A mass spectrometer weighs molecules. It turns them into ions in a vacuum and measures how they move in electric or magnetic fields[1]. For peptides it does two jobs. It confirms that a peptide has the mass it should have, and, if the peptide is broken into pieces inside the instrument, it reads the amino acid sequence from the masses of those pieces.
During the 1990s mass spectrometry displaced Edman degradation as the main way to identify and sequence proteins, because it is far more sensitive, fragments a peptide in seconds rather than hours or days, does not need a pure sample, and is not stopped by a blocked amino terminus[2]. But it has blind spots that follow directly from what it measures. Anything that does not change the mass, such as swapping leucine for isoleucine or an L-amino acid for its D mirror image, is invisible to the mass alone.
| What it measures | Mass-to-charge ratio (m/z) of ions |
|---|---|
| Getting peptides into the gas phase | Electrospray ionisation (ESI) or matrix-assisted laser desorption/ionisation (MALDI) |
| Reading a sequence | Tandem MS (MS/MS): the peptide ion is broken, mostly at peptide bonds, into b and y ions |
| Usual way to identify a peptide | Matching its fragment spectrum to sequences in a protein database |
| Cannot tell apart by mass | Leu and Ile; L- and D-amino acids; Gly-Gly and Asn; Gly-Ala and Gln |
Getting a peptide into the gas phase
A peptide cannot simply be boiled into a mass spectrometer. Large polar molecules “cannot be vaporized without extensive, even catastrophic, decomposition”, and for decades this kept biological molecules out of reach of the instrument[1]. Two “soft” methods, developed in the late 1980s, solved the problem, and in 2002 John Fenn and Koichi Tanaka shared half of the Nobel Prize in Chemistry “for their development of soft desorption ionisation methods for mass spectrometric analyses of biological macromolecules”[3].
Electrospray
In electrospray ionisation a solution of the peptide flows out of a fine needle held at a few kilovolts. The liquid breaks into a spray of charged droplets, the solvent evaporates, and as each droplet shrinks its surface charge grows until it tears apart into smaller droplets. The cycle repeats until peptide ions leave the droplets and enter the gas[1]. Fenn's group was candid that the exact way the final ions are released had “certainly not been proved”[1].
The defining feature of electrospray is that molecules pick up several charges, usually protons. How many depends on the molecule. Gramicidin S gave singly and doubly charged ions, cyclosporin A, a peptide of similar size, only singly charged ones, and bradykinin a strong triply charged peak. The authors linked this to the two basic residues in the ring of gramicidin S and the basic residues at both ends of bradykinin[1]. Because the instrument measures mass divided by charge, multiple charging brings even large proteins into the range of an ordinary analyser: bovine insulin, 5,730 Da, appeared with 4 to 6 charges between m/z 950 and 1450[1]. Electrospray works straight from a flowing liquid, so it can be coupled directly to liquid chromatography; in proteomics, peptides are eluted from a capillary column 50 to 150 µm wide at about 100 nanolitres per minute straight into the spray[2].
MALDI, and who invented it
In matrix-assisted laser desorption/ionisation the peptide is mixed with a large excess of a small light-absorbing compound, the matrix, and dried on a metal plate. A laser pulse vaporises the matrix, which carries the peptide with it into the gas phase, and the peptide ends up mostly singly protonated[2]. Because each peptide gives one main peak, MALDI spectra of mixtures are easy to read.
The history is less tidy than the Nobel citation. Tanaka's group at the Shimadzu Corporation mixed proteins with an ultra-fine cobalt powder, particles of about 300 Å, suspended in glycerol, and used a nitrogen laser to record lysozyme, chymotrypsinogen and carboxypeptidase A, up to 34 kDa[4]. In the same year, Michael Karas and Franz Hillenkamp in Münster published spectra of proteins up to about 67 kDa using an organic matrix, nicotinic acid, which absorbs strongly at the wavelength of their laser[5]. The matrices used today are of that second kind, small aromatic acids[2]: α-cyano-4-hydroxycinnamic acid, for example, favours peptides of 800 to 1800 Da, and sinapinic acid larger ones[6].
Reading the mass
A peptide of mass M carrying z protons appears at m/z = (M + z × 1.00728) / z. Steen and Mann's example is a peptide of 1232.55 Da that, doubly protonated, is seen at (1232.55 + 2 × 1.0073) / 2 = 617.28[2]. Peptides made by trypsin usually carry two protons in electrospray; longer peptides, or those with extra basic residues such as histidine, can carry more[2].
The charge can be read from the peak itself. Each peptide signal is a small cluster of peaks, because about 1% of carbon atoms are 13C, and the members of the cluster are one mass unit apart. On the m/z scale they are therefore 1 apart for a singly charged ion and 0.5 apart for a doubly charged one[2]. Seeing those peaks separately takes resolution: an ion trap barely resolves them for doubly charged ions, a time-of-flight instrument with a resolution of 10,000 separates them cleanly, and Fourier-transform ion cyclotron resonance instruments exceed 100,000 with a mass accuracy of a few parts per million[2]. The first peak of the cluster, containing only 12C, is the monoisotopic mass, which is why it is the number to compare with a calculated value; the peptide calculator page explains how that value is built from residue masses.
For Leu-enkephalin, Tyr-Gly-Gly-Phe-Leu, the calculated monoisotopic mass is 555.2693 Da. Its singly protonated ion, [M+H]+, should appear at m/z 556.2766 and the doubly protonated ion at m/z 278.6419.
A correct mass is good evidence but not proof of a correct peptide. It is consistent with the expected composition, but it does not establish the peptide's identity, and it says nothing about the order of the residues. Any rearrangement of the same residues has the same mass, and so do the look-alikes described below.
Breaking it to read the sequence
To read the order, the peptide is broken. In 1986 Donald Hunt's group at the University of Virginia described the method that became standard. A triple-quadrupole instrument used its first quadrupole to pick out one protonated peptide from a mixture, passed it into a second quadrupole filled with argon, where collisions broke it “primarily at the various amide linkages”, and used the third quadrupole to weigh the pieces[7]. They sequenced peptides from fractions containing 10 to 15 peptides each, without further purification, and a whole protein digest took no more than 4 or 5 days[7]. This is tandem mass spectrometry, MS/MS; the selected ion is the precursor, the pieces are product ions[2].
The names of the pieces
By the early 1980s different laboratories named the same fragments differently, so that, as Peter Roepstorff and Johan Fohlman wrote, reading a paper “first requires careful studies of the nomenclature used”. Their one-page proposal of 1984 gave a single scheme. A peptide backbone can break at three bonds in each residue; the fragment is called A, B or C if the charge stays on the N-terminal piece and X, Y or Z if it stays on the C-terminal piece, and the number counts the residues in the fragment from its own end[8]. Hydrogens moved during the cleavage were marked with apostrophes, so the most common C-terminal ion in the spectra of the time was written Y″[8]. Today the letters are written in lower case, and the scheme is usually credited to Roepstorff, Fohlman and Biemann[2].
In collision-induced fragmentation the useful ions are b and y, from cleavage of the amide bond[2]. A b ion holds the N-terminal residues and is commonly thought to be a ring structure, a protonated oxazolone; a y ion is a shortened peptide with its own C-terminus and water[6]. For singly charged ions:
- bi = sum of the first i residue masses + 1.00728
- yj = sum of the last j residue masses + 18.01056 + 1.00728
Two simple checks follow. The difference between neighbouring ions in a series is the mass of one residue; in Leu-enkephalin b3 − b2 = 57.02, glycine, and b4 − b3 = 147.07, phenylalanine. And the b and y ions from the same bond add up to the protonated peptide plus one more proton, 556.28 + 1.01 = 557.28, because each piece was counted with a proton but the precursor carried only one. Reading a series of such differences is how a sequence is extracted from a spectrum[2].
Other ions help. Loss of carbon monoxide from a b ion gives an a ion, 27.9949 lower. In spectra of doubly charged tryptic peptides it is normally seen only as an a2/b2 pair at the low-mass end[2]. For MALDI time-of-flight instruments, some of which fragment with much higher energy, the Mascot search program by default also looks for a whole series of a ions[6]. Small immonium ions, single residues cut at both sides, reveal which amino acids are present: m/z 86 for leucine or isoleucine, 120 for phenylalanine and 136 for tyrosine[7], all three of which Leu-enkephalin contains.
Why spectra are not tidy ladders
A real spectrum rarely shows every b and y ion with equal strength. Hunt's group noticed that where the charge ends up depends on residues with high gas-phase basicity: lysine, arginine, histidine, tryptophan and proline[7]. The idea was later formalised as the mobile proton model: to break a backbone bond, a proton has to move to it[6]. Several rules follow, and all of them are used when spectra are checked by eye.
- Arginine holds on to its proton. A singly charged peptide containing arginine has almost no mobile proton and gives few fragments; with two charges, one proton is free and the spectrum becomes rich[6].
- Trypsin helps. Trypsin cuts after arginine and lysine, so every tryptic peptide except the last ends in a basic residue. That gives peptides of a good size and fragment spectra that are rich in information and easy to interpret[2]; on quadrupole and quadrupole-time-of-flight instruments the y ions dominate[2].
- Proline and aspartate are hot spots. The bond on the N-terminal side of proline breaks easily, because the proline amide nitrogen takes up a proton readily, and a y ion starting at proline often dominates the spectrum[7]. Cleavage after aspartate is also favoured, especially when no mobile proton is available. Cleavage on the other side of either residue is energetically unfavourable[2][6].
- Side chains lose small molecules. Fragments containing serine or threonine lose water (18); those containing arginine, lysine, glutamine or asparagine lose ammonia (17); phosphoserine and phosphothreonine lose phosphoric acid, and methionine loses methanethiol[7].
- The instrument matters. Quadrupole ion traps, for example, discriminate against product ions below about 30% of the precursor m/z, so immonium ions and small sequence ions are generally missing from their fragment spectra[6].
Electron capture and electron transfer dissociation, ECD and ETD, break peptides another way. Adding an electron to a multiply charged peptide cleaves the N–Cα bond instead of the amide bond, giving c and z• ions, and the peptide backbone can be broken while modifications such as glycosylation stay attached[6].
From spectrum to sequence
Reading a sequence directly from a spectrum, de novo, is harder than it looks. Spectra are incomplete, and a gap of 114 between two peaks could be one asparagine or two glycines (57 each) with the middle peak missing. Experts can interpret at least parts of spectra by hand, but in 2004 Steen and Mann judged computer programs “as yet, unreliable” at the task[2].
The way round this, worked out in the early 1990s, is to stop reading and start matching. Jimmy Eng, Ashley McCormack and John Yates described the approach in 1994; it is the basis of the Sequest program[2]. For each spectrum, the program finds every stretch of every protein in the database whose mass matches the precursor within 0.05% or at least 1 mass unit, predicts the fragment ions that stretch would give, and scores how well the prediction overlaps the measured spectrum[9]. The scale of the search shows why the fragments are needed: for one peptide of 1734 Da the database held 384,398 different sequences of that mass, within ±1[9]. A clear gap between the best and second-best scores usually meant the answer was right[9].
Matching has its own limits. It can only find sequences that are in the database, so large-scale work needs organisms with sequenced genomes[2]. Short peptides match poorly; in Eng's tests, peptides of at least 10 residues gave a high-probability identification[9], and Steen and Mann warn that peptides of fewer than seven residues may match more than one protein[2]. Their main sources of false identifications are proteins identified from a single peptide and the use of peptides that are not fully tryptic; to set score thresholds, the same spectra can be searched against a reversed, “nonsense” database[2]. And many spectra are never identified at all: a recent review cites an analysis of a public data repository in which, on average, 75% of spectra remained unidentified, and sums up the problem in one line, “you can only find what you explicitly look for”[10].
What the mass cannot tell apart
Mass spectrometry distinguishes residues by mass, so anything with the same atoms is a problem. Even a single amino acid substitution changes the mass of a peptide, with one exception, leucine and isoleucine[2]. The cases below matter both in proteomics and when checking a synthetic peptide.
| Pair | Mass difference | What it means |
|---|---|---|
| Leu and Ile | 0 (same formula) | Not separated by ordinary fragmentation; Hunt's group wrote “Lxx” for either[7]. Side-chain fragments formed at high collision energy can distinguish them[6]. |
| L- and D-amino acid | 0 (mirror images) | A peptide with one residue racemised has exactly the same mass and composition. Mass alone cannot show it; see D-amino acids. |
| Gly-Gly and Asn | 0 (both C4H6N2O2) | A missing peak in a series can make two glycines look like one asparagine[2]. |
| Gly-Ala and Gln | 0 (both C5H8N2O2) | Listed as a look-alike pair in Hunt's 1986 table[7]. |
| Lys and Gln | 0.036 | Same nominal mass, 128. The instruments of the 1980s and 1990s could not separate them by fragmentation, so lysine was revealed by acetylation, which adds 42 to each free amino group[7][9]. |
Differences calculated from monoisotopic residue masses.
For a synthetic peptide this means that a certificate showing the right mass supports the expected composition and catches errors that change it, such as a missing or extra residue, which shifts the mass by that residue's weight. It does not confirm that every residue has the intended handedness, or that leucine and isoleucine are in the right places. Those need other methods.
Peaks that are not the peptide
Not every peak next to the expected mass is a different peptide. Some are the same peptide with something attached or removed, often during sample handling.
- Adducts. Instead of a proton, an ion may carry a sodium or potassium ion[1]; this puts it 21.98 or 37.96 above the protonated form.
- Oxidation. Methionine is easily oxidised to the sulfoxide (+16) or sulfone (+32) during preparation, and tryptophan, cysteine and histidine can be too[6].
- Deamidation. In deamidation, the side-chain amide of asparagine or glutamine is converted to a carboxyl group, turning them into aspartate or glutamate; the mass rises by about 1 (calculated 0.984). It can happen during preparation, in acidic MALDI matrices or during electrospray[6]; see stability and storage.
- Reactions with reagents. Free cysteine can pick up acrylamide from gels (+71) or 2-mercaptoethanol (+76)[6].
- Other proteins. Keratin from dust, hair or wool, for example when working without gloves, can give stronger peaks than the protein being studied[2].
Some mass changes are real chemistry rather than artefacts. Each disulfide bond removes two hydrogen atoms, 2.016, so a peptide with two disulfides is 4.03 lighter than the same chain with free cysteines. Modifications such as phosphorylation appear in fragment spectra as characteristic losses, for example 98 from phosphoserine[2].
How much, not only what
The height of a peak is a poor guide to quantity. Complete digestion of a protein should ideally give all its peptides in equal amounts, yet their signals can differ by orders of magnitude, because of how easily the protease reaches them, how soluble they are and how efficiently they ionise[2]. Because these effects are reproducible, the same peptide can be compared between samples[2].
For accurate numbers, the trick is a twin that differs only in mass. A known amount of the same peptide made with stable isotopes, such as 13C or 15N, behaves identically in the instrument but appears at a slightly higher mass, and the ratio of the two peaks gives the amount[2]. A shift of at least three mass units is desirable so that the two isotope clusters do not overlap[2]. In SILAC, cells are grown with a heavy amino acid such as 13C-arginine so that the whole proteome of one population is labelled and can be compared with another[2].
In short
- Mass spectrometry measures mass-to-charge ratios. Peptides are brought into the gas phase intact by electrospray, which gives multiply charged ions and couples directly to chromatography, or by MALDI, which gives mostly singly charged ions.
- The Nobel Prize for soft ionisation went to Fenn and Tanaka in 2002, but MALDI as used today, with a small organic acid as matrix, follows the approach of Karas and Hillenkamp rather than Tanaka's metal powder.
- In tandem MS a selected peptide ion is broken, mostly at peptide bonds, into b ions (N-terminal pieces) and y ions (C-terminal pieces). Neighbouring ions differ by one residue mass, and each b and y pair adds up to [M+H]+ plus one proton.
- Fragmentation is uneven: basic residues hold protons, and bonds before proline and after aspartate break more easily.
- Most peptides are identified by matching spectra to sequences in a database, which only finds sequences that are there; de novo reading remains harder.
- Mass alone cannot tell leucine from isoleucine, an L-amino acid from its D form, Gly-Gly from Asn or Gly-Ala from Gln. A correct mass supports the expected composition; it does not show order or handedness.
References
- Fenn JB, Mann M, Meng CK, Wong SF, Whitehouse CM (1989). Electrospray ionization for mass spectrometry of large biomolecules. Science 246(4926):64–71. DOI: 10.1126/science.2675315
- Steen H, Mann M (2004). The abc's (and xyz's) of peptide sequencing. Nature Reviews Molecular Cell Biology 5(9):699–711. DOI: 10.1038/nrm1468
- The Royal Swedish Academy of Sciences (2002). The Nobel Prize in Chemistry 2002. Press release, 9 October 2002. nobelprize.org
- Tanaka K, Waki H, Ido Y, Akita S, Yoshida Y, Yoshida T, Matsuo T (1988). Protein and polymer analyses up to m/z 100 000 by laser ionization time-of-flight mass spectrometry. Rapid Communications in Mass Spectrometry 2(8):151–153. DOI: 10.1002/rcm.1290020802
- Karas M, Hillenkamp F (1988). Laser desorption ionization of proteins with molecular masses exceeding 10,000 daltons. Analytical Chemistry 60(20):2299–2301. DOI: 10.1021/ac00171a028
- Wysocki VH, Resing KA, Zhang Q, Cheng G (2005). Mass spectrometry of peptides and proteins. Methods 35(3):211–222. DOI: 10.1016/j.ymeth.2004.08.013
- Hunt DF, Yates JR 3rd, Shabanowitz J, Winston S, Hauer CR (1986). Protein sequencing by tandem mass spectrometry. Proceedings of the National Academy of Sciences of the USA 83(17):6233–6237. DOI: 10.1073/pnas.83.17.6233
- Roepstorff P, Fohlman J (1984). Proposal for a common nomenclature for sequence ions in mass spectra of peptides. Letter to the editors. Biomedical Mass Spectrometry 11(11):601. DOI: 10.1002/bms.1200111109
- Eng JK, McCormack AL, Yates JR (1994). An approach to correlate tandem mass spectral data of peptides with amino acid sequences in a protein database. Journal of the American Society for Mass Spectrometry 5(11):976–989. DOI: 10.1016/1044-0305(94)80016-2
- Frédérique V, Appeltans S, Askenazi M, Valkenborg D (2026). Defining spectral quality in mass spectrometry-based proteomics: a retrospective review. Mass Spectrometry Reviews 45(5):914–928. DOI: 10.1002/mas.21933