Welcome to the third installment of my report from the 2012 Keystone conference on the microbiome and innate immunity. Tuesday’s morning and evening sessions were titled, respectively, “Complexity in microbiota-host interactions” and “Innate immune protection against microbiota” (the evening session was a joint session between the two conferences).
– The Tuesday morning session began with a talk entitled “Systems approaches to investigate host-microbiota metabolic interactions” given by Jeremy Nicholson. Nicholson started by noting that research on the human microbiome uses either a bottom up approach; which focuses on organisms and cells, involves a closed system, is very controlled, and uses very exact models; or a top down approach; which uses supraorganisms, involves an open system, is less controlled, and uses probabilistic modeling.
He then went on to note that most human diseases share genetic mechanisms, but most human diseases are not genetic in the Mendelian sense, which puts limitations on genome wide association studies (GWAS). For example the “fat gene” only accounts for 1.47% of the variance in BMI between people. Because there is so much mounting evidence that the microbiota can play significant roles in human diseases, his goal is to figure out how human and microbiome gen network interact.
Both of these networks are influenced by environmental factors (which he called the “exposome”, which I find abhorrent). Together these networks are able to perform combinatorial metabolism. For example, cholic acid is produced by the liver and co-metabolized by the microbiota. Bile acids are also metabolized by the microbiota. Gastric bypass surgery has a very large impact on the microbiota. It is associated with an expansion of gamma-Proteobacteria and massive changes in bile acid metabolism. It turns out the fecal water of rats that have gastric bypasses is highly cytotoxic, which may be a result of the altered microbiota metabolism.
Because the microbiota can affect metabolism, there’s a good change that differences in the microbiota can change the way people metabolize, and thus react, to drugs. For example, Nicholson looked at Tylenol metabolism by sampling subjects’ urine. It turns out, people are either strong or weak sulfaters. Microbiotas that produce high levels of 4-cresol lead to system sulfate loss during Tylenol metabolism.
One exciting aspect of the microbiota for pharmacologists is that the microbiota probably has many many more drugable targets than the human genome. For example, one group was able to alleviate a cancer drug’s toxicity by inhibiting a bacterial enzyme. To look for things like this, Nicholson is doing Metabolome wide association studies (MWAS). Doing these in conjunction with geographical metabolic mapping revealed that people’s metabolism was highly influenced by were they lived, and part of this geographical distinction was attributable the microbiome. He also found a microbial metabolic signature for obesity, as well as significant differences in this signature between males and females. Then he used the word “nutriome” and I passed out from disgust for a minute (is nutrition just not cutting it anymore?!). When I came to, he was explaining that microbial metabolites can affect human blood pressure, and their production relies on a balance of a microbial metabolic network that is affected by diet and other factors. Despite his prediliction for -omes, it was a fairly interesting talk. I think the idea of targeting bacterial products to affect drug metabolism and or human disease is intriguing.
– Next up was Jens Walter with a talk titled “Evolutionary mechanisms that shape host-microbiota interactions and their consequences”. The focus of this talk was on the possible evolutionary mechanisms that could old be at work to result in our mutualism with out microbiota (as opposed to all our associated microbes being parasitic). The three mechanisms he lists are (1) host investment in symbionts control, (2) vertical evolutionary transmission, (3) stringent constraints, no options outside of the host. The first point he made was that species that co-speciate tend to have mutualistic relationships and strong interdependencies whereas microbes that are not host specific at all are only conditionally beneficial to the host. The problems with this argument is (a) co-speciation is NOT the same as coevolution. They are not mutually exclusive at all, but co-speciation can occur without coevolution and vice versa and (b) strong interdependencies between species are not necessarily born of mutualistic interactions. I remember a talk at the beneficial microbes conference in Miami in October 2010 were one of the speakers presented evidence that an obligate relationship between a fly and in intracellular bacterium could be just as well if not better explained by an intial antagonistic relationship than an intial mutualism.
Anyway, there is a bit of evidence for a core humane microbiota and Walter did some quantitative genetics to see if there were any host genes associated with particular microbial taxa. To do this, he considered each microbial taxon as an individual trait and found 13 regions of the host genome that were significantly associated with particular microbial taxa.
He then talked about how different strains of Lactobacillus reuteri have associations with different hosts. All strains, though, are able to colonize mice, however, when co-colonized, the rodent strains out compete the others. Additionally, the L. reuteri phylogeny looks a lot like the phylogeny of its vertebrate hosts. Again, this is evidence for co-speciation, or possibly co-strainification, but there’s no evidence that the hosts are experiencing any selection from the different L. reuteri strains.
– I must start my description of this next talk with an apology. My notes for this one aren’t so great, and I can’t remember any of the missing information, so I’ll just be giving the gist of the whole talk.
Thaddeus Stappenbeck’s talk was titled “Host-microbe interactions that play a role in the pathogenesis of inflammatory bowel disease”. Stappenbeck used a dnKO mouse model which is a KO for IL-10R2 and a dominant negative for T regulatory cells. These mice all die from colitis, but it can be prevented with antibiotics. They screened for colitogenic microbes by treating the dnKO mice with antibiotics, halting them and then gavaging with different bacteria to see which ones caused colitis again. They found that cultured anaerobes induced colitis, specifically many species of Bacteroides, the most effective of which was B. thetaiotaomicron (B. theta). Enterobacteriaceae were enriched in dnKO mice, but they were unable to illicit colitis after treatment with antibiotics. It appears access to the host mucosa and the expression of sulfatases by B. theta is required for the induction of colitis, but it’s not apparent why.
Stappenbeck also talked about how a mutation in a gene associated with autophagy causes goblet cells to be unable to secret mucus. This gene is associated with IBD in GWAS, and it might be because without proper mucus secretion, the bacteria have easy access to the epithelium (much more on this from later talks).
– The last full talk of the morning was given by Andrew Gewirtz. His talk, entitled “TLR5: protecting the gut from its BFF” was a vey nice complement to Ruth Ley’s talk I described in Monday’s post. In the case of this talk’s title, BFF stands for “Best Frenemy Forever”. This because, while it’s largely good, the microbiota cannot be completely trusted. To illustrate this, Gewirtz focused on both IBD and metabolic syndrome. Both of these diseases involve the remodeling of adipose tissue and both have seen increased incidence in the past few decade.
The Gewirtz group intially predicted that knocking out TLR5 function would reduce the tendency for commensal bacteria to induce inflammation, thus protecting mice from colitis. However, about 10% of the KO mice developed spontaneous colitis. It turns out, in these mice, bacteria were infiltrating the mucus to the epithelium and activating TLR4. While the majority of the T5KO mice had no evidence of colitis, they all had an elevated innate immune response, and the non colitis ones were much more susceptible to developing it than WT mice. Anti IL10 antibodies caused colitis in the T5KO mice, but not the WT. Furthermore, non colitic T5KO mice had increased weight gain (~20%), and transplanting T5KO embryos to WT females brought colitis down to just marginally above controls, but weight gain still occurred.
It turned out the mice were fat because they were actually eating more than WT mice. Calorie restriction reduced metabolic syndrome, but there was still insulin resistance, probably due to inflammation. Antibiotics treated the metabolic syndrome, and germ free T5KO mice didn’t have the mutant phenotype. All the T5KO mice had altered microbiota, with an expansion of Proteobacteria. Like the Ley lab found, the microbiota of the KO mice was much more variable through time than the WT mice. Additionally, the WT microbiota are more diverse, suggesting that it true that for the microbiota that a more diverse community is a more stable community.
– The afternoon’s joint session started with a talk from Laura Hooper entitled “Immune defense of the intestinal epithelial surface”. The question that drives Hooper’s research is How do we exist with 100 trillion bacteria and not get sick? Antimicrobial proteins play a role in this phenomenon.
RegIII-gamma is a type C lectin that preferentially kills Gram positive bacteria. It targets Gram positives by binding to the carbohydrate portion of peptidoglycan which is exposed on Gram positives but under the LPS in Gram negatives. RegIII-gamma disrupts membranes, which they determined using dye leakage assays on liposomes. It does this by forming an oligomeric pore in membranes. She then described the work her lab did to characterize the 3D structure of the protein, which I’m not going to go into here.
Hooper then went on to describe the spatial separation of host and bacteria in the gut. The gut is coated with a 150 micron layer of mucus. She had a very lovely picture showing that there is a very consistent gap of 50 microns between the epithelium and the bacterial community (why this is will be covered by the next talk, and I should mention there were some exceptions, as there always are). This 50 um zone she referred to as the “demilitarized zone”. It turns out MyD88 is required to maintain this DMZ. Deep sequencing reveals that the bacterial community in the lumen is no different in composition or numbers in WT and MyD88 -/- mice. The mucin barrier is not compromised in MyD88 mutants, in fact it trends toward thicker. RegIII-gamma is MyD88 dependent and both MyD88 -/- and RegIII-gamma -/- mice have increased bacteria invasion to the epithelium. IgA and Th1 production is increased in these mutants to compensate, but it doesn’t appear to be effective enough. The biggest puzzle in these findings is why, if RegIII-gamma is Gram positive specific, why don’t Gram negatives regularly penetrate to the epithelium?
– The last full talk, “Good barriers encourage appropriate microbiota”, was given by Gunnar Hansson. He started by giving a good physical description of the mucosal layers in mammals. Starting from the posterior, the colon has two layers, a bottom layer that is very dense and firmly attached to the epithelium, with a looser, less dense outer layer. The inner layer is about 50 microns thick (remember the size of the DMZ?) and the outer is about 100. The small intestine only has a single layer that is much like the outer layer of the colon. The stomach again has two layers, the inner denser and firmly attached, while the outer is less dense and loosely attached. He demonstrated these properties using tissue explants.
He then went into great detail about the biochemistry of mucin formation, which I will not even try to describe here. Suffice it to say that muc2 is the primary protein component and what gives it most of its special properties. The important aspect here is that when secreted by colonic or gastric epithelial cells, the muc2 proteins are arranged in flat, rigid sheets of rings that pack densely together. A new layer starts on the bottom of the inner mucosal layer, and ove 1-2 hours it travels to the interface with the outer layer were proteins break it up, making it less dense and rigid, forming the looser outer layer. The inner layer is impenetrable by bacterial sized beads, but the outer is not. As somewhat of a side note, he mentioned that wild forest mice have much thicker mucus layers than lab reared mice, an important caveat for users of these model organisms.
GF and colitic mice have much thinner mucus layers which are much more penetrable by bacteria. Additionally, oligosaccharides are likely attachment sites for bacteria in the outer layer. Human glycosylation is uniform in humans, but diverse in mice which sheds some doubt on the ability to fully “humanize” mice with human microbiota transplants as talked about by Jeff Gordon. Bacteria degrade the O-glycans on the mucin, and the rate at which they do so, and the length of these glycans may be very important. O-glycan deficient mice experience invasion of bacteria to the epithelium, and longer ones can keep them from invading.
So overall it seems the best barrier is a thick, glycosylated mucus chock full of antimicrobial peptides and IgA, but as I’ll talk about in the next post, complete containment of the bacteria away from the epithelium is not good either…