sign in sign up
<<
Home , Anaerostipes, Bacteroidaceae, Insulitis, Intestinal permeability, Megasphaera, Prevotella

Diabetes Care Volume 41, November 2018 2385

Gut Microbiota Differs in Isabel Leiva-Gea,1 Lidia Sanchez-Alcoholado,´ 2 Composition and Functionality Beatriz Mart´ın-Tejedor,1 Daniel Castellano-Castillo,2,3 Between Children With Type 1 Isabel Moreno-Indias,2,3 Antonio Urda-Cardona,1 and MODY2 and Healthy Francisco J. Tinahones,2,3 Jose´ Carlos Fernandez-Garc´ ´ıa,2,3 and Control Subjects: A Case-Control Mar´ıa Isabel Queipo-Ortuno~ 2,3 Study

Diabetes Care 2018;41:2385–2395 | https://doi.org/10.2337/dc18-0253

OBJECTIVE is associated with compositional differences in . To date, no microbiome studies have been performed in maturity-onset diabetes of the young 2 (MODY2), a monogenic cause of diabetes. Gut microbiota of type 1 diabetes, MODY2, and healthy control subjects was compared. PATHOPHYSIOLOGY/COMPLICATIONS RESEARCH DESIGN AND METHODS This was a case-control study in 15 children with type 1 diabetes, 15 children with MODY2, and 13 healthy children. Metabolic control and potential factors mod- ifying gut microbiota were controlled. Microbiome composition was determined by 16S rRNA pyrosequencing. 1Pediatric , Hospital Materno- Infantil, Malaga,´ Spain RESULTS 2Clinical Management Unit of Endocrinology and Compared with healthy control subjects, type 1 diabetes was associated with a Nutrition, Laboratory of the Biomedical Research significantly lower microbiota diversity, a significantly higher relative abundance of Institute of Malaga,´ Virgen de la Victoria Uni- Ruminococcus Veillonella Blautia Streptococcus versityHospital,Universidad de Malaga,M´ alaga,´ , , , , and genera, and a Spain lower relative abundance of Bifidobacterium, Roseburia, Faecalibacterium, and 3Centro de Investigacion´ BiomedicaenRed(CIBER)´ Lachnospira. Children with MODY2 showed a significantly higher abun- de Fisiopatolog´ıa de la Obesidad y Nutricion,´ In- dance and a lower Ruminococcus and Bacteroides abundance. Proinflammatory stituto Salud Carlos III, Madrid, Spain cytokines and lipopolysaccharides were increased in type 1 diabetes, and gut per- Corresponding author: Jose´ Carlos Fernandez-´ Garc´ıa, [email protected]. meability (determined by levels) was significantly increased in type 1 Received 5 February 2018 and accepted 26 diabetes and MODY2. The PICRUSt analysis found an increment of genes related to August 2018. lipid and amino acid metabolism, ABC transport, lipopolysaccharide biosynthesis, This article contains Supplementary Data online arachidonic acid metabolism, processing and presentation, and chemokine at http://care.diabetesjournals.org/lookup/suppl/ signaling pathways in type 1 diabetes. doi:10.2337/dc18-0253/-/DC1. I.L.-G. and L.S.-A. contributed equally to this CONCLUSIONS work. Gut microbiota in type 1 diabetes differs at taxonomic and functional levels not only © 2018 by the American Diabetes Association. in comparison with healthy subjects but fundamentally with regard to a model of Readers may use this article as long as the work nonautoimmune diabetes. Future longitudinal studies should be aimed at eval- is properly cited, the use is educational and not for profit, and the work is not altered. More infor- uating if the modulation of gut microbiota in patients with a high risk of type 1 mation is available at http://www.diabetesjournals diabetes could modify the natural history of this autoimmune disease. .org/content/license. 2386 Gut Microbiota in Type 1 Diabetes and MODY2 Diabetes Care Volume 41, November 2018

Type 1 diabetes is a multifactorial immune- caused by a heterozygous inactivating diseases or infectious diseases or un- mediated disease characterized by the mutation in the glucokinase (GCK) gene dergoing treatment with antibiotics, pre- progressive loss of -producing b- (14). To date, no studies have evaluated biotics, or or any other medical cells in the islets of Langerhans in the the gut microbiome structure in MODY2 treatment that could potentially influ- . The causes that lead to the patients. Nevertheless, MODY2 is a highly ence intestinal microbiota 3 months be- appearance of type 1 diabetes have not attractive model to assess the relation of fore inclusion. yet been fully identified, with genetic gut microbiota with type 1 diabetes, as The parents of all the participants factors playing a major role; however, MODY2 is not normally associated with completed a structured interview to ob- environmental factors are also closely obesity, a glycemic control similar to tain health status, lifestyle aspects, and linked, such as birth delivery mode (1), type 1 diabetes can be achieved, and, dietary habits. Patients with type 1 di- diet in early life (cow milk proteins or importantly, its cause is not of autoim- abetes and MODY2 were instructed to -containing cereals) (2), and wide- mune origin (15). follow a standard diabetic diet, contain- spread usage of antibiotics (3), all fac- We hypothesize that if the fecal micro- ing 40–50% of calories from carbohy- tors closely related to gut microbiota. biota in type 1 diabetes differs from that drates, 20–30% from fat, and 20% from Recent studies have associated the of MODY2, the gut microbiota profile protein. Dietary intake patterns were microbiome with the development of could constitute a novel associated risk determined from a food frequency ques- type 1 diabetes; animal studies have factor for the type 1 diabetes autoim- tionnaire. demonstrated a close link between in- mune process. On the contrary, if the The written informed consents of the testinal microbiota and type 1 diabetes fecal microbiota were similar and differ- children’s guardians or parents were in BioBreeding diabetes-prone rats (4) ent from the microbiota of healthy con- obtained. The sampling and experimen- and in nonobese diabetic mice (5). Also, trol subjects, this would indicate that tal processes were performed with the in children with type 1 diabetes, auto- differences in intestinal microbiota could approval of the local Ethics Committee immune positivity has been related to be attributed to per se. of the Regional Hospital of Malaga.´ changes in microbiota composition (6,7). Therefore, the aim of this case-control Laboratory Measurements In a previous study, we found that in study is to evaluate the gut microbiota Serum glucose, cholesterol, triglycerides, comparison with healthy control sub- profile, functional capacity, low-grade and interleukin-1b (IL-1b) and IL-10 cyto- jects, children with type 1 diabetes pre- inflammation, and gut permeability be- kines were measured by ELISA as previ- sented large significant differences in tween patients with type 1 diabetes and ously described (17). Concentrations of IL-6, the relative abundance of predominant MODY2 and healthy control subjects. IL-13, and tumor necrosis factor (TNF)-a phyla, families, and genera (8). Poten- were quantified by ELISA assay kits (Thermo tially, the altered microbiota profile in RESEARCH DESIGN AND METHODS Fisher Scientific) in serum samples ac- type 1 diabetes may be associated with This study was a case-control study, in- cording to the instructions of the man- alterations in the gut immune system, cluding 15 children with type 1 diabetes, ufacturer. The detection limits were as such as increased gut permeability (9). 15 children with MODY2, and 13 healthy follows: 7.8–500 pg/mL for IL-6, 1.6–100 Recent studies have shown that com- control children, all under 18 years old, pg/mL for IL-13, and 15.6–1,000 pg/mL mensal are crucial for the ma- of Caucasian origin and with the same for TNF-a. turing and functioning of the mucosal geographical location. immune system. Moreover, an impaired Type 1 diabetes was diagnosed accord- DNA Extraction, Pyrosequencing of 16S integrity of the intestinal barrier with ing to the criteria of the American Di- rRNA Sequences, and Bioinformatic an increase in permeability has been abetes Association (16) and the positivity Analysis described in both animal models and of at least two persistent, confirmed anti- Study participants collected their fecal human type 1 diabetes studies (10). islet (anti-insulin autoan- samples in a sterile and hermetically Therefore, given that intestinal microbes tibodies, GAD autoantibodies, or tyrosine sealed receptacle provided by the re- may affect intestinal permeability, intes- phosphatase autoantibodies). MODY2 search team. Fecal samples were col- tinal ecology could also play a crucial role children were diagnosed by suggestive lected in the morning of the day of in the development of type 1 diabetes clinical history, negative anti-islet auto- sample receipt and were immediately (11). On the other hand, zonulin, a phys- antibodies, and positive genetic testing. refrigerated in household freezers and iological modulator of intercellular tight Healthy control subjects were children transported to the laboratory during the junctions, increases gut permeability and with negative anti-islet autoantibodies, following 4 h. Frozen fecal samples were macromolecule absorption, and previous matched to children with type 1 diabetes transported with ice to avoid important studies claim a role for zonulin as a capital andMODY2forage,sex,race,BMI,mode changes of temperature that might cause regulator of intestinal barrier function in of delivery, and duration of breastfeed- bacterial DNA degradation and were the genesis of metabolic disorders (12). ing. In addition, patients with type 1 subsequently stored at 280°C in the Maturity-onset diabetes of the young diabetes and MODY2 were controlled laboratory until analysis. No DNA stabi- (MODY) is a genetic form of diabetes that by HbA1c levels. Patients with type 1 lizers were added to the fecal samples. accounts for 1–2% of all diabetes cases diabetes were undergoing treatment DNA was extracted from the fecal in Europe and is associated with specific with multiple doses of insulin, whereas samples using the QIAamp DNA Stool loss-of-function mutations with charac- MODY2 patients were drug naive. Exclu- Mini Kit (Qiagen, Hilden, Germany) ac- teristic phenotypes (13). The most com- sion criteria to participate in this study cording to the protocol of the manufac- mon presentation of MODY is MODY2, included acute or chronic inflammatory turer. Amplification of genomic DNA was care.diabetesjournals.org Leiva-Gea and Associates 2387

performed using bar-coded primers that Mann-Whitney U test, and differences respectively) (Fig. 1A and B). Neverthe- targeted the V2–V3 regions of the bac- among the three groups were ana- less, analysis of similarities with permu- terial 16S rRNA gene. Amplification, se- lyzed using the Kruskal-Wallis test with tations revealed no significant differences quencing,andbasicanalysiswereperformed Bonferroni post hoc test. The Spearman between MODY2 and healthy control using a GS Junior 454 platform according to correlation coefficient was calculated to subjects (P = 0.27), as demonstrated by the protocols of the manufacturer and a estimate the linear correlations between the two principal component scores, which Titanium chemistry apparatus (Roche Ap- variables. A multiple linear regression accounted for 14.50% and 26.34% of plied Science, Indianapolis, IN). The 454 analysis was performed to identify which total variation (Fig. 1C). pyrosequencing data sets were analyzed by bacteria taxa were independent predic- Quantitative Insights into Microbial Ecology tors for serum zonulin, LPS, inflammatory -Based Comparisons of (QIIME) 1.8.0 software as previously de- mediators, and HbA1c levels in each study Fecal Microbiota at the Phylum, Family, scribed (17). PICRUSt analysis was used to group. Values were considered to be and Level predict metagenome function by picking statistically significant when P # 0.05. The dominant phyla of all groups were operational taxonomic units (OTUs) against and , followed the Greengenes database as previously RESULTS by Proteobacteria and Actinobacteria. described (17,18). The statistical analysis Diet and Anthropometric and Individuals with type 1 diabetes showed fi was performed in R 3.3.3. P values were Biochemical Measurements a signi cant increase in the abundance corrected for multiple comparisons using All study participants had a similar physical of Bacteroidetes (64.69% type 1 diabe- the Benjamini-Hochberg method (17). activity and dietary profile. No significant tes, 39.85% MODY2, 49.85% healthy con- differences in the consumption patterns trol subjects; P , 0.001, FDR-adjusted Intestinal Permeability of wheat, rice, vegetables, fish, or meat P , 0.001) and a significant decrease in The plasma level of zonulin was deter- were found between study groups. the abundance of Firmicutes (19.60% mined by ELISA using commercial kits (Im- The main clinical and biochemical char- type 1 diabetes, 31.15% MODY2, 29.69% munodiagnostik AG, Bensheim, Germany). acteristics of the study groups are shown healthy control subjects; P , 0.001, FDR- The detection limit was 0.22 ng/mL. , in Table 1. As expected, glucose and HbA1c adjusted P 0.001) and Actinobacteria fi (1.02% type 1 diabetes, 2.27% MODY2, Limulus Amebocyte Lysate Assays levels were signi cantly higher in chil- Serum concentrations of lipopolysacchar- dren with type 1 diabetes and MODY2 (no 2.82% healthy control subjects; P =0.003, ides (LPS) were measured by endotoxin differences between them) when com- FDR-adjusted P = 0.007) when compared assay, based on a Limulus amebocyte ex- pared with healthy children. No other with MODY2 and healthy control subjects. tract with a chromogenic Limulus amebo- significant differences were observed, in- The frequency of Bacteroidetes was sig- fi cyte lysate assay (QCL-1000; Lonza Group cluding breastfeeding time or mode of ni cantly lower in MODY2 than in healthy , Ltd.) according to the instructions of the delivery. However, children with type 1 control subjects (FDR-adjusted P 0.001). fi manufacturer. The sensitivity limit for the diabetes had significantly higher levels of Proteobacteria was signi cantly lower in assay was 0.02 endotoxin units (EU)/mL. IL-1b, IL-6, TNF-a, and LPS and signifi- type 1 diabetes when compared with cantly lower levels of IL-10 and IL-13 than healthy control subjects (1.68% type 1 Statistical Analysis MODY2 and healthy control subjects. diabetes, 3.06% healthy control subjects; Given the exploratory nature of this study P , 0.001, FDR-adjusted P , 0.001), but (no previous studies evaluating the differ- Characterization of Fecal Microbiota not regarding MODY2 (1.68% type 1 di- ences in gut microbiota between type 1 Pyrosequencing abetes, 1.80% MODY2; P = 0.699, FDR- diabetes, MODY2, and healthy control The Chao index (community richness) of adjusted P . 0.05). The remainder of the subjects), a priori sample size estimation each group suggested a similar bacterial bacterial population belonged to five other was not performed. richness in the fecal samples from the phyla that had a relative abundance ,1% The relative abundances of each OTU three study groups (199.42 6 52.26 type (Supplementary Fig. 1 and Supplemen- (taxa) were compared by a Wilcoxon 1 diabetes, 195.08 6 41.62 MODY2, tary Data). In addition, the Firmicutes- signed rank test with a continuity cor- 218.65 6 51.05 healthy control sub- to-Bacteroidetes ratio was significantly rection using the Explicet software pack- jects; P . 0.05). Despite a significantly lower in type 1 diabetes than in MODY2 age specifically addressed to analyze lower Shannon index (microbiota diver- and healthy control subjects (0.30% type microbiome data. All the resulting P sity) in type 1 diabetes (4.76 6 0.42) and 1 diabetes, 0.76% MODY2, 0.58% healthy values were then adjusted for multiple MODY2 (4.78 6 0.47) in comparison with control subjects; P , 0.001, FDR-adjusted comparisons via the Benjamini-Hochberg healthy control subjects (5.16 6 0.33) P , 0.001). false discovery rate (FDR) correction (P , 0.05), no differences in a-diversity A total of 16 families were identified (FDR-corrected P value of ,0.05). a-and were found between type 1 diabetes among the fecal samples analyzed. b-diversities were achieved by QIIME, and MODY2 (P = 0.850). Within the Bacteroidetes, two differ- a-diversity using a nonparametric Stu- Regarding the b-diversity of gut micro- ent families were significantly higher in dent t test using a default number of biota, weighted UniFrac principal coor- type 1 diabetes when compared with Monte Carlo permutations of 999, and dinates analysis (PCoA) showed that MODY2 and healthy control subjects: b-diversity with the analysis of similarities children with type 1 diabetes had a Bacteroidaceae (71.48% type 1 diabetes, statistical method with 99 permutations. different pattern of clustering when 55.43% MODY2, 59.62% healthy control Differences in the clinical characteristics compared with MODY2 and healthy subjects; P , 0.001, FDR-adjusted P , between two groups were analyzed using control subjects (P =0.03andP =0.02, 0.001) and Rikenellaceae (15.26% type 1 2388 Gut Microbiota in Type 1 Diabetes and MODY2 Diabetes Care Volume 41, November 2018

Table 1—Baseline anthropometric and biochemical variables Healthy control Type 1 diabetes MODY2 subjects (n = 13) (n = 15) (n = 15) P Male/female, n 7/6 7/8 7/8 Vaginal delivery/cesarean section, n 8/5 10/5 10/5 Age (years) 12.25 6 2.92 12.56 6 3.59 13.06 6 3.20 0.654 BMI (kg/m2) 17.35 6 1.82 17.89 6 2.01 18.23 6 1.90 0.430 Age of onset of diabetes (years) 7.35 6 1.76 6.91 6 1.40 0.455 Duration of diabetes (years) 5.68 6 1.84 6.10 6 1.97 0.551 Breastfeeding time (months) 6.58 6 2.32 6.41 6 2.81 6.54 6 3.2 0.911 Birth weight (kg) 3.19 6 0.45 3.28 6 0.38 3.22 6 0.55 0.249 Weight (kg) 37.35 6 9.0 38.32 6 8.92 36.91 6 7.72 0.765 a b b HbA1c (%) 4.47 6 0.21 6.26 6 0.38 6.11 6 0.33 0.001 a b b HbA1c (mmol/mol) 25.3 6 2.3 44.9 6 4.2 43.3 6 3.6 0.001 Triglycerides (mg/dL) 52.67 6 9.43 53.50 6 10.15 53.88 6 9.88 0.843 Cholesterol (mg/dL) 153.88 6 14.64 153.62 6 16.87 154.5 6 17.9 0.920 IL-1b (pg/mL) 83.21 6 28.25b 119.41 6 27.12a 89.45 6 25.31b 0.004 IL-10 (pg/mL) 126.67 6 9.87b 81.03 6 9.97a 121.28 6 5.46b 0.001 IL-6 (pg/mL) 85.71 6 6.28b 109.89 6 6.50a 88.98 6 7.49b 0.001 TNF-a (pg/mL) 164.31 6 16.78b 373.46 6 90.65a 169.54 6 18.3b 0.001 IL-13 (pg/mL) 50.15 6 8.37b 22.83 6 5.47a 45.46 6 7.31b 0.001 LPS (EU/mL) 0.49 6 0.10b 1.10 6 0.40a 0.56 6 0.38b 0.001 Values are presented as means 6 SD unless otherwise specified. P value was based on Kruskal-Wallis test. Mann-Whitney U testwasusedtocomparethe clinical characteristics between two groups. Different superscript letters (a,b) next to values in a row indicate that the means of the different groups are significantly different (P , 0.05, Bonferroni post hoc test).

diabetes, 7.01% MODY2, 11.19% healthy subjects; P = 0.002, FDR-adjusted P = adjusted P , 0.001). Prevotella was control subjects; P , 0.001, FDR-adjusted 0.015) was significantly higher in compar- significantly increased in MODY2 and P , 0.001). (2.65% type 1 ison with type 1 diabetes and MODY2. In type 1 diabetes when compared with diabetes, 15.41% MODY2, 1.25% healthy Actinobacteria, a significant enrichment healthy control subjects (1.95% type 1 control subjects; P , 0.001, FDR-adjusted of Bifidobacteriaceae (2.71% type 1 di- diabetes, 8.32% MODY2, 1.42% healthy P = 0.003) was significantly higher in type abetes, 4.50% MODY2, 5.68% healthy control subjects; P , 0.001, FDR- 1 diabetes than in healthy control sub- control subjects; P =0.004,FDR- adjusted P = 0.005). Regarding Firmi- jects, and Prevotellaceae was significantly adjusted P = 0.017) was found in healthy cutes, the relative abundances of four higher in MODY2 than in type 1 diabetes control subjects when compared with genera were significantly higher in type (P , 0.001, FDR-adjusted P , 0.001). In MODY2 and type 1 diabetes. Finally, for 1 diabetes than in MODY2 and healthy the Firmicutes, another three families the Proteobacteria families, a significant control subjects: Ruminococcus (17.19% were significantly higher in type 1 di- increase of Enterobacteriaceae (25.20% type 1 diabetes, 5.74% MODY2, 8.85% abetes compared with MODY2 and type 1 diabetes, 15.03% MODY2,13.04% healthy control subjects; P , 0.001, FDR- healthy control subjects: Ruminococcaceae healthy control subjects; P =0.006,FDR- adjusted P = 0.002), Blautia (15.50% (38.19% type 1 diabetes, 25.74% MODY2, adjusted P = 0.03) was found in type 1 type 1 diabetes, 5.73% MODY2, 3.74% 30.84% healthy control subjects; P , diabetes when compared with MODY2 healthy control subjects; P , 0.001, FDR- 0.001, FDR-adjusted P , 0.001), and healthy control subjects, but no adjusted P =0.003),Veillonella (21.59% (31.94% type 1 diabetes, significant differences were found for type 1 diabetes, 12.33% MODY2, 7.20% 20.33% MODY2, 18.03% healthy control Alcaligenaceae (62.18% type 1 diabetes, healthy control subjects; P , 0.001, FDR- subjects; P , 0.001, FDR-adjusted P = 49.03% MODY2, 57.07% healthy control adjusted P =0.006),andStreptococcus 0.003), and Streptococcaceae (1.93% type subjects; P = 0.019, FDR-adjusted P = (4.86% type 1 diabetes, 2.64% MODY2, 1 diabetes, 0.96% MODY2, 0.56% healthy 0.267) (Supplementary Fig. 2 and Supple- 1.47% healthy control subjects; P =0.003, control subjects; P , 0.001, FDR-adjusted mentary Data). FDR-adjusted P = 0.028). In addition, four P = 0.004). No significant differences at Twelve genera were differentially genera were significantly lower in type 1 the Firmicutes family level were found abundant between study groups. For diabetes and MODY2 than in healthy con- between MODY2 and healthy control the Bacteroidetes genera, the type 1 trol subjects: Lachnospira (5.34% type 1 subjects except in the abundance of diabetes group was significantly en- diabetes, 7.15% MODY2, 15.25% healthy Ruminococcaceae (P, 0.001, FDR-adjusted riched with sequences attributed to the control subjects; P , 0.001, FDR-adjusted P , 0.001). In healthy children, only genus Bacteroides (72.21% type 1 di- P = 0.012), Roseburia (1.35% type 1 dia- (22.1% type 1 diabetes, abetes, 52.41% MODY2, 58.45% healthy betes, 4.16% MODY2, 6.99% healthy con- 27.95% MODY2, 42.0% healthy control control subjects; P , 0.001, FDR- trol subjects; P , 0.001, FDR-adjusted care.diabetesjournals.org Leiva-Gea and Associates 2389

Figure 1—Clustering of fecal bacterial communities according to the different study groups by PCoA using weighted UniFrac distances. Each point corresponds to a community coded according to the child group. The percentage of variation explained by the plotted principal coordinates is indicated on the axes. A: Type 1 diabetes (red dots) vs. healthy control subjects (blue dots). B: Type 1 diabetes (blue dots) vs. MODY2 (red dots). C:MODY2 (blue dots) vs. healthy control subjects (red dots). PC1, principal coordinate 1; PC2, principal coordinate 2.

P = 0.015), Anaerostipes (2.15% type 1 in the Proteobacteria genera, Sutterella Supplementary Data). A cladogram diabetes, 2.97% MODY2, 5.79% healthy and Enterobacter were significantly in- showing differences in fecal microbiota control subjects; P , 0.001, FDR-adjusted creased in type 1 diabetes when com- between study groups is shown in P = 0.023), and Faecalibacterium (4.21% pared with MODY2 and healthy control Fig. 2. type 1 diabetes, 8.08% MODY2, 13.26% subjects (61.30% type 1 diabetes, 49.32% healthy control subjects; P , 0.001, FDR- MODY2, 57.07% healthy control subjects Serum Zonulin Levels in Type 1 adjusted P = 0.004). In Actinobacteria, only [P = 0.002, FDR-adjusted P = 0.027] and Diabetes, MODY2, and Healthy the Bifidobacterium was significantly in- 16.18% type 1 diabetes, 8.25% MODY2, Control Subjects creased in healthy control subjects when 6.99% healthy control subjects [P , 0.001, Patients with MODY2 showed higher compared with type 1 diabetes (1.93% FDR-adjusted P = 0.003], respectively). serum zonulin levels (4.80 6 1.41 ng/mg type 1 diabetes, 6.75% healthy control No significant differences between protein) when compared with patients subjects; P , 0.001, FDR-adjusted P = study groups were found in the rela- with type 1 diabetes (3.94 6 1.44 ng/mg 0.017), but not regarding MODY2 (6.07% tive abundance of Desulfovibrio, Hae- protein, P = 0.02) and healthy control MODY2, 6.75% healthy control subjects; mophilus,andBilophila (FDR-adjusted subjects (3.21 6 1.24 ng/mg protein, P , P = 0.503, FDR-adjusted P . 0.05). Finally, P . 0.05) (Supplementary Fig. 3 and 0.001). Significant differences in serum 2390 Gut Microbiota in Type 1 Diabetes and MODY2 Diabetes Care Volume 41, November 2018

Figure 2—Cladogram showing differentially abundant taxa of the fecal microbiota in type 1 diabetes, MODY2, and healthy controls. Linear discriminant analysis effect size analysis (P , 0.05 for Kruskal-Wallis test) was used to validate the statistical significance and the effect size of the differential abundances of taxa in the study groups. The diameter of each circle is proportional to its abundance.

zonulin levels were also found between zonulin levels and HbA1c were found increase in Bacteroides (P = 0.002, b = type 1 diabetes and control subjects in children with type 1 diabetes and 0.995,r2 =0.94)andVeillonella(P=0.021, (P , 0.05). MODY2, but not in healthy control sub- b = 0.636, r2 = 0.93) and the decrease in jects (Table 2). In addition, significant Faecalibacterium (P = 0.041, b = 20.671, 2 The Relationship Between Gut correlations between serum levels of r = 0.94) and Roseburia (P = 0.038, 2 Microbiota Composition, Serum IL-1b, IL-6, TNF-a, IL-10, IL-13, and LPS b = 20.694, r = 0.93) in type 1 diabetes

Zonulin, HbA1c,Inflammatory and gut microbiota composition were and the rise in Prevotella in MODY2 (P = 2 Mediators, and LPS found in type 1 diabetes (Supplementary 0.002, b = 0.682, r = 0.90) were asso- Several significant correlations between Table 1). ciated with the increment in serum zonulin the relative abundance of specific bac- The linear regression analysis including levels. On the other hand, the increase teria at different taxa levels and serum all the bacterial groups showed that the in the abundance of Blautia (P = 0.043, care.diabetesjournals.org Leiva-Gea and Associates 2391

Table 2—Correlation between gut microbiota composition and serum zonulin and HbA1c levels in children with type 1 diabetes and MODY2

Zonulin HbA1c Type 1 diabetes MODY2 Type 1 diabetes MODY2 Ruminococcus 20.162 (P = 0.55) 20.540 (P = 0.036) Ruminococcus 20.362 (P = 0.18) 20.584 (P = 0.025) Roseburia 20.320 (P = 0.031) 20.732 (P = 0.48) Firmicutes-to- 20.561 (P = 0.029) 20.332 (P = 0.45) Bacteroidetes ratio Prevotella 0.560 (P = 0.37) 0.798 (P = 0.037) Blautia 0.559 (P = 0.038) 0.740 (P = 0.79) Faecalibacterium 20.703 (P = 0.027) 20.547 (P = 0.49) Streptococcus 0.068 (P = 0.018) 0.441 (P = 0.12) Bacteroides 0.739 (P = 0.004) 0.350 (P = 0.090) Veillonella 0.570 (P = 0.033) 0.704 (P = 0.04) Correlations are reported as Spearman r (r), and P values are given in parentheses. Statistical significance was set at a P value of ,0.05. b = 0.469, r2 = 0.93) and the decrease in addition, pathways of lipid and carbohy- and healthy control subjects. Also, we re- the Firmicutes-to-Bacteroidetes ratio (P = drate metabolism in type 1 diabetes had a port that gut permeability, determined 0.001, b = 20.947, r2 = 0.91) in patients significant enrichment in the proportion of by serum zonulin levels, is significantly in- with type 1 diabetes were associated with genes related to arachidonic acid and creasedinMODY2andtype1diabetes HbA1c levels, whereas in MODY2 patients, propanoate metabolism (P = 0.002 and when compared with healthy control sub- only the decrease in Ruminococcus (P = P = 0.003, respectively), in comparison jects. Another key finding in this study is a 0.003, b = 20.877, r2 = 0.92) was asso- with MODY2 and healthy subjects (Fig. 3). significantly lower diversity of the dom- ciated with HbA1c levels. In addition, the Metagenomic comparison between study inant bacterial community in type 1 di- increase in the relative abundance of groups showed that gene families linked abetes and MODY2 when compared with Bacteroides (P = 0.006, b = 0.991, r2 = to amino acid metabolism, such as amino healthy control subjects. 0.95) and Veillonella (P = 0.012, b = acid–related enzymes (P = 0.010), arginine The higher loss of diversity in patients 0.825, r2 = 0.95) and the decrease in and proline metabolism (P = 0.006), and with type 1 diabetes when compared Bifidobacterium (P = 0.039, b = 20.654, valine, leucine, and isoleucine biosynthe- with healthy control subjects might be r2 = 0.94), Roseburia (P = 0.032, sis (P = 0.020), were significantly increased related to the autoimmune process. In a b = 20.675, r2 = 0.95), and Faecalibac- in type 1 diabetes, whereas genes for previous study evaluating type 1 diabetes terium (P = 0.023, b = 20.678, r2 = 0.94) glutation metabolism (P = 0.005) were sig- markers, a lower microbial diversity was in type 1 diabetes were associated with nificantly depleted. Conversely, in MODY2 found in fecal samples of children with serum IL-1b levels. Finally, in type 1 and healthy control subjects, in compar- at least two positive disease-associated diabetes, the increase in the abundance ison with type 1 diabetes, a significant autoantibodies than in samples from of Bacteroides (P = 0.007, b = 0.632, r2 = increase in genes related to glycolysis/ -negative children matched 0.85) and the decrease in Roseburia (P = gluconeogenesis (P= 0.018), pentose phos- for age, sex, early feeding history, and 0.029, b = 20.518, r2 = 0.85) were asso- phate pathway (P = 0.015), and butanoate HLA genotyping (19). Also, in longitudinal ciated with serum IL-6 and TNF-a levels, metabolism (P = 0.023), as well as in en- studies from birth with children at risk for and the increase in Streptococcus (P = ergy metabolism genes such as sulfur (P = type 1 diabetes, a decrease of microbial 0.014, b = 0.616, r2 = 0.82) and the 0.014), nitrogen metabolism (P = 0.012), diversity occurred just before the appear- decrease in Bifidobacterium (P = 0.009, and oxidative phosphorylation (P =0.018), ance of anti-islet cell antibodies and sub- b = 20.904, r2 = 0.82) were associated was detected (Fig. 3). sequent onset of type 1 diabetes (20). with serum IL-10 and IL-13 levels. Regard- Finally, when compared with MODY2 We demonstrate that there are clear ing the serum levels of LPS, only the in- and healthy control subjects, gut micro- differences in the gut microbiota profile crease in the abundance of Veillonella (P = biota from patients with type 1 diabetes of type 1 diabetes, MODY2, and healthy 0.006, b =0.887,r2 = 0.83) was associated was significantly enriched with genes for control subjects, given that in the OTU- with the levels in type 1 diabetes. antigen processing and presentation (P = based PCoA plot analysis, patients with 0.010), chemokine signaling pathways type 1 diabetes clustered separately when Functional Differences in the Gut (P = 0.001), LPS biosynthesis (P = 0.008), compared with MODY2 patients and Microbiota of Study Groups bacterial invasion of epithelial cells (P = healthy control subjects. Accordingly, chil- The PICRUSt analysis indicated that genes 0.017), and ABC transporters (P = 0.016) dren with type 1 diabetes showed a signi- for energy (P = 0.015) and carbohydrate (Fig. 3). ficant increase in the relative abundance metabolism (P = 0.014) were significantly of Bacteroides, Ruminococcus, Veillonella, depleted in type 1 diabetes in comparison CONCLUSIONS Blautia, Enterobacter,andStreptococcus with MODY2 and healthy control subjects. In this study comparing the bacterial flora genera and a decrease in the relative Nevertheless, lipid metabolism functions in type 1 diabetes, MODY2, and healthy abundance of Bifidobacterium, Roseburia, (P = 0.008), together with amino acid control subjects, we show that type 1 Faecalibacterium,andLachnospira,when metabolism functions (P =0.013),were diabetes is associated with different gut compared with MODY2 and healthy con- overrepresented in type 1 diabetes when microbial composition and functional trol subjects. On the other hand, MODY2 compared with the other groups. In profiling, in comparison with MODY2 was related to a significant increase in 2392 Gut Microbiota in Type 1 Diabetes and MODY2 Diabetes Care Volume 41, November 2018

Figure 3—Predicted functional composition of metagenomes based on 16S rRNA gene sequencing data of type 1 diabetes (T1DM), MODY2, and healthy control subjects. Heat map of differentially abundant Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways identified in the three study groups. The values of color in the heat map represent the normalized relative abundance of KEGG pathways.

Prevotella abundance and a significant microbiota in type 1 diabetes (regardless main genus leading to type 1 diabetes– decrease in Ruminococcus and Bacter- of ethnicity, age, and geography), most associated dysbiosis (21,22). oides, when compared with type 1 di- published studies (including the present Bacteroides, a Gram-negative bacte- abetes and healthy control subjects. one) identify Bacteroides (acetate- and rium, could contribute to chronic inflam- Despite the great variability of intestinal propionate-producing bacteria) as the mation by the impairment of the barrier care.diabetesjournals.org Leiva-Gea and Associates 2393

function of the epithelial cell layer (23). between zonulin levels and the relative Also, we report a significantly lower In our study, subjects with type 1 diabetes, abundance of Bacteroides and Veillonella, Firmicutes-to-Bacteroidetes ratio in type when compared with MODY2 and and a significant negative correlation with 1 diabetes and a negative correlation be- healthy control subjects, presented sig- Faecalibacterium and Roseburia,aswell tween this ratio and HbA1c. This has been nificantly higher levels of LPS and proin- as a significant positive correlation be- previously demonstrated by our group flammatory cytokines IL-1b, IL-6, and tween zonulin and Prevotella in patients (8) and by other authors who have TNF-a along with a significant decrease with MODY2. Zonulin regulates intestinal reported a decline in Firmicutes and in anti-inflamatory citokines IL-10 and permeability by modulating intercellular an increase in Bacteroidetes abundance IL-13. Moreover, the significant correla- tight junctions. A possible mechanism is over time in the gut microbiome until the tions found between gut microbiota and based on bacterial and micro- development of type 1 diabetes (41). host serum levels of LPS and cytokines organism toxins that may be sensed by The PICRUSt analysis demonstrated indicate that the significant depletion of molecules related to epithelial cell tight that several microbial functions were Faecalibacterium, Roseburia,andBifidobac- junctions like zonulin, altering their ac- significantly over- or underrepresented terium, reported to exert anti-inflammatory tivity and consequently increasing gut between study groups, due to important effects and strengthen gut barrier function permeability and bacterial translocation differences in bacteria composition. (through cytokine production modulation (21,31). However, zonulin levels alone do Thus, when compared with healthy con- and butyrate production, respectively) not show impaired epithelial integrity, trol subjects and MODY2, gut microbiota (24,25), together with the significant in- and in accordance with other authors, in type 1 diabetes showed a depleted crease of Veillonella (a lactate-utilizing the impaired epithelial integrity found abundance of genes involved in meta- and propionate-producing bacteria with in the patients with type 1 diabetes might bolic pathways such as carbohydrate and proinflammatory capacity) and Bacter- be caused by the binding of Veillonella (a energy metabolism. Conversely, there oides, could raise paracellular permeabil- lactate-utilizing bacteria) to immature was an increase in genes related to lipid ity and low-grade inflammation in type 1 cells in colonic crypts able to ferment and amino acid metabolism, ABC trans- diabetes (26,27). This situation could al- glucose to lactate, which are pushed to port, LPS biosynthesis, arachidonic acid low luminal antigens to escape from the the luminal surface and form poor tight metabolism, antigen processing and gut and promote islet-directed autoim- junctions (32). Also, in type 1 diabetes, presentation, and chemokine signaling mune responses. Therefore, the lower there is an increased intestinal permeabil- pathways related to inflammation and abundance of anti-inflammatory bacteria ity that may affect absorption of antigens immune response. The relative abun- unable to regulate epithelial integrity capable of attacking and damaging pan- dance of genes associated with a given could be associated with the intestinal creatic b-cells (33–35). Thus, the signif- pathway may indicate an increased met- immune activation in type 1 diabetes icant increase in gut permeability may abolic capacity of the gut microbiota with (28,29). be an important player in the develop- regard to this pathway. Interestingly, a On the other hand, we have found a ment of type 1 diabetes. Moreover, higher level of arachidonic acid metab- significant increase in the relative abun- some authors have stated that a leaky olism (inflammatory intermediate) in dance of Prevotella in MODY2 when gut could be involved in the develop- type 1 diabetes gut microbiota might compared with type 1 diabetes and ment of type 1 diabetes complications, be the result of a growth of proinflam- healthy control subjects. Given that as high serum LPS activity has been matory pathobionts in the gut (42). Prevotella is an important succinate associated with features of metabolic Our study has certain limitations but producer and mucin degrader bacteria, syndrome, visceral fat mass, and the also some important strengths. The lim- this could suggest a lack of mucin on the progression of kidney disease in type 1 itations include the relatively small sam- intestinal epithelial layer of MODY2 diabetes (36–38). ple size (mainly caused by the very low subjects, disturbing the protection of On the other hand, we have shown prevalence of MODY2 in children), which the host mucosal surfaces. Moreover, that the relative abundance of Blautia may not be enough for detecting differ- Prevotella-produced succinate is a bac- and the Firmicutes-to-Bacteroidetes ra- ences between low-abundance microbes terial metabolite that leads to inhibition tio were positively correlated with HbA1c that may be of relevance or to assess of hepatic glucose output and improves in type 1 diabetes, and in MODY2, HbA1c overall differences between type 1 dia- glycemic control and energy metabolism levels were negatively correlated with betes and MODY2. Another limitation is through the activation of intestinal glu- the abundance of Ruminococcus.Butyrate- the inherent nature of the study, a cross- coneogenesis (30). Accordingly, Spegel´ producing intestinal bacteria such as Ru- sectional design, where only an associ- et al. (13) suggested that in MODY2, minococcus and Blautia could play an ation and not a cause can be inferred, and despite the shift in insulin secretion, met- importantrole in blood glucose regulation where potential changes of the gut mi- abolic control remains intact probably due and lipid metabolism, as shown by fecal crobioma from a healthy state to a gut to the existence of compensatory mech- transplantation studies (39). Some au- pattern potentially boosting autoimmu- anisms external to b-cells. Therefore, we thors have reported that Blautia is pos- nity in type 1 diabetes were not evaluated. suggest that the significant increase in itively correlated with serum glucose, Finally, although PICRUSt gives functional the abundance of Prevotella found in HbA1c levels, and the number of type 1 information of potential importance, it is a MODY2 patients could be related to the diabetes autoantibodies, suggesting that limitation compared with shotgun meta- maintenance of glycemic control. Blautia might influence the development genomics analysis. On the other hand, In addition, we have found in type 1 of type 1 diabetes through the regulation the strengths of our study lie in the diabetes a significant positive correlation of T-cell differentiation (20,40). careful design, the inclusion of patients 2394 Gut Microbiota in Type 1 Diabetes and MODY2 Diabetes Care Volume 41, November 2018

with MODY2, the well-matched cohorts Author Contributions. I.L.-G. conceived the 10. Bosi E, Molteni L, Radaelli MG, et al. In- (including age, mode of delivery, breast- study, developed the experimental design, ac- creased intestinal permeability precedes clinical quired and selected all samplesusedinthisstudy, feeding time, antibiotics use, BMI, and onset of type 1 diabetes. Diabetologia 2006;49: compiled the database, performed statistical 2824–2827 glycemic levels), and the next-generation analysis and data interpretation, and wrote the 11. Mathis D, Benoist C. The influence of the sequencing of the microbiome. manuscript and provided critical revision. L.S.-A. microbiota on type-1 diabetes: on the threshold In conclusion, our data suggest that performed laboratory assays and wrote the man- of a leap forward in our understanding. Immunol gut microbiota in type 1 diabetes not only uscript. B.M.-T. acquired and selected all samples Rev 2012;245:239–249 used in this study, compiled the database, per- ł˛ ł differs at the taxonomic level regarding 12. Zak-Go ab A, Koce ak P, Aptekorz M, et al. formed statistical analysis and data interpretation, Gut microbiota, microinflammation, metabolic pro- MODY2 and healthy subjects but also at and wrote the manuscript and provided critical file, and zonulin concentration in obese and normal the functional level, involving different revision. D.C.-C. performed laboratory assays, weight subjects. Int J Endocrinol 2013;2013:674106 metabolic pathways. compiled the database, and performed statistical 13. SpegelP,EkholmE,TuomiT,GroopL,Mulder´ Type 1 diabetes was characterized by analysis and data interpretation. I.M.-I. performed H, Filipsson K. Metabolite profiling reveals nor- laboratory assays, compiled the database, per- a less diverse gut microbiota profile, with mal metabolic control in carriers of mutations in formed statistical analysis and data interpretation, the glucokinase gene (MODY2). Diabetes 2013; Bacteroidetes dominating at the phylum and wrote the manuscript. A.U.C. acquired and 62:653–661 level and an increased proportion of selected all samples used in this study, compiled 14. Nyunt O, Wu JY, McGown IN, et al. Inves- proinflammatory bacteria. Therefore, the the database, and performed statistical analysis tigating maturity onset diabetes of the young. type 1 diabetes gut microbiota profile and data interpretation. F.J.T. conceived the study, Clin Biochem Rev 2009;30:67–74 developed the experimental design, and wrote the was associated with impaired epithelial 15. Fajans SS, Bell GI. MODY: history, genetics, manuscript and provided critical revision. J.C.F.-G. pathophysiology, and clinical decision making. integrity, low-grade inflammation, and conceived the study, developed the experimental Diabetes Care 2011;34:1878–1884 autoimmune response. On the other hand, design, compiled the database, performed statis- 16. American Diabetes Association. Classification gut microbiota in MODY2 was charac- tical analysis and data interpretation, and wrote the and diagnosis of diabetes. Sec. 2. In Standards of manuscript and provided critical revision. M.I.Q.-O. terized by a dominance of succinate- Medical Care in Diabetesd2017.DiabetesCare conceived the study, developed the experimental 2017;40(Suppl. 1):S11–S24 producer and mucin-degrading bacteria, design, acquired and selected all samples used in 17. Sanchez-Alcoholado L, Castellano-Castillo D, potentially modulating glucose metabo- this study, performed laboratory assays, compiled Jordan-Mart´ ´ınez L, et al. Role of gut microbiota on lism in the intestine of the host and in- the database, performed statistical analysis and cardio-metabolic parameters and immunity in cor- fluencing systemic energy homeostasis. data interpretation, and wrote the manuscript and onary artery disease patients with and without type-2 provided critical revision. All authors read and ap- diabetes mellitus. Front Microbiol 2017;8:1936 Our results provide evidence of a dif- proved the final manuscript. J.C.F.-G. and M.I.Q.-O. fi 18. Bhute S, Pande P, Shetty SA, et al. Molecular ferent microbiota pro le and function- are the guarantors of this work and, as such, had full characterization and meta-analysis of gut mi- ality in children with type 1 diabetes, not access to all the data in the study and take re- crobial communities illustrate enrichment of only in comparison with healthy subjects sponsibility for the integrity of the data and the Prevotella and Megasphaera in Indian subjects. but fundamentally with regard to a accuracy of the data analysis. Front Microbiol 2016;7:660 model of nonautoimmune diabetes, 19. de Goffau MC, Luopajarvi¨ K, Knip M, et al. Fecal microbiota composition differs between References suggesting a potential role of gut micro- children with b-cell and those biota in the autoimmune process in- 1. Khashan AS, Kenny LC, Lundholm C, Kearney without. Diabetes 2013;62:1238–1244 volved in type 1 diabetes. Given that PM, Gong T, Almqvist C. Mode of obstetrical 20. Kostic AD, Gevers D, Siljander H, et al.; delivery and type 1 diabetes: a sibling design DIABIMMUNE Study Group. The dynamics of the most studies to date have shown that study. Pediatrics 2014;134:e806–e813 intestinal microbiota, rather than being human infant gut microbiome in development 2. Rewers M, Ludvigsson J. Environmental risk and in progression toward type 1 diabetes. Cell involved in the initiation of the disease factors for type 1 diabetes. Lancet 2016;387: Host Microbe 2015;17:260–273 – process of type 1 diabetes, might be in- 2340 2348 21. Mej´ıa-Leon´ ME, Petrosino JF, Ajami NJ, volved in the progression from b-cell 3. Hansen CH, Krych L, Nielsen DS, et al. Early Dom´ınguez-Bello MG, de la Barca AM. Fecal mi- life treatment with vancomycin propagates autoimmunity to the clinical disease (43), crobiota imbalance in Mexican children with Akkermansia muciniphila and reduces diabetes type 1 diabetes. Sci Rep 2014;4:3814 future longitudinal studies should be incidence in the NOD mouse. Diabetologia 2012; 22. Brown CT, Davis-Richardson AG, Giongo A, – aimed at evaluating if the modulation of 55:2285 2294 et al. Gut microbiome metagenomics analysis gut microbiota in patients with a high risk 4. Roesch LF, Lorca GL, Casella G, et al. Culture- suggests a functional model for the develop- fi of type 1 diabetes could modify the nat- independent identi cation of gut bacteria cor- ment of autoimmunity for type 1 diabetes. PLoS related with the onset of diabetes in a rat model. ural history of this autoimmune disease. One 2011;6:e25792 ISME J 2009;3:536–548 23. Tlaskalova-Hogenov´ aH,St´ epˇ ankov´ aR,´ 5. Kriegel MA, Sefik E, Hill JA, Wu HJ, Benoist C, Kozakov´ a´ H, et al. The role of gut microbiota Mathis D. Naturally transmitted segmented fil- (commensal bacteria) and the mucosal barrier Acknowledgments. The research group be- amentous bacteria segregate with diabetes pro- in the pathogenesis of inflammatory and auto- longs to the Centro de Investigacion´ Biomedica´ tection in nonobese diabetic mice. Proc Natl immune diseases and : contribution of en Red (CIBER, CB06/03/0018) of the Instituto Acad Sci U S A 2011;108:11548–11553 germ-free and gnotobiotic animal models of hu- de Salud Carlos III. 6. Vaarala O. Is the origin of type 1 diabetes in man diseases. Cell Mol Immunol2011;8:110–120 Funding. This study was supported by the the gut? Immunol Cell Biol 2012;90:271–276 24. Sokol H, Pigneur B, Watterlot L, et al. Fae- Ministerio de Educacion,´ Cultura y Deporte 7. He C, Shan Y, Song W. Targeting gut micro- calibacterium prausnitzii is an anti-inflammatory (FPU13/04211 to D.C.-C.), the Miguel Servet biota as a possible therapy for diabetes. Nutr Res commensal bacterium identified by gut micro- Type I program from the Instituto de Salud Carlos 2015;35:361–367 biota analysis of Crohn disease patients. Proc Natl III (cofoundedby theFondoEuropeodeDesarrollo 8. Murri M, Leiva I, Gomez-Zumaquero JM, et al. Acad Sci U S A 2008;105:16731–16736 Regional) (CP16/00163 to I.M.-I. and CP13/00065 Gut microbiota in children with type 1 diabetes 25. Tamanai-Shacoori Z, Smida I, Bousarghin L, to M.I.Q.-O.), and the Servicio Andaluz de Salud differs from that in healthy children: a case- et al. Roseburia spp.: a marker of health? Fu- (research contract B-0033-2014 to J.C.F.-G.). control study. BMC Med 2013;11:46 ture Microbiol 2017;12:157–170 Duality of Interest. No potential conflicts of 9. Vaarala O. Leaking gut in type 1 diabetes. Curr 26. Bischoff SC, Barbara G, Buurman W, et al. interest relevant to this article were reported. Opin Gastroenterol 2008;24:701–706 Intestinal permeability–a new target for disease care.diabetesjournals.org Leiva-Gea and Associates 2395

prevention and therapy. BMC Gastroenterol 33. Watts T, Berti I, Sapone A, et al. Role of the 38. Lassenius MI, Ahola AJ, Harjutsalo V, 2014;14:189 intestinal tight junction modulator zonulin in the Forsblom C, Groop PH, Lehto M. Endotoxins 27. Burger-van Paassen N, Vincent A, Puiman PJ, pathogenesis of type I diabetes in BB diabetic- are associated with visceral fat mass in type 1 et al. The regulation of intestinal mucin MUC2 prone rats. Proc Natl Acad Sci U S A 2005;102: diabetes. Sci Rep 2016;6:38887 expression by short-chain fatty acids: implications 2916–2921 39. Puddu A, Sanguineti R, Montecucco F, for epithelial protection. Biochem J 2009;420:211– 34. Gulden¨ E, Wong FS, Wen L. The gut micro- Viviani GL. Evidence for the gut microbiota 219 biota and type 1 diabetes. Clin Immunol 2015; short-chain fatty acids as key pathophysiological 28. Li N, Hatch M, Wasserfall CH, et al. Butyrate and 159:143–153 molecules improving diabetes. Mediators In- type 1 diabetes mellitus: can we fix the intestinal 35. Lee AS, Gibson DL, Zhang Y, Sham HP, Vallance flamm 2014;2014:162021 leak? J Pediatr Gastroenterol Nutr 2010;51:414–417 BA, Dutz JP. Gut barrier disruption by an enteric 40. Qi CJ, Zhang Q, Yu M, et al. Imbalance of 29. Vaarala O. Gut microbiota and type 1 di- bacterial pathogen accelerates insulitis in NOD fecal microbiota at newly diagnosed type 1 di- abetes. Rev Diabet Stud 2012;9:251–259 mice. Diabetologia 2010;53:741–748 abetes in Chinese children. Chin Med J (Engl) 30. DeVadderF,Kovatcheva-DatcharyP,ZitounC, 36. Nymark M, Pussinen PJ, Tuomainen AM, 2016;129:1298–1304 Duchampt A, Backhed¨ F, Mithieux G. Microbiota- Forsblom C, Groop PH, Lehto M; FinnDiane Study 41. Giongo A, Gano KA, Crabb DB, et al. Toward produced succinate improves glucose homeostasis Group. Serum lipopolysaccharide activity is as- defining the autoimmune microbiome for type via intestinal gluconeogenesis. Cell Metab 2016; sociated with the progression of kidney disease 1 diabetes. ISME J 2011;5:82–91 24:151–157 in Finnish patients with type 1 diabetes. Diabe- 42. Candela M, Biagi E, Soverini M, et al. Mod- 31. Vaarala O. Human intestinal microbiota and tes Care 2009;32:1689–1693 ulation of gut microbiota dysbioses in type 2 type 1 diabetes. Curr Diab Rep 2013;13:601–607 37. Lassenius MI, Pietilainen¨ KH, Kaartinen K, diabetic patients by macrobiotic Ma-Pi 2 diet. 32. Lopez CA, Miller BM, Rivera-Chavez´ F, et al. et al.; FinnDiane Study Group. Bacterial endotoxin Br J Nutr 2016;116:80–93 Virulence factors enhance Citrobacter roden- activity in human serum is associated with dysli- 43. Knip M, Siljander H. The role of the intestinal tium expansion through aerobic respiration. pidemia, , obesity, and chronic microbiota in type 1 diabetes mellitus. Nat Rev Science 2016;353:1249–1253 inflammation. Diabetes Care 2011;34:1809–1815 Endocrinol 2016;12:154–167