecific binding was blocked with 5% bovine serum albumin for 30 min. The sections were then incubated with primary antibodies at 4°C overnight, followed by incubation with fluorophore-conjugated secondary antibodies at room temperature for 1 h. The nuclei were counterstained with 4′,6-diamidino-2-phenylindole (DAPI) for 5 min, and images were captured using a fluorescence microscope. Relative fluorescence intensities were quantified with ImageJ and normalized to those of the control group.

To assess the potential systemic toxicity of MN sustained-release tablets, major organs (heart, liver, spleen, lung, and kidney) were collected 1 week after administration and subjected to histological analysis after staining with H&E.

Transcriptome sequencing and data analysis

Total RNA was isolated with TRIzol reagent (Invitrogen, CA, USA) in strict accordance with the supplier’s instructions. The purity and concentration of the RNA samples were measured using a NanoDrop 2000 spectrophotometer (Thermo Scientific, USA), while RNA integrity was examined with an Agilent 2100 bioanalyzer (Agilent Technologies, Santa Clara, CA, USA). Libraries were prepared using the VAHTS Universal V6 RNA-seq Library Prep Kit in accordance with the vendor’s standard protocol. All sequencing and downstream bioinformatics processing data were performed by OE Biotech Co., Ltd (Shanghai, China).

The libraries were sequenced on an Illumina NovaSeq 6000 system to yield 150 bp paired-end reads. Initial processing of the raw FASTQ files was performed with fastp to filter out low-quality sequences, yielding in high-confidence clean reads. Clean reads were mapped to the reference genome using HISAT2; FPKM values for each gene were computed, and raw read counts were obtained with HTSeq-count. Biological consistency among replicates was assessed by principal component analysis (PCA) implemented in R (v3.2.0).

Differentially expressed genes (DEGs) were subjected to Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses on the basis of the hypergeometric distribution. Significant terms were identified using R (v3.2.0), and graphical representations, including bar graphs, chord plots, and bubble charts, were produced in the same environment.

GSEA was carried out with GSEA software and a predefined gene set. Genes were ordered by the magnitude of differential expression between the two sample types, and notable enrichment of the gene set at the upper or lower extremes of the ranked list was evaluated.

Western blot analysis of grafted adipose tissue

To evaluate the expression of proteins associated with lipid metabolism, angiogenesis, and inflammatory signaling, grafted adipose tissue was harvested 7 days after autologous fat transplantation. Briefly, tissues were homogenized in RIPA lysis buffer containing protease and phosphatase inhibitors and centrifuged to collect the supernatants. Protein concentrations were determined using a BCA assay.

Equal amounts of protein were separated by SDS–PAGE and transferred to PVDF membranes. After blocking with 5% nonfat milk for 1 h at room temperature, the membranes were incubated overnight at 4°C with primary antibodies against PPARγ, VEGF, phosphorylated p65 (p-p65), total p65, and β-actin. After being washed, the membranes were incubated with HRP-conjugated secondary antibodies for 1 h at room temperature. The protein bands were visualized using an enhanced chemiluminescence (ECL) system.

The band intensities were quantified using ImageJ software and normalized to those of β-actin. Activation of the NF-κB pathway was evaluated by calculating the ratio of p-p65 to total p65. All experiments were performed in triplicate.

Statistical analysis and reproducibility

The experiments were performed in triplicate. The data are presented as the mean ± SD. Statistical analyses were performed using GraphPad Prism (v8.0 or v9.5.1). Student’s t-test was used to analyses differences between two independent groups; one-way or two-way analysis of variance with Tukey’s post hoc test was used to analyse differences among multiple groups. ImageJ was used to quantify the photographs and SEM images using pixel-to-mm/μm calibration. Graphs and figures were created with GraphPad Prism and BioRender. Statistical significance was defined as P < 0.05, indicating that the results were unlikely due to chance. This approach ensured robust, reproducible evaluation of the experimental outcomes across datasets and visual representations.

Results

A synergistic macrophage–inflammation–lipid axis in lipid-associated pathologies

In order to address what the common pathophysiological mechanisms underlying lipid-related disorders are, we first analysed gene expression profiles from three representative lipid-associated pathologies: steatohepatitis (GSE33814), atherosclerosis (GSE100927), and pancreatitis (GSE194331). The heatmap analysis indicated clearly different gene expression signatures to segregate the control and the disease group in the three datasets (Figure 2a). Pathway enrichment analysis showed that the three diseases have overlapping inflammatory pathways (e.g. NF-κB and IL17) and metabolic pathways (e.g. lipid metabolism), which were accompanied by the highly upregulated disease-associated genes (Figure 2b). We also detected disease-specific alterations in lipid metabolism: non-alcoholic steatohepatitis (NASH) is characterized by suppression of fatty acid degradation and bile secretion pathways; atherosclerosis shows prominent enrichment of the “lipid and atherosclerosis” pathway; and pancreatitis exhibits concomitant activation of both the “lipid and atherosclerosis” and “non-alcoholic fatty liver disease” pathways. All these findings illustrate that lipid dysregulation represents a shared feature in the pathogenesis of these disorders, pointing to potential common therapeutic targets.

Figure 2.

Bioinformatics analyses, including heatmap visualization, KEGG enrichment and correlation analyses, revealed the expression characteristics and interactions of genes related to lipid metabolism, inflammation and macrophage function in multiple lipid-associated diseases. (a) DEGs are visualized in heatmaps for three lipid-related disease cohorts: nonalcoholic steatohepatitis (GSE33814), atherosclerosis (GSE100927), and pancreatitis (GSE194331), with darker shades indicating higher expression levels, respectively. (b) Subsequent KEGG enrichment analysis highlighting dysregulated signaling pathways, depicted using a color gradient corresponding to the adjusted P values. (c) Correlation heatmap illustrating connections among lipid metabolism, inflammation, and macrophage markers; here, the circle size reflects the magnitude of the correlation, and the asterisks indicate statistical significance

To quantify the interdependencies among key phenotypic axes, we performed ssGSEA to calculate macrophage infiltration, inflammatory activity, and lipid-related gene signature scores for individual samples in the three disease models. We did the correlation analysis to show that the level of lipids, the inflammatory status and the score of macrophage infiltration are strongly and significantly correlated in all the models (Figure 2c). The circular node in the correlation plot shows the strength of the pairwise correlation, while color denotes the correlation direction (range: –1 to 1), with the size and color of the node proportional to the effect size. For all three diseases, there are strong positive correlations between “lipid” and “inflammation”, “lipid” and “macrophage infiltration” and “inflammation” and “macrophage infiltration” (all P < 0.001, ^***).

To validate the synergistic macrophage-inflammation crosstalk on adipose tissue lipid-associated injury, we analysed the adipose graft samples acquired on Days 3 and 7 post-transplantation. ssGSEA of inflammatory activity and macrophage infiltration showed that both the inflammatory signature scores and macrophage infiltration scores were markedly elevated in the adipose grafts compared with those in the control group (Figure S2a). In agreement, pairwise correlation analysis across all adipose graft samples showed an extremely strong positive linear correlation between inflammatory activity and the macrophage infiltration score (Pearson’s R = 0.9, P < 2.2e-16) (Figure S2b), which verifies the tightly coordinated upregulation of inflammation and macrophage accumulation in lipid-rich adipose tissue after transplantation.

These results show that lipid imbalance induces inflammatory responses, allows for macrophage recruitment/activation, and worsens tissue injury, and that a synergistic macrophage-inflammation-lipid axis might be at play at the center of different lipid-related diseases.

Synthesis and characterization of the modified silk fibroin hydrogels

On the basis of the understanding of the abovementioned pathological axis, we have started to develop a multifunctional biomaterial platform which is able to respond to oxidative stress, to regulate inflammation and to promote tissue repair. First, APBA was covalently grafted on the methacryloyl-modified silk fibroin (SilMA) by an EDC/NHS-catalyzed amidation reaction. The product was then mixed with polyvinyl alcohol (PVA) to prepare the SilMA–APBA–PVA composite hydrogel. APBA quickly condenses with the hydroxyl groups on PVA to form dynamic borate ester bonds and induce the spontaneous gelation in 30 s at room temperature (Figure 3a and b). This simple and efficient synthesis approach allows the synergistic regulation of the functional group modification and the physical gel behavior of the material.

Figure 3.

Chemical synthesis, structural morphology, penetration performance, mechanical properties, and drug release behavior of the vascular endothelial growth factor-loaded modified silk fibroin/grooved poly(lactic-co-glycolic acid)–microneedles(VEGF@mSF/gPLGA–MNs). (a) Schematic illustration of modified silk fibroin (SF) chemical modification and hydrogel synthesis. (b) Photographs of silk fibroin methacryloyl (SilMA), SilMA–3-acrylamidophenylboronic acid (SilMA–APBA), and the hydrogel formed immediately after rapidly mixing SilMA–APBA with PVA. (c) ¹H nuclear magnetic resonance (NMR) spectrum of SilMA–APBA–PVA. (d) Fourier transform infrared (FTIR) spectra of PVA, SilMA, APBA, and SilMA–APBA–PVA. (e) X-ray photoelectron spectroscopy (XPS) survey scans of PVA, SilMA, APBA, and SilMA–APBA–PVA. (f) Representative image of the VEGF@mSF/gPLGA–MNs. (g) Scanning electron microscopy (SEM) images of gPLGA–MNs; porous structures formed after water bath treatment, and MNs with nested mSF shells. Scale var: 50 μm; 5 μm. (h) Confocal laser scanning microscopy (CLSM) images of MN cores labeled with sodium fluorescein (green) and shells labeled with rhodamine B (red) (RB). Scale bar: 100 μm. (i) Mechanical strength of the VEGF@mSF/gPLGA–MNs. (j) Representative images of VEGF@mSF/gPLGA–MNs, their application to mouse skin, and haematoxylin and eosin (H&E) staining of local skin sections from mice after MN application, demonstrating successful penetration into the dorsal skin. Scale var: 25 μm; 100 μm. (k) Cumulative release of VEGF under physiological conditions and in the presence of high concentrations of H₂O₂. (l) Time-dependent changes in the relative reactive oxygen species (ROS) level in the tissue, cumulative release of VEGF, degradation rate of the mSF shell, and oil adsorption rate of the VEGF@mSF/gPLGA–MNs over a 7-day treatment cycle, revealing the sequential cascade therapeutic profile of the composite MNs

Spectroscopic analyses confirmed that APBA worked as a cross-linker, covalently bridging the PVA and SilMA to form a network stabilized by hydrogen bonds and ROS-sensitive borate ester bonds (SilMA–APBA–PVA). The ¹H NMR spectrum (Figure 3c) of SilMA–APBA–PVA shows that the aromatic proton peaks from the benzene ring of APBA appear at δ 7.0–8.0 ppm and the characteristic hydroxyl proton signals (δ 4.5–5.5 ppm), typical of the boronic acid-diol complexes, were detected, confirming that APBA was incorporated and the borate ester were formed. The FTIR spectra are shown in (Figure 3d). A characteristic peak corresponding to the ester carbonyl group (C=O) of the methacryloyl moiety is observed at approximately 1720 cm⁻¹ in the SilMA spectrum, whereas the amide I and II bands of the silk fibroin backbone are located at approximately 1640 cm⁻¹ and 1530 cm⁻¹, respectively. In the spectrum of the SilMA–APBA–PVA composite, a peak attributed to the B-O stretching vibration of APBA appears at approximately 1340 cm⁻¹. Also, a broad and intense O-H adsorption band emerged within the 3200—3500 cm⁻¹ range, which indicates the possibility of the incorporation of PVA and the formation of a cross-linked network facilitated by the formation of boronate ester and hydrogen bonds. XPS survey scans (Figure 3e) showed a B 1 s signal (190—195 eV). The presence of the B 1 s signal in the SilMA–APBA–PVA composite confirms the successful grafting of APBA, and its retention verifies the integration of APBA in the PVA-based network. In conclusion, the NMR, FTIR and XPS results confirmed the successful synthesis of the SilMA–APBA–PVA composites.

Preparation, characterization, and functional validation of the VEGF-loaded modified silk fibroin/grooved PLGA core–shell microneedles

On the basis of the above hydrogel platform, we designed and prepared a core–shell structured MN patch for functional integration and targeted therapy, i.e. for simultaneous growth factor delivery and local lipid management to improve the microenvironment after fat transplantation and to promote graft survival. In brief, the patch is composed of a mSF shell loaded with VEGF and a porous, grooved PLGA core forming the composite VEGF@mSF/gPLGA–MNs structure (Figure 3f and g).

The morphological characterization showed that the MN array was arranged in a regular 9 × 9 array with a uniform spatial distribution (Figure 3f). SEM (Figure 3g) showed that the tip of the gPLGA core adopted a porous honeycomb structure after it was treated in a water bath, while after it was encapsulated by the mSF shell, the core were very sharp and uniform conical MNs with the needle height of about 1000 μm and the base diameter of about 430 μm, and the structure was very good. To check the fabrication reproducibility, the gPLGA–MNs prepared in the independent fabrication batches were examined by SEM (Figure S3c). In general, the MN morphology, the groove architecture, and the porous tip structures were comparable in the batch, which indicates the good manufacturing reproducibility and the structural uniformity. To check the core-shell structure, the fluorescent dyes sodium fluorescein (green) and rhodamine (red) were encapsulated in the shell and core, respectively, and the cross sections were observed by CLSM (Figure 3h). The results showed that the shell uniformly covered the core, so that the subsequent controlled drug release and the mechanical stability could be ensured.

From the mechanical tests, it was shown that a single MN can support a maximum compressive force of 1.13 N, which is far beyond the minimum threshold of 0.1 N for the skin to be punctured (Figure 3i). After being applied to mouse skin, the MN array became completely embedded in the tissue, resulting in the formation of regular microchannels (Figure 3j). The analysis of the histological section confirmed that the drug has been effectively delivered into the epidermis and dermis. In vitro drug release experiments showed that the MNs that we used with H₂O₂ will have a faster rate of VEGF release in the first stage, while the control group that we have not treated has more persistent sustained release behavior (Figure 3k). This stimulus-responsive release characteristic comes from the ROS-sensitive phenylboronic acid bonds in the mSF shell, and this system will enable on-demand controlled drug release kinetics.

To determine whether the structural integrity of the MN system could be maintained after the implantation to exert the designed therapeutic functions, the in vivo stability of the gPLGA core, the lipid adsorption capability, the ROS regulation capacity and the function dynamics of the MN system were systematically investigated. As shown in (Figure S3a), the gPLGA–MNs had negligible mass loss over the whole 7 days of the observation. This means the gPLGA has slow degradation kinetics and excellent structural stability under physiological conditions. The SEM shows that the architecture of the MN is still kept on Day 7 without evident collapse or structural damage (Figure S3b). It means the gPLGA core can provide a stable structural platform in the very important early stage after the fat transplantation. The lipid removal capacity of the MN core was then evaluated in vivo. When the MN cores were retrieved 7 days after the implantation, the surfaces of the MN cores showed a clear amount of lipid accumulation, especially in the region of the tip and the shaft (Figure S4a). In a quantitative analysis, we found that the oil was adsorbed more by the gPLGA–MNs than by the non-grooved PLGA–MNs (Figure S4b). Thus, we can conclude that the porous tip and the grooved sidewall architecture really help the lipid sequestration and the retention in the graft microenvironment.

In order to know whether the MN system could regulate the oxidative stress in the microenvironment and therefore exert the designed sequential therapeutic functions after implantation, we quantified the tissue ROS levels in vivo. As shown in (Figure S5), the ROS level in the control group increased much more shortly after transplantation, while VEGF@mSF/gPLGA–MNs markedly attenuated the ROS accumulation and accelerated the ROS level normalization, which means that the mSF shell scavenges the ROS. In order to correlate these biological effects with material behavior, we analysed the dynamic change of the ROS levels, VEGF release, shell degradation and lipid adsorption (Figure 3l). It can be observed that the mSF shell underwent rapid degradation during the first 3 days, and it coincides with the sustained VEGF release and effective suppression of ROS accumulation. After shell degradation, the lipid adsorption by the exposed gPLGA core progressively increases and becomes the main function of the system. These findings reveal a programmed spatiotemporal transition from early-stage ROS scavenging and angiogenesis promotion to late-stage lipid removal and microenvironmental remodeling, enabling sequential intervention against multiple pathological processes that compromise fat graft survival.

Biocompatibility of the VEGF@mSF/gPLGA–MNs

Once we confirmed the functional properties of the materials, we evaluated the biocompatibility of the VEGF@mSF/gPLGA–MNs system systematically at the cellular, blood and systemic levels. After in vitro induction and differentiation, the 3T3L1 cells showed in the cytoplasm many transparent, very refractive lipid droplets; after Oil Red O staining (Figure S6), these droplets were bright red granular deposits, which is a direct reflection of the extent of the adipogenic differentiation.

After 1, 3 (Figure 4a and b) and 5 (Figure S7a and b) days of exposure to mSF–MNs, gPLGA–MNs, or mSF/gPLGA–MNs extracts, live/dead staining showed that both 3 T3-L1-induced adipocytes and HUVECs still kept the normal morphology as those of the control groups. The cells were displaying intense green fluorescence (living cell) and little red fluorescence (dead cells), meaning of negligible acute cytotoxicity. Quantitative CCK-8 assays (Figure 4c and d) also showed no significant different of relative survival of the two kinds of cells on Days 1, 3, and 5, and the cell viability is still high throughout the observation period, thus confirming the minimal short-term cytotoxic effects.

Figure 4.

Cell viability, proliferation, and hemolysis assays and in vivo histological and routine blood tests demonstrating the biocompatibility and biosafety of the composite MNs. (a and b) Fluorescence microscopy images of live/dead staining of 3T3-L1 adipocytes and HUVECs in the control (Ctrl), modified silk fibroin microneedles (mSF–MNs), grooved poly(lactic-co-glycolic acid)–microneedles (gPLGA–MNs), and modified silk fibroin/grooved poly(lactic-co-glycolic acid)–microneedles (mSF/gPLGA–MNs) groups on Days 1 and 3. Scale bar: 100 μm. (c and d) Relative proliferation rates of 3 T3-L1 cells and human umbilical vein endothelial cells (HUVECs) in the Ctrl, mSF–MNs, gPLGA–MNs, and mSF/gPLGA–MNs groups on Days 1, 3, and 5. (e) Hemolysis rates in the Ctrl, negative control (NC), mSF–MNs, gPLGA–MNs, mSF/gPLGA–MNs, and VEGF@mSF/gPLGA–MNs groups. (f) Haematoxylin and eosin (H&E) -stained histological sections of the heart, liver, spleen, lung, and kidney, demonstrating biosafety in the Ctrl and VEGF@mSF/gPLGA–MNs groups for in vivo applications. Scale bar: 50 μm. (g) Complete blood count and serum biochemical analyses of mice in the Ctrl, mSF–MNs, mSF/gPLGA–MNs, and VEGF@mSF/gPLGA–MNs groups, including hematological indicators and liver and renal function parameters

All the tested materials (mSF–MNs, gPLGA–MNs, and mSF/gPLGA–MNs) showed a hemolysis rate close to that of the negative control (NC). Hemolysis assays (Figure 4e) showed that the hemolysis rate of the VEGF@mSF/gPLGA–MNs is comparable with that of the NC and well below the 5% safety threshold. These results confirmed that the incorporated VEGF and the other constituent materials have the excellent blood compatibility, i.e. the red blood cells (RBCs) are not damaged too much. These results support that the incorporated VEGF and the other constituent materials are suitable for the biomedical applications in which the contact with the blood is needed.

Systemic safety was evaluated by the histological analysis of the major organs (heart, liver, spleen, lung, and kidney) of control mice and those treated with mSF–MNs, mSF/gPLGA–MNs (Figure S8) and VEGF@mSF/gPLGA–MNs (Figure 5f). All the groups showed normal tissue architecture, without signs of inflammation, necrosis and pathological remodeling. H&E-stained sections revealed well-organized myocardial fiber structure in the heart, orderly hepatocyte arrangement in the liver, intact lymphoid follicles in the spleen, clearly defined alveolar architecture in the lung, and normal renal tubules and glomeruli in the kidney.

Figure 5.

Computational simulations of vascular endothelial growth factor (VEGF) release and lipid transport behavior within the core–shell MN system. (a) Computational simulations of drug release under H₂O₂/Ctrl conditions, showing time-lapse pseudocolor images of drug diffusion from loaded MNs. (b–e) Single MN and its top, middle, and bottom surface cross-sectional planes, with corresponding drug release curves for each plane. (f–i) Single MN core and its tip and groove segments, and liquid increase curves and velocities for each part. (j) Pseudocolor images of liquid height increases from the computational simulations, comparing grooved poly(lactic-co-glycolic acid)–microneedles (gPLGA–MNs) and poly(lactic-co-glycolic acid)–microneedles (PLGA–MNs). Statistical significance: ^*P < 0.05, ^****P < 0.0001

Hematological and serum biochemical profiling (Figure 4g) showed no statistically significant intergroup differences among the four cohorts (Ctrl, mSF–MNs, mSF/gPLGA–MNs, and VEGF@mSF/gPLGA–MNs) for all the detected parameters, both hematological indices [white blood cell (WBC), RBC, hematocrit (HCT), mean corpuscular hemoglobin (MCH), platelet (PLT), hemoglobin (HGB), lymphocyte (LYM), mean platelet volume (MPV), neutrophil (NEU), monocyte (MON)] and liver/kidney function markers [alanine transaminase (ALT), aspartate transaminase (AST), alkaline phosphatase (ALP), creatinine (Crea), and blood urea nitrogen (BUN)]. These results showed that all the MN formulations did not negatively affect hematopoietic function or systemic hepatic or renal physiological homeostasis in vivo.

Collectively, these results demonstrated the excellent cytocompatibility, hemocompatibility, and systemic biocompatibility of the mSF–MNs, gPLGA–MNs, mSF/gPLGA–MNs, and VEGF-loaded mSF/gPLGA–MNs, supporting their potential for future biomedical applications.

Computational and experimental evaluation of directional lipid adsorption and inflammation regulation by the gPLGA–MNs cores

Computational simulations are essential for elucidating the dynamic behaviors of drug delivery and oil adsorption by MN patches. By integrating the experimental data with the simulation results using COMSOL Multiphysics software, we obtained a comprehensive understanding of the system, thereby facilitating the informed optimization of device design and performance. These simulations focused on two key aspects. First, they captured the sustained and ordered release of the drug across different skin layers over time, which is critical for determining the changes in drug concentration during the responsive degradation of the drug-loaded shell; additionally, these concentration profiles were visualized using color maps of spatiotemporal transport dynamics. Second, the simulations investigated how the core structure, specifically, a grooved shaft with a porous tip (gPLGA–MNs) versus a smooth surface (PLGA–MNs), influenced oil adsorption via capillary action, leading to a progressive increase in the liquid level within the core over the corresponding time course. This comparative analysis highlighted the role of the core architecture in modulating adsorption kinetics and informed the rational design of MN systems for combined drug delivery and oil removal applications.

There was a pronounced difference between the H₂O₂ group and the control in simulations of drug release. As shown in (Figure 5a) (0 to 120 h), the H₂O₂ group showed very quick and large expanses of concentration gradients of the top, the middle and the bottom of the surface, while the control group displayed slower diffusion. (Figure 5b) is the schematic of a single MN and its constructed simulation model, which depicts the position of drug release that was detected and the diffusion way. Analyses of the surfaces (Figure 5c–e) revealed that the top surface in the H₂O₂ group exhibited a rapid early increase in drug concentration followed by a gradual decline, indicating rapid release into the tissue surrounding the midsection of the MN. Drug release in the control group is very low for most of the time of the observation. Similar to the trend of the middle and the bottom surfaces, the H₂O₂ environment can make the drug release and penetrate to the MN shell earlier and deeper, so both the superficial release and the bulk phase transport efficiency will be increased.

The simulations of the core structure showed how the different designs have an influence on the liquid adsorption and the mass transport. (Figure 5f) depicts the detailed MN architecture (needle shaft and tip), and (Figure 5j) presents the time series from 0 to 60 s to evaluate oil adsorption performance: for the gPLGA–MNs we see more efficient and rapid oil adsorption (evidenced by a broader range of color-rise coverage), while for the PLGA–MNs the uptake is more restricted and slower. These differences on adsorption behavior are in the same sense of the results of the quantitative data analysis. (Figure 5g) quantifies the increases in liquid height and velocity for both designs: gPLGA–MNs displayed significantly greater increases in height and speed, particularly when the liquid reached the grooved region of the needle shaft. In the segmented analysis (Figure 5h and i), we see that the rate of the liquid height increase also in the shaft and in the tip of the gPLGA–MNs is much greater than that of the PLGA–MNs, which means that the combination of the grooved shaft and the porous tip increases the liquid adsorption capacity. These data confirmed that gPLGA–MNs allow faster and more efficient oil adsorption and transport and is an effective design for a localized therapeutic application.

In order to study the functional mechanisms of this MN system in terms of the function of lipid management and regulation of inflammation, we paid attention to the directional oil adsorption capacity of the core of gPLGA–MNs and the regulating effect of the inflammatory microenvironment. The structure was analysed, in which regularly spaced grooves were found on the sidewalls of gPLGA–MNs, which formed the paths of the capillary-driven directional flow of the liquids, especially the lipid-containing liquids (Figure 6b).

Figure 6.

Lipid adsorption capacity of the MNs and their effects on macrophage polarization, cell viability and intracellular ROS levels under oxidative stress. (a) Time series images showing directional lipid adsorption by grooved poly(lactic-co-glycolic acid)–microneedles (gPLGA–MNs), poly(lactic-co-glycolic acid)–microneedles (PLGA–MNs), and grooved modified silk fibroin microneedles (gmSF–MNs) using a human adipose tissue-derived lipid fraction containing triglycerides, free fatty acids, and endogenous lipid components. (b) Schematic diagram showing the directed adsorption of oil red O-labeled lipids onto the MN core. (c and d) Mean time and mean speed of the increase in lipid height to the base plate for gPLGA–MNs and PLGA–MNs. (e) Schematic of gPLGA–MNs altering the RAW264.7 cell microenvironment via lipid adsorption. (f) Flow cytometry (CD86-PE/CD206-FITC) analysis of RAW264.7 cell polarization in the negative control (NC), Ctrl, and gPLGA–MNs groups. (g) Proportion of M1 macrophages. (h and i) Fluorescence microscopy images and bar graphs of live/dead staining of 3T3-L1 cells under H₂O₂ induction, showing cell viability in the Ctrl, mSF–MNs, gPLGA–MNs, and mSF/gPLGA–MNs groups. Scale bar: 100 μm. (j and k) Fluorescence microscopy images and bar graph of Hoechst/DCFH-DA staining of 3T3-L1 cells under H₂O₂ induction, showing intercellular reactive oxygen species (ROS) levels in the Ctrl, mSF–MN, gPLGA–MN, and mSF/gPLGA–MN groups. Scale bar: 100 μm. Statistical significance: ^*P < 0.05, ^**P < 0.01, ^***P < 0.001, ^****P < 0.0001

In order to better resemble the lipid-rich microenvironment that is experienced during the fat graft remodeling, a human adipose tissue-derived lipid fraction that contains triglycerides, free fatty acids, and endogenous lipid components was used, and the migration behavior was also dynamically followed using Oil Red O as a visual tracer (Figure 6a). Over the specified time intervals (0, 10, 20, 30, 40, and 50 s), both the gPLGA–MNs and PLGA–MNs groups exhibited a stepwise increase in liquid level height. Owing to the capillary effect generated by the groove structure, the liquid started to increase in the gPLGA–MNs group as early as approximately 10 s. At this time, the liquid moved upwards along the grooves, causing the entire liquid level to rise more rapidly. By 50 s, complete wetting of the substrate was achieved in the gPLGA–MNs group (bottom of the culture dish), whereas the PLGA–MNs group remained partially wetted. Quantitative analysis of the average time to contact the substrate (Figure 6c) confirmed that the gPLGA–MNs required significantly less time, indicating markedly accelerated adsorption. Similarly, gPLGA–MNs exhibited a greater average rate of liquid surface increase (Figure 6d), highlighting their superior adsorption kinetics. Furthermore, despite possessing an identical grooved architecture, the gmSF–MNs exhibited markedly less lipid adsorption than the gPLGA–MNs did during the observation period (Figure 6a). Moreover, to investigate directional liquid adsorption by the gPLGA–MNs core, we used water-soluble (oil-insoluble) sodium fluorescein and observed by fluorescence microscopy that the dye particles migrated toward the MN tips, indicating directional liquid uptake toward this region. Collectively, these findings indicate that both groove-mediated capillary transport and PLGA-mediated lipophilicity contribute to lipid adsorption, while their combination in gPLGA–MNs results in the highest adsorption efficiency.

To investigate the regulation of inflammation, gPLGA–MNs were used to adsorb SF/gPLGA–MNs exogenous lipids from a culture dish, and the treated dish was then used to culture the Raw264.7 macrophages (Figure 6e). Flow cytometric analysis of macrophage polarization markers, namely, CD86 (M1 phenotype) and CD206 (M2 phenotype), revealed that compared with that in the NC group, the lipids significantly increased the proportion of M1 macrophages (CD86⁺CD206⁻) in the control group. After gPLGA–MNs treatment, the proportion of M1-polarized macrophages was significantly reduced (Figure 6f and g). These findings indicate that the gPLGA–MNs core can actively modulate macrophage immune responses, suppress excessive inflammatory reactions, and thereby exert protective anti-inflammatory effects.

Modified SF/gPLGA–MNs scavenge ROS and protect mitochondria in vitro

In order to explore the in vitro ROS scavenging and mitochondrial protective effect of mSF/gPLGA–MNs, the 3T3L1 adipocytes were first exposed to a H₂O₂ concentration gradient to set up the oxidative stress model (Figure S9a and b). According to the results of the cell viability assays, 480 μM H₂O₂ was chosen as the induction dose in the following experiments.

The Live/dead staining performed in the presence of 480 μM H₂O₂ (Figure 6h) showed a higher cell death rate in the control group (H₂O₂ only) than in the NC group (untreated control). When the cells were treated with the extracts from the mSF–MNs, gPLGA–MNs and mSF/gPLGA–MNs, the H₂O₂-mediated cell death was markedly reduced, in particular for the cells in which the mSF–MNs and mSF/gPLGA–MNs extracts were used (Figure 6i). This shows that the mSF confer a potent cytoprotective activity, probably by scavenging the ROS. The intracellular ROS levels were quantified by the DCFH-DA fluorescence staining (Figure 6j). Compared with the NC group, the control group showed an elevated ROS level with stronger fluorescence intensity. This confirms the successful induction of oxidative stress. In contrast, mSF–MNs and mSF/gPLGA–MNs significantly reduced the intracellular ROS level (Figure 6k). This is in agreement with the decrease in cell death, and it supports a direct link between a reduction in the ROS level and cell survival under oxidative stress.

Further analyses of mitochondrial function by JC-1 staining (Figure 7a) showed that in the control group, there was a marked decrease in the JC1 aggregate/monomer ratio, which is a sign of loss of the MMP. The treatment with mSF–MNs, gPLGA–MNs, or mSF/gPLGA–MNs partially restored this ratio, in a more marked way for the mSF–MNs and the mSF/gPLGA–MNs groups; the fluorescence intensity distribution curves confirm that mSF-containing formulations afforded stronger MMP protection. Quantitative analysis (Figure 7f) showed that the JC1 aggregate/monomer ratios are significantly higher in the mSF–MNs and in the mSF/gPLGA–MNs groups than in the control group, so that they confirm that these formulations are able to preserve mitochondrial function.

Figure 7.

Multiple detection methods demonstrate that the composite MNs alleviate oxidative damage by regulating intracellular reactive oxygen species (ROS) levels, the mitochondrial membrane potential (MMP) and mitochondrial morphology. (a) Fluorescence microscopy images of 3T3-L1 cells stained with JC-1 following H₂O₂ induction. The images depict MMP levels in different treatment groups: Ctrl, modified silk fibroin microneedles (mSF–MNs), grooved poly(lactic-co-glycolic acid)–microneedles (gPLGA–MNs), and modified silk fibroin/grooved poly(lactic-co-glycolic acid)–microneedles (mSF/gPLGA–MNs) . Scale bar: 20 μm. (b) Flow cytometry results showing the ROS scavenging capacity of 3T3-L1 cells. (c) Confocal laser scanning microscopy (CLSM) images of 3T3-L1 cells co-stained with a mitochondrial tracker (MitoTracker) and 4′,6-diamidino-2-phenylindole (DAPI). These images illustrate mitochondrial morphology in H₂O₂-treated cells from each experimental group. Scale bar: 10 μm; 2 μm. (d) Quantitative analysis of cellular ROS scavenging levels based on flow cytometry data. (e) Transmission electron microscopy (TEM) images showing the ultrastructural features of mitochondria in 3T3-L1 cells after H₂O₂ treatment across different groups. Scale bar: 1 μm; 200 μm. (f) Quantitative analysis of the JC-1 aggregate-to-monomer ratio, which is an indicator of MMP, in each group. Statistical significance: ^*P < 0.05, ^**P < 0.01, ^***P < 0.001, ^****P < 0.0001

Flow cytometry (Figure 7b) revealed a rightward shift of the ROS fluorescence peak in the control group, which is consistent with ROS accumulation, whereas the mSF–MNs and mSF/gPLGA–MNs treatments shifted the peak leftward, indicating ROS clearance. From the statistics (Figure 7d) we see that these groups attained >70% ROS scavenging efficiency, i.e. they have a strong antioxidant capacity.

As can be seen from the MitoTracker staining (Figure 7c), the mitochondria in the NC group were intact and uniformly stained, while in the control group, after exposure to H₂O₂, the mitochondria were fragmented and weakly stained. The staining of the mitochondria was kept intact in the mSF–MNs and mSF/gPLGA–MNs, and the pattern of staining is very close to the NC group. This has been confirmed by TEM (Figure 7e): the mitochondria in the control group swell and the cristae are disrupted when the cells have been under the H₂O₂ stress, while the mitochondria in mSF-containing treatments alleviated the damage and better preserved the structural integrity of the mitochondria with the cristae well-defined.

To conclude, the mSF/gPLGA–MNs system demonstrated excellent ROS clearance capacity, preserved the MMP, and promoted cell survival in vitro, confirming its ability to effectively mitigate oxidative stress-induced cell damage and providing a potential strategy for treating oxidative stress-related diseases.

VEGF@mSF/gPLGA–MNs promote endothelial cell migration and angiogenesis in vitro

In order to test the in vitro angiogenic potential of the VEGF-loaded MNs, the HUVECs were treated with the extracts of the four formulations, mSF–MNs, gPLGA–MNs, mSF/gPLGA–MNs and VEGF@mSF/gPLGA–MNs.

Three angiogenesis-related assays were performed to evaluate the proangiogenic effects of these extracts. In the scratch wound healing assay, the VEGF@mSF/gPLGA–MNs extract promoted cell migration. In this group, the scratched area remained much smaller than the areas in the control and all non-VEGF-loaded groups at 12, 24 and 48 h after seeding (Figure 8a). Quantitative analysis of the wound closure rates also confirmed these results (Figure 8b). In the tube formation assay, HUVECs treated with the VEGF@mSF/gPLGA–MNs extract formed more extensive capillary-like networks and the greatest degree of branching complexity (Figure 8c). Compared with all the other groups, this group had significantly greater total tube length and node count (Figure 8d and e). It confirms that the delivered VEGF facilitate the endothelial network formation. Furthermore, in the Transwell migration assay (Figure 8f), the number of migrating cells that traversed the membrane was much greater in the VEGF@mSF/gPLGA–MNs group. Statistical analysis of the relative migration index (Figure 8g) shows a very significant increase, which indicates the potent efficacy of this platform in stimulating chemotactic migration.

Figure 8.

In vitro cell experiments confirmed that the MN materials significantly promoted the migration and angiogenic capacity of human umbilical vein endothelial cells (HUVECs). (a) Representative images of HUVECs scratch wound healing assays in the Ctrl, modified silk fibroin microneedles (mSF–MNs), grooved poly(lactic-co-glycolic acid)–microneedles (gPLGA–MNs), modified silk fibroin/grooved poly(lactic-co-glycolic acid)–microneedles (mSF/gPLGA–MNs) and VEGF-loaded modified silk fibroin/grooved poly(lactic-co-glycolic acid)–microneedles (VEGF@mSF/gPLGA–MNs) groups, showing wound closure from 0–48 h. Scale bar: 200 μm. (b) Quantitative analysis of the scratch wound closure rate (residual wound area, %). (c) Representative images from HUVECs tube formation assays for the Ctrl, mSF–MNs, gPLGA–MNs, mSF/gPLGA–MNs and VEGF@mSF/gPLGA–MNsgroups at 8 h. Scale bar: 100 μm. (d) Quantitative analysis of total tube length in the tube formation assay. (e) Quantitative analysis of the number of branching points in the tube formation assay. (f) Representative images from HUVECs transwell migration assays for the Ctrl, mSF–MNs, gPLGA–MNs, mSF/gPLGA–MNs and VEGF@mSF/gPLGA–MNs groups at 24 h. Scale bar: 100 μm. (g) Quantitative analysis of the number of migrated cells in the Transwell migration assay. Statistical significance: ^*P < 0.05, ^**P < 0.01, ^***P < 0.001, ^****P < 0.0001

In summary, we conclude that the VEGF@mSF/gPLGA–MNs system has the advantage in all three important vascularization processes, cell migration, lumen formation and chemotactic migration in vitro. These findings suggest its potential utility in promoting vascular regeneration and have laid the groundwork for its application in tissue repair and regenerative medicine.

VEGF@mSF/gPLGA–MNs increase fat graft retention and promote tissue remodeling

In order to explore the in vivo functionality of the VEGF@mSF/gPLGA–MNs, we set up the mouse autologous fat transplantation model and also did longitudinal assessments at 1, 4, and 12 weeks (which could represent the early, middle, and late postoperative stages, respectively). We included 4 treatment groups for comparison: the control, mSF–MNs, mSF/gPLGA–MNs, and VEGF@mSF/gPLGA–MNs. We systematically investigated the effects of such treatments on the fat graft survival, the structural integration, and the tissue remodeling. The gross observation of the dorsal skin almost immediately after transplantation shows that in all the groups there is a slight elevation at the graft site (Figure S10). The subsequent histological and imaging evaluations show the distinct trajectories of graft evolution among the experimental cohorts. Compared with the control group, the mSF–MNs, mSF/gPLGA–MNs, and VEGF@mSF/gPLGA–MNs groups exhibited progressive improvements in graft volume retention, vascularization, and extracellular matrix organization, which indicates that the MN-based treatments are associated with the improved graft outcomes.

Overall, the results of the morphological and ultrasound evaluations (Figure 9a) showed marked temporal changes in the appearance and the internal structure of the fat grafts in the groups. At 1 week post-surgery, the graft volumes were similar in all the groups. At 4 weeks the VEGF@mSF/gPLGA–MNs group had more tissue and was fuller; the newly formed blood vessels were clearly made part of the graft, the high-frequency ultrasound showed uniform internal echoes, and there was also a clear set of high-echo areas that seemed to be suggestive of viable perfused tissue. On the contrary, in the control and mSF–MNs groups, the echo was not uniform, and there were larger anechoic cystic spaces that corresponded to necrotic regions, which is consistent with the reduced viability of the tissue. At 12 weeks, the most abundant and compact adipose-like tissue had formed in the VEGF@mSF/gPLGA–MNs group; the boundaries of the graft were well-defined, and the structure was in the layer form that is consistent with mature and organized adipose regeneration. Compared with the pale and underdeveloped control group, both the VEGF@mSF/gPLGA–MNs and the mSF/gPLGA–MNs groups appeared darker and fuller, which also indicates an increase in angiogenesis and the vitality of the tissue. On the contrary, in the control group, the tissue volume was smaller, the internal structure was the most heterogeneous, and there were persistent anechoic cystic spaces, which means that in that group there was ongoing liquefaction and necrosis. Together, the macroscopic and ultrasound findings show that compared with no treatment and the single-component control, VEGF@mSF/gPLGA–MNs treatment improved fat graft volume retention, vascularization, and tissue maturity while the necrotic degeneration was also reduced.

Figure 9.

Gross observation, ultrasound detection and histological staining were used to assess the long-term retention and structural integrity of and collagen deposition in adipose grafts after different treatments in vivo. (a) Macroscopic and ultrasound images of transplanted adipose tissue at weeks 1, 4, and 12 in the Ctrl, modified silk fibroin microneedles (mSF–MNs), modified silk fibroin/grooved poly(lactic-co-glycolic acid)–microneedles (mSF/gPLGA–MNs), and VEGF-loaded modified silk fibroin/grooved poly(lactic-co-glycolic acid)–microneedles (VEGF@mSF/gPLGA–MNs) groups. (b) Changes in transplanted adipose tissue mass over time for each group. (c) Quantification of the cross-sectional area of transplanted adipose tissue at 4 weeks and 12 weeks across different treatment groups. (d) quantification of the ratio of the cystic area to the cross-sectional area of transplanted adipose tissue at 4 weeks and 12 weeks across different treatment groups. (e) Quantitative analysis of the area of collagen deposition in adipose tissue at weeks 1, 4, and 12. (f) Masson’s trichrome staining images of transplanted adipose tissue at 1, 4, and 12 weeks in the Ctrl, mSF–MNs, mSF/gPLGA–MNs, and VEGF@mSF/gPLGA–MNs groups. Scale bars: 100 μm. Statistical significance: ^*P < 0.05, ^**P < 0.01, ^***P < 0.001, ^****P < 0.0001

The superior performance of the VEGF@mSF/gPLGA–MNs was also confirmed by the quantitative analyses. The mass of the transplanted adipose tissue in this group (Figure 9b) was better preserved over time and was obviously greater than those in the control group and the mSF–MNs group at 12 weeks. The mSF/gPLGA–MNs group was also better than the control group, but the retention rate in this group is inferior to that of the VEGF@mSF/gPLGA–MNs group. Similarly, the adipose tissue cross-sectional area (Figure 9c) in the VEGF@mSF/gPLGA–MNs group was significantly larger than that in the control group and the mSF–MNs group at 4 and 12 weeks, which means that the adipose survival rates are greater. Ratio analysis of the oil cyst area (inversely indicator of adipose survival, i.e. a lower ratio reflects greater cell viability) (Figure 9d) at 4 weeks showed that the VEGF@mSF/gPLGA–MNs group had a lower ratio than the other groups did. At 12 weeks, the oil cyst ratio was significantly lower in the VEGF@mSF/gPLGA–MNs group than in the control and mSF–MNs groups, which suggests that adipose survival is increased and also has a less risk of the liquefaction and necrosis.

Histological analysis by Masson’s trichrome staining (Figure 9f) revealed that in the control group, continuous collagen deposition was detected over time, suggesting that fibrous encapsulation occurred; in the mSF–MNs and mSF/gPLGA–MNs groups, the collagen was more uniformly distributed and the fibers were finer; however, collagen deposition was the lowest in the VEGF@mSF/gPLGA–MNs group and the adipose tissue structure was best preserved, indicating the lowest degree of fibrosis and best tissue integration. Quantitative analysis (Figure 9e) revealed that at 4 and 12 weeks, the collagen area in the VEGF@mSF/gPLGA–MNs group was significantly smaller than that in the other groups.

Overall, VEGF@mSF/gPLGA–MNs increased the retention rate and quality of transplanted fat, promoted vascularization, strengthened the inhibition of liquefaction, necrosis and fibrosis, and therefore led to superior tissue retention and long-term survival.

Transcriptomic profiling reveals the mechanisms underlying VEGF@mSF/gPLGA–MN-mediated fat graft remodeling

In order to understand the mechanism of the improvements of the fat graft survival and the remodeling induced by VEGF@mSF/gPLGA–MNs, we did the RNA sequencing (RNA-seq) of the transplanted adipose tissue. Three biological replicates were taken from the control and the VEGF@mSF/gPLGA–MNs groups. The PCA showed the clear separation between the two groups, which implied that the two groups have different transcriptional profiles (Figure 10a). The volcano plot analysis showed a total of 1935 DEGs, and the 1099 were significantly upregulated genes and 836 were significantly downregulated genes (Figure 10b). The heatmap clustering showed that the expression pattern of the top 50 most DEGs was inversely between the two groups (Figure 10c).

Figure 10.

Transcriptomic analyses, including principal component analysis (PCA), volcano plots, heatmaps and functional enrichment, reveal the changes in gene expression and related signaling pathways in fat grafts after MN intervention. (a) PCA plot showing the separation among all samples. (b) Volcano plot comparing the VEGF-loaded modified silk fibroin/grooved poly(lactic-co-glycolic acid)–microneedles (VEGF@mSF/gPLGA–MNs) group with the Ctrl group; red dots indicate upregulated genes, and blue dots indicate downregulated genes. (c) Heatmap of differentially expressed genes (DEGs) between the VEGF@mSF/gPLGA–MNs and Ctrl groups; red indicates upregulation, and blue indicates downregulation. (d) Gene Ontology (GO) functional enrichment analysis of upregulated DEGs. (e) GO functional enrichment analysis of downregulated DEGs. (f) Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis of both up- and downregulated DEGs. (g) Chord diagrams illustrating the relationships between enriched KEGG terms and their associated genes

To elucidate the underlying molecular mechanism, we performed GO and KEGG pathway enrichment analyses on the DEGs. GO analysis revealed that the upregulated genes were significantly enriched in biological processes, including positive regulation of cell migration, angiogenesis, lipid metabolism, negative regulation of inflammation, and browning of adipocytes (Figure 10d), suggesting the enrichment of pathways associated with angiogenesis, modulation of adipocyte metabolism, and those that foster a microenvironment supportive of tissue survival. In contrast, the downregulated genes were predominantly associated with immune effector processes, inflammatory responses, and lipid catabolism (Figure 10e), suggesting their potential involvement in immune regulation and lipid homeostasis maintenance. KEGG enrichment and chord diagram analyses corroborated these findings (Figure 10f and g). Notably, the PPAR signaling pathway, which is crucial for adipogenesis, was significantly enriched, whereas key inflammatory pathways (IL-17, NF-κB, toll-like receptor, and TNF signaling) exhibited general decreases in gene expression.

To summarize, we showed by our transcriptomic data that VEGF@mSF/gPLGA–MNs have the angiogenic properties of VEGF, the protective effect of mSF and the immunomodulatory function of gPLGA. All these coordinated effects have a molecular microenvironment which gives for the angiogenesis, the inhibition of inflammation and the stimulation of adipogenesis, which increases the survival of the fat graft and promotes remodeling.

To validate the transcriptomic findings at the protein level, proteins in key inflammation- and metabolism-related signaling pathways were examined by western blotting. Consistent with the RNA-seq results, VEGF@mSF/gPLGA–MNs significantly increased the expression of PPARγ and VEGF (Figure S11a–c), indicating increased lipid metabolism and tissue regeneration. In contrast, the NF-κB pathway was markedly suppressed, as evidenced by reduced p-p65 expression and a decrease in the p-p65/p65 ratio (Figure S11d and e). These results confirm that gPLGA–MNs modulate inflammatory and metabolic pathways at the level of transcription and of the protein, and in this way support the therapeutic role of the lipid removal-mediated microenvironmental remodeling.

VEGF@mSF/gPLGA–MNs increase fat graft survival by promoting angiogenesis

In order to test the proangiogenic effect of VEGF@mSF/gPLGA–MNs on mice and the survival status of transplanted adipose tissue on the basis of the above RNA-seq analysis results, we also did immunofluorescence staining and histological analysis on skin and adipose tissues harvested at 1, 4, and 12 weeks after surgery (Figure 11a). At 1 week, the CD31 fluorescence signals in the skin and fat tissues were weak in all the groups, but the VEGF@mSF/gPLGA–MNs group presented increased CD31 expression. Furthermore, perilipin fluorescence (green, indicating lipid droplets) in the adipose tissue of this group was greater than that in the other groups. After 4 weeks, the CD31 signals had increased in all the treatment groups, yet the VEGF@mSF/gPLGA–MNs group maintained the highest fluorescence intensity in both the skin and fat. Perilipin fluorescence in this group did not decrease markedly, whereas the other groups exhibited fusion and vacuolization of lipid droplets. At 12 weeks, although the CD31 and perilipin signals had increased overall, the VEGF@mSF/gPLGA–MNs group still displayed strong fluorescence along with a regular fat tissue morphology and uniformly sized lipid droplets, suggesting sustained angiogenesis and adipocyte stability.

Figure 11.

Immunofluorescence and histological staining images showing the effects of VEGF-loaded modified silk fibroin/grooved poly(lactic-co-glycolic acid)–microneedles (VEGF@mSF/gPLGA–MNs) on angiogenesis and adipocyte maintenance in transplanted adipose tissue at different time points. (a) Immunofluorescence images of CD31 and perilipin in transplanted adipose tissue and the overlying skin at weeks 1, 4, and 12 in the Ctrl, modified silk fibroin microneedles (mSF–MNs), modified silk fibroin/grooved poly(lactic-co-glycolic acid)–microneedles (mSF/gPLGA–MNs), and VEGF@mSF/gPLGA–MNs groups. Scale bar: 100 μm; 50 μm. (b and c) Quantitative analysis of the CD31⁺ vessel area in skin and adipose tissue at weeks 1, 4, and 12. (d) Quantitative analysis of the perilipin⁺ adipocyte area in subcutaneous adipose tissue at weeks 1, 4, and 12. (e) Haematoxylin and eosin (H&E) staining images of transplanted adipose tissue and the overlying skin sections at weeks 1, 4, and 12 in the Ctrl, mSF–MNs, mSF/gPLGA–MNs, and VEGF@mSF/gPLGA–MNs groups. Scale bars: 200 μm (main images); 50 μm (insets). Statistical significance: ^*P < 0.05, ^**P < 0.01, ^***P < 0.001, ^****P < 0.0001

We analysed the CD31-positive area in skin and adipose tissues (Figure 11b and c) in order to quantify these observations. For the skin, after 1 week, compared with the other groups, the VEGF@mSF/gPLGA–MNs group had a significantly greater percentage of CD31-positive area in the skin; this pattern is still at a very high level at 4 weeks, whereas the other groups have an increasing trend but are all lower. After 12 weeks, although all the groups have an increase in the CD31-positive area, the VEGF@mSF/gPLGA–MNs group still has significantly greater fluorescence intensity than the VEGF-free controls do. A similar trend is detected in adipose tissue. With respect to the perilipin-positive area in adipose tissue (Figure 11d), the proportion is significantly greater in the VEGF@mSF/gPLGA–MNs group at 1 week. At 4 weeks, the fluorescence intensity is decreased in all the groups, but the reduction is milder in the VEGF@mSF/gPLGA–MNs group. After 12 weeks, the perilipin expression is remained highest in the VEGF@mSF/gPLGA–MNs group.

Complementary H&E staining revealed histological alterations at each time point (Figure 11e). In the control group, the adipose tissue showed progressive vacuolation, fusion, and necrosis. The vascular structures appeared at 12 weeks only. Compared with the control group, in the mSF–MNs group, the survival is a little better, but still, they show the fusion and vacuolation. In the mSF/gPLGA–MNs group, at 4 weeks there is an early sign of repair, new cell growth and remodeling; at 12 weeks, the adipocytes are enlarged, and the tissue structure is more regular. The VEGF@mSF/gPLGA–MNs group shows the most evident tissue remodeling: at 1 week, the skin and adipose tissue show extensive tissue remodeling accompanied by an increase in the vascular structures; at 4 weeks, the small vascular structures are consistent with the neovascularization in the adipose tissue; at 12 weeks, the tissue architecture is more complex, with increased integration and vascularization. All these show the positive impact of VEGF-loaded MNs on tissue repair and angiogenesis.

VEGF@mSF/gPLGA–MNs increase adipose tissue remodeling and modulate the immune microenvironment in transplanted fat grafts

Finally, we performed the immunofluorescence staining and quantitative analysis of the samples collected at 1, 4, and 12 weeks after transplantation to evaluate the effects of different MNs formulations on adipose tissue remodeling and the immune microenvironment, on the basis of the previous RNA-seq results. The immunofluorescence staining images (Figure 12a) showed that there were marked differences in UCP-1 and PPARγ expression among the control, mSF–MNs, mSF/gPLGA–MNs and VEGF@mSF/gPLGA–MNs groups. Compared with the other groups, the VEGF@mSF/gPLGA–MNs group seemed to have consistently stronger UCP-1 and PPARγ signals over time, which suggested more browning-associated features and adipogenic activity. The quantitative analysis confirmed these observations (Figure 12b and c); at each time point, the proportions of UCP1⁺ and PPARγ⁺ areas in the VEGF@mSF/gPLGA–MNs group were greater than those in other groups, although these proportions were also much greater in the mSF–MNs and mSF/gPLGA–MNs groups than in the control group.

Figure 12.

Quantification of the immunofluorescence intensity to characterize the regulatory effects of the MN system on adipocyte browning and macrophage polarization within fat grafts over time. (a) Immunofluorescence images of UCP-1 and peroxisome proliferator-activated receptor gamma (PPARγ) in transplanted adipose tissue at 1, 4, and 12 weeks in the Ctrl, modified silk fibroin microneedles (mSF–MNs), modified silk fibroin/grooved poly(lactic-co-glycolic acid)–microneedles (mSF/gPLGA–MNs), and VEGF-loaded modified silk fibroin/grooved poly(lactic-co-glycolic acid)–microneedles (VEGF@mSF/gPLGA–MNs) groups. Scale bar: 100 μm. (b) Quantitative analysis of the UCP-1⁺ area in adipose tissue at weeks 1, 4, and 12. (c) Quantitative analysis of the PPARγ⁺ area in adipose tissue at weeks 1, 4, and 12. (d) Quantitative analysis of the CD86⁺ area in adipose tissue at weeks 1, 4, and 12. (e) Quantitative analysis of the CD206⁺ area in adipose tissue at weeks 1, 4, and 12. (f) Immunofluorescence images of CD206 and CD86 expression in transplanted adipose tissue from the Ctrl, mSF–MN, mSF/gPLGA–MN, and VEGF@mSF/gPLGA–MN groups at 1, 4, and 12 weeks. Scale bars: 100 μm. Statistical significance: ^*P < 0.05, ^**P < 0.01, ^***P < 0.001, ^****P < 0.0001

To measure how different MNs formulations regulate the immune microenvironment, macrophage polarization was assessed by immunofluorescence staining for CD86⁺ (M1 proinflammatory) and CD206⁺ (M2 anti-inflammatory) macrophages (Figure 12f). At 1 week after surgery, the macrophages in the mSF/gPLGA–MNs and VEGF@mSF/gPLGA–MNs groups (both of which contained a gPLGA–MNs core) were all CD206-positive, while those in the control and mSF–MNs groups were all CD86⁺. At 4 weeks, in the control and mSF–MNs groups, CD86⁺ expression in macrophages decreased, and CD206⁺ expression increased; moreover, in the mSF/gPLGA–MNs and VEGF@mSF/gPLGA–MNs groups, the CD206⁺ macrophage population continued to increase and the CD86⁺ macrophage population gradually decreased. By 12 weeks, the macrophages in both the control and mSF–MNs groups were characterized by mainly CD206⁺, while in the mSF/gPLGA–MNs and VEGF@mSF/gPLGA–MNs groups the levels of both CD206⁺ and CD86⁺ macrophages were lower than they were at earlier time points. The lowest levels of both markers were in the VEGF@mSF/gPLGA–MNs group.

These observations were corroborated by the quantitative analysis (Figure 12d and e). At 1 week, the CD86⁺ level was significantly greater in the control group than in all the other groups, whereas the CD206⁺ level was markedly greater in the VEGF@mSF/gPLGA–MNs group. This trend persisted at 4 weeks, when CD206⁺ expression in the VEGF@mSF/gPLGA–MNs group plateaued. After 12 weeks, the VEGF@mSF/gPLGA–MNs group displayed the lowest levels of both CD206⁺ and CD86⁺ macrophages. These results demonstrate that the mSF/gPLGA–MNs and VEGF@mSF/gPLGA–MNs formulations, particularly the latter, can effectively modulate the polarization of macrophages toward an anti-inflammatory phenotype over time, thereby shifting the immune microenvironment toward a pro-regenerative state and promoting adipose tissue remodeling with enhanced brown fat characteristics.

Discussion

Adipose tissue reconstruction is very important during esthetic and reconstructive surgery [50]. However, the high resorption rate (40%–60%) of transplanted fat remains a major clinical challenge [51]. A major contributor to graft failure is delayed revascularization caused by early ischemia–hypoxia, followed by sustained inflammation triggered by the release of necrotic lipids, which collectively form a vicious cycle leading to graft failure [12, 52]. To address these challenges, we developed a multifunctional core–shell MN system (VEGF@mSF/gPLGA–MNs). Through a spatiotemporally coordinated and multitarget intervention strategy, this system simultaneously intervenes in two key pathological stages of fat grafting, early vascular insufficiency and late lipid-driven inflammation, which improves graft retention and the tissue integration of transplanted adipose tissue.

The “vascularization first, lipid removal later” sequence is dictated by the evolving pathology of fat grafts. Initially, ischemia–hypoxia and excessive ROS accumulation dominate the early pathological microenvironment, directly damaging endothelial cells and driving adipocyte necrosis. The early restoration of perfusion and redox balance is therefore necessary for graft stabilization. During subsequent remodeling, necrotic adipocytes release large amounts of extracellular lipids, fueling lipotoxicity, macrophage activation, and chronic inflammation. Lipid removal before adequate revascularization may provide incomplete therapeutic benefits, as ongoing ischemia leads to the continuous generation of new lipid debris. More critically, promoting angiogenesis within an unresolved lipid-rich and proinflammatory niche may not necessarily result in functional vascular maturation and could contribute to vascular instability, persistent inflammation, and aberrant remodeling, as reported for lipid-associated pathological vascular remodeling and neovascularization [53–55]. Nonetheless, functional tissue repair ultimately requires the coordinated interplay of angiogenesis, immune modulation, and metabolic reprogramming rather than isolated vascular promotion [56, 57]. Hence, staging vascularization prior to lipid removal allows the microenvironment to first acquire sufficient vascular and metabolic support before lipid clearance is initiated, which was achieved with our system through ROS-responsive VEGF release from the mSF shell followed by lipid adsorption by the exposed gPLGA core.

While most recent studies have utilized MNs as simple drug delivery carriers, for example, to deliver VEGF or anti-inflammatory agents, these systems, despite benefiting from localized retention, generally lack responsiveness to changes in the pathological microenvironment and the ability to actively clear harmful substances [58, 59]. In contrast, our core–shell architecture enables coupling ROS-responsive VEGF release with synergistic groove–pore lipid adsorption. We systematically verified the spatiotemporal degradation characteristics of the core–shell structure both in vitro and in vivo. The mSF shell underwent rapid degradation within the first 3 days after implantation, which coincided with the acute phase of oxidative stress in fat grafts, whereas the gPLGA core remained structurally stable and became progressively exposed following shell degradation. This programmed degradation behavior establishes a temporally coordinated therapeutic sequence characterized by “angiogenesis first and lipid clearance later”, which closely matches the dynamic pathological progression of fat graft remodeling. The mSF shell not only serves as a VEGF reservoir but also scavenges ROS, contributing to the mitigation of early oxidative stress. This design offers more precise release logic and superior microenvironment adaptability compared with conventional VEGF delivery systems that rely solely on material-based sustained release (e.g. PLGA microspheres or hydrogels) [60, 61]. Furthermore, the lipid-adsorbing function of the gPLGA core differs from that of recently reported nanoparticle strategies that modulate lipid metabolism [34, 62, 63]. Whereas those approaches depend primarily on cellular uptake and metabolic reprogramming [63], our system directly removes extracellular lipid debris through physical adsorption and capillary transport. By eliminating pathological lipids within the local microenvironment, the system reduces the phagocytic burden on local macrophages and helps prevent M1 macrophage polarization, which is typically induced by lipid overload [11, 14]. Crucially, by removing the excessive lipid burden that drives chronic inflammation, our system may help shift the immune microenvironment toward a state conducive to graft survival [64, 65].

The superior lipid clearance performance of gPLGA–MNs results from the integration of the lipophilicity of PLGA with groove-mediated capillary transport. The lower lipid adsorption observed with the gmSF–MNs and non-grooved PLGA–MNs indicates that efficient lipid clearance depends on the combined effects of PLGA (lipophilicity) and groove-mediated transport, thereby reducing local lipotoxic stress and promoting tissue regeneration. This is significant because balanced macrophage polarization toward the M2 phenotype has been shown to support fat graft retention through multiple mechanisms: M2 macrophages promote angiogenesis by releasing proteases and activating paracrine signaling with vascular endothelial cells [66, 67], and these cells also secrete anti-inflammatory mediators (e.g. IL-10 and TGF-β) that mitigate inflammation and promote tissue repair [68]. Consequently, this physical lipid clearance strategy mitigates lipid-induced M1 macrophage polarization, thereby improving graft integration and providing a versatile therapeutic paradigm for lipid overload-driven pathologies.

Our study demonstrated that the mSF shell of the system rapidly responded to ischemia and hypoxia in the microenvironment, releasing VEGF to effectively promote angiogenesis in the early stage. Moreover, the ROS-scavenging capacity of the mSF shell significantly alleviates oxidative stress-induced mitochondrial damage in adipocytes, preserving cellular structural and functional integrity. Following shell degradation, the grooved and porous gPLGA core is exposed and actively adsorbs lipids released from necrotic adipocytes via capillary action, thereby attenuating macrophage overactivation and chronic inflammation triggered by local lipid overload. This “first promote angiogenesis, then adsorb lipids” temporally sequential design overcomes the limitations of conventional single-target interventions and achieves dynamic modulation of the fat graft microenvironment. Notably, transcriptomic analysis revealed the molecular mechanisms underlying the effects of the VEGF@mSF/gPLGA–MNs, extending our understanding from structural regeneration to functional recovery of the adipose graft. RNA-seq revealed that the VEGF@mSF/gPLGA–MNs significantly upregulated genes associated not only with angiogenesis but also with the positive regulation of lipid metabolism and adipocyte browning (a key thermogenic and metabolic function) [69] and suppressed pathways related to inflammatory responses, findings that were corroborated by western blot analyses showing increased PPARγ and VEGF expression together with reduced NF-κB activation.

These findings provide systems-level evidence that the MN system orchestrates multipathway synergy, including the activation of PPAR signaling (central to adipogenesis and lipid homeostasis) and the suppression of key inflammatory cascades, to create a molecular microenvironment conducive to adipose tissue survival [66, 70]. This coordinated functional shift implies that regenerated fat tissue achieves physiological lipid metabolism and tissue remodeling beyond volume repair, supporting microenvironment-targeted therapies for comprehensive structural and functional regeneration of adipose tissue despite the need for further validation of full functional restoration.

Accumulating evidence indicates that lipid accumulation drives lipotoxicity [71], with an excess of lipid droplets in fat grafts constituting a key yet frequently overlooked driver of sustained inflammation and fibrosis [14, 72] that decisively contributes to graft failure. Unlike previous strategies that indirectly modulate lipotoxicity, such as pharmacologically inhibiting lipolysis or skewing the macrophage phenotype, our system directly removes extracellular lipids through an integrated material and microscale physical design [35, 73–75]. This direct clearance strategy experimentally reduces the phagocytic burden on macrophages, promotes their polarization toward a pro-repair M2 phenotype, and suppresses key inflammatory pathways in situ. Thus, our work shifts the therapeutic paradigm from merely suppressing inflammatory symptoms to eliminating the underlying cause (pathological lipid debris), offering both a novel engineered solution and deeper mechanistic insight into lipotoxicity in fat graft failure.

Nevertheless, this study has several limitations that also suggest clear directions for future research. First, all the experiments were conducted in small animal models, which differ from humans in terms of skin thickness, metabolic profiles, and fat volume. Clinical translation may require adjusting the MN dimensions to accommodate thicker human skin, potentially increasing invasiveness and leading to the formation of more noticeable local traces [76, 77]. Moreover, the lipid-adsorbing function of the MNs may be spatially biased toward the superficial layer of the graft adjacent to the skin. However, since adipocyte necrosis and insufficient vascularization also predominantly occur in this region, such spatial limitations may align with the pathological distribution [78–80].

Second, although the current fabrication process demonstrated good batch-to-batch reproducibility, further optimization of scalable manufacturing strategies is needed to ensure structural consistency and product quality during future clinical translation. Finally, the current responsive design is triggered primarily by ROS; integrating multiple biological cues, such as pH or specific enzymes, in the future could enable more precise and dynamic regulation [81, 82]. Beyond fat grafting, these findings also suggest the potential applicability of this platform to other lipid-associated pathological conditions, such as atherosclerotic plaques and lipid-rich chronic inflammatory lesions.

Conclusions

In summary, we developed a core–shell MN system that integrates ROS scavenging, angiogenesis promotion, and active lipid removal and demonstrated its ability to significantly increase fat graft survival through multitarget synergistic effects. This work not only provides an improved engineering solution to the clinical challenge of fat grafting but also, more importantly, exemplifies a novel therapeutic concept: the management of lipotoxic inflammatory diseases through spatiotemporally regulated local physical intervention. Future research will focus on the clinical translation of this system and explore its potential applications in other lipid-related pathological conditions.

Acknowledgements

The authors thank the other members of their laboratory for their valuable insights and technical assistance.

Author contributions

Hengyu Wu (Conceptualization [equal], Data curation [equal], Formal analysis [equal], Investigation [equal], Validation [equal], Visualization [equal], Writing—original draft [equal]), Ganghua Yang (Conceptualization [equal], Data curation [equal], Investigation [equal], Methodology [equal], Software [equal], Writing—original draft [equal]), Yuanzheng Zhu (Conceptualization [equal], Methodology [equal]), Yanglong Zhu (Investigation [equal], Methodology [equal], Software [equal]), Haowen Kang (Investigation [equal], Software [equal]), Jiahui Wu (Investigation [equal], Software [equal]), Minchen Zhang (Software [equal], Validation [equal]), Xinghong Zeng (Investigation [equal], Software [equal]), and Yangyan Yi (Conceptualization [equal], Funding acquisition [equal], Supervision [equal], Writing—review & editing [equal])

Ethics approval and consent to participate

All animal procedures were approved by the Animal Ethical and Welfare Committee of Nanchang University, China (Approval No. NNCULAE-20250524001) and were conducted in accordance with institutional guidelines for the care and use of experimental animals.

Conflicts of interest

All authors declare no conflicts of interest in this work. None of the authors has a financial interest in any of the products or devices mentioned in this article.

Funding

This work was supported by grants from the National Natural Science Foundation of China (82460448), the Natural Science Foundation of Jiangxi Province (20242BAB26140), and the Jiangxi Province Key Laboratory of Precision Cell Therapy (2024SSY06241).