For some time now I’ve been bewitched by the microbiome–those 100 trillion passengers that make our bodies their vessel (here’s a piece from the New York Times last year, and a long essay from last month). But I was especially intrigued by a paper that came out today in Nature. Scientists found they could sort people into just three distinct gut microbiomes, much like they can sort people into four blood types. Here’s my story in the Times, which will appear in tomorrow’s edition.
One thing that you won’t find in the article is some intriguing speculation I indulged in with the scientists I interviewed. Researchers have clearly demonstrated that microbes can influence their host’s behavior. They release molecules in the gut that travel into the blood and then into the brain. The bacteria that live in obese mice can make ordinary mice voracious. My fellow Discover blogger Ed Yong has written about how the microbiome can steer the development of mice to become more or less anxious as adults. In an upcoming review called “The Mind-Body-Microbial Continuum,” a team of microbiome experts ponder how our microbes might play a role in psychological disorders such as autism and attention deficit disorder. So it’s reasonable to wonder if our “bug type” influences our personality, much as the brain parasite Toxoplasma appears to.
I’ll be talking more about my story Thurdsay morning at 6:30 am and 8:30 am EST on the public radio show The Takeaway.
Update #1: Here’s the link to my appearance on the The Takeaway.
Update #2: Here’s a figure from the paper, representing the separation of the three types. It’s helpful when reading the skeptical comments from Joshua Shapiro of Princeton below.
Here’s the accompanying caption: “Phylogenetic differences between enterotypes. a–c, Between-class analysis, which visualizes results from PCA and clustering, of the genus compositions of 33 Sanger metagenomes estimated by mapping the metagenome reads to 1,511 reference genome sequences using an 85% similarity threshold (a), Danish subset containing 85 metagenomes from a published Illumina data set (b) and 154 pyrosequencing-based 16S sequences (c) reveal three robust clusters that we call enterotypes. IBD, inflammatory bowel disease. Two principal components are plotted using the ade4 package in R with each sample represented by a filled circle. The centre of gravity for each cluster is marked by a rectangle and the coloured ellipse covers 67% of the samples belonging to the cluster. IBD, inflammatory bowel disease. d, Abundances of the main contributors of each enterotype from the Sanger metagenomes. See Fig. 1 for definition of box plot. e, Co-occurrence networks of the three enterotypes from the Sanger metagenomes. Unclassified genera under a higher rank are marked by asterisks in b and e.”
April 21st, 2011 at 1:02 am
this reminded me of another paper i read about that is also mentioned in nature, about specific enzymes found only in japanese people (associated with the bacteriodes bacterium, which happens to be the basis for one of the enterotypes):
http://www.nature.com/nature/journal/v464/n7290/edsumm/e100408-14.html
April 21st, 2011 at 6:40 am
[...] Excellent summaries of this research can be found on Ed Yong’s Not Exactly Rocket Science blog (“Divided by language, united by gut bacteria—people have three common gut types”) and in an article by Carl Zimmer in The New York Times (“Bacteria Divide People Into 3 Types, Scientists Say”). Carl Zimmer has further comments on his own blog, The Loom (“Blood type, meet bug type: My new story for the New York Times”). [...]
April 21st, 2011 at 10:46 am
I find the quoted numbers for microbes versus number of human cells intriguing . I struggle with the numbers that I am by a factor of 10X or perhaps higher by count composed more of microbes than human cells. what is the ratio by weight? opens new questions about what it means to be human or presumably any species. take away the microbes and what is left?
April 21st, 2011 at 5:00 pm
Fascinating! Also viruses, which outnumber bacteria by 100 to 1 in the body, play important roles in the human microbiota. See, for example ‘Going Viral: Exploring the Role Of Viruses in Our Bodies’ by
Elizabeth Pennisi, Science 25 March 2011. DOI: 10.1126/science.331.6024.1513
April 22nd, 2011 at 8:29 am
Thank you Carl for your nice articles on the human microbiome. They are always very interesting and perfectly translate to the general public what our research focuses on. I would like to reproduce below the comment we posted this morning (Paris time) under the New York Times article (currently under moderation).
The enterotypes paper is the result of a wide European collaboration. The EU-funded Project MetaHIT (metagenomics of the human intestinal tract) is behind the enterotypes discovery. Many fields of expertise, teams, labs, researchers are gathered in order for this collaboration to publish such exciting results. We believe that it is more than normal that the whole consortium be thanked and its name, MetaHIT, put forward.
Congratulations to all the contributors of this wonderful adventure that is our collaboration! Stay tuned for more, it s only a beginning!
Visit our website for more information http://www.metahit.eu & Follow us on twitter, @MetaHIT
Do not hesitate to get in touch with us, we’ll be more than happy to give further details on our activities.
April 22nd, 2011 at 6:44 pm
I wonder how much microbiome change as we age leads to our getting fatter as we get older. I used to be able to eat like a horse and not gain weight, even without much exercise. Now, even with vigorous hikes around twice a week, I find gaining weight effortless. Might it be that the mix of the microbiome changes as we get older, or maybe some new bug finally manages to colonize our guts, and that might be part of the obesity epidemic we are seeing these days?
April 23rd, 2011 at 3:24 pm
I’m pretty skeptical of the claim that there are exactly three types. Identifying the number of clusters in a dataset is a very hard problem, and accuracy depends a lot on things like the number of samples, the number of dimensions, and the relationships among the measured variables. Adding more samples could easily demonstrate that there are more clusters than they claim, or fewer. Indeed, if you look at the supplementary data, in the largest cohort there appears to be nearly equal support for 2-4 samples. I think that this paper does demonstrate that there are correlations among different species of gut bacteria, but that is not the same thing as saying that there are N distinct types, and in fact it makes many of the conclusions that they draw from their simulations suspect. They simulate with uncorrelated normal distributions for the bacterial species populations, when it is clear that correlations exist, and many of the distributions are decidedly not normal.
I am also a bit annoyed by depiction in the paper of the clusters, with lines drawn from each point to the center of the cluster. Aside from making for a very busy plot that obscures the actual data points (as does writing labels over the data), it really trains the eye to accept the conclusion of clustering uncritically; points that might fall between the inferred clusters are made to look like they are clearly members of one cluster or the other, when in fact there would be quite a bit of uncertainty in their assignment. They also might represent underrepresented areas of the distribution, which would lead to changes in the conclusions about the number of types. Amplifying that concern is the fact that the authors state that many of the individuals were chosen “for diversity.” That decision might have resulted in the artificial appearance of clusters if individuals at the extremes are overrepresented.
I don’t want to sound too negative, as I think there are a lot of really good and interesting things in the paper. But I think that comparing “enterotypes” to blood types is inaccurate and misleading. If the types do exist, they are much fuzzier than blood types, and I don’t think we know yet how many there really are.
April 25th, 2011 at 2:20 pm
Although I have not studied the paper in detail, my superficial reading inclines me to agree with Joshua’s comments above. Delving into the supplementary info, the basis for the “3 enterotypes” claim in the MetaHIT Nature paper is the Calinski-Harabasz index applied to a K-medians clustering approach. This is widely-used and reasonable, but not exactly bulletproof. The determination of the best value of “K” in a “K clusters” approach is the topic of a lot of statistical methodology, and it can depend quite a lot on one’s initial assumptions and underlying model. Given the interest in this topic, I would hope to see some statistics that could e.g. provide confidence levels or error bars on K (rather than the assertion that “K=3″); I would also hope for discussion of the underlying model and its assumptions. Instead we have “K=3″ and some visually persuasive plots, which do (deliberately or not) tend to disguise the fact that these clusters are spread out, they rub shoulders, some could be split, and they are plotted in an abstract two-dimensional principal component projection (i.e. PCA) of a much higher-dimensional space of species counts.
IMO, there is nothing here (yet) that could be said to be as unambiguous as the serologic tests for blood types. It does seem to be true that there is some broad clustering, but it’s not clear to me that all of those clusters are discrete or well-defined enough to warrant the label “enterotypes”.
Having said this, I don’t mean to diminish the paper at all. It is absolutely fascinating and fundamental work, rich with data, and I will certainly be re-reading it with my students many times over.
On a somewhat related note, on the statistical methodology side of things: PCA is in my opinion a pretty blunt instrument. This high-dimensional space of counts-of-each-known-bacterial-species is not just a vector of arbitrary counts; it has structure (the phylogenetic tree of those species). I hope that metagenomics researchers will start using variations on PCA that reflect this structure, such as Erick Matsen’s “edge PCA”:
http://matsen.fhcrc.org/pplacer/demo/pca.html
April 27th, 2011 at 12:19 pm
This is interesting…”microbes can influence their host’s behavior”…If that’s true, first of all being aware of this possibility sets great importance for each of us…This could be an answer to “why me” questions…Different races or nationalities having different group of microbes due to living, eating and geographical conditions…this is interesting too…in order to explain certain differences between cultures and behaviours…
However, research in this area is still a baby…We will see hopefully in following years, what role microbes or viruses have on human behavior both physically and mentally…