Polybacterial intracellular macromolecules shape single-cell inflammatory profiles in upper airway epithelia

Abstract

Mucosal epithelial cells of the upper airways are continuously exposed to microbes throughout life. Specialized niches such as the anterior nares and the tooth are especially susceptible to dysbiosis and chronic inflammatory diseases. Here, we reanalyzed our v1-Human Periodontal Atlas, identifying polybacterial signatures (20% Gram-positive; 80% Gram-negative) and distinct responses of bacterial-associated epithelia. Fluorescence microscopy detected numerous persistent polybacterial intracellular macromolecules (PIMs) within human oral keratinocytes (HOKs), including bacterial rRNA, mRNA, and glycolipids. PIM levels directly correlated with enhanced receptor-ligand signaling in vivo. Inflammatory "keratokines" targeting immune cells were synergistically upregulated in lipopolysaccharide-challenged HOKs, while endogenous lipoteichoic acid (LTA) correlated with CXCL1/8 expression in vitro and in vivo. Application of Drug2Cell suggested altered drug efficacy predictions based on PIM detection-agnostic of disease state. CXCL1/8 expression again correlated with LTA in epithelial cells of the nasal cavity, oropharynx, and trachea. Thus, PIMs shape epithelial single-cell profiles across upper airway mucosae.

Fig. 1. Intracellular bacterial signatures across periodontal cell types.

a We previously generated an integrated periodontal atlas comprising scRNAseq data from four independent studies. b We performed metagenomic reannotation on this atlas using unmapped reads that were mapped back to bacterial sequences. All cell types were classified as having no or containing one (monobacterial) or multiple (polybacterial) intracellular bacterial RNA reads. c Remapping the bacterial reads across all structural cell types revealed numerous gram-positive and gram-negative species. Staphylococcus (S.) aureus, Stenotrophomonas maltophilia, Modestobacter mariunus, Porphyromonas (P.) gingivalis, and Prevotella sp. were the most frequently mapped. d Considering disease states of only keratinocytes (health, gingivitis, and periodontitis), cells contained no, monobacterial and polybacterial intracellular bacterial reads. e, f Bacteria as a cell state and gingivitis and periodontitis as clinical states all had impacts on gene upregulation, but all 3 independent states e shared 42 upregulated genes, including f transcription factors, enzymes, RNA and DNA binding proteins, and structural, signaling, chromatin, Golgi/nuclear, and ubiquitin markers. Abbreviations: KCs Keratinocytes, ILCs innate lymphoid cells, LECs lymphatic endothelial cells, VECs vascular endothelial cells. Image in c created using InteractiVenn.

Fig. 2. Human gingival keratinocytes (HOKs) contain bacterial rRNA signal, maintained over multiple passages.

RNAscope on P2 HOKs detected bacterial rRNA in a basal and b suprabasal human oral keratinocytes. c We employed 3D super-resolution imaging to detect intact bacteria in P2 HOKs and in saliva. 3D reconstruction of wheat germ agglutinin (WGA) and SYTO9 staining of HOKs revealed no intracellular intact bacteria. Staining of saliva samples, on the other hand, revealed bacteria on the surfaces of cells. d Microbiome analysis of HGKs, considering as-received (passage [P]2) and multiple passages (P4) thereafter revealed the HOKs contained similar proportions of gram-positive and gram-negative bacterial species across passages. e Multiplexing anti-lipoteichoic acid (LTA) and anti-lipopolysaccharide (LPS) antibodies with bacterial rRNA probes fimA (P. gingivalis) and spa (S. aureus) revealed intracellular bacterial components were maintained at P4. f LTA and S. aureus spa counts had a quasilinear relationship compared to LPS and P. gingivalis fimA. Scale bars: a, b 5 µm; c 10 µm; e 5 µm.

Fig. 3. Intracellular bacterial signal affects keratinocyte signaling and correlates with intracellular 16S signal and LPS in vitro challenge of human gingival keratinocytes (HGKs).

a Receptor-ligand analysis showed that in health and periodontitis, signaling by KCs without intracellular bacterial signal decreased, while signaling by KCs with intracellular bacterial signal remained constant. The breadth of signaling to by KCs with intracellular bacterial signal to LECs and VECs expanded. b In health and periodontitis, states with no bacteria, monobacterial, and polybacterial signals impacted keratokine expression: polybacterial signatures resulted in a shift towards CXCL17, CXCL8, IL36G, IL1B, and CCL20; only CXCL3 was a hallmark of monobacterial signature. c P4 HOKs were cultured with and without LPS (P. gingivalis). These cells were assessed for 16S signal and keratokines predicted to increase in disease states using RNA FISH in three rounds. d HOKs were imaged in xyz using Nyquist-optimized parameters to show correlation of 16S+ burden and keratokine production. This happens without Porphyromonas gingivalis LPS (d, e) yet increases in HOKs with LPS, highlighting this phenomenon for CCL28, CXCL17, and CXCL8. f Quantification of all keratokines considering 16S signals per cell. For CXCL17, CXCL8, IL36G, IL1B, CCL20, TNFSF15, IL6, and CXCL8, intracellular bacterial signature elicits mRNA expression without LPS challenge. Only CCL28, which is known to be chemoattractant to T and B cells, is downregulated with increasing bacterial burden. Binning cells in 16S groups with and without LPS challenge, CXCL17 and CXCL8 expression demonstrate a synergy between LPS challenge, polybacterial intracellular signal, and keratokine expression. Scale bar: d, e 10 μm. p < 0.005, paired Student’s T test (f). Illustration in c created with BioRender (https://www.biorender.com).

Fig. 4. Intracellular bacterial signal in situ correlates with keratokine signal in periodontitis.

a We used our RNAscope HiPlex keratokine panel on human periodontitis biopsies and imaged the junctional, sulcular, gingival margin, and attached gingival epithelial regions. Both low- and high-16S keratinocytes were observed in all regions of the epithelium. At a single-cell level, 16S signal correlated with keratokine upregulation, shown here for CXCL17, CCL28, and CXCL8. b Across the gingival epithelial regions, quantification of intracellular 16S signal revealed statistically significant associations between bacterial burden and CXCL17, CCL28, and CXCL8. Scale bars: a 10 µm. b p < 0.005, paired Student’s T test.

Fig. 5. Gram-positive bacterial glycolipids correlate with keratokine signaling in vitro and in vivo.

a Multiplexing anti-LTA and anti-LPS antibodies with keratokines CXCL8 and CXCL1 in HOKs revealed synergistic upregulation of keratokines with increasing intracellular LTA. b We quantified per-cell correlation of bacterial glycolipids and keratokines and found that CXCL1/8 both increased linearly with increasing LTA counts. On the other hand, keratokine counts had little correlation with increasing LPS counts. c In periodontitis tissues, we observed epithelial-wide signatures of bacterial components and CXCL1/8. In the disease zone of the sulcus, we found LPS- and LTA-low and -high cells whose CXCL1/8 directly correlated with bacterial component presence. d We binned keratinocytes in health and periodontitis based on their LTA and LPS counts, site agnostic of the epithelium. For both LTA and LPS, regardless of health or disease status, we found no statistically significant difference between the number of cells with less than 50 counts, cells with 50 or greater but less than 100 counts, and cells with greater than 100 counts. e We quantified the correlation of keratokine counts in health and periodontitis tissues with intracellular LTA and LPS counts. In both health and periodontitis, increasing CXCL1/8 counts had higher correlation with LTA than with LPS. LTA lipoteichoic acid, LPS lipopolysaccharide. Scale bars: a 5 µm; c 20 µm.

Fig. 6. Drug2Cell analysis of the impact of intracellular bacterial components (IBCs) on predicted drug targets.

a Impact of intracellular bacterial components on drug target prediction through Drug2Cell. Drugs were organized by either drug category or whether the drug was shared between or unique to keratinocytes (KCs) with or without bacteria. Drug categories included keratinocyte-specific, antibiotic, monoclonal antibody, chemotherapeutic, or anti-inflammatory. KCs with bacterial signal had higher predicted targets than KCs without. Ker. Keratinocyte-specific, Anti-infl. anti-inflammatory; rest: see Fig. 1.

Fig. 7. Correlation of PIMs and single-cell inflammatory signatures in other upper airway niches and trachea.

Multiplexed detection using anti-lipoteichoic acid (LTA) and anti-lipopolysaccharide (LPS) antibodies, alongside RNA probes for CXCL8 and CXCL1, revealed cytokine upregulation in association with increasing intracellular LTA levels in healthy human a nasal epithelium, b tonsillar epithelium, c tracheal epithelium, d nasal glands, and e tracheal glands. Both low-count (yellow arrows) and high-count (red arrows) bacterial glycolipid–positive epithelial cells were observed across all five sites. Quantitative analysis of cytokine transcript counts as a function of intracellular bacterial glycolipids demonstrated a positive correlation between LTA levels and CXCL8/CXCL1 expression in all tissues. Scale bar: 30 µm.