Application of the adverse outcome pathway (AOP) concept to structure the available in vivo and in vitro mechanistic data for allergic sensitization to food proteins

The multiple facets of intestinal permeability and epithelial handling of dietary antigens were previously reviewed [15]. The following sections intend to capture the current understanding of the role of transport in sensitization induction as schematically outlined in Fig. 2.

Impairment of tight junctions may be implicated in sensitization induction. Paracellular transport is mainly determined by pore size in tight junction and only concerns molecules of MW < 600 Da [16, 17]. Proteins that are transported via this paracellular route are not exposed to lysosomes in the enterocyte and therefore are not degraded [18, 19]. Passage of intact protein may allow sensitization of immune cells in the subepithelial compartment (Fig. 2a).

Several in vitro studies suggest that gluten (gliadin), kiwi (Act d 1) and peanut (Ara h 2) allergens facilitate absorption by modulating tight junctions [20‐23]. Act d 1, in analogy with house dust mite (HDM) Der p 1 [24], may cause a protease-dependent disruption of the tight junctions of the epithelial cell (EC) layer (Fig. 2a) For Ara h 2 the effect on the tight junctions and barrier integrity appears to involve endocytosis and downregulation of the zink finger protein A20, a crucial gatekeeper preserving tissue homeostasis with ubiquitin-regulatory activities [21].

Among various intestinal innate immune cells, mast cells (MCs) play critical roles in maintaining gut immune homeostasis. Studies on sensitized individuals revealed that the integrity of the tight junctions is decreased by mast cell-derived mediators like tryptase and IL-4 [25‐30]. In murine models, the importance of MCs for the development of food allergies has also been clearly shown. Mice presensitized with alum and OVA develop allergic diarrhoea after oral OVA inoculation [31]. Inhibitors of serotonin and histamine as well as depletion of MCs by an anti-c-kit monoclonal antibody reduce the occurrence of allergic diarrhoea. Therefore, MCs as well as serotonin and histamine derived from them are involved in the allergic diarrhoea in this model [31]. These murine and human data clearly show the importance of MCs in the gut homeostasis and the effect of MC activity by allergens resulting in inflammatory responses in mucosal compartments [32]. However, in the context of food allergy, the importance of MCs has only been described in already sensitized animals/individuals. To date there is no data available supporting the role of mast cells in sensitization induction to food proteins as depicted in Fig. 2b.

Transcellular transport mechanisms provide intact protein with free passage to the sub-epithelial compartment. Transcellular transport of proteins and peptides can occur via different mechanisms as illustrated in Fig. 2a, b. Most dietary proteins or peptides are absorbed actively by ECs in non-sensitised persons by endocytosis at the apical membrane and undergo transcytosis toward the lamina propria (LP). An estimate of 70–90% of the transported protein is degraded by intracellular enzymes into amino acids and peptides [33, 34].

In vitro studies showed that intact soybean allergen P34 and bovine β-lactoglobulin cross the epithelial barrier by transcytosis, while tryptic fragments of β-lactoglobulin followed para- and transcellular routes [35, 36]. Thus, protein resisting proteolysis in the gastrointestinal tract is transported intact through the intestinal epithelial barrier, and may enhance the risk for sensitisation induction.

In sensitized persons, T_h2 or innate lymphoid cells (ILC)-2-derived IL-13 with other inflammatory mediators (e.g. IFN-γ, TNF-α and IL1-β) stimulate transport via the CD23 receptor, which is overexpressed by enterocytes in sensitised persons. Antigen-IgE complexes are transcellularly transported by the CD23 receptor [18]. A role for CD23-mediated transport in sensitization induction must still be documented.

In other studies, receptor-mediated transcellular transport via the CD23 receptor appears to protect proteins from lysosomal degradation [18, 19], however as explained above, CD23-mediated transport has only been described in already sensitized individuals.

M cells transport protein directly to the underlying immune cells. M cells are specialized ECs covering the Peyer’s, caecal and colonic patches. These cells deliver intact proteins and particles directly to the gut associated lymphoid tissues (GALT), and therefore may play an important role in the regulation of immunological reactions to dietary antigens [37, 38].

The current perception is that particulate or aggregated antigens are taken up by M cells, resulting in antigen-specific local or systemic immune responses [39, 40]. This perception is supported by a study in mice showing that pasteurization shifted the transport of soluble β-lactoglobulin and α-lactalbumin from transcytosis through enterocytes to transport of protein aggregates by M cells to the Peyer’s patches, and higher IgE and cytokine (IL-5, IL-13, IFN-γ, IL-10) levels [41]. However, another study demonstrating the presence of soluble protein as well as protein bodies of digested peanut in the cytoplasm of M cells within intestinal Peyer’s patches in BALB/c mice [42] suggests that additional factors may contribute to the induction of adverse responses.

Monocytes and DCs sample protein directly from in the gut lumen. Both cells can sample the gut lumen using their pseudopodia. Tight junctions are opened and after sampling new tight junctional complexes between adjacent ECs are formed. Luminal sampling is upregulated in inflammatory conditions [43, 44]. However, information is lacking whether protein transport occurs via these cells.

The role of goblet cells in protein up-take and sensitization induction seems limited. In healthy human ileum and jejunum goblet cells can transport small peptides (10 kDa) and particles across the intestinal wall [45]. A role for these cells in uptake of food proteins has not yet been documented, nor their role in sensitization induction to food proteins.

There is growing evidence ascribing food proteins the capacity to induce sensitization when contacting the skin, and potentially the respiratory tract. How these proteins acquire access to the cutaneous immune cells that drive sensitization of the gastrointestinal tract is not sufficiently understood yet. In contrast, the mechanistic understanding of protein-induced sensitization of the respiratory tract is growing [14].

Evidence for gastrointestinal tract sensitization following cutaneous exposure exists. Mice exposed via different routes (intragastrically, cutaneously, intranasally or sublingually) to α-lactalbumin in the presence of cholera toxin were all sensitized, however sensitization via the skin resulted in the highest IgE levels [46]. Supporting evidence emerged from studies using hazelnut protein [47] and ovalbumin [48] as model proteins. In each case, oral as well as intragastric challenge led to anaphylactic symptoms after skin sensitization.

Even though human skin appears to be less penetrable than murine skin, epidemiological studies in human populations seem to confirm the skin as a relevant route for food sensitization induction and allergy in man.

High cutaneous exposure to peanut increases the risk of peanut allergy in human [49]. The use of peanut-protein containing skin creams was found to correlate with peanut allergy in children. These creams were mostly used to relieve eczema, i.e. damaged skin with increased permeability for e.g. proteins [50]. The importance of the quality of the skin barrier was demonstrated by the impact of filaggrin mutation/expression on human skin integrity and permeability, and its correlation with increased incidence of sensitization [51]. In Japan, at least 1800 individuals were sensitized following application of a facial soap containing hydrolysed wheat proteins (HWP). An epidemiological relationship was documented between wheat allergy and contact exposure to HWP in Japanese women [52]. Cutaneous sensitization to latex correlates with allergic symptoms upon ingestion of cross-reacting foods e.g. avocado, banana and kiwi. However, it cannot be excluded that oral exposure to these foods may lead to allergic reactions to cutaneous latex exposure [53].

Against the evidence in favor of the skin being a route for sensitization to food and food proteins stand data from studies in mice and humans showing that desensitization can be achieved to food and aero-allergens by applying antigen cutaneously [46].

Evidence for gastrointestinal tract sensitization following respiratory exposureis more ambiguous. The contribution of the respiratory tract to sensitization induction of the gastrointestinal tract is less clear, despite a growing understanding of how proteins may acquire access to immune cells and trigger inflammation [14].

Dunkin et al. [46] presented data suggesting that mice are sensitized by exposure of the respiratory tract to α-lactalbumin in the presence of cholera toxin. However, respiratory sensitization was less effective than cutaneous sensitization as judged from the IgE levels induced by subsequent oral challenges.

In humans, the second most significant risk factor associated with life-threatening asthma is a history of an asthma attack precipitated by food. A clear link between the pathophysiology of food and respiratory allergy is not yet established, but two hypotheses are currently pursued: (i) effector cells resulting from gastral sensitization populate the respiratory tract; (ii) chronic inhalation of food particles results in direct sensitization of the respiratory tract [54]. These hypotheses are not mutually exclusive. The oral allergy syndrome is currently believed to result from respiratory sensitization to pollen (e.g. Bet v 1 protein) in conjunction with oral exposure to apple or carrot, containing cross-reactive Bet v 1 homologues. However, it cannot be excluded that oral ingestion of apple or carrot, results in sensitization and subsequent cross-reactivity to inhaled pollen [55]. There is also evidence suggesting that respiratory sensitization to food proteins not necessarily develops into food allergy. Indeed, to it has been reported that bakers with occupational asthma following exposure to wheat flower can safely consume wheat products [56].

It is widely accepted that sensitization involves factors from gut epithelium which are activated during cellular stress. Such factors include radical oxygen species (ROS), Th2 driving mediators (e.g. IL-1, IL-18, IL-25, IL-33, TSLP) and mucus. There is limited availability of mechanistic data on pathways driving the allergic sensitization via the intestinal mucosa, however, much can be learned from pathways involved in skin and lung sensitization.

The role of ROS in allergen-induced skin sensitization was reviewed [72]. Evidence for a role of ROS in respiratory sensitization induction is provided by allergens with proteolytic activity (e.g. HDM allergen Der p 1). There is evidence that cysteine proteases may induce stress and elevated ROS levels e.g. through cleavage and activation of the Protease-Activated Receptor (PAR) 2 [24], and inactivation of lung surfactant proteins (e.g. SP-A, SP-D). Consequently, tight junction proteins (e.g. ZO-1 and occluding) become accessible for protease activity, leading to increased transepithelial access of allergens to immune cells [73, 74]. If and how this information can be translated to gastrointestinal tract is not yet clear, but it is clear that its epithelial lining is used to deal with protease activity involved in digestion.

ROS activation is also implicated in the sensitization to pollen allergens exerting [NAD(P)H] oxidase activity, an enzyme activity also found in mitochondria and driving intracellular ROS production [75]. Endocytosis of this pollen-derived enzyme increases ROS production and results in degradation of endogenous hyaluronic acid (HA), and TLR2 and TLR4 activation. This may be of relevance as a variety of studies implicate MyD88-mediated TLR2 and TLR4 signalling in the induction of T_h2 immune responses leading to allergic respiratory inflammation and promotion of T_h17 responses [14, 76].

ROS production may be induced directly via TLR activation. The HDM allergen Der p 2 reveals sequential homology to the MD-2-related lipid-recognition (ML) domain family, which is a well-characterized member of the TLR4 signalling complex. There is compelling evidence demonstrating that Der p 2 exhibits the same function as ML. Thus, the intrinsic adjuvant activity of MD-2 homologous allergens and their lipid cargo is likely to have wide generality as a mechanism underlying sensitization induction [77].

Activation of TLR requires phosphorylation by c-Src signals and activation of phosphoinositide 3-kinase and phospholipase C γ for activating NF-κB and chemokine expression leading to lymphocyte recruitment to the lung and increase in mucus production. Subsequently, Ca²⁺-dependent proteases cleave the transmembrane proteins occludin and e-cadherin on ECs promoting transmigration of leukocytes [78]. NFкB activation is causally related to increased release by ECs of IL-33, IL-25 and TSLP, endogenous danger factors (e.g. high-mobility group box-1 (HMGB-1), uric acid and ATP, DC activation and migration, and the induction of ovalbumin and Der p 2 sensitization [79]. The release of ATP and uric acid drives the activation of the NLRP3 inflammasome complex resulting in cleavage of pro-IL-1β to mature IL-1β through caspase 1. IL-1β creates a pro-inflammatory micro-environment with the production of IL-6 and chemokines that mobilize neutrophils and enhance T_h17 cell differentiation [80]. Uric acid may play an important role in T_h2 skewing [81].

The limited mechanistic data suggest that similar TLR-dependent and -independent pathways may drive allergic sensitization via the intestinal mucosa. Epithelium-derived factors (IL-33, TSLP, IL-25) identified in studies focusing on HDM-induced sensitization were observed during\ intestinal peanut sensitization in mice, of which it was shown that IL-33 is important for OX40L expression on DC and the development of peanut allergy (see KE 2) [82]. Furthermore, α-amylase inhibitor from cereals activates TLR4 mediated processes [83]. Kong et al. demonstrated that uric acid is a critical signal for the induction of peanut allergy [84].

PRR, TLR and ROS signalling pathways are also active in endothelial cells, macrophages, fibroblasts as well as DCs. These cell types may therefore contribute to the induction of inflammation and sensitization by employing the same mechanisms as described for ECs.

In vitro data revealed that the allergen Pru p 3, but not the non-allergenic LPT 1 variant, crosses the epithelial barrier and induces production of the T_h2 skewing cytokines TSLP, IL-25, and IL-33. In addition, Pru p 3, but not LPT 1, triggered the expression of inflammatory cytokines such as IL1β, IL6, and IL10 in a co-culture of Caco-2 cells and human peripheral blood mononuclear cells. The highest induction was observed for IL1β, which is related to a T_h2 response and antibody production [85, 86].

The importance of EC-derived IL-33 for in vivo DC activation was demonstrated using a murine model for intragastric induced peanut allergy. It was observed that peanut administered in the presence of cholera toxin increased the expression of OX40L, a co-signaling molecule required for proper T cell activation [82]. In IL-33 receptor knock out mice this increase was not observed. TSLP and IL-25 protein levels were increased in duodenal tissue of peanut allergic mice, but TSLP or IL-25 receptor (IL-17-RB) knock out (KO) mice revealed no driving role for IL-25 or TSLP in peanut sensitization. These data show that IL-33 is sufficient for the observed increase in OX40L expression, expansion of ILCs and the development of peanut allergy [82].

DCs, among other cells, can sample food protein directly from the gut lumen using pseudopodia. Alternatively, they can come in direct contact with intact protein after paracellular or transcellular transport through the ECs, or by M cell mediated transport. Driven by microenvironment changes and contact with the protein, DCs upregulate the expression of MHC II, costimulatory molecules such as CD54, CD80, and CD86, and receptors that are essential for migration (e.g. CCR7) [87].

While proteins are recognized to drive the activation of DCs in an inflammatory context, the observed suppression of LPS-induced IL-12 production suggest that HDM and peanut allergens may be capable of suppressing T_h1 driving DC responses [82].

Eosinophils constitute about 20% of the leukocytes in the LP and appear to be required for the induction of CCR7 and CD86 expression on CD103⁺DC. Hence eosinophils may contribute to activation and instructed migration of DCs from the LP to the MLN [88, 89]. Upon eosinophil depletion T_h2 priming and induction of peanut allergy was abolished in an IL-4 independent fashion [89]. While supporting data were presented, these data need to be further substantiated to strengthen the case for eosinophils playing a key role in the induction of food sensitization.

Molecular profiling of human T cell pathways indicate that the signal supplied by the T cell receptor (TCR) (Fig. 3, event 1) only participates in the development of the T_h1 cell phenotype, whereas formation of the T_h2 phenotype depends on the CD28 signalling, even in the absence of specific TCR activation [102] (Fig. 3, events 2). Stimulation of CD28 activates NF-κB signalling and the expression of GATA3, a signal favouring T_h2 differentiation [103] (Fig. 3, event 3).

In the process of T_h1/T_h2 polarization also CTLA-4 (CD152) appears important. CTLA-4 mediated dephosphorylation of ‘linker for activation of T cells’ (LAT) results in TCR signal inhibition, decreased GATA3 expression, and inhibition of CD28 mediated T_h2 differentiation [104, 105] (Fig. 3). During sensitization, both CD28 and CTLA-4 molecules may interact with either CD80 or CD86 on the surface of APCs [106, 107] (Fig. 3, events 2). Following exposure to an allergen CD86 expression increases faster than for CD80, suggesting that CD86 is required for initiating the immune response, whereas CD80 has a more regulatory function [106]. Kuchroo et al. [108] showed that CD28–CD86 interaction relates to T_h2 cell responses, whereas CD28–CD80 interaction favours T_h1 cells responses. Thus, compounds (e.g. allergens) affecting the kinetics of CD86 and CD80 expression on APCs and CD28 and CTLA-4 expression on T cells, may favour a T_h2 response.

The abundance of GATA3 in T_h2 cells (Fig. 3, event 3) triggered the hypothesized that naïve CD4⁺T cells are “naturally” programmed for switching into T_h2 cells and that T_h2 cell differentiation is repressed by T-bet during T_h1 cell differentiation in healthy individuals [109] (Fig. 3, event 4). The importance of GATA3 for the T_h2 cell phenotype is well established [110]. It promotes the IL-4/IL-5/IL-13 gene translation (Fig. 3, event 5) by regulating methylation of histones which grants access for transcription factors (e.g. c-Maf) that activate IL-4 gene expression. GATA3 also binds specific DNA sites controlling activation (e.g. IL-5 and IL-13) or inhibition (e.g. T-bet, IFN-γ and IL-12Rβ2) of cytokine expression [110, 111].

IL-2 from activated CD4⁺T cell plays a crucial role in the T_h2 cell polarization [112] (Fig. 3, event 6). Activation of the IL-2 signalling via STAT5 results in an early and IL-4-independent induction of the IL-4Rα subunit with formation of a functional IL-4 type I receptor, the signalling pathway for IL-4 [113]. IL-2/STAT5 may also participate in the early initiation of the IL-4 gene expression [114]. The mechanisms driving the establishment of IL-4 signalling by activated CD4⁺T cells resembles OX40L/OX40 signalling [115]. OX40 (CD134) expression is induced on the surface of naïve T cells hours after exposure to antigen [116]. Allergen-induced T_h2 polarization requires only endogenous IL-4 and is controlled by the OX40L/OX40 signalling pathway [115]. OX40 signalling sustains TCR, CD28 and IL-2R expression [116].

The contribution of TSLP, IL-25 and IL-33 to T_h2 polarization is recognized in a murine asthma model. Influence of TSLP on the T_h2 response development is indirect and occurs by inhibiting IL-12 secretion and by inducing OX40L co-signalling in DCs. The presence of IL-25R and IL-33R on CD4⁺T cells suggests that IL-25 and IL-33 exert a direct effect on these cells. In vitro IL-25 promotes T_h2 differentiation and IL-4 expression through NFATc1 and IL-4/STAT6-dependent mechanisms [117]. This observation was not confirmed by a murine model for HDM allergy [82]. Epithelial IL-33 affects OX40L expression by DCs and activates signalling pathways that are relevant for the T_h2 polarization (e.g. ERK, MAPKs and NF-κB) via IL-33R. Abolishing IL-33 signalling in allergic mice decreased the production of specific IgE with 50% [82, 118]. However, whether IL-33 is inducing or maintaining a T_h2 phenotype cytokine remains to be clarified.

Activation and differentiation of naïve CD4⁺T cells may be under hormonal and metabolic control. A key event for both immune and metabolic signalling is mTOR protein kinase (PK) activity. A HDM/OVA mouse model for asthma showed that mTORC1 activity relates to T_h1 and T_h17, while mTORC2 relates to T_h2 differentiation [119]. A role of mTOR PK, if any, in sensitization seems plausible. Inhibition of mTOR PK prevents activation of STATs and T-bet expression, known to be involved in sensitization, and drives iT_reg cell differentiation by blocking polarization signals even in the presence of the T_h1 and T_h2 polarization cytokines. A leptin-dependent increase of the mTOR signalling is one of the causes of iT_reg cell developmental disturbances resulting in food allergies [119, 120].

Synthesis of specific IgE requires two different levels of qualitative changes in the DNA of immunoglobulin (Ig) genes. Somatic hypermutation (SHM) in the variable regions of heavy and light chain leads to changes in the affinity to antigen. Direct or sequential Ig class-switch recombination (CSR) constitutes strictly controlled intrachromosomal DNA fragment deletion (class segments) in the constant region of heavy chain (IGH) locus [121]. In atopic dermatitis patients the CSR for IgE antibodies may occur in a direct way (IgM > IgE) or an indirect one (IgM > IgG > IgE) [122‐124].

The surface B cell antigen-recognizing receptor (BCR) identifies the antigen in its native form. BCR stimulation triggers a complex cascade of signalling events leading to B cell activation [125] (Fig. 4). The CD19/CD81/CD21 complex on B-cells is important for B cell activation with CD81 playing a crucial role in B-cell/T-cell communication through the MHCII-TCR (Fig. 4). Activated B cells establish contact with T_h2 cells by interaction of the CD80/86–CD28 and CD40–CD40L molecules (Fig. 4).

Effective induction of the activation-induced cytidine deaminase (AID) expression requires a cooperation between the CD40/TRAFs/NFκB signalling pathway, IL-4 and IL-13 as well as CD40 and IL-4 activation of HoxC4 in the B cell nucleus (Fig. 4). CD40 signalling in combination with IL-4 and IL-13 activates the synthesis of GLTɛ. Activation of the Iɛ promoter by transcription factors Pax5, PU.1, C/EBP, AP1, and E-box for E2A enhances GLTɛ transcription. GLTɛ synthesis is regulated by IL-4 only, since IL-4 depletion is sufficient to block Ig CSR to IgE [126‐128]. Ig CSR to IgE may occur also in the absence of CD40 signalling [122‐124]. BAFF and APRIL molecules on DCs activate surface BAFFR and TACI molecules on B cells, and stimulate NFκB through a pathway involving NIK (NFκB-inducing kinase) and p52 activation (Fig. 4). IL-4-secreting mast cells, eosinophils, basophils, and γδ T cells, are alternative inducers of CD40 independent IgE synthesis [104, 128].

That IgE synthesis is tightly regulated is suggested by the minute amounts of serum IgE as compared to IgG, even in allergy or parasitic diseases. One regulatory mechanism involves rapid capturing and subsequent endocytosis by cells expressing a high affinity receptor FcɛRI (e.g. on mast cells, basophils or eosinophils), and to a lesser extent by the cells expressing a low affinity receptor FcɛRII/CD23 (e.g. intraepithelial cells (IECs) or B cells). Due to technical challenges the mechanistic understanding of IgE synthesis regulation and Ig class switch at cellular level is limited, especially in humans [122, 129]. IgE switching may result in suboptimal polyadenylation of 3′-untranslated region of membrane IgE resulting in less stable mRNA for membrane IgE (mIgE) and lower expression of mIgE BCRs [129]. A plasma membrane molecule specifically involved in the negative feedback regulation of IgE-CSR is CD23 (FcεRII), which plays a role in in vivo sensitization induction by cysteine proteases (e.g. HDM Der p 1) [130]. In the presence of high levels of IgE or IgE-antigen complexes membrane-bound CD23 (mCD23) suppresses the production of IgE. Endogenous or exogenous (e.g. by Der p 1) proteolytic cleavage of mCD23 renders CD23 soluble (sCD23) which reinforces the production of IgE by inhibiting binding of IgE to mCD23. In combination with CD21 mediated signalling, sCD23 enhances B cell proliferation and transformation into IgE + plasma cells. Three cytokines may also be involved in down-regulation of IgE synthesis. IL-21 represses IL-4 stimulated GLTɛ synthesis in a STAT6-independent manner. IFN-γ increases IgG1-CSR over IgE-CSR, whereas TGF-β induces transcriptional factor ID2 (inhibitor of DNA binding-2) that suppresses the E2A binding with Iɛ promoter [131, 132].