Early host–microbe interaction in a peri‐implant oral mucosa‐biofilm model

The host‐microbe relationship is pivotal for oral health as well as for peri‐implant diseases. Peri‐implant mucosa and commensal biofilm play important roles in the maintenance of host‐microbe homeostasis, but little is known about how they interact. We have therefore investigated the early host‐microbe interaction between commensal multispecies biofilm (Streptococcus oralis, Actinomyces naeslundii, Veillonella dispar, Porphyromonas gingivalis) and organotypic peri‐implant mucosa using our three‐dimensional model. After 24 hr, biofilms induced weak inflammatory reaction in the peri‐implant mucosa by upregulation of five genes related to immune response and increased secretion of IL‐6 and CCL20. Biofilm volume was reduced which might be explained by secretion of β‐Defensins‐1, ‐2, and CCL20. The specific tissue reaction without intrinsic overreaction might contribute to intact mucosa. Thus, a relationship similar to homeostasis and oral health was established within the first 24 hr. In contrast, the mucosa was damaged and the bacterial distribution was altered after 48 hr. These were accompanied by an enhanced immune response with upregulation of additional inflammatory‐related genes and increased cytokine secretion. Thus, the homeostasis‐like relationship was disrupted. Such profound knowledge of the host‐microbe interaction at the peri‐implant site may provide the basis to improve strategies for prevention and therapy of peri‐implant diseases.

immune response, which is protective (Darveau, 2010). This controlled immune response is necessary to maintain oral health (Darveau, 2010;Hajishengallis, 2015;Ingendoh-Tsakmakidiset al., 2019). However, little is known about the initiation and maintenance of homeostasis in the early interaction between commensal biofilm and mucosa (Meyle & Chapple, 2015). Various factors (e.g., immunodeficiency, environmental factors, keystone pathogens) are able to induce dysregulation of the host-microbe homeostasis (Lamont & Hajishengallis, 2015). This dysbiosis is accompanied by an increased inflammatory reaction and a shift in the microbiome, which can lead to oral diseases such as peri-implantitis (Berglundh et al., 2018;Meyle & Chapple, 2015). Peri-implantitis is highly prevalent, progresses rapidly and is difficult to treat (Belibasakis, 2014;Dreyer et al., 2018). It would therefore be desirable to know more about host-microbe homeostasis at the implant site in oral health, as this could help to improve strategies for the prevention and therapy of peri-implant diseases. Therefore, our aim was to investigate the interaction between an early commensal multispecies biofilm and peri-implant mucosa, using our 3D peri-implantmucosa-biofilm model (Ingendoh-Tsakmakidiset al., 2019), in order to get new insights into the early interaction between hosts and microbes.

| Multispecies biofilm formation
The commensal multispecies biofilms were formed as previously described (Kommerein et al., 2017). Briefly, equal volumes of the bacterial cultures were mixed in BHI/vitamin K to achieve a final optical density (600 nm) of 0.01 for each species. The multispecies biofilms were grown on glass cover slips (18 mm diameter, thickness 1, Thermo Scientific Menzel) in 6-well plates for 48hr under anaerobic conditions (80% N 2 , 10% H 2 , 10% CO 2 ) at 37 C.
2.3 | Co-culture of the peri-implant mucosa with the multispecies biofilm After the assembly of the peri-implant mucosa models and the formation of multispecies biofilms, both were washed separately with phosphate-buffered saline (PBS). The multispecies biofilms were placed on spacers and on the integrated titanium disk of the periimplant mucosa model. The same setting without mucosa was used to cultivate the controls of multispecies biofilm. The control mucosa models were exposed to a glass cover slip without biofilm. All samples were cultured after being submerged in co-culture medium (3:1 DMEM (P04-03591, Pan-Biotech) and Ham's F-12 (P04-14559, Pan-Biotech), 5 μg/ml insulin, 0.4 μg/ml hydrocortisone, 2 × 10 −11 M 5-triiodo-L-thyronine, 1.8 × 10 −5 M adenine, 5 μg/ml transferrin, 10 −10 M cholera toxin, 2 mM L-glutamine, 10% v/v FBS, 10% v/v BHI/vitamin K) for 24 or 48 hr at 37 C in a humidified 5% CO 2 atmosphere. After 24 hr, the medium was replaced once.

| Quantitative and qualitative biofilm analysis
After co-culture, the multispecies biofilms were washed once with PBS.
They were then stained-either with the LIVE/DEAD BacLight Bacterial Viability Kit (Life Technologies) or by using fluorescence in situ hybridization (FISH) (Kommerein et al., 2017;Kommerein, Doll, Stumpp, & Stiesch, 2018). For FISH, the biofilms were fixed and subsequently permeabilised for 10 min with 1 μg/ml lysozyme. Hybridization was performed with 4 μM of each 16S rRNA probe (Eurogentec), as listed in Appendix Table S1. Both live/dead and FISH-stained biofilms were imaged by a confocal laser scanning microscope (CLSM, Leica TCS SP8, Leica Microsystems). For each sample, five z-stack images were acquired with a z-step size of 1 μm. The Imaris ×64 8.4 software package (Bitplane AG) was used to reconstruct 3D images and to calculate the volume. The viable (SYTO9; green), dead (propidium iodide; red), and colocalised (SYTO9 + propidium iodide; green + red) parts of the biofilms were quantified using live/dead staining. The co-localised part was defined as dead biofilm part. The volume proportions of each bacterial species in the multispecies biofilm were quantified using FISH-stained biofilms.

| Histology of the peri-implant mucosa
The histological analysis of the peri-implant mucosa was performed as previously described (Ingendoh-Tsakmakidiset al., 2019). Briefly, the tissues with integrated implant were embedded in Technovit 9,100 and ground to 22-36 μm slides. Finally, the slides were stained according to van Gieson.
2.6 | RNA extraction and microarray data analysis RNA extraction and microarray data analysis were performed as previously described (Ingendoh-Tsakmakidiset al., 2019) and stored in detail in the GEO database (GSE136274

| Quantification of cytokines and human β-Defensins
Cytokine and human β-Defensin (hBD) levels were measured in the collected supernatants. Cytokines CXCL1, IL-10, IL-1β, IL-6, CXCL8, CCL2, CCL20, and TNF-α were quantified using a Human Chemokine Bio-Plex kit (Bio-Rad). These proteins were measured by the Luminex-based multiplex technique according to the manufacturer's instructions. Concentrations were calculated by the Bio-Plex Manager 6.0 using the standard curve with five-parameter logistic (5-PL) regression curves. Cytokine CXCL2, and hBD-1, -2, and -3 were measured using enzyme-linked immunosorbent assays (ELISAs). The ELISA kits were purchased from PeproTech and performed according to the manufacturer's protocol. The concentrations of CXCL2, and hBD-1 to -3 were calculated using a fourparameter logistic (4-PL) equation resulting from the standard curve.

| Statistical analysis
The statistical analysis was performed using Prism 8 (GraphPad). All the results were analysed using the two-way analysis of variance (ANOVA) with Bonferroni correction. Statistical differences were considered significant at p < .05. The number of individual experiments is stated in the figure legends.

| Reduction of biofilm volume after periimplant mucosa exposure
The commensal multispecies biofilms were cultivated with the periimplant mucosa models or in the same setting without mucosa as control for 24 or 48 hr in a humidified 5% CO 2 atmosphere. The biofilm volume was determined by live/dead staining. Exposure to the mucosa significantly reduced the biofilm volume to a similar level after 24 and 48 hr (Figure 1b), which indicates that the biofilm volume was reduced within the first 24 hr. However, the viable part of the biofilm was significantly increased after 48 hr of co-cultivation ( Figure 1c).

| Microbial shift in the multispecies biofilm after 48 hr peri-implant mucosa exposure
The bacterial distribution in the multispecies biofilms was determined by FISH staining. After 24 hr, S. oralis volume dominated the control biofilms, followed by A. naeslundii, V. dispar, and P. gingivalis (Figure 2b).
This order did not change after further 24 hr, but there was a sig-

| Damage and detachment of the peri-implant mucosa after 48 hr biofilm exposure
After 24 or 48 hr exposure to the biofilm, the impact on the periimplant mucosa morphology was investigated histologically. Twenty four hours exposure had no visible effect on the peri-implant mucosa ( Figure 3a). However, after 48 hr biofilm challenge, the mucosa was slightly damaged and began to detach from the implant (Figure 3a).
Bacterial colonisation on the epithelium was detected after 24 and 48 hr (Figure 3b). In contrast to the sporadic colonisation after 24 hr, bacterial colonisation was increased and spread out over the whole epithelium after 48 hr.
3.4 | Increased transcriptional response of the peri-implant mucosa after 48 hr biofilm exposure The transcriptional activity of the peri-implant mucosa was measured by microarrays. After 24 hr, only five genes were differentially expressed in the peri-implant mucosa exposed to biofilm-in comparison to the unexposed mucosa-and all of these were upregulated ( Figure S1A). Most of the genes were involved in the cytokinecytokine receptor interaction pathway (Table 1). Forty eight hours exposure to the biofilm led to regulation of 122 genes in the periimplant mucosa: 14 were up-and 108 downregulated ( Figure S1B).
Most of the upregulated genes were involved in TNF signalling and cytokine-cytokine receptor interaction pathway (Table 1). Downregulated genes were related to different signalling pathways and to cell adhesion (ECM-receptor interaction, focal adhesion) (Table 1).

| Increased pro-inflammatory cytokine secretion after 48 hr biofilm exposure
The hBD, cytokine and chemokine levels in the supernatants were determined by ELISA or a Luminex-based multiplex assay (Figure 4).
Biofilm exposure significantly altered cytokine and chemokine secretion. After 24 hr, IL-6 and CCL20 secretions were increased compared to control, whereas IL-1β, TNF-α, and CCL20 secretions were higher after 48 hr. The CXCL2 level was decreased compared to control at both time points. The constitutive CXCL1, CXCL8, CCL2, hBD-1, and hBD-2 secretion was not enhanced by addition of the biofilm independently from time. hBD-3 and IL-10 secretion was below the detection limit of both ELISA and Luminex-based multiplex assays (data not shown).

| DISCUSSION
The innate immune response triggered by commensal bacteria supports oral health (Darveau, 2010) and can protect from periimplantitis, a highly prevalent disease (Dreyer et al., 2018). However, little is known about the host-microbe interaction at the implant site during the early phase. Therefore, our aim was to investigate the interaction of a peri-implant mucosa and a commensal multispecies The peri-implant mucosa secreted CCL20, hBD-1, and -2, all of which possess antimicrobial activities (Hans & Madaan Hans, 2014;Yang et al., 2003). Both antimicrobial peptides hBD-1 and -2 are expressed in healthy gingival tissue (Dale et al., 2001). hBD-1 is con- tion of CCL20 were induced in the peri-implant mucosa after 24 hr exposure to the commensal biofilm. The protective antimicrobial response of the tissue might explain the reduction in the biofilm volume in this study. The control of commensal biofilm overgrowth contributes to oral health (Hans & Madaan Hans, 2014). Biofilm volume decreased without changes in live/dead or in the distribution of bacterial species, which still corresponded to the native early plaque (Kommerein et al., 2017). A stable microbial community is associated with oral health (Kilian et al., 2016). Taken together, in the first phase of biofilm exposure to the peri-implant mucosa, a symbiotic relationship was created between bacteria and the host that was similar to the homeostasis in oral health.
In contrast, 48 hr biofilm exposure to the peri-implant mucosa gave rise to a different reaction from both tissue and biofilm.
F I G U R E 3 Histological sections of the peri-implant mucosa after exposure to either multispecies biofilm or co-culture medium without bacteria for 24 or 48 hr. (a) Implant-mucosa interface is shown. The mucosa was intact in both controls and after 24 hr biofilm exposure. After 48 hr biofilm exposure, the mucosa was detached from the implant and the epithelium was damaged. (b) Bacterial colonisation after 24 or 48 hr. Arrows indicate the bacteria. The ground sections were stained according to van Gieson. Representative pictures of four peri-implant mucosa models for each condition Histological results showed that the epithelial barrier was disrupted and the mucosa was detached from the implant, as is characteristic of peri-implant diseases (Valente & Andreana, 2016). In response to the biofilm, the peri-implant mucosa downregulated genes related to focal adhesions, which are important for the epithelial barrier (Handfield, Baker, & Lamont, 2008) and its attachment to titanium (Pendegrass et al., 2015). Hence, this downregulation might contribute to the epithelial disruption at the implant surface. In addition, the observed enhanced bacterial colonisation and epithelial barrier damage might be promoted by P. gingivalis-induced downregulation of the PI3K-Akt signalling pathway and genes related to cell adhesion. P. gingivalis, which was a viable part of the biofilm, is able to attenuate the PI3K-Akt signalling pathway (Nakayama, Inoue, Naito, Nakayama, & Ohara, 2015) and to disrupt cell-cell junctions at the levels of gene expression and protein content (Abe-Yutori, Chikazawa, Shibasaki, & Murakami, 2017;Katz, Sambandam, Wu, Michalek, & Balkovetz, 2000). Both promote its colonisation and invasion of mucosal tissue. This epithelium penetration and damage is an important step in the pathogenesis of oral diseases (Groeger, Doman, Chakraborty, & Meyle, 2010). Consequently, the effect on the F I G U R E 4 Cytokine, chemokine and hBD secretion by peri-implant mucosa after exposure to either multispecies biofilm or co-culture medium without bacteria for 24 or 48 hr. The cytokines IL-6,IL-1β, TNF-α, and the chemokines CCL20, CXCL8, CCL2, and CXCL1 were measured by the Luminex-based multiplex technology and hBD1, hBD2, and CXCL2 by ELISA. The Tukey box plots represent the measured data points (18 or 24) of 6 to 8 peri-implant mucosa models for each condition. BF, biofilm; PIM, periimplant mucosa. *p < .05; **p < .01 peri-implant mucosa indicated that host-microbe homeostasis was disrupted after 48 hr.
After 48 hr, the pro-inflammatory response of the organotypic mucosa was enhanced compared to 24 hr, by additional upregulation of genes related to inflammation and by elevated secretion of CCL20, IL-1β, and TNF-α. Several clinical studies have found that IL-1β levels were significantly higher in patients with periodontitis compared to healthy patients and that they correlated significantly with clinical parameters of periodontitis (Jaedicke et al., 2016). Moreover, excessive production of IL-1β and TNF-α can lead to tissue destruction and loss of attachment of connective tissue (Graves & Cochran, 2003).
Consequently, the enhanced and uncontrolled pro-inflammatory response of the peri-implant mucosa leads to the development of mucosal inflammation with tissue destruction (Meyle & Chapple, 2015). These were further signs of disruption of the hostmicrobe homeostasis after 48 hr.
Bacterial distribution was altered in the biofilm after 48 hr exposure to the peri-implant mucosa. This provides an additional sign of homeostasis disruption, as the relative proportion and diversity of species varies during the development of oral diseases (Kilian et al., 2016). The alteration in bacterial distribution, with the increase in the proportion of V. dispar, was probably caused by the release of antimicrobial peptides or tissue breakdown products from the periimplant mucosa, and/or the presence of P. gingivalis, because all have been shown to influence microbial composition (Hajishengallis, 2014;Hajishengallis et al., 2011;Langfeldt et al., 2014). V. dispar is an early coloniser and associated with oral health (Avila, Ojcius, & Yilmaz, 2009). An increase of this species in dysbiosis has not yet been described. However, there is some evidence that this species is potentially pathogenic. There are several explanations of how the initially commensal multispecies biofilm suddenly disrupts the relationship similar to hostmicrobe homeostasis. According to the keystone pathogen hypothesis (Hajishengallis, Darveau, & Curtis, 2012), P. gingivalis, which was present in the multispecies biofilm, can induce a disruption of the homeostasis. It is known that this species, even at very low-colonisation level, is able to induce inflammatory destruction in the presence of commensal bacteria and to remodel the microbial composition (Hajishengallis et al., 2011), as underlined by our results. A further explanation, why the commensal biofilm triggered disruption of homeostasis could be the lack of immune cells in our model. In the in vivo host-microbe interaction, immune cells have an important role in detection of pathogen components as well as in homeostasis (Darveau, 2010). Neutrophils play an important role in the initial reaction in the oral cavity and form a wall between the biofilm and epithelium (Pöllänen, Laine, Ihalin, & Uitto, 2012). The lack of neutrophils through disease or chemical induction invariably causes periodontitis (Darveau, 2010). These clinical situations together with our observations suggest that the induced early innate immune response of the mucosal non-immune compartment is not sufficient to protect the tissue without the support of immune cells. Consequently, a commensal biofilm could cause tissue destruction in the absence of immune cells.
In conclusion, this study demonstrated that an early commensal multispecies biofilm induced a protective pro-inflammatory response in the peri-implant mucosa within the first 24 hr, thus, maintaining host-microbe homeostasis and oral health. However, after further 24 hr incubation, the relationship similar to host-microbe homeostasis was disrupted. Our study suggested that various factors (V. dispar, P. gingivalis, immune cells) could be involved in the disruption or maintenance of homeostasis. Future investigations will expand the understanding of the pathogenesis of peri-implantitis and help to develop new strategies to prevent and treat peri-implant diseases, as should eventually lead to improved implants.

ACKNOWLEDGMENTS
Microarray data were generated and the 026652QM_RCUG_HomoSapiens microarray was developed by the Research Core Unit Genomics at Hannover Medical School. We would like to thank Dr. Oliver Dittrich-Breiholz and Heike Schneider for support and advice in their use. Ground sections were performed at HIK