Oral host-microbe interactions investigated in 3D organotypic models

Abstract

The oral cavity is inhabited by abundant microbes which continuously interact with the host and influence the host’s health. Such host-microbe interactions (HMI) are dynamic and complex processes involving e.g. oral tissues, microbial communities and saliva. Due to difficulties in mimicking the in vivo complexity, it is still unclear how exactly HMI influence the transition between healthy status and disease conditions in the oral cavity. As an advanced approach, three-dimensional (3D) organotypic oral tissues (epithelium and mucosa/gingiva) are being increasingly used to study underlying mechanisms. These in vitro models were designed with different complexity depending on the research questions to be answered. In this review, we summarised the existing 3D oral HMI models, comparing designs and readouts, discussing applications as well as future perspectives.

Keywords:

• Organotypic model
• oral mucosa
• host-microbe interaction
• oral microbiota
• biofilm

1. Introduction

Oral health and disease status is greatly influenced by dynamic interactions between microbes and the host. Health-related host-microbe interactions (HMI), which are maintained by a variety of factors such as dietary and hygienic interventions e.g. tooth brushing, benefit the microbial stability as well as the oral barrier functions (Moutsopoulos and Konkel 2018). Disrupting such HMI may result in dysbiotic microbial communities and dysfunctional host responses, leading to the onset and progression of local and even systemic pathologies e.g. periodontal diseases, cardiovascular infection, rheumatoid arthritis and Alzheimer’s disease (Hajishengallis 2015). The close correlation between oral microbial communities and the host status has been illustrated by studies investigating specific health or disease conditions using DNA sequencing, –omics technologies, microscopy, genetically-modified organisms and animal models (Hajishengallis 2015; Lamont et al. 2018; Valm 2019). However, it is still challenging to study the oral cavity as a whole system in order to understand how dynamic processes of HMI influence health and disease including transitional stages between homeostasis, resilience, pathogenesis and recovery.

Organotypic models are in vitro cell-culture systems representing three-dimensional (3D) structures and barrier functions of the host tissue. They have been developed to study the HMI of gut and skin (Barrila et al. 2018; Aguilar et al. 2021) and recently also the oral cavity – probably the most complex niche in a human body. Its complexity comes from the large number of diverse niches consisting of soft (e.g. oral mucosa) and hard (e.g. tooth) tissues, multi-species microbial communities (e.g. aerobic and anaerobic), as well as saliva and the gingival crevicular fluid (GCF) (Lamont et al. 2018; Mark Welch et al. 2019; Valm 2019). Previously, 2D oral HMI models were used to study specific microbial exposures, contributing to our knowledge on virulence factors of certain microbes and the corresponding host response (Lamont et al. 2018). However, most of these models used a single type of host cells (e.g. primary, immortalised or cancer-originated keratinocyte, fibroblast etc.) in combination with a single microbial species, which may lead to confounding results as they only represent selected HMI. Clinical studies avoid such disadvantages, however ethical rules only allow non- or less-invasive interventions and assessments, limiting the possibility of looking into the underlying mechanisms. To date, mechanistic insights into oral HMI have been mainly provided using animal models. By comparing germ-free mice with specific pathogen free mice, researchers showed for the first time that the absence of oral microbiota led to insufficient neutrophils infiltration in the periodontal junctional epithelium, thereby affecting the maintenance of periodontal health (Zenobia et al. 2013; Kirchner and LeibundGut-Landmann 2021). The invention of transgenic mice expressing human genes has greatly increased the translational effectiveness in studying human mechanisms (Schmidt et al. 2010). However, such models can only represent human situations to a certain extent when considering the intrinsic differences between human and mice in terms of the host immunity and the colonised microbiota (Dutta and Clevers 2017). Based on current understanding, reconstructing the in vivo complexity of the human oral cavity is challenging but is the key to mimicking the native oral HMI. Therefore, organotypic models are being increasingly appreciated as they are reconstructed using human-based material in a way that resembles the human oral physiology whilst allowing mechanistic investigations.

The aim of this review was to summarise the existing organotypic oral HMI models and to categorise these models by the level of complexity and the resemblance to the in vivo situation. Based on this, model applications and future perspectives were discussed.

2. Varying complexity of 3D co-culture models for studying oral HMI

To mimic oral HMI, ideally a host model should be able to harbour a viable microbial community, allow exchange of metabolites and nutrients with microbes, process microbial signals, respond accordingly, and eventually reach homeostasis within a certain time. From the view of the host, oral HMI may involve hard tissues as well as soft tissues (Mark Welch et al. 2019). As the hard tissues (e.g. enamel) contain comparatively low organic components, the corresponding HMI have been studied mainly focussing on biofilms formed on inorganic surfaces and their metabolism, composition and structures, which have been extensively reviewed (Dongari-Bagtzoglou 2008; Brown et al. 2019; Valm 2019). This review will focus on studies investigating HMI on soft tissues (oral epithelium/mucosa) specifically with the use of 3D organotypic models.

The earliest organotypic host models used tissue explants embedded in or cultured on top of different types of 3D gels to form an organ-like structure (Simian and Bissell 2017). For oral 3D models, this tissue is obtained by directly excising a part of oral mucosa from a human or animal (Ohnemus et al. 2008). However, due to ethical concerns as well as the limited availability of human oral mucosa and tremendous variability between donors (Sacks 1996), researchers started to develop in vitro reconstructed organotypic models as a scalable and reproducible substitute. These models were reconstructed to mimic important biological features of oral epithelium or oral mucosa (gingiva), such as the 3D tissue architecture and the communication between different types of host cells as well as the extracellular matrix (ECM). Organotypic oral models have shown their value in discovering novel drug targets, testing efficacy of drug delivery systems, and validating the biosafety and biocompatibility of graft materials for oral-maxillofacial surgery. These models have been extensively reviewed elsewhere (Moutsopoulos and Konkel 2018; Gibbs et al. 2019; Takahashi et al. 2019). Yet the oral microbiota, in spite of its huge impact on the host (patho)physiology, has not been broadly incorporated into these models. In the studies that have begun to do so, researchers have employed various representatives for microbes in order to mimic HMI of various oral niches. Previously, infection-related oral HMI models have been exclusively reviewed (Tabatabaei et al. 2020). As more is understood about the dynamic oral HMI and their influence on the transition between health and disease in the oral cavity, a review on the existing oral HMI models will help to realise to what extent the current models have captured the in vivo oral HMI in general, highlighting the potentials as well as discussing their limitations for future studies.

For this review, a literature search was conducted using the following filter in PubMed: time: 1990–2022; language: English; the major key words were: in vitro, oral, host-microbe interaction, 3D oral mucosa, 3D oral model, engineered oral mucosa, oral mucosa equivalents, reconstructed human oral mucosa, reconstructed human gingiva, oral bacteria, multi-species oral biofilm, oral microbe. After a manual check, 57 studies that specifically used 3D organotypic models were selected. Due to the varying model complexity, these studies were categorised and reviewed based on their different experimental designs regarding (1) host model, (2) microbial exposure and (3) the methods used to evaluate HMI.

2.1. Host model

2.1.1. Air-lifted transwell models

Most of the organotypic models used in studying oral HMI were cultured in transwell (Figures 1 and 2). The transwell model consists of two compartments: an upper culture compartment (a transwell insert with a porous membrane) hangs in a lower compartment (a petri dish or a single well of a multi-well culture plate). The upper compartment accommodates the host model while the lower compartment contains medium that supplies nutrition to the host model via the porous membrane. This setup enables the host model to be exposed to the air from above, whilst allowing the culture medium from below to promote the host model to differentiate into a stratified squamous oral epithelium i.e. reconstructed human oral epithelium (RHE). In addition to RHE, a reconstructed human oral mucosa (RHM) can be achieved by incorporating a fibroblast-populated ECM beneath the epithelium, resembling the oral mucosa which is composed of epithelium and lamina propria (Figures 1 and 2). The transwell-cultured RHE and RHM do not only represent the basic morphology and barrier function of oral epithelium and mucosa respectively (reviewed in (Bierbaumer et al. 2018)), they have also been further optimised to incorporate (i) local and recruited immune cells and mediators (Schaller et al. 2004; Bao, Belibasakis, et al. 2015; Bao, Papadimitropoulos, et al. 2015; Brown et al. 2019; Ollington et al. 2021), (ii) flow systems resembling the saliva flow or the GCF (Diaz et al. 2012; Bao, Belibasakis, et al. 2015; Bao, Papadimitropoulos, et al. 2015), and (iii) hard tissues e.g. teeth (Gursoy et al. 2012; Pöllänen et al. 2012) and restorative materials e.g. fillings and implants (Ingendoh-Tsakmakidis et al. 2019; Souza, Bertolini, Thompson, Barão, et al. 2020).

Figure 1. Modelling oral HMI using 3D organotypic models. Transwell models (left) mimic the stratified and differentiated oral epithelium (RHE) or the bi-layered oral mucosa (epithelium and lamina propria, RHM) and can be exposed to microbes both topically (direct application via epithelium) or systematically (via culture medium and transwell membrane). Spheroid models (right) mimic the bi-layered oral mucosa by allowing spontaneous attachment of epithelial cells to the pre-formed ECM-embedded fibroblasts. Spheroid models can only be exposed to microbes topically (indicated in the schematic diagram for spheroid model) via the surrounding culture medium (pictures reprinted from Bugueno et al. 2018 under Creative Commons Attribution 4.0 International License: https://creativecommons.org/licenses/by/4.0/). RHE: reconstructed human oral epithelium, RHM: reconstructed human oral mucosa, HMI: host-microbe interactions.

Figure 2. Varying complexity of current 3D oral HMI models. Oral HMI models are reconstructed with varying complexity. The Sankey plot illustrates the five main factors contributing to the model complexity, represented by the columns from the left to the right: the host system, host type, microbe type, species and the co-culture time. Of all the 57 studies included in this review, most (56) used transwells to culture RHE and RHM. Then these different types of host models were exposed to either bacteria, fungi or virus that were respectively composed of different numbers of species (from 1 species to ≥ 3 species). The co-cultures were then kept for certain periods of time varying from less than 24 h to more than 48 h. RHE, reconstructed human oral epithelium. RHM, reconstructed human oral mucosa. RHE/RHM + extra: RHE or RHM incorporated with extra elements, such as a flow, immune cells, tooth surfaces or dental implant materials. Time > 48 h + others: studies with microbial exposure longer than 48 h or exposure time not specified (for six studies with viral exposure).

With the transwell setup, HMI can be achieved in two ways: the host model is exposed to microbes via the epithelium side (environmental) or via the lamina propria side (systemic) (Figure 1). A remarkable advantage for recreating HMI in such 3D context is that some key outcomes, such as microbial invasiveness and the networking among different types of host cells, can be better represented and evaluated than in conventional 2D models. Microbial exposures that have been tested in transwell models were aerobes or facultative anaerobes. They allow easy cultivation in co-culturing with host cells, as air exposure is necessary to stimulate the stratification and differentiation of epithelium (Figure 2 and Table 1). Since strict anaerobes require an anaerobic environment which is not favourable for host cells, they have been rarely incorporated into transwell-cultured HMI models even though many strict anaerobes have been indicated to be closely associated with oral diseases. To date, behaviours of strict anaerobes have been mostly investigated without the presence of host cells. Interestingly, recent evidence indicated that transwell-cultured RHM may have potentials for studying oral HMI with anaerobes as they can host multi-species microbial consortiums where strict anaerobes have a better chance of survival. For example, one strict anaerobe Porphyromonas gingivalis was found with increased relative abundance in a multi-species biofilm after co-culturing with RHM for five days (Li et al. 2021). The transwell-cultured RHE and RHM are so far the most used 3D host models in studying oral HMI since they represent the construct of oral epithelium or mucosa and they can be exposed to complex microbial communities in physiologically relevant ways.

Table 1. Microbial exposures used in 3D oral co-culture models.

Number of species	Microbe	Ref
1	A. actinomycetemcomitans	(Dickinson et al. 2011; Paino et al. 2012; Ahlstrand et al. 2017; Beklen et al. 2019; Ingendoh-Tsakmakidis et al. 2019)
1	C. albicans	(Schaller et al. 2002; Claveau et al. 2004; Mostefaoui et al. 2004; Schaller et al. 2004; Tardif et al. 2004; Villar et al. 2005; Dongari-Bagtzoglou and Kashleva, 2006; 2006; Lermann and Morschhauser 2008; Decanis et al. 2009; Rouabhia et al. 2011; Semlali et al. 2011; Silva et al. 2011; Yadev et al. 2011; Whiley et al. 2012; Sobue et al. 2018)
1	C. glabrata	(Schaller et al. 2002; Silva et al. 2011; Sobue et al. 2018)
1	C. tropicalis	(Schaller et al. 2002)
1	F. nucleatum	(Gursoy et al. 2010; Dickinson et al. 2011; Gursoy et al. 2012; Pöllänen et al. 2012)
1	P. gingivalis	(Andrian et al. 2004; 2007; Groeger et al. 2010; Dickinson et al. 2011; Wayakanon et al. 2013; Pinnock et al. 2014; Bugueno et al. 2018)
1	S. mitis	(Sobue et al. 2018; Shang et al. 2020)
1	S. mutans	(Rahimi et al. 2018)
1	S. oralis	(Sobue et al. 2018; Ingendoh-Tsakmakidis et al. 2019)
1	S. salivarius	(Mostefaoui et al. 2004)
1	S. gordonii	(Dickinson et al. 2011)
1	Virus e.g. EBV, HPV	(Makielski et al. 2016; Phyu et al. 2017; Israr et al. 2018)
2	C. albicans, C. glabrata	(Silva et al. 2011)
2	C. albicans, S. oralis	(Diaz et al. 2012; Bertolini et al. 2015; Xu et al. 2016; Souza, Bertolini, Thompson, Barão, et al. 2020; Souza, Bertolini, Thompson, Mansfield, et al. 2020)
2	C. albicans, S. gordonii	(Souza, Bertolini, Thompson, Barão, et al. 2020)
2	C. albicans, S. mitis	(Souza, Bertolini, Thompson, Barão, et al. 2020)
2	C. albicans, S. sanguinis	(Souza, Bertolini, Thompson, Barão, et al. 2020)
2	C. albicans, S. aureus	(de Carvalho Dias et al. 2018)
2	HPV18, EBV	(Makielski et al. 2016)
3	S. mitis, S. intermedius, S. oralis	(Brown et al. 2019)
4	S. oralis, A. naeslundii, V. dispar, P. gingivalis	(Mikolai et al. 2020)
4	S. sanguinis, S. gordonii, A. viscosus, A. odontolyticus	(Morse et al. 2018)
5	C. albicans, S. sanguinis, S. gordonii, A. viscosus, A. odontolyticus	(Morse et al. 2018)
7	F. nucleatum, C. rectus, V. dispar, A. oris, S. anginosus, S. oralis, P. intermedia	(Belibasakis et al. 2013; Thurnheer et al. 2014; Belibasakis et al. 2015; Bostanci et al. 2015)
7	S. mitis, S. intermedius, S. oralis, F. nucleatum, F. nucleatum spp. vincentii, A. naeslundii, V. dispar	(Brown et al. 2019)
10	S. mitis, S. intermedius, S. oralis, F. nucleatum, F. nucleatum spp. vincentii, A. naeslundii, V. dispar, P. gingivalis, P. intermedia, A. actinomycetemcomitans	(Brown et al. 2019)
10	F. nucleatum, C. rectus, V. dispar, A. oris, S. anginosus, S. oralis, P. intermedia, P. gingivalis, T. forsythia, T. denticola	(Belibasakis et al. 2013; Thurnheer et al. 2014; Belibasakis et al. 2015; Bostanci et al. 2015)
11	P. intermedia, A. actinomycetemcomitans, C. rectus, V. dispar, F. nucleatum, S. oralis, T. denticola, A. oris, S. anginosus, T. forsythia, P. gingivalis	(Bao, Belibasakis, et al. 2015; Bao, Papadimitropoulos, et al. 2015)
Microcosm cultured from oral swab		(De Ryck et al. 2014)
Microcosm cultured from saliva		(Buskermolen et al. 2018; Shang et al. 2018; 2019)

2.1.2. Spheroid model

Spheroids are another type of 3D model used to mimic the oral HMI. Despite the extensive use of spheroids in studying gut HMI (Barrila et al. 2010), there is only one study demonstrating the interaction between P. gingivalis and an oral mucosa spheroid model (Bugueno et al. 2018). In general, spheroids are cultured initially from a droplet of fibroblast-embedded ECM, and then epithelial cells and optional immune cells are added to form the exterior epithelium (Figure 1). The submerged culture environment of spheroids can be static or fluidic (using microfluidic systems such as a bioreactor (Barrila et al. 2018)). Both conditions favour the cultivation of strict anaerobes such as P. gingivalis. Within the submerged environment, microbes can spontaneously attach to the epithelial surface of spheroids, mimicking the in vivo microbial attachment (e.g. in the gingival crevice). However, spheroid models have two main disadvantages to be used to mimic oral HMI. First, it is difficult to standardise the size, the epithelial differentiation and polarisation of spheroids, as indicated by spheroid gut models (Barrila et al. 2010, 2018). Second, the compact form of spheroids means that microbes can only interact with the host via the surface but not via the lamina propria, while both routes are available in vivo as well as with transwell models.

2.1.3. Organ-on-a-chip (OoC)

The two above mentioned 3D host models focus on mimicking the spatial structure of oral mucosa. In addition to this, the use of state-of-the-art OoC further ensures a microfluidic environment by perfusing culture chamber(s) via micro-channels. This mimics dynamic interactions between tissues and organs, thus enabling investigations on interactive processes and systemic effects (Delon et al. 2019; Koning et al. 2021). Gut, lung and skin OoC models are already in use to investigate human physiology and pathology including local HMI (Vahav et al. 2020; Baddal and Marrazzo 2021; Moysidou and Owens 2021). Only until recently, specific oral niches have been simulated using OoC: a salivary gland tissue chip (Song et al. 2021), a tooth-on-a-chip (França et al. 2020; Rodrigues et al. 2021; Soares et al. 2021) and a RHM gingiva-on-a-chip with incorporated Langerhans cells in the epithelium (Koning et al. 2021). These OoC have shown various advantages for screening drugs and testing material and substance compared to conventional monolayer cultures. For example, salivary gland OoC preserved phenotypes of acinar cells regarding the polarised morphology, Ca²⁺ signalling and secretory function which were rapidly lost in monolayer acinar cell cultures (Song et al. 2021). Therefore, they are suitable for screening radioprotective compounds that can be used for protecting salivary glands during radiation therapy for head and neck cancers. To the best of our knowledge, no OoC has been ever exposed to microbes. This is probably due to the technical difficulties of incorporating living microbes into the fluidic system. Nevertheless, OoC provide great potentials for future oral HMI studies to investigate the attachment and formation of biofilms on a host surface as well as continuous exchange of nutrients and metabolites between the host and the microbes with the presence of a dynamic flow.

2.2. Microbe exposure

2.2.1. Single-species

Even though the oral microbiota is composed of more than 700 species (Aas et al. 2005), the most practical and selected way to represent it in vitro is by using single bacterial, fungal, or viral species cultivated in their specific optimal conditions (Figure 2). Although simplifying the community to a specific species decreases the physiological relevance, it efficiently reduces the technical difficulties and allows us to exclusively investigate how a single microbe may interact with the host in every possible way. Until now, oral microbes that have been tested in single species in 3D HMI models are those known in 2D models or in vivo for playing key roles in influencing the host or interacting with other microbial species, e.g. Candida albicans, Fusobacterium nucleatum, Aggregatibacter actinomycetemcomitans and Streptococcus group (Table 1). Previously in 2D models, these fast-growing microbes (usually applied in single species) were found to be problematic in co-culturing as they often overgrew within 12 h, leading to acidified and nutrient-depleted environments where host cells were not able to maintain viability let alone to respond properly (Mostefaoui et al. 2004; Gursoy et al. 2010; Pinnock et al. 2014). While most 3D models managed to maintain the co-culturing with single-species microbes for at least 24 h without showing significant damage on the host histology and metabolic activity, especially when they were exposed to oral microbes with low or moderate virulence (Table 2).

Table 2. Readouts of the host-microbe interactions in 3D co-culture models.

Setup of co-cultures			Microbe				Host
Microbe	Host	Time	Growth	Colonisation / Invasion	Cellular Internalisation	Composition / metabolites	Tissue integrity	Viability	Proliferation	Apoptosis	Cell juntion	AMP	Cytokine*	Transcription	Proteomics	Reference
A. actinomycetemcomitans ATCC 29522 LPS	Mucosa	24h	–	–	–	–	–	–	–	–	–	–	↑	–	–	(Bedran et al. 2014)
A. actinomycetemcomitans D7S	Mucosa	8h	–	–	–	–	↘	–	↘	↓	–	–	↑	–	–	(Paino et al. 2012; Ahlstrand et al. 2017)
A. actinomycetemcomitans D7S	Mucosa	24h	–	–	–	–	↓	–	–	–	–	–	–	–	–	(Beklen et al. 2019)
C. albicans AC-1 (healthy isolate)	Epithelium	24h	–	↗	–	–	→	→	–	–	–	–	↗	–	–	(Whiley et al. 2012)
C. albicans NCYC 1467 (stomatitis)	Epithelium	24h	–	↑	–	–	↓	→	–	–	–	–	↑	–	–
C. albicans CAI4	Mucosa	24h	–	–	–	–	–	–	–	–	–	↑	–	↑	–	(Rouabhia et al. 2011)
C. albicans CAF2-1	Epithelium	24h	–	↑	–	–	↓	↓	–	–	–	→	↗	–	–	(Yadev et al. 2011)
	Mucosa	24h	–	↑	–	–	↓	↓	–	–	–	↑	↑	–	–
C. albicans clinical strain (candidiasis)	Mucosa	24h	–	–	–	–	–	–	–	–	–	↑	↑	↑	–	(Decanis et al. 2009)
C. albicans clinical strain (candidiasis)	Mucosa	24h	–	–	–	–	–	–	–	–	–	–	↑/↓	↑/↓	–	(Claveau et al. 2004)
C. albicans LAM-1	Mucosa	48h	–	–	–	–	–	–	–	–	–	→	↑	→	–	(Tardif et al. 2004)
C. albicans SC5314	Epithelium	48h	↑	↑	–	–	↓	↓	–	–	–	–	–	–	–	(Lermann and Morschhauser 2008)
C. albicans SC5314	Mucosa	24h	–	↑	–	–	↓	↓	↓	–	–	–	–	–		(Dongari-Bagtzoglou and Kashleva, 2006)
C. albicans SC5314	Epithelium	24h	–	–	–	–	↓	↓	–	–	–	–	↑	↑	–	(Schaller et al. 2004)
	Epithelium-PMN	24h	–	–	–	–	↘	↘	–	–	–	–	↑↑	↗	–
C. albicans SC5314	Mucosa	48h	↑	↑	–	–	–	–	–	–	–	–	–	–	–	(Villar et al. 2005)
C. albicans DAY25	Mucosa	48h	↑	↑	–	–	–	–	–	–	–	–	–	–	–
C. albicans SC5314 (WT)	Epithelium	48h	–	↑	–	–	↓	–	–	–	–	–	–	–	–	(Dongari-Bagtzoglou and Kashleva, 2006)
C. albicans SC5314 (WT)	Mucosa	48h	–	↑	–	–	↓	–	–	↑	–	–	↗	–	–
C. albicans rbt4/rbt4 (KO)	Mucosa	48h	–	↑	–	–	↓	–	–	–	–	–	↗	–	–
C. albicans efg1/efg1/cph1/cph1 (KO)	Mucosa	48h	–	→	–	–	→	–	–	–	–	–	→	–	–
C. albicans rim101/rim101 (KO)	Mucosa	48h	–	→	–	–	→	–	–	–	–	–	↗	–	–
C. albicans SC5314	Epithelium	36h	–	↑	–	–	↓	–	–	–	–	–	↑↑	–	–	(Schaller et al. 2002)
C. tropicalis DSM4959	Epithelium	36h	–	→	–	–	→	–	–	–	–	–	↑	–	–
C. glabrata clinical strain (candidiasis)	Epithelium	36h	–	→	–	–	→	–	–	–	–	–	↑	–	–
C. albicans SC5314	Mucosa	24h	–	–	–	–	–	–	–	–	–	→	↑/↓	→	–	(Semlali et al. 2011)
C. albicans SC5314 treated with KSL-W	Mucosa	24h	–	–	–	–	–	–	–	–	–	↓	↑/↓	↓	–
F. nucleatum ATCC 25586 biofilm	Mucosa	24h		↑	–	–	↓	↓	↓	–	–	–	–	–	–	(Gursoy et al. 2010)
F. nucleatum ATCC 25586 biofilm	Mucosa-dentin	24h	–	–	–	–	→	–	→	→	–	↑	–	–	–	(Gursoy et al. 2012; Pöllänen et al. 2012)
F. nucleatum AHN 9508 biofilm	Mucosa-dentin	24h	–	–	–	–	→	–	→	↓	–	↑	–	–	–
P. gingivalis ATCC 33277	Sphiroid mucosa	24h	–	↑	↑	–	↓	–	–	↑	↓	–	↑	–	–	(Bugueno et al. 2018)
P. gingivalis ATCC 33277	Epithelium	24h	–	↑	↑	–	↓	–	–	–	–	–	↑	↑	–	(Andrian et al. 2004; 2007)
P. gingivalis KDP128 (KO)	Epithelium	24h	–	↑	↗	–	↓	–	–	–	–	–	↑	↑	–
P. gingivalis clinical strain A245br	Mucosa	18h	–	↑	–	–	–	–	–	–	–	–	–	–	–	(Wayakanon et al. 2013)
P. gingivalis NCTC 11834/W50	Mucosa	4h	↑	↑	↑	–	↓	–	–	–	–	–	↑/↓	–	–	(Pinnock et al. 2014)
S. mitis LMG 14557	Mucosa	24h	→	↑	–	–	→	→	–	–	–	–	→	↑	→	(Shang et al. 2020)
S. mutans UA-159	Submerged oral mucosa-on-a-chip	24h	–	↑	–	–	→	–	–	–	↓	–	–	–	–	(Rahimi et al. 2018)
HSV-1 F strain	Mucosa	1w	–	↑	–	–	↓	–	–	–	–	–	–	–	–	(Syrjänen et al. 1996)
KSHV	Mucosa	?	–	↑	–	–	–	–	–	–	→	–	–	–	–	(Chen et al. 2020)
EBV	Mucosa	Prior	–	–	–	–	↓	–	↓	–	–	↑	–	–	–	(Nawandar et al. 2015)
HPV16	Mucosa	Prior	–	–	–	–	–	–	–	–	–	–	–	√	–	(Israr et al. 2018)
A. actinomycetemcomitans VT1169	Epithelium	24h	–	↑	–	–	–	–	–	↑	→	↑	–	–	–	(Dickinson et al. 2011)
F. nucleatum ATCC 25586	Epithelium	24h	–	↑	–	–	–	–	–	↑	→	↑	–	–	–
P. gingivalis ATCC 33277	Epithelium	2h	–	↑	↑	–	–	–	–	↑	→	→	–	–	–
P. gingivalis ATCC 33277	Epithelium	24h	–	↑	↑	–	–	–	–	→	→	→	–	–	–
S. gordonii DL-1	Epithelium	24h	–	→	–	–	–	–	–	↑	→	↑	–	–	–
A. actinomycetemcomitans JP2 strain	Mucosa-implant	24h	–	–	–	–	↓	–	–	–	→	–	↗	↗	–	(Ingendoh-Tsakmakidis et al. 2019)
S. oralis DSM 20627	Mucosa-implant	24h	–	–	–	–	→	–	–	–	→	–	↘	↘	–
C. albicans	Mucosa	48h	–	↑	–	–	↓	→	–	–	–	–		↑		(Mostefaoui et al. 2004)
S. salivarius ATCC 25975	Mucosa	48h	–	–	–	–	→	→	–	–	–	–	↑	↑
C. albicans SC5314	Mucosa-flow	24h	–	↑	–	–	↓	–	–	↑	–	–	–	–	–	(Diaz et al. 2012)
C. albicans SC5314 ● S. oralis 34	Mucosa-flow	24h	–	↑	–	–	↓	–	–	↑	–	–	–	–	–
C. albicans SN425 (WT)	Mucosa	16h	↑↑	↑	–	–	↓	↓	–	–	↓	–	–	–	–	(Bertolini et al. 2015; Xu et al. 2016)
C. albicans CJN2412 (KO)	Mucosa	16h	↑	↗	–	–	↓	↘	–	–	–	–	–	–	–
C. albicans CJN2412 ● S. oralis 34	Mucosa	16h	↑↑	↑	–	–	↓	↓	–	–	↓	–	–	–	–
C. albicans SC5314	Mucosa-implant	16h	–	↑	–	–	–	↓	–	–	–	–	↑	–	–	(Souza, Bertolini, Thompson, Barão, et al. 2020)
C. albicans SC5314 ● S. oralis 34	Mucosa-implant	16h	–	↑	–	–	–	↓	–	–	–	–	↑	–	–
C. albicans SC5314 ● S. gordonii CH1	Mucosa-implant	16h	–	↑↑	–	–	–	↓	–	–	–	–	↑	–	–
C. albicans SC5314 ● S. mitis ATCC 49456	Mucosa-implant	16h	–	↑	–	–	–	↓	–	–	–	–	↑	–	–
C. albicans SC5314 ● S. sanguinis SK36	Mucosa-implant	16h	–	↑	–	–	–	↓	–	–	–	–	↑	–	–
C. albicans SC5314	Mucosa	24h	→	↑	–	–	–	–	–	–	–	–	–	–	–	(Souza, Bertolini, Thompson, Mansfield, et al. 2020)
C. albicans SC5314 ● S. oralis 34 (WT)	Mucosa	24h	↑	↑	–	–	–	–	–	–	–	–	–	–	–
C. albicans SC5314 ● S. oralis ΔgtfR (KO)	Mucosa	24h	↑	↑	–	–	–	–	–	–	–	–	–	–	–
C. glabrata D1 or AE2 (oral)	Epithelium	12h	–	→	–	–	–	↘	–	–	–	–	–	–	–	(Silva et al. 2011)
C. albicans 324LA/94	Epithelium	12h	–	↑	–	–	–	↓	–	–	–	–	–	–	–
C. albicans 324LA/94 ● C. glabrata D1 or AE2	Epithelium	12h	–	↑↑	–	–	–	↓	–	–	–	–	–	–	–
C. albicans SC5314	Mucosa	16h	–	↑	–	–	→	↓	–	–	–	–	–	–	–	(de Carvalho Dias et al. 2018)
S. aureus ATCC 25923	Mucosa	16h	–	→	–	–	↓	↘	–	–	–	–	–	–	–
C. albicans SC5314 ● S. aureus ATCC 25923	Mucosa	16h	–	↑	–	–	↓	↓	–	–	–	–	–	–	–
HPV18	Epithelium	Prior	–	–	–	–	→	–	↑	–	–	–	–	–	–	(Makielski et al. 2016)
EBV	Epithelium	Prior	–	→	–	–	→	–	↑	–	–	→	–	–	–
HPV18 ● EBV	Epithelium	prior	–	↑	–	–	→	–	↑	–	–	↑	–	–	–
C. albicans ATCC 90028	Epithelium	12h	–	–	–	–		↓	–	–	–	–	–	–	–	(Morse et al. 2018)
	Mucosa	12h	–	–	–	–	↓	↘	–	–	–	–	–	–	–
Dental biofilm (S. sanguinis ATCC 10556, S. gordonii ATCC 10558, A. viscosus ATCC 15987, A. odontolyticus NCTC 9935)	Epithelium	12h	–	–	–	–		→	–	–	–	–	–	–	–
	Mucosa	12h	–	–	–	–	→	→	–	–	–	–	–	–	–
C. albicans ATCC 90028 ● Dental biofilm (S. sanguinis ATCC 10556, S. gordonii ATCC 10558, A. viscosus ATCC 15987, A. odontolyticus NCTC 9935)	Epithelium	12h	–	–	–	↑		↓↓	–	–	–	–	–	–	–
	Mucosa	12h	–	–	–	↗	↓	↓	–	–	–	–	–	–	–
S. oralis ATCC 9811 ● A. naeslundii DSM 43013 ● V. dispar DSM 20735 ● P. gingivalis dsm 20709 biofilm	Mucosa-implant	24h	↑	↗	–	→	→	–	–	–	–	→	↗	→	–	(Mikolai et al. 2020)
		48h	↓	↑	–	changed	↓	–	–	–	–	↘	↗	↑	–
		24h	–	–	–	–	↓	–	↓	↑	–	–	→	–	–
Subgingival biofilm include red-complex (C. rectus OMZ 697, F. nucleatum OMZ 596, P. intermedia ATCC 25611, V. dispar ATCC 17748, A. oris OMZ 745, S. anginosus OMZ 871, and S. oralis SK 248, P. gingivalis ATCC 33277, T. forsythia OMZ 1047, T. denticola ATCC 35405)	Epithelium	3h	–	–	–	–	–	→	–	–	↑	–	↑	√	–	(Belibasakis et al. 2013; Thurnheer et al. 2014; Belibasakis et al. 2015; Bostanci et al. 2015)
		24h	–	–	–	→[C]	↘	→	–	–	↓	–	→	√	√
		48h	–	↑↑	–	→[C]	↓	–	–	–	–	–	–	–	√
Subgingival biofilm exclude red-complex (C. rectus OMZ 697, F. nucleatum OMZ 596, P. intermedia ATCC 25611, V. dispar ATCC 17748, A. oris OMZ 745, S. anginosus OMZ 871, and S. oralis SK 248)	Epithelium	3h	–	–	–	–	–	→	–	–	↑	–	↗	√	–
		24h	–	–	–	↓[C] streptococci	↘	→	–	–	↓	–	→	√	√
		48h	–	↑	–	↑[C] streptococci	↓	–	–	–	–	–	–	–	√
Subgingival biofilm (C. rectus OMZ 398, F. nucleatum OMZ 598, P. intermedia ATCC 25611, V. dispar ATCC 17748, A. oris OMZ 745, S. anginosus OMZ 871, and S. oralis SK 248, P. gingivalis W50 33277, T. forsythia OMZ 1047, T. denticola ATCC 35405, A. actinomycetemcomitans JP2)	Mucosa-flow, Mono-Mac-6	24h	–	↑	–	changed	↓	–	–	–	–	–	↑	–	√	(Bao et al. 2015; 2015)
Healthy-associated biofilm (S. mitis NCTC 12261, S. intermedius DSM 20753, S. oralis NTCC 11427)	Epithelium	24h	–	–	–	–	–	→	–	–	–	–	→	→	–	(Brown et al. 2019)
Gingivitis-associated biofilm (S. mitis NCTC 12261, S. intermedius DSM 20753, S. oralis NTCC 11427, F. nucleatum ATCC 10596, F. nucleatum spp. vincentii DSM 19507, A. naeslundii DSM 17233, V. dispar NCTC 11831)	Epithelium	24h	–	–	–	–	–	→	–	–	–	–	↑	↑	–
	Epithelium-PBMC	24h	–	–	–	–	–	–	–	–	–	–	–	√	–
	Epihelium-PBMC/MoCD14+	24h	–	–	–	–	–	–	–	–	–	–	–	√	–
Periodontitis-associated biofilm (S. mitis NCTC 12261, S. intermedius DSM 20753, S. oralis NTCC 11427, F. nucleatum ATCC 10596, F. nucleatum spp. vincentii DSM 19507, A. naeslundii DSM 17233, V. dispar NCTC 11831, P. gingivalis W83, P. intermedia DSM 20706, A. actinomycetemcomitans ATCC 4371)	Eptihelium	24h	–	–	–	–	–	→	–	–	–	–	↑	↑	–
Microcosm cultured from oral swab	Mucosa	72h	↓	–	–	↑ [C]	→	→	–	–	–	–	–	–	–	(De Ryck et al. 2014)
Saliva-derived microcosm-commensal	Mucosa	24h	–	–	–	–	→	–	–	–	–	↑	↑	↑	–	(Buskermolen et al. 2018; Shang et al. 2019)
Saliva-derived microcosm-gingivitis	Mucosa	24h	–	–	–	–	→	–	–	–	–	↑	↗	↗	–
Saliva-derived microcosm-cariogenic	Mucosa	24h	–	–	–	–	→	–	–			↑	→	→	–
Saliva-derived microcosm-commensal	Mucosa	7d	↓	↑	–	–	↗	→	→	–	–	↑	↑	↑	–	(Shang et al. 2018)

√: multiple markers investigated, showed differential change, [C]: with regards to composition, h: hour, d: day, w: week, -: not applicable, arrows suggest regulation: significant upregulation ↑/↑↑, significant downregulation ↓/↓↓, trending of upregulation ↗, trending of downregulation ↘, →: no significant regulation, ↑/↓: both up and downregulation were observed, Mucosa: reconstructed human mucosa (RHM), Epithelium: reconstructed human epithelium (RHE).

2.2.2. Dual-species and multi-species

Compared to single species, dual- or multi-species microbial exposures are theoretically more representative to be used in studying oral HMI, as they reflect internal interactions between microbial species. Such interactions can be synergistic or antagonistic, which will ultimately contribute to the HMI (Falsetta et al. 2014; Pasman et al. 2022; Perpich et al. 2022). For example, a synergistic relationship between C. albicans and S. gordonii was found to enhance the antimicrobial resistance of the dual-species biofilm to clindamycin in vitro (Montelongo-Jauregui et al. 2019). However, not much has been studied regarding how inter-species interactions affect the HMI as well as the host response. Among the 57 studies included in this review, 7 different combinations of dual species were investigated using 3D HMI models. Interestingly, 6 out of 7 studies used C. albicans, in combination with either non-albicans Candida species or bacterial species (Figure 2 and Table 1). These studies showed versatile roles of C. albicans in the microbial community as well as in the oral HMI: certain dual-species combinations formed stable biofilms on top of 3D host models, contributing to the survival of commensal bacteria (e.g. C. albicans and S. gordonii (Souza, Bertolini, Thompson, Barão, et al. 2020)), while other combinations resulted in enhanced detrimental effect on the host: decreased cell viability, interrupted cell junctions, increased apoptosis as well as significantly upregulated transcription and secretion of multiple inflammatory cytokines (Table 2). Moreover, in the only study investigating dual viral exposure (HPV18 and EBV), no significant damage on the host was observed compared to single viral infection (Table 2).

Multi-species biofilms used in studying oral HMI in 3D models can be divided into two types (Figure 2, 9 out of 57 studies, details in Table 1). One is synthetic biofilms consisting of varying numbers (3–11) of in-vivo prevalent and/or well-studied oral microbes: Streptococcus, Actinomyces, Veillonella, Fusobacterium, Prevotella, Campylobacter, Candida and periodontitis-associated Treponema denticola, Tannerella forsythia and P. gingivalis. An advantage of synthetic biofilms is that they can be modified to reach the desired species richness, species evenness and biomass before being co-cultured with the host model. Thus profiles of these biofilms can be related to a specific host response. Brown et al. observed that RHM responded differently to multi-species oral biofilms composed of 3-, 7- or 10-species after a 24 h co-culturing. Using the unexposed RHM as a reference, the authors described no significant changes in CXCL8 secretion when RHM were exposed to a 3-species commensal biofilm. In comparison, increased transcription and release of CXCL8 were found with the exposure of either a 7-species gingivitis biofilm or a 10-species periodontitis biofilm (Brown et al. 2019).

The other type of multi-species biofilms used in 3D oral HMI models is clinical samples-derived biofilms from e.g. saliva (salivary microbiota) or cheek swabs (buccal mucosa microbiota) (Table 1). In theory, these biofilms should resemble the in vivo oral microbiota the most in terms of the composition and diversity. Yet to what extent can these biofilms maintain their representativeness still needs to be investigated: the community properties may be significantly affected by different in vitro co-culturing conditions. There are only two studies where biofilm compositions were analysed post HMI and altered community structure as well as decreased diversity were indeed observed (De Ryck et al. 2014; Li et al. 2021).

2.3. Evaluation methods and outcome

2.3.1. Microbe

Changes in properties of a microbial community, such as the 3D biofilm structure, composition, diversity and metabolism, are important indicators when evaluating how relevant an HMI model is to the native oral HMI (Tables 2 and 3). However currently, it is technically difficult to monitor these microbial characteristics in real time within the 3D construct. In the existing 3D oral HMI models, researchers have used two methods to assess such characteristics: (1) without host: biofilms were cultured and assessed in a separate model without host, under similar conditions to the HMI model, or (2) with host: the entire co-culture tissue was harvested for visualising microbes, retrieving living microbial cells or extracting microbial DNA at certain time points. With the first method, it is possible to extensively study biofilm composition and viability (Ingendoh-Tsakmakidis et al. 2019; Mikolai et al. 2020), but these results may not represent the true nature of biofilm activities during HMI. For instance, a 4-species biofilm was cultured either by itself or with a peri-implant RHM for 48 h (Mikolai et al. 2020). The biofilm composition was assessed in the biofilm-alone culture: three out of the four species were found with changed ratio: S. oralis and V. dispar increased, A. naeslundii decreased and the ratio of P. gingivalis did not change. With the second method, biofilms were investigated at specific HMI time points for microbial colonisation, gene expressions, metabolites and compositions (for multi-species biofilms). Studies using the second method have demonstrated microbial colonisation on the host tissue by light microscopy, confocal laser scanning microscopy (CLSM) and transmission electron microscopy (TEM) in combination with immunohistochemistry (IHC) staining or fluorescence in situ hybridisation (FISH) of tissue sections (Table 3). Some studies even detected invaded microbes in deeper tissue or intracellularly e.g. C. albicans and P. gingivalis (Whiley et al. 2012; Pinnock et al. 2014). With single-species microbial exposure, co-culturing with host models was shown to regulate expression of microbe-specific genes and production of key metabolites. For example, co-culturing with RHE or RHM was found to up-regulate several genes of C. albicans (ALS1, ALS3, EPA1) which are involved in C. albicans hyphal growth (Morse et al. 2018; Swidergall et al. 2021). With synthetic multi-species biofilms, HMI in vitro significantly influenced biofilm composition and metabolites. Thurnheer et al. inoculated an RHE with a 7-species biofilm consisting of similar amounts of each species (individual bacterial pre-cultures were adjusted to the same optical density and mixed at equal ratio), and found all of these species had different viable bacterial cell counts after 24 h or 48 h (Thurnheer et al. 2014). Bao et al. showed that co-culturing with an RHM for 24 h decreased viable bacterial cell counts of multiple species in an 11-species subgingival biofilm (Bao, Papadimitropoulos, et al. 2015). In a follow-up study, they described that the profile of bacterial proteins in the biofilm-RHM co-culture supernatants were different from those secreted by the biofilm when cultured alone (Bao, Belibasakis, et al. 2015).

Table 3. Methods and biomarkers used to evaluate host-microbe interactions in 3D co-culture models.

Microbe

Host

Microbial viability Colony forming unit (CFU) (Pinnock et al. 2014; Shang et al. 2018; 2020; Souza, Bertolini, Thompson, Mansfield, et al. 2020; Zhang et al. 2022) CFU/biomass of parallel microbe-only culture (Ingendoh-Tsakmakidis et al. 2019,Paino et al. 2012) Live/dead bacterial staining (De Ryck et al. 2014; Mikolai et al. 2020) Intercellular survival assay (Pinnock et al. 2014) Biofilm morphology FISH (Dickinson et al. 2011; Silva et al. 2011; Dabija-Wolter et al. 2012; Diaz et al. 2012; Pinnock et al. 2014; Thurnheer et al. 2014; Makielski et al. 2016; Xu et al. 2016; Bugueno et al. 2018; Buskermolen et al. 2018; Shang et al. 2018; Sobue et al. 2018; Mikolai et al. 2020; Shang et al. 2020; Souza, Bertolini, Thompson, Barão, et al. 2020; Souza, Bertolini, Thompson, Mansfield, et al. 2020; Zhang et al. 2022) or other stainings (Toludine blue, Periodic Acid-Schiff (PAS)) and/or microscopy (light microscopy, CLSM, TEM) (Bugueno et al. 2018; Schaller et al. 2004; Gursoy et al. 2010; Pinnock et al. 2014; Mikolai et al. 2020; Thurnheer et al. 2014; Dickinson et al. 2011; Dongari-Bagtzoglou and Kashleva 2006; Yadev et al. 2011; Schaller et al. 2002; Andrian et al. 2004; Wayakanon et al. 2013) Metabolic products Biofilm matrix (a-glucans (Souza, Bertolini, Thompson, Mansfield, et al. 2020), mucins (De Ryck et al. 2014)) Gene expression related with microbial virulence (Morse et al. 2018) Proteomic analysis on biofilm lysate (Bao, Belibasakis, et al. 2015) Others: short-chain fatty acids, Lactate, pH (De Ryck et al. 2014) Microbial activity Internalisation of host molecule (IL-1β) detected via immuno-electron microscopy (Paino et al. 2012) Microbial composition DGGE profiling (De Ryck et al. 2014) Sequencing (De Ryck et al. 2014; Li et al. 2021)

Tissue histology Staining (H&E, van Gieson, Masson trichrome) and/or microscopy (light microscopy, scanning electron microscopy (SEM), TEM) (Schaller et al. 2002; Dongari-Bagtzoglou and Kashleva 2006; Andrian et al. 2004; Shang et al. 2020; Mostefaoui et al. 2004; Schaller et al. 2004; Villar et al. 2005; Dongari-Bagtzoglou and Kashleva 2006; Yadev et al. 2011; Gursoy et al. 2010; Diaz et al. 2012; Pöllänen et al. 2012; Whiley et al. 2012; Pinnock et al. 2014; Bertolini et al. 2015; Makielski et al. 2016; Bugueno et al. 2018; Buskermolen et al. 2018; de Carvalho Dias et al. 2018; Morse et al. 2018; De Ryck et al. 2014; Shang et al. 2018; Sobue et al. 2018; Ingendoh-Tsakmakidis et al. 2019; Beklen et al. 2019; Mikolai et al. 2020; Souza, Bertolini, Thompson, Barão, et al. 2020; Zhang et al. 2022) Thickness of the epithelium (Paino et al. 2012; Shang et al. 2018) Tissue viability Cell counting (Mostefaoui et al. 2004; 2004; De Ryck et al. 2014) Lactate dehydrogenase (LDH) release (Dongari-Bagtzoglou and Kashleva 2006; Gursoy et al. 2010; Lermann and Morschhauser 2008; Silva et al. 2011; Yadev et al. 2011; Belibasakis et al. 2013; De Ryck et al. 2014; Xu et al. 2016; de Carvalho Dias et al. 2018; Morse et al. 2018; Brown et al. 2019; Souza, Bertolini, Thompson, Barão, et al. 2020) MTT (Whiley et al. 2012; De Ryck et al. 2014; Shang et al. 2020; Zhang et al. 2022) Expression of biomarkers (related function) Transcription RT-PCR (Schaller et al. 2002; Claveau et al. 2004; Mostefaoui et al. 2004; Schaller et al. 2004; Tardif et al. 2004; Semlali et al. 2011; Andrian et al. 2007; Decanis et al. 2009; Rouabhia et al. 2011; Belibasakis et al. 2013; 2015; Bugueno et al. 2018; Israr et al. 2018; Shang et al. 2018; Brown et al. 2019; Ingendoh-Tsakmakidis et al. 2019; Shang et al. 2019; Mikolai et al. 2020; Shang et al. 2020) IHC staining (related function) Ki67 (proliferation) (Gursoy et al. 2010; Paino et al. 2012; Pöllänen et al. 2012; De Ryck et al. 2014; Shang et al. 2018; Ingendoh-Tsakmakidis et al. 2019; Shang et al. 2020; Zhang et al. 2022) Cytokeratin 5/7/10/14/19, pan-cytokeratin (differentiation) (Bugueno et al. 2018; Beklen et al. 2019; Ingendoh-Tsakmakidis et al. 2019; Zhang et al. 2022) Vimentin, phalloidin (cytoskeleton) (Bugueno et al. 2018) Laminin V (ECM) [33] Annexin-V, caspase-3, APAF-1 (apoptosis) (Dickinson et al. 2011; Diaz et al. 2012; Pöllänen et al. 2012; Bertolini et al. 2015) E-cadherin, claudin (cellular attachment) (Dongari-Bagtzoglou and Kashleva, 2006; De Ryck et al. 2014; Xu et al. 2016; Bugueno et al. 2018; Ingendoh-Tsakmakidis et al. 2019) hBD2, hBD3, LL-37, Elafin (antimicrobial response) (Yadev et al. 2011; Gursoy et al. 2012; Buskermolen et al. 2018; Shang et al. 2018) IL-6, CXCL8, IL-1 (pro-inflammatory response) (Ingendoh-Tsakmakidis et al. 2019) Integrin β1 (cell surface receptor) (Bugueno et al. 2018) Calpain-1 (cell activity) (Xu et al. 2016) Western blotting Caspase-3 p11 (apoptosis) (De Ryck et al. 2014) IL-18, ICE (Tardif et al. 2004) TLRs (Shang et al. 2020) Secretion ELISA, Luminex (Schaller et al. 2002; Claveau et al. 2004; Mostefaoui et al. 2004; Schaller et al. 2004; Tardif et al. 2004; Semlali et al. 2011; Dongari-Bagtzoglou and Kashleva, 2006; Andrian et al. 2007; Decanis et al. 2009; Dickinson et al. 2011; Rouabhia et al. 2011; Andrian et al. 2004; Yadev et al. 2011; Paino et al. 2012; Belibasakis et al. 2013; Ahlstrand et al. 2017; Whiley et al. 2012; Buskermolen et al. 2018; Shang et al. 2018; Sobue et al. 2018; Brown et al. 2019; Ingendoh-Tsakmakidis et al. 2019; Mikolai et al. 2020; Shang et al. 2020; Souza, Bertolini, Thompson, Barão, et al. 2020; Zhang et al. 2022) Cytokine immunoblot assay (Pinnock et al. 2014) Proteomic analysis (Bao, Belibasakis, et al. 2015; Bostanci et al. 2015) Others TEER (cellular attachment) (Dickinson et al. 2011; Rahimi et al. 2018)

2.3.2. Host

In comparison to the limited results on microbes, host responses have been much more investigated in current 3D oral HMI models regarding two main outcomes: the tissue integrity (viability, histology, permeability etc.) and the expression of functional biomarkers (Table 3). Different from 2D models, HMI in 3D models are recreated in a physiologically relevant way, mimicking the native process of oral HMI starting from the topical microbial exposure on the epithelium. Furthermore, the HMI co-culture time, has been extended from the previous maximum 24 h in 2D HMI models, to longer than 24 h in 3D models without significant damage to the host tissues (Stathopoulou et al. 2010). With such 3D construct and extended co-culture time, not only the colonising and invading microbes can be visualised, but also the corresponding tissue histology and barrier function-related biomarkers, in situ as well as secreted, can be investigated. These host outcomes were then correlated with the corresponding microbial exposure to determine how closely these models could mimic the native oral HMI.

When exposed to low-virulent microbes, 3D host models were found to maintain intact tissue histology while showing upregulated pro-inflammatory responses and antimicrobial peptide (AMP) expression. For example, when exposed to single oral resident bacteria of low or moderate virulence e.g. S. mitis, S. gordonii and F. nucleatum, RHM maintained viability and showed protective responses after 24 h (Dickinson et al. 2011; Gursoy et al. 2012; Pöllänen et al. 2012; Shang et al. 2020; Zhang et al. 2022). With regards to multi-species biofilms, RHM also remained viable and responsive when exposed to a saliva-derived commensal biofilm over 24 h with increased AMP expression (Elafin, human beta defensing (hBD) 2, hBD3), up-regulated toll-like receptor signalling pathways and increased cytokine production (Buskermolen et al. 2018; Shang et al. 2018; Shang et al. 2019). These findings are in line with some known characteristics of a healthy gingiva: it constantly releases defensive molecules such as hBDs, essential in maintaining health (Gursoy and Kononen 2012).

In contrast, with specific detrimental microbial exposures, 3D models may end up with excessive tissue damage (Table 2). These findings represent the significant tissue damage occurred during in vivo oral infections e.g. periodontitis (Sedghi et al. 2021). Such responses were often observed in vitro as disrupted tissue integrity, decreased expression of defensive biomarkers as well as drastic changes in cytokine release in comparison to the intra-experimental ‘healthy’ controls. One study exposed RHM to different strains of C. albicans and several highly invasive strains showed a more detrimental effect than other less virulent and knock-out strains (increased LDH release; increased expression of apoptotic markers such as Caspase-3, Annexin V; increased DNA cleavage; decreased expression of cellular junction markers as indicated in Table 2) (Dongari-Bagtzoglou and Kashleva, 2006). In other studies, after 24 h exposure to two pathogenic biofilms with phenotypes representing either gingivitis (high proteolytic activity) or cariogenic biofilms (high lactic acid production), the viability and histology of an RHM was not affected but remarkably less AMP and cytokines were expressed (Buskermolen et al. 2018; Shang et al. 2018, 2019). However, not all well-known pathogens can cause such extreme host damage. For example, exposure to P. gingivalis did not cause disrupted cell junctions nor increase AMP expression in RHE and RHM (Dickinson et al. 2011; Bugueno et al. 2018). Such insignificant changes in 3D host models after P.gingivalis exposure may support the known characteristic of P. gingivalis for being able to evade immune surveillance (Zheng et al. 2021).

Cytokine secretion is one of the most evaluated host responses in 3D oral HMI models. However, outcomes of different studies cannot be compared directly but need to be interpreted individually, due to differences in model complexity or experimental design. Commonly investigated cytokines are IL-6, IL-8 (CXCL8), IL-1α, IL-1β, TNF-α, CXCL1, CXCL2, CCL2, CCL5, CCL20, MMPs and TIMPs as they are previously found to be involved in the native oral HMI (Dickinson et al. 2011; Bao, Belibasakis, et al. 2015; Abhyankar et al. 2019; Diesch et al. 2021). Pro-inflammatory cytokines e.g. IL-6, TNF-α were upregulated by most microbial exposures in 3D oral HMI models regardless of the experimental design, suggesting their important roles in maintaining host-microbe homeostasis as well as in defending against microbe invasion. In addition, regulation of some cytokines have been associated with specific microbial exposures, such as C. albicans – MMP-2, TIMP-2 (Claveau et al. 2004) and IL-10 (Schaller et al. 2004), S. oralis – CCL2 (Ingendoh-Tsakmakidis et al. 2019), A. actinomycetemcomitans – CXCL1 and CCL2 (Ingendoh-Tsakmakidis et al. 2019), P. gingivalis – IL-1α, TNF-α, CXCL10, CCL2, CCL5 and TIMP-2 (Pinnock et al. 2014).

Changes in intercellular junctions in epithelium is another important host outcome. However, results from the current 3D oral HMI models were inconclusive about how intercellular junctions were affected by microbe exposure, due to different experimental design and readouts. The most used assessments were trans-epithelial electrical resistance (TEER) (Dickinson et al. 2011; Rahimi et al. 2018; De Rudder et al. 2020) and microscopic visualisation of junctional components such as E-cadherin, Claudins and desmosomes after IHC or directly by TEM (Belibasakis et al. 2015; Xu et al. 2016; Bugueno et al. 2018; Ingendoh-Tsakmakidis et al. 2019). One study used IHC and showed that the expression of E-cadherin in RHM was not affected by either S. oralis or A. actinomycetemcomitans biofilms (Ingendoh-Tsakmakidis et al. 2019). In two other studies, the expression of E-cadherin decreased in RHM when exposed to C. albicans alone and to C. albicans in combination with S. oralis (Bertolini et al. 2015; Xu et al. 2016). TEER did not change in RHE after single-species exposure with S. gordonii or three other bacterial species (Dickinson et al. 2011), but decreased TEER was observed when a submerged RHM was exposed to another Streptococcus species – S. mutans (Rahimi et al. 2018). Such discrepancies in outcomes may be caused by the use of different readout methods as well as the different model complexity.

In conclusion, host outcomes, especially the change of tissue integrity, appeared to correlate well with the type of corresponding microbial exposure. However, these results should be carefully interpreted as they only represent one specific HMI condition. In particular, comparisons should only be made within the same study with strict controls such as unexposed condition. Although currently, a large amount of data are available regarding host outcomes from 3D as well as 2D oral HMI studies, we need to keep in mind that the difference described in each study is relative and may vary significantly depending on the level of model complexity. Most importantly, due to lack of relevant in vivo human healthy and disease data it is difficult to correlate the above in vitro findings with the complex host-microbe interactions in the living dynamic oral cavity and therefore interpretation of results still needs to be done with caution.

3. Influencing factors

3.1. Type of microbial exposure

While reviewing related HMI models, we noticed a couple of factors that may greatly affect the outcomes. First and the most significant is the type of microbial exposure (e.g. the taxonomy and the number of species). In animal models, 2D in vitro models, and clinical studies, oral microbes were generally correlated at the species level with their commensal or pathogenic influence on the host. However, host responses in 3D HMI models were found to differ greatly depending on the strain level of microbes, e.g. RHM was shown to respond differently to strains of C. albicans with different pathogenic behaviours (Dongari-Bagtzoglou and Kashleva, 2006; Gursoy et al. 2010). Also, the number of included species in a microbial exposure affected the host response to a great extent. When a multi-species exposure was compared with single-species exposures in the same study, the host responses were found to be more pronounced with multi-species exposure (Bertolini et al. 2015; Xu et al. 2016; Morse et al. 2018; Souza, Bertolini, Thompson, Barão, et al. 2020). Interestingly, all studies comparing single-species versus multi-species exposures in 3D oral HMI models similarly focussed on investigating C. albicans: whether the addition of bacteria to Candida (Candida single-species vs. Candida-bacteria multi-species) would result in more host damage. Indeed, Candida-bacteria exposure led to higher microbial colonisation and invasion as well as lower host viability, higher apoptosis, decreased expression level of cellular junctions, and upregulated release of (pro-)inflammatory cytokines.

However, if we review different 3D HMI models for the regulation of a specific host outcome, models using multi-species microbial exposures do not necessarily present more significant difference (compare to the unexposed condition in the same study) than models using single-species exposure. This may result from a combination of factors, such as, with different microbial species in combination with different host models or co-culture periods. For example, expressions of hBDs appeared to be affected by all above-mentioned factors: within 24 h, C. albicans up-regulated hBD2 in both RHM and RHE (Yadev et al. 2011) and two strains of F. nucleatum increased hBD3 in an RHM (Bedran et al. 2014), while a 4-species peri-implant biofilm (S. oralis, A. naeslundii, V. dispar, P. gingivalis) maintained but did not significantly regulate the secretion of hBD1 or hBD2 in a peri-implant RHM (Mikolai et al. 2020). Similarly, with one commonly investigated chemokine CXCL8: single species of commensals (S. salivarius (Mostefaoui et al. 2004), S. gordonii (Dickinson et al. 2011)) and opportunistic pathogens (F. nucleatum, C. albicans (Souza, Bertolini, Thompson, Barão, et al. 2020; Dongari-Bagtzoglou and Kashleva, 2006; Yadev et al. 2011; Schaller et al. 2002), P. gingivalis (Andrian et al. 2004; Pinnock et al. 2014)), as well as a 7-species gingivitis-associated and a 10-species periodontitis-associated biofilm (Brown et al. 2019) were all able to up-regulate CXCL8 expression. However, CXCL8 was not increased by a 4-species peri-implant biofilm in RHM (Mikolai et al. 2020) or a 3-species healthy-associated biofilm in a monocytes-incorporated RHE (Brown et al. 2019). CXCL8 secretion was even found to decrease after the exposure to a 7-species subgingival biofilm or this biofilm plus the periodontitis-related red-complex (P. gingivalis, T. forsythia, T. denticola) in RHE (Belibasakis et al. 2013; Thurnheer et al. 2014; Bostanci et al. 2015). The transcription of CXCL8 in an RHE was up-regulated after 3 h exposure to a 7-species subgingival biofilm, while the secretion of CXCL8 decreased after 24 h exposure to the same biofilm (Belibasakis et al. 2013; 2015). Furthermore, the host transcription and protein synthesis may not always correspond, e.g. increased transcription of IL-1 converting enzyme (ICE) but decreased protein expression were found when an RHE was exposed to C. albicans and measured every 6 h over a 48 h co-culturing (Tardif et al. 2004).

All in all, the outcomes from current 3D oral HMI models are interpreted mainly based on the known commensal or pathogenic phenotypes of the microbial exposure. However, given that microbes may behave differently from in vivo to in vitro and from one type of host model to another, ultimately, more factors should be taken into consideration as the model complexity increases, such as the setup of the host model, the co-culture time and environment, and the markers chosen as readouts.

3.2. Source of host cells

In contrast to the significant influence of microbial exposures, the reported host responses were more consistent and less influenced by the source of host cells. Cancer cell lines (Schaller et al. 2002; Lermann and Morschhauser 2008; Silva et al. 2011; Yadev et al. 2011; Whiley et al. 2012; De Ryck et al. 2014; Pinnock et al. 2014; Morse et al. 2018; Souza, Bertolini, Thompson, Barão, et al. 2020), immortalised cell lines (Dongari-Bagtzoglou and Kashleva, 2006; 2006; Groeger et al. 2010; Gursoy et al. 2010; Diaz et al. 2012; Paino et al. 2012; Pöllänen et al. 2012; De Ryck et al. 2014; Bao, Belibasakis, et al. 2015; Bao, Papadimitropoulos, et al. 2015; Xu et al. 2016; de Carvalho Dias et al. 2018; Morse et al. 2018; Rahimi et al. 2018; Beklen et al. 2019; Ingendoh-Tsakmakidis et al. 2019) and primary cells (Andrian et al. 2004; Mostefaoui et al. 2004; Villar et al. 2005; Dongari-Bagtzoglou and Kashleva 2006; Andrian et al. 2007; Dickinson et al. 2011; Semlali et al. 2011; Pinnock et al. 2014; Thurnheer et al. 2014; Bostanci et al. 2015; Bugueno et al. 2018; Brown et al. 2019) were the three major host cell sources used in 3D oral HMI models. After exposing to certain microbes e.g. C. albicans, similar results were found in host models using either primary keratinocytes or keratinocyte cell lines, regarding their tissue histology, LDH release, hBD2 expression and TNF-α, IL-1β and CXCL8 secretions (Dongari-Bagtzoglou and Kashleva, 2006; Yadev et al. 2011). RHM reconstructed with either primary gingival cells or immortalised TERT cell lines demonstrated similar responses to commensal microbial exposures: they remained viable and showed protective responses e.g. activated toll-like receptor (TLR) signalling pathways, increased AMP expression (Elafin, hBD2, hBD3) and cytokine release (IL-6, CXCL8, CXCL1 and CCL20) (Shang et al. 2018, 2019, 2020). When exposed to a 5-species pathogenic biofilm, two RHM reconstructed using different keratinocyte cell lines (TR146 or FNB6) showed similar levels of tissue damage and their damage was less than that of a commercial RHE after the same biofilm exposure (Morse et al. 2018). Another study suggested slightly higher P. gingivalis invasion occurred in 3D models using H375 cell line than using primary keratinocytes, but the difference was not significant (Pinnock et al. 2014). In contrast to bacterial exposure, the host response to viral exposures greatly depended on the source of host cells and the method of exposure. For instance, exposing to HSV-1 resulted in a lower viability in the host model reconstructed using the immortalised epithelial cell line HaCaT than a host model using the other epithelial cell line HMK (Turunen et al. 2014).

The complexity of 3D host model was another important factor resulting in different host responses to the same microbial exposure, such as RHE (only keratinocytes) versus RHM (keratinocytes and fibroblasts). Two RHM, although reconstructed with different sources of host cells (cell line or primary cells), both showed significantly higher upregulation of IL-1β and CXCL8 to C. albicans exposure than an RHE (Yadev et al. 2011). After 24 h of P. gingivalis exposure, secretion of IL-6 and CXCL8 were not modified in an RHE (Dickinson et al. 2011), while both were increased in RHM exposed to P. gingivalis for the same period of 24 h (Andrian et al. 2007; Pinnock et al. 2014).

Taken together, the type of microbial exposure is probably the most important factor deciding the host response in 3D oral HMI models. Also, models of different complexity may have different host outcomes to the same microbial challenge. In contrast, using different sources of host cells (i.e. cell lines versus primary cells) had little impact on HMI outcomes in these models.

4. Applications

Given the enhanced physiological relevance, 3D host models are being more and more valued for their usage in both fundamental and clinical purposes. The current oral (dental) field still heavily relies on animal models and 2D cell models for mechanistic studies as well as pre-clinical tests of oral hygiene products and medicines e.g. to treat salivary secretory disorders, oral wound healing or oral cancers (Kim et al. 2020; Olek et al. 2021; Toma et al. 2021). However, results of these models were not always translatable to the corresponding human situation since they do not fully represent human oral (patho)physiology, and therefore some results were indicated by the authors as having limited correlation to the clinical situation (Moharamzadeh et al. 2012; Janjić et al. 2020). In comparison, human-based 3D oral models are scalable and reproducible platforms for a variety of applications, enabling investigations on important barrier functions of the human oral mucosa such as the tissue histology and permeability (Nikaido et al. 2019). Such applications have focussed on (i) modelling disease such as oral cancer (Almela et al. 2018) and infections (De Ryck et al. 2014; Jackson et al. 2020), (ii) developing alternatives for gingival graft (Dragan et al. 2017) and (iii) testing therapeutic substances for antimicrobial effects, biosafety, biocompatibility and beneficial effects on the oral barrier (Bierbaumer et al. 2018; Nikaido et al. 2019; Lin et al. 2020).

With microbes being increasingly recognised for affecting a variety of host events in the oral cavity (Lamont et al. 2018), researchers have expanded the use of 3D models by incorporating microbes to study oral infection, wound healing, and substances of potential anti-microbial effect. The substances which have been studied in 3D oral HMI models include a nystatin suspension (Ohnemus et al. 2008; Melkoumov et al. 2013), a synthetic decapeptide KSL-W (Semlali et al. 2011), polymersome-encapsulated metronidazole, doxycycline and gentamicin (Wayakanon et al. 2013), atmospheric-pressure cold plasma, penicillin and fluconazole (Delben et al. 2016) and 5-fluorouracil (De Ryck et al. 2014; Sobue et al. 2018; Hong et al. 2019). Also, there are increasing interests on studying peri-implantitis in 3D oral HMI models incorporated with dental implant abutments (Mikolai et al. 2020; Souza, Bertolini, Thompson, Barão, et al. 2020). Just as importantly, the participation of immune cells in the human oral physiology and pathology are being more and more investigated in 3D oral HMI models with increased model complexity and physiological relevance (Koning et al. 2021; Ollington et al. 2021).

5. Future perspectives

Understanding oral HMI is important for maintaining oral health as well as for preventing disease. For this purpose, 3D oral HMI models are being increasingly appreciated as a platform for studying underlying mechanisms and testing new medications, dental materials and oral hygiene products (Métris et al. 2022). Future perspectives are realised by their outstanding potentials to incorporate key players in native oral HMI e.g. immune cells and saliva flow which are missing in most of the current models. For example, only four studies have incorporated immune cells into 3D oral HMI models. These studies illustrated the important role of immune cells in the oral HMI, including polymorphonuclear leukocytes (PMN) in oral candidiasis (Schaller et al. 2004); peripheral blood mononuclear cells (PBMC) and CD14+ monocytes during the exposure of gingivitis-associated biofilms (Brown et al. 2019), and monocytes (Mono-Mac-6) in a periodontal pocket model (Bao Papadimitropoulos, et al. 2015). However, since these models were reconstructed differently by e.g. using different types of immune cells, different types of local gingival cells as well as different microbial exposure it is not possible to make any direct comparisons between the models. In addition, the physiological relevance can be further improved by combining oral models with models representing other organs or tissues, such as various types of organ-on-a-chip and vascular equivalents. Recently, we have combined an RHM with reconstructed human skin (RHS) containing Langerhans cells in the HUMMIC-2-Organ-Chip to investigate systemic immune-toxicity (skin rash) to dental materials (Koning et al. 2021). With further incorporation of microbes in the future, such studies will provide important insights into how a specific oral condition affects the entire body, for example to explore mechanisms underlying the systemic influence of pathological oral conditions via the oral-gut-brain axis (Belstrom 2020; Diesch et al. 2021).

From a practical angle, an ideal oral HMI model is expected to include the following aspects. First, a microbial exposure that contains multiple species representing a meaningful HMI condition in vivo e.g. commonly seen species of commensal oral microbiota or species that relate to a specific oral disease. Second, a robust host equivalent that allows colonisation of microbes while maintaining its viability and responsiveness. Third, the in vitro HMI is maintained for a meaningful period of time in order to mimic the dynamic processes of HMI in vivo. Just as importantly, a clearly stated control condition is essential in order to interpret results in different HMI models: to which HMI element can the observed discrepancies be correlated, e.g. different types of host model or microbial exposure. In the present review we have summarised the current 3D oral HMI models based on the use of organotypic RHM and RHE and living microbes. Based on the various model complexity, we categorised and compared these models for their capability in representing the native oral microbes as well as the host. Yet we must be aware that these models of different experimental designs may result in confounding results when being compared with each other. This is not only caused by the varying model complexity but also due to the limitation of reductionist models in general. First, the synthetic microbial communities (biofilms) in 3D HMI models do not completely mirror their in vivo counterparts in terms of the complex biofilm structure, diverse composition and the dynamic metabolism under constant influence from the host and the surrounding environment. Second, some missing elements of the host and from the environment may lead to discrepancies in host responses, such as the lamina propria consisting of fibroblasts, immune cells, hard tissues, the flow of saliva and GCF, and the long-term maintained HMI period. Last but not least, exposing a sterile organotypic host model to microbes does not exactly mimic the natural situation as the in vivo oral tissues have already been primed and trained by resident microbes since the time of birth, or even before birth. The future 3D oral HMI models should be optimised to include the basal influence from these resident microbes as well. Overall, the ultimate aim of developing 3D oral HMI models is to provide human-representing platforms for fundamental studies as well as for pre-clinical tests. Therefore, improving the model complexity based on known physiology of the oral cavity as well as the oral microbiota is key, and information to improve these models needs to be continuously obtained from relevant in vitro models and importantly from clinical data.

Acknowledgement

We would like to thank Unilever R&D including the Safety and Environmental Assurance Centre for critically reviewing this manuscript and for the financial support.

Additional information

Funding

The author(s) reported there is no funding associated with the work featured in this article.