Altered Duodenal Microbiota Composition in Celiac Disease Patients Suffering From Persistent Symptoms on a Long-term Gluten-free Diet

Pirjo Wacklin PhD; Pilvi Laurikka; Katri Lindfors PhD; Pekka Collin MD; Teea Salmi MD; Marja-Leena Lähdeaho MD; Päivi Saavalainen PhD; Markku Mäki MD; Jaana Mättö PhD; Kalle Kurppa MD; Katri Kaukinen MD


Am J Gastroenterol. 2014;109(12):1933-1941. 

In This Article


Clinical Data

Altogether, 18 patients with persistent gastrointestinal symptoms and 18 with no current symptoms were selected for intestinal microbiota analyses. Microbiota profiling was successful for all samples except one sample from each study group, and the final results were thus available for 34 patients. There were no differences between the two groups in any demographic, clinical, serological, or histological parameters except that the patients with persistent symptoms were younger and had, by definition, higher GSRS scores than the patients with no symptoms (Table 1). More detailed information about GSRS sub-dimension scores is presented in Supplementary Table S2 The age difference was mainly caused by two young patients (27 and 31 years old) in the Persistent symptoms group. Celiac disease-associated genotypes were tested for 22 out of the 34 patients, all of whom had the human leukocyte antigen DQ2 or DQ8. In the Persistent symptoms group, eight patients were reported to suffer from gastrointestinal disorders that were previously diagnosed by health-care providers: four with gastroesophageal reflux, one with diverticulosis, two with lactose intolerance, and one with irritable bowel syndrome (IBS). One patient in the No symptoms group had a previous diagnosis of gastroesophageal reflux. In each case the disease was considered mild by the patient, and there were no differences in the use of nonsteroid anti-inflammatory drugs or PPIs between the groups (Table 1). None of the participants were found to have H. pylori infection. Controlled hypothyroidism was reported in four patients with persistent symptoms and three patients with no current symptoms. No other gastrointestinal or autoimmune diseases, e.g., motility disorder or diabetes, were reported by the patients or found in medical records.

Mucosal Microbiota Composition

The microbiota analysis indicated that the celiac disease patients with persistent symptoms on GFD had altered duodenal microbiota composition in comparison with patients without symptoms. Microbiota compositional differences were assessed by several clustering methods, and all of them gave congruent results at the genera level and mostly similar at the OTU level. According to an constrained clustering method, redundancy analysis, microbiota composition based on bacterial genera (P=0.01) or OTUs (P=0.03) differed significantly between the groups (Figure 1 and Supplementary Figure S1A online An unconstrained clustering method, multidimensional scaling, also supported the detected differences between the study groups (Figure 2 and Supplementary Figure S1B Similarly, the intestinal microbiota structure differed between the study groups by weighted Unifrac based on genera (P<0.001) or OTUs (P<0.001). According to unweighted Unifrac, the membership of duodenal bacterial genera (P=0.02), but not OTUs (P=0.25), varied statistically significantly between the study groups. There was a similar trend in hierarchical clustering based on the parsimony analysis of genera but this was not significant (P=0.07). Hierarchical clustering of OTUs was not statistically significant (P=0.71).

Figure 1.

Redundancy analysis (RDA) plot indicating alterations in duodenal microbiota composition in celiac disease patients with persistent symptoms (closed circles) in comparison with patients without symptoms on gluten-free diet (open circles). PCA, principal component analysis. RDA was based on observed genera. Triangles indicate centroids of the study groups.

Figure 2.

Multidimensional scaling (MDS) plot indicating differences in duodenal microbiota composition in gluten-free diet-treated celiac disease patients with persistent symptoms (closed symbols) and treated patients without symptoms (open symbols). The celiac disease patients with an additional gastrointestinal disease are indicated by squares and patients without other gastrointestinal symptoms are indicated by circles. The arrows indicate the samples of two youngest patients (see text for details). MDS was based on observed genera.

A further analysis of microbiota showed that the celiacs suffering from persistent symptoms had a lower relative abundance of Bacteroidetes (15% vs. 25%, P=0.01) and Firmicutes (33% vs. 46%, P=0.05) and a higher relative abundance of Proteobacteria (40% vs. 21%, P=0.04) compared with patients with no symptoms on GFD (Figure 3). The difference in microbiota composition was also detected at the genus level in altogether seven taxa, including the highly abundant genus Prevotella (Table 2). The most abundant bacterial genera in both study groups are presented in Supplementary Table S1

Figure 3.

The differences in average abundances of bacterial phyla in gluten-free diet-treated celiac disease patients suffering from persistent symptoms and treated patients without symptoms.

The duodenal microbiota richness, measured as a number of detected genera or OTUs, was reduced in patients with persistent symptoms on GFD. On average, 32 genera and 72 OTUs per sample (in total, 117 genera and 721 OTUs) were detected in patients in the Persistent symptoms group, whereas the No symptoms group had on average 37 genera and 106 OTUs per sample (in total, 131 genera and 1,016 OTUs). Furthermore, the microbiota richness based on genera (P=0.05) or OTUs (P=0.007) was lower in symptomatic compared with nonsymptomatic patients assessed by c2m randomization test (Figure 4 and Supplementary Figure S1C Similarly, rarefaction curves based on OTUs and genera indicated lower richness in patients with persistent symptoms (Supplementary Figures S1D and S2, although the difference was significant only for OTUs. No difference was detected on inverse Simpson or Shannon diversity indices that take account of both the richness (number of species) and the abundances and evenness of species. As the richness is part of diversity estimates, they usually correlate well. Thus, the similar diversity between the groups indicated that in the Persistent symptom group the richness was decreasing and the abundances of the dominating species were getting more even (the evenness was increasing). This suggests that the abundance of different species was affected differently in subjects with persistent symptoms. A decreasing rate of genera and OTUs at the end of the rarefaction curves (Supplementary Figures S1D and S2 and Good's coverage for genera of over 97.8% in all samples demonstrated that the sequencing effort was adequate to capture most of the bacterial diversity.

Figure 4.

The reduced richness of duodenal bacterial genera in gluten-free diet-treated celiac patients suffering from persistent symptoms in comparison with treated patients without symptoms.

To study the possible contribution of the additional gastrointestinal diseases on observed microbiota alterations, microbiota analyses were repeated by excluding nine patients with a diagnosis of other gastrointestinal disease from the analyses. The results of this smaller data set (n=25, including 9 patients from the Persistent symptoms and 16 patients from the No symptoms groups) were mostly in accordance with those seen in all 34 patients (Supplementary Figure S3 The relative abundances of Firmicutes and Bacteroidetes were lower and the relative abundance of Proteobacteria was higher in the patients with persistent symptoms, although only the abundance of Bacteroidetes (P=0.03) was statistically significant. In addition, similar to the whole data set, microbiota composition differed between the groups by redundancy analysis (P=0.06) (Supplementary Figure 3A,B and by multidimensional scaling, but not by Unifrac and parsimony analyses. Based on the OTUs, microbial richness was reduced in the Persistent symptoms group, even after exclusion of the patients with other gastrointestinal diseases (Supplementary Figure S3D Richness based on genera (P=0.21) did not differ between the groups in analysis excluding celiac disease patients with an additional gastrointestinal disease (Supplementary Figure S3C