Decoding the role of macrophages in periodontitis and type 2 diabetes using single-cell RNA-sequencing

Abstract

Macrophages are resident myeloid cells in the gingival tissue which control homeostasis and play a pivotal role in orchestrating the immune response in periodontitis. Cell heterogeneity and functional phenotypes of macrophage subpopulations in periodontitis remain elusive. Here, we isolated gingival tissue from periodontitis-affected and healthy sites of patients with and without type 2 diabetes mellitus (T2DM). We then used single-cell RNA-sequencing (scRNA-seq) to define the heterogeneity of tissue-resident macrophages in gingival tissue in health vs. periodontitis. scRNA-seq demonstrated an unforeseen gene expression heterogeneity among macrophages in periodontitis and showed transcriptional and signaling heterogeneity of identified subsets in an independent cohort of patients with periodontitis and T2DM. Our bioinformatic inferences indicated divergent expression profiles in macrophages driven by transcriptional regulators CIITA, RELA, RFX5, and RUNX2. Macrophages in periodontitis expressed both pro-inflammatory and anti-inflammatory markers and their polarization was not mutually exclusive. The majority of macrophages in periodontitis expressed the monocyte lineage marker CD14, indicating their bone marrow lineage. We also found high expression and activation of RELA, a subunit of the NF-κB transcription factor complex, in gingival macrophages of periodontitis patients with T2DM. Our data suggested that heterogeneity and hyperinflammatory activation of macrophages may be relevant to the pathogenesis and outcomes of periodontitis, and may be further augmented in patients with T2DM.

Keywords: macrophages; periodontitis; single-cell RNA-sequencing; type 2 diabetes.

FIGURE 1

Cellular heterogeneity in periodontitis-affected gingival tissue versus health. (A) Major clusters and respective cell-type assignments in UMAP plot of 5 periodontitis-affected and 5 healthy gingival tissue samples. Clusters are assigned to indicated cell types by differentially expressed genes. (B) Heatmap shows genes (columns) that are differentially expressed genes (log-normalized, scaled) across 15 cell clusters (rows). Yellow: high expression; Purple: low expression. (C) Frequency (%) of cellular subtypes in each cellular cluster in periodontitis versus health

FIGURE 2

Lineage markers and differentially expressed genes in Mφ in gingival tissue. (A) Feature plots showing the RNA expression (log-normalized) of lineage marker genes used for cell-type classification (CD3D, CD79A, CD14, CD68, GATA2, MS4A1, CD68, FCGR3A) in the same embedding as Figure 1A. (B–J) Violin plots of different pro-inflammatory and anti-inflammatory genes in periodontitis-affected tissue and health. Red: periodontitis-affected tissue; Gray: Health

FIGURE 3

Pathway enrichment analysis and network analysis of differentially expressed genes in Mφ in gingival tissue. (A) Gene ontology enrichment analysis of Mφ in periodontitis-affected tissue versus health. (B) Protein interaction network analysis of periodontitis-affected tissue. p-Values are adjusted by Benjamini–Hochberg false discovery rate (FDR) adjustment <.01. (C, D) Gene set enrichment analysis of overexpressed transcripts in Mφ in periodontitis-affected tissue based on the ENCODE TF ChIP-seq database. p-Values are adjusted by Benjamini–Hochberg false discovery rate (FDR) adjustment <.01. (E) Gene set enrichment analysis of over-expressed proteins in Mφ in periodontitis-affected tissue based on TRRUST transcription factors

FIGURE 4

Sub-clustering of Mφ based on the transcriptional profile. (A) Sub-clusters in UMAP plot of periodontitis-affected Mφ. Clusters are assigned to indicate sub-clusters by differentially expressed genes. (B) Heatmap shows genes (columns) that are differentially expressed (log-normalized, scaled) across 15 cell clusters (rows). Yellow: high expression; Purple: low expression. (C) Feature plots showing the RNA expression (log-normalized) of Mφ markers in the same embedding as Figure 4A. (D–H) Gene set enrichment analysis of over-expressed transcripts in Mφ in periodontitis-affected tissue based on Gene Ontology (Molecular Function). p-Values are adjusted by Benjamini–Hochberg false discovery rate (FDR) < .01

FIGURE 5

Re-clustering of Mφ in periodontitis-affected tissue based on transcriptional landscape. (A) Macrophages in periodontitis-affected tissue were clustered based on the DoRothEA algorithm. (B, C) Violin plots demonstrating two Mφ-specific sub-clusters (cluster 3 and cluster 8) expressing pan-Mφ markers (CD11b and LYZ). (D, E) Violin plots demonstrating over-expression of Mφ activation markers (CD86, CD163) in Mφ clusters (cluster 3 and cluster 8) isolated from periodontitis-affected tissue. (F, G) Top transcriptional factors expressed in two sub-clusters of Mφ in periodontitis-affected tissue. Data presented as log 2-fold change for significantly activated transcription factors with p-value of <.01 after FDR adjustment

FIGURE 6

(A) Dot plot of predicted interactions of macrophages with other cells in the periodontitis-affected tissue, p-values are indicated by circle size. The expression levels of all the interacted genes were indicated by colors as presented by the scales on the right. (B) Ligand receptor heat map of periodontal tissue. The heat map is presented based on the log of the number of interactions

FIGURE 7

Expression of RELA, RFX5, RUNX2, and CIITA in periodontitis with and without T2DM. (A) Mφ were isolated with CD11b magnetic beads from periodontitis-affected tissue in patients with (n = 7) or without T2DM (n = 8). qRT-PCR was used to identify the expression of RELA in Mφ of patients with and without T2DM. 18S was used as a normalizer. The 2^−ΔΔCT method was used for relative expression. (B, C) Confocal images demonstrating RELA nuclear and cytoplasmic expression in Mφ in unstimulated condition and after Porphyromonas gingivalis LPS challenge (10 nM for 24 h). Nucleolus and cytoplasmic fluorescent intensities were quantified by ImageJ. (C–E) qRT-PCR was used to identify the expression of RFX5, RUNX2, and CIITA in Mφ of patients with and without T2DM. 18S was used as a normalizer. The 2^−ΔΔCT method was used for relative expression. Data are presented as mean and 95% confidence interval. *p < .05