Network medicine modeling of the m⁶A regulatory landscape identifies a KLF6–WTAP axis as a therapeutic target in pulmonary fibrosis

Abstract

Background

Idiopathic pulmonary fibrosis (IPF) is sustained by multicellular circuits linking endothelial activation, fibroblast remodeling, and immune crosstalk. However, how N⁶-methyladenosine (m⁶A) regulation is embedded within these networks and how such network-level regulators can be prioritized as actionable nodes relevant to clinical pharmacology and safety remains unclear.

Methods

Guided by a computational modelling and network medicine framework, we integrated single-cell RNA-seq with spatial transcriptomics to systematically profile 23 canonical m⁶A regulators in pulmonary fibrosis and to map their coupling to immune, cytokine, and extracellular-matrix (ECM) programs. CellChat-based ligand–receptor inference was used to reconstruct intercellular communication, while hdWGCNA co-expression modules and pseudotime trajectories resolved intracellular program architecture and dynamic transitions. Key nodes were further interrogated experimentally. WTAP function was evaluated via shRNA-mediated silencing in primary human lung fibroblasts and fibroblast-specific conditional deletion in a bleomycin (BLM)–induced mouse fibrosis model. Immunofluorescence, MeRIP-qPCR, ChIP-qPCR, luciferase reporter assays, RT-qPCR, and western blotting were used to validate WTAP expression, upstream regulation, and downstream m⁶A-linked effects.

Results

Network modelling highlighted IGF2BP3-associated sprouting angiogenesis with strengthened adhesion/chemokine signaling in endothelial cells and identified HNRNPA2B1 as a marker of pro-inflammatory macrophage states characterized by enhanced MDK and ITGB2 axes. In fibroblasts, WTAP emerged as a central m⁶A writer connecting an ECM/contractile module to a metabolic-to-mechanical transition along pseudotime. Spatial mapping and immunofluorescence confirmed elevated WTAP in fibroblast-enriched fibrotic regions. Functionally, WTAP silencing attenuated TGF-β–induced α-SMA and collagen III expression, reduced proliferation/migration, and lowered global m⁶A levels. Mechanistically, ChIP-qPCR and promoter reporter assays supported KLF6-dependent transcriptional activation of WTAP, and WTAP was associated with methylation-linked regulation of MYC, NR4A3, and IGFBP5. In vivo, fibroblast-specific WTAP deletion improved survival, preserved lung mechanics, and diminished collagen burden in BLM-treated mice.

Conclusions

This study establishes a multi-omics network medicine map of m⁶A regulation in IPF and nominates the KLF6–WTAP axis as a central, potentially targetable hub coordinating pathogenic stromal programs. The framework provides a systems-level basis for target prioritization and for evaluating fibrosis-related safety liabilities and pharmacovigilance signals in clinical pharmacology.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12967-026-07876-x.

Introduction

Idiopathic pulmonary fibrosis (IPF) is a chronic and progressive interstitial lung disease characterized by relentless extracellular-matrix (ECM) accumulation, architectural distortion, and respiratory failure [1, 2]. Although current anti-fibrotic therapies can slow functional decline, IPF remains incurable, reflecting an incomplete understanding of the multicellular programs governing injury, aberrant repair, and scar maturation [3, 4].

Single-cell RNA sequencing has revolutionized our view of pulmonary fibrosis, uncovering epithelial plasticity, profibrotic macrophage subsets, activated endothelium, and diverse fibroblast and myofibroblast populations that collectively construct the fibrotic niche [5, 6]. Yet, how post-transcriptional regulation, particularly N⁶-methyladenosine (m⁶A) RNA modification, coordinates these cell states and their intercellular communication remains poorly understood.

m⁶A is the most abundant internal modification of eukaryotic mRNAs and selected non-coding RNAs. It is installed by a “writer” complex composed of METTL3/METTL14 and the adaptor WTAP, removed by demethylases (FTO, ALKBH5), and interpreted by reader proteins (YTH and IGF2BP families) that collectively regulate mRNA splicing, export, stability, and translation [7]. Through this machinery, m⁶A serves as a dynamic layer of epitranscriptomic control in development and disease.

Emerging evidence indicates that WTAP is not merely a structural adaptor but a key regulatory hub in fibrogenesis and tissue remodeling. WTAP-associated methylation has been shown to promote fibroblast proliferation, migration, and ECM synthesis [8, 9], while its loss suppresses TGF-β-induced activation and attenuates organ fibrosis [10–12]. In cardiac and hepatic fibrosis, WTAP knockdown reduces collagen deposition and inflammatory signaling [13], underscoring its cross-organ profibrotic role. Mechanistically, WTAP interacts with METTL3/METTL14 to facilitate translation of pro-adhesive and ECM-related transcripts via m⁶A modification [14]. Together with reader proteins such as IGF2BP3 that stabilize angiogenic and adhesion-related mRNAs [15, 16], this network couples transcriptional and post-transcriptional control of migration, vascular remodeling, and inflammation.

In IPF, aberrant endothelial activation, fibroblast reprogramming, and macrophage–stromal crosstalk act in concert to sustain fibrogenesis; recent single-cell atlases capture these axes and their peripheral immune correlates [6, 17], yet the system-level coupling between m⁶A regulators, particularly WTAP, and these multicellular circuits remains uncharted.

To address this, we integrated single-cell transcriptomics with network modeling to delineate how m⁶A regulators distribute across lung cell types and shape profibrotic communication. We profiled 23 canonical m⁶A regulators across a fibrotic-lung atlas, linked them to immune, cytokine, and ECM pathways, and combined CellChat-based ligand–receptor inference [18] with hdWGCNA module mapping [19], pseudotime reconstruction and spatial transcriptomics to capture dynamic fibroblast and endothelial state transitions.

Complementary in vitro and in vivo experiments further validated that WTAP silencing attenuates TGF-β–induced fibroblast activation and that fibroblast-specific WTAP deletion alleviates bleomycin-induced pulmonary fibrosis. Mechanistically, we identify KLF6 as an upstream transcriptional activator of WTAP, whereas WTAP, as part of the writer complex, promotes m⁶A methylation of downstream profibrotic transcripts, thereby linking epitranscriptomic regulation to matrix remodeling.

Together, these findings integrate single-cell epitranscriptomics, communication-network analysis, and functional validation to reveal WTAP-driven m⁶A rewiring of endothelial, mesenchymal, and immune axes in pulmonary fibrosis, providing a conceptual and experimental framework for targeting WTAP-centered m⁶A–communication–ECM circuits as potential therapeutic strategies for IPF.

Methods

Data acquisition and quality control

Single-cell RNA-seq data were downloaded from GEO (GSE128033, fibrotic lung). Spatial transcriptome sequencing data were downloaded from GEO (GSE248082 and GSE285246). Raw matrices were processed in R 4.2.2 with Seurat v4.3. Cells were retained if nFeature_RNA > 500, nCount_RNA > 1,000, and percent.mt < 10%; genes detected in < 3 cells were removed. Ambient RNA and doublets were mitigated using SoupX (default contamination estimate) and DoubletFinder (expected doublet rate 2–5% per lane; pK tuned by BCmvn). Outliers in UMI/gene counts (> 3 IQR from median) and percent. ribo > 50% were excluded. Counts were normalized as log1p(CPM × 10⁴); where indicated, SCTransform was applied. A fixed random seed (set.seed = 11337) was used throughout. After filtering, a total of 26,326 high-quality cells were retained for downstream analyses. The RNA-seq data (TPM) and corresponding clinical data of GSE70866, GSE93606, GSE110147, GSE150910, GSE47460 and GSE53845 datasets were obtained from the GEO database.

Integration, clustering, and visualization

For multiple donors/runs, samples were SCTransformed and integrated using anchor-based RPCA (FindIntegrationAnchors, dims = 1:30). Highly variable genes (~ 2,000–3,000 HVGs) were used for PCA, graph construction (FindNeighbors, dims = 1:30, k.param = 20), and Leiden/Louvain clustering (FindClusters, resolution 0.6–1.2, chosen by clustree stability and silhouette). Global views used t-SNE embeddings (RunTSNE, dims = 1:30; perplexity = 30). Cell-cycle scores (S and G2/M) were computed and regressed when noted.

Cell-type annotation

Clusters were annotated by canonical markers and FindAllMarkers DEGs: macrophages (LYZ, CSF1R, MARCO), fibroblasts (COL1A1, DCN, PDGFRA), endothelial cells (PECAM1, KDR), T cells (CD3D), smooth muscle cells (ACTA2, TAGLN), B cells (MS4A1), mast cells (TPSAB1), and epithelial subtypes (FOXJ1, KRT5, SCGB1A1). Calls were cross-checked against lung atlases/PanglaoDB.

m⁶A regulator set

A curated panel of 23 m⁶A regulators was analyzed: writers/adaptors (METTL3, METTL14, WTAP, RBM15, RBM15B, VIRMA/KIAA1429, ZC3H13), erasers (FTO, ALKBH5), readers/auxiliaries (YTHDF1/2/3, YTHDC1/2, IGF2BP1/2/3, HNRNPC, HNRNPA2B1, FMR1, LRPPRC, ELAVL1). HGNC symbols were harmonized.

Spatial transcriptomics analysis

Individual ST samples were pre-processed with Seurat. To assess the spatial distribution of clusters identified in the scRNA-seq discovery cohort, ST and scRNA-seq expression matrices were integrated using CellTrek, incorporating spatial coordinates.

Pathway scoring and regulator–pathway correlations

Per-cell pathway activities (GO terms and curated angiogenesis/adhesion/ECM sets) were computed using GSVA/ssGSEA on log-normalized counts. Scores were z-scaled within each lineage. Spearman correlations between regulator expression and pathway scores were computed within cell types; Benjamini–Hochberg FDR was controlled at 5%. Heatmaps/bubble plots display significant associations.

Regulator-high vs. regulator-low contrasts

Within ECs, fibroblasts, or macrophages, cells were split at the median of the focal regulator (IGF2BP3, WTAP, HNRNPA2B1). DEGs used Wilcoxon with |log₂FC| ≥ 0.25, min.pct ≥ 0.10, FDR < 0.05. Enrichment used fgsea / clusterProfiler (reporting NES and FDR).

Co-expression network analysis

Lineage-specific networks were built with hdWGCNA on metacells (k-nearest pooling, size ~ 20–50). Signed adjacency with soft threshold (β = 8) targeted scale-free topology. Modules were defined with minModuleSize = 50–100, mergeCutHeight = 0.25. Module eigengenes were correlated (Spearman; FDR 5%) with regulators and pathway scores. Hub genes were identified by kME ≥ 0.6–0.7. GO/KEGG enrichment was performed at FDR 5%.

Transcription factor regulon analysis

Transcriptional regulators were inferred with the SCENIC workflow (SCENIC v1.2.4; pySCENIC v0.11.2). Briefly, GENIE3 (v1.22.0) was used to infer TF–target links from log-normalized counts; RcisTarget (v1.20.0) identified enriched TF-binding motifs and candidate regulons; AUCell (v1.22.0) quantified per-cell regulon activity. Analyses were run per lineage to limit spurious cross-lineage co-variation. Within endothelial cells (ECs), cells were split at the median IGF2BP3 expression (IGF2BP3-high vs. -low; same split rule as in “Regulator-high vs regulator-low contrasts”). Group differences in regulon activity were assessed by two-tailed Wilcoxon rank-sum tests with Benjamini–Hochberg correction (FDR 5%). To connect TF regulation with network structure, regulon target sets were overlapped with lineage-specific hdWGCNA modules (e.g., the EC black module); enrichment was evaluated by Fisher’s exact test (BH-adjusted P < 0.05). Enriched TFs/regulons were annotated for angiogenesis, migration, and ECM-related functions using GO terms and literature curation. Visualization included regulon-by-cell activity heatmaps (pheatmap v1.0.12) and TF–target subnetworks (Cytoscape v3.10.1 / ggplot2 v3.4.4).

Pseudotime inference (fibroblasts)

Trajectories were reconstructed with Monocle 2 using HVGs plus fibroblast DEGs as ordering genes. DDRTree reduction and orderCells were applied; the root was set in the WTAP-high / metabolic-homeostasis subcluster based on pathway scores. Gene and pathway trends along pseudotime were smoothed (loess); sensitivity checks with Monocle 3 yielded concordant branching.

Cell–cell communication

Intercellular signaling was inferred with CellChat (human database v1.1+). For IGF2BP3-high vs. -low ECs and HNRNPA2B1-high vs. -low macrophages, we computed information flow per pathway and outgoing/incoming strengths per cell type. Group differences used permutation tests (n = 1,000) with BH-FDR 1–5%. Visualizations included circle networks, pathway heatmaps, chord diagrams, and ligand–receptor dot plots; only LR pairs with evidence score > 0.1 and FDR < 0.05 were reported.

Specimen collection

Human lung tissue specimens were collected between June 2017 and September 2019 at Yangpu Hospital of Tongji University. Six fibrotic lung tissues were obtained from patients with a confirmed diagnosis of idiopathic pulmonary fibrosis (IPF), whereas four non-fibrotic control lung tissues were obtained from lung donors. All human tissue samples were acquired after written informed consent was obtained from the participants or their legally authorized representatives.

Inclusion criteria for the IPF cohort were as follows: male patients older than 50 years with preoperative high-resolution computed tomography (HRCT) demonstrating a usual interstitial pneumonia (UIP) pattern, characterized by subpleural and basal-predominant reticulation with honeycombing, with or without traction bronchiectasis, and minimal ground-glass opacities. Postoperative histopathological confirmation of UIP was required, including patchy interstitial fibrosis with identifiable fibroblastic foci. Patients with alternative causes of interstitial lung disease were excluded, including familial ILD, occupational/environmental exposure–related ILD, connective tissue disease–associated ILD, and drug-induced lung toxicity.

The study protocol involving human specimens was reviewed and approved by the Ethics Committee of Yangpu Hospital of Tongji University (Approval No. LL-2025-SCI-011), and all procedures were conducted in accordance with the Declaration of Helsinki.

Cell culture and transfection

Primary human lung fibroblasts (HPF) were obtained from the Cell Bank of the Chinese Academy of Sciences. Cells were cultured in FM medium (Cat. No. 2301) supplemented with 10% fetal bovine serum (Yamei, Cat.No.CY201P) and maintained at 37 °C in a humidified atmosphere containing 5% CO₂ to ensure exponential growth. For stable transfection, HPF cells were transduced with lentiviral constructs targeting WTAP, which were designed and synthesized by GIMA Corporation (China). Briefly, HPF cells were detached, resuspended in complete medium, and plated at a density of 1 × 10⁴ cells per well in 6-well plates containing 2 ml of culture medium. After cell attachment, viral particles were added to each well according to the manufacturer’s protocol and incubated for 24 h. Subsequently, the medium was replaced with fresh complete medium, and cells were cultured for an additional 48 h. Stably transfected cells were selected using puromycin until resistant colonies were obtained. After establishing stable knockdown cell lines, fibroblasts were stimulated with recombinant human TGF-β1 (Proteintech, Cat No. HZ-1011) at a final concentration of 10 ng/ml. All in vitro experiments were performed in compliance with RIVER experimental guidelines.

Protein extraction and western blot

Total cellular proteins were isolated from HPF cells using RIPA lysis buffer (Solarbio, China) and quantified by the BCA Protein Assay Kit (Beyotime, China). Equal amounts of protein (30 µg per lane) were resolved on 10% SDS–PAGE gels and subsequently transferred to PVDF membranes. The membranes were blocked with 5% bovine serum albumin (BSA) and incubated overnight at 4 °C with primary antibodies against WTAP (Proteintech, Cat No. 60188-1-Ig), Collagen III(Proteintech, Cat No. 22734-1-AP), KLF6 (Proteintech, Cat No. 14716-1-AP), NR4A3 (Proteintech, Cat No. 55405-1-AP), MYC(Proteintech, Cat No. 10828-1-AP), GAPDH (Proteintech, Cat No. 60004-1-Ig), α-SMA (Proteintech, Cat No. 14395-1-AP). After washing, the membranes were incubated for 1 h at room temperature with HRP-conjugated secondary antibodies (1:1000). Protein bands were visualized by ECL chemiluminescence (Thermo Fisher Scientific), and GAPDH served as a loading control. Band intensities were quantified using ImageJ software.

Cell proliferation and migration assays

For the EdU incorporation assay, cells grown in 24-well plates were processed using the Click-iT EdU Imaging Kit (Beyotime, China) according to the manufacturer’s protocol to evaluate DNA synthesis activity.

Besides, HPF cells were seeded in 6-well plates and scratched using a 200 µL pipette tip for wound healing assay. After washing with PBS, cells were incubated in DMEM with 1% FBS. Wound closure was imaged under an inverted microscope at indicated intervals. For transwell assay, boyden chambers with 8-µm pore filter inserts in 24-well plates (BD Biosciences, USA) were used. After the indicated treatments, cells were harvested, and 2 × 10⁴ cells in serum-free medium were seeded into the upper chamber; the lower chamber contained medium with 20% FBS as a chemoattractant. After incubation, cells that migrated to the underside of the membrane were fixed, stained, and counted at ×200 magnification in 10 random fields per insert.

Dot blot assays

Total RNA was isolated from HPFs using TRIzol reagent (Beyotime, Cat. No. R0016). The extracted RNA was serially diluted to final concentrations of 400 ng/µL, 200 ng/µL, and 100 ng/µL. Samples were denatured at 95 °C for 5 min, immediately cooled on ice, and then spotted onto a nylon transfer membrane (Yamei, Cat. No. WJ002S) using a vacuum filtration system.

Following UV crosslinking, the membranes were blocked and subsequently incubated overnight at 4 °C with an anti-m⁶A antibody (Proteintech, Cat No. 68055-1-Ig). After washing, membranes were treated with an HRP-conjugated secondary antibody for 1 h at room temperature, and signal detection was performed using an ECL chemiluminescent kit on a FUSION FX SPECTRA imaging platform. A parallel membrane was stained with methylene blue (MB) to verify equal RNA loading.

Immunofluorescence analysis

For immunofluorescence staining, fibroblasts were seeded onto circular glass coverslips and cultured until reaching approximately 40–50% confluence. Cells were fixed with 4% paraformaldehyde for 20 min, permeabilized with 0.1% Triton X-100, and blocked with 5% bovine serum albumin (BSA) for 1 h at room temperature. Human and mouse lung tissues were fixed in 4% paraformaldehyde at 4 °C overnight, incubated in 30% sucrose, and embedded in OCT compound for cryosectioning. Frozen tissues were sectioned at a thickness of 10 μm. For antigen retrieval, sections were incubated in sodium citrate buffer at 95 °C for 15 min. After blocking, cells or tissue sections were incubated with primary antibodies at 4 °C overnight, followed by fluorophore-conjugated secondary antibodies for 2 h at room temperature. Nuclei were counterstained with DAPI (Solarbio), and fluorescence images were acquired using an epifluorescence microscope. Primary antibodies were as follows: WTAP, Ki-67, and α-SMA (all from Proteintech; 1:500), followed by incubation with fluorophore-conjugated secondary antibodies (SouthernBiotech; 1:200).

RNA isolation and RT-qPCR

Total RNA was isolated using the RNA Isolater Total RNA Extraction Reagent (R401-01, Vazyme) following the manufacturer’s protocol. Complementary DNA (cDNA) was synthesized from 1 µg of total RNA using the PrimeScript RT Master Mix (Perfect Real Time) (RR036A, Takara). Quantitative PCR was carried out with TB Green Premix Ex Taq II (Tli RNaseH Plus) (RR820A, Takara) on a LightCycler 96 instrument. Relative transcript abundance was determined by the 2^–ΔΔCt method after normalization to 18 S rRNA (or GAPDH where indicated). The following primers were synthesized by BioAsia (Hyderabad, India) in supplementary table.

Methylated RNA immunoprecipitation quantitative polymerase chain reaction (MeRIP-qPCR)

Frozen samples were homogenized in a non-denaturing lysis buffer containing 0.5% NP-40. RNA fragmentation was omitted due to primer design considerations. Specifically, the SRAMP m⁶A modification site prediction tool was first employed to identify potential m⁶A-enriched regions within target transcripts, and primers were subsequently designed to amplify these predicted sites. Total RNA was then incubated with Dynabeads™ magnetic beads (Cat. No. 10001D, Thermo Fisher Scientific) conjugated with either an anti-m⁶A antibody (Cat. No. 68055-1-Ig, Proteintech) or an IgG antibody as a negative control (Cat. No. 98136-1-RR, Proteintech). After immunoprecipitation, DNase I (Cat. No. 13653-1-AP, Proteintech) was applied at room temperature to eliminate residual DNA. The bead-bound complexes were subsequently treated with Proteinase K (Cat. No. HY-K0010, MCE) to digest proteins and release the immunoprecipitated RNA. Finally, 1 ng of input RNA and m⁶A-enriched RNA were subjected to reverse transcription followed by quantitative PCR (RT-qPCR) using the primers described above. The corresponding sequences of gene primers are provided in Supplementary table.

Chromatin immunoprecipitation (ChIP) assay

Fibroblasts were fixed with 1% formaldehyde for 10 min at room temperature, followed by quenching with 125 mM glycine for 5 min. Cells were then washed twice with PBS, harvested in lysis buffer, and subjected to sonication on ice using an ultrasonic homogenizer (35% amplitude, 5 min) to generate chromatin fragments with an average size of 200–800 bp. An aliquot of each sonicated lysate was used to verify DNA concentration and fragment size.

For immunoprecipitation, chromatin extracts were incubated with ChIP-grade anti-KLF6 antibody together with Protein A/G agarose beads (Thermo Fisher Scientific, Cat# 20421) at 4 °C overnight with gentle rotation. After extensive washing, immune complexes were treated with proteinase K at 60 °C for 2 h and RNase at 37 °C for 1 h to remove proteins and RNA, respectively. DNA was subsequently purified using a PCR purification kit (Thermo Fisher Scientific, Cat# K0702). Enrichment of target genomic regions was quantified by qPCR using primers listed in Supplementary Table, and ChIP signals were normalized to input DNA.

Luciferase reporter assay

HPF cells were transfected with 100 ng WTAP promoter-driven luciferase reporter plasmid (CMV-C-Flag) alone, or co-transfected WTAP promoter-driven luciferase reporter plasmid and 200 ng KLF6 plasmid (CMV-KLF6-C-Flag). And two groups both transfected 20 ng of Renilla luciferase plasmid as an internal control. Forty-eight hours after transfection, luciferase activity was measured using a dual-luciferase reporter assay system (Yeasen Biotechnology, China). Firefly luciferase activity was normalized to Renilla luciferase activity to account for transfection efficiency.

Animal experiments

Fibroblast-specific Wtap conditional knockout (cKO) mice were generated by crossing Col1a2-2 A-Cre mice with Wtap^fl/fl mice. C57BL/6 wild-type mice and fibroblast-specific Cre mice (Col1a2-2 A-Cre, Cat. No. NM-KI-215043) were purchased from Shanghai Model Organisms Center, Inc., while Wtap-floxed mice (Strain No. S-CKO-12549) were obtained from Cyagen. All mice were housed under specific pathogen-free conditions with a controlled environment (approximately 21 °C, 40–60% humidity), a 12-hour light/dark cycle, and ad libitum access to standard chow and water. Male mice aged 8–10 weeks were used for all experiments.

To establish the pulmonary fibrosis model, mice were intratracheally administered bleomycin (BLM) at a dose of 2.0 mg/kg, and lung tissues were harvested 21 days after treatment. All animal experimental procedures were performed in strict accordance with the guidelines of the Animal Center of Tongji University and were approved by the Institutional Animal Care and Use Committee (Approval No. LL-2025-SCI-010).

Primary lung fibroblast isolation and culture

Lung tissues were obtained from 8-week-old male Wtap cKO mice. To initiate dissociation, 1 mL of an enzymatic cocktail, comprising neutral protease (5 U/mL), DNase I (0.33 U/mL), collagenase type I (200 U/mL), and elastase (4 U/mL), was administered via intratracheal instillation. After a 20-minute digestion at room temperature, the lung lobes were mechanically disrupted in DMEM containing 15% fetal bovine serum (FBS) to generate a single-cell suspension. The mixture was passed through sequential 100 μm and 40 μm cell strainers to eliminate debris. Following centrifugation, the cell pellet underwent erythrocyte lysis and was subsequently resuspended in complete growth medium (DMEM with 15% FBS). Cells were cultured in 10 cm dishes with an initial medium change at 48 h. Homogeneous fibroblast populations were achieved after 2–3 serial passages. For downstream applications, fibroblasts were seeded onto glass coverslips in 24-well plates at a density of 5 × 10⁵ cells per well and maintained in DMEM supplemented with 15% FBS and 1% penicillin-streptomycin. Cell health and morphology were consistently monitored using phase-contrast microscopy.

Histological staining and hydroxyproline content assay

H&E staining followed the standard protocol. After being washed by water, lung slices were sequentially stained with hematoxylin (Abcam, ab150678) for 2 min and eosin (Sigma, HT110280) for 3 min. Following the dehydration and clearing steps, slices were sealed with neutral balsam. Stained images were captured by Olympus SLIDEVIEW VS200 digital slide scanner. The hydroxyproline content was measured using a kit purchased from Nanjing Jiancheng Bioengineering Institute (Nanjing, China) according to the manufacturer’s protocol.

Pulmonary function analysis and arterial blood gas analysis

At the end of the experiment, the mice were anesthetized by intraperitoneal injection of 50 mg/kg pentobarbital sodium (50 mg/kg, Sinopharm, Beijing, China), and then were tracheostomized and ventilated. Subsequently, the animals were positioned in the EMMS (Electro-Medical Measurement Systems, Hants, UK, http://www.electromedsys.com/pulmonary.html) Resistance and Compliance (R&C) system to measure pulmonary resistance and static compliance according to the manufacturer’s instructions.

Survival analyses

The survival curves were plotted using the Kaplan–Meier method and analyzed with the Log-rank Mantel-Cox test. Spearman’s correlation coefficient analysis was used to determine the correlations between pulmonary functional parameters. P values < 0.05 were considered statistically significant. N.S., not significant, ∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001, ∗∗∗∗P < 0.0001.

Statistics and reproducibility

Unless stated, tests were two-sided with FDR = 5%. Correlations report Spearman’s ρ and FDR; GSEA reports NES and FDR; CellChat reports Δ information flow and permutation p/FDR. Analyses ran in R 4.2.2; figures were generated with Seurat/ggplot2/ComplexHeatmap. Data are available at GEO: GSE128033. P < 0.05 was considered statistically significant and the levels of significance are denoted as follows: *P < 0.05; **P < 0.01; ***P < 0.001.

Results

Cell-type-specific expression patterns of m⁶A regulators and their potential roles in pulmonary fibrosis

To elucidate the functional landscape of m⁶A regulators in pulmonary fibrosis, we analyzed single-cell RNA sequencing (scRNA-seq) data derived from fibrotic lung tissues. After stringent quality control, 26,326 cells were retained for downstream analysis. Using t-distributed stochastic neighbor embedding (t-SNE) and Uniform manifold approximation and projection (UMAP), we visualized cellular heterogeneity and identified major clusters, including macrophages, fibroblasts, endothelial cells, T and B cells, smooth muscle cells, mast cells, and epithelial subtypes such as ciliated, AT1 and AT2 cells, basal airway, and club cells (Fig. 1A, Figure S1A).

Fig. 1.

Single-cell landscape of m⁶A regulators in pulmonary fibrosis and their functional associations. (A) t-SNE visualization showing major immune and structural cell populations, including epithelial subtypes (ciliated, basal airway, AT1, AT2 and club cells). (B) Widespread expression of m⁶A-related genes across multiple cell types. (C) Average expression heatmap of 23 m⁶A regulators highlighting cell-type heterogeneity. (D) Comparison of clustering results based on all genes (left) versus m⁶A-regulator features (right). (E) Single-gene expression maps for representative regulators (FTO, METTL3, RBM15B, WTAP). (F) Intersection analysis between m⁶A regulators and canonical cell-type marker genes. (G) Correlation bubble plot linking m⁶A regulators to immune, cytokine, and extracellular matrix (ECM) pathways

Across these populations, m⁶A regulators showed widespread yet heterogeneous expression (Fig. 1B), suggesting that m⁶A modification functions as a pervasive regulatory layer in the fibrotic lung microenvironment. Comparative profiling of 23 m⁶A regulators highlighted clear cell-type preferences (Fig. 1C).

To test whether this m⁶A landscape encodes lineage identity or instead reflects shared transcriptional states, we contrasted clustering based on the full transcriptome with clustering using only m⁶A regulators. Transcriptome-wide features sharply separated canonical cell types, whereas m⁶A-only clustering yielded more convergent groupings (Fig. 1D), indicating that m⁶A patterns track state-level regulation rather than lineage boundaries.

Consistent with this interpretation, t-SNE feature plots of representative regulators (FTO, METTL3, RBM15B, and WTAP) revealed discrete expression “hotspots” rather than cell-type-restricted domains (Fig. 1E). Supplementary maps across all regulators confirmed this broad yet uneven distribution at single-cell resolution (Figure S1C). Intersection analysis between m⁶A regulators and canonical marker genes showed minimal overlap (Fig. 1F). A bubble plot documented cell-type-restricted patterns of canonical markers, average expression and percent expressed, serving as contextual evidence for marker-based identities (Figure S1B).

Given the central roles of immune activation, cytokine signaling, and extracellular matrix (ECM) remodeling in fibrosis, we next examined correlations between m⁶A regulators and disease-relevant pathways. Correlation maps uncovered significant, cell-type-specific associations of several regulators with immune, cytokine, and ECM processes (Fig. 1G). Supplementary heatmaps validated these relationships (Figure S1D). Notably, ALKBH5, YTHDF2, and IGF2BP family members displayed strong positive correlations with inflammatory and ECM-related genes in macrophages and fibroblasts (Figure S1E), suggesting their potential involvement in shaping the inflammatory milieu and promoting matrix remodeling.

Overall, m⁶A regulators were broadly expressed but showed cell-type–specific associations with immune, cytokine, and ECM programs.

IGF2BP3 associates with endothelial proliferation and migration involved in sprouting angiogenesis

In endothelial cells, IGF2BP3 showed the strongest positive association with sprouting angiogenesis, proliferation, and migration scores (Fig. 2A-B). Consistently, the majority of angiogenesis-related genes were positively associated with IGF2BP3 expression, including the representative upregulation of FN1, and IGF2BP3-high endothelial cells showed enrichment of immune- and adhesion-related GO terms (Figure S2A–C).

Fig. 2.

IGF2BP3 links endothelial activation to sprouting angiogenesis. (A) Correlations between m⁶A regulators and angiogenesis-related pathways in endothelial cells. IGF2BP3 shows the strongest positive association with sprouting angiogenesis and endothelial migration. (B) Correlation between IGF2BP3 expression and pathway enrichment scores. (C) Violin plots showing upregulation of LGALS1 and BGN in IGF2BP3-high ECs. (D) Volcano plot showing differential gene expression between IGF2BP3-high and IGF2BP3-low ECs. (E) GO and GSEA enrichment: upregulated genes enriched in membrane/antigen-presentation processes; downregulated genes in detoxification and structural components. (F) Spatial HE staining and spatial feature plots of the cell type, IGF2BP3, and pathway enrichment scores in IPF tissue sections. (G) hdWGCNA module detection in endothelial cells; the black module is highlighted. (H) Module–pathway correlations showing that the black module aligns with sprouting angiogenesis and endothelial migration/proliferation

Stratification of endothelial cells (ECs) by IGF2BP3 expression revealed markedly elevated levels of adhesion- and extracellular matrix (ECM)-remodeling–associated genes, such as LGALS1 and BGN, in the IGF2BP3-high group (Fig. 2C). Differential expression analysis further identified the upregulation of angiogenic and ECM-related genes (FN1, DKK2, TAGLN, TMEM100, TIMP1, RGS5, CTGF, BGN, and LGALS1) in IGF2BP3-high ECs, whereas the chemokines CX3CL1 and CXCL10 were preferentially expressed in the low group (Fig. 2D).

Gene Ontology (GO) and Gene Set Enrichment Analysis (GSEA) revealed that IGF2BP3-high ECs upregulated genes involved in membrane trafficking, secretion, and antigen presentation, while genes related to detoxification and structural maintenance were downregulated (Fig. 2E; Figure S2D). Additionally, spatial deconvolution of four IPF spatial transcriptomic datasets with CellTrek revealed predominant IGF2BP3 expression in endothelial cells; subsequent GSVA showed significant enrichment of sprouting angiogenesis and endothelial-cell migration in this compartment (Fig. 2F). Weighted gene co-expression network analysis (WGCNA) further identified multiple EC co-expression modules (Fig. 2G; Figure S2E). Among them, the black module exhibited the strongest positive correlations with angiogenesis, endothelial migration, and proliferation (Fig. 2H). Functional enrichment of this module highlighted pathways related to angiogenesis regulation, endothelial development, and the response to fluid shear stress (Figure S2F).

These data are consistent with an IGF2BP3-high endothelial state coupled to integrin–ECM programs and sprouting angiogenesis.

IGF2BP3 enhances endothelial intercellular communication through adhesion- and angiogenesis-related signaling

To explore IGF2BP3’s impact on ECs cellular communication, we compared networks between IGF2BP3-high and IGF2BP3-low groups. Pathway-level analysis of relative information flow showed predominant elevations in the IGF2BP3-high group (Fig. 3C).

Fig. 3.

IGF2BP3 enhances endothelial intercellular communication and activates adhesion/angiogenic signaling. (A) Bar plots comparing the number of inferred interactions and total interaction strength between IGF2BP3-high and IGF2BP3-low EC networks. (B) Heatmaps showing the differences in the number or intensity of interactions between all cell types in IGF2BP3-high group and IGF2BP3-low group. Orange represents an increase in IGF2BP3-low group compared to IGF2BP3-high group, while green represents a decrease. The colored bar graph at the top represents the total value of each column displayed in the heatmap (incoming signal). The colored bar graph on the right represents the total value of each row (outgoing signal). (C) Comparison of several signaling pathways in IGF2BP3-high and IGF2BP3-low groups. (D) Circular network views of representative pathways (COLLAGEN, LAMININ, CXCL, VEGF) in IGF2BP3-high vs. IGF2BP3-low groups. (E) Dot-plot of ligand–receptor pairs between ECs and other cell types. (F) Dot-plot highlighting significant ligand–receptor pairs (p < 0.01), enriched for COL4A1/2–CD44 and LAMININ–integrin axes (e.g., LAMA5–ITGA1/ITGB1; LAMB2–ITGA6/ITGB1/ITGB4). (G) Expression of selected ligand–receptor genes in IGF2BP3-high vs. IGF2BP3-low endothelial cells (ITGB1, ITGA2, LAMB2, VEGFB). Distributions are shown as half-violin/box plots (line = median). P values from two-sided Wilcoxon rank-sum tests were adjusted by Benjamini–Hochberg FDR (*** q < 0.001, ** q < 0.01, * q < 0.05)

Network visualizations further illustrated denser COLLAGEN, LAMININ, CXCL, and VEGF signaling circuits centered on ECs in the high group (Fig. 3D), while dot plots highlighted widespread EC–stromal connectivity (Fig. 3E). ECs communicated extensively with immune cells and stromal cells through the LAMININ (including LAMC1-CD44, LAMB2-(ITGA2 + ITGB1), LAMB2-(ITGA7 + ITGB1), LAMB2-(ITGA6 + ITGB1)), CXCL (CXCL3-ACKR1), and VEGF (VEGFB-VEGFR1) pathways (Fig. 3E and F; Figure S3A–B).

To directly compare expression between groups, we plotted the abundances of representative genes. LAMB2 (a laminin subunit) and its receptors ITGB1/ITGA2, as well as VEGFB, showed higher expression in IGF2BP3-high ECs than in IGF2BP3-low ECs (Fig. 3G). Together, these findings demonstrate that IGF2BP3 amplifies endothelial intercellular communication via matrix–integrin and pro-angiogenic signaling, thereby facilitating vascular remodeling and immune engagement in fibrotic lungs.

WTAP is associated with profibrotic programs and module-level remodeling in fibroblasts

To assess the role of m⁶A regulators in fibroblast activation, we correlated their expression with fibrosis-relevant biological processes. WTAP showed the strongest positive associations with fibroblast activation, migration, TGF-β response, and growth factor signaling (Fig. 4A). Consistently, WTAP expression correlated positively with the response to fibroblast growth factor (r = 0.35, P = 8.2 × 10⁻¹⁹) and the fibroblast proliferation score(r = 0.48, P = 4.4 × 10⁻³⁷) at the single-cell level (Fig. 4B). Expanded correlation profiles confirmed positive links between WTAP and fibrogenic genes such as ZFP36L1, TSC22D3, EGR1, THBS1, and CASP7 (Figure S4A).

Fig. 4.

WTAP associates with profibrotic activation and module remodeling in fibroblasts. (A) Correlation bubble plot linking m⁶A regulators to fibroblast functional programs. (B) Scatter plots showing positive correlations between WTAP expression and response to fibroblast growth factor and fibroblast proliferation. (C) Violin plots comparing WTAP-high and WTAP-low fibroblasts for representative genes. (D) Volcano plot of differentially expressed genes between groups. (E) GO molecular function enrichment for WTAP-high cells. (F) GSEA showing enrichment of fibroblast proliferation, cell adhesion, and wound response. (G) Spatial feature plots of the cell type, WTAP, and pathway enrichment scores in IPF tissue sections. (H) hdWGCNA dendrogram identifying co-expression modules in fibroblasts. (I) Module eigengene expression across fibroblast cells. (J) Network of the turquoise module and top kME hub genes. (K) Functional enrichment of top turquoise-module genes. (L) Heatmap showing correlations between modules and fibroblast pathways. (M) Regulatory network linking transcription factors (TFs) to m⁶A writers and readers. (N) Global cell–cell communication network comparing WTAP-high versus WTAP-low states; edge thickness reflects the number of ligand–receptor interactions per cell type pair. (O) Differential heatmaps summarizing changes in interaction number and strength by sender–receiver pairs, with marginal bars showing totals across lineages

Stratification of fibroblasts revealed that WTAP-high cells exhibited elevated MYC, NR4A3, KLF6, and IGFBP5 (Fig. 4C; Figure S4B). Within fibroblasts, NR4A3, MYC, and KLF6 are closely linked to profibrotic activation. MYC directly promotes pulmonary fibroblast proliferation and myofibroblast differentiation in IPF models, and its inhibition mitigates fibrosis [20]. KLF6 participates in profibrotic signaling: in lung fibrosis contexts it is engaged by TGF-β1 pathways, and in human fibroblasts KLF6 cooperates with Sp1 to enhance collagen biosynthesis [21, 22]. Although NR4A3 is less well characterized in lung fibrosis, it is a TGF-β–inducible immediate-early TF that can promote fibroblast activation in other models, which is consistent with our WTAP-high signature [23]. Differential expression analysis further identified upregulation of inflammatory and ECM-related genes (CXCL2, IL6, PTX3, COL11A1) in the WTAP-high group, whereas WTAP-low cells preferentially expressed homeostatic markers (Fig. 4D). A collagen-focused analysis further pinpointed ligand–receptor pairs underlying these shifts: dot-plots highlighted significant COL1A1/1A2– (ITGA1/ITGB1, CD44, SDC4) interactions across lineages (Figure S4C–D), with denser networks under WTAP-high and marked attenuation of these pairs in WTAP-low (Figure S4E–F).

Functional enrichment indicated that WTAP-high fibroblasts were engaged in ribosome biogenesis, ubiquitin protein ligase binding, and cadherin and ribonucleoprotein complex binding (Fig. 4E). GSEA revealed enrichment in fibroblast proliferation, cell–cell adhesion, and wound-healing processes (Fig. 4F). Spatial deconvolution proved the same results (Fig. 4G).

High-dimensional WGCNA identified a turquoise module comprising ECM and contractile genes (COL1A1, COL1A2, LGALS1, SPARC, TIMP1, BGN; Fig. 4H–J), enriched for EMT, collagen fibril organization, and Wnt signaling (Fig. 4K). We focused on the turquoise module because, across all three fibroblast subclusters, its eigengene showed the strongest positive association with profibrotic traits (Fig. 4L). Integrative TF–m⁶A mapping highlighted KLF6 as a key upstream TF that preferentially targets the m⁶A writer arm, WTAP, thereby linking transcriptional control to the writer complex (Fig. 4M). Subsequently, we further confirmed the positive correlation between KLF6 and WTAP in patients with IPF across multiple bulk transcriptomic datasets as well as within fibroblast subpopulations (Figure S5A-C). The ST revealed significant colocalization between WTAP and KLF6 (Figure S5D) as well. In addition, cell–cell communication analysis showed that the WTAP-high state was associated with increased intercellular connectivity, with both the total number of ligand–receptor interactions and the overall interaction strength elevated across multiple lineages, most notably between fibroblasts, macrophages/smooth-muscle cells and basal airway cells (Fig. 4N–O).

Together, multi-method analyses consistently linked WTAP-high fibroblasts to ECM/contractile modules and increased intercellular connectivity.

WTAP marks a fibroblast pseudotime shift from metabolic homeostasis to adhesion/migration and ECM remodeling

Reclustering resolved four fibroblast subclusters (F1–F4) (Fig. 5A).

Fig. 5.

WTAP marks a fibroblast pseudotime shift from metabolic homeostasis to adhesion/migration and ECM remodeling. (A) Reclustering of fibroblasts into subclusters F1–F4. (B) Pseudotime projection of fibroblast differentiation trajectories. (C) Trajectory inference illustrating F1–F4 transition path. (D) Violin plots of WTAP expression and phenotypic-switching scores across subclusters. (E) Representative pathways (respiration, growth-factor response, phenotypic switching) mapped along pseudotime. (F) Segmented pseudotime heatmaps and GO enrichment of dynamic modules. (G) Correlation of WTAP with metabolic and adhesion-related terms. (H) Pseudotime expression dynamics of WTAP, JUN/JUNB, SDHC, PPP1R15A, and TSPO. (I) Correlation bubble plot showing WTAP association with transcriptional and metabolic regulators

Pseudotime analysis revealed a continuous trajectory from F1 → F3/F2 → F4 (Fig. 5B–C), consistent with established single-cell lineage frameworks. WTAP expression and a phenotypic-switch signature peaked in early-stage (F1) fibroblasts and declined along pseudotime (Fig. 5D).

Functional trends shifted from oxidative phosphorylation toward cytoskeletal remodeling and adhesion/proliferation programs (Fig. 5E-F; Figure S6A-B), and WTAP expression correlated positively with adhesion/migration pathways and negatively with oxidative phosphorylation (Fig. 5G).

Temporal gene expression confirmed early mitochondrial activity (SDHC, TSPO) followed by upregulation of proliferation and adhesion genes (JUN, JUNB, MYC, KLF4) (Fig. 5H–I; Figure S6C–D).

Collectively, these findings identify WTAP as a dynamic marker of fibroblast pseudotime progression, capturing the shift from metabolic homeostasis to adhesion-driven ECM remodeling, a process underlying matrix deposition and scar maturation in pulmonary fibrosis.

HNRNPA2B1 marks a pro-inflammatory macrophage state and amplifies MDK/ITGB2-centered crosstalk

Among m⁶A regulators, HNRNPA2B1 showed the strongest positive associations with macrophage activation, chemotaxis, migration, differentiation, and inflammatory-response signatures (Fig. 6A). At single-cell resolution, HNRNPA2B1 expression correlated with macrophage chemotaxis and activation/migration pathway scores (Fig. 6B). Consistent with this, macrophages with high HNRNPA2B1 expression showed decreased levels of IFNGR1, TRIB1(genes linked to inflammatory/activated programs) (Fig. 6C).

Fig. 6.

HNRNPA2B1 marks a pro-inflammatory macrophage state and amplifies MDK/ITGB2-centered crosstalk. (A) Correlation of m⁶A regulators with macrophage functional signatures. (B) Scatter plots showing HNRNPA2B1 expression versus chemotaxis, activation, and migration pathway scores. (C) Violin plots showing decreased IFNGR1 and TRIB1 in HNRNPA2B1-high macrophages. (D) Volcano plot showing differential expression between HNRNPA2B1-high and -low macrophages. (E) GO enrichment of HNRNPA2B1-high macrophages. (F) GSEA enrichment plots of macrophage activation and defense-response terms. (G) Spatial feature plots of the cell type, HNRNPA2B1, and pathway enrichment scores in IPF tissue sections. (H) hdWGCNA turquoise module network of HNRNPA2B1 co-expressed genes. (I) Module–trait correlation heatmap linking turquoise module to inflammatory functions. (J) Cell–cell interaction networks in HNRNPA2B1-high vs. -low macrophages. (K) Overall pathway strength comparison for key signaling routes. (L) Chord diagram of MDK signaling (all senders) in the high state. (M) Upregulated MDK signaling partners in HNRNPA2B1-high. (N) Dot-plot of selected ligand–receptor pairs across sender–receiver lineages; color = communication probability; dot size = significance. (O) Expression distributions (high vs. low) for MDK, ITGB2, ICAM1, PECAM1; significance indicated

Differential expression analysis between HNRNPA2B1-high and -low macrophages revealed a transcriptional profile characteristic of activated macrophages, with marked upregulation of GRN, MARCO, FABP4, UBB, ALDH2, and HLA-DRA, and downregulation of the epithelial marker SFTPC (Fig. 6D).

Functionally, Enrichment analyses highlighted programs related to translation regulation, cytoskeletal remodeling, and antigen presentation, consistent with an activated macrophage state (Fig. 6E–F).

Spatial transcriptomic datasets with CellTrek and GSVA revealed predominant HNRNPA2B1 expression in macrophage cells with significant enrichment of positive regulation of macrophage activation of immune response in this compartment (Fig. 6G).

To define the network-level context of these changes, hdWGCNA identified a turquoise module strongly enriched for inflammatory and ECM-interacting genes, including TSPO, GRN, CD68, ACP5, and LGALS3 (Fig. 6H), whose eigengene aligned positively with proliferation, differentiation/activation, and inflammatory-response traits, whereas the blue module showed opposite trends (Fig. 6I).

To investigate the impact of HNRNPA2B1 on macrophage communication networks, we compared the total number of cell–cell interactions between the HNRNPA2B1-high and HNRNPA2B1-low groups. Macrophages in the HNRNPA2B1-high group exhibited relatively increased communication with several other cell types (Fig. 6J, Figure S7A). In the HNRNPA2B1-high state, signaling-family patterns showed selective increases: MDK signals were broadly distributed with stronger activity toward endothelial, fibroblast, and T-cell recipients; the ITGB2 (β2-integrin) axis intensified across multiple lineages, most prominently toward fibroblasts, endothelial cells, T cells, macrophages and smooth-muscle cells; PECAM1 activity increased mainly within endothelial circuits and in macrophage to endothelial interactions; ADGRE5 and GRN strengthened toward fibroblast/epithelial recipients; and SAA showed higher activity mainly towards club cells, smooth-muscle cells (Fig. 6K).

Chord diagram analysis revealed that multiple epithelial and stromal populations, including basal airway cells, club cells, AT2 cells, and fibroblasts, contributed substantially to MDK secretion. Notably, differential analysis demonstrated that the upregulation of MDK signaling was predominantly driven by basal airway cells and fibroblasts. MDK derived from these cell populations preferentially interacted with Syndecan family receptors, integrin complexes, and NCL across multiple target cell types, suggesting that HNRNPA2B1 may promote microenvironmental remodeling through enhancement of MDK-mediated intercellular communication. (Fig. 6L-M; Figure S7B). As a heparin-binding cytokine, MDK is known to promote leukocyte recruitment and vascular remodeling during chronic inflammation [24], consistent with our findings.

Selected ligand–receptor pairs showed higher communication probabilities in the high group, including MDK–SDC2/SDC4/NCL/LRP1, ITGB2–ICAM1, PECAM1–PECAM1, SAA1–FPR2, GRN–SORT1, and ADGRE5–CD55 (Fig. 6N), in parallel with higher expression of MDK, ITGB2, ICAM1, and PECAM1 (Fig. 6O). In parallel, the β2-integrin (ITGB2)–ICAM adhesion axis was broadly amplified across lineages, consistent with increased ITGB1/ITGB2 expression (Figure S7C–D).

Together, these results identify HNRNPA2B1 as a key regulator of pro-inflammatory macrophage states and intercellular communication. Through coordinated activation of the MDK and ITGB2 axes, HNRNPA2B1-high macrophages establish a chemokine-rich, adhesive communication network that reinforces inflammatory and fibrotic microenvironments in the lung.

WTAP promotes TGF-β–induced fibroblast activation through m⁶A-associated mechanisms

Firstly, immunofluorescence assays confirmed that WTAP expression was elevated in IPF patients (Fig. 7A). To elucidate the function of WTAP in fibroblast activation, we firstly validated the knockdown efficiency of WTAP using western blotting and RT-qPCR (Fig. 7B-C, Figure S8A). Then we assessed α-SMA and Collagen III expression in lung fibroblast cell lines following WTAP knockdown, both under basal conditions and after TGF-β stimulation. As expected, TGF-β markedly elevated α-SMA and Collagen III levels, whereas WTAP depletion substantially blunted this upregulation (Fig. 7D). Immunofluorescence analysis further confirmed the attenuation of α-SMA and ki-67 expression after WTAP loss (Fig. 7E). In addition, EdU assays revealed that TGF-β enhanced fibroblast proliferation and viability, effects that were reversed by WTAP silencing (Fig. 7F, Figure S8B). Wound healing and Transwell assays demonstrated that TGF-β promoted fibroblast migration, which was significantly reduced following WTAP knockdown (Fig. 7G-H, Figure S8C). Moreover, dot blot assays showed that global m⁶A methylation levels increased upon TGF-β treatment, whereas WTAP knockdown downregulated m⁶A methylation, suggesting that WTAP modulates TGF-β–induced fibroblast activation through m⁶A-associated mechanisms (Fig. 7I). Bioinformatics analysis revealed that WTAP influenced the expressions of genes (MYC, KLF6, NR4A3, IGFBP5), and to explore this impact, we conducted western blot assays and found that IGFBP5, NR4A3, and MYC exhibited increased expression in activated fibroblasts, whereas WTAP knockdown attenuated this effect; in contrast, KLF6 expression was not significantly affected (Fig. 7J, Figure S8D). To investigate whether m⁶A modification is involved in the regulation of the above target genes, we used the SRAMP m⁶A modification site prediction tool to predict the m⁶A sites in these target genes and found that the above target genes were rich in m⁶A modifications. Then, we designed primers for m⁶A modification sites in target genes and conducted MeRIP-qPCR to verify the m⁶A modification. The m⁶A modification levels of MYC, IGFBP5 and NR4A3 in TGF-β-induced fibroblasts were lower in sh#WTAP-transfected cells than in control cells, suggesting that their m⁶A were upregulated by WTAP. Consistent with the WB results, KLF6 did not exhibit significant changes. These finding suggests that KLF6 is unlikely to function as a downstream effector regulated by WTAP, which is in agreement with the SCENIC analysis presented in Fig. 4M (Fig. 7K). As shown in Fig. 7L, there are four putative KLF6 binding sites in the promoter region of WTAP according to the JASPAR transcription factor binding profile database [25]. We then confirmed that KLF6 binds to the − 1325 site in the promoter region of WTAP by chromatin IP assay (Fig. 7M). Dual-luciferase reporter assay further confirmed that KLF6 significantly enhances WTAP promoter activity, supporting KLF6 as an upstream transcriptional regulator of WTAP. Moreover, WB result of eight collected IPF tissue specimens validated WTAP expression was correlated positively with the expression of KLF6 (Fig. 7N).

Fig. 7.

WTAP promotes fibroblast proliferation and migration in pulmonary fibrosis through m⁶A methylation. (A) Representative immunofluorescence images of the merged photos in normal and IPF samples. (B) Representative western blot images of WTAP. (C) qRT-PCR result of the transfection efficiency. (D) Knockdown of WTAP significantly reduced the expression of α-SMA and Collagen III in TGF-β–stimulated fibroblasts and reversed their upregulation induced by TGF-β. (E) Representative immunofluorescence images to detect WTAP, Ki-67 and α-SMA expression in TGF-β-induced groups at 48 h. Scale bars, 20 μm. (F) Cell proliferation was measured by EdU staining. (G-H) Fibroblasts migration ability was measured by the wound healing and transwell assay. (I) Dot blot assay using an anti-m⁶A antibody in TGF-β-induced fibroblasts, shWTAP and NC groups. MB staining was included as a loading control. P < 0.05 was considered significant. (J) Western blot images of IGFBP5, KLF6, NR4A3 and MYC levels. (K) MeRIP-qPCR analysis of MYC, IGFBP5, KLF6 and NR4A3 in TGF-β-induced fibroblasts. (L) Putative KLF6 binding site on the promoter region of WTAP. (M) KLF6 binds to the promoter region of WTAP in TGF-β induced fibroblasts. Chromatin-IP was performed using KLF6 antibody or control IgG. Values are percentage of input. (O) KLF6 activates WTAP promoter activity in dual-luciferase reporter assay. (O) Western blotting analysis and correlation analyses of WTAP expression with expression of KLF6 in 8 freshly collected human IPF samples. *p < 0.05; **p < 0.01; ***p < 0.001 compared to the corresponding groups

Fibroblast-specific deletion of WTAP alleviates bleomycin-induced pulmonary fibrosis in mice

To determine the functional role of WTAP in pulmonary fibrosis, we generated fibroblast-specific WTAP conditional knockout (cKO) mice. Bleomycin (BLM) was administered to induce pulmonary fibrosis (Fig. 8A). ChIP–qPCR of lung chromatin from WTAP-deficient mice showed specific enrichment of KLF6 at the Wtap promoter compared with IgG, and this occupancy was further increased in bleomycin (BLM)–treated mice relative to control WTAP-deficient mice (Fig. 8B). Survival analysis revealed that a large proportion of WTAPfl/fl mice succumbed to severe fibrosis following bleomycin (BLM) administration, resulting in a survival rate of only 46.67% by day 21. In contrast, 78.57% of WTAP-cKO mice survived to the same endpoint, indicating that fibroblast-specific WTAP deletion confers protection against BLM-induced mortality (Fig. 8C).

Fig. 8.

Fibroblast-specific deletion of WTAP attenuates bleomycin-induced pulmonary fibrosis. (A) Schematic illustration of fibroblast-specific WTAP conditional knockout (cKO) mice generated by crossing WTAP^fl/fl mice with Col1a2-Cre transgenic mice and subsequent induction of pulmonary fibrosis by bleomycin (BLM). (B) Increased KLF6 occupancy at the Wtap promoter in WTAP-deficient mice after bleomycin challenge. (C) Kaplan–Meier survival curves showing the survival rate of WTAP^fl/fl and WTAP-cKO mice following BLM treatment. (D-E) Pulmonary function assessment of lung static compliance (Cst) (D) and inspiratory capacity (IC) (E) in WTAP^fl/fl and WTAP-cKO mice on day 21 after BLM challenge. (n = 4 per group). (F) Quantification of hydroxyproline content in lung tissues from WTAP^fl/fl and WTAP-cKO mice treated with BLM (n = 6 per group). (G) Representative hematoxylin and eosin (H&E) staining of lung sections showing fibrotic progression in WTAP^fl/fl and WTAP-cKO mice on day 21 post-BLM. Scale bar, 100 μm. (H-I) Representative immunofluorescence staining of COL1A1 and α-SMA (ACTA2) and corresponding quantitative analysis of fluorescence intensity in WTAP^fl/fl and WTAP-cKO lungs after BLM exposure (n = 6 per group). Scale bar, 50 μm. Data are presented as mean ± s.e.m. Statistical significance was determined by unpaired two-tailed Student’s t-test or log-rank (Mantel–Cox) test for survival analysis. P < 0.05 was considered significant. *p < 0.05; **p < 0.01; ***p < 0.001 compared to the corresponding groups

Pulmonary function tests demonstrated marked reductions in lung static compliance and inspiratory capacity in BLM-treated WTAPfl/fl mice, whereas WTAP-cKO mice exhibited substantial preservation of respiratory function following BLM exposure (Figs. 8D-E). Consistently, fibroblast-specific WTAP deficiency resulted in significantly lower hydroxyproline content compared with WTAP+/+ controls after BLM treatment (Fig. 8F), indicating a reduction in collagen deposition. Histological analysis further confirmed that fibrotic progressing was markedly attenuated in WTAP-cKO lungs on day 21 after BLM challenge (Fig. 8G). Moreover, compared with WTAPfl/fl controls, WTAP-cKO lungs displayed reduced expression of fibrotic markers, including COL1A1 and α-SMA (Fig. 8H).

In line with our bioinformatics and in vitro analyses, these in vivo findings demonstrate that fibroblast-specific WTAP ablation mitigates BLM-induced pulmonary fibrosis by limiting fibrotic progression and preserving lung function.

Discussion

This study delineates how the m⁶A machinery orchestrates the multicellular programs that underlie pulmonary fibrosis. Although m⁶A regulators are broadly expressed across lung lineages, they exhibit cell type–specific functional coupling, suggesting that disease-defining phenotypes emerge from lineage-restricted combinations of writers/adaptors and readers. This view aligns with the canonical model in which the METTL3/METTL14/WTAP complex deposits m⁶A at consensus motifs, whereas readers such as the IGF2BP and YTH families stabilize or degrade selected mRNAs to rapidly remodel cellular outputs in growth, adhesion, angiogenesis, and inflammation.

Within endothelial cells, we observed a strong association between IGF2BP3 and sprouting, migratory, and proliferative signatures, accompanied by increased information flow along the LAMININ, COLLAGEN, VEGF, PECAM1, and chemokine axes. These network-level gains are consistent with the ability of IGF2BP readers to stabilize pro-angiogenic and adhesion-related transcripts, providing a quantitative link between m⁶A decoding and endothelial activation. Experimental studies have shown that IGF2BP3 recognizes and stabilizes m⁶A-modified VEGFA transcripts, thereby promoting endothelial cell migration and tube formation. For instance, a previous study demonstrated that RBM15/YTHDF2/IGF2BP3-mediated m⁶A modification of VEGFA mRNA contributes to angiogenesis in hepatocellular carcinoma [26]. In fibrotic lungs, aberrant microvascular remodeling and endothelial–stromal crosstalk coincide with immune recruitment and matrix expansion. Our findings suggest that IGF2BP3 amplifies this coupling by enhancing adhesive and chemokine-rich communication, thereby potentiating pericyte/smooth-muscle and immune engagement [27, 28]. Recent syntheses emphasizing m⁶A-mediated control of endothelial homeostasis and vascular pathology further reinforce the plausibility and therapeutic tractability of this node [29] and [30] provided mechanistic insights linking m⁶A writers/readers to vascular remodeling and disease progression.

Having defined the endothelial node, we next focused on fibroblasts, the principal effectors of matrix consolidation, with WTAP emerging as a central writer governing this transition. In the fibroblast compartment, WTAP, an adaptor within the writer complex, shows strong connectivity to an ECM/contractile module enriched for epithelial–mesenchymal transition, collagen fibril organization, and migration. Pseudotime reconstruction aligns WTAP with a metabolic-to-mechanical switch, marking the transition from oxidative metabolism toward adhesion, motility, and ECM assembly, hallmarks of fibroblast-to-myofibroblast differentiation during scar maturation [31]. Mechanistically, our data support a KLF6-WTAP regulatory axis, and indicate that WTAP promotes m⁶A enrichment on MYC, NR4A3, and IGFBP5, thereby sustaining TGF-β–driven fibroblast activation.

Functional perturbation in lung fibroblast models demonstrates that dampening writer activity attenuates TGF-β–induced activation: in human and murine lung fibroblasts, silencing METTL3 reduces α-SMA/Collagen expression and inhibits fibroblast-to-myofibroblast transition, and in vivo METTL3 knockdown alleviates bleomycin-induced fibrosis [8, 11]. Across organs, WTAP promotes fibroblast proliferation/migration and matrix programs in the heart, and its depletion blunts these responses [9], supporting a writer-centric, m⁶A-associated mechanism of fibroblast activation.

In vivo model appraisals underscore the translational considerations of bleomycin injury [32, 33]. While fibroblast-specific WTAP knockout in lung has not yet been published to our knowledge, convergent evidence from lung METTL3 perturbation and cross-organ WTAP studies supports writer-associated control of matrix remodeling.

Consistent with these findings, machine-learning analyses in IPF identify WTAP as a top fibrosis-related regulator and nominate candidate small molecules [34]. Cross-organ reviews converge on a conserved “writer-centric” paradigm for fibrosis, implicating WTAP/METTL3/METTL14 in ECM remodeling [35, 36]. Analogous m⁶A-high/ECM-high states have been synthesized in skin scarring and wound healing frameworks [30], supporting a generalizable metabolic-to-mechanical transition downstream of writer activity.

In contrast to WTAP’s role in fibroblast remodeling, HNRNPA2B1 marks a pro-inflammatory macrophage state characterized by upregulation of antigen-presentation, actin-binding, and translation-related programs, along with strengthened communication through Midkine (MDK) and ITGB2 axes. MDK, a heparin-binding cytokine, promotes leukocyte recruitment and vascular remodeling during chronic inflammation [37], whereas ITGB2 (within LFA-1/Mac-1) mediates firm adhesion and trans-endothelial migration via ICAM interactions [38].

A systems perspective emerges from integrating intercellular signaling with intracellular module capacity. Integrating CellChat with hdWGCNA links communication partners to the intracellular programs that enable heightened signaling, positioning m⁶A regulators within both network layers. This elevates IGF2BP3, WTAP, and HNRNPA2B1 from correlates to network-positioned, testable regulators.

Several limitations should be noted. The inferred communication maps reflect steady-state transcript levels and predicted ligand–receptor compatibility rather than protein localization, secretion kinetics, or site-specific m⁶A occupancy. Pseudotime ordering is inferential and may conflate parallel trajectories in highly plastic populations. Future studies should integrate long-read single-cell transcriptome/epitranscriptome sequencing to identify transcript-specific m⁶A sites, and apply spatial transcriptomics to anchor communication edges in situ. Establishing causality and defining therapeutic windows will require CRISPRi/a or targeted degradation in primary human cells, lung organoids, and humanized mouse models. Given the pleiotropy of m⁶A regulators, precise temporal control and cell-restricted delivery(potentially via lipid nanoparticles or transient editing)will be essential [39, 40].

Translationally, the m⁶A–communication–ECM map motivates multi-node strategies, co-targeting IGF2BP3-linked endothelial activation, WTAP-centered fibroblast programs alongside TGF-β or noncanonical WNT5A–JNK–ROCK–ITGAV blockade [41]; and HNRNPA2B1-driven macrophage adhesion could be attenuated by pairing m⁶A modulation with MDK antagonism or ITGB2/ICAM inhibition [37]. These proposals align with the emerging paradigm that IPF progresses through coupled endothelial–immune–mesenchymal circuits, advocating for multi-node rather than single-axis interventions [41].

In summary, our data nominate a KLF6-WTAP-m⁶A (MYC/NR4A3/IGFBP5) axis that couples fibroblast activation to endothelial–immune modules, suggesting that targeting WTAP-centered m⁶A regulation may rebalance multicellular communication and slow IPF progression.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

GSEA

Gene set enrichment analysis

DEGs

Differentially expressed genes

RT-qPCR

Quantitative real-time polymerase chain reaction

WB

Western blotting

GO

Gene ontology

KEGG

Kyoto Encyclopedia of Genes and Genomes

scRNA-seq

single-cell RNA sequencing

IPF

Idiopathic pulmonary fibrosis

m⁶A

N⁶-methyladenosine

ECM

Extracellular-matrix

t-SNE

t-distributed stochastic neighbor embedding

MDK

Midkine

BLM

Bleomycin

ITGB2

β2-integrin

WGCNA

Weighted gene co-expression network analysis

EC

Endothelial cell

LR

Ligand–receptor

Author contributions

CYX and SSC were responsible for the research design and conceptualization. SSC curated the data and performed formal analyses. ZHZ acquired funding and provided resources for the study. CYX and SSC conducted the investigation and developed the methodology. CYX and SSC administered the project, and BL contributed to the software used for data processing. SQZ validated the results, while YNL contributed to the visualization of the data. CYX and SSC wrote the original draft of the manuscript, and both CYX and SSC reviewed and edited the manuscript. CYX, SSC, and BL were consulted for molecular pathology of IPF, while SQZ and ZFW supervised and supported the research. All authors have read and approved the final manuscript, and therefore, have full access to all the data in the study and take responsibility for the integrity and security of the data.

Funding

This work was funded by Shanghai Clinical Pharmacy Key Specialty Construction Project Support project (Shanghai Health and Pharmaceutical Administration) (201809).

Data availability

All sequencing data could be downloaded at (http://www.ncbi.nlm.nih.gov/geo).

Declarations

Ethics approval and consent to participate

This study was approved by the Ethics Committee at the Yangpu Hospital, Tongji University (Shanghai, China). Written informed consent was obtained from all participants. The animal experiments were executed in line with the documents of the Animal Center of Tongji University (Shanghai, China).

Consent for publication

All authors approved the final manuscript and consent to its publication.

Conflict of interest

The authors have declared that no competing interests exists.