Mining the human intestinal microbiota for biomarkers associated with metabolic disorders

After birth, our gastrointestinal (GI) tract is colonized by a highly complex assemblage of microbes, collectively termed the GI microbiota, that develop intimate interactions with our body. Recent evidence indicates that the GI microbiota and its products may contribute to the development of obesity and related diseases. This, coupled with the current worldwide epidemic of obesity, has moved microbiome research into the spotlight of attention. Although the main cause of obesity and its associated metabolic complications is excess caloric intake compared with expenditure, differences in GI tract microbial ecology between individuals might be an important biomarker, mediator or even new therapeutic target. Nevertheless, it is currently still unclear which bacterial groups play a role in the development of the metabolic syndrome in humans. This might partly be explained by: 1. Biological factors such as the heterogeneity in genotype, lifestyle, diet; and the often complex aetiology of human disease of which the metabolic syndrome is no exception. 2. Technological factors, such as the use of miscellaneous incompatible methods to assess the gut microbiota, often enumerating specific groups rather than using broad 16S rRNA gene surveys or metagenomics. 3. Studies vary greatly in the populations considered, their designs, and the degree of control for potential confounding factors such as lifestyle and diet. Nevertheless, recent research on this matter has shown a conceptual shift by focusing on more homogenous subpopulations, based on stricter control over variables such age range or through the use of both anthropometric (weight, total body fat) as well as biochemical variables (insulin resistance, hyperlipidaemia) to define groups. Perturbations in microbial diversity and community structure in adults with overweight and obesity may be partly due to long-term dietary habits or physiological changes in these subjects. As such, exploring the association between the gut microbiota and variation in BMI and weight in early life, prior to or close to the onset of overweight, might provide additional insights into these processes. Therefore, we studied the fecal microbiota of 295 six-seven year old children from the KOALA Birth Cohort, living in the south of the Netherlands. This age range is relatively uncharted microbiota territory. We found that its composition seems to conform to tot same ecosystem rules as that of adults. The bimodal distribution pattern of several bacterial groups as well as their co-correlating groups that were reported previously, including Uncultured Clostridiales II (UCII), Prevotella spp. and Dialister were confirmed. Furthermore, one of the previously described bimodal groups (Uncultured Clostridiales I) was shown before to exhibit very clear shifting state probabilities associated with ageing, where the high abundance state was mainly observed above 40 years of age. This was corroborated as no support for bimodality of this group was observed in the children included in the study described here. A large part of the variation in microbiota composition was explained by the abundance of aforementioned groups in contrast to the anthropometric outcomes, suggesting that in this group of healthy children within a relatively normal weight range, weight and associated parameters were not major drivers of overall genus-level microbial composition or vice versa. Hereafter, multiple linear and logistic regression models with rigorous adjustment for confounders were applied to investigate individual microbiota features association with weight related anthropometric outcomes. Previously reported parameters such as diversity, richness and Bacteroidetes to Firmicutes ratio, were not significantly associated with any of the outcomes. Nevertheless, the abundance of several specific bacterial taxa; Akkermansia, Sutterella wadsworthia et rel. and Bryantella formatexigens et rel. and the dichotomous abundance state of the bi-modally distributed UCII was consistently associated with weight-related outcomes. Other biochemical features of the metabolic syndrome have been associated with the gut microbiome. Mainly rodent studies have indicated that antibiotic treatment may improve glucose homeostasis and metabolic impairments. Therefore, the effects of gut microbiota manipulation by antibiotics (7d administration of amoxicillin, vancomycin or a placebo) on tissue-specific insulin sensitivity, energy metabolism, gut permeability and inflammation in 57 obese, pre-diabetic men from the same geographical region, were investigated. Vancomycin decreased bacterial diversity and significantly reduced well known butyrate- producing Firmicutes from Clostridium clusters IV and XIVa and bacterial groups involved in bile acid metabolism. These changes occurred concomitantly with altered plasma and fecal concentrations of these metabolites. In adipose tissue, gene expression of oxidative pathways was upregulated by antibiotics, whereas immune-related pathways were downregulated by vancomycin. However, antibiotic treatment had no significant effects on tissue-specific insulin sensitivity, energy/substrate metabolism, postprandial hormones and metabolites, systemic inflammation, gut permeability and adipocyte size. Importantly, despite a still considerably altered microbial composition at eight weeks follow-up, energy harvesting, adipocyte size and whole-body insulin sensitivity (HOMA-IR) remained unaltered. Overall these data indicate that interference with adult microbiota by antibiotic treatment for 7 days had no clinically relevant impact on metabolic health in obese humans. These data are in contrast with several rodent studies as well as a human intervention. The present study, which was well-powered and placebo-controlled, indicates that the previously reported vancomycin-induced effects on human peripheral insulin sensitivity are probably of minor physiological significance. The aforementioned group that was relatively homogeneous with regards to phenotype was combined with another cohort with similar phenotypical characteristics (obese, male and pre-diabetic) from another region of the Netherlands, to investigate whether tissue specific insulin sensitivity, as measured by the golden standard hyperinsulinemic-euglycemic clamp technique, is related to a specific microbial pattern. Remarkably, despite the fact that both cohorts were constructed based on comparable recruitment strategies, the average microbiota composition in both cohorts showed pronounced differences. Firstly, we found no consistent and significant association between liver, adipose tissue or skeletal muscle insulin sensitivity and the microbiota in both cohorts. Nevertheless, Random Forests classifiers using microbiota composition as predictors revealed taxa associated with fasting glucose concentrations and HbAc1 but only in one cohort. The top microbial features distinguishing classes were different Proteobacteria and groups involved in butyrogenesis, such as Faecalibacterium prausnitzii, Roseburia intestinalis, and Eubacterium rectale and related species, for fasting glucose levels. For HbAc1 these taxa were Oscillospira guillermondii, Sporobacter termitidis, Lactobacillus gasseri and Peptococcus niger and related species. The striking cohort-specific observations suggest that the relation between microbiota composition and type 2 diabetes mellitus as well as other characteristics of the metabolic syndrome is very dependent on the selected cohort of patients and their respective baseline microbiota composition. Similar observations have been made by other researchers as well. It could be that differences in microbiota composition are not associated with the insulin resistance phenotype when the overweight and/or obese state of the patient is already established, as is the case for our metabolic syndrome patients. In the latter case we cannot exclude that the composition of the fecal microbiota may play a role in the worsening of insulin sensitivity in an early stage in the development from a lean towards an overweight/obese phenotype. Furthermore, the observation of a subgroup- specific microbiota only observed in one of the cohorts might indicate an alternative state of microbiota composition driven by yet unknown forces. Nevertheless, this study clearly demonstrated that cohort-specific microbiota differences hamper finding a consensus biological interpretation between cross-sectional studies. This, combined with the complexity of individual disease pathogenesis, as well as the individual-specific differences in microbiota composition, may explain the inconsistency in observations between different studies concerning the identification of signature microbes for obesity, irritable bowel syndrome and other diseases. Besides the biological drivers for cohort specific inconsistencies in identified microbial biomarkers, there are also technological factors. Although high-throughput sequencing of short, hypervariable segments of the 16S ribosomal RNA (rRNA) gene has transformed the methodological landscape describing microbial diversity within and across complex biomes, evidence is increasing that methodology rather than the biological variation is responsible for observed sample composition and distribution. Large meta-analyses would aid in elucidating whether the basis for these observed inconsistencies is biological, technical or maybe a combination of both. To facilitate these meta-analyses of microbiota studies we developed NG-Tax, a pipeline for 16S rRNA gene amplicon sequence analysis that was validated with different Mock Communities (MC). NG-Tax demonstrated high robustness against choice of region and other technical biases associated with 16S rRNA gene amplicon sequencing studies. The analysis of α- and β-diversity of these MC confirmed conclusions guided by biology rather than the methodological aspects. This pipeline was applied to biological samples to monitor the developing communities an in vitro gut model (TIM-2) fed either with a normal diet, or modified versions from which the carbohydrate (MPLC) or protein fraction was diluted (LPMC) for 72h. In combination with global metatranscriptomics and metabolomics this revealed that each diet produced distinct microbial communities and temporal patterns and ratios of metabolites. The microbiota in reactors fed diets containing normal carbohydrate levels were enriched in members of the genera Prevotella, Subdoligranulum, Blautia and Bifidobacterium, all associated with carbohydrate fermentation. In turn, the microbiota in the reactors fed the MPLC diet, containing ten-fold less carbohydrates, was enriched in the genus Bacteroides, which is associated with diets rich in protein and animal fat. This setup allows researchers to study the (trophic) interactions and task division within a community and how they are impacted by diet-related factors under controlled conditions, which may assist in defining causal links between specific diet-derived parameters microbial groups and their activities. In conclusion, currently it seems that GI microbiota based biomarkers associated with metabolic impairments and anthropometric variables associated with the metabolic syndrome are cohort specific or possibly individual, which could partly be due to the use of incompatible analytical approaches. Nevertheless, there is growing evidence that human health is a collective property of the human body and its associated microbiome and thus requires to study the interface of two very complex systems, i.e. on one side the extraordinary coding capacity, high inter-individuality and complex dynamics of the microbiome and on the other side the multifactorial individual nature of human disease. In light of these observations the manifestation of individual dynamics of the microbiota with the host when homeostasis is lost seems plausible and likely.