Commensal and pathogenic biofilms differently modulate peri‐implant oral mucosa in an organotypic model

Abstract The impact of oral commensal and pathogenic bacteria on peri‐implant mucosa is not well understood, despite the high prevalence of peri‐implant infections. Hence, we investigated responses of the peri‐implant mucosa to Streptococcus oralis or Aggregatibacter actinomycetemcomitans biofilms using a novel in vitro peri‐implant mucosa‐biofilm model. Our 3D model combined three components, organotypic oral mucosa, implant material, and oral biofilm, with structural assembly close to native situation. S. oralis induced a protective stress response in the peri‐implant mucosa through upregulation of heat shock protein (HSP70) genes. Attenuated inflammatory response was indicated by reduced cytokine levels of interleukin‐6 (IL‐6), interleukin‐8 (CXCL8), and monocyte chemoattractant protein‐1 (CCL2). The inflammatory balance was preserved through increased levels of tumor necrosis factor‐alpha (TNF‐α). A. actinomycetemcomitans induced downregulation of genes important for cell survival and host inflammatory response. The reduced cytokine levels of chemokine ligand 1 (CXCL1), CXCL8, and CCL2 also indicated a diminished inflammatory response. The induced immune balance by S. oralis may support oral health, whereas the reduced inflammatory response to A. actinomycetemcomitans may provide colonisation advantage and facilitate later tissue invasion. The comprehensive characterisation of peri‐implant mucosa‐biofilm interactions using our 3D model can provide new knowledge to improve strategies for prevention and therapy of peri‐implant disease.

shift in species composition of oral biofilms incorporating more pathogenic bacteria (Graves, Correa, & Silva, 2019;G. Hajishengallis, 2014;G. Hajishengallis & Lamont, 2016). As a result, in people carrying dental implants, peri-implant diseases might develop (G. N. Belibasakis, 2014;Berglundh et al., 2018). The reversible inflammation of the soft tissue around the implant is termed "peri-implant mucositis." The more severe form, which is termed peri-implantitis, is irreversible and additionally characterised by loss of bone supporting the implant (Berglundh et al., 2018). Moreover, peri-implant diseases are characterised by high prevalence. A recent meta-analysis showed that 26% of patients with an implant function ≥5 years develop peri-implantitis (Dreyer et al., 2018).
One reason could be that dental implants are missing Sharpey's fibres and the periodontal ligament leading to a reduced physical barrier of the oral mucosa against bacterial invasion (G. N. Belibasakis, 2014). In order to expand the knowledge about the interaction of the peri-implant mucosa and oral microbiome, physiologically relevant in vitro models are required. The invivo situation is much better reflected in three dimensional (3D) organotypic models (Antoni, Burckel, Josset, & Noel, 2015).
In order to study the soft-tissue-implant interface, Chai et al. developed an organotypic oral mucosa with an integrated implant. However, their model did not include an oral biofilm, which is a key element of the peri-implant area (Chai et al., 2010;Chai et al., 2013;Chai, Brook, Palmquist, van Noort, & Moharamzadeh, 2012). To the best of our knowledge, an in vitro model to study the interactions between all three components, implant material, organotypic oral mucosa, and biofilm, is absent.
Balanced immune response maintains the host-microbe homeostasis and confers oral health. The oral health-associated symbiotic microbial community consists mainly of gram-positive Streptococcus spp. and Actinomyces spp., and dozens of less studied species are present (G. Hajishengallis, 2015;Mombelli, Müller, & Cionca, 2012;Szafranski et al., 2015). The commensal Streptococcus oralis belongs to the initial colonizer and is one of the predominant Streptococcus spp. in the early biofilm (Diaz et al., 2006) and consequently should have a considerable impact on oral homeostasis. However, little is known about the mechanisms by which S. oralis interacts with the host. This knowledge would help to elucidate the role of this microbe in host-microbiome homeostasis beyond biofilm initiation. The opportunistic pathogen Aggregatibacter actinomycetemcomitans is genetically diverse (Kittichotirat, Bumgarner, & Chen, 2016) and can be detected at periimplant disease sites (Rams, Degener, & van Winkelhoff, 2014;van Winkelhoff & Wolf, 2000). It expresses various virulence factors and has different strategies to evade host innate defence mechanisms, for example, migration through the epithelium, and binding of different human proinflammatory cytokines (T. Ahlstrand et al., 2017;Dickinson et al., 2011;Herbert, Novince, & Kirkwood, 2016). However, the overall impact including transcriptional response of A. actinomycetemcomitans on the oral mucosa remains unclear. Deciphering of how commensal and pathogenic bacteria, that is, S. oralis and A. actinomycetemcomitans, impact mucosal homeostasis would help to understand peri-implant pathogenesis and to develop new therapeutic options.
The first aim of the present study was to develop an in vitro periimplant mucosa-biofilm model combining the main three components: the organotypic oral mucosa, an implant material, and an oral biofilm ( Figure 1). The second aim was to expand the knowledge about the species-specific effect of commensals and opportunistic pathogens on the mucosal tissue, by studying the impact of either S. oralis or A. actinomycetemcomitans biofilms on the peri-implant mucosa in our unique organotypic model.

| Characterisation of the peri-implant mucosa model
The assembly of the three-dimensional peri-implant mucosa models had duration of 25 days. Briefly, a titanium disk (implant material) was integrated into collagen-embedded human gingival fibroblasts (HGFs).
On the top of the fibroblast-collagen gel, oral keratinocytes (OKF6/ TERT-2) were added around the titanium and differentiated. The morphology of the peri-implant mucosa model was evaluated by van Gieson staining and immunohistochemistry in order to confirm that the mucosal structure reflected the previously published engineered human oral mucosa (Dongari-Bagtzoglou & Kashleva, 2006). The organotypic oral mucosa consisted of a differentiated stratified epithelium and the underlying connective tissue, including the HGF. Four different layers of the epithelium, the stratum basale, stratum spinosum, stratum granulosum, and the superficial keratinised layer were seen to be similar to native human gingival tissue ( Figure S1A). The suprabasal epithelial layer was stained by cytokeratin 13, the basement membrane by collagen IV, and the keratinized superficial cells by cytokeratin 10 (Figure S1B-D). Sporadic proliferating cells were also detected-mainly at the basal layer-through Ki67 staining ( Figure S1E). E-cadherin and claudin staining confirmed the tight epithelial barrier (Figure S1F-G).
FIGURE 1 Schematic illustration of the peri-implant mucosa-biofilm model. The organotypic oral mucosa with an integrated implant was developed in culture inserts. Spacers with a ring form were placed around the tissue model, which allowed the disposition of the Streptococcus oralis or Aggregatibacter actinomycetemcomitans biofilm on top of the implant. Spacers and implant material have the same height keeping the biofilm planar The organotypic oral mucosa was attached to the implant and created an intact implant-mucosa interface ( Figure 2). If titanium disks free of fibroblasts were inserted, the epithelial cells grew apically along the titanium disk, deep into the collagen. This apical epithelial migration created an elongated junctional epithelium covering a considerable area of the implant surface (Figure 2a-c). In contrast, the use of fibroblast-colonized titanium disks hindered such deep epithelial cell migration into the collagen (Figure 2d-f). The staining at the upper part of the implant is related to the fibroblasts, which grew around it prior to insertion into the tissue. The peri-implant mucosa model with a fibroblast-colonized titanium disk built an intact mucosa-implant interface, with only minimal apical epithelial migration along the titanium disk and was used for the following co-cultures.

| S. oralis and A. actinomycetemcomitans biofilms formation
The developed peri-implant mucosa should be challenged with either S. oralis or A. actinomycetemcomitans cells grown as biofilms. Therefore, reproducible and viable sessile communities of these two species were 2.3 | Histology of the peri-implant mucosa after biofilm challenge Peri-implant mucosa was exposed to either S. oralis or A. actinomycetemcomitans biofilm for 24 hr, and the effect was evaluated with histological analysis. Co-cultures with the biofilms resulted in an intact implant-mucosa interface (Figure 4d,g). The epithelium located directly at the implant was slightly loosened after challenge with the S. oralis biofilm (Figure 4d,e). In contrast, the epithelium at a distance from the implant was histologically similar (Figure 4f Immunohistological staining for adherent junctions (E-cadherin) and proinflammatory factors (IL-6, CXCL8, and TNF-α) was similar for the control and for a tissue exposed to the S. oralis biofilm. However, the intensity of claudin staining for tight junctions appeared slightly diminished after co-culture with the S. oralis biofilm ( Figure S2).

| Transcriptional response of the peri-implant mucosa to biofilms
Transcriptional activity of the mucosa was measured by microarrays after 24 hr exposure to biofilms. After co-culture with the S. oralis biofilm, 83 genes were differentially expressed in the peri-implant mucosa compared with the unexposed tissue. Thirty six genes were upregulated whereas 47 genes were downregulated ( Figure 5a). Most of the upregulated genes belonged to the heat-shock proteins 70 (HSP70). These genes are involved in mitogen-activated protein kinase signalling and antigen processing and presentation pathways (Table 1). In addition, some genes from the chemokine signalling pathway (i.e., CCL20, CCL8, and PIK3R5) were upregulated. Genes coding for the invariant alpha  Table S3). No particular pathway was upregulated. The pathway analysis of the downregulated genes revealed the PI3K-Akt signalling pathway (Table 3), including genes related to this signal transduction (MDM2, IL2RG, TLR4, and F2R).

| Cytokine secretion
The cytokine levels in the collected supernatants were measured by using a Luminex-based multiplex assay. The results showed that S. oralis biofilm challenge caused a significant increase in TNF-α level in the peri-implant mucosa compared with the unchallenged tissue ( Figure 6). In contrast, the levels of IL-6, CXCL8, and CCL2 were significantly reduced after stimulation with the S. oralis biofilm. Challenge of the peri-implant mucosa with the A. actinomycetemcomitans biofilm led to significant lower levels of CXCL1, CXCL8, and CCL2 ( Figure 6). The secretion level of CXCL2 was not affected by any of the studied biofilms.
Development of a successful prevention or therapy strategy requires comprehensive understanding of the host-microbe interactions at the peri-implant site. Here, by applying an organotypic model, we investigated the impact of either commensal S. oralis or the pathogenic  (Handfield et al., 2005;Hasegawa et al., 2007). The weak transcriptional response to commensal bacteria supports the adaptive coevolution theory of commensal bacteria with the oral mucosa (Handfield, Baker, & Lamont, 2008;Hooper & Gordon, 2001). The overall response to S. oralis at the transcriptional level was related to protective response. Pathways related to tissue protection were upregulated (Table 1)   Histological sections of the peri-implant mucosa-biofilm model after 24 hr. An overview of the implant-mucosa interface is shown for the control (a), Streptococcus oralis (d) and Aggregatibacter actinomycetemcomitans (g) biofilm challenged groups. An intact implant-mucosa interface was observed in the control (b) at higher magnification. The epithelium at the implant site was slightly loosed, after S. oralis biofilm challenge (e), whereas an intact implant-mucosa interface was observed after co-culture with the A. actinomycetemcomitans biofilm (h). The adjacent tissues of the control (c), S. oralis (f) and A. actinomycetemcomitans (i) biofilm challenged group were intact. The ground sections were stained according to van Gieson. Representative pictures of three independent experiments. So = S. oralis and Aa = A. actinomycetemcomitans. Scale bars: 50 μm FIGURE 5 Heat maps of the global gene expression profiles from the peri-implant mucosa comparing the control and biofilm challenged groups after 24 hr. Results for the Streptococcus oralis (a) and the Aggregatibacter actinomycetemcomitans (b) biofilm are shown in separate heat maps. The heat maps show the hierarchical clustering of the experimental groups and the differentially regulated genes. Data from two-three independent experiments and duplicates were used. Red indicates upregulation and green downregulation after biofilm co-culture. So = S. oralis and Aa = A. actinomycetemcomitans   , 2003). In addition, the adaptive immune response was suppressed as indicated by the downregulation of antigen presentation and processing (Table 2). These might lead to a state of unresponsiveness-with decreased both humoral and cell-mediated immune response (Han et al., 2003;Hasegawa et al., 2007). Hyporesponsiveness induced by commensals probably plays a role in protection from tissue destruction induced by inflammatory response (Pollanen et al., 2012). Compared with the S. oralis biofilm, transcriptional response to A. actinomycetemcomitans was broader without targeting pathways. Upregulated genes were related to DNA damage, DNA repair, and cell division suggesting general stress response. Analysis of the downregulated genes revealed a single enriched pathway: the PI3K-Akt signalling ( Table 3). Attenuation of this pathway by P. gingivalis can promote its invasion and colonisation of the mucosal tissue (Nakayama, Inoue, Naito, Nakayama, & Ohara, 2015). Similarly, our observed changes may promote colonization and survival of A. actinomycetemcomitans. In summary, the transcriptional profiles of the peri-implant mucosa revealed a tissue protective response to the S. oralis biofilm and a stress response to the A. actinomycetemcomitans biofilm.
The classical proinflammatory cytokine IL-6 and the neutrophil recruiting chemokines CXCL8 and CCL2 were found at lower levels in the supernatants after challenge with the S. oralis biofilm ( Figure 6).
Corresponding to our results, different studies could show that commensal bacteria reduce the proinflammatory cytokines, IL-6 and CXCL8 (Cosseau et al., 2008;Hasegawa et al., 2007;Twetman et al., 2009;Zhang, Chen, & Rudney, 2008). Therefore, S. oralis biofilm might attenuate the proinflammatory response, which is consistent with our observations on gene expression. TNF-α was increased in response to the S. oralis biofilm ( Figure 6). It is one of the main inflammation mediators (Groeger & Meyle, 2015) and is present at low levels in the gingival crevicular fluid in healthy patients (Darveau, 2010;Petkovic-Curcin, Matic, Vojvodic, Stamatovic, & Todorovic, 2011). Probably, cytokines controlled by commensal bacteria are involved in limiting biofilm development and consequently in maintaining gingival health (Darveau, 2010;Dickinson et al., 2011;Rouabhia, 2002). After challenge with the FIGURE 6 Cytokine levels in the co-culture supernatants of the peri-implant mucosa-biofilm model after 24 hr. The groups challenged with the Streptococcus oralis or Aggregatibacter actinomycetemcomitans biofilm were compared with the control groups for their CXCL1, CXCL2, IL-6, CXCL8, CCL2, and TNF-α levels. The cytokines were measured using the luminex technology and a Bio-Plex Kit. The Box and Whiskers graphs with Tukey error bars represent the measured data points. The S. oralis biofilm group includes 42 measured data points from 14 samples and four independent experiments. The A. actinomycetemcomitans biofilm group includes 29 measured data points from seven samples and two independent experiments. So = S. oralis and Aa = A. actinomycetemcomitans. The statistical significance was determined using the Mann-Whitney method, with P = .05. Single asterisk indicates P < .05 and double asterisk P ≤ .01 A. actinomycetemcomitans biofilm, the levels of CXCL1, CXCL8, and CCL2 were lower than in the control ( Figure 6). Previously, it was found that A. actinomycetemcomitans can sense and bind cytokines; among them was CXCL8 (T. Ahlstrand et al., 2017;T. Ahlstrand et al., 2018; T. Ahlstrand, Kovesjoki, Maula, Oscarsson, & Ihalin, 2019). Lower metabolic activity of biofilms induced by cytokine binding could lead to higher resistance (A. Paino et al., 2011)  In conclusion, our novel peri-implant mucosa-biofilm model promises enormous experimental potential to investigate the interaction of three key components: mucosa, biofilm, and implant. Our 3D model reflected that commensal streptococci induce a balanced immune response of the soft tissue including specific transcriptional response and attenuated pro-inflammatory cytokines. This subtle effect could preserve the oral health. Furthermore, the colonization advantage of opportunistic pathogens by suppression of inflammatory reaction could favour dysbiosis. We showed that species-specific molecular reactions of the peri-implant mucosa to biofilm can be successfully studied in our peri-implant mucosa-biofilm model.
The influence of various implant materials and surface functionalisation on biofilm formation and tissue reaction are additional factors that will be analyzed in the future. Accordingly, the findings will provide new opportunities for future strategies of disease prevention and treatment as well as for implant improvement.

| Co-culture of the peri-implant mucosa with the biofilms
The co-cultures were conducted in AL-medium without any antibiotics. The peri-implant oral mucosa model was used and washed with PBS prior to co-culture. Either the 72-hr-old S. oralis or A. actinomycetemcomitans biofilm was washed five times with PBS and was placed on spacers with the biofilm side facing the periimplant oral mucosa model with direct contact to the integrated titanium disk (Figure 1). The co-cultures were performed for 24 hr at 37°C in a 5% CO 2 humidified atmosphere.

| RNA extraction and microarray data analysis
The supernatants were collected after co-culture for subsequent analysis of the secreted cytokines. The tissues were stored in RNAlater™

| Histological analysis
The peri-implant oral mucosa models were fixed in a 4% buffered formalin solution for 24 hr. The samples were watered, dehydrated by using an ethanol gradient, and embedded in Technovit 9100. The embedded samples were either cut into 5-μm slides for implant-free sections or were grinded to 22-36-μm slides for the peri-implant ground sections according to the cutting-grinding technique by Donath K. Prior staining, the Technovit 9100 was removed by rinsing the slides in acetone. Afterwards, the slides were rehydrated by using an ethanol gradient. Finally, the slides were stained according to van Gieson or specific antibodies. For van Gieson staining, the slides were rinsed for 10 min in ferric haematoxylin, and then washed once with tap water and twice with hydrochloric acid alcohol. After rinsing in tap water for 10 min, they were added into the van Gieson solution for 3 min, subsequently washed with 96% ethanol, 100% ethanol, xylol, and finally mounted. For immunohistochemical staining, the slides were washed with distilled water, washing buffer, antigen retrieval buffer, and washing buffer. The slides were incubated overnight with the primary antibody. All primary antibodies were against

| Statistical analysis
All presented data were derived from two to three independent experiments.
Statistical evaluation of the cytokine levels was performed using GraphPad Prism 7. A Mann-Whitney test was used to analyze the statistical differences between the controls and biofilm groups. Differences were considered statistically significant at P < .05.