Mass spectrometry is sensitive enough to detect compounds at extremely low concentrations, but before a sample reaches the instrument, it needs to go through a transformation that determines much of what the data will actually show. Sample preparation in metabolomics is where the analysis begins, and where a lot can go wrong [1].
Why Sample Preparation Matters in Metabolomics
Mass spectrometers usually are not used to run samples directly but are coupled to chromatography, either gas chromatography (GC) or liquid chromatography (LC). Therefore, the sample has to be compatible with both the separation system and the ionization source. This compatibility requirement drives everything.
For reversed-phase (RP) LC, the mobile phase starts as mostly water, often 90–100% aqueous. If a sample is prepared in a strongly hydrophobic solvent, when injected, it can precipitate on the column head – the metabolites “crash out” of solution before they can be retained and separated. The practical rule is that injection solvent polarity has to be close to or weaker than the initial mobile phase.
For GC, the solvent generally has to be volatile enough to evaporate rapidly at the injector temperature, typically 250–350°C, so highly aqueous solvents are usually unsuitable. Common choices include ethyl acetate, methanol, or hexane depending on what is being extracted. For derivatization-based GC-MS, widely used for amino acids, organic acids, and sugars, samples are typically dried completely and reconstituted in a derivatization reagent such as BSTFA or MSTFA, which converts hydroxyl and carboxyl groups into trimethylsilyl derivatives to improve volatility and chromatographic behavior.
Ion suppression is the other major constraint. Salts and detergents are destructive in ESI-based LC-MS: they compete with analytes for ionization at the spray interface, reducing sensitivity and making quantification unreliable. Phosphate buffers, Tris, and common biological surfactants such as SDS are all problematic.
Large molecules – proteins, high molecular weight lipids, and polysaccharides, all cause problems at both ends: they suppress signal in the source and clog the LC column. In GC, they burn in the injector at high temperatures and generate a persistent chemical background that obscures real peaks. Any sticky, gummy matrix residue accumulates in the tubing and valves of the chromatography system, deposits on the ion transfer components in the mass spectrometer, and causes carryover between samples.
In LC-MS/MS-based clinical or high-throughput work, carryover from inadequate sample preparation can become a source of false positive signals.
Removing the Matrix During Metabolomics Sample Preparation
The first job of metabolomics sample preparation is removal of everything that’s not soluble.
Protein precipitation is the most common first step for biofluid analysis: adding two to three volumes of cold methanol or acetonitrile to plasma or serum denatures and precipitates proteins, which are then removed by centrifugation. The supernatant is taken for analysis.
This is fast and inexpensive, but it is a blunt instrument — precipitation removes proteins but does not selectively retain or concentrate any particular class of metabolites.
For solid samples such as tissue, feces, or bacterial pellets, mechanical or chemical homogenization is necessary before liquid-phase extraction can occur. Without complete homogenization, it may be possible to analyze primarily surface material, which is not representative of the bulk sample.
Metabolite Extraction: Getting Metabolites Out
Once the matrix is handled, the actual metabolite extraction begins, and this is where the chemistry gets specific.
Solvent selection is driven by polarity matching: nonpolar metabolites: lipids, fat-soluble vitamins, hydrophobic secondary metabolites, partition into organic solvents like hexane, dichloromethane, or ethyl acetate. Polar metabolites: amino acids, organic acids, nucleotides, most central carbon metabolites, require aqueous or mixed solvents like methanol:water or acetonitrile:water.
Bacterial biofilm and mucus, for example, can be extracted with ethyl acetate, which efficiently recovers a broad range of secondary metabolites, including many polyketides, terpenoids, and nonribosomal peptides.
Intracellular metabolites present a different problem: the cell membrane has to be disrupted before extraction can work. Freeze-thaw cycling is one approach. Repeated rapid freezing in liquid nitrogen or a -80°C freezer followed by thawing causes ice crystal formation that can rupture membranes mechanically, releasing cytoplasmic contents. Bead beating and sonication are alternatives for more robust cell walls, such as those found in yeast, bacteria, or plant tissue.
Some metabolites are genuinely difficult to extract because their biology works against simple solvent-based recovery. Folate (Vitamin B9) is a useful example.
In blood, folate is present in plasma and inside erythrocytes, where it occurs largely in polyglutamate forms.
A common approach is to first treat the sample with acidified conditions (typically dilute ascorbic acid at low pH) to denature the binding proteins and release folate, combined with heat treatment to lyse red blood cells. But folate itself is water-soluble with multiple glutamate conjugates, so recovery then relies on aqueous partition. The problem is compounded by folate’s instability: it oxidizes rapidly, so extraction has to be done quickly under antioxidant conditions and protected from light.
A metabolite that is protein-bound, cell-sequestered, pH-sensitive, and light-labile requires several simultaneous accommodations just to measure accurately [2].
How Sample Preparation Introduces Bias in Metabolomics
Sample preparation is both art and science in the way that a good chromatographic method is: the underlying principles are well understood, but translating them to a real matrix with heterogeneous composition, unknown interfering compounds, and analytes spanning ten or more orders of magnitude of concentration requires judgment at every step. Those judgments propagate directly into the data.
The clearest example is pH during extraction.
Acidic conditions protonate acidic metabolites: fatty acids, bile acids, phenolic compounds, rendering them neutral and driving their partition into organic solvent. Basic conditions do the opposite, deprotonating acids and keeping them in the aqueous phase while neutralizing amines and making them more organically extractable.
Additionally, low pH in the mobile phase improves ESI ionization and peak shape for basic metabolites in positive mode, compounding the analytical bias. The net result is that low pH extraction and positive mode analysis together strongly favor one portion of the metabolome; switching to high pH and negative mode shifts the window substantially toward acidic compounds.
Targeted vs. Untargeted Metabolomics Sample Preparation
In GC-MS acidifying the sample will lead to partitioning of volatile acids into headspace to make them more accessible for GC-MS.
For targeted quantitative work, where it is known exactly what is being measured and there exists a validated method for that specific analyte, this is manageable. You optimize extraction for your target and accept that other things are not well-recovered.
For untargeted metabolomics, where the goal is to detect and compare as much of the metabolome as possible without prior compound assumptions, it is a fundamental problem. Two datasets from identical biological samples, processed with acid-based versus base-based extraction, will show substantially different metabolite profiles, not because the biology is different, but because the chemistry selects for different compound classes. These datasets are not directly comparable, which is a major source of irreproducibility in the metabolomics literature [3].
Approaches to Reduce Sample Preparation Bias
Several approaches try to address this. For example, QuEChERS (Quick, Easy, Cheap, Effective, Rugged, Safe) was originally developed for pesticide residue analysis in food matrices, uses a combination of acetonitrile extraction followed by salt-induced phase partitioning and dispersive SPE cleanup to achieve reasonably broad, pH-neutral coverage across compound classes [4].
Biphasic extractions – classically Bligh-Dyer or Folch for lipidomics, produce a distinct organic and aqueous layer simultaneously. The organic phase captures nonpolar lipids and hydrophobic metabolites; the aqueous phase retains polar, water-soluble compounds.
Running both layers for analysis gives complementary coverage that neither alone provides, at the cost of additional analytical runs.
Dynamic Range in Lipidomics Sample Preparation
Lipid analysis adds a particularly acute version of a general metabolomics problem: dynamic range.
In plasma lipidomics, phosphatidylcholines (PCs) are the dominant lipid class, present at high micromolar concentrations and accounting for a disproportionate share of the total ion signal in positive-mode ESI. Bioactive lipid mediators: prostaglandins, leukotrienes, hydroxy-fatty acids, and other oxylipins, may be present at concentrations in the low picomolar range, meaning the difference between the most and least abundant lipids of biological interest can span six to nine orders of magnitude [5].
No single LC-MS injection can capture both ends of this range simultaneously without bias. The high-abundance species saturate the detector and suppress ionization of low-abundance ones before the instrument can measure them.
Solid-Phase Extraction for Low-Abundance Lipids
Solid-phase extraction (SPE) is the standard solution in this case. By passing sample extract through an appropriate sorbent (C18 for broad lipid capture, mixed-mode sorbents for more selective retention) the high-abundance compounds such as phospholipids can be washed away under conditions where low-abundant ones (e.g. oxylipins) are retained.
The retained fraction is then eluted with a solvent that recovers the analytes of interest in a concentrated, cleaner form. This accomplishes two things simultaneously: it removes the dominant interfering compounds and pre-concentrates the low-abundance targets, improving both detection limit and linearity.
The choice of elution solvent is itself a tunable parameter – different solvents or solvent gradients can fractionate the retained compounds by polarity, providing an additional layer of selectivity before the chromatographic separation even begins.
Practical Considerations for Metabolomics Sample Preparation
None of this should be paralyzing. Sample preparation in metabolomics is governed by chemistry, and chemistry is solvable with proper understanding and planning accordingly.
For untargeted analysis, running complementary extractions in parallel (e.g. one aqueous, one organic) covers substantially more chemical space than any single method and reduces the systematic blind spots that any individual protocol introduces. Internal standards spiked before extraction, covering a range of chemical classes and polarities, allows to track extraction efficiency and correct for variability between samples.
Modern mass spectrometers are sensitive enough that even an imperfect extraction often leaves sufficient analyte for detection. The harder problem is not detection but quantification and cross-sample comparison: concentrations measured across samples extracted with slightly different methods, or even the same method under slightly different conditions (temperature, incubation time, centrifugation speed), require careful normalization.
Detection of a compound in an untargeted run tells us that it is present. Saying how much is present, and comparing that amount reliably across samples, is a separate and more demanding problem that starts with extraction.
Metabolomics Sample Preparation at Arome Science
At Arome Science, sample preparation is part of every metabolomics project we take on. We maintain validated extraction protocols for common sample types: plasma, urine, stool, bacterial cultures, tissue, and can develop custom workflows when the matrix or target analytes require it.
