Porphyromonas gingivalis bypasses epithelial barrier and modulates fibroblastic inflammatory response in an in vitro 3D spheroid model

Porphyromonas gingivalis-induced inflammatory effects are mostly investigated in monolayer cultured cells. The aim of this study was to develop a 3D spheroid model of gingiva to take into account epithelio-fibroblastic interactions. Human gingival epithelial cells (ECs) and human oral fibroblasts (FBs) were cultured by hanging drop method to generate 3D microtissue (MT) whose structure was analyzed on histological sections and the cell-to-cell interactions were observed by scanning and transmission electron microscopy (SEM and TEM). MTs were infected by P. gingivalis and the impact on cell death (Apaf-1, caspase-3), inflammatory markers (TNF-α, IL-6, IL-8) and extracellular matrix components (Col-IV, E-cadherin, integrin β1) was evaluated by immunohistochemistry and RT-qPCR. Results were compared to those observed in situ in experimental periodontitis and in human gingival biopsies. MTs exhibited a well-defined spatial organization where ECs were organized in an external cellular multilayer, while, FBs constituted the core. The infection of MT demonstrated the ability of P. gingivalis to bypass the epithelial barrier in order to reach the fibroblastic core and induce disorganization of the spheroid structure. An increased cell death was observed in fibroblastic core. The development of such 3D model may be useful to define the role of EC–FB interactions on periodontal host-immune response and to assess the efficacy of new therapeutics.


Microtissues formation.
To generate MTs, the hanging-drop culture method was used. MT is characterized by a 3D spheroid structure where cells are in direct contact, allowing cell-to-cell and cells-to-extracellular matrix components (ECM) interactions. This method can be used to co-culture two (or more) different cell populations to elucidate the role of cell-cell or cell-ECM interactions in a 3D environment 10,12 (Fig. 1B). It also allows the addition of very small quantities of any biological agent or drug in the cell culture medium. Initially, 3 different concentrations of FBs (3 × 10 3 , 7 × 10 3 and 10 × 10 3 ) were cultured in suspension within a droplet of 40 μl of cells and medium in a 3D culture plate (GravityPLUS TM 3D Culture, InSphero AG, Zürich, Switzerland) for 5 days, until the first true spheroid was visualized. Then, 20 μl of Defined Keratinocyte-SFM basal medium (KSFM) supplemented medium containing ECs (1.5 × 10 3 , 5 × 10 3 and 7 × 10 3 ) was added in each droplet of the suspension for 5 more days respectively (Fig. 1B,C). After testing different MT sizes, a concentration of 7 × 10 3 FBs + 5 × 10 3 ECs was selected for further experiments. Every 2 days, 5 μl of both cell media were added to maintain the viability and physiological conditions of the cells, preventing the drop from desiccation and consequent disaggregation. Infection with P. gingivalis and stimulation with its LPS and heat-killed P. gingivalis. Regarding monolayer infection, 2 × 10 5 cells were plated in each well of a 24-wells plate. At the day of the experiment, cells were washed twice with PBS and were either infected for 24 h with P. gingivalis at a multiplicity of infection (MOI) of 100 or stimulated with ultrapure P. gingivalis LPS (1 μg/ml) (InvivoGen, San Diego, CA, USA). To ensure that effects are induced by P. gingivalis invasion, after 2 h of infection, the medium was removed and replaced by the same volume of bacteria free culture medium.
Regarding MT infection and stimulation, MTs were first collected from 3D culture plate. Then, MTs were infected with P. gingivalis at MOI 100, heat-killed P. gingivalis or stimulated with its LPS (1 μg/ml) (InvivoGen, San Diego, CA, USA). To inactivate bacteria, P. gingivalis was incubated for 15 minutes at 85 °C prior to the experiment. The MTs suspended in a 96-Well plate (CytoOne, Orsay, France) were infected for 2 to 24 h. As described for monolayer cells infection, after 2 h, the medium was replaced by a bacteria-free culture medium. To ensure P. gingivalis invasion, after 2 h of contact with cells or MTs, cell medium was replaced with a P. gingivalis-free medium. Furthermore, an antibiotic protection assay has been performed (Supplementary Figure 1) that confirmed the invasiveness of P. gingivalis in both cell types and MT.
RNA isolation and reverse transcription. Total RNA from gingival samples and cells were extracted using the High Pure RNA Isolation Kit (Roche Diagnostics, Meylan, France) according to the manufacturer's instructions. The total RNA concentration was quantified using NanoDrop 1000 (Thermofisher, Illkirch, France). 25 ng of RNA from each sample were used for Reverse transcription. Reverse transcription was performed using iScript Reverse Transcription Supermix (Biorad, Miltry-Mory, France) according to the manufacturer's instructions. (B) Microtissue spheroids formation diagram. Firstly, FBs were seeded in a droplet of culture medium for 5 days. After the formation of a spheroid, ECs were seeded over it. After 5 more days, two-cell types MTs were constituted. During all this procedure, cell culture medium was changed every 2 days. (C) Morphological characteristics of 1-Step and 2-Steps cell culture technique for MT spheroids. Three different cell concentrations (10 × 10 3 FBs + 7 × 10 3 ECs/8 × 10 3 FBs + 5 × 10 3 ECs/3 × 10 3 FBs + 1.5 × 10 3 ECs) were tested. In the 1-step procedure, both cell types were seeded simultaneously in a droplet while for the 2-steps procedure, each cell type was seeded in a time-specific manner. (D) Apoptosis quantification. Annexin V/IP staining of MTs at different cell concentrations showed a slight increase of cell death mainly located at the core of the MT, especially, in the biggest MT. Quantitative real-time PCR analysis. To quantify mRNA expression, qPCR was performed on the cDNA samples. PCR amplification and analysis were performed with CFX Connect TM Real-Time PCR Detection System (Biorad, Miltry-Mory, France). Amplification reactions were performed using iTAq Universal SYBR Green Supermix (Biorad, Miltry-Mory, France). Beta-actin was used as endogenous RNA control (housekeeping gene) in all samples. Primer sequences related to Bcl-2, Bax-1, Integrin β1, Apaf-1, Tnf-α, Il-6, Il-8, Col IV and β-actin were purchased from Qiagen, (Courtaboeuf Cedex, France) and ThermoFischer (Illkirch, France) (Supplementary Table 1). Expression level was calculated after normalization to the housekeeping gene expression.
Immunofluorescence. Immunofluorescence has been performed on sections of MTs and the tissues har-   Statistical analysis. Statistical analysis was performed using pair-wise Anova test and post-hoc Tukey's test. Statistical significance level was considered for p < 0.05. Data were analyzed using PRISM 6.0 (GraphPad, La Jolla, CA, USA). All experiments have been performed at least in triplicate (biological and technical triplicates).

Morphological characteristics of formed MT.
To generate gingiva-like MTs, the hanging-drop method was used to sequentially seed the FBs in the first step followed by ECs as a second layer (Fig. 1B). This sequential seeding allows the formation of well-organized MT exhibiting two distinct layers at contrary to simultaneous seeding of both cell types (Fig. 1C). Different cell concentrations for FBs and ECs were also tested ( Fig. 1D) to assess impact on MT size and cell death at the core of the MT. The increase of the number of cells seeded at each step increased MT size. However, this increased size (until 200-300 μm) did not induce significant cell death at the core of the MT as observed with immunostaining with Annexin V (Fig. 1D). Formed MT exhibited a well-defined spatial organization where ECs were organized in an external cellular multilayer while FBs constituted the core of the MT ( Fig. 2A,B). This distinct cellular pattern was confirmed by immunostaining with E-cadherin, pan cytokeratin and cytokeratin 14 observed at the periphery, while vimentin was mainly detected at its core ( Fig. 2I-K,N,O). In addition, laminin αV was localized mainly at the ECs layer (Fig. 2M).
The expression and localization of the same proteins were evaluated in an experimental periodontitis model ( Fig. 3C-H). In healthy site, collagen IV and integrin β1 were localized from the epithelium-connective tissue interface until the alveolar bone border (Fig. 3C-H). Laminin αV was mainly detected at the epithelium-connective tissue interface (Fig. 3M,N). Vimentin was used as a connective tissue marker and was mainly expressed in connective tissue (Fig. 3K,L). These results confirmed the relevance of the 3D model to evaluate the mechanisms involved at the epithelial-fibroblast interface during periodontitis. P. gingivalis influenced epithelial barrier integrity. Invasion of P. gingivalis was evaluated by immunofluorescence in both 3D and monolayer culture. In spheroid model, invasion of MT by P. gingivalis was detected after 24 h (Fig. 2R,S). Interestingly, P. gingivalis was not only detected in the epithelial compartment but also within the fibroblastic core, demonstrating the ability of P. gingivalis to bypass the epithelial barrier and to disseminate within MT (Fig. 2S), hence, confirming the capability of P. gingivalis to invade both cell types (Fig. 2U-X observed in situ (Fig. 3O,P). Interestingly, localization of inactivated P. gingivalis (HPg) was not observed within the core of the MT highlighting the role of invasion in this process ( Fig. 2L,P,T).

Infection of MTs with P. gingivalis induced disorganization and destruction of the spheroid structure.
To evaluate the effect of P. gingivalis infection on MT structure, 3D morphometric changes were followed after infection (Fig. 4) and compared to uninfected MT at 6 and 24 h. SEM analysis showed that P.gingivalis was able to adhere and colonize the external epithelial surface of the MT and to be internalized  Fig. 4C). Furthermore, bacterial infection induced EC exfoliation (Fig. 4D). At 6 and 24 h, infected MT exhibited a collapsed shape associated with signs of cell injury and cell death illustrated by the detection of apoptosis (Fig. 4G,H). This observed collapsed shape seemed to be associated with the destruction of the core of the MT. Interestingly, at 24 h, infected MT exhibited signs of total destruction and some FBs could be observed even at the surface of the MT (Fig. 4H). Such destruction was not observed when MT were only stimulated with P. gingivalis-LPS even though structural modifications could be observed at the surface (Fig. 4F).
Changes induced by infection on structural organization of the MT were also observed using TEM. Uninfected MT displayed several hallmarks of a fibroblastic core surrounded by a healthy epithelial barrier (Fig. 5A-D). Several epithelial junctions have been observed such as desmosomes, adherens junctions and long tight junctions between ECs (Fig. 5C,D). However, infection of MT by P. gingivalis induced cellular stress highlighted by the presence of apoptotic bodies correlated with bacterial invasion of ECs at 2 h (Fig. 3E-H). The effect of the infection was also visible on the integrity of the epithelial barrier illustrated by the reduction of the number of cell junctions, especially located at the most external and superficial part of ECs (Fig. 5H). At 24 h, the full disruption of the epithelial barrier, multiple apoptotic cells and an acute cellular stress in all strata of the MT could be observed ( Fig. 5M-P). This was confirmed by immunofluorescence staining of integrin β1 which demonstrated a decreased expression at the epithelial level in response to infection (Fig. 6J,K). This pattern of expression was directly correlated to the P. gingivalis invasion pattern that has been shown previously by immunofluorescence and SEM (Figs 3M-O and 4).

MT collapse correlated with P. gingivalis-induced apoptosis.
To determine how P. gingivalis affects the different compartments of the MT, a focus was made on induced cell death using Annexin V/ IP staining. In comparison with the uninfected MT, at 24 h, a strong Annexin V staining was observed within the fibroblastic  (Fig. 6B) explaining the observed collapse (Figs 4 and 5). Interestingly, a differential expression of apoptosome APAF-1 and caspase-3 was simultaneously observed (Fig. 6C-H). Herein, in P. gingivalis-infected MT, an increased APAF-1 and caspase-3 expression was observed in FBs that emphasized the role of APAF-1 apoptosome in the activation of apoptosis. At contrary, APAF-1 detection within the ECs layer in infected MT was low (Fig. 4D,F).

Infection with P. gingivalis modulated differentially the inflammation and apoptosis associated gene expression.
To evaluate the pattern of gene expression associated with inflammation, apoptosis and epithelial integrity, mRNA expression in MT in response to P. gingivalis infection was measured and compared to that of monolayer cell culture. Regarding inflammation, infection of MT increased TNF-α, IL-6 and IL-8 expression (Fig. 7A-C). Interestingly, same trend of results was observed in monolayer cell cultures. Effect on apoptosis has also been evaluated focusing on Bcl-2/Bax-1 and Apaf-1. In ECs, infection with P. gingivalis displayed anti-apoptotic effects through a decrease of Bax-1 and Apaf-1 and an increase of anti-apoptotic Bcl-2 expression in EC (Fig. 7D-F). In FBs, P. gingivalis infection induced opposite effects as it increased the expression of pro-apoptotic markers (Fig. 7D-F). The gene expression of MT was closer to that of FBs. The breakdown of the epithelial barrier was also confirmed as mRNA expression of integrin-β1 was significantly decreased by infection (Fig. 7G).
To evaluate if the mRNA expression pattern observed in infected MT was comparable to that expressed in diseased tissues, mRNA expression of the same genes was evaluated in human gingival tissues harvested from healthy (H) and chronic periodontitis (CP) patients (Fig. 8). Analysis revealed that the expression of TNF-α, IL-6, IL-8, Bax-1 and Apaf-1 mRNAs was significantly increased in the CP group (TNF-α, 4.7-folds; IL-6, 3.85-folds; IL-8, 2.5-folds; Bax-1, 3-folds; Apaf-1, 3.9-folds) (p < 0.05) (Fig. 8A-D,F) while the anti-apoptotic Bcl-2 expression was decreased (1.9-folds) (Fig. 8E). As observed in MT, mRNA expression of integrin β1 was also increased significantly in CP (Fig. 8G). These results confirmed that the modulation of mRNA expression by P. gingivalis in MT is similar to the one observed in human diseased tissues.

Discussion
In this study, we designed a spheroid 3D in vitro model of gingival tissue and assessed the destructive effects associated with P. gingivalis infection. The well-organized 3D structure of the synthesized MT mimicked the epithelial-fibroblastic barrier and allowed investigation of the P. gingivalis-induced effects associated with bacterial invasion and to focus on the epithelial barrier disruption and inflammation. It has already been described that the synthesis of 3D MT using hanging-drop method is a reliable technique to generate homogenous organs or tissue-like structures at low-cost without requiring specific growth factors or complicated technical procedures. This type of spheroid model has been used previously to determine the physiopathological processes associated with diseases such as cancers, cellular organization during embryogenesis, drug testing and also as potential therapeutics such as demonstrated in the context of dental pulp regeneration 10,17,34-36 . The use of this model is of interest to evaluate the effects associated with the infection by P. gingivalis or by other periodontal pathogens while taking into account the epithelio-fibroblastic interactions. This coupling in 3D maximizes cell-to-cell communication and signaling that is critical to cell function. Furthermore, the phenotype or function of cells grown in 3D is more complex and closer to the functions of human tissues compared to the cells grown in 2D 37 . Hypoxia can be a physiologically relevant phenomenon considered as a major characteristic of 3D microenvironments both in vitro and in vivo. Oxygen concentration in 3D tissues depends on the balance between oxygen delivery and consumption 38 . In cancer models, hypoxic conditions need to be included in the 3D scaffold design and this can be done by strictly regulating the size of the MT. It has been shown that oxygen can diffuse across 100-200 μm of tissue thickness and it is generally desirable to maintain the optimal aggregate size of approximately 250 μm to prevent hypoxia 39 . In this study, the number of cells seeded was standardized to an ideal size for which the cells in the core of the uninfected MT remained healthy and unaffected by hypoxia. However, a modification of the protocol of MT synthesis may be done, through an increase in the number of the cells seeded, to evaluate the effect of hypoxia on cells response to infection as this phenomenon was demonstrated to be deleterious to oral health 40 . Such 3D model may also be interesting in the future to evaluate the effects of anti-infectious or anti-inflammatory drugs while considering tissular penetration.
Several other 3D models of oral mucosa have already been developed 9,[41][42][43] . Most of them are collagen-based oral mucosal models (OMM) and displayed efficiently the characteristics of para-keratinized tissue. Interestingly, the feasibility of combining such engineered mucosa with engineered alveolar bone was also established 43 . The use   of such models confirmed the need of 3D models to investigate inflammatory processes and also host-response to pathogens, such as P. gingivalis, as differential effects between 3D and 2D models were highlighted, especially, regarding inflammatory cytokines (IL-8, IL-1α) secretion 9 . In our 3D MT model, same discrepancies related to the mRNA expression of inflammatory markers were also observed in comparison with monolayer cultures. Furthermore, the gene expression of the non-infected and P. gingivalis-infected MT was similar to that measured in healthy and diseased gingival tissues respectively, confirming the similarity between MT and gingival tissue. It should be mentioned that one limitation of the MT model is the absence of keratinization that could be obtained and controlled in OMM after exposure of the cell layers to air 9,43,44 .
In MT, the external localization of ECs mimics the epithelial barrier, a key element of the host-defense and innate immune response due to its implication in host-bacterial crosstalk 45 . At the junctional epithelium, ECs are connected to each other by a variety of specialized transmembrane proteins including tight junctions, adherens junctions, gap junctions and desmosomes 27,28 that constitute a thin structure. Adhesion structures are the key to maintain epithelial integrity and are also involved in the coordination of several signaling and trafficking molecules, thereby, regulating cell differentiation, proliferation, and polarity [46][47][48] . In our 3D model, ECs were characterized by the presence of E-cadherin observable at the periphery of the MT and localized at the epithelial level, confirming an architecture similar to that of the junctional epithelium. Following P. gingivalis infection, we observed in vivo and in MT, the bacterial ability to invade ECs, as demonstrated previously, and also to bypass the epithelial barrier to reach the underlying tissue. Invasion is a rapid process and is accompanied by cytoskeletal modifications, calcium ion fluxes, modulation of MAP kinase and apoptotic pathways, and downregulation of IL-8 expression 49,50 contributing in vivo to bacterial persistence and progression of the chronic aspects of periodontitis 51 . Several studies demonstrated that P. gingivalis induces proteolysis of adhesion structure components and modulates related gene expression 52,53 . Here, we focused on integrins due to their involvement in binding and invasion of P. gingivalis as demonstrated in several cell types 54,55 . In this study, P. gingivalis strain 33277 was used as it is a well-described invasive strain exhibiting several virulence factors such as fimbriae and our laboratory has significant experience using this strain 24,56,57 . Future studies should explore the precise role of each virulence factor in the invasion process and inflammatory response observed in MT model.
Modulation of apoptosis is a key feature of innate immune response subversion elicited by P. gingivalis, especially at the epithelial level 33,58 . In MT, the expression of apoptotic markers (caspase-3, Bax-1/Bcl-2, Apaf-1) was differentially measured according to cell type as observed previously 33 . In MT, APAF-1 related apoptosome was mostly expressed in fibroblastic core and its expression within epithelial layer was decreased following P. gingivalis infection. The inhibition of apoptosis in ECs is a well-described mechanism involved in P. gingivalis pathogenesis. Several studies demonstrated the targeting of cells with a rapid turnover, such as junctional epithelial cells, by P. gingivalis 59 and its ability to decrease epithelial apoptosis. Therefore, the use of MT model will be useful to evaluate indirect impact of P. gingivalis on connective tissue following ECs stimulation.
In conclusion, this 3D in vitro model of gingival tissue may be used to analyze pathophysiological processes involved in periodontitis especially molecular mechanisms related to either innate immune response, role of bacterial virulence factors occurring at the epithelium-connective tissue interface and therapeutic properties of potential anti-inflammatory or anti-infective drugs. The potential of adding a third cell type at the core of the MT, such as osteoblasts, may be of interest to investigate the influence of soft tissue infection and inflammation on soft tissue-bone crosstalk.