The Immunology of Food Allergy

Within the clinical realm of allergy, food allergy is receiving an increasing amount of attention, mirroring its increasing prevalence both nationally and internationally. Current estimates put food allergy as affecting up to 15 million people within the United States (1). Therapeutically, these patients are dependent on a difficult avoidance approach, with injectable epinephrine as a life-saving option in case of accidental exposure. This has been shown to significantly affect quality of life (2), and recent advances in understanding the mechanisms behind food allergy have been fueled by the desire to develop improved therapies.

In considering such mechanisms, we focus on three processes that may be important: oral tolerance, sensitization to food allergens, and anaphylactic reactivity to these food allergens. Finally, an emerging concept of “nonresponsive tolerance,” where anaphylactic reactivity does not occur or is lost despite evidence for IgE-associated sensitization is highlighted.

Oral tolerance to egg proteins was first described >100 y ago (3). This natural phenomenon, where ingested food proteins do not elicit a specific immune response, is also observed in humans (4), but the necessary mechanisms remain unclear. Despite gastrointestinal enzymes degrading food and the physical barrier of the intestinal mucosa, immune surveillance of food Ags and establishment of tolerance mechanisms are clearly occurring. Several reviews addressed possible routes of Ag sampling and presentation (5–7), including sampling by dendritic cells (DCs) across the epithelial layer, presentation by M cells or goblet cells to DCs, or soluble Ag directly traversing the epithelium through paracellular or transcellular routes.

Key cells seem important for oral tolerance and the maintenance of regulatory (Foxp3⁺) T cell (Treg) populations (Fig. 1). CD11c⁻CD11b⁺F4/80⁺ macrophages exhibit an anti-inflammatory gene signature and produce IL-10 (8). Additionally, two distinct subsets of tolerance-associated CD11c⁺ cells reside in the intestinal lamina propria, expressing either CX₃CR1 or CD103 (9). CX₃CR1 knockout mice show diminished IL-10 production and Treg populations, as well as a lack of oral tolerance in a food allergy model (10). In contrast, CX₃CR1⁺CD103⁻ cells have been implicated in intestinal inflammation (11).

Most evidence supports a role for CX₃CR1⁻CD103⁺ DCs in tolerance. These cells exhibit lymph-node homing where they activate naive T cells (9, 12) and promote a FoxP3⁺ Treg phenotype, a process requiring both TGF-β and retinoic acid (13–15). Retinoic acid imprints the gut-homing receptors CCR9 and α₄β₇ onto both Tregs (16) and IgA-secreting B cells (17), an event that also seems to contribute to oral tolerance (16). CD103⁺ DCs also use IDO for tolerance, and loss of IDO function drives T cells toward a Th1 or Th17 phenotype, limiting Tregs and oral tolerance (18). Recent findings also show that MUC2, a mucin secreted by intestinal goblet cells, supports the anti-inflammatory potential of these CD103⁺ DCs (19).

IPEX (immune dysregulation, polyendocrinopathy, enteropathy, X-linked) patients with mutations in the FOXP3 locus (20) develop severe food allergy, as well as a plethora of other disorders, including autoimmunity, enteropathy, and atopic dermatitis (21), indicating the importance of Tregs in tolerance. Foxp3-mutant mice (scurfy) and DEREG mice, in which Tregs can be deleted upon diphtheria toxin treatment, also have been used to demonstrate the importance of Tregs in allergic responses (22, 23). Expression of CCR9 and α₄β₇ on Tregs is necessary for tolerance, because these molecules support gut homing (10, 16). Although previous reviews summarized the effects of Ag concentration in oral tolerance (5, 6) [i.e., low doses drive Tregs, whereas high doses yield anergy and deletion of T cells (24)], most evidence points toward Treg-associated low-dose tolerance as being critical in food allergy. We (25) showed previously that loss of oral tolerance to peanut was associated with diminished Treg responses but also that high-dose Ag feeding could overcome allergic responses.

In food allergy, the immune response is clearly biased toward a type 2 cytokine–associated phenotype. Why specific food Ags trigger this response remains unclear, although some food Ags possess the potential to stimulate innate immune responses. For example, the peanut allergen Ara h1 binds to CD209 on DCs (26), and milk sphingomyelin activates type 2 cytokine responses from invariant NKT cells (27).

Changes in microbial flora have been associated with allergic sensitization, with several lines of evidence supporting protection by specific bacteria and their products, likely through sustaining intestinal Treg populations (reviewed in Ref. 28). Mice with decreased commensal bacteria colonies, which includes either germ-free or antibiotic-treated mice, exhibit increased food-allergen sensitivity (29), high serum IgE, and increased circulating basophils (30). Interestingly, mice with enhanced signaling through IL-4Rα who display profound allergic sensitization and anaphylaxis to food Ags (31) also exhibit an altered microbiota that can be normalized by Treg transfer (32). Consequently, the interaction of host immunity and commensal microbiota seems bidirectional, with host immune responses not just responding to bacterial presence but also shaping the bacterial flora toward that associated with pathology.

The critical mechanisms responsible for allergic sensitization are beginning to be elucidated. Generally in allergy, epithelial production of TSLP, IL-25, and IL-33 has become a key area of interest (33). However, a recent study (34) of these cytokines using a cholera toxin–driven oral peanut model showed that only IL-33 was required for sensitization. IL-33 can increase mucosal permeability (35) and promote Th2 skewing by DCs (36). Interestingly, although constitutive IL-33 expression occurs in epithelial cells, increasing evidence supports the potential for inducible expression by several immune cells, including DCs, that is sufficient for subsequent Th2 immunity, as was shown in helminth infection and for IgG immune complexes (37, 38). However, the key producer of IL-33 in food allergy remains to be determined.

At the level of Ag presentation, several mechanisms that may participate in tipping the balance from tolerance to sensitization have been described. Binding of OX40 ligand to OX40, TIM4 to TIM1, and jagged to notch on DCs and naive T cells, respectively, can regulate T cell differentiation from Tregs toward Th2, as previously reviewed (39). Environmental interactions may drive this differentiation; for example, staphylococcal enterotoxin B (SEB) can break tolerance and promote food allergy (25, 40), and it was shown mechanistically to induce TIM4 expression on DCs that is necessary for Th2 skewing (40). Th2-associated responses also can occur if Tregs are deleted (10) or become dysfunctional, as induced by SEB (41). In contrast, some innate signals also may protect against sensitization, because TLR9^−/− mice have impaired IgE and IgA responses, resulting in reduced anaphylaxis to peanut (42).

Intestinal penetrance by allergens also may enhance allergic sensitization (43). On intestinal epithelial cells, IL-4 can induce upregulation of the low-affinity IgE receptor, CD23, which binds Ag-specific IgE and facilitates Ag uptake (44). This potential mechanism may explain why large or low-solubility Ags traverse the epithelium and elicit systemic responses (45). Similarly, alterations in tight-junction integrity may allow Ag penetrance. For example, deficiency in the desmosomal ICAM desmoglein-1 was shown to elicit profound allergic responses (46), whereas its expression is reduced in tissues of patients with eosinophilic esophagitis (47).

Recent interest also has focused on the skin as a potential route for sensitization, because food allergy often associates with eczema in patients (1). Barrier integrity also may be important here, because filaggrin-deficient mice, which exhibit weak epithelial barrier function, become sensitized to proteins on the skin (48), and epicutaneous sensitization is sufficient to promote anaphylaxis upon oral challenge (49). Although very few studies have defined specific genes associated with food allergy, it is interesting to note that mutations in desmoglein-1 (47), filaggrin (50), and TSLP (51) were shown to be associated with food allergy or eosinophilic esophagitis in human cohorts, because these molecules all regulate skin homeostasis. However, it is unclear whether these associations relate to food allergy or eczema, because these diseases are often coincident in children, and the number of genes associated with food allergy alone remains relatively limited (52).

The mechanisms of anaphylaxis––the hallmark of food-allergy reactivity––are generally biphasic: an acute reaction occurs immediately after allergen exposure, followed by a late-phase reaction several hours later. Symptoms occurring during the acute reaction are due to release of preformed mediators, whereas the late-phase response involves influx of inflammatory cells. Clinically, heterogeneity in responses is observed, with some patients experiencing either the acute or late-phase reaction and others experiencing both the acute and late-phase reactions (53). In addition to clinical heterogeneity, anaphylactic responses can be elicited through multiple mechanisms.

First shown in 1997 by Miyajima et al. (54), both IgE and IgG can play a role in anaphylaxis in the mouse. IgE functions via its high-affinity receptor, FcεRI, which is highly expressed on mast cells and basophils (55). FcεRI^−/− mice do not respond in a passive IgE-mediated systemic anaphylaxis model (56) and have reduced responses in models of allergic diarrhea and food allergy (57–59). IgG has several receptors: the high-affinity FcγRI and FcγRIV and the low-affinity FcγRIIB and FcγRIII. All of these receptors are expressed on several cell types involved in anaphylaxis, including mast cells, basophils, neutrophils, and macrophages. Using a model of systemic anaphylaxis, Strait et al. (56) showed that inhibition of FcγRII/III abolished temperature drops associated with shock in IgG-mediated, but not IgE-mediated, anaphylaxis. Similarly, Jönsson et al. (60) used knockout mice to show that FcγRIV is necessary for systemic anaphylaxis. Although these pathways have been differentially defined using these passive models, both Abs appear to participate in active food allergy: Arias et al. (61) showed that IgE^−/− and IgG₁^−/− mice were only partially protected from peanut-induced anaphylaxis, but blockade of IgG₁ in IgE^−/− mice completely abolished the response; similarly, FcRγ^−/− mice, which lack the common chain for both the IgE and IgG receptors, were protected (62). Importantly, recent studies using humanized mice supported the potential anaphylactic functions of IgG via human receptors (63).

Histamine, platelet-activating factor (PAF), and 5-hydroxytryptamine (5-HT; serotonin) are all sufficient to induce early-phase anaphylaxis (64, 65). Several groups also looked at the necessity for each of these mediators in anaphylaxis, and there appears to be heterogeneity here also.

Histamine, produced from both mast cells and basophils, is a well-established mediator necessary for anaphylaxis (56, 66). In IgE-mediated systemic anaphylaxis, histamine synthesis, as well as histamine H1 and H2 receptors, is necessary for responses (66, 67), and blockade of these receptors is therapeutically beneficial in patients with acute allergic reactions (68).

Additionally, PAF and 5-HT were shown to contribute to anaphylaxis (56, 58, 61–63). Several inflammatory cells make PAF, including macrophages/monocytes, mast cells, basophils, neutrophils, and platelets. Although associated with platelet activation, PAF also influences vascular permeability, leukocyte recruitment, and leukocyte activation (69). Studies using models of allergic diarrhea, food allergy, or systemic anaphylaxis models showed that responses may be due to either PAF and histamine (56, 61) or PAF and 5-HT (58, 63).

Although other mast cell– and basophil-derived mediators have been implicated in food allergy, their role is less defined. These include other preformed mediators (e.g., tryptase, chymase, and heparin), lipid mediators [e.g., PGD₂, LTC₄, LTD₄, and LTE₄ (70)], and several cytokines. IgE activation of mast cells has the potential to generate several cytokines that were shown to direct late-phase inflammation, including release of preformed TNF and synthesis of IL-33 (71, 72). TNF was shown to be necessary for late-phase recruitment of neutrophils (71), as well as for a late-phase increase in PAF in the serum (73). The IL-33 receptor, ST2, is necessary for IgE-triggered tissue inflammation (72). IL-33 does not directly cause mast cell degranulation (74), but it promotes expression of several cytokines and chemokines, including IL-6 and IL-13, from mast cells and eosinophils (72, 75). Similarly, IL-9 can both stimulate and be produced by mast cells (76). IL-9 was shown to be critical for the initiation and severity of food-associated anaphylaxis by promoting intestinal mastocytosis (77, 78).

Largely from murine studies of passive sensitization models, mast cells, basophils, macrophages, and neutrophils were shown to contribute to anaphylactic shock responses. Four distinct pathways of response seem to be possible: a “classic” pathway involving IgE, FcεRI, mast cells, and histamine; an “alternative” pathway mediated by IgG₁, FcγRIII, macrophages, and PAF (79); an IgG–basophil–PAF pathway (80); and an IgG–neutrophil–PAF pathway via FcγRIV activation (Fig. 2) (60).

In active sensitization models, IgE, FcεRI, and mast cells are responsible for inducing allergic diarrhea (58, 59). Although both allergen-specific IgE and IgG Abs are increased by sensitization, only FcεRI (58, 59), and not FcγRII/III (58), is required. Interestingly, the diarrhea response seems to be mediated by a combination of PAF and 5-HT. In contrast, the mast cell responses that are key in anaphylactic food allergy models (with contributions from macrophages and basophils) occur via IgE- and IgG-dependent mechanisms requiring both histamine and PAF (57, 61, 81). Recently, the necessity for basophils in peanut anaphylaxis also was defined (82). Interestingly, the pathways to systemic anaphylaxis models may relate to the Ag dose required to trigger each mechanism, because small doses activate the classical pathway, and large doses activate the alternative pathway (56).

Clinical studies showed that the incidence of food allergen–specific IgE is 10-fold greater than the incidence of food allergy (83), suggesting an additional level of tolerance regulation above that of simply preventing immunological priming toward Th2 and IgE. Furthermore, in patients with Stat3 mutations leading to hyper-IgE syndrome, anaphylactic reactivity to food allergens is actually diminished (84). Recent work showed that Tregs can suppress IgE-primed mast cell degranulation to Ag exposure via OX40/OX40 ligand interactions (85). In food allergy, we demonstrated that Treg transfer could suppress anaphylaxis and restore intestinal Th17 homeostasis by enhancing mast cell–derived IL-6 (41). Interestingly, this cytokine-mediated process was OX40 independent and instead was mediated via TGF-β (41). Additionally, Tregs can downregulate FcεRI on mast cells in vitro (86). This emerging form of active tolerance, occurring despite the presence of an Ag-specific IgE–primed immune system, seems distinct from Ag desensitization, which is associated with internalization of FcεRI and IgE and altered Syk activation (87, 88).

Immunologically, food allergy is a disease with much left to determine. The mechanisms of tolerance, both in terms of what prevents most people from developing responses, as well as why some individuals outgrow or never develop food allergies despite sensitization, remain unclear. Similarly, the environmental and genetic influences over sensitization are just becoming understood. Importantly, studies from animal models are showing that the mechanisms of anaphylactic reactions may well be heterogeneous in terms of routes of exposure, cell types involved, and the mediators responsible for symptoms. A better understanding of this heterogeneity will be crucial in developing future therapies.

The authors have no financial conflicts of interest.