Written and maintained by CASRAI Editorial Board
Last updated
A lipidomics mass spectrometry experiment is shaped by two decisions that are locked in before a single spectrum is acquired. The first is chemical: which extraction protocol pulls the lipidome out of a plasma, tissue, or cell sample without systematically discarding some classes over others. The second is instrumental: whether lipids are infused directly into the source (shotgun lipidomics) or separated by chromatography first, and in which acquisition mode. Neither choice is reversible after the run, and both interact with a third question — how much structural confidence a given identification actually earns, and how to report that honestly. This guide covers all three in the order they are actually decided.
Why lipid extraction chemistry is the first fork in the road
Lipids are not a single chemical class — the lipidome spans deeply hydrophobic species (triglycerides, cholesteryl esters, sterols) to markedly more polar ones (lysophospholipids, sphingosine-1-phosphate, free fatty acids). A single extraction protocol has to recover across that whole polarity range, or the sample is already biased toward whichever classes partition most efficiently before a spectrum is ever collected. The classical approach is biphasic liquid-liquid extraction: a chloroform/methanol/water (or equivalent) mixture is added to the homogenized sample and allowed to separate into an aqueous phase (carrying polar metabolites and salts), a protein interphase, and an organic phase (carrying the bulk of the lipidome). Which extraction chemistry you use determines which phase the lipids end up in, how much of each class survives the partition, and how easy the organic layer is to recover cleanly.
Three extraction chemistries and the trade-off each one makes
Three protocols dominate practice, and they are not interchangeable defaults — each makes a different trade-off between recovery, throughput, and solvent handling.
- Folch extraction (chloroform:methanol, 2:1, followed by a water wash) uses a relatively large solvent-to-sample ratio and is generally regarded as giving the most complete lipid recovery of the classical methods, which is why it remains a reference method for lipid-rich tissue. The lipid-bearing organic phase settles at the bottom of the tube, under the aqueous layer and the protein interphase — recovering it cleanly means pipetting through or around that interphase without disturbing it, which is awkward to automate.
- Bligh & Dyer extraction uses a smaller, more tightly optimized chloroform:methanol:water ratio in a single tube, originally developed for lower-lipid-content samples. It is faster and uses less solvent than Folch, at some cost to recovery in lipid-rich matrices, and shares the same bottom-organic-phase pipetting problem.
- Matyash (MTBE) extraction replaces chloroform with methyl-tert-butyl ether and inverts the phase geometry: the lipid-bearing organic phase settles on top, with the protein pellet and aqueous phase below it. That single change is the practical reason MTBE extraction has become the default for high-throughput and automated lipidomics — the organic phase can be pipetted straight off the top without touching the interphase, and MTBE is non-chlorinated, which simplifies solvent disposal and reduces the toxicity handling burden chloroform-based methods carry.
As a selection rule: automated or high-sample-count workflows generally favor MTBE for the top-phase pipetting alone; a study where maximizing recovery of low-abundance classes matters more than throughput may still favor Folch’s larger solvent ratio. Monophasic (single-step, no phase separation) protocols also exist for very limited sample material, trading some class coverage and quantitative rigor for speed and a smaller required sample size.
Add class-representative internal standards before extraction, not after
Extraction efficiency is not uniform across the lipidome — it varies by class, and even between individual species within a class. That means an internal standard mix has to be spiked into the sample before homogenization and extraction, not added to the finished extract just before injection. A standard added after extraction can only correct for instrument response drift; it cannot correct for a lipid class that was systematically under-recovered during the extraction step itself, because by the time the standard is added that loss has already happened. A lipidomics standard mix that spans the major classes being quantified (commonly a set of stable-isotope-labeled or odd-chain lipids, one or more per class) is what makes class-level recovery correction possible at all. This is the same logic that governs pooled quality-control sample design in untargeted metabolomics — anything meant to correct for a systematic effect has to enter the workflow upstream of where that effect happens, not downstream of it.
One practical efficiency note: because biphasic extractions partition polar metabolites into the aqueous phase and lipids into the organic phase from the same homogenate, a single extraction can support both an untargeted metabolomics and a lipidomics workflow run from one sample — worth planning for at the internal-standard-spiking stage if both analyses are planned.
Choosing an acquisition strategy: shotgun versus LC-MS
Once the extract exists, the next decision is whether it goes into the source directly or through a chromatographic separation first.
- Shotgun (direct infusion) lipidomics infuses the extract straight into the ion source with no chromatography, typically on a high-resolution instrument running survey scans plus class-targeted precursor-ion or neutral-loss scans. It is fast, requires essentially no method development, and is well suited to relative quantitation across large sample cohorts. The trade-off is real: shotgun acquisition cannot resolve isomeric or isobaric lipid species that differ only in acyl-chain regiochemistry or double-bond position, since there is no retention-time axis to separate them, and it is more exposed to ion suppression because the entire lipidome reaches the source at once rather than being spread out over a chromatographic run.
- LC-MS lipidomics separates the extract chromatographically before it reaches the source — reversed-phase separation resolves species within a class largely by acyl-chain length and unsaturation, while HILIC separates primarily by headgroup class (see HILIC method development and how LC and MS are coupled). Retention time becomes a second identification axis on top of m/z, and spreading elution over time reduces the simultaneous ion-suppression load shotgun acquisition carries. The cost is longer run time and real method-development effort.
Within either strategy, the data-dependent (DDA) versus data-independent (DIA) acquisition trade-off familiar from other MS-based -omics applies here too — DDA selects a limited number of the most abundant precursors per cycle for fragmentation, giving deep but somewhat stochastic MS2 coverage, while DIA fragments broad, fixed m/z windows systematically, trading reproducibility for a harder deconvolution problem (see the same trade-off worked through in more depth in DDA vs. DIA for proteomics). Lipidomics adds one further acquisition decision that most other -omics applications don’t: ionization polarity. Choline-headgroup classes (phosphatidylcholine, sphingomyelin) ionize and fragment informatively in positive mode, producing a characteristic phosphocholine headgroup fragment; anionic classes (phosphatidylserine, phosphatidylinositol, phosphatidic acid, free fatty acids) are addressed far more effectively in negative mode. A method aiming for full-lipidome coverage commonly runs both polarities, either as separate injections or via fast in-run polarity switching — a single-polarity method should be a deliberate scope decision, not an oversight.
Instrument choice follows from what you just decided
Targeted quantitation of a defined lipid panel — a fixed ceramide panel, for example — suits a triple quadrupole running scheduled multiple-reaction-monitoring transitions per species, the same configuration logic covered in triple quadrupole LC-MS/MS setup. Untargeted or discovery lipidomics, where the point is annotating as much of the lipidome as possible rather than quantifying a known list, benefits from the accurate-mass survey scans a high-resolution QTOF or Orbitrap platform provides, paired with data-dependent MS2 — see how much resolving power you actually need and the direct QTOF vs. triple quadrupole comparison for the procurement-level version of this decision.
Reporting identifications: the annotation-confidence conventions specific to lipids
What a lipidomics acquisition actually measures does not always support the level of structural detail a lipid name implies, and the field has a standardized way to say so. The LIPID MAPS shorthand nomenclature (Liebisch et al., Journal of Lipid Research, 2020) defines nested tiers of structural resolution, and each one corresponds to a specific amount of MS evidence:
- Species (sum composition) level — e.g.
PC 34:1: total acyl carbons and total double bonds for the class, with no claim about which individual chains make up that sum. This is the level an accurate-mass MS1 survey alone supports, because MS1 alone cannot distinguish which chain combination produced a given mass. - Molecular species level — e.g.
PC 16:0_18:1: the individual acyl chains are resolved, but their position on the glycerol backbone and the position of each double bond are not. Reaching this level requires diagnostic MS2 fragment ions that reveal the individual chains, not just the summed composition. - sn-position level — e.g.
PC 16:0/18:1: full glycerol-backbone regiochemistry resolved. Routine DDA/DIA fragmentation does not generally deliver this; it needs specialized fragmentation approaches or supporting chromatographic evidence most workflows don’t generate as a matter of course.
A further tier — double-bond position — is reachable with specialized techniques such as ozone-induced dissociation or Paternò-Büchi derivatization, but these are not standard acquisition-method components and shouldn’t be assumed available by default. The practical discipline this framework enforces: report an identification at the level your actual MS/MS evidence supports, and no further. A survey-scan-only untargeted run should publish species-level IDs (PC 34:1), not a molecular-species or sn-position notation implying resolution the acquisition never generated. That same “don’t claim more precision than the method delivered” discipline runs through analytical reporting generally — see, for a parallel example from a different technique, how isotope ratio MS reporting conventions handle the same problem.
Frequently asked questions
What’s the practical difference between shotgun and LC-MS lipidomics?
Shotgun infuses the extract directly with no chromatography — fast and simple, but unable to separate isomeric or isobaric species and more exposed to ion suppression. LC-MS adds a chromatographic separation step first, which resolves more isomers and reduces simultaneous ion suppression, at the cost of run time and method development.
Does the extraction method matter if the goal is untargeted profiling rather than quantitation?
Yes. Even a purely qualitative untargeted survey is still shaped by which classes the extraction chemistry recovers efficiently — a method that under-recovers a given class will under-represent it in the resulting profile regardless of how good the acquisition afterward is.
Can one extraction support both lipidomics and metabolomics from the same sample?
Often yes. Biphasic extractions (Folch, Bligh & Dyer, MTBE) partition polar metabolites into the aqueous phase and lipids into the organic phase from the same homogenate, so a single extraction step can feed both a metabolomics and a lipidomics workflow — worth planning for when internal standards for both analyses are spiked in.
What does “PC 34:1” tell me that “PC 16:0/18:1” tells me more of?
PC 34:1 only states the total acyl carbon count and double bond count summed across both chains — several real chain combinations could produce that sum. PC 16:0/18:1 additionally resolves which two chains those are and which glycerol position each occupies, a higher confidence level that requires more specific MS/MS evidence to support.
Do I need to acquire in both ionization polarities?
For full-lipidome coverage, generally yes — choline-headgroup classes fragment informatively in positive mode and anionic classes in negative mode, so a single-polarity method is scoped to only part of the lipidome by default, which is fine if that’s a deliberate choice but not otherwise.
