The paper in Nature is here: http://go.nature.com/2FIwXvy
Where does our gut microbiome come from? In recent years this has become a much sought after question. While we know that our microbiomes are similar to those of our parents, we don’t know if this is because we share the same genes or because we lived in the same house when growing up. This question may even have practical implications. For example, if our genes exert a strong influence on our microbiome, then microbiome transplants may require finding a genetically-compatible donor, which complicates the deployment of these techniques. Hence, the classic question of nature-versus-nurture has deep clinical implications with respect to our microbiome.
We were therefore very excited by the opportunity to investigate this nature-versus-nurture question hands-on, based on a cohort of ~1000 Israeli individuals with genetic and microbiome profiles that we collected as part of our large-scale personalized nutrition project. Our cohort is especially interesting due to Israel’s unique history: Most of Israel’s population originates from recent immigration from around the world, yet the current population shares very similar diet and lifestyle habits. This setting approximates the ideal lab experiment, which would consist of placing genetically diverse individuals under the same environmental conditions and observing how this influences their microbiome.
With great excitement we began analyzing the data, expecting an initial barrage of promising discoveries, and preparing ourselves to be good scientists and remain skeptical at these initial results until we could validate them.
Certain that we would find strong links between genetics and microbiome, we began the analysis, but to our great surprise (and strong disappointment!) we did not find any significant associations. We were sure that we got it all wrong. At this stage we started questioning the quality of our data and thought perhaps there are mix-ups of samples. After a lengthy analysis process that involved matching microbiome samples to their matching subject by matching human DNA traces in stool samples, we were convinced that there were no such mix-ups. Unbeknownst to us, these investigations would eventually form an important part of our paper. The moment of realization that this is an important finding was when we realized that in our data, we can we successfully predict the ancestral origins of our participants with their DNA but not with their microbiome. In other words, while closely ancestral individuals have more similar DNA, closely ancestral individuals don’t have more similar microbiome.
We were able to establish that ancestry is not associated with the microbiome and we next wanted assess the overall microbiome heritability. For this we contacted Julia Goodrich (who completed her PhD at the lab of Prof. Ruth Ley) and asked her approval to re-analyze 2,252 twins from the UK. Our re-analysis established that although some gut bacteria our clearly heritable, the overall microbiome heritability is likely close to 2%. Hence, the overall human genetics-microbiome associations are indeed small. So this then became the story, as this is the story that the data insisted on telling us…
But this lack of signal is in fact a blessing, because our results provide an optimistic outlook for the microbiome field in general. For example, in the case of microbiome transplant, it indicates that genetic compatibility between a donor and a host is unlikely to be an important factor affecting efficacy.
Although our initial study did not turn out as expected, along the way we made exciting discoveries. As part of our data consistency investigation we devised the microbiome analogue of genetic heritability, i.e., a statistical technique to quantify the association of the microbiome with traits such as obesity and blood glucose levels. To our amazement, we found that the association of many traits with the gut microbiome is often comparable, and sometimes stronger, than the association with host genetics. This result is also very promising for microbiome-based therapeutic purposes. We also found that we could use the gut microbiome composition to substantially improve prediction of traits such as blood glucose levels, which may turn out to be useful for early screening of individuals at pre-diabetes risk. These results ended up being a core part of our paper. A more general take home message that we went with was that as always, be open-minded and follow your data!
Written with Omer Weissbrod, Dafna Rothschild, and Elad Barkan