Periodontal disease was induced in C57BL/6J mice by placing a ligature around the maxillary left second molar²³; gingival defect developed after Day 3 following ligature placement (Figure 1A and B). Three-dimensional (3D) reconstruction of micro-computed tomography (microCT) images (Figure 1C) further revealed a reduction in alveolar bone height starting from Day 3, which progressively increased on Day 7 (Figure 1D—figure supplement 1). Overall, the observed pattern of periodontal tissue degradation was consistent with that reported in previous studies^{24, 25}.

Left-side palatal gingiva tissue was harvested from mice on Day 0 (i.e., healthy gingiva without ligature placement) and on Days 1, 4, and 7 after ligature placement, and gingival cells were dissociated for single-cell RNA sequencing (scRNA-seq)^{22, 26}. On Days 0 and 1, the major cell types identified included epithelial cells expressing cadherin 1 (Cdh1)²⁷ and type XVII collagen (Col17a1)²⁸, fibroblasts expressing type I collagen (Col1a1)²⁹ and lumican (Lum)³⁰, B cells expressing membrane spanning 4 domains A1 (Ms4a1)³¹ and cluster of differentiation 79A (Cd79a)³², T cells expressing epsilon subunit of T cell receptor complex (Cd3e)³³ and cluster of differentiation 5 (Cd5)³⁴, and myeloid cells expressing lysozyme 2 (Lyz2)³⁵ and integrin subunit alpha M (Itgam)³⁶ (Figure 1E and F—figure supplement 2). On Days 4 and 7, an additional endothelial cell fraction expressing selectin P (Selp)³⁷ and selectin E (Sele)³⁸ emerged (Figure 1E and F—figure supplement 2), suggesting increased inflammatory neovascularization with the progression of periodontal inflammation. The proportion of B cells was increased on Day 1, and the proportion of myeloid cells increased progressively from Day 4 to 7 (Figure 1F). In addition, the proportion of fibroblasts was increased on Day 7 (Figure 1F).

The infiltration of proinflammatory neutrophils into the gingiva has been extensively characterized in this mouse model of periodontitis^{11, 39}. Here, we found that macrophages expressing cluster of differentiation 86 (Cd86)⁴⁰ and integrin subunit alpha X (Itgax)⁴¹ were predominant on Day 0 in the myeloid cell fraction from healthy gingiva (Figure 2A and B— figure supplement 3). In contrast, the proportion of neutrophils expressing CXC motif chemokine receptor (Cxcr2)⁴² and G0/G1 switch gene 2 (G0s2)⁴³ increased after ligature placement and during periodontitis development from Day 1 to 7 (Figure 2A and B—figure supplement 3), suggesting continuous neutrophil infiltration into the early and established gingival lesion. Moreover, after ligature placement, gingival neutrophils upregulated the expression of triggering receptor expressed on myeloid cells 1 (Trem1), indicating that these cells are activated and participating in the amplification of inflammatory signals^{44, 45} (Figure 2C). Strikingly, the Trem1-expressing activated neutrophils also show upregulation of matrix metalloproteinase 9 (Mmp9; Figure 2C)—a protein associated with extracellular matrix degeneration within gingival tissue⁴⁶. We further detected expression of tumor necrosis factor alpha (Tnfa) and transforming growth factor beta 1 (Tgfb1) in both macrophages and neutrophils on Day 0 and after ligature placement (Figure 2D). Collectively, these myeloid cell behaviors are consistent with those reported in prior studies on periodontitis development, thus validating our scRNA-seq data.

Myeloid cells are also known to stimulate other immune cells through the expression of CC motif chemokine ligands (CCLs) and CXC motif chemokine ligands (CXCLs), many of which are associated with periodontitis development⁴⁷. Our scRNA-seq analysis revealed that macrophages expressed Ccl2, Ccl9, Cxcl4, and Cxcl6, and neutrophils expressed Ccl3, Ccl4, Ccl6, Cxcl2, and Cxcl3 throughout periodontitis development (Figure 2E and F). These data suggest that chemokines and cytokines produced by macrophages and neutrophils in inflamed tissue may amplify and polarize the immune response toward chronic gingival inflammation.

We previously identified two distinct subpopulations of gingival fibroblasts, differentiated by expression of type XIV collagen (Col14a1)²², and these were also detected in our current scRNA-seq data from Day 0 to 7 (Figure 3A). Gene ontogeny (GO) enrichment analysis of the Col14a1-expressing fibroblast subpopulation revealed expression of major gene clusters related to immune regulation, including “Regulation of leukocyte migration” (Figure 3B). This immune regulatory phenotype appears to be unique to Col14a1-expressing fibroblasts, which are therefore referred to as ‘fibroblasts activated to guide leukocyte migration’ or AG fibroblasts. The other subpopulation of Col1a1-expressing fibroblasts appeared to keep typical fibroblastic features and is thus referred to as KT fibroblasts (Figure 3C). An additional fibroblast subpopulation expressing smooth muscle aortic actin 2 (Acta2) was detected on Day 7, and these were identified as myofibroblasts⁴⁸ (MF) (Figure 3A). We found that the proportions of KT and AG fibroblasts were equal on Day 0 and on Days 1 and 4 after ligature placement (Figure 3E and F). However, on Day 7, the proportion of AG fibroblasts decreased, and the MF fraction emerged (Figure 3E and F).

To further characterize AG fibroblasts and their “Regulation of leukocyte migration” phenotype, we analyzed expression of CCL and CXCL chemokines within fibroblast subpopulations in our scRNA-seq dataset¹³. Results show that AG fibroblasts activated expression of Ccl8, Ccl11, Ccl19, Cxcl1, Cxcl11, and Cxcl12 immediately after ligature placement on Days 1 and 4 (Figure 4A). Chemokine expression levels then decreased on Day 7, suggesting that AG fibroblasts might initiate early leukocyte migration into gingival tissue. Given that activation of Toll-like receptors (TLR) is known to increase chemokine expression, we further assessed expression of TLRs in fibroblast subpopulations. We found that AG fibroblasts displayed temporal activation of Tlr2, Tlr3, and Tlr4 expression on Days 1 and 4 (Figure 4B). Similarly, the TLR downstream genes Myd88, Irak1, Map3k7, and Rela were also expressed on Days 1 and 4 (Figure 4B). These data suggest that AG fibroblasts may guide the establishment of an early inflammatory environment within the gingiva and thereby promote periodontitis pathogenesis.

To validate the presence of AG fibroblasts, Day 1 gingiva and periodontal tissue histological sections were subjected to immunohistochemistry with antibodies against COL14A1 and CXCL12 (Figure 4C). We detected COL14A1- and CXCL12-positive AG fibroblasts localized near gingival epithelial cells in the connective tissue papillae and free gingiva, as well as in the cervical zone of the periodontal ligament (PDL) space (Figure 4C). We note that this localization pattern of AG fibroblasts appears to be highly suitable for early immune surveillance during periodontitis pathogenesis.

The interaction between chemokine ligands and their receptors has been extensively studied. Here, to evaluate the interaction between chemokine ligands strongly expressed by AG fibroblasts and chemokine receptors in innate immune cells, we matched AG fibroblast-expressed chemokines to expression of their putative receptors in myeloid cells. Results show that AG fibroblasts may regulate macrophages via the expression Ccl8 and Ccl11, which encode chemokines that can interact with CC chemokine receptors (CCRs) encoded by Ccr2 and Ccr5 in macrophages. Similarly, gene expression signatures suggest that AG fibroblast-mediated neutrophil regulation may occur through CCL8–CCR1, CXCL1–CXCR2, and CXCL12–CXCR4 interactions (Figure 5A and B). Notably, all chemokine–receptor pairs are expressed throughout periodontitis development, although expression of factors mediating the interaction between AG fibroblasts and macrophages was decreased on Day 7.

Intercellular communication may occur via other ligand–receptor interactions, which induce downstream target gene expression in recipient cells. Here, we performed NicheNet analysis to identify potential interactions between gingival fibroblasts and myeloid cells, revealing a trend toward increasing ligand–target interactions between fibroblasts and myeloid cells from Day 1 to Day 7 (Figure 5C and D). On Day 1, ligand expression was present primarily in AG fibroblasts (Figure 5E), whereas on Day 7, MF fibroblasts also exhibited prominent ligand expression (Figure 5F). Analysis of target gene expression revealed that both macrophages and neutrophils expressed receptors capable of interacting with AG fibroblast ligands on Day 1 (Figure 5G). In contrast, on Day 7, receptor expression was almost exclusively present in neutrophils, suggesting these cells are primarily targeted by MFs and AG fibroblasts at later stages of periodontitis development (Figure 5H).

We next analyzed expression of T cell-related genes in our scRNA-seq dataset and found that the Day 0 healthy gingiva exclusively contained Th cells expressing Cd4⁴⁹. However, after ligature placement, cytotoxic T (Tc) cells expressing Cd8⁴⁹ emerged, and on Day 7, Cd4^- Zbtb16⁺ ILCs and Cd4⁺Foxp3⁺ regulatory T (Treg) cells⁵⁰ were also detected (Figure 6A and B). The ILCs expressed Nfil3, a basic leucine zipper (bZIP) transcription factor required for ILC development⁵¹ (Figure 6C). However, to our surprise, expression of Rorγ, Il17a, and Il17f was only detected in ILCs (Figure 6C), indicating that this gingival subpopulation is predominantly composed of ILC3s.

ILC3s and Th17 cells share similar regulatory functions. However, the role of ILC3s in osteoimmunology has not been fully deciphered. Macrophage-colony stimulating factor (M-CSF), encoded by Csf1, and receptor activator of nuclear factor kappa-Β ligand (RANKL), encoded by Tnfsf11, are known to play critical roles in osteoclast induction. We found that in gingival tissue, AG fibroblasts and neutrophils are the primary cellular sources of Csf1; ILC3s also express this gene, albeit at lower levels (Figure 6E). Expression of Tnfsf11 was detected in AG fibroblasts and ILC3s (Figure 6G). These observations suggest that ILC3s may function as inducers of inflammatory osteoimmune responses, similar to Th17 cells, leading to osteoclastogenesis in periodontitis downstream of AG fibroblast activation.

We next examined the role of ILC3s in periodontitis pathogenesis by measuring ligature-induced gingival defects and alveolar bone resorption in Rag2^-/- mice, which lack functional B, Th, and Tc cells, and Rag2–IL-2 receptor common gamma (Il2rg) double-knockout mice (Rag2^-/-γc^-/-), lacking all lymphocytes, including ILCs. After ligature placement, gingival defects developed similarly in wild-type (WT), Rag2^-/-, and Rag2^-/-γc^-/- mice (Figure 7A and B). However, we found that alveolar bone loss was decreased in Rag2^-/- mice and nearly eliminated in Rag2^-/-γc^-/- mice (Figure 7C and D—figure supplement 4A). MicroCT image analysis further indicated a better perseveration of alveolar bone structure in Rag2^-/-γc^-/- mice, relative to the other groups (Figure 7C—figure supplement 4B). Histologically, osteoclastic resorption lacunae were observed on the alveolar bone surface at the cervical PDL and tooth apex PDL zones only in WT mice (Figure 7E and F). In addition, we detected a significant decrease in the number of tartrate-resistant acid phosphatase (TRAP)-positive osteoclasts in the cervical PDL zone of Rag2^-/-γc^-/- mice and in the apical PDL zone of both Rag2^-/- and Rag2^-/-γc^-/- mice (Figure 7G and H). Collectively, these data suggest that ILC3s, not Th17 cells, are responsible for cervical alveolar bone resorption in the mouse periodontitis model, a pathological phenotype consistent with human disease.

Based on our present data, we hypothesize that ILC3s within the gingival tissue play a pathological role in alveolar bone resorption in our mouse model. We therefore aimed to identify the cells that promote ILC3 development in mice. Similar to Th17 cells, ILC3 development was shown to be triggered by IL-6 and IL-23a^{52, 53}. A survey of our scRNA-seq data identified epithelial cells, fibroblasts, and myeloid cells as the source of Il6 in gingival tissue (Figure 8A). On Days 1 and 4 following ligature placement, AG fibroblasts primarily expressed Il6, whereas neutrophils became the predominant source on Day 7. We further found that gingival epithelial cells comprise at least four different subpopulations, plus an additional group displaying an epithelial–mesenchymal transition (EMT) phenotype on Day 7; Il6 was expressed by several of these epithelial subpopulations, including the EMT subgroup (Figure 8A—figure supplement 5A–D). Il23a was also detected in AG fibroblasts and myeloid cells (Figure 8B), with expression present in all epithelial cell subsets at various points in periodontitis development (Figure 8B— figure supplement 5D).

Lastly, we evaluated potential interactions between chemokine ligands expressed by AG fibroblasts and neutrophils and chemokine receptors in innate immune cells. Our data suggest the presence of a chemokine–receptor association between ILC3s and both AG fibroblasts (Figure 8C) and neutrophils (Figure 8D), although interactions with other innate immune cells are also possible. NicheNet analysis further identified potential ligand–target gene associations between lymphocytes, including ILC3s, and both fibroblasts (Figure 8E) and myeloid cells (Figure 8F). Ligand expression was more prominent in AG fibroblasts than in other fibroblast subpopulations (Figure 8G) and elevated in neutrophils relative to macrophages (Figure 8H). Additionally, target gene expression was detected in ILC3s, although it was not specific to these cells (Figure 8I, J). Thus, in total, our data suggest a regulatory role for a newly identified AG fibroblast subpopulation in the gingiva, which appears to orchestrate chronic gingival inflammation, at least in the early stages, and to promote alveolar bone resorption via stimulation of neutrophils and ILC3s.

All protocols for animal experiments were reviewed and approved by the University of California Los Angeles (UCLA) Animal Research Committee (ARC# 2003-009) and followed the Public Health Service Policy for the Humane Care and Use of Laboratory Animals and the UCLA Animal Care and Use Training Manual guidelines. C57BL/6J WT, Rag2^-/-, and Rag2^-/-γc^-/- mice were purchased from the Jackson Laboratory (Bar Harbor, ME, USA). Animals had free access to regular rodent diet and water and were maintained in standard housing conditions with 12-h light/dark cycles in the Division of Laboratory Animal Medicine at UCLA.

A silk thread was gently tied around the left maxillary second molar of 8- to 12-week-old female WT, Rag2^-/-, and Rag2^-/-γc^-/- mice under general inhalation anesthesia with isoflurane (Henry Schein, Melville, NY, USA). On Days 1, 3, and 7 after ligature placement, WT mice were randomly chosen and euthanized by 100% CO₂ inhalation. WT mice without ligature placement were used as Day 0 pre-periodontitis control. Rag2^-/- and Rag2^-/-γc^-/- mice were euthanized on Day 7. The palatal tissue was digitally photographed, and maxillae were harvested and fixed in 10% buffered formalin (Thermo Fisher Scientific, Waltham, MA, USA). The gingival defect area was measured from digital photographs, using the ImageJ Java-based image-processing program (NIH, Bethesda, MD, USA), and normalized to the circumferential area of the maxillary first molar. Fixed maxillae were subjected to microCT imaging at an energy level of 60 kV and 166 µA, and 3D images were reconstructed from microCT scans (Skyscan 1275: Bruker, Billerica, MA, USA). Alveolar bone loss was assessed at three sites (mesiobuccal cusp, distobuccal cusp, and distal cusp) of the first molar, two sites (mesiobuccal cusp and distobuccal cusp) of the second molar, and one site (buccal cusp) of the third molar by measuring the distance from the cementoenamel junction (CEJ) to the alveolar bone crest (ABC) on the buccal and palatal side of the alveolar bone. Total bone loss was calculated from the six-site total CEJ–ABC distance. The bone volume/total volume (BV/TV) ratio, bone surface, trabecular number (Tb.N), and trabecular thickness (Tb.Th) in the buccal side of alveolar bone of the second molar were determined using the proprietary analysis program (CTan: Bruker, Billerica, MA, USA). Statistical analysis to assess differences among multiple experimental groups was performed using two-way analysis of variance (ANOVA) with Tukey’s multiple-comparison test; p < 0.05 was considered to be statistically significant.

On Days 1, 4, and 7 after ligature placement, mice were euthanized by 100% CO₂ inhalation. And maxillary gingival tissues (n = 4 per group) were harvested from freshly isolated mouse maxillae.

Cell Ranger was used to align reads, generate feature–barcode matrices, and perform clustering and gene expression analyses on the scRNA-seq data, and the output from this program was analyzed using the R-program Seurat (https://satijalab.org/seurat/). Cells with <2,400 genes detected or >1% mitochondrial gene expression were filtered out as low-quality cells. Individual gene counts for each cell were divided by the total gene counts for that cell and multiplied by a scale factor of 10,000; natural-log transformation was then applied to the counts. The FindVariableFeatures function was used to select 2,000 variable genes with default parameters, and the ScaleData function was used to scale and center the counts in the dataset. Principal component analysis and Uniform Manifold Approximation and Projection dimensional reduction were performed on variably expressed genes. The cluster markers were found using the FindAllMarkers function, and cell types were manually annotated based on the cluster markers. Cell types were assigned based on expression of cell marker genes, and gene expression within different cell types was displayed using dot plots and violin plots.

Fixed maxillae were decalcified in 10% EDTA (Sigma–Aldrich) for 3 weeks and then embedded in paraffin. Histological cross-sections were stained with hematoxylin and eosin (HE) and evaluated on a light microscope. Adjacent paraffin sections (4 µm) were subjected to a heat-induced epitope retrieval procedure and then immunohistochemically stained with polyclonal antibodies to COL14A1 (#PA5-49916, Thermo Fisher Scientific) and CXCL12 (#PA5-30603, Thermo Fisher Scientific) at a 1:100 dilution, followed by secondary antibody application, diaminobenzidine staining, and methylene blue counterstaining.

Using maxillary cross-sections of WT, Rag2^-/-, and Rag2^-/-γc^-/- mice, osteoclasts were evaluated by TRAP staining using a commercially available kit (Acid Phosphatase TRAP kit, Sigma–Aldrich), according to the manufacturer’s instructions. TRAP-positive cells were counted in the 0.01-mm² area. Statistical analysis to assess the differences among multiple experimental groups was performed using two-way ANOVA with Tukey’s multiple-comparison test; p < 0.05 was considered to be statistically significant.

Annotation and visualization of GO terms were performed by Metascape (http://metascape.org/gp/index.html#/main/step1). The top-100 differentially expressed genes in each population were input and filtered with the term “immune”. Filtered genes were then input, and only “biological process” gene sets were retrieved from the GO database.

NicheNet (v.1.0.0, https://github.com/saeyslab/nichenetr) was used to predict interactions between cell types. In brief, the integrated Seurat object containing each cell subpopulation was input into the NicheNet Seurat wrapper. Sender cells and receiver cells were determined, and interactions between active ligands expressed by sender cells and target receptors expressed by receiver cells were predicted based on information in signaling and ligand–receptor databases.