None of the three routine colorimetric assays measures peptide. Lowry and BCA measure how many electrons a sample donates in a copper-dependent oxidation; Bradford measures how much anionic dye a sample can hold. Each converts that chemistry into a mass through a slope taken from a standard protein, and each responsive residue carries a weight unrelated to its share of the sample’s mass. Run one sample through two of these assays against the same standard and the answers can differ by more than the effect you were trying to measure.
What the copper assays are counting
The Lowry reaction is a copper-catalysed oxidation of peptides and proteins by phosphomolybdic/phosphotungstic acid. Legler and co-workers studied it on defined peptides. From peptides without reducing side chains, two reducing equivalents were transferred under ideal conditions to Mo6+/W6+ for each unit of tetradentate copper complex, with concomitant formation of an imino peptide; tyrosine and tryptophan side chains contributed four additional reducing equivalents [1]. The unit of reaction is the copper chelate, not the milligram of material.
Two further findings bear on synthetic sequences. Oxidation of proline-containing peptides was greatly impaired as judged from colour formation, due to the interference of the imino acid with complex formation [1]. And the clean two-electron stoichiometry is not a general property of the assay: from the time course of blue colour development with tetraalanine the authors inferred an initial oxidation, in under 5 s, to a Cu3+ peptide complex followed by slow absorbance changes, especially above 0.2 mM, so the two-electron stoichiometry has to be considered only as a limiting case for peptide concentrations below 0.02 mM [1]. A proline-rich and a proline-free sequence of equal mass do not sit on one curve.
BCA keeps the copper step and replaces the Folin step. Bicinchoninic acid forms an intense purple complex with cuprous ion in an alkaline environment, and the assay monitors the Cu1+ produced when protein reacts with alkaline Cu2+ in the biuret reaction; against Lowry, the original report demonstrated greater tolerance toward commonly encountered interferences such as nonionic detergents and simple buffer salts, and claimed the method maintains the high sensitivity and low protein-to-protein variation associated with the Lowry technique [2].
Wiechelman, Braun and Fitzpatrick took the reaction apart. Cysteine, tryptophan, tyrosine and the peptide bond are each capable of reducing Cu2+ to Cu+ [3]. Temperature decides how much you collect: electrochemical studies and the magnitude of the colour changes observed at 37 °C indicate that tryptophan, tyrosine and the peptide bond are not completely oxidised at that temperature, and at 60 °C significantly more colour formation is observed for those three groups [3]. A standard curve is transferable only within one temperature.
The finding that matters most for short sequences is the next one. Studies with di-, tri- and tetrapeptides and with proteins indicate that the extent of colour formation is not the sum of the contributions of the individual colour-producing functional groups [3]. A peptide’s expected response is therefore not computable from its composition. The same group showed that compounds bearing functional groups similar to those of cysteine, cystine, tyrosine or tryptophan react with the bicinchoninic acid reagent [3], so residual small molecules from a synthesis or a storage buffer can register as analyte.
Time, temperature and range in the BCA assay
Cortés-Ríos and colleagues explored the kinetics and the analytical response of the assay to free amino acids, peptides containing tryptophan and tyrosine, and proteins. Kinetic profiles were characterised by the absence of plateaus, with behaviours depending on the type of the sample, and contributions to the BCA index elicited by oxidation products generated at the side chains of tryptophan and tyrosine, and by pre-oxidised β-casein, evidenced complex reaction mechanisms [4]. The practical conclusion is narrower than the mechanism: the BCA index was not modulated by the incubation time for responses producing absorbance intensities at 562 nm higher than 0.1, on which basis they proposed the assay can be applied at 15 min rather than the 30 min at 37 °C indicated in manufacturer specifications [4]. Below that absorbance the timing discipline is not optional.
Range is the other axis. Rogatsky reported that the BCA assay’s linearity range and its sensitivity are counterbalancing factors, that the true linear range may be considerably narrower than 20–2000 µg/ml, and that expanding the range up to 2000 µg/ml comes together with unavoidable heavy negative biases at low protein concentrations — biases that only exacerbate with longer incubation time and/or increased sample to working reagent ratio [5]. The remedy proposed there was a two-step incubation and calibration. A wide curve is a decision to be inaccurate at its bottom end, where a dilute peptide stock tends to land.
Bradford is an arginine counter
Coomassie brilliant blue G-250 exists in three forms — cationic, neutral and anionic — and although the anion is not freely present at the dye reagent pH, it is this form that complexes with protein [6]. Two statements from Compton and Jones set the boundary of the method: dye binding requires a macromolecular form with certain reactive functional groups, and interactions are chiefly with arginine rather than primary amino groups, with the other basic residues (His, Lys) and the aromatic residues (Trp, Tyr, Phe) giving slight responses, the binding attributed to Van der Waals forces and hydrophobic interactions [6]. Interference by bases, detergents and other compounds is explained in the same paper through their effects upon the equilibria between the three dye forms [6] — an interferent does not merely add absorbance, it changes how much binding-competent anion exists.
For a synthetic peptide both statements are hostile. The signal is dominated by arginine content, so two peptides of equal mass and unequal arginine content cannot be read off one curve, and an arginine-free sequence is quantified almost entirely on residues the authors characterise as giving slight responses. The requirement for a macromolecular form is a floor that composition cannot buy a way past: a short sequence, however basic, may not present the dye with what it binds to.
Friedenauer and Berlet measured how far the response moves when the solution around it changes, using 16 different proteins and peptides of Mr 1140 to 146,000. Triton X-100 at a final concentration of 0.008% (v/v) increased the sensitivity of the Bradford assay toward all proteins of Mr 5700 or higher by an average 33%, with increases ranging from +11% with myelin basic protein to +128% with aprotinin, and the relative range of absorbance of proteins and the deviations from bovine serum albumin decreased by approximately 25% [7]. Urea at 0.16 M final concentration increased sensitivity by 14%; SDS at 0.004% (w/v) final concentration lowered sensitivity by an average 75% from that of the control assay [7]. A nonionic detergent at eight parts per hundred thousand more than doubles one species’ response and barely moves another’s.
The readout itself is not linear. Zor and Selinger showed that intrinsic non-linearity compromises the sensitivity and accuracy of the standard 590 nm measurement, that under standard assay conditions the ratio of the absorbances 590 nm over 450 nm is strictly linear with protein concentration, and that this ratio increases accuracy and improves sensitivity about 10-fold, permitting quantitation down to 50 ng of bovine serum albumin; the ratiometric procedure also permits assay in the presence of up to a 35-fold weight excess of sodium dodecyl sulfate over bovine serum albumin [8]. Both the linearity and that detergent tolerance are properties of the two-wavelength readout they introduced, not of the single-wavelength 590 nm measurement the paper describes as the standard one.
The standard is the assay
Sapan, Lundblad and Price reviewed Coomassie Blue G-250 dye binding, the Lowry assay, the bicinchoninic acid assay and the biuret assay, comparing several assays on one sample population. They suggest that for characterising a protein preparation the most critical issue in using a chromogenic protein assay is the selection of a standard for its calibration, and that it is crucial that the standard be representative of the sample; where composition cannot be matched, they judge it preferable to use an assay not sensitive to the composition of the protein, such as a micro-Kjeldahl technique, quantitative amino acid analysis or the biuret assay [9]. Whatever method is adopted as the gold standard for a given protein, that method needs to be used routinely for calibration [9].
The penalty for ignoring this has been measured. Rossi and colleagues compared Bradford, Lowry and Non-Interfering assays using bovine serum albumin as standard against quantitative amino acid analysis, described there as the most accurate currently available method for protein quantification. In all three assays the colour yield — optical density per mass of protein — of the test material was markedly different from that of BSA, with a ratio of approximately 4 for the Bradford assay and highly variable between preparations, and approximately 0.7 for the Lowry and Non-Interfering assays, which the authors took as highlighting the need to calibrate the standard against material quantified by amino acid analysis [10]. Read against one BSA curve, a single sample therefore reports several-fold more protein by Bradford than by Lowry, in opposite directions from the truth.
Those ratios belong to their analyte. Generalized modules for membrane antigens are complex biological structures containing lipid and polysaccharide as well as protein, and 4 and 0.7 are not peptide numbers. What transfers is the shape of the failure: the value an instrument prints is the true mass multiplied by the ratio of the analyte’s colour constant to the standard’s, and nothing in the assay reveals that ratio.
Why a BSA curve misreports a short peptide
Biuret-based protein assays are theoretically applicable to peptide measurement, but Kapoor and co-workers reported a high level of interpeptide variation, determined largely by peptide hydrophobicity [11]. Their remedy was to remove the variable rather than correct for it: this variation in peptide reactivity can be significantly reduced by heat-denaturation of peptides at 95 °C for 5 min in the presence of 0.1 M NaOH containing 1% (w/v) SDS, prior to incubation for 30 min at 37 °C in BCA standard working reagent, a modification they present as allowing an accurate, rapid and economical estimation of the peptide concentration within an unknown sample [11]. Treat that as a diagnosis as much as a protocol: if pre-denaturation changes the answer, the untreated assay was reporting something other than mass alone.
Fears, Petrovykh and Clark tried to establish a uniform methodology for measuring peptide concentrations in water across a series of model GG–X–GG, X5 and X10 oligopeptides, assessing UV spectroscopy of peptides having a C-terminal tyrosine, the BCA protein assay, and amino acid analysis. Due to shortcomings or caveats associated with each of the different methods, none were effective at measuring concentrations across the entire range of representative model peptides, and in general reliable measurements were within 30% of the nominal concentration based on the weight of as-received lyophilised peptide [12]. That is the honest error bar on short sequences of known composition.
The structural reason is arithmetic plus non-additivity. Backbone contribution scales with the number of peptide bonds, and a chain of n residues carries n − 1 of them, so backbone reducing sites per unit mass fall as the chain shortens; the side-chain contributions then do not add, as the di-, tri- and tetrapeptide result established [3]. A BSA slope encodes that protein’s residue census and backbone density. Applied to a decapeptide of different aromatic and cysteine content, it is an assumption, not a calibration.
Interferences, and why a matched blank does not fix them
Brown, Jarvis and Hyland catalogued what breaks the BCA assay during purification. Glucose, mercaptoethanol and dithiothreitol elicit a strong absorbance at 562 nm when combined with the BCA working reagent, and that absorbance appears to be identical to the normal response elicited by protein [13]. Ammonium sulfate and certain ampholytes instead diminish the protein-induced colour development and shift the wavelength of the colour response [13]. Both types of interference were eliminated by selectively precipitating protein with deoxycholate and trichloroacetic acid prior to reaction with bicinchoninic acid [13]. A reducing agent in the tube is not an interference you subtract; it is signal indistinguishable from the analyte.
Lipids do the same. Kessler and Fanestil found the bicinchoninic acid technique to produce erroneously high values for protein when common membrane phospholipids were included in the assay: phospholipids produced an absorbance peak similar to that produced by protein, this absorbance was linear with concentration, the slope varied for individual phospholipids, and the combined absorption of phospholipid and protein was not strictly additive [14].
Non-additivity defeats the obvious correction: the colour formed by compounds bearing functional groups similar to those of cysteine, cystine, tyrosine or tryptophan, in the presence of bovine serum albumin, cannot be compensated for by using a reagent blank containing an identical concentration of the interfering compound [3]. The move is removal — buffer exchange or precipitation — not arithmetic.
Where colorimetry sits among the options
Ranked by how much they assume, the three routine ways of putting a number on a peptide stock are not equivalent. Amino acid analysis assumes least: it was described in one head-to-head comparison as the most accurate currently available method for protein quantification [10], and appears among the composition-insensitive alternatives recommended when a matched standard is unavailable [9]. Absorbance at 280 nm assumes a coefficient, but an explicit and calculable one. The absorbance of a protein at 280 nm depends on the content of Trp, Tyr and cystine, and on the basis of 116 measured ε values for 80 proteins the ε(280) of a folded protein in water is best predicted by ε(280) (M−1 cm−1) = (#Trp)(5,500) + (#Tyr)(1,490) + (#cystine)(125); these values are quite reliable for proteins containing Trp residues and less reliable for proteins that do not, and the authors’ own recommendation is that the best approach is to measure rather than predict ε [15].
Colorimetry assumes a coefficient that no additive rule computes from sequence. The colour constant is empirical, sequence-specific, non-additive in its parts [3], temperature-dependent [3], time- and concentration-dependent [1, 4, 5], and movable by residual detergent [7]. That is why it is the third option, not the first.
A direct alternative exists where the sequence carries the right residue. Wiśniewski and Gaugaz evaluated tryptophan fluorescence for total protein determination in whole tissue lysates and for peptide quantification in protein digests, reporting a simple, sensitive and direct method, fully compatible with SDS and other lysis solutes, particularly suitable for determination of peptide content in diluted samples, and non-destructive in that the whole sample can be recovered [16]. Their universal-standard argument rests on a population average: tryptophan content varied only a little between mouse tissues, 1.16 ± 0.08% of total amino acids, and was similar in human cells at 1.19 ± 0.06%, on which basis free tryptophan can be used as a universal standard [16]. A single defined peptide is precisely the case that average does not describe; free tryptophan calibrates it only if the conversion is done from the known sequence rather than from a proteome-wide mean.
What the number licenses
A colorimetric reading supports relative comparison between samples measured on one plate, against one standard, with one reagent lot, at one temperature and one incubation time.
It does not support an absolute mass unless the colour constant of the actual peptide has been fixed against a method insensitive to composition [9]. It does not distinguish peptide from free amino acids [3, 4], reducing excipients or lipid [13, 14]. It does not transfer between assays, since one sample read on two chemistries against the same standard can return numbers differing several-fold [10]. And a concentration quoted without naming the assay, the standard, the temperature and the incubation time is not a measurement anyone else can reproduce.
The workable procedure is to stop treating the assay as absolute. Anchor one lot of one peptide against amino acid analysis, derive the colour constant for that sequence, then use that constant with that assay under fixed conditions for every subsequent measurement of that peptide — the discipline stated as adopting one method as the gold standard for a given protein and using it routinely for calibration [9]. Change the sequence and the constant is void.