Structure Elucidation of Peptide Degradation Products

A stressed sample throws a new peak, and the extracted mass says the species is one oxygen heavier than the parent, or one dalton heavier. That number constrains elemental composition and nothing else. It says an oxygen was added somewhere, or that an amide became a carboxylate somewhere. It does not say which residue, and for two of the commonest peptide degradation routes it does not even narrow the answer to a single chemical species. Everything that follows is the work of turning a constraint into a structure.

Asparagine deamidation shows the problem at its worst. The reaction runs through a succinimide and delivers a mixture — L- and D-succinimidyl, L- and D-aspartyl and L- and D-isoaspartyl forms, with L-isoaspartyl dominating — and spontaneous isomerisation of an existing L-aspartyl residue produces the same product set ^[1]. A single deamidated peak is therefore consistent with an aspartyl or an isoaspartyl product, in either configuration, at any of the asparagines in the sequence.

That shift is also the one most easily manufactured by the instrument. The monoisotopic peak of the deamidated form sits 19.34 mDa below the ¹³C peak of the amidated form. In mouse brain tissue lysate analysed by nanoLC on a quadrupole-Orbitrap, database searches run with a precursor mass tolerance wider than 5 ppm misassigned deamidated peptides; assignments were correct when mass measurement accuracy was within ±5 ppm ^[2]. Wide tolerances turn an isotope peak into a degradant.

And the two products you most want to separate carry no mass difference at all. Isoaspartate is an isomer of aspartate in which the C_β is incorporated into the backbone, lengthening it by one methylene unit ^[3]. No resolving power separates them on precursor mass.

Why "+16" is not an answer

Oxygen addition is chemically promiscuous, and three residues show why a mass alone will not settle it. Methionine gives the sulfoxide, recognisable in low-energy CID by the characteristic loss of methanesulfenic acid (CH₃SOH, 64 Da) from the side chain ^[4]. Histidine gives 2-oxo-histidine: when Zn²⁺ insulin was exposed to Cu²⁺/ascorbate-induced metal-catalysed oxidation in vitro, MS/MS sequencing located 2-oxo-His at both B-chain histidines, with a relative ratio of oxidised His10(B) to His5(B) of 2.8 ± 1.3 ^[5]. Tryptophan is worse, because it yields several products at once — 5-hydroxytryptophan, oxindolylalanine, kynurenine and N-formylkynurenine — and the first two of those are isomers of one another ^[6].

Accurate mass narrows the composition. Only fragmentation buys you the residue, and only a well-chosen fragmentation buys you the chemical form.

Localising the modification, and what the fragments will not survive

Localisation needs a pair of backbone fragments that bracket the candidate residue and that still carry the mass shift when they arrive at the detector. The second half of that sentence is where most localisation attempts fail.

Methionine sulfoxide is the clean example. The 64 Da methanesulfenic acid loss that makes the modification easy to spot is the same reaction that strips it off the fragment you wanted to weigh. For human atrial natriuretic peptide the predominant CID pathway of the molecular ion was that neutral loss. ECD of the same peptide instead produced many sequence-informative products, including cleavages inside the disulfide-bonded ring, and localised the sulfoxide directly without the usual disulfide reduction and alkylation ^[4]. The general lesson: a diagnostic neutral loss can confirm that a modification is present while consuming the fragmentation that would have placed it.

For isoaspartate the fragmentation itself is the assay. ECD of model peptides and their isoaspartyl analogues produced c_n•+58 and z_l−n−57 ions found only in the isoaspartyl spectra, attributed to cleavage of the C_α–C_β backbone bond that splits the isoaspartyl residue between the two fragments; a complementary feature specific to aspartyl residues was neutral loss of the aspartic acid side chain from the charge-reduced species. The same work notes that earlier CAD-based methods depended on comparison with spectra of aspartyl standards ^[3]. The equivalent diagnostic pair appears under electron transfer dissociation, reported as c+57 and z−57, is relatively abundant, and defines both the presence and the position of isoaspartyl residues ^[7]. The two reports label the c-type ion differently, so match the convention your search software uses before concluding that a diagnostic peak is absent.

Instrument choice then sets what you can conclude in the negative. On the lower-resolution ion trap used for the ETD work, the aspartic acid diagnostic peak was difficult to detect because of interference from arginine side-chain fragment ions; aspartyl residues were still evident in the backbone cleavages and could be inferred from the absence of the isoaspartyl diagnostic ions ^[7]. Inferring aspartate from absence is weaker evidence than detecting it, and it is only as good as your fragment coverage at that position.

Applied bottom-up, the signature c_n+57 and z_l−n−57 ions located isoaspartate formation in an immunoglobulin gamma (IgG1) in a single chromatographic run, and that report treats the signature ions as more conclusive than the two softer lines of evidence it compares them against: relative peak intensity, and the decrease in hydrophobicity of the isoaspartate-containing peptide on elution ^[8].

One caution about the ion series you lean on. Protonated b ions rearrange. Low-energy CID of the b₅ fragments of five pentapeptide amides built from the same five residues in different orders gave nearly the same dissociation patterns, and protonated cyclo-(YAGFL) gave the same fragments at nearly identical abundances under similar experimental conditions — evidence that the linear b₅ ion cyclises, reopens at various amide bonds and fragments as a mixture of linear structures. For the b₅ ion of FLYAG-NH₂ this was documented as complete loss of sequence information ^[9]. A modification site resting on one b ion is a hypothesis.

Chromatographic behaviour as corroboration

Isomeric degradants that MS cannot separate often do separate on a column, which is useful and routinely over-read.

The tryptic heptapeptide spanning residues 128–134 of human growth hormone was synthesised together with the forms in which L-Asp was replaced by D-Asp, L-isoAsp, D-isoAsp or the L-cyclic imide, and the set was used as standards. On a C18 column with a shallow acetonitrile gradient in trifluoroacetic acid/water the elution order was D-isoAsp, L-isoAsp, L-Asp, D-Asp, L-cyclic imide. Capillary electrophoresis in uncoated capillaries with aspartic acid/cyclodextrin buffers at low pH gave a different order: D-Asp, L-Asp, L-isoAsp, D-isoAsp, L-cyclic imide. When human growth hormone was degraded by incubation at high temperature, only the L-forms of aspartic and isoaspartic acid were found in that fragment ^[10].

How transferable is such an order? For the peptide SLNGEWR (residues 54–60 of E. coli β-galactosidase), all three components of the deamidation mixture separated on 300 Å C18 sorbent with trifluoroacetic acid as ion-pairing modifier, and the βAsp/Asn/Asp elution order was confirmed by MALDI MS and MS/MS. Across a large pool of peptide retention data that order is normally observed, especially where the −Asn-Gly− sequence is surrounded by hydrophobic residues. But moving to 100 Å pore-size sorbent, or switching the modifier to formic acid, increased βAsp retention relative to the Asn/Asp pair, so the order can sometimes be different ^[11].

Retention order is therefore corroborating evidence tied to a specific stationary phase and modifier. It becomes identifying only when an authentic standard has been run on your own system under your own conditions.

Orthogonal confirmation

An enzyme can supply the specificity that mass cannot. Protein isoaspartate methyltransferase (PIMT, EC 2.1.1.77) selectively methylates isoaspartyl residues; hydrolysing the resulting methyl ester in ¹⁸O-water regenerates isoAsp with ¹⁶O replaced by ¹⁸O, converting an invisible isomerisation into a 2 Da shift that ordinary CID and standard data-analysis algorithms can recognise. Detection and site assignment were demonstrated on isoAsp peptides from a monoclonal antibody and on β-delta sleep-inducing peptide ^[12]. The label is placed by a reaction that discriminates the isomer, so the evidence does not depend on interpreting a fragmentation pathway.

Amino acid analysis answers a narrower question than most people expect. Peptides are hydrolysed, preferably in gas phase, with 6 M HCl at 110 °C for 20–24 h and the liberated amino acids resolved by ion-exchange chromatography with post-column ninhydrin derivatisation, giving quantitation typically better than 5%. Depending on hydrolysis conditions tryptophan is destroyed, cysteine likewise unless derivatised, and glutamine and asparagine are deamidated to glutamic acid and aspartic acid ^[13]. Read that list against your candidate structure: acid hydrolysis converts asparagine to aspartate, so amino acid analysis cannot confirm or exclude a deamidation degradant, and it destroys the tryptophan whose oxidation state you were asking about. What it does confirm well is gross composition — a residue that has been lost, or one that has been gained.

Independent synthesis of the proposed structure is the definitive test, and both of the isomer problems above were settled that way. The five aspartyl forms of the growth hormone heptapeptide existed as authentic standards before any elution order could be assigned to them ^[10]. For tryptophan oxidation, panels of five synthetic peptides per sequence — carrying Trp, 5-HTP, Kyn, NFK or Oia at the same position — were required to show that the modified position is identifiable from the y-series, that NFK- and Kyn-containing peptides give an intense signal at m/z 174.1 usable in a precursor ion scan, and that the isomeric 5-HTP and Oia residues can be told apart by the intensity ratio of signals at m/z 130.1 and 146.1 in both ESI- and MALDI-MS/MS, and by v- and w-ions under high collision energy CID on a MALDI-TOF/TOF ^[6]. Diagnostic ions like these are only diagnostic because someone made both isomers and looked.

Where the degradant is present at low level, isolate it first. In a stress study of the synthetic peptide ganirelix, run under acidic and alkaline hydrolytic, photolytic, thermal and oxidative conditions, four major degradants were separated with resolution >1.5 on a C18 column, then fractionated and desalted by flash chromatography before LC-MS-QTOF analysis. Two of the four turned out to be diastereomers of the parent, one a deamination product and one the result of insertion of an additional amino acid residue into the sequence ^[14]. Two of those four have the same mass as the parent peptide; a mass-shift-driven workflow would never have found them, and the isolation step is what made the structural assignment possible.

Deciding whether the degradant is yours or your method's

Before assigning a structure, establish that the species existed before you opened the tube. Peptides containing −Asn-Gly− showed roughly 70–80% deamidation after a standard overnight (~12 h) tryptic digestion at 37 °C, and the synthetic SLNGEWR parent had a half-life of about 8 h under those digestion conditions ^[11]. A peptide-mapping protocol can generate the very degradant it is measuring.

Two controls address this directly. Digesting in ¹⁸O-water distinguishes deamidation occurring during digestion, which gives a +3 Da shift, from pre-existing deamidation, which gives +1 Da; monitoring three rapidly deamidating peptides released from calmodulin, ribonuclease A and lysozyme showed that a fast (~4 h) trypsin digestion generated no additional detectable peptide deamidation ^[15]. Separately, the succinimide intermediate is unstable, and that instability is what makes its analysis difficult. At low pH it is stable and can be characterised by accurate mass measurement and tandem MS to confirm its identity and localise its modification site; a hydrophobic interaction chromatography method built on those assignments then separates and quantifies unmodified protein, deamidated products and succinimide ^[16].

When a degradant must be identified rather than tracked

Tracking a peak by relative area is enough when the peak is reproducible and the question is how much of it there is. Identification becomes necessary in four situations, each of which appears above.

First, when the degradant is isobaric with the parent or with another degradant, so peak area cannot be attributed to a species without an independent assignment — the aspartyl/isoaspartyl pair, and the ganirelix diastereomers ^[14].

Second, when the assignment determines which chemistry you are trying to suppress. A +16 Da peak that is methionine sulfoxide and a +16 Da peak that is 2-oxo-histidine point at different reactions with different susceptibilities, and the site preference within a single residue type can itself be pronounced ^[5].

Third, when the peak may be an artefact of the analytical procedure. Deciding that requires knowing what the species is, because the control experiments — isotope labelling of the digest, or a pH-shifted digestion — are chosen against a specific chemical hypothesis ^{[11, 15, 16]}.

Fourth, when the degradant is to serve as a marker in a stability-indicating method. A peak used as a reference or system-suitability marker must have a fixed identity before the method that reports it can mean anything; the ganirelix method quantifies related substances down to 0.04% of the test concentration, and that figure is only interpretable because the four degradants behind it were structurally assigned ^[14].

Outside those cases, a well-resolved, reproducible unknown peak can reasonably be tracked and left unidentified while more urgent questions are answered. Inside them, the mass shift is where the work starts, not where it ends.