Epidermal growth factor receptor signaling suppresses αvβ6 integrin and promotes periodontal inflammation and bone loss

Periodontal diseases (PDs) are inflammatory conditions that affect the supporting structures of the teeth, resulting in destruction of alveolar bone and loss of the periodontal attachment (Kinane et al., 2017). The prevalence of PDs is high, with 10 to 15% of adult population worldwide suffering from the severe forms of the disease (Petersen and Ogawa, 2005; Petersen and Ogawa, 2012). These diseases also contribute to systemic inflammation and thus to systemic diseases, including cardiovascular disease, cerebrovascular disease, Alzheimer's disease, rheumatoid arthritis and cancer (Cullinan and Seymour, 2013; Dominy et al., 2019). During the PD process, junctional epithelium, which seals the gingiva to the tooth enamel in the healthy periodontal condition, transforms into pocket epithelium that proliferates into the connective tissue, allowing infiltration of bacteria and their virulence factors into the periodontal and other tissues (Larjava et al., 2011). Multispecies bacterial biofilms then accumulate deeper between the tooth and the pocket epithelium causing further periodontal inflammation, tissue degradation and alveolar bone loss. Thus, the function of junctional epithelial cells, as the first line of defence, may play a key role in PD. However, the molecular pathways in the host pocket and junctional epithelial cells that could regulate the ‘inflammatory’ host response in PD still remain largely elusive.

In the inflammatory infiltrate, the outcome of the cytokine response is regulated by the balance of the pro-inflammatory and anti-inflammatory cytokine levels (Dutzan et al., 2009). Transforming growth factor-β1 (TGF-β1), together with interleukin-10 (IL-10), serves as a major balancing anti-inflammatory cytokine in PD (Dutzan et al., 2009). The central anti-inflammatory surveillance role of TGF-β1 has been evidenced in studies with TGF-β1-knockout mice, which die soon after birth from massive infiltration of lymphocytes and macrophages into many organs (Shull et al., 1992; Kulkarni et al., 1993).

Integrin αvβ6 is an exclusively epithelial receptor with limited expression in adult tissues (Koivisto et al., 2018). It plays a major role in activating latent TGF-β1 bound to extracellular matrix or from the cell surface of immune cells by binding to the TGF-β1 latency-associated peptide, and thus has an essential role in innate anti-inflammatory surveillance in vivo (Annes et al., 2003; Taylor, 2009; Wang et al., 2012). We have shown previously that αvβ6 integrin is highly expressed in the junctional epithelium where it colocalizes with TGF-β1 (Ghannad et al., 2008). However, in human PD, αvβ6 integrin is strongly diminished in the pocket epithelium (Ghannad et al., 2008; Haapasalmi et al., 1995). Patients with mutations in the ITGB6 gene, which encodes the rate-limiting subunit for αvβ6 integrin heterodimer formation, have been reported to suffer from severe PD (Ansar et al., 2016). Furthermore, β6 integrin-null mice not only spontaneously develop PD but exhibit severe inflammation in the skin and lungs (Bi et al., 2019; Ghannad et al., 2008; Huang et al., 1996), emphasizing the critical importance of the αvβ6 integrin-mediated localized activation of TGF-β1 in the anti-inflammatory surveillance in vivo. Recently, we have also shown that the presence of αvβ6 integrin in the junctional epithelium may provide protection against periodontal inflammation through positive regulation of the AIM2 inflammasome and the anti-inflammatory IL-10 expression (Bi et al., 2019). Therefore, αvβ6 integrin may play a critical role in the maintenance of periodontal health.

The human ITGB6 promoter contains a binding site for TGF-β1 signaling mediator Smad3 (Xu et al., 2015). Consequently, endogenous TGF-β1 is required for sustained ITGB6 expression in gingival epithelial cells (GECs) and exogenously added TGF-β1 strongly upregulates its expression, indicating that there is a mutual positive feedback loop between these two molecules (Ghannad et al., 2008; Bi et al., 2017). However, signaling pathways that suppress ITGB6 expression in epithelial cells are not well understood. Other cytokines associated with PD, such as IL-1β, tumor necrosis factor-α and IL-6, lipopolysaccharides from the common periodontal pathogens Porphyromonas gingivalis, Tannerella forsythia and Treponema denticola and synthetic bacterial components, including triacylated lipopeptide Pam3CSK4, flagellin, iE-DAP (a peptidoglycan dipeptide) and lipoteichoic acid, all failed to suppress ITGB6 expression in these cells (Ghannad et al., 2008; Bi et al., 2017). However, extracts of mature, multispecies oral bacterial biofilms could specifically and dose dependently suppress the β6 integrin mRNA and protein expression in GECs, and promote the expression of inflammatory cytokines, such as IL-1β, potentially by interfering with the anti-inflammatory TGF-β1 signaling (Bi et al., 2017). FSL-1, a synthetic diacylated lipopeptide derived from Mycoplasma salivarium (a common component of oral microflora) downregulated ITGB6 expression in a similar manner (Bi et al., 2017). Moreover, suppression of β6 integrin expression alone by RNA interference stimulated IL-1β expression in these cells, suggesting a shared regulatory pathway (Bi et al., 2017).

In the present study, we explored the cellular signaling pathways contributing to the bacterial biofilm- and FSL-1-induced ITGB6 downregulation in GECs. We show that the suppression of β6 integrin and the upregulation of inflammatory cytokine expression were mediated by epidermal growth factor receptor (EGFR) signaling and required extracellular signal-regulated kinase 1 and 2 (ERK1/2, also known as MAPK3 and MAPK1, respectively; hereafter denoted ERK) activity. This downregulation was mediated by EGFR- and ERK-dependent phosphorylation, and inhibition of the TGF-β1 signaling mediator Smad3, which in turn attenuates TGF-β1-mediated maintenance of β6 integrin expression. We also show that blocking EGFR activity reduces alveolar bone loss and the inflammatory response in a ligature-induced experimental PD mouse model, supporting the functional relevance of EGFR in the etiology of PD in vivo.

Mitogen-activated protein kinases (MAPKs), including ERK, p38 MAPK family proteins (hereafter denoted p38) and c-Jun N-terminal kinases (JNKs), regulate a variety of cellular processes in epithelial cells, such as cell proliferation, adhesion, differentiation and migration (Huang et al., 2004b; Lee et al., 2016; Ihermann-Hella et al., 2014). Many oral bacteria activate MAPK pathways (Chung and Dale, 2004; Hasegawa et al., 2007). We investigated, therefore, whether MAPKs are activated by the biofilm extract in the GECs. Cells exposed to biofilm extract showed time-dependent, transient phosphorylation of ERK, p38 and JNK1 (also known as MAPK8) that peaked at ∼30 min of exposure (Fig. 1A), indicating activation of these three protein kinases families. Since MAPKs are commonly activated downstream of EGFR in epithelial cells (Shi et al., 2016), we also tested EGFR activation by the biofilm extract. This also showed activation-dependent Y1068 phosphorylation after a 30-min biofilm extract exposure (Fig. 1A). The synthetic bacterial lipopeptide FSL-1, which we have previously shown to regulate ITGB6 and IL1B expression similarly to biofilm (Bi et al., 2017), also induced MAPK and EGFR activation (Fig. 1A), indicating a shared mechanism of activation between this lipopeptide and the biofilms.

Next, we tested whether the suppression of β6 integrin expression by the biofilm extract involves EGFR and MAPKs by using chemical inhibitors [AG1478 for EGFR receptor kinase, PD184352 for the ERK upstream activator MEK1 (also known as MAP2K1), JNK-IN-8 for JNK1, SB203580 for p38 MAPKs] or a function-blocking anti-EGFR antibody. We found that biofilm-induced downregulation of ITGB6 was completely prevented by inhibition of either the ERK or JNK pathway, whereas inhibition of p38 had no effect (Fig. 1B). A chemical inhibitor of the EGFR kinase and a neutralizing antibody against EGFR ligand binding also blocked the biofilm-induced ITGB6 suppression (Fig. 1B), indicating that ITGB6 suppression by biofilm is MAPK- and EGFR-dependent in GECs. We then studied whether these pathways also regulate biofilm-induced IL1B stimulation, a key pro-inflammatory cytokine in PD. Inhibition of ERK, p38 or the EGFR pathway reduced IL1B upregulation by 40–70%, whereas inhibition of JNK had no effect (Fig. 1B). The results suggest that EGFR and ERK are part of the common mechanism involved in the regulation of both ITGB6 and IL1B by biofilm. ERK activation by biofilm was blocked in the presence of EGFR inhibitors, suggesting that ERK is activated downstream of EGFR (Fig. 1C).

To explore whether endogenously expressed EGFR ligands might play a role in EGFR–ERK-mediated regulation of ITGB6 and IL1B by biofilm, we first determined whether biofilm treatment regulates the expression of the major epithelial cell EGFR ligands, including heparin-binding EGF-like growth factor (HB-EGF), TGFα, amphiregulin and epiregulin (Chu et al., 2005). Cell treatment with either biofilm extract or FSL-1 significantly induced mRNA expression of all these ligands (Fig. 2A). While the induction of TGFA was modest, based on the real-time quantitative PCR (RT-qPCR) Cq values, it was already highly expressed in GECs prior to their treatment with biofilm or FSL-1 (Table S2). Consequently, we tested whether HB-EGF and TGFα, the two most highly expressed EGFR ligands, could regulate ITGB6 and IL1B in these cells similarly to biofilm. Both growth factors induced a similar, dose-dependent downregulation of ITGB6 and upregulation of IL1B (Fig. 2B), and also suppressed the expression of β6 integrin at the protein level (Fig. 2C), suggesting that biofilm-induced autocrine EGFR signaling is involved in mediating the effects of biofilm on ITGB6 and IL1B.

Integrin β6 expression in normal epithelial cells is primarily maintained by TGF-β1–Smad signaling, whereas in cancer cells other transcription factors, such as Stat3, may also play a role (Bi et al., 2017; Ghannad et al., 2008; Sullivan et al., 2011; Xu et al., 2015). Therefore, we tested the effect of chemical inhibitors of TGF-β1 receptor kinase (SB431542), Smad3 (SIS3) and Stat3 (Stattic) on the basal GEC ITGB6 expression. Inhibition of either the TGF-β1 receptor or Smad3 similarly reduced ITGB6 expression by ∼50%, whereas Stat3 inhibitor had no effect (Fig. 3A), suggesting that β6 integrin expression in GECs is largely dependent on TGF-β1 and Smad3. Treatment with TGF-β1 induced an ∼2–3-fold increase in ITGB6 expression in these cells that was dose dependently reduced by the addition of HB-EGF or TGFα (Fig. 3B,C), suggesting that EGFR ligands may interfere with TGF-β1–Smad3 regulation of ITGB6 expression.

EGFR–ERK signaling can inhibit Smad3 activity via ERK-mediated phosphorylation of the Smad3 linker region; the main ERK target being the S208 based on the EGF dose curve, time course of phosphorylation induction and other criteria (Matsuura et al., 2005). Therefore, we investigated whether biofilm, FSL-1 and the EGFR ligand TGFα induce phosphorylation of this residue in GECs. Treatment with any of these agents induced Smad3 S208 phosphorylation that peaked at 30 min (Fig. 3D). This phosphorylation was ERK-dependent, as a MEK1 inhibitor that prevented ERK activation also blocked the biofilm- and EGFR ligand-induced Smad3 phosphorylation (Fig. 3E,F), suggesting that the inhibitory Smad3 phosphorylation is indeed involved in the EGFR–ERK-mediated suppression of ITGB6 expression.

To confirm the essential role of Smad3 S208 phosphorylation in the biofilm- and EGFR ligand-mediated ITGB6 downregulation, GECs were transduced with recombinant adenoviruses to overexpress either the wild-type or S208A mutant of Smad3 that cannot be phosphorylated at this residue. Both the wild-type- and S208A mutant-transduced cells overexpressed Smad3 by ∼6-fold compared to cells transduced with viral particles carrying the empty vector only (Fig. 4A,B). Similarly, overexpression of either Smad3 construct resulted in an ∼3-fold induction of β6 integrin protein (Fig. 4A,B), indicating Smad3 regulation of ITGB6 expression. Treatment of cells expressing the S208A mutant of Smad3 exhibited significantly attenuated β6 integrin downregulation by biofilm or EGFR ligand treatment compared to cells overexpressing wild-type Smad3, indicating that phosphorylation of Smad3 S208 was indeed involved in the EGFR–ERK-mediated suppression of β6 integrin expression (Fig. 4C,D).

We have recently demonstrated aggravated bone loss and inflammation in mice in an experimental ligature PD model compared to nonligatured control animals (Bi et al., 2019); also shown in Fig. 5 (panels A and B for control; E and F for ligature) and Fig. 6 (panels A–D for control; I–L for ligature), respectively. In order to test whether this bone loss and inflammation could be alleviated by EGFR inhibition, we first screened an array of chemical EGFR inhibitors currently in clinical use or in clinical trials as cancer drugs (gefitinib, afatinib, erlotinib, lapatinib, dacomitinib and canertinib) (Roskoski, 2019) as well as the tyrphostin AG1478 that was used in the cell culture experiments, for their efficacy in inhibiting HB-EGF-induced ITGB6 suppression in GECs. All of these inhibitors dose dependently blocked β6 integrin downregulation by HB-EGF (Fig. S1). Next, we tested selected inhibitors in the ligature model, namely local application of AG1478 directly to the silk ligature and systemic administration of gefitinib by gavage. As evidenced by the micro-computed tomography (µCT) images, both inhibitors reduced periodontal bone loss compared to the ligature only treatment (Fig. 5C,D compared to E,F, respectively; Fig. 5G,H). The bone loss at the root furcation was also significantly reduced in the gefitinib-treated animals (Fig. 5G). Histological staining of decalcified tissue sections of the mouse jaw showed high levels of inflammatory cell accumulation in the ligatured animals in the absence of EGFR inhibitors (Fig. 6I–L), whereas in the presence of either inhibitor, the inflammatory cell infiltration was significantly reduced (Fig. 6E–H,M,N). Collectively, the data show that EGFR inhibition significantly reduces both bone loss and periodontal inflammation in the experimental mouse periodontitis model, suggesting that EGFR inhibition is potentially an effective treatment for periodontitis.

Expression of αvβ6 integrin in the junctional epithelium is essential for periodontal health (Koivisto et al., 2018). In PD, αvβ6 integrin is downregulated, but the mechanism is not well understood. Our previous studies have shown that TGF-β1–Smad3 signaling sustains the expression of this receptor in healthy junctional epithelium and that oral bacterial biofilm extracts suppress β6 integrin mRNA and protein expression in GECs by interfering with this pathway (Bi et al., 2017). The data from the present study point to substances that activate EGFR and MAPK pathways as the likely culprits for downregulation of αvβ6 integrin during initiation and progression of PD. During this process, EGFR–ERK signaling may be further sustained by the induction of autocrine EGFR ligands. It seems also likely that both the β6 integrin downregulation and the pro-inflammatory IL-1β stimulation are regulated through these same signaling pathways, as biofilm-induced cytokine responses were also prominently attenuated by blockers of EGFR and ERK signaling. While MAPK pathways have previously been indicated in the GEC immune response and in their pro-inflammatory cytokine stimulation by oral bacteria (Watanabe et al., 2001; Krisanaprakornkit et al., 2002; Huang et al., 2004a; Hasegawa et al., 2007), the critical involvement of EGFR–ERK signaling in αvβ6 integrin suppression by oral bacterial biofilm components is a novel finding. Illustration of the putative signaling pathways regulating αvβ6 integrin downregulation and inflammatory cytokine upregulation is presented in Fig. 7.

FSL-1 represents the N-terminal lipopeptide of the cell membrane lipoprotein LP44 of the common oral bacterium M. salivarium (Shibata et al., 2000). We have previously shown that FSL-1 suppressed ITGB6 and upregulated ILB1 expression similarly to the biofilm extracts in GECs, whereas various other bacterial components failed to suppress ITGB6 expression (Bi et al., 2017; Ghannad et al., 2008). Here, we show that it also activates EGFR–MAPK signaling and induces autocrine EGFR ligand expression in a similar manner to the biofilm extract. Toll-like receptors (TLRs) play a key role in oral pathogen signaling and can transactivate EGFR (Koff et al., 2008; Ebersole et al., 2013; Shaykhiev et al., 2008). FSL-1 is typically considered an agonist for TLR2 and TLR6 (Kang et al., 2009). However, blocking the function of this receptor with specific antibodies did not diminish the ability of either FSL-1 or the biofilm extract to suppress GEC ITGB6 expression (Bi et al., 2017), suggesting alternative recognition sites on host cell surface. Thus, the mechanism of how FSL-1 activates EGFR in GECs remained to be determined. Interestingly, the N-terminal peptide of lipoprotein p37 from another mycoplasma strain, M. hyorhinis, has been shown to form a cell surface complex with EGFR and annexin A2, inducing EGFR activation in cultured human gastric cancer cells (Duan et al., 2014a; Duan et al., 2014b), suggesting a direct interaction with EGFR is also a possibility.

Both biofilm extract and EGFR ligands were able to attenuate the stimulatory TGF-β1 effect on ITGB6 expression in GECs (this study and Bi et al., 2017), suggesting EGFR-mediated repression of TGF-β1–Smad3 signaling. EGFR signaling inhibits Smad3 activity via ERK-mediated phosphorylation of the Smad3 linker region, with S208 being the primary ERK target site (Matsuura et al., 2005). In our study, both biofilm extract and EGFR ligands induced ERK-dependent Smad3 S208 phosphorylation, and overexpression of the phosphorylation-defective S208A mutant of Smad3 significantly attenuated the ability of biofilm or EGFR ligands to suppress β6 integrin expression, highly suggesting that ERK-mediated Smad3 S208 phosphorylation is an essential mechanism for ITGB6 downregulation. EGFR–ERK signaling might also attenuate TGF-β1–Smad3 functions by other means, for example by ERK-mediated phosphorylation and stabilization of the Smad2/3 co-repressor TGIF (Yang et al., 2008; Liu et al., 2014). While both EGFR ligands and biofilm extract increased TGIF protein levels in GECs, this increase was not affected by EGFR or ERK inhibition (L.K., unpublished results), suggesting that this was not the primary mechanism for EGFR–ERK-mediated ITGB6 suppression in these cells, at least in vitro. The role of TGIF accumulation in the suppression of αvβ6 integrin expression in PD needs to be further studied in vivo. It should be noted that TGF-β1 itself can also activate ERK (Zhang, 2009). However, the TGFβ receptor inhibitor SB431542 had no inhibitory effect on biofilm-induced ERK activation (L.K., unpublished results), making it unlikely that at least the initial ERK activation was mediated by TGFβR. Unraveling the crosstalk between the EGFR and TGF-β1 pathways will be important in understanding how αvβ6 integrin participates in the regulation of the inflammatory response and to identify novel targets for the treatment of PD.

Epithelial tight junctions have been shown to be compromised in periodontal lesions and, consequently, translocation of harmful bacteria into gingival tissue is significantly increased in these lesions compared to healthy sites, eliciting a defensive host inflammatory response that can induce alveolar bone loss (Choi et al., 2014; Jiao et al., 2014). As shown in airway epithelia, increased epithelial permeability may result in chronic autocrine EGFR receptor activation and a wound healing response that can contribute to the tissue remodeling that is a hallmark of PD (Vermeer et al., 2003). PD may thus result from the capacity of the bacterial components to redirect host responses towards those that are chronic and detrimental to the health of tissues and cells, resulting in PD as collateral damage from a dysregulated host response (Ebersole et al., 2013). Therefore, PD may represent a survival response gone awry, and periodontal health could be restored by rebalancing the TGF-β1–EGFR equilibrium.

In addition to its anti-inflammatory functions, TGF-β1 is a negative regulator of epithelial cell proliferation (Ten Dijke et al., 2002; Bretones et al., 2015). Integrin αvβ6-mediated TGF-β1 activation limits epithelial cell growth, and the loss of this receptor may lead to epithelial hyperplasia (Ludlow et al., 2005; Xie et al., 2012). Conversely, EGFR–ERK signaling promotes epithelial cell growth and migration (Sun et al., 2015; Wang et al., 2019). Therefore, the extension of pocket epithelium both downwards and into the gingival connective tissue during the progression of human PD may be caused by an imbalance between the TGF-β1 and EGFR signaling in the pocket epithelium. Increased EGFR expression and activation has been detected in several types of lesions that exhibit aberrant epithelial cell proliferation and saw tooth-like epithelial penetration into the underlying connective tissue, such as oral leukoplakia, oral lichen planus and skin psoriasis (Wang et al., 2019; Cortes-Ramirez et al., 2014; Ribeiro et al., 2012). Interestingly, patients with psoriasis undergoing cancer treatment with EGFR inhibitors showed remarkable therapeutic improvement of their chronic psoriasis lesions, indicating aberrant EGFR activity as a contributing factor in this skin condition (Wang et al., 2019). However, EGFR expression and activation in PD is not well studied. EGFR expression is low or absent in healthy junctional epithelium, whereas some studies have reported increased expression in inflamed gingiva and in the pocket epithelium of PD compared to healthy junctional epithelium (Nordlund et al., 1991; Irwin et al., 1991). The EGFR activation status and autocrine epithelial EGFR ligand expression in the pocket epithelium will be further studied but it is likely to be responsible for the promotion of the increase in cell proliferation and apical migration of junctional epithelium that are observed in the progression of PD similar to other conditions of epithelial hyperplasia.

We have shown that EGFR inhibition provided a significant reduction in periodontal inflammation and alveolar bone loss in the experimental mouse periodontitis model. EGFR inhibition may thus provide an attractive treatment target for human PD. Several EGFR inhibitors have already been approved for human use as a treatment for epithelial cancers (Roskoski, 2019). Although their systemic administering can produce side effects (Guggina et al., 2017), their local use for restoring the αvβ6 integrin levels in the junctional epithelium could be used to treat periodontal inflammation and bone loss in patients with PD.

Reagents used for cell treatment were: FSL-1 (Pam2CGDPKHPKSF; a synthetic diacylated lipopeptide derived from Mycoplasma salivarium; Abcam, Cambridge, MA); EGFR ligands TGFα (GenWay Biotech, San Diego, CA) and HB-EGF (R&D Systems, Minneapolis, MN) and TGF-β1 (EMD Millipore, Etobicoke, ON); and the EGFR inhibitors, function-blocking anti-EGFR antibody (#GR13L, clone 225, Calbiochem, San Diego, CA), tyrphostin AG1478 (Sigma-Aldrich Canada, Oakville, ON), gefitinib and afatinib (both Toronto Research Chemicals, Toronto, ON, Canada), erlotinib (Cayman Chemicals, Ann Arbor, MI), lapatinib ditosylate (Phoenix Pharmaceuticals, Burlingame, CA), dacomitinib and canertinib dihydrochloride (both LC Laboratories, Woburn, MA). Other cellular signaling inhibitors were: PD184352 (an inhibitor of ERK activation; Axon Medchem BV, Groningen, The Netherlands), JNK-IN-8 (a JNK inhibitor) and SB431542 (a TGF-β receptor inhibitor; both Selleckchem Chemicals, Houston, TX), SB203580 (a p38 MAPK inhibitor; New England Biolabs, Whitby, ON), SIS3 (Smad3 inhibitor; BioVision, Milpitas, CA) and Stattic (Stat3 inhibitor; R&D Systems). Primary antibodies used for western blotting were: anti-pERK1/2 (T202/Y204; #4370; 1:2000), anti-p38 MAPK (#9228; 1:1000) and anti-pEGFR (Y1068; #3777; 1:1000) from Cell Signaling Technology (Danvers, MA); anti-ERK1/2 (sc-154-G; 1:500), anti-JNK1 (sc-571; 1:1000) and anti-EGFR (sc-373746; 1:500) from Santa Cruz Biotechnology (Dallas, TX); anti-active p38 MAPK (V121A; 1:2000) and anti-active JNK (V7931; 1:5000) from Promega (Madison, WI); anti-β6 integrin (AF2389; 1:3000) from R&D Systems; anti-β-tubulin (MAB3408; 1:2000) from EMD Millipore and Smad3 (ab283790; 1:2000) from Abcam; and anti-phospho-Smad3 S208 produced as described (Matsuura et al., 2005; used at 0.15 µg/ml). All PCR primers were synthesized by Integrated DNA Technologies (Coralville, IA) and are listed in Table S1.

Procedures involving human biofilm donors were reviewed and approved by the Clinical Research Ethics Board at the University of British Columbia (Protocol number H17-02478). Oral bacterial biofilms were grown, and biofilm extracts prepared as previously described in detail (Bi et al., 2016). Briefly, subgingival plaque from the first and second upper molars was collected from healthy volunteers in brain heart infusion broth (Difco, Detroit, MI), dispersed, seeded on collagen-coated (bovine dermal type I collagen; Cohesion, Palo Alto, CA) sterile hydroxyapatite discs (Clarkson Chromatography Products, Williamsport, PA) and incubated under anaerobic conditions (AnaeroGen; Oxoid, Cambridge, UK) at 37°C. The growth medium was changed once a week. After 21 days, the discs were rinsed once with phosphate-buffered saline (PBS; pH 7.4), and the bacteria then dispersed in PBS by pipetting, snap-frozen, ground over liquid nitrogen using a mortar and pestle and sonicated with Branson Sonifier 250 on ice. The non-solubilized bacterial debris was removed by centrifugation at 10,000 g for 10 min and the supernatant stored at −20°C for later use. The total biofilm extract protein concentration was quantified with Bio-Rad protein assay reagent (Bio-Rad Laboratories, Hercules). The biofilm extract was heated for 5 min at 95°C before being used for cell treatment as previously described (Bi et al., 2017).

The spontaneously immortalized human gingival epithelial cell line (GEC; Mäkelä et al., 1998) was maintained in a humidified incubator at 37°C in an atmosphere of 5% CO₂ in Dulbecco's modified Eagle's medium (DMEM, Gibco, Life Technologies, Inc., Burlington, ON) supplemented with 23 mM sodium bicarbonate, 20 mM HEPES, 1% antibiotics (50 µg/ml streptomycin sulfate, 100 U/ml penicillin; Gibco) and 10% heat-inactivated fetal bovine serum (FBS; Gibco) as previously described (Bi et al., 2017).

GECs were seeded into cell culture dishes (5×10⁵ cells/cm²) and allowed to grow to confluency for 2 days in their complete growth medium, rinsed once with PBS and switched to FBS-free DMEM for 2 h in the presence or absence of cellular signaling inhibitors (2 µM PD184352, 1 µM JNK-IN-8, 20 µM SB203580, 300 nM AG1478, 2 µg/ml anti-EGFR, 15 µM SB431542, 3 µM SIS3 or 10 µM Stattic), prior to adding biofilm extract (60 µg/ml of protein), FSL-1 (100 ng/ml), TGFα (1–10 ng/ml), HB-EGF (1–10 ng/ml) or TGF-β1 (2 ng/ml) to the cells. The cells were further cultured for 32 h or 48 h, rinsed with PBS and used for RNA isolation and RT-qPCR or for western blotting, respectively, as described in more detail below. To investigate cellular signaling pathway activation, the cells were grown, pre-treated with inhibitors and then stimulated as above for 0–120 min, and their phosphorylated protein to total protein expression ratio was determined by western blotting and subsequent quantification.

The RNA isolation and RT-qPCR were performed as previously described (Bi et al., 2016). Briefly, total RNA was extracted using the NucleoSpin RNA II kit (Macherey-Nagel, Bethlehem, PA) and RNA concentration measured by spectrophotometry. Then, 1 µg of total RNA was reverse-transcribed into cDNA using an iScript Select cDNA synthesis kit (Bio-Rad). The cDNA was diluted to a concentration for which the threshold-cycle value was well within the range of its standard curve. 5 µl of cDNA was mixed with 10 µl of 2×iQ SYBR Green I Supermix (Bio-Rad) and 5 pmol of primers in each well of the 96-well plate. Asparagine-linked glycosylation 9 (ALG9), beta-2-microglobulin (B2M) and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) were used as reference genes to conduct amplification reactions of the target genes in triplicates. Real-time PCR amplification was performed with the CFX96 system (Bio-Rad). The running program was 3 min at 95°C, followed by 45 cycles of 15 s at 94°C, 20 s at 60°C and 20 s at 72°C. The data were analyzed by CFX Manager Software Version 2.1 (Bio-Rad). PCR detection level was adjusted to ≥1.5-fold to reflect significant change. The mean Cq values lower than 33 were included in the gene expression analyses.

Western blotting was performed as previously described (Bi et al., 2016). The cells were washed once with PBS and lysed in 1x Laemmli sample buffer. The cell lysates were separated by SDS/PAGE and transferred onto Hybond Protran membrane (Amersham, Little Chalfont, Buckinghamshire, UK). The membranes were incubated in Odyssey Blocking Buffer (LI-COR Biosciences, Lincoln, NE) for 1 h, followed by incubation with primary antibodies for overnight at +4°C. After incubation with species-appropriate IRdye-conjugated secondary antibodies (LI-COR) for 1 h, the blots were washed and scanned with LI-COR Odyssey Infrared Imaging system. The results were analyzed with Odyssey software (version 3.0).

The recombinant adenoviruses were produced using AdEasy Adenoviral Vector System (Stratagene, La Jolla, CA). Briefly, the LZRSΔ-IRES-GFP retroviral vectors carrying the full-length human Smad3 cDNA or its mutant form with a mutation of serine to non-phosphorylatable alanine residue at position 208 (S208A) (Matsuura et al., 2005) were digested with EcoRI and blunted with Klenow fragment. The sequences encoding wild-type or S208A Smad3 were then ligated into SalI-digested, Klenow fragment-blunted pShuttle-CMV to generate pShuttleSmad3 and its mutant, respectively. The resulting plasmid constructs, and pShuttle as an empty vector control, were linearized with PmeI, dephosphorylated and then co-transformed into Escherichia coli BJ5183 strain with pAdEasy-1 to generate their corresponding recombinant viral plasmids. These recombinant plasmids were further amplified in E. coli XL10-Gold strain and verified with PacI digestion. The plasmids were then transfected into human HEK293 embryonal kidney cells with Lipofectamine (Invitrogen, Carlsbad, CA, USA) to generate the respective replication-deficient human serotype 5 adenoviruses. Subsequently, these viruses were propagated and amplified in the HEK293 cells, pooled and recovered in FBS-free DMEM through four cycles of freeze thawing. Viral titers were determined by plaque forming units (pfu).

To verify adenoviral overexpression of Smad3, GECs were trypsinized, and suspended in serum-free DMEM. The cells were mixed with the recombinant adenoviruses (10 pfu/cell), gently shaken at room temperature for 30 min and then plated at 5×10⁵ cells/cm². After a 3-hour incubation and attachment at 37°C (5% CO₂), FBS was added to 10%. After 48 h, the cells were analyzed for their Smad3 and β6 integrin levels by western blotting. In a set of experiments, the adenovirus-transduced cells were further treated with EGFR ligands and biofilm extract for 48 h and analyzed for β6 integrin expression by western blotting.

GECs were seeded into 96-well plates (5×10⁵ cells/cm²) in their complete growth medium for 24 h, washed once with PBS and treated with TGF-β1 (2 ng/ml) for 24 h in DMEM containing 1% FBS. The cells were then washed with PBS and incubated in FBS-free medium for 2 h in the presence or absence of EGFR inhibitors (0.003–3 µM), followed by addition of HB-EGF (5 ng/ml) for 48 h. The cells were fixed with formaldehyde fixative [PBS containing 4% (v/v) formaldehyde and 5% (w/v) sucrose] for 20 min, permeabilized in PBS containing 0.5% Triton X-100 for 5 min, blocked with Odyssey Blocking Buffer for 1 h and then incubated with anti-β6 integrin antibody for overnight at +4°C, followed a species-appropriate IRdye800-conjugated secondary antibody (LI-COR) and CellTag700 (LI-COR) staining for 1 h. The wells were then washed and scanned with LI-COR Odyssey Infrared Imaging system. The results were analyzed with Odyssey software (version 3.0) relative to the cell numbers (CellTag700) in each well.

All animal experiments were reviewed and approved by the University of British Columbia Committee on Animal Care (Protocol number A16-0034). A total of 24 mice were used for the experiment (12 mice each for local treatment group and systemic treatment group; 6 mice for treatment with drug and 6 control mice without drug in each group). To induce experimental periodontitis, silk sutures (6-0; Ethicon, San Lorenzo, PR, USA) were tied around the left second maxillary molars on 9-week-old FVB/NHsd mice as we have done previously (Bi et al., 2019). The right-side maxillary molar of the same animal was left untied as a control. For local EGFR inhibition, the silk was first soaked in ethanol with or without tyrphostin AG1478 (10 µM) (Johns et al., 2003) at +4°C for overnight. The inhibitor-infused sutures were changed every second day for 14 days. For systemic EGFR inhibition, another set of mice were gavaged with 1% Tween-80 in PBS (pH 7.4) with or without gefitinib (100 mg/kg body weight/dose) (Zhao et al., 2013) every second day for 14 days. The silk sutures were tied a day after the first gavage and retained in place until the end of the experiment. Maxilla specimens were then collected and used for analysis of bone loss and histological assessment by micro-computed tomography (μCT) and histochemistry, respectively.

Maxilla specimens were fixed in 4% formaldehyde (v/v) in PBS for 24 h at +4°C and then transferred into 2% formaldehyde in PBS for µCT scanning (10 µm voxel size; Scanco CT 40; Scanco, Wayne, PA). Alveolar bone loss was evaluated from the horizonal-angulated µCT images with Microview Software (Parallax, Ilderton, ON, Canada). The distances from the cemento-enamel junction to the alveolar bone crest in mesial and distal sites of the ligatured second molar were measured and averaged.

After µCT scanning, the maxilla specimens were decalcified in PBS containing 2% formaldehyde and 0.4 M EDTA for 5 weeks. The specimens were paraffin-embedded and sectioned (8 µm) in the mesio-distal direction followed with hematoxylin and eosin staining. A sliding visual score was used to score the level of inflammation between the first and second molars (0, no inflammation; 1, mild; 2, moderate; 3, severe). The scores from three investigators that were blind to the experimental conditions were averaged.

All cell experiments were separately repeated for at least three times. Statistical comparisons between two experimental groups were analyzed by unpaired, two-tailed Student's t-test. Multiple comparison tests were performed using one-way ANOVA followed by post hoc comparison with Tukey–Kramer multiple comparisons test or least significant difference test (LSD). Statistical significance was set at P<0.05. Statistical significance for RT-qPCR data (performed in triplicates) was calculated using log2-transformed data.