Aberrations in gut microbiota are associated with metabolic disorders, including obesity. However, whether shifts in the microbiota profile during obesity are a characteristic of the phenotype or a consequence of obesogenic feeding remains elusive. Therefore, we aimed to determine differences in the gut microbiota of obese-prone (OP) and obese-resistant (OR) rats and examined the contribution of this microbiota to the behavioral and metabolic characteristics during obesity. We found that OP rats display a gut microbiota distinct from OR rats fed the same high-fat diet, with a higher Firmicutes-to-Bacteroidetes ratio and significant genera differences. Transfer of OP but not OR microbiota to germ-free (GF) mice replicated the characteristics of the OP phenotype, including reduced intestinal and hypothalamic satiation signaling, hyperphagia, increased weight gain and adiposity, and enhanced lipogenesis and adipogenesis. Furthermore, increased gut permeability through conventionalization resulted in inflammation by proinflammatory nuclear factor (NF)-κB/inhibitor of NF-κB kinase subunit signaling in adipose tissue, liver, and hypothalamus. OP donor and GF recipient animals harbored specific species from Oscillibacter and Clostridium clusters XIVa and IV that were completely absent from OR animals. In conclusion, susceptibility to obesity is characterized by an unfavorable microbiome predisposing the host to peripheral and central inflammation and promoting weight gain and adiposity during obesogenic feeding.
The overabundance of energy-dense foods in developed westernized societies has transformed obesity from an American burden to a worldwide epidemic, with grave health and socioeconomic consequences. Recent advancements in DNA sequencing techniques and metagenomic profiling have established a link between the trillions of microbial inhabitants of the gut (i.e., gut microbiota) and the development of obesity-related metabolic dysregulations (1). Gut microbiota are significantly altered in humans and animal models of obesity, with reduction in bacterial diversity (2) as well as overall compositional shifts, such as a reduced abundance of Bacteroidetes and a proportional increase in Firmicutes phylum (3–6). Gut microbiota regulate several host metabolic functions, and microbial dysbiosis is associated with altered energy homeostasis (1). Obesity is characterized by an enrichment in genes encoding enzymes responsible for extracting calories from otherwise indigestible polysaccharides (2,4). In addition to increased energy harvest, the microbial metabolic byproducts, such as short-chain fatty acids (FAs), modulate secretion and gene expression of gut peptides controlling satiety, such as glucagon like peptide-1 (GLP-1) and peptide YY (PYY), by acting on G-coupled protein receptors (GPRs) in intestinal enteroendocrine cells, suggesting a role for gut microbiota in modulating satiation (7). Obesity and high-fat (HF) feeding are also associated with intestinal, systemic, and adipose tissue inflammation (1). Intestinal inflammation is an early consequence of HF feeding, present before the onset of obesity and insulin resistance (8), supporting a direct, causal role for HF-induced microbiota changes in the development of obesity.
Conventionalization studies have demonstrated the contribution of the gut microbiota to the development of metabolic disease because the metabolic phenotype is transmissible by gut microbiota transplantation (3,4,9). These early studies investigated the effect of gut microbiota using genetic, transgenic, or HF-fed models of obesity, none of which are an accurate reflection of human obesity that encompasses the interaction between genes and the environment (3,4,9,10). Therefore, in the current study, we examined the role of the gut microbiota in obese-prone (OP) and obese-resistant (OR) rats, a model that closely mimics the characteristics of the human obese phenotype, including a polygenic mode of inheritance, whereby some, but not all, individuals are susceptible to weight gain when exposed to an obesogenic environment (10). First, by using both chow- and HF-fed OP and OR rats, we determined whether microbiota shifts solely result from consumption of an obesogenic diet or are a manifestation of the obese phenotype. Second, we transplanted microbiota from OP and OR rats fed a HF diet into germ-free (GF) mice to examine the capacity of microbiota to influence phenotype and associated metabolic and molecular changes, including gut–brain satiation signaling, lipid storage, and systemic and hypothalamic inflammation.
Research Design and Methods
All experiments were done in accordance with the European Guidelines for the Care and Use of Laboratory Animals.
OP-CD (n = 7) and OR-CD (n = 4) rats from Charles River Laboratories (Wilimington, MA) were used. Animals were housed individually in a temperature-controlled vivarium with 12:12-h light/dark cycle (lights on at 0700). Starting at 8 weeks of age, rats were fed a HF diet (4.2 kcal/g; D12334B; Research Diets, New Brunswick, NJ) for 12 weeks, until sacrifice (11). Additional OP (n = 5) and OR (n = 4) rats, kept in the same housing conditions, were maintained on chow throughout and used for microbiota analysis.
Male C57BL/6J GF mice (n = 20) from our GF colonies (Animalerie Axénique de Micalis, Jouy-en-Josas, France) were used for inoculation studies. Two groups (n = 10 each) were housed separately in two Trexler-type isolators (Igenia, France), with animals housed individually in polycarbonate cages with cedar bedding.
GF mice (n = 10 for each phenotype) were conventionalized (CV) with fecal microbiota from one OP and OR donor rat, both maintained on the HF diet. Feces from each donor were freshly collected, quickly diluted, and homogenized in liquid casein yeast medium (1:100 w/v) in anaerobic conditions. GF mice (12 weeks old) were inoculated immediately thereafter by oral gavage (250 µL) and maintained on standard chow (2.83 kcal/g; R03, Safe Diets) for 4 weeks in separate gnotobiotic isolators. Afterward, half of each group was switched to a HF diet (4.73 kcal/g; D12451, Research Diets) for 8 weeks, while the remaining half was fed chow. All animals were killed and tissue collected for analyses.
Quantitative RT-PCR and Western Blotting
For all experiments, protein and RNA extraction, and subsequent Western blotting and quantitative (q)PCR were performed as previously described (11). Briefly, RNA was extracted by using TRiZol, quantified, 10 μg RNA was reverse transcribed, and cDNA was diluted fivefold for qRT-PCR using TaqMan gene expression assays (Applied Biosystems). Data are expressed as the relative mRNA normalized to β-actin and analyzed according to the 2–ΔΔCT method.
For Western blotting, tissues were thawed on ice and suspended in 1 mL radioimmunoprecipitation assay buffer containing protease inhibitors (Sigma-Aldrich, Lyon, France). Cells were lysed, homogenized, and centrifuged for 20 min at 14,000g at 4°C. After quantifying, soluble protein (25–100 µg) was run on SDS-PAGE gels containing 8–12% acrylamide, transferred to nitrocellulose membranes, and probed with antibodies (Santa Cruz Biotechnology and Abcam). Immune complexes were detected by chemiluminescence and quantified by scanning densitometry using ImageJ software against β-actin (cytosolic proteins) or Ras-related nuclear protein (nucleus proteins) as internal controls.
Plasma was analyzed for glucose, triglycerides, total cholesterol, and total HDL by an AU 400 automated biochemical analyzer (Olympus). Gut hormones and cytokines were determined in duplicate using a mouse gut hormone panel or mouse cytokine magnetic bead panel (Millipore), measured with Luminex technology (St. Antoine, Paris, France), following the manufacturers’ instructions.
Adipose Tissue Immunodetection
Adipose tissue was fixed with 4% formaldehyde overnight, stored in 75% ethanol, embedded in paraffin, and 4-µm-thick microtome cut sections were processed using standard procedures. For macrophage infiltration, sections were incubated overnight at 4°C with primary anti-F4/80 antibody (sc:71088, 1:100; Santa Cruz) followed by 1-h incubation with secondary antibody (goat antirat IgG, sc-2041, 1:200), then developed with DAB (Dako Kit-K3465) for ∼5 min. The number of F4/80+ cells per microscope field was counted and divided by the total number of adipocytes in the field (percentage macrophages-to-adipocyte), with five to eight fields counted per animal sample (n = 4 animals per group). For immunofluorescence of tumor necrosis factor-α (TNF-α), sections were incubated overnight at 4°C with primary, anti–TNF-α antibody (D2D4 #11948, 1:100; Cell Signaling Technology), followed by 1-h incubation at room temperature in the dark with secondary anti-rabbit IgG antibody (Alexa Fluor 488 Conjugate #4412, 1:200; Cell Signaling). Images were acquired with an inverted confocal microscope LSM510 with ×40 oil-immersed objective using AxioVision SE64 Rel.4.8 software. The corrected total cell fluorescence (CTCF) was computed by ImageJ software using uniformly sized adipocyte images and handled based on the mean fluorescence detected in the internal controls sections. CTCF = integrated density – (area of selected cell × mean fluorescence of background readings). Background readings were run to control for false negatives or positives of the first and secondary antibody.
Immunofluorescence for GLP-2 Enteroendocrine Cells and Intestinal Macrophage Infiltration
For GLP-2, sections were incubated overnight at 4°C with primary anti–GLP-2 (C-20) (sc-7781, 1:100; Santa Cruz), followed by 1-h incubation at room temperature in the dark with secondary antibody (anti-goat IgG-CFL 488: sc-362255, 1:200; Santa Cruz). GLP-2 containing EEC was quantified by counting total villi and GLP-2+–stained cells throughout the entire length of the section (>20 nonoverlapping microscopic areas) for each animal (n = four per group). For macrophage infiltration, sections were incubated with mouse anti-F4/80+ antibody and revealed with immunofluorescence secondary antibody coupled with goat anti-mouse IgG (H&L)–Alexa Fluor 647 (1:100). Data were expressed as the number of F4/80+ macrophages per high-power field (HPF: visible area of a slide under magnification). Between three and six fields were counted per sample (eight samples) for the number of F4/80+ macrophages per HPF. All images were acquired with an inverted confocal microscope LSM510 with ×40 oil-immersed objective and processed by AxioVision SE64 Rel.4.8 software. Controls were run for false negatives or positives of the first and secondary antibody.
Total DNA was extracted from 200 mg caecal content from 20 rats (chow: OP = 5, OR = 4; HF: OP = 7, OR = 4) and 16 mice (CVOP = 9, CVOR = 7) (12). Microbiota composition was assessed by 454 pyrosequencing (GS FLX TI technology; Genoscreen, Lille, France) targeting the V3-V4 region of the bacterial 16S rRNA gene (V3 forward: 5′TACGGRAGGCAGCAG3′; V4 reverse: 5′GGACTACCAGGGTATCTAAT3′). Sequences were trimmed for barcodes, PCR primers, and binned for a minimal sequence length of 300 pb, a minimal base quality threshold of 27, and a maximum homopolymers length of 6. Resulting sequences were assigned to the different taxonomic levels, from phylum to genus using the Ribosomal Database Project (release 10, update 26;release 11, update 1) (13). Sequences were further clustered into operational taxonomic units (OTUs) or phylotypes at 97% of identity using Quantitative Insights Into Microbial Ecology (QIIME) pipeline (14) and cdhit (15). OTUs were assigned to the closest taxonomic neighbors and relative bacterial species using Seqmatch and Blastall.
Data are expressed as mean ± SEM. Unless otherwise noted, statistics were performed by GraphPad Prism 5 software or R software (packages ade4, randomForest, and lattice). For statistical analysis among groups during HF feeding, a t test with Welch’s corrections was used, and ANOVA with Bonferroni post hoc test was used for analyses between groups and diet conditions. Principal component analysis was computed based on bacterial genus composition and statistically assessed by a Monte Carlo rank test. The Wilcoxon test was applied for differences in bacterial composition. Significance was accepted at P < 0.05. For heat maps representation, log10-transformation was applied on the bacterial relative abundance data matrix, which allowed visualizing similarities or differences between samples that affect members of the community that may make up less than 1% of the relative abundance in a sample. Machine learning techniques, using the random forest classifier, were applied to relative abundance data of distributed bacterial genera (16).
OP and OR Rats Have Different Gut Microbiota Profiles Only During HF Feeding
After 12 weeks of HF feeding, OP rats had increased 24-h food consumption, weighed significantly more, and had increased adiposity relative to OR rats (Supplementary Fig. 1). Among the 198,611 obtained 16S rRNA gene sequences, 121,887 passed quality filters and were further assigned to taxonomic levels, from phylum to bacterial species and OTUs. As highlighted by the principal component analysis, OP and OR microbiota did not differ during chow feeding. However, 12 weeks of HF feeding resulted in a differing microbiota of the HF-fed rats compared with their chow-fed counterparts, and more interestingly, a divergence between the OP and OR phenotypes during the HF-feeding period (Fig. 1A). There were no differences at the phyla level during chow feeding, but HF feeding resulted in significantly increased Bacteroidetes and Proteobacteria percentages and decreased Firmicutes in both groups. Moreover, during HF feeding, OP rats had significantly less bacteria from Proteobacteria phyla (Fig. 1B) and a trend toward more bacteria belonging to the Firmicutes (P = 0.06667), with an average Firmicutes-to-Bacteroidetes ratio higher in OP rats than in OR rats (1.67 vs. 1.28). During chow feeding, very few differences were observed at deeper taxonomic levels, except for bacteria belonging to the genus Clostridium cluster XIVb and Flavonifractor. HF feeding resulted in many significant genus differences between chow- and HF-fed animals (Supplementary Table 1). More interesting is that specifically during HF feeding, within the Firmicutes phylum, OP rats exhibited significantly higher proportions of Ruminococcus, Oscillibacter, and Alistipes genera (Fig. 1C) compared with HF-fed OR rats. On the other hand, Bacteroidetes, β-Proteobacteria (Parasutterella) and γ-Proteobacteria (Escherichia/Shigella) were significantly more represented in OR rats. This is intriguing, given that high levels of Escherichia coli were also noticed in humans undergoing weight loss after gastric bypass (5). Supervised classification with random forest classifier analysis demonstrated the variable importance of genera predicted to belong to OP and OR phenotypes fed the HF diet, with an estimate error rate of 20% (17% for OP and 25% for OR). The main predictive genera for phenotype discrimination were unclassified Ruminococcaceae, unclassified Porphyromonadaceae, Clostridium cluster XIVb, Escherichia, Parasutterella, and Alistipes (Supplementary Fig. 1). On the contrary, for animals fed the chow diet, the estimate error rate of supervised classification was higher at 56%, with a 75% error rate for OP classification and 40% for OR classification. Most important, genera in data classification were related to Flavonifractor, Clostridium cluster XIVb, Johnsonella, unclassified Clostridiales, and Lachnospiraceae incertae sedis, although bacterial genera discriminatory power was weak in chow-fed rats. Interestingly, some specific bacterial species were differentially represented between OP and OR rats (Fig. 1D; see Supplementary Table 1 for all species differences), and in particular, Bacteroides vulgatus was reduced in OP rats, similar to obese children (17).
Gut Microbiota Transfer Replicates Signatures of OP Phenotype
To determine the overall strength of the OP and OR microbiota profiles in the promotion and generation of their metabolic phenotypes, we inoculated GF-C57BL/6J mice (n = 10/phenotype) with microbiota from an OP or OR donor, termed CVOP and CVOR, respectively, that were fed chow or the HF diet. Strikingly, phenotype and behavioral differences between OP and OR rats were reliably transferred to CVOP and CVOR animals maintained on the HF diet. After 2 weeks, HF-CVOP mice gained significantly more weight than their chow counterparts, and after 7 weeks, were heavier than the HF-CVOR mice (Fig. 2A). More importantly, mice inoculated with OP microbiota had a significantly greater adiposity index (∼30% increase) than CVOR mice during HF feeding but not chow feeding. Similar to OP rats, 24-h food intake of CVOP mice was increased only during HF feeding (Fig. 2B). In addition, homeostasis model assessment of insulin resistance (Fig. 2C) and circulating leptin and insulin levels were significantly increased in HF-CVOP animals (Fig. 2D), as were triglyceride and glycemia levels (Supplementary Fig. 2), features all associated with metabolic syndrome.
Similar to OP donors, we found that hyperphagia in HF-CVOP mice was associated with reduced plasma GLP-1 and PYY (Fig. 2E) as well as decreased intestinal PYY and GLP-1 protein expression (Supplementary Fig. 2). Furthermore, gene and protein expression of intestinal GPRs, which mediate nutrient-induced satiety peptide secretion (7,18,19), was increased in OP compared with OR rats, an effect transferred to GF recipients, irrespective of their maintenance diet (Fig. 2F).
OP Microbiota Enhances Adipogenesis and Lipogenesis
Gut microbiota alter expression of host genes regulating adipogenesis and lipogenesis (20). In adipose tissue, HF-CVOP mice exhibited increased FA synthetase (FAS) and a reduced phosphorylated-acetyl-CoA carboxylase (ACC)–to–total ACC ratio, indicating increased capacity for de novo FA synthesis. Furthermore, sterol regulatory element-binding protein-1c (SREPB-1c), a key transcription factor of glucose-induced hepatocyte lipogenesis and activator of ACC and FAS, was significantly upregulated in CVOP mice (Fig. 3A and B). In addition, angiopoietin-like 4 (Angptl4), an inhibitor of lipoprotein lipase, was reduced in HF-OP donors and their GF recipients (Fig. 3A and Supplementary Fig. 4). This is significant, because others have shown that intestinal Angptl4 is directly regulated by gut microbiota, consistent with our observations of reduced intestinal Angptl4 in CVOP animals (data not shown) (20). As expected, all of these changes in lipogenic markers and enzymes resulted in increased adipocyte hyperplasia in CVOP mice (Fig. 3E). Furthermore, we found identical trends in the livers of HF-CVOP mice, where FAS and SREPB-1c were increased, but phosphorylated (p)-ACC/ACC and Angptl4 were both reduced (Fig. 3A and B). In addition, peroxisome proliferator–activated receptor γ (PPARγ), a transcription factor of adipogenesis, was increased in CVOP mice (Fig. 3C). In the liver, PPARα protein levels were significantly decreased, contributing to reduced hepatic FA oxidation and increased circulating triglycerides. Finally, protein expression of adipose and hepatic FA transporters, cluster of differentiation 36 (CD36), and FA binding protein (FABP) were increased in OP and CVOP compared with OR rats and CVOR mice during HF feeding but not chow feeding (Fig. 3D and Supplementary Fig. 4), supporting its role in increased intracellular shuttling of FAs and lipid accumulation in adipose and hepatic tissue.
“Obese” Microbiota Alters Tight Junction Proteins and Increase Inflammation
During HF feeding but not chow feeding, OP and CVOP animals both exhibited altered tight junction protein levels. Specifically, zonula occludens protein-1 (ZO-1) and occludens levels were decreased in distal intestinal epithelial cells, whereas phosphorylation of myosin light chain, a mechanism that results in cytoskeleton contraction and disruption of tight junction integrity (21), was increased in HF-OP and -CVOP animals (Fig. 4A and Supplementary Fig. 5). Furthermore, jejunal and colonic TNF-α, a biomarker of intestinal inflammation, was increased in HF-CVOP animals (Fig. 4C), likely due to enhanced intestinal macrophage infiltration (Fig. 4E and F). Impairment in markers of tight junction disruption of HF-CVOP mice by gut microbiota was further associated with significant decreases in L cells immunostained with GLP-2, a trophic gut hormone shown to regulate gut permeability (22), in the ileum and colon, which contain the largest amounts of L cells (Fig. 4D). HF-fed OP rats also had higher circulating levels of the inflammatory markers TNF-α, monocyte chemotactic protein-1 (MCP-1), macrophage inflammatory protein-1 α (MIP-1α), interleukin (IL)-6, and IL-1α (Supplementary Fig. 5). CV with OP gut microbiota replicated increases in plasma TNF-α, MIP-1α, IL-6, and IL-1α in OP compared with OR recipients fed the HF diet (Fig. 4B).
“Obese” Microbiota Increases Adipose Tissue and Liver Inflammation
We observed increased macrophage infiltration in CVOP compared with CVOR mice fed the HF diet (37.1% vs. 27.4%, P < 0.05; Fig. 5A), as well as increased total F4/80 and CD3 (Supplementary Fig. 6), markers of macrophage and T-cell infiltration, respectively. Activation of Toll-like receptor 4 (TLR4) regulates inflammatory process and, as such, TLR4 mRNA levels were increased in adipose tissue of HF-fed OP donors and their CVOP recipients (Fig. 5B and Supplementary Fig. 6), as were the expression of inflammatory chemokines TNF-α, plasminogen activator inhibitor-1 (PAI-1), and IL-6 (Fig. 5C and F and Supplementary Fig. 6). Similarly, HF feeding resulted in upregulation of inflammatory gene expression TNF-α and IL-6 in the liver of OP donor rats and their GF mice recipients (Fig. 5D and Supplementary Fig. 6).
Obese animals exhibited an increase in the inflammatory nuclear factor (NF)-κB/inhibitor of κβ kinase (IKKβ) pathway. Specifically, phosphorylation/activation of the p65 subunit of NF-κB at its ser536 residue and of IKKβ at residues ser177 and ser181 was significantly increased in adipose and hepatic tissue of OP and CVOP animals during HF feeding but not chow feeding (Fig. 5E and Supplementary Fig. 6). Activation of IKKβ allows NF-κB to dissociate and enter the nucleus and induce gene expression of inflammatory factors. IKKβ activation is accomplished by activation of TLR4, TNF receptor, and IL-1 receptor (23). Here we demonstrate that TLR4 and its TNF-α ligand are increased in adipocytes.
“Obese” Microbiota Increases Hypothalamic Inflammation and Satiety Neuropeptide Expression
CV of GF mice with OP microbiota significantly increased expression of hypothalamic IL-6, TNF-α, and TLR4 genes compared with CVOR during HF feeding (Fig. 6A). When examining the effect of CV on hypothalamic energy-regulating peptides, we found that CVOP mice fed the HF diet had reduced proopiomelanocortin and increased Agouti-related peptide and neuropeptides Y (Fig. 6B). These changes were reliably transferred from OP donors characterized by decreased anorexigenic and increased orexigenic peptides (24,25), and are the first demonstration of the ability of the gut microbiota to alter central nervous system (CNS) energy homeostatic-signaling proteins.
Microbiota-Related Phenotype Is Preserved and Transferred to GF Mice
Although specific microbial phyla present in the OP and OR donors were also detected in recolonized GF mice, phyla differences were not replicated, because the average Firmicutes-to-Bacteroidetes ratio was not different in CVOP (1.22) versus CVOR (1.45) mice. However, bacterial genera distribution differed between HF-fed CVOP and CVOR mice (Fig. 7A), and several genera differences between OP and OR donors were replicated in GF recipients (Fig. 7B; see Supplementary Fig. 1 for supervised classification analysis). Clostridium cluster IV and Oscillibacter from the Clostridiaceae family and some unclassified Ruminococcaceae were significantly more represented in OP and CVOP compared with OR animals. Furthermore, we observed 25 bacterial molecular species (or OTUs) that were specific to OP and CVOP but were totally absent from OR and CVOR animals (Fig. 7C and Supplementary Table 7). As such, the 25 OTUs all clustered within the Firmicutes phylum and were related to 10 bacterial isolates, mostly from Clostridium cluster XIVa, Clostridium cluster IV, and Oscillibacter. Clostridium cluster XIVa, part of the Eubacterium rectale-C. coccoides group contain the currently recognized butyrate-producing bacteria in the gut (26), resulting in more efficient energy extraction from the diet. Furthermore, Clostridia strains from Clostridium cluster XIVa and Clostridium cluster IV have been shown to evoke a proinflammatory cytokine response in vitro (27), whereas Oscillibacter has been positively correlated with gut permeability (28). Therefore, the potential ability of these specific bacterial strains to increase energy harvest and promote intestinal inflammation could explain the increases in adiposity and gut permeability and the subsequent systemic inflammation in OP donors and their GF recipients fed a HF diet.
Our studies provide new evidence demonstrating that 1) OP and OR phenotypes are associated with distinct and differing gut microbial communities during HF feeding that are not present during chow feeding and that 2) transfer of OP microbiota replicates the obese phenotype of the donor as well as associated differences in the chemosensory, metabolic, and neural dysregulations. Although some studies suggest the presence of a specialized “obese” microbiota capable of increased energy storage, most are confounded by the observation that obesity often results from an obesogenic, western diet known to rapidly alter the gut microbiota (29), thus making it difficult to ascertain the influence of the host metabolic phenotype versus the diet on microbial composition during the obese state. Indeed, HF feeding of humanized gnotobiotic mice results in a rapid shift in the microbiome and its metabolic pathways preceding increased adiposity (29). Similarly, genetically OR mice exhibit decreased Bacteroidetes and increased Firmicutes during HF feeding, emphasizing that diet, and not host phenotype, may be the main determinant of microbiota shifts (30). However, our results clearly demonstrate that the “obese” gut microbiota profile is not a mere result of HF feeding but instead is unique and conserved to the obese state, because chow-fed OP and OR rats exhibited similar microbiota profiles that diverged during HF feeding. Therefore, this distinct gut microbiota, with phyla-, genera-, and species-specific differences, is a signature of the obese host phenotype not only of HF feeding. This finding is consistent with results from a recent study demonstrating that transplantation of gut microbiota from twins discordant for obesity into GF mice resulted in increased body mass and adiposity of mice receiving the obese cotwins compared with the lean cotwin, whether the groups were maintained on a diet low or high in saturated fat (31).
We observed a high Firmicutes-to-Bacteroidetes ratio in OP rats; however, not all studies replicated the low Bacteroidetes levels in obesity (32), suggesting that differences at the genus and species levels play a greater role. Indeed, we found high levels of bacteria from the Ruminococcus genus in OP rats, similar to that observed in obese humans and HF-fed mice (33,34). Ruminococcus is phylogenetically heterogenous, and most species fall under several Clostridium clusters, including Clostridium clusters IV and XIVa. As such, C leptum (cluster IV) has been associated with both obesity and weight loss (35,36) and was increased in OP and CVOP animals. Further, OTUs only present in OP and their GF recipients were assigned to Clostridium cluster XIVa, which is directly correlated with fat pad mass and BMI (35,37,38) and contains bacterial species known to break down polysaccharides, promoting monosaccharide absorption, enhanced lipogenesis, and lipid storage (4,20). In agreement with this, OP microbiota CV resulted in increase in adipogenic and lipogenic enzyme expression in the liver and adipose tissue as well as increased FA transporters, providing a more efficient transfer and storage of fermented byproducts. Furthermore, we demonstrate, for the first time, the ability of cross-species transfer (rat to mouse) of the obese phenotype. Interestingly, OP and OR bacterial phyla-level differences were not found in CV animals, indicating that although overall phyla shifts may be indicative and representative of the obese state, they are not the main determinants in shaping the metabolic profile of the host recipient, emphasizing the role of more specific and detailed taxonomic differences in the development of obesity.
Obesity is characterized by increase in gut paracellular permeability likely promoting metabolic endotoxemia (22,39). In this regard, we found that OP donors and GF recipients exhibited disruptions in tight junction protein levels, indicative of increased gut permeability (39). This was associated with systemic inflammation, likely through lipopolysaccharide or saturated FAs, as indicated by increased adipose macrophage recruitment and hepatic NF-κB/IKKβ inflammatory-signaling pathways (40,41).
Our currents results provide new evidence showing that aberrations in microbiota during obesity result in alterations in peripheral and central satiety signaling that promote hyperphagia and weight gain. Gut microbiota modulate expression of intestinal nutrient receptors (7,11), and here we demonstrate that obesity-associated differences in GPRs were replicated in GF recipient mice, affecting the ability of enteroendocrine cells to sense and respond to intestinal nutrients. In line with this, intestinal and circulating satiety peptide levels were both reduced in OP and CVOP animals in addition to reduced L-cell number. Changes in gut peptide signaling and inflammation may also be secondary to increased adiposity and/or increased fat consumption, possibly a result of increased energy harvest from OP microbiota, although the relationship between diet, obesity, and energy extraction is much more complex than previously thought (42). Interestingly, HF feeding causes rapid microbiota shifts occurring before changes in adiposity (29), suggesting that direct microbe host cross talk influences intestinal-signaling mechanisms that precede and result in hyperphagia and promote weight gain. Indeed, certain bacterial species or their metabolic byproducts have been shown to influence gut peptide signaling by directly altering expression of signaling proteins of enteroendocrine cells or by activating enteroendocrine cells through GPR signaling, respectively (7,43–45). Nevertheless, future studies using pair-fed animals to control for dietary fat intake, or time course experiments assessing the development of obesity with changes in microbiota and intestinal satiation signaling, would aid in delineating their effects.
There is a growing appreciation of the bidirectional interaction between gut microbiota and the brain, with microbiota modulating CNS activity through endocrine and neural pathways (46). Impaired hypothalamic activity, as well as diet-induced hypothalamic inflammation, has been causally linked with the development of obesity (47,48). Interestingly, hypothalamic inflammatory signaling occurs rapidly within a few days of HF -feeding (49) that corresponds with changes in diet-induced microbiota shifts (29). Therefore, it is plausible that CNS inflammation is a direct result of aberrations in gut microbiota. Here we show, for the first time, the ability of the gut microbiota to alter hypothalamic anorexigenic and orexigenic peptides. Although this may be due to increased hypothalamic inflammation, which directly alters leptin and insulin resistance in these neurons and causes a shift in neuropeptides, it is possible that direct signaling pathways between the microbes and the brain may also affect central functions. For example, peripheral endotoxins and cytokines evoke neural activation through vagal afferents (50,51), whereas the stress-reducing effects of prebiotics require intact vagal signaling (52). Furthermore, several bacteria have the capacity to produce neurometabolites, such as γ-aminobutyric acid, serotonin, and dopamine (46). Nevertheless, the current study clearly shows that “obese” microbiota are associated with hypothalamic energy-regulating peptides that promote energy intake and adiposity.
In conclusion, this study demonstrates the broad and extensive contribution of the “obese” gut microbiota to the modulation of complex molecular-signaling machinery responsible for host metabolism, energy storage, intestinal nutrient sensing, and inflammatory pathways, Taken together, it demonstrates that humans susceptible to obesity may harbor a disadvantageous gut microbiome that exacerbates adiposity during HF feeding and could be used as a potential marker for susceptibility to obesity in humans. Further, by identifying and characterizing specific bacterial groups or species that play a major role in the promotion and perpetuation of obesity, these findings open the potential for future therapeutic treatments that target these organisms and alleviate the metabolic dysregulation of the host.
Acknowledgments. The authors thank Gary Schwartz, from Albert Einstein College of Medicine, and Jeffrey Felton, from Western University of Health Sciences, for their editorial input, and the Animalerie Axénique de Micalis team of l’Institut National de la Recherche Agronomique–Jouy-en-Josas, and Timothy Swartz, from Institut Cochin (INSERM), for assisting with the behavioral experiments.
Funding. The study was supported by INRA through a scientific package awarded to M.C. and by the Romanian National Program PN-II-ID-PCE-2012-4-0608 No. 48/02.09.2013, “Analysis of novel risk factors influencing control of food intake and regulation of body weight.”
Duality of Interest. No potential conflicts of interest relevant to this article were reported.
Author Contributions. F.A.D. and M.C. designed the study, researched data, and wrote the manuscript. Y.S. researched data and reviewed the manuscript. P.L. analyzed metagenomic data and reviewed the manuscript. F.D. and B.L. researched data. J.D. reviewed the manuscript. M.C. 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.
Prior Presentation. These data were partly presented in abstract form at the 2012 Experimental Biology Conference, San Diego, CA, 23 April 2012.