Fecal and Duodenal Microbiota in Pediatric Celiac Disease

Background and Objective: The gut microbiota plays a role in regulating the host immunity. Therefore, alterations in gut microbiota (or dysbiosis) have been investigated in several gastrointestinal diseases, including Celiac Disease (CD). The aim of this study is to summarize the main characteristics of the gut microbiota in pediatric CD. Methods: We performed a systematic review to retrieve the available studies investigating the gut microbiota in pediatric CD patients and controls. In detail, after the screening of >2,200 titles from the medical literature, 397 articles were assessed for eligibility based on the abstracts: of those, 114 full-text original articles were considered as eligible according to the aim of this systematic review. Results: The final search output consisted of 18 articles describing the gut microbiota of CD children and including one or more control groups. Eleven pediatric studies provided information on the duodenal microbiota and as many investigated the fecal microbiota; three articles analyzed the microbiota on both fecal and duodenal samples from the same cohorts of patients. Conclusion: Due to the heterogeneity of the experimental procedures and study design, it is not possible to evidence any specific celiac signature in the fecal and/or duodenal microbiota of CD children. However, some specific components of the fecal microbiota and, in detail, Bifidobacterium spp. (e.g., Bifidobacterium longum) may deserve additional research efforts, in order to understand their potential value as both probiotic therapy and diagnostic/prognostic biomarker.


INTRODUCTION
The microbial communities naturally colonizing the gastrointestinal (GI) tract, namely, gut microbiota, are an essential and physiological component of a healthy human body. In fact, the "whole" deriving from this symbiotic interaction between microbiota and host, has been defined as a "superorganism, " where the former contributes to several metabolic, physiologic, inflammatory, and immunologic functions for the latter (1,2).
In detail, the microbiota plays a role in regulating the host immunity by influencing the development and homeostasis of the gut epithelial layer and mucosal-associated lymphoid tissue. Therefore, alterations in gut microbiota (or dysbiosis) have been explored and investigated in several GI disorders, including Celiac Disease (CD).
Indeed, CD is a systemic immune-mediated condition, characterized by a variable pattern of GI and extra-GI clinical manifestations, but clearly defined by the presence of gluten-dependent atrophic (small bowel) enteropathy associated with a consistent serological panel (positivity for anti-tissue transglutaminase antibody and/or anti-endomysium antibody) (3,4). Importantly, gluten can trigger CD in a minority of patients who are carriers of specific HLA-DQ genotypes (HLA-DQ2 and HLA-DQ8): indeed, both these environmental and genetic factors respectively represent necessary, but not sufficient, conditions to develop CD (5).
Each GI tract is characterized by a peculiar microbiota, in terms of qualitative and quantitative composition. In general, the bacterial density progressively increases along the GI tract, ranging from 10 3 units per gram of luminal content in the duodenum up to values of 10 12 order in the colon; concomitantly, the bacterial diversity (in terms of number of different bacterial species) gradually increases from the proximal to the distal GI tracts as well (2,6).
The knowledge on gut microbiota has greatly improved in the last few years, since culture-independent approaches became available, such as the analysis of bacterial 16S ribosomal RNA (rRNA) gene. Among bacteria (indeed, also yeast and viruses are part of gut microbiota), the dominant phyla are Firmicutes and Bacteroidetes: the former phylum mainly includes Clostridium spp., Bacillus spp., Lactobacillus spp., Ruminococcus spp., and Enterococcus spp.; Bacteroidetes phylum is predominantly composed of Bacteroides spp. and Prevotella spp. The remaining phyla (Actinobacteria, Fusobacteria, Proteobacteria, and Verrucomicrobia) represent around 10% of the bacterial gut population; among them, the most studied genus is Bifidobacterium spp., which belongs to Actinobacteria phylum (7)(8)(9).
The inter-individual variability of gut microbiota is high: indeed, its composition is influenced by a multitude of factors, such as genetics, age, diet, hygiene level, and medication exposure (first of all, antibiotics). As mentioned above, deviations in the composition of gut microbiota may also be related to specific pathological conditions, but it is still unclear if these are cause or effect of the comorbid disease, due to the limitations of the available studies so far (10,11).
In this systematic review, we summarized the main characteristics of the gut microbiota in pediatric CD patients.

Protocol
The PRISMA guidelines were used for this systematic review (12). This systematic review includes original articles (such as experimental studies, randomized control trials, case-control studies, cohort studies, and observational studies), providing information on gut microbiota in CD children and age-matched controls. Indeed, the primary endpoint was the description of gut microbiota in pediatric patients affected with CD. This systematic search did not include review and abstract papers.

Search Strategy
The literature search strategy of this review consisted of two stages: (i) an extensive search in five databases, namely, Scopus, Cochrane Central Register of Controlled Trials, PubMed, Ovid, and Web of Science, by using relevant keywords, as described in Table 1; (ii) a search of reference lists from the collected and related articles identified at the previous stage.

Data Extraction
After a critical reading of the articles, data extraction was done by one investigator and then checked by a second investigator following these inclusion criteria: any original articles in which gut microbiota composition was studied in CD patients with clear description of methods and outcomes. In detail, the following items were extracted from each study: first author's last name, publication year, country of origin, study population details, age group, sample type, aims of the study, intervention type, number of participants, and analytical methods to study microbiota.

RESULTS
The results of each step of the literature search according to the PRISMA guidelines are schematically summarized in Figure 1.
In detail, the literature search resulted in 1,143 papers from Scopus, 44 from the Cochrane Central Register of Controlled Trials, 337 from PubMed, 82 from Ovid, and 509 from Web of Science. Eighty-two additional articles were identified through the references of the previous papers. After the screening by title, the duplicated records were removed, and 397 articles were assessed for eligibility based on the abstracts: of those, 114 full-text original articles were considered as eligible according to the aim of this systematic review. The final search output consisted of 18 articles describing the gut microbiota of CD children and including one or more control groups (13)(14)(15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30). All the articles finally included in this systematic review are listed in Table 2, where the study population, participants' age group, sample type(s), aim of the study, type of intervention, sample size, and detailed laboratory methods are summarized. Importantly, studies focused on HLA-DQ predisposition, which analyzed the gut microbiota in children before becoming celiac (thus, with no microbiome analysis in CD pediatric patients after the diagnosis), were excluded. Moreover, only studies including a control group to be compared to CD children, were included. The specific findings resulting from these selected studies are summarized in Tables 3 and 4, which refer to the gut microbiota studies on duodenal biopsy and stool samples, respectively. In detail, these tables provide the main findings in terms of overall bacterial abundance and/or bacterial diversity and/or specific bacterial composition, according to the variable aims of each study.

DISCUSSION
In the landscape of immune-mediated non-communicable disorders, CD is the only disease for which the necessary HLA genetic background and environmental trigger are known. Briefly, the dietary intake of gluten triggers the development of CD in some of those individuals who are carriers of the specific HLA-DQ allelic variants coding DQ2 and DQ8 heterodimers (DQA1 * 0501-DQB1 * 02 and DQA1 * 0301-DQB1 * 0302, respectively) (4,5). Overall, 30-40% of the general population in Europe and North America carry this HLA-DQ

Database
Search strategy SCOPUS (TITLE-ABS-KEY ("celiac disease" OR "coeliac") AND TITLE-ABS-KEY ("gut microbiota" OR "bacteria" OR "microbes" OR "microbiome" OR "microbiota dysbiosis" OR "metagenome" OR "metabolomics" OR "fecal microbiota" OR "intestinal microbiota" OR "duodenal microbiota") PUBMED ("celiac disease" [mh] OR "coeliac" [mh]) AND ("gut microbiota" OR "bacteria" OR "microbes" OR "microbiome" OR "microbiota dysbiosis" OR "metagenome" OR "metabolomic" OR "fecal microbiota" OR "intestinal microbiota" OR "duodenal microbiota"), showing the search details: "celiac disease" [mh] AND ("gut microbiota" [All Fields] OR "bacteria" [All Fields] OR "microbes" [All Fields] OR "microbiome" [All Fields] OR "microbiota dysbiosis" [All Fields] OR "metagenome" [All Fields] OR "metabolomics" [All Fields] OR "fecal microbiota" [All Fields] OR "intestinal microbiota" [All Fields] OR "duodenal microbiota" [All Fields]) COCHRANE *"celiac disease" OR "coeliac" in Title Abstract Keyword AND "gut microbiota" OR "bacteria" OR "microbes" OR "microbiome" OR "microbiota dysbiosis" OR "metagenome" OR "metabolomics" OR "fecal microbiota" OR "intestinal microbiota" OR "duodenal microbiota" in Abstract - ( genotype, but only 3% of them actually develop CD during the life (despite the dietary exposure to gluten foods), which thus corresponds to around 1% prevalence in the general population (31,32). Therefore, these HLA-DQ genes and the gluten dietary intake, even if necessary factors, are not sufficient to develop CD; other concomitant (and currently unknown or not welldefined) environmental agents, non-HLA genetic aspects, and maybe epigenetic mechanisms, are supposed to play a critical role in determining which individuals will become celiac within a much larger HLA-predisposed and gluten-exposed population. In this regard, the gut microbiota has been considered among the factors potentially affecting or modulating the risk of developing CD, through its interplay with the intestinal epithelium and/or the host immune system (33,34). In this systematic review, we aimed at summarizing and describing the main characteristics of the gut microbiota in CD children, compared to non-celiac controls. In order to achieve this purpose, we analyzed the pediatric studies investigating both fecal samples and duodenal specimens, which present substantial differences, of course. Indeed, as mentioned, the gut microbiota greatly varies along the GI tract, due to the different environmental conditions, in terms of pH, oxygen tension, substrates availability, host secretion, and intestinal motility (35). In the duodenum (which is characterized by a relatively acidic pH, high level of oxygen, and rapid transit time), facultative anaerobic and rapidly growing bacteria able to adhere to the epithelium in the mucus layer, are more likely to survive. On the contrary, the fecal samples are more representative of the colonic environment, which presents more favorable conditions for bacterial survival and, thus, is characterized by a more abundant and diverse microbiota, including mainly anaerobes (36,37).
Nadal et al. (14) mainly found that Bacteroides spp., Prevotella spp., and Escherichia coli populations were significantly more abundant in CD patients with active disease than in controls. Recently, Di Biase et al. (30) showed a "total dominance" of Enterobacteriaceae in the duodenal flora of active CD children; to follow, Bacteroidetes spp. and Streptococcus spp. turned out to be the second most represented category. However, no control group was included for this analysis. Schippa et al. (21) described a significantly higher microbiota biodiversity in CD children, compared to controls; in terms of bacterial composition, these authors highlighted a significant difference in the prevalence of Bacteroides vulgatus (85 vs. 20%) and E. coli (95 vs. 20%) between these two groups. In terms of phyla, as described by Sánchez et al. (27), Firmicutes and Proteobacteria were respectively more and less abundant in children with CD, compared to their controls; importantly, these results were also confirmed with respect to CD patients on gluten-free diet (non-active CD group).
Actually, two of the aforementioned studies also included non-active CD children [Nadal et al. (14) and Schippa et al. (21)]: both found no significant differences between these groups of children and their controls, in terms of bacterial proportions and biodiversity, respectively. Overall, all these three studies on pediatric CD indicated that the bacterial deviations could be normalized after the appropriate gluten-free diet. Conversely, Collado et al. (17) described significantly higher number of Bacteroides and Clostridium leptum in both treated and untreated (active) CD children compared to controls; however, a statistically significant difference between active CD children from one side, and both treated CD and control children, on the other side, was reported for Staphylococcus spp. and E. coli. Three additional studies included adult CD patients, in addition to children. Actually, Kalliomäki et al. (24) compared CD children to healthy children and non-active CD adults: they found no difference in bacterial counts among these three groups. Again, in the study by Nistal et al., there was no significant difference between active CD and healthy controls, considering the 36 different genera of known bacteria detected by their analysis. Importantly, this study provided a direct comparison between CD children and adults: even though both types of patients were colonized by bacteria mainly belonging to the Firmicutes, Proteobacteria, and Bacteroidetes phyla, the bacterial abundance and diversity were significantly lower in CD children than in CD adults (23). Cheng et al. also reported that the overall composition and diversity of gut microbiota was comparable between CD children and healthy controls; however, the specific profiles of eight bacterial groups were significantly different: in detail, Prevotella spp. (and, in particular, Prevotella melaninogenica), Haemophilus spp., and Serratia spp. (and, in particular, S. marcescens) were more abundant in CD children, whereas Prevotella oralis, Ruminococcus bromii, Papillibacter cinnamivorans, Proteus spp., and Clostridium stercorarium were less abundant (26).
Importantly, several studies focused the attention on some specific bacteria and, in detail, Bifidobacterium spp. and Lactobacillus spp. As for the former group, in both studies by Collado et al. (16), a significantly lower number was reported

Fecal sample
Comparison of fecal microbiota in CD and control children.
The fecal bacteria were analyzed using different plate culture media. Cellular morphology, Gram staining, biochemical test and antibiotic susceptibility analysis, were used to identify bacterial colonies. Total counts of bacteria were expressed as log of the number of colonies forming units (CFU) per gram of wet feces. Observational 1) n = 20 2) n = 10 3) n = 8 Fluorescent in situ hybridization (FISH) and flow cytometry with oligonucleotide probes was used to identify bacterial groups. 3 Sanz et al.
Fecal sample

Identification of difference in Bifidobacterium and
Lactobacillus groups in microbiota of celiac children and controls.
Observational 1) n = 10 2) n = 10 The microbiome analysis was performed by denaturing gradient gel electrophoresis (DGGE) analyses for Bifidobacterium spp. and Lactobacillus spp. using universal and specific primers. 4 Collado et al.  Evaluation of the relationships between fecal microbiota composition and immunoglobulin-coated bacteria in untreated and treated CD children.
The oligonucleotide probes were used for the fluorescent in situ hybridization (FISH) analysis identifying bacteria colonizing gut. Flow cytometry was used to identify immunoglobulin-coated bacteria, through EPICS ® XL-MCL flow cytometer. Evaluating the difference in the composition of microbiota and metabolome between treated CD children and healthy controls.
The microbiota and some subgroups (e.g., Bifidobacteria and Lactobacilli) were analyzed by PCR (universal primers targeting V6-V8 regions of the 16S rRNA) and denaturing gradient gel electrophoresis (DGGE). 11 Nistal et al. (Continued) Frontiers in Pediatrics | www.frontiersin.org Staphylococci were isolated from fecal samples and identified by PCR using the primers for Staphylococcus isolates and DNA sequencing with an ABI PRISM-3130XL Gene Analyzer.
Frontiers in Pediatrics | www.frontiersin.org Bacteroides-Prevotella bacteria and E. coli were significantly more abundant in biopsy specimens of CD patients with active disease than in controls. These results were not significantly different between controls and symptom free CD patients. Bacteroides-Prevotella, Streptococcus-Lactococcus, and E. coli were dominant groups in active CD patients, while in non-active CD patients the dominant groups were Streptococcus-Lactococcus and Clostridium histolyticum; in control children, Clostridium histolyticum, Bifidobacterium, and Faecalibacterium prausnitzii were the most abundant groups. 2 Collado et al.
Bifidobacterium longum was one of the most frequently detected, then it were the following frequently detected species: Bifidobacterium breve, Bifidobacterium bifidum, Bifidobacterium catenulatum, and Bifidobacterium lactis. In active CD group and control patients' biopsy samples more prevalent species was Bifidobacterium lactis. The prevalence of total Bifidobacterium was statistically significantly higher in controls than in active CD group. 3 Collado et al.
The highest total bacterial counts (not statistically significant) were in the untreated CD group, followed by treated CD patients and, finally, controls. Clostridium coccoides group was more prevalent in controls than in the other two groups (treated, not treated CD). Lactobacillus spp. was more prevalent in untreated CD patients and in controls, compared to treated CD group. Akkermansia muciniphila was more prevalent in untreated CD patients compared to treated CD. Bacteroides and Clostridium leptum groups were more prevalent in CD groups compared to controls.
Staphylococcus and E. coli groups were more prevalent in untreated CD than treated CD and controls. Bifidobacterium spp. was more prevalent in controls than in untreated CD patients. Lactobacillus group levels were significantly lower in treated CD control groups than in untreated CD. 4 Sánchez et al. (19) Bacteroides distasonis, Bacteroides fragilis/Bacteroides thetaiotaomicron, Bacteroides uniformis, and Bacteroides ovatus were higher in controls than in CD patients. Bacteroides vulgatus was higher in controls than in treated CD patients. Bacteroides dorei was more frequent in active CD patients than in treated CD patients and controls. Bifidobacterium adolescentis and Bifidobacterium animalis were higher in active CD than in treated CD and controls. A higher Lactic Acid Bacteria (LAB) diversity was described in treated CD and control patients than in active CD group. Weissella spp. and Lactobacillus fermentum were more common in treated CD than in controls and active CD patients. 5 Schippa et al. (21) Statistically significant differences between CD patients and controls were respectively found in the prevalence of Bacteroides vulgatus (85 vs. 20%), and E. coli (95 vs. 20%). Additionally, significant differences were found in the prevalence of Bacteroides vulgatus (80 vs. 90%) and Clostridium coccoides group (50 vs. 90%) between in active and inactive CD patients, respectively. There was no statistically significant difference in the prevalence of Bifidobacterium spp. between CD patients and control group, and between active and inactive CD patients. Overall, bacterial population was more diverse in duodenal mucosa of CD group compared to the control group.
6 Di Cagno et al. (22) The Lactobacillus plantarum was present in duodenal biopsies of treated CD patients and healthy controls. Bifidobacteria group were not present in duodenal biopsies of these groups. 7 Kalliomäki et al. (24) There was no statistically significant difference in the bacterial species between the following groups and species: Bifidobacterium genus, Bifidobacterium adolescentis, Bifidobacterium catenulatum group, Bifidobacterium longum subsp infantis, Bifidobacterium longum, Staphylococcus aureus, Bacteroides-Prevotella-Porphyromonas group, Bacteroides fragilis group, Streptococcus genus, and Lactobacillus group. The Bacteroides fragilis group, Bifidobacterium catenulatum group, Bifidobacterium longum, and Streptococcus genus were detected in a small number of samples. However, Lactobacillus group, Staphylococcus aureus, and B. longum subsp infantis were not detected at all. 8 Nistal et al. (23) Ninety eight percent of the sequenced bacterial species from the proximal small intestine of adults belonged to the following phyla: Firmicutes (38%), Proteobacteria (29%), Bacteroidetes (17%), Actinobacteria (10%), and Fusobacteria (4%). Majority (99%) of the sequences among children belonged to Proteobacteria (38%), Firmicutes (34%), Bacteroidetes (13%), Actinobacteria (4%), Deinococcus-Thermus (2.7%), Fusobacteria (2.9%), and unknown phylum sequences (5%). Sequences from genera such as Streptococcus, Prevotella, Neisseria, Haemophilus, Granulicatella, and Acinetobacter were present in at least 60% of children. Streptococcus and Prevotella genera sequences were more prevalent among healthy children compared to untreated CD children. 9 Cheng et al. (26) In both healthy controls and CD groups, major bacterial groups were found: Members of phylum Proteobacteria were higher in active CD patients than in controls and non-active CD. Members of phylum Firmicutes were lower in active CD than in controls and non-active CD groups. Members of phylum Actinobacteria were higher in active CD vs. non-active CD patients. Klebsiella oxytoca isolates were more prevalent in active CD than in controls. Staphylococcus epidermidis and Staphylococcus pasteuri were more prevalent in active CD group than in controls and inactive CD group. Streptococcu anginosus and Streptococcus mutans were more abundant in controls than in active and non-active CD. Streptococcus mitis group was higher in non-active CD than in active CD. Actinomyces odontolyticus was higher in active CD compared to inactive CD.
Overall, there was a dominance of Enterobacteriaceae (87.4%) in duodenal samples of CD patients. Sub-dominance of Bacteroidetes (4.8%) and Streptococcus (3%) was observed in several samples. The evaluation of the composition of the Enterobacteriaceae cluster was not possible due to the methodology used in the study. However, it was found that 50% of the samples had Enterobacteriaceae belonging to the genus Proteus.
in CD (and, in particular, untreated) children than in controls; on the contrary, Di Cagno et al. (18) found no significant representation of Bifidobacteria in their CD and control groups, overall. No significant differences were reported in the prevalence of Bifidobacterium spp. between CD patients and controls, and between active and inactive CD, in the study by Schippa et al.  Bacteroides, Clostridium, and Staphylococcus spp. were the dominant bacterial groups in CD patients, and their levels were higher than in controls. Dominant bacterial groups in controls were Bifidobacterium and Enterobacteriaceae. Bacteroides-Prevotella, Clostridium histolyticum, and Clostridium coccoides groups sulfate-reducing bacteria (SRB) and Atopobium group were significantly higher in CD patients than in controls. 2 Sanz et al.
The fecal microbiota was more diverse in CD patients than in controls. In celiac children one to six different Lactobacillus groups were present; species other than Lactobacillus species were dominant in comparison with controls. The Lactobacillus casei group was more prevalent in control group than in celiac patients. Bifidobacterium species were significantly higher and diverse in controls than in CD patients. Controls combined both infant-and adult-type Bifidobacterium species, while CD patients mainly showed infant-type species (Bifidobacterium bifidum and Bifidobacterium infantis). Bifidobacterium longum, Bifidobacterium pseudocatenulatum, and Bifidobacterium dentium were higher in controls, while the Bifidobacterium bifidum was prevalent in CD group. Bifidobacterium dentium and Bifidobacterium adolescentis were not detected in any CD samples. 3 Collado et al.
Fecal samples showed higher numbers of bifidobacteria than duodenal samples for every analyzed group of bacteria. Bifidobacterium adolescentis was detected more frequently in non-active CD than in active CD and controls. Bifidobacterium dentium was significantly more prevalent in non-active CD than in controls. The most predominant Bifidobacterium species in fecal samples were Bifidobacterium longum, Bifidobacterium catenulatum, and Bifidobacterium bifidum. Total Bifidobacterium levels were significantly higher in controls than in active and non-active CD patients. 4 Collado et al.
Total bacterial counts were significantly lower in control children than in untreated and treated CD patients. Staphylococcus spp. were less prevalent in controls than in untreated and treated CD patients. E. coli was significantly higher in untreated and treated CD patients than in controls.
Bacteroides and Clostridium leptum groups were significantly higher in both untreated and treated CD patients than in controls. Bifidobacterium spp. counts were significantly higher in controls than in untreated and treated CD patients. Lactobacillus spp. counts were significantly different only between treated CD patients and controls.

Di Cagno et al. (18)
Bacteroides and Clostridium were higher in treated CD and untreated CD than in controls. The ratio of lactic acid bacteria and Bifidobacterium to Bacteroides and enterobacteria, was lower in treated CD. This ratio was even lower in untreated CD. Lactobacillus brevis, Lactobacillus rossiae, and Lactobacillus pentosus were detected only in treated CD patients and controls. Lactobacillus fermentum, Lactobacillus delbrueckii subsp. bulgaricus, and Lactobacillus gasseri were observed only in several fecal samples of healthy controls. Compared to controls the composition of Bifidobacterium species varied for treated CD, while in untreated CD patients there was more variance of these species.

6
De Palma et al. (20) Gram-positive bacterial population was more common in active CD patients compared to GFD celiac group. Bifidobacterium population was lower in fecal samples of untreated CD patients than in controls. Ig-A coated Bacteroides-Prevotella were higher in controls than in treated and untreated CD patients. Clostridium histolyticum, Clostridium lituseburense, and Faecalibacterium prausnitzii groups were lower in untreated CD than in controls. E. coli, Staphylococcus, Lactobacillus-Enterococcus, and sulfate-reducing bacteria were similar in all three groups studied.
Staphylococcus epidermidis was lower in the control group than in active and non-active CD groups. Staphylococcus haemolyticus was higher in active CD group than in controls. Staphylococcus aureus was lower in active CD children compared to non-active CD and controls. Staphylococcus spp. was more diverse in active CD patients than in all the other patients. 9 Pisarello et al.
Lactobacilli were less abundant in the CD patients on GFD than in control children. The enterobacteria population had a trend to increase in CD group compared to healthy controls. Still, there was no significant differences in total counts of the aerobic and anaerobic among two studied groups.
10 Quagliariello et al. (29) Lactobacillus spp. was more abundant in controls than in CD patients. Bacteroides fragilis was higher in CD group than in controls. No significant differences were detected in levels of Bacteroides. After the treatment with probiotics, the enterobacteria were higher in the controls compared to CD group. Later, after 3 months of probiotic treatment, the levels of enterobacteria declined in the probiotic group.
A statistically significant lower relative abundance was detected for Bacteroides/Prevotella cluster in CD group (10.2%) compared to the control group (15.6%). A statistically significant lower abundance in the cluster Akkermansia was found in CD group (0.7%) compared to control patients (4.2%) and, as for the cluster Staphylococcaceae, in CD group (0.9%) compared to healthy controls (2%). The relative abundances of bacterial clusters of duodenal and fecal microbiota were not statistically significantly correlated.
sample size, since all these studies included around or no more than 30 children, probably due to technical and ethical issues concerning the invasive procedures required to analyze the duodenal microbiota. This aspect may have also affected the selection and inclusion of an appropriate control group, of course: actually, whereas in the studies by Collado et al. (13) and Nadal et al. (14) these control children are not better defined than as "without gluten intolerance" and "without known food intolerance, " only Schippa et al. (21) disclosed the reason why they underwent upper GI endoscopy, namely, "functional dyspepsia." Therefore, the unclear description and, anyway, the impossibility to compare these findings with completely healthy children represent additional obstacles to the appropriate interpretation of the microbiota findings obtained from the duodenal mucosa. These specific research limitations can be lessened by studying the microbiota in fecal samples, whose collection does not imply any invasive procedure, of course. Importantly, this aspect makes this analysis much more attractive as potential non-invasive diagnostic and/or prognostic biomarkers for pediatric CD.
To start with, Collado et al. (13) showed significantly lower total bacterial counts in stools from the control group compared to treated or untreated CD children, but this specific information was not provided by other studies. In terms of microbiota general diversity, Sanz et al. (15) found it significantly higher in CD children than in controls, but Di Cagno et al. (18) could not confirm this finding.
Some studies mainly analyzed the relative composition of the fecal microbiota. In general, De Palma et al. (20) investigated both active and treated CD children, in addition to healthy controls: they reported a significantly decreased gram+/gramratio in both groups of CD children, compared to controls; interestingly, treated CD children showed intermediate values between active CD children and healthy controls, but the difference was statistically significant only with the latter group, suggesting that this specific aspect was not completely restored by the gluten-free diet. In terms of phyla, these authors reported that the Bacteroides/Prevotella cluster was significantly more expressed in untreated CD patients than in healthy controls; this finding was confirmed by Collado et al. (16,17) and Quagliariello et al. (29).
However, most researchers investigated some specific bacteria of the fecal microbiota, mainly Bifidobacterium spp. and Lactobacillus spp., which differed in terms of relative species abundance between CD patients and healthy controls (15,22,28). Sánchez et al. (25) focused on Staphylococcus spp. and observed some peculiarities in species diversity and abundance in CD children; however, the increased presence of S. epidermidis isolates carrying genes conferring resistance to methicillin suggested that these changes may be induced by a greater exposure to antibiotics rather than CD itself. However, Di Biase et al. (30) reported a lower abundance of Staphylococcus spp. in CD children, as well as a significant reduction of the Bacteroides/Prevotella cluster, which is in contrast with some of the aforementioned studies.
Therefore, even though the available fecal microbiota studies in CD children are more than those investigating the duodenal microbiota, their clinical aims and settings were quite heterogeneous: indeed, they provided a qualitative and quantitative information that is not immediately comparable, and the results are often conflicting because of that, probably.
Importantly, 3 of these 11 articles on fecal microbiota also provided data related to duodenal biopsies from the same cohorts of patients (16,17,30). In detail, Collado et al. discussed the relation between fecal and duodenal microbiota: they found significant correspondences for Bifidobacterium spp. among all three groups (active CD, treated CD, and controls), whereas other bacterial groups did not correlate in all groups. As regards specifically untreated CD children, Bacteroides spp., Staphylococcus spp., Lactobacillus spp., C. leptum group, Clostridium coccoides group, E. coli, and Akkermansia muciniphila showed a significant correlation between these two types of samples (16,17). Unfortunately, Di Biase et al. (30) did not analyze such a correlation; however, about this study, it may be worth to emphasize the fact that A. muciniphila was found to be significantly less abundant in CD children's stool, along with Bacteroides spp. and the Staphylococcus group, as previously mentioned. Although Collado et al. (13) described a significant duodenalfecal relation for that microorganism, they did not actually find statistically significant differences in its content among their study groups. In an additional study, they focused on Bifidobacterium spp. and, once again, they compared duodenal biopsies and fecal samples in the same cohort of patients. The fecal microbiota resulted to include significantly higher numbers of bifidobacteria than the duodenal samples. Interestingly, they found a statistically significant correlation between these sample types in all groups as regards the total number of bifidobacteria and, in terms of species, only Bifidobacterium longum significantly correlated in both samples and all groups. B. longum was the most frequent and abundant species in this study and, importantly, was significantly different among all study groups, both in stools and duodenum, in addition to be the only one to significantly correlate between both types of samples, as previously explained (16).
Indeed, B. longum was the object of some preliminary investigations as potential therapeutic resource and prognostic biomarker. In a double-blind, randomized, and placebocontrolled trial, Olivares et al. (38) showed that the administration of the CECT 7347 strain of B. longum was associated with a significant reduction of the Bacteroides fragilis group in the microbiota of CD patients, in addition to some clinical benefit in terms of anthropometric parameters. Some recent analyses derived from a large perspective cohort study (PROFICEL), comparing the fecal microbiota in children at HLA-DQ genetic risk of CD before the appearance of the disease (exactly, at 4 and 6 months of life), highlighted some microbiota differences between those children who eventually developed CD and those who did not. Interestingly, high-risk genotypes for CD (referring to HLA-DQ2 and HLA-DQ8 heterodimers) were associated to a lower amount of Bifidobacterium spp. and, specifically, B. longum (39)(40)(41). More in general, several microbial species and related metabolites (with inflammatory and immunological properties) have been recently suggested as potentially specific to CD, through multi-omics analysis (42).
Even though the interplay between HLA-DQ (and, in detail, HLA-DQB1 * 02, which is the most frequent allelic variant in CD children) and intestinal microbiota must be precisely elucidated yet, these preliminary observations might provide the background to plan further studies to assess the risk of developing CD in gluten-exposed population and, potentially, to consider additional non-invasive diagnostic tools and prognostic markers for CD (37,43,44).

CONCLUSION
Due to the heterogeneity of the experimental procedures and design of these studies, it is not possible to evidence any specific and absolute celiac signature in the fecal and/or duodenal microbiota of CD children. However, some specific components of the fecal microbiota (e.g., Bifidobacterium spp.) may deserve additional research efforts to understand the potential application as both probiotic therapy and/or diagnostic/prognostic biomarker.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author/s.