Gut microbiota in celiac disease: microbes, metabolites, pathways and therapeutics

Abstract

Introduction:

Current evidence supports a vital role of the microbiota on human health outcomes, with alterations in an otherwise healthy balance linked to chronic medical conditions such as celiac disease (CD). Recent advances in microbiome analysis allow for unparalleled profiling of the microbes and metabolites. With the growing volume of data available, trends are emerging that support a role for the gut microbiota in CD pathogenesis.

Areas covered:

In this article, the authors review the relationship between factors such as genes and antibiotic exposure on CD onset and the intestinal microbiota. The authors also review other microbiota within the human body (such as the oral microbiota) that may play a role in CD pathogenesis. Finally, the authors discuss implications for disease modification and the ultimate goal of prevention. The authors reviewed literature from PubMed, EMBASE, and Web of Science.

Expert opinion:

CD serves as a unique opportunity to explore the role of the intestinal microbiota on the development of chronic autoimmune disease. While research to date provides a solid foundation, most studies have been case-control and thus do not have capacity to explore the mechanistic role of the microbiota in CD onset. Further longitudinal studies and integrated multi-omics are necessary for investigating CD pathogenesis.

1. Introduction

Celiac disease (CD) is a multifactorial autoimmune condition that is characterized by T-cell mediated small intestinal damage triggered by gluten ingestion. More than three million people in the United States are affected by CD, and there is a 1–2% worldwide prevalence [1]. Though CD was once considered a disease of childhood, more recently it has been recognized that CD may develop at any age, and that a wide range of symptoms may be present at diagnosis. Many individuals with CD present with gastrointestinal symptoms, such as abdominal pain, bloating, diarrhea, and weight loss, while others develop more subtle, non-specific signs, such as anemia, headache, rash, or joint pain; others with CD remain asymptomatic [2]. While blood work (such as tissue transglutaminase IgA) is utilized for screening for CD, endoscopy with duodenal biopsy remains the gold standard for diagnosis in North America. At this time, a strict lifelong gluten-free diet remains the only treatment option [2].

Gluten is a composite of multiple proteins found in wheat and is made of up gliadin and glutenin proteins. Individuals with CD mount an immune response to gliadin as well as the proteins hordeins and secalins, which are present in barley and rye, respectively [3]. These proteins are rich in prolines and glutamines, which the human intestine can only partially digest [4,5]. The resulting incompletely digested peptide fragments from these proteins are transported from the small intestinal lumen via transcellular and paracellular pathways, and these fragments trigger responses from both the innate and adaptive immune systems [6]. Through the innate immune system, gliadin, hordein and secalin peptides induce the secretion of epithelial growth factor and IL-15, which leads to mucosal changes such as the proliferation of enterocytes and changes in vesicular trafficking [7].

Studies support a strong hereditary component to CD; it affects 10–20% of first degree relatives, and 75–80% of monozygotic twins but only 10% of dizygotic twins [8,9]. The adaptive immune response is shaped largely by genetics, specifically by the human leukocyte antigen (HLA) DQ haplotype (DQ2, DQ8), which is the strongest risk factor to develop CD. Individuals who are homozygous for HLA-DQ2 are considered to have high risk genetics for developing CD, as HLA-DQ2 homozygosity confers a 25–30% risk of developing early-onset CD [10–12]. The HLA-DQ2 and HLA-DQ8 structures contain positively charged pockets which bind to specific gluten peptides that have been deamidated by transglutaminase 2 [13]. The HLA molecules then present these peptides to CD4+ T-cells, which activate a downstream adaptive immune response [14,15]. Of note, more recent genome-wide association studies have identified >100 non-HLA-related genes that are associated with CD and likely confer some degree of risk to developing CD [16–18]. For instance, one study identified multiple genes involved in innate immunity and cell adhesion that were upregulated in individuals with CD [19].

The increase in intestinal permeability and the activation of the innate and adaptive immune systems by gluten is central to the pathophysiology of CD. An additional independent player in this system is zonulin. Zonulin is a family of endogenous proteins, which regulate the tight junctions in the small intestinal wall to allow the passage of larger particles [20]. Research shows that gluten activates zonulin in both healthy and non-healthy individuals, however, in patients with CD it triggers an abnormal increase in intestinal permeability [20,21]. This leads to additional exposure to dietary and environmental antigens, thus further activating the innate and adaptive immune systems.

Exposure to gluten and genetic predisposition are necessary but not sufficient to develop CD. While most diets contains a significant amount of gluten and 30% of the US population carries the necessary genetics, only 3% of these individuals ultimately develop CD [22]. Furthermore, CD can develop at any age, suggesting factors other than gluten and compatible genetics contribute to this loss of tolerance and increased intestinal permeability. The intestinal microbiota is one factor that has been implicated in the loss of tolerance to dietary gluten.

This review aims to summarize available data on associations between the gut microbiota and CD. The literature search was run using PubMed, EMBASE, and Web of Science using the following terms: “microbiome and CD”, “microbiota and CD”, “gut microbiome and CD”, “gut microbiota and CD”, “intestinal microbiome and CD”, and “intestinal microbiota and CD”. The search was limited to articles written in English.

2. Current understanding of the development of the human intestinal microbiota early in life

The intestinal microbiota is comprised of over 10¹⁴ microorganisms, including bacteria, viruses, fungi, protists and archaea [23]. Together, these microorganisms express over two million genes, significantly exceeding the number of genes expressed by the human genome (approximately 23,000 genes) [24,25]. The intestinal microbiota plays many vital roles, including digestion of insoluble fiber, vitamin production, and bile acid metabolism [26]. Additionally, the intestinal microbiota plays a critical role in the development of tolerance and the immune system. Recent research suggests that alterations in the intestinal microbiota likely contribute to many chronic immune disorders, such as obesity, food allergies, and inflammatory bowel disease, in addition to CD [27–29].

Though data is conflicting, some studies propose that the colonization of intestinal microbiota begins in utero, as bacteria have been isolated from meconium which is often passed before first nutritional intake after birth [30]. The infant microbiota is further shaped by the delivery method; multiple studies suggest that a vaginal birth leads to initial colonization with microbes from the birth canal (including higher prevalence of Bacteroides and bifidobacteria) compared to cesarean delivery, which can lead to colonization of the infant’s gut microbiota with skin microbiota (such as Staphylococcus) [31–33].

The infant’s intestinal microbiota’s composition and function is then further influenced by the source of nutrition over the first several months of life. In addition to nutrients for the infant, breast milk contains a unique microbiota and includes prebiotics known as human milk oligosaccharides (HMOs). HMOs promote the growth of bifidobacteria while preventing the growth of pathogenic organisms such as Clostridium difficile and Escherichia coli [34,35]. Formula feeding, conversely, seems to promote increased bacterial diversity, as well as an increased prevalence of C. difficile, E. coli, and Bacteroides fragilis [36,37].

Once solid foods are introduced into the diet (traditionally around 6 months of age), the infant intestinal microbiota begins shifting toward a more diverse system [38]. Another shift occurs when the infant transitions from breast milk or formula to mainly solid food, which typically occurs by 12 months of age [39]. The intestinal microbiome then stabilizes by approximately 2 years of age and represents what is considered a mature intestinal microbiome. The two major phyla of the mature intestinal microbiome are Firmicutes and Bacteroidetes, together comprising approximately 90% of the gut microbiota [40,41]. Multiple environmental exposures such as antibiotic use, genetics, probiotics, and nutrition can also alter the infant and the adult microbiome [42–45].

3. Mechanisms through which the gut microbiota contributes to celiac disease pathogenesis

Given that CD is a disease characterized by inflammation in the small intestine, it is reasonable to postulate that the local microenvironment, which is heavily influenced by the microbiota, plays a significant role in disease pathogenesis and loss of tolerance to dietary gluten. There are many proposed mechanisms through which the gut microbes are implicated in the pathogenesis of CD. Some bacteria express epitopes that mimic gliadin, and thus may potentially trigger a host immune response [46]. Lipopolysaccharide, a molecule expressed in the outer membrane of gram negative bacteria, may also trigger both the innate and adaptive immune systems through IL-15 production [7]. Other bacteria, such as Pseudomonas aeruginosa, may synergize with gluten to cause increased mucosal inflammation [47]. Gastrointestinal viral infections may activate the innate immune system through toll-like receptor 3 [48]. The microbiota may also modify the digestion of gluten, producing either immunogenic or tolerogenic gluten peptides, and thus influences antigen development. Additionally, the microbiome can directly affect intestinal permeability through zonulin release, and by promoting epithelial mucosal maturation. Intestinal microbes also modulate the immune system through the production of proinflammatory or anti-inflammatory peptides, metabolites and cytokines [49].”

4. Celiac disease and the microbiota: genetics

The HLA-DQ haplotypes remain one of the most important predictors of developing CD, and as a result significant research has focused on the impact of genetics on the gut microbiota in subjects genetically predisposed to CD [10]. Data from multiple prospective cohort studies suggests that HLA-DQ haplotype affects the composition of the early intestinal microbiota.

Olivares et al. noted a correlation between HLA-DQ genotype and differences in the early intestinal microbiota. Using 16S rRNA sequencing, they found that breastfed infants at genetic risk of CD had a decreased abundance of Bifidobacterium species present in their feces. Additionally, these at-risk infants had an increased abundance of Proteobacteria as well as an unclassified Enterobacteriaceae species [50]. Of note, a decrease in abundance of Bifidobacterium species has also been seen in other autoimmune conditions, such as rheumatoid arthritis, which may suggest a protective role against inflammation [51].

Other commensal intestinal microbes, Bacteroides species, have been linked to intestinal inflammation, including in inflammatory bowel disease [52]. In infants with a genetic risk of CD, Sanchez et al. specifically evaluated for Bacteroides species using PCR and denaturing gradient gel electrophoresis. They found that infants with a high genetic risk demonstrated a higher prevalence of B. vultagus, whereas low genetic risk infants exhibited a higher prevalence of B. ovatus, B. plebeius, and B. uniformis [53]. Similarly, in a longitudinal study utilizing FISH technology, De Palma et al. found a higher Bacteroides-Prevotella proportion in high risk infants, as well as a significantly higher abundance of Gram-negative bacteria, E. coli, Streptococcus-Lactococcus, E. rectale-C. coccoides, sulphate-reducing bacteria, C. lituseburense and C. histolyticum [54].

A number of prospective cohort studies have also investigated the dynamics of the gut microbiota in genetically susceptible infants. For example, a study using subjects from the PROFICEL cohort (a prospective longitudinal cohort study that followed infants at genetic risk of developing CD over a 5-year period in Spain) compared stool samples of infants with different feeding practices (breastfed or formula fed) at 7 days, 1 month, and 4 months old and reported Enterotoxigenic E. coli (ETEC) to be more prevalent in infants with high risk genetics [55].

In another prospective longitudinal cohort study, Sellitto et al. followed 34 breastfed infants at genetic risk for developing CD and collected stool at 8 time points over the first 2 years after birth. Utilizing 16S sequencing, this study found that subjects carrying high risk genetics had an increased prevalence of Firmicutes and Proteobacteria and a decreased overall prevalence of Actinobacteria and Bacteroidetes. Interestingly, it was also noted that the stool microbiota of these genetically at risk infants did not stabilize by 12 or 24 months old, contrary to what had been described in other studies in healthy infants [56].

CDGEMM (Celiac Disease, Genomic, Environmental, Microbiome and Metabolomic Study) is an ongoing large prospective longitudinal study that our research group is conducting, which aims to enroll 500 infants at birth [57]. This international (United States, Italy, and Spain) study seeks to investigate the role that the gut microbiota and metabolome play in the pathogenesis of CD, utilizing a multi-omics analysis to include environmental, genetic, and biologic data from participants to determine risk factors for the development of CD [57,58]. In a recent study [59], the CDGEMM investigative team utilized metagenomic sequencing to compare fecal samples from birth, 3 months, and 4–6 months for infants at standard and high genetic risk of CD. Infants with standard and high risk of CD also demonstrated increased abundances of Bacteroides and Enterococcus species, respectively. When compared with individuals with no genetic risk of CD, the microbiota of infants with a high and standard risk of CD were associated with a decreased abundance of Streptococcus, Coprococcus, Veillonella, Parabacteroides and C. perfringens [59]. This is in contrast to prior microbiome studies in other autoimmune conditions such as autoimmune hepatitis and Behcet’s disease, where Veillonella and Clostridium were found in higher abundance [60,61].

5. Celiac disease and the microbiota: perinatal factors

The onset of CD may be as early as 12–18 months of age, thus it is reasonable to suspect that the initial engraftment of the intestinal microbiome plays a pivotal role in loss of tolerance to dietary gluten [62]. Many factors, including birth mode, infant feeding type, and antibiotics have been implicated.

Association between birth mode and the risk of developing of CD remains controversial. One retrospective cohort study found an association between cesarean section and increased risk of CD, though this has not been consistently replicated [63]. Another retrospective cohort study found an association between acute but not elective cesarean section deliveries with CD, while a third prospective cohort study found only an association between elective cesarean sections and CD [64,65]. Furthermore, two large interventional double-blind placebo controlled prospective studies found that vaginal delivery was not protective against developing CD [10,12]. Until recently, there had been no studies specifically exploring the microbiome composition and CD in the context of delivery mode. For that reason, our group recently evaluated this utilizing the CDGEMM cohort [59]. We found that infants born via cesarean section demonstrated an increased abundance of Enterococcus faecalis and a decreased abundance of multiple species of Bacteroides (such as B. vulgatus and B. dorei) and Parabacteroides compared to those born vaginally [59]. Intriguingly, Parabacteroides has been shown in murine models to reduce the severity of intestinal inflammation otherwise induced by dextran sulfate sodium, thus suggesting a protective role against an immune response [66].

While initial data seemed to suggest that formula feeding was a risk factor for the development of CD [67], subsequent large double-blind, placebo controlled prospective studies found that breast feeding does not appear to significantly impact the development of CD [10,12,68]. In a subsequent study evaluating the microbiome in patients at risk of CD, breastfeeding has been found to support the growth of Bifidobacterium longum, Bifidobacterium breve, and C. leptum, which parallels the characteristic microbiota found in breastfed infants who are otherwise healthy [69]. Another study found that that formula-fed infants demonstrated an increased prevalence of C. perfringens and C. difficile in the stool [55]. Our prospective CDGEMM study found similar trends. However, in addition to infants fed exclusively on either breast milk or formula, we also evaluated infants who were fed both breastmilk and formula. Any amount of formula (whether exclusively or in combination with breastmilk) correlated with a decreased abundance of Bifidobacterium breve, while exclusively formula fed infants also demonstrated an increased abundance of Bifidobacteirum adolescentis, Ruminococcus gnavus and Lachnospiraceae [59]. Ruminococcus may promote activation of the immune system, as it has been linked to an increased incidence of allergic disease in infants [70]. Similarly, Lachnospiraceae may increase intestinal inflammation and contribute to diseases such as diabetes [71].

When evaluating the impact of antibiotics on the gut microbiota, the mother’s exposure to antibiotics during pregnancy and the infant’s exposure to antibiotics after delivery must both be considered. While there is some evidence in healthy mothers that antibiotic exposure during pregnancy influences the infant intestinal microbiota, there has not been comparable data in mothers with CD [36]. A cohort study performed by Marild et al. reported no statistically significant association between maternal antibiotic use during pregnancy and development of CD in the offspring [72]. Mixed data exists regarding antibiotic exposure early in life and subsequent risk of developing CD. A case-control study from Sweden suggested an association between antibiotic use and CD, as well as with degree of inflammation on duodenal biopsy and with positive celiac serologies [73]. Similarly, a study from Denmark suggested a dose-dependent relationship between the cumulative dose of antibiotics prescribed during the first year of life and the risk of later development of CD [74]. This parallels the results of a population-based birth cohort study documenting a dose-dependent increased risk of CD with antibiotic (especially cephalosporins) use over the first year of life [75]. However, multiple other studies found no association between antibiotic use and CD risk [76,77].

The impact of early antibiotic exposure on the microbiota was prospectively explored in the first 6 months after birth in patients at risk for CD [59]. In this study, antibiotic use at birth (administered to the mother during delivery or to the infant at birth) was associated with a decrease in the abundance of Bifidobacterium merycicum and Strepotococcus lutetiensis in infants exposed to antibiotics, as well as an increased abundance of Bacteroides thetaiotaomicron, Propionibacterium and Subdoligranulum species [59]. Fittingly, an increased abundance of B. thetaiotaomicron has previously been associated with antibiotic (especially amoxicillin) use [78]. Further studies exploring the impact of antibiotics on the specific intestinal microbiota are needed to further elucidate the alterations in the microbiota that may precede loss of tolerance and onset of CD.

6. Celiac disease and the microbiota: infectious exposures

In considering the gut microbiota, many studies have evaluated pathogenic viruses that cause acute infection and may disrupt the normal balance of the gut microbiota. Several studies have suggested that an increased number of gastrointestinal illnesses during the first 6–18 months of life is associated with an increased risk of CD later in life [76,79,80]. In theory, these infections lead to alterations in the gut microbiota, and contribute to increased intestinal permeability, which may lead to loss of tolerance and onset of CD autoimmunity. Many specific enteric infections such as rotavirus, enterovirus, adenovirus type 12, Orthoreovirus, and Candida albicans have been evaluated [81–85]. Mixed data exists on an association with rotaviruses, which primarily affect the small intestine and were the most common cause of diarrheal infections in infants and young children prior to introduction of a rotavirus vaccine in the early 2000s [86]. A study by Stene et al. found that frequent rotavirus infections were a predictor of a higher risk of CD autoimmunity [87]. However, a recent study reported a 1.5% increase in the prevalence of CD over the past 25 years despite the introduction of the rotavirus vaccine in the intervening time period, which argues against any increased risk [88]. Enteroviruses have also been linked to an increased risk of CD [81]. A prospective cohort study by Lindfors et al. found that between the ages of one and two years old, the cumulative number of enteroviral exposures increased the risk of CD autoimmunity [81]. Adenovirus type 12, which has amino acid sequence homology to a specific gliadin protein, has been shown to have a positive association with the later development of CD in multiple studies [82,83]. Also implicated is Orthoreovirus, a virus that causes both respiratory and gastrointestinal infections and is proposed to stimulate inappropriate immune activation, which subsequently leads to loss of tolerance to gliadin and increased intestinal inflammation and permeability [84]. Finally, C. albicans expresses a hyphal wall protein (hyphal wall protein 1) that is structurally similar to several gliadin epitopes, and has been proposed as a possible trigger for CD [85]. However, despite these associations, no studies evaluating the role of gastrointestinal infection on the microbiota in CD have been completed to date, though studies performed in otherwise healthy children with acute gastroenteritis suggest that these infections may lead to alterations in the microbiota, including alterations in microbial diversity [89,90].

7. Celiac disease and the intestinal microbiota

To date, most studies evaluating CD and the gut microbiota have been case-control, comparing individuals with active CD (on a gluten-containing diet) to those with CD in remission (on a gluten-free diet) or to healthy controls. The studies focus on the duodenal and fecal microbiota.

Many of the early cross-sectional studies were performed with samples obtained in Sweden between 1985–1995, when changes in recommended infant feeding practices (specifically, flour additives in infant formula and timing of gluten introduction with solid foods) led to an increase in annual CD incidence up to 4 times higher than the periods immediately before and after [91]. These studies utilized scanning electron microscopy to evaluate samples of patients with untreated, treated, and challenged CD with controls, and researchers found rod-shaped bacteria that were frequently associated with CD mucosa [92].

Alterations in the microbiota associated with CD are characterized by lower abundance of beneficial bacteria within the microbiome. Many Bifidobacterium species are considered to be beneficial for overall gut health, with demonstrated anti-inflammatory properties as well as inducing IL-10 production and promoting an appropriate Th1/Th2 balance [93,94]. Accordingly, numerous studies have focused on Bifidobacterium and have generally shown this species to be present in lower abundance in patients with untreated and treated (with a gluten-free diet) CD compared to non-CD controls [95–98]. One study found a reduced diversity of Bifidobacterium species overall despite a higher abundance of B. bifidum in adults with active CD when compared with healthy controls [99]. Similarly, Lactobacillus species are also considered beneficial to intestinal health, through multiple mechanisms including the secretion of anti-inflammatory cytokines and by modification of the Th1 immune response [100]. Research suggests that Lactobacillus species are less prevalent in individuals with both active and inactive CD when compared to healthy controls [96,101,102].

Many case-control studies have also identified shifts in bacteria that have been associated with other chronic medical conditions, such as inflammatory bowel disease [29]. While Bacteroides species make up a significant portion of the total normal microbiota, specific species (such as B. fragilis and B. vulgatus) are often associated with an increase in intestinal permeability and mucin degradation [103–105]. Not surprisingly, some Bacteroides species have been found in higher prevalence in patients with CD when compared to healthy controls [96–98,106]. One study specifically evaluated different Bacteroides species, and found higher levels of B. fragilis but lower levels of B. ovatus in patients with active CD when compared to healthy controls [107].

E. coli has also been associated with dysbiosis, by mechanisms such as increased intestinal permeability and mucin degradation [108,109]. Multiple studies evaluating both intestinal biopsy and fecal samples found higher concentrations of E. coli in patients with active CD and CD in remission when compared with healthy controls [96,97,106]. Sanchez et al. explored this idea further and found E. coli with higher virulence-gene carriage in fecal samples of children with both active CD and CD in remission [110].

While Staphylococcus species are a common component of many human microbiota sites(such as the intestinal and skin microbiota), studies show that with increased prevalence, the toxins and metabolites produced by Staphylococcus may tip the balance of normal intestinal homeostasis towards altered intestinal junctional integrity and barrier function [111]. Two different studies comparing active CD, CD in remission, and non-CD controls found an increased abundance of the Staphylococcaceae family (in one study, specifically Staphylococcus epidermidis and Staphylococcus pasteuri [112]) in duodenal biopsies and fecal samples of active CD patients compared with healthy controls [97,112]. These results were supported by other studies, which found an increase in virulence-gene carriage (methicillin-resistance gene, mecA) of Staphylococcus epidermidis in fecal samples of individuals with active CD [112,113].

Many additional distinctions between CD patients and controls have been noted, though are not observed across different studies. For instance, Clostridium histolyticum, Clostridium lituseburense, Faecalibacterium prausnitzii, the Streptococcaceae family (including Streptococcus anginosus and Streptococcus mutans), Dorea, Akkermansia, Firmicutes, and Actinobacteria have all been noted in lower abundance in individuals with CD [98,107,112,114,115]. Conversely, several other bacteria have been found in higher abundance in individuals with active CD compared to controls, including Clostridium leptum, Klebsiella oxytoca, Helicobacter, and the Neisseria family [97,112,114–116]. Similarly, the Proteobacteria phylum seems to exacerbate intestinal damage in the presence of gluten, which is potentially secondary to an altered intestinal mucus layer [117,118].

It should be noted that while many trends have been noted in these case-control studies, there have been multiple studies that reported no significant difference between the gut microbiota of subjects with CD and subjects without CD [119–124]. This includes one study by Rintala et al., which evaluated the fecal microbiota of 27 Finish children at high genetic risk of developing CD [125]. In their study, no significant difference was found in the fecal microbiota at 9 months and 12 months between controls and those children who developed CD before 4 years of age [125]. Similarly, mixed data exists surrounding certain bacteria such as Prevotella, where one study found higher abundance but another study revealed lower abundance in patients with CD [98,114]. Additional data and further studies are required to further elucidate what roles, if any, these bacteria play in the pathogenesis of CD.

While the aforementioned studies provide a crucial foundation for our understanding of the microbiota’s connection with CD pathogenesis, there are limitations inherent in the case-control studies. These studies can suggest associations, but cannot discriminate between cause and effect. This implies that any differences noted between CD patients and healthy controls may represent changes either secondary to active inflammation seen with CD, or might instead be the cause of the loss of tolerance and increased intestinal permeability preceding autoimmunity.

To address this, Garcia-Santisteban et al. recently presented a two-sample Mendelian randomization analysis (a method to improve causal inference in observational studies) with the microbiota as the exposure and CD as the outcome. Using this approach, they identified a number of associations between the intestinal microbiota, host genetics and single nucleotide polymorphisms (SNPs) associated with CD. Of note, all of the identified SNPs in this study were associated with the phyla Firmicutes and Proteobacteria, both of which had been shown in prior case-control studies to be linked to CD. Further studies, such as this one, will likely enrich our understanding of the interaction between the human host and the microbiota [126].

From a more fundamental point of view, prospective studies are necessary to answer many of the critical cause and effect questions. These studies will provide an understanding of shifts in the microbiota over the time both prior to and during the onset of CD; this information can then be used to address dysbiosis and thus prevent the onset of CD. The PROFICEL study is one notable prospective cohort study that followed infants at genetic risk of developing CD. One study using subjects from this cohort compared the fecal microbiota of these infants at 4 and 6 months old The researchers found that infants who did not develop CD later in the study timeframe (5 years) had increased bacterial diversity over time and also had higher levels of Firmicutes and B. longum. Conversely, those who later developed CD had higher levels of B. breve and Enterococcus species [127]. In another notable prospective study using subjects from the All Babies in Southeast Sweden (ABIS) cohort, Huang et al. compared 32 total children (including 16 subjects who progressed to CD and 16 healthy controls matched for HLA genotype and breastfeeding duration) with fecal samples obtained at 1, 2.5, and 5 years of age. They noted a higher alpha diversity in CD progressors at 1 year of age, as well as higher levels of Firmicutes in these patients. Interestingly, this group also compared the number of immunoglobulin A (IgA) coated bacteria between the two groups, as IgA plays a significant role in modulating the intestinal immune response. Children who later progressed to CD had a higher number of bacteria coated with IgA, which suggests a higher immune and possibly inflammatory response [128]. Case-control and prospective studies are summarized in Table 1.

Table 1:

Overview of Fecal and Duodenal Microbiota Studies in CD

Reference	Study Population	Study Design	Samples	Microbiota Profiling Method	Significant Findings
Sanz et al. [94]	Children (n = 20) Untreated CD Control	Case-control	Fecal	DGGE	Untreated CD: ↓ Bifidobacterium species diversity
Nadal et al. [95]	Children (n = 28) Untreated CD Treated CD Control	Case-control	Duodenal biopsy	FISH	Untreated CD: ↑ total bacteria and Gram negative bacteria Untreated CD: ↑ abundance Bacteroides and E. coli Untreated and treated CD: ↓ Lactobacillus–Bifidobacterium to Bacteroides–E. coli ratio
Sanchez et al. [109]	Children (n = 31) Untreated CD Treated CD Control	Case-control	Fecal	VRBD agar, API20E system, PCR	Untreated and treated CD: ↑ number of virulence genes and E. coli
Collado et al. [96]	Children (n = 78 fecal, n = 41 biopsy) Untreated CD Treated CD Control	Case-control	Fecal, duodenal biopsy	RT-PCR	Untreated and treated CD: ↓ Bifidobacterium levels (feces) Untreated and treated CD: ↑ levels of Bacteroides (feces, duodenal biopsies), and Clostridium leptum (feces, duodenal biopsies) Untreated CD: ↑ abundance of E. coli (feces, duodenum), and Staphylococcus (feces, duodenum)
Ou et al. [123]	Children (n = 63) Untreated CD Control	Case-control	Duodenal biopsy	16S rRNA	No statistically significant difference
De Palma et al. [53]	Children (n = 20) High genetic risk Intermediate and low genetic risk	Prospective	Fecal	FISH	High risk: ↑ abundance Bacteroides-Prevotella, Gram-negative bacteria, E. coli, Streptococcus-Lactococcus, E. rectale-C. coccoides, sulphate-reducing bacteria, C. lituseburense and C. histolyticum
Di Cagno et al. [101]	Children (n = 21) Untreated CD Treated CD Control	Case-control	Fecal	PCR-DGGE, 16S rRNA	Controls: ↑ percentage of Lactobacillus
De Palma et al. [97]	Children (n = 62) Untreated CD Treated CD Control	Case-control	Fecal	FISH	Untreated CD: ↓ Bifidobacterium, Clostridium histolyticum, C. lituseburense and Faecalibacterium prausnitzii Untreated CD: ↑ abundance of Bacteroides-Prevotella group
Schippa et al. [105]	Children (n = 30) Untreated CD Treated CD Control	Case-control	Duodenal biopsy	16S rDNA, TTGE	Untreated and treated CD: B. vulgatus and E. coli detected more often
Sanchez et al. [52]	Children (n = 75) High genetic risk Low genetic risk Breastfeeding Formula feeding	Prospective	Fecal	PCR, DGGE	High genetic risk: ↑ prevalence Bacteroides vulgatus Low risk: ↑ prevalence B. ovatus, B. plebeius, and B. uniformis Formula feeding: ↑ Bacteroides species diversity
Di Cagno et al. [175]	Children (n = 34) Treated CD Control	Case-control	Fecal, duodenal biopsy	DGGE	Control: ↑ levels of Lactobacillus, Enterococcus, and Bifidobacteria (fecal)
Nistal et al. [98]	Adults (n = 32) Untreated CD Treated CD Control	Case-control	Fecal	DGGE	Treated CD: ↓ diversity of Lactobacillus and Bifidobacterium Untreated CD: ↑ presence of Bifidobacteirum bifidum
Sanchez et al. [112]	Children (n = 60) Untreated CD Treated CD Control	Case-control	Fecal	PCR ABI PRISM-3130XL Gene Analizer	Untreated and treated CD: ↑ Staphylococcus epidermidis carrying the mecA gene
Sanchez et al. [106]	Children (n = 58) Untreated CD Treated CD Control	Case-control	Fecal	Schaedlr agar, 16S rRNA	Untreated and treated CD: ↑ levels of Bacteroides fragiles but ↓ levels of Bacteroides ovatus
Nistal et al. [176]	Children (n = 13) Untreated CD Control Adults (n = 15) Untreated CD Treated CD Control	Case-control	Duodenal biopsy	16S rRNA	Bacterial richness ↓ in children versus adults Untreated CD (adults, children): ↓ Prevotella and Streptococcus
Kalliomaki et al. [118]	Children (n = 19) Untreated CD Control Adults (n = 6) Treated CD	Case-control	Duodenal biopsy	qPCR	No significant differences
Sellitto et al. [55]	Children (n = 34) High genetic risk Control (non-genetic risk)	Prospective	Fecal	16S rRNA, qPCR	High risk: ↑ prevalence of Firmicutes and Proteobacteria, with overall ↓ prevalence of Actinobacteria and Bacteroidetes High risk: microbiome did not stabilize by 2 years of age
De Palma et al. [68]	Children (n = 164) Breastfeeding Formula feeding	Prospective	Fecal	qPCR	Breastfeeding: supports the growth of Bifidobacterium longum, Bifidobacterium breve, and Clostridium leptum
Sanchez et al. [111]	Children (n = 57) Untreated CD Treated CD Control	Case-control	Duodenal biopsy	16S rRNA	Active CD: ↑ abundant Proteobacteria, Klebsiella oxytoca, Staphylococcus epidermidis, Staphylococcis pasteuri Active CD: ↓ abundant Streptococcaceae family Control: ↑ abundance S. anginosus and S. mutans
Cheng et al. [119]	Children (n = 19) Untreated CD Control	Case-control	Duodenal biopsy	qRT-PCR, HITChip	No significant differences
Wacklin et al. [177]	Adults (n = 51) Untreated CD with GI symptoms, dermatitis herpetiformis, or anemia Controls	Case-control	Duodenal biopsy	PCR-DGGE	GI symptoms or anemia: ↓ microbial diversity and ↑ Proteobacteria (compared to DH and control)
de Meij et al. [120]	Children (n = 42) Untreated CD Control	Case-control	Duodenal biopsy	16S-23S ISPRO PCR	No significant differences
Olivares et al. [49]	Children (n = 22) High genetic risk Low genetic risk Breastfeeding Formula feeding	Case-control	Fecal	16S rRNA, RT-PCR	High risk: ↑ Proteobacteria and unclassified Enterobacteriaceae with ↓ in Bifidobacterium Formula fed: ↑ prevalence of Clostridium perfringens and C. difficile
Nistal et al. [121]	Adults (n = 18) Untreated CD Control	Case-control	Duodenal biopsy	16S rRNA	No significant differences
D’Argenio et al. [114]	Adults (n = 41)	Case-control	Duodenal biopsy	16S rRNA	Untreated CD: ↑ abundance Proteobacteria and ↓ abundance Firmicutes and Actinobacteria Untreated and treated CD: ↑ abundance Neisseria genus (esp. Neisseria flavescens)
Olivares et al. [126]	Children (n = 20) Developed CD later Control	Prospective	Fecal	16S rRNA	Controls: ↑ Firmicutes families with time (not seen in CD); Controls: ↑ abundance Bifidobacterium longum Developed CD: ↑ proportions of Bifidobacterium breve and Enterococcus spp.
Olivares et al. [54]	Children (n = 127) High genetic risk Intermediate or low genetic risk	Prospective	Fecal	16S rRNA	High risk: ↑ enterotoxigenic E. coli (ETEC)
Rintala et al. [124]	Children (n = 27) Developed CD later Control (Both groups at high genetic risk)	Prospective	Fecal	16S rRNA	No significant difference
Bodkhe et al. [113]	Adults (n = 62) Untreated CD Healthy first degree relatives Control	Case-control	Fecal, duodenal biopsy	16S rRNA	Untreated CD: ↑ abundance amplicon sequence variant Megasphaera and Helicobacter (duodenal biopsies)
Huang et al. [127]	Children (n = 32) Developed CD later Control	Prospective	Fecal	16S rRNA	Developed CD: ↑ alpha diversity at 1 year of age, ↑ levels of Firmicutes, ↑ IgA coated bacteria
Leonard et al. [58]	Children (n = 31) High genetic risk Standard genetic risk No genetic risk Vaginal delivery Cesarean section (CS) delivery Breastfeeding Formula feeding	Prospective	Fecal	Meta-genomic sequencing	High and standard genetic risk: ↓ abundance Streptococcus, Coprococcus, Veillonella, Parabacteroides and Clostridium perfringens CS: ↓ abundance Bacteroides (such as B. vulgatus and B. dorei) and Parabacteroides Formula (exclusive or with BM): ↓ abundance of Bifidobacterium breve Exclusively formula: ↑ abundance of Bifidobacteirum adolescentis, Ruminococcus gnavus and Lachnospiraceae

DGGE = denaturing gradient gel electrophoresis; FISH = fluorescence in situ hybridization; VRBD = violet red bile dextrose; qPCR = quantitative polymerase chain reaction; TTGE = temporal temperature gradient gel electrophoresis

Additional prospective, longitudinal studies are necessary to understand the shifts in the intestinal microbiome prior to the onset of CD.

8. Celiac disease and non-intestinal microbiota: oropharyngeal, breast milk, and blood

It is worth noting that the majority of research to date focuses on the intestinal microbiota in CD evaluating either the mucosa or stool. However, there are other distinct microbiomes within the human body that must be considered, including the oropharynx, breast milk, and blood microbiota.

The oropharyngeal microbiota is perhaps the most studied microbiota in CD outside of duodenal and fecal samples. In a comparison between patients with CD in remission, refractory CD, healthy controls, and individuals with functional gastrointestinal complaints, one study found that the oral microbiota of both groups with CD showed significant differences in the microbiota, including higher salivary lactobacilli. Further, this study found increased levels of salivary gluten-degrading enzymes in patients with CD, which theoretically could explain differences in tolerance to dietary gluten between healthy controls and CD patients [129]. Two other studies found increased Neisseria species prevalence in active CD, with an associated increase in proinflammatory cytokines [130,131]. See Table 2 for summary of alterations in the oropharyngeal microbiota in untreated CD when compared to the stool and duodenal microbiota.

Table 2:

Alterations in the Microbiota in Untreated Celiac Disease

Sample Type	Increased in Untreated CD				Decreased in Untreated CD
	Phylum		Species		Phylum	Species
Stool	Bacteroidetes Proteobacteria Firmicutes Actinobacteria		Bacteroides spp. Bacteroides-Prevotella group Bacteroides fragilis E. coli Clostridium leptum Staphylococcus spp. Bifidobacterium bifidum	[96] [97] [106] [96] [96] [96] [98]	Bacteroidetes Firmicutes Actinobacteria	Bacteroides ovatus Faecalibacterium prausnitzii Clostridium histolyticum Clostridium lituseburense Bifidobacterium spp.	[106] [97] [97] [97] [96,97]
Duodenal Biopsy	Bacteroidetes Proteobacteria Firmicutes		Bacteroides spp. Bacteroides vulgatus Neisseiria spp. N. flavescens E. coli Klebsiella oxytoca Clostridium leptum Staphylococcus spp. Staphylococcus epidermidis Staphylococcus pasteuri	[95] [105] [114] [114] [95,96,105] [111] [96] [96] [111] [111]	Bacteroidetes Firmicutes	Prevotella Streptococcus	[175] [175]
Oropharyngeal	Proteobacteria	[130]	Neisseria spp.	[129,130]

Although the gastrointestinal tract represents the largest and most diverse microbiota in the human body, breast milk has also been shown to have a unique though less concentrated microbiota. As discussed earlier, breast feeding alone has not been shown to protect against the development of CD [10,12]. However, breast milk does play a critical role in shaping the infant immune system, by providing an incredible array of antibodies, microbes, metabolites, and prebiotics (human milk oligosaccharides (HMOs)) [132,133]. The first study to look at the composition of breast milk in the context of CD compared 12 heathy mothers with 12 mothers with CD [134]. This study utilized PCR-based techniques and identified a decreased abundance of Bifidobafcterium species in CD mothers, and researchers found no significant difference in HMOs [134]. A recent study from the PreventCD cohort compared the breast milk microbiota composition for mother’s whose children did and did not develop CD. They found increased abundance of Methylobacterium komagatae, Methylocapsa palsarum, and Bacteroides vulgatus in the breast milk of mother’s whose children who later developed CD [135]. Further studies are needed to better understand the interplay between the microbiota and metabolome, as well as the role of HMOs in CD pathogenesis.

Finally, recent data suggest that human blood, once considered a sterile fluid, instead contains a unique microbiota of non-pathogenic bacteria that are difficult to culture [136,137]. As one of the proposed sources of the blood microbiota is through damaged intestinal epithelium, it is interesting to consider a link to CD, a disease characterized by increased intestinal permeability [137]. Serena et al. compared the blood and fecal microbiomes of active CD, CD in remission, and healthy patients [138]. Though there were no statistically significant differences between the three groups, they did note a trend that patients with CD seemed to have a decrease in hematic Bifidobacterium and Sphingomonas with an increase in Bacillales [138].

9. Celiac disease and the intestinal metabolome

When studying the gut microbiota, it is imperative to also evaluate the physiological state of the intestinal environment. This is addressed by the metabolite profiling of fecal samples as a proxy for the gut (the “gut metabolome”); analysis involves identifying the diverse set of small molecules present in that sample, such as amino acids, sugars, vitamins, and antibiotics. Metabolomic analysis of fecal samples thus reveals the functional activity of the gut microbiota. The most studied microbially-derived metabolites in the gut are short chain fatty acids (SCFAs), which are synthesized by commensal bacteria as end products of complex carbohydrate fermentation that are otherwise indigestible by host human enzymes. In addition to serving as the main energy source for human colonocytes, SCFAs are also implicated in several immunomodulatory functions. For example, they help condition regulatory T-cells (Tregs), which are important for immune system modulation and tolerance of self-antigens [139]. Multiple studies have identified differences in SCFA patterns between patients with untreated and treated CD and healthy controls, including SCFAs such as butyrate [140,141]. SCFAs are produced by microbes in both the large and small intestine [142]. Butyrate plays a large role in helping T-cells differentiate into Tregs, and it has been the focus of metabolomic studies in CD [143]. A study evaluating intestinal organoids developed from duodenal biopsies from CD and non-CD patients found that organoids from CD patients showed increased intestinal permeability and increased secretion of proinflammatory cytokines when exposed to gliadin [18]. However, when treated with microbial by-products such as butyrate, lactate, and polysaccharide A, there was reduced intestinal permeability as well as decreased cytokine production in the CD patient organoids [18]. This suggests that SCFAs may play a role in modulating intestinal response to gliadin [18]. Similarly, research suggests that patients with CD have an imbalance between FoxP3 (a protein that plays a fundamental role in Treg cell differentiation) isoforms, favoring a non-functional spliced form [144]. This imbalance has been linked to imbalance of butyrate and lactate levels within the intestinal environment [144].

Our prospective CDGEMM cohort has also revealed additional patterns of fecal metabolome alterations in infants at risk of developing CD beyond SCFAs [59]. We found increased levels of butanoic acid in infants with a high genetic risk, increases in multiple metabolites (including butanoic acid, glycolic acid, oxalic acid, and hydroxyphenlacetic acid) in infants born via cesarean section, increases in sucrose and threonine in exclusively formula fed infants, and a decrease in sucrose in infants exposed to antibiotics [59]. Another recent study compared the fecal microbiota and metabolome of untreated and treated CD, and reported changes in the metabolome that were consistent with changes in dietary (macronutrient) intake [145]. These results provide unique insights into the continuously shifting gut metabolome and will provide crucial information for our understanding of the changes that occur prior to CD onset. However, further research is needed to mechanistically link gut microbes and metabolites. For example, while polysaccharide A is known to be involved in decreased intestinal permeability, some studies D report a higher abundance of B. fragilis, which is paradoxically a polysaccharide A producer, in individuals with CD [18,107].

In addition to the fecal metabolome, other researchers have studied alterations in the blood metabolome of celiac patients compared to non-celiac controls. For example, Huang et al. evaluated plasma metabolites in 10 children who progressed to CD (3 who progressed to CD prior to the time point the plasma was collected) and 9 controls [128]. Out of 387 metabolites studied, 19 were significantly different between the two groups. The most altered metabolites that were increased in those children who developed CD include TDCA (a proinflammatory conjugated bile acid), glucono-D-lactone, and isobutyryl-L-carnitine [128].

10. Pursuit of microbial candidate for disease immunomodulation

Further understanding of the role of the gut microbiota in CD pathogenesis will ultimately lead to novel approaches for prevention of CD through the early correction of detrimental alterations in the microbiota, prior to the onset of increased intestinal permeability. In the meantime, many studies have focused on treating active CD through the manipulation of the gut microbiota, either by using probiotics to alter the microbial balance or by harnessing naturally occurring gluten-degrading enzymes [146–167].

10.1. Probiotics

Probiotics are defined by the World Health Organization as “live microorganisms which when administered in adequate amounts confer a health benefit on the host” [168]. Few pathologies have shown benefits from probiotic administration, including decreasing antibiotic-associated diarrhea as well as inflammation from pouchitis in inflammatory bowel disease [169–171]. Multiple studies have been conducted to examine the effects of probiotic administration in CD [146–158].

There are a number of preclinical and animal studies of probiotics in CD. Multiple studies have focused on bifidobacteria, which appear to be beneficial in CD. In one study utilizing a murine model of CD, researchers observed a decrease in proinflammatory cytokines such as TNF-alpha with the administration of B. longum CEC 7347 [147]. They also observed a subsequently lower level of damage to jejunal epithelium in these animal models [147]. A separate preclinical study, evaluating this same strain of Bifidobacterium on monocyte-derived dendritic cells, showed a decrease in IFN-production with stimulation from gliadin, as well as an increase in IL-10 release [148]. In another preclinical study, B. longum NCC2705 modulated the immune response through the production of a serine protease inhibitor, which ultimately attenuated the histologic damage with gliadin-administration in NOD/DQ8 mice [149]. B. infantis has shown beneficial effects in CD and has been noted in a preclinical study to decrease Paneth cells based on electronic microscopy of duodenal biopsies of CD patients [153]. Paneth cells play a large role in the innate immune system in response to noxious stimuli, such as gliadin [153,172]. Additional preclinical studies have investigated the role for Lactobacillus as a probiotic in CD. L. casei (in one study, specifically L. casei DN-114001) appears beneficial in CD in several preclinical studies, via mechanisms such as reducing TNF-alpha secretion and decreasing villous blunting [154–156]. Additionally, one preclinical study found that the administration of Lactobacillus species degraded gluten peptides and ultimately reduced immunogenicity [157].

A limited number of probiotics have been or are being tested in clinical studies. One clinical study found that B. longum ES1 and B. bifidum ES2 downregulated proinflammatory changes from the Th1 pathway typically activated in CD [146]. In two other clinical studies, two different strains of B. breve (BR03, B632) also appear beneficial in CD as data suggest that they lead to a decrease in proinflammatory cytokines (such as TNF-alpha), as well as positive alterations in the SCFA profiles [150,151]. Another clinical study found that despite administration of these B. breve strains, there was no noted corresponding rise in fecal levels of Bifidobacterium; however, there was evidence of modulation of the intestinal microbiota, characterized by reduction in the Firmicutes-Bacteroides ratio [152]. In one randomized double-blind placebo-controlled study, two lactobacillus strains (L. plantarum HEAL9 and L. paracasei 8700:2) were administered for six months to at children at risk of developing CD; when compared to children receiving placebo, these children were found to have alterations in their lymphocytes, though there was no significant difference in their tissue transglutaminase levels [173]. Finally, a blend of probiotic bacteria including Streptococcus thermophilus, B. breve, B. longum, B. infantis, L. acidophilus, L. plantarum, L. paracasei, and L. delbrueckii subsp. bulgaricus, which has demonstrated efficacy in other gastrointestinal inflammatory conditions such as ulcerative colitis, has not shown any changes in the fecal microbiome in a clinical study in patients with CD [158,174]. Interestingly, a recent systematic review and meta-analysis examining probiotic use by patients with CD found that ingestion of probiotics was associated with an improvement in GI symptoms [175]. One prospective cohort study found that, in general, any probiotic use (regardless of specific formulation) during the first year of life was not associated with CD or CD autoimmunity [176].

Further information is likely forthcoming, as there are multiple clinical trials are currently underway investigating the role of probiotics both in the prevention and treatment of CD.

10.2. Gluten-degrading enzymes

As noted above, gluten, which is comprised of gliadins and glutenins, is incompletely digested by the human gastrointestinal system since the major digestive enzymes (such as trypsin, pepsin, elastase, and chemotrypsin) are unable to cleave the proline-rich residues within the molecules [4,5]. However, multiple microorganisms have demonstrated promising glutenase capabilities in vitro, through the hydrolysis of these proline- and glutamine-rich residues [177–179].

One study globally evaluated the functional capabilities of duodenal biopsies and found that of the 114 bacterial strains isolated, 85 (75%) grew on media with gluten as the sole nitrogen source, 31 (27%) had extracellular activity against gluten, and 27 (24%) strains had the ability to breakdown a specific immunogenic peptide in CD (the 33-mer peptide) [159]. Similarly, another study isolated 144 bacterial strains from fecal samples, and the researchers found that 94 (65%) demonstrated some ability to break down gluten, 61 (42%) had extracellular activity against gluten, and 11 (7.6%) strains displayed peptidase activity against the 33-mer peptide [160]. In another effort to identify specific gluten-degrading bacteria, Fernandez-Feo et al. investigated bacteria normally found in the human oropharynx. They found multiple bacteria, including Rhotia spp., Actinomyces odontolyticus, Neisseria mucosa, and Capnocytophaga sputigena, which displayed the ability to degrade the toxic epitopes of gliadin associated with CD [161].

It is worth noting that while bacteria may degrade gliadin, unless the specific immunogenic peptides are cleaved, the peptides can still cross the intestinal mucosal barrier and trigger an immunogenic response. In one study, Caminero et al. colonized germ-free mice with microbes isolated from duodenal samples isolated from individuals with CD and healthy controls [157]. They found that Pseudomonas aeruginosa (isolated from CD patients in a prior study from the group [122]) cleaved gluten and produced peptides that were able to more easily translocate across the intestinal mucosa and trigger a T-cell response. Conversely, Lactobacillus species (found more often in controls) could cleaved gluten peptides and the peptides produced by P. aeruginosa thereby decreasing the immunologic response and thus suggesting a protective role [157]. A later study by this group also found that amylase trypsin inhibitors (which are present in gluten-containing grains) may increase the inflammatory response secondary to gluten exposure, though Lactobacilli may degraded the amylase trypsin inhibitors and modulate this response [162]. Other in vitro studies also suggest that Bifidobacterium may degrade immunogenic peptides in gluten, thus modulating the inflammatory response in CD [163,164]. Similarly, in a murine model, Saccharomyces boulardii KK1 demonstrated the ability to hydrolyze gliadin, with a subsequent decrease in local cytokine production [165].

Gluten-degrading enzymes are an attractive pharmacologic target, as they would allow individuals with CD to consume gluten without risking exposure to the gliadin epitopes that trigger the immune response. While there are many commercially available glutenase products currently on the market, these agents have not successfully gone through the drug development pathway. One trial found no significant difference in patients receiving the gluten degrading enzyme latiglutinase when compared with placebo in evaluating improved symptoms and reducing villous atrophy; however, in a subgroup of patients with worse symptoms and seropositivity, they did find a decrease in symptoms with latiglutinase administration [166,180]. Further clinical studies with latiglutuinase are currently underway. Another peptidase under investigation, Kuma030, has shown degradation of gliadin and subsequent decreased activation of T cells and the immune system in vitro [167]. Multiple clinical trials for Kuma030 (NCT03409796) and other peptidases are currently underway. It should be pointed out that the kinetics of many of these enzymes may not prevent the activation of the innate immune system (that occurs in matter of minutes) once exposed to gluten.

11. Conclusions

Research available to date strongly favors a relationship between the gut microbiota and CD, with many alterations that may promote increased inflammation and intestinal permeability. Most studies are case-control and though they present thought-provoking associations, they are unable to comment on whether differences seen are secondary to or the cause of CD. Further research is needed to explore the role of specific environmental factors, such as infectious exposures, nutrition and antibiotics, on alterations to the microbiome in individuals at genetic risk of developing CD, with subsequent monitoring to determine which individuals ultimately progress to CD. Prospective cohort studies will play an essential role in investigating the state of the intestinal microbiota immediately preceding CD onset, and towards identifying any causal relationship between the microbiota and metabolites and the pathogenesis of CD. This data is essential for a “multi-omics”, individualized approach to CD, in which genetics, dietary exposures, and environmental factors and their effect on the gut microbiota can be evaluated and ultimately manipulated to prevent the onset of CD.

12. Expert opinion

CD provides us a unique opportunity to link the microbiome composition and function to the pathogenesis of a variety of noninfective chronic inflammatory diseases. In the past, genetic predisposition and exposure to environmental triggers were considered necessary and sufficient to develop chronic inflammatory diseases, including CD. However, several epidemiological studies, including prospective longitudinal cohort studies suggest that, despite the fact that the introduction of the environmental trigger, gluten, causing CD is easily determined, the initial exposure does not coincide with the onset of the disease as onset can occur years after gluten exposure. This observation suggests that other factors are at play in causing the break of tolerance of gluten and onset of CD. Given the growing evidence of the potential epigenetic role of the gut microbiota in changing host functions, particularly immune response, many investigators have been focusing on studying the role of the gut microbiota in contributing to the pathogenesis of CD. Early studies, including ours, were limited by their design (cross-sectional, case-control design) and by technical limitations that only allowed one to study microbiota composition at high hierarchy levels (genus or phylum) without gaining any insight into its capability to affect the host immune response. More recently, the shift from 16S sequencing to shotgun metagenomic analysis of the gut microbiota allowed researchers to gain more granular details from the microbiota composition at the strain level and, most importantly, of its potential function in influencing host key metabolic pathways. However, the shortfall of cross-sectional studies still limits our capability to mechanistically link these studies to diseases pathogenesis. Thus, despite state-of-the-art studies examining the link between CD and the microbiota, it remains difficult to interpret mechanistic connections and challenging to reconcile differences in specific components of the microbiota. Technical differences and the capability to detect microbiota components at the strain level and most importantly availability of robust metadata are all key elements to be taken into consideration for proper interpretation of data published thus far. Furthermore, case-control studies suffer from the limitation of identifying associations rather than causation, thus providing assumptions related to the role of the microbiota in CD pathogenesis. Therefore, the most logical approach to move from association to causation is to design cohort surveillance allowing for the surveillance of the microbiota composition and function before, during, and after the onset of the disease. Some chronic inflammatory diseases are not amenable to this kind of study design, given that their onset is either open ended (for example type 1 diabetes) or they occur later in life (like multiple sclerosis or rheumatoid arthritis) thus requiring infeasibly long prospective surveillance. Conversely, conditions that occur mostly early in life would be more amenable to birth cohort surveillance. As shown by several birth cohort studies, the vast majority of the expected cases of CD occurs within the first 3–4 years of life [10,12]. Therefore, it is obvious that longitudinal studies from birth are feasible for a disease like CD where disease onset occurs early in life, and that these studies offer the best approach to establish a potential pathogenetic role of the gut microbiota in CD onset. Current availability of a variety of technologies set the stage for an ideal study design in which robust metadata, extensive biobanking, proper handling of specimens, and integration with other “-omic” analyses (including metabolomic, genomic, and proteomic analyses) will provide the ideal roadmap to capitalize on microbiota studies. The CD-GEMM represents a model not only to study the microbiota in the context of this multi-omic approach, but also to provide the proper tools for disease pathogenesis modeling by using artificial intelligence and machine learning. Indeed, integrating our current knowledge of CD pathogenesis, ongoing birth cohort studies, extensive metadata collection and state-of-the-art machine learning approaches will lead to disease modeling aimed at disease interception by offering novel potential targets for microbiota manipulation, thus allowing for CD primary prevention.

Article highlights:

• Celiac disease (CD) is a lifelong autoimmune condition in genetically susceptible individuals caused by a sensitivity to dietary gluten. Disease onset can occur at any age.

• The gut microbiome likely plays an important role in CD onset.

• The oropharyngeal, breast milk and blood microbiota are distinct from the gut microbiota, and also likely play important roles in CD onset.

• Genetics and environmental factors, such as antibiotic use and nutrition, heavily influence the composition of the gut microbiome.

• Many tools aimed at manipulating the microbiota, including probiotics and gluten-degrading enzymes, are currently being investigated in the context of CD.

• A multi-omics approach and longitudinal studies are necessary to further understand the role of the gut microbiota in CD pathogenesis.