Extinction Coefficients and A280 in Peptide Quantification

An absorbance reading at 280 nm is not a measurement of peptide. It is a measurement of indole, phenol and disulfide — the side chains of tryptophan and tyrosine, and the cystine bridge — converted into a peptide concentration by an assumed proportionality. Every failure mode follows from that substitution. The chromophores are a small and unevenly distributed subset of the molecule; they respond to what surrounds them; and anything else in the cuvette that absorbs or scatters near 280 nm is counted as peptide.

Beer-Lambert, and which of its terms you actually know

The concentration of a purified protein in solution is most conveniently and accurately measured by absorbance spectroscopy, absorbance being a linear function of molar concentration through A = ε × l × c, where ε is the molar absorption coefficient and l the cell path length ^[1]. The absorbance is the only term you measure. The other two you assert.

Path length stops being trivial the moment you leave a 10 mm cuvette. Short path lengths are used with high-concentration targets to keep absorbance values inside the optimal range, and NIST developed Standard Reference Material 2082 as a path length standard for the newer generation of microvolume spectrophotometers and short-path cuvettes, used to calculate path length values accurately and so improve the accuracy of subsequent measurements on those instruments ^[2]. Report a concentration from a microvolume drop without having characterised that instrument’s effective path length and the number carries an unquantified multiplicative error.

Where the extinction coefficient comes from

For a peptide available in small amounts, measuring ε directly is often impractical, which is why sequence-based calculation exists. Gill and von Hippel presented a method for calculating molar extinction coefficients for proteins at 280 nm from amino acid composition alone, accurate to ±5% in most cases, calibrated against 18 normal globular proteins whose coefficients were accurately known ^[3].

Pace and co-workers refined the arithmetic on a larger base. The absorbance of a protein at 280 nm depends on the content of Trp, Tyr and cystine, and from 116 measured ε values for 80 proteins they concluded that the ε at 280 nm of a folded protein in water is best predicted by ε(280) (M^-1 cm^-1) = (#Trp)(5,500) + (#Tyr)(1,490) + (#cystine)(125) ^[4]. Note the referent of the third term: 125 M^-1 cm^-1 per cystine, that is per disulfide bond, not per cysteine residue. Reduce the peptide and that term no longer describes your spectrum.

An independent regression reached similar per-residue values by another route. Mach, Middaugh and Lewis used matrix linear regression, with a mapping of average absolute deviations between experimental and calculated values, to obtain coefficients at 280 nm of 5,540 M^-1 cm^-1 for tryptophan and 1,480 M^-1 cm^-1 for tyrosine residues in an average protein, defined by experimentally determined extinction coefficients for more than 30 proteins; these give a significant improvement over the commonly used values from solutions of model compounds in guanidine-HCl, and the authors frame the result cautiously, as a method that may offer acceptable accuracy in the initial estimation of molar absorptivities of globular proteins ^[5].

Two features of those calibration sets matter to anyone weighing out a 10-mer. All were built from whole proteins, so a short synthetic peptide sits outside the population the coefficients were fitted to. And Pace and colleagues did not present the predicted value as the endpoint: their ε(280) values are described as quite reliable for proteins containing Trp residues and less reliable for those that do not, and their recommendation is to measure rather than predict ε, the Edelhoch method being the best route to measuring it ^[4].

Folding, denaturants and the neighbouring residues

A calculated coefficient describes a chromophore in a defined environment, and a folded peptide does not provide that environment. Pace and colleagues measured chromophore ε values in water, propanol, 6 M guanidine hydrochloride and 8 M urea and compared them with the averages the same chromophores show inside proteins: for Trp, the average ε values for the proteins are less than those measured in any of those solvents, while for Tyr the average protein values are intermediate between those in 6 M GdnHCl and those in propanol ^[4]. This is why ε is measured in denaturant rather than buffer — the denatured state is the one in which the chromophore contributions are defined.

The size of the folding term has been examined directly. Hilario and co-workers calculated extinction coefficients from the measured absorbance of model compounds of the amino acid chromophores and compared these unfolded-state values with experimental concentrations determined by enzymatic digestion; across a range of proteins the experimentally determined coefficient for the native proteins was consistently 1.05 times the calculated value for the unfolded proteins, with good accuracy and precision under well-controlled conditions, and the differences between predicted and experimental coefficients scattered randomly, indicating no systematic bias among the proteins measured ^[6]. A five percent folding term is small next to pipetting error, but it moves in one direction rather than averaging out.

The local environment also moves the band positions, which is what makes the residue-by-residue sum an approximation rather than an identity. Deconvoluting near-ultraviolet second-derivative spectra into component tryptophan, tyrosine and phenylalanine contributions across 16 proteins, Mach and Middaugh found the maximum spectral shift between solvent-exposed model compounds and side chains entirely buried in an apolar protein core to be approximately 5 nm for tyrosine, 4 nm for tryptophan and 2 nm for phenylalanine; for all three, no consistent correlation between absolute band position and average solvent accessibility was observed, which suggests a significant influence of other local effects, electrostatics among them ^[7]. Burial alone therefore does not predict how far a given residue’s band shifts from its model-compound position.

When the sequence has no chromophore at all

The 280 nm method inherits a hard boundary from its own mechanism. Absorbance there is useful because the molar absorptivity can be predicted directly from sequence, but the method is only applicable to proteins that contain tryptophan or tyrosine residues; if a protein contains neither, A₂₈₀ cannot be used to determine its concentration in solution ^[8]. A great many synthetic peptides fall on the wrong side of that line, and no dilution, longer path length or more sensitive detector recovers a signal from a chromophore that is not there.

The partial cases are more dangerous than the empty ones, because they still return a number. Routinely used assays for protein quantification, including the Bradford assay and ultraviolet absorbance at 280 nm, usually seriously misestimate the concentrations of intrinsically disordered proteins because of their distinct and variable amino acid composition; comparing four globular and nine disordered proteins against elemental analysis as the gold standard, Contreras-Martos and colleagues found the ninhydrin assay and a commercial fluorescent protein assay to give reliable data on disordered-protein quantity, while noting that because such proteins can show extreme variation in composition and physical features not necessarily covered by their examples, even those techniques should only be used following standardisation ^[9]. They traced the consequences through circular dichroism spectrum deconvolution and receptor-ligand affinity determination, both of which take the concentration as an input.

Small disulfide-rich peptides sit in the same territory for a different reason. Comparing gravimetric, ultraviolet and NMR-based quantification across a representative set of cyclic disulfide-rich peptides, a class of roughly 14 to 29 residues, Conibear, Daly and Craik conclude that gravimetric and UV absorbance methods should be used with caution for small peptides and that all methods should be carefully validated; for routine work they recommend comparing the analytical reversed-phase HPLC trace, or absorbance at 214 nm, with a standard peptide solution quantified by amino acid analysis ^[10].

205 and 214 nm: counting the backbone instead of the side chains

Every peptide bond absorbs in the far ultraviolet, so a wavelength chosen there measures something every peptide has. Anthis and Clore validated a method for predicting molar absorptivity at 205 nm directly from sequence, fitting a backbone contribution of 2,780 ± 168 M^-1 cm^-1 from data on eight different polypeptides, with side-chain increments on top ^[8]. The backbone dominates but does not monopolise the signal: across their test cases its contribution ranged from 59.4% to 80.8%. With the sequence-specific treatment they report errors in concentration of at most 8.2%, and under 4% for all but two cases; using instead a non-sequence-specific estimate of ε₂₀₅ = 31 mL mg^-1 cm^-1, the errors were in general much larger, up to 19.0% ^[8]. The wavelength is workable for aromatic-free sequences; what it does not tolerate is a generic per-mass coefficient.

At 214 nm the same logic has been calibrated for chromatographic work. Kuipers and Gruppen measured molar extinction coefficients of the 20 amino acids and of the peptide bond at 214 nm in the presence of acetonitrile and formic acid, reporting 923 M^-1 cm^-1 for the peptide bond; tryptophan is approximately 30 times higher, while phenylalanine, tyrosine and histidine are approximately six times higher ^[11]. Two residues behave counter-intuitively. Proline as an individual amino acid has a negligible molar extinction coefficient, but when present in the peptide chain, except at the N terminus, it absorbs approximately three times more than a peptide bond; methionine is similar to the peptide bond, and all other amino acids are much lower ^[11]. Validated against several proteins and peptides, most measured and calculated coefficients were in good agreement, which is what permits quantitative comparison of peptides eluting from reversed-phase HPLC once identified by mass spectrometry ^[11].

The cost of moving down the spectrum is that the matrix starts absorbing with you. Anthis and Clore note that many solutes will exhibit some absorbance at 205 nm, an effect even more pronounced at lower wavelengths, and the scale is not subtle: the undiluted buffers used for two of their samples had A₂₀₅ values of 23.74 and 27.45 ^[8]. That the 214 nm coefficients were determined specifically in the presence of acetonitrile and formic acid ^[11] makes the same point constructively — a far-ultraviolet coefficient belongs to a solvent system. At 280 nm a buffer blank is a formality; at 205 nm the blank is half the measurement, and reducing agents, organic acids and any aromatic additive have to be accounted for before the number means anything.

Scattering, and why the baseline above 320 nm is not empty

Aggregates add no absorbance at 280 nm; they redirect light out of the beam, and the detector cannot distinguish that loss from absorption. The wavelength dependence is the diagnostic. Under the Rayleigh approximation, light scattering of spherical solutes is proportional to λ^-4, so apparent absorbance rises steeply toward shorter wavelengths and persists into the region where peptide chromophores are silent ^[12]. Porterfield and Zlotnick used exactly this, correcting nucleoprotein spectra with a two-point approximation based on the apparent absorbance at 340 nm and 360 nm, wavelengths at which there is no significant absorbance by either protein or nucleic acid, so that what is measured there is essentially scattering ^[12]. The same reasoning underlies the habit of inspecting the 320 to 350 nm region before trusting an A₂₈₀: absorbance that has not returned to baseline there says part of the 280 nm signal is particulate.

Extrapolating from two points assumes one scattering regime, which larger particles do not obey. Pilkington and colleagues describe a fitting routine that corrects full protein ultraviolet spectra for both Rayleigh and Mie scattering, and on samples displaying various degrees of natural and artificially induced scattering they show it can aid in the determination of protein concentrations and can also be employed in the spectral analysis of structural changes accompanied by alterations in scatter intensity ^[13]. The corollary is that the scattering term is itself information. Ultraviolet absorption spectroscopy can give an initial insight into the concentration and aggregation state of protein samples ^[14]; a dilution series whose apparent ε drifts with concentration is reporting association, not chemistry.

Nucleic acid and the other things that absorb near 280 nm

Nucleobases absorb far more strongly than peptide chromophores and their bands overlap yours. Stacked purines and pyrimidines absorb with a maximum at 260 nm and a 260/280 ratio of about 2.0, whereas protein absorbance peaks near 280 nm with a characteristic shoulder at 290 nm and a 260/280 ratio of about 0.6 ^[12]. A ratio drifting up from 0.6 in a peptide preparation is evidence of a co-purifying absorber — but light scattering also increases the 260/280 ratio for large complexes ^[12], so a raised ratio does not by itself identify nucleic acid.

Where the contaminant has to be identified rather than merely detected, derivative methods separate the components. Using near-ultraviolet second-derivative spectra in a matrix least-squares multicomponent analysis, Mach, Middaugh and Lewis detected tryptophan, tyrosine, phenylalanine and phenol in DNA samples of absorbance 0.1 at absorbance levels below 0.002, with proteins detectable below 0.003, and determined the contents of protein-DNA mixtures with less than 1% error and DNA-RNA mixtures with an error of approximately 2% ^[15]. Phenol in that list is worth noting: residual extraction solvents and aromatic scavengers absorb in the same window as your tyrosines.

Relating the ultraviolet number to a mass-balance number

A UV concentration and an amino acid analysis concentration answer different questions, and the calibration runs one way. The molar absorption coefficient of a protein is itself usually based on concentrations measured by dry weight, nitrogen, or amino acid analysis ^[4] — the spectroscopic number is downstream of a mass-balance number, never independent of one. Validating a UV method for a peptide therefore ends at a destructive assay.

Hydrolysis-based cross-checks exist in more than one detection format. Tchipilov, Meyer and Weller hydrolysed protein samples to amino acids and quantified aromatic residues by quantitative ¹H NMR against aromatic internal standards, finding that a pH of 12 minimised signal overlap among the four aromatic amino acids and reporting recovery of about 97% for histidine, phenylalanine and tyrosine from an amino acid standard at approximately 1 mM; acidic hydrolysis of a certified reference material of bovine serum albumin then gave recoveries of 98% ± 2% for Phe and 88% ± 4% for Tyr, at 16 g/L or 250 µM ^[16]. The gap between those last two figures is the instructive part: the same hydrolysis that liberates one aromatic residue almost quantitatively recovers another less completely, so a mass-balance number carries residue-specific loss of its own.

The practical hierarchy follows. Anchor one lot against a destructive method — amino acid analysis, elemental analysis, or hydrolysis with quantitative NMR — use that lot to fix the empirical ε, and use ultraviolet absorbance thereafter as the fast, non-destructive transfer standard. Where a peptide’s chromophore content is too low for 280 nm to be dependable, the route recommended by Conibear and colleagues is the same idea in chromatographic form: compare the analytical reversed-phase HPLC trace or A₂₁₄ against a standard peptide solution that has itself been quantified by amino acid analysis ^[10]. Grimsley and Pace likewise set out protocols for calculating ε for a folded or unfolded protein from the average values of the contributing chromophores, with a more sensitive method for proteins containing few if any tryptophan or tyrosine residues ^[1].

What the reading licenses

An A₂₈₀ taken above a clean baseline, in a buffer of known absorbance, with the 320 to 350 nm region flat, licenses a concentration whose uncertainty is at best that of the coefficient you used — on the order of ±5% for a Trp-containing sequence inside the calibration population ^[3], less reliable for a Trp-free one ^[4], plus whatever the path length contributes. It does not license a statement about purity: an aromatic-bearing deletion sequence, an oxidised variant and the intact peptide all absorb, and the spectrum sums them. It does not license identity or disulfide connectivity, although the cystine term does mean a reduced and an oxidised sample of the same material do not share a coefficient ^[4]. And it does not license a mass. Converting the molar figure into the milligrams that left the balance requires the net peptide content of the solid, which no wavelength in the ultraviolet reports.