Protein extinction coefficient and A280 concentration

Put 1 mg/mL of hen egg-white lysozyme in a cuvette and it absorbs about 2.65 at 280 nm.^[1] The same mass of human ubiquitin, by its sequence, should absorb about 0.17. Both samples hold the same amount of protein, yet one reads fifteen times higher, because lysozyme carries six tryptophans and ubiquitin has none. Any instrument setting that treats an absorbance of 1 as 1 mg/mL gets at least one of them badly wrong.

The number that resolves this is the extinction coefficient: how strongly one protein absorbs at 280 nm per unit of concentration. For most proteins it is predicted from the sequence by counting tryptophans, tyrosines and disulfide bonds, and the extinction coefficient calculator does that count and the concentration arithmetic from a pasted sequence and an A280 reading. The prediction is accurate to a few percent for a typical protein with tryptophan, and it can miss by more than 15% for one without.^[1]

The Beer-Lambert law

Absorbance is proportional to how many absorbing molecules the light passes through, which is the concentration multiplied by the distance traveled. The constant that links them is the molar extinction coefficient $\varepsilon$:

$$A_{280} = \varepsilon \, c \, l$$

Here $A_{280}$ is the absorbance at 280 nm, which has no units, $c$ is the molar concentration and $l$ is the path length in centimeters. For the equation to balance, $\varepsilon$ must carry units of $\text{M}^{-1}\,\text{cm}^{-1}$, the same as $\text{L}\,\text{mol}^{-1}\,\text{cm}^{-1}$. It is the absorbance a 1 M solution would give in a 1 cm cuvette, extrapolated from the dilute solutions actually measured.

Pace and colleagues call the same quantity the molar absorption coefficient, and others call it the molar absorptivity; the three names are interchangeable.^[1][5] Rearranged for concentration, with a dilution factor $d$ to get back to the undiluted sample, it becomes the equation used on every A280 reading:

$$c = \frac{A_{280}}{\varepsilon \, l} \times d$$

Absorbances add. Every molecule in the cuvette that absorbs at 280 nm contributes to the reading, and the equation assumes all of it belongs to the protein. Most of the practical problems with A280 come from that assumption.

Why only tryptophan, tyrosine and cystine count

At 280 nm, protein absorbance comes from the aromatic side chains of tryptophan and tyrosine, plus a small contribution from disulfide bonds.^[5] Phenylalanine absorbs at shorter wavelengths, and a free cysteine thiol does not absorb appreciably above 260 nm, while cystine, the disulfide-bonded pair, does.^[4] The peptide backbone absorbs far more strongly, but at much shorter wavelengths, with a maximum near 190 nm.^[5]

The three chromophores are not equal. On the values Pace's group recommends, one tryptophan absorbs as much as 3.7 tyrosines or 44 disulfide bonds.^[1] That makes tryptophan the residue that decides how a protein reads at 280 nm, and it is also the least common amino acid, at about 1.1% of residues in Swiss-Prot. A protein that happens to have none, like ubiquitin or calmodulin, is left with a weak signal from a handful of tyrosines.

From Edelhoch to the Pace equation

The idea of predicting the coefficient from composition goes back to Harold Edelhoch, who in 1967 measured blocked model compounds of tryptophan and tyrosine in 6 M guanidine hydrochloride, a denaturant that unfolds proteins so each chromophore sits in the same solvent.^[3][4] Stanley Gill and Peter von Hippel showed in 1989 that the same model values could predict the coefficients of native proteins too. Calibrated against 18 globular proteins, their method was accurate to about ±5% in most cases.^[2]

Nick Pace and colleagues refined it in 1995. They gathered 116 measured coefficients for 80 proteins and fitted the values that best reproduce them for folded proteins in water:^[1]

$$\varepsilon_{280} = 5500\,n_{\text{Trp}} + 1490\,n_{\text{Tyr}} + 125\,n_{\text{cystine}}$$

These are the values used by Expasy ProtParam and by the ProteinIQ extinction coefficient calculator.^[4] The equation needs only the three counts, which the amino acid composition tool also reports.

The fitted values differ from those of the model compounds, and the difference says something about proteins. Pace's group measured the model compounds in four solvents and found that tryptophan inside a folded protein absorbs less than tryptophan in any of them, while tyrosine falls between its values in guanidine hydrochloride and propanol.^[1]

Chromophore	Water	8 M urea	6 M GdnHCl	Propanol	Folded proteins (fitted)
Tryptophan	5,630	5,635	5,685	6,075	5,500
Tyrosine	1,215	1,300	1,285	1,680	1,490
Cystine	110	115	125	135	125

How accurate the prediction is

Across all 116 coefficients, the Pace equation's average deviation from the measured values was 3.8%. The split by tryptophan content is what matters in practice: 3.2% for the 93 entries with at least one tryptophan, and 6.5% for the 23 without. Only one tryptophan-containing protein missed by more than 10%, and its measurement carried an unusually large scattering correction. Seven of the 23 tryptophan-free entries missed by more than 10%, and bovine pancreatic trypsin inhibitor (BPTI) missed by 17.8%.^[1]

The reason is that tyrosine is more sensitive to its surroundings than tryptophan. In BPTI the average tyrosine absorbs about 1,270 $\text{M}^{-1}\,\text{cm}^{-1}$ and in insulin about 1,290, against 1,490 in ribonuclease A. Three ketosteroid isomerase variants that each have a single tyrosine and nothing else at 280 nm measured 1,310, 1,640 and 1,740.^[1] When a whole coefficient rests on a few tyrosines, one unusual environment moves it by tens of percent.

Reduced or oxidized cysteines

Because only bonded cysteines absorb, the coefficient depends on whether the protein's cysteines are paired. The oxidized value counts every possible disulfide, $\lfloor n_{\text{Cys}}/2 \rfloor$ of them; the reduced value counts none. ProtParam reports both for this reason.^[4]

For a protein rich in tryptophan the choice barely matters. Lysozyme's four disulfides add 500 to a coefficient of 37,470, about 1.3%. For a protein with no tryptophan and several disulfides, the cystine term is a larger share of a smaller number: 5.6% for ribonuclease A and 6.6% for human serum albumin, which has 17 disulfides and one free cysteine.

Use the oxidized value for secreted proteins, antibodies and anything else folded with its native disulfides, and the reduced value for cytoplasmic proteins and samples in DTT, TCEP or another reducing agent. When you know the protein forms fewer bonds than the maximum, count only those.

Protein	Trp / Tyr / Cys	ε reduced (M⁻¹ cm⁻¹)	ε oxidized (M⁻¹ cm⁻¹)	MW (Da)	E0.1% oxidized	From Trp
Human ubiquitin (P0CG48, 1 to 76)	0 / 1 / 0	1,490	1,490	8,565	0.174	0%
Human calmodulin (P0DP23, 2 to 149)	0 / 2 / 0	2,980	2,980	16,706	0.178	0%
Bovine ribonuclease A (P61823, 27 to 150)	0 / 6 / 8	8,940	9,440	13,690	0.690	0%
BPTI (P00974, 36 to 93)	0 / 4 / 6	5,960	6,335	6,518	0.972	0%
Human serum albumin (P02768, 25 to 609)	1 / 18 / 35	32,320	34,445	66,472	0.518	16%
Hen lysozyme (P00698, 19 to 147)	6 / 3 / 8	37,470	37,970	14,313	2.653	87%

The lysozyme prediction, 2.653, sits on the 2.65 that Pace's group averaged from five literature measurements.^[1] The four proteins without tryptophan are the ones whose predictions deserve the least trust.

Molar and mass extinction coefficients

The molar coefficient gives concentration in mol/L, which binding and kinetics experiments need. Formulation, yields and gel loading are planned in mg/mL, so labs often use a mass coefficient instead: the absorbance of a 1 mg/mL solution in a 1 cm path, written $E^{0.1\%}$ or Abs 0.1%. Dividing the molar coefficient by the molecular weight in daltons gives it:^[4]

$$E^{0.1\%}_{280} = \frac{\varepsilon_{280}}{M_r} \qquad c_{\text{mg/mL}} = \frac{A_{280}}{E^{0.1\%}_{280}\, l} \times d$$

Older literature and many protein data sheets use $E^{1\%}$, the absorbance of a 1% solution, which is 10 mg/mL. It is ten times $E^{0.1\%}$, so lysozyme's $E^{1\%}$ is about 26.5. Mixing the two up is the classic factor-of-ten error in A280 work. To turn a published $E^{1\%}$ into a molar value, multiply by the molecular weight and divide by 10.

The molecular weight has to describe the same molecule as the coefficient. The molecular weight calculator gives the polypeptide mass from a sequence, and glycans add mass but no absorbance at 280 nm, so the result is protein mass rather than glycoprotein mass. For an IgG, whose mass is about 150 kDa, use all four chains together; Protein Parameters reports both coefficients alongside mass, pI and stability indices for the same sequence.

Worked example: human serum albumin

Say you have purified human serum albumin, diluted 10 µL of the stock into 190 µL of buffer, a 1:20 dilution, and read the diluted sample in a 1 cm cuvette: $A_{280} = 0.450$ and $A_{330} = 0.012$. Albumin has one tryptophan, 18 tyrosines and 17 disulfides, so its oxidized coefficient is:

$$\varepsilon_{280} = 5500(1) + 1490(18) + 125(17) = 34{,}445\ \text{M}^{-1}\,\text{cm}^{-1}$$

A protein solution should not absorb at 330 nm, so the reading there is light scattered by aggregates or particles. Scattering rises steeply toward shorter wavelengths, and Pace's group corrected for it by multiplying the 330 nm absorbance by 1.929, which is $(330/280)^4$, the Rayleigh scattering ratio.^[1] The corrected absorbance is $0.450 - 1.929 \times 0.012 = 0.427$.

The concentration of the stock follows from the rearranged Beer-Lambert law:

$$c = \frac{0.427}{34{,}445 \times 1} \times 20 = 2.48 \times 10^{-4}\ \text{M} = 248\ \mu\text{M}$$

Multiplying by the molecular weight, 66,472 g/mol, gives 16.5 g/L, which is 16.5 mg/mL. Dividing the corrected absorbance by $E^{0.1\%} = 0.518$ and multiplying by 20 gives the same answer directly.

Calculation	Stock (mg/mL)	Error
Oxidized coefficient, scattering corrected	16.5	Reference
Oxidized coefficient, no scattering correction	17.4	+5.4%
Reduced coefficient, scattering corrected	17.6	+6.6%
Reduced coefficient, no scattering correction	18.5	+12.4%
Absorbance of 1 taken as 1 mg/mL	8.5	−48%
E1% (5.18) used as if it were E0.1%	1.65	−90%

No single mistake in the table is dramatic, but two small ones in the same direction already put the answer 12% high. The generic 1 mg/mL assumption and the unit mix-up are far worse, and neither shows up as anything odd on the instrument.

When A280 gives the wrong answer

Every limitation of A280 is a case where the absorbance is not all protein, or where the coefficient is not right for the protein.

Proteins with little or no tryptophan

Without tryptophan, the coefficient rests on tyrosines whose individual values vary by tens of percent, as the BPTI and ketosteroid isomerase numbers show. Expasy warns that the error can exceed 10% for such proteins.^[4] The signal is also weak: calmodulin at 1 mg/mL reads about 0.18, so a dilute sample sits close to the noise of the instrument.

A protein with neither tryptophan nor tyrosine has a coefficient of zero, or 125 per disulfide at most, and A280 cannot measure it. Nicholas Anthis and Marius Clore proposed reading at 205 nm instead, where the peptide backbone absorbs, and predicting that coefficient from the sequence as well. Their sequence-based values kept concentration errors below 4% for most of the eight polypeptides they tested, against up to 19% for the generic 205 nm value of 31 mL mg⁻¹ cm⁻¹. Buffers absorb far more at 205 nm, so blanks need more care.^[5]

Nucleic acid contamination

Nucleic acids absorb strongly at 280 nm too, and they absorb even more at 260 nm. Pure RNA has an $A_{260}/A_{280}$ ratio of about 2.0 and pure DNA about 1.8, while a purified protein reads around 0.57.^[7][6] A protein sample whose ratio climbs above 0.6 is carrying something that absorbs like a nucleic acid, and the extra absorbance at 280 nm counts as protein.

If the protein contributes absorbance with a ratio of 0.57 and RNA with a ratio of 2.0, the measured ratio tells you how much of $A_{280}$ belongs to RNA, because absorbances add. The share $f$ is $(r - 0.57)/(2.0 - 0.57)$, and the protein concentration is overestimated by a factor of $1/(1 - f)$:

Measured A260/A280	Share of A280 from nucleic acid	Protein concentration overestimated by
0.6	2%	2%
0.7	9%	10%
0.8	16%	19%
1.0	30%	43%
1.2	44%	79%

The protein's own ratio is not exactly 0.57: protein samples in the Implen note read between 0.585 and 0.636.^[6] Below about 0.65 the table cannot separate contamination from composition. Above 0.7, the right fix is another purification step, such as a nuclease treatment or an ion exchange column, rather than a correction factor.

Scattering, cofactors and other absorbers

Aggregates and dust scatter light at every wavelength and show up as absorbance above 320 nm. Pace's group found that the most widely used coefficient for ribonuclease A, 9,800 $\text{M}^{-1}\,\text{cm}^{-1}$, was about 4% higher than their scattering-corrected measurements, and suggested that many literature values had never been corrected.^[1] They filtered stock solutions through a 0.2 µm filter and corrected any remaining absorbance above 325 nm with the 330 nm rule used in the worked example.

The sequence-based method also assumes the protein has no other chromophore. Gill and von Hippel excluded proteins with prosthetic groups that absorb in the near-UV, such as catalase, hemoglobin and peroxidase.^[4] The same applies to fluorescent labels, bound nucleotides and absorbing buffer components: blank with the exact buffer, and for labeled proteins subtract the dye's contribution at 280 nm using the supplier's correction factor.

Finally, the coefficient has to belong to the construct in the cuvette. A His tag adds no absorbance, but a fusion partner can add tryptophans and tyrosines, and a signal peptide or propeptide that is cleaved during expression should not be counted. Paste the expressed sequence, not the UniProt precursor.

Measuring the coefficient instead

Pace's group concluded that measuring the coefficient is better than predicting it, and found the Edelhoch method the most convenient accurate way to do it.^[1] Two identical dilutions of the same stock are made, one in buffer and one in 6 M guanidine hydrochloride. In the denaturant every chromophore is exposed to solvent, so its coefficient is the model-compound value from Table 1, and the concentration of both samples follows from the unfolded spectrum:

$$\begin{aligned} \varepsilon_{\text{native}} &= \varepsilon_{\text{GdnHCl}} \times \frac{A_{\text{native}}}{A_{\text{GdnHCl}}} \\[4pt] \varepsilon_{\text{GdnHCl}} &= 5685\,n_{\text{Trp}} + 1285\,n_{\text{Tyr}} + 125\,n_{\text{cystine}} \end{aligned}$$

Measured this way, Pace's coefficients for seven well-characterized proteins agreed with literature values from other methods to within 2% on average.^[1] Measuring pays off for a protein without tryptophan, for a reagent whose concentration feeds every later experiment, or whenever the prediction and another assay disagree.

Extinction coefficient tools on ProteinIQ

The extinction coefficient calculator predicts reduced, oxidized or user-specified disulfide coefficients from FASTA or CSV input, and with an A280 reading, path length and dilution returns the concentration in M, µM and mg/mL for each sequence. Its page walks through the input options and the meaning of each output column.

Protein Parameters reports the same coefficients in a ProtParam-style profile with molecular weight, pI, instability index and GRAVY. The molecular weight calculator gives the mass needed to convert between molar and mg/mL values, and the amino acid composition tool lists the tryptophan, tyrosine and cysteine counts behind every coefficient. If your sequence is not yet in FASTA format, any of them will take a plain one-letter sequence.

Frequently asked questions

What are the units of the extinction coefficient?

The molar extinction coefficient is in $\text{M}^{-1}\,\text{cm}^{-1}$, equivalent to $\text{L}\,\text{mol}^{-1}\,\text{cm}^{-1}$. The mass coefficient $E^{0.1\%}$ is in $(\text{mg/mL})^{-1}\,\text{cm}^{-1}$, often written $\text{mL}\,\text{mg}^{-1}\,\text{cm}^{-1}$, and $E^{1\%}$ refers to a 10 mg/mL solution, so it is ten times larger.

How do I calculate protein concentration from A280?

Divide the absorbance by the molar extinction coefficient and the path length, then multiply by the dilution factor: $c = A_{280}/(\varepsilon\,l) \times d$. Multiply the molar result by the molecular weight to get mg/mL, or divide the absorbance by $E^{0.1\%}$ directly.

What is a normal extinction coefficient for a protein?

There is no typical value, because it scales with size and with tryptophan content. Molar coefficients range from zero for a protein with no tryptophan or tyrosine to more than 200,000 $\text{M}^{-1}\,\text{cm}^{-1}$ for an IgG antibody. Mass coefficients vary less but still widely: among the six proteins in Table 2 they run from 0.17 for ubiquitin to 2.65 for lysozyme.

Can I use A280 if my protein has no tryptophan?

Yes, if it has tyrosines, but expect larger errors. In Pace's data set, predictions for proteins without tryptophan were off by 6.5% on average and by more than 10% for about a third of them. Measuring the coefficient by the Edelhoch method, or reading at 205 nm with a sequence-based coefficient, gives a more reliable number.^[1][5]

What A260/A280 ratio means a protein sample is pure?

A purified protein reads around 0.57, and real samples land slightly above that depending on composition.^[6] A ratio of 0.7 means roughly a tenth of the 280 nm absorbance comes from nucleic acid, and 1.0 means almost a third.

Should I use the reduced or oxidized extinction coefficient?

Match the state of the protein in the cuvette. Use the oxidized value when the protein folds with its native disulfides, as secreted proteins and antibodies do, and the reduced value for cytoplasmic proteins or any sample containing a reducing agent. For proteins with several tryptophans the difference is about 1%.