Single-cell and Spatial Transcriptomic Profiling Reveal that LAPTM5-mediated Ferroptosis in Macrophages Induces Fibroblast Dysfunction and Amplifies Periodontal Inflammation

Abstract

Periodontitis represents a persistent inflammatory condition marked by gradual damage to the gingival connective tissues and alveolar bone. Programmed cell death (PCD) is essential for preserving immune balance and includes various forms such as apoptosis, pyroptosis, necroptosis, and ferroptosis. Nevertheless, the principal PCD pathway that contributes periodontal inflammation has not been clearly identified. In this work, we integrated multiple publicly available single-cell RNA sequencing datasets and applied diverse gene set scoring approaches to systematically characterize the dynamic expression of PCD-associated genes in periodontitis-affected tissues. Ferroptosis emerged as one of the important PCD pathways, mainly occurring in macrophages. Cell-cell communication and spatial analyses suggested that ferroptotic macrophages were located adjacent to fibroblast-enriched regions and may interact with fibroblasts through Galectin and OSM signaling pathways. Functionally, ferroptotic macrophages suppressed fibroblast proliferation and migration while amplifying their pro-inflammatory response, underscoring their pivotal role in sustaining chronic inflammation. Furthermore, machine-learning analyses identified LAPTM5 as one of the ferroptosis-associated hub genes in macrophages. Knockdown of LAPTM5 in macrophages suppressed ferroptosis and subsequently attenuated fibroblast inflammatory responses. Collectively, our study highlights ferroptosis as one of the key pathogenic mechanisms in periodontitis and identifies LAPTM5-driven macrophage ferroptosis as a key driver of fibroblast dysfunction and chronic inflammation, providing potential therapeutic insights for restoring periodontal immune balance.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1007/s10753-026-02559-7.

Introduction

Periodontitis is a common, long-standing inflammatory condition of the tooth-supporting apparatus, characterized by progressive breakdown of the gingival tissues, periodontal ligament and alveolar bone, which can ultimately result in tooth loss and contribute to systemic disease. Despite progress in debridement, regenerative techniques, and host-modulatory therapy, achieving long-term periodontal stability remains challenging due to difficulties in post-treatment maintenance and the high risk of disease recurrence. The persistence of dysregulated host responses—rather than microbial infection alone—has emerged as the key factor driving disease progression [1, 2]. Excessive oxidative stress, pro-inflammatory cytokine release, and impaired resolution responses collectively result in a self-sustaining inflammatory microenvironment that promotes continuous tissue damage [3]. However, the cellular mechanisms underlying this irreversible inflammatory amplification remain incompletely understood.

Macrophages are key regulators of innate and adaptive immunity within the periodontal milieu. These cells display substantial functional plasticity, shifting between pro-inflammatory M1-like and inflammation-resolving M2-like phenotypes [4]. In healthy gingival tissues, macrophages maintain homeostasis through phagocytosis of apoptotic cells and secretion of anti-inflammatory mediators. In contrast, during periodontitis, they become hyperactivated under the influence of bacterial components and inflammatory cytokines, releasing excessive reactive oxygen species (ROS), TNF-α, and IL-1β [5]. Such sustained activation not only amplifies inflammation but also increases the susceptibility of macrophages to various forms of stress-induced cell death, thereby fueling a vicious inflammatory cycle [6, 7]. Programmed cell death (PCD) is a fundamental biological process that regulates immune cell turnover and tissue remodeling. Beyond classical apoptosis, emerging modes of PCD such as pyroptosis, necroptosis, and ferroptosis have been increasingly implicated in inflammatory disorders [8, 9]. Ferroptosis, a type of regulated necrosis, features iron-driven lipid peroxidation, glutathione exhaustion, and glutathione peroxidase 4 (GPX4) inhibition [10, 11]. Growing data underscore its significant role in chronic inflammatory conditions by inducing oxidative damage and intensifying immune reactions [12, 13]. In the context of periodontitis, macrophages are particularly susceptible to ferroptotic stress due to their high iron content and active oxidative metabolism [6]. Studies have demonstrated that lipopolysaccharide (LPS) markedly elevates intracellular Fe²⁺ levels in RAW264.7 macrophages and triggers ferroptosis by upregulating key ferroptosis-associated molecules, such as acyl-CoA synthetase long-chain family member 4 (ACSL4) and transferrin receptor protein 1 (TFR1). The ferroptotic process subsequently activates damage-associated molecular patterns (DAMPs) and promotes the release of pro-inflammatory mediators, thereby amplifying oxidative stress and exacerbating periodontal inflammation [14]. Moreover, ROS generated during ferroptosis negatively influences the osteogenic potential of bone marrow-derived mesenchymal stem cells (BMSCs), thereby impeding bone repair and aggravating bone loss in apical periodontitis models [15]. However, the precise cellular sources and molecular mechanisms through which ferroptosis drives the pathogenic microenvironment in periodontitis remain poorly defined.

In this study, we combined single-cell RNA sequencing (scRNA-seq) with spatial transcriptomics to map PCD patterns in periodontitis and applied multiple computational approaches alongside experimental validation. Our analyses indicated that ferroptosis represents a highly active PCD pathway in diseased periodontal tissues and is primarily enriched in macrophages. We further uncovered a distinct crosstalk between ferroptotic macrophages and fibroblasts, in which macrophage ferroptosis suppresses fibroblast proliferation and migration while intensifying their inflammatory responses. In addition, machine-learning analysis identified LAPTM5 and AHR as two ferroptosis-associated hub genes within macrophages. Experimental validation confirmed that knockdown of LAPTM5 in macrophages attenuated ferroptosis, which subsequently alleviated the inflammatory response and functional impairment in co-cultured fibroblasts. Collectively, our findings suggest that LAPTM5-mediated macrophage ferroptosis contributes to fibroblast dysfunction and inflammatory amplification, thereby promoting periodontal pathology.

Methods

Data Collection and Processing

Transcriptomic profiles were acquired from the GEO repository for consolidated examination. The scRNA-seq cohort was constructed by combining three datasets—GSE152042 (2 healthy and 2 inflamed gingival samples), GSE164241 (13 healthy and 8 periodontitis samples), and GSE161267 (5 healthy gingival samples)—to provide a comprehensive representation of both healthy and diseased gingival tissues. For bulk transcriptomic validation, three independent datasets were included. The microarray dataset GSE16134, consisting of 69 healthy and 241 periodontitis gingival samples, served as a large external validation cohort. In addition, two bulk RNA-seq datasets, GSE173078 and GSE223924, were integrated as an additional validation cohort, comprising 22 healthy and 22 periodontitis gingival samples in total. To minimize technical variation between datasets, batch effects were adjusted using the removeBatchEffect function in the limma package. The batch-corrected expression matrix was subsequently used for visualization analyses. To further resolve the spatial organization of cell populations within periodontitis lesions, two Visium spatial transcriptomic samples (GSM6258256 and GSM6258257) from dataset GSE206621 were included, enabling in situ assessment of cellular localization and microenvironmental structure. All datasets included in this study are summarized in Table S1.

scRNA-seq Analysis

Single-cell RNA sequencing (scRNA-seq) data were processed with the Seurat toolkit (version 4.3.1). Quality control filtering excluded cells expressing fewer than 200 or more than 5,000 genes, along with those where mitochondrial gene content exceeded 30%. The SCTransform method was applied for data normalization, and principal component analysis (PCA) was then performed for dimensionality reduction. To mitigate potential batch effects across samples, the Harmony algorithm was applied to the PCA embeddings. Harmony is specifically designed for the integration of multiple single-cell datasets and efficiently corrects dataset-specific batch effects while preserving biological variation across cell populations [16, 17]. Cells were subsequently clustered using the FindNeighbors and FindClusters functions based on the Harmony-corrected dimensions, and visualized in two-dimensional space using UMAP. Cluster-defining marker genes were identified using the FindAllMarkers function (log2FC > 0.25; pct > 0.25), and each cluster was biologically annotated through comparison with established lineage-specific markers.

Gene Set Scoring in scRNA-seq

To quantify pathway activity at single-cell resolution, five scoring strategies were used: AUCell, UCell, singscore, ssGSEA, and AddModuleScore. AUCell measures the AUC of ranked expression patterns; UCell relies on normalized rank-based statistics; singscore generates a rank-based score reflecting gene-set activation; ssGSEA compares gene-set expression relative to the entire transcriptome; and AddModuleScore calculates average expression of gene modules with subsequent normalization.

Metabolic Pathway Scoring

Metabolic pathway activity was quantified using the scMetabolism R package [18], which provides curated KEGG- and Reactome-based metabolic gene sets. Scores for each cell were generated with the RunSCMetabolism function using the default VISION method. The resulting matrix of metabolic activities was visualized using bubble plots to highlight differences across major pathways.

GSVA and GSEA Enrichment Analysis

To reveal biological pathways associated with different risk categories, we performed gene set variation analysis (GSVA) and gene set enrichment analysis (GSEA). GSVA was carried out using the GSVA R package, while GSEA was conducted with clusterProfiler. Gene sets were downloaded from the MSigDB repository (https://www.gsea-msigdb.org/gsea/msigdb). Pathways with statistical significance at P < 0.05 were retained.

Cell-cell Communication Analysis

CellChat was used to infer potential intercellular communication changes. Following the standard workflow, we quantified ligand-receptor interactions using the default CellChatDB database. Overexpressed ligands and receptors in each cell population were identified to predict cell type-specific signaling relationships and enhanced interaction patterns.

Analysis of Transcription Factor Regulon Activity Using SCENIC

To characterize transcription factor (TF) regulatory programs, we implemented the SCENIC workflow using the log-normalized expression matrix generated from Seurat as input. Co-expression modules were first inferred to identify candidate TF-target gene relationships. These modules were then refined through motif enrichment analysis based on the hg38-tss-centered-10 kb cis-regulatory motif database, allowing the construction of high-confidence regulons for each TF. AUCell scoring was used to quantify regulon activity per cell by ranking transcript abundance and computing enrichment values. The activity scores were subsequently binarized using SCENIC’s default threshold to distinguish active from inactive TF regulons. The resulting regulon activity matrix was used for downstream clustering, visualization, and comparative analyses of TF regulatory states among different cellular populations.

Pseudotime Trajectory Analysis

Cell-state progression was reconstructed using the Monocle2 package. A cell-by-gene matrix was generated, and cells were arranged along a continuous pseudotime trajectory to identify developmental states. To further estimate the differentiation potential of macrophages, CytoTRACE analysis was conducted using the CytoTRACE package in R. CytoTRACE predicts cellular differentiation states based on gene expression diversity and transcriptional profiles, where higher CytoTRACE scores indicate less differentiated cells, whereas lower scores correspond to more differentiated cells. To characterize dynamic transcriptional changes during macrophage state transitions, differentially expressed genes were identified along the pseudotime trajectory. Functional enrichment analyses of these genes were subsequently performed to explore stage-specific biological pathways associated with macrophage differentiation.

Spatial Data Processing

Spatial transcriptomic data were processed in Seurat, with low-quality spots (containing fewer than 10 genes) being excluded. SCTransform normalization was applied while regressing out mitochondrial and ribosomal gene signals. Dimensionality reduction and clustering were performed to define spatial transcriptional domains, visualized through UMAP and Seurat’s spatial mapping functions. SpatialDimPlot was used to display cluster distribution, and SpatialFeaturePlot assessed gene-expression patterns across sections.

Spatial Deconvolution Analysis

To estimate the cellular composition of each spatial spot, we used cell2location (v0.1.4) with scRNA-seq data as a reference. Cell2location employs a Bayesian framework to integrate single-cell and spatial transcriptomics data, enabling the inference of cell-type-specific abundance across spatial locations. The model was trained for 30,000 iterations to ensure convergence. This approach accounts for technical variability while preserving biological signals, thereby enabling robust and high-resolution mapping of cell-type distributions within the tissue.

Spatial Interaction Analysis

To investigate spatially resolved cell-cell interactions, we applied mistyR. This approach integrates spatial context with gene expression by modeling transcriptional activity through a multiview machine-learning framework, enabling the quantification of neighborhood-dependent regulatory influences and the identification of spatially organized signaling networks.

Identification of Macrophage Ferroptosis-related Modules and Genes using hdWGCNA

To identify co-expression signatures linked to macrophage ferroptosis, we constructed a single-cell gene co-expression network using the hdWGCNA workflow. Genes expressed in ≥ 5% of macrophages were retained, and metacell matrices were generated with the MetacellsByGroups function to reduce noise and enhance network robustness. For metacell construction, we used K = 25 for the K-nearest neighbor algorithm and max_shared = 10, which limited the number of shared cells between metacells to preserve their distinct identity. The Seurat object was restricted to macrophage clusters, after which the TestSoftPowers function was used to determine the appropriate soft-thresholding power to approximate scale-free topology. Network construction and module detection were conducted using ConstructNetwork, following the hdWGCNA protocol (https://smorabit.github.io/hdWGCNA/articles/basic_tutorial.html). To further evaluate the robustness and reproducibility of the identified co-expression modules, module stability and preservation analyses were performed using the modulePreservation function implemented in the WGCNA framework. In brief, a permutation-based preservation test was applied to assess whether the module structure identified in the reference dataset was reproducible in resampled datasets. The preservation statistics were summarized using the Zsummary score, which integrates multiple preservation metrics including density and connectivity. According to standard WGCNA criteria, modules with Zsummary > 10 were considered strongly preserved, 2 < Zsummary ≤ 10 moderately preserved, and Zsummary ≤ 2 not preserved. In addition, medianRank statistics were calculated to complement Zsummary, where lower medianRank values indicate stronger module preservation and are less dependent on module size.

Ferroptosis Subtype Identification and Functional Characterization

Unsupervised consensus clustering was performed using ConsensusClusterPlus to classify samples based on ferroptosis-related gene expression. The optimal cluster number was determined by examining cumulative distribution function (CDF) distributions and consensus matrices, and PCA was used to confirm subtype separation. GSVA, GSEA and ssGSEA were used to assess functional pathways and immune infiltration patterns across ferroptosis subtypes. Group differences were compared using the Wilcoxon rank–sum test and displayed with boxplots.

Machine Learning Algorithms

To screen for key biomarkers and build a predictive scoring model, LASSO regression and random forest (RF) algorithms were applied. LASSO uses an L1 penalty to shrink less informative coefficients, reducing overfitting and retaining essential features. The optimal λ value was selected by minimizing binomial deviance. RF was used to rank variable importance based on mean decrease in Gini impurity, providing complementary evidence for key predictors.

Tissue Sample Collection

Gingival tissues were collected from periodontally healthy individuals and patients with periodontitis at the Stomatology Hospital, Zhejiang University School of Medicine. Healthy gingival tissues were obtained from individuals undergoing third molar extraction who presented clinically healthy periodontal conditions, defined as probing depth ≤ 3 mm, no bleeding on probing, and no clinical attachment loss. Periodontitis tissues were obtained from patients diagnosed with periodontitis according to the 2018 World Workshop classification of periodontal diseases. These patients exhibited probing depth ≥ 5 mm, clinical attachment loss ≥ 3 mm, bleeding on probing, and radiographic evidence of alveolar bone loss. Inflamed gingival tissues were collected from periodontal pocket regions during flap surgery. Exclusion criteria included systemic illness, pregnancy or lactation, recent antibiotic or immunosuppressive treatment (within the preceding 3 months), and smoking or alcohol abuse. Each specimen was immediately divided into two portions: one fixed in 4% paraformaldehyde and paraffin-embedded, and the other preserved in RNAlater for RNA extraction. Written informed consent was obtained from all participants, and all procedures were approved by the Medical Ethics Committee of the Stomatology Hospital, Zhejiang University School of Medicine (Approval No. I20250130 and 2025-118 (R)).

Immunohistochemical (IHC) and Immunofluorescence (IF) Staining

Gingival tissues were fixed in 4% paraformaldehyde, paraffin-embedded, and sectioned at 4 µm. For IHC, sections were deparaffinized, rehydrated, and subjected to citrate buffer-based antigen retrieval (pH 6.0, 95 °C, 10 min). Endogenous peroxidase was blocked using 3% H₂O₂, followed by 5% BSA blocking. Primary antibodies were applied overnight at 4 °C, followed by HRP-conjugated secondary antibodies and DAB detection. Hematoxylin was used for nuclear counterstaining. For IF staining, sections underwent similar preparation and were incubated with Alexa Fluor 488- or 594-conjugated secondary antibodies (Invitrogen, USA), with DAPI staining of nuclei. Fluorescence images were captured using an upright microscope.

Cell Lines and Primary Cell Culture

Mouse-derived Cell Lines

RAW264.7 murine macrophages and L929 mouse fibroblasts were maintained under standard culture conditions. Cells were incubated at 37 °C in a humidified atmosphere containing 5% CO₂. RAW264.7 cells were cultured in DMEM/F-12, whereas L929 cells were maintained in DMEM. All media contained 10% fetal bovine serum and 100 IU/mL penicillin-streptomycin. Cells were passaged at 70-80% confluence, and culture media were replaced every 2-3 days to sustain optimal growth.

Human Gingival Fibroblasts (HGFs)

HGFs were isolated from healthy gingival tissues obtained from donors undergoing routine dental procedures. Briefly, gingival tissues were rinsed with PBS containing penicillin–streptomycin to remove blood and debris, and then cut into small fragments (~ 1 mm³) under sterile conditions. The tissue fragments were placed onto culture dishes and allowed to adhere for approximately 4 h. The culture flasks were then gently inverted to facilitate attachment of the tissue explants. After attachment, DMEM supplemented with 10% fetal bovine serum and 100 IU/mL penicillin-streptomycin was added. Cells were maintained at 37 °C in a humidified atmosphere containing 5% CO₂, and the medium was replaced every 2-3 days. HGFs migrated out from the tissue explants and were expanded in culture. HGFs between passages 3-6 were used for subsequent experiments.

Human Monocyte-derived Macrophages (HMDMs)

Human primary monocytes were isolated from the peripheral venous blood of healthy volunteers using a monocyte isolation kit strictly according to the manufacturer’s instructions. Briefly, peripheral blood was diluted with PBS at a 1:1 ratio, layered onto the separation medium, and centrifuged at 400 × g for 30 min at room temperature. The monocyte-enriched layer was carefully collected, washed twice with PBS, and resuspended in RPMI 1640 medium supplemented with 10% FBS and 1% penicillin–streptomycin. To induce monocyte differentiation into macrophages, recombinant human macrophage colony-stimulating factor (M-CSF) was added to the culture medium at a final concentration of 50 ng/mL. The cells were seeded in culture plates and incubated at 37 °C with 5% CO₂ for 7 days, with half of the medium replaced every 2-3 days.

Macrophage-fibroblast Co-culture Model

To further investigate the interaction between macrophages and fibroblasts, we utilized a Transwell co-culture system (Corning, New York, USA) [19]. Two species‑specific models were established: (1) murine RAW 264.7 macrophages co‑cultured with L929 fibroblasts, and (2) HMDMs co‑cultured with HGFs. In both systems, macrophages were seeded in the upper chamber and fibroblasts in the lower wells at a macrophage‑to‑fibroblast ratio of 1:4. Prior to co‑culture, macrophages were pre‑stimulated for 24 h with 1000 ng/mL LPS in the presence or absence of the ferroptosis inhibitor Fer‑1. After stimulation, the medium was replaced with fresh complete medium, and co‑culture was performed for 24 h. Fibroblasts were then harvested for subsequent analyses of proliferation, migration, and expression of inflammation‑related genes.

Cell Transfection

To investigate whether LAPTM5 regulates ferroptosis in macrophages, small interfering RNAs (siRNAs) targeting LAPTM5 and a negative-control siRNA were synthesized (Table S2; GenePharma, Shanghai, China). RAW 264.7 macrophages and HMDMs were seeded into 6-well plates and cultured until reaching 60–70% confluence. Lipofectamine™ 2000 (Invitrogen) was mixed with LAPTM5 or control siRNA (final concentration: 25 nM) following the manufacturer's protocol. Cells were harvested 24-48 h post-transfection for downstream analyses.

CCK-8 Assay

Cell viability was assessed using the CCK-8 kit (Dojindo, Japan). Two independent experiments were performed to evaluate: (1) the direct effect of ferroptosis on macrophage viability, and (2) the indirect effect of macrophage ferroptosis on fibroblast viability. To evaluate the direct effect of ferroptosis on macrophage viability, Raw 264.7 were seeded in 96-well plates (1 × 10⁴ cells/well) and treated with LPS in the presence or absence of Fer-1 for 24 h. To evaluate the indirect effect of macrophage ferroptosis on fibroblast viability, fibroblasts were seeded in 96-well plates (5 × 10³ cells/well) and treated with the corresponding macrophage-conditioned medium for 24 h. 10 µL of CCK-8 solution was then added, incubated for 2 h, and absorbance was measured at 450 nm. Viability was calculated relative to untreated controls. All experiments were performed in triplicate.

Wound-healing Assay

To determine whether macrophage ferroptosis affects fibroblast migration, a scratch-wound assay was conducted. Fibroblasts were cultured in 6-well plates until they reached ~ 90% confluence, after which a linear scratch was created using a sterile 200-µL pipette tip. Detached cells were removed by washing with PBS, and serum-free medium—containing conditioned medium from macrophages undergoing ferroptosis or the corresponding controls—was added. Images were taken at 0 h and 24 h using an inverted microscope. The wound area was quantified with ImageJ.

RNA Extraction and Quantitative Real-time PCR (RT-qPCR)

Total RNA from tissues or cultured cells was isolated using TRIzol (Invitrogen). cDNA synthesis was performed using 1 µg RNA and the Vazyme reverse-transcription kit. RT-qPCR was conducted using SYBR Green Master Mix (TaKaRa) on a QuantStudio system (Applied Biosystems). GAPDH served as the reference gene, and relative expression was calculated using the 2^ − ΔΔCT method. Reactions were run in triplicate. Primer sequences are listed in Table S3.

Enzyme-linked Immunosorbent Assay (ELISA)

Cytokine levels in supernatants were quantified using commercial ELISA kits. Supernatants were collected, centrifuged (1,000 × g, 5 min, 4 °C), and added to pre-coated 96-well plates along with standard dilutions. After incubation and washing, HRP-conjugated detection antibodies were applied, followed by TMB substrate. The reaction was stopped, and absorbance at 450 nm was measured. Concentrations were obtained from standard curves.

Assessment of Lipid Peroxidation and Labile Fe²⁺ in Macrophages

Lipid peroxidation is a hallmark event in ferroptosis. To assess this process in macrophages, BODIPY 581/591 C11 staining (Thermo Fisher Scientific) was employed. The BODIPY 581/591 C11 probe undergoes a fluorescence emission shift from red to green upon oxidation. Red fluorescence indicates non-oxidized lipids, while green fluorescence corresponds to oxidized lipid peroxides. Cells were seeded in glass-bottom dishes at a density of 1 × 10⁴ cells per well and treated with either LPS or control medium for 24 h. Subsequently, the cells were incubated with 10 μM BODIPY 581/591 C11 probe and Hoechst stain in the dark at 37 °C for 30 min. Fluorescence images were captured using an upright Olympus FV3000 confocal laser scanning microscope. In parallel, intracellular labile ferrous iron (Fe²⁺) levels were assessed using FerroBrite™ Green (AAT Bioquest), a live-cell-permeable fluorescent probe sensitive to Fe²⁺ imbalance during ferroptosis. Under ferroptotic conditions, intracellular Fe²⁺ accumulation leads to a reduction in FerroBrite™ Green fluorescence intensity, enabling real-time monitoring of iron dysregulation. After 24 h of LPS treatment, cells were incubated with 10 μM FerroBrite™ Green working solution under the same conditions, and fluorescence was captured using the FITC channel on the same microscope.

Ferrostatin-1 Treatment

Ferroptosis inhibition was achieved by treatment with Ferrostatin-1. For all experiments requiring ferroptosis blockade, cells were pretreated with Ferrostatin-1 (1‑2 μM) for 1 h before stimulation and maintained in the presence of Ferrostatin-1 throughout the treatment period. The same volume of vehicle was added to control groups.

Statistical Analysis

Data processing and statistical calculations were executed in R (4.3.1). Comparisons of continuous data were performed with Student’s t-test or Wilcoxon test; and categorical comparisons used chi-square or Fisher’s exact test. FDR adjustment was applied to correct for multiple comparisons. Pearson correlations were computed for association analyses. Statistical significance was set at P < 0.05 (two-tailed).

Results

Single-cell Transcriptomic Analysis Reveals Major Cellular Composition and Heterogeneity in Periodontitis

To gain a comprehensive understanding of the cellular composition in human gingival tissues, we merged three publicly available scRNA-seq datasets (GSE161267, GSE152042, and GSE164241), encompassing 20 healthy and 10 periodontitis gingival samples. Quality control was then performed on the combined dataset to ensure consistent filtering criteria across all samples (Figure S1A-B), yielding 112,087 high-quality cells, including 12,546 cells from GSE152042, 96,702 cells from GSE164241, and 2,839 cells from GSE161267. To eliminate the batch effect between different datasets, we performed batch effect correction using Harmony, which effectively eliminated technical variations across datasets while preserving biological heterogeneity (Figure S1C-D). Finally, we identified 14 distinct cell populations after integrating the three scRNA-seq datasets (Fig. 1A; Figure S1E). Representative marker genes were used to annotate major cell populations, including COL1A1, DCN, and LUM for fibroblasts; AQP1, CLDN5, VWF and PECAM1 for endothelial cells; KRT5 and KRT14 for epithelial cells; ACTA2 and TAGLN for vascular mural cells; C1QA, C1QB, and CD163 for macrophages; CD79A and MS4A1 for B cells; MGB1 and IGHG1 for plasma B cells; NAMPT, G0S2 and SOD2 for neutrophils; IRF7 and SOX4 for plasmacytoid dendritic cells (pDCs); CLEC9A and CD1C for myeloid dendritic cells (mDCs); CPA3 and TPSAB1 for mast cells; TOP2A and MKI67 for proliferative cells; and TRAC, CD3D and NKG7 for T/NK cells (Fig. 1B). Figure S1F highlighted several genes that are highly expressed in each cell types. Next, we quantified the relative abundance of each cell population within the periodontal microenvironment and compared their proportions between healthy and periodontitis samples to explore inflammation-associated cellular remodeling. Overall, fibroblasts, endothelial cells, and immune cells constituted the major cellular components in both conditions (Figure S1G). Notably, cell proportion analysis revealed a substantial expansion of macrophages, neutrophils, and plasma cells in periodontitis, accompanied by a relative reduction in fibroblast and endothelial cell fractions—findings consistent with inflammatory infiltration and tissue destruction (Figure S1H). Collectively, these results delineate the key stromal and immune subpopulations that constitute the periodontal microenvironment, providing a comprehensive cellular atlas that highlights the distinct remodeling of immune and structural compartments in periodontitis.

Fig. 1.

Ferroptosis represents a major form of PCD in periodontitis. (A) UMAP visualization showing distinct cell clusters within gingival tissues. (B) The heatmap summarizing canonical marker expression patterns across major cell types. (C) Comparison of 18 PCD pathways across major cell populations using five independent scoring algorithms. (D) Violin plots comparing ferroptosis scores between healthy and periodontitis samples across major cell populations. (E) GSEA of the bulk RNA-seq dataset validating ferroptosis enrichment in periodontitis. (F-G) Representative IHC staining of GPX4 and ACSL4 in human gingival tissues. (H) UMAP visualization segregating cells into high- and low-ferroptosis subgroups based on enrichment scores. (I) Cell-cell communication networks inferred by CellChat highlighting distinct signaling patterns between high- and low-ferroptosis groups. Red indicates signaling pathways that are upregulated in the high-ferroptosis group, whereas blue denotes pathways that are enhanced in the low-ferroptosis group. (J) Scatter plot illustrating the overall outgoing and incoming interaction strengths of each cell type in high- versus low-ferroptosis subgroups. (K) Bar plots summarizing the total number and overall strength of intercellular interactions in both groups. (L) Ranking of signaling pathways by relative information flow between high- and low-ferroptosis conditions

Ferroptosis as a Key PCD Pathway in Periodontitis Linking Inflammation and Metabolism

To clarify the key patterns of PCD in periodontitis, we systematically evaluated 18 PCD signatures across major cell populations using multiple independent scoring frameworks. Across various scoring methods and integrated analyses, ferroptosis consistently ranked among the top enriched PCD pathways (Fig. 1C; Figure S2A-E). Correlation analyses among the different algorithms revealed strong positive associations, supporting the robustness of our findings (Figure S2F). Meanwhile, we observed pronounced alterations in ferroptosis scores across multiple cell types (Fig. 1D). Consistently, GSEA based on the bulk RNA-seq dataset GSE16134 revealed a significant upregulation of ferroptosis-related gene signatures in periodontitis samples (NES = 1.5077, P = 0.0161) (Fig. 1E). IHC staining showed reduced GPX4 and increased ACSL4 expression in periodontitis gingival tissues compared with healthy controls, collectively indicating enhanced ferroptosis activity in diseased samples (Fig. 1F-G).

To explore the transcriptional heterogeneity associated with ferroptotic activity across the periodontal microenvironment, all cells were subsequently classified into high- and low-ferroptosis subgroups based on the ferroptosis score (Fig. 1H). Cell-cell communication analysis revealed that both the number and intensity of intercellular interactions were markedly increased in the high-ferroptosis group (Fig. 1I, K). In particular, macrophages, neutrophils, and fibroblasts displayed enhanced incoming and outgoing signaling strength, suggesting a ferroptosis-associated amplification of inflammatory crosstalk (Fig. 1J). Further ligand-receptor network analysis revealed that the high-ferroptosis subgroup exhibited a more active and complex communication landscape, characterized by an increased number of ligand-receptor interactions. Notably, multiple inflammation- and oxidative stress-related signaling pathways were upregulated, including Galectin [20–23], Oncostatin M (OSM) [24, 25], IL1, IL6, and CXCL (Fig. 1L; Figure S3A). Additionally, to further investigate metabolic differences between the subgroups, we assessed metabolic pathway activity and found significant upregulation of metabolism-related pathways in the high-ferroptosis group (Figure S3B). Collectively, these results indicate that enhanced ferroptotic activity profoundly remodels both inflammatory and metabolic signaling networks within periodontal tissues, creating a pro-oxidative, cytokine-rich microenvironment that promotes persistent inflammation.

Ferroptotic Macrophages Exhibit Enhanced Inflammatory Signaling and Metabolic Reprogramming

Our analysis revealed that macrophages exhibited the highest ferroptosis activity among all cell populations, which was elevated in periodontitis compared with healthy samples (Fig. 2A-B). To validate this observation and characterize macrophage ferroptosis, we performed IF staining and in vitro experiments. Double staining for CD68 with GPX4 and ACSL4 showed reduced GPX4 and increased ACSL4 expression in CD68⁺ macrophages from periodontitis samples, indicating enhanced ferroptosis (Fig. 2C-D). To replicate the periodontal inflammatory milieu, we stimulated RAW264.7 macrophages with Porphyromonas gingivalis (P. gingivalis)-derived LPS to establish an in vitro inflammatory model. Ferroptosis activity was evaluated using intracellular iron probes and lipid peroxidation assays. Upon LPS stimulation, we observed a significant accumulation of oxidized phospholipids—hallmark executioners of ferroptosis—compared with controls (Fig. 2E). Concurrently, intracellular Fe²⁺ levels were monitored using the live-cell permeable probe FerroBrite™ Green, whose fluorescence intensity decreases in response to Fe²⁺ dysregulation during ferroptosis. LPS stimulation consistently led to a pronounced reduction in FerroBrite™ Green signal (Fig. 2F). CCK-8 assays further showed that LPS significantly reduced macrophage viability, which was partially rescued by the ferroptosis inhibitor Fer-1 (Fig. 2G). In parallel, LPS treatment suppressed the expression of the ferroptosis inhibitor Gpx4 but enhanced the expression of the ferroptosis promoter Acsl4 (Fig. 2H-I). Collectively, these findings demonstrate that P. gingivalis LPS robustly induces ferroptosis in macrophages, characterized by reduced cell viability, lipid peroxidation, iron dysregulation, and altered expression of key ferroptosis-related genes. Subsequently, to elucidate the biological mechanisms underlying macrophage ferroptosis, we stratified macrophages into subgroups with high and low ferroptosis scores. GSVA and GSEA were then carried out to compare transcriptional programs between the two groups. Pathway enrichment analyses based on Hallmark, GO, and KEGG databases revealed that ferroptosis-high macrophages were enriched in multiple pathways associated with inflammation, immune activation, and iron metabolism remodeling, including complement activation, IL6-JAK-STAT3 signaling, inflammatory response, regulation of leukocyte migration, NOD-like receptor signaling and iron ion homeostasis (Figure S4A-C). GSEA confirmed that ether lipid metabolism, cytokine–receptor interactions, and chemokine pathways were significantly elevated in ferroptosis-high macrophages, indicating strong immune activation coupled with metabolic reprogramming (Figure S4D-F). To further dissect the transcriptional regulation underlying ferroptosis heterogeneity, we performed TF activity ranking in low- and high-ferroptosis macrophages using SCENIC algorithm (Fig. 2J-K). Low-ferroptosis macrophages were characterized by higher activity of RFX7, PAX5, ZNF580, WRNIP1 and SETDB1. In contrast, high-ferroptosis macrophages showed markedly elevated activity of HIF1A, NR1H3, HIVEP2, CEBPB, and BRF1—transcription factors known to be involved in metabolic and inflammatory regulation [26–29]. This distinct TF landscape suggests a reinforced pro-inflammatory program in ferroptosis-high macrophages. Subsequently, to evaluate the functional outcome of this transcriptional program, we performed pseudotime trajectory analysis, and utilized the CytoTRACE package to define the root state. The results revealed a dynamic transition of macrophages with varying ferroptosis activity. Initially, macrophages displayed higher ferroptosis activity, which decreased as macrophages transitioned along the pseudotime axis (Figure S4G-H). Along this trajectory, we observed a gradual decrease expression of key inflammatory mediators, including IL-1β, CXCL3, CXCL8, CCL3, and CCL4, accompanied by a progressive decline in pathways related to leukocyte cell–cell adhesion, positive regulation of leukocyte activation, and cytokine production (Figure S4I-J). Consistent with this trajectory, ferroptosis-high macrophages exhibited significantly elevated expression of canonical M1-associated markers, including IL-1β, CCL3, CCL4, IRF1, and SOCS3, along with a higher M1 polarization score as quantified by AUCell analysis (Fig. 2L; Figure S4K). Metabolic profiling indicated that ferroptosis-high macrophages were more actively engaged in multiple metabolic pathways compared to their ferroptosis-low counterparts (Figure S4L). In summary, our integrated analysis demonstrates that macrophage ferroptosis in periodontitis is associated with a distinct pro-inflammatory transcriptional landscape, characterized by specific TF activation, a pronounced M1-like hyper-inflammatory state, and concomitant metabolic reprogramming, collectively amplifying the inflammatory microenvironment.

Fig. 2.

Transcriptomic and regulatory landscape of macrophages with distinct ferroptosis activity. (A) Dot plot showing the expression patterns of ferroptosis-related genes across distinct cell types. (B) Violin plot showing ferroptosis scores in macrophages from healthy and periodontitis samples. (C-D) Representative IF images of gingival tissues showing co-staining of CD68 with GPX4 and ACSL4 in control and periodontitis samples. (E) Lipid peroxidation in RAW264.7 macrophages assessed by BODIPY 581/591 C11 staining. (F) FerroBrite™ Green staining showing intracellular labile Fe²⁺ dysregulation and its quantification in RAW264.7 macrophages following stimulation with LPS. (G) Cell viability of RAW264.7 macrophages measured by CCK-8 assay following LPS stimulation with or without Fer-1 at indicated time points. (H-I) Relative mRNA expression of Gpx4 and Acsl4 in RAW264.7 macrophages following stimulation with increasing concentrations of P. gingivalis LPS using qPCR. (J-K) SCENIC analysis identifies TFs that are differentially activated between high- and low-ferroptosis macrophages. (L) Violin plots and AUCell scoring showing increased M1-associated gene expression (IL1β, IRF1, SOCS3) and elevated M1 polarization in ferroptosis-high macrophages. ns, not significant; *p < 0.05; **p < 0.01; ***p < 0.001

Single-cell and Spatial Transcriptomics Revealing the Interaction Between Ferroptotic Macrophages and Fibroblasts

To explore the potential role of ferroptosis in macrophage-mediated intercellular communication, we performed CellChat analysis comparing macrophage subsets with high and low ferroptosis activity. The results suggested that high-ferroptosis macrophages exhibit enhanced intercellular communication and increased interaction strength with multiple cell types compared with low-ferroptosis macrophages (Fig. 3A-C). Specifically, high-ferroptosis macrophages exhibited greater outgoing signals to epithelial cells, endothelial cells, fibroblasts, vascular endothelial cells, neutrophils, and mDCs (Figure S5A). Conversely, high-ferroptosis macrophages also received more signals from these cell populations (Figure S5B). Notably, within the stromal and immune compartments, fibroblasts emerged as the primary communication partners of high-ferroptosis macrophages. Further pathway analysis indicated that high-ferroptosis macrophages may preferentially signal to fibroblasts through the Galectin and OSM pathways, both of which are associated with inflammatory regulation (Fig. 3D-E). In turn, fibroblasts were predicted to signal back via IL-6 and chemokine-related pathways (Figure S5C-D). Among the Galectin family, Galectin-9 was selected for validation due to its prominent signaling activity in the ligand-receptor analysis (Figure S5E). ELISA assays showed that LPS stimulation significantly increased the secretion of OSM and Galectin-9, which was partially attenuated by the ferroptosis inhibitor Fer-1, suggesting that ferroptosis contributes to their production (Fig. 3F-G). To further examine their spatial relationship, we performed spatial deconvolution using cell2location to infer cell-type composition within spatial transcriptomic spots. The results suggested co-localization of high-ferroptosis macrophages and fibroblasts in periodontitis tissues (Fig. 3H-I). Consistently, mistyR analysis indicated spatially inferred interactions between these two cell types (Fig. 3J-K; Figure S5F-G). IF staining supported this finding, revealing enhanced co-localization of CD68-labeled macrophages and Vimentin-labeled fibroblasts under inflammatory conditions (Fig. 3L). To investigate whether ferroptotic macrophages promote fibroblast inflammation, a Transwell co‑culture system was established, with macrophages seeded in the upper chamber and L929 fibroblasts cultured in the lower chamber. CCK-8 assays and scratch wound-healing assays demonstrated that macrophages stimulated with P. gingivalis-derived LPS significantly suppressed the proliferative capacity and migratory ability of L929 fibroblasts. Notably, these inhibitory effects were partially reversed by treatment with the ferroptosis inhibitor ferrostatin-1 (Fig. 3M-N). Furthermore, qPCR analysis revealed that LPS‑stimulated macrophages markedly upregulated the expression of various inflammatory factors and chemokines in L929 fibroblasts, including Cxcl1, Cxcl2, Ccl2, Il1b and Il6 (Fig. 3O). ELISA further confirmed that LPS‑treated macrophages induced elevated secretion of CXCL1 and IL‑1β from L929 cells, an effect that was significantly attenuated by Ferrostatin‑1 (Fig. 3P-Q). To further validate whether macrophage ferroptosis directly regulates fibroblast function independent of inflammatory stimuli, macrophages were treated with the ferroptosis inducer Erastin and subsequently co-cultured with L929 fibroblasts using a Transwell system. Compared with control macrophages, Erastin-treated macrophages significantly impaired the proliferative capacity and migratory ability of L929 fibroblasts, as demonstrated by CCK-8 and wound-healing assays (Figure S6A-B). Moreover, qPCR analysis revealed that Erastin-treated macrophages increased the expression of pro-inflammatory cytokines and chemokines in L929 fibroblasts, including Cxcl1, Cxcl2, Il1b and Il6 (Figure S6C). Consistently, ELISA confirmed elevated secretion of CXCL1 and IL‑1β in fibroblasts co-cultured with Erastin-treated macrophages (Figure S6D-E). Together, these findings, together with the LPS-based experiments, demonstrate that macrophage ferroptosis promotes fibroblast dysfunction and amplifies inflammatory responses, whereas inhibition of ferroptosis alleviates fibroblast impairment and reduces their pro‑inflammatory activity.

Fig. 3.

Macrophage ferroptosis exacerbates fibroblast dysfunction and inflammatory activation. (A) Scatter plot illustrating the overall incoming and outgoing interaction strengths of each cell population inferred by CellChat. (B-C) Heatmaps summarizing the number and overall strength of intercellular interactions among cell types, highlighting distinct communication patterns between high- and low-ferroptosis macrophages. (D-E) Pathway-specific communication networks showing enhanced Galectin, OSM signaling in the high-ferroptosis macrophage group. (F-G) ELISA assay showing OSM and Galectin-9 levels in macrophage supernatants under control, LPS, and LPS + Fer-1 conditions. (H-I) Spatial mapping showing the spatial association between high-ferroptosis macrophages and fibroblasts in gingival tissues from GSM6258257 and GSM6258256. (J-K) mistyR analysis indicating spatially inferred communication interactions between these cell types in both samples. (L) IF staining of the macrophage marker CD68 and the fibroblast marker vimentin in gingival samples. (M) CCK-8 assay assessing the proliferation of fibroblasts co-cultured with untreated macrophages (NC), LPS-stimulated macrophages (LPS), or LPS-stimulated macrophages treated with the ferroptosis inhibitor Ferrostatin-1 (LPS + Fer-1). (N) Representative images and quantification of scratch wound-healing assays showing fibroblast migration following co-culture with macrophages under the indicated conditions. Dashed lines denote wound margins at 0 h and 24 h. (O) Relative mRNA expression of inflammatory cytokines and chemokines (Cxcl1, Cxcl2, Ccl2, Il1b and Il6) in fibroblasts after co-culture with NC, LPS, or LPS + Fer-1–treated macrophages, as determined by qPCR. (P-Q) ELISA analysis of CXCL1 and IL-1β secretion by fibroblasts following co-culture with macrophages under the indicated conditions. *p < 0.05; **p < 0.01; ***p < 0.001

Identification of Ferroptosis-associated Hub Genes in Macrophages Using hdWGCNA

To identify key regulatory networks associated with macrophage ferroptosis, we applied hdWGCNA to construct a single-cell gene co-expression network based on macrophage subsets and to detect modules correlated with ferroptosis activity. A total of 3,610 macrophages were included for network construction. Scale-free network topology was ensured by choosing a soft-thresholding power of 7 (Fig. 4A). Hierarchical clustering of genes yielded 11 co-expression modules, each represented by a distinct color (Fig. 4B). The yellow and blue modules showed the largest positive correlations with ferroptosis scores (correlation coefficients = 0.61 and 0.63, respectively; Fig. 4C), according to module-trait correlation analysis. To further evaluate the robustness of the identified modules, module preservation analysis was performed using both Zsummary and medianRank statistics. Consistent results from these two metrics indicated that the yellow and blue modules showed high preservation and stability, supporting the robustness of the constructed gene co-expression network (Figure S7A-B). The top ten hub genes within these two modules, along with their intra‑modular protein–protein interaction networks, are displayed in Figure S7C and Figure S7D-E, respectively, while correlations among all modules are shown in Figure S7F. Together, these results highlight the yellow and blue co‑expression modules as core transcriptional programs linked to ferroptosis in macrophages, providing a molecular framework for subsequent ferroptosis‑based subtyping and functional studies in periodontitis.

Fig. 4.

Identification of ferroptosis‑associated macrophage gene modules and molecular subtypes in periodontitis. (A) Determination of the optimal soft-thresholding power for constructing a scale-free co-expression network. (B) Hierarchical clustering dendrogram of macrophage genes with color-coded module assignments based on topological overlap. (C) Correlation heatmap showing the associations between hdWGCNA modules and ferroptosis scores. (D) Consensus clustering of bulk RNA-seq samples based on ferroptosis-related module genes identified by hdWGCNA. (E) CDF curve used to determine the optimal number of clusters (k = 2). (F) The delta area plot was used to determine the optimal number of clusters for ferroptosis-related core gene subclusters, identifying k = 2 as the optimal solution. (G-I) GSVA based on the HALLMARK, GO, and KEGG databases revealing distinct biological pathways enriched between ferroptosis subtypes. (J-L) GSEA plots highlighting key pathways upregulated in each ferroptosis subtype. (M-N) Comparison of immune cell infiltration levels (M) and immune function scores (N) between the two ferroptosis subtypes. *p < 0.05; **p < 0.01; ***p < 0.001

Ferroptosis-related Clustering Identifies Distinct Inflammatory and Immune Phenotypes in Periodontitis

To further investigate the heterogeneity of ferroptosis‑related molecular features in periodontitis, we performed consensus clustering using the expression profiles of the top 20 genes from the yellow and blue modules. Evaluation of the CDF curve and delta area plot supported k = 2 as the optimal solution, yielding two distinct subtypes including Cluster 1 and Cluster 2 (Fig. 4D-F). GSVA enrichment analysis based on Hallmark, GO and KEGG gene sets revealed striking biological differences between these subtypes. Cluster 2 exhibited strong enrichment of immune- and inflammation-related pathways, including interferon-α/γ responses, IL6-JAK-STAT3 signaling, lymphocyte chemotaxis, and chemokine signaling, consistent with a highly activated inflammatory state. In contrast, Cluster 1 was preferentially enriched in oxidative phosphorylation, glycolysis, protein secretion, and other metabolic homeostasis pathways, reflecting a comparatively metabolically balanced phenotype (Fig. 4G-I). GSEA further confirmed that Cluster 2 exhibited significant activation of leukocyte migration, lymphocyte chemotaxis, chemokine signaling, Toll-like receptor signaling, and interferon signaling pathways (Fig. 4J-L). Immune cell infiltration analysis indicated that Cluster 2 harbored higher levels of activated CD8⁺ T cells, macrophages, effector memory CD4⁺ T cells, and central memory CD4⁺ T cells (Fig. 4M). In addition, ssGSEA of immune-related functional signatures revealed that Cluster 2 exhibited elevated levels of chemokine receptor (CCR) signaling, inflammation-promoting functions, cytolytic activity, MHC class I expression, parainflammation, and type I interferon responses (Fig. 4N), reflecting a markedly enhanced inflammatory and immune-activation profile. Collectively, these findings demonstrate that ferroptosis-related molecular subtypes define distinct immune–metabolic states in periodontitis, which may differentially influence local inflammatory balance and tissue homeostasis.

Machine Learning Identifies LAPTM5 and AHR as Ferroptosis-associated Hub Genes

To identify ferroptosis-related key genes in macrophages, we performed differential expression analysis between macrophage populations with high and low ferroptosis activity. A total of 379 differentially expressed genes were identified using a threshold of |log₂FC|> 0.25 and adjusted p < 0.05 (Table S4). By intersecting these genes with 40 ferroptosis-associated genes previously identified through hdWGCNA, we obtained 13 overlapping genes for further machine learning-based feature selection (Figure S8). To pinpoint the most relevant ferroptosis-related regulators, we applied both LASSO regression and RF modeling. RF analysis ranked gene importance and highlighted two genes with importance scores > 10 (Fig. 5A; Figure S9A), while LASSO regression identified 10 candidate genes strongly associated with ferroptosis (Fig. 5B; Figure S9B). The intersection of results from both methods consistently identified LAPTM5 and AHR as the most robust hub genes related to ferroptosis in macrophages (Fig. 5C). To evaluate their expression patterns and diagnostic value, we analyzed the training dataset GSE16134 and an independent validation cohort comprising two bulk RNA-seq datasets (GSE173078 and GSE223924). In both datasets, LAPTM5 and AHR were significantly upregulated in periodontitis gingiva compared with healthy controls (p < 0.001) (Fig. 5D, Figure S9D-E). ROC analysis demonstrated strong diagnostic performance for LAPTM5 (AUC = 0.838) and moderate performance for AHR (AUC = 0.688) in the training set (Figure S9C), with similar diagnostic potential observed in the validation cohort (Figure S9F-G). qPCR validation in clinical samples further confirmed their increased expression in diseased tissues (Fig. 5E). Next, GSEA was used to investigate functional pathways connected to these two genes in order to further clarify their biological activities. LAPTM5 expression was positively correlated with pathways involved in chemokine signaling, leukocyte transendothelial migration, and cell adhesion molecules (CAMs), suggesting a role in inflammatory cell recruitment and immune activation. In contrast, AHR was enriched in cell cycle regulation, P53 signaling, and NOD-like receptor pathways (Fig. 5F). We next assessed their correlations with immune cell infiltration. LAPTM5 exhibited broad positive correlations with most immune cell types, suggesting a pervasive pro-inflammatory role. In contrast, AHR showed specific correlations with macrophages, plasma B cells, Th2 cells, and activated CD4⁺ T cells (Fig. 5G). We further assessed the correlations of LAPTM5 and AHR with core ferroptosis regulators including GPX4 [30], ACSL4 [31], HMOX1 [32], LPCAT3 [33], and NCOA4 [34]. LAPTM5 expression exhibited positive correlations with pro-ferroptotic genes ACSL4, LPCAT3, and HMOX1. Similarly, AHR showed a positive correlation with ACSL4 and NCOA4 but a negative correlation with the anti-ferroptotic factor GPX4. These coordinated expression patterns suggest that both LAPTM5 and AHR may contribute to a cellular state permissive to ferroptosis (Fig. 5H-I). Subsequently, we investigated the cellular localization of these two genes within gingival tissues. Single-cell transcriptomic analysis revealed that LAPTM5 was predominantly expressed in macrophages and neutrophils (Fig. 5J; Figure S9H), while AHR was mainly found in macrophages, fibroblasts, vascular murals, melanocytes and endothelial cells (Fig. 5K; Figure S9I). Notably, both genes showed markedly elevated expression in macrophages from periodontitis lesions compared to healthy tissues (Fig. 5L). Spatial transcriptome mapping further confirmed regional co-localization of LAPTM5 and AHR with multiple ferroptosis-related genes (Fig. 5M). Collectively, our comprehensive analysis identifies LAPTM5 and AHR as core ferroptosis-related regulators that may contribute to periodontitis pathogenesis by modulating immune responses and facilitating the ferroptosis process.

Fig. 5.

Identification and validation of ferroptosis-related hub genes in periodontitis. (A-B) RF and LASSO regression analyses identifying key ferroptosis-associated genes in periodontitis. (C) Venn diagram showing the intersection of genes obtained from RF and LASSO analyses, yielding two hub candidates, LAPTM5 and AHR. (D) Boxplots showing mRNA expression of LAPTM5 and AHR in periodontitis tissues compared with healthy controls in the training dataset. (E) qPCR analysis revealing the expression levels of LAPTM5 and AHR in clinically collected healthy (n = 8) and periodontitis (n = 8) gingival samples. (F) GSEA identifying pathways significantly enriched in relation to LAPTM5 and AHR expression. (G) Heatmap showing the association between LAPTM5 expression and immune cell infiltration patterns. (H-I) Correlation analysis illustrating relationships between LAPTM5 or AHR and well-established ferroptosis-regulatory genes documented in previous studies. (J-K) UMAP plots visualizing the expression and cellular localization of LAPTM5 and AHR across major gingival cell populations. (L) scRNA-seq analysis revealed the expression patterns of LAPTM5 and AHR in macrophages derived from healthy gingiva and periodontal inflammatory lesions. (S) Spatial transcriptomic mapping confirming the co-localization of LAPTM5 and AHR with ferroptosis-associated genes (GPX4, ACSL4, SLC7A11) in gingival tissues. *p < 0.05; **p < 0.01; ***p < 0.001

LAPTM5⁺ Macrophages Drive Pro-inflammatory Fibroblast Crosstalk

The above findings show that LAPTM5 demonstrates a more concentrated, highly enriched profile within macrophages than AHR. Therefore, we prioritized LAPTM5 for further experimental validation to elucidate its specific function in macrophage ferroptosis. To elucidate the intercellular communication network regulated by LAPTM5 in macrophages, we classified macrophages into LAPTM5⁺ and LAPTM5⁻ subsets according to their LAPTM5 expression and analyzed their communication profiles using CellChat. Compared with LAPTM5⁻ macrophages, LAPTM5⁺ macrophages exhibited markedly greater numbers and intensities of intercellular interactions, highlighting their enhanced regulatory influence within the inflammatory microenvironment of periodontitis (Fig. 6A). We then examined the specific communication patterns between LAPTM5⁺ macrophages and other cell populations. The results revealed increased interaction strength and frequency between LAPTM5⁺ macrophages and fibroblasts, epithelial cells, and endothelial cells, with the most pronounced enhancement observed in macrophage-fibroblast crosstalk (Fig. 6B-E). This finding is consistent with our earlier observation of spatial co-localization between ferroptotic macrophages and fibroblasts, suggesting a functional coupling between these two cell types. Pathway-level analysis further revealed that LAPTM5⁺ macrophages may primarily signal to fibroblasts via the TNF, Galectin, and Tumour necrosis factor-like weak inducer of apoptosis (TWEAK) pathways (Figure S10A-C). TWEAK is a macrophage-expressed member of the TNF superfamily that can cause gingival fibroblasts to produce adhesion molecules and pro-inflammatory cytokines [35]. These results imply that LAPTM5 may enhance pro-inflammatory responses from fibroblasts by means of signaling mediated by macrophages. Conversely, we also observed that fibroblasts signal back to macrophages via complement and chemerin pathways (Figure S10D-E), the latter of which has been reported to promote M1 polarization while suppressing M2 activation [36, 37]. Spatial transcriptomic analysis further suggested co-localization between LAPTM5⁺ macrophages and fibroblast-enriched regions (Fig. 6F-G). Consistently, mistyR analysis across multiple spatial contexts (intra, juxta, and para) in both samples suggested relatively stronger spatially inferred interaction potentials between these two cell types (Fig. 6H-M). Collectively, our results establish LAPTM5 as a central hub in macrophages that drives crosstalk with fibroblasts through multiple inflammation-related pathways, which may contribute to sustained fibroblast activation and chronic inflammation in periodontitis.

Fig. 6.

LAPTM5⁺ macrophages enhance intercellular crosstalk with fibroblasts in periodontitis. (A) Scatter plot illustrating the overall incoming and outgoing interaction strengths of each cell population inferred by CellChat. (B-C) Heatmaps summarizing the number and overall strength of intercellular interactions among different cell types. (D-E) Circle plots depicting differential signaling networks between LAPTM5⁺ and LAPTM5⁻ macrophages. (F-G) Spatial transcriptomic mapping showing the spatial distribution of LAPTM5⁺ macrophages and fibroblasts in gingival tissues from GSM6258257 and GSM6258256. (H-J) mistyR analysis across three spatial contexts (intra, juxta, and para) in GSM6258257. (K-M) mistyR analysis across three spatial contexts (intra, juxta, and para) in GSM6258256

To confirm whether LAPTM5 mediates ferroptosis in macrophages, we first performed IF staining on gingival tissue samples. The results showed co-localization of LAPTM5 with the macrophage marker CD68 (Fig. 7A). In vitro, LPS stimulation of RAW264.7 macrophages led to a dose-dependent increase in LAPTM5 expression, further supporting its functional relevance in macrophages (Fig. 7B). To determine the role of LAPTM5 in ferroptosis, we knocked down LAPTM5 expression in macrophages using siRNA. Among the tested siRNAs, si-LAPTM5-1 demonstrated the highest knockdown efficiency by qPCR and was selected for subsequent experiments (Fig. 7C). In LAPTM5-knockdown macrophages, we observed an upregulation of Gpx4 mRNA alongside a decrease in Acsl4 expression (Fig. 7D-E). Consistent with this molecular profile, silencing LAPTM5 significantly attenuated lipid peroxidation (Fig. 7F) and intracellular labile Fe²⁺ dysregulation (Fig. 7G-H) in LPS-stimulated macrophages, indicating that LAPTM5 contributes to the induction of ferroptosis. Furthermore, when fibroblasts were co-cultured with LAPTM5-silenced macrophages, they exhibited enhanced proliferative (Fig. 7I) and migratory capacities (Fig. 7J-K). At the molecular level, qPCR analysis revealed downregulation in the mRNA expression of multiple cytokines and chemokines, including Cxcl1, Ccl2, Il1b and Il6 (Fig. 7L). Consistently, ELISA measurements of the fibroblast culture supernatant confirmed a reduction in the secretion of CXCL1 and IL‑1β following co‑culture with LAPTM5‑silenced macrophages (Fig. 7M-N).

Fig. 7.

LAPTM5 mediates macrophage ferroptosis and exacerbates fibroblast inflammation and dysfunction in murine cells. (A) Representative IF images of gingival tissues from healthy controls and periodontitis patients showing the co-localization of LAPTM5 with the macrophage marker CD68. (B) qPCR showing the relative Laptm5 mRNA expression in RAW264.7 macrophages following stimulation with increasing concentrations of LPS. (C) qPCR analysis of Laptm5 mRNA expression in RAW264.7 macrophages transfected with different LAPTM5 siRNAs, demonstrating knockdown efficiency. (D-E) Relative mRNA expression of Gpx4 and Acsl4 in RAW264.7 macrophages transfected with control siRNA (si-NC) or LAPTM5 siRNA (si-LAPTM5). (F) Representative images of lipid peroxidation in RAW264.7 macrophages transfected with si-NC or si-LAPTM5, as assessed by BODIPY 581/591 C11 staining. (G) Intracellular labile Fe²⁺ dysregulation detected by FerroBrite™ Green staining in LPS-stimulated RAW264.7 macrophages with or without LAPTM5 knockdown. (H) Quantification of relative FerroBrite™ Green fluorescence intensity. (I) CCK-8 assay showing the proliferative capacity of L929 fibroblasts co-cultured with RAW264.7 macrophages transfected with si-NC or si-LAPTM5. (J) Representative images of scratch wound-healing assays of L929 fibroblasts co-cultured with si-NC– or si-LAPTM5–treated RAW264.7 macrophages at 0 h and 24 h. (K) Quantification of wound closure percentage in the scratch assay. (L) qPCR analysis of inflammatory cytokine and chemokine mRNA expression (Cxcl1, Cxcl2, Ccl2, Il1b and Il6) in L929 fibroblasts co-cultured with si-NC- or si-LAPTM5-treated RAW264.7 macrophages. (M–N) ELISA quantification of CXCL1 and IL-1β levels in the supernatants of L929 fibroblasts following co-culture with RAW264.7 macrophages transfected with si-NC or si-LAPTM5. ns, not significant; *p < 0.05; **p < 0.01; ***p < 0.001

To validate these findings in a human context, peripheral blood was collected from healthy donors, and monocytes were isolated using a monocyte isolation kit and subsequently differentiated into macrophages with M-CSF (Fig. 8A). HGFs were used as the fibroblast model. LAPTM5 expression was upregulated in human macrophages following LPS stimulation (Fig. 8B). Among the tested siRNAs, si-LAPTM5-1 exhibited the highest knockdown efficiency and was selected for subsequent experiments (Fig. 8C). Consistent with the murine results, LAPTM5-silenced in human macrophages increased GPX4 expression and decreased ACSL4 levels (Fig. 8D-E), accompanied by reduced intracellular Fe²⁺ accumulation and lipid peroxidation (Fig. 8F-G). Furthermore, co-culture with LAPTM5-silenced human macrophages enhanced the proliferative and migratory capacities of HGFs, while attenuating their inflammatory activation (Fig. 8H-I; Figure S11). Specifically, the expression and secretion of multiple pro-inflammatory cytokines and chemokines, including IL-1β, IL-6, IL-8, CXCL1, CXCL2, and CXCL6, were markedly reduced (Fig. 8I). Consistently, ELISA analysis confirmed decreased secretion of CXCL1 and IL-1β in HGFs (Fig. 8J-K). These findings further support a conserved role of LAPTM5 in regulating macrophage ferroptosis and fibroblast inflammatory responses.

Fig. 8.

Functional validation of LAPTM5 in regulating macrophage ferroptosis and fibroblast responses in human cells. (A) Schematic illustration of human monocyte-derived macrophage differentiation from peripheral blood. (B) Relative expression of LAPTM5 in human macrophages following LPS stimulation. (C) siRNA-mediated knockdown efficiency of LAPTM5. (D-E) Relative mRNA expression of GPX4 and ACSL4 in human macrophages transfected with si-NC or si-LAPTM5. (F) Intracellular labile Fe²⁺ levels detected by FerroBrite™ Green staining in human macrophages with or without LAPTM5 knockdown, along with quantification of fluorescence intensity. (G) Representative images of lipid peroxidation in human macrophages transfected with si-NC or si-LAPTM5, assessed by BODIPY 581/591 C11 staining. (H) CCK-8 assay showing the viability of human macrophages treated with LPS and si-LAPTM5. (I) qPCR analysis of inflammatory cytokine and chemokine mRNA expression (CXCL1, CXCL2, CXCL6, IL1β, IL6, and IL-8) in fibroblasts co-cultured with human macrophages transfected with si-NC or si-LAPTM5. (J-K) ELISA quantification of CXCL1 and IL-1β levels in fibroblast supernatants following co-culture with human macrophages. ns, not significant; *p < 0.05; **p < 0.01; ***p < 0.001

Discussion

Recently, PCD has gained increasing attention for its role in shaping immune responses and tissue remodeling in various chronic inflammatory diseases [38]. Systematic reviews have further highlighted the involvement of multiple PCD pathways—including ferroptosis, pyroptosis, and necroptosis—in the progression of periodontitis, highlighting their consistent association with inflammatory exacerbation, alveolar bone loss, and disease progression [39]. However, its specific involvement in the pathogenesis of periodontitis remains to be fully elucidated. In this study, we systematically evaluated the activity of multiple PCD pathways in periodontitis by integrating various gene set scoring methods and identified ferroptosis as one of the major PCD pathways associated with the disease. Mechanistically, we observed increased ferroptotic activity in macrophages, which in turn suppressed the proliferation and migration of fibroblasts while enhancing their inflammatory activation, collectively contributing to the amplification of inflammation and exacerbation of periodontal tissue destruction. Through integrated analysis using hdWGCNA and machine learning approaches, we identified LAPTM5 as a key regulatory gene governing macrophage ferroptosis. Functionally, inhibition of LAPTM5 in macrophages attenuated ferroptosis, restored fibroblast function, and suppressed the acquisition of a pro-inflammatory fibroblast phenotype, highlighting LAPTM5 as a promising therapeutic target for mitigating inflammation and tissue destruction in periodontitis.

Our results identify ferroptosis as one of the key PCD pathways involved in periodontitis, highlighting its previously underappreciated role in disease progression. Given its links to oxidative stress, lipid metabolism, and inflammatory amplification, ferroptosis may help explain hallmark features of periodontitis, including persistent oxidative stress, iron accumulation, and sustained immune activation. Supporting this notion, Torres et al. [40] found that ferroptosis coincides with clinical disease progression, characterized by increased ferroptosis inducers (SNCA, NMI, STAT1) and reduced suppressors (GCLC, HSPB1, CD44) in progressive periodontitis sites. Furthermore, we found that macrophages exhibited the highest ferroptosis activity among periodontal cells. Experimentally, LPS stimulation promoted lipid peroxidation in macrophages, accompanied by decreased GPX4 expression and increased ACSL4 expression. CCK8 assays further indicated that LPS reduced macrophage viability, whereas co-treatment with a ferroptosis inhibitor partially rescued this effect, confirming the induction of ferroptotic features. Notably, ferroptotic macrophages displayed elevated expression of inflammatory mediators with a more pronounced M1-like phenotype, in line with previous findings [41]. Through the process of ferroptosis, macrophages may release DAMPs, lipid peroxidation products, and pro-inflammatory mediators, thereby perpetuating oxidative stress and amplifying local inflammatory responses within the periodontal microenvironment. Therefore, targeting macrophage ferroptosis may represent a potential strategy to restore immune homeostasis and alleviate chronic inflammation within the periodontal microenvironment.

Another key finding is the crosstalk between ferroptotic macrophages and fibroblast dysfunction, revealing an immune-stromal axis in periodontitis. Our results show that ferroptotic macrophages can spatially co-localize with fibroblasts and may act as upstream drivers of fibroblast inflammatory activation. Pharmacological inhibition of macrophage ferroptosis attenuated this effect. Notably, studies have shown that fibroblasts themselves can also exhibit increased susceptibility to ferroptosis during periodontitis progression [40]. Consistently, our gene set scoring analysis revealed elevated ferroptosis activity in fibroblasts from periodontitis samples. These observations raise the possibility of a self-amplifying feed-forward loop: ferroptotic macrophages promote fibroblast inflammatory activation and dysfunction, which may in turn increase fibroblast susceptibility to ferroptosis, thereby amplifying inflammation and tissue damage. Receptor–ligand analysis identified Galectin and OSM as potential mediators linking ferroptotic macrophages to fibroblasts. ELISA confirmed that ferroptotic macrophages secreted higher levels of Galectin-9 and OSM, which was reversed by Fer-1. Galectin-9 is elevated in periodontal tissues of severe periodontitis patients, potentially linking to ferroptosis activity [42, 43]. Knockdown of Galectin-9 inhibits TNF-α–induced fibroblast inflammatory responses. Notably, ferroptotic cells have been reported to release Galectin-13, which binds to CD44 and inhibits SLC7A11 membrane localization in neighboring cells, thereby promoting ferroptosis propagation. OSM may represent another key pathway mediating macrophage–fibroblast crosstalk. In rheumatoid arthritis, macrophage‑derived OSM activates pathogenic synovial fibroblast subpopulations via the oncostatin M receptor (OSMR), thereby driving inflammation and joint destruction [44]. Similarly, in the tumor microenvironment, OSM from macrophages reprograms cancer‑associated fibroblasts (CAFs) toward a pro‑inflammatory, tumor‑promoting phenotype through the upregulation of inflammatory gene networks [45]. Together, these findings suggest that Galectin and OSM signaling may serve as key molecular links connecting ferroptotic macrophages to fibroblast activation, contributing to inflammatory amplification in periodontitis.

Through hdWGCNA analysis, we identified gene modules highly correlated with macrophage ferroptosis activity scores. The putative gene set was further honed through differential expression analysis between high- and low-ferroptosis groups. Integrating these results with LASSO and RF models highlighted LAPTM5 and AHR as the most stable ferroptosis-associated regulators in macrophages. These two genes are highly expressed in periodontitis tissues, show strong correlations with key ferroptosis regulators such as GPX4, ACSL4, and SLC7A11, and demonstrate regional spatial co-localization with them. LAPTM5 is a lysosome-localized transmembrane protein predominantly expressed in hematopoietic and immune cells. Early studies identified its pivotal role in inflammatory signaling within macrophages. Glowacka et al. demonstrated that LAPTM5 is essential for full activation of NF-κB and MAPK pathways downstream of Toll-like and TNF receptors, and that its deficiency markedly reduces IL-6 and TNF-α secretion [46]. Subsequent research has expanded the functional spectrum of LAPTM5 beyond classical inflammation. Sun et al. reported that LAPTM5 deficiency disrupts endosomal–lysosomal trafficking and autophagic flux, thereby attenuating macrophage activation and metabolic responsiveness [47]. More recently, Zhong et al. uncovered a regulatory link between LAPTM5 and ferroptosis in vascular pathology, showing that LAPTM5 cooperates with WDFY4 to modulate the CDC42/mTOR/4EBP1/SLC7A11 axis, promoting lipid peroxidation and ferroptotic cell death in endothelial cells [48]. Consistent with this concept, our single-cell and experimental data demonstrated that knockdown of LAPTM5 suppressed ferroptotic activity in macrophages, as evidenced by reduced lipid peroxidation and oxidative stress. Functionally, co-culture with LAPTM5-knockdown macrophages alleviated the impairment of fibroblast proliferation and migration induced by ferroptotic macrophages, and also attenuated their inflammatory responses. Overall, LAPTM5 appears to be a critical regulator of macrophage ferroptosis and inflammatory crosstalk with fibroblasts, contributing to periodontal tissue destruction, suggesting that LAPTM5 may serve as a potential therapeutic target for periodontitis.

The ligand-activated transcription factor known as the aryl hydrocarbon receptor (AHR) is essential for controlling redox homeostasis, immunological responses, and cellular metabolism. AHR directly controls the expression of SLC7A11, the cystine/glutamate antiporter essential for the glutathione-dependent antioxidant defense system and closely linked to ferroptosis. Kou et al. showed that by suppressing SLC7A11 expression and encouraging lipid peroxidation, pharmacological or genetic suppression of AHR dramatically increases erastin-induced ferroptosis [49]. Similarly, Liao et al. reported that the AHR-SLC7A11 signaling axis, which sustains the GSH/GPX4-dependent antioxidant defense, confers resistance to ferroptosis in epithelial and tumor cells, suggesting that AHR can act as a suppressor of ferroptotic death in specific cellular contexts [50]. Although direct evidence linking AHR to ferroptosis in macrophages remains limited, macrophage iron overload and lipid peroxidation are known to be major drivers of inflammation in diverse diseases. Therefore, AHR-mediated regulation of SLC7A11 may influence macrophage susceptibility to ferroptosis and, consequently, modulate inflammation within the periodontal microenvironment.

Despite these findings, several limitations should be acknowledged. First, our study demonstrated that ferroptotic macrophages promote the dysfunction and pro-inflammatory activation of gingival fibroblasts. However, the precise molecular mechanisms underlying this intercellular crosstalk remain to be fully elucidated. In particular, whether macrophage-derived lipid peroxidation products, cytokines, or exosomal factors directly trigger fibroblast inflammation requires further experimental validation. Second, while LAPTM5 was experimentally confirmed to enhance macrophage ferroptosis and subsequent fibroblast activation, the mechanistic pathways by which LAPTM5 regulates lysosomal iron metabolism and redox homeostasis were not explored in detail. Third, beyond mechanistic insights, future investigations into LAPTM5-targeted therapeutic compounds are warranted to enhance the translational and clinical potential of this target. Fourth, although AHR was identified as a ferroptosis-related candidate gene in our computational analysis, its functional role in macrophage ferroptosis was not explored in the present study. Therefore, further studies are warranted to elucidate the precise function and underlying mechanisms of AHR in this process. Finally, our in vitro models, including P. gingivalis LPS-stimulated macrophages and macrophage-fibroblast co-culture, represent simplified systems that cannot fully recapitulate the complex periodontal microenvironment, which involves polymicrobial biofilms, hypoxia, iron dysregulation, and diverse immune-stromal-microbial interactions. Future studies using more physiologically relevant models and in vivo experiments are needed to further validate our findings.

Conclusion

This study reveals ferroptosis as one of the important forms of PCD in periodontitis, primarily occurring in macrophages. Through integrated single-cell and spatial transcriptomic analyses, we demonstrate that ferroptotic macrophages are spatially positioned near fibroblast-enriched regions. Functionally, ferroptotic macrophages inhibit fibroblast proliferation and migration while amplifying their inflammatory activation. Machine-learning analysis identified LAPTM5 and AHR as hub genes associated with macrophage ferroptosis. Subsequent functional validation revealed that knockdown of LAPTM5 effectively suppressed ferroptosis in macrophages and alleviated inflammatory responses in fibroblasts. Collectively, these findings uncover a macrophage-fibroblast interaction axis driven by LAPTM5-mediated ferroptosis, providing new insight into the mechanisms of chronic periodontal inflammation and highlighting ferroptosis modulation as a potential therapeutic avenue for restoring immune–stromal homeostasis in periodontal disease.

Supplementary Information

Below is the link to the electronic supplementary material.

Author Contributions

Erli Wu and Peng Deng conceptualized and designed the study. Erli Wu, Xuan Yin, Yitong Chen, Jing Hu, Chongcheng Qiu, Chen Yang, Qian Chen, Rui Zhang, Peng Wang, Mengyuan Zhang, Misi Si, and Wei Shao performed the experiments and drafted the initial manuscript. Erli Wu and Qian Liu conducted the bioinformatics analyses and prepared the figures. Qianming Chen and Lijie Fan provided important intellectual input and technical support. Peng Deng critically reviewed and revised the manuscript. All authors reviewed and approved the final manuscript.

Funding

This work was supported by the National Natural Science Foundation of China (No. U25A6003, 82370918, 82330029, 82571774 and 82401100), Department of Science and Technology of Zhejiang Province (No. 2024C03193), and Key Research and Development Program of Zhejiang Province (No. 2025C02100).

Data Availability

The datasets supporting the conclusions of this article are available in the GEO repository: the single-cell RNA sequencing datasets are available under accession numbers GSE164241, GSE152042, and GSE161267; the spatial transcriptomics dataset is available under accession number GSE206621; and the bulk RNA-seq datasets are available under accession numbers GSE16134, GSE173078 and GSE223924.

Declarations

Ethics Approval and Consent to Participate

This study was performed in accordance with the principles of the Declaration of Helsinki. Ethical approval was obtained from the Medical Ethics Committee of the Stomatology Hospital, Zhejiang University School of Medicine (Approval No. I20250130 and No. 2025-118 (R)) and the Ethics Committee of the College and Hospital of Stomatology, Anhui Medical University (Approval No. 83240163). All participants provided informed consent prior to sample collection.

Consent for Publication

Not applicable.

Clinical Trial Number

Not applicable.

Competing interests

The authors declare no competing interests.