Lipoaspirate-derived secretome restores redox homeostasis to attenuate pathological scar formation

Abstract

Background

Pathological scars are characterized by persistent fibroblast activation and excessive extracellular matrix (ECM) deposition. Although oxidative stress and dysregulated NRF2 signaling contribute to TGF-β1-mediated fibrosis in internal organs, their roles in cutaneous pathological scarring remain unclear. Secretome-based cell-free therapies have shown regenerative and antifibrotic potential, but whether lipoaspirate-derived secretome can restore fibroblast redox homeostasis and thereby attenuate pathological scarring is unknown.

Methods

Lipoaspirate fluid obtained during standard tumescent liposuction was processed by 100-kDa ultrafiltration to generate lipoaspirate-derived secretome (LA), while secretome from adipose-derived stromal cell (ADSC) culture supernatant served as a comparator (CS). LA and CS were characterized by nanoparticle tracking analysis, transmission electron microscopy, and immunoblotting for EV markers. In vivo efficacy was evaluated in a rabbit ear scar model with weekly intradermal LA injection, followed by gross, histological, collagen, and qPCR assessments. Comparative proteomic profiling of LA and CS was performed using data-independent acquisition LC–MS/MS with enrichment analysis, and paired human scar and normal skin samples were analyzed by single-cell RNA sequencing. In vitro, TGF-β1-stimulated fibroblasts were treated with LA or CS, and NRF2 involvement was assessed using the inhibitor ML385 or Nrf2-targeting siRNA. Redox balance, NRF2 signaling, and profibrotic responses were assessed by fluorometric assays, qPCR, and immunoblotting.

Results

LA showed a substantially higher particle yield than CS and contained abundant extracellular vesicles. Weekly intradermal LA injections reduced scar formation and improved collagen organization in the rabbit ear model. Comparative proteomics of LA versus CS highlighted cytoprotective pathways, including glutathione metabolism and NRF2-associated antioxidant signaling. Analyses of human scar tissues revealed elevated NOX4 expression, increased 4-HNE, and impaired NRF2-associated antioxidant signaling in scar fibroblasts. In vitro, both LA and CS attenuated TGF-β1-driven profibrotic fibroblast activation, whereas LA showed greater antioxidant activity, including stronger suppression of oxidative stress-related responses and improved GSH/GSSG balance. Pharmacological inhibition and genetic silencing of NRF2 partially reversed the antioxidant and antifibrotic effects of LA.

Conclusion

Lipoaspirate-derived secretome (LA) is a clinically accessible cell-free therapeutic candidate for pathological scarring. LA restores redox homeostasis, in part through NRF2-associated antioxidant signaling, attenuates TGF-β1-driven profibrotic fibroblast activation, and improves scar remodeling in vivo. These findings support LA as a promising cell-free strategy for attenuating pathological scar formation.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s13287-026-05123-0.

Background

Pathological scar formation, including hypertrophic scars and keloids, is a common and clinically challenging consequence of trauma, burns, and surgical procedures [1]. Aberrant fibroblast activation is a central driver of this pathology, characterized by excessive proliferation, myofibroblast differentiation, and disproportionate deposition of extracellular matrix (ECM) components such as collagen type I, collagen type III, and fibronectin [2, 3]. Although a variety of surgical and nonsurgical treatments are available, their clinical efficacy remains limited and recurrence is frequent [4].

Transforming growth factor-β1 (TGF-β1) is a key upstream regulator of fibrotic responses [5]. In addition to inducing profibrotic gene expression, TGF-β1 promotes excessive reactive oxygen species (ROS) production, often through activation of NADPH oxidase 4 (NOX4) [6, 7]. In turn, ROS can further amplify TGF-β1 signaling. In cell-free systems, ROS generated by ionizing radiation or metal-catalyzed ascorbate reactions can directly convert recombinant latent TGF-β1 to its active form. This oxidative activation appears specific to the TGF-β1 isoform and has been linked to modification of a key methionine residue (Met253) in the latency-associated peptide (LAP-β1) [8]. Beyond activating latent TGF-β1, ROS can also enhance its expression and secretion. In multiple cell types, ROS upregulates TGF-β1 via NF-κB, AP-1, and p38/JNK/ERK signaling and is also associated with epithelial–mesenchymal transition and endothelial-to-myofibroblast transition [9–12]. These observations support the existence of a TGF-β1-ROS feed-forward loop that sustains profibrotic signaling and promotes downstream ECM accumulation in fibrosis [13]. Although this mechanism has been extensively characterized in cardiac, hepatic, pulmonary, and renal fibrosis [6, 14–16], its contribution to cutaneous pathological scarring remains incompletely understood. Nevertheless, increased NOX4 expression has been observed in keloids [17], pointing to a potential involvement of oxidative stress in pathological scarring.

Nuclear factor erythroid 2-related factor 2 (NRF2) is a master regulator of the cellular antioxidant response. Upon activation, NRF2 translocates to the nucleus and induces the expression of cytoprotective genes—including HMOX1, SLC7A11, NQO1, SOD1, SOD2, GCLM, and GCLC, and thereby enhancing glutathione metabolism and alleviating oxidative stress [18]. NRF2 activation has been shown to counteract fibrotic processes by attenuating TGF-β1-driven ROS elevation and subsequent ECM overproduction [19]. Therefore, restoring NRF2-mediated redox homeostasis may represent a rational strategy to disrupt the TGF-β1–ROS profibrotic loop and suppress pathological scar formation.

In recent years, cell-free regenerative therapies based on the secretome have received increasing attention. The secretome comprises bioactive factors released by cells into the extracellular environment, including soluble growth factors, cytokines, and extracellular vesicles (EVs) [20]. Increasing evidence suggests that many of the therapeutic effects of stem cell transplantation are mediated mainly through paracrine signaling rather than direct differentiation [21], indicating that the secretome is a key mediator of these effects.

The secretome of adipose-derived stromal cells (ADSCs) has shown promising regenerative and antifibrotic potential [22], and may even promote wound healing more effectively than ADSCs themselves [23]. As an important component of the secretome, ADSC-derived extracellular vesicles (EVs) have also been widely studied. However, their effects on scar formation remain uncertain. Although ADSC-EVs have been reported to inhibit hypertrophic scarring [24], another study suggests that they may enhance collagen deposition at later stages of wound healing, potentially favoring scar hypertrophy [25]. Notably, ADSC-EVs can alleviate oxidative stress-related fibroblast injury via the KEAP1/NRF2 axis [26], and EVs from NRF2-overexpressing ADSCs were found to reduce ROS accumulation and suppress fibrosis in diabetic wound models [27]. Together, these findings suggest that the anti-scarring effects of ADSC-EVs remain controversial, whereas NRF2 signaling may play an important role in scar modulation.

The composition and function of the secretome are highly dependent on cell type and microenvironment and can be reshaped by preconditioning [28]. Liposuction may represent a form of in situ preconditioning by exposing adipose tissue to transient hypoxia and mechanical shear stress. Unlike conditioned medium from cultured ADSCs, lipoaspirate fluid is collected directly during routine liposuction and may contain stress-responsive cytoprotective factors reflecting both the natural adipose tissue environment and liposuction-related stress. Bellei et al. reported that the extracellular fraction of lipoaspirate fluid attenuated oxidative stress-induced senescence in fibroblasts and improved their paracrine function [29], suggesting that the cell-free liquid component of lipoaspirate may have independent regenerative value. In addition, EVs isolated from lipoaspirate fluid have also been shown to improve graft survival, modulate fibrotic responses [30], promote wound healing [31], and suppress inflammation [32]. These findings suggest that lipoaspirate fluid, which is often discarded as clinical waste, may represent an accessible source of bioactive cell-free material.

In this study, we investigated whether an imbalance characterized by oxidative stress overload and impaired NRF2-associated antioxidant signaling contributes to pathological scar formation, and whether lipoaspirate-derived secretome (LA) can suppress scar formation. We show that scar fibroblasts exhibit elevated NOX4 expression together with impaired NRF2-associated antioxidant signaling. LA improves redox homeostasis and attenuates TGF-β1-associated oxidative stress and profibrotic fibroblast activation in vitro, and reduces scar formation in vivo. LA also activates NRF2 signaling, and loss-of-function experiments using pharmacological inhibition and genetic silencing indicate that NRF2 contributes substantially, but not exclusively, to its antioxidative and antifibrotic effects. These findings identify LA as a readily obtainable, clinically translatable, acellular therapeutic candidate for scar management.

Materials and methods

Cell culture and treatment

Fibroblasts (FBs) were purchased from GuangZhou Jennio Biotech Co., Ltd. (China) and maintained in high-glucose DMEM (DMEM-H) supplemented with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin (P/S, v/v) at 37 °C in a humidified incubator with 5% CO₂. For mechanistic studies, FBs were treated with TGF-β1 (10 ng/mL) and/or secretome concentrates derived from lipoaspirate fluid or ADSC-conditioned medium (LA/CS; 100 µg/mL; prepared as described below), in the presence or absence of the NRF2 inhibitor ML385 (5 µM; vehicle control).

ADSC isolation and culture

Adipose-derived stromal cells (ADSCs) were isolated from the adipose fraction of lipoaspirate collected under sterile conditions following standard tumescent liposuction. The fluid fraction was reserved for LA preparation, and ADSCs were isolated as previously described [33, 34]. ADSCs were cultured in DMEM-H supplemented with 10% FBS and 1% P/S (v/v) at 37 °C with 5% CO₂, and passages 2–5 (P2–P5) were used for subsequent experiments.

Processing of lipoaspirate fluid derived-secretome (LA)

Fresh lipoaspirate was centrifuged at 1,000 × g for 10 min at 4 °C to remove intact cells and large debris. After centrifugation, the upper adipose layer was collected for ADSC isolation (as described above), while the middle aqueous fraction was retained for secretome extraction.

To remove residual debris, larger particles, and potential microbial contaminants, the fluid was sequentially filtered through 0.8 μm, 0.45 μm, and 0.22 μm sterile syringe filters.

The clarified fluid was then concentrated using 100 kDa molecular-weight-cutoff ultrafiltration tubes (Millipore, Cat. no. UFC9100) at 4,000 rpm for 20 min at 4 °C, yielding a lipoaspirate-derived secretome concentrate. The final concentrate (~ 400–500 µL) was aliquoted and stored at − 80 °C.

Processing of ADSC culture supernatant derived secretome (CS)

ADSCs were expanded to P2–P5 and, at 80–90% confluence, were washed with PBS and incubated in DMEM-H containing exosome-depleted FBS (Shanghai Nonin Biological Technology; Cat. no.C3801) for 48 h to allow collect ADSC-derived secretome while minimizing interference from serum-derived vesicles. The conditioned medium was collected and processed as described for LA, including sequential filtration and 100-kDa ultrafiltration.

Nanoparticle tracking analysis (NTA)

Particle size distribution and concentration were determined using a ZetaView system (Particle Metrix, Germany). Samples were diluted in PBS to fall within the instrument’s optimal measurement range and analyzed under identical acquisition settings across samples. Particle concentrations were calculated from multiple video positions per sample (≥ 11 positions) and averaged for quantification.

Transmission electron microscopy (TEM)

Secretome samples were diluted in PBS (LA was diluted 1:100 due to high particle abundance) and adsorbed onto carbon-coated copper grids for 1–2 min. Grids were then negatively stained with 2% uranyl acetate and imaged using a JEM-1400 transmission electron microscope (JEOL, Japan) operated at 120 kV.

IncuCyte live-cell imaging proliferation assay

Fibroblast proliferation was assessed using a live-cell imaging platform (IncuCyte S3, Sartorius). Fibroblasts were seeded in 96-well plates at 4 × 10³ cells/well in complete medium and allowed to attach overnight. Cells were then serum-starved for 12 h for cell-cycle synchronization and subsequently treated with vehicle, TGF-β1 (10 ng/mL), LA (100, 200, or 300 µg/mL), CS (100, 200, or 300 µg/mL), or the indicated combinations.

Plates were placed into the IncuCyte system, and phase-contrast images were acquired automatically every 4 h for up to 48 h using a 10× objective. For each well, four non-overlapping fields were imaged and analyzed. Cell proliferation was quantified as percent confluence using the Incucyte integrated analysis software (IncuCyte^®2024 A).

To account for baseline differences in seeding density, confluence values at each time point were normalized to the corresponding 0 h confluence for each well and expressed as a fold change relative to 0 h (0 h = 1.0). For statistical comparisons, normalized confluence at selected time points (e.g., 12, 24, 36, and 48 h) was used, and group differences were analyzed by one-way ANOVA followed by post hoc tests or by Student’s t test where appropriate, as detailed in the Statistical analysis section.

Intracellular ROS measurement

Intracellular ROS levels were assessed using the Reactive Oxygen Species Assay Kit (Beyotime, China; Cat. no. S0033S) according to the manufacturer’s instructions. Fibroblasts were assigned to the following groups: CTRL, CS, LA, H₂O₂, CS + H₂O₂, and LA + H₂O₂. Cells in the CS- and LA-related groups were treated with CS or LA (100 µg/mL). For oxidative stress induction, cells were exposed to 100µM H₂O₂ for 2 h, with CS or LA pretreatment for 6 h in the combination groups. After treatments, cells were incubated with DCFH-DA working solution, washed, and imaged under a fluorescence microscope using the FITC channel. Fluorescence intensity was quantified in ImageJ and normalized to the CTRL group to obtain relative ROS levels.

RNA isolation and quantitative real-time PCR (qPCR)

Total RNA was extracted from cultured fibroblasts and rabbit scar tissues using TRIZOL reagent (15596018CN; Invitrogen, Thermo Fisher Scientific, USA) according to the manufacturer’s instructions. cDNA was synthesized using a TIANGEN reverse transcription kit (TIANGEN, Beijing, China; Cat. no. KR116-02). qPCR was performed using a SYBR Green master mix (QIAGEN, Hilden, Germany; Cat. no. 1129280) on a Bio-Rad CFX real-time PCR system (Bio-Rad, USA). The qPCR primer sequences are listed in Table 1.

Table 1.

Primer sequences used for qPCR

Gene name	forward	reverse
Human-GAPDH	GGAGCGAGATCCCTCCAAAAT	GGCTGTTGTCATACTTCTCATGG
Human-COL1A1	GAGGGCCAAGACGAAGACATC	CAGATCACGTCATCGCACAAC
Human-COL3A1	TTGAAGGAGGATGTTCCCATCT	ACAGACACATATTTGGCATGGTT
Human-CTGF	AAAAGTGCATCCGTACTCCCA	CCGTCGGTACATACTCCACAG
Human-TBP	TGCACAGGAGCCAAGAGTGAA	CACATCACAGCTCCCCACCA
Human-HPRT1	GACCAGTCAACAGGGGACAT	CCTGACCAAGGAAAGCAAAG
Human-NQO1	CCTGCCATTCTGAAAGGCTGGT	GTGGTGATGGAAAGCACTGCCT
Human-NOX4	GCCAGAGTATCACTACCTCCAC	CTCGGAGGTAAGCCAAGAGTGT
Human-SLC7A11	TCCTGCTTTGGCTCCATGAACG	AGAGGAGTGTGCTTGCGGACAT
Human-HMOX1	CCAGGCAGAGAATGCTGAGTTC	AAGACTGGGCTCTCCTTGTTGC
Rabbit-GAPDH	AGTATGATTCCACCCACGGC	GATGGCCTTCCCGTTGATGA
Rabbit-TGFB1	CCAAGTGGACATCAACGGGA	ATGTTGAGCCCGTTCCAGAG
Rabbit-CTGF	CACCCGGGTTACCAATGACA	GCTCAAACTTGACCGGCTTG
Rabbit-COL1A1	TGGATTGACCCCAACCAAGG	GAACTGGAAGCCATCGGTCA
Rabbit-COL3A1	GTGAACCCGGTAAGAACGGT	CTGGTTTCCCATCACTGCCT
Rabbit-FN1	TGCACAGACCATACTGTTTTGG	AAGGGGAAGTGGCATAAGGC
Rabbit-LOX	GTGAAAAACCAGGGCACGTC	TGGGGTTGACGCTAACCTTC
Rabbit-ACTA2	AGCACTGTCAGGAATCCCGT	CAGCTCTTGGAGCATCGTCT
Rabbit-HMOX1	GCCACCAAGGAGGTACACAT	GCTTGTTGCGCTCTATCTCC
Rabbit-NQO1	TTCAACCCCGTCCTTTCCAG	ACATGGAGGCGTACGTGTAG
Rabbit-SLC7A11	TCACCATTGGCTACGTGCT	GCCACAAAGATCGGAACTGCT

Western blot analysis

Total protein from LA and CS concentrates and cultured fibroblasts was extracted using RIPA buffer (Applygen, China; Cat. no. C1053) supplemented with protease inhibitors (Applygen, China; Cat. no. P1265). For fibroblasts, nuclear and cytoplasmic proteins were separated using a Nuclear and Cytoplasmic Protein Extraction Kit (Beyotime, China; Cat. no. P0028). Protein concentrations were determined by the BCA assay, and equal amounts of protein were subjected to SDS–PAGE and transferred onto PVDF membranes (Millipore). Membranes were blocked with 5% non-fat milk and incubated with primary antibodies followed by HRP-conjugated secondary antibodies. Signals were developed using ECL reagents and imaged with a Bio-Rad ChemiDoc™ system. Band intensities were quantified using ImageJ and normalized to appropriate loading controls (GAPDH/β-actin/vinculin for total/cytoplasmic proteins and Lamin B for nuclear proteins).

Intracellular GSH/GSSG ratio detection

Intracellular glutathione levels were measured using a GSH/GSSG Assay Kit (Beyotime, China; Cat. no. S0053) following the manufacturer’s instructions. Fibroblasts were seeded in 6-well plates at 2 × 10⁵ cells/well, allowed to adhere for 24 h, and serum-starved for 12 h. Cells were then treated with TGF-β1 (10 ng/mL), LA (100 µg/mL), ML385 (5 µM), or the indicated combinations for 24 h. Total GSH and GSSG were quantified using a microplate reader (412 nm), and the GSH/GSSG ratio was calculated as an index of intracellular redox status.

Rabbit ear scar model and in vivo treatment

All animal procedures were approved by the Institutional Animal Care and Use Committee of Peking University Health Science Center (approval no. DLASBE0417). This study is reported in accordance with the ARRIVE guidelines 2.0. Twelve female New Zealand white rabbits (2.5–3.0 kg) were included in a paired, within-animal design. For each rabbit, the left ear received vehicle (tumescent solution) and the right ear received LA, allowing paired comparison within the same animal.

On day 0, rabbits underwent unilateral inguinal liposuction under general anesthesia by intravenous injection of sodium pentobarbital at 30 mg/kg via the marginal ear vein. After disinfection, approximately 10 mL of tumescent solution (500 mL normal saline supplemented with 15 mL of 2% lidocaine and 1 mL of 0.1% epinephrine) was infiltrated into the groin fat pad, followed by manual aspiration using a 10-mL sterile syringe [35]. The collected lipoaspirate fluid was processed for LA secretome preparation as described above.

A rabbit ear scar model was established by creating four 7-mm full-thickness excisional wounds on the ventral surface of each ear, with removal of the epidermis, dermis, and perichondrium [36]. No secondary re-injury procedure was performed. Wounds were dressed and allowed to heal, and re-epithelialization was typically completed by day 14 [37], at which point intradermal injections were initiated. The left ear received tumescent solution (the solvent used for LA), whereas the right ear received LA secretome (40 µg per injection). The tumescent solution was used as the injection vehicle and contained lidocaine and epinephrine, which may have helped reduce injection-related pain during each intradermal injection. Injections were administered into the scar dermis once weekly.

Rabbits were euthanized at 1, 2, 3, or 4 weeks after the first injection (n = 3 rabbits per time point). Euthanasia was performed by intravenous injection of an overdose of sodium pentobarbital at 100 mg/kg via the marginal ear vein, after which full-thickness scar tissues were harvested for downstream analyses. For quantitative analysis, values from the four scars on each ear were averaged, and the individual rabbit was considered the biological replicate for paired comparison between left and right ears.

Histological staining and quantitative assessment of scar morphology

Excised rabbit ear scar tissues and human scar specimens were fixed in 4% paraformaldehyde, dehydrated, paraffin-embedded, and sectioned. Rabbit ear sections were subjected to hematoxylin and eosin (H&E), Masson’s trichrome, and Sirius Red staining according to the manufacturers’ instructions. For oxidative stress assessment, 4-hydroxynonenal (4-HNE) immunofluorescence staining was performed using an Alexa Fluor 488-conjugated anti-4-HNE antibody (Bioss, Beijing, China; Cat. No. bs-6313R-BF488; 1:400), followed by DAPI nuclear counterstaining. Human scar sections, including keloid and hypertrophic scar tissues, were additionally subjected to immunofluorescence staining for NOX4 using an anti-NOX4 antibody (ABclonal, Cat. No. A22149; 1:400) and for 4-HNE as described above.

Images were acquired for quantitative analysis and analyzed using ImageJ software. For morphometric analysis of rabbit ear scars, measurements were performed at the central and thickest portion of the healed wound on H&E-stained sections, with adjacent normal dermis used as the reference tissue. The scar elevation index (SEI) was calculated as follows: SEI = (total thickness of scar tissue) / (thickness of adjacent normal dermis) [38]. For immunofluorescence analysis, the mean fluorescence intensity (MFI) of 4-HNE or NOX4 within the region of interest was quantified by an investigator blinded to group allocation.

Proteomic profiling and bioinformatic analysis

Proteins were extracted in lysis buffer containing protease and phosphatase inhibitors, clarified by centrifugation (12,000 × g, 4 °C, 10 min), and quantified by BCA assay. Proteins were reduced, alkylated, diluted, and digested with trypsin, and the resulting peptides were desalted on Strata X SPE columns.

Peptides were analyzed by nanoLC–MS/MS on a Vanquish Neo nano-UPLC coupled to an Orbitrap Astral mass spectrometer operated in data-independent acquisition (DIA) mode. Full MS scans were acquired at a resolution of 240,000 over m/z 380–980, and DIA MS/MS spectra were acquired at a resolution of 80,000 using HCD. DIA data were processed with DIA-NN (v1.8) against the Homo sapiens UniProt/Swiss-Prot database (Homo_sapiens_9606_SP_20231220.fasta) using a reverse decoy strategy. Trypsin/P specificity with up to one missed cleavage was assumed, carbamidomethyl (Cys) and N-terminal Met excision were set as fixed modifications, and peptide and protein identifications were filtered at FDR < 1%.

Identified proteins were cross-referenced with ExoCarta to annotate extracellular vesicle–associated proteins. Label-free relative protein abundances were used for quantitative comparisons between groups. Fold change (FC) was calculated from mean intensities, and statistical significance was assessed using two-sided Student’s t tests on log2-transformed intensities. Proteins with FC > 1.5 or FC < 1/1.5 and P < 0.05 were considered differentially expressed; Functional enrichment analyses were performed using Fisher’s exact test with annotations from GO, KEGG, Reactome, and WikiPathways. For GO enrichment, the input list comprised proteins upregulated in LA relative to CS with FC > 1.5 and P < 0.05. For pathway enrichment (KEGG/Reactome/WikiPathways), to focus on robustly induced signals, we used the subset of upregulated proteins with FC > 2 relative to CS.

Single-cell transcriptomic analysis

Pathological scar (PS) tissue, specifically keloid tissue, and matched normal skin (NS) tissue were obtained from two patients with approval from the Peking University Third Hospital Medical Science Research Ethics Committee (approval no. M20250686). Single-cell suspensions were prepared and libraries were constructed using the 10x Genomics platform. Sequencing data were processed with Cell Ranger, yielding 16,591 (N1), 11,397 (P1), 15,852 (N2), and 10,216 (P2) cells for downstream analyses.

Downstream analyses were performed in R (v4.4.0) using Seurat (v5.1.0). Cells with < 300 or > 7,500 detected genes, or > 10% mitochondrial transcripts were removed, and potential doublets were excluded using DoubletFinder (v2.0.4). Datasets were integrated using Harmony (v1.2.3), followed by identification of highly variable genes and PCA. Clustering was performed on the integrated embedding using a shared nearest-neighbor graph, and clusters were visualized using UMAP.

Differentially expressed genes were identified with Seurat FindAllMarkers (Wilcoxon rank-sum test with Bonferroni correction). Cell types were annotated based on canonical markers from the literature [39–41], supported by CellMarker [42] and PanglaoDB [43]. For pathway-level quantification, Reactome gene sets related to ECM organization and NRF2/antioxidant signaling were retrieved via msigdbr (v25.1.1) and scored using Seurat AddModuleScore. Canonical NRF2 target genes were curated from the literature and visualized by heatmaps within fibroblasts across conditions.

SiRNA transfection

Fibroblasts were seeded in 6-well plates at a density of 4.5 × 10^5 cells per well and cultured to approximately 60% confluence before transfection. Cells were transfected with Nrf2 siRNAs or negative control siRNA (sequences listed in Table 2) using the CALNP™ RNAi In Vitro Transfection Kit (D-Nano Therapeutics, China) according to the manufacturer’s instructions, with a final siRNA concentration of 30 nM. After 24 h, total RNA was collected for qPCR analysis, and the transfection medium was replaced with complete culture medium. After 48 h, protein lysates were harvested for immunoblotting to assess knockdown efficiency. The siRNA exhibiting the highest silencing efficiency was selected for subsequent experiments. For functional assays, transfected cells were then treated with TGF-β1 (10 ng/mL) and LA (100 µg/mL) under the same conditions described in Section  "Cell culture and treatment".

Table 2.

Sequences of Nrf2 siRNAs and negative control siRNA used in this study

siRNA	Sense (5’ to 3’)	Antisense (5’ to 3’)
CTRL siRNA	UUCUCCGAACGUGUCACGUTT	ACGUGACACGUUCGGAGAATT
Nrf2 siRNA1	CCAGUUGACAGUGAACUCATT	UGAGUUCACUGUCAACUGGTT
Nrf2 siRNA2	CCCGUUUGUAGAUGACAAUTT	AUUGUCAUCUACAAACGGGTT
Nrf2 siRNA3	CGCUCAGUUACAACUAGAUTT	AUCUAGUUGUAACUGAGCGTT

Statistical analysis

Unless otherwise specified in the figure legends, all data are presented as mean ± standard deviation (SD). Statistical analyses were performed using GraphPad Prism (version 10.1.2). Differences among multiple groups were assessed using one-way analysis of variance (ANOVA) followed by Tukey’s post hoc multiple comparison test. For pairwise comparisons, a paired Student’s t test was used for comparisons between matched wounds from the same rabbit in the rabbit ear in vivo experiments, whereas an unpaired Student’s t test was used for other two-group comparisons, where applicable. A two-tailed P < 0.05 was considered statistically significant. Significance levels are indicated as: P < 0.05 (*), P < 0.01 (**), P < 0.001 (***), P < 0.0001 (****).

Results

Lipoaspirate-derived secretome shows higher particle yield and EV-marker enrichment than cell-culture supernatant-derived secretome

Nanoparticle tracking analysis (NTA) (Fig. 1A-C) showed that LA had a significantly higher particle concentration and a smaller mean particle diameter (134.56 ± 60.50 nm) compared with CS (186.00 ± 105.42 nm) (P < 0.05), consistent with the abundant vesicle-like nanoparticles observed by transmission electron microscopy (Fig. 1D).

Fig. 1.

Characterization of the lipoaspirate-derived secretome (LA) and the cell-culture supernatant–derived secretome (CS). A–C Nanoparticle tracking analysis (NTA) showing particle size distribution and concentration in LA and CS. LA displayed a higher particle concentration and smaller mean diameter than CS (mean ± SD; *P < 0.05). D Representative TEM images showing round, cup-shaped vesicular structures in CS (1) and LA (2). The LA sample was diluted 1:100 in PBS before imaging because of its high particle concentration. E Western blot analysis of EV-associated markers (30 µg/lane)

Western blot analysis further confirmed the presence of classical EV-associated markers (CD63, ALIX, and TSG101) in the LA fraction (Fig. 1E). Under equal protein loading (30 µg per lane), these markers were barely detectable in the CS fraction, likely owing to its extremely low particle yield. As expected from the characteristics of ultrafiltration, the LA preparation also contained minor non-vesicular components, as indicated by weak detection of the endoplasmic-reticulum marker Calnexin.

Together, these data indicate that lipoaspirate-derived LA represents a secretome concentrate with substantially higher particle yield and stronger vesicle-associated marker signals than secretome derived from conventional ADSC culture supernatants (CS).

LA treatment alleviates rabbit ear hypertrophic scarring and attenuates oxidative stress in vivo

In the rabbit ear scar model (Fig. 2A), LA treatment produced a clear therapeutic benefit. Gross inspection showed that LA-treated scars became progressively flatter and less erythematous from W2 onward, with sustained improvement through W4 (Fig. 2B). Consistently, the scar elevation index (SEI) was comparable between groups at W1, but was significantly reduced in the LA group at W2–W4 (Fig. 2C), indicating attenuated scar overgrowth.

Fig. 2.

LA reduces scar overgrowth and improves collagen architecture in vivo. A Schematic overview of unilateral inguinal lipoaspirate for LA preparation and its subsequent intradermal application to rabbit ear scars. B Representative macroscopic images of rabbit ear scars from Day 0 through Week 4 (W1–W4) after treatment initiation, showing reduced scar size, elevation, and erythema in LA-treated ears compared with controls. C H&E staining of rabbit ear scars, with black lines indicating the maximal scar thickness and the thickness of adjacent normal dermis used for SEI calculation, and corresponding quantification of the scar elevation index (SEI). Data are shown as mean ± SD (n = 3) and analyzed by paired Student’s t-test. D Masson’s trichrome staining at W4, illustrating dense, compact collagen bundles in control scars and looser, more dermis-like collagen organization in LA-treated scars. E Sirius Red staining under polarized light at W4 and quantification of the relative proportions of collagen I (red/orange birefringence) and collagen III (green birefringence). F Heatmap of fibrosis-associated and oxidative stress-responsive gene expression in scars from W1 to W4. Values represent fold change (LA vs. CTRL) determined by qPCR. G Representative immunofluorescence images of 4-HNE immunofluorescence staining of rabbit ear scars (CTRL and LA), with DAPI counterstaining, and corresponding quantification of 4-HNE mean fluorescence intensity (MFI). Data are shown as mean ± SD (n = 3 Rabbits). *, **, ***, **** indicate P < 0.05, P < 0.01, P < 0.001 and P < 0.0001 respectively

Histological findings corroborated these macroscopic changes. At W4, Masson’s trichrome staining revealed a clear architectural difference: LA-treated scars exhibited a more dermis-like structure with finer, loosely organized collagen fibers and better-preserved adnexal structures, whereas scars treated with tumescent fluid remained hypercellular with densely packed, disorganized collagen bundles (Fig. 2D). Sirius Red staining further suggested a shift toward a more physiological collagen composition, as evidenced by an increased proportion of collagen III and a reduced collagen I/III ratio (Fig. 2E).

At the molecular level, LA consistently suppressed fibrosis-associated transcripts, including FN1, COL1A1, COL3A1, LOX, and TGF-β1, from W1 to W4 (Fig. 2F). ACTA2 showed a transient increase at W1 and W2, but subsequently decreased under LA treatment. In addition, the oxidative stress-responsive genes HMOX1, NQO1, and SLC7A11 were also downregulated following LA treatment, suggesting attenuation of oxidative stress-related responses within scar tissue.

To further evaluate oxidative damage within scar tissues, we performed 4-HNE immunofluorescence staining at W4. CTRL scars displayed strong 4-HNE signals across the scar dermis, whereas LA-treated scars exhibited markedly weaker staining. Quantification confirmed a significant reduction in 4-HNE mean fluorescence intensity (MFI) in the LA group compared with CTRL (Fig. 2G), indicating that LA attenuated lipid peroxidation and oxidative stress in vivo, consistent with its anti-fibrotic effects observed at the gross, histological, and transcriptional levels.

Together, these data indicate that LA limits scar overgrowth, facilitates more physiological ECM remodeling, and alleviates oxidative stress in vivo.

Proteomic profiling reveals antioxidant enrichment and NRF2-associated signatures in LA compared to CS

To gain insight into the molecular basis of LA’s in vivo activity, we performed comparative proteomic profiling of LA and CS protein concentrates. Cross-referencing identified proteins with ExoCarta indicated that both preparations were EV-enriched, with 86% of proteins annotated as exosome-related (Fig. 3A). Unsupervised clustering clearly separated LA from CS, consistent with distinct secretome compositions (Fig. 3B).

Fig. 3.

Proteomic profiling and transcriptional validation highlight NRF2-associated antioxidant enrichment in LA. A Pie chart showing the overlap of identified proteins with the ExoCarta database, indicating extracellular vesicle (EV) enrichment of the LA secretome. B Heatmap of differentially expressed proteins between LA and CS samples. C Gene Ontology enrichment analysis of proteins upregulated in LA (fold change > 2), shown for Biological Process, Cellular Component, and Molecular Function categories. D Pathway enrichment analysis of LA-upregulated proteins across KEGG, WikiPathways, and Reactome databases, highlighting terms related to glutathione metabolism, KEAP1–NFE2L2 (NRF2) signaling, and cellular response to chemical stress. E qPCR validation of NRF2-associated targets (HMOX1, NQO1, SLC7A11) and NOX4 expression in fibroblasts under the indicated treatments. Data are presented as mean ± SD (n = 3) and analyzed by one-way ANOVA with Tukey’s post hoc test. *, **, ***, **** denote P < 0.05, 0.01, 0.001, and 0.0001, respectively; ns, not significant. F DCFH-DA–based measurement of intracellular ROS, including representative fluorescence images and quantification of relative fluorescence intensity normalized to the CTRL group. *, ** denote P < 0.05 and P < 0.01 versus CTRL, and ###, #### denote P < 0.001 and P < 0.0001 versus the H₂O₂ group

GO analysis of proteins upregulated in LA relative to CS supported a cytoprotective, stress-adaptive profile (Fig. 3C). In the Biological Process and Molecular Function categories, enrichment converged on redox homeostasis and antioxidant activities, together with metabolic processes and protein quality control, consistent with an enhanced capacity to buffer oxidative stress. In the Cellular Component category, enriched terms were mainly related to vesicle-associated luminal compartments and proteasome and peptidase complexes, suggesting enrichment of vesicle-linked components and proteasome-mediated proteolysis.

Pathway enrichment analysis was performed using proteins significantly upregulated in LA relative to CS (fold change > 2) across KEGG, WikiPathways, and Reactome. Enriched pathways consistently highlighted antioxidant and cytoprotective programs, including glutathione metabolism, NRF2 signaling, KEAP1–NFE2L2 regulation, and cellular responses to chemical stress. (Fig. 3D).

Guided by these findings, we next examined NRF2-related antioxidant responses in fibroblasts under TGF-β1 stimulation. LA induced higher expression of HMOX1, NQO1, and SLC7A11 than CS. LA also reduced TGF-β1–induced NOX4 expression, while CS produced a more modest effect (Fig. 3E).

To further assess the functional consequences of these changes, we quantified intracellular ROS accumulation (Fig. 3F). Both CS and LA reduced basal ROS, and upon H₂O₂ exposure, LA pre-treatment resulted in a more substantial reduction in ROS accumulation than CS. Together, these data indicate that, compared with CS, LA is enriched in antioxidant proteins and NRF2-associated signatures and exerts stronger antioxidant effects in fibroblasts, accompanied by greater induction of antioxidant response genes and suppression of NOX4 expression.

Single-cell analysis and tissue validation reveal NOX4 upregulation, impaired NRF2-associated antioxidant programs, and increased oxidative damage in human pathological scars

Single-cell transcriptomic profiling delineated the major cellular constituents of pathological scars and normal skin, including fibroblasts, vascular endothelial cells, lymphatic endothelial cells, immune cells, and epidermal lineages (Fig. 4A). To map potential sources of oxidative stress in scar tissue, we profiled NOX4 expression across major cell types. NOX4 was elevated in fibroblasts as well as vascular and lymphatic endothelial cells in pathological scars (Fig. 4B). Given that fibroblasts represent the principal ECM-producing effector cells in skin fibrosis, we next focused on fibroblasts to examine the relationship between redox imbalance and fibrotic activation.

Fig. 4.

Single-cell transcriptomics reveals NOX4 enrichment and impaired NRF2 antioxidant programs in human pathological scar fibroblasts.A UMAP visualization of integrated single-cell RNA-seq data from paired pathological scar (PS) and normal skin (NS) samples, with major cell populations annotated using canonical markers and database-assisted references. B The proportion of NOX4-expressing cells across major cell types in PS and NS. C-D Fibroblast-focused comparison of pathway activity based on AddModuleScore using Reactome gene sets, including REACTOME_KEAP1_NFE2L2_PATHWAY and REACTOME_EXTRACELLULAR_MATRIX_ORGANIZATION. E Heatmap showing expression patterns of representative NRF2-responsive antioxidant genes in fibroblasts from PS versus NS

Within fibroblasts, module score analysis showed reduced NRF2 antioxidant signaling in pathological scars (Fig. 4C). In parallel, the ECM-related gene set score was markedly increased in scar-derived fibroblasts (Fig. 4D). Consistent with this pattern, a broad panel of NRF2-associated antioxidant genes was coordinately downregulated in fibroblasts from pathological scars compared with normal skin, whereas KEAP1, a negative regulator of NRF2 signaling, was upregulated (Fig. 4E). These transcriptional features suggest impaired antioxidant capacity in scar-associated fibroblasts.

To further assess oxidative status in human scar tissue, we examined normal skin (NS) and pathological scar (PS) samples by NOX4 immunohistochemistry and 4-hydroxynonenal (4-HNE) immunofluorescence staining. In NS samples, NOX4 staining was weak, with limited positive signals in the epidermal basal layer and little to no staining in fibroblast-associated regions of the dermis. In PS samples, NOX4 staining was stronger and more broadly distributed in both the epidermis and dermis, including fibroblast-rich areas (Fig. 5A). Similarly, 4-HNE fluorescence was weak or barely detectable in NS, whereas PS samples showed stronger 4-HNE signals in both epidermal and dermal regions (Fig. 5B).

Fig. 5.

Histological detection of NOX4 and 4-HNE in normal skin and pathological scar tissues. A Representative immunohistochemical staining of NOX4 in normal skin (NS) and pathological scar (PS) tissues. B Representative immunofluorescence staining of 4-hydroxynonenal (4-HNE, green) in NS and PS tissues. Nuclei were counterstained with DAPI (blue). Insets show higher-magnification views of the boxed areas

Together, these findings indicate that pathological scar-associated fibroblasts display features consistent with excessive oxidative stress, accompanied by NOX4 upregulation and impaired NRF2-associated antioxidant programs, with additional histological evidence of increased oxidative damage in scar tissue.

Effects of LA and CS on TGF-β1–driven profibrotic activation in fibroblasts

To compare the antifibrotic activity of LA and CS, we first monitored fibroblast proliferation under TGF-β1 stimulation. IncuCyte live-cell imaging showed that TGF-β1 markedly reduced proliferation compared with untreated controls (Fig. 6A), consistent with a shift toward a profibrotic, matrix-producing phenotype. Co-treatment with LA significantly alleviated this growth suppression from 12 h onward, whereas CS produced a weaker and less consistent effect (Fig. 6B).

Fig. 6.

Effects of LA and CS on TGF-β1–associated fibroblast growth suppression and ECM gene induction. A Real-time proliferation curves acquired by IncuCyte live-cell imaging. Confluence was normalized to baseline (0 h). B Quantification of normalized confluence at 12, 24, 36, and 48 h. Data are shown as mean ± SD (n = 6). One-way ANOVA with Tukey’s post hoc test; *, **, *** indicate P < 0.05, 0.01 and 0.001, respectively. C qPCR analysis of ECM-related genes. Data are presented as mean ± SD and analyzed by one-way ANOVA followed by Tukey’s test. Data are shown as mean ± SD and analyzed by one-way ANOVA with Tukey’s post hoc test

We next examined ECM-related gene expression. qPCR analysis showed that LA or CS alone did not increase basal expression of COL1A1, COL3A1, or CTGF, indicating that neither secretome is fibrogenic under resting conditions. In contrast, TGF-β1 robustly upregulated these genes, confirming fibroblast activation. Both LA and CS attenuated TGF-β1–induced induction of COL1A1 and COL3A1, with no statistically significant difference between the groups. Notably, LA showed a significantly stronger inhibitory effect than CS on CTGF expression under the same conditions (Fig. 6C).

NRF2 activation by LA restores redox homeostasis and attenuates profibrotic signaling

To assess the role of NRF2 signaling in the antifibrotic and redox-modulating effects of LA, we evaluated glutathione homeostasis and NRF2 pathway readouts in fibroblasts treated with TGF-β1 in the presence or absence of the NRF2 inhibitor ML385.

TGF-β1 markedly decreased the intracellular GSH/GSSG ratio, whereas LA substantially restored this redox imbalance (Fig. 7A). Importantly, this restorative effect was largely attenuated by ML385, indicating that NRF2 contributes to the redox-improving action of LA.

Fig. 7.

LA-mediated redox regulation and profibrotic markers under ML385-mediated NRF2 blockade. A GSH/GSSG ratio in fibroblasts treated with TGF-β1, LA, and/or the NRF2 inhibitor ML385. B Nuclear and cytosolic NRF2 protein levels assessed by western blotting after TGF-β1 and/or LA treatment. C Representative immunofluorescence images showing NRF2 localization (green) and nuclei (DAPI, blue) under the indicated treatments, with quantification of the nuclear-to-cytoplasmic NRF2 fluorescence intensity ratio. Scale bar, 20 μm. D qPCR analysis of NRF2-responsive genes under the indicated treatments. E HO-1 protein abundance and densitometric quantification in response to TGF-β1, LA, and/or ML385; Data are shown as mean ± SE (n = 3). F qPCR analysis of ECM- and ROS-related genes under the indicated treatments. G COL1 protein expression and densitometric quantification under the indicated treatments; data are shown as mean ± SE (n = 3). Unless otherwise indicated, data are presented as mean ± SD (n = 3). Statistical significance was determined by one-way ANOVA with Tukey’s post hoc test; *, **, ***, **** indicate P < 0.05, 0.01, 0.001, and 0.0001, respectively

Consistent with enhanced NRF2 signaling, nuclear–cytoplasmic fractionation followed by western blotting showed increased nuclear NRF2 in the TGF-β1 plus LA group compared with TGF-β1 alone (Fig. 7B). Immunofluorescence analysis further confirmed a higher nuclear-to-cytoplasmic NRF2 fluorescence ratio in the TGF-β1 plus LA group (Fig. 7C).

At both the transcriptional and protein levels, LA enhanced NRF2 pathway readouts, and these effects were attenuated by ML385 (Fig. 7D, E). Specifically, LA increased the expression of the NRF2 target genes HMOX1, NQO1, and SLC7A11 (Fig. 7D), accompanied by elevated HO-1 protein abundance (Fig. 7E), and these responses were reduced in the presence of ML385. In parallel, LA attenuated the TGF-β1-induced upregulation of profibrotic and redox-related genes, including COL1A1, COL3A1, NOX4, and CTGF, and this transcriptional effect was partially reversed by ML385 (Fig. 7F). However, although LA decreased COL1 protein levels under TGF-β1 stimulation, ML385 did not clearly restore COL1A1 protein expression in LA-treated fibroblasts despite attenuating NRF2 pathway activity (Fig. 7G).

Together, these results support the involvement of NRF2 signaling in the redox-modulating and antifibrotic effects of LA.

Genetic silencing of Nrf2 partially attenuates the antioxidative and antifibrotic effects of LA

To genetically validate the involvement of NRF2 signaling, three independent siRNAs targeting Nrf2 and a control siRNA were tested in fibroblasts. All three siRNAs reduced Nrf2 expression to varying degrees at both the mRNA and protein levels, with Nrf2-siRNA3 showing the highest knockdown efficiency (Fig. 8A, B). Therefore, Nrf2-siRNA3 was selected for subsequent experiments.

Fig. 8.

Genetic validation of NRF2 involvement in the effects of LA in TGF-β1-treated fibroblasts. A, B Three independent siRNAs targeting Nrf2 and a control siRNA were transfected into fibroblasts. Knockdown efficiency was evaluated by RT-qPCR and western blotting. C–I Fibroblasts transfected with control siRNA or Nrf2-siRNA3 were treated with TGF-β1 and/or LA under the indicated conditions, and the expression of NRF2-responsive genes and fibrosis-/ROS-related genes was analyzed by qPCR. J, K Protein expression of HO-1, NQO1, COL1A1, and α-SMA was assessed by western blotting, with densitometric quantification. L, M Intracellular ROS levels were detected by DCFH-DA staining and quantified by fluorescence intensity. Data are presented as mean ± SD (n = 3). Statistical significance was determined by one-way ANOVA with Tukey’s post hoc test; *, **, ***, and **** indicate P < 0.05, 0.01, 0.001, and 0.0001, respectively

Silencing Nrf2 attenuated the transcriptional effects of LA on most antioxidant- and fibrosis-related genes. In control siRNA-transfected cells, LA increased the expression of NRF2-responsive genes, including NQO1 and SLC7A11, while suppressing the TGF-β1–induced upregulation of profibrotic genes such as NOX4, COL1A1, COL3A1, and CTGF (Fig. 8C–I). These effects were markedly weakened after Nrf2 knockdown, although they were not completely abolished.

Notably, HMOX1 behaved differently from the other NRF2 target genes. Even in Nrf2-silenced cells, HMOX1 expression remained significantly induced under combined TGF-β1 and LA treatment (Fig. 8D), indicating that HMOX1 regulation under these conditions is not exclusively dependent on NRF2. This finding further suggests that the induction of HMOX1 by LA is only partially mediated by NRF2 and may also involve additional signaling pathways.

Western blot analysis further supported the transcriptional findings (Fig. 8J, K). In control siRNA-transfected fibroblasts, LA increased the protein expression of the NRF2 downstream targets HO-1 and NQO1, while reducing the TGF-β1-induced accumulation of the fibrotic markers COL1A1 and α-SMA. Following Nrf2 knockdown, the LA-induced upregulation of NQO1 was markedly attenuated, and the inhibitory effects of LA on COL1A1 and α-SMA protein expression were also significantly weakened. In contrast, HO-1 protein expression remained elevated in response to LA even in Nrf2-silenced cells, consistent with the mRNA data and suggesting that HO-1 may also be regulated by additional NRF2-independent mechanisms under these conditions.

Intracellular ROS production was assessed using the DCFH-DA fluorescent probe (Fig. 7L, M). In control siRNA-transfected fibroblasts, H₂O₂ markedly increased ROS accumulation, whereas LA substantially attenuated this effect. In contrast, after Nrf2 knockdown, the ROS-lowering effect of LA was significantly weakened, with ROS levels remaining elevated despite LA treatment.

Taken together, these findings demonstrate that Nrf2 knockdown partially abrogated the antioxidative and antifibrotic effects of LA at the mRNA, protein, and intracellular ROS levels, supporting a substantial contribution of NRF2 to LA activity. The partial preservation of these responses following Nrf2 silencing, along with sustained HMOX1/HO-1 induction, further indicates the involvement of additional NRF2-independent pathways.

Discussion

Our study identifies lipoaspirate fluid as an unexpectedly rich source of a bioactive, acellular secretome and demonstrates that a simple ultrafiltration workflow can convert this routinely discarded surgical by-product into a therapeutic preparation that mitigates pathological scar formation. Compared with secretome derived from conventional ADSC culture supernatants, LA displayed a distinct and more complex bioactive composition, together with stronger anti-fibrotic effects in fibroblast assays. Because adipose tissue is exposed to transient ischemia, hypoxia, and mechanical shear during aspiration [44], we speculate that liposuction may induce a stress-adaptive secretory response that shapes the unique molecular milieu captured in lipoaspirate fluid.

From a translational perspective, LA offers clear practical advantages over conventional cell culture-derived secretome products. Its preparation does not require in vitro expansion, conditioned-medium collection, or prolonged manufacturing workflows. Instead, LA can be generated directly from freshly obtained lipoaspirate fluid by ultrafiltration within about an hour. Routine abdominal liposuction can generate approximately 3–5 L of aspirate, and in our workflow only ~ 15 mL is sufficient for a single LA preparation. When normalized to the same starting volume, lipoaspirate fluid yielded secretome concentrates that were approximately four orders of magnitude greater than those obtained from ADSC culture supernatants. This difference directly impacts feasibility for repeat dosing, batch production, and standardization.

Additionally, ultrafiltration is expected to enrich a heterogeneous mixture of acellular components rather than a single class of bioactive particles. Immunoblotting detected APOA1 in LA (Fig.S1), a principal component of high-density lipoproteins (HDL). In vivo, APOA1 self-assembles into HDL nanoparticles that undergo continuous remodeling during maturation [45]. HDL has also been reported to carry circulating miRNAs in plasma [46] and to exert anti-inflammatory, antioxidant, and anti-atherosclerotic effects [47, 48]. The presence of APOA1 in LA likely reflects contributions from interstitial fluid and/or minor bleeding during aspiration. Together, these observations suggest that LA contains HDL-associated and other non-vesicular constituents alongside vesicle-associated components, further supporting its nature as a complex and heterogeneous secretome preparation.

LA showed clear therapeutic benefit in vivo. In a rabbit ear scar model, intradermal LA administration reduced scar elevation and improved dermal architecture and collagen organization, accompanied by a reduced collagen I/III ratio and downregulation of profibrotic markers including TGF-β1, LOX, COL1A1, COL3A1, FN1, and CTGF. Notably, LA also attenuated oxidative stress in scar tissues, as evidenced by decreased 4-HNE immunofluorescence intensity.

Our data suggest that modulation of redox homeostasis may contribute to the antifibrotic effects of LA. NRF2 signaling is a key protective regulator in many fibrotic disorders [49, 50]. Single-cell analysis indicated that pathological scar fibroblasts exhibit a redox imbalance, characterized by elevated NOX4 expression together with suppressed NRF2 antioxidant signaling and reduced expression of canonical NRF2 target genes. Comparative proteomic profiling of LA versus CS revealed enrichment of cytoprotective pathways, including glutathione metabolism and NRF2-associated stress responses. Together, these findings support restoration of redox balance as one component of LA-mediated scar attenuation.

In fibroblasts, LA promoted NRF2 nuclear accumulation, improved glutathione redox buffering, reduced intracellular ROS, and induced antioxidant genes including HMOX1, NQO1, and SLC7A11. LA also attenuated TGF-β1-driven profibrotic activation, consistent with disruption of the TGF-β1–ROS feed-forward loop. Both pharmacologic inhibition of NRF2 with ML385 and genetic silencing of NRF2 by siRNA blunted LA-induced antioxidant responses and partially reversed suppression of profibrotic transcripts, indicating that NRF2 contributes importantly to the actions of LA in fibroblasts. However, because these interventions only partially reversed the antifibrotic effects of LA, its actions are unlikely to be fully explained by NRF2-dependent redox regulation alone and may also involve additional pathways.

Interestingly, regulation of antioxidant target genes appeared to be context dependent. Whereas LA induced HMOX1, NQO1, and SLC7A11 in cultured fibroblasts, these genes were reduced in scar tissue after LA treatment together with lower 4-HNE levels, suggesting that resolution of oxidative stress in vivo may reduce the need for sustained compensatory stress-response signaling at the tissue level.

This study has several limitations. First, LA is a tissue-derived, ultrafiltration-enriched mixture rather than a single defined entity, and the dominant cellular sources and specific active components responsible for its bioactivity remain unclear, which may complicate its standardization and clinical translation. Second, evidence for NRF2 involvement was supported in vitro by both pharmacologic inhibition and siRNA-mediated silencing in fibroblasts; However, the in vivo experiments did not include direct NRF2 gain- or loss-of-function approaches, limiting mechanistic validation in scar tissue. Third, the in vivo work relied on within-animal comparisons of LA versus vehicle control, and no CS group was included because of ethical and feasibility constraints in rabbits. In addition, the in vivo sample size was small, which limit the robustness of these findings. Furthermore, in the absence of re-injury, this model represented a simple scar model rather than a hypertrophic scar model and may not fully recapitulate the pathological features of hypertrophic scarring. The limited availability of rabbit-reactive antibodies also restricted pathway-level validation in scar tissue. Future studies should therefore incorporate larger sample sizes, improved modeling approaches, additional comparison groups, and animal models that permit genetic or pharmacologic validation of the underlying mechanisms in vivo. Lastly, due to ethical constraints, the number of patients from whom both scar and adjacent normal skin samples could be obtained was limited, resulting in a small sample size for single-cell analysis. This may affect the generalizability of our findings. Although our analysis revealed an imbalance of oxidative stress and NRF2 signaling in scar tissue, consistent with studies in other fibrotic organs, larger cohorts are needed to validate these results. Additionally, the patterns of oxidative stress and NRF2 activation may vary across scar locations due to differences in local skin tension. Future studies should involve larger patient cohorts and genetically modified animal models to further investigate these findings and better address the limitations of this study.

Conclusion

Our study identifies lipoaspirate-derived secretome (LA) as an acellular therapeutic approach that mitigates pathological scar fibrosis by rebalancing redox homeostasis and restraining profibrotic fibroblast activation. Mechanistically, LA appears to exert these effects through NRF2-associated antioxidant responses, suppression of NOX4-related oxidative stress, and likely additional redox- and fibrosis-regulating pathways, thereby helping to interrupt the TGF-β1–ROS feed-forward circuit and promote improved scar remodeling in vivo. Furthermore, single-cell transcriptomic analysis of human keloids, together with histological observations in human pathological scars, supports the presence of oxidative stress-related damage and compromised NRF2-related antioxidant capacity in scar fibrosis. Collectively, these findings position LA as a promising cell-free therapeutic strategy for targeting redox dysregulation and fibrotic remodeling in pathological scars.

Abbreviations

ADSC

Adipose-derived stromal cells

CS

ADSC culture supernatant–derived secretome

ECM

Extracellular matrix

EV(s)

Extracellular vesicle(s)

GSH/GSSG

Reduced glutathione/oxidized glutathione

LA

Lipoaspirate-derived secretome

NOX4

NADPH oxidase 4

NRF2

Nuclear factor erythroid 2–related factor 2

ROS

Reactive oxygen species

SEI

Scar elevation index

TGF-β1

Transforming growth factor-beta 1

Authors’ contributions

Xin Huang : Conceptualization, Formal analysis, Investigation, Methodology, Project administration, Software, Supervision, Validation, Visualization, Writing – original draft. Writing – review & editing. Pengbing Ding : Methodology, Resources, Software. Zhixuan Sun : Methodology, Software. Haibo Xiang : Methodology, Project administration. Muqian Wei : Methodology, Project administration. Hongsen Bi: Conceptualization, Funding acquisition, Supervision, Validation. Zhenmin Zhao : Conceptualization, Data curation, Funding acquisition, Supervision, Validation, Writing – review & editing.

Funding

This study was supported by Beijing Municipal Nature Science Foundation (Grant No. 7252151).

Data availability

The datasets supporting the conclusions of this article are included within the article and its additional files.

Declarations

Ethics approval and consent to participate

The study protocol involving human participants and/or human tissue was approved by the Medical Science Research Ethics Committee of Peking University Third Hospital (approval no. M20250915, A study on the collection of adipose-derived stem cells and extracellular vesicles from lipoaspirate obtained from healthy donors, approved on 20 October 2025; and approval no. IRB00006761-M20250686, Mechanistic study of YAP lactylation in regulating Treg cell function in pathological scars and its crosstalk with fibroblasts, approved on 4 August 2025). Written informed consent was obtained from all relevant patients prior to participation. All animal procedures were approved by the Institutional Animal Care and Use Committee (IACUC) of Peking University Health Science Center (approval No. DLASBE0417; protocol title: Extracellular Vesicles Derived from Liposuction Aspirate Inhibit Hypertrophic Scar Formation; approved on 18 April 2025).

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.