Celiac disease (CeD) is a chronicimmune-mediated enteropathyprecipitated by to dietary gluten in genetically predisposed individuals (1). Current diagnosis is based on demonstrating the enteropathy in small intestinal biopsies where histologic examination shows villous atrophy, crypt hyperplasia and intraepithelial lymphocytosis, and the presence of circulating CeD-specific antibodies to tissue transglutaminase (tTG), deamidated gliadin peptides (DGP), and endomysium (EMA). In children who have symptoms suggestive of CeD, a strongly positive tTG antibody (tTGA) titre and a CeD-associated HLA genotype, the diagnosis of CeD may be possible without the need for small intestinal biopsy (2). Since the 1950s when gluten was identified as the causative trigger of CeD, a strict and lifelong gluten-free diet (GFD) has been the mainstay of treatment.

While CeD is common around the globe and is rising in prevalence in many populations, it is frequently undetected in clinical practice (3). Symptomatic and untreated disease is associated with elevated morbidity and mortality and impaired quality of life (4–7). The clinical presentation is broad and includes gastrointestinal upset, chronic fatigue, nutrient deficiencies, poor growth, and failure to thrive. Extra-intestinal manifestations are common, and while they were once considered more frequent in adults than children with CeD, recent data suggests the frequency is similar in CeD children, although the types and rates of recovery differ (8). In children, extra-intestinal manifestations include short stature, anemia, delayed puberty, dental enamel hypoplasia, reduced bone density, oral ulcers, liver and biliary disease, and dermatitis herpetiformis. Poor growth and anemia tend to be the most common and there is a correlation with more severe histological damage at diagnosis compared to children with a gastrointestinal presentation or screen-detected cases (9). Insidious effects of undiagnosed CeD in children include behavioral disturbances and reduced educational performance (10).

CeD is also associated with an increased risk of autoimmune diseases including Hashimoto's thyroiditis, Graves' disease and type 1 diabetes (T1D) (11–14). A large Danish population study revealed the prevalence of autoimmune diseases was 16.4% among CeD patients compared with 5.3% in the general population in 2016 (12). Approximately 5% of patients with CeD have T1D and ~6% of T1D patients have CeD (15). In Northern Sardinia, a population with a high prevalence of CeD (16), patients with autoimmune thyroiditis had a 4-fold greater prevalence of CeD than the general population and while iron deficiency was present in almost half, none had gastrointestinal symptoms (17). The co-occurrence of autoimmune diseases supports the concept of shared genetic and immune pathways contributing to immune dysregulation and loss of self-tolerance, however it remains unclear whether CeD directly leads to other autoimmune disease and whether early diagnosis and treatment with a GFD alters this risk (18). Next generation sequencing of the HLA region shows that extended class II haplotypes differ between populations (19); this may partially explain regional differences in the degree of association between CeD and autoimmune disease. The strong association of CeD with autoimmunity, especially T1D and autoimmune thyroid disease, supports screening for these conditions in CeD patients.

The consumption of gluten containing foods is necessary for CeD to develop. Gluten is the viscoelastic protein that remains after washing dough and is composed of alcohol-soluble gliadins and alcohol-insoluble glutenins. The rheological properties of gluten allow it to impart a light and extensible texture to food making it highly favored in the food industry. Modern wheat gluten arises from a hexaploid genome, making it heterogeneous and more genetically complex than the human genome. Similar proteins rich in glutamine and proline (thus, the collective term prolamins) are found in barley and rye and termed hordeins and secalins, respectively, and are also toxic in CeD. Avenin, the prolamin component in oats, is phylogenetically distinct from wheat, barley and rye prolamin. Oats are considered safe for consumption by most people with CeD, although adverse immune and clinical effects have been reported (20–22) and some cultivars may be more immunogenic (23). Several expert reviews have concluded that further research into the toxicity of oats in CeD is needed (24, 25).

Why CeD develops in some people remains unanswered. Epidemiologic and prospective observational studies implicate a range of environmental factors that impact disease development. A role for the microbiome in the development of tolerance and disease pathogenesis is also emerging. Genetic and immunological studies have revealed the importance of key HLA and non-HLA susceptibility genes in disease development and a long-lived pathogenic population of gluten-specific T cells targeting certain gluten peptides (T cell epitopes). The emerging story of CeD development is one where environmental factors increase the risk of CeD in genetically predisposed individuals by shaping the immunologic context in which gluten is presented and shifting the balance from gluten tolerance to reactivity, and that this may be in part mediated through microbiome-host interactions. Contemporary clinical issues of importance include expediting the detection and diagnosis of CeD, improving and quality of life and health outcomes for those diagnosed, and developing treatments that are more effective and less burdensome than the current approach of a lifelong and strict GFD. This article will review current knowledge about CeD development and pathogenesis and how this knowledge is being applied to explore novel diagnostic, treatment and preventative approaches.

CeD is a global disease that has been reported in Western and Eastern Europe, North America, South America, Asia, Oceania, and Africa (3, 26). It appears to be relatively uncommon in southeast Asia and sub-Saharan Africa. Recent reports from China suggest the illness could be substantially unrecognized there however more biopsy-based studies are required (27). In a systematic review and meta-analysis, the global seroprevalence and biopsy-confirmed prevalence of CeD was estimated to be 1.4% and 0.7%, respectively (3). The sero-prevalence of CeD in the US from National Health and Nutrition Examination Surveys (NHANES) was 0.7% and, consistent with a range of population studies from around the globe, showed that most cases remain undiagnosed in the community (28). Since CeD is frequently undiagnosed an active case-finding approach is considered best practice.

The prevalence of CeD varies with sex, age, and geographic location with the frequency of predisposing HLA haplotypes in the general population and per-capita wheat consumption the two major determinants of prevalence. There is a modest gender bias in favor of females (29). Familial clustering in CeD is common with 10% of first-degree relatives of a CeD sufferer affected. The high concordance rate for monozygotic twins (~80%) compared to HLA-identical siblings (~30%) and dizygotic twins (~10%) underscores the importance of both genetic factors (HLA and non-HLA genes) and the environment in CeD risk (30).

The high prevalence of CeD observed among populations living in areas with a high consumption of wheat products are highly suggestive for involvement of dietary gluten in CeD development (26). Although intake of gluten is necessary for CeD to develop, it does not solely explain why not all genetically-predisposed individuals consuming gluten develop CeD and why the disease can develop later in life despite many decades of gluten intake. Significant differences in the prevalence of CeD between people of similar genetic background and wheat intake living in nearby regions (for example, Finland and Russian Karelia) is strong evidence that the risk of CeD is influenced by other factors apart from genetic susceptibility and wheat consumption (31). Indeed, population studies have implicated a range of environmental factors associated with CeD risk (summarized in Table 1). Heterogeneity of study design has yielded conflicting results in the search for triggers in CeD. We have limited our review to focus on the association between CeD and diet, infections, antibiotic use, and delivery mode.

Microbial colonization occurs at birth and shapes the development of the mucosal and systemic immune system and the intestinal barrier. These host-microbe interactions continue throughout life, and a disruption of these interactions, through altered bacterial composition or functions, have been hypothesized to increase the risk of a range of autoimmune or inflammatory diseases such as CeD. Altered microbiota composition in patients with CeD may represent an environmental modifier of CeD development.

An early study described the presence of rod-shaped bacteria in duodenal biopsies of Swedish children with CeD born during the epidemic, which weren't observed in biopsies of control children (80), or in children born following the epidemic (81). The bacteria were subsequently identified as Clostridium spp, Prevotella spp, and Actinomyces spp, and their presence was suggested to be a risk factor for CeD that contributed to the increase in disease incidence in Sweden from 1985-1995 (81). Subsequent clinical studies have described differences in both fecal and duodenal microbial composition in children and adults with active compared with treated CeD, or healthy controls (82). While no specific microbial signature has been described for CeD, many groups have described increases in the proportions of Bacteroides and members of Proteobacteria, and decreases in Lactobacillus and Bifidobacterium (83, 84). In addition, CeD patients suffering from persistent symptoms were shown to have increased abundance of Proteobacteria compared to those who were asymptomatic (85). While these studies suggest an association between altered microbial composition and development of CeD, studies exploring mechanisms and causality are lacking. Moreover, whether alterations in the microbial composition are a cause or consequence of small intestinal inflammation has not been fully elucidated.

Recent studies have suggested that the microbiota from CeD patients may harbor more pathogenic or pro-inflammatory bacteria. Reports of CeD diagnosis following Campylobacter jejuni infection (86) suggest that bacterial infections could precede CeD development. Escherichia coli clones isolated from CeD patients expressed a higher number of virulent genes compared to those isolated from healthy controls (87). Similarly, the presence of virulent genes were higher in Staphylococcus spp and in Bacteroides fragilis strains isolated from CeD patients compared to healthy controls (88, 89). Importantly, strains isolated from CeD patients were more pro-inflammatory in vitro and stimulated altered DC morphology, characteristic of DC maturation, increased pro-inflammatory cytokine production, and altered epithelial barrier integrity. Similarly, Neisseria flavescens, a member of Proteobacteria, was identified in the duodenum of active CeD patients but not from control subjects and induced an inflammatory phenotype in human and murine DCs (90).

In contrast to the above studies, bacterial infections may also protect against CeD development. Some studies indicate an inverse relationship between the presence of Helicobacter pylori and CeD in both adults and children (41–43) whereas other studies have shown a positive or no association (44, 45). Mechanisms underlying this association have not been elucidated and inconsistencies across studies may relate to differences in techniques used to determine H. pylori status or H. pylori virulence. Less virulent strains may exacerbate the mucosal response in CeD whereas more virulent strains may provide protection against CeD (45, 91).

Functional differences in the microbiota could also affect metabolic processes important in CeD pathogenesis. The gastrointestinal tract harbors diverse bacteria that participate in gluten metabolism in vitro and this may differ between healthy individuals and those with CeD (92–94). As most studies have measured microbial composition in active or treated CeD compared to healthy controls it is difficult to determine whether functional differences are present prior to disease onset.

To gain insight into the potential role of microbial factors in CeD development, previous studies have profiled the fecal microbial composition of genetically at-risk children. High-risk children were shown to harbor a different microbiota compared to children who were at low genetic risk for CeD (95, 96), suggesting that the high-risk genotype may influence early gut microbiota composition. Infants at the highest risk for CeD had a higher prevalence of enterotoxigenic E. coli compared to those at low or intermediate risk for CeD (97). In addition, in a cohort of 164 infants, those at risk for CeD had lower numbers of Bifidobacterium spp and B. longum and increased numbers of B. fragilis and Staphylococcus spp. The differences in Bacteroides and bifidobacteria were attenuated by breastfeeding (98). At-risk children that later developed CeD were recently shown to have an altered microbial trajectory that coincided with immune changes. These changes were suggestive of a “premature maturation” of the gut microbiota in children who went on to develop CeD (99). On the other hand, the fecal microbiota of at-risk infants who went on to develop CeD was similar at 9–12 months to those infants that remained healthy by the age of four (100). Whether the duodenal microbial composition or function is altered in at-risk individuals that go on to develop CeD needs to be investigated further in larger clinical trials.

Diet and environment also determine gut microbiota composition (101, 102), highlighting the complexity of delineating the influence of genotype and environment on shaping the microbiota. Larger clinical trials where both the composition and function of the microbiota is studied in at-risk individuals and followed over time are needed to help understand gene-microbe interactions in CeD development.

While environmental factors are important for CeD development a notable feature of CeD is its high heritability and strong HLA association (103). This strong genetic association reflects the central role of CD4+ T cells as the HLA molecules associated with CeD bind specific gluten peptides that activate T cells (104). Ninety percent of Caucasian CeD patients possess the HLA-DQ2.5 haplotype (encoded by the DQA1^*05:01 and DQB1^*02:01 alleles) either in cis or trans positions, and the remaining carry either HLA-DQ8 (encoded by the DQA1^*03:01 and DQB1^*03:02 alleles), HLA-DQ2.2 alone (encoded by the DQB1^*02:02 allele) or HLA-DQ7 alone (encoded by the DQA1^*05:01 allele). Less than 1% of CeD patients lack these HLA haplotypes (105) and their absence can be exploited in the clinical setting to assist in excluding a diagnosis of CeD.

A “gene-dose effect” related to the number of copies of the DQB1^*02 allele has been reported to affect CeD risk, clinical phenotype and patient responses to a T cell targeted therapy. The presumed basis for this effect is that gluten presented by APCs in HLA-DQ2.5 homozygous (i.e., two copies of DQB1^*02) individuals can induce at least a 4-fold higher T-cell response compared with gluten presented by APCs in HLA-DQ2.5 heterozygous (i.e., one copy of DQB1^*02) individuals (106). The CeD risk in HLA-DQ2.5 homozygous patients is ~2.5 and 5 times that conferred by HLA-DQ2.5 heterozygosity and lower risk HLA groups, respectively (107). A prospective Italian study (Celiac Disease and Age at Gluten Introduction study; CELIPREV) followed newborns with a family history of CeD and showed the risk of CeD autoimmunity (positive CeD-serology panel) at 10 years of age was far higher among children who were HLA-DQ2.5 homozygous (or who had two copies of DQB1^*02 than among those who were HLA-DQ2.5 heterozygous or HLA-DQ8 (38 vs. 19%, P = 0.001), as was the risk of overt CeD (26 vs. 16%, P = 0.05) (34). In this cohort, 80% of those in whom CeD developed did so during the first 3 years of life. In the TEDDY study following 6403 US and European genetically at risk children at for CeD, the risks of CeD autoimmunity and confirmed CeD by age 5 were 11 and 3%, respectively in the heterozygous children and 26 and 11%, respectively, in those who were homozygous (108).

In addition, HLA-DQ2.5 homozygosity has been associated with a more severe CeD phenotype with earlier disease onset, greater villous atrophy, diarrhea, and lower hemoglobin at presentation, and a slower rate of villous healing on a GFD (109), plus a higher rate of refractory (non-responsive) CeD (110). In a recent clinical trial of an immunotherapy targeting gluten-specific CD4+ T cells, CeD subjects who were HLA-DQ2.5 homozygous were more likely to experience gastrointestinal symptoms following systemic administration compared to those who were heterozygous (111).

The contribution of non-HLA genes to CeD risk susceptibility is much less strong (OR < 1.5) than the HLA-associated haplotypes (OR >5) but collectively are significant. More than 70 candidate genes in over 40 non-HLA loci have been implicated in CeD heritability (112–117). These loci encode proteins involved in a range of immune pathways affecting T and B cell activation, chemokine receptor activity and cell migration, cytokine binding, thymic differentiation of CD4+ and CD8+ T cells, stress pathways and innate immunity. Only one gene has been shown to be gut specific (RGS1), underscoring the systemic nature of immune dysregulation in CeD (114). To date there is no evidence to implicate specific alleles encoding gastrointestinal proteases or tTG. Mirroring the frequent disease co-occurrence, there is substantial overlap between genetic risk factors for CeD and those of autoimmune diseases such as rheumatoid arthritis, multiple sclerosis and T1D (112, 115, 118, 119). Furthermore, despite a much weaker clinical association with CeD, there is overlap of genetic risk loci for inflammatory bowel disease such as Crohn's disease (18, 120). Intriguingly, 90% of the identified risk loci map to non-coding regions such as promoter regions, enhancers or non-coding RNA genes, suggesting that regulation of gene expression rather than changes at the protein-coding level are more important for CeD susceptibility and development (121).

The role of CD4+ helper T cells in CeD was confirmed with the isolation of pro-inflammatory gluten-specific CD4+ T cells from intestinal tissue of CeD patients (122) (Figure 1). These pathogenic T cells have a Th1 phenotype characterized by production of IFN-γ and TNF-α (123) and almost all are HLA-DQ2- and/or DQ8-restricted (122, 124, 125). Gluten peptides that have been post-translationally modified by the enzyme tTG in a process called deamidation can effectively activate these T cells (126, 127). Deamidation converts specific glutamine residues to glutamate and this modification enhances the gluten peptide's binding affinity to disease-associated HLA dimers (128–130). Deamidation is crucial in converting poorly immunogenic wild-type gluten peptides to highly immunogenic antigens for CD4+ T cells. The structural requirements that generate effective binding of gluten peptides to HLA-DQ2 or DQ8 and T cells via the T cell receptor (TCR) have been further elucidated in structural studies (131, 132) and assessment of the biased use of TCR genes (133, 134).

Studies of T cells isolated from the intestine of CeD patients, or from their blood after short-term oral gluten challenge, have been used to define the gluten peptides (specifically, the T cell epitopes) immunogenic in CeD. Most studies have focused on the 90% of CeD patients who are HLA-DQ2.5 and a range of immunodominant T cell epitopes have been defined (135). Less is known about the nature of the T cell response to gluten in CeD patients without HLA-DQ2.5. The most immunogenic gluten epitopes for CeD patients with HLA-DQ2.5 after wheat ingestion reside in α-gliadin and ω-gliadin (136). Much of the field has focused on the immunogenicity of T cell epitopes in α-gliadin, specifically those encompassed within a protease resistant 33mer (137). However, gluten challenge studies show that the most immunogenic peptides induced by gluten ingestion depends on whether wheat, rye or barley is ingested, and that a sequence from ω-gliadin (encompassing the T cell epitopes DQ2.5-glia-ω1 and DQ2.5-glia-ω2) is dominant irrespective of which grain is consumed (136). Despite the multitude of immunogenic peptides, just three peptides from wheat and barley appear to recapitulate most of the response to gluten in CeD patients with HLA-DQ2.5. Interestingly, after oat ingestion, about 8% of CeD patients have detectable T cells specific for avenin peptides that share close sequence homology with barley hordein, suggesting that cross-reactive T cells may mediate immune responses following oats ingestion in some CeD patients (21).

While early work suggested T cells from children with CeD displayed a different pattern of reactivity to gluten compared to adults (138), more recent studies show gluten-specific T cells in blood induced by oral wheat challenge (139), or expanded from the small intestine during active disease (140), share the same specificity for deamidated, immunodominant T cell epitopes across all ages. The same gluten-specific T cell clonotypes persist in patients' blood and intestinal tissue up to several decades and share the same TCR gene use motifs in CeD patients from Norway, Finland and Australia (134, 139, 141, 142). Their stability over such long periods of time may be maintained by ongoing gluten exposure as inadvertent gluten intake is common in CeD even when a strict gluten-free diet is attempted (143, 144).

Measurement of tTGA is a useful screening test for CeD as the titer reflects disease activity caused by gluten, however a direct role in disease pathogenesis is less clear. As tTGA are anti-angiogenic they may contribute to some extra-intestinal manifestations of CeD (145). As tTGA are detectable in intestinal tissue prior to the typical enteropathy of CeD and predict future disease onset, antibody production is likely to occur early in disease development (146). However, approximately 50% of children with positive CeD serology normalize their levels despite ongoing gluten exposure, suggesting that gluten immunity that leads to a pathogenic response is not necessarily fixed once it has commenced (147). Supporting this concept is the observation that almost 20% of adults with CeD diagnosed during childhood who elect to resume gluten intake have no evidence of active disease (148).

Production of tTGA appears reliant on T cells, as antibody formation to tTG and DGP is strictly dependent on the presence of CeD-associated HLA types as well as gluten. This supports the idea that tTG-specific B cells internalize tTG in complex with gluten peptides and present gluten-derived peptides to gluten-specific T cells, effectively amplifying the T cell response. These T cells then provide the required “help” to the B cell, resulting in production of tTGA and DGP antibodies (149, 150).

In active CeD, a large number of plasma cells can be found in the intestinal lesion and tTG-specific plasma cells made up a large proportion of them (5–25%) (151). There is much that still needs to be understood about how these antibody producing cells are selected and mature. tTG can form covalently linked multimers with itself that readily bind gluten peptides and can be taken up by tTG-specific BCR transduced cells and activate gluten-specific T cells with increased capacity compared to tTG monomers (152). Immunoglobulin expressed on B cells could act as substrates for tTG, in particular IgD, resulting in BCR/tTG cross-linking (153). As B cell epitopes are in close proximity to immunodominant T cell epitopes (154) and react to a higher degree to deamidated peptides (155) future work needs to examine the B cell and T cell interaction during the gluten-specific immune response.

The events that culminate in the histological changes of CeD are incompletely understood. Gluten T cell epitopes cluster in regions of high proline making them resistant to the effect of gastrointestinal proteases (156). These peptides may pass across the epithelium via transcellular (157–159) and paracellular (receptor or antibody mediated) pathways (160). Increased epithelial permeability, a feature of active CeD, may be mediated by a direct effect of gliadin acting via the chemokine receptor CXCR3 in intestinal epithelial cells on tight junctions (160). DCs are presumed to play a major role in presentation of gluten peptides to CD4+ T cells (161), however little is known about their identity, where this presentation occurs, and the extent to which this role is undertaken by gluten-specific B cells. Activated CD4+ T cells produce large amounts of IFN-γ, that may induce cytotoxicity of intraepithelial lymphocytes (IELs), as well as IL-21, which plays a role in T-cell-dependent B cell responses (162). IL-17 producing CD4+ T cells have been reported in untreated CeD (163) but their role and that of IL-17A production in CeD pathogenesis is less clear (162, 164). IL-15 and IFN-α feature prominently in the inflamed tissue in CeD patients (165, 166). IELs are believed to play an important effector role in mediating destruction of enterocytes in CeD in a TCR-independent manner. When activated by stress signals on intestinal epithelial cells such as HLA-E and MIC-A (167), IELs express high levels of NK activating receptors such as NKG2D and CD94/NKG2C and adopt a cytolytic phenotype capable of destroying enterocytes (168). IL-15 plays a key role by upregulating the activating NKG2D receptor and acting as a co-stimulatory molecule. The effect is to ‘license' cytotoxic IELs with the ability to kill intestinal epithelial cells expressing the stress-induced MIC molecules. Adaptive immunity to gluten and epithelial stress where cytotoxic IELs have acquired an activated NK cell phenotype may both be required for villous atrophy to develop in CeD (169).

The causes of epithelial stress that trigger IEL activation and transformation in CeD are not known. Some hypothesize it is driven by gluten itself or other stimuli such as those resulting from a microbe-host interaction (Figure 2). The innate immune system is a pre-programmed form of host defense that responds rapidly to stimuli. Responses are triggered when pattern recognition receptors, for example, toll-like receptors on macrophages, bind molecules with conserved structures. An innate immune stimulatory effect of a gliadin sequence (A-gliadin p31-43) has been reported, but this work has not been replicated (170). No other gluten peptides activating innate immunity in humans have been defined. Gliadin may function as a stress signal for the activation of MICA expression only in the initial stages of disease and decline to baseline levels once the inflammatory lesion is established. In contrast to the critical role of HLA-restricted, gluten-specific CD4+ T cells in CeD pathogenesis, the relative contribution of innate immunity to disease has not been established in genetic or functional studies and further research in this area is required.

Although gluten consumption is common in the Western diet most individuals who possess HLA susceptibility for CeD develop immunologic tolerance to it as they do for other food proteins. Why a small percentage fail to develop or lose tolerance to gluten remains unresolved. Tregs are an important immune component contributing to intestinal homeostasis and when functioning normally inhibit pathogenic responses to dietary antigens and help maintain oral tolerance. This role suggests they may be important in CeD, however identifying and studying this cell population has been limited by the technical challenge of identifying a population of cells truly gluten-specific and functionally suppressive in vivo. These challenges have meant that studies of Treg function in CeD have been conflicting, with some indicating reduced suppressive function (171–173) and some showing it is retained (174, 175). Further studies that definitively isolate true gluten-specific Tregs and assess their function in CeD are needed.

Interestingly, the immunosuppressive cytokines IL-10 and TGF-β can be detected in high amounts in the CeD intestinal lesion but appear insufficient to counter active disease (176, 177). The high levels of pro-inflammatory cytokines such as IFN-γ and IL-15 may render this anti-inflammatory response inadequate (178, 179). IL-15 is a pleiotropic cytokine that may promote inflammation through several pathways, including driving the accumulation of cytotoxic IELs in the CeD lesion, interfering with the suppressive activity of Tregs (175), impairing TGF-β signaling (176), or allowing activation of disease-specific CD4+ T cells (180). IL-15 also drives the expansion of aberrant IEL clones that can lead to the development of enteropathy-associated T cell lymphoma. The multiple pro-inflammatory effects of IL-15 make it a rational target for therapeutic blockade, and clinical trials of anti-IL-15 biologic agents are underway in refractory CeD (181).

Additional insights into molecular pathways altered in loss of tolerance may be gained by study of gene expression profiles in CeD patients. Analysis of blood-derived non-gluten stimulated CD4+ T cells in CeD showed significant upregulation of the IFN-γ gene and reduced expression of a network of BACH2 regulated genes (182). BACH2 is a transcription factor that plays an important immunoregulatory role in inhibiting T effector cell development and promoting development of Tregs. Advances in techniques such as RNAseq now affords the opportunity to analyze multiple transcriptional and phenotypic features of antigen-specific effector T cells and Tregs to provide insights at the single cell level into these heterogeneous populations.

The findings on Treg numbers in the CeD lesion suggest that the defect in tolerance is not simply due to defects in numbers or recruitment of suppressive cells, but that the pro-inflammatory environment characteristic of CeD overcomes the tolerogenic milieu that normally maintains balance and inhibits abnormal immune responses. However, further studies on disease-relevant Tregs from the small intestine of CeD patients that assess antigen-specificity, function in vivo and the factors that impact their function are required.

Our view of CeD has evolved from a gastrointestinal illness to an immune disease characterized by the presence of specific HLA genes, CD4+ T cells responding to specific gluten peptides, circulating antibodies to tTG and systemic clinical manifestations. Aside from the fact it is driven by an exogenous, dietary antigen, the genetic and immunologic basis for CeD overlaps with that of traditional autoimmune diseases. While HLA susceptibility and wheat consumption and are major determinants of disease development it is apparent that non-HLA genes and a range of environmental factors are important for disease development. Prospective studies have established that the timing of gluten introduction and breastfeeding do not impact the development of CeD. More results from multicenter, prospective longitudinal studies are needed to understand the long-term effects of a high amount of gluten intake and to identify if other environmental exposures might trigger the disease.

Furthermore, although in vitro and in vivo studies suggests there are host-microbe or gluten-microbe interactions that promote gluten-specific immune responses, larger clinical trials where both the composition and function of the microbiota is studied in at-risk individuals and followed over time are needed to help understand gene-microbe interactions in CeD development. These kinds of studies may provide insight into the microbial events leading to CeD development that could be targeted as preventative or therapeutic strategies.

Finally, understanding how tolerance to gluten is lost in CeD is a fundamental question that needs more study. Insights into disease relevant pathways may come from analysis of the genome and gene expression in CeD, and epigenetic studies are needed to examine the impact of environment on gene expression and disease development. While novel therapies for CeD have not yet been tested in children, emerging studies on the role of environmental factors and the microbiome and how they might impact gluten immunity and tolerance may one day make disease prevention possible. For now, as primary prevention of CeD is a highly attractive, but as yet unrealized goal, the focus must be on driving expeditious diagnosis and treatment in symptomatic children and adults.

JT-D wrote the first draft of the manuscript. DA and HG wrote sections of the manuscript. All authors contributed to manuscript revision, read, and approved the submitted version.