Native mass spectrometry asks a narrow question and answers it well: what is the mass of the intact assembly, and therefore how many copies of what does it contain. Everything difficult about the technique follows from two facts. The assembly has to survive a trip out of water, through a vacuum interface and into a detector. And the instrument does not report a mass — it reports a mass-to-charge ratio, from which mass and charge have to be disentangled.
The method is aimed at preserving and determining the native structure, composition and stoichiometry of biomolecules and their complexes from solution after transfer into the gas phase, and its practical enemy is heterogeneity. When a sample carries several coexisting stoichiometries, bound ligands or lipid populations, tens or even hundreds of peaks can overlap in a single spectrum, which makes assignment of ion mass, charge, composition and structure very challenging [1]. What follows covers keeping noncovalent contacts intact, why charge reduction helps, what a stoichiometry number establishes, and the caveat governing the whole exercise — the measurement happens in vacuum.
What has to survive, and what a protocol does about it
Mass spectrometry of complexes gives lower than atomic structural resolution. What it gives instead is mass, and from mass, stoichiometry and composition; fragile noncovalent interactions within protein complexes can be preserved in the gas phase under gentle ionization and transfer conditions [2]. The word doing the work there is gentle. Ionization energy, source pressure and voltage, and the collision energy applied downstream all trade signal against survival, and the settings that give a clean, narrow peak are frequently the settings that have already broken something.
A general protocol exists, published in a form that can be carried out on commercially available electrospray instruments without extensive modification. It begins with preparing suitable solutions for electrospray, follows the complex through each stage of transmission until detection, and extends to deliberate dissociation into multiple subcomplexes to generate interaction networks. The critical experimental factors that determine success are highlighted throughout [3].
Buffer exchange into a volatile electrolyte, usually by gel filtration or repeated dilution and concentration in a centrifugal device, is the standard first step; so is nanoelectrospray introduction from a pulled capillary. Neither is a finding. Both are convention, adopted because nonvolatile salts do not evaporate with the solvent and instead decorate every peak.
The solution you spray is not the solution you prepared
Almost all native electrospray work uses aqueous ammonium acetate, and it is a widely repeated misconception that ammonium acetate buffers a solution at neutral pH. A buffer consists of a weak acid and its conjugate weak base. Ammonium and acetate are not a conjugate acid/base pair, so they do not constitute a buffer at pH 7. Dissolving the salt in water does give pH 7, but that pH is highly labile. Ammonium acetate does buffer around pH 4.75, the pKa of acetic acid, and around pH 9.25, the pKa of ammonium — a buffering range covering the weak acid pKa plus or minus one pH unit. On that reasoning, neutral ammonium acetate solutions sprayed in positive ion mode were predicted to acidify to around pH 4.75 plus or minus one unit in the electrospray plume. The additive remains useful: it is a volatile electrolyte that can mimic the solvation properties proteins experience under physiological conditions, and the recommendation was terminological — call it ammonium acetate solution, not buffer [4].
That prediction was later examined directly. Using bulk evaporation experiments under electrospray-mimicking conditions together with molecular dynamics simulations that allowed proton transfer, one group described the droplet chemistry. At the tip of the Taylor cone, electrolytically generated protons are neutralized by acetate, leaving ammonium as the net charge carrier in weakly acidic nascent droplets. As the droplets evaporate they lose water along with the ammonia and acetic acid produced by proton transfer; ammonia leaves faster because it is more volatile, so acetic acid accumulates. With residual acetate, that acetic acid forms an acetate buffer which stabilized the average droplet pH at 5.4 plus or minus 0.1, as governed by the Henderson-Hasselbalch equation. The authors took the success of native electrospray work in the literature to imply that this drop of about 1.6 units from an initially neutral solution can be tolerated by most biomolecular analytes on the short time scale of the electrospray process [5].
The practical reading is not that ammonium acetate is a bad choice. It is that the pH your complex experiences during ionization is not the pH you measured in the tube, and a complex whose assembly state is pH-sensitive in the range between 7 and roughly 5 is one whose native spectrum needs independent corroboration.
Charge reduction
Conventional native electrospray can disrupt labile interactions, and one of the standard countermeasures is to add a charge-reducing reagent to the solution before spraying — triethylamine, trimethylamine oxide and imidazole among them. Lowering the charge acquired during electrospray reduces the Coulombic repulsion that leads to dissociation, and such reagents may also lower the internal energy of the ions through evaporative cooling. A survey of imidazole derivatives across a soluble protein complex, a membrane protein complex in detergent, and lipoprotein nanodiscs with and without embedded peptides found that hydrophobic substituents at the 2 position on imidazole can significantly improve both charge reduction and gas-phase stability relative to existing reagents [6].
There is a second benefit that is purely arithmetic. Consider a 64,000 Da complex. At 16+ and 17+ its peaks fall near m/z 4,001 and 3,766 — about 235 Th apart. Reduce the same complex to 10+ and 11+ and the peaks fall near m/z 6,401 and 5,819, about 582 Th apart. Fewer charges push the ion to higher m/z and spread adjacent charge states farther apart, which is what helps when several species of different mass populate overlapping charge-state envelopes. The cost is that the analyzer must transmit and resolve ions at that higher m/z, which not every instrument does equally well.
Reading a stoichiometry off the spectrum
A native spectrum does not display the mass directly. It displays a series of peaks from the same species carrying different numbers of protons, and the mass is reconstructed from the series. Two adjacent peaks suffice. If a peak at m/z 4,001.0 and the next at m/z 3,765.7 belong to the same species one charge apart, the charge state of the higher-m/z peak is (3,765.7 − 1.0) ÷ (4,001.0 − 3,765.7), which is 16. The neutral mass follows as 16 × (4,001.0 − 1.0) = 64,000 Da. Copy number is then a division: that mass against a subunit mass calculated from sequence, provided the whole mass is subunits.
That proviso is where the interesting work is. A residual mass the subunit count does not explain is a cofactor, a lipid, a modification, retained solvent, or salt. Adducts set the resolution floor. A sodium ion retained in place of a proton adds about 22 Da to the neutral mass, which on a 20+ ion is a shift of only about 1.1 Th, whereas a 1,000 Da ligand on the same ion shifts it by 50 Th. It takes roughly forty-five unresolved sodium adducts to smear a peak across the same m/z window that a single such ligand would move it — so adduct heterogeneity is a peak-width problem long before it becomes a mass-assignment problem, and the width is what destroys the ability to count small ligands.
Two limits are worth stating plainly. Mass cannot distinguish molecules of identical composition and different arrangement, which is as true of a complex as it is of a peptide. And a mass consistent with a given stoichiometry is consistent with it, not proof of it — a hexamer and a dimer-of-trimers weigh the same.
Nonspecific binding: the artifact that looks like a result
When a protein and a ligand share a droplet, some of the complexes that reach the detector were assembled by evaporation rather than by affinity. The standard correction adds a reference protein to the nanoelectrospray solution which does not bind specifically to any solution component. Nonspecific binding then announces itself as reference-protein-plus-ligand complexes in the spectrum, and the fraction of reference protein carrying ligand gives a quantitative measure of the nonspecific contribution, so that errors introduced into association constants can be corrected. The method rests on the assumption that the fraction of proteins engaging in nonspecific binding is set by the number of free ligand molecules in the offspring droplets and is independent of the size and structure of the protein or complex — an assumption the authors showed to be generally valid, and demonstrated on two protein-carbohydrate systems where nonspecific binding was prevalent [7].
A follow-up study using pairs of proteins and small charged biomolecules that bind neither found that nonspecific binding of charged molecules during electrospray is a statistical process, independent of protein structure and size, but that the kinetic stabilities of those nonspecific interactions are sensitive to the nature of the protein ions. Distributions that appeared nonequivalent between two proteins were traced to differences in how far the nonspecific complexes underwent in-source dissociation; adding imidazole, which protects complexes from in-source dissociation, made the distributions equivalent in every case tested. The conclusion was conditional: the reference protein method can be applied to ionic ligands provided in-source dissociation of the nonspecific interactions is minimized [8].
A related trap concerns intensity ratios. The difference between the charge-state distributions of free protein and bound complex is commonly blamed on nonspecific binding or in-source dissociation, and is unwelcome because the association constant read from peak intensities then varies strongly with charge state. One analysis argued this difference can arise with no loss of complex integrity at all, simply because ligand binding changes the protein's solvent-accessible surface area in solution; the charge-state shifts across several protein-ligand systems were rationalized with a simple model based on the charged residue mechanism, and the authors argued the effect should be considered rather general [9].
Taking the complex apart
Stoichiometry answers how many. Topology — which subunit touches which — requires breaking the assembly in the instrument and reading the pieces. The caveat is that the released subunits have to reflect the native arrangement for the exercise to mean anything, and many proteins tend to favor unfolding on collision with neutral gas, the commonest activation method, which typically limits what collision-induced dissociation reveals about quaternary structure without further manipulation. Activation by collision with a surface behaves differently: subcomplexes released after surface collision have been consistent with native quaternary structure for several protein systems studied, including a chaperone approaching megadalton mass, an outcome attributed to the fast, energetic deposition a massive target permits [2].
Gas-phase dissociation also partitions charge unevenly, and the asymmetry is measurable. One group defined an asymmetric charge partitioning factor to quantify it and applied it to monomer ejection from a cytochrome c dimer and a beta-amylase tetramer, where roughly 60 to 70 percent of the precursor charge ended up on the ejected monomer, corresponding to factors of 1.38 and 2.51 respectively [10]. The two numbers differ because the baseline differs: an even split would put half the precursor charge on a dimer's ejected monomer but only a quarter on a tetramer's, so the same measured percentage represents a much larger departure in the tetramer. Charge asymmetry is not itself an error, but it means the charge a fragment carries is not a reliable readout of the size of the piece it came from.
The solution-phase caveat
Electrospray is gentle enough to move nonvolatile biomolecules into the gas phase without breaking covalent bonds, and in some cases noncovalent bonding is maintained well enough to characterize the solution structure of large complexes and native enzyme-substrate binding. But the gaseous environment can ultimately cause dramatic structural alterations. A synthesis of a range of studies proposed a stepwise evolution of native structure spanning picoseconds to minutes during and after transfer into the gas phase, which can involve side-chain collapse, unfolding, and refolding into new, non-native structures; control of individual experimental factors allows optimization for particular objectives [11]. Read carefully, that is a statement about time scales: what survives depends on how long the ion spends where.
Sample handling upstream can also set the answer. Most native mass spectrometry of membrane proteins relies on detergent solubilization, and detergents can promote local lipid rearrangement and can cause fluctuations in the oligomeric state of protein complexes — which is a direct threat to a stoichiometry measurement, since the observed oligomer may be a property of the detergent. A protocol that ejects complexes directly from sonicated lipid vesicles without detergents was developed to avoid this, using the observed masses together with omics-based strategies to assign subunit stoichiometry as well as cofactor and lipid binding [12].
How far the numbers track solution measurements has been tested head to head. Comparing nanoelectrospray mass spectrometry, surface plasmon resonance and isothermal titration calorimetry on human carbonic anhydrase I with eight characterized sulfonamide inhibitors whose affinities span more than four orders of magnitude, agreement between the three methods was excellent for two compounds, within experimental error; good for two more, varying by approximately one order of magnitude; and poor for others, sulpiride among them. Two of the reported mass spectrometry values illustrate the range: 5 plus or minus 1 nM for ethoxzolamide and 145.7 plus or minus 10.0 micromolar for sulfanilamide, about 29,000-fold apart [13]. Agreement is therefore system-dependent rather than assured — the honest summary for a technique whose measurement conditions are not the conditions being modeled.
What is not settled
Several things in this workflow are convention rather than established fact, and it is worth saying which. Ammonium acetate is chosen for volatility and for solvation properties resembling physiological ones, not because it holds pH; the pH it delivers during spraying was determined to be about 5.4 in one study of the droplet chemistry [5], and whether that generalizes across concentrations, additives and source geometries is not established.
Deeper down, many fundamental questions about the properties of gas-phase proteins remain open, and further research is needed to answer them — a point made explicitly by practitioners of surface-induced dissociation even while reporting that their subcomplexes matched native topology [2]. Ion-mobility cross sections measured on the dissociation products give insight into the dissociation pathways, which the authors write may mimic disassembly pathways — the modal verb is theirs. A native stoichiometry is a strong constraint on what a complex can be. It is not, by itself, a solution-phase result.