The NIH/3T3 mouse embryo fibroblast, HFL-1 human fetal lung fibroblast and RAW264.7 mouse monocyte macrophage leukemia cell lines were obtained from the National Collection of Authenticated Cell Cultures. NIH/3T3 and RAW264.7 cells were cultured in complete DMEM (Gibco; Thermo Fisher Scientific, Inc.), while HFL-1 cells were maintained in complete Ham's F-12K medium (Procell Life Science & Technology Co., Ltd.). All media were supplemented with 10% fetal bovine serum and 1% penicillin-streptomycin (Gibco; Thermo Fisher Scientific, Inc.). The cultures were incubated at 37°C with 5% CO₂.

A comprehensive analysis was performed using the GEO database (13) to explore gene expression differences between normal and IPF-affected lung tissues. The focus was on the GSE110147 dataset, which includes expression profiles from 11 normal lung tissue samples (age 52±18 years; Male/Female 7/4) and 22 samples from patients with IPF (age 62±6 years; Male/Female 14/5). Detailed patient information can be found in the referenced literature (14). Differential expression analysis was performed using GEO2R (13) to compare gene expression across sample groups. A significance threshold was set with a Log₂ fold change of 1 and a P-value of 0.05. The Benjamini and Hochberg method was applied to adjust P-values for multiple comparisons, controlling the false discovery rate. A volcano plot was generated using the bioinformatics cloud platform (https://www.bioinformatics.com.cn) for data visualization.

For in vitro studies, NIH/3T3 fibroblasts were treated with or without 5 ng/ml recombinant TGF-β1 (R&D Systems, Inc.) for 12 h prior to RNA collection for transcriptomic analysis (n=3 per group). Transcriptomic data for these samples were derived from our prior study (15) and were re-analyzed in the context of the present work. For in vivo studies, a total of six specific pathogen-free (SPF) male C57BL/6J mice (aged 6-8 weeks, weight 22-24 g) were used. Mice were purchased from Guangdong Vital River Laboratory Animal Technology Co., Ltd. and housed in the SPF-grade animal facility at the Shenzhen Institute of Technology, Hong Kong Polytechnic University. Housing conditions were maintained at a temperature of 22±2°C, a 12-h light/dark cycle and a humidity of 50±10%. Mice were randomly assigned to either a control group or a bleomycin (BLM)-induced fibrosis model group. After anesthesia with 3-4% isoflurane, pulmonary fibrosis was induced in the model group by intratracheal administration of 2.5 mg/kg BLM (Shanghai Macklin Biochemical Technology Co., Ltd.), while control mice received an equivalent volume of sterile saline. Three weeks post-treatment, lung tissues were harvested for transcriptomic analysis (n=3 per group). Transcriptome sequencing was conducted by Wekemo Tech Group using the Illumina platform (Illumina, Inc.). Gene expression levels were quantified in Fragments Per Kilobase of transcript per Million mapped reads and differential expression analysis was performed using the DESeq2 R package v1.26.0 (16). Genes were considered differentially expressed if their adjusted P-values were <0.05 and Log₂ fold change >1.

SPF male C57BL/6J mice were randomly assigned to either a control group (saline; n=3) or a pulmonary fibrosis model group (BLM 2.5 mg/kg intratracheal; n=3). Lungs were harvested 14 days post-exposure, preserved in sCelLive solution (Singleron Biotechnologies GmbH), minced, and digested with sCelLive dissociation solution using the PythoN™ System (Singleron Biotechnologies GmbH). Cell suspensions were filtered through a 40 μm mesh, red blood cells were lysed using GEXSCOPE RCLB (Singleron Biotechnologies GmbH), and cell viability was assessed via Trypan Blue staining (Bio-Rad Laboratories, Inc.) at room temperature for 2-3 min. Single-cell suspensions (2×10⁵ cells/ml) were processed on the Matrix System (Singleron Biotechnologies GmbH) for barcoding, reverse transcription, cDNA amplification, and library construction using GEXSCOPE Kits (Singleron Biotechnologies GmbH). Libraries (4 nM) were sequenced on an Illumina NovaSeq 6000 (150 bp PE; Illumina, Inc.). Raw reads were trimmed using Cutadapt v1.17 (17), aligned to GRCm38 with STAR v2.6.1a (18), and Unique Molecular Identifier (UMI)/gene counts were generated using featureCounts v2.0.1 (19). Using Scanpy v1.8.2 (20), cells were filtered based on gene count (<200 or top 2%), UMI count (top 2%) and mitochondrial content (>50%); genes detected in fewer than 5 cells were excluded, resulting in 76,995 cells retained for analysis. Potential doublets were identified and filtered using Scrublet v0.2.3 (21). Counts were normalized, log-transformed, and the top 2000 variable genes were subjected to Principal Component Analysis. Louvain clustering (resolution=1.2, 32 clusters) using the top 20 PCs was visualized via Uniform Manifold Approximation and Projection (UMAP). Differentially expressed genes (DEGs) were identified with Seurat's (v3.2.3) FindMarkers (22) using the Wilcoxon test (expressed in >10% of cluster cells, mean logFC >0.25). Clusters were annotated using canonical markers (DEGs/literature) and heatmaps. Cell types were assigned via the SynEcoSys database (www.synecosys.com).

Total proteins were isolated from cells or tissue samples using PIPA lysis buffer (Thermo Fisher Scientific, Inc.). Protein concentration was determined using the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific, Inc.) according to the manufacturer's instructions. Equal amounts of protein (20 μg per lane) were separated by SDS-PAGE [Shenzhen Dakewe Bio-engineering Co., Ltd.; cat. no. 8012011; suitable for the separation and identification of (10~250) kDa proteins, without the need to adjust the separation gel concentration based on protein size]. Proteins were then transferred onto PVDF membranes (MilliporeSigma). After blocking for 60 min at room temperature with 5% non-fat milk, the membranes were incubated overnight at 4°C with primary antibodies. Antibodies for Collagen type I α 1 chain (COL1A1; cat. no. 72026; 1:1,000), Phospho-STAT6 (Tyr641; cat. no. 56554S; 1:1,000) and α-smooth muscle actin (α-SMA; cat. no. 19245; 1:1,000) were obtained from Cell Signaling Technology, Inc. The HAS2 (cat. no. bs-11290R; 1:1,000) antibody was purchased from BIOSS and HRP-conjugated β-tubulin (cat. no. HRP-66240: 1:20,000) and STAT6 (cat. no. 51073-1-AP; 1:5,000) antibodies were obtained from Proteintech Group, Inc. Membranes were incubated for 1 h at room temperature with an HRP-conjugated secondary antibody (Cell Signaling Technology, Inc.; 1:2,000) following TBS-T (containing 0.05% Tween-20) washing. Protein bands were detected using Immobilon Western Chemiluminescent substrate (MilliporeSigma) and imaged with a Tanon 5200 Multi automatic chemiluminescence system (Tanon Science and Technology Co., Ltd.).

RNA extraction, cDNA synthesis and qPCR were performed according to the manufacturers' protocols. Total RNA was extracted from cells (1-2×10⁶) or tissue samples using the RNA extraction kit (Accurate Biology). RNA was then reverse transcribed into cDNA using the First-Strand cDNA synthesis kit (TransGen Biotech Co., Ltd.). The SYBR Green Premix qPCR Kit (Accurate Biology) was used to set up the real-time PCR system, and analysis was performed on the Roche LightCycler 480 II (Roche Diagnostics, Ltd.). The thermal cycling conditions were as follows: Initial denaturation at 95°C for 30 sec; followed by 40 cycles of denaturation at 95°C for 5 sec, annealing/extension at 60°C for 30 sec. Data were analyzed using the 2^−ΔΔCq method for relative quantification (23), with normalization to GAPDH mRNA levels. Primer sequences for the target genes are provided in Table SI.

Cell culture supernatants and mouse serum were collected and analyzed using ELISA assay kits according to the manufacturer's instructions. HA content was measured with the HA ELISA kit (Beijing Solarbio Science & Technology Co., Ltd.; cat. no. SEKH-0509), while TGF-β1 levels were assessed using the Mouse TGF-β1 ELISA Kit (Proteintech Group, Inc.; cat. no. KE10005).

Primary lung fibroblasts were isolated from C57BL/6J mice in both the control (saline-treated) and BLM-induced fibrosis model groups, following an established protocol with minor modifications (24). Briefly, lung tissues were aseptically collected, minced, and subjected to sequential enzymatic digestion with a collagenase/dispase cocktail (1 mg/ml) followed by 0.25% Trypsin-EDTA. After RBC lysis, the cell suspension was resuspended in fibroblast growth medium (DMEM with 10% FBS, 1% penicillin/streptomycin, and amphotericin B) and plated. Explanted tissue fragments were retained in the culture dish to facilitate fibroblast outgrowth. Cells were maintained in a humidified incubator at 37°C with 5% CO₂ and passaged once they reached confluence. Fibroblasts at passage 2 were harvested for protein analysis.

HA was purchased from Shanghai Macklin Biochemical Technology Co., Ltd. To generate low molecular weight HA, the HA sample was subjected to sonication (60 W, 20 sec each time, a total of 3 times) while kept on ice. The molecular size of the HA after sonication was assessed by agarose gel electrophoresis as previously described (25). Briefly, samples alongside HA molecular weight standards (10-200, 400-800, 1,500-2,500 kDa) were separated on a 1% agarose gel, followed by staining with StainsAll (MilliporeSigma) at room temperature with gentle agitation overnight for visualization. To assess macrophage polarization into M1 and M2 subtypes, RAW264.7 macrophages were stimulated with HA in vitro. After stimulation for the indicated time period, RAW264.7 cells (1-5×10⁶) were stained with fluorophore-conjugated antibodies targeting canonical surface markers: Phycoerythrin (PE)-labeled anti-CD86 (M1 marker; BioLegend, Inc.; cat. no. 159204) and fluorescein isothiocyanate (FITC)-labeled anti-CD206 (M2 marker; BioLegend, Inc.; cat. no. 141704). Cells were incubated with antibodies in the dark on ice for 30 min, washed twice with cold PBS, and analyzed using a Beckman CytoFlex flow cytometer (Beckman Coulter, Inc.). FlowJo v10 software (BD Biosciences) was used to calculate the percentage of CD86⁺/CD206⁺ subpopulations and quantify phenotypic shifts.

To investigate the functional role of HAS2, the present study employed both gain-of-function and loss-of-function approaches in NIH/3T3 fibroblasts. The lentivirus for overexpression and the shRNA plasmid for knockdown were both designed, constructed and provided by Shanghai GeneChem Co., Ltd. For HAS2 overexpression, a lentiviral construct (GV492 vector) carrying the full-length mouse HAS2 coding sequence (GenBank: NM_008216.3) was used. Lentiviral particles (titer: 2.00×10⁹ TU/ml) were transduced into cells at a multiplicity of infection (MOI) of 50. After 16 h, the medium was replaced with fresh culture medium. Stably transduced cells were selected with 2 μg/ml puromycin starting 72 h post-transduction. Following initial selection, the puromycin concentration was reduced to a maintenance level of 1 μg/ml for continued culture and expansion of the transduced population. For HAS2 knockdown, a short hairpin RNA (shRNA) targeting the sequence 5'-CCTGCCAAGATGTTTGCAATT-3' was cloned into a GV493 lentiviral vector. A non-targeting scramble shRNA (5'-TTCTCCGAACGTGTCACGT-3') served as the negative control. Plasmid (2.5 μg) was transfected into NIH/3T3 cells using Lipofectamine^® 3000 (Invitrogen; Thermo Fisher Scientific, Inc.) according to the manufacturer's protocol. Cells were processed for subsequent experiments 24 h post-transfection.

High throughput virtual screening was used to screen natural products that can target HAS2 protein. Protein Preparation: The three-dimensional structure of human HAS2 (AlphaFold ID: AF-Q92819-F1) was downloaded from the AlphaFold database (26) and processed using the Protein Preparation Wizard module of Schrödinger Maestro v11.4. The protein was hydrogenated and energy-optimized. Compound Preparation: The 2D structures of compounds from a Natural Product Library (3,995 compounds, MedChemExpress) were hydrogenated and energetically optimized using the LigPrep module of Schrödinger Maestro v11.4 to generate 3D structures for virtual screening. Virtual screening was conducted with the Virtual Screening Workflow module for molecular docking. The prepared compounds were imported for molecular docking using the Glide module, which involved geometric and energetic matching of receptor and ligand. The compounds were screened against HAS2 using standard precision mode. The binding affinities and structures of the top 200 compounds from the natural product library were reviewed based on molecular docking scores for output.

The top 200 compounds, based on docking scores, were converted to SMILES format and analyzed for ADME parameters using the SwissADME website (http://www.swissadme.ch/). Optimal compounds were selected based on criteria such as lipophilicity (XLOGP3 between −0.7 and +5.0), molecular weight (150-500 g/mol), polarity (TPSA between 20 and 130 Ų), solubility (log S ≤-6), saturation (at least 0.25 fraction of sp³ hybridized carbons), and flexibility (maximum of 9 rotatable bonds).

To elucidate the molecular binding mechanism between OG and HAS2, including key interaction residues, dynamic stability and binding free energy, molecular docking and MDS were conducted using GROMACS v2023.1 (www.gromacs.org). The structures of OG (PubChem CID: 12315192) and HAS2 (PDB ID: Q92819) were retrieved from PubChem (https://pubchem.ncbi.nlm.nih.gov/) and AlphaFold (26), respectively. The systems were solvated in a water box with a 1.0 nm cutoff, using the MMFF and Charmm36 force fields for the protein and TIP3P for water. Electrostatic interactions were calculated with PME, and covalent bonds were constrained using the LINCS algorithm (27). Counter ions (Na⁺/Cl⁻) were added to neutralize the systems, and energy minimization was carried out using the steepest descent method. After equilibration under NVT conditions, a 100 ps position restraint MDS was performed before production runs in the NPT ensemble with a 2 fsec timestep. Temperature and pressure coupling were achieved using V-rescale and the Parrinello-Rahman barostat. The trajectories were analyzed using VMD v1.9.3 (28), PyMOL v3.1.0 (29) and Origin v2024 (30). Hydrogen bonds were analyzed with the VMD Hydrogen Bonds plugin (28), with stable hydrogen bonds defined as having ≥50% occupancy. The ligand-protein binding free energy was estimated using the MM/PBSA method in GROMACS (31).

To validate the binding affinity and kinetic parameters of OG (MedChemExpress; Purity: 99.57%) to immobilized HAS2, as predicted by computational simulations, SPR analysis was performed using a Biacore T200 system (Cytiva). The CM5 sensor chip (Fc2 channel) was activated with a 400 mM EDC/100 mM NHS mixture (10 μl/min, 800 sec), followed by covalent immobilization of recombinant human HAS2 (100 μg/ml in 10 mM NaAc, pH 4.0) to achieve 10,993 response units. The chip was deactivated by injecting 1 M ethanolamine hydrochloride into both sample channels at 10 μl/min for 800 sec. The running buffer was prepared by mixing PBS-P with 5% DMSO. OG was prepared in seven concentrations (1,250, 625, 312.5, 156.3, 78.2, 39.1, 19.6 and 0 μM) and introduced into the Fc1-Fc2 channels at 30 μl/min for 60 sec to allow binding, followed by a 90-sec dissociation phase.

CETSA was performed on cell lysates following the established protocol (32,33). CETSA was employed to confirm the direct binding of OG to HAS2 via ligand-induced thermal stabilization. NIH/3T3 cells were pretreated with 5 ng/ml TGF-β1 for 24 h to upregulate endogenous HAS2 expression, followed by cell collection, washing with cold PBS, and lysis (with a proteinase inhibitor) through three freeze-thaw cycles. Lysates were centrifuged at 20,000 x g for 20 min at 4°C to isolate soluble fractions, which were then adjusted to 4 μg/μl (as determined by BCA assay). The lysates were divided into two groups: 100 μM OG treatment or DMSO control (1 h at room temperature). Aliquots were heated at 3°C intervals (37-70°C) for 3 min, cooled for 3 min at room temperature, and re-centrifuged (4°C; 20,000 x g; 20 min). The supernatants were denatured with 5X loading buffer (100°C for 10 min) and analyzed by western blotting to assess protein stability.

The effect of OG on NIH/3T3 and HFL-1 cell proliferation was evaluated using the CCK-8 assay. Cells were plated at 3,000 cells per well in a 96-well plate, treated with eight drug concentration gradients (0-100 μM), and incubated for 48 h. Afterward, 10 μl of CCK-8 solution (Biosharp Life Sciences) was added to each well, followed by an incubation period of 1-4 h. Absorbance at 450 nm was recorded using a microplate reader.

Part 1: Time-course analysis of HAS2 and HA dynamics. To delineate the temporal dynamics of HAS2 expression and HA accumulation during fibrotic progression, a cohort of 35 SPF male C57BL/6J mice (6-8 weeks old) was used. Mice were randomly assigned to seven groups (n=5 per group): a saline control group (Day 0) and BLM-treated groups sacrificed at 3, 5, 7, 10, 14, and 21 days post-instillation. Pulmonary fibrosis was induced via a single intratracheal administration of BLM (2 mg/kg) under isoflurane anesthesia (3-4% for induction, 1-1.5% for maintenance); control mice received an equal volume of saline. At each designated endpoint, mice were anesthetized with intraperitoneal injection of pentobarbital sodium (50 mg/kg). After anesthesia induction, sacrifice was performed by exsanguination. Death was confirmed by the absence of spontaneous breathing, heartbeat, and corneal reflex, along with pupillary dilation. Bronchoalveolar lavage fluid (BALF) and lung tissues were then collected for subsequent analysis.

Part 2: Therapeutic efficacy assessment of OG. SPF male C57BL/6J mice (6-8 weeks old) were randomly assigned to six groups (n=8/group) using a computer-generated sequence: i) naïve control (saline vehicle), ii) BLM-induced fibrosis model (untreated), iii) OG-Low (25 mg/kg/day OG), iv) OG-Medium (50 mg/kg/day OG), v) OG-High (100 mg/kg/day OG) and vi) positive control (pirfenidone 100 mg/kg/day). The mice were housed in the SPF-grade animal facility at Shenzhen Glorybay Biotech Co., Ltd., under controlled conditions maintained at a temperature of 22±2°C, a 12-h light/dark cycle and a humidity of 50±10%. Pulmonary fibrosis was induced in groups 2-6 via intratracheal administration of BLM under isoflurane anesthesia, while control mice received an equivalent volume of saline. Therapeutic interventions were administered daily via oral gavage (0.1 ml/10 g body weight) starting 24 h post-induction, and continued for 14 days, with Groups 1 and 2 receiving volume-matched saline. After two weeks of treatment, anesthesia was induced by an intraperitoneal injection of sodium pentobarbital (50 mg/kg), followed by sacrifice via exsanguination. Serum and lung tissues were then collected for further analysis.

The left lungs of the mice were preserved in 10% formalin at room temperature for 24 h. The tissues were then dehydrated through a graded ethanol series (70, 80, 95 and 100%), cleared in xylene, and embedded in paraffin. Thin slices of ~4-5 μm were prepared for subsequent staining procedures. Lung injury and inflammation were assessed using hematoxylin and eosin (H&E) staining (performed at room temperature; hematoxylin for 5-8 min, eosin for 1-3 min) to visualize cellular morphology and inflammation. Masson's trichrome staining was performed at room temperature using a commercial kit (Solarbio, Beijing, China) according to the manufacturer's instructions to evaluate collagen deposition in lung tissue samples.

To evaluate the systemic and pulmonary exposure of OG following oral administration, OG concentrations in serum and lung tissue homogenates were measured in BLM-induced fibrotic mice treated with OG at three dose levels (25, 50 and 100 mg/kg/day). After 14 days of treatment, blood samples were collected and centrifuged at 2,000 x g for 15 min at 4°C to obtain serum. The right lung lobes were excised, weighed and homogenized in ice-cold saline using a tissue homogenizer with zirconia beads until a uniform suspension was achieved. OG concentrations in serum and lung homogenates were quantified using a validated ultra-high performance liquid chromatography-tandem mass spectrometry (UHPLC-MS/MS) method. Briefly, the samples were pretreated by protein precipitation with acetonitrile, followed by nitrogen evaporation and reconstitution in 20% methanol. Chromatographic separation was performed on a Waters ACQUITY UPLC HSS T3 column (2.1×100 mm ×1.8 μm; Waters China, Ltd.) with gradient elution using acetonitrile and 0.1% formic acid in water. Mass detection was conducted in negative electrospray ionization mode with multiple reaction monitoring (MRM) of the transitions m/z 285.1-123.2 for OG and m/z 271.2-107.8 for the internal standard (arbutin). The method demonstrated good linearity (0.0408-4.08 μg/ml), specificity, and recovery (85-120%) in both matrices.

All data are presented as mean ± standard deviation (SD). All reported replicates are independent biological replicates. Statistical analysis was performed using GraphPad Prism (Dotmatics) to identify significant differences. The normality of data distribution was assessed with the Shapiro-Wilk test. For comparisons between two groups, a two-tailed Student's t-test was applied to normally distributed data; otherwise, an appropriate nonparametric test was used. For comparisons among multiple groups, one-way analysis of variance (ANOVA) followed by Tukey's post hoc test was used when the data met normality and homogeneity of variance assumptions; otherwise, the nonparametric Kruskal-Wallis test followed by Dunn's post hoc test was applied. P<0.05 was considered to indicate a statistically significant difference.

To systematically identify conserved molecular signatures of IPF pathogenesis, comparative transcriptomic analyses were performed across human clinical samples, animal models and in vitro systems. Bioinformatic analysis of the GSE110147 dataset from the GEO database revealed significant dysregulation of 3,499 genes in patients with IPF compared with healthy controls, with 2,170 genes upregulated and 1,329 downregulated (Fig. 1A). A parallel investigation of a BLM-induced pulmonary fibrosis murine model showed similar transcriptional alterations, with 1,058 genes markedly upregulated and 737 downregulated in fibrotic lung tissues relative to healthy controls (Fig. 1B). In complementary cellular experiments, TGF-β1-stimulated myofibroblast differentiation in NIH/3T3 cells resulted in substantial transcriptomic remodeling, with 1,077 genes upregulated and 1,483 downregulated compared with untreated fibroblasts (Fig. 1C). Cross-species comparative analysis identified 17 evolutionarily conserved genes that were consistently upregulated across all three experimental systems (human IPF samples, murine fibrosis models, and TGF-β1-activated myofibroblasts). This core signature of 17 intersecting genes (Fig. 1D) represents potential conserved therapeutic targets for pulmonary fibrosis intervention.

To explore cellular heterogeneity in fibrotic pathogenesis, single-cell transcriptomic profiling was performed on pulmonary tissues from BLM-challenged mice and saline-treated controls. After quality filtering, 76,995 cells were retained and clustered into 32 distinct cell populations (Fig. 1E). The per-sample UMAP embeddings, shown in Fig. S1, demonstrated consistent cell distribution patterns across individual mice within each group. The marker genes defining each of these 32 clusters are listed in Table SII. Cell type annotation was validated through heatmap visualization of canonical marker gene expression for each major cell type (Fig. S2). These clusters were subsequently annotated into 10 major cell types (Fig. 1F), including B cells (15,558 cells), endothelial cells (7,238 cells), epithelial cells (4,383 cells), fibroblasts (2,430 cells), mast cells (368 cells), mononuclear phagocytes (24,214 cells), mural cells (645 cells), neutrophils (5,319 cells), plasmacytoid dendritic cells (4,551 cells) and T cells (12,289 cells). Comparative UMAP visualization revealed conserved clustering patterns for all 10 annotated cell types across control and fibrotic lungs (Fig. 1G). Cell type-specific differential expression analysis identified significant HAS2 upregulation exclusively in the fibroblast populations of BLM-treated mice (Fig. 1H), while expression remained stable across other lineages. This cell type-specific dysregulation corroborated prior bulk transcriptome findings and localized HAS2 hyperactivity to pathogenic fibroblast subsets.

Bulk transcriptomic and single-cell RNA sequencing analyses clearly indicated that HAS2 is highly expressed in both murine models of pulmonary fibrosis and myofibroblasts. To define the dynamic profile of HAS2 and its product HA during fibrosis development, the present study performed a longitudinal analysis in BLM-challenged mice. Lung tissues and BALF were collected across a time course encompassing the inflammatory and fibrotic phases (Fig. 2A). Western blot analysis revealed that the expression of the fibrotic marker COL1A1 was markedly increased from day 7 onward, peaking at days 14 and 21. HAS2 protein levels exhibited a parallel escalating trend, demonstrating a strong temporal correlation with collagen deposition (Fig. 2B). Notably, HA levels in BALF increased sharply, reaching a peak at day 5 and remained highly elevated throughout the subsequent observation period until day 21 (Fig. 2C), identifying HA as an early-sustained pathogenic component preceding full fibrotic consolidation. To further validate the cellular source of HAS2 upregulation and its association with fibrotic activation, the present study isolated primary lung fibroblasts from control and BLM-treated mice (Fig. 2D). Western blot analysis of these primary cells demonstrated a significant concomitant increase in the protein levels of both COL1A1 and HAS2 in fibroblasts derived from fibrotic lungs compared with those from control lungs (Fig. 2E). Similar results were observed in the TGF-β1-induced myofibroblast model (Fig. 2F). Following stimulation with TGF-β1, the expression of fibrotic markers COL1A1 and ACTA2 in NIH/3T3 fibroblasts was markedly upregulated, indicating their differentiation into myofibroblasts (Fig. 2G). Concomitantly, both HAS2 protein and mRNA levels were markedly increased (Fig. 2G and H). Furthermore, HA levels were markedly increased in the culture medium of myofibroblasts, further supporting the association of HAS2 activity with enhanced HA production (Fig. 2I).

To investigate the role of HAS2 in macrophage M2 polarization, shRNA was used to knock down the HAS2 gene in fibroblasts (Fig. 3A). TGF-β1 stimulation markedly increased HA production in fibroblasts. However, when HAS2 was knocked down, the TGF-β1-induced HA production was notably inhibited (Fig. 3B). Subsequently, TGF-β1-stimulated HAS2 knockdown fibroblasts were used to condition the medium for macrophage cultures, followed by flow cytometric analysis of the macrophages (Fig. 3C). The conditioned medium from TGF-β1-stimulated fibroblasts markedly induced M2 polarization in macrophages; however, this effect was markedly reduced when HAS2 was knocked down (Fig. 3D). These results suggested that HAS2 is critical for HA synthesis in response to TGF-β1 stimulation and that downregulating HAS2 impairs the ability of myofibroblasts to induce macrophage M2 polarization.

Given their critical role in the fibrotic response, the influence of HA on macrophage polarization was investigated. Agarose gel electrophoresis revealed that the molecular weight of sonicated HA fell within the 10-200 kDa range, confirming it as low-molecular-weight HA (Fig. S3). This sonicated HA was then applied to RAW264.7 mouse monocyte-macrophage cells, and flow cytometry was used to assess macrophage polarization. The results indicated that HA treatment promoted the polarization of RAW264.7 cells towards the M2 macrophage phenotype, as evidenced by increased CD206 expression (Fig. 4A). HA also induced TGF-β1 production in macrophages, further supporting its role in M2 polarization (Fig. 4B). Following HA stimulation, phosphorylation levels of STAT6 progressively increased, suggesting that HA-induced macrophage M2 polarization occurs through the activation of the STAT6 signaling pathway (Fig. 4C). CD44, the receptor through which HA binds to cells, was implicated in this process. Inhibition of CD44 markedly reduced HA-induced CD206 expression in macrophages (Fig. 4D), as well as suppressed TGF-β1 expression and STAT6 phosphorylation levels (Fig. 4E and F). To confirm the necessity of STAT6 activation, pharmacological inhibition of STAT6 (AS1517499) effectively blocked HA-induced STAT6 phosphorylation (Fig. 4G). To validate this signaling axis in a more physiologically relevant context, a fibroblast-macrophage co-culture system was employed (Fig. 4H). Conditioned medium from TGF-β1-stimulated NIH/3T3 fibroblasts strongly induced M2 polarization in RAW264.7 macrophages. This effect was substantially diminished when macrophages were co-treated with either a CD44 inhibitor or a STAT6 inhibitor (Fig. 4I). These results demonstrate that HA induces macrophage M2 polarization by binding to CD44 on macrophages and activating the STAT6 pathway.

A high-throughput virtual screening of natural compounds was conducted to identify small molecules that can bind to the HAS2 protein (Fig. 5A and B). After the initial screening, the top 200 compounds were selected based on docking scores for further analysis. These compounds were then evaluated using ADME parameters to identify those with favorable drug-like properties. Among the 200 compounds, only OG met all the ADME screening criteria, highlighting its potential as a HAS2-targeting compound with favorable drug-like properties (Fig. 5C). The molecular formula of OG is shown in Fig. 5D and its docking score with HAS2 is-7.325. The physicochemical properties of OG include: XLOGP3 (-0.99), molecular weight (286.28 g/mol), TPSA (119.61 Ų), log S (-1.02), Fraction Csp3 (0.54), and 3 rotatable bonds. To assess the binding affinity between HAS2 and OG, an SPR experiment was performed (Fig. 5E). The association rate constant (Ka) was 20.66 M⁻¹sec⁻¹, and the dissociation rate constant (Kd) was 0.006278 sec⁻¹. The dissociation constant was calculated to be 3.039×10⁻⁴ M, confirming a direct binding interaction between HAS2 and OG. Further validation using CETSA showed that OG enhanced HAS2 protein stability under rising temperatures compared with the control group, reinforcing the direct interaction between OG and HAS2 (Fig. 5F). The SPR and CETSA experiments confirm that OG binds directly to the HAS2 protein.

A CCK-8 assay was performed to assess the cytotoxic effects of OG on murine NIH/3T3 and human HFL-1 fibroblast cell lines. The results (Fig. 5G) indicated that OG was non-toxic to these cell lines at concentrations ranging from 0 to 100 μM. TGF-β1-stimulated NIH/3T3 and HFL-1 cells were then treated with varying concentrations of OG, and ELISA was performed to measure HA content in the cell culture supernatant. The results showed that OG effectively and concentration-dependently inhibited HA secretion, with half-maximal inhibitory concentrations (IC₅₀) of 28.19 μM and 31.11 μM for NIH/3T3 and HFL-1 cells, respectively (Fig. 5H). Additionally, lentiviral vectors were used to overexpress the HAS2 protein in NIH/3T3 cells (Fig. 5I). HAS2 overexpression markedly increased HA production, whereas OG treatment effectively inhibited this increase (Fig. 5J). These results suggested that OG targets HAS2 to suppress HA production.

MDS confirmed stable binding between OG and HAS2 (Fig. 6A), with hydrophobic interactions and hydrogen bonds formed with key residues (Fig. 6B and C). Hydrogen bond occupancy analysis identified ASP313 (137.60%) and ASP121 (110.10%) as primary stable binding sites, with an average of five hydrogen bonds maintained throughout the >40 nsec simulation (Fig. 6D). Hydrogen bond occupancy was calculated as the percentage of simulation frames in which a specific atom pair formed a hydrogen bond. Occupancies >100% indicate that the residue formed multiple, concurrent hydrogen bonds with the ligand within individual frames, reflecting a high-density interaction interface. A complete list of hydrogen bonds and a time-resolved occupancy heatmap are provided in Table SIII and Fig. S4. Conformational stability was confirmed by low root-mean-square fluctuations (RMSF) values across 100 nsec, indicating minimal protein flexibility, except for residues near positions 30, 70, 400, 470 and 500 (Fig. 6E). Root-mean-square deviation (RMSD) analysis showed stable binding conformations (0.1-0.4 nm) after the initial 10 nsec, with consistent radius of gyration and solvent-accessible surface area values, confirming no major structural changes (Fig. 6F-H). Free energy landscape analysis identified the lowest energy binding conformation at 24.12 ns (39.954 kJ/mol), further supporting the stable OG-HAS2 interaction (Fig. 6I).

To further investigate the role of OG in macrophage M2 polarization, fibroblasts were treated with OG during TGF-β1 stimulation. The conditioned medium was collected and used to culture macrophages, followed by ELISA and flow cytometric analysis (Fig. 7A). The results showed that conditioned medium from TGF-β1-stimulated fibroblasts markedly induced macrophage M2 polarization; however, this effect was markedly reduced with OG treatment. Conditioned medium from OG-treated fibroblasts showed markedly reduced capacity to induce TGF-β1 release from macrophages (Fig. 7B). Similarly, the ability of this conditioned medium to stimulate CD206 expression in macrophages was substantially impaired (Fig. 7C). These results demonstrated that OG inhibited the ability of myofibroblasts to induce macrophage M2 polarization.

A BLM-induced mouse model was used to assess the anti-fibrotic effects of OG on the lungs. Histological analysis with H&E and Masson's trichrome staining showed that OG treatment mitigated pulmonary inflammation and fibrosis progression in the mouse model (Fig. 7E). Immunoblotting of lung tissues demonstrated that OG inhibited the expression of fibrotic markers, including COL1A1 and α-SMA (Fig. 7F). Furthermore, OG markedly reduced the levels of HA and TGF-β1 in the mice (Fig. 7G and H). Quantitative analysis of OG concentrations in serum and lung tissues revealed dose-proportional increases in both compartments (Fig. 7I). This dose-dependent tissue distribution of OG paralleled its graded therapeutic efficacy, establishing a clear exposure-response relationship across treatment groups.