‘Reverse Metabolomics’ Enables Biological Discoveries From Repository Data

December 21, 2023

By Deborah Borfitz

December 21, 2023 | In a turnaround of the usual metabolomics study approach, University of California San Diego scientists are homing in on the microbial-produced molecules that interact with cells and affect human health. A couple million such molecules are likely awaiting discovery in this fashion, according to Pieter C. Dorrestein, PhD, professor at Skaggs School of Pharmacy and Pharmaceutical Sciences at UC San Diego. 

Hundreds of never-before-seen molecules were found in the first application of the “reverse metabolomics” strategy and the data were used to identify a new metabolic signature for inflammatory bowel disease (IBD), as reported recently in Nature (DOI: 10.1038/s41586-023-06906-8). The molecules could potentially serve as a biomarker for diagnosing IBD or as a therapeutic target to help treat the disease.  

The reverse metabolomics technique combines organic synthesis, data science, and mass spectrometry, and can be applied to different sample types—tissue, plasma, urine, or fecal—to reveal the bodily distribution of molecules being secreted by the microbiome, Dorrestein says. In the latest study, data from fecal samples were used to plot peak area abundances. 

It marks the first time repository-scale metabolomics information has been leveraged to make biological discoveries, says Dorrestein, although that has been the norm in the sequencing community since the 1990s. Researchers first synthesized candidate molecules they thought the microbes could potentially be making and then, after creating barcoded identifiers (“fragmentation spectra”) for each of them using mass spectrometry, went searching for those barcodes in public repositories. 

The problem up to now is that most of the barcodes are unannotated, meaning a description of the contents of any given sample is largely absent. Only 14% of the metabolomics data is currently annotated, and therefore reusable, despite extensive efforts to make the data public over the past eight years, he points out. 

Reverse metabolomics enables extensive knowledge building about the roles of specific metabolites, their association with different diseases and interventions, and in which organ system or biofluids they might be distributed, continues Dorrestein. Those sorts of questions can now be addressed without having to collect the data one study at a time.  

Why this hasn’t been done since the 1960s, when mass spectrometry first started being used for quantitative measurements of metabolites, is anyone’s guess, he says. “We need a similar moonshot as we had for sequencing of the human genome, but for metabolites... we [in the metabolomics community] are just lagging behind.” 

‘Hypothesis of Associations’

The barcode signatures of metabolites from roughly 2,800 studies are now in the public domain, and reverse metabolomics provides a way to organize the data so that scientists can start to understand their functional role, says Dorrestein. To that end, his lab synthesized about 2,400 of these molecules and generated a corresponding barcode and began searching for a match. They started with the molecules that were relatively easy to produce, a population that likely numbers in the “hundreds of thousands.”   

This exercise makes it possible, for the first time, to employ data science on information that has been amassing in public repositories since 2016. Researchers can summarize the specific barcodes they observe in relationship to the metadata (e.g., disease state and treatment) affiliated with the people who have contributed samples, he explains, and from that build a “hypothesis of associations.” 

In this way, Dorrestein and his colleagues identified many connections to both diabetes and IBD in their latest study. When comparing the metabolomic signatures of samples from different patient populations, they found a particularly strong association between a synthesized class of microbial molecules, called bile amidates, and IBD. Certain bile amidates were elevated exclusively in patients with Crohn’s disease when they were experiencing symptoms.  

Most of the studies available in the public domain were deposited there by the precision nutrition initiative of the Crohn’s & Colitis Foundation, he says, enabling investigators to look at the relative abundance of these molecules in the data and improve their confidence in the suspected IBD metabolic signature. It also became clear that in IBD patients treated with antibiotics, levels of the molecules tended to disappear. This supported a hypothesis that the microbial metabolites were being produced by the gut microbiota. 

The research team cultured 200 of these organisms to show they were indeed microbiome-derived, Dorrestein adds, which was confirmed by another lab (Erin Baker, University of North Carolina) using ion mobility mass spectrometry. They also had the strong connection seen between the metabolites and Crohn’s disease replicated in additional patient cohorts from the Broad Institute and Penn State.  

One class of molecules found included 139 bile acids previously undescribed in the literature, says Dorrestein. However, earlier papers had shown that some related bile acids regulate immune cells. 

Investigators consequently decided to conduct in vitro experiments where the top dozen metabolites differentiated in IBD, particularly Crohn’s disease, were tested against T cell regulation. As it turned out, several of the molecules were indeed able to regulate T cells, providing a foundation for future studies in humans, he reports. One microbial compound produced a six-fold increase of a key cytokine known to be involved in the pathogenesis of Crohn’s disease. 

In addition to seeing certain metabolites uniquely increased in Crohn’s disease, as reported in the paper, different but related metabolites may be increased in ulcerative colitis, says Dorrestein. That suggests a diagnostic approach could be developed using these classes of molecules to better distinguish one from the other, which is a known clinical need. 

Straightforward Science

Untargeted metabolomics, whereby scientists collect a lot of data and then try to interpret it, is one of the big challenges in the field, says Dorrestein. Interpretation requires standards and most of them—particularly for microbial metabolites—aren’t commercially available. A typical metabolomics study can therefore only characterize about 10% of the molecular data from a human microbiome sample. 

Dorrestein’s lab used combinatorial chemistry to create mixtures of different classes of molecules in one pot and then used metabolomics to separate them. In this way they get the barcode signatures of individual metabolites, providing reference data with which to begin their repository search, he explains.  

Reverse metabolomics is straightforward science, adds Dorrestein. Researchers elsewhere have adapted these methods already, for example, to look at traumatic injury in infants. “We just put this approach on the map, and now hopefully others will follow suit... [and] once that starts to pick up, we will uncover so much new biology that we’ll have to rewrite our biochemistry textbooks.” 

In the clinic, some of these metabolites could potentially be used to diagnose IBD and differentiate ulcerative colitis from Crohn’s disease, as well as serve as medications for treating symptoms of those diseases, he says. The medications might alternately come in the form of molecules or microbes.  

Since metabolites are often regulated by dietary patterns, it may be that patients in the future will be prescribed specific microbial-associated dietary habits (e.g., removing pepper from the diet), continues Dorrestein. Or perhaps the enzymes responsible for producing these molecules could be targeted. 

Once it is understood how these molecules function, therapies may need to be developed to down-regulate specific T cells using the metabolite that upregulates them, he adds. For individuals lacking the enzymes in their microbiome that produce that molecule, personalized treatment strategies would need to shift accordingly. 

Incomplete Puzzle

Among the follow-up studies underway in Dorrestein’s lab is one investigating the possibility of controlling production of metabolites and changing the inflammatory response to pathogens. He and his team will also be evaluating the impact of dietary patterns on the production of different molecules. In the future, the function of molecules could conceivably be controlled through diet and better healthy lifestyle choices. 

Dorrestein is the scientific co-founder of three startups—Ometa Labs, enabling data science with mass spectrometry by private companies that in-license technology developed at UC San Diego; Arome, which provides metabolomic services and will likely offer some of the standardized compounds; and Enveda, a drug discovery company focused largely on identifying the active ingredient in medicinal plants for different disease indications. 

The overarching goal here is to mechanistically understand the molecules that constitute much of the human body, says Dorrestein. “For every human gene there are about 150 microbial genes, and they contribute a lot to our metabolism.” As with most of the molecules that come from food, the structures and activities of the microbial ones produced by metabolic processes remain largely unexplored. 

A picture of how much the microbiome matters is suggested by a mouse study where Dorrestein found that as much as 70% of the body’s chemistry was different depending on where researchers looked for the presence or absence of microbiota, including distant organs like the brain controlling how people think and behave. “And that is not to mention the nutrient availability from different diets,” Dorrestein says. 

“If we don’t have all the puzzle pieces, how are we ever going to understand how biology really functions?” It is going to be complicated, he says, but getting to that moonshot is going to require taking inventory of all molecules that exist in humans.