We have demonstrated that infants genetically susceptible for CD (DQ2+/DQ8+/DQ2–8+) in which gluten introduction in their diet was delayed from 4–6 months of age to 12 months of age showed a decrease in immune response to gluten and a lower incidence of CD autoimmunity. Therefore, our results suggest that a delayed introduction of gluten in the diet of genetically susceptible infants can at least delay the onset of the disease. This observation was accomplished by using a prospective clinical study that included a dietary intervention with randomized and double-blind allocation to two diet groups (gluten-containing and gluten-free diets between the ages of 6 months to 1 year). This unique longitudinal study design allowed for the prospective collection of biological specimens that included stool samples, as well as behavioral and diet metadata.
In this study, we tested the effect of early exposure to gluten (6 months of age) or late exposure (12 months of age) on the immune response to gluten and development of CD autoimmunity in 26 genetically susceptible infants. We showed that delayed exposure to gluten has a positive effect on prolonging gluten tolerance and delaying onset of CD autoimmunity. During early childhood, antibodies can fluctuate between positive and negative, in line with the fluctuation seen in the incidence of AGA IgG in this study 
. This study and the resulting observations are important because, despite the significant progress made in understanding the adaptive immunological aspects of CD pathogenesis, the early steps following intestinal mucosal exposure to gliadin that lead to the loss of tolerance and the development of the autoimmune process are still largely unknown. Increasing evidence in the literature seems to suggest a dysfunctional cross talk between innate and adaptive immunity as the key pathogenic element in the autoimmune process of the disease 
. Recent retrospective studies also suggest that this dysfunctional cross talk could be influenced by the timing of gluten introduction into the diet, but more importantly, the studies suggest a potential role of the GI microbiota composition in subjects genetically susceptible to CD 
CD incidence is markedly increasing 
, along with other immune-mediated disorders such as inflammatory bowel diseases (IBD), asthma or atopy 
. The rapidity of the increase in disease rates could never be solely explained by changes in genetic make-up 
. This hypothesis is supported by our recent prospective study on a single American cohort followed since 1974 indicating that CD autoimmunity in these subjects doubled between 1974 (1 of 501 subjects) and 1989 (1 of 219 subjects), thus excluding the genetic component as the cause of this increased prevalence 
. Rather, alterations in host-commensal microbial interactions could have a pivotal role in the development of autoimmune disorders by triggering increased immune stimulation, epithelial dysfunction and enhanced mucosal permeability 
. The colonization of the GI tract, i.e. the succession of microbial communities that are established in the GI tract starting at birth, has been identified as a key factor influencing the risk of autoimmune and food-related diseases 
. The involvement of intestinal colonization in the maturation of immune responses is well characterized in animal models, showing that both the intestinal-associated immune system and systemic immunity mature upon stimulation by the GI microbiota. Germ-free animals show extensive defects in the development of gut-associated lymphoid tissues and in antibody production, and have fewer and smaller Peyer's patches and mesenteric lymph nodes 
. Furthermore, it has been shown that germ-free animals have impaired development and maturation of isolated lymphoid follicles 
. Recent studies also suggest that intestinal bacteria interact with the mammalian immune system to direct the differentiation of both pro- and anti-inflammatory T cell populations 
. Because the microbiota has marked influences on the immune system, we have hypothesized that deviations from the “normal” development of the microbiota may alter the outcome of immune development and potentially predispose individuals to inflammatory diseases.
Our longitudinal study design allowed us to characterize the GI microbial colonization processes from 7 days to 24 months of age in DQ2+
infants from two intervention groups. One of the major finding of our study is that unlike in infants without a family history of CD and without genetic susceptibility to CD 
, the GI tract microbiota in DQ2+
infants appears to be lacking significant numbers of member of the phylum Bacteroidetes. The GI tract microbiota in these DQ2+
infants does not stabilize nor resembles adult microbiota at 1 year of age, and this characteristic remains at 24 months of age. A comparison with the study of Palmer et al. 
highlights these major differences. In non-susceptible infants, the GI microbiota composition changes with different life stages, the most important of which is the introduction of solid food around 6 months of age. Overall the microbial ecosystem in each healthy baby achieves stability converging toward a profile more similar to that of an adult in the first year of life 
with the level of Bacteroidetes ranging from a few percent to over 50% by 1 year of age. A recent paper by Koenig et al. 
who described the GI colonization process in one infant for 2.5 years, further support the results of Palmer et al.
. In that study, Bacteroidetes comprised between 40 and 60% of the communities from 6 months to 2.5 years of age and the GI communities appeared to stabilize at 1 year of age 
. It is evident that the colonization process in these infants with a higher abundance of Proteobacteria and Bacteroidetes throughout the first year of life and a lower abundance of Firmicutes, is dramatically different from those in the present study who are DQ2+
and with a family risk of CD. A complete clustering analysis of GI microbial communities from non-susceptible infants and the communities in this study revealed two major groups () separating infants by HLA genetic background and family history of CD. The result is highly influenced by the high abundance of Firmicutes and low abundance of Bacteroidetes in infants genetically susceptible for CD (mostly time points D, E, F and H [>6 months]), and by the high abundance of Proteobacteria in the earlier time points in the Palmer et al. study 
The ratio of Firmicutes over Bacteroidetes is known to vary throughout the lifespan. The ratio is lower in the first year of life, becomes higher in adulthood, and decreases in elderly 
. Bacteroidetes have also been found to delineate profound differences between African and European children 
, with significantly higher levels of Bacteroidetes in African than European children. While certainly influenced by diet, a higher abundance of Bacteroidetes in African children appear to be protective against pathogens and other gastrointestinal diseases 
and indicates a potential impact of the Western diet on the colonization and the establishment of the GI microbiota in Europe and other developed countries. The beneficial role of members of the phylum Bacteroidetes
, including Bacteriodes fragilis
, has been previously demonstrated 
. B. fragilis
has been shown to establish a cross-talk between the GI microbiota and the intestinal epithelium 
. B. fragilis
, through the production of a polysaccharide capsule, directly induces the development of FoxP3+ regulatory T cells, which in turn produces anti-inflammatory cytokine directly in the gut 
. The lack of Bacteroidetes could represent a major predisposing deficiency in infants genetically susceptible for CD, since it has been reported that quantitative and/or qualitative defects of FoxP3+ regulatory T cells affect immune tolerance surveillance and, therefore may lead to the autoimmune response typical of CD 
Our findings of decreased Bacteroidetes abundance in children at risk of CD are at odds with previous reports showing higher Bacteroidetes representation in CD children 
. However, substantial differences between our study design and the approach used by other investigators in regards to microbiota analysis (fluorescent in situ hybridization coupled with flow cytometry), age of children studied (older children), and clinical characteristics (children already affected by CD) make previous findings not directly comparable to our results. Indeed, to our knowledge, our study is the first prospective report on infants at risk of CD that analyzes the microbiota dynamic over time starting from birth, using state-of-the-art microbiome 16S rRNA short amplicon pyrosequencing combined with quantitative PCR and 1
H NMR spectroscopy. The key finding of our report is the lack of maturation of the gut microbiota within the first 2 years of life in infants at risk of CD characterized by a relative absence of Bacteroidetes and a parallel high abundance of Firmicutes. It is important to note that differences in stool sampling and storage, DNA extraction and purification methods, 16S rRNA gene PCR primer pairs, and sequencing methods between all studies, including this and the Palmer et al.
could potentially explain some of the variations observed in the composition of the GI microbial community. However, it is unlikely that the DNA extraction method used in this study is responsible for the observed low abundance of Bacteroidetes, as high abundance of this phylum was detected in stool samples () from adult subjects (HLA DQ2/8- and some HLA DQ2/8+
on gluten free diet for more than 24 months). Furthermore, the method is commonly used to extract stool samples in our laboratory and members of the phylum Bacteroidetes and other phyla are often identified (data not shown).
Metabolomic profiling of selected subjects and samples revealed that few metabolites appeared important and accompanied the succession of the microbial taxa colonizing the GI during the first two year of life. The GI metabolome during the first 6 months of life reflects the infants' diet of exclusively milk comprised mainly of polysaccharides and other sugars. This unique metabolic profile is very similar in all infants. However, once solid food is introduced at 6 months of age, a major shift occurs and the SCFA succinate, acetate, propionate and butyrate are found in the feces. By 24 months, butyrate and acetate are the main SCFA present in the metabolome of these infants. A recent study found Bacteroidetes associated with the presence of butyrate, and acetate and most strongly with propionate, while Firmicutes was negatively associated with these SCFAs 
. The role of these SCFAs is thought to be protective and the results of the breakdown of complex plant polysaccharides by Bacteroidetes 
. Propionate was not found in high level in the feces of the infants studied compared to other published studies 
. This finding correlates with the concordant lack of Bacteroidetes in the GI microbiota of infants genetically susceptible for CD. One can envision that the high abundance of Firmicutes and the low abundance of Bacteroidetes in these infants results in lower levels of SCFAs in the GI tract and a diminished GI health and/or a predisposition for CD or other autoimmune diseases. More quantitative methods and a higher number of subjects would be needed to confirm this finding. The metabolic profiles are consistent across the infants and cluster by age (). This finding supported the concept that while GI microbiota may differ in microbial species composition and abundance, they conserve a functional core, whether it is as conserved gene content 
or as observed in this study in conserved metabolic output. Our metabolomic analysis did not reveal any consistent differences between infants from each dietary intervention group (). However, because one of the infants (subject 8) was diagnosed with CD at 24 months, we had the opportunity to prospectively examine both the GI microbiota colonization process and its associated metabolome. Interestingly, between 6 and 12 months of age, the metabolome of subject 8 contained high levels of lactate (), which correlated with the presence of high levels of Lactobacillus
spp. in the GI microbiota (Figure S3
). Two more subjects showed elevation of lactate in their metabolome profile. Subject 15 (group A) showed elevated and sustained lactate levels within an interval period similar to subject 8 (from 6 to 10 month of age) (). Interestingly, subject 15 developed type 1 diabetes (T1D), another autoimmune disease, at 22 months. The second subject with elevated lactate (subject 20, group B) showed only a very transient increase in lactate at age 8 months which returned to baseline levels similar to other infants by 10 months of age (). Based on these results, it is tantalizing to hypothesize that a decrease in Lactobacillus
spp., with a subsequent decrease in lactate production during a crucial time of maturation of mucosal immunity functions (between 6 and 12 months of age) can leads to loss of tolerance to non-self antigens (gluten in case of CD, unknown antigen(s) in case of T1D) in genetically susceptible individuals. Larger studies with more cases are necessary to support this hypothesis.
Subject 8 also experienced reduced bacterial richness during the intervention period compared to the other infants, which showed increased richness with time (Figure S4
). While this was observed in only one infant, it introduces the possibility of discovering potential biomarkers that could be predictive to the development of autoimmunity in CD; a longitudinal study design is essential to achieve this goal.
In summary, infants genetically susceptible for CD may benefit from delayed exposure to gluten from 6 months of age to at least 12 months of age. While the molecular mechanisms underlying the benefit are yet unknown, it might be related to a lack of maturity of the GI microbiota in these infants, and we hypothesize that the introduction of gluten in an immature GI microbiota could trigger or accelerate the development of autoimmunity. By combining 16S rRNA gene short amplicon pyrosequencing, quantitative PCR and 1H NMR spectroscopy to analyze the microbiota of infants with genetic predisposition for CD over the first two years of life, we have characterized the GI colonization process and its metabolic output in infants genetically susceptible for CD. The data presented here, while derived from a relatively small number of subjects, suggest significant differences between the developing microbiota of infants with a genetic predisposition for CD and those from infants with a non-selected genetic background. Furthermore, the metabolic output of the GI microbiota in these infants while similar to one another within age groups might reflect a potential dysbiosis of the GI microbiota and lead to less than optimal cross-talk with the host to promote health. Interestingly, one of the infants in the study was diagnosed with CD at 24 months of age. The retrospective analyses of the GI microbiota and metabolomic data suggest that potential specific biomarkers might be identified that would be predictive for autoimmune development in subjects genetically at risk, possibly leading to the development of potential interventions during the pre-clinical phase of the disease to arrest the loss of tolerance to gluten and, therefore, to prevent the onset of CD autoimmunity.