This study was carried out in accordance with the recommendations of the Institutional Ethics Committee of West China Hospital of Stomatology with written informed consent from all subjects. All subjects gave written informed consent in accordance with the Declaration of Helsinki. The protocol was approved by the Institutional Ethics Committee of West China Hospital of Stomatology (WCHSIRB-ST-2014-091). Periodontitis tissues were obtained from patients (n = 5) diagnosed as chronic periodontitis undergoing gingivectomy. Healthy gingival tissues were harvested from patients (n = 3) without any periodontal diseases undergoing crown lengthening surgery. The included patients were diagnosed as chronic periodontitis in West China Hospital of Stomatology accordingly to classification of the periodontal diseases (Armitage, 1999). They had neither periodontal treatment nor antibiotic use for at least 3 months. The exclusion criteria included diabetes mellitus, pregnancy, liver, or kidney dysfunction, autoimmune diseases and infectious stomatitis (oral candidosis, herpetic stomatitis). The patient information was shown in Supplementary Table 1.

This study was carried out in accordance with the recommendations of the Guide for the Care and Use of Laboratory Animals from the National Institutes of Health. The protocol acquired the approval of the Institutional Animal Care and Use Committee, Cedars Sinai Medical Center (IACUC003638). The mice were housed at the Cedars Sinai Medical Center Animal Facility. 4–8 weeks old female Balb/c mice were obtained from the Harlan Laboratories (Madison, WI, USA). The experimental periodontitis model was established according to previous studies (Yokoyama et al., 2011; Cheng et al., 2015a). One hour before euthanization, the mice were injected into the peritoneal cavities with pimonidazole hydrochloride (Hypoxyprobe™-1; HPI, Inc., Burlington, MA, USA; Shi et al., 2013) at a dose of 60 mg/kg.

For human specimens, the IHC detection was conducted using antibodies anti-human HIF-1α (Abcam, Cambridge, MA, USA), anti-human NLRP3 (LifeSpan BioSciences, Seattle, WA, USA), anti-human cleaved-caspase-1 (Biorbyt Ltd., Cambridge, UK) and anti-human IL-1β (Abcam; mainly detect mature IL-1β). Images were captured by Nikon Eclipse 80i (Nikon Instruments Inc., Melville, NY, USA) and shown in Figure 1 and Supplementary Figure 1. For mouse tissues, the IHC was detected using anti-mouse NLRP3 (LifeSpan BioSciences, Seattle, WA, USA), anti-mouse IL-1β (R&D Systems Inc., Minneapolis, MN, USA; detect pro- and mature IL-1β). Images were captured by Aperio® AT2 (Leica Biosystems, Wetzlar, Germany), and shown in Figure 4 and Supplementary Figure 2. The IHC scores were determined as described previously (Kreisberg et al., 2004).

The method was a modification of a previously described method (Lin et al., 2013). The antigen retrieved tissues incubated in the following sequences: the endogenous biotin-blocking kit (Molecular Probes Inc., Eugene, OR, USA) according to the instruction; PBS containing 2% bovine serum albumin (BSA) at 37°C for 30 min; equilibration buffer (KeyGEN.Co.Ltd., Nanjing, Jiangshu, China) for 10 min; the TdT reaction mix (rTdT enzyme and biotin-11-dUTP in equilibration buffer; KeyGEN. Co.Ltd.,) for 60 min; streptavidin-conjugated QD (QD-SA) 605 (Invitrogen, Carlsbad, CA, USA) for 30 min. After adequate wash, the specimens were incubated in 2% BSA at 37°C for 30 min; anti-human cleaved-caspase-1 (Biorbyt Ltd.,) for overnight at 4°C; 2% BSA at 37°C for 10 min; biotin-conjugated secondary antibody at 37°C for 30 min; 2% BSA at 37°C for 20 min; QD-SA 525 at 37°C for 30 min. Finally, the slides were mounted in 496-diamidino-2-phenylindole (DAPI) (Applygen Technologies Inc., Beijing, China) for 5 min. The images were taken by using a fluorescence microscope (Olympus Fsx100, Olympus Corporation, Tokyo, Japan).

The human gingival tissue was cut into pieces, and then digested in 3 mg/mL collagenase type I (Hyclone, Logan, UT, USA) for 1 h at 37°C. Cells were cultured in Dulbecco's modified Eagle's medium (Gibco, Grand Island, NY, USA) adding 10% fetal calf serum (Biowest, France) plus 100 U/mL penicillin and 100 μg/mL streptomycin (Hyclone). Cells between passages 3–7 were used. Mouse gingival fibroblasts were cultured from gingiva obtained from 6 to 8 weeks old female Balb/c mice (Harlan Laboratories). The gingival tissue was cut and cultured in Alpha Modifications Minimum Essential Medium with 10% fetal bovine serum plus 100 U/mL penicillin and 100 μg/mL streptomycin (all from Cambrex, Walkersville, MD, USA) in a humidified atmosphere of 5% CO₂ at 37°C. Cells between passages 3–6 were used.

Human gingival fibroblasts were serum-starved for 24 h then treated with 1 μg/mL E. coli LPS [Escherichia coli LPS (O111:B4; Sigma-aldrich, St Louis, MO, USA)] (Herath et al., 2011) or 10 μg/mL P. gingivalis LPS (Invivogen, San Diego, CA, USA) (Abe-Yutori et al., 2017) at 2 or 20% O₂ for 6 h. Mouse gingival fibroblasts were serum-starved for 24 h then were treated with 1 μg/mL E. coli LPS at 2% O₂ for 1, 2, 4, and 24 h. Cells were collected and proteins were extracted in lysis buffer (Keygen Biotech Inc., Nanjing, China) and 20–30 μg of the total proteins were loaded and separated by using 10% SDS-PAGE. The following proteins were detected using antibodies, anti-human, -mouse NLRP3 (LifeSpan BioSciences, Seattle, WA, USA); anti-human, -mouse caspase-1 (Santa Cruz Biotechnology, Inc., Santa Cruz, CA, USA); anti-human IL-1β (Abcam; detect pro- and mature IL-1β) and anti-mouse IL-1β (Novus Biologicals, Inc., Littleton, CO, USA; detect pro- and mature IL-1β) were used. The proteins were visualized using an enhanced chemiluminescence kit (Millipore Inc., Darmstadt, Germany). The bands were analyzed by using Quantity One (Bio-Rad).

Mouse gingival fibroblasts were serum-starved for 24 h then stimulated with 1 μg/mL E. coli LPS for 24 h. Total mRNA was extracted using RNase mini kit (Giagen, Hilden, Germany). The total RNA were reversely transcript to cDNA by (Bio-Rad Laboratories, Inc., Hercules, CA, USA). The mRNA expression of IL-1β was measured by RT-PCR, which was performed on Bio-radS1000™ Thermal Cycler (Bio-Rad Laboratories) using GoTaq Green Master Mix (VWR International, Radnor, PA, USA). Primer sequences were forward ACCTAGCTGTCAACGTGTGG; reverse TCAAAGCAATGTGCTGGTGC.

Data were expressed as mean ± SD. The statistical significance of differences among groups was assessed using Student's t-test or one-way ANOVA by SPSS 16.0 (IBM Corp. New York, NY, USA). A difference was considered significant if P < 0.05.

We found hypoxia to be common in human periodontitis afflicted gingival tissues, as evidence by the expression of the transcription factor, HIF-1α (hypoxia-inducible factor α) (Ng et al., 2011; Figure 1A). Conversely, healthy gingiva presented a negative staining of HIF-1α, suggesting that periodontal inflammation induced a shift from normoxic phase to hypoxic phase. As hypoxia and elevated reactive oxygen are associated with inflammasome activation, we tested its presence in patient gingiva. In response to bacterial infection, the NLRP3 inflammasome would be assembled for the caspase-1 activation. The subsequent activation of caspase-1 promotes the release of the proinflammatory cytokine IL-1β in contributing to the inflammatory response (Kim and Jo, 2013). In this study, we examined the expressions of NLRP3, cleaved-caspase-1 (active), and IL-1β expression, each of which were upregulated in human periodontitis gingival tissue, compared to health gingiva (Figure 1B). In support of caspase-1-mediated cell death, pyroptosis, we found co-expressions of cleaved-caspase-1 and Tunel in the periodontitis tissues. The co-staining supported that activated caspase-1 was associated with cell death (Figure 2A). Gingival fibroblasts also suffered from cell death in periodontitis (Figure 2B). The hypoxic environment, usually accompanying inflammation, is a potential pathological factor in itself (Eltzschig and Carmeliet, 2011). Although clinical evidence and research support P. gingivalis as a key periodontal pathogen, we speculated that hypoxia can be involved in periodontal cell death process.

There is a longstanding paradox that P. gingivalis LPS reveals low inflammatory potency though it is the “keystone pathogen.” The paradox might be explained by the synergistic infection of P. gingivalis and commensal microbial community (Darveau et al., 2012). LPS is the principal component of Gram-negative bacteria that activates the innate immune system. Here E. coli LPS, a kind of PAMP, was used to study the effect of Gram-negative microbiota. Even though E. coli LPS does not come from oral bacteria, it has been used in stimulating in vivo and in vitro periodontal inflammation (Gürkan et al., 2009; Lu et al., 2017). We also simulated normoxic and hypoxic conditions to study the effects of two kinds of LPS in vitro.

Our data showed that P. gingivalis LPS down-regulated NLRP3, IL-1β precursor (pro-IL-1β) and mature IL-1β under normoxia. However, under moderate hypoxia (2% O₂), P. gingivalis LPS enhanced NLRP3 expression, cleaved-caspase-1, and mature IL-1β (Figure 2). Further, the protein level of pro-IL-1β was increased (Figure 2), suggesting a NF-kB transcriptional activity. The E. coli LPS induced pro-IL-1β expression and maturation of IL-1β, which was associated with cleaved-caspase-1, under hypoxic condition. But mature IL-1β was also upregulated compared to control under normoxic condition (Figure 3). Therefore, E. coli LPS was able to increase mature IL-1β in both normoxic and hypoxic phases. E. coli LPS-induced inflammation would be a candidate reason in the shift from normoxic phase to hypoxic phase.

To verify the assumption that E. coli LPS may induce hypoxia, E. coli LPS-induced periodontitis model was chosen in this experiment. We found evidence of hypoxia in the periodontitis model, as determined by hypoxyprobe localization. The results showed that E. coli LPS was capable of changing normoxic phase (the control) to hypoxic phase (the periodontitis model) in vivo. The results also suggested that Gram-negative microbiota may prepare a favorable environment for P. gingivalis and helped to propel caspase-1 activation.

In the experimental periodontitis model, NLRP3 and IL-1β were also enhanced, similar to the results of human periodontitis specimen (Figure 4B). The mechanism of pro-IL-1β and mature IL-1β generation was tested in mouse gingival fibroblasts with E. coli LPS at 2% O₂. We found that NLRP3 up regulation at 2 h of treatment was short lived (Figure 5A). Yet, cleaved-caspase-1 was enhanced by hypoxia and E. coli LPS at 1 h and 2 h. But by 4 h, E. coli LPS had little effect on either NLRP3 or cleaved caspase-1 expression. E. coli LPS were associated with elevated IL-1β transcriptional expression. Caspase-1 activation was the mechanism of mature IL-1β production in mouse gingival fibroblasts (Figures 5B,C). Together, the results in mouse in vivo and in vitro models confirmed that E. coli LPS induced hypoxic phase, which subsequently participated in the caspase-1 activation and mature IL-1β generation.

When infection initiates, NOD-like receptors (NLRs) are able to assemble a multiprotein complex (inflammasome), which activated caspase-1 and even induced pyroptotic (Vande Walle and Lamkanfi, 2016). Some studies have provided clues that inflammasome and caspase-1 be involved in the initiation and progression of periodontitis. For example, much higher NLRP3 inflammasome components were detected in periodontitis tissues than from healthy gingiva (Huang et al., 2015; Xue et al., 2015). The NLRP3 inflammasome increased the secretion of IL-1β, IL-6, and TNF-α in human periodontal ligament fibroblasts (Lu et al., 2017). Our study verified that NLRP3, cleaved-caspase-1 and IL-1β were enhanced in periodontitis tissues (Figure 1). Caspase-1 activation finally led to pyroptotic cell death (Figure 2). Usually, caspase-1 activation function as a host defense mechanism in eliminating pathogens (Yang et al., 2015). Caspase-1 induced human leukocytes lysis and IL-1β secretion under the attack of leukotoxin produced by A. actinomycetemcomitans (Kelk et al., 2008, 2011). However, elevated or inappropriate caspase-1 activation can lead to inflammation and excessive cell death (Simon and van der Meer, 2007). Caspase-1 is involved in the pathogenesis of inflammatory diseases, including inflammatory bowel disease, neurodegenerative diseases, and endotoxic shock (Bergsbaken et al., 2009). It is therefore reasonable to consider caspase-1 be involved in the pathogenesis of periodontitis.

Hypoxia is a key factor of propelling caspase-1 activation. In macrophages, hypoxia increased the NLRP3, caspase-1 activation and mature IL-1β secretion (Folco et al., 2014). In hepatocellular carcinoma cells, hypoxia induces caspase-1 activation, cleavage and release of proinflammatory cytokines, IL-1β and IL-18 (Yan et al., 2012). Our data for the first time demonstrated the importance of hypoxia in the mechanism of caspase-1 activation and IL-1β maturation. The effect of E. coli LPS was enhanced by hypoxia on gingival fibroblasts. Furthermore, hypoxia reversed the effect of P. gingivalis on NLRP3, caspase-1 activation and production of mature IL-1β in vitro. Unlike IL-6 and TNF-α, IL-1β has a unique post-translational modification and secretion mechanism. IL-1β is produced as a proprotein, which is proteolysed to its active form by caspase 1. Then the plasma-membrane pores produced by pyroptosis are helpful for the secretion (Piccioli and Rubartelli, 2013). It suggested that caspase-1 and pyroptosis is important in producing and releasing IL-1β in periodontitis. However, the secretory IL-1β was very low in this study, suggesting that the secretory process requires further investigation (Supplementary Figure 3).

Therefore, the shift from normoxic phase to hypoxic phase may be a potential milestone in periodontitis, from which the inflammation would be exaggerated. The Gram-negative microbiota, as we assumed, could influence the phase conversion. In this study, E. coli LPS successfully induced a hypoxic environment in the experimental periodontitis model. As a facultative anaerobe, E. coli LPS induced the caspase-1 cascade under normoxia, which might function as a defense mechanism in eliminating pathogens. However, E. coli LPS also increased IL-1β and thus induced hypoxia in the gingiva in a chronic process.

As an anaerobe, P. gingivalis scarcely survives under normorxia. Correspondingly, hypoxic condition could be a more favorable environment for P. gingivalis. This pattern of oxygen tension should accompany bacterial toxin in mediating inflammatory responses. Previous studies showed a paradox that P. gingivalis was strongly associated with periodontitis but was not an apparent inducer of inflammation (Darveau et al., 2012). P. gingivalis LPS revealed an unusually low inflammatory effect compared with E. coli LPS and LPS from many other Gram-negative bacteria (Munford and Varley, 2006). The environment factor, hypoxia, was a reasonable explanation for the paradox. It was established that P. gingivalis-induced periodontitis requires the presence of the commensal microbiota (Hajishengallis et al., 2011). A model of 11 subgingival biofilm (including P. gingivalis) and gingival tissues also up-regulated IL-1β secretion under hypoxia (Bao et al., 2015). Now we replenish that the effect of E. coli LPS (represent LPS from Gram-negative bacteria) helps to prepare a hypoxic environment, which exaggerate the inflammatory effect of P. gingivalis. The assumption may be further verified by using germ-free mice, which supplies an ideal model to study the synergistic effects of E. coli LPS and P. gingivalis LPS.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.