Subject enrollment

The inclusion criteria for the LoS patient group were as follows: (i) females aged 18–60 years old, (ii) newly diagnosed patients with LoS who have not received steroid or fat grafting treatment, and (iii) presence of facial contour deformities and planning to undergo AFG surgical treatment at the department of plastic surgery in Peking Union Medical College Hospital. The exclusion criteria included: (i) suffering from cancer or infectious diseases and (ii) suffering from other skin-affecting diseases. As controls, healthy individuals who underwent liposuction surgeries for cosmetic purposes were also recruited. The two groups were required to be matched in terms of sex, age, body mass index (BMI), and ethnicity.

Liposuction and sample collection

Subcutaneous adipose tissue specimens were collected from the thighs of healthy donors or the nonlesional thighs of patients with linear scleroderma lesions on the face. Both groups underwent the same liposuction procedures performed by the same plastic surgery division. Before liposuction, general anesthesia was administered. To prepare for liposuction, a tumescent solution was injected into the thigh. The solution contained 0.025% lidocaine and 1:1,000,000 epinephrine in 1000 ml of fluid. Adequate infiltration was confirmed by tissue blanching and moderate tension. Liposuction was performed using 3 mm blunt-tip cannulas and 20 ml syringes in the deep fat layer. The syringes were then left undisturbed for 10 min to allow separation of the adipose tissue from the tumescent fluid. The harvested adipose tissues were transferred to 50 ml sterile centrifuge tubes containing 20 ml of Dulbecco’s modified Eagle’s medium (10569044, Gibco) cell culture medium and kept on ice during transportation.

SVF isolation

To prepare the fat tissue for SVF isolation, it underwent several washes with Hank’s balanced salt solution (14025126, Gibco). Thereafter, it was digested at 37 °C for 30 min using 0.15% collagenase I (17100017, Gibco) supplemented with 4% penicillin–streptomycin (15140122, Gibco). Following centrifugation at 1000 rpm for 10 min, the resulting cell pellet was resuspended in high-glucose Dulbecco’s modified Eagle’s medium (10569044, Gibco) containing 10% fetal bovine serum (10099141, Gibco). The suspension was then filtered through a 100-μm strainer and centrifuged at 1500 rpm for 5 min. Subsequently, the obtained cell suspension was resuspended in Hank’s balanced salt solution, and red blood cell lysis buffer was added at room temperature for 5 min to eliminate red blood cells. Another round of centrifugation was performed, and the cell pellet was resuspended in Hank’s balanced salt solution with 0.04% bovine serum albumin (A1933-5G, Sigma) before being filtered through a 40-μm strainer. Finally, the cells were centrifuged and resuspended in Dulbecco’s phosphate buffered saline without Ca2+ and Mg2+ (14190144, Gibco). To assess cell concentration and viability, the resulting single-cell suspension was incubated with an equal volume of AOPI Staining Solution (Logos Biosystems) and analyzed using a LUNA-FL Fluorescence Cell Counter (Logos Biosystems).

Single-cell transcriptomic sequencing

The Chromium Single-Cell 3′ Reagent Kit v3 (10 × Genomics, USA) was used for various steps, including single-cell gel bead-in-emulsion (GEM) making, post-GEM-RT cleanup, barcoding, cDNA amplification, and cDNA library construction, following the manufacturer’s protocol. Library sequencing was carried out using the NovaSeq 6000 system (Illumina, USA).

Preprocessing of the scRNA-seq data

The sample demultiplexing, barcode processing, and unique molecular identifier (UMI) counting were performed using the official software Cell Ranger (v3.0.2). The FASTQ data was demultiplexed using “cellranger mkfastq,” and the resulting gene-barcode matrix for each library was generated with the “cellranger count” pipeline. To reduce noise, genes expressed in fewer than three cells were excluded, and potential poor-quality cells were removed based on their expressed gene number, UMI count sum, and percentage of mitochondrial genes. Scrublet (v0.2.3) was used to predict and remove doublets. The quality control thresholds for each sample are provided in Additional file 1: Table S1. Cells with enriched hemoglobin gene expression were identified and removed as potential red blood cells. The UMI counts for each cell were normalized to 10,000 and log-transformed. For each sample, the “FindVariableFeatures” function of Seurat (v3.1.0) was applied to select 2000 features (genes). The canonical correlation analysis in Seurat was used to integrate the datasets and correct potential batch effects under default settings. Linear regression was used to regress out variation from mitochondrial gene proportion, S phase score, G2M phase score, and UMI count. Principal component analysis (PCA) was used for linear dimensional reduction, with the first 30 principal components used to construct the neighborhood graph of the cells (Fig. S1). The uniform manifold approximation and projection (UMAP) algorithm was used to embed the graph in a two-dimensional space. The cells were then clustered using Louvain clustering (resolution = 0.6) implemented in Seurat.

Cellular annotation through reference mapping

Cellular annotation was performed using the data transfer workflow of Seurat with the scRNA-seq data of SVF that has previously been published by our lab [11] as a reference. Briefly, the function FindTransferAnchors was used to find anchors between two datasets. Then, the function TransferData was used to classify the query cells based on reference data.

Differential compositional testing

To detect statistically credible alterations in cellular composition derived from the single-cell dataset, we used a Bayesian approach implemented in scCODA (v0.1.9) [13] (reference_cell_type = “automatic,” Hamiltonian Monte Carlo sampling method with default settings).

Differential expression analysis

The identification of differentially expressed genes (DEGs) between groups was performed using the FindMarkers function of Seurat. The test utilized for this analysis was the likelihood-ratio test with the “bimod” option. Genes were considered significantly different in expression between the two groups if they showed an absolute log2-fold change exceeding 0.25 and an adjusted P-value lower than 0.05.

Gene set enrichment analysis

We first ranked all the expressed genes using the Signal2Noise method, which normalized the mean difference between LoS and the control (CTRL) with the standard deviation. The resulting ranked list of genes was subsequently imported into the GSEA software (version 4.0.1). A statistically significant threshold was established with an FDR q value below 0.05. For the analysis, the precompiled REACTOME pathways from MSigDB (version 7.0) were utilized. The visualization of the results was accomplished using the EnrichmentMap plugin incorporated in Cytoscape (version 3.7.0).

High-dimensional weighted gene coexpression network analysis

The high-dimensional weighted gene coexpression network analysis (hdWGCNA) was conducted using the hdWGCNA R package (v0.2.04) [14] with default parameters. This package has been specifically designed to analyze high-dimensional scRNA-seq data. Within each module, the hub genes were determined by identifying the top 25 genes ranked by eigengene-based connectivity (kME).

Intercellular communication analysis

The CellChat (v1.6.1) package [15] was utilized to deduce interactions between cell types within each group and detect changes in intercellular communication through comparative analysis. Briefly, the approach involved identifying overexpressed ligands or receptors for each cell type and quantifying potential interactions between any two cell types using a communication probability value. Significant interactions (P-value < 0.05) were identified by employing a permutation test, where cell type labels were randomly permuted, and the communication probability was recalculated. The use of pattern recognition techniques helped detect prominent incoming and outgoing signal patterns for each cell type in each condition. Network centrality analysis was used to infer the main sources and targets of the signaling network for a particular pathway. The identification of significantly altered signaling pathways was achieved through a comparative analysis of the overall information flow within each signaling pathway by assessing the sum of communication probability among all pairs of cell types in the inferred network.

Fluorescence-activated cell sorting

To eliminate dead cells, freshly isolated SVF was suspended in phosphate-buffered saline (PBS) containing 1:1000 Zombie NIR dye (423105, BioLegend) and incubated at room temperature in the dark for 15 min. After washing with cell staining buffer (420201, BioLegend), the cells were resuspended and incubated with cellular surface antibodies, including PE anti-human CD140a (PDGFRα) antibody (323506, BioLegend) and APC anti-human CD55 antibody (311312, BioLegend), at room temperature in the dark for 30 min. The cells were washed twice with cell staining buffer by centrifugation at 350×g for 5 min. Flow cytometry data were obtained using the LSRFortessa flow cytometer (BD Biosciences) and analyzed using the FlowJo software (BD Biosciences). The CD55high ASCs were finally sorted using fluorescence-activated cell sorting (FACS) on Beckman Moflo Astrios EQ (Beckman-Coulter).

Bleomycin-induced mice models for skin fibrosis and ASC treatment

To compare the anti-fibrotic effects of CD55high and CD55low ASCs in patients with LoS, we established a mouse model using bleomycin (BLM) induction to mimic the skin fibrosis and subcutaneous adipose tissue loss observed in human patients. To avoid potential immune rejection from xenografts, partially immunodeficient Balb/c nude mice were used since the ASCs used for treatment were isolated from healthy human donors. A total of 20 female Balb/c nude mice aged 6 weeks were randomly assigned to one of four groups: PBS control, model control, CD55high ASC subcutaneous injection, and CD55low ASC injection groups. During model construction, all mice received daily subcutaneous injections of 20 μg of BLM (HY-17565, MedChemExpress, NJ, USA) dissolved in 100 μl of PBS in the lower back for 30 days, except for the PBS control group, which received the same volume of PBS without BLM. The successful establishment of skin fibrosis models was confirmed by skin appearance and histological staining. A day after the last injection of BLM, freshly isolated CD55high ASCs and CD55low ASCs were subcutaneously injected into the lower back of the mice at a concentration of 1 × 104 cells per animal, suspended in 100 μl of PBS. The PBS control and model control groups received 100 μl of PBS without ASCs at the same site. After 30 days, the mice were sacrificed, and skin samples from the lower back were collected for further experiments. All 20 mice survived the treatments and were included in further analysis.

In vivo fluorescence imaging

Three mice were injected subcutaneously with bleomycin to induce a scleroderma skin fibrosis model, as described above. ASCs at a concentration of 5 × 107 cells/mL were labeled with 2.5 µmol/L DiD by incubating them at 37 °C for 20 min. The DiD-labeled ASCs were then injected subcutaneously into the lower back of the mice at a dose of 5 × 106 cells per animal. Fluorescence signal intensity was quantified using a PerkinElmer IVIS Lumina III system at days 0, 2, 7, 14, 21, 28, 35, and 46. Relative signal intensity was determined by comparing the fluorescence signal at each time point with that at day 0.

Paraffin section preparation

Skin samples from the mice were fixed overnight at 4 ℃ in a 4% paraformaldehyde solution (G1101, Servicebio, Wuhan, China) and subsequently embedded in paraffin. Deparaffinization of 5-μm sections was carried out using xylene, followed by rehydration in alcohol and rinsing in distilled water before further staining.

Hematoxylin and eosin staining

The rehydrated sections were stained using the hematoxylin and eosin (HE) staining kit (G1005, Servicebio, Wuhan, China) following the manufacturer’s protocol. After staining, the sections were dehydrated and covered with neutral balsam mounting medium (10004160, Sinopharm, Shanghai, China). Scanning of the stained sections was performed using the Pannoramic SCAN (3DHISTECH) system. The thickness of the epidermis, dermis, and subcutaneous adipose tissue was measured from six random regions per section using ImageJ software.

Masson’s trichrome staining

The rehydrated sections were stained using a Masson’s trichrome staining kit (G1006, Servicebio, Wuhan, China) following the manufacturer’s protocol. After staining, the sections were scanned using the Pannoramic SCAN (3DHISTECH) system. The blue-stained collagen area was quantified using ImageJ software from three random regions per section. The collagen volume fraction (CVF) was then calculated using the formula: CVF (%) = collagen area / full area × 100%.

Immunohistochemistry staining

The rehydrated sections were subjected to antigen retrieval by heating in sodium citrate buffer (pH 6.0). To eliminate endogenous peroxidase, 3% H2O2 was used. Subsequently, the sections were blocked with 3% bovine serum albumin (GC305010, Servicebio, Wuhan, China) for 30 min and incubated with primary antibodies overnight at 4℃. After washing off unbound primary antibodies, horseradish peroxidase (HRP)-conjugated secondary antibodies were applied, followed by DAB treatment and hematoxylin staining. The primary antibodies used included TGFβ1 (GB14154-50, Servicebio, Wuhan, China), Adiponectin (21613-1-AP, Proteintech, Wuhan, China), and PPARγ (16643-1-AP, Proteintech, Wuhan, China). The sections were dehydrated, covered with mounting medium, and scanned. Integrated optical density (IOD) values were measured from six random regions per section using ImageJ.

Western blotting

Mice skin samples were lysed using cold RIPA buffer (P0013, Beyotime Biotechnology, Shanghai, China) supplemented with protease/phosphatase inhibitors (P1045, Beyotime Biotechnology, Shanghai, China) using a tissue grinder. The lysed samples were then centrifuged at 12,000g for 10 min to remove debris. Protein concentrations were determined using the Thermo Scientific Pierce BCA Kit (23225, Thermo Scientific). For western blotting analysis, 20–40 μg of protein per sample was loaded onto a 10% sodium dodecyl-sulfate (SDS)–polyacrylamide gel and separated by electrophoresis. The proteins were then transferred onto 0.45 μm PVDF membranes and blocked with 5% nonfat milk (P0216, Beyotime Biotechnology, Shanghai, China) in Tris-buffered saline with tween-20 (TBST, ST671, Beyotime) for 1 h at room temperature. Primary antibodies against collagen I (1:2000, ab260043, Abcam) and α-SMA (1:2000, 14976082, Invitrogen) were incubated overnight at 4℃. Glyceraldehyde 3-phosphate dehydrogenase (GAPDH; 1:2000, 2118L, Cell Signaling Technology) was used as a control. After washing off the unbound antibodies with TBST, the membranes were incubated with secondary antibodies conjugated with HRP for 1 h at room temperature. The protein bands were detected using enhanced chemiluminescence (ECL; P0018A, Beyotime Biotechnology, Shanghai, China), and the intensity of the bands was quantified using ImageJ.

Statistical analysis

Statistical analyses were performed using GraphPad Prism (version 9). Two-tailed Wilcoxon rank-sum test or Student’s t-test was used to compare two groups of data, while one-way analysis of variance (ANOVA) was used to analyze data among multiple groups. Following ANOVA, a post hoc test was performed using Tukey’s honest significant difference (HSD). A P-value of less than 0.05 was considered statistically significant, denoted by *, **, ***, and **** to indicate P-values of less than 0.05, 0.01, 0.001, and 0.0001, respectively.