For the past two years, the metabolomics community has debated the extent to which the “dark metabolome” – the large proportion of unidentified peaks in untargeted mass spectrometry experiments – can be explained by technology artifacts, instead of yet-to-be-discovered biology.
Gary Siuzdak and Martin Giera ignited the debate with their 2024 paper arguing, having mined the METLIN MS/MS database, that the dark metabolome is largely made up of fragment ions generated during electrospray ionization (ESI) via in-source fragmentation (ISF). Then in 2025, Pieter Dorrestein and Yasin El Abiead responded with evidence from biological samples showing that many unannotated features persist even after accounting for ISF, indicating they cannot be dismissed as mere artifacts without additional analysis.
Now, Richard Zare, Xiaowei Song and colleagues have shown that microdroplets formed during electrospray ionization can promote a wide array of chemical reactions, producing artifact ions that are frequently misassigned as metabolites – and potentially accounting for many signals that cannot be explained by in-source fragmentation alone. Importantly, these artifact ions may themselves undergo in-source fragmentation, further expanding the apparent “unknown” chemical space.
A simple question
The team’s interest in the problem began in early 2025 with a simple question: why do metabolomics experiments detect far more ions than expected? Even if each of roughly 2,000 human metabolites produced several common adduct and isotope peaks in electrospray ionization, the total would only reach around 18,000 ions. In practice, however, a direct injection of a biological sample can produce more than 50,000 signals – and far more in LC-MS experiments.
“This huge discrepancy cannot be explained simply by impurities or background ions,” say Zare, Professor of Chemistry at Stanford University, USA, and Song (lead author, who recently left Zare’s research group to assume an Assistant Professor position at Tsinghua University in Shenzhen, China). “It made us wonder what these extra peaks are, where they come from, and how they are formed.”
Zare’s group had long studied microdroplet chemistry and knew that the electrospray ionization process generates microdroplets. “Microdroplets, particularly their air-water interfaces, are rich in reactive species such as reactive oxygen and nitrogen species (ROS/RNS),” say Zare and Song. “They can easily react with metabolites to cause artifact ions.” The researchers reasoned that metabolites could also be activated in these droplets to form reactive intermediates – cations, anions, or radicals – which could then interact to create new structures, potentially helping to explain some of the long-mysterious signals in the “dark metabolome.”
To test their hypothesis, the researchers designed a series of controlled electrospray experiments using both individual metabolite standards and defined mixtures. By spraying known compounds into high-resolution mass spectrometers, they could determine whether additional ions appeared even when the starting chemistry was fully known.
In one key experiment, the team electrosprayed a solution containing 70 well-characterized metabolites and analyzed the resulting spectra using ESI-MS. To focus on meaningful signals, they removed background peaks associated with solvents and instrument materials and ignored very low-intensity signals by applying a minimum detection threshold. They then compared the number of observed peaks with the number expected from simple adduct formation and isotope patterns.
The results suggested that far more ions were present than conventional ionization chemistry alone could explain.
Setting the detection threshold to 1,000 and subtracting background ions derived from solvents and tubing, the researchers observed 1,744 ions in the mass spectrum. Yet only 113 peaks corresponded to simple proton or sodium adducts of the original metabolites – just 5.3 percent of the total ion count and about half of the total signal intensity.
“We did not anticipate that such a dramatic difference would be found,” say Zare and Song.
The missing piece?
The researchers also considered whether the unexplained peaks could simply be the result of in-source fragmentation (ISF): a well-known phenomenon in which metabolite ions break apart inside the mass spectrometer and produce additional fragment signals. To distinguish between fragmentation and chemistry occurring before ions enter the instrument, the team systematically varied parameters that affect the electrospray plume, including spray voltage and the distance between the emitter and the MS inlet.
They found that increasing the distance – effectively giving the droplets more time to travel through air before entering the instrument – increased the abundance of many of the additional ions. This behavior is consistent with chemical reactions occurring in the droplets themselves, rather than fragmentation inside the instrument.
“In-source fragmentation is indeed a very well-known phenomenon that has been appreciated for many years,” say Zare and Song. “However, ISF is metabolite structure and MS configuration-dependent and can only explain part of total unknown ions – and their concentration must be lower than that of the original metabolite ions. Our work demonstrates that metabolites in spraying microdroplets undergo chemical changes before entering the inlet of the mass spectrometer. So, this work complements and extends the ISF explanation – which alone cannot account for the myriad of ions we observed, particularly those of larger mass than those of the initial metabolites in the solution.”
“We believe that microdroplet-induced dark reactions provide a break-through direction to significantly resolve the ‘dark metabolome’ problem,” they add. “If there is any unresolved piece in this puzzle, it is the fact that there are still a lot of unexplored microdroplet reactions, which may generate more new structures.”
“From a personal perspective, this paper is quite satisfying,” says Gary Siuzdak, who wasn’t involved in the research. “Our work on the in-source fragmentation of over 900,000 authentic standards showed why individual metabolites can produce families of peaks, but it never fully explained why untargeted LC-MS datasets appear so chemically crowded to begin with. The realization that spray-based ionization can generate new molecules through microdroplet-mediated reactions – and that these products can themselves undergo in-source fragmentation – helps connect those missing pieces.
“Together, microdroplet chemistry and ISF form a coupled process that naturally explains the apparent explosion of signals in LC-MS data,” he adds. “Seeing these two mechanisms integrated provides a more complete, coherent picture of what we are actually measuring, and why empirical recognition of source-derived ions is so critical for keeping metabolomics grounded in biology.”
Context matters
But not everyone is convinced that microdroplet chemistry can explain such a large proportion of the dark metabolome – particularly in biological samples.
“It is beautiful work and certainly serves as an additional caution when doing metabolomics using MS/MS reference libraries,” says Pieter Dorrestein. “However, observing reactions in microdroplets, while impressive, is like testing a wave in a swimming pool and assuming it behaves the same in the ocean – context changes everything.”
“For me, it is difficult to assess the relevance of observations obtained under non-standard conditions for LC-MS experiments in biological metabolomics studies, such as placing the spray nozzle at unusually large distances from the MS inlet and altered energetics of ion transfer.”
Dorrestein also notes that in LC-based workflows the physicochemical context during ionization differs substantially from experiments that mix analytes together at high concentrations, because compounds are separated before entering the electrospray.
“When chromatographic separation is used, intermolecular reactions that couple two distinct molecules (excluding solvent-derived processes or transformations such as oxidations) would require both reactants to elute at sufficiently high concentrations at the same retention time for them to react. If they are not present together, the reaction cannot take place.”
As an example, Dorrestein points to the reported acetylation of histamine. “This would require acetate and histamine to co-elute if the reaction were occurring post-chromatography within electrospray droplets, which is unlikely under typical chromatographic conditions,” he says. “This is not addressed in the Zare paper and the methods lacked sufficient detail to fully assess what they did in the reanalysis of a public LC-MS dataset.”
He also argues that there are “fairly straightforward” ways to evaluate whether a detected ion originates from a species present prior to chromatography or from a post-column reaction. “Extraction of all relevant ion signals followed by chromatographic peak-shape and retention-time correlation analysis allows one to assess whether features co-vary in a manner consistent with a common origin, some of which we recently wrote a perspective on,” he says. “This type of analysis was not reported by Zare et al. in the reanalysis of the LC data, meaning it is impossible to determine whether the observed species arise from in-droplet chemistry or were already present before chromatography.”
Consequently, Dorrestein suggests the findings could be particularly relevant for techniques that lack chromatographic separation, “such as most imaging methods, given that it will be harder to distinguish whether or not the molecules were already part of the sample or a result of in-droplet or gas-phase chemistry – even ion mobility will not help,” he says. He adds that other approaches – including MALDI, nanoDESI, and DESI – may also need to consider the possibility of such reactions, particularly if instrument settings resemble those used in the study.
“How much impact in-droplet chemistry has under typical instrument conditions will require further research,” he says.
“Even if the prevalence of specific microdroplet reactions depends on experimental context, the broader implication remains that electrospray sources are chemically active environments in which both newly formed species and their associated fragment ions can expand the number of signals observed in untargeted LC-MS datasets.”
Siuzdak, however, suggests that even if the extent of these effects varies by workflow, the underlying principle may still hold. “Even if the prevalence of specific microdroplet reactions depends on experimental context, the broader implication remains that electrospray sources are chemically active environments in which both newly formed species and their associated fragment ions can expand the number of signals observed in untargeted LC-MS datasets,” he says.
There are also real stakes to this debate. Siuzdak points to the risk that students and postdocs may spend significant time pursuing molecular identities that do not correspond to real biological entities when provisional annotations are interpreted too strongly. Dorrestein, meanwhile, has noted that the discussion has already influenced funding and publication outcomes, with grants and manuscripts focused on new metabolite discovery being rejected, potentially affecting early-career researchers.
A warning and a roadmap
To help researchers interpret their data more reliably, Zare’s team also compiled a catalogue of the microdroplet reactions most likely to occur during electrospray ionization. The researchers grouped these “dark reactions” into five broad categories based on how the reactive intermediates form: electron impact-like ionization, unimolecular dissociation, intermolecular reactions, radical reactions, and solvent-driven transformations such as methylation or acetylation.
“We summarized all possible microdroplet reaction roles that have been observed to happen to various metabolites to the best of our knowledge,” say Zare and Song. “Metabolomics researchers can then use these rules to take a retrospective analysis of their previous untargeted data.”
The researchers also explored whether the formation of these artifact ions could be reduced. To do so, they developed a modified, softer spray configuration known as air-flow-assisted spray ionization (AFASI). In this setup, the electrospray emitter and mass spectrometer inlet are enclosed in a sealed chamber connected to a vacuum pump, and argon was supplied instead of nitrogen as the chamber’s atmosphere.
The design creates a drier environment that suppresses reactions at the air-water interface while also shortening the time droplets spend traveling toward the instrument inlet. “This reduces the interaction between gas and metabolites across the microdroplet interface,” the researchers explain, helping minimize the formation of dark ions during analysis.
Taken together, the findings provide both a warning and a roadmap for metabolomics researchers. “This work provides caution to researchers in untargeted metabolomics about what they are really observing from a mass spectrum,” say Zare and Song. “Researchers can know how to use microdroplet reaction rules to predict and recognize the possible artifacts, and how to minimize dark ion formation by carefully changing ionization conditions.”
Beyond metabolomics, the study also highlights the broader potential of microdroplet chemistry. Because reactions in microdroplets can rapidly generate large numbers of new molecular structures, the researchers suggest the phenomenon could be harnessed for applications such as combinatorial chemistry or fragment-based drug discovery. “Given n precursors as input, you can expect no less than n² possible new chemicals as the output in an ultrafast way,” they say.
Newsletters
Receive the latest analytical science news, personalities, education, and career development – weekly to your inbox.
