Trace Metals in Peptide Work: ICP-MS Elemental Impurity Analysis

A peptide that looks clean by reversed-phase HPLC can still carry enough copper, palladium or iron to change what happens the next time it goes into a tube. Trace metals act catalytically, at concentrations far below anything a UV chromatogram resolves, and they arrive by routes an organic-purity assay is structurally blind to: a palladium-mediated deprotection, a copper-catalysed ligation, a nickel column, an afternoon in a stainless-steel vessel. Inductively coupled plasma mass spectrometry closes that blind spot. The reason it needs a matrix-specific protocol is that a peptide is a carbon-dense organic sample, and carbon is not a spectator in an argon plasma.

Where the metal enters

Palladium is the first suspect in any sequence built with allyl-type orthogonal protection or elaborated by cross-coupling. A pilot study that quantified residual palladium in reaction products after progressively more rigorous purification reported that significant levels can remain in isolated products following chromatographic purification, and that only a subsequent metal-scavenging step reliably reduced them to a low level ^[9]. That work was done on small lead-like screening compounds, not peptides, so the absolute figures do not transfer — but two things do. Silica chromatography alone is not a reliable metal-removal step. And the authors framed the study as filling a gap, noting that quantification of metal remaining in reaction products is rarely reported despite the known and potentially deleterious effects of low-level metal impurities in biological assays ^[9].

Copper enters through CuAAC, and nothing in the standard protocol takes it back out. In the standard bioconjugation protocol, ascorbate is used as reducing agent to maintain the required cuprous oxidation state; because those conditions produce reactive oxygen species, five equivalents of a copper-binding ligand are used with respect to metal, and the ligand both accelerates the reaction and serves as a sacrificial reductant, protecting the biomolecules from oxidation ^[10]. Note what the ligand does and what it does not. It manages copper's redox behaviour during the reaction. It does not remove copper afterwards. The click chemistry guide in this library treats ligand-to-copper ratio as a control on oxidative damage; residual copper in the isolated conjugate is a separate measurement and a separate problem. ICP-MS has also been turned on the reaction itself: a modified on-line ICP-MS/MS platform was used to screen Cu(I) intermediates in CuAAC with and without ligand protection, and the authors report that the unligated Cu(I) intermediates are very active and extremely difficult to observe using other existing techniques ^[11].

Nickel comes off immobilised-metal affinity media. Work developing a hydroxynaphthol blue leaching assay reported that significant amounts of Ni²⁺, up to 20 mM, were co-purified with a target protein — the cytochrome bc₁ complex — when histidine was used to elute from Ni-NTA resin ^[12]. That figure belongs to a protein eluted under competitive histidine elution, not to a synthetic peptide and not to IMAC generally; the paper's own framing is that there are few studies of metal leaching from affinity resin, and that it matters particularly when eluants stronger than imidazole are used ^[12]. The transferable idea is that an eluent chosen for its ability to displace your construct from the metal is, by construction, capable of displacing the metal from the resin.

Iron, chromium and nickel come from process contact surfaces. In biotherapeutic formulations held against 316L stainless steel, those three elements were the ones monitored by ICP-MS because they are the major components of the alloy, and among the factors tested, metal chelators produced the strongest effect in facilitating metal leaching; leachables decreased as fill volume per unit contact surface area increased, reaching a plateau at approximately 3 mL/cm^{2 [13]}. That study was run on protein formulations rather than synthetic peptides, and the chelator finding is the one worth carrying across: adding a chelator to suppress metal-catalysed chemistry can raise the total metal a solution extracts from its container.

Why a trace concentration is a mechanistic problem

Fenton oxidation by Fe(II)/H₂O₂ of several Met-, Tyr- and His-containing model peptides produced evidence for a metal-bound reactive oxygen species selectively oxidising methionine to methionine sulfoxide, which the authors read as indicating a Met-specific oxidant and arguing against the involvement of freely diffusible hydroxyl radicals ^[14]. In the model peptide Nor-Gly-His-Met-NH₂, which carries a conformationally restricted norbornyl methionine analogue, Fenton oxidation gave a 2:1 preference for sulfoxide formation at the C-terminal Met over the N-terminal Nor residue, while incubation with H₂O₂ alone gave 1:1 ^[14]. Read the framing before generalising: that paper presents the hydroxyl-radical picture of Fenton chemistry and the surface-accessibility picture of methionine oxidation as commonly held beliefs which its results complicate, not as settled descriptions it confirms. The working consequence is that oxidation selectivity in a metal-contaminated peptide is not predictable from solvent accessibility alone.

The electron bookkeeping matters for what you will see downstream by LC-MS. Two-electron oxidation of methionine in aerobic Fenton reactions with [Fe(II)(EDTA)]^2- and H₂O₂ was shown to be the consequence of two consecutive one-electron transfer reactions carried out by either free or complexed hydroxyl radicals, followed by reaction of an intermediary sulfur-nitrogen bonded radical cation with O₂; in the model peptide Met-Met the resulting azasulfonium diastereomers could be chemically isolated before hydrolysis to sulfoxide ^[15]. The iron in that system was supplied as its EDTA complex, which is worth noting for anyone who assumes chelation is deactivation.

Sequences that are their own chelators

Some peptides do not merely encounter metal; they collect it. N-terminal Xaa-Zaa-His tripeptides, the ATCUN motifs, bind Cu(II) ions in exchangeable form, and appear in serum albumin and other proteins involved in copper metabolism ^[17]. LC-ICP MS direct competition experiments found that ATCUN peptides and the truncated peptides Aβ4-16 and Aβ11-15 bind Cu(II) with affinity similar to that of human serum albumin ^[17]. The methodological caveat from the same work is the one to carry into your own measurements: the Cu(II)-binding affinity (log K_d) of several ATCUN peptides, determined by different methods and different authors, varies by more than three orders of magnitude, a spread the authors attribute to peptide chemical properties or to methodological differences ^[17]. Binding constants for a metal-chelating peptide are not interchangeable across techniques, and a value lifted from one method does not predict how much metal your sequence will strip out of a buffer.

Getting an organic matrix into the plasma

Sample preparation for ICP-based determination of elemental impurities spans dissolution in aqueous or organic solvents, extraction, wet digestion and combustion methods, and direct solid analysis ^[1]. For a peptide the practical choice is closed-vessel acid digestion, and the parameter that tells you whether it worked is residual carbon content rather than how clear the digest looks. Using a single reaction chamber, digestions of whole milk powder and bovine liver with 2.0 mol L^-1 nitric acid solution plus concentrated hydrogen peroxide at 240 °C gave residual carbon contents of 0.825% and 1.50% and residual acidities of 1.04 and 0.618 mol L^-1 respectively, parameters the authors describe as fully compatible with further measurement by ICP-OES or ICP-MS ^[5]. Those are certified reference materials of animal origin, not peptides, and the figures are specific to that matrix and heating programme; what transfers is the pairing of a dilute-acid digestion with an explicit residual-carbon acceptance criterion, and the demonstration that dilute nitric acid can be sufficient for efficient digestion of organic samples.

Two published validations show what else a digest has to survive. A method developed for a tablet matrix digested the sample in the presence of a large quantity of excipients, titanium dioxide in particular, and measured Cd, Pb, As, Hg, Co, V and Ni in collision cell mode with kinetic energy discrimination ^[4]. A method developed for an active pharmaceutical ingredient used a stabilising matrix to overcome a low spike recovery problem encountered with osmium, which allowed As, Cd, Hg, Pb, V, Cr, Ni, Mo, Cu, Pt, Pd, Ru, Rh, Os and Ir to be determined in a single analysis ^[3]. Both are small-molecule matrices. A peptide digest differs in that the analyte itself contributes sulfur from Cys and Met, and phosphorus if the sequence is phosphorylated.

Polyatomic interference and what the cell can and cannot fix

A digested organic sample delivers carbon, nitrogen and residual acid anions into an argon plasma, and those elements recombine into polyatomic ions that land on analyte masses. Collision and reaction cells attenuate them. In a multi-element food method, four influential cell factors — hexapole bias, quadrupole bias, helium flow and hydrogen flow — were optimised through experimental design methodology specifically to reduce spectroscopic interferences on V, Cr, Fe, Co, Ni, As and Se ^[6]. The same work treated the other class of problem differently: non-spectroscopic interferences, which the authors note may severely disturb the analysis of matrices containing large amounts of non-target elements, were significantly reduced by a limited decrease in nebuliser flow rate rather than by a specific, time-consuming dilution ^[6]. Keep the two categories apart when you troubleshoot. Cell gas addresses spectral overlap. It does nothing about a plasma whose ionisation conditions the matrix has changed.

Carbon is not a neutral background

Carbon changes sensitivity, and it does so selectively. A systematic study across B, Be, P, S, Zn, As, Se, Pd, Cd, Sb, I, Te, Os, Ir, Pt, Au and Hg, varying plasma settings, sample delivery approach and carbon source (glycerol, citric acid, potassium citrate, ammonium carbonate), found that P, As, Se, Sb, Te, I, Au and Hg sensitivities are always higher for carbon-containing solutions, while other elements showed effects that varied with the conditions; the enhancement was attributed to charge transfer reactions involving carbon-containing charged species, with C⁺ alongside CO⁺, CO₂⁺, C₂⁺ and ArC⁺ present in the plasma ^[7]. Two operational consequences follow. A calibration curve prepared in clean dilute acid does not describe the same plasma as a carbon-bearing peptide digest for those elements, so either match carbon between standards and samples or drive residual carbon down until the effect is negligible — which is precisely why the residual-carbon figure from the digestion step is a method parameter and not a footnote. And because the effect is element-specific, a single global correction factor applied across the panel is the wrong shape of fix.

Internal standards, and the matrix effects they can and cannot correct

Internal standardisation here is not a formality. A study of the ICP-MS matrix effect found that the magnitude of signal suppression or enhancement depends in a regular way on mass number, so accurate correction for non-spectral interferences is only possible using an internal standard with mass number close to that of the analyte; a close-mass internal standard also improved precision, while the ionisation energy of the internal standard appeared to be of no or only secondary importance ^[8]. The conclusion drawn was explicit: when a number of elements over a considerable mass range are to be determined, several internal standards have to be used ^[8]. Practice follows. The tablet method above ran ²⁰⁹Bi⁺, ¹¹⁵In⁺ and ⁸⁹Y⁺ together, and reported relative standard deviations of recovery in precision of not more than 1.8% and linear correlation coefficients above 0.9997 ^[4]. A single internal standard bracketing a fifteen-element panel is a design error, not a shortcut.

Detection limit and quantitation limit are different objects

In that validated tablet method the quantitation limits were 0.25, 0.25, 0.75, 1.5, 2.5, 5 and 8 µg L^-1 for Cd, Pb, As, Hg, Co, V and Ni respectively, with recoveries at the limit of quantitation ranging from 90.5% to 106.4% and accuracy recoveries from 91.8% to 103.6% ^[4]. Read the structure of that claim rather than the numbers. The quantitation limit is established by demonstrated recovery and precision at that concentration in that matrix, not by extrapolating some multiple of a blank standard deviation. Osmium in the API method is the sharpest illustration of the difference: the obstacle there was not sensitivity but spike recovery, and it took a stabilising matrix to fix ^[3]. An element can be detectable and still not be quantifiable, and the failure mode is chemical, not electronic.

Those figures are also concentrations in the measured solution. Converting them into a limit on the peptide requires the dilution factor of the digest, and for an organic matrix that factor is not free: it is bounded by how much residual carbon and residual acid the introduction system and plasma will tolerate. A sample that has to be diluted twenty-fold to keep the plasma stable has a twenty-fold worse limit on a mass basis than the solution figure implies. Carry the digestion blank through every vessel, reagent and transfer the sample sees, and report the limit you demonstrated in your own digest rather than the instrument's figure in clean acid.

The ICH classes as a reporting frame

The ICH Q3D guideline is described as changing the control of elemental impurities from the historical heavy metals test to a scientific-based risk assessment and testing by modern analytical instrumentation such as ICP-OES and ICP-MS ^[2]. Two of the studies drawn on here anchor to that framing. One tested Class 1 and Class 2a elements and found that ICP-OES had its own limitations, especially for arsenic and lead at low concentrations, and so developed an ICP-MS method for those two ^[2]. The other validated Cd, Pb, As, Hg, Co, V and Ni as its panel ^[4], while a full compendial panel extends to fifteen elements including Cr, Mo, Cu and the platinum-group metals ^[3]. The numeric limits attached to the classes are exposure-based and belong to product assessment rather than to a bench method; what a methods laboratory takes from the framework is the discipline it enforces — fix the element panel in advance from the actual process sources in your own route, and demonstrate quantitation for each element on that panel in your own matrix. That discipline matters more than it should have to, because the development of a general method accurate for all elemental impurities, and the establishment of an official method, are still open challenges ^[1].

What a total-metal number licenses

Digestion destroys the information you may most want. After acid digestion the answer is total element per unit mass. It does not say whether the copper was held at an N-terminal histidine site, adsorbed to the vessel wall, or free in the buffer, and it does not report oxidation state. ICP-based systems applied to speciation analysis are treated as a separate category for exactly that reason ^[1]. If the question is association rather than burden, put a separation in front of the plasma: the ATCUN competition experiments resolved which species held the copper because the liquid chromatography step carried the complexes intact through to the ICP source ^[17].

The same coupling turns the technique into a quantification method. A precolumn isotope dilution nanoHPLC-ICPMS approach quantified sulfur-containing peptides by adding a ³⁴S-labelled, species-unspecific sulfate spike directly to the chromatographic eluents; interference-free detection of the ³²S and ³⁴S isotopes was achieved by eliminating O₂⁺ ions in a collision cell using xenon gas at 130 µL min^-1, giving a sulfur detection limit of 45 µg L^-1 corresponding to 1–2 pmol of individual peptides, with 103% recovery and 2.1% RSD on a standard peptide solution ^[16]. That method was demonstrated on sulfur-containing peptides in tryptic digests of human serum albumin and a yeast protein under those nanoHPLC conditions, and the picomole figure is the per-peptide equivalent of a sulfur limit rather than a universal peptide detection limit. It is still the clearest statement of the underlying principle: the plasma does not care which molecule the atom arrived in. That indifference is what makes ICP-MS a contamination assay, and — with a spike and a separation in front of it — a way of counting peptide that never touches a chromophore.