In vitro models have important limitations but they enable the preliminary screening of the effects that bacterial cells or fractions might have on different components of the immune response (Kobayashi et al., 2017). Most in vitro models based on immune cells employ peripheral blood mononuclear cells (PBMCs). In this way, whole cells of B. longum, B. breve, B. bifidum, and B. animalis subsp. lactis strains demonstrated capacity to induce dendritic cell (DC) maturation, and a species/strain-dependent T cell polarization response (Medina et al., 2007; López et al., 2010; Nicola et al., 2016). These studies revealed that, while B. animalis and many B. longum strains induced the production of the modulatory cytokine IL10 to varying degrees, the greatest strain-dependent differences were displayed in TNFα and INFγ production (Figure 1). Stimulation of PBMCs with subcellular fractions of bifidobacteria, including cytoplasmic, surface extracts, and supernatants, has also allowed the identification of molecular determinants of the elicited effects. For instance, a trypsin-labile cytoplasmic fraction of a B. bifidum strain was identified as the effector of CD8⁺ T cell activation; and supernatants of B. breve BB99 and B. longum 1941 exerted a regulatory T cell induction (Mouni et al., 2009). PBMC models are thus useful to identify desirable immune profiles in probiotic strain screenings (Liu et al., 2016).

Other in vitro models differentiate DCs, a specialized type of antigen presenting cells, from monocytes. DCs are regarded as the main guardians of the intestinal mucosa and are important in initiating the microbiota–immune system cross-talk. Their pattern recognition receptors (PRRs) interact with specific microbial-associated molecular patterns (MAMPs), which orchestrated molecular cascades that will determine the nature of the immune response (Hoarau et al., 2008; Wittmann et al., 2013). In vitro differentiated DCs allowed the identification of specific domains of a B. bifidum surface protein and the exopolysaccharide (EPS) of B. longum 35624, as the effectors of the immune responses elicited by the strains (Guglielmetti et al., 2014; Schiavi et al., 2016). DC models have also been used to predict the anti-inflammatory potential of bifidobacterial strains/molecules in specific population groups; for instance, bifidobacteria improved antigen uptake and processing by DC from Crohn’s disease patients (Strisciuglio et al., 2015).

Other in vitro models using immune cells employ murine splenocytes (Tanabe et al., 2008; Srutkova et al., 2015), macrophage-like cell lines (He et al., 2002; Lee et al., 2012; Mokrozub et al., 2015), or cells isolated from the gut-associated lymphoid tissues (GALT) (Hidalgo-Cantabrana et al., 2014), although they have not been widely used to examine the immunomodulation potential of bifidobacteria and thus their utility to predict immune responses is yet to be confirmed.

The immunomodulation potential of bifidobacteria has also been studied on enterocytes including Caco-2 or HT29 cell models (Bahrami et al., 2011; Chichlowski et al., 2012; Khokhlova et al., 2012; Arboleya et al., 2015; Sánchez et al., 2015; Luongo et al., 2017). Although the immune response of epithelial cells is much more limited than the one exerted by specialized immune cells, enterocytes are more directly exposed to the intestinal milieu and are considered to play a key role in initiating the bifidobacteria–host interactions.

Beyond that, co-culture systems employing both immune and intestinal cells have also been implemented to study microbial–host interactions and promise to overcome some of the limitations of single cell type models (Duell et al., 2011). However, few studies have used them on bifidobacteria (Pozo-Rubio et al., 2011). The application of organized multicellular systems like intestinal organoids for these kinds of studies is envisaged (Noel et al., 2017).

Germ-free (GF) and conventional in vivo models, including healthy and disease-induced models, have shed light on the immune modulation capability of bifidobacterial strains including live and heat-killed cells (Sugahara et al., 2017). Screening of a large collection of gut symbionts on GF mice identified a B. adolescentis strain which induces a robust Th17 response, albeit not inducing intestinal inflammation (Tan et al., 2016). However, immune responses may vary strongly depending on the health status of the host, as the human sera of Clostridium difficile patients were shown to be more reactive against B. longum extracts than that of healthy individuals (Górska et al., 2016).

In vivo models of intestinal diseases have demonstrated the potential of B. bifidum and B. animalis strains to restore immune markers and intestinal barrier in low chronic inflammation models (Philippe et al., 2011). Similarly, B. longum CECT 7347 attenuated the production of inflammatory cytokines and the CD4⁺ T cell-mediated immune response in a gliadin-induced enteropathy model (Laparra et al., 2012). Other disease models, described below, have been tested in literature. In food-allergy models, a vesicle-derived protein from B. longum (Kim et al., 2016) and a B. animalis (Ezendam et al., 2008) strain, administered during lactation, exerted immunomodulatory effects. In a gut model, B. longum strain 51-A reduced inflammation (Vieira et al., 2015). Finally, in obesity models, B. pseudocatenulatum restored the lymphocyte–macrophage balance and B. adolescentis IM38 improved high-fat-diet induced colitis inhibiting NF-κB activation (Moya-Pérez et al., 2015; Lim and Kim, 2017). Furthermore, the role of bifidobacteria in responsiveness to immunotherapy has recently been suggested. Accordingly, using tumor models in mice, Bifidobacterium administration was shown to improve tumor-specific immunity and response to therapy through augmented DC function, opening new avenues to exploit the bifidobacterial-immune dialog in the context of this disease (Sivan et al., 2015).

Finally, non-murine in vivo models, like pig models, are very attractive for the study of microbe–host interactions due to the similarities in the gastrointestinal function and development between pigs and humans. In this context, bifidobacterial administration in neonatal piglets has been shown to increase the production of intestinal IL-10 (Herfel et al., 2013), and to improve B and T cell responses following rotavirus vaccination (Vlasova et al., 2013; Kandasamy et al., 2014; Ishizuka et al., 2016). In addition, colonization with a combination of lactobacilli and bifidobacteria in non-vaccinated gnotobiotic piglets reduced the severity of rotavirus infection, while in vaccinated animals enhanced Th1 (Chattha et al., 2013). Thus, in vivo models closer to humans are valuable to study the immunomodulatory potential of certain strains, should be “particularly in the context of pig models in order to study pre-term birth and necrotizing enterocolitis (NEC)” (Oosterloo et al., 2014).

Different immunoreactive proteins from two B. longum strains have been identified in mono-colonized mice, rabbit, and human sera, revealing that the effects are strain and host specific (Górska et al., 2016) and emphasizing the need to further support in vitro immunomodulatory effects in clinical trials. A summary of human studies that focus on the immunomodulatory effects of bifidobacterial consumption in multiple disorders, in some of which gut microbial ecology dysbiosis and altered immune profiles coexist, is presented in Table 1.

Bifidobacterial proteins are one of the targets of human immunoglobulins, notably IgA, which is secreted into the gut lumen in order to control the commensal microbiota populations. Up to six different extracellular proteins from the strains B. longum subsp. longum NCIMB 8809, B. bifidum LMG 11041^T, and B. animalis subsp. lactis IPLA 4549 were recognized by pooled sera from healthy individuals, or Inflammatory Bowel Disease (IBD) patients (Hevia et al., 2014). Perhaps the best known example of an immunomodulatory protein is the extracellular serpin secreted by B. longum subsp. longum. Serpin stands for serine protease inhibitor and includes different families that share the ability to bind and irreversibly inactivate proteases. The gene coding for serpin is not widely distributed among the genus Bifidobacterium, being present in up to nine species so far (Turroni et al., 2010a). More precisely, the targets of serpin secreted by B. longum are two important pro-inflammatory proteases: human neutrophil and pancreatic elastases (Ivanov et al., 2006), proteases that have been shown to induce the serpin gene through a two-component regulatory system (Alvarez-Martin et al., 2012). Limiting the local action of these proteases suggests a role of bifidobacteria in the maintenance of gut homeostasis.

Other well-known protein structures with an immunomodulatory action are pili, which self-assemble on the bifidobacteria surface in the form of filaments and have a primary function of adherence to the intestinal surface (Turroni et al., 2014). Lower levels of IL10 and higher levels of TNFα were detected in the murine cecum mucosa as a response to the presence of a Lactococcus lactis strain, genetically modified for producing B. bifidum pili. This response was not observed in the wild-type strain, suggesting a specific interaction of these structures with the gastrointestinal mucosa (Turroni et al., 2013). Another protein with an immunomodulatory effect is the peptidoglycan hydrolase TgaA, a surface-associated protein in B. bifidum, which was shown to induce IL2 production in monocyte-derived dendritic cell (MoDC), the key cytokine in T_reg cell expansion (Zelante et al., 2012; Guglielmetti et al., 2014). Finally, our own work has revealed the presence of immunomodulatory peptides encrypted in the sequences of bifidobacteria proteins. In this sense, a peptide contained within the sequence of the protein translocase subunit SecA of B. longum DJ010A triggered a marked Th17 response when incubated with human PBMCs (Hidalgo-Cantabrana et al., 2017b).

EPSs are carbohydrate polymers that are synthesized and exhibited in the bifidobacterial surface (Hidalgo-Cantabrana et al., 2014). Although the exact molecular mechanisms have not been described so far, EPSs have a great impact on the host immune function (Hidalgo-cantabrana et al., 2012). In a murine model, the EPS-producing strain B. breve UCC2003 was associated with increases in the mucosal levels of the pro-inflammatory IL12, INFγ, and TNFα which turned out to protect against Citrobacter infection (Fanning et al., 2012). Murine J77A.1 macrophages challenged with the EPS produced by strain B. longum BCRC 14634 increased the production of the anti-inflammatory cytokine IL10 when compared to basal conditions, and when challenged with lipopolysaccharide, the presence of the EPS was linked to lower levels of the pro-inflammatory cytokine TNFα (Wu et al., 2010). It is noteworthy that the rhamnose-rich, high-molecular weight EPS isolated from the strain B. animalis subsp. lactis IPLA-R1 was able to increase IL10 production in a PBMC model and to decrease the TNFα production in human colonic biopsies (Hidalgo-Cantabrana et al., 2015). Moreover, the administration of strain IPLA-R1 to Wistar rats was associated with higher serum levels of TGFβ and lower serum levels of the pro-inflammatory interleukin IL6 (Salazar et al., 2014).

EPS produced by specific bifidobacteria strains have been shown as molecules able to prevent exacerbated pro-inflammatory responses. B. longum subsp. longum 35624 is a strain which has shown clinical efficacy in Irritable Bowel Syndrome, a human condition cursing with chronic mucosal inflammation (Altmann et al., 2016). The anti-inflammatory effects elicited by this strain were shown to rely in its surface-associated EPS, which prevented expansion of the pro-inflammatory Th17 response compared to an exopolysaccharide-negative mutant derivative (Schiavi et al., 2016).

Finally, recent data on a mouse model of pathological cell shedding, EPS from B. breve UCC2003 appeared to confer protective effect through MyD88-dependent signaling (Hughes et al., 2017). Diversity of gene clusters responsible for EPS biosynthesis is high among bifidobacterial species/strains (not to mention variations in the level of EPS production) and this diversity may hold tremendous potential for strain-specific immune responses.

Bifidobacteria possess genomes with high G+C proportions, and un-methylated CpG motifs derived from them can interact with the TLR 9 present on immune cells. Several publications have reported on the immunomodulatory activity of bifidobacterial DNA. CpG motifs have in one case been linked to a promotion of the T_h1 response, dedicated to fight intracellular pathogens such as viruses (Ménard et al., 2010). Another work described an oligodeoxynucleotide derived from the B. longum BB536 strain able to inhibit anti-ovalbumin–IgE titres in a murine model of type-I allergic response after ovalbumin injection (Takahashi et al., 2006).