OBJECTIVE

Previous studies have demonstrated an association between gut microbiota composition and type 1 diabetes (T1D) pathogenesis. However, little is known about the composition and function of the gut microbiome in adults with longstanding T1D or its association with host glycemic control.

RESEARCH DESIGN AND METHODS

We performed a metagenomic analysis of the gut microbiome obtained from fecal samples of 74 adults with T1D, 14.6 ± 9.6 years following diagnosis, and compared their microbial composition and function to 296 age-matched healthy control subjects (1:4 ratio). We further analyzed the association between microbial taxa and indices of glycemic control derived from continuous glucose monitoring measurements and blood tests and constructed a prediction model that solely takes microbiome features as input to evaluate the discriminative power of microbial composition for distinguishing individuals with T1D from control subjects.

RESULTS

Adults with T1D had a distinct microbial signature that separated them from control subjects when using prediction algorithms on held-out subjects (area under the receiver operating characteristic curve = 0.89 ± 0.03). Linear discriminant analysis showed several bacterial species with significantly higher scores in T1D, including Prevotella copri and Eubacterium siraeum, and species with higher scores in control subjects, including Firmicutes bacterium and Faecalibacterium prausnitzii (P < 0.05, false discovery rate corrected for all). On the functional level, several metabolic pathways were significantly lower in adults with T1D. Several bacterial taxa and metabolic pathways were associated with the host’s glycemic control.

CONCLUSIONS

We identified a distinct gut microbial signature in adults with longstanding T1D and associations between microbial taxa, metabolic pathways, and glycemic control indices. Additional mechanistic studies are needed to identify the role of these bacteria for potential therapeutic strategies.

Type 1 diabetes (T1D) is a common chronic disease in children and adolescents. The incidence of T1D has been rapidly rising in the past decade, especially in young children (1). While a genetic predisposition for T1D exists, this rapid increase in the prevalence of the disease and the fact that <10% of genetically susceptible individuals will eventually develop T1D are suggestive of a large contribution of environmental factors to disease pathogenesis. These may include viral infections and nutritional factors (2). Gut microbiota composition has also been highlighted as a possible risk factor, with several studies in humans and animal models implicating its potential role in disease pathogenesis (35). These observations have further led to the “balanced signal” hypothesis, stating that microbiome composition may promote or inhibit T1D development (6). Several suggested mechanisms for the possible influence of the gut microbiome on T1D pathogenesis include immunological deregulation mediated by gut dysbiosis, as there is evidence that the microbiome plays an important role in the development and maturation of the immune system (7), and gut leakiness, as structural mucosal alterations and gut dysfunction was observed in both human and animal studies on T1D (8).

In recent years, a rapidly growing number of studies have investigated the role of the gut microbiome in T1D (9). However, most studies focused on disease pathogenesis, while only a few studies thus far have investigated the microbiome composition of individuals with a longstanding diagnosis, and those were mostly conducted on small cohorts and used a variety of computational analysis methods (10,11). In addition, while evidence on the regulating roles of the microbiome in normal and impaired glycemic response is accumulating in both animal models and humans (12), little is known on the role of the microbiome in glycemic control in individuals with longstanding T1D. Here, we analyzed microbial composition and function in a cohort of individuals with T1D who were at least 1 year following diagnosis and the associations between microbial taxa, functional pathways, and glycemic indices in individuals with T1D.

Study Design

We conducted a prospective clinical cohort originally designed to study the postprandial glycemic responses (PPGRs) of individuals with T1D. Full details on recruitment and the study protocols are specified in a companion paper by Shilo et al. (13), focused exclusively on modeling the PPGR in individuals with T1D. In brief, on the first day of the study, participants were invited to a study initiation meeting at the medical center. In this meeting, a physician authorized participation and acquired informed consent, anthropometric measurements were obtained, and blood tests were drawn and analyzed in the hospital’s laboratories. Health and lifestyle questionnaires were completed by the participants. Throughout the 2 weeks of study participation, participants used a proprietary smartphone app (www.personalnutrition.org), to log, in real-time, food intake, sleep times, physical activity, and medication intake with the exception of insulin, which was recorded in the continuous subcutaneous insulin infusion devices. Participants were asked to follow their normal routine and dietary habits, with the exception of seven standardized meals. Participants were asked to provide one microbiome sample collected during the 2 weeks of study participation.

Participant Recruitment

Enrollment and recruitment were conducted in three medical centers in Israel between March 2017 and April 2019 (Fig. 1). The inclusion criteria for the study included age between 3 and 70 years old (13). However, as previous studies demonstrated that the interpersonal variation in the composition of the bacterial communities is significantly greater among children (14) and as a large variation exists in clinical phenotypes between children and adults (15), we choose to include only adults (18–70 years old) in the analyses presented here. Additional inclusion criteria were >1 year following T1D diagnosis, using continuous glucose monitoring (CGM) and continuous subcutaneous insulin infusion devices simultaneously, and a capability to work with a mobile phone app daily for the recording of the dietary intake. Exclusion criteria included an active inflammatory or neoplastic disease, pregnancy, and antibiotic use 3 months prior to participation in the study. Participants who reported a diagnosis of celiac disease were excluded from all microbiome analyses since several studies previously showed that celiac disease is correlated with a change in gut microbial composition (16).

Study Population

Overall, 142 individuals with T1D were recruited to the study, and 124 participants provided a stool sample. Seven participants reported a diagnosis of celiac disease and were therefore excluded from microbiome analyses, resulting in 117 individuals. From them, 74 were >18 years and were included in the analyses (Fig. 1). The average age was 32.3 ± 14.4 years (median 26 [interquartile range 21–43] years), and average disease duration was 14.6 ± 9.6 years (median 12 [interquartile range 7.8–18.3] years). Mean HbA1c level was 7.3 ± 1% (56.3 ± 10.9 mmol/mol) (see Supplementary Table 1 for mean values of all blood test results at study initiation). The mean BMI value was 25.1 ± 4 kg/m2. Of the 74 participants, 33 (44.6%) had at least one additional comorbidity. The most common comorbidities were hypothyroidism (12 participants [16.22%]) and hyperlipidemia (10 participants [13.51%]). Thirty-nine participants (52.7%) consumed additional medications apart from insulin during the study. The most common medications were levothyroxine (12 participants [16.22%]), oral contraceptives (8 participants, 10.21%), and antilipidemic drugs (8 participants, 10.21%) (see Supplementary Table 2 for a full list of medical conditions and medications consumed by the participants during the study). Cohort characteristics are presented in Table 1. Of the 74 individuals, 73 logged meals in real-time during the 2 weeks of study participation (see a companion paper by Shilo et al. [13]). Total energy intake was 1,666,610 kcal (22,830 per person). Average carbohydrate, fat, and protein consumption was 43 ± 1%, 38 ± 7%, and 17 ± 4% from the total energy, respectively.

Cohort Matching

To compare between the composition of the microbiome in adults with T1D and healthy adults, we used gut metagenomic profiles obtained from Israeli adults (17). Of 35,304 Israeli adults who submitted their sample between 13 January 2017 and 1 May 2021, 14,012 were excluded due to a different sequencing method, and 13,295 were excluded due to the presence of one of the following metabolic, gastrointestinal, or systemic diseases: type 2 diabetes, T1D, gestational diabetes, prediabetes, impaired glucose tolerance or impaired fasting glucose, metabolic syndrome, fatty liver disease, morbid obesity, inflammatory bowel disease, Crohn disease, ulcerative colitis, undetermined colitis, pancreatic diseases, celiac disease, irritable bowel syndrome, diverticulosis, hepatitis or other liver diseases, cholangitis or other bile-related diseases, HIV, autoimmune diseases, and cancer. Next, individuals with T1D were matched by age to healthy control subjects in a 1:4 ratio (1 adult with T1D to 4 healthy control subjects), resulting in 296 healthy individuals who were included as control subjects (Fig. 1 and Table 1). Comparisons between individuals with T1D and healthy control subjects were done using the linear discriminant analysis (LDA) method (18). False discovery rate (FDR)–corrected P values were computed following the Benjamini-Hochberg procedure and were used at the rate of 0.1.

Stool Sample Collection and Genomic DNA Extraction

Participants entering the study received a verbal explanation from the study coordinators and detailed printed instructions for stool collection. Microbiome sampling was done using a swab and an OMNIgene-GUT (OMR-200; DNA Genotek) stool collection kit. Each participant was requested to collect stool via one swab and one separate OMNIIgene-GUT kit. However, only samples collected by OMNIIgene-GUT kits were sequenced and analyzed since it has the advantage of maintaining DNA integrity in typical ambient temperature fluctuations and since samples of the control group were collected only by the OMNIIgene-GUT kits. Collected samples were immediately stored in a home freezer (−18°C) and transferred in a provided cooler to our facilities where they were stored at −80°C (−20°C for OMNIIgene-GUT kits) until DNA extraction. Samples from adults with T1D were sequenced between April 2019 and August 2019, and samples from healthy control subjects were sequenced between April 2019 and May 2021. All samples analyzed in this study were sequenced using the same sequencing methods, including sequencing protocols of DNA extraction, library preparation, and sequencing machine. Control samples demonstrated that performing the process on different days had no effect on the results when the sequencing protocols were kept the same.

Metagenomic DNA was purified using MagAttract PowerSoil DNA extraction kit (QIAGEN) optimized for the Tecan automated platform. Next-generation sequencing libraries were prepared using Nextera DNA library prep (Illumina) and sequenced on a NovaSeq sequencing platform (Illumina). Sequencing was performed with 100 base pair single end reads with a depth of 10 million reads per sample. We filtered metagenomic reads containing Illumina adapters, filtered low-quality reads, and trimmed low-quality read edges. We detected host DNA by mapping with Bowtie 2 (19) to the human genome with inclusive parameters and removed those reads. Bacterial relative abundance (RA) estimation was performed by mapping bacterial reads to species-level genome bins (SGB) representative genomes (20). We selected all SGB representatives with at least five genomes in a group, and for these representative genomes kept unique regions as a reference data set. Mapping was performed using Bowtie 2 (19), and abundance was estimated by calculating the mean coverage of unique genomic regions across the 50% most densely covered areas, as previously described (21). Feature names include the lowest taxonomy level identified. In addition, we also estimated the RA of bacterial groups, such as Akkermansia, Alistipes, Roseburia, Eubacterium, and Faecalibacterium prausnitzii as a summation of the abundances of SGBs belonging to the relevant species by National Center for Biotechnology Information classification.

Microbial Biodiversity Indices and Functional Analysis

Microbiome α-diversity was calculated by the Shannon diversity index. Richness was calculated as a number of species in the sample detected with an abundance of at least 1e-4. Comparison between microbial indices and RA of microbial taxa were performed using Mann-Whitney U test. HUMAnN2 v2.8.2 (22) was used to integrate taxonomic information with functional profiles.

Associations With Clinical Phenotypes

We used several indices to analyze the association between clinical and microbial features and measures of glycemic control. These included fasting glucose, HbA1c level, and lipids measured by a blood test at study initiation and indices calculated based on CGM measurements during the 2 weeks of study participation, available for 73 participants. CGM-derived features included the percentage of the time spent in hypoglycemia and hyperglycemia defined as glucose values <70 mg/dL (3.9 mmol/L) and >180 mg/dL (10 mmol/L), respectively, time in range, defined as time spent in glucose values between 70 and 180 mg/dL (3.9–10 mmol/L) (23), and coefficient of variation (CV) as a measure of glucose variability (24). For the 73 participants who also logged meals throughout the study period, PPGRs were calculated (see in a companion paper by Shilo et al. [13]). Pearson correlations between the clinical phenotypes, RA converted to a log space of microbial taxa, and metabolic pathways were calculated.

T1D Prediction Model Based on Microbial Features

To evaluate the discriminative power of microbial composition for T1D, we constructed a prediction model based on XGBoost (25), which solely takes microbiome features as inputs. This model can capture nonlinear interactions between bacteria and was previously shown to outperform other methods for the classification of human microbiome data (26). The mean and SD of the receiver operating characteristic curve were computed by using the curves that were generated in fivefold cross-validation. In addition, we verified that when randomly swapping the target labels, the performances reflected a random prediction, hence an area under the curve (AUC) very close to 0.5, as an additional control. We analyzed feature attributes using SHAP (SHapley Additive exPlanation) to explore model interpretability. SHAP values represent the average change in the model’s output upon conditioning on a specific feature (27).

Ethical Approval

The study was approved by Rambam Medical Center Institutional Review Board (IRB), Tel Hashomer Hospital IRB, Shamir Medical Center IRB, and Weizmann Institute of Science IRB. All participants signed written informed consent forms. All identifying details of the participants were removed prior to the computational analysis. The trial was registered as NCT02919839 at https://clinicaltrials.gov/.

Data and Resource Availability

Metagenomic sequencing data that support the findings of this study are available. Clinical data cannot be shared due to restrictions by informed consent. The data set is available at https://data. mendeley.com/datasets/bcz47mhvc3/1. Analysis code is available at https://github.com/Nastyagodneva/T1D_Micro biome.

Correlations Between Microbial Strains, Functional Pathways, and Clinical Phenotypes

We first sought to explore the associations between microbial features, functional pathways, and clinical parameters (Fig. 2). Several bacterial taxa were significantly associated with glycemic indices, including a negative correlation between the relative abundance of Prevotellaceae species SGB592 and SGB1340 and HBA1c level (r = −0.35) and a positive correlation between Enterobacterales species (SGB2483) and glucose average (r = 0.41; P < 0.05, FDR corrected for all). Species from the Clostridiaceae family (SGB1422) were positively correlated with time in range (r = 0.38). Several associations between microbial taxa and lipids were also observed: Faecalibacterium prausnitzii species (SGB15339) were negatively correlated with total cholesterol levels (r = −0.41), and species from the Clostridiales order and Firmicutes class (SGB1421 and SGB1451) were negatively correlated with triglyceride levels (r = −0.4). In addition, several metabolic pathways were significantly associated with glucose average, including pathways relating to aromatic acid biosynthesis (COMPLETE-ARO-PWY, r = 0.42), chorismate biosynthesis from 3-dehydroquinate (PWY-6163, r = 0.39), and chorismate biosynthesis I (ARO-PWY, r = 0.42). In contrast, an inverse correlation was observed between the pyrimidine nucleobases salvage pathway (PWY-7208, r = −0.41) (Supplementary Fig. 1) and glucose average (P < 0.05, FDR corrected for all). No statistically significant associations were found between nutritional parameters and bacterial taxa.

Microbiome Composition in Individuals With T1D

To improve our understanding of the composition of the gut microbiome in T1D, individuals with T1D were compared with healthy controls (1:4 matching by age, see Research Design and Methods). Overall, 74 adults with T1D were compared with 296 healthy adults. There were no statistically significant differences in sex, weight, or BMI between groups. As expected, healthy adults had significantly lower levels of HbA1c (Table 1). Microbial α-diversity was not significantly different between the groups (Fig. 3C), aligned with previous studies (11,28) but in contrast with others, reporting a lower diversity in individuals with T1D (29,30). In addition, species richness and the ratio of Firmicutes-to-Bacteroidetes of taxonomic profiles were not significantly different between the groups (Supplementary Table 3). Linear discriminant analysis showed a total of 17 bacterial taxa with significantly higher LDA scores in individuals with T1D and 15 bacterial taxa with significantly higher LDA scores in healthy adults (Fig. 3A and D and Supplementary Table 4). Bacterial species with significantly higher scores in individuals with T1D included Prevotella copri, Eubacterium siraeum, and Alistipes inops, and several species with a higher score in healthy adults, including Firmicutes bacterium, Alistipes putredinis, Faecalibacterium prausnitzii, and Ruminococcus gnavus (P < 0.05, FDR corrected). Dimensionality reduction techniques, including principal component analysis (PCA), in which the principal coordinate combination with the greatest contribution rate was PC1 = 7.7%, PC2 = 4.1%, and t-distributed stochastic neighbor embedding (t-SNE), did not reveal visually distinctive differences between individuals with T1D and control subjects (Supplementary Fig. 2). On the functional level, when comparing metabolic pathways, several metabolic pathways, including l-glutamate and l-glutamine biosynthesis, l-ornithine de novo biosynthesis, and superpathway of hexuronide and hexuronate degradation, were significantly lower in adults with T1D (P < 0.05, FDR corrected).

Classification of Individuals With T1D by Microbial Features

We next analyzed our ability to distinguish individuals with T1D from control subjects based solely on microbiome features. We constructed a prediction model based solely on microbial fetures and used cross-validation schemes for validation of the model (see Research Design and Methods). The discrimination performance of the model had an AUC of 0.89 ± 0.03 and permutations P < 0.001 (Fig. 3B). The most impactful microbial taxa for the prediction were Prevotella copri, which impacted the model toward the prediction of T1D, and Ruminococcus, which impacted the model toward the prediction of a healthy state (Supplementary Fig. 3).

In this study, we profiled the gut microbiome composition in adults with longstanding T1D and identified several associations between bacterial taxa, metabolic pathways, and the glycemic control of the host. While a growing body of evidence, mainly originating from studies on animal models, suggests that gut microbiota has a causal impact on host glycemic control (31) that may be mediated by mechanisms such as modulation of incretin secretion, short-chain fatty acid production, metabolism of bile acid, and regulation of adipose tissue (32), data regarding the role of the microbiome in the glycemic control of individuals with T1D are still sparse.

Here, bacterial taxa and metabolic pathways that were significantly associated with glycemic indices of the host included Enterobacterales species and pathways relating to aromatic acid and chorismate biosynthesis, which were correlated with glucose average, Prevotellaceae species that were inversely correlated with HbA1c level, and the pyrimidine nucleobases salvage pathways that were inversely correlated with glucose average (P < 0.05, FDR corrected for all). Several small-scale studies previously showed different associations, including a study of 12 Chinese subjects with T1D (33) that demonstrated an inverse correlation of the abundance of Faecalibacterium and HbA1c levels, and a study conducted in Brazil that included 20 individuals with T1D and demonstrated a correlation between the relative abundances of Bacteroidetes, Lactobacillales, and Bacteroides dorei and HbA1c levels (34). Importantly, the correlations observed in this study were not strong, and further studies integrating multiomic data, including metagenomic, metatranscriptomic, and metaproteomic, along with high-quality clinical and nutritional data, are needed in order to identify the potential role of these bacteria and metabolic pathways and their influence on the host’s glycemic control.

We identified a distinct gut microbial signature in adults with longstanding T1D compared with healthy adults. By using an expanded reference set (20) for the first time in individuals with T1D, as well as a relatively large control group, we show a total of 17 bacterial taxa with significantly higher LDA scores in T1D and 15 bacterial taxa with significantly higher LDA scores in control subjects (Fig. 3). Although dimensionality reduction analyses did not reveal visually distinctive differences (Supplementary Fig. 2) and the diversity and richness were not statistically different between groups, we were able to devise a model that accurately distinguishes between adults with T1D and healthy control subjects using only microbiome features (AUC = 0.89 ± 0.03) (Fig. 3B). Interestingly, the most impactful microbial taxa for the prediction were Prevotella copri, which impacted the model toward the prediction of T1D, and Ruminococcus, which impacted the model toward the prediction of a healthy state (Supplementary Fig. 2), aligned with the results of the LDA analysis, showing higher scores for Prevotella copri in T1D and Ruminococcus gnavus in healthy adults (Fig. 3).

Previous studies (9) reported various results regarding the taxonomic composition of the gut microbiome in individuals with T1D compared with healthy control subjects and their interpretation is challenging due to a large heterogeneity in both study population and analytic approaches. It is also worthy to note that that gut microbiota in T1D was previously shown to differ at taxonomic and functional levels compared not only with healthy subjects but also with nonautoimmune diabetes (35). The most common findings in individuals with T1D included alterations in the following bacterial species: Bacteroides, Streptococcus, Clostridium, Bifidobacterium, Prevotella, Staphylococcus, Blautia, Faecalibacterium, Roseburia, and Lactobacillus (36). In the largest human cohort to date, no particular taxa was associated with the T1D development, but the microbiome of control children was found to contain more genes related to fermentation and biosynthesis of short-chain fatty acid (SCFA) compared with children who eventually developed T1D (4). An additional study also reported a decrease in SCFA producers in individuals with longstanding T1D. Moreover, it was previously shown that feeding NOD mice with SCFA-rich (butyrate and acetate) diets had substantial effects on their immune system and a protective effect from the development of diabetes (37). In this cohort, when comparing metabolic pathways, we found several metabolic pathways, including l-glutamate and l-glutamine biosynthesis, l-ornithine de novo biosynthesis, and superpathway of hexuronide and hexuronate degradation, that were significantly lower in adults with T1D (P < 0.05, FDR corrected). To the best of our knowledge, these findings have not been previously described in individuals with T1D, and their role should be further explored in future work.

The strength of our study includes a relatively large sample size compared with previous studies, the integration of data on glucose measurements obtained from CGM devices, and the expanded reference set we used.

The greatest limitation of our study is its observational nature. Further studies are needed in order to attribute causality to the gut microbiome alterations we describe, as currently, whether these taxa are a cause or an effect of the disease remains unclear. In addition, although the sample size of the cohort is relatively large, it may still be insufficient to reach robust associations with clinical phenotyping. Finally, several additional factors may influence the composition of the gut microbiome. For example, nutritional habits may differ between individuals with T1D compared with healthy individuals. While we did not have detailed nutritional data on our control group, macronutrient distribution in the T1D cohort was very similar to healthy adults in Israel as measured in a previous study performed by our group (38). In this study, healthy individuals logged meals during 1 week and consumed an average of 46 ± 8% carbohydrate, 36 ± 7% fat, and 15 ± 3% protein from the total energy, compared with an average of 43 ± 1% carbohydrate, 38 ± 7% fat, and 17 ± 4% protein consumed by the T1D cohort. Moreover, in the group of individuals with T1D, no associations were found between nutritional parameters and bacterial taxa. Medication consumption may also influence microbial composition, and we therefore excluded individuals with antibiotic use 3 months prior to participation. While other types of medications, such as proton pump inhibitors, may also have an effect (39), they were only consumed by a very small percentage of our cohort (Supplementary Table 2). Family kindred may also have a pronounced effect on the structural and functional composition of the gut microbiome (40). However, none of the adults with T1D included in this study were family members sharing the same household. Microbiome composition is also heavily influenced by geographic location (14), and therefore, additional studies are needed in order to determine whether our findings can be generalized to non-Israeli populations.

In conclusion, our study highlights a distinct gut microbial composition in individuals with longstanding T1D compared with healthy individuals. We identified unknown associations between microbial taxa, metabolic pathways, and clinical phenotypes and note the importance of expanding the gut microbiome reference set, as it allows us to also identify associations with unclassified bacterial strains that may play a part in disease pathogenesis. Our findings provide a foundation for additional large-scale analyses of the gut microbiome in individuals with T1D in order to identify host–microbe interactions and to identify the causal role of these bacterial taxa for the development of novel therapeutic strategies in T1D.

Clinical trial reg. no. NCT02919839, clinicaltrials.gov

See accompanying article, p. 502.

This article contains supplementary material online at https://doi.org/10.2337/figshare.17430323.

Acknowledgments. The authors thank the Segal group members for fruitful discussions.

Funding. This work is supported by the Israel Science Foundation (ISF) (grant no. 3-14762). E.S. is supported by the Crown Human Genome Center, Larson Charitable Foundation New Scientist Fund, Else Kroener Fresenius Foundation, White Rose International Foundation, Ben B. and Joyce E. Eisenberg Foundation, Nissenbaum Family, Marcos Pinheiro de Andrade and Vanessa Buchheim, Lady Michelle Michels, Aliza Moussaieff, and grants funded by the Minerva foundation with funding from the Federal German Ministry for Education and Research and by the European Research Council and the Israel Science Foundation.

These funding sources had no role in the design of this study and will not have any role during its execution, analyses, interpretation of the data, or decision to submit results.

Duality of Interest. E.S. is a paid consultant for DayTwo. No other potential conflicts of interest relevant to this article were reported. No pharmaceutical manufacturers or companies from the industry contributed to the planning, design, or conduct of the trial.

Author Contributions. S.S. and A.G. conceived the project, designed and conducted the analyses, interpreted the results, and wrote the manuscript. M.R. provided data and interpreted the results. T.K. conceived the project, designed the analysis, and interpreted the results. Y.B. designed the analysis and interpreted the results. D.K. and T.K. designed and conducted the analyses. N.B. designed the analysis and interpreted the results. B.C.W. and Y.G.-G. coordinated and designed data collection. M.C., N.Z.L., N.S., N.G., N.L., and S.K. provided data and interpreted the results. A.W. conceived the project and directed sample sequencing. O.P.-H. and E.S. conceived the project, designed and conducted the analyses, interpreted the results, and supervised the project and analyses. All authors reviewed and approved the manuscript and vouch for the accuracy and completeness of the data. E.S. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

1.
Ehehalt
S
,
Dietz
K
,
Willasch
AM
;
Baden-Württemberg Diabetes Incidence Registry (DIARY) Group
.
Epidemiological perspectives on type 1 diabetes in childhood and adolescence in Germany: 20 years of the Baden-Württemberg Diabetes Incidence Registry (DIARY)
.
Diabetes Care
2010
;
33
:
338
340
2.
Ilonen
J
,
Lempainen
J
,
Veijola
R
.
The heterogeneous pathogenesis of type 1 diabetes mellitus
.
Nat Rev Endocrinol
2019
;
15
:
635
650
3.
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
4.
Vatanen
T
,
Franzosa
EA
,
Schwager
R
, et al
.
The human gut microbiome in early-onset type 1 diabetes from the TEDDY study
.
Nature
2018
;
562
:
589
594
5.
Paun
A
,
Yau
C
,
Danska
JS
.
The influence of the microbiome on type 1 diabetes
.
J Immunol
2017
;
198
:
590
595
6.
Burrows
MP
,
Volchkov
P
,
Kobayashi
KS
,
Chervonsky
AV
.
Microbiota regulates type 1 diabetes through Toll-like receptors
.
Proc Natl Acad Sci U S A
2015
;
112
:
9973
9977
7.
Sommer
F
,
Bäckhed
F
.
The gut microbiota--masters of host development and physiology
.
Nat Rev Microbiol
2013
;
11
:
227
238
8.
Secondulfo
M
,
Iafusco
D
,
Carratù
R
, et al
.
Ultrastructural mucosal alterations and increased intestinal permeability in non-celiac, type I diabetic patients
.
Dig Liver Dis
2004
;
36
:
35
45
9.
Zheng
P
,
Li
Z
,
Zhou
Z
.
Gut microbiome in type 1 diabetes: a comprehensive review
.
Diabetes Metab Res Rev
2018
;
34
:
e3043
10.
Stewart
CJ
,
Nelson
A
,
Campbell
MD
, et al
.
Gut microbiota of Type 1 diabetes patients with good glycaemic control and high physical fitness is similar to people without diabetes: an observational study
.
Diabet Med
2017
;
34
:
127
134
11.
de Groot
PF
,
Belzer
C
,
Aydin
Ö
, et al
.
Distinct fecal and oral microbiota composition in human type 1 diabetes, an observational study
.
PLoS One
2017
;
12
:
e0188475
12.
Suez
J
,
Shapiro
H
,
Elinav
E
.
Role of the microbiome in the normal and aberrant glycemic response
.
Clin Nutr Exp
2016
;
6
:
59
73
13.
Shilo
S
,
Godneva
A
,
Rachmiel
M
, et al
.
Prediction of personal glycemic responses to food for individuals with type 1 diabetes through integration of clinical and microbial data
.
Diabetes Care
2022
;
45
:
502
511
14.
Yatsunenko
T
,
Rey
FE
,
Manary
MJ
, et al
.
Human gut microbiome viewed across age and geography
.
Nature
2012
;
486
:
222
227
15.
Gerstl
EM
,
Rabl
W
,
Rosenbauer
J
, et al
.
Metabolic control as reflected by HbA1c in children, adolescents and young adults with type-1 diabetes mellitus: combined longitudinal analysis including 27,035 patients from 207 centers in Germany and Austria during the last decade
.
Eur J Pediatr
2008
;
167
:
447
453
16.
Caio
G
,
Volta
U
,
Sapone
A
, et al
.
Celiac disease: a comprehensive current review
.
BMC Med
2019
;
17
:
142
17.
Rothschild
D
,
Leviatan
S
,
Hanemann
A
,
Cohen
Y
,
Weissbrod
O
,
Segal
E
.
An atlas of robust microbiome associations with phenotypic traits based on large-scale cohorts from two continents
.
30 May 2020 [preprint]. BioRxiv 2020.05.28.122325
18.
Segata
N
,
Izard
J
,
Waldron
L
, et al
.
Metagenomic biomarker discovery and explanation
.
Genome Biol
2011
;
12
:
R60
19.
Langmead
B
,
Salzberg
SL
.
Fast gapped-read alignment with Bowtie 2
.
Nat Methods
2012
;
9
:
357
359
20.
Pasolli
E
,
Asnicar
F
,
Manara
S
, et al
.
Extensive unexplored human microbiome diversity revealed by over 150,000 genomes from metagenomes spanning age, geography, and lifestyle
.
Cell
2019
;
176
:
649
662.e20
21.
Korem
T
,
Zeevi
D
,
Suez
J
, et al
.
Growth dynamics of gut microbiota in health and disease inferred from single metagenomic samples
.
Science
2015
;
349
:
1101
1106
22.
Franzosa
EA
,
McIver
LJ
,
Rahnavard
G
, et al
.
Species-level functional profiling of metagenomes and metatranscriptomes
.
Nat Methods
2018
;
15
:
962
968
23.
Gabbay
MAL
,
Rodacki
M
,
Calliari
LE
, et al
.
Time in range: a new parameter to evaluate blood glucose control in patients with diabetes
.
Diabetol Metab Syndr
2020
;
12
:
22
24.
DeVries
JH
.
Glucose variability: where it is important and how to measure it
.
Diabetes
2013
;
62
:
1405
1408
25.
Chen
T
,
Guestrin
C
.
XGBoost: a scalable tree boosting system
. In
Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining - KDD ’16
.
New York, NY
:
ACM Press
;
2016
, p.
785
794
26.
Wang
X-W
,
Liu
Y-Y
.
Comparative study of classifiers for human microbiome data
.
Med Microecol
2020
;
4
:
100013
27.
Lundberg
SM
,
Erion
GG
,
Lee
S-I
.
Consistent individualized feature attribution for tree ensembles
.
21 June 2018 [preprint]. arXiv:1802.03888
28.
Alkanani
AK
,
Hara
N
,
Gottlieb
PA
, et al
.
Alterations in intestinal microbiota correlate with susceptibility to type 1 diabetes
.
Diabetes
2015
;
64
:
3510
3520
29.
de Goffau
MC
,
Fuentes
S
,
van den Bogert
B
, et al
.
Aberrant gut microbiota composition at the onset of type 1 diabetes in young children
.
Diabetologia
2014
;
57
:
1569
1577
30.
Kostic
AD
,
Gevers
D
,
Siljander
H
, et al.;
DIABIMMUNE Study Group
.
The dynamics of the human infant gut microbiome in development and in progression toward type 1 diabetes
.
Cell Host Microbe
2015
;
17
:
260
273
31.
Meijnikman
AS
,
Gerdes
VE
,
Nieuwdorp
M
,
Herrema
H
.
Evaluating causality of gut microbiota in obesity and diabetes in humans
.
Endocr Rev
2018
;
39
:
133
153
32.
Gérard
C
,
Vidal
H
.
Impact of gut microbiota on host glycemic control
.
Front Endocrinol (Lausanne)
2019
;
10
:
29
33.
Huang
Y
,
Li
S-C
,
Hu
J
, et al
.
Gut microbiota profiling in Han Chinese with type 1 diabetes
.
Diabetes Res Clin Pract
2018
;
141
:
256
263
34.
Higuchi
BS
,
Rodrigues
N
,
Gonzaga
MI
, et al
.
Intestinal dysbiosis in autoimmune diabetes is correlated with poor glycemic control and increased interleukin-6: a pilot study
.
Front Immunol
2018
;
9
:
1689
35.
Leiva-Gea
I
,
Sánchez-Alcoholado
L
,
Martín-Tejedor
B
, et al
.
Gut microbiota differs in composition and functionality between children with type 1 diabetes and MODY2 and healthy control subjects: a case-control study
.
Diabetes Care
2018
;
41
:
2385
2395
36.
Jamshidi
P
,
Hasanzadeh
S
,
Tahvildari
A
, et al
.
Is there any association between gut microbiota and type 1 diabetes? A systematic review
.
Gut Pathog
2019
;
11
:
49
37.
Mariño
E
,
Richards
JL
,
McLeod
KH
, et al
.
Gut microbial metabolites limit the frequency of autoimmune T cells and protect against type 1 diabetes
.
Nat Immunol
2017
;
18
:
552
562
38.
Zeevi
D
,
Korem
T
,
Zmora
N
, et al
.
Personalized nutrition by prediction of glycemic responses
.
Cell
2015
;
163
:
1079
1094
39.
Weersma
RK
,
Zhernakova
A
,
Fu
J
.
Interaction between drugs and the gut microbiome
.
Gut
2020
;
69
:
1510
1519
40.
Heintz-Buschart
A
,
May
P
,
Laczny
CC
, et al
.
Integrated multi-omics of the human gut microbiome in a case study of familial type 1 diabetes
.
Nat Microbiol
2016
;
2
:
16180
Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More information is available at https://www.diabetesjournals.org/journals/pages/license.