Background
Celiac disease (CD) is a common chronic immune-mediated, inflammatory disorder of the small intestine induced by intolerance to gluten-containing dietary products [
1,
2]. When a CD patient consumes gluten, an inflammatory cascade occurs in the small intestinal mucosa, eventually resulting in an active disease that is characterized by villous atrophy, crypt hyperplasia and increased numbers of lymphocytes in the lamina propria [
1,
3]. Untreated CD is manifested by gastrointestinal symptoms, malabsorption and even malnutrition [
1,
3] and also by extra-intestinal symptoms such as dermititis herpetiformis (skin rash) in some individuals [
3]. Both genetic predisposition and environmental factors are considered to be involved in the development of CD [
4]. Individuals who carry the alleles human leukocyte antigen (HLA)-DQ2 or HLA-DQ8 have an increased risk of developing the disease, but only less than one tenth of them eventually get CD, indicating that other genetic factors and/or environmental factors are also important in the pathogenesis [
5]. Recently, 13 new CD risk loci were identified, bringing the number of known CD loci to 40 and giving a refined picture of the genetic risk of CD [
6]. Most of these loci contain candidate genes of immunological function, but the pathways leading from genetic predisposition to an affected person are poorly understood [
6,
7]. The intestinal microbiota is a major dictator of the antigen milieu of enterocytes, and it may have a role in the CD pathogenesis.
Gut colonization starts immediately after birth, and considerable microbiota maturation takes place during the first years of life followed by a gradual microbiota succession until young adulthood [
8‐
10]. A balanced commensal microbiota contributes to the physiological development of the gut and the maturation of the immune system; thereby, alterations in the intestinal microbiota could play a role in the onset of different diseases, including CD [
3,
11]. Increased bacterial diversity and changes in several bacterial groups in the microbiota of pediatric CD patients have been reported in several studies [
12‐
16]. However, other recent studies have failed to show major microbiota differences between children with and without CD [
17‐
19]. Two recent studies have addressed the microbiota in infants with a genetic predisposition to CD [
4,
11]. Both of these reported the microbiota of predisposed infants to be different from that of non-predisposed, but while Sellitto et al. reported a reduction or lack of
Bacteroides in predisposed infants [
4], De Palma et al. found that
Bacteroides fragilis and staphylococci were increased and bifidobacteria were reduced in genetically susceptible infants [
11]. Although the idea that the microbiota is involved in the etiology of CD has been addressed in numerous studies, the results on specific CD-associated microbiota changes remain inconclusive.
Recently, we demonstrated that pediatric CD patients have decreased duodenal expression of Toll-like receptor 2 (TLR2) and the negative regulator of Toll-receptor signaling (Tollip), and increased expression of TLR9 and interleukin 8 (IL-8), which is a marker of intestinal inflammation [
18]. TLRs are a family of pattern recognition receptors, which recognize conserved microbe-associated motifs such as lipopolysaccharides (LPS, ligand for TLR4), lipoproteins and lipoteichoeic acids (LTA, ligand for TLR2), flagellin (TLR5) and nucleic acid motifs (TLR3, 7 and 9). Intestinal epithelial homeostasis is dependent on the activation of TLRs at adequate levels in order to keep defense against microbes balanced and to avoid an excessive inflammatory response to gut commensals [
20,
21]. Further, TLR signaling is known to affect epithelial barrier function by having a bearing on the expression of tight junction proteins, mucus protein mucin 2 and antimicrobial peptides such as RegIII-γ [
20,
21]. Thus, our previous results suggested that microbiota may play a role in the etiology of CD, but we were unable to reveal aberrancies in the duodenal mucosal microbiota of pediatric CD by targeting selected bacterial groups by quantitative PCR (qPCR) [
18].
The objectives of the present study were to comprehensively characterize the total duodenal mucosal microbiota and to re-evaluate the possible microbiota differences in pediatric CD patients and healthy controls by using a high-throughput bacterial phylogenetic microarray (HITChip). Further, the expression of nine mucosa-associated genes, IL-10, interferon-gamma (IFN-γ), tumor necrosis factor alpha (TNF-α), a tight junction protein zonula occludens-1 (ZO-1), a gap junction protein connexin-43 (Cx43), a mucus protein mucin 2 (MUC2), an antimicrobial peptide RegIIIγ, a chemokine CXCL16 and its receptor CXCR6 were measured by using a quantitative reverse transcription-PCR (qRT-PCR).
Discussion
Human small intestine, especially duodenum, is a distinctive environment for microbial life because of the excretion of digestive enzymes and bile, and it harbors a microbiota that is different from the one in the large intestine [
39]. The HITChip analysis showed that representatives of the genus
Streptococcus constitute 29% and 32% of the total duodenal microbiota in HC and CD subjects, respectively, and that the signal obtained for all Bacilli (phylum-like group) predominantly resulted from streptococci. This is in line with the previous findings [
39,
40]. In addition to Bacilli, Proteobacteria, Bacteroidetes and Clostridium cluster XIVa were found to be the most abundant phylum-like groups in the duodenum. These bacteria have previously been found to be dominant groups in the ileum, distal duodenum and proximal jejunum [
17,
39,
41] and in the duodenum [
19]. Interestingly, within Proteobacteria a single most abundant genus-like group was found to be
Sutterella wadsworthensis et rel. both in CD and HC subjects, which suggests that it belongs to the normal duodenal microbiota. Similarly, Mukhopadhya et al. detected
S. wadsworthensis with an equally high frequency of 84 to 86% in the colonic biopsies of healthy adults (n = 64) or ulcerative colitis patients (n = 69), showing that the species belongs to the normal microbiota of intestinal mucosa and is unlikely to have a role in IBD [
42]. Genus
Aquabacterium belonging to the Proteobacteria also had a relatively high abundance (~9.5%, prevalence 100% in both CD and HC) in the duodenum of both groups of children. Previously,
Aquabacterium has been detected in human colonic mucosal biopsy [
43], but to our knowledge, this is the first time that
Aquabacterium has been described as an abundant inhabitant of the human small intestine.
In the PCA and RDA analysis, HC and CD subjects did not cluster separately. Furthermore, none of the 65 genus-like bacterial groups was found to be significantly different in abundance between HC and CD. Thus, the overall duodenal microbiota composition seems comparable between HC and CD, which is in line with the results obtained by Ou et al [
17] and Nistal et al [
19] using small intestinal biopsies. Moreover, the bacterial diversity was also found to be comparable between the study groups. Previously, bacterial diversity assessed by PCR-T/DGGE has been observed to be increased in children with CD [
13,
16]. It is noteworthy that PCR-D/TGGE analysis detects only the most abundant bacteria and therefore may strongly underestimate microbiota diversity in complex communities. Our results from high-throughput microbiota profiling, like the results by Nistal et al [
19], give a more in depth view of the duodenal microbiota regarding the bacterial groups inhabiting duodenal mucosa and the overall diversity. In addition, several studies have found differences in specific bacterial groups between HC and CD [
12,
14,
15,
44]. The most consistent findings from these studies were that children with CD have increased counts of
Bacteroides and reduced counts of bifidobacteria either in feces or duodenal epithelium [
12,
14,
15] which were not found to differ between CD and HC in this study.
As the overall microbiota profile, diversity or individual genus-like groups did not show a significant difference between HC and CD, random forest was used to explore whether a sub-population bacterial profile could be associated with the health status. A profile of eight bacterial groups was found to distinguish HC from CD. The abundance of
Prevotella melaninogenica, and the total abundance of
Prevotella spp (Table
1) were found to be higher in CD, which is in line with the results of Ou et al. [
17]
Haemophilus et rel. were also found to be enriched, although insignificantly as an individual bacterial group, in CD children by Nistal et al. [
19].
Serratia spp. was also found to be present in higher abundance in CD. The phylogenetic microarray targets mainly
S. marcescens, which is considered as an opportunistic pathogen able to cause invasive infections (sepsis, meningitis, pneumonitis) in neonates [
45,
46]. These bacteria may impair the intestinal integrity, but their possible role in CD remains to be elucidated.
In the distinctive profile detected by random forest
, P. oralis, R. bromii,
P. cinnamivorans,
Proteus and
C. stercorarium groups are increased in the HC group.
R. bromii is of particular interest, because it acts as key species for fiber/resistant starch degradation in the intestine [
47] feeding butyrate producing bacteria including
P. cinnamivorans[
48]. Butyrate is a major source of energy to the enterocytes and acts as a regulator of gene expression, inflammation and differentiation in host cells [
49]. Previously,
R. bromii has been detected at increased level in healthy subjects as compared to Crohn’s disease patients [
50], indicating its potential role for benefiting the gut mucosal homeostasis.
Finally, it should be taken into account that the highly individual-specific microbiota-profiles may have a strong impact on the results within small study groups. Moreover, children in the HC group had healthy duodenal mucosa, but had gastrointestinal complaints or other reasons for gastroscopy and the possibility of microbiota alterations in these children as compared to children devoid of any symptoms can´t be excluded. Therefore, the profile comprising eight genus-like bacterial groups, which showed significant difference between CD and HC should be verified in future studies.
Similar average proportion of Gram-positive and Gram-negative bacteria, which are the carriers of LTA and LPS, respectively, was found in CD and HC. Our previous findings from the same cohort showed that the expression of TLR2 is higher in HC [
18]. As the LTA content is comparable between HC and CD, the net signaling through TLR2 is presumably higher in HC.
In vitro, TLR2 stimulation of intestinal epithelial cells (IECs) has been shown to increase Cx43 synthesis, the apical reorganization of ZO-1 and trans-epithelial resistance, which reflects the strength of tight junctions between IECs and barrier function [
51‐
53]. Although no significant difference was found in the expression ZO-1 and Cx43 between HC and CD, the expression of ZO-1 was significantly lower in CD as compared to T-CD. Moreover, there was a significant positive correlation between the expressions of ZO-1 and TLR2 in HC further supporting earlier findings, which demonstrated more permeable epithelial barrier in CD as compared to HC due to the decreased expression of tight junction proteins [
54]. Negative correlation was found between the abundance of
S. moorei et rel. and the expression of Cx43. Previously this bacterium has been associated with oral cavity diseases [
55] and its possible down-regulatory effect of Cx43 may facilitate invasion and colonization. In CD, however,
S. moorei et rel. seems to be irrelevant as it was found to be as abundantly present in HC and CD.
CXCL16 has been shown to function as a scavenger receptor in antigen-presenting cells where it mediates adhesion and phagocytosis of both Gram-positive and Gram-negative bacteria [
56]. In addition, it works as a chemokine for CXCR6-expressing cells such as natural killer T cells and T helper 1 (Th1) - polarized CD4 T cells [
57]. Since both microbiota alterations and Th1-polarized inflammation have been linked to the pathogenesis of celiac disease, we decided to evaluate the expression of both CXCL16 and its receptor CXCR6 in this study. We found that the expression of cytokine CXCL16 was found to be comparable in CD and HC and T-CD, whereas the expression of its receptor CXCR6 was higher in CD and T-CD as compared to HC. CXCL16 has dual functions as a transmembrane adhesion molecule and a soluble chemokine [
58]. Both the membrane-bound form of CXCL16 and its receptor CXCR6 have been found to be expressed not only by dendritic cells/macrophages and T cells respectively, but also by IECs [
58‐
60]. Previously, increased intestinal CXCL16 expression has been observed in the colonic biopsies of Crohn´s disease patients due to immune cell infiltration [
60]. In this study, an increased expression of CXCR6 was not only observed in the inflamed mucosa of CD but also in T-CD without immune cell infiltration suggesting an altered expression in epithelial cells. However, increased expression of CXCR6 in mucosa associated immune cells cannot be excluded when studying whole biopsies with different cell types. Diegelmann et al observed that
in vitro stimulation of CXCR6 activates several distinct signaling pathways in IECs and they suggested that the CXCL16-CXCR6 chemokine-receptor system contributes to the integrity of epithelium and the regulation of mucosal innate and adaptive immune systems [
60]. To our knowledge our data report for the first time an increased duodenal expression of CXCR6 in CD subjects in whom it may have an important role in the mucosal immunity.
The average genomic GC content (GC%) of the total microbiota and the proportion of high-GC% bacteria was found to be comparable between the HC and CD groups. High GC content of a bacterial genome correlates with a higher number of potentially immunostimulatory CpG motifs in the genome (Kant R, de Vos WM, Palva A, Satokari R, unpublished results) [
29,
30] and thus, the GC% was taken to reflect the load of TLR9 ligands in the microbiota. Consequently, HC and CD are likely to harbor similar load of TLR9 ligands. Our previous results from the same cohort showed that the expression of TLR9 in duodenum is higher in CD [
18]. Thus, the signaling through TLR9 is presumably higher in CD as compared to HC due to the equally high load of TLR9 ligands and higher expression of TLR9. In immune cells, TLR9 stimulation is known to trigger Th1 type immune responses [
61‐
63]. Further, a recent
in vitro study demonstrated that apical TLR9 stimulation of IECs leads to an increased expression of IFN-γ and IL-10, but not TNF-α from peripheral blood mononuclear cells (PBMCs) on the basolateral side [
64]. In CD patients, increased mucosal expression of IL-10 and IFN-γ is well documented and has been associated with the activity of gliadin-reactive T-cells upon the gluten/gliadin stimulation either
in vivo or
in vitro[
65‐
69]. We found that the expression of IL-10 and IFN-γ in the duodenal mucosa was increased not only in CD but also in T-CD with gluten-free diet, suggesting for an additional gluten/gliadin-independent route of stimulation. Similarly to previous studies [
65,
68,
69], TNF-α expression was found to be unaffected in CD and T-CD as compared to HC. The ratio of IL-10 to IFN-γ transcripts was strongly reduced in both CD and T-CD as compared to HC suggesting that although IL-10 expression was also increased there is a significant inclination towards a Th1 response both in untreated and treated CD. Since both CD and T-CD had increased expression of TLR9 [
18], we hypothesize that the increased TLR9 signaling in the small intestine may contribute to the persistent activation of Th1 (IFN-γ) signaling pathway markers in the small intestine found in CD children despite gluten-free diet treatment [
70]. Due to the limitations in studying whole biopsies with different cell types we cannot conclude which of the proposed signaling routes i.e. the direct TLR9 stimulation of immune cells or the stimulation of IECs with subsequent triggering of immune cells would possibly encompass more into the observed cytokine profile.
Competing interests
The authors declare that they have no competing interest.
Authors’ contributions
Study concept and design: MK, WdV, RS; collection of clinical samples: MK, HL; acquisition of data: JC, MK, HH, HL, RS; analysis and interpretation of data: JC, MK, HH, JS, RS; statistical analysis: JC, JS; drafting of the manuscript: JC, MK, RS; critical reading and revision of the manuscript: JC, MK, AP, WdV, JS, RS; obtained funding: MK, AP, RS; administrative: MK, AP, RS; study supervision: MK, JS, RS. All authors read and approved the final version for publication.