A weighed portion of lyophilised peptide is not a weighed portion of peptide. In a candidate certified reference material for angiotensin II, the trifluoroacetate counter-ion alone accounted for nearly 25% by mass [1]. Two custom synthetic peptides sourced commercially, DR and DRVYI, carried mannitol at 20 and 43% w/w as an undeclared constituent — a highly polar, ultraviolet-transparent adulterant, and generic LC-based quality procedures depend largely on the detection mode and can be blind to impurities of that kind [2]. Across twelve peptide solutions made up by nominal weight, salt formation, weighing errors and other pre-analytical pitfalls produced differences of up to 85% between the indicated and the actual concentration, enough to turn activity data false positive or false negative [3]. Content is the quantity that scales everything downstream, and it is not what an area-percent chromatogram reports.
Why an integral is a content measurement
NMR is quantitative by definition, because the intensity of a resonance line is directly proportional to the number of resonant nuclei giving rise to it [4]. Nothing in that proportionality is analyte-specific: no molar absorptivity, no ionisation efficiency, no detector response factor sits between the spins and the number. That is what makes the technique a primary ratio method of measurement [5]. A single spectrum containing the analyte and a weighed internal standard of established purity gives the analyte's mass fraction from two integrals, two proton counts, two molar masses and two weights, with no reference standard of the analyte itself required [6]. Detection is close to universal, and absolute qNMR with flexible calibration captures analytes that frequently escape other detectors, water and sorbents among them [7].
Internal, external and electronic referencing
An internal standard is weighed into the same tube as the analyte and sees the same field, pulse, probe and temperature, so instrumental terms cancel in the ratio. Reported choices in this space include acesulfame potassium, applied to valine and to the peptide T5 [8], and gemcitabine hydrochloride, applied to the cyclic octapeptide octreotide acetate [6]. The standard needs a resolved signal, established purity, solubility in the chosen solvent and chemical inertness towards the analyte — and the last of those is not automatic. Maleic acid, a routine qNMR standard, was found to influence the quantification of captopril in deuterium oxide because of a chemical reaction between them; the reaction products were isolated by HPLC and structurally elucidated as two pairs of new diastereoisomers [9]. Internal standard, integration, pH effect and uncertainty were all treated as parameters to be examined rather than assumed in that work [9].
External referencing avoids the interaction hazard by never mixing the two materials. PULCON correlates the absolute intensities of two spectra measured under different solution conditions, so the reference spectrum can be recorded on a separate sample; it is easy to implement, applies on all NMR spectrometers, and needs no special hardware or software [10]. The acronym expands to pulse length based concentration determination, and the method works from 1D spectra without an internal reference compound [11]. Because the calibration travels through the pulse length, the 90-degree pulse is not a setup convenience here but a measured quantity, and it belongs to the tube in front of you. PULCON is most accurate when the spectrum of interest contains a resolved resonance; applied to a region of overlapping resonances, an error arises whenever the exact number of contributing lines is unknown, and that error degrades precision [11]. The same work reduces that uncertainty by filtering classes of resonances or extending PULCON to two-dimensional spectra, which pins down the number of contributing lines [11]. For a peptide, overlap is the normal condition, so the caveat is the operative one unless one of those routes is taken. The electronic alternative, ERETIC, substitutes a synthesised reference signal for a chemical one, and has been used to quantify the solution concentration of modified cyclic peptides where authentic standards were not available [12].
Relaxation delay: why a structural acquisition biases the integral
A proton spectrum acquired for structure elucidation typically repeats every second or two. That is adequate for chemical shifts and inadequate for integrals. Longitudinal recovery is exponential in T1, so a repetition time short compared with the longest T1 in the tube leaves signals partially saturated; and because T1 differs between an aromatic proton, a methyl group and the internal standard's own protons, the saturation is unequal. The ratio you then measure is a relaxation ratio wearing the costume of a stoichiometric one. The cost of leaving the method as a whole uncontrolled — not relaxation delay alone — is on record: the absence of a precise protocol covering the measurement procedure together with spectral processing and evaluation allowed quantitative investigations of identical samples in different laboratories to differ by as much as 90% relative to gravimetric reference values [4].
The remedy is procedural rather than clever: measure T1 for the analyte signal you intend to integrate and for the standard signal, then set the repetition time from the larger. Recent work makes that possible even without a deuterated solvent, embedding binomial-like pulses in an inversion-recovery sequence so that T1 can be measured under no-D conditions and the correct repetition time used for high-accuracy measurement [13]. Published methods handle relaxation delay as an optimisation target alongside the number of scans and the pulse angle, not as a default; that was the sequence followed for octreotide acetate before validation for selectivity, stability, linearity, precision and robustness [6]. The same acquisition parameters — relaxation delay, number of scans and windowing — were optimised in a purity assay that additionally used Lorentz/Gauss deconvolution in post-processing to separate impurity and noise contributions from the integrated region of the quantitative proton [14]. The validated protocol behind the published uncertainty figure covers instrument-specific parameters and the data processing and evaluation routines as explicitly as it covers acquisition [4]; digital resolution, phasing and baseline correction are part of the method, not matters of operator taste.
Choosing a signal in a crowded peptide spectrum
qNMR was rarely applied to compounds as large as peptides for one blunt reason: the peaks are too crowded to select an isolated one [8]. Three tactics recur. The first is to hunt for a resonance that already sits clear of the envelope — the same logic that led to selecting a single proton at 4.37 ppm in one purity assay specifically to avoid interfering signals [14]. The second is deconvolution of the integrated region, which separates impurity and noise from the quantitative proton rather than pretending they are absent [14].
The third is to model the spectrum instead of eyeballing it. Quantum-mechanical iterative full spin analysis (HiFSA) fits complete multiplet structure and yields identity and purity information concurrently; the resulting profiles are independent of instrument or software tools and work at any magnetic field strength. The approach was demonstrated on single amino acids and on peptides of increasing size, including the octapeptide angiotensin II and the nonapeptide oxytocin, and pairing it with absolute or 100% qHNMR enables quantification of mixtures and determination of peptide conformer populations [15]. It was HiFSA that unambiguously identified both the mannitol and the peptides in the adulterated custom syntheses and confirmed the quantitative result, on a contaminant that quantitative proton NMR detected even at 60 MHz on a benchtop instrument [2].
Whichever signal you settle on, a related peptide impurity resonating in the same place is counted as analyte. Both qNMR and hydrolysis-based isotope-dilution LC-MS/MS are susceptible to interference from related peptide impurities, which have to be accurately measured and accounted for separately [1].
The exchangeable-proton problem
Backbone amide NH protons, hydroxyls and several side-chain NH groups exchange with solvent, and their exchange rates are structural, not incidental. The rate of exchange of peptide-group NH hydrogens with the hydrogens of aqueous solvent is sensitive to neighbouring side chains through both inductive and steric blocking effects; all twenty naturally occurring amino acids were characterised in dipeptide models, the additivity of nearest-neighbour effects was tested and confirmed in oligo- and polypeptides, and temperature effects were considered [16]. An amide NH integral is therefore a signal whose intensity is set by sequence context and conditions rather than by how much peptide is in the tube.
The standard countermeasure is to delete those protons deliberately. Removing the unwanted exchangeable peaks by proton exchange isolates the quantitative protons of the peptide and is what extended qNMR from small compounds to peptides in the first place [8]; deuterium oxide was used for the same purpose to isolate quantitative signals in the crowded octreotide acetate spectrum [6]. The cost is that the amide region stops being available to integrate, and the exchange has to be driven to completion, because residual NH intensity is intensity you have not budgeted for.
What water suppression does to the integrals near it
When the sample must be run in a protonated or partly protonated aqueous solvent, the solvent resonance has to be suppressed, and suppression deforms the spectrum around it. Limited dynamic range, losses due to relaxation, and proximity to the solvent peak — where quantitation starts failing — are the studied limits of acquisitions made in the presence of a large solvent signal [13]. Binomial-like sequences produced the most robust and reliable results in the majority of scenarios examined, but they carry secondary suppression notches; the notch positions are adjustable, which means they can be moved off your quantified signal, and that they must be checked [13].
Presaturation, jump-and-return, WATERGATE and excitation sculpting all affect the quantitative assessment of analytes containing exchangeable protons [17]. Pulses designed to eliminate the water signal while preserving exchangeable-proton intensities were reported to allow direct quantification of urea — a small molecule, not a peptide — in urine, kidney tissue extract and plasma, a measurement described as difficult by standard experiments [17]. The practical reading for a peptide assay is narrower than the metabolomics case but points the same way: if your quantified protons are non-exchangeable and far from the water line, suppression is an inconvenience; if they are near it, or if you are leaning on exchangeable protons at all, the integral is provisional until you have compared it against an acquisition without suppression.
Uncertainty, and what the number is worth
National and international round-robin testing of a validated single-pulse proton protocol, on three five-component model mixtures, put the maximum combined measurement uncertainty at 1.5% for a 95% confidence interval, and that applies to determinations of molar ratios and of amount fractions alike [4]. That is what a controlled qNMR measurement can deliver; it is not what an uncontrolled one delivers. When solvent suppression is in use the budget grows, and a full uncertainty estimate has to include the allowances that only become relevant under suppression [13].
A peptide example shows the pieces assembled. For a candidate certified reference material for angiotensin II, proton qNMR and isotope-dilution LC-MS/MS amino acid analysis were each corrected for related peptide impurities and combined through a Bayesian statistical approach that used mass-balance results as prior knowledge, with the mass-balance leg resting on a validated fluorine-19 qNMR measurement of the TFA counter-ion. The three orthogonal methods agreed, and the final purity value assignment was 691 ± 9 mg/g at k = 2 [1]. Read the magnitude as well as the uncertainty: just under 70% peptide by mass in a material good enough to be a reference-material candidate.
Against amino acid analysis and area percent
Amino acid analysis hydrolyses the peptide and quantifies the liberated residues, which is a genuine content measurement of the hydrolysate. qNMR instead allows quantitation of intact peptides, thereby eliminating potential bias due to hydrolysis [1]. Hydrolysis conditions must be chosen against residue stability: various protecting agents were tested to keep tryptophan intact in a derivatisation-free fluorescence method that quantifies tyrosine, phenylalanine and tryptophan to better than 95% accuracy and returned 98.6% on a protein reference standard [3]. On time and precision, qNMR of the peptide T5 achieved simpler operation, an analysis time of 1 h against 3 days, and a smaller coefficient of variation, 0.36% against 0.93%, compared with amino-acid-based isotope-labelled mass spectrometry performed via a hydrolysis reaction [8].
Area percent is the weakest of the three claims, because an ultraviolet detector is selective and responds unequally to different compounds, which introduces probable error into a purity assessment built on it [14]. The clearest published magnitude for that error is not a peptide result: for a commercial ginsenoside Rg1 chemical reference substance, a small molecule, qNMR returned 90.34 ± 0.21% where the manufacturer's HPLC-UV figure was greater than 98.0%, and HPLC-UV, HPLC-ELSD and LC-MS all gave significantly higher purity values than qNMR, their accuracy depending on how much sample was loaded and on the properties of the impurities [14].
The peptide evidence is more nuanced and should not be flattened into a slogan. In a multi-laboratory study on the nonapeptide oxytocin comparing an HPLC assay, qNMR and amino acid analysis, the HPLC assay run against the same peptide bulk material as the standard showed the lowest inter-laboratory variability — though that coefficient of variation was calculated without counting the uncertainty attached to the mass-balance purity assignment of the standard itself [18]. The same study characterised proton qNMR as a direct measurement of the peptide against an internal standard, not difficult to perform under common laboratory conditions, and concluded that its simpler operation and shorter analytical time make it worth further exploration as a primary method for peptide reference standard value assignment [18]. For octreotide acetate, F, Student's t and non-parametric tests at the 95% confidence level found no significant difference between the validated qNMR assay and the pharmacopoeial HPLC method [6].
What the measurement licenses
A qNMR content figure gives the mass fraction of the analyte in the material you weighed, traceable through the internal standard's certified purity and your balance, on four conditions: the integrated signal is correctly assigned, fully relaxed, free of co-resonant impurity, and undistorted by solvent suppression. It counts constituents that other detectors miss, including water and sorbents [7], and in at least one documented case a bulk excipient that no ultraviolet trace would have shown [2]. It does not by itself establish that the sequence is correct; that needs the spin-system-level analysis HiFSA supplies alongside the quantitation [15]. It does not resolve a related peptide impurity sitting under your quantified proton, and that interference must be measured and accounted for independently, exactly as it must in the hydrolysis-based route [1]. And it is orthogonal to chromatography rather than a replacement for it [7]. Run qNMR as the content assay, and let area percent describe the impurity distribution it is actually competent to describe.