The role of the intestinal microbiota as a regulator of autoimmune diabetes in animal models is well-established, but data on human type 1 diabetes are tentative and based on studies including only a few study subjects. To exclude secondary effects of diabetes and HLA risk genotype on gut microbiota, we compared the intestinal microbiota composition in children with at least two diabetes-associated autoantibodies (n = 18) with autoantibody-negative children matched for age, sex, early feeding history, and HLA risk genotype using pyrosequencing. Principal component analysis indicated that a low abundance of lactate-producing and butyrate-producing species was associated with β-cell autoimmunity. In addition, a dearth of the two most dominant Bifidobacterium species, Bifidobacterium adolescentis and Bifidobacterium pseudocatenulatum, and an increased abundance of the Bacteroides genus were observed in the children with β-cell autoimmunity. We did not find increased fecal calprotectin or IgA as marker of inflammation in children with β-cell autoimmunity. Functional studies related to the observed alterations in the gut microbiome are warranted because the low abundance of bifidobacteria and butyrate-producing species could adversely affect the intestinal epithelial barrier function and inflammation, whereas the apparent importance of the Bacteroides genus in development of type 1 diabetes is insufficiently understood.

Type 1 diabetes (T1D) is caused by the destruction of the pancreatic β-cells in genetically susceptible individuals. The disease is considered to be immune mediated, and the appearance of circulating autoantibodies against β-cells is seen years before the diagnosis along with a significant reduction in β-cell mass (1,2). Environmental factors associated with the activation of the gut immune system, such as early exposure to dietary antigens (cow’s milk and gluten), have been associated with the induction of this process (35). The role of the gut immune system in the pathogenesis of T1D has been supported by studies showing an immunological link between the pancreas and the gastrointestinal tract. It has been demonstrated that oral antigens are capable of activating antigen-specific T cells in pancreatic lymph nodes (6) and that the interaction between endothelium and T cells is controlled by shared homing receptors in inflamed islets and in the gut (7). The development of autoimmune diabetes in animal models is regulated by factors affecting the function of the gut immune system, such as dietary factors and microbial stimuli, which further affect the intestinal mucosal barrier and immune responsiveness (8). The effects of intestinal microbes may not be restricted to barrier mechanisms, but gut microbiota seems to play a key role in the regulation of the T-cell populations in the gut, including regulatory T cells, T helper 1, and T helper 17 cells (9).

Several animal studies indicate that alterations in the intestinal microbiota are associated with the development of autoimmune diabetes. Nonobese diabetic mice lacking MyD88, an essential signal transducer in Toll-like receptor signaling, did not have development of diabetes (10), which emphasizes the role of intestinal microbiota as a regulator of autoimmune diabetes. There are differences in the gut microbiota between bio-breeding (BB) diabetes-prone (DP), and diabetes-resistant rats before the diagnosis of diabetes. Antibiotics also can prevent autoimmune diabetes in BB-DP rats (11). Furthermore, it has been reported that stool from BB diabetes-resistant rats contained more probiotic-like bacteria, whereas Bacteroides, Eubacterium, and Ruminococcus were more prevalent in BB-DP rats (12). Lactobacillus johnsonii prevented diabetes when administered to BB-DP rats (13). There are only a few studies of the intestinal microbiota in relation to T1D in humans, but the results of a follow-up study including four children with development of T1D suggested that the Bacteroidetes-to-Firmicutes ratio increased over time in those children with eventual progression to clinical T1D, whereas it decreased in children who remained nondiabetic (14).

The aim of this study was to compare the composition of the gut microbiota between children with β-cell autoimmunity and autoantibody-negative children matched for age, sex, HLA risk genotype, and early feeding history using pyrosequencing as the method of choice.

The current study included 18 children with HLA-conferred susceptibility to T1D who had development of signs of progressive β-cell autoimmunity, i.e., tested positive for at least two diabetes-associated autoantibodies (cases). Eighteen control children were matched for age, sex, and HLA-DQB1 genotype, as well as for the time of exposure to and the type of infant formula. The characteristics of the children recruited to the gut microbiota study are shown in Table 1 and are shown in detail in Supplementary Table 1. The study subjects were recruited from the study population of two intervention trials performed in Finland: the Trial to Reduce IDDM in the Genetically at Risk (TRIGR) pilot (n = 20) or the Finnish Dietary Intervention Trial for Prevention of Type 1 Diabetes (FINDIA) pilot study (n = 16) (3,15). In the TRIGR pilot study, autoantibody positivity was monitored until the age of 10 years and in the ongoing FINDIA pilot study and follow-up time for autoantibodies varied from 3 to 6 years. Eighteen of 26 children who had development of at least two autoantibodies in these intervention studies but who had not yet had progression to overt T1D participated in the current study. The fecal samples from the study subjects recruited from the FINDIA and TRIGR studies were similarly collected between March 2009 and February 2010. Fecal samples from children were collected using stool collection vials and immediately stored in home freezers (−20°C). Families delivered the frozen sample to the study center, and the sample was stored at −80°C until processing. The fecal samples were collected at a point in time when the study subjects did not have gastroenteritis and had not received any antibiotic treatment during the past 3 months. Four children had developed T1D after the fecal samples were collected. The control children remained negative for T1D and for all four autoantibodies analyzed. The study was approved by the ethics committees of the participating hospitals and the families gave their written informed consent.

TABLE 1

Characteristics of the study subjects

Characteristics of the study subjects
Characteristics of the study subjects

In the TRIGR pilot study, infants with a first-degree relative affected by T1D were randomized to receive either a regular cow’s milk formula (Enfamil; Mead Johnson, Evansville, IN) or an extensively hydrolyzed casein-based test formula (Nutramigen; Mead Johnson) until the age of 6–8 months. In the FINDIA study, infants were randomized to receive a standard cow’s milk–based formula (Tutteli; Valio, Helsinki, Finland), a whey-based hydrolyzed formula (Peptidi-Tutteli), or a whey-based FINDIA formula from which bovine insulin was removed. In both studies exclusive breastfeeding was encouraged.

Autoantibody assays.

Insulin autoantibodies, autoantibodies against the 65-kDa isoform of glutamic acid decarboxylase, and autoantibodies against the protein tyrosinase phosphatase–related IA-2 molecule (IA-2A) were measured by specific radiobinding assays, and islet cell antibodies were measured by a standard immunofluorescence assay as described previously (16). Six out of 18 cases tested positive for four autoantibodies, seven were positive for three autoantibodies, and five were positive for two autoantibodies (Table 1).

HLA genotyping.

HLA genotyping was performed according to the screening protocol in the TRIGR and FINDIA studies. The initial HLA-DQB1 typing for risk-associated (DQB1*02, DQB1*03:02) and protective (DQB1*03:01, DQB1*06:02, and DQB1*06:03) alleles was complemented with DQA1 typing for DQA1*02:01 and DQA1*05 alleles in those with DQB1*02 without protective alleles or the major risk allele DQB1*03:02. This two-step screening technique is based on the hybridization of PCR products with lanthanide-labeled probes detected by time-resolved fluorometry as described previously (17,18).

DNA extraction.

Total DNA was extracted from 0.25 g fecal sample using the repeated bead beating method described in detail by Yu et al. (19), with a number of modifications. In brief, four 3-mm glass beads were added during the homogenization step, whereas 0.5-mm glass beads were not used at all. Bead beating was performed using a Precellys 24 (Bertin Technologies, Montigny le Bretonneux, France) at 5.5 ms−1 in three rounds of 1 min each with 30-s pauses at room temperature in between. The incubation temperature after the bead beating was increased from 70°C to 95°C. Importantly, protein precipitation with 260 μL ammonium acetate was performed twice. Elution of DNA from the purification columns was performed twice. Columns from the QiaAmp Stool kit were replaced by those from the QIAamp DNA Mini Kit (Qiagen, Hilden, Germany).

Pyrosequencing.

From each sample, the 16S rRNA genes were amplified using a primer set corresponding to primers 27F-degS (20) and 534-R (21). These PCR primers target the V1, V2, and V3 hypervariable regions of the 16 S rRNA; 27-degS was chosen in particular because it appears to provide a more complete assessment of actinobacterial abundance (20). Pyrosequencing was performed using a Roche FLX Genome Sequencer at DNAvision (Liège, Belgium) using their standard protocol (22).

Sequencing quality control.

Pyrosequencing produced a total of 461,874 reads of 16S rDNA. Sequences were assigned to samples according to sample-specific barcodes. Using the Galaxy Tools web site (23), SFF files from the 454 Genome Sequencer FLX were converted into FASTA files and FASTA quality files. FASTA formatted files contained an average (± SD) of 12,830 ± 4,888 reads per sample. The RDP pyrosequencing pipeline (24) (RDP 10 database, update 17) was subsequently used to check the FASTA sequence files for the same criteria as described by De Filippo et al. (22) and to check that the average experimental quality score was at least 20. After this quality check, the FASTA files contained an average (± SD) of 8,024 ± 3,136 high-quality reads.

Classification.

Taxonomy (phylum, family, and genus level) was assigned using RDP classifier 2.01 (25). Richness and diversity analyses were performed as described by De Filippo et al. (22). Identification to the species level was performed using ARB software (26). For this, SSU reference database (SSURef_106_ SILVA_19_03_11) was downloaded from the SILVA web site (27). From this database, only sequences of cultured and identified isolates were used. From these sequences, a “PT-server” database was built, which was subsequently used to find the closest match for each of our high-quality sequences imported from the FASTA files. For this, the “search next relatives of listed species in PT-server” function was applied with the following settings: oligo length, 12; mismatches, 0; match score, relative; and minimum score, 10. The average (± SD) match score was 75.2 ± 18.5. Sequences that were identified as being from different strains but belonging to the same species were grouped together. Species that represented >0.005% of all sequences were taken along for statistical analysis, together representing 99.2 ± 0.35% of all high-quality FASTA reads per sample.

Assay of IgA and calprotectin in stool samples.

A beaker in the bottom cap of the extraction device was filled with thawed and homogenized stool sample, avoiding seeds and grains (∼100 mg). The extraction tube was filled with 4.9 mL extraction buffer and vortexed for 30 s. Mixing was continued in a shaker at 1,000 rpm for 3 min. The particles were allowed to settle before 10-min centrifugation at 10,000g at room temperature. Supernatant was collected and stored at −20°C.

IgA values were measured with the modified ELISA method described by Lehtonen et al. (28). Calprotectin levels in stool samples were determined using Calprolab calprotectin ELISA tests (Lot CALP-Pilot3; Calpro AS, Lysaker, Norway); 20 µL extract was mixed with 980 µL of sample dilution buffer.

Statistical analysis.

Principal component analysis (PCA) was performed to find clusters of similar groups of samples or species. PCA is an ordination method based on multivariate statistical analysis that maps the samples in different dimensions. All tests were performed with PASW Statistics 18 (SPSS, Chicago, IL). Initial analysis of the samples showed that bacterial populations were most often not normally distributed. Mann-Whitney U and Spearman ρ and χ2 tests were used. All tests were two-tailed. P < 0.05 was considered to indicate statistical significance.

Comparison of sequence diversity at phylum, family, and genus levels.

The analysis of the high-quality reads of all samples at the phylum level showed that Firmicutes (58.1%), Actinobacteria (36.2%), and Bacteroidetes (3.4%) were the most dominant phyla. On the family level, the Bifidobacteriaceae (32.8%; Actinobacteria), the Lachnospiraceae (18.4%; Firmicutes), and Ruminococcaceae (17.1%; Firmicutes) were the most common. On the genus level, the Bifidobacterium genus was the most frequent (34.2%). The most interesting finding was that the Bacteroidetes phylum, the Bacteroidaceae family (2.5%), and the Bacteroides genus (3.1%) were more common in autoantibody-positive children than in autoantibody-negative peers (4.6 vs. 2.2%, 3.5 vs. 1.5%, and 4.3 vs. 2.0%, respectively; P = 0.035, 0.022, and 0.031, respectively; Mann-Whitney U test).

Species level PCA.

PCA analysis of the species level revealed various correlations with β-cell autoimmunity. The first principal component (PC1) (46.0%) correlated positively with a number of important short-chain fatty acid–producing species (P ≤ 0.01 for all species; Spearman ρ test), such as Bifidobacterium adolescentis (11%), Faecalibacterium prausnitzii (5.6%), Clostridium clostridioforme (2.3%), and Roseburia faecis (0.94%), as shown in Fig. 1. This PC1 was inversely related to the number of diabetes-associated autoantibodies in children (Fig. 1; P = 0.018; Spearman ρ test). The children with four autoantibodies had a significantly lower PC1 score, i.e., they had lower numbers of short-chain fatty acid producers than the control children (P = 0.008; Mann-Whitney U test). PC1 score was especially lower in IA-2A–positive children than in children negative for IA-2A (P = 0.004; Mann-Whitney U test).

FIG. 1.

The association of PC1 with autoantibody positivity was demonstrated as an inverse correlation between PC1 and number of β-cell autoantibodies in the study cohort (P = 0.018; Spearman ρ test) and as a difference in PC1 score between the children positive for four autoantibodies and control children negative for autoantibodies (P = 0.008; Mann-Whitney test). Control children are indicated by being positive for no autoantibodies and cases are positive for two, three, or four autoantibodies (x-axis). Correlations between PC1 and bacterial groups are shown in right panel.

FIG. 1.

The association of PC1 with autoantibody positivity was demonstrated as an inverse correlation between PC1 and number of β-cell autoantibodies in the study cohort (P = 0.018; Spearman ρ test) and as a difference in PC1 score between the children positive for four autoantibodies and control children negative for autoantibodies (P = 0.008; Mann-Whitney test). Control children are indicated by being positive for no autoantibodies and cases are positive for two, three, or four autoantibodies (x-axis). Correlations between PC1 and bacterial groups are shown in right panel.

Close modal

A remarkable chevron-like distribution of dots in the second PC (PC2) and third PC (PC3) provided another important observation illustrated in Fig. 2. PC2 showed an inverse correlation with the abundance of B. adolescentis, but a positive correlation with the abundance of Bifidobacterium pseudocatenulatum (P = 1 × 10−9 and P = 1 × 10−6; Spearman ρ test). B. adolescentis and B. pseudocatenulatum represented the two most commonly identified species (11.0 and 9.1%, respectively). The PC2 score differed significantly between the children from the TRIGR study representing older children and the FINDIA children who were younger (P = 0.001; Mann Whitney U test). B. adolescentis was the most common species (15.8%) among the children in the TRIGR pilot study, and their samples were clustered in the left leg of the chevron-like distribution, whereas B. pseudocatenulatum was most frequent (15.8%) in the children from the FINDIA study, and their samples were clustered in the right leg. PC3 was inversely related to the abundance of both B. adolescentis and B. pseudocatenulatum (P = 0.023 and P = 0.002; Spearman ρ test), and it is characterized by the sum of B. adolescentis and B. pseudocatenulatum (P = 1 × 10−18; Spearman ρ test). Various bacteria were inversely associated with the combined count of these two bifidobacteria, especially the members of the Clostridium cluster XI (P = 6 × 10−4).The apex, encompassed by a circle in Fig. 2, can be described as comprising those samples in which the combined abundance of B. adolescentis and B. pseudocatenulatum is <12%. The children with β-cell autoimmunity were over-represented in the apex when compared with control children (10/18 vs. 4/18; P = 0.040; χ2 test).

FIG. 2.

The distribution of children positive for β-cell autoantibodies (case subjects with open symbols) and autoantibody-negative children (control subjects with filled symbols) according to PC2 (x-axis) and PC3 (y-axis). PC2 shows an inverse correlation with the abundance of B. adolescentis and a positive correlation with B. pseudocatenulatum, whereas PC3 shows an inverse correlation with both B. adolescentis and B. pseudocatenulatum. Indicated percentages indicate the prevalence of B. adolescentis and B. pseudocatenulatum, respectively. Age of the children is associated with PC2, whereas autoantibody positivity is associated with PC3. The children 3 to 7 years of age (circle) from the FINDIA pilot have higher numbers of B. pseudocatenulatum and lower numbers of B. adolescentis than the children from the TRIGR pilot 11 to 14 years of age (square). The main correlations with PC2 and PC3 are depicted with vectors in the top left. Children with a dearth of both B. adolescentis and B. pseudocatenulatum (sum <12%) are encompassed by the dashed circle near the top and the number of autoantibody-positive children is higher than the number of control subjects at the apex (10/18 vs. 4/18; P = 0.040; χ2 test).

FIG. 2.

The distribution of children positive for β-cell autoantibodies (case subjects with open symbols) and autoantibody-negative children (control subjects with filled symbols) according to PC2 (x-axis) and PC3 (y-axis). PC2 shows an inverse correlation with the abundance of B. adolescentis and a positive correlation with B. pseudocatenulatum, whereas PC3 shows an inverse correlation with both B. adolescentis and B. pseudocatenulatum. Indicated percentages indicate the prevalence of B. adolescentis and B. pseudocatenulatum, respectively. Age of the children is associated with PC2, whereas autoantibody positivity is associated with PC3. The children 3 to 7 years of age (circle) from the FINDIA pilot have higher numbers of B. pseudocatenulatum and lower numbers of B. adolescentis than the children from the TRIGR pilot 11 to 14 years of age (square). The main correlations with PC2 and PC3 are depicted with vectors in the top left. Children with a dearth of both B. adolescentis and B. pseudocatenulatum (sum <12%) are encompassed by the dashed circle near the top and the number of autoantibody-positive children is higher than the number of control subjects at the apex (10/18 vs. 4/18; P = 0.040; χ2 test).

Close modal

Species-level analysis.

The association of single bacterial species with autoantibody positivity is shown in Table 2. Roseburia faecis (0.94%) was more abundant in autoantibody-negative than autoantibody-positive children (P = 0.009; Mann-Whitney U test), whereas Clostridium perfringens (0.03%) were more abundant in children with β-cell autoimmunity than in those without (P = 0.18; Mann-Whitney U test). The Bacteroides genus was associated with autoantibody positivity, but only a few of its many members were observed to be related on the species level. Interestingly, the correlation of the Bacteroides genus with autoantibody positivity was significant only in males, not in females (P = 0.005 vs. 0.655, respectively).

TABLE 2

Association between single bacterial species and signs of β-cell autoimmunity

Association between single bacterial species and signs of β-cell autoimmunity
Association between single bacterial species and signs of β-cell autoimmunity

In children from the FINDIA study, it is furthermore noteworthy that Eubacterium hallii (6.0%) was strongly inversely related to the number of β-cell autoantibodies (P = 0.002; Spearman ρ correlation test).

IgA and calprotectin levels.

Calprotectin levels, a marker of intestinal inflammation, were not observed to be correlated with autoantibody positivity. Clostridium orbiscindens (0.47%) correlated inversely with calprotectin levels (P = 0.004; Spearman ρ test). Lower levels of calprotectin were seen in children with the lower-risk HLA genotype, i.e., (DR3)DQB1*02-DQA1*05, when compared with the children with moderate-risk or high-risk HLA risk genotypes (P = 0.001 and 0.005; Mann-Whitney U test). Fecal IgA levels did not differ between autoantibody-positive and autoantibody-negative children but were inversely correlated with age (P = 0.007; Mann-Whitney U test).

Diversity analyses.

The analysis of the 90, 95, and 97% operational taxonomic similarity levels showed that the diversity was significantly higher in the children in the TRIGR pilot cohort than in the children in the FINDIA pilot cohort (P < 0.001; Mann-Whitney U test). Furthermore, there was a trend that the diversity per sample was higher in autoantibody-negative children than in the autoantibody-positive children (Supplementary Table 2). This difference in diversity, measured in the number of observed operational taxonomic units and the Chao index, was significantly different at the 95% taxonomic similarity level when comparing the autoantibody-positive children and their corresponding control subjects in the TRIGR cohort (P = 0.028 and 0.034, respectively; Mann-Whitney U test).

We observed significant differences in the composition of fecal microbiota between children positive for at least two diabetes-associated autoantibodies and autoantibody-negative children. Because the major histocompatability complex genotype may affect the intestinal microbiota composition (29), we matched the case and control subjects not only for age, sex, and early feeding history but also for their HLA class II risk genotype of T1D. Our results thus should not be secondary to HLA-related differences in gut microbiota.

We found that the score reflecting the abundance of several lactate- and butyrate-producing bacteria, i.e., PC1, was inversely related to the number of β-cell autoantibodies in children; the lowest levels were observed in children positive for three or four autoantibodies. The bacteria, which correlated with PC1 included B. adolescentis, an acetate- and lactate-producing bacterium, R. faecis, a member from Clostridium cluster XIVa, which produces butyrate using acetate (30), and F. prausnitzii, a member from Clostridium cluster IV, an acetate utilizing butyrate-producing bacterium (30,31) with anti-inflammatory properties (32). In addition to this, we found that in children from the FINDIA study, E. hallii, an acetate- and lactate-utilizing and butyrate-producing bacterium from Clostridium cluster XIVa (31), was inversely related to the number of β-cell autoantibodies. A recent study that included four pairs of cases with development of T1D and autoantibody-negative control subjects suggested that higher proportions of butyrate-producing and mucin-degrading bacteria were observed in control subjects compared with case subjects (33). Because lactate can be further metabolized to butyrate, the producers of lactate may contribute to the net production of butyrate; bifidobacteria represents a prime example of this (31). PCA analysis revealed that a low abundance (<12%) of the sum of the two most common bifidobacteria, B. adolescentis and B. pseudocatenulatum, was associated with β-cell autoimmunity.

Butyrate is thought to be beneficial because it is the main energy source for colonic epithelial cells (34). Furthermore, butyrate has been shown to regulate the assembly of tight junctions and gut permeability (35). In animal models of autoimmune diabetes, increased gut permeability precedes the development of diabetes, and environmental factors, which modulate the permeability, modulate the disease incidence. Although butyrate treatment during the weaning period in BB-DP rats did not prevent autoimmune diabetes, it modulated the gut inflammatory response (36). Both gut permeability and inflammation have been linked to the development of T1D in humans (37). In children with T1D, subclinical small intestinal inflammation with T-cell activation has been reported previously (37), but there are no studies of children at risk for T1D, i.e., autoantibody-positive children. We did not find increased fecal calprotectin or IgA levels in β-cell autoantibody–positive children, which does not support the presence of significant intestinal inflammation in preclinical T1D, although differences in the composition of fecal microbiota were observed.

Our results suggest that the changes characterized by low numbers of butyrate producers are related to the late phase of prediabetes, i.e., positivity for multiple autoantibodies and IA-2A. The correlation of certain bacterial findings with the number of positive autoantibodies could indicate a role of dysbiosis as a regulator of β-cell autoimmunity in the progression of the autoimmune process toward β-cell destruction and clinical disease. It should be emphasized that our findings demonstrate only microbial changes and functional studies are needed to prove causality between these kinds of changes and β-cell autoimmunity. The results thus are tentative and support the animal studies in which causality has been demonstrated.

There is increasing evidence that the Bifidobacterium genus in the human gut plays an important role in maintaining health, both within the gastrointestinal tract and in the rest of the body (38), and this also is supported by the current observations, although we did not provide any data on functional changes related to the observed bacterial diversity. It has been shown that besides contributing to the production of butyrate, bifidobacteria inhibit bacterial translocation (3941). In contrast, bacterial translocation was found to be enhanced by the Bacteroides genus and, in particular, the Bacteroides fragilis group, which includes B. fragilis, B. ovatus, and B. vulgatus (40). A role of the Bacteroides genus in the development of autoimmune diabetes has been implicated in animal models and in humans (11,12,14,33). In our study, the Bacteroides genus and C. perfringens, which is known to be associated with increased gut permeability and inflammation via its production of several toxins (42), were positively associated with β-cell autoimmunity. Bifidobacteria, however, might inhibit the translocation or the growth of the B. fragilis group and C. perfringens as they compete for space/adherence (43) and nutrients (44), and they enhance the intestinal epithelial barrier function (45) by increasing the thickness of the mucus layer (46,47). Accordingly, our findings actually may be interrelated because a low abundance of bifidobacteria might favor the growth of Bacteroides (40).

Concerning the biological importance of the Bacteroides genus, it should be noted that its abundance as reported here obtained via pyrosequencing (∼3%) greatly underestimates its actual predominance within the gut of these children. The reason for this is that the primer pair used in this study, which was chosen because it provides a more complete assessment of actinobacterial abundance (20), very likely caused an overestimation of the amount of the Bifidobacterium genus and possibly caused an underestimation of the Bacteroides genus. We confirmed our suspicions using fluorescent in situ hybridization (data not shown); the Bifidobacterium genus had been overestimated by a factor of two and the Bacteroides genus was underestimated by a factor of six.

In agreement with a previous study in which the diversity decreased with increasing age in the four children with development of T1D (14), the microbial diversity was lower in autoantibody-positive children when compared with autoantibody-negative children, especially in the children aged 12 to 14 years from the TRIGR pilot study. The samples from the children of the TRIGR and FINDIA studies were collected at the same time using the same protocol for transportation and storage. Thus, the findings are not influenced by differences in the storage time or transportation history of the samples.

In conclusion, our study emphasizes the importance of bifidobacteria and lactate- or butyrate-producing species in general in relation to the development of β-cell autoimmunity. Bifidobacteria not only supply butyrate-producing species with lactate and acetate but also enhance the intestinal epithelial barrier function by modulating the gut mucosa, thereby possibly preventing the translocation of for example Bacteroides, which were, in turn, confirmed to be more highly abundant in children with β-cell autoimmunity. Because only children with at least two autoantibodies were included in this study, it was not possible to demonstrate that dysbiosis has a role in the initiation of β-cell autoimmunity. The pyrosequencing method does have its limitations because it does not give direct data on the functionally important changes of the microbiota, which should be demonstrated if pathogenetic importance is considered. One also should be careful with making conclusions based on PCA of the data on the bacterial diversity. Our results only indicate that alterations in the intestinal bacterial diversity are seen at the prediabetes stage of the disease, and these changes precede the development of T1D. Large birth cohort studies that include time series of samples and methodology to study the functional changes in the intestinal microbiota are needed for the evaluation of the role of gut microbiota in the development of β-cell autoimmunity and T1D.

Clinical trial reg. nos. NCT00570102 and NCT01055080, clinicaltrials.gov.

Supported by Sigrid Juselius Foundation and the Päivikki and Sakari Sohlberg Foundation.

No potential conflicts of interest relevant to this article were reported.

M.C.d.G. and K.L. analyzed the data. M.C.d.G., K.L., and O.V. wrote the manuscript. M.C.d.G., G.W.W., and H.J.H. were responsible for the microbiological studies. K.L. and T.R. coordinated the study recruitment and sample collection. M.K. contributed to the study recruitment and edited the manuscript. M.K. and T.H. were responsible for the autoantibody analyses. J.I. was responsible for HLA typing. L.O. performed the fecal IgA determinations. S.H. performed the fecal calprotectin tests. O.V. was responsible for the study design. O.V. is the guarantor of this work and, as such, had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

1.
Eisenbarth
GS
.
Type I diabetes mellitus. A chronic autoimmune disease
.
N Engl J Med
1986
;
314
:
1360
1368
[PubMed]
2.
Knip
M
,
Siljander
H
.
Autoimmune mechanisms in type 1 diabetes
.
Autoimmun Rev
2008
;
7
:
550
557
[PubMed]
3.
Knip
M
,
Virtanen
SM
,
Seppä
K
, et al
Finnish TRIGR Study Group
.
Dietary intervention in infancy and later signs of beta-cell autoimmunity
.
N Engl J Med
2010
;
363
:
1900
1908
[PubMed]
4.
Norris
JM
,
Barriga
K
,
Klingensmith
G
, et al
.
Timing of initial cereal exposure in infancy and risk of islet autoimmunity
.
JAMA
2003
;
290
:
1713
1720
[PubMed]
5.
Vaarala
O
,
Knip
M
,
Paronen
J
, et al
.
Cow’s milk formula feeding induces primary immunization to insulin in infants at genetic risk for type 1 diabetes
.
Diabetes
1999
;
48
:
1389
1394
[PubMed]
6.
Turley
SJ
,
Lee
JW
,
Dutton-Swain
N
,
Mathis
D
,
Benoist
C
.
Endocrine self and gut non-self intersect in the pancreatic lymph nodes
.
Proc Natl Acad Sci USA
2005
;
102
:
17729
17733
[PubMed]
7.
Hänninen
A
,
Nurmela
R
,
Maksimow
M
,
Heino
J
,
Jalkanen
S
,
Kurts
C
.
Islet beta-cell-specific T cells can use different homing mechanisms to infiltrate and destroy pancreatic islets
.
Am J Pathol
2007
;
170
:
240
250
[PubMed]
8.
Vaarala
O
.
The gut as a regulator of early inflammation in type 1 diabetes
.
Curr Opin Endocrinol Diabetes Obes
2011
;
18
:
241
247
[PubMed]
9.
Gaboriau-Routhiau
V
,
Rakotobe
S
,
Lécuyer
E
, et al
.
The key role of segmented filamentous bacteria in the coordinated maturation of gut helper T cell responses
.
Immunity
2009
;
31
:
677
689
[PubMed]
10.
Wen
L
,
Ley
RE
,
Volchkov
PY
, et al
.
Innate immunity and intestinal microbiota in the development of Type 1 diabetes
.
Nature
2008
;
455
:
1109
1113
[PubMed]
11.
Brugman
S
,
Klatter
FA
,
Visser
JT
, et al
.
Antibiotic treatment partially protects against type 1 diabetes in the Bio-Breeding diabetes-prone rat. Is the gut flora involved in the development of type 1 diabetes?
Diabetologia
2006
;
49
:
2105
2108
[PubMed]
12.
Roesch
LF
,
Lorca
GL
,
Casella
G
, et al
.
Culture-independent identification of gut bacteria correlated with the onset of diabetes in a rat model
.
ISME J
2009
;
3
:
536
548
[PubMed]
13.
Valladares
R
,
Sankar
D
,
Li
N
, et al
.
Lactobacillus johnsonii N6.2 mitigates the development of type 1 diabetes in BB-DP rats
.
PLoS ONE
2010
;
5
:
e10507
[PubMed]
14.
Giongo
A
,
Gano
KA
,
Crabb
DB
, et al
.
Toward defining the autoimmune microbiome for type 1 diabetes
.
ISME J
2011
;
5
:
82
91
[PubMed]
15.
Vaarala
O
,
Ilonen
J
,
Ruohtula
T
, et al
.
Removal of bovine insulin from cow's milk formula and early initiation of beta-cell autoimmunity in the FINDIA Pilot Study
.
Arch Pediatr Adolesc Med
2012
;PMID:22393174
16.
Kukko
M
,
Virtanen
SM
,
Toivonen
A
, et al
.
Geographical variation in risk HLA-DQB1 genotypes for type 1 diabetes and signs of beta-cell autoimmunity in a high-incidence country
.
Diabetes Care
2004
;
27
:
676
681
[PubMed]
17.
Laaksonen
M
,
Pastinen
T
,
Sjöroos
M
, et al
.
HLA class II associated risk and protection against multiple sclerosis-a Finnish family study
.
J Neuroimmunol
2002
;
122
:
140
145
[PubMed]
18.
Sjöroos
M
,
Iitiä
A
,
Ilonen
J
,
Reijonen
H
,
Lövgren
T
.
Triple-label hybridization assay for type-1 diabetes-related HLA alleles
.
Biotechniques
1995
;
18
:
870
877
[PubMed]
19.
Yu
Z
,
Morrison
M
.
Improved extraction of PCR-quality community DNA from digesta and fecal samples
.
Biotechniques
2004
;
36
:
808
812
[PubMed]
20.
van den Bogert
B
,
de Vos
WM
,
Zoetendal
EG
,
Kleerebezem
M
.
Microarray analysis and barcoded pyrosequencing provide consistent microbial profiles depending on the source of human intestinal samples
.
Appl Environ Microbiol
2011
;
77
:
2071
2080
[PubMed]
21.
Wu GD, Lewis JD, Hoffmann C, et al. Sampling and pyrosequencing methods for characterizing bacterial communities in the human gut using 16S sequence tags. BMC Microbiol 2010;10:206
22.
De Filippo
C
,
Cavalieri
D
,
Di Paola
M
, et al
.
Impact of diet in shaping gut microbiota revealed by a comparative study in children from Europe and rural Africa
.
Proc Natl Acad Sci USA
2010
;
107
:
14691
14696
[PubMed]
23.
Goecks
J
,
Nekrutenko
A
,
Taylor
J
Galaxy Team
.
Galaxy: a comprehensive approach for supporting accessible, reproducible, and transparent computational research in the life sciences
.
Genome Biol
2010
;
11
:
R86
[PubMed]
24.
Cole
JR
,
Wang
Q
,
Cardenas
E
, et al
.
The Ribosomal Database Project: improved alignments and new tools for rRNA analysis
.
Nucleic Acids Res
2009
;
37
(
Database issue
):
D141
D145
[PubMed]
25.
Wang
Q
,
Garrity
GM
,
Tiedje
JM
,
Cole
JR
.
Naive Bayesian classifier for rapid assignment of rRNA sequences into the new bacterial taxonomy
.
Appl Environ Microbiol
2007
;
73
:
5261
5267
[PubMed]
26.
Ludwig
W
,
Strunk
O
,
Westram
R
, et al
.
ARB: a software environment for sequence data
.
Nucleic Acids Res
2004
;
32
:
1363
1371
[PubMed]
27.
Pruesse
E
,
Quast
C
,
Knittel
K
, et al
.
SILVA: a comprehensive online resource for quality checked and aligned ribosomal RNA sequence data compatible with ARB
.
Nucleic Acids Res
2007
;
35
:
7188
7196
[PubMed]
28.
Lehtonen
OP
,
Gråhn
EM
,
Ståhlberg
TH
,
Laitinen
LA
.
Amount and avidity of salivary and serum antibodies against Streptococcus mutans in two groups of human subjects with different dental caries susceptibility
.
Infect Immun
1984
;
43
:
308
313
[PubMed]
29.
Toivanen
P
,
Vaahtovuo
J
,
Eerola
E
.
Influence of major histocompatibility complex on bacterial composition of fecal flora
.
Infect Immun
2001
;
69
:
2372
2377
[PubMed]
30.
Duncan
SH
,
Barcenilla
A
,
Stewart
CS
,
Pryde
SE
,
Flint
HJ
.
Acetate utilization and butyryl coenzyme A (CoA):acetate-CoA transferase in butyrate-producing bacteria from the human large intestine
.
Appl Environ Microbiol
2002
;
68
:
5186
5190
[PubMed]
31.
Duncan
SH
,
Louis
P
,
Flint
HJ
.
Lactate-utilizing bacteria, isolated from human feces, that produce butyrate as a major fermentation product
.
Appl Environ Microbiol
2004
;
70
:
5810
5817
[PubMed]
32.
Sokol
H
,
Pigneur
B
,
Watterlot
L
, et al
.
Faecalibacterium prausnitzii is an anti-inflammatory commensal bacterium identified by gut microbiota analysis of Crohn disease patients
.
Proc Natl Acad Sci USA
2008
;
105
:
16731
16736
[PubMed]
33.
Brown
CT
,
Davis-Richardson
AG
,
Giongo
A
, et al
.
Gut microbiome metagenomics analysis suggests a functional model for the development of autoimmunity for type 1 diabetes
.
PLoS ONE
2011
;
6
:
e25792
[PubMed]
34.
Hague
A
,
Butt
AJ
,
Paraskeva
C
.
The role of butyrate in human colonic epithelial cells: an energy source or inducer of differentiation and apoptosis?
Proc Nutr Soc
1996
;
55
:
937
943
[PubMed]
35.
Peng
L
,
Li
ZR
,
Green
RS
,
Holzman
IR
,
Lin
J
.
Butyrate enhances the intestinal barrier by facilitating tight junction assembly via activation of AMP-activated protein kinase in Caco-2 cell monolayers
.
J Nutr
2009
;
139
:
1619
1625
[PubMed]
36.
Li
N
,
Hatch
M
,
Wasserfall
CH
, et al
.
Butyrate and type 1 diabetes mellitus: can we fix the intestinal leak?
J Pediatr Gastroenterol Nutr
2010
;
51
:
414
417
[PubMed]
37.
Vaarala
O
,
Atkinson
MA
,
Neu
J
.
The “perfect storm” for type 1 diabetes: the complex interplay between intestinal microbiota, gut permeability, and mucosal immunity
.
Diabetes
2008
;
57
:
2555
2562
[PubMed]
38.
Mitsuoka
T
.
Bifidobacteria and their role in human health
.
J Ind Microbiol Biotechnol
1990
;
6
:
263
267
39.
Duffy
LC
.
Interactions mediating bacterial translocation in the immature intestine
.
J Nutr
2000
;
130
(
Suppl
):
432S
436S
[PubMed]
40.
Romond
MB
,
Colavizza
M
,
Mullié
C
, et al
.
Does the intestinal bifidobacterial colonisation affect bacterial translocation?
Anaerobe
2008
;
14
:
43
48
[PubMed]
41.
Wang
ZT
,
Yao
YM
,
Xiao
GX
,
Sheng
ZY
.
Risk factors of development of gut-derived bacterial translocation in thermally injured rats
.
World J Gastroenterol
2004
;
10
:
1619
1624
[PubMed]
42.
Smedley
JG
 3rd
,
Fisher
DJ
,
Sayeed
S
,
Chakrabarti
G
,
McClane
BA
.
The enteric toxins of Clostridium perfringens
.
Rev Physiol Biochem Pharmacol
2004
;
152
:
183
204
[PubMed]
43.
Stecher
B
,
Hardt
WD
.
The role of microbiota in infectious disease
.
Trends Microbiol
2008
;
16
:
107
114
[PubMed]
44.
Gibson
GR
,
Willems
A
,
Reading
S
,
Collins
MD
.
Fermentation of non-digestible oligosaccharides by human colonic bacteria
.
Proc Nutr Soc
1996
;
55
:
899
912
[PubMed]
45.
Liévin
V
,
Peiffer
I
,
Hudault
S
, et al
.
Bifidobacterium strains from resident infant human gastrointestinal microflora exert antimicrobial activity
.
Gut
2000
;
47
:
646
652
[PubMed]
46.
Kleessen
B
,
Hartmann
L
,
Blaut
M
.
Fructans in the diet cause alterations of intestinal mucosal architecture, released mucins and mucosa-associated bifidobacteria in gnotobiotic rats
.
Br J Nutr
2003
;
89
:
597
606
[PubMed]
47.
Kleessen
B
,
Blaut
M
.
Modulation of gut mucosal biofilms
.
Br J Nutr
2005
;
93
(
Suppl 1
):
S35
S40

Supplementary data