Can peptide mapping detect a D-amino acid or a scrambled sequence that co-elutes with the main peak?

Accepted answer

Your reasoning is right on both counts and the conclusions differ. Mapping can catch a transposition, cannot catch an epimer, and there is a real orthogonal method for the epimer case that a commercial lab will run but that has a floor on its sensitivity you need to know about.

Transpositions: MS/MS does work

Collision-induced dissociation of a protonated peptide cleaves the amide backbone and produces b ions (charge retained on the N-terminal piece) and y ions (charge on the C-terminal piece). Each consecutive pair in a series differs by exactly one residue mass, so the series reads the sequence out. Swap two non-identical residues and the ladder shifts at the swap and back again.

Take the semaglutide Glu-C fragment 16-21, VSSYLE, mass 696.33. Its y-ion series is y1 = Glu, y2 = Leu-Glu, y3 = Tyr-Leu-Glu, and so on. If Leu20 and Tyr19 were transposed, y2 and y3 change: y2 becomes 311.13 (Tyr-Glu, using Tyr residue mass 163.06) rather than 261.16 (Leu-Glu, using Leu 113.08). A 50 Da difference in one ladder rung, unambiguous.

Where it breaks in practice, and these are real:

• Isobaric residues. Leu and Ile have identical residue masses (113.0841) and no CID ladder can distinguish them, ever. Gln and Lys differ by 0.0364 Da and need high-resolution fragments. Semaglutide has Leu20, Leu32 and Ile29; tirzepatide has three Ile and two Leu. A Leu-to-Ile transposition among those is undetectable by MS/MS.
• Incomplete ladders. Real spectra do not fragment evenly. Proline suppresses cleavage on its N-terminal side, which is why the tirzepatide C-terminal region with its PPP motif fragments badly. A gap in the ladder is a region where a transposition would be missed.
• Large precursors. A 3841 Da tryptic fragment gives a crowded, poorly populated CID spectrum. Use a second enzyme rather than fragmenting a monster.

So: MS/MS catches transpositions well within the regions it covers, subject to the isobaric-residue floor. That is a real capability that intact mass does not have.

Epimers: mapping is blind, and here is what is not

You are correct. A D-residue changes no bond, no formula, no fragment mass. Retention time can shift slightly — diastereomeric peptides sometimes separate on RP-HPLC because the altered backbone geometry changes the exposed hydrophobic surface — but the shift is unpredictable, frequently under the resolution of the method, and sometimes zero. Relying on it is not a method.

The real method is chiral amino acid analysis after total hydrolysis, and the classical version is Marfey's method:

1. Hydrolyse the peptide completely, typically 6 M HCl at 110 degrees for 24 hours under vacuum or nitrogen, giving free amino acids.
2. Derivatise with Marfey's reagent, 1-fluoro-2,4-dinitrophenyl-5-L-alaninamide (FDAA), which is itself chiral. D and L amino acids become diastereomers of the derivative rather than enantiomers.
3. Separate on ordinary reversed-phase HPLC with UV detection at 340 nm. Diastereomers separate on an achiral column, which is the whole trick.
4. Quantify D as a percentage of D plus L for each residue.

The critical limitation: acid hydrolysis itself racemises amino acids. Twenty-four hours in 6 M HCl at 110 degrees typically produces 1 to 4% D from a pure L peptide, varying by residue — Asp and Ser are among the worst, Ile and Val among the best. That artefactual racemisation is the detection floor. You cannot credibly report 0.5% D-Asp from a standard hydrolysate, because the hydrolysis makes more than that.

Mitigations that competent labs use: shorter hydrolysis at higher temperature, deuterated acid (DCl in D2O, where artefactual racemisation incorporates deuterium and can be distinguished by mass), gas-phase hydrolysis, or enzymatic hydrolysis, and always a parallel hydrolysis of an authentic all-L reference standard to establish the blank. A result reported without that parallel control is uninterpretable.

Alternatives: chiral GC-MS of N-trifluoroacetyl amino acid esters on a Chirasil-Val column, or LC on a teicoplanin or crown-ether chiral stationary phase. Same hydrolysis problem, different separation.

The special case worth its own paragraph: iso-aspartate

Iso-Asp is the other mass-silent isomer and it has a specific detection route that does work. The aspartimide intermediate that forms during Fmoc synthesis reopens either at the alpha carbonyl (giving normal Asp) or at the beta (giving a backbone linked through the side chain, one extra CH2 in the main chain). Same mass, same composition.

Electron-transfer dissociation produces c and z ions rather than b and y, and at an iso-Asp site the c and z ions shift by 57.02 Da in opposite directions relative to the normal-Asp peptide — a diagnostic pair that CID does not produce. There is also an enzymatic assay: protein L-isoaspartyl methyltransferase specifically methylates iso-Asp using S-adenosylmethionine, and you quantify the S-adenosylhomocysteine released.

Semaglutide has one Asp (position 15) and tirzepatide has two (9 and 15), so this is not academic for either.

Will a commercial lab do it?

Chiral amino acid analysis, yes, at contract-research pricing, and you must ask for it by name and supply several hundred micrograms — one or two vials. ETD for iso-Asp needs an ETD-capable instrument and an analyst who has done it, which narrows the field. Neither is on the standard menu at the independent services; Janoshik, Medutest and PeptideMeter run purity, identity, content and endotoxin, and mapping where offered is CID-based.

Which means, realistically: the stereochemical question is open on essentially every vial in this market, and the correct posture is to know that rather than to imagine a purity number closed it.

edited 20 Apr 2026 by Dr_Colm_Fitzhenry — updated for the 2026 guidance change