PROTOCOL

Using circular dichroism spectra to estimate protein

secondary structure
Norma J Greenfield
Department of Neuroscience and Cell Biology, Robert Wood Johnson Medical School, 675 Hoes Lane West, Piscataway, New Jersey 08854-8021, USA. Correspondence
should be addressed to N.J.G. ([email protected]).

Published online 25 January 2007; doi:10.1038/nprot.2006.202

Circular dichroism (CD) is an excellent tool for rapid determination of the secondary structure and folding properties of proteins that
© 2007 Nature Publishing Group http://www.nature.com/natureprotocols

have been obtained using recombinant techniques or purified from tissues. The most widely used applications of protein CD are to
determine whether an expressed, purified protein is folded, or if a mutation affects its conformation or stability. In addition, it can
be used to study protein interactions. This protocol details the basic steps of obtaining and interpreting CD data, and methods for
analyzing spectra to estimate the secondary structural composition of proteins. CD has the advantage that measurements may be
made on multiple samples containing r20 lg of proteins in physiological buffers in a few hours. However, it does not give the
residue-specific information that can be obtained by x-ray crystallography or NMR.

INTRODUCTION
The rapid characterization of new proteins is of great importance circularly polarized light and the CD of optically active molecules
for the fields of proteomics and structural genomics. Circular with animated graphics (http://www.enzim.hu/~szia/cddemo/).
dichroism (CD) is an excellent method for rapidly evaluating the CD is an excellent method of determining the secondary struc-
secondary structure, folding and binding properties of proteins. ture of proteins. When the chromophores of the amides of the
Briefly, CD is defined as the unequal absorption of left-handed and polypeptide backbone of proteins are aligned in arrays, their optical
right-handed circularly polarized light. A beam of light has time- transitions are shifted or split into multiple transitions as a result of
dependent electric and magnetic fields associated with it. If the light ‘exciton’ interactions (see Ref. 3 for a recent review). The result is
is polarized by passing through suitable prisms or filters, its electric that different structural elements have characteristic CD spectra
field, E, will oscillate sinusoidally in a single plane. When viewed (Fig. 1a). For example, a-helical proteins have negative bands at
from the front, the sinusoidal wave can be visualized as the resultant 222 nm and 208 nm and a positive band at 193 nm (Ref. 4).
of two vectors of equal length, which trace out circles, one that Proteins with well-defined antiparallel b-pleated sheets (b-helices)
rotates clockwise (ER) and the other that rotates counterclockwise have negative bands at 218 nm and positive bands at 195 nm
(EL). The two circularly polarized waves have physical existence. (Ref. 5), whereas disordered proteins have very low ellipticity above
The waves are 90 degrees out of phase with each other and can be 210 nm and negative bands near 195 nm (Ref. 6). The collagens are
separated using a variety of prisms or electronic devices that utilize a unique class of proteins, which have three chains that wrap
Pockel’s effect1. When asymmetric molecules interact with light, together in a triple helix. Each strand has a conformation resem-
they may absorb right- and left-handed circularly polarized light to bling that of poly-L-proline7 in an extended helical conformation
different extents (hence the term circular dichroism) and also have in which all of the bonds are trans to each other (poly-L-proline II).
different indices of refraction for the two waves. The result is that Charged polypeptides, such as poly-L-glutamate or poly-L-lysine at
the plane of the light wave is rotated and that the addition of the ER neutral pH (originally thought to be in random coil conformation),
and EL vectors results in a vector that traces out an ellipse and the have a similar extended poly-L-proline II-like conformation8–10.
light is said to be elliptically polarized. CD is reported either in The spectra of some representative proteins, with widely varying
units of DE, the difference in absorbance of ER and EL by an conformations, are shown in Figure 1b. Because the spectra of
asymmetric molecule, or in degrees ellipticity, which is defined as proteins are so dependent on their conformation, CD can be used
the angle whose tangent is the ratio of the minor to the major axis
of the ellipse. [y], the molar ellipticity in deg cm2 dmol–1 ¼
3,298DE. For illustrations of the phenomena of CD see Beychok
(Ref. 2). There is also a website that illustrates the production of a 1 α-helix b
8e + 4
2 Antiparallel β 8e + 4
1 Myoglobin
6e + 4 3 Extended
2 LDH
4 Collagen (triple helix) 6e + 4
Figure 1 | CD spectra of polypeptides and proteins with representative
[θ], deg cm2 dmol –1

[θ], deg cm2 dmol –1

3 Chymotrypsin
4e + 4 5 Collagen (denatured) 4 Bence Jones
secondary structures. (a) CD spectra of poly-L-lysine at pH 11.1 in the (1, 4e + 4 2
black) a-helical and (2, red) antiparallel b-sheet conformations and at pH 5.7 2e + 4
2e + 4
in the (3, green) extended conformations5 and placental collagen in its (4, 0e + 0 4
blue) native triple-helical and (5, cyan) denatured forms64. (b) CD spectra of 0e + 0

representative proteins with varying conformations: 1 (black), sperm whale –2e + 4 5 2 3

–2e + 4
myoglobin; 2 (green), chicken heart lactate dehydrogenase; 3 (red), bovine –4e + 4 3 1 1
–4e + 4
a-chymotrypsin and 4 (cyan), human Bence Jones protein REI light chain, which 4 180 200 220 240
–6e + 4
is a human immunoglobulin light chain of k type. Spectra are from data sets 190 200 210 220 230 240 250 Wavelength, nm
supplied by Dr. W.C. Johnson (Oregon State University, Corvallis, Oregon, USA). Wavelength, nm

2876 | VOL.1 NO.6 | 2006 | NATURE PROTOCOLS

 PROTOCOL

to estimate the structure of unknown proteins and monitor with that collected under the same conditions in cells with
conformational changes due to temperature, mutations, heat, 0.1-cm or higher pathlengths.
denaturants or binding interactions. Although CD does not give
the secondary structure of specific residues, as do x-ray crystal- Preparation of buffers
lographic and NMR structural determinations, the method has the Buffers for CD spectroscopy must not contain any materials that
advantage that data can be collected and analyzed in a few hours on are optically active and should be as transparent as possible. The
solutions of samples containing r20 mg of protein in aqueous total absorbance of the sample, including the buffer and cell, should
buffers under physiological conditions. Secondary structure can be below one for high-quality data. Samples in which the protein is
also be estimated from Fourier transform infrared spectroscopy dissolved in water alone have the highest transparency, but some
of proteins11–13 and Raman spectroscopy14, but in some cases the proteins denature in the absence of salt. The low wavelength cutoffs
measurement must be made on protein films or in deuterium of several buffers commonly used for CD measurements (buffer +
oxide. 0.1 mg ml–1 protein in 0.1-cm cuvettes) are given in Table 1.
© 2007 Nature Publishing Group http://www.nature.com/natureprotocols

In addition to the intrinsic CD of the protein backbone, when Oxygen absorbs light below 200 nm; for optimum transparency,
ligands with chromophores bind to proteins they may develop strong buffers should therefore be prepared with glass-distilled water or
extrinsic CD bands that can also be used to follow binding. The the water should be degassed before use. Samples for CD analysis
aromatic chromophores of proteins, which have bands in the near- should be free of particulate matter. They can be filtered through
UV region, are often in asymmetric environments and can be used to filters (e.g., 0.1–0.2 mm) such as those sold by Millipore (http://
examine whether mutations change the tertiary structure of proteins. www.millipore.com) or centrifuged at 100,000g.
Key considerations in designing and implementing CD experi-
ments are discussed here. Preparation of proteins and peptides
Samples for CD spectroscopy must be at least 95% pure by the
Choosing CD cuvettes criteria of high-performance liquid chromatography (HPLC),
CD spectra are collected in high-transparency quartz cuvettes mass spectroscopy or gel electrophoresis. For secondary structure
(cells). Both rectangular and cylindrical cells are available, with measurements, sample concentrations may range from 0.005–
pathlengths ranging from 0.01–1 cm. Water-jacketed cylindrical 5 mg ml–1 depending on the pathlength of the cell. The most
cells are available for CD machines that do not have temperature- difficult part of obtaining high-quality CD data is the correct
regulated cell holders. Most cylindrical cells and 0.1-cm rectangular determination of protein concentration. The method of Lowry15
cells have relatively low birefringence and give reasonably straight and the Bradford dye binding methods16 give different results
baselines, but all cells are different and baseline spectra must always for different proteins and must not be used for determining
be collected. Rectangular cells with pathlengths 40.2 cm may have protein concentrations for CD.
high birefringence due to strain. Cells designed for fluorescence, The most accurate method of determining protein concentration
in which all four sides are made of the same material, usually is quantitative amino acid analysis, using the concentrations of the
have lower birefringence than cells where two sides are frosted. stable amino acids (e.g., alanine and lysine) to calculate the
Rectangular cells with pathlengths o0.1 cm often have a very concentration of the intact protein. This method is very sensitive
small total sample volume and a very small surface area facing the and can be performed on aliquots of the actual CD samples;
light beam of the CD machine. It is important that the light however, many laboratories do not have the equipment to do
beam be very tightly focused if these cells are used, because large their own measurements. Concentration can be determined using
artifacts are produced if the light does not go directly through the published molar extinction coefficients if they are available; the
sample. Dismountable short-pathlength cells are available, but the protein should be dialyzed or desalted into the CD buffer imme-
pathlength can be affected by the viscosity of the sample, and they diately before the spectrum is obtained and filtered though
tend to leak if temperatures are changed. Their use is not recom- 0.1–0.2 mm filters to reduce light scattering. The spectrum
mended. If they are necessary to collect data at low wavelengths, the of the CD buffer alone must be subtracted from the spectrum
scale should be checked by comparing data at higher wavelengths of the sample.

TABLE 1 | Properties of buffers.

BLower wavelength
Buffer limit, nm*
10 mM potassium phosphate, 100 mM potassium fluoride 185
10 mM potassium phosphate, 100 mM (NH4)2SO4 185
10 mM potassium phosphate, 50 mM Na2SO4 185
10 mM potassium phosphate, 100 mM KCl 195
20 mM sodium phosphate, 100 mM NaCl 195
Dulbecco’s phosphate buffered saline (PBS): 9.33 mM potassium phosphate, 136 mM NaCl, 200
2.7 mM KCl, 0.6 mM MgCl2, 0.9 mM CaCl2
2 mM Hepes, 50 mM NaCl, 2 mM EDTA, 1 mM dithiothreitol 200
50 mM Tris, 150 mM NaCl, 1 mM dithiothreitol, 0.1 mM EDTA 201
*The lower limit values are typical for solutions containing B 0.1 mg ml–1 protein in 0.1-cm cells. Below the lower wavelength cutoffs the dynode voltages rapidly increase, the signal to noise is poor and
the ellipticity is not a linear function of the path length of the cell. DMSO and formamides have high absorbance and cannot be used for CD measurements. Many organic solvents, e.g. trifluoroethanol,
hexafluoropropanol and hexane are transparent to 185 nm and below but will change the structure of proteins and polypeptides. Buffers can contain up to 20% glycerol, but measurements can only be made
to 200 nm at this concentration.

NATURE PROTOCOLS | VOL.1 NO.6 | 2006 | 2877

 PROTOCOL

Three different methods of rapid and accurate protein quanti- well-folded proteins (see later). However, most programs using
fication, which are independent of protein composition, are data based on protein spectra do not correctly analyze the
described in this protocol (Step 1, Options A–C). Because CD is conformation of proteins with a majority of pure b-helices, such
usually carried out on samples with relatively low concentrations, it as found in some synthetic polypeptides and in amyloids, or many
is best to prepare a filtered stock solution of the protein of B1 mg fibrous proteins such as collagen or coiled-coil proteins. For
ml–1 in the CD buffer, determine its protein concentration and analysis of the spectra of these samples, more accurate results will
dilute the stock for the CD measurements. be obtained using constrained least-squares fitting analysis pro-
None of these rapid methods will work if the protein is grams (e.g. LINCOMB (Ref. 40)) with polypeptide-based reference
highly stable and the protein is not fully unfolded in 6 M guanidine sets5,20,21, or the programs K2D (Ref. 38) and CONTIN (Ref. 27).
HCl or in 3% NaOH (wt/vol). In these cases the protein can Table 3 compares the results obtained using various methods for
be hydrolyzed and the amino acid composition quantified the conformational analyses of the proteins shown in Figure 1b,
using ninhydrin17; the protein may be subjected to Kjeldhal and for the analysis of the conformation of two polypeptides, one
© 2007 Nature Publishing Group http://www.nature.com/natureprotocols

digestion followed by determination of ammonium sulfate using with a pure a-helical conformation and one with a pure b-pleated
Nessler’s reagent18; or the protein may be ashed with 72% HClO4 in sheet conformation. The effects of wavelength range on the good-
water at 210 1C and the nitrogen determined with Berthelot ness of fits for the programs listed in Table 3 have been compared
phenol–hypochorite reagent as described by Jaenicke19. The elsewhere26. The effect of the nature of the protein database on the
phenol–hypochorite method has high sensitivity, needing only goodness of the estimates for SELCON3, CONTIN and CDSSTR
0.05–10 mg of protein in the sample, and has an error within of have been described in detail36.
0.01 mg, but it and the micro-Kjeldhal method cannot be used
for samples dissolved in buffers containing nitrogen, such as Linear regression. Linear regression fits the spectrum of an
Tris-HCl. A protocol for the Kjeldhal method is provided in unknown protein by comparison to the spectra of a set of fixed
Supplementary Methods. standards5. It is useful for evaluating the effects of mutations,
Some proteins, such as collagen and collagen fragments, fold very ligands and solvents on protein conformation. Suitable standards
slowly. If the folding properties are not known the proteins may be include polypeptides with known conformations5,20,21, which are
pre-folded for several days on ice in the refrigerator, or for several essential for the analysis of nonglobular polypeptides and fibrous
hours at 25 1C before performing CD experiments. proteins; standard spectra extracted from a database of known
proteins using the method of least squares5,22,24,25,41,42, which are
Data analysis useful for analyzing globular proteins; and standard spectra
There are many different methods to analyze CD spectra to extracted from a database of protein with known conformations
estimate secondary structure; the most commonly used are using the convex constraint algorithm (CCA)40. Either constrained
described here. Links to sites where the software can be obtained least-squares or nonconstrained multilinear regression (MLR)
and descriptions of the sites are given in Table 2. In addition, some can be used.
of the software is available in the Supplementary Software. In constrained least-squares fits, the sum of the contribution of
All methods of analyzing CD spectra assume that the spectrum each spectrum is constrained to equal one: yl ¼ SeiSli, Sei ¼ 1 and
of a protein can be represented by a linear combination of the ei Z 1. Least-squares analysis can be performed using the program
spectra of its secondary structural elements, plus a noise term, LINCOMB (Ref. 40), which is available in the Supplementary
which includes the contribution of aromatic chromophores and Software. It can also be done by loading a database of reference
prosthetic groups. spectra into a commercial graphics program such as SigmaPlot
yl ¼ Sei Sli + noise ð1Þ (Systat Software Inc.; http://www.systat.com/products/sigmaplot/),
with the wavelength in the first column, the ellipticity values of the
where yl is the CD of the protein as a function of wavelength; ei is references (e.g., a-helix, antiparallel b-sheet, parallel b-sheet, turn
the fraction of each secondary structure, i; and Sli is the ellipticity and disordered) in columns 2–6 and the unknown protein in
at each wavelength of each ith secondary structural element. In column 7, and fitting the data in column 7 to the equation yli ¼
constrained fits the sum of all the fractional weights, ei, must equal SeiSi in which SSi is constrained to equal 1 and ei is constrained to
one, and all of the fractional contributions must be greater than or be Z0. An example of the needed equations is given in SigmaPlot
equal to zero. format in Supplementary Equations. The use of commercial
There are two general classes of methods to evaluate protein graphics programs is recommended because they quickly find the
conformation. The first uses standards of polypeptides, with best fit to the data with all of the fractional components being
defined compositions in known conformations, which have been positive and one can easily plot the best fit to the unknown data.
determined by x-ray scattering of films or by IR in solution5,20,21. The major advantages of this approach are the use of an invariant
The second uses the spectra of proteins that have been characterized database, which is useful for direct comparisons, and that better fits
by x-ray crystallography as standards. These are then compared to are obtained than with nonconstrained least squares methods.
the spectra of unknown proteins using least-squares analysis22–26; However, there are no good single standards for b-turns.
ridge regression27; singular value decomposition (SVD)17; SVD In nonconstrained least-squares fit the sum of the conformations
with variable selection28–31; the self-consistent method32–36; or is not constrained to equal one20.
neural network analysis37–39.
All of the listed methods described here are useful for determin- yl ¼ Sei Sli + C ð2Þ
ing the a-helical content of globular proteins. The protein-based A DOS-based program for calculating nonconstrained least-
analyses are superior when analyzing the conformation of globular, squares fits is MLR. It is available in the program in Supplementary

2878 | VOL.1 NO.6 | 2006 | NATURE PROTOCOLS

 PROTOCOL

TABLE 2 | Sources of circular dichroism analysis software.

Website Software Operating systems
CDPro SELCON3 Windows 95, 98, XP;
http://lamar.colostate.edu/Bsreeram/CDPro/ListPro.htm LINUX, UNIX
Advantages: Data conversion program included. Superior fits of data on CONTIN
globular proteins. Source code available and can be compiled for use with LINUX CONTINLL
or UNIX machines. CDSSTR

Circular Dichroism at UMDNJ LINCOMB MS-DOS, Windows 95,

http://www2.umdnj.edu/cdrwjweb/. These programs are also in the file CD.Zip in MLR 98, XP
SUPPLEMENTARY MATERIALS.
Advantages: Data conversion programs are supplied. Data collected at any range SELCON
of wavelengths can be analyzed. Peptide references for least squares analyses SELCON2
© 2007 Nature Publishing Group http://www.nature.com/natureprotocols

are included. CONTIN

Disadvantages: Programs are not user friendly. Data conversion and text editing necessary. VARSLC
K2D
CCA
CONTIN CONTIN LINUX, Source Code
http://s-provencher.com/index.shtml
http://s-provencher.com/pages/contin.shtml
Disadvantages: Programs must be compiled by user.

CCA + the CD Spectrum Analyser System CCA Windows 95, 98

http://www2.chem.elte.hu/protein/programs/cca/
Advantage: Windows operating system.

DICROPROT LINEAR REGRESSION Windows 95, 98, XP

http://dicroprot-pbil.ibcp.fr/ SELCON2
ftp://ftp.ibcp.fr/pub/C_DICROISM/ SELCON3
Advantages: Easy to use. Every method uses the same input format. Output files CONTIN
are shown on the screen immediately. Note: Programs to convert data to the DicroProt VARSLC
‘dic’ format are in SUPPLEMENTARY MATERIALS.
Disadvantages: The package does not have a constrained least squares analysis module
where the fractional spectral components are constrained to be positive. Many of the
programs do not run unless the CD data is in the range of 260–178 nm.
DICHROWEB CONTINLL Online
http://www.cryst.bbk.ac.uk/cdweb/html/home.html SELCON3
Advantages: On-line analysis available. Accepts input directly from the output of many CDSSTR
different CD machines. VARSLC
K2D
K2D K2D Online, DOS, Windows
http://www.embl-heidelberg.de/%7Eandrade/k2d.html 95, 98, XP
Advantages: Simple to use.

SOMCD SOMCD Online

http://geneura.ugr.es/cgi-bin/somcd/index.cgi
Advantages: Update of K2D algorithm. Analyzes turns as well as a-helix and b-sheets
and uses data from 240–190 nm as well as from 240–200 nm.

Software. It also can be evaluated using graphics programs spectra of a large database of proteins with known conformations.
with curve-fitting routines. An example of such a routine is also In this method the contribution of each reference spectrum is kept
given in Supplementary Equations. The advantages of this small unless it contributes to a good agreement between the
approach are that there is no need to know protein concentration, theoretical best fit curve and the raw data. This method results in
invariant standards are used and it is useful for analyzing difference relatively good estimates of a-helices and b-sheets. Different
spectra. However, it is the least accurate method. references are used for every fit; this is an advantage for obtaining
the best fits of the data, but it complicates the quantitative analysis
Ridge regression (CONTIN). CONTIN (Ref. 27) fits the CD of of the effect of a mutations or denaturant because a different set of
unknown proteins by comparison to a linear combination of the standards is used for each analysis.

NATURE PROTOCOLS | VOL.1 NO.6 | 2006 | 2879

 PROTOCOL

Singular value decomposition. SVD (Ref. 17) extracts basis curves provides very good estimates of the structure of globular proteins.
with unique nodes from a set of spectra of proteins with known However, SELCON3 gives poor estimates of turns compared to
structures. The basis curves are each characterized by a mixture of SELCON1 and SELCON2, VARSLC and CDNN. All the versions of
secondary structures and are then used to analyze the conformation SELCON are unsuitable for the analysis of polypeptides and protein
of unknown proteins. The sum of weights is not constrained to fragments.
equal 1. This method provides the best estimates of a-helical
content of proteins. However, it provides terrible estimates of Neural networks (CDNN, K2D, SOMCD). A neural network is an
b-sheets and turns if data are not collected to at least 184 nm. It is artificial intelligence program used to find correlations in data. Two
also unsuitable for the analysis of polypeptides and protein fragments. widely used programs are CDNN (Ref. 37) and K2D (Ref. 38).
CDNN analyzes data to determine helix, antiparallel and parallel
Variable selection (VARSLC, CDSSTR). In the variable selection b-structure, turns and remainder and K2D determines helix, total
method28 an initial large database of standard spectra from proteins b-structure and remainder. SOMCD is a variation of K2D that uses
a larger reference set to train the network and also analyzes turns39.
© 2007 Nature Publishing Group http://www.nature.com/natureprotocols

with known spectra and secondary structures is created. The

protein structure is then determined using SVD. Some of the A neural network is first trained using a set of known proteins,
protein spectra are then eliminated systematically to create new so that the input of the CD at each wavelength results in the
databases with a smaller number of standards. SVD is performed output of the correct secondary structure. The trained network is
using all of the reduced data sets, and those fulfilling selection then used to analyze unknown proteins. K2D gives a good estimate
criteria for a good fit are averaged. Two versions of the program of the helical and sheet contents of both proteins and polypeptides.
are available: VARSLC (also called Varselec) and CDSSTR. This However, the K2D program does not estimate turns. CDNN is
method provides superior fits of the conformation of globular not suitable for the analysis of polypeptides, and it currently is not
proteins. However, it does not always give fractional content of being distributed.
various conformations that add up to 100%, and it can be very slow
if a large number of reference sets are used and more than two Convex constraint algorithm. The CCA algorithm44 deconvolutes a
spectra are eliminated from the combinations. It is also unsuitable set of spectra into a desired number of basis spectra, which, when
for the analysis of polypeptides and protein fragments. recombined, generate the entire data set with a minimum deviation
between the original data set and the reconstructed curves. It is very
Self-consistent method (SELCON). In SELCON, the spectrum of useful to determine how many different states contribute to the
the protein to be analyzed is included in the basis set and an initial changes in CD as a function of ligand or temperature26,45–47. The
guess is made for the unknown structure as a first approximation. method was developed to estimate protein conformation but is
The resulting matrix equation is solved using the SVD algorithm, poorer than least squares, SVD or neural net analysis.
and the initial guess is replaced by the solution. The process is The protocol described here will cover setting up CD machines
repeated until self-consistency is attained. The program works well to collect data; procedures for obtaining high-quality, reproducible
for globular proteins. The original program, SELCON (Ref. 32), data; and methods to analyze CD spectra to estimate the secondary
evaluates a-helix, antiparallel and parallel b-sheets, turns and structures of proteins. For more detailed information there are
remainder. SELCON2 (Ref. 43) is modified to use a data set several reviews that discuss CD spectroscopy in detail, including the
where the poly-L-proline II conformation is evaluated in addition preparations of buffers, protein samples and cuvettes, and convert-
to a-helix, total b-sheets and turns. In the version on the CDpro ing raw data to molar and mean residue ellipticity48–53, and analysis
website (see Table 2), SELCON3 (Ref. 36), the a-helix and b-sheet of data to yield secondary structure information3,26,50,54. Three
conformations in globular protein structures are divided into additional protocols will cover (i) the determination of the thermo-
regular and distorted fractions by considering a certain number dynamics of protein folding from CD data collected as a function
of terminal residues in a given helical or strand segment to be of temperature47; (ii) determination of the free energy of
distorted. The number of a-helical and b-strand segments and their folding and binding constants from CD data collected as a
average length in a given protein are estimated from the fraction of function of denaturants, osmolytes or ligands55; and (iii) determi-
distorted helical and strand conformations relative to the total helix nation of the kinetics of folding from CD data collected as a
and strand content. The main advantage of this method is that it function of time56.

MATERIALS
REAGENTS EQUIPMENT
. Proteins to be analyzed, dissolved in an appropriate buffer . Circular dichroism instrument (see EQUIPMENT SETUP)
All of the following reagents are available from Sigma–Aldrich . Circular dichroism cuvettes (Hellma Cells Inc. (http://www.HellmaUSA.com
(http://www.sigmaaldrich.com) or http://www.hellma-worldwide.de); NSG Precision Cells Inc.
. Guanidine HCl (Step 1, options A and B (cat. no. G7153) (http://www.nsgpci.com/) or Luzchem Research Inc. (http://www.
. Sodium hydroxide (NaOH) reagent grade pellets (Step 1, options A–C; luzchem.com/products/)).
cat. no. 22146) . Spectrophotometer (Step 1, options A–C)
. Sodium carbonate (Step 1, option C; cat. no. 222321) . Microtiter plate reader (Step 1, option C)
. Sodium citrate (Step 1, option C; cat. no. S1804) . Microtiter plates (Step 1, option C)
. BSA Cohn Fraction V (BSA) (Step 1, option C; cat. no. A3059) . 0.1–0.2 mm filters (Millipore; http://www.millipore.com)
. Copper sulfate (Step 1, option C; cat. no. C1297)

REAGENT SETUP TIMING 30 min to 2 h
. Whatman #1 filter paper, diam. 11 inches; Step 1, option C; ! CAUTION Solutions of guanidine HCl, NaOH, and Benedict’s reagent
cat/ mp/ Z240087) are caustic. Wear gloves.

2880 | VOL.1 NO.6 | 2006 | NATURE PROTOCOLS

 PROTOCOL

6 M Guanidine HCl, pH 7.1 Dissolve 57.3 mg guanidine HCl in B90 ml Camphor-10-sulfonic acid (Sigma-Aldrich cat. no. C2107; http://www.
glass-distilled or de-ionized water. Adjust the pH to 7.1 by adding either 1 M sigmaaldrich.com), 1 mg ml–1 in a 1-cm cell, has an A285 of 0.149 and an
HCl or solid NaOH pellets. Adjust the volume to 100 ml. Keeps indefinitely in ellipticity band with a peak at 291 nm of 335 millidegrees. In addition, the ratio
a glass bottle with a plastic cap. of ellipticity of CSA at 192.5–290.5 nm should be between 2.05 and 2.08. These
6 M Guanidine HCl, pH 12.5 Dissolve 57.3 mg guanidine HCl in B90 ml numbers correspond to a De of 2.36 at 290.5 nm and 4.90 at 192.5 nm17,30.
of water. Adjust the pH to 12.5 by adding NaOH pellets. Adjust the volume to Cleaning cuvettes This procedure can be completed in 10 min or can be done
100 ml. This solution should be checked before use and the pH adjusted if overnight. Cuvettes for CD measurements must be clean and dry. Quartz cells
necessary to 412 by adding solid NaOH pellets, because carbon dioxide can be cleaned by soaking in: mild detergent solutions available from several cell
absorbed from the air will lower the pH. manufacturers, such as Hellma; a mixture of 30% concentrated HCl and 70%
6 M Guanidine HCl, pH 6.5 Dissolve 57.3 mg guanidine HCl in B90 ml water. ethanol; or concentrated nitric acid. Protein residues can be dissolved using
Adjust the pH to 6.5 by adding either 1 M HCl or solid NaOH pellets. Adjust the 6 M guanidine HCl. After soaking, cells should be rinsed with water and ethanol
volume to 100 ml. Keeps indefinitely in a bottle with a plastic cap. and either dried by suction using an aspirator or blown dry with nitrogen
Benedict’s reagent Combine 50 g of sodium carbonate with 86.5 g of sodium or compressed air that has passed through a filter to remove impurities.
citrate in 300 ml of water. Dissolve by stirring on a hot plate. Filter through If residual proteins are not removed by the previous cleaning agents,
Whatman no. 1 filter paper. Add 8.63 g of CuSO4 dissolved in 50 ml of water. Dilute filling the cells with Chromerge (a mixture of potassium chromate in
© 2007 Nature Publishing Group http://www.nature.com/natureprotocols

to 500 ml. This solution lasts about 1 year in a brown bottle at room temperature. concentrated sulfuric acid) (Fisher Scientific cat. no. C577-12; https://
3% NaOH Dissolve 3 g of NaOH pellets in 100 ml of water. www1.fishersci.com) and immediately rinsing out with water and drying
BSA, 1 mg ml–1 Dissolve 2 mg of BSA in 1 ml of water and filter through a usually is effective. This method is the best method for cleaning cells with
Millipore filter. The A280 of a 1 mg mll solution of BSA is 0.68. Dilute to a final 0.01- or 0.02-cm pathlengths.
concentration of 1 mg ml–1. ! CAUTION Nitric acid, HCl and Chromerge are very caustic and will burn
EQUIPMENT SETUP holes in lab coats and damage clothing. Wear gloves.
CD machines For data collection from B700–175 nm, machines can be obtained Starting the CD machine This procedure requires 30 min. CD machines have
from Applied Photophysics Ltd. (http://www.photophysics.com), Aviv Biomedical very powerful lamps that promote the ionization of oxygen to ozone. Ozone is
(http://www.avivbiomedical.com/), Jasco Inc. (http://www.jascoinc.com/) or Olis toxic and also will quickly destroy the mirrors in the optics of the machines.
(http://www.olisweb.com/products/cd/). For data collection at lower wavelengths Most CD machines must be flushed with nitrogen to remove oxygen before the
there are CD machines than use beamline radiation from synchrotrons: the machine is turned on and during operation. Nitrogen sources include tanks of
NSLS Brookhaven, USA (beamlines U9b and Ul1); ISA in Aarhus, Denmark prepurified nitrogen, which last B5 h, and high-pressure liquid nitrogen tanks,
(beamline UV_1); the SRS Daresbury, UK (beamline CD12); HSRC/HiSOR, which produce gas and will last 1–2 weeks depending on usage. Commercial
Hiroshima, Japan; BESSY2 in Berlin, Germany; the BSRF in Beijing, China; nitrogen should be free of oxygen and most other impurities, but some
and the NSRL in Hefei, China. m CRITICAL CD machines must be calibrated manufacturers suggest using a trap to remove impurities for added safety. If you
on a regular basis to check that the ellipticity values and wavelengths are are new to CD and there is no one to teach you the operation of your specific
correct52,57. A commonly used calibration standard, crystallized CSA, (1S)-(+)- machine, call the manufacturer and ask for their start-up protocol.

PROCEDURE
Determination of protein concentration TIMING 0.5–1 h

1| Determine the concentration of the protein stock solutions. The following three simple spectroscopic methods give rapid,
accurate measurements of protein concentration and are independent of protein composition, provided the protein is unfolded
in 6 M guanidine HCl or 3% NaOH. Options A (Refs. 58, 59) and B (Ref. 58) require that the protein have tyrosines or
tryptophans. Option C (Ref. 60) does not work with collagen-like proteins with high proline contents.
(A) Determination of protein concentration from the difference spectrum of the protein dissolved in 6 M guanidine at
pH 12.5 versus pH 7.1
(i) Pipette exactly the same volume (0.4–1 ml depending on the sample volume of the cuvette) of each solution into
two cuvettes with 1-cm pathlengths, and scan the baseline from 320–270 nm with the pH 7.1 solution in the reference
compartment and the pH 12.5 solution in the sample compartment.
m CRITICAL STEP The dilutions of the samples must be exactly the same in the reference and sample compartments.
(ii) Add exactly the same volume (e.g., 10–100 ml) of protein solution to each cuvette and obtain the difference spectrum.
Correct the spectrum for the baseline.
(iii) Calculate the concentration of protein. The molar concentration (in mol liter–1) of protein in the cuvette ¼ A/(2,357Y +
830W) (Ref. 59), where A is the absorbance at 294 nm, Y is the number of tyrosines and W is the number of tryptophans.
Correct the measured concentration for dilution. The mean residue concentration can be calculated by multiplying the
molar concentration by the number of amino acids in the protein. The number of milligrams per milliliter of protein is
calculated by multiplying the molar concentration by the molecular weight. If the proteins are denatured by the guanidine
solution, the difference method should give one band at 294 nm.
? TROUBLESHOOTING
(B) Determination of protein concentration from the aromatic spectrum determined in 6 M guanidine HCl, pH 6.5
(i) Run a baseline spectrum of two cells with an equal volume of guanidine HCl in each side.
(ii) Add a small aliquot of protein solution to the sample side and an equal volume of the buffer to the reference side.
m CRITICAL STEP The protein must be free of scattering material and 2-mercaptoethanol or DTT. The oxidized form of
these compounds absorb strongly at 280 nm, and the oxidation rates are faster in solutions containing protein than in
plain buffers, so it is difficult to subtract their contributions.
(iii) Collect the spectrum of the protein between 350–250 nm, and calculate the protein concentration using the formulas:
e288 ¼ 4; 815W+385Y+75C and e280 ¼ 5; 690W+1; 280Y+120C

NATURE PROTOCOLS | VOL.1 NO.6 | 2006 | 2881

 PROTOCOL

where W, Y and C are the numbers of tryptophans, tyrosines and cystines (oxidized) per mole of protein and e288 and
e280 are the molar extinction coefficients of the protein at 288 nm and 280 nm, respectively. The protein concentrations
(in mol liter–1) are the absorbance value at 288nm/e288 and 280nm/e280. The concentrations determined at 288 and
280 nm should agree.
? TROUBLESHOOTING
(C) Determination of protein concentration using a microbiuret procedure
(i) Aliquot protein samples in buffer (e.g., 0, 0.025, 0.05 and 0.1 ml) and add the buffer to a final volume of 0.1 ml in small,
clean test tubes.
(ii) Prepare a standard curve containing 0.02, 0.04, 0.06, 0.08 and 0.10 ml of BSA diluted to a final volume of 0.1 ml for
concentrations of 0.2, 0.4, 0.6, 0.8 and 1.0 mg ml–1, respectively.
(iii) Add 0.5 ml of 3% NaOH and 0.02 ml of Benedict’s reagent to the standards and samples. Mix with a vortex mixer.
(iv) Allow to stand at least 15 min for the color to develop.
© 2007 Nature Publishing Group http://www.nature.com/natureprotocols

(v) Read the absorbance at 330 nm (y-axis).

(vi) Plot the standard curve of absorbance as a function of protein concentration in mg ml–1 (x-axis). Correct the absorbance
of each unknown sample for the contribution of the buffer, and read the protein concentration from the curve. Correct
for dilution. This assay can be done in microtiter plates using half the volume of each reagent. The plates can be read at
340 nm, although the color intensity is lower at this wavelength than at 330 nm. The microbiuret method, using freshly
prepared Benedict’s reagent, should be linear for protein concentrations ranging between 0 and 1.5 mg ml–1, with the
absorbances ranging from B0.2 for the blank (100 ml buffer) + 0.5 ml of 3% NaOH and 0.02 ml of Benedict’s reagent to
B0.5–0.6 for the sample with 100 ml of BSA, 1.5 mg ml–1.
? TROUBLESHOOTING

Sample preparation TIMING 10–30 min
2| Prepare the protein samples. For typical measurements in a 0.1-cm cell, depending on the buffer (see Table 1), make
solutions of 0.05–0.2 mg ml–1 protein. For measurements in 0.01- to 0.02-cm cells use 0.2–1 mg ml–1 protein, and for
1-cm cells use 0.005–0.01 mg ml–1 protein.

Equipment preparation TIMING 30–40 min
3| Turn on the nitrogen, and flush the optics compartment for 15–20 min before starting the machine (see manufacturer’s
suggested time).
! CAUTION Nitrogen displaces oxygen. Operate a CD machine only in a well-ventilated room. Do not close the door. If a tank of
liquid nitrogen begins to vent because pressure has built up, leave the room and allow the excess nitrogen to disperse before
reentering.
4| Turn on the water supply or circulating water bath chiller to the lamp housing, if the lamp is water cooled. If a water
supply is used, make sure the filter is clean. If a bath is used, make sure the water is clean.
m CRITICAL STEP Avoid using ethylene glycol in the water in the circulating bath, because it can damage the pumps.
5| Turn on the circulating water bath for temperature control. If the temperature is controlled using a temperature-regulated
cell compartment that requires a heat sink, set the temperature of the bath to 20 or 25 1C. If the temperature is controlled
using water-jacketed cells, set the temperature of the bath to the desired temperature (see Step 9).
m CRITICAL STEP Be sure that the water is circulating before turning on the power to the thermal regulators, or you can burn out
the heating units.
? TROUBLESHOOTING
6| Turn on the lamp, if it has a separate switch, before you turn on the power to the rest of the CD machine or computers.
m CRITICAL STEP Firing the lamp may cause a voltage surge that can destroy electronic boards or computers in some machines if
they are turned on before the lamp is lit.
? TROUBLESHOOTING

7| Turn on the CD machine and computer, and start the CD collection program.
? TROUBLESHOOTING
8| Set the data path of the operating program to store your data.

9| Set the desired temperature. For a previously uncharacterized protein, collect spectra at multiple temperatures and
correlate the spectra with some measurement of the activity or the protein (e.g., enzyme activity, or ability to bind ligands or
antibodies). Typically, preliminary spectra of proteins are collected at 4, 25, 37, 50, 60 and 70 1C. Once a stability range has

2882 | VOL.1 NO.6 | 2006 | NATURE PROTOCOLS

 PROTOCOL

been established, data may be collected at more closely spaced intervals to determine whether there are spectral changes
indicative of folding intermediates using the CCA algorithm44 or SVD47,61,62.

Collecting CD spectra for the determination of secondary structure TIMING 2–4 h for collection of five samples and
five baselines at a single temperature, depending on wavelength range and interval and number of repeat spectra
10| Set the desired equilibration time. Usually globular proteins reach equilibrium within 2 min , but some proteins (e.g., collagen)
can require literally days to fold (see preparation of proteins). If unsure of the folding time, incubate the CD sample on ice for several
days before starting the CD measurements for measurements at 4 1C or at 25 1C for measurements near room temperature.
11| Set the half-bandwidth between 1 and 1.5 nm. These values give spectra with good signal-to-noise ratios and adequate
spectral resolution, because UV bands are broad (Fig. 2a,b).
12| Set the wavelength range: from 260–185 nm for 0.1–0.2 mg ml–1 protein samples in transparent buffers (see Table 1) in
© 2007 Nature Publishing Group http://www.nature.com/natureprotocols

0.1-cm cells; from 260–178 nm for 0.2–0.8 mg ml–1 samples in 0.01-cm cells and from 260–200 nm for 0.01–0.02 mg ml–1
proteins in 1-cm cells.
13| Set the wavelength interval, 0.5 nm for samples with signal-to-noise ratios 420:1 or 0.1, 0.2 or 0.25 nm for samples
with low ellipticity. Data collected at 0.10–0.5 nm intervals with half-bandwidths of 1–1.5 nm will give well-defined spectra
(see Fig. 2a,b, where data were collected using a half-bandwidth of 1.5 nm and a wavelength spacing of 0.25 nm).
m CRITICAL STEP Only use intervals which have reciprocals that are integers, because most data analysis programs have databases
with ellipticity values collected at 1-nm intervals.
14| Set the time for data collection at each point (i.e., signal averaging time). For samples with a concentration of
B0.1 mg ml–1 in a transparent buffer, collecting for 1 s at each point should be sufficient. If collecting replicate spectra,
at intervals from 0.1–0.5 nm, 0.5 s per point should be adequate. If the protein concentration is low or the buffer has
high absorbance, increase the averaging time. The relative signal-to-noise ratio will increase as the square root of signal
averaging time.
15| Set the instrument time constant. For routine collection of CD spectra this should be 100 ms. For rapid collection of data
(e.g., in measurements of stopped-flow CD) the time constant should be decreased to r100 ms (no greater than one-tenth the
data averaging time at each point).
16| Set the instrument to record the ellipticity and the photomultiplier tube (PMT) voltage. When light hits the photomultiplier
of the CD machine a current is induced. Most CD machines maintain constant current by raising the voltage as the amount of
light decreases. As it scans to lower wavelengths, the absorbance will increase and the PMT voltage will rise. The signal-to-noise
ratio will greatly diminish once the PMT voltage exceeds 500 V, and the data often become very noisy and unreliable.

Figure 2 | Circular dichroism spectra of lysozyme

in 10 mM sodium phosphate pH 7.0. (a) The
a 20 b 15 c 1,000 0.1-cm cell
1 Buffer Air
15 10
spectra of air (black ), water (red), buffer (green) 800
Dynode Voltage
Millidegrees
Millidegrees

10 5
and three replicate spectra of lysozyme, 0.09 mg
ml–1 in a 0.1-cm cell (orange, blue and magenta). 5 0 600
(b) The spectra of buffer (black, red, green) and 0 –5
400
three replicate spectra each of lysozyme at 0.41 –5 –10
–1
mg ml (cyan, purple, brown) and 0.83 mg ml –1
–10 –15 200
concentration (blue, magenta, orange) in a 0.01- 180 200 220 240 260 180 200 220 240 260
0
0
0
0
0
0
0
0
0
18
19
20
21
22
23
24
25
26

cm cell. (d) The change in the photomultiplier Wavelength, nm Wavelength, nm Wavelength, nm

tube dynode voltage as a function of wavelength
for the conditions illustrated in Figure 1a and b d 20,000 e 20,000 f 20,000
0.1-cm cell LINCOMB1
and for 0.41 mg ml–1 lysozyme in a 0.1-cm cell. 15,000 0 41 mg ml –1
15,000 15,000
[θ], deg cm2 dmol –1

[θ], deg cm2 dmol –1

[θ], deg cm2 dmol–1

The mean residue ellipticity of lysozyme in a 10,000 10,000 10,000

–1
0.1-cm cell (black), 0.41 mg ml (raw data not 5,000 5,000 5,000
shown); (cyan) 0.09 mg ml–1; and in a 0.01-cm 0 0 0
–1
cell (green), 0.41 mg ml ; (red), 0.83 mg ml . –1
–5,000 –5,000 –5,000
(e) The mean residue ellipticity of lysozyme (black –10,000 –10,000 –10,000
circles) fitted using the method of least squares: –15,000 –15,000 –15,000
using a peptide data base20 (blue); four basis 180 190 200 210 220 230 240 180 190 200 210 220 230 240 180 190 200 210 220 230 240
spectra extracted from 17 proteins (black)26; five Wavelength, nm Wavelength, nm Wavelength, nm

basis spectra extracted from 33 proteins

(orange)26 and CONTIN (magenta)27; K2D (cyan)38 and SELCON2 (orange)43. (f) The mean residue ellipticity of lysozyme (black circles) fitted using the CDPro
Package33: SELCON3 (magenta), CDSSTR (cyan) and CONTIN (black). Note: Lysozyme was obtained from Sigma (L6876) and dissolved in sodium phosphate, 10
mM. The protein concentration was determined using the published extinction coefficient of e1% of 26.5 (eM ¼38.2  103) (Ref. 65) for comparison with
previous data5. Data were obtained on an Aviv Model 215 spectrometer (Aviv Biomedical).

NATURE PROTOCOLS | VOL.1 NO.6 | 2006 | 2883

 PROTOCOL

(On an Aviv instrument the PMT voltage is called the dynode voltage, whereas on a Jasco it is called HT voltage, on an Applied
Photophysics machine it is called the detector gain and on an Olis machine it is called the PMT HV). For protein concentrations
ranging from 0.05–0.1 mg ml–1 in 0.1-cm cells in 10 mM phosphate buffer, the signal-to-noise ratios are usually better than
10:1 over a wavelength range of 260–185 nm (Fig. 2a) with dynode voltages below 500 V (Fig. 2c). With 0.01-cm cells and
0.4–0.8 mg ml–1 concentrations, the signal-to-noise ratio of the data is usually 430:1 between 260–178 nm with dynode
voltages o500 V (Fig. 2b,c). When the dynode voltage (Fig. 2c) exceeds 500 V (e.g., below 185 nm for data collected in a
0.1-cm cell at B0.1 mg ml–1 concentration), the signal-to-noise ratio usually becomes very poor (Fig. 2a), although the signal
may still be linear as a function of concentration on some instruments up to 700 V.
17| Determine the spectrum of the blank. Fill the cell with water and determine its spectrum. Suitable buffers should have
no ellipticity, but their increased absorbance as compared with water decreases the signal-to-noise ratio. The spectrum of the
cell containing water should be relatively flat but may be displaced from that of pure air because of the birefringence of the
© 2007 Nature Publishing Group http://www.nature.com/natureprotocols

cell (Fig. 2a).

? TROUBLESHOOTING
18| Collect a spectrum of the buffer to make sure that it does not have any ellipticity due to dichroic components or a high
absorbance leading to a very poor signal-to-noise ratio and possibly false peaks. Note that phospholipids are asymmetric and have
CD bands. If samples are suspended in phospholipids it is essential that the spectrum of the blank contains the same concentration
of protein-free lipids. The spectrum of the buffer and water should overlay each other, within the experimental error, but the
spectrum of the buffer usually has a lower signal-to-noise ratio than the spectrum of water at low wavelengths (see Fig. 2a).
? TROUBLESHOOTING
19| Clean the cell, fill with protein solution and collect the CD spectrum. It is best to collect data two or three times for new
samples to make sure that the sample is at equilibrium and the signal is not changing as a function of time. Many CD machines
can collect multiple spectra automatically. If the protein is at equilibrium, replicate scans of the protein solutions should overlay
each other and not drift as a function of time (Fig. 2a,b).
? TROUBLESHOOTING
20| If the data sets overlay each other one, average the data sets. For the most accurate estimates of protein secondary
structure, data should be collected to 178 nm or lower wavelengths, in 0.01–0.02 cm cells. because data collected at
o200 nm may have low signal-to-noise ratios, one should collect three to five scans and average them.
21| Save the raw data on the hard drive, floppy disk or other media.
m CRITICAL STEP Always immediately save the data to prevent loss if there is a power failure.
22| Smooth the spectra of the sample and blank. Most CD machines have built-in smoothing algorithms, and some will
automatically pick the best smoothing parameters; refer to the manual. The smoothing algorithms that are used depend on the
manufacturer. If the machine uses Savitsky–Golay smoothing63 and data are collected at 0.5-nm intervals, a polynomial order of
3 and a smoothing window of 15 points usually gives a good fit. If data is collected at shorter wavelength intervals, increase
the number of points. Some smoothing protocols give estimates of the goodness of smoothing by calculating whether the
difference between the raw and smoothed data has the statistical characteristics of noise.
23| Check that the data have not been oversmoothed by subtracting the smoothed curve from the raw data. The points should
be evenly distributed around zero. Some CD machines will automatically calculate the residuals from the smoothing, and these
can be viewed on the spectrometer in its data viewing mode.
? TROUBLESHOOTING

24| Subtract the smoothed baseline from the smoothed spectrum of the sample. The ellipticity for most proteins should be
close to zero between 260 and 250 nm.
? TROUBLESHOOTING
25| Convert the data to mean residue ellipticity or De using the formulas:
Ellipticity, [y], in deg cm2 dmol–1 ¼ (millidegrees  mean residue weight)/(pathlength in millimeters  concentration
in mg ml–1) or
[y] ¼ millidegrees/(pathlength in millimeters  the molar concentration of protein  the number of residues)
De ¼[y]/3,298
The mean residue weight of a peptide is the molecular weight divided by the number of backbone amides (number of amino
acids – 1 if the protein is not acetylated). It is B115 for most proteins if the molecular weight of the sample is not known, but
it should be calculated directly if the molecular weight of the protein is known for accurate results. If the protein or peptide is

2884 | VOL.1 NO.6 | 2006 | NATURE PROTOCOLS

 PROTOCOL

TABLE 3 | Comparison of various methods to estimate the secondary structure of representative proteins and polypeptides.
METHOD X-Ray LINCOMB* MLR* CONTIN SELCON VARSLC K2D CDNN
Lowest wavelength, nm 200 178 200 178 200 190 200 178 200 178 200 200 180
Myoglobin a-helix 78 96 93 89 97 67 89 73 79 76 74 74 83 84
b-sheet 0 0 0 0 0 0 0 3 0 0 0 8 3 2
Turn 10 4 5 8 3 16 0 13 20 4 18 ND 9 9
Lactate a-helix 37 46 40 63 42 46 40 41 39 40 39 55 42 44
dehydrogenase b-sheet 14 21 29 15 33 7 39 12 27 15 28 11 13 10
Turn 25 15 11 13 9 26 1 22 27 17 13 15 14
Chymotrypsin a-helix 10 15 21 33 28 11 9 7 15 24 16 12 19 20
b-sheet 38 25 14 6 0 16 32 16 26 0 11 33 29 50
Turn 26 10 16 5 9 44 32 13 16 42 39 ND 21 22
© 2007 Nature Publishing Group http://www.nature.com/natureprotocols

Bence Jones protein a-helix 3 0 0 0 0 6 0 3 9 14 0 3 13 16

b-sheet 50 43 40 68 40 42 81 47 34 7 69 50 39 22
Turn 24 25 29 17 28 25 10 23 40 28 17 ND 22 22
Lowest wavelength, nm 200 190 200 190 200 190 200 178 200 178 200 200 180
Poly(lys-leu)n a-helix 0 0 9 0 2 12 9 31 24 31 30 5 34 39
In 0.5 M NaF, pH 7 b-sheet 100 89 65 89 99 73 91 24 39 51 55 89 18 13
Turn 0 11 26 7 0 0 0 26 26 1 3 ND 16 13
Lowest wavelength, nm 200 190 200 190 200 190 200 190 200 190 200 190 190
Poly-L-lysine 0.01% a-helix 100 97 100 - - 100 100 99 89 100 96 100 92 94
helical form, pH 11.1 b-sheet 0 3 0 - - 0 0 0 1 6 18 0 1 1
In water Turn 0 0 0 - - 0 0 1 6 13 15 0 7 7
*The polypeptide reference set of Brahms and Brahms20 for a-helix, b-structure, generic b-turn and disordered were used as standards for the evaluation the helical contents of the four proteins. The polypeptide
reference set of Reed and Reed21 for a-helix, b-structure, b-turns (type 1 and type 2) and disordered were used for the evaluation of the spectra of poly(lys-leu) in the b-structure conformation and poly-L-lysine in
the a-helical conformation.

monomeric and does not aggregate under the experimental conditions used to collect the CD data, the spectra collected at
different protein concentrations and pathlengths should give the same mean residue ellipticities (Fig. 2d).
? TROUBLESHOOTING
26| Save the corrected data of each sample in separate text files (not binary) as mean residue ellipticity, [y] (y-axis) as a
function of wavelength (x-axis), or in files with the ellipticity values, which also have information about the starting
wavelength, ending wavelength and data interval so that it can be converted into formats that can be used to estimate
protein conformation.
m CRITICAL STEP Save the data in text (ASCII) format. Binary files are in proprietary formats that can only be accessed by the
correct CD machine and cannot be edited by a text editor or imported into a spreadsheet or converted to formats that can be
used for analysis programs.
27| If data have been collected at different wavelength ranges, e.g. 260–190 nm in 0.1-cm cells and 200–178 nm in 0.01-cm
cells, and the data agree between 200 and 190 nm, combine the data containing the mean residue ellipticity into a single file
using a text editor. If the data do not agree with each other, make sure the proper baseline has been subtracted for each sample
and that the pathlengths of the cells are correct. If they still do not agree, repeat the measurements.
m CRITICAL STEP For data analysis, ensure there is a single entry for each wavelength in the final file, or it will confuse the data
analysis programs.

Data analysis TIMING 2–16 h
28| Analyze the data using appropriate methods, four of which are provided here as (A) through (D). To obtain the most
accurate results, use as many methods as possible.
(A) Evaluating the secondary structure of globular proteins using data collected between 260 and 178 nm

TIMING B10–15 min per spectrum for data conversion, B10–15 min per spectrum for analysis
The CDPro package or the online analysis programs at DicroWeb are recommended for the most accurate estimates of secondary
structure. The DicroProt suite of programs uses older versions of the programs available at DicroWeb and in CDPro, but the results
are similar to those of the more modern versions. It is easy to convert data to both the CDPro and the DicroProt ‘dic’ format,
and the user interface is simple for both programs. A simple program to convert CD data in columns of ellipticity as a function
of wavelength to the dic format, CONVERT.EXE, is in Supplementary Software. Converting data to the formats used by the
CDPro package and analyzing it using SELCON, CONTINLL and CDSSTR will require B10–15 min per spectrum. Converting data
to the dic format and then analyzing it by all the programs in the DicroProt package including VARSLC, SELCON2 and SELCON3,
CONTIN and K2D requires B10–15 min per spectrum.

NATURE PROTOCOLS | VOL.1 NO.6 | 2006 | 2885

 PROTOCOL

(B) Evaluating the secondary structure of globular proteins using data truncated below 260 nm or above 178 nm

TIMING B10 min per spectrum per program
For data sets collected over truncated wavelength ranges (e.g. 240–200 nm), use the programs SELCON, VARSLC, K2D, CONTIN or
LINCOMB (with a database extracted from proteins as references). These programs are all included in Supplementary Software,
which also includes conversion programs that will edit the reference sets so that they match the range of data in the sample
files. Reference sets for LINCOMB are also included. Converting raw CD data and running each program will require B10 min per
spectrum per program, because each of these older versions of the program has a different format for its input files, and the CD
spectrum must be converted separately for each program.


(C) Analyzing the effect of mutations on the helical or b-sheet content of a protein TIMING B20 min per spectrum
Use the constrained method of least squares (e.g., LINCOMB) with a fixed reference set so that the same standards will be used to
analyze the wild-type and mutant proteins. This method should also be used for quantifying the conformational changes in response
to addition of a ligand. Analysis of data using the LINCOMB or MLR programs is slow and takes about 20 min per spectrum, because
© 2007 Nature Publishing Group http://www.nature.com/natureprotocols

the programs are DOS based and use the old DOS graphic screens, which take time to display the results. In addition, one must first
convert the data to the correct format, then run the program and finally examine the results with a text editor.


(D) Analysis of the conformation of polypeptides or short protein fragments TIMING 10–15 min per spectrum
Use LINCOMB (with polypeptide references), CONTIN or K2D. These programs are in Supplementary Software.
? TROUBLESHOOTING

TIMING
Summary of procedure for collecting and analyzing one CD spectrum
Protein sample preparation and concentration determination: 0.5–1 h (can be done in advance)
Machine setup: 0.5 h
Data collection (assuming that three replicate data sets are collected for both the protein sample and the reference between
260 and 185 nm at 0.5 nm intervals with a 0.5 s signal averaging at each point): 0.5 h
Averaging the data, correcting the data for the baseline, smoothing the data and converting the data to mean residue
ellipticity: 10 min
Converting the data to CDPro format and analyzing the data with SELCON3, CONTINLL and CDSSTR: 10 min
Cleaning the cuvette between and after spectra collection: 10 min
When collecting data on multiple samples, the data correction and analysis can be done on the first sample while data on
subsequent samples are being collected. Overall, this procedure can be completed in approximately 2–2.5 h.

? TROUBLESHOOTING
Troubleshooting advice can be found in Table 4.

TABLE 4 | Troubleshooting table.

Step Problem Solution
1(A) Difference spectra have multiple peaks between 280 and 300 nm The protein is not fully denatured by guanidine HCl. Try a different
method.
1(B) Analysis at 280 and 288 nm gives different results The protein is not fully denatured, or the sample has nonprotein
contaminants. Try a different method.
1(C) Standard curves are nonlinear Make fresh Benedict’s reagent or try a lower concentration range for
the standard.
5 The water is not circulating through the sample compartment The inlet or outlet lines may be clogged. Replace the fluid in the tank
to make sure it is clean. Try reversing the flow to the temperature
controller to dislodge any blockage. You may have to have the unit
repaired.
6 The lamp does not turn on Some machines require the lamp base to be warmed up before firing.
Wait a minute and try again. Make sure that the ignition switch is not
stuck. If this does not work, turn off the instrument. Disconnect
the power switch, and check that the lamp is not burned out and that
the leads to the lamp are connected properly. Each machine is
different, so consult the manual. You may need a service call.
7 The machine does not turn on First make sure everything is plugged in. If the machine is
controlled by a computer, make sure that it is turned on and that the
monitor is also turned on. Consult the manual to see how to start the
collection program if it does not start automatically when the
machine is turned on.

2886 | VOL.1 NO.6 | 2006 | NATURE PROTOCOLS

 PROTOCOL

TABLE 4 | Troubleshooting table (continued).

Step Problem Solution
17 The spectrum of the blank has a very large ellipticity (i) Make sure that the cell is not under pressure, caused by, e.g., cell
adapters being too tight or the water pressure being too high if a
water-jacketed cell is being used. (ii) Try recleaning the cell, or try a
different cell. (iii) Check that the light is going directly through the
cell by setting the wavelength to 500 nm and putting a piece of
paper before or after the cell holder to visualize the light path.
! CAUTION Do not look at the light beam in the UV region.
18 The spectrum of the buffer does not agree with that of water Make sure that the buffer does not have any optically active
components, such as glutamate or ATP, and that the buffer is
© 2007 Nature Publishing Group http://www.nature.com/natureprotocols

transparent in the wavelength range of interest.

19 A. Successive scans of the same solution do not overlay A. (i) Allow the solution to equilibrate for a longer time. (ii) Check
each other the spectrum of air to make sure the lamp is stable and the baseline is
not drifting. If there is a problem, contact the manufacturer.
B. There is very low ellipticity B. Check the concentration of the sample. If it was determined by
absorbance at 280 nm or a dye binding technique, it could be too
low. Oxidized DTT and 2-mercaptoethanol absorb at 280 nm, and dye
binding is very dependent on the composition of a protein.
C. The signal-to-noise ratio is very low C. (i) Make sure that the half-bandwidth is set to at least 1 nm and
that the slits are open. (ii) Check the absorbance of the sample + cell
on a spectrometer. Signal-to-noise ratio decreases when the absor-
bance is 41. (iii) Measure the ellipticity of a standard solution of
CSA, and make sure that the instrument is calibrated properly. (iv)
Check the age of the lamp; most lamps begin to deteriorate above
1,000 h of use. (v) Mirrors in CD machines deteriorate with time. If
the machines are old or used heavily, the mirrors in the optics may
require replacement. Call the manufacturer.
D. The dynode voltage does not increase as the wavelengths D. This result may be caused by ‘stray’ light. (i) Make sure that the
decrease, and the CD bands below 200 nm have much lower half-bandwidth is not 42 nm. (ii) Try lowering the half-bandwidth.
magnitude that those at higher wavelength (iii) If the problem persists, contact the manufacturer to make sure
that the slits are functioning properly, the optics are properly aligned
and the light source is focused on the sample.
23 When the smoothed spectra are subtracted from the raw data, Try either lowering or raising the number of points or the degree of
the noise is not distributed evenly around zero the polynomial used for smoothing.
24 The ellipticity is flat between 260 and 250 nm but is greater Sometimes the baselines of CD machines will drift slightly when the
or less than zero machine is first turned on. The data can be corrected for the drift by
adding or subtracting a constant so that the ellipticity between 260
and 250 nm is at zero. For highest accuracy, the spectrum of the
baseline should be taken again to make sure that the displacement is
due to instrumental drift and that the displacement is constant over
the entire wavelength range.
25 The mean residue ellipticity seems much too high or too low. Check that the protein concentration is correct and that the data
(Typically proteins will have ellipticity maxima or minima were divided by the correct pathlength to calculate the mean
between 222 and 200 nm ranging from B5,000 to –45,000 deg residue ellipticity.
cm2 dmol–1 (Fig. 1).
23 The mean residue ellipticity changes as a function of protein The protein may not be monomeric or may be aggregating.
concentration Try to determine the oligomeric state of the protein using
methods such as ultracentrifugation, light scattering or native
gradient gel electrophoresis.
28 When analyzed by the curve fitting programs, the sum The protein concentration probably is not correct. Try a different
of the fractions of each conformation are much larger or method of determining the concentration.
much less than 1.

NATURE PROTOCOLS | VOL.1 NO.6 | 2006 | 2887

 PROTOCOL

TABLE 5 | Analyses of the structure of lysozyme from the CD data in Figure 2.

Fraction of each secondary structure

Basis Path Conc. Wavelength Helix Helix Helix b b b

Method sets Length, cm mg ml1 range, nm regular ends total regular ends total Turns Other
CDPro Packagea
X-Rayb 0.20 0.22 0.42 0.02 0.05 0.07 0.30 0.22

CDSSTR 30 0.01 0.41 260–178 0.19 0.14 0.33 0.11 0.08 0.19 0.20 0.28
CDSSTR 30 0.01 0.83 260–178 0.19 0.14 0.32 0.11 0.08 0.18 0.21 0.28
CDSSTR 30 0.1 0.085 260–185 0.20 0.13 0.34 0.14 0.08 0.22 0.19 0.25
CDSSTR 30 0.1 0.41 260–200 0.20 0.14 0.34 0.12 0.07 0.20 0.20 0.25

CONTIN 30 0.01 0.41 260–178 0.18 0.17 0.35 0.18 0.17 0.35 0.18 0.17
© 2007 Nature Publishing Group http://www.nature.com/natureprotocols

CONTIN 30 0.01 0.83 260–178 0.18 0.17 0.35 0.07 0.07 0.14 0.27 0.23
CONTIN 30 0.1 0.085 260–185 0.16 0.16 0.32 0.10 0.08 0.18 0.25 0.24
CONTIN 30 0.1 0.41 260–200 0.21 0.21 0.42 0.00 0.04 0.04 0.27 0.26

CONTINLL 30 0.01 0.41 260–178 0.19 0.16 0.34 0.08 0.07 0.15 0.23 0.27
CONTINLL 30 0.01 0.83 260–178 0.19 0.17 0.36 0.07 0.06 0.13 0.24 0.27
CONTINLL 30 0.1 0.085 260–185 0.17 0.15 0.32 0.11 0.08 0.18 0.24 0.25
CONTINLL 30 0.1 0.41 260–200 0.18 0.15 0.33 0.12 0.07 0.19 0.22 0.25

SELCON3 30 0.01 0.41 260–178 0.17 0.14 0.30 0.11 0.08 0.19 0.23 0.28
SELCON3 30 0.01 0.83 260–178 0.16 0.14 0.30 0.12 0.08 0.20 0.23 0.27
SELCON3 30 0.1 0.085 260–185 0.16 0.13 0.29 0.11 0.08 0.18 0.23 0.28
SELCON3 30 0.1 0.41 260–200 0.17 0.14 0.31 0.09 0.07 0.16 0.24 0.27

AVERAGE 0.18 0.16 0.34 0.10 0.08 0.17 0.23 0.26

STDEV 0.02 0.02 0.04 0.04 0.03 0.07 0.03 0.03

Other methods Total helix Total b Turns Other

X-RAYc 0.39 0.11 0.34 0.16

Method of constrained least squares fits with peptide referencesd

Peptide set 1 4 0.01 0.41 240–178 0.26 0.29 0.02 0.42
Peptide set 1 4 0.01 0.41 240–200 0.31 0.26 0.02 0.40
Peptide set 2 5 0.01 0.41 240–190 0.24 0.31 0.06 0.38
Peptide set 3 3 0.01 0.41 240–208 0.28 0.15 ND ND
Average 0.27 0.25 0.03 0.40
Stdev 0.03 0.07 0.02 0.02

Method of constrained least squares with references extracted from sets of protein spectrae
Protein set 1 4 0.01 0.41 240–178 0.30 0.22 0.22 0.26
Protein set 2 5 0.01 0.41 240–178 0.27 0.22 0.32 0.20

Methods that use singular value decomposition to deconvolute the reference data setsf
SVD 33 0.01 0.41 260–178 0.27 0.15 0.17 0.25
VARSLC 33 0.01 0.41 260–178 0.28 0.18 0.18 0.27
SELCON 17 0.01 0.41 260–178 0.31 0.18 0.28 0.24
SELCON2 17 0.01 0.41 260–178 0.32 0.16 0.24 0.29

Ridge regressiong
CONTIN 16 0.01 0.41 240–190 0.30 0.30 0.23 0.18

Methods using neural network analysesh

K2D 18 0.01 0.41 240–200 0.33 0.14 ND ND
CDNN 17 0.01 0.41 260–180 0.32 0.20 0.18 0.27
CDNN 17 0.01 0.41 260–200 0.32 0.18 0.17 0.31

Average off all the methods using protein references excluding those in the CDProa package
Average 0.30 0.19 0.22 0.25
Stdev 0.02 0.05 0.05 0.04
aThe CDPro Package36 contains the programs CONTIN (Ref. 27), CONTINLL (Ref. 36), SELCON3 (Ref. 36) and CDSSTR (Ref. 31). bThe secondary structure of lysozyme was determined by the method of Kabsch and
Sander (Ref. 64) and separated into classes of regular a and b helices and helices found at the ends of helical segments, or short fragments36. cThe secondary structure of lysozyme was analyzed to total helices and
turns by the method of Kabsch and Sander (Ref. 64). dThe peptide references sets were 1, a-helix, b-structure, generic turn and other)20; 2, a-helix, b-structure, type 1 turn, type 2 turn, and others21; and 3 poly-L-
lysine in the a-helical, b-structure and extended conformations5. eProtein data sets containing 33 proteins29 or 17 proteins32 were deconvoluted into 5 (a-helix, antiparallel b, parallel b, turn and other) or
4 reference spectra (a-helix, total b-structure, turn and other) by the method of least squares22. fThe programs for extracting data from variations of the method of singular value decomposition analysis were:
SVD (Ref. 17), VARSLC (Ref. 28), SELCON (Ref. 33) and SELCON2 (Ref. 43). gThe data were analyzed by the original CONTIN method using 16 proteins27. hThe neural net programs employed were K2D (Refs. 37,38)
and CDNN.

2888 | VOL.1 NO.6 | 2006 | NATURE PROTOCOLS

 PROTOCOL

ANTICIPATED RESULTS
Typical results for the study of a globular protein (chicken hen’s egg white lysozyme) are given in Figure 2, and the results of
the analysis of the spectra to give the secondary structure are given in Table 5. The data collection of the 10 individual spectra
used to create Figure 2, including washing the cells between measurements, smoothing the data and correcting the data for the
baseline and protein concentration, took B4 h, and the 30 analyses summarized in Table 5 took B12 h, including converting
the data to the appropriate formats, running the calculations and tabulating the results.
All of the methods used to analyze protein spectra should give reasonable estimates of a-helical content (Table 5). The four
programs included in the CDPro package (SELCON3, CONTIN, CONTINLL and CDSSTR) should give results that are almost identical
to each other and relatively independent of the lower limit wavelength range from 200–178 nm (Table 5). Similar secondary
contents should be also obtained with SELCON and SELCON2, CDNN and K2D (Table 5). When proteins are analyzed using the
method of least squares, basis sets extracted from proteins should give good fits, but some peptide data sets tend to
overestimate b-structure and underestimate turns if data lower than 208 nm are fitted5.
© 2007 Nature Publishing Group http://www.nature.com/natureprotocols

Some of the analysis programs output the fitted curve to the raw data. It should be noted that the programs giving the best
fits to the data do not necessarily give the best estimates of protein conformation, because better fits will be obtained when
more variables are used. Therefore, the fits obtained using CONTIN, which fits protein CD data by a large number of reference
spectra, almost always gives a perfect fit to the raw data as compared with methods that use fewer reference sets. Representa-
tive fits using programs with graphical output are shown in Figure 2e and f.

SUPPLEMENTARY MATERIALS
Supplementary materials include equations to estimate secondary structure of proteins fitted them to basis sets using
contrained and unconstrained least squares fits. The equations are in SigmaPlot format in Supplementary Equations.
Supplementary Methods has the procedure for estimating protein concentration using a micro-Kjeldahl procedure.
Supplementary Software has DOS-based versions of SELCON1, SELCON2, MLR, LINCOMB, VARSLC, K2D and the CCA algorithm.
They run under MS-DOS and in a command box under Microsoft Windows 95, 98 and XP. Also provided are programs from
converting data from Aviv, Jasco and text files to the formats needed for each program.

Note: Supplementary information is available via the HTML version of this article. 11. Nafie, L.A. Infrared and Raman vibrational optical activity: theoretical and
experimental aspects. Annu. Rev. Phys. Chem. 48, 357–386 (1997).
ACKNOWLEDGMENTS I thank M.A. Andrade, W.C. Johnson, Jr., N. Sreerama, R.W. 12. Jackson, M. & Mantsch, H.H. The use and misuse of FTIR spectroscopy in the
Woody, S. Venyaminov, and especially the late Gerald D. Fasman for supplying determination of protein structure. Crit. Rev. Biochem. Mol. Biol. 30, 95–120
versions of their software for analyzing CD data for use, analysis and distribution. (1995).
The research was supported by a National Institutes of Health grant, GM-36326 to 13. Cooper, E.A. & Knutson, K. Fourier transform infrared spectroscopy investigations
N.J.G. and to Sarah E. Hitchcock-DeGregori and by the Circular Dichroism Facility of protein structure. Pharm. Biotechnol. 7, 101–143 (1995).
at Robert Wood Johnson Medical School (UMDNJ). 14. Pelton, J.T. & McLean, L.R. Spectroscopic methods for analysis of protein
secondary structure. Anal. Biochem. 277, 167–176 (2000).
COMPETING INTERESTS STATEMENT The author declares that she has no 15. Lowry, O.H., Rosebrough, N.J., Farr, A.L. & Randall, R.J. Protein measurement
competing financial interests. with the Folin phenol reagent. J. Biol. Chem. 193, 265–275 (1951).
16. Bradford, M.M. A rapid and sensitive method for the quantitation of microgram
Published online at http://www.natureprotocols.com quantities of protein utilizing the principle of protein–dye binding. Anal.
Rights and permissions information is available online at http://npg.nature.com/ Biochem. 72, 248–254 (1976).
reprintsandpermissions 17. Hennessey, J.P. Jr. & Johnson, W.C. Jr. Information content in the circular
dichroism of proteins. Biochemistry 20, 1085–1094 (1981).
1. Velluz, L., Legrand, M. & Grosjean, M. Optical Circular Dichroism: Principles, 18. Lang, C.A. Simple microdetermination of Kjeldahl nitrogen in biological material.
Measurements, and Applications (Verlag Chemie Academic Press Inc., New York Anal. Chem. 30, 1692–1694 (1958).
and London, 1965). 19. Jaenicke, L. A rapid micromethod for the determination of nitrogen and
2. Beychok, S. Circular dichroism of biological macromolecules. Science 154, 1288– phosphate in biological material. Anal. Biochem. 61, 623–627 (1974).
1299 (1966). 20. Brahms, S. & Brahms, J. Determination of protein secondary structure in solution
3. Sreerama, N. & Woody, R.W. Computation and analysis of protein circular by vacuum ultraviolet circular dichroism. J. Mol. Biol. 138, 149–178 (1980).
dichroism spectra. Methods Enzymol. 383, 318–351 (2004). 21. Reed, J. & Reed, T.A. A set of constructed type spectra for the practical estimation
4. Holzwarth, G. & Doty, P. The ultraviolet circular dichroism of polypeptides. of peptide secondary structure from circular dichroism. Anal. Biochem. 254,
J. Am. Chem. Soc. 87, 218–228 (1965). 36–40 (1997).
5. Greenfield, N. & Fasman, G.D. Computed circular dichroism spectra for the 22. Saxena, V.P. & Wetlaufer, D.B. A new basis for interpreting the circular dichroic
evaluation of protein conformation. Biochemistry 8, 4108–4116 (1969). spectra of proteins. Proc. Natl Acad. Sci. USA 68, 969–972 (1971).
6. Venyaminov, S., Baikalov, I.A., Shen, Z.M., Wu, C.S. & Yang, J.T. Circular dichroic 23. Chen, Y.H., Yang, J.T. & Chau, K.H. Determination of the helix and b form of
analysis of denatured proteins: inclusion of denatured proteins in the reference proteins in aqueous solution by circular dichroism. Biochemistry 13, 3350–3359
set. Anal. Biochem. 214, 17–24 (1993). (1974).
7. Bovey, F.A. & Hood, F.P. Circular dichroism spectrum of poly-L-proline. 24. Chang, C.T., Wu, C.S. & Yang, J.T. Circular dichroic analysis of protein
Biopolymers 5, 325–326 (1967). conformation: inclusion of the b-turns. Anal. Biochem. 91, 13–31 (1978).
8. Tiffany, M.L. & Krimm, S. Effect of temperature on the circular dichroism spectra 25. Yang, J.T., Wu, C.S. & Martinez, H.M. Calculation of protein conformation from
of polypeptides in the extended state. Biopolymers 11, 2309–2316 (1972). circular dichroism. Methods Enzymol. 130, 208–269 (1986).
9. Woody, R.W. Circular dichroism and conformation of unordered polypeptides. 26. Greenfield, N.J. Methods to estimate the conformation of proteins and
Adv. Biophys. Chem. 2, 31–79 (1992). polypeptides from circular dichroism data. Anal. Biochem. 235, 1–10 (1996).
10. Rucker, A.L. & Creamer, T.P. Polyproline II helical structure in protein unfolded 27. Provencher, S.W. & Glöckner, J. Estimation of globular protein secondary structure
states: lysine peptides revisited. Protein Sci. 11, 980–985 (2002). from circular dichroism. Biochemistry 20, 33–37 (1981).

NATURE PROTOCOLS | VOL.1 NO.6 | 2006 | 2889

 PROTOCOL

28. Manavalan, P. & Johnson, W.C. Jr. Variable selection method improves the 46. Safar, J., Roller, P.P., Gajdusek, D.C. & Gibbs, C.J. Jr. Thermal stability and
prediction of protein secondary structure from circular dichroism spectra. Anal. conformational transitions of scrapie amyloid (prion) protein correlate with
Biochem. 167, 76–85 (1987). infectivity. Protein Sci. 2, 2206–2216 (1993).
29. Toumadje, A., Alcorn, S.W. & Johnson, W.C. Jr. Extending CD spectra of proteins to 47. Greenfield, N.J. Using circular dichroism, collected as a function of temperature,
168 nm improves the analysis for secondary structures. Anal. Biochem. 200, 321– to determine the thermodynamics of protein unfolding and binding interactions.
331 (1992). Nat. Protoc. doi: 10.1038/nprot.2006.204 (2006).
30. Toumadje, A. & Johnson, W.C. Jr. Effects of relative band intensity on prediction 48. Adler, A.J., Greenfield, N.J. & Fasman, G.D. Circular dichroism and optical
of protein secondary structure from CD. Anal. Biochem. 211, 258–260 (1993). rotatory dispersion of proteins and polypeptides. Methods Enzymol. 27, 675–735
31. Johnson, W.C. Analyzing protein circular dichroism spectra for accurate secondary (1973).
structures. Proteins 35, 307–312 (1999). 49. Rosenkranz, H. Circular dichroism of helical and nonhelical proteins. A review of
32. Sreerama, N. & Woody, R.W. A self-consistent method for the analysis of protein the limits of the methods for calculating secondary structure. Z. Klin. Chem. Klin.
secondary structure from circular dichroism. Anal. Biochem. 209, 32–44 (1993). Biochem. 12, 222 (1974).
33. Sreerama, N. & Woody, R.W. Protein secondary structure from circular dichroism 50. Johnson, W.C. Jr. Protein secondary structure and circular dichroism: a practical
spectroscopy. Combining variable selection principle and cluster analysis with guide. Proteins 7, 205–214 (1990).
neural network, ridge regression and self-consistent methods. J. Mol. Biol. 242, 51. Woody, R.W. Circular dichroism. Methods Enzymol. 246, 34–71 (1995).
497–507 (1994). 52. Kelly, S.M., Jess, T.J. & Price, N.C. How to study proteins by circular dichroism.
© 2007 Nature Publishing Group http://www.nature.com/natureprotocols

34. Sreerama, N., Venyaminov, S.Y. & Woody, R.W. Estimation of the number of a- Biochim. Biophys. Acta 1751, 119–139 (2005).
helical and b-strand segments in proteins using circular dichroism spectroscopy. 53. Miles, A.J. & Wallace, B.A. Synchrotron radiation circular dichroism spectroscopy
Protein Sci. 8, 370–380 (1999). of proteins and applications in structural and functional genomics. Chem. Soc.
35. Sreerama, N., Venyaminov, S.Y. & Woody, R.W. Estimation of protein secondary Rev. 35, 39–51 (2006).
structure from circular dichroism spectra: inclusion of denatured proteins with 54. Venyaminov, S.Y. & Yang,, J.T. in Circular Dichroism and the Conformational
native proteins in the analysis. Anal. Biochem. 287, 243–251 (2000). Analysis of Biomolecules (ed. Fasman, G.D.) (Plenum Press, New York and London,
36. Sreerama, N. & Woody, R.W. Estimation of protein secondary structure from 1996).
circular dichroism spectra: comparison of CONTIN, SELCON, and CDSSTR methods 55. Greenfield, N.J. Determination of the folding of proteins as a function of
with an expanded reference set. Anal. Biochem. 287, 252–260 (2000). denaturants, osmolytes or ligands using circular dichroism. Nat. Protoc.
37. Böhm, G., Muhr, R. & Jaenicke, R. Quantitative analysis of protein far UV circular doi: 10.1038/nprot.2006.229 (2006).
dichroism spectra by neural networks. Protein Eng. 5, 191–195 (1992). 56. Greenfield, N.J. The use of circular dichroism to study the kinetics of protein
38. Andrade, M.A., Chacón, P., Merelo, J.J. & Morán, F. Evaluation of secondary folding and unfolding. Nat. Protoc. doi: 10.1038/nprot.2006.244 (2006).
structure of proteins from UV circular dichroism spectra using an unsupervised 57. Miles, A.J., Whitmore, L. & Wallace, B.A. Spectral magnitude effects on the
learning neural network. Protein Eng. 6, 383–390 (1993). analyses of secondary structure from circular dichroism spectroscopic data.
39. Unneberg, P., Merelo, J.J., Chacón, P. & Morán, F. SOMCD: method for evaluating Protein Sci. 14, 368–374 (2005).
protein secondary structure from UV circular dichroism spectra. Proteins 42, 58. Edelhoch, H. Spectroscopic determination of tryptophan and tyrosine in proteins.
460–470 (2001). Biochemistry 6, 1948–1954 (1967).
40. Perczel, A., Park, K. & Fasman, G.D. Analysis of the circular dichroism spectrum of 59. Mihalyi, E. Numerical values of the absorbances of the aromatic amino
proteins using the convex constraint algorithm: a practical guide. Anal. Biochem. acids in acid, neutral, and alkaline solutions. J. Chem. Eng. Data 13, 179–182
203, 83–93 (1992). (1968).
41. Chen, Y.H. & Yang, J.T. A new approach to the calculation of secondary structures 60. Goa, J. A micro biuret method for protein determination. Scand. J. Clin. Lab.
of globular proteins by optical rotatory dispersion and circular dichroism. Invest. 5, 218–222 (1953).
Biochem. Biophys. Res. Commun. 44, 1285–1291 (1971). 61. Konno, T. Conformational diversity of acid-denatured cytochrome c studied by a
42. Chen, Y.H., Yang, J.T. & Martinez, H.M. Determination of the secondary structures matrix analysis of far-UV CD spectra. Protein Sci. 7, 975–982 (1998).
of proteins by circular dichroism and optical rotatory dispersion. Biochemistry 11, 62. Melberg, S.G. & Johnson, W.C. Jr. Changes in secondary structure follow the
4120–4131 (1972). dissociation of human insulin hexamers: a circular dichroism study. Proteins 8,
43. Sreerama, N. & Woody, R.W. Poly(pro)II helices in globular proteins: identification 280–286 (1990).
and circular dichroic analysis. Biochemistry 33, 10022–10025 (1994). 63. Savitsky, A. & Golay, M.J.E. Smoothing and dIfferentiation of data by simplified
44. Perczel, A., Hollósi, M., Tusnády, G. & Fasman, G.D. Convex constraint analysis: a least squares procedures. Anal. Chem. 36, 1627–1639 (1964).
natural deconvolution of circular dichroism curves of proteins. Protein Eng. 4, 64. Bentz, H., Bächinger, H.P., Glanville, R. & Kühn, K. Physical evidence for the
669–679 (1991). assembly of A and B chains of human placental collagen in a single triple helix.
45. Greenfield, N.J. & Hitchcock-DeGregori, S.E. Conformational intermediates in the Eur. J. Biochem. 92, 563–567 (1978).
folding of a coiled-coil model peptide of the N-terminus of tropomyosin and 65. Steiner, R.F. Structural transitions of lysozyme. Biochim. Biophys. Acta 79, 51–63
a-tropomyosin. Protein Sci. 2, 1263–1273 (1993). (1964).

2890 | VOL.1 NO.6 | 2006 | NATURE PROTOCOLS