The growing burden of food allergy suggests that environmental factors are driving its rising prevalence. The community of symbiotic micro-organisms residing in our gastro-intestinal tract (the gut ‘microbiota’) reflect aspects of our environmental exposure and may have the capacity to mediate some drivers towards the development of food allergy. Observational studies from the turn of the century highlighted that infants who carry a significant imbalance of micro-organisms amongst their gut microbiota (termed ‘dysbiosis’) may be at risk of developing atopic dermatitis and allergic sensitization [1]. Animal models have elucidated several mechanisms by which gut microbiota may influence the development of allergic disease. An increasing number of studies in humans report altered gut microbiota characteristics amongst children in relation to their food allergy status, however, none have utilized population-based designs or adjusted for clinical confounders such as atopic dermatitis or microbial confounders such as antibiotic exposure. Here, we review whether current studies are able to support the hypothesis that gut microbiota dysbiosis primes their host towards developing food allergy and look at how future work may more robustly answer this question.

Children and adults living with immunoglobulin E (IgE) mediated food allergy risk life threatening reactions if they eat their respective food allergens. IgE-mediated food allergy, therefore, necessitates on-going rigorous avoidance of food allergens, the carriage of emergency medication at all times and living with the prospect of suffering a serious allergic reaction. These factors impair the quality of life of children with food allergy more than those living with either rheumatoid arthritis or diabetes [7 - 9].

The prevalence of IgE-mediated food allergy is rising amongst economically advantaged countries, with the greatest prevalence being demonstrated amongst young children. Population-based studies show that 7-10% of children in their first three years demonstrate signs of food allergy when challenged, and that this falls towards 4.2% amongst school-aged children as a proportion of milk and egg allergy resolves [10 - 12]. Clinical activity relating to food allergy has been rising amongst Western countries, with a six to seven fold increase in admission frequencies due to food allergy and anaphylaxis being reported over 20 years in the UK [13, 14]. In Australia, fatal food anaphylaxis has increased 9.7% per year (p = 0.04) with food-induced anaphylaxis admissions increasing 10% per year between 1997 to 2013 [15]. Therefore, the prevalence of food allergy has grown to an appreciable public health problem amongst western countries [16].

The rising burden of food allergy is also likely driven by the increasing prevalence of atopic dermatitis [18]. A systematic review has established that eczema increases the risk of children developing food allergy [17]. Dermatitis arises before food allergy, and more severe disease is associated with greater risk of food sensitization showing a clear dose response effect [10, 18]. It is therefore crucial that studies investigating relationships between the microbiota and food allergy examine whether this relationship is confounded by atopic dermatitis.

The hygiene hypothesis was born out of two studies highlighting inverse relationships between allergic disease and micro-organism exposure. The first was published in 1976 and found lower prevalence of asthma, eczema and urticaria amongst the Métis families of northern Saskatchewan in Canada who showed greater carriage of helminthes [19]. The more widely acknowledged study from the British National Child Development Study reported that the risk of hay fever and atopic dermatitis were inversely associated with the number of siblings [20]. At the time, David Strachan postulated that children living in larger families were protected from developing atopic dermatitis and allergies by greater exposure to pathogens.

This hypothesis led to a large body of work exploring how the prevalence of atopic dermatitis, asthma and rhinitis varied with respect to ‘hygiene-related’ epidemiological factors, such as mode of delivery (caesarean section versus normal vaginal delivery), living with older siblings, attending communal childcare facilities, antibiotic exposure and farming. For instance, living on a farm and attending communal child care facilities were shown to demonstrate associations with a reduced risk of asthma, rhinitis and atopic dermatitis [21 - 26]. Some studies have used molecular microbiological techniques to identify pathogens which may have a role in raising the risk of children developing allergic disease. For example, quantitative real-time Polymerase Chain Reaction was used to analyse faceal samples collected at one month of age from the 952 participants of the Child, Parent and Health: Lifestyle and Genetic Constitution (KOALA) Birth Cohort Study in the Netherlands. The authors demonstrated that colonisation with and increasing numbers of Escherichia coli were associated with a greater risk of parent reported atopic dermatitis at 2 years (P_trend 0.02), whilst colonization with Clostridium difficile was significantly associated with atopic dermatitis, recurrent wheeze and atopic sensitization [27].

Around the turn of this century, some culture-based studies introduced the idea that an equilibrium between key constituents was of more importance than the presence of pathogens alone. Atopic Estonian and Swedish children demonstrated less frequent colonization with enterococci during their first month (72% vs. 96%, p < 0.05) and with bifidobacteria during the first year of life (17 to 39% vs 42 to 69%, p < 0.05) [6]. Such observations prompted Graham Rook to postulate the ‘Old Friends Hypothesis’, suggesting that the less-harmful agents associated with rural, hunter-gatherer lifestyle provide immunological protection from allergic disease [28. 29].

‘Next-generation’ sequencing technologies now allow a hypothesis-free approach to surveying the relative abundance and potential function of constituent identities amongst microbiome samples. Next-generation sequencing has become increasingly affordable over the last ten years, and allows parallel sequencing of millions of DNA reads extracted from microbiota samples. ‘16S’ techniques rely on amplification of bacterial DNA strands identified by universal bacterial primers annealing to well conserved 16S rRNA gene regions within the bacterial genome. With this technology, complementary strands of DNA are built from these 16S rRNA gene primers by successively adding individual fluorescently labelled nucleotides, alongside which a camera records which order of coloured nucleotides were incorporated to determine the sequence [30]. The major alternative non-16S rRNA gene sequencing approach aims to sequence all the genetic material within a sample (whether of human, bacterial, archaeal, eukaryotic or viral origin). These metagenomic methodologies such as shot-gun sequencing allow comparison of relative proportions of gene groups, enabling relative abundance comparisons of DNA reads indicating micro-organism identities and also functional capabilities of microbiome samples [31].

Next-generation sequencing technology allows the identification of around 80% more bacterial strains in faecal microbiota than conventional culture-based methods, highlighting its complexity and diversity [32]. Strong associations have been demonstrated between gut microbiota characteristics and insulin sensitivity and obesity [33 - 35]. Partly through this novel technology, the hygiene hypothesis has metamorphosed into the ‘biodiversity hypothesis’, which proposes that the diversity of the gut microbiota is of greater importance than identifying individual bacterial strains and that this aids the development of a regulatory T cell, cytokine and complement network, which protects against the development of allergic disease [36].

These hypotheses postulate that different routes and characteristics of micro-organism exposure may influence the development of food allergy. However, can we be sure that environmental exposures significantly influence the gut microbiota in the first place?

Studies using next-generation sequencing technology have shown that mode of delivery influences newborn colonization patterns, with those born by Caesarean carrying maternal skin-origin microbiota across oral, nasopharyngeal, skin and intestinal compartments [37]. Antibiotic studies have shown that microbiota manifest a recovery after therapy, however with difference in relative abundance of taxa [38]. Porcine animal models demonstrated striking differences in their gut microbiota and host-immune response, depending on whether they were sow-reared, indoors or outdoors [39]. The introduction of new food into the infant diet supports the development and diversification of gut microbiota, which in turn alters the immunological function of gut epithelia and the host-immune response as a whole [40, 41]. This is particularly pertinent in the light of recent randomised trials showing that early complementary feeding with food allergens such as peanut and egg have the capacity to prevent their respective food allergies [10, 42]. Thus, a range of environmental factors are associated with alteration of host gut microbiota characteristics.

A systematic review has collated a range of evidence reporting microbiota-related environmental factors which are associated with the development of food allergy [43] . The largest case control study investigated 16,237 cases of cow’s milk allergy with age- and sex-matched controls, using a Finnish registry of state-funded hypoallergenic formula prescription. The authors reported that being born by caesarean section was associated with an increase in milk allergy diagnosis (adjusted Odds Ratio 1.18 [1.10–1.27]) [44]. The same study also showed that having a greater number of siblings was associated with protection from food allergy (p_trend < 0.01 across six categories of previous number of deliveries). A second population-based birth cohort of 5,276 one year old infants that diagnosed egg allergy using a challenge-based algorithm also demonstrated a significant protective association with having siblings (adjusted Odds Ratio 0.72 [0.62–0.83] per sibling) [45]. These relationships support the hypothesis that microbiota-related environmental factors influence the development of food allergy. However, have mechanisms been described by which the gut microbiota may mediate these associations?

There is good animal model data elucidating different mechanisms by which the gastro-intestinal microbiota may influence the development of food allergy.

Germ-free mice, grown without any environmental exposure to microbes, have a strong disposition towards developing food allergy. These germ-free mice show that lacking exposure to environmental microbiota results in a profoundly under-developed immune system, with less local and lymphoid IgA and a paucity of plasma cells in their gastro-intestinal mucosa. They also suffer a dramatic reduction in their secondary lymphoid tissue, with smaller mesenteric lymph nodes that have no germinal centres and fewer Peyer’s patches [46]. The intraperitoneal administration of ovalbumin promotes sensitization and allergy to ovalbumin amongst mouse models. Regular oral feeding with ovalbumin before this intraperitoneal administration amongst wild-type mice with a normal gut microbiota prevents sensitization, and assures oral tolerance to ovalbumin. However, sensitised germ-free mice are unable to produce the Interferon-γ and IgG2a associated with tolerance induction, and instead uniformly produce high levels of IgE and much higher levels of Interleukin-4 (IL-4) in particular on ovalbumin challenge [47]. It therefore appears that the gut microbiota facilitates the development of oral tolerance.

A critical view may suggest that germ-free mice are not adequately representative of human pathophysiology. Antibiotic exposure may be considered a reasonable surrogate for an environmental factor associated with reduced gut microbiota abundance. Numerous studies show that rearing wild-type mice with commensal microbiota, treating them with a combination of broad-spectrum antibiotics followed by peanut sensitization and challenge procedures causes increased IgE levels and increased population of circulating basophils [48 - 50]. Both gradual re-population of commensal gut microbiota after antibiotic treatment and the introduction of clostridia strains by gavage are associated with attenuation of this allergic immune-phenotype [48, 50]. Colonisation with a consortium of clostridia (predominantly composed of members of Clostridium clusters XIVa, XIVb, and IV) has been shown to induce the production of IL-22 in the colon of antibiotic-treated mice, block food allergen sensitization and prevent the clinical responses analogous to peanut-induced anaphylaxis amongst pathogen-free murine models [48].

Some groups have demonstrated that the generation of colonic regulatory T cells are specifically promoted by clostridia species from the human gut microbiota, and that these cells allow attenuation of the allergic response [51, 52]. Others have demonstrated successful tolerance induction and regulatory B cell promotion using allergen immunotherapy amongst pathogen-free murine models when combined with administration of Clostridium butyricum [53]. It may be that clostridia inhibit the development of allergy by inhibiting the systemic absorption of key food allergens such as peanut, lowering its availability in the blood stream [48]. These models suggest that clostridia species may fulfil the ‘Old Friends’ Hypothesis for best supporting the development of oral tolerance to food allergens.

Alternatively, dietary substrates or their resulting metabolites such as short-chain fatty acids may influence the development of food allergy, rather than the presence of specific groups of bacteria such as clostridia species. Diets high in fibre, for instance, appear to impact on the relative balance of different constituent groups amongst the developing gut microbiota. This has been reported when comparing gut microbiota from children across different countries and also when adults and mice take part in dietary intervention studies [54 - 56]. Fibre is fermented into short-chain fatty acids by the intestinal microbiota, which interact with G-protein receptors on both colonic epithelial and immune cells [57]. Murine models of inflammatory bowel disease have shown that clostridia species are able to ferment dietary fibre and increase the luminal concentration of short-chain fatty acids such as butyrate proprionate, which in turn stimulate the production of colonic T-regulatory cells [58]. Dietary fibre has the ability to induce tolerance to peanut when compared to a diet without any fibre. Murine models fed fibre demonstrated a greater proportion of CD103+ dendritic cells and gut homing T-regulatory cells in their mesenteric lymph nodes, and did not develop sensitization or symptoms of anaphylaxis on peanut challenge [59]. The group showed that their dietary fibre increased the ability of CD103+ dendritic cells to covert vitamin A to retinoic acid, thereby promoting the differentiation of naive T cells into T-regulatory cells.

Animal model work has also shown that more diverse gut microbiota facilitates oral tolerance induction. A diverse gut microbiota stimulates T-regulatory cells to produce IL-10 and TGF-β, and has been shown to repress the development of Th2 allergic responses [60, 61]. The microbiota may induce these immunological responses through specific immune modulators. For example, Bacteroides fragilis, a common commensal in the human gut microbiota, carries Polysaccharide A, a key molecule capable of stimulating CD4 ⁺ CD45Rb and down-regulating inflammatory cytokine cascades [62, 63]. On the other hand, other teams have shown that the diversity of gut microbiota influences the capacity of gut microbiota to facilitate oral tolerance, rather than the carriage of particularly immunologically active moieties. The introduction of eight altered Schaedler commensal gut microbiota strains to five wild-type germ-free mouse strains induces de novo generation of mucosal CD4⁺CD25⁺ Foxp3⁺ cells, allowing the local production IL-10 [64]. By removing these T-regulatory cells from a model of oral tolerance, we see a relapse into allergy. Specific pathogen-free mice receiving PC61 anti-CD25 monoclonal antibody are no longer able to support tolerance after oral β-lactoglobulin gavage, and instead demonstrate raised β-lactoglobulin specific IgE and reduced ability to suppress IL-5 and IL-13 production from splenic preparations [65]. Perhaps most importantly, the capacity for mucosal tolerance induction in mice depends upon the diversity of commensal constituents amongst their gut microbiota. Germ-free mice, and those inoculated with single or dual commensal strains manifest raised IgE levels. However, increasing the diversity of their commensal gut microbiota reduces this predisposition, such that just less than half of animals inoculated with eight species have undetectable IgE, and all of those administered 40 commensal strains demonstrate undetectable IgE [66].

These models show that specific characteristics of bacteria, their diversity amongst the gut microbiota and specific dietary substrates may all support the development of oral tolerance. However, such animal models are considerably less complex than the gut microbiota associated with infants. So we next evaluate whether these hypotheses may be further substantiated amongst young children.

There are now eight studies exploring associations between the gut microbiota and food allergy or sensitization in children, five of which utilised next-generation sequencing data. Unfortunately, none of the birth cohort studies were large enough to examine associations with challenge proven food allergy and they resorted to reporting food sensitization alone.

Three prospective birth cohort studies have been able to collect faecal samples in early infancy and follow up participants to determine their clinical status at a later point, however each using food sensitization as a clinical outcome rather than diagnosis of food allergy itself. The highest quality study in this area investigated the largest number of participants, with a combined total of 324 infants recruited from Sweden, England and Italy, to ascertain associations between food sensitization and gut microbiota characteristics across Europe. Six faecal samples were collected from each participant throughout their first year for conventional aerobic and anaerobic culture. No significant relationship was demonstrated between the age of faecal colonization with any cultured species and food sensitization at 18 months, and diversity of the microbiota was not assessed [67]. The second study originated from the Dutch KOALA cohort mentioned previously in which Bifidobacterium species, Escherichia coli, Clostridium difficile, Bacteroides fragilis group and Lactobacillus species were quantified from stool samples at 1 month of age using real-time Polymerase Chain Reaction and were tested for food sensitization at 1, 2 and 6-7 yrs of age. Clostridium difficile was more often found in the faecal samples of those sensitized to food later in childhood, but only significantly so amongst those with a family history of allergic disease (1.68 (1.07-2.65), p<0.05) [68]. The most recent prospective study was published by the team running the Canadian Healthy Infant Longitudinal Development (CHILD) cohort, who requested faecal samples from 166 children in their cohort at three and 12 months of age, and assessed participants’ food sensitization at 1 year. The authors were careful to exclude two infants who already demonstrated food sensitization at the time of their faecal sample at 3 months to ensure the analysis was entirely prospective rather than cross-sectional. However, the CHILD group also excluded participants with ‘microbiota-disrupting exposures’ which included any children who had received antibiotics, formula or were born by caesarean, leaving only 38 children in the analysis. Each quartile increase in overall gut microbiota richness at 3 months was associated with a 55% reduction in risk for food sensitization (adjusted Odds Ratio 0.45, 0.23–0.87) [69].

Four studies have used cross-sectional designs to investigate the gut microbiota of children with established food allergy, and one further assessing those with food sensitization. Two studies originated from Spanish infants aged two to 12 months who were diagnosed with IgE-mediated cow’s milk allergy by milk challenge at a tertiary referral centre, matched to healthy age/sex-matched controls attending hospital for routine appointments. The first study ascertained through conventional aerobic and anaerobic culture-based techniques that infants with cow’s milk allergy demonstrated a higher number of total colony-forming units of bacteria (p=0.002) and of anaerobes (p=0.002), whereas a lower number of yeasts (p=0.001) [70]. The same study team used 10 fluorescent in situ hybridization probes to characterise their relative abundance of faecal microbiota constituents. Clostridium coccoides (p<0.01) and Atopobium cluster species (p<0.01) were significantly more common amongst faecal microbiota from infants with cow’s milk allergy, but there were no differences with regard to Bifidobacteria, Lactobacilli and Bacteroides [71]. Furthermore, the children with milk allergy yielded increased faecal concentrations of short-chain fatty acids such as butyrate. None of these results support the animal model work suggesting clostridia taxa and short chain fatty acids may protect against the food allergy phenotype.

Two cross-sectional studies have used next-generation sequencing to assess faecal microbiota of children with established milk allergy. The first to be published showed that infants with cow’s milk allergy have significantly more diverse faecal microbiota (Shannon’s index, healthy=1.7±0.8 vs milk allergy=2.6±0.4, p<0.0001) and also significantly more even (Pielou’s evenness; healthy = 0.52 ± 0.2 versus milk allergy = 0.6 ± 0.3, p<0.05) faecal microbiota than age-matched healthy controls [72]. The large cross-sectional study amongst children with established milk allergy assessed faecal samples from 226 American children and compared the diversity of their faecal microbiota according to whether their milk allergy resolved by 8 years. Faecal samples from milk allergic children enrolling between 3 and 6 months of age demonstrated a greater diversity and enrichment of Clostridia and Firmicutes phyla if they grew out of their milk allergy by 8 years, supporting the putative benefits of clostridia [73].

A further two cross-sectional studies have assessed gut microbiota by next-generation sequencing amongst food allergy patient case series. The first enrolled 79 infants from clinical practice, 34 of whom were diagnosed with food allergy through skin and serum testing for sensitization and oral food challenge and 45 demonstrated negative test results and were controls. This clinical study showed no difference in diversity between the two groups, although their control population was selected from a similar cohort as their patients [74]. At the family level of classification, Clostridiaceae type 1 organisms were more prevalent amongst infants with food allergy than controls (p<0.02). The second study examined the gut microbiota amongst 45 children aged between 6 and 24 months of age according to their food sensitization measured by any specific IgE over 0.35 kUA/ml to milk, egg white, wheat, peanut, soy or gluten. Food sensitised children carried a lower Shannon diversity of both the total microbiota (p=0.01) and the bacterial phylum Bacteroidetes (p=0.02), and significantly lower overall gut microbiota richness (p=0.04) as measured by the Chao1 diversity index [75]. However, the class of clostridia and the majority of clostridia derived genera were more abundant in the children with food sensitization than controls (p<0.01). Interestingly, this work supports diversity hypothesis over the clostridia line of enquiry.

These eight studies do not provide good support for current animal model data, but neither are they robust in their clinical design. One of the prospective cohort studies suggests that lesser gut microbiota richness may predispose towards food sensitization, although two of the cross-sectional studies of established milk and food allergy respectively report increased diversity amongst allergic cases. Furthermore, a number of these studies find increased clostridia groups of constituents or species amongst cases with food allergy, which is counter to the animal model literature. However, thesse studies amongst children have significant shortcomings. None assess infants’ gut microbiota before they have developed allergy diagnosed by food challenges. The case control studies assess participants’ gut microbiota at the time of clinical diagnosis, conflating correlation with causation. Any priming of the gut microbiota towards food allergy must occur well before the relevant children present to specialist allergy centres. None of these studies control for potential confounders such as dietary choices, antibiotic exposure or mode of delivery, each of which may influence the development of allergic disease and are likely to have independent relationships with gut microbiota characteristics. Studies comparing diversity indices, relative abundance of clostridia or short chain fatty acid concentrations without parallel investigation of potential bias may be significantly flawed. It is therefore too early to draw together threads of cohesive support amongst the studies of children where consistency is currently lacking.

Future studies should use a population-based approach to recruit infants early in life, and prospectively assess their dermatitis, sensitization and challenge-proven food allergy status, alongside likely confounders such as antibiotic usage and dietary intake. Gut microbiota characteristics would be most sensitively explored within a study capable of adjusting for the relative strength of associations between epidemiological exposures, rather than amending their sample size post-hoc according to assumed ‘microbiota-disrupting exposures’ or comparing highly heterogeneous groups of children [69, 73]. Observational studies are underway amongst well known randomised control trials such as the Enquiring About Tolerance (EAT) Study, which will help understand the influence of the microbiome in the context of population-based research [10]. Commercial treatment trials have the opportunity to assess potential confounders such as atopic dermatitis and microbiological confounders such as antibiotic usage to increase the rigour of their findings and develop greater understanding for designing better treatments. Thus far, microbiome studies in humans have not been able to adequately support or refute the respective mechanisms proposed by experimental models linking the gut microbiota to food allergy phenotypes. Nonetheless, next-generation sequencing is becoming more affordable for their use amongst large scale studies. More robust insights will be achieved through using more than one microbiological technique amongst study participants. In particular, researchers should consider utilizing both metagenomic and 16S rRNA gene sequencing in parallel with additional assessment of relevant bio-markers, to raise the integrity of research in this area.

Forthcoming consensus statements relating to the methodological design of microbiome studies will forge progress in this field, although the application to population-scaled research for young infants continues to require investigation. [Costea PI, et al. Nat Biotechnol. 2017 - standards in fecal metagenomic processing].

Our understanding of the gut microbiome is growing. Early studies suggest it has the capacity to influence the development of atopic outcomes and mechanistic data offers insights about its potential impact. However, studies amongst children vary greatly in design and the resulting data amongst infants do not yet consolidate a single pathway of relevance. Through careful investigation using a number of microbiological methodologies and judicious assessment of co-factors, the role of the gut microbiota may be further evaluated and its role in the development of food allergy be fully appreciated.

Not applicable.