Distinct alterations of gut morphology and microbiota characterize accelerated diabetes onset in nonobese diabetic mice

The rising prevalence of type 1 diabetes (T1D) over the past decades has been linked to lifestyle changes, but the underlying mechanisms are largely unknown. Recent findings point to gut-associated mechanisms in the control of T1D pathogenesis. In nonobese diabetic (NOD) mice, a model of T1D, diabetes development accelerates after deletion of the Toll-like receptor 4 (TLR4). We hypothesized that altered intestinal functions contribute to metabolic alterations, which favor accelerated diabetes development in TLR4-deficient (TLR4−/−) NOD mice. In 70–90-day-old normoglycemic (prediabetic) female NOD TLR4+/+ and NOD TLR4−/− mice, gut morphology and microbiome composition were analyzed. Parameters of lipid metabolism, glucose homeostasis, and mitochondrial respiratory activity were measured in vivo and ex vivo. Compared with NOD TLR4+/+ mice, NOD TLR4−/− animals showed lower muscle mass of the small intestine, higher abundance of Bacteroidetes, and lower Firmicutes in the large intestine, along with lower levels of circulating short-chain fatty acids (SCFA). These changes are associated with higher body weight, hyperlipidemia, and severe insulin and glucose intolerance, all occurring before the onset of diabetes. These mice also exhibited insulin resistance–related abnormalities of energy metabolism, such as lower total respiratory exchange rates and higher hepatic oxidative capacity. Distinct alterations of gut morphology and microbiota composition associated with reduction of circulating SCFA may contribute to metabolic disorders promoting the progression of insulin-deficient diabetes/T1D development.

The rising prevalence of type 1 diabetes (T1D) over the past decades has been linked to lifestyle changes, but the underlying mechanisms are largely unknown. Recent findings point to gutassociated mechanisms in the control of T1D pathogenesis. In nonobese diabetic (NOD) mice, a model of T1D, diabetes development accelerates after deletion of the Toll-like receptor 4 (TLR4). We hypothesized that altered intestinal functions contribute to metabolic alterations, which favor accelerated diabetes development in TLR4-deficient (TLR4 ؊/؊ ) NOD mice. In 70 -90-day-old normoglycemic (prediabetic) female NOD TLR4 ؉/؉ and NOD TLR4 ؊/؊ mice, gut morphology and microbiome composition were analyzed. Parameters of lipid metabolism, glucose homeostasis, and mitochondrial respiratory activity were measured in vivo and ex vivo. Compared with NOD TLR4 ؉/؉ mice, NOD TLR4 ؊/؊ animals showed lower muscle mass of the small intestine, higher abundance of Bacteroidetes, and lower Firmicutes in the large intestine, along with lower levels of circulating short-chain fatty acids (SCFA). These changes are associated with higher body weight, hyperlipidemia, and severe insulin and glucose intolerance, all occurring before the onset of diabetes. These mice also exhibited insulin resistance-related abnormalities of energy metabolism, such as lower total respiratory exchange rates and higher hepatic oxida-tive capacity. Distinct alterations of gut morphology and microbiota composition associated with reduction of circulating SCFA may contribute to metabolic disorders promoting the progression of insulin-deficient diabetes/T1D development.
Type 1 diabetes (T1D) 3 is characterized by absolute insulin deficiency resulting from progressive autoimmune destruction of pancreatic beta cells (1,2). Epidemiological data indicate that not only the prevalence of obesity-related type 2 diabetes (T2D) but also of T1D is continuously increasing (3). In this context, accelerated childhood body weight gain (4) and insulin resistance (5) along with impaired glucose tolerance (6) associate with a higher risk for or earlier onset of T1D (7). Moreover, recent studies suggest that tissue-specific abnormalities of energy metabolism could underlie the insulin resistance not only in T2D but also in rodent models and humans with T1D (8,9).
Toll-like receptor 4 (TLR4), a dominant member of the family of pattern recognition receptors (10,11), is an attractive candidate to link diabetes-promoting immunologic and metabolic processes. TLR4 was initially identified as a receptor for lipopolysaccharides (LPS), which are integral components of the cell wall of Gram-negative bacteria (12). Based on this feature, TLR4 contributes to the activation of host defense against microorganisms (12). Meanwhile, however, increasing evidence suggests that TLR4 is also involved in the induction of autoimmunity, including development of T1D (13,14). The presence of TLR4 on insulin target tissues, such as liver, adipose tissue, and skeletal muscle (15), supports its relevance for diabetes-associated metabolic disorders such as obesity and insu-lin resistance (16). In fact, elevated TLR4 expression correlates with increased systemic concentrations of free fatty acids (FFA) and glucose (17) and with the degree of insulin resistance in obesity (18,19). However, the mechanisms underlying TLR4dependent modulation of the progression of insulin-deficient/ T1D are not yet clear.
Recent findings suggest an impact of the intestinal microbiome on the onset and progression of T1D. Patients with T1D have a less diverse and stable gut microbiome than glucosetolerant humans (20,21). Moreover, gut microbiome dysbiosis in early childhood associates with an increased risk of progression to T1D (22,23). Intestinal bacteria are able to modulate the host's metabolism by various mechanisms, specifically by the release of short-chain fatty acids (SCFA), a set of highly-abundant fermentation products derived from the bacterial degradation of complex macrofibrous dietary components (24,25). SCFA are potent modulators of glucose and energy metabolism (26) and thereby might affect the development of T1D (27).
These findings are in line with the concept of control of the pathogenesis of insulin-deficient/T1D by TLR4 expression status via modification of morphological and functional properties of the gut. Here, we tested the hypothesis that TLR4-dependent alterations of intestinal function affect metabolic homeostasis thereby contributing to accelerated progression of insulin-deficient diabetes. To this end, we used the TLR4-deficient (TLR4 Ϫ/Ϫ ) nonobese diabetic (NOD) mouse strain, which shows enhanced insulitis and therefore can serve as a model of accelerated human T1D development (14), that allows us to examine disease-relevant metabolic processes on a genetic background predisposing to insulin-deficient diabetes (28).

Altered gut morphology and gut microbiota in NOD TLR4 ؊/؊ mice
Detailed analyses of gut morphology showed a proportional reduction of both longitudinal and circular muscle layers of the small intestine of NOD TLR4 Ϫ/Ϫ mice ( Fig. 2A) resulting in lower total muscle thickness of the gut segment (Fig. 2B). The TLR4 expression status neither affected the dimensions of small intestinal villi and crypts (Fig. 2, C-E) nor the length and weight of the gut segments (Fig. 2, F and G) or weight of gut contents (Fig. 2H).
Effects of TLR4 deletion on gut microbiome composition were assessed using bacterial DNA isolated from small intestine, cecum, and colon. The gut segments contained comparable microbial biomass in NOD TLR4 ϩ/ϩ and NOD TLR4 Ϫ/Ϫ mice as estimated from the ratio of isolated bacterial DNA/pellet of gut contents (Fig. 2I). Employing next generation sequencing (NGS) revealed an increased abundance of Bacteroidetes and a decrease of Firmicutes in the cecum and colon of NOD TLR4 Ϫ/Ϫ mice compared with the corresponding gut segments of NOD TLR4 ϩ/ϩ animals ( Fig.  2J and Table S1). In line with the high quantity of bacteria in Figure 1. Accelerated body weight development and higher fat mass in the prediabetic state as well as younger age and higher blood glucose levels at diabetes onset in NOD TLR4 ؊/؊ mice. Food intake (A), body weight (B), and body composition (C) were determined in normoglycemic 70 -90-day-old female TLR4-expressing (TLR4 ϩ/ϩ ) and TLR4-deficient (TLR4 Ϫ/Ϫ ) NOD mice. The age at diabetes onset (D) was defined as the first of 2 consecutive days with blood glucose levels above 250 mg/dl. Initial blood glucose concentrations (E) were documented at diabetes manifestation. Data are presented as means Ϯ S.E. (n ϭ 9 -13 per group). *, p Ͻ 0.05; ***, p Ͻ 0.001 by Student's t test.

Gut microbiota and diabetes progression
cecum and colon, these gut segments exhibited the highest ␣-diversity (Fig. 2K).
Assessing intestinal inflammation from frequency and distribution of CD3 ϩ lymphocytes in the gut wall revealed no differences between NOD TLR4 ϩ/ϩ and NOD TLR4 Ϫ/Ϫ mice (Fig. S1, A and B), which resembled the distribution pattern in corresponding tissue samples of age-and sex-matched normal healthy C57BL10 mice (data not shown). NOD TLR4 Ϫ/Ϫ mice had a higher activity of matrix metalloproteinase-2 (MMP-2), a marker of extracellular matrix integrity (Fig. S1C), but unchanged levels of other mediators of systemic inflammation (Fig. S1, D and E). Mucispirillum content correlated with plasma glucagon like peptide (GLP)-2 only in NOD TLR4 Ϫ/Ϫ mice, again indicating impaired gut integrity of NOD TLR4 Ϫ/Ϫ mice (Fig. S2).

Higher LPS levels and altered lipid and SCFA profiles in NOD TLR4 ؊/؊ mice
As gut wall integrity and intestinal microbiome composition strongly affect the availability of bacterial products in the host's circulation, we determined the levels of plasma components reflecting absorption of gut-derived bacterial products and lipids. When compared with NOD TLR4 ϩ/ϩ mice, NOD TLR4 Ϫ/Ϫ mice showed higher plasma concentrations of LPS (Fig. 3A), a cell wall component of Gram-negative bacteria that can be incorporated into chylomicrons. NOD TLR4 Ϫ/Ϫ mice also had higher levels of triglycerides (TG) and FFA than NOD TLR4 ϩ/ϩ mice in plasma (Fig. 3, B and C). We next measured circulating levels of fetuin A in serum, an abundant liver-derived glycoprotein, acting as an adaptor to enable the interaction of FFA with TLR4 (29). However, both NOD TLR4 ϩ/ϩ and NOD TLR4 Ϫ/Ϫ mice showed high but comparable fetuin A concentrations (Fig. 3D).
As intestinal bacteria-derived SCFA are potent modulators of host energy metabolism, we determined their concentrations in the lumen of small intestine, cecum, and colon using GC-MS. Highest SCFA concentrations were found in the two distal gut segments with acetic acid, propionic acid, and butyric acid as the most abundant species (Fig. 4). No differences were observed in SCFA concentrations in gut segments from TLR4 ϩ/ϩ and NOD TLR4 Ϫ/Ϫ mice.

Impaired energy homeostasis and glucose metabolism in NOD TLR4 ؊/؊ mice
Metabolic phenotyping during three dark and two light phases revealed lower energy expenditure during light phases, occurring independently of the TLR4 expression status (Fig.  6A). Physical activity followed a similar pattern, showing comparable low levels during the light phase and 3-4-fold increases in the dark phase (Fig. 6B). Interestingly, NOD TLR4 Ϫ/Ϫ mice had greater maximum physical activity than NOD TLR4 ϩ/ϩ mice during the dark phase. All mice also exhibited the typical pattern of circadian rhythmicity of the respiratory exchange rate (RQ) with maximum values in the dark phases (Fig. 6, C and D). NOD TLR4 Ϫ/Ϫ mice had consistently lower RQ values than NOD TLR4 ϩ/ϩ mice (Fig. 6D), suggesting increased utilization of dietary fat as energy source.
Intraperitoneal glucose tolerance tests (ipGTT) revealed severe glucose intolerance with elevation of both glucose peak (Fig. 6E) and overall glucose appearance, as measured from the area under the glucose concentration curve, in NOD TLR4 Ϫ/Ϫ mice (Fig. 6F). Of note, peak glucose levels occurred 25 min later, pointing to delayed glucose absorption in the TLR4-deficient mice. Serum insulin concentration, measured at the end of the ipGTT, was not appropriately increased in NOD TLR4 Ϫ/Ϫ , indicating impaired glucose-dependent insulin secretion (Fig. 6G). To assess insulin sensitivity in more detail, the animals underwent an intraperitoneal insulin tolerance test (ipITT), which showed marked whole-body insulin resistance in NOD TLR4 Ϫ/Ϫ mice (Fig. 6H). In addition, in line with hepatic insulin resistance these mice had higher HOMA-IR values (Fig. 6I). To examine glucose-responsive insulin secretion, we isolated islets of Langerhans from pancreata of the mice. These studies revealed comparable dose-dependent glucose- Gut microbiota and diabetes progression stimulated insulin release in NOD TLR4 ϩ/ϩ and NOD TLR4 Ϫ/Ϫ mice (Fig. 6J).

Higher hepatic oxidative capacity but unchanged muscle oxidative capacity in NOD TLR4 ؊/؊ mice
Reduced RQ and glucose tolerance along with higher plasma lipids in NOD TLR4 Ϫ/Ϫ mice indicate abnormal energy metabolism. For the direct measurement of substrate-dependent oxidation, we employed high-resolution respirometry on samples from soleus muscle and liver to quantify the mitochondrial oxygen fluxes via complex I and complexes I and II and to assess maximum respiratory capacity (Fig. 7). In soleus muscle, TLR4 expression status did not affect oxygen fluxes (Fig. 7A). In contrast, livers from NOD TLR4 Ϫ/Ϫ mice showed 60 -86% higher rates of oxygen flux in complex I and complexes I and II as well as maximal respiratory capacity compared with those obtained from NOD TLR4 ϩ/ϩ mice (Fig. 7B). This indicates hepatic mitochondrial adaptation to greater glucose and lipid availabil-ity as reported previously in insulin-resistant rodent models and humans (8,30).

Discussion
These data show that profound alterations of gut morphology and microbiota and tissue-specific abnormalities of energy metabolism associate with altered availability of SCFA and precede accelerated diabetes progression in a model of T1D. This study also suggests a complex relationship between TLR4-dependent intestinal abnormalities and altered inter-organ communication by an imbalance between SCFA and FFA resulting in disease-promoting conditions in the prediabetic state.
The earlier onset of diabetes and higher blood glucose levels at disease onset in NOD TLR4 Ϫ/Ϫ mice confirmed the diabetes-accelerating effect of TLR4 deficiency in this NOD mouse model (14). This study provides evidence that TLR4 deficiency promotes the development of diabetes under conditions of

Gut microbiota and diabetes progression
early alterations of gut microbiota composition and gut morphology before disease onset.
The TLR4 expression status determines important structural and functional properties of the intestinal system. By controlling the proliferation of smooth muscle cells, this receptor is involved in the regulation of muscle tissue expansion (31). As intestinal smooth muscle cells express functional TLR4 (32), TLR4 deficiency may foster decreasing the thickness of the intestinal muscle layer as observed in the NOD TLR4 Ϫ/Ϫ mice in this study. Reduced intestinal muscle thickness is known to impair gut motility (33). Moreover, the contractile function of intestinal smooth muscle cells also depends on the appropriate expression of TLR4 (34). Thus, TLR4 deficiency likely promotes prolongation of the intestinal transit time of gut contents (35) as reported recently in metabolically healthy C57BL6 mice carrying a homozygous TLR4 defect (36).
Consequently, reduced gut motility and prolonged intestinal transit time cause alterations of the digestive process and the accumulation of absorbable food components (37), thereby modulating the survival conditions for distinct bac-terial communities. A recent study demonstrated such an association between delayed intestinal transit and alterations of gut microbiome composition (38). In this study, NOD TLR4 Ϫ/Ϫ mice showed an imbalance of the two most abundant bacterial phyla when compared with NOD TLR4 ϩ/ϩ mice. Bacteroidetes were increased while Firmicutes were decreased in the distal gut segments of normoglycemic TLR4-deficient mice at an age of 70 -90 days, i.e. before the age of diabetes-onset. This finding corresponds to observations in patients with T1D and in children at increased risk for T1D showing a predominance of distinct Bacteroidetes species in their gut microbiome prior to the clinical onset of auto-immunity (21). Of note, some previous studies suggested an involvement of gut microbiota in the development of insulin-deficient diabetes (3,39,40).
In this study, NOD TLR4 Ϫ/Ϫ mice further exhibited signs of impaired barrier function of the intestinal wall, indicated by increased activity of MMP2, a dominant member of the metalloproteinase family involved in tissue remodeling (41). Of note, expression of metalloproteinases is also up-regulated in the gut  Figure 6. Effect of the TLR4 expression status on energy balance and glucose homeostasis of NOD mice. Energy expenditure (A), physical activity (B), and RQ (C and D) were monitored in normoglycemic 70 -90-day-old female TLR4-expressing (TLR4 ϩ/ϩ ) and TLR4-deficient (TLR4 Ϫ/Ϫ ) NOD mice over three dark and two light phases. Glucose tolerance was determined by ipGTT (E and F), and the serum insulin concentrations of the animals were quantified by ELISA (G). Insulin tolerance was determined by ipITT (H). HOMA-IR was calculated from fasting blood glucose and insulin levels (I). Glucose-stimulated insulin release was determined from isolated cultivated islets of NOD TLR4 ϩ/ϩ and NOD TLR4 Ϫ/Ϫ mice (J). Data are presented as means or means Ϯ S.E. (n ϭ 3-10 per group). *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001 by Student's t test or one-way ANOVA.

Gut microbiota and diabetes progression
of patients with T1D pointing to disturbed tissue integrity (42). Furthermore, NOD TLR4 Ϫ/Ϫ mice had higher plasma concentrations of bacterial LPS, indicative of an enhanced translocation of the toxin from the gut lumen to the circulation. The increased abundance of Bacteroidetes, which are mainly composed of Gram-negative species, might further contribute to the elevated peripheral LPS levels in these animals. Bacterial products, including LPS, can reduce self-tolerance thereby promoting the development of organ-specific autoimmunity (43). In line, augmented translocation of bacterial products from the gut lumen into the circulation may particularly contribute to the accelerated progression of beta-cell damage and insulindeficient diabetes observed in NOD TLR4 Ϫ/Ϫ mice (14, 39). Indeed, the 70 -90-day-old prediabetic NOD TLR4 Ϫ/Ϫ mice exhibited impaired glucose-induced insulin secretion during the ipGTT.
Intestinal bacteria release SCFA, which can enter the circulation thereby representing a possibly important link between gut microbiome and host metabolism (24,26). Interestingly, NOD TLR4 Ϫ/Ϫ mice exhibited profound changes in the composition of gut bacteria, dominated by a decrease in the abundance of Firmicutes and an increase of Bacteroidetes. However, this shift in the intestinal microbial community did not result in altered SCFA composition in the gut lumen of these mice. Nevertheless, NOD TLR4 Ϫ/Ϫ mice showed lower plasma SCFA levels with a most pronounced decline in butyrate. As butyrate stabilizes tight junctions in gastrointestinal epithelia, lower butyrate concentration may further contribute to destabilizing the gut barrier in NOD TLR4 Ϫ/Ϫ mice (44). Reduced SCFA levels in blood but not in the gut of NOD TLR4 Ϫ/Ϫ mice may result from altered peripheral consumption of SCFA. These findings further suggest that SCFA in the circulation rather than in the gut lumen are decisive for the progression of insulindeficient diabetes. Of note, a recent study demonstrates the relevance of peripheral but not fecal SCFA for regulation of insulin sensitivity and lipolysis in humans (45). Moreover, the observation of reduced peripheral SCFA concentrations in NOD TLR4 Ϫ/Ϫ mice with accelerated diabetes development are in line with results of a study in children on the protective effect of SCFA in early-onset T1D (27).
The key metabolic effects of SCFA result from binding to specific FFA receptors, which are involved in the control of intracellular lipid turnover (46). Activation of FFA receptors stimulates fatty acid oxidation and inhibits lipolysis resulting in reduced plasma FFA concentrations and decreased body weight (47). In the NOD TLR4 Ϫ/Ϫ mice, lower plasma SCFA were associated with increases in plasma FFA, TG, body weight, and fat mass despite unaltered food uptake in the prediabetic period. Augmented availability of (long-chain) FFA, mainly mediated by the diacylglycerol/protein kinase C pathway (26,48), is most likely responsible for the marked whole-body insulin resistance in NOD TLR4 Ϫ/Ϫ mice. This view is supported by the finding that NOD TLR4 Ϫ/Ϫ mice exhibit low-insulin sensitivity along with lower butyrate, but higher FFA levels than the insulin-sensitive NOD TLR4 ϩ/ϩ mice.
The elevation of circulating FFA may result from excessive production and/or impaired mitochondrial oxidation (49). Interestingly, NOD TLR4 Ϫ/Ϫ mice featured increased oxidative capacity only in liver and not in skeletal muscle. Altered substrate availability might have contributed to the increased hepatic oxygen flux, supporting our recent findings of augmented hepatic mitochondrial respiration in prediabetic NOD mice (8) and in severely obese humans (30). Of note, this will promote hepatic oxidative stress and challenge hepatocellular anti-oxidant defense mechanisms and further worsen energy metabolism and insulin resistance (30). Recent studies confirm abnormalities of hepatic mitochondrial function and energy homeostasis also in patients with overt T1D (49). In addition, the liver-derived protein, fetuin A, could have served as a mediator between hyperlipidemia and metabolic balance, due to its described function as adaptor mediating FFA binding to TLR4 interfering with insulin signaling (50). However, fetuin A levels were independent of the TLR4 expression status and tissuespecific mitochondrial function in these mice. Likewise, hepatic mitochondrial adaptation occurs independently of fetuin A in obese humans (30). Thus, these data exclude a relevant role of fetuin A-FFA-TLR4 interaction in the observed metabolic changes.
The profound metabolic disturbances in normoglycemic NOD TLR4 Ϫ/Ϫ mice support the view that the metabolic disturbances mediated by TLR4 deficiency affect the pathogenesis

Gut microbiota and diabetes progression
of the disease already at an early, prediabetic stage. According to the model outlined in Fig. 8, NOD TLR4 Ϫ/Ϫ mice exhibit an altered composition of gut bacteria characterized by an increased abundance of Bacteroidetes and a decrease of Firmicutes. The animals further show impaired gut morphology due to muscular atrophy and increased MMP2 activity. Increased peripheral levels of (long-chain) FFA and TG further point to an altered absorption of nutrients and an impaired gut barrier in NOD TLR4 Ϫ/Ϫ mice. Altered substrate availability as a consequence of hyperlipidemia results in impaired insulin sensitivity associated with increased respiratory activity of hepatic mitochondria. Decreased peripheral levels of SCFA, particularly butyrate, which are linked to lower body weight and improved glycemic control, are insufficient to counteract the impairment of insulin sensitivity and to normalize glucose homeostasis. Under conditions of impaired insulin sensitivity, an increased demand for insulin might impose further "metabolic" stress on beta cells that are already exposed to enhanced inflammatory stress mediated by the rapidly-progressing insulitis as observed in prediabetic NOD TLR4 Ϫ/Ϫ mice (14). This accelerates betacell failure and thereby progression to insulin-deficient diabetes in NOD TLR4 Ϫ/Ϫ mice.
In conclusion, NOD TLR4 Ϫ/Ϫ mice allowed us to identify the associations between distinct alterations of gut morphology and microbiome composition with metabolic disorders promoting the progression of insulin-deficient diabetes/T1D. Accelerated diabetes development likely results from altered hepatic respiratory activity and impaired insulin sensitivity. Circulating SCFA may serve as important mediators in these processes. Taken together, these findings identify TLR4-dependent metabolic pathways as promising targets for intervention strategies aimed at the preservation of beta-cell function in individuals at risk of T1D.

Experimental procedures
Animals NOD mice were from the breeding colony at the German Diabetes Center. TLR4-expressing C57BL/10ScSn (C57BL/10 TLR4 ϩ/ϩ ) and C57BL/10ScCr mice lacking TLR4 expression (C57BL/10 TLR4 Ϫ/Ϫ ), due to a spontaneous deletion of the TLR4-encoding region (51), were from the Max-Planck-Institute for Immunobiology and Epigenetics, Freiburg, Germany. The TLR4 defect was backcrossed onto the NOD background for more than 12 generations, and heterozygous littermates were intercrossed to generate NOD TLR4 ϩ/ϩ and TLR4 Ϫ/Ϫ animals (14). Unless otherwise indicated, the animals had free access to water and standard diet (Ssniff M-Z Extrudat, 4.5% fat; SSNIFF Spezialdiäten GmbH, Soest, Germany). All experiments were performed with co-housed female mice at the age of 70 -90 days derived from breeding pairs carrying a heterozygous TLR4 defect. Female mice were monitored for the development of diabetes until 220 days of age. Animals were considered diabetic with blood glucose levels (measured on an EPOS Analyzer 5060, Eppendorf, Hamburg, Germany) exceeding 250 mg/dl on 2 consecutive days. All experiments were conducted in accordance with the Principles of Laboratory Animal Care (GV-SOLAS, Society for Laboratory Animal Science) and were approved by the ethics committee on animal welfare of the State of North Rhine Westphalia.

Indirect calorimetry and body composition
Mice were individually placed in an 8-chamber indirect calorimetry system (PhenoMaster, TSE Systems, Bad Homburg, Germany). After 24 h of adaptation, parameters of indirect calorimetry (RQ), physical activity (interruption of IR light beam), and the intake of water and food were simultaneously analyzed for each mouse over 48 h (8). Body composition of mice was assessed by an Echo MRI body composition analyzer system (EchoMRI TM -100 System, Echo Medical Systems, Houston, TX) allowing the quantification of fat and lean body mass (52).

Glucose and insulin tolerance tests
The ipGTT was performed in 70 -90-day-old animals. It was commenced after 6 h of fasting by injecting 2 g/kg body weight glucose and measuring the glucose levels in blood obtained from the tail vein before (baseline) and 5, 15, 30, 60, 90, and 120 min after injection. The ipITT was performed in 70 -90-day- Insulin-sensitive metabolic processes modulate the development of T1D, specifically the accelerated progression of diabetes in the NOD TLR4 Ϫ/Ϫ mouse resulting from altered microbiota, which lead to impaired metabolic homeostasis, thereby promoting insulin-deficient diabetes.

Glucose-stimulated insulin secretion from isolated islets
Pancreatic islets were isolated from normoglycemic female mice by collagenase digestion (Serva, Heidelberg, Germany), as described previously (54). The islets were exposed to increasing glucose concentrations (2.8 -28 mM) in Hanks' balanced salt solution for 24 h (37°C, 5% (v/v) CO 2 ). The amount of insulin released from the cultivated islets was measured by ELISA (Mouse Insulin ELISA kit, Mercodia, Uppsala, Sweden).

Morphological analysis and immunofluorescence staining of intestinal tissue
Paraformaldehyde-fixed frozen thin sections of the small intestine were stained with hematoxylin and eosin. Images were acquired using a Leica DMRBE inverted microscope equipped with an Olympus DP73 digital color camera. CellSens imaging software version 1.7 (Olympus Life Science, Hamburg, Germany) was used for morphometric analyses of muscle thickness, villi length and width, as well as crypt length. Paraformaldehyde-fixed thin sections of small intestinal tissue were incubated with a polyclonal rabbit anti-CD3 antibody (DAKO, Hamburg, Germany) followed by incubation with a secondary Alexa Fluor 488 -conjugated donkey anti-rabbit antibody (Life Technologies, Inc., Darmstadt, Germany) and Hoechst 33342 dye (Sigma, Steinheim, Germany) for nuclei staining.

Assessment of MMP-2 activity
MMP-2 activity was assessed by gelatin zymography in homogenates of intestinal tissue, as described previously (55). Collagenase served as a positive control for proteinase activity. Band intensities were quantified using the software Fiji (National Institute of Health, Bethesda, MD). Resulting values were used as relative arbitrary units for MMP activity.

High-resolution respirometry
Ex vivo mitochondrial function was measured by highresolution respirometry (Oxygraph-2k, Oroboros Instruments, Innsbruck, Austria) as described previously (8). Briefly, fresh liver and soleus muscle samples were permeabilized by saponin treatment, and defined respiratory states were obtained by the following multiple substrate-inhibitor titration protocol: 2 mM malate, 10 mM pyruvate, 10 mM glutamate, and 2.5 mM ADP for state 3 respiration of complex I; 10 mM succinate for combined state 3 respiration of complex I and II; 10 M cytochrome c for mitochondrial membrane integrity check; carbonyl cyanide p-trifluorome-thoxyphenylhydrazone (FCCP) (stepwise increments of 0.25 M up to the final concentration of 1.25 M) for maximal respiratory capacity, namely state u, and 2.5 M antimycin A. Addition of cytochrome c did not increase oxygen consumption indicating integrity of the outer mitochondrial membrane after saponin permeabilization. The results of the analyses were expressed as pmol/(s ϫ mg) after correction for mitochondrial copy numbers as described (8).

Extraction of genomic DNA and profiling of the 16S rRNA gene by NGS
Genomic DNA was isolated from gut contents by applying the DNA stool mini kit (Qiagen, Hilden, Germany) according to the manufacturer's protocol. DNA concentrations and the OD 260/280 ratio for quality control were determined using the NanoDrop 1000 spectrophotometer (Thermo Scientific, Waltham, MA). A minimum OD 260/280 ratio of 1.9 was regarded as sufficient for further analyses.
The bacterial DNA samples were profiled by sequencing of the V4 region of the 16S rRNA gene on an Illumina MiSeq (Illumina RTA version 1.17.28; MCS version 2.5) using 515 forward and 806 reverse primers designed for dual indexing and the V2 kit (2 ϫ 250 bp paired-end reads). For details of DNA isolation, next generation sequencing, and evaluation, see supporting information (Extraction of genomic DNA and profiling of the 16S rRNA gene by Next Generation Sequencing (NGS)).

Measurement of SCFA
SCFA were extracted from plasma samples and from samples of the contents of ileum, cecum, and colon and analyzed as described previously (56 -58). The acids were quantified using GC-MS. Adequate internal standards of the SCFA were included in all steps of the preparation and analysis procedures. SCFA concentrations were quantified from calibration curves of SCFA that underwent the same preparation and analysis procedures.

Statistical analyses
Data are expressed as mean values Ϯ S.E., and statistical analysis was performed using nonlinear regression, ANOVA with Bonferroni post hoc analysis, or Student's t test. Differences were considered statistically significant at p Ͻ 0.05. All statistical analyses were performed using the Prism software package version 4 (GraphPad Software, San Diego, CA).