The paper in Microbiome is here: go.nature.com/2zrSqap
Gut microbiota is highly variable (e.g., two healthy individuals could harbor totally different gut microbiota compositions) and susceptible to technical and analytical procedure selections. As a result, dysbiosis patterns revealed by different studies could vary dramatically. Taking Firmicutes/Bacteroidetes ratio (F/B) as an example, F/B is quite popular in microbiome studies as it was first discovered to increase in obesity and the explanation is that Firmicutes are better at converting non-absorbable materials in the colon for our body to re-use and hence increase our body’s energy absorption. Nevertheless, as more and more obesity-microbiome studies being reported, it becomes unclear whether this ratio is increased, unchanged or decreased in obesity because all possibilities were reported, and there are scientifically sound explanations to support each of these possibilities. Similar examples could be easily found for other diseases and their microbiome markers, which confuses us regarding the relationship between human gut microbiota alterations and our health and diseases.
Some hypotheses could be used to explain why dysbiosis patterns vary between different studies, e.g., due to technical variations, strain-level differences, function redundancies and etc. Our hypothesis is that human gut microbiota exhibits regional variation, and this variation could confound our analysis of disease signatures, resulting in different dysbiosis patterns and disease models. The Guangdong Gut Microbiome Project (GGMP) is one ideal dataset to put this hypothesis into test. The project selected 14 districts in Guangdong, then randomly selected three neighborhoods per district, then two communities per neighborhood, and finally 45 households per community. A total number of 7,009 individuals were involved in our final analysis.
A large regional variation in human gut microbiota was identified in GGMP, as the proportion of human gut microbiota variations explained by districts was almost five-fold higher than age, a commonly recognized gut microbiota co-variate. Surprisingly, using gut microbiota could accurately identify whether an individual resides in a specific district, further supporting the existence of regional variations in human gut microbiota. Meanwhile, the variations of metabolic disease microbiota markers were larger between host locations than between disease and healthy individuals and disease models built in one location failed when used elsewhere. These results supported our hypothesis.
So how could we identify dysbiosis patterns if they are confounded by regional variations? The answer could be to integrate data from multiple locations and analyze on population-level. We showed in our Nature Medicine paper (https://doi.org/10.1038/s41591-018-0164-x) that consistent biomarkers could be identified for some diseases and we fully illustrated the dysbiosis patterns for metabolic syndrome in our Microbiome paper (doi to be available). The dysbiosis patterns for MetS on population-level is phylogenetically conserved, manifested as OTUs from Bacteroidetes and Ruminococcaceae were negatively associated with MetS while OTUs from Proteobacteria and Firmicutes (other than Ruminococcaceae) were positively associated with MetS. We formulated a MetS index to represent the overall dysbiosis extent and this index is always higher in MetS subjects regardless of their place of residence or economic status, indicating its general applicability. Based on the MetS index, we further identified that gut microbiota dysbiosis, together with sedentary lifestyle, is correlated with MetS prevalence. This study, for the first time, connected gut microbiota dysbiosis, lifestyles and MetS epidemiology, even though more studies are needed to understand their full interactions.
Our study clearly showed the regional variations in human gut microbiota and indicated that microbiome-based diagnosis and prognosis of metabolic diseases should be tested in a geographical context rather than based on small-scaled case/control studies. We also showed that using population-level analysis could help us understand how gut microbiota alters under diseases more accurately.