Humans are host to a complex flora of microorganisms that colonize the oral mucosa and gingiva. More than 700 species of commensal and pathogenic, aerobic and anaerobic bacteria inhabit the oral cavity (Aas et al., 2005). The healthy oral cavity is a result of a delicate balance between the host and these microbes.

Streptococcus gordonii is a Gram-positive facultative anaerobic bacterium which has adapted to a commensal role during a long evolutionary relationship with the host. S. gordonii is one of the initial colonizers of dental plaque, and is also one of the streptococcal groups which dominates the oral mucosa of healthy individuals (Díaz et al., 2012). However, a subset of S. gordonii can escape the host restraint mechanisms and initiate oral infectious diseases. For example, by adhering to Candida albicans hyphae, the two microbes form bacteria-fungal biofilms which are highly antibiotic resistant and invasive (Montelongo-Jauregui et al., 2019; Salvatori et al., 2020). S. gordonii is also considered to be a direct precursor of periodontal disease inhabiting the subgingival crevice (Cook et al., 1998; Kuboniwa and Lamont, 2010). For these reasons, S. gordonii is referred to as an opportunistic commensal (Boman, 2000; Dixon et al., 2004).

In contrast, Aggregatibacter actinomycetemcomitans is strongly associated with the pathogenesis of aggressive periodontitis (Mans et al., 2006; Casarin et al., 2010). It is a Gram-negative facultative anaerobic coccobacillus that exerts specific effects on the host mucosal immune system and is associated with the development of oral infectious diseases (Henderson et al., 2002; Herbert et al., 2016). Numerous virulence factors are produced by A. actinomycetemcomitans that affect the gingival epithelium and trigger the onset of periodontitis, including lipopolysaccharide (LPS), leukotoxin, cytolethal-distending toxin, collagenase and outer membrane protein (Gholizadeh et al., 2017). A. actinomycetemcomitans is also considered to play an important role in the initiation of biofilm that is potentially pathogenic (Díaz and Kolenbrander, 2009).

When bacteria invade into the oral mucosa barrier, the host innate immunity is rapidly activated as a first line of defense, with gingival epithelial cells playing an important role (Dickinson et al., 2011; Moutsopoulos and Konkel, 2018). Bacteria can bind to host receptors e.g., toll-like receptors (TLRs) on the epithelial cell membrane, leading to a cascade of intracellular pathways which may ultimately result in the release of pro-inflammatory cytokines and chemokines such as interleukin (IL)-6, IL-8, CCL2, CCL5, CCL20 and CXCL1 (Schaller et al., 2002; Teles et al., 2009; Dickinson et al., 2011; Lundmark et al., 2019). The level of these functional molecules, which in turn determines the extent and type of immune reaction has been shown to indicate the severity of mucosa tissue damage (Teles et al., 2009; Diesch et al., 2021). To maintain health, it is critical that the host strikes a complex balance between tolerating (and even benefitting from) resident commensals and suppressing potential pathogens (Lamont et al., 2018; Moutsopoulos and Konkel, 2018). Therefore, studying how the oral barrier responds to different bacterial challenges will contribute greatly to understanding mechanisms behind this balance.

Animal models have been extensively used to study the interaction between host and oral microbes. For example, A. actinomycetemcomitans has been repeatedly verified for its pathogenic roles during the onset and development of periodontitis in rodent periodontitis models, such as the rat infection model, lipopolysaccharide injection model, and even in an advanced humanized mouse model (Graves et al., 2012; Buschhart et al., 2020; Rojas et al., 2021; Wang et al., 2021). However, using experimental animals to study human oral host-microbe interactions has limited translational effectiveness due to differences between humans and animals in the mucosal immunity as well as oral microbiota (Elliott David et al., 2005; Hyde et al., 2014). Furthermore, when preparing periodontitis models, imprecise administration of microbes, risk of mechanical trauma, failure to induce periodontitis in spite of repetitive injection can be problematic. Although oral bacteria, like S. gordonii and A. actinomycetemocomitans, have been broadly investigated for their host influence in vitro, most of these studies used 2D keratinocyte cultures to represent the host (Hasegawa et al., 2007; Sliepen et al., 2009; Stathopoulou et al., 2010; Dickinson et al., 2011). Little is known about how these bacteria affect the host in a human-representative context. Therefore, there is an urgent need to develop human physiologically relevant in vitro models to investigate host-microbe interactions.

In the past, we have developed a reconstructed human gingiva (RHG) consisting of a stratified and differentiated gingival epithelium on a fibroblast-populated collagen hydrogel which functions as the lamina propria (Kosten et al., 2015). RHG is cultured exposed to the air from above with culture medium being supplied from below, which enables topical application of chemicals and bacteria similar to the in vivo situation (Shang et al., 2020). Using this RHG model, we have investigated host TLR expression and cytokine secretion in response to saliva-derived multi-species biofilms (Buskermolen et al., 2018; Shang et al., 2019), as well as to an oral resident bacterium Streptococcus mitis in combination with metal sensitizers related to oral allergy (Shang et al., 2020). Even though this co-culture model represents complex host-microbe interactions in comparison to conventional 2D monolayer models, it still has the major limitation in that bacterial cells were no longer viable after 24 hours. This reduction in viability was most likely caused by the addition of antibiotics (e.g., penicillin-streptomycin (PS)) to the culture medium (Waksman, 1953; Georgopapadakou and Liu, 1980). Antibiotics were added to the culture medium to reduce the risk of bacterial infection caused by bacteria spilling over from the surface of RHG via the sides of the transwell into the culture medium below. Taken together, a new strategy which can maintain the viability of topically applied bacterial cells without increasing the risk of extensive bacterial infection is needed to study the host-microbe interactions for an extended period of time.

The first aim of this study was to develop a model to investigate interactions between the RHG and living oral bacteria over an extended period of time of 3 days. The impact of two essential technical steps on RHG viability were investigated: a) the impact of topically applied agar as a vehicle to keep the bacteria localized on the surface of RHG thus preventing bacteria from spilling over into the culture medium via the edges of the transwell, and b) the impact of the presence and absence of PS on the host-microbe co-culture over a period of 3 days. The second aim of this study was to investigate differences in innate cytokine immune responses when RHG is exposed to a commensal and a pathogenic bacterium for 3 days. In this study, S. gordonii represented the commensal and A. actinomycetemcomitans represented the pathogenic bacteria.

Here we describe a technical advance which enables co-culture of RHG together with living commensal S. gordonii or pathogenic A. actinomycetemcomitans bacteria for a duration of 3 days. This development enables investigation of dynamic human host–microbe interactions over an extended period in vitro. Literature reviews describing work from other groups show that the few studies performed mainly report RHG exposed to a single- or dual-species of oral bacteria for maximally 48 hours (Dongari-Bagtzoglou and Kashleva, 2006; Tabatabaei et al., 2020). Most of these reports described a disrupted tissue integrity, decreased viability and increase in apoptosis of host tissues within 48 hours (Andrian et al., 2004; Gursoy et al., 2010; de Carvalho Dias et al., 2018; Morse et al., 2018; Beklen et al., 2019). One report described a model where a multi-species biofilm was co-incubated with RHG for 3 days without affecting the viability of RHG (De Ryck et al., 2014). However, in that model, the biofilms were cultured on a transwell membrane and not in direct contact with the RHG. In contrast, our model describes for the first time, a direct contact interaction between host and bacteria, which is similar to in vivo. We demonstrated that the viabilities of both host and bacteria, and tissue integrity could be preserved for the entire 3-day culture period.

Our technical advance was achieved by introducing 2 key changes into our previously reported studies. The first being that the bacteria were applied topically to RHG in a soft, low concentration of agar containing sufficient nutrients to maintain viability or promote growth of bacterial cells and which also solidified thus keeping the bacteria localized on the surface of RHG. In one of our previous studies, bacteria were applied in suspension which could spill over into the RHG culture medium via the edges of the transwell and therefore PS was required to prevent contamination so that a prolonged co-culture period of 7 days was reached. However, this resulted in a rapid decrease to 10% of bacterial viability within 24 hours and eventually death of all bacteria soon afterwards (Shang et al., 2018). This motivated our second protocol adaption which was to omit PS from the RHG culture medium from the time that bacteria were topically applied. We expected that this would permit optimal conditions for stable culture of bacteria on RHG within the soft agar gel. Indeed, our results showed when RHG-bacteria co-cultures were grown in the presence of PS, the viable counts decreased considerably but when grown without PS, the bacterial viability were unaffected. This finding also indicated that the PS added to the cell culture medium can diffuse through the RHG into the agar containing the bacteria and reduced bacterial viability, which inevitably affects the studied host-microbe interactions.

It has been described by others that PS, when added throughout the culture period, has a detrimental effect on keratinocytes. In particular, these antibiotics suppressed keratinocyte proliferation and differentiation in monolayer culture as well as the formation of a multilayered reconstructed epidermis (Neftel and Hübscher, 1987; Llobet et al., 2015; Nygaard et al., 2015). In our study, we did not observe any differences in RHG viability, proliferation, differentiation or inflammatory cytokine release when cultured in the presence or absence of PS or when agar was present. However it should be noted that, in contrast to these studies which only had an epidermis, our RHG also has a fibroblast populated hydrogel compartment which will substantially support growth of the epithelium (El-Ghalbzouri et al., 2002).

Having established a stable model for investigating host-microbe interactions, we next went on to show that commensal S. gordonii and pathogenic A. actinomycetemcomitans exerted differential responses in RHG. Whilst neither bacterium significantly influenced RHG viability or tissue histology, differential effects on (pro-)inflammatory cytokine and chemokine secretion and expression of AMPs were observed. S. gordonii exposure resulted in a negligible inflammatory response whereas A. actinomycetemcomitans strongly stimulated cytokine and chemokine secretion indicating that the later had initiated TLR signaling and activation of the inflammazone, which, in vivo, would be expected to combat invasion of the pathogen into the host tissues in order to maintain homeostasis (Wells et al., 2011; McClure and Massari, 2014; Shang et al., 2019). Previous studies have shown that virulence factors, such as lipopolysaccharide or genomic DNA of A. actinomycetemcomitans were able to induce proinflammatory cytokine production in human gingival fibroblasts. These factors were A. actinomycetemcomitans specific, which might explain the RHG differential expression observed between A. actinomycetemcomitans and S. gordonii (Patil et al., 2006; Soto-Barreras et al., 2017)

So far, only one study has examined the differential inflammatory responses of gingival epithelial cells to 24 hours exposure of S. gordonii and A. actinomycetemcomitans (Dickinson et al., 2011). This study showed that both bacteria induced IL-6 and IL-8 secretion, but the levels of induction by A. actinomycetemcomitans were lower than S. gordonii. The different IL-6 and IL-8 secretion patterns in Dickinson et al. and our study may be due to the different host model used: the previous study used multi-layer epithelial cells whereas this study used a 3D organotypic model consisting of epithelial and fibroblast cells (Dickinson et al., 2011; Shang et al., 2018). Moreover, the bacterial strains used in these two studies might also play a role. The differential host responses to commensals and pathogens have been the focus of several studies. However, the findings were inconsistent. Ingendoh-Tsakmakidis and Shang showed pathogenic biofilms induced lower cytokine secretion of RHG, as compared to commensal biofilms, whereas Brown found the opposite trend: the “health-associated” Streptococcus biofilms had little inflammatory effect on the RHG tissue but the “gingivitis-associated” biofilms induced significant inflammatory response, which is in line with the result of current study (Shang et al., 2019; Ingendoh-Tsakmakidis et al., 2019; Brown et al., 2019). Taken together, the contrasting results are likely due to different RHG models and different types of biofilms, single- or multi-species, employed in the studies.

Moreover, the expression of AMPs (Elafin and HBD-2) increased in the upper regions of RHG epithelium after exposure to both bacteria, in line with our previous studies (Buskermolen et al., 2018; Shang et al., 2018). However, substantially more Elafin secretion was observed after commensal S. gordonii exposure compared to the pathogenic A. actinomycetemcomitans. This finding is supported by another report, where a periodontal pathogen Porphyromonas gingivalis significantly reduced Elafin gene expression and secretion in gingival keratinocytes (Yin et al., 2010). This would indicate that the commensal bacteria can initiate a more protective response from RHG than the pathogenic bacteria. Importantly, our results also show that the observed inflammatory response from living A. actinomycetemcomitans (without PS) is more than just a question of TLR activation by bacterial cell membrane components (e.g., Lipopolysaccharides) since when the same number of heat killed bacteria were applied to RHG the inflammatory cytokine response was significantly lower.

As with all complex in vitro models our host-microbe model does have limitations and there is still room for further improvement. In this study, the RHG was constructed using TERT immortalized gingiva keratinocyte and fibroblast cell lines. In the past the cell line model was extensively compared with the RHG constructed from primary cells and was found to show excellent correlation with regards to histology, biomarker expression and cytokine secretion (Buskermolen et al., 2016). Since the use of primary cells is accompanied with extensive limitations such as availability and logistics of receiving fresh tissue biopsies, small biopsies resulting in limited scalability (low number of replicate RHG within an experiment), and high risk of infection from oral microbiota, it was decided to use the TERT RHG model in this study. However, this is accompanied by the fact that our results represent a single donor for the keratinocytes and a non-matched single donor for the fibroblasts and therefore donor variation is not possible to study in the current TERT cell line model. Notably, immune cells are absent from the current host-microbe interaction model. Since antigen presenting cells, such as Langerhans cells, play an important role in the host’s response to microbes, these immune cells will now be incorporated into the model. Already we have been able to show that MUTZ-3 derived Langerhans Cells incorporated into RHG maintain their functionality and plasticity as shown by their ability to mature and migrate from the epithelium to the fibroblast hydrogel upon exposure to contact sensitizers (Kosten et al., 2016).

In this study, we used a single species for the microbial challenge which is less complex than the in vivo microbiome, and therefore the RHG response resembles a conditioned host reaction. Furthermore, the amount of S. gordonii and A. actinomycetemcomitans applied for practical reasons needs to be carefully interpreted as in vivo these bacteria constitute part of the complex oral microbiome with differential proportions which also vary among oral sites (Mark Welch et al., 2019). For future studies, it would be most interesting to expose RHG to a mixture of S. gordonii and A. actinomycetemcomitans, as enhanced virulence has been reported for this mixture.

In conclusion, we describe here a novel model to investigate human host-microbiome interactions which is stable for at least 3 days in aerobic culture conditions. Such a model has potential to investigate multi-species biofilms and the influence of antibiotics and probiotics in the future and importantly to replace current animal models which do not adequately represent human host-microbe interactions.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

This study did not involve any living humans or animals, neither did it involve biological samples derived from humans and therefore no ethical consent was required for this study. It only involved immortalized human cell lines and bacterial strains which were obtained as described in Methods.

YZ, methodology, investigation, formal analysis, data curation, and writing – original draft. LS, methodology, investigation, formal analysis, and writing – original draft. SR, methodology and investigation. BPK, writing – review and editing. SG, conceptualization, investigation, writing – original draft, writing – review and editing, supervision. DD, conceptualization, methodology, investigation, writing – original draft, writing – review and editing, and supervision. All authors gave their final approval and agreed to be accountable for all aspects of the work.

YZ is supported by China Scholarship Council (Number: 201808440350) for the research.

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

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.