Deciphering metabolism, one microbe at a time
"Small molecules produced by gut microbes can influence human physiology. An atlas of metabolic outputs of diverse gut microbes offers new ways to decipher the microbial mechanisms behind their production."
The microorganisms in our gut can have far-reaching effects — on our liver, arteries and potentially even on our behaviour. One way these microbes exert their effects is through the generation or consumption of small molecules, termed metabolites. Measuring metabolite levels, an approach called metabolomics, has led to ever-increasing recognition of their importance. And yet only rarely do we understand the underlying mechanisms driving these levels: namely, which microbes, enzymes and interactions are involved in the production and uptake of a specific metabolite. This task is further hindered by the complexity of microbial communities such as the gut microbiome, studies of which have to take into account the large number of microbes, the interactions between them, their diverse metabolic capabilities and several hard-to-measure non-microbial factors, such as host diet. Writing in Nature, Han et al. present a comprehensive approach to addressing this major challenge, by carrying out metabolic and genetic analyses of a broad set of microbes commonly found in the human gut.
The authors’ approach (Fig. 1) has been facilitated by notable technical advances. Using liquid chromatography–mass spectrometry (LC–MS), a technique that quantifies metabolites on the basis of their polarity, mass and charge, Han and colleagues compiled a reference database of 833 metabolites that are relevant to microbial metabolism. They confirmed that these metabolites are detectable in biological samples, and that their measurement is consistent in several types of sample, such as faeces or blood, and quantifiable over a wide range of concentrations. The authors also developed an analysis pipeline that enables compound identification and statistical analysis. With this infrastructure in place, Han et al. measured metabolite levels in thousands of samples from in vitro cultures of 178 microbial strains grown separately in multiple media types, and from various tissues taken from mice whose intestines were colonized by the same strains, either alone or in communities of five or six species.
Having compiled an atlas of single-microbe in vitro metabolic outputs, Han and colleagues set out to address a long-standing question: to what degree is the evolutionary relationship between two microbes (their phylogeny) related to their metabolic capacity? The authors show that, although the two generally correspond, this correspondence is not perfect. For example, Han et al. report that Clostridium sporogenes and Clostridium cadaveris, two closely related species, have strikingly different metabolic profiles. By contrast, Atopobium parvulum and Catenibacterium mitsuokai, two phylogenetically distant species, have similar metabolic profiles.
Furthermore, although the authors could identify some strong species-specific associations with the production of particular metabolites, such as production of the molecule tyramine by Enterococcus faecalis, metabolomic profiles were insufficient to independently distinguish between members of different species. A machine-learning algorithm trained to identify a species on the basis of its metabolomic profile was correct only about 30% of the time, and even members of different genera or families were not well separated by the algorithm (in such analyses, it achieved an accuracy of approximately 70%). These results raise a note of caution regarding typical microbiome analyses, which often rely on microbial-abundance estimates at the genus and species levels, and thus might miss crucial metabolic aspects of the microbial community.
Although the correspondence between phylogeny and metabolism is imperfect, the authors present an analytical approach using the association between specific genes and metabolic outputs to obtain insights into microbial metabolism. Han and colleagues paired their metabolomic analyses with analyses of bacterial genomes to uncover the genes responsible for unexplained metabolic capacities. The authors identify a previously unknown mechanism by which microbes of the phylum Bacteroidetes utilize the amino acids glutamine and asparagine. Nevertheless, the spe genes responsible for producing the molecules putrescine and agmatine in several species are not present in three species of Fusobacterium that the authors found to produce these molecules — a result that demonstrates the limitations of this analytical method.
Han and colleagues conclude by turning to the most challenging aspect of their approach: assessing the correspondence between in vitro and in vivo metabolic output. Strains with a prominent metabolic capacity, such as Citrobacter portucalensis, which produces agmatine from the amino acid arginine, maintained some of this capacity both in culture and in mice. In some cases, this led to effects reaching beyond the gut (systemic effects). For example, agmatine levels were increased in the urine of mice if the animals were colonized by C. portucalensis.
However, such a high level of correspondence between in vitro and in vivo data was not observed for the overall metabolic output, as tested by the authors for two strains. For these strains, there was only a moderate correlation between the in vitro metabolic profile of the strain and the profile measured from the intestines or faeces of a mouse colonized by it. Furthermore, no correlation was found between the in vitro profiles of these strains and the blood or urine profiles of mice colonized by them. This was the case despite the simplified ‘mono-colonization’ scenario, in which each mouse harboured only a single microbial strain, without other members of the bacterial community and the complex effects that arise from the interactions between them6.
These results highlight a major challenge left in the wake of this impressive endeavour, which is to use this extensive atlas of metabolic measurements, taken in simplified settings, to provide accurate models of complex community metabolism. This could be done experimentally — for example, by extending the work performed by Han et al. to assess combinatorial co-cultures — or by harnessing various computational and mathematical methods7,8. Future work could further validate the utility of this data set for studying the human gut microbiome; extend the data set to strains that are found in, and have probably adapted to, a specific host9; and expand the data set to microbes and metabolites that are relevant to other human-associated microbial communities, such as the vaginal and skin microbiomes.
Han and colleagues provide useful resources for the research community, including an extensive metabolomics data set consisting of thousands of samples, web resources with which to explore it and analytical approaches for studying microbial metabolism. Moreover, this work provides a truly open-source technical resource, with protocols, analysis pipelines and an extensive metabolite reference library, which the authors demonstrate to be applicable, with minimal calibration, to different machines. This resource could be used by others as they pursue similar experimental set-ups, thereby promoting the democratization of metabolomics. Altogether, this work lays a foundation for future work seeking to decipher microbial metabolism — an important step towards new therapeutics that target the microbiome.
Read more at:
Figure 1 | An approach to studying microbial metabolism.a, Han et al.5 generated a reference library of 833 small molecules (metabolites) relevant to gut-microbial metabolism, and used it to assess the metabolic output generated by 178 bacterial strains commonly found in the human gut. The authors thereby generated an atlas of metabolic outputs, as shown in this hypothetical example. Darker blue on the heat map indicates higher levels of production. b, The authors present a range of approaches for studying microbial metabolism using these data. They investigated the correspondence between the evolutionary relationships of different microbes (their phylogeny) and their metabolic output. Phylogeny and metabolism generally correspond; however, the authors found some exceptions and divergences. c, Han and colleagues further show that a parallel comparison between microbial genomes and their metabolic outputs could suggest genes responsible for unexplained metabolic capacities. d, The authors also investigated the correspondence between in vitro and in vivo microbial metabolism, identifying many metabolites produced in both contexts.