Identification and characterization of fibroblast-related biomarkers and pro-inflammatory subpopulations in periodontitis by integrated transcriptomic and single-cell analysis

Abstract

Periodontitis is a common inflammatory disease that severely impairs oral health and may contribute to systemic complications such as cardiovascular diseases. Fibroblasts play a crucial role in the progression and repair processes of periodontitis. Exploring related biomarkers and pro-inflammatory fibroblast (PIF) subpopulations may offer potential therapeutic strategies for periodontitis. We integrated transcriptomic and single-cell RNA sequencing (scRNA-seq) data to identify hub fibroblast genes via Least Absolute Shrinkage and Selection Operator (LASSO) regression. Diagnostic models and nomograms were developed and validated based on these genes. Fibroblast molecular patterns were defined using ConsensusClusterPlus package, and pseudotime analysis was performed with Monocle package. Fibroblast-related genes (FRGs) and PIF subsets were confirmed by quantitative real-time PCR (qRT-PCR), immunohistochemistry (IHC), and fluorescence in situ hybridization (FISH). By integrating bulk transcriptomic and single-cell RNA sequencing data, six FRGs (XBP1, SELL, ST6GAL1, IGHG1, CD79A, and PIM2) were identified using LASSO regression to construct a diagnostic model. The area under the curve (AUC) values for the model were 0.894, 0.919, and 0.746 in the GSE10334, GSE16134, and GSE173078 datasets, respectively. A predictive nomogram was also developed and validated. Based on these FRGs, periodontitis samples were clustered into three groups. Enrichment analysis revealed involvement of classical inflammatory pathways such as KRAS/p53 and IL-6/JAK/STAT. Immune infiltration analysis showed that cluster 1 exhibited a pro-inflammatory microenvironment. CXCL13⁺ fibroblasts were significantly increased in periodontitis tissues. qRT-PCR confirmed significantly elevated expression of SELL, CD79A, and ST6GAL1 in periodontitis tissues. IHC demonstrated increased SELL expression in periodontitis samples, and FISH analysis revealed a marked increase of CXCL13⁺ fibroblasts in periodontitis tissues, consistent with the bioinformatics findings. This study not only confirms the critical role of fibroblasts in periodontitis but also identifies FRGs and a PIF subset, CXCL13⁺ fibroblasts. Our findings provide novel insights into the pathogenic mechanisms of fibroblasts in periodontitis and may inform the development of targeted diagnostic and therapeutic strategies.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-37385-2.

Introduction

Periodontitis is a prevalent chronic non-communicable disease, initiated by microbial dysbiosis and aggravated by dysregulated host immune responses. It can lead to irreversible tissue destruction, tooth loss, and may contribute to systemic complications such as cardiovascular diseases^1–3. In periodontal tissues, fibroblasts are the primary connective tissue cells responsible for maintaining tissue structure and homeostasis. Gingival and periodontal ligament fibroblasts play crucial roles in the pathogenesis of periodontitis by interacting with bacterial products and inflammatory cytokines, thereby participating in inflammation, tissue destruction, and repair processes⁴.

Within the periodontal microenvironment, fibroblasts respond to bacterial products and pro-inflammatory signals by secreting cytokines, chemokines, and matrix-degrading enzymes, promoting immune cell infiltration and connective tissue breakdown^5–7. PIF have been identified as key drivers in periodontitis progression, primarily through amplifying inflammatory responses and tissue damage. Specifically, periodontitis is a chronic inflammatory disease characterized by persistent inflammation of gingival tissues, where fibroblasts, as the main structural cells, actively contribute to pathology by secreting inflammatory mediators rather than serving merely as passive scaffolds. For instance, scRNA-seq studies have identified a PIF subpopulation that is significantly expanded in periodontitis patients (from 1.9% to 4.6%). These cells promote the release of inflammatory mediators such as IL-6 and CXCL10 by activating the TNF-α/NF-κB signaling pathway, thereby exacerbating periodontal inflammation and tissue damage⁸.

Moreover, the role of PIGFsextends beyond their intrinsic inflammatory effects; they interact with other immune cells through cell-cell communication. Communication analyses have revealed that PIGFs form interactive networks with monocytes, T cells, and other immune populations, enhancing the expression of lymphocyte-attracting chemokines such as CXCL8 and CXCL10^8,9. This leads to increased leukocyte infiltration and amplification of inflammation in periodontal tissues. These findings indicate that pro-inflammatory fibroblasts modulate the immune microenvironment and act as key pathological drivers in periodontitis. Functional experiments further demonstrated that knockdown of PIGF-associated genes, such as MME and TSPAN11, significantly inhibited fibroblast proliferation, migration, and inflammatory cytokine release, confirming these genes as potential therapeutic targets for periodontitis⁸. Therefore, identifying fibroblast-associated biomarkers and pro-inflammatory fibroblast subpopulations in periodontitis may provide valuable insights for future personalized therapeutic strategies.

In this study, we integrated publicly available bulk and scRNA-seq datasets to systematically investigate the role of fibroblasts in periodontitis. Our objectives were to identify fibroblast-related gene signatures, evaluate their diagnostic potential, and characterize fibroblast heterogeneity across different periodontal tissues. Notably, we identified a novel PIF subpopulation-CXCL13⁺ fibroblast. Our findings provide new insights into the pathogenic role of fibroblasts in periodontitis and may offer references for the development of targeted diagnostic and therapeutic strategies.

Materials and methods

Data source

Gene expression datasets were obtained from the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo/) using the GEOquery R package (version 2.70.0)¹⁰. Specifically, the datasets GSE10334¹¹, GSE16134¹², and GSE173078 were included for analysis. In addition, the scRNA-seq dataset GSE171213 was also retrieved from GEO. This dataset includes periodontal tissues of healthy humans (HC) as well as periodontitis patients with or without treatment(PD or PDT). The datasets included in this study were selected based on the following criteria: (1) relevance to the research question, (2) completeness of data without missing values, and (3) publication in peer-reviewed sources or broad recognition within the field. Exclusion criteria included outdated datasets or those with insufficient sample sizes to ensure statistical validity. These criteria were applied to ensure the robustness and reliability of our findings. Detailed clinical information is provided in Table S1.

Estimation of fibroblast infiltration abundance

scRNA-seq data were processed using the Seurat R package (version 5.0.1)¹³. Cell clustering was performed using the FindClusters function, and the optimal resolution (0.5) was determined using the clustree function to categorize cells into distinct clusters. Dimensionality reduction was conducted using the RunTSNE function to visualize fibroblast proportions across tissues of different origins. Fibroblast infiltration abundance in the GSE10334, GSE16134, and GSE173078 datasets was quantified using the MCPcounter R package¹⁴. Based on the median fibroblast infiltration score, samples were stratified into high- and low-risk groups.

Screening of FRGs in periodontitis patients

Differentially expressed genes (DEGs) were identified in the GSE173078, GSE16134, and GSE10334 datasets using the limma package¹⁵, with the thresholds set at |log₂ fold change| ≥ 1.0 and P < 0.05. Volcano plots were generated to visualize the DEGs. For scRNA-seq data, the FindAllMarkers function was employed to identify marker genes of fibroblast subsets, applying the same cutoff criteria (|log₂fold change| ≥ 1.0 and P < 0.05). FRGs were then determined by intersecting the DEGs from the periodontitis datasets with the fibroblast marker genes using the VennDiagram package. Subsequently, LASSO regression analysis was performed on the intersecting genes using the glmnet package (version 4.1.7)¹⁶, with the number of iterations set to 10. The results of the LASSO regression were visualized using a prognostic risk model plot and a coefficient trajectory plot.

Construction of a diagnostic model based on FRGs and development and validation of a predictive nomogram

Based on the FRGs identified through LASSO regression, a multivariate logistic regression analysis was performed to construct a diagnostic model. The predictive performance of the model in both the training and testing sets was evaluated by plotting receiver operating characteristic (ROC) curves using the pROC package¹⁷. A predictive nomogram was then developed using the nomogramEx package, and a calibration curve was generated to assess the consistency between predicted probabilities and actual outcomes. To further evaluate the clinical utility and net benefit of the nomogram, decision curve analysis (DCA) was conducted.

Construction of fibroblast-related molecular patterns in periodontitis

To identify distinct molecular subtypes of periodontitis based on FRGs, consensus clustering was performed on the GSE10334 datasets using the ConsensusClusterPlus package¹⁸. The number of clusters was set between 2 and 9, with 50 iterations performed by randomly sampling 90% of the total samples in each iteration. The clustering algorithm was set to “km” (k-means), and the distance metric was defined as “euclidean.” DEGs between the identified clusters were determined using the limma package¹⁵, with cutoffs of |log2FC| ≥ 1.0 and P < 0.05. To investigate the functional characteristics of the DEGs, gene set enrichment analysis (GSEA) was conducted using the GSEABase package¹⁹. Enrichment results were used to elucidate the distinct biological processes and signaling pathways associated with each fibroblast-related disease subtypes.

Analysis of the immune infiltration microenvironment and effective immunotherapy response in different clusters

The proportions of infiltrating immune cells in different patterns of the GSE10334 dataset were estimated using multiple deconvolution algorithms, including CIBERSORT, EPIC, ESTIMATE, MCP-counter, QUANTISEQ, and TIMER, via the IBOR package²⁰. A heatmap was generated to visualize the immune landscape across patterns. In addition, the CIBERSORT algorithm was applied independently to provide a detailed assessment of immune cell infiltration patterns. The Wilcoxon rank-sum test was used to evaluate differences in immune cell infiltration between the identified clusters.

Identification, clustering, and trajectory analysis of fibroblast subsets

Fibroblast subsets were extracted from the scRNA-seq data and clustered using the FindClusters function, as described previously. The resolution parameter (set to 0.1) for defining distinct cell clusters was determined using the clustree function. Dimensionality reduction was performed with the RunTSNE function to visualize the distribution and proportions of fibroblast populations across tissues of different origins. The AUCell package²¹ was applied to calculate gene set activity scores for each cell based on FRGs, enabling the assessment of functional heterogeneity among fibroblast subtypes. To further explore the potential differentiation trajectories of fibroblast subsets, pseudotime analysis was conducted using the monocle package²².

Tissue sample collection

A total of 28 healthy gingival tissue samples and 21 periodontitis tissue samples were collected from healthy volunteers and periodontitis patients, respectively, at Fujian Provincial Hospital. This study was approved by the Ethics Committee of Fujian Provincial Hospital (No. K2025-02-045). All research procedures were conducted in accordance with relevant guidelines and regulations, and written informed consent was obtained from all participants and/or their legal guardians prior to sample collection.

qRT-PCR

Total RNA was extracted from tissue samples using RNA extraction reagent (Servicebio, China, G3013). Complementary DNA (cDNA) was synthesized from total RNA using a reverse transcription kit (Servicebio, China, G3337) according to the manufacturer’s instructions. qRT-PCR was performed using a real-time fluorescence PCR detection system (Bio-Rad, Hercules, CA, USA). Relative gene expression levels were calculated using the 2^−ΔΔCT method, with GAPDH serving as the internal control. All experiments were conducted in triplicate. Primer sequences are listed in Table 2.

IHC

Paraffin sections were dewaxed and rehydrated using an eco-friendly dewaxing solution followed by a graded ethanol series. Antigen retrieval was performed in EDTA buffer (pH 9.0) using microwave heating, followed by PBS rinsing. Endogenous peroxidase activity was blocked by incubation with 3% H₂O₂ at room temperature, after which sections were blocked with 3% BSA. Primary antibody incubation was performed overnight at 4 °C with the following antibodies: anti-SELL (1:100, Servicebio, China, GB11632), anti-IGHG1 (1:100, Thermo Fisher Scientific, USA, AB_2719156), anti-XBP1U (1:200, Proteintech Group, China, 25997-1-AP), anti-ST6AL1 (1:500, Proteintech Group, China, 14355-1-AP), anti-PIM2 (1:500, ImmunoWay Biotechnology, USA, YT3729), and anti-CD79A (1:200, Proteintech Group, China, 22349-1-AP). After PBS washing, sections were incubated with the appropriate secondary antibody at room temperature, followed by DAB chromogenic development and hematoxylin counterstaining. Sections were dehydrated and cleared through graded ethanol and xylene, then mounted. Under bright-field microscopy, positive staining appeared brownish-yellow, while cell nuclei were stained blue. IHC staining was quantified using the H-score method²³.

FISH

The CXCL13 probe was obtained from Wuhan Servicebio Technology Co., Ltd. (China), with the specific probe sequence provided in Table S3. FISH combined with immunofluorescence double-labeling was performed as follows: Tissues were rinsed in PBS, immediately fixed in in situ hybridization fixative for ≥ 12 h at 4 °C, and trimmed into ~ 3 mm-thick blocks. After dehydration, clearing, and paraffin embedding, 4 μm sections were prepared and baked, followed by dewaxing and rehydration. Antigen retrieval was performed at 90 °C for 40 min, followed by proteinase K digestion at 40 °C for 5 min. Pre-hybridization was carried out at 37 °C for 1 h, after which sections were hybridized overnight with the specific probe. Post-hybridization washing was performed using SSC buffer (Servicebio, China, G3015) at varying concentrations; if nonspecific binding was high, washing conditions were adjusted accordingly. Subsequent hybridization steps with branched and signal probes were performed at the appropriate temperatures, followed by washing. Nuclei were counterstained with DAPI, and sections were mounted using an antifade solution. Images were captured under a fluorescence microscope. Fluorescence signals of CXCL13⁺ fibroblasts were quantified using the Cell Counter and co-localization plugins in ImageJ. For each section, five randomly selected high-power fields were analyzed. Co-localization with fibroblast markers was confirmed using dual-channel fluorescence, and the counts were averaged for each sample to compare healthy and periodontitis tissues.

Statistical analysis

All data were processed using R software (Version 4.2.2) and GraphPad Prism 8. For two-group comparisons, normally distributed data were analyzed by Student’s t-test; non-normal data used the Mann-Whitney U test. For three or more groups, the Kruskal-Wallis test was applied. All tests were two-sided unless stated otherwise, with p < 0.05 considered significant.

Result

Fibroblast characteristics in periodontitis

We analyzed the scRNA-seq dataset GSE171213 using the Seurat package, identifying 17 distinct cell clusters, which were subsequently annotated into eight major cell types based on marker genes and a combined approach of SingleR-based automatic annotation and manual curation. These cell types included B cells, endothelial cells, epithelial cells, fibroblasts, mast cells, monocytes, neutrophils, and T/NK cells (Fig. 1A). Among these, fibroblasts were characterized by high expression of signature genes such as LUM, FGF7, and COL1A1 (Fig. 1B). Comparative analysis of cell-type proportions across different tissue samples revealed a marked reduction in fibroblast subpopulation proportions in PDT, suggesting a potential role for fibroblast modulation in therapeutic response (Fig. 1C). To further quantify fibroblast abundance, we applied the MCP-counter algorithm to three independent transcriptomic datasets (GSE10334, GSE16134, and GSE173078). Across all datasets, fibroblast infiltration levels were significantly elevated in periodontitis tissues compared to healthy controls. Moreover, samples stratified by high fibroblast infiltration abundance were associated with a higher proportion of periodontitis patients, further supporting a link between fibroblast activity and periodontitis (Fig. 1D-E).

Fig. 1.

(A) t-SNE plots showing single-cell clustering and cell-type annotation results; (B) t-SNE plots displaying marker genes of fibroblast subpopulations; (C) Pie chart showing the proportion of different cell subpopulations and bar plot illustrating their distribution across tissue sources (HC vs. PD vs. PDT); (D–F) Boxplots and stacked bar charts showing fibroblast infiltration in healthy gingival tissues and periodontitis tissues in the GSE10334 (D), GSE16134 (E), and GSE173078 (F) datasets.

Identification of fibroblast hub genes

First, we utilized the limma R package to analyze differential gene expression between periodontitis and healthy periodontal tissues across three datasets: GSE173078, GSE16134, and GSE10334. This analysis identified 526, 149, and 192 DEGs, respectively. Subsequently, the FindAllMarkers function was applied to the GSE171213 dataset to identify 752 marker genes specific to fibroblast subsets, using the same threshold criteria (|log₂ fold change| ≥ 1.0 and P < 0.05). By intersecting the DEGs from the periodontitis datasets with the fibroblast marker genes, we identified seven fibroblast-associated genes closely linked to periodontitis: XBP1, SELL, ST6GAL1, ADA2, IGHG1, CD79A, and PIM2. These seven candidate genes were subjected to LASSO regression analysis to further evaluate their diagnostic relevance in the GSE10334 dataset. A LASSO regression model was constructed, and both the LASSO coefficient trajectory plot (Fig. 2E) and the optimal lambda model plot (Fig. 2F) were generated for visualization. The final model retained six FRGs: XBP1, SELL, ST6GAL1, IGHG1, CD79A, and PIM2. The expression profiles of these six genes across different periodontitis datasets are visualized in the heatmap shown in Figure S1, highlighting their consistent differential expression. These six genes were ultimately identified as hub fibroblast-associated feature genes in periodontitis. To evaluate their diagnostic potential, we calculated the area under the ROC curve for each dataset. The six-gene signature achieved AUC values of 0.894 in GSE10334, 0.919 in GSE16134, and 0.746 in GSE173078, indicating robust diagnostic performance across datasets.

Fig. 2.

(A–C) Volcano plots showing DEGs between healthy gingival tissues and periodontitis tissues in the GSE10334 (A), GSE16134 (B), and GSE173078 (C) datasets. (D) Venn diagram illustrating the overlap between DEGs from GSE10334, GSE16134, and GSE173078, and fibroblast subpopulation DEGs from scRNA-seq dataset. Plots of prognostic risk models (E) and variable trajectories (F) from the LASSO regression model.

Nomogram construction and validation

A nomogram incorporating the six fibroblast hub genes was constructed to predict the risk of periodontitis progression in GSE10334, GSE16134, and GSE173078 datasets (Fig. 3A, D, G). The calibration curves of the nomogram for the three datasets are shown in Fig. 3B, E, and H. To further evaluate the clinical utility of the nomogram, DCA was conducted (Fig. 3C, F, I). In the GSE10334 and GSE16134 datasets, the nomogram model demonstrated consistent advantages in identifying patients at risk of disease deterioration. Specifically, in both GSE10334 and GSE16134, the DCA curves indicated that when the threshold probability for disease progression ranged from 20% to 78%, the nomogram provided greater net clinical benefit and could effectively assist in clinical decision-making.

Fig. 3.

(A–C) ROC curves of the LASSO model in the GSE10334, GSE16134, and GSE173078 datasets. (D) Diagnostic nomogram based on fibroblast-related hub genes in the GSE10334 dataset, with corresponding (E) calibration curve and (F) DCA.

Three distinct clusters were identified based on the expression patterns of FRGs

Subsequently, an unsupervised consensus clustering approach was employed to classify samples in the GSE10334 dataset into distinct fibroblast-related patterns based on the expression levels of the FRGs. Consensus matrix analysis indicated that k = 3 was the optimal number of clusters, with strong intra-cluster correlations observed for each sample (Fig. 4A–C). A heatmap illustrated the expression profiles of the six diagnostic FRGs across the three clusters, showing that these genes were highly expressed in the C2 cluster (Fig. 4D). Volcano plots displaying the DEGs among the clusters are shown in Fig. 4E, G, and I. To further explore the molecular mechanisms underlying fibroblast heterogeneity in periodontitis samples, we performed GSEA on the DEGs between clusters. For the comparison between C1 and C2 (Fig. 4F), significant enrichment was observed in pathways associated with hormonal and environmental stress responses, such as estrogen response, UV response, and the KRAS/p53 signaling pathway. In the comparison between C2 and C3 (Fig. 4J), DEGs were predominantly enriched in pathways including the unfolded protein response, IL-6/JAK/STAT signaling, and classic inflammatory pathways. For the C1 vs. C3 comparison (Fig. 4H), enrichment was mainly detected in pathways related to the cell cycle, myeloid cell proliferation, and epithelial–mesenchymal transition (EMT).

Fig. 4.

Identification of fibroblast-related clusters in periodontitis. Consensus clustering in the GSE10334 dataset based on FRGs. Three distinct clusters were identified according to the (A) consensus matrix, (B) CDF curve, and (C) CDF delta area curve for k = 3, with the index increased from 2 to 9. (D) Heatmap displaying the expression profiles of the six diagnostic FRGs across the three clusters. Volcano plots and GSEA results are shown for comparisons between cluster 1 and cluster 2 (E, F), cluster 1 and cluster 3 (G, H), and cluster 2 and cluster 3 (I, J).

Immune landscape characterization across fibroblast-related clusters

To elucidate the immune landscape associated with fibroblast-related molecular subtypes, we performed a comprehensive immune infiltration analysis in the GSE10334 cohort using the IOBR R package. As shown in Fig. 5A, there were significant differences in immune cell infiltration patterns among the three clusters. The C1 cluster was characterized by a pro-inflammatory immune microenvironment. Specifically, it exhibited elevated infiltration of T cell subsets (e.g., CD8 + T cells, follicular helper T cells) and myeloid-derived immune cells (e.g., M1 macrophages, dendritic cells), along with an increased immuneScore, indicating an active cytotoxic immune profile. In contrast, the C2 cluster demonstrated an immunosuppressive microenvironment, as evidenced by increased infiltration of regulatory T cells (Tregs) and M2 macrophages, as well as an elevated StromalScore, suggesting immune evasion and stromal activation. The C3 cluster displayed a relatively quiescent or immune-balanced phenotype. It showed lower infiltration of effector T cells (e.g., CD8 + T cells, Th1 cells) compared to C1, and reduced levels of immunosuppressive cells (e.g., Tregs, Th2 cells) compared to C2. Both ImmuneScore and StromalScore in C3 were intermediate between those of C1 and C2. Notably, C3 was enriched in resting or memory T cell subsets, such as memory CD4 + T cells, suggesting a non-inflammatory, homeostatic immune microenvironment characterized by resting immune populations, intermediate-polarized myeloid cells, and immunological equilibrium. Further analysis using the CIBERSORT algorithm confirmed the heterogeneity in immune cell composition across clusters. Significant differences were observed in the infiltration levels of multiple immune cell types—including follicular helper T cells, macrophages, and dendritic cells—highlighting the diverse immune microenvironments among the three fibroblast-associated clusters. Violin plot analysis further supported these findings (Fig. 5B). Treg cell abundance was significantly higher in the C2 cluster compared to C1 and C3, whereas C1 showed the lowest Treg infiltration (Fig. 5C). M1 macrophage levels were highest in C1 and lowest in C2 (Fig. 5D), aligning with their respective pro- and anti-inflammatory profiles. CD8 + T cell infiltration was markedly elevated in C1 but diminished in both C2 and C3 (Fig. 5E). Likewise, activated dendritic cells (DCs) were most abundant in C1 and least in C2 (Fig. 5F).

Fig. 5.

(A) Heatmap showing immune infiltration patterns across different subtypes in the GSE10334 cohort, as estimated by multiple algorithms. (B) Immune cell abundance of 22 cell types in different patterns, calculated using the CIBERSORT algorithm. Significant differences were observed among patterns for (C) T cells regulator, (D) Macrophages M1, (E) T cells CD8⁺, and (F) Dendritic cells activated.

Functional heterogeneity and dynamic transitions of fibroblast subpopulations in periodontitis

To further elucidate the role of fibroblasts in the development and progression of periodontitis, we performed an in-depth analysis of fibroblast subpopulations in the GSE171213 single-cell RNA sequencing dataset. Using the t-distributed stochastic neighbor embedding (t-SNE) algorithm for dimensionality reduction and visualization, we identified five distinct fibroblast clusters at a resolution of 0.1. Based on the differential expression of characteristic marker genes, these clusters were annotated as five fibroblast subtypes: CD34⁺ fibroblasts, COL6A2⁺ fibroblasts, CXCL13⁺ fibroblasts, IGHG2⁺ fibroblasts, and RGS5⁺ fibroblasts (Fig. 6A). Functional enrichment analysis revealed substantial functional heterogeneity among these subpopulations. CD34⁺ fibroblasts were significantly enriched in pathways related to muscle contraction, vascular development, and cardiac contraction regulation. COL6A2⁺ fibroblasts were enriched in leukocyte chemotaxis and cell migration pathways. CXCL13⁺ fibroblasts exhibited functions associated with B cell receptor signaling, complement activation (immune regulation), and extracellular matrix organization (matrix remodeling) (Fig. 6B). We next performed pseudotime trajectory analysis to explore the developmental relationships among fibroblast subtypes. The inferred trajectory suggested that CXCL13⁺ fibroblasts are positioned near the beginning of the pseudotime path, while RGS5⁺ fibroblasts occupy the terminal end. This positioning indicates a potential early-activated or precursor-like state for CXCL13⁺ fibroblasts (Fig. 6C). To assess the functional activity of each fibroblast subtype, we utilized the AUCell R package to calculate activity scores based on the previously identified six fibroblast-associated gene signatures. We observed distinct fibroblast subtype compositions depending on tissue condition, with significant variability in AUC activity scores (Fig. 6D–F). IGHG2⁺ fibroblasts were predominantly present in healthy periodontal tissues. CXCL13⁺ fibroblasts were significantly increased in periodontitis tissues. Following treatment, COL6A2⁺ and RGS5⁺ fibroblasts became the dominant subpopulations. Furthermore, fibroblast activity scores were markedly lower in periodontitis tissues compared to both healthy and post-treatment tissues (Fig. 6G). Among all subtypes, IGHG2⁺ fibroblasts exhibited the highest fibroblast activity score, indicating their potentially critical role in maintaining periodontal homeostasis (Fig. 6H).

Fig. 6.

(A) t-SNE plots showing five fibroblast clusters and their corresponding subtypes. (B) Network diagram of functional enrichment analysis for each fibroblast subtype. (C) Developmental trajectory of all fibroblast subtypes. (D) Pseudotime ordering of fibroblast differentiation. (E) t-SNE plots illustrating the distribution of fibroblast subtypes across different tissue origins (HC vs. PD vs. PDT). (F) t-SNE plots showing AUC scores, based on fibroblast-related hub genes, across fibroblast subtypes from different tissue origins (HC vs. PD vs. PDT). (G) Bar plots displaying the proportion of fibroblast subtypes in tissues from different origins (HC vs. PD vs. PDT). Violin plots illustrating the comparison of AUC scores, calculated based on fibroblast-related hub genes, across (H) different tissue origins and (I) different fibroblast subtypes.

Experimental verification of fibroblast-associated hub genes

To validate the bioinformatics findings, we further examined FRGs in clinical samples using qRT-PCR and IHC. Compared with healthy periodontal tissues, qRT-PCR analysis showed significantly higher expression levels of SELL, CD79A, and ST6GAL1 in periodontitis tissues, whereas IGHG1, PIM2, and XBP1 showed no significant differences (Fig. 7A). Consistently, IHC results revealed that only SELL expression was markedly elevated in the periodontitis group compared to healthy controls (Fig. 7B–D).

Fig. 7.

Expression analysis of FRGs in periodontitis and control tissues. (A) mRNA expression levels of FRGs in periodontitis versus control groups. (B) IHC of SELL. Histograms showing the proportion of positive cells (C) and H-score (D) comparisons between periodontitis and control groups. *p < 0.05; **p < 0.01; ***p < 0.001.

FISH validation of elevated CXCL13⁺ fibroblasts in periodontitis

Based on the analysis of fibroblast subpopulation proportions across different tissues, we observed a significant increase of CXCL13⁺ fibroblasts in periodontitis tissues. To validate this finding, FISH was performed on clinical samples to compare the expression of CXCL13⁺ fibroblasts between periodontitis and healthy tissues. Co-localization analysis demonstrated a markedly higher presence of CXCL13⁺ fibroblasts in periodontitis tissues compared to healthy controls (Fig. 8A-C).

Fig. 8.

Visualization of fibroblast and CXCL13⁺ cell distribution in healthy and periodontitis tissues. (A) CXCL13 FISH combined with vimentin immunofluorescence staining. Green signals indicate vimentin-positive fibroblasts, red signals represent CXCL13 RNA detected by FISH, and nuclei are stained blue with DAPI. (B) Quantitative analysis of the density of cells co-expressing vimentin and CXCL13 signals in healthy versus periodontitis tissues. (C) Proportion of cells double-positive for vimentin and CXCL13 in healthy and periodontitis tissues. *p < 0.05.

Discussion

Periodontitis is a prevalent inflammatory disease characterized by the destruction of the supporting structures of teeth, significantly impacting oral health and overall quality of life. This condition not only leads to tooth loss but also poses severe health risks, including associations with systemic diseases such as and cardiovascular conditions^2,24,25. The economic burden of periodontitis on healthcare systems is substantial, as it requires ongoing management and treatment efforts. Current treatment modalities, including scaling and root planing, surgical interventions, and antibiotic therapies, often fall short, resulting in incomplete resolution of inflammation and disease recurrence^26,27. Therefore, there is a pressing need for innovative diagnostic and therapeutic approaches to understand and manage periodontitis effectively.

Previous studies have established the roles of fibroblasts in tissue homeostasis and inflammation, highlighting their potential as biomarkers for disease progression^28,29. In this study, integrative bioinformatic analyses of gene expression identified six fibroblast-associated hub genes related to periodontitis, culminating in the development of a predictive diagnostic model. These findings may provide critical insights into the potential molecular mechanisms underlying periodontitis and open new avenues for therapeutic intervention. Further validation of these fibroblast-associated hub genes by qRT-PCR and IHC strengthened the clinical relevance of our results. Compared with healthy controls, genes such as SELL, CD79A, and ST6GAL1 were significantly upregulated in periodontitis tissues, supporting their potential roles as diagnostic and prognostic biomarkers. SELL encodes the L-selectin protein, which is expressed on the surface of leukocytes. By recognizing ligands on vascular endothelial cells, L-selectin mediates the initial adhesion and rolling of leukocytes at sites of inflammation^30,31. Although SELL is a classical leukocyte marker, previous studies have shown that fibroblasts can express certain adhesion molecules, including SELL³², under specific conditions, contributing to interactions within the inflammatory microenvironment. In our study, scRNA-seq data revealed SELL expression in fibroblast subpopulations, and IHC confirmed high SELL levels in fibroblasts within periodontitis tissues. These findings suggest that the elevated SELL expression in periodontitis may originate, at least in part, from specific fibroblast subsets. Activated leukocytes can secrete cytokines, such as TGF-β and IL-6, which may indirectly regulate fibroblast proliferation, differentiation, and extracellular matrix production. ST6GAL1 catalyzes α-2,6 sialylation and exerts tissue-specific effects on fibroblasts. In rheumatoid arthritis synovial fibroblasts, the pro-inflammatory cytokine TNF-α suppresses ST6GAL1 expression via transcriptional regulation, leading to a marked reduction in cell-surface α-2,6 sialylation and driving the transition toward a pro-inflammatory phenotype³³. In contrast, during skin repair, ST6GAL1 enhances α-2,6 sialylation, promotes fibroblast growth factor-2 (FGF-2) expression, and downregulates transforming growth factor-β1 (TGF-β1) and type I collagen (COL-I), thereby facilitating tissue repair³⁴.

Based on fibroblast hub genes, we classified periodontitis into three distinct clusters. Immune infiltration analysis revealed unique immune cell dynamics in each cluster, suggesting that fibroblast-associated subpopulations may shape the inflammatory microenvironment. For example, the pro-inflammatory C1 cluster exhibited a marked increase in CD8⁺ T cells and M1 macrophages, indicative of an enhanced immune response, whereas the C2 cluster was enriched in regulatory T cells and M2 macrophages, reflecting a more immunosuppressive state. This duality underscores the complex interplay between fibroblasts and immune cells, implying that targeting specific immune subpopulations could improve therapeutic outcomes in periodontitis patients. Gene set enrichment analysis further revealed associations with inflammatory signaling pathways, including IL-6/JAK/STAT and KRAS/p53, both of which play pivotal roles in mediating inflammatory responses and cellular stress^35,36. These findings suggest that fibroblast-derived signaling may represent a key driver in the pathogenesis of periodontitis.

Distinct periodontal fibroblast subpopulations may differentially contribute to pro-inflammatory or reparative processes. Single-cell sequencing has identified ICAM1⁺ fibroblasts as a key pro-inflammatory subset in periodontitis. This population exhibits high expression of chemokines such as CCL2 and CXCL1 via the NF-κB pathway, thereby recruiting macrophages and modulating neutrophil-driven inflammation³⁷. In addition, senescent gingival fibroblasts in periodontitis display hyperactivation of the mTOR pathway, leading to an imbalance in fibroblast activation protein (FAP) and osteonectin (OLN) expression. This imbalance promotes inflammatory polarization of macrophages and osteoclast differentiation, exacerbating bone resorption³⁸. Therefore, we further analyzed distinct fibroblast subtypes in periodontitis. Our study highlights a previously unreported CXCL13⁺ fibroblast subpopulation that is markedly enriched in periodontitis tissues, suggesting a specific role in shaping the inflammatory microenvironment. In contrast, IGHG2⁺ fibroblasts were predominantly observed in healthy tissues, implying a potential role in maintaining periodontal homeostasis. Importantly, tissue-level validation using FISH confirmed CXCL13 expression in fibroblast subsets, supporting the notion that these cells actively contribute to inflammation regulation. These findings provide new insights into fibroblast heterogeneity in periodontal disease and may inform future studies on tissue repair and inflammation modulation. Some hub genes (CD79A and IGHG1) are traditional B-cell markers, suggesting possible signal contamination or subpopulation-specific expression. However, FISH confirmed high expression of CXCL13 in fibroblast subsets, indicating they may reflect fibroblast functional states in the inflammatory microenvironment.

This study has several limitations that warrant consideration. Firstly, while we conducted a comprehensive analysis of gene expression and fibroblast dynamics in periodontitis, the sample size remains relatively small, which may limit the generalizability of our findings. Additionally, despite utilizing multiple datasets for validation, potential batch effects could influence the results, leading to variability in gene expression profiles. Furthermore, while our experimental validations provided support for key hub genes, not all findings were experimentally confirmed, which may affect the robustness of the conclusions drawn. Lastly, the complexity of fibroblast heterogeneity and their functional roles in periodontal disease necessitates further investigation to elucidate their contributions to disease pathology and therapeutic responses. Large-scale clinical studies will be necessary to translate these observations into practical applications.

This study not only confirms the critical role of fibroblasts in periodontitis but also identifies FRGs and a PIF subset, CXCL13⁺ fibroblasts. Our findings provide novel insights into the pathogenic mechanisms of fibroblasts in periodontitis and may inform the development of targeted diagnostic and therapeutic strategies.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

T.R.H. contributed to the study concept and design. M.H. wrote the first draft of the manuscript. C.P.L. and Y.L. supervised and oversaw the study. Z.Y.W. contributed to the statistical analysis. T.R.H. supervised the study. All authors have read and approved the final version of the manuscript and agreed to be accountable for all aspects of the research in ensuring that the accuracy or integrity of any part of the work is appropriately investigated and resolved.

Data availability

The datasets analyzed for this study can be found in the TCGA database (http://www.cancer.gov/tcga), the Gene Expression Omnibus (https://www.ncbi.nlm.nih.gov/geo/).

Declarations

Competing interests

The authors declare no competing interests.

Ethics approval and consent to participate

Ethics approval and consent to participate The studies involving human participants were reviewed and approved by the Ethics Committee of Fujian Provincial Hospital. The patients/participants provided their written informed consent to participate in this study.