Biophysical Chemistry 133 (2008) 66 – 70

http://www.elsevier.com/locate/biophyschem

Predicting ultraviolet spectrum of single stranded and double stranded

deoxyribonucleic acids
Andrey V. Tataurov, Yong You, Richard Owczarzy ⁎
Integrated DNA Technologies, 1710 Commercial Park, Coralville, IA-52241, USA
Received 10 December 2007; accepted 15 December 2007
Available online 23 December 2007

Abstract

Synthetic oligodeoxynucleotides are widely used in many biological, biochemical and biophysical applications. The concentration,
composition and structure of DNA are often determined from its ultraviolet spectrum. Although parameters for use with the nearest-neighbor
model for prediction of extinction coefficients of single stranded DNAs at 260 nm were published some time ago, similar parameters for other
wavelengths or for use with DNA duplexes have not been reported. Practical formulae and parameters for prediction of UV spectra,
hypochromism and peak wavelengths were experimentally determined for both single stranded and double stranded oligodeoxynucleotides in the
range from 215 to 310 nm. The accuracy of predictions made using the nearest-neighbor model and the base composition model was determined
and compared. The spectrum of any DNA oligomer can be calculated using a Microsoft Excel® application that is available in the Appendix A.
© 2007 Elsevier B.V. All rights reserved.

Keywords: Extinction coefficient; Oligodeoxynucleotides; Hypochromism; Nearest-neighbor model; DNA; Ultraviolet spectrum

1. Introduction published [3] although this value is crucial for determination of

concentration of double stranded DNA oligomers.
Near ultraviolet spectra of DNAs and RNAs are often In this paper, the UV spectra of a large number (~200) of
collected to determine concentrations [1–4], base composition synthetic DNAs were analyzed. We report methods and empirical
[5,6] and secondary structure of nucleic acids [7,8]. Published parameters for prediction of: (1) duplex hypochromism at
models and parameters allow calculations of the oligonucleotide 260 nm; (2) the wavelength of maximum absorbance; and (3)
extinction coefficient, ɛ260, at 260 nm [1,3]; a complete set of extinction coefficients for any single stranded or double stranded
nearest-neighbor (n-n) parameters for other wavelengths has not DNA oligomers in the near UV region with 1 nm resolution.
been reported.
The extinction coefficient of double stranded DNA (dsDNA) 2. Experimental and theoretical methods
is less than the sum of the extinction coefficients for the
composing two single stranded DNAs (ssDNA) [9]. This 2.1. Base composition model
hypochromic effect of nucleotides is attributed to dipole-induced
dipole interactions between transition moments in stacked Several models have been proposed that relate an intensive
neighboring bases [10]. Empirical parameters for accurate physical property, P, of an oligonucleotide to its structure and
predictions of oligonucleotide hypochromism have not been base sequence [11,12]. The base composition model (zeroth-
neighbor model) takes into account the number or frequency of
occurrences of bases, but neglects interactions between neighbor
⁎ Corresponding author. Tel.: +1 319 626 8459; fax: +1 319 626 9621. bases or base pairs,
E-mail address: [email protected] (R. Owczarzy). X
URL's: http://biophysics.idtdna.com, http://www.owczarzy.net P¼ fi P i ð1Þ
(R. Owczarzy). i¼A;C;G;T

0301-4622/$ - see front matter © 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.bpc.2007.12.004
 A.V. Tataurov et al. / Biophysical Chemistry 133 (2008) 66–70 67

This sum is computed over all kinds of bases present in single were obtained. In our analysis, the design matrices had no
strands or over all types of base pairs in double stranded nucleic singular values indicating that all parameters Pij were unique
acids. Properties of each base or base pair, Pi, are determined and no parameter could be obtained from a linear combination of
from fit to experimental data. The fractions of bases, fi, are the remaining parameters [13]. The SVD was implemented
calculated from the oligonucleotide sequence, using Excel Add-in, Matrix.xla package, version 2.3 (Foxes
Team, Leonardo Volpi, http://digilander.libero.it/foxes).
N Ni
fi ¼ Pi ¼ ð2Þ The accuracy of predictions was statistically compared using
Ni Nb reduced χ2r and the root mean square deviation (rmsd) between
i¼A;C;G;T
predicted and experimentally measured values [14].
Ni is the number of i type base in the oligonucleotide sequence. If
the physical property P is extensive (e.g., hybridization free 2.4. Experimental methods
energy, extinction coefficient) the frequencies fi are replaced
with the number of bases in Eq. (1) [11]. All spectra were recorded at 25 °C in 1.000 ± 0.001 cm
pathlength quartz cuvettes using a single-beam, diode-array
2.2. Nearest-neighbor model Hewlett-Packard HP8453 spectrophotometer. Absorbance and
wavelength calibration was regularly verified with reference
The nearest-neighbor (first-neighbor) model broadens the dichromate and holmium oxide solutions (Starna Cells, Atasca-
base composition model and takes into account interactions dereo, CA). Temperature was controlled by a Peltier cuvette
between adjacent bases or base pairs. However, any interactions holder. Absorbance data at 1 nm intervals were processed using
over a longer range (e.g., between next nearest-neighbor bases) UV–Visible ChemStation and Excel 2003. This work employed
are neglected. The intensive physical property P is obtained by the sets of DNA oligomers from a previous study that explored the
summing properties of all types of nearest-neighbor subunits effects of sodium cations on Tm [15]. The experimental data set of
(doublets) present in the nucleic acid (e.g., AA, AC, AG), single stranded oligodeoxynucleotides consisted of 202
sequences ranging in length from 6 to 30 bases and in GC
X
P¼ fij Pij ð3Þ content from 0 to 80%. The spectra of these single strands were
i;j¼A;C;G;T measured in a low salt buffer (0.1 mM sodium phosphate, 0.5 mM
NaCl, 0.01 mM Na2EDTA, pH = 7.0) while strand concentrations
When DNA is composed from four native bases, there are 20 and ranged from 0.5 to 11 μM.
12 unique Pij parameters for single stranded and double stranded The data set of DNA duplex oligomers contained 86 unique
oligomers, respectively [11]. The nearest-neighbor parameters sequences ranging in length from 10 to 60 base pairs and in GC
include initiation (end) interactions, which are denoted using a content from 20 to 80%. The UV spectra of duplexes were
fictitious base “E” for the end of an oligonucleotide [12]. collected in a high salt buffer (100 mM NaCl, 10 mM sodium
We employed both the nearest-neighbor and base composition phosphate, 1 mM Na2EDTA, pH = 7.0, total [Na+] = 119 mM)
models and evaluated their ability to predict accurately various where all duplexes were stable. The helix–coil transitions and
features of UV spectra for single stranded and double stranded stabilities of all duplexes were previously determined using UV
oligodeoxynucleotides. The base composition model was pre- melting experiments [15].
ferred because of the lower number of parameters. However, if
these parameters did not provide accurate predictions, the nearest- 3. Results and discussion
neighbor parameters are reported.
3.1. Hypochromism of DNA
2.3. Determination of parameters
When two single stranded oligodeoxynucleotides, S1 and
Once the property P is measured for a set of oligonucleotides, S2, hybridize and form a DNA duplex, D, absorbance decreases
Eqs. (1) and (3) are constructed for each sequence. To decrease [9]. The extinction coefficient of the duplex, εD, can be
the impact of experimental errors on parameters there are usually estimated from the extinction coefficients of both strands, εS1,
more sequences than unknown parameters, i.e., the resulting εS2, and hypochromicity, h,
system of linear equations is overdetermined,
eD ¼ ð1  hÞ  ðeS1 þ eS2 Þ ð5Þ
Y Y
F  P ij ¼ P exp ð4Þ The value of 1 − h is the fraction of absorbance that remains
Y Y when the single strands anneal to form the duplex,
P ij is the column vector of parameters and P exp is the column
vector of the experimentally measured properties, one for each AD
1h¼ ð6Þ
sequence. The design matrix F has a row for each oligonucleo- ðAS1 VS1 þ AS2 VS2 Þ=ðVS1 þ VS2 Þ
tide and a column for each parameter (base or nearest-neighbor
unit). The overdetermined Eq. (4) was solved using the singular where AD is the absorbance of the duplex solution. AS1, AS2 are
value decomposition method (SVD), so that the squares of absorbance values of the single strand S1 and S2 solutions,
residuals were minimized [13]. Least-squared fitted parameters respectively. Volumes VS1 and VS2 of S1 and S2 strand solutions
68 A.V. Tataurov et al. / Biophysical Chemistry 133 (2008) 66–70

were mixed to make the duplex sample. Absorbance of 86 Table 1

duplexes and composing single strand solutions were measured at Parameters used in Eq. (8)
260 nm, neutral pH and at constant temperature of 25 °C. Parameter λA λC λG λT
Hypochromicity was calculated from Eq. (6). Since hypochro- Peak wavelength (nm) 256.7 265.7 247.3 265.1
mism originates from interactions between neighbor bases, it is
not surprising that hypochromism is sequence dependent. Fig. 1
shows that hypochromism decreases with increasing GC content.
A similar dependence was reported earlier [16], however, their stranded DNA sequence. The rmsd value of 0.7 nm was observed
hypochromism was calculated from the difference between absor- for the set of 202 sequences, which is similar to the resolution of
bance values at low and high temperatures [9]. We determine the spectrophotometer (1 nm). The peak wavelengths in Table 1
hypochromism at a constant temperature. The values of h at 260 nm are from 2 to 6 nm smaller than the values observed for
were fit to base composition model, where hypochromicity is pro- deoxynucleotide monophosphates [18]. A similar hypsochromic
portional to the fraction of GC and AT base pairs, shift was reported for AAA trinucleoside diphosphate [19].
hð260nmÞ ¼ fAT  0:287 þ fGC  0:059 ð7Þ 3.3. Prediction of ultraviolet spectrum
The 29% hypochromicity for dA − dT base pair is similar to 22–
24% hypochromicity values observed for rAnrUn duplexes (n = It has been well established that both absorbance and
4−7) [17]. extinction coefficients, ε, of DNA oligomers vary with
Since experimental errors of hypochromicity are about 0.02, wavelength [1]. These properties are predicted accurately from
the nearest-neighbor model did not significantly improve the nearest-neighbor model. The n-n ε parameters have been
accuracy of hypochromicity predictions. The rmsd of 0.014 previously published for single stranded oligonucleotides at
for the n-n model is comparable with rmsd of 0.02 obtained from 260 nm [1,3]. Using these parameters, the extinction coefficient
Eq. (7). This result is consistent with measurements of genomic is traditionally calculated from the following formula [1–3],
DNA samples where hypochromicity has been approximated b 1 b 1
X
N X
N
reasonably well by a linear function of GC content [9]. e260 ¼ ei;iþ1  ei ð9Þ
i¼1 i¼2

3.2. Prediction of peak wavelength The εi,i + 1 are the extinction coefficients for doublets of
nucleotides, the εi is the extinction coefficient for the nucleotide
The peak wavelength of single stranded DNA oligomers, i, and Nb is the number of nucleotides in the oligomer sequence.
λpeak, where absorbance reaches local maximum is sequence The second term of Eq. (9) subtracts extinction coefficients of
dependent. Both the nearest-neighbor model [12] and the base internal bases, which are counted twice in the first term. The
composition model predicted the peak wavelength with similar parameters are listed in the first two columns of Table 2. The
accuracy. The peak wavelength can therefore be computed format of Eq. (9) differs from the usual format of the nearest-
using the base composition model, neighbor model, Eq. (3) [11],
X
kpeak ¼ fA  kA þ fC  kC þ fG  kG þ fT  kT ð8Þ e260 ¼ Nij enn
ij ð10Þ
i;j¼A;C;G;T;E
where fA, fC, fG and fT are the fractional contents of adenine,
cytosine, guanine and thymine bases, respectively. Eq. (8) and the The initiation (end) interactions (EA, AE, EG, etc.) present at
four parameters shown in Table 1 apply to any native single each oligonucleotide terminus are included in Eq. (10). The
parameters of Eq. (9) can be converted to parameters of the
simpler Eq. (10). The converted parameters are listed in the last
two columns of Table 2. Eqs. (9) and (10) are mathematically
equivalent and provide the same extinction coefficient for any
oligodeoxynucleotide sequence. However, Eq. (10) better
demonstrates that the total number of nearest-neighbor units
(doublets) is Nb + 1 for an oligonucleotide that is Nb bases long,
X
Nij ¼ Nb þ 1 ð11Þ
i;j¼A;C;G;T;E

The extinction coefficient, ελ, at any wavelength λ can be

calculated from ε260 and the ratio of absorbance at wavelength λ
to the absorbance at 260 nm,

Ak
k ¼
R260 ð12Þ
Fig. 1. Hypochromicity values decrease with increasing GC base pair content. A260
 A.V. Tataurov et al. / Biophysical Chemistry 133 (2008) 66–70 69

ek ¼ R260
k  e260 ð13Þ

This method allows easy updating of ελ values whenever more

accurate parameters for ε260 are introduced in the literature [3].
We have measured experimentally the ratios Rλ260 for both
single stranded and double stranded DNA oligomers. The
nearest-neighbor model was found to provide a substantially
better fit of these ratios (rmsd of Rλ260 = 0.007 for dsDNA) than
the base composition model (rmsd of Rλ260 = 0.011) for both
ssDNAs and dsDNAs. The ratio Rλ260 is an intensive physical
variable and depends therefore on the fraction of specific
nearest-neighbor doublets,
X
k ¼
R260 fij Rnn
ij ð14Þ
i;j¼A;C;G;T;E Fig. 2. Comparison of predicted (solid lines) and experimentally measured
(symbols) spectra for two single stranded DNA oligomers. The GC base pair
content of sequences is 20% (open circle) and 80% (open triangle).
where the fraction of a particular nearest-neighbor doublet is
defined as,

N Nij errors. Since negative Rλ260 are not observed experimentally,

fij ¼ Pij ¼ ð15Þ these negative values should be set to zero. Fig. 2 demonstrates
Nij Nb þ 1
i;j¼A;C;G;T;E
the accuracy of prediction for two oligodeoxynucleotides of very
different base composition. Excellent agreement between
The Rijn-n are sequence dependent nearest-neighbor parameters experimentally measured and predicted spectra is seen. When
for doublets of nucleotides, e.g. AA, TG, EA. Twenty and twelve Rλ260 values were predicted for our dataset using the n-n pa-
unique parameters were determined for ssDNA and dsDNA, rameters, overall root mean square deviations of 0.013 and 0.007
respectively, at each wavelength. These Rijn-n values are reported were observed for single stranded and double stranded DNAs,
in the Appendix A. Some bases (e.g., dA) do not absorb above respectively.
295 nm. Under these conditions their Rijn-n parameters and Eq.
(14) may predict small negative Rλ260 values due to experimental 3.4. Software overview

The Excel spreadsheet provided in the Appendix A can be

used to calculate the UV spectrum of any oligodeoxynucleotide.
Table 2 Visual Basic for Applications macros must be enabled in order to
Two equivalent sets of parameters for the extinction coefficient of use these functions (set security to medium level in the Excel
oligodeoxynucleotides at 260 nm submenu Tools\Macro\Security and restart the Excel program).
Traditional format (Eq. (9)) New format (Eq. (10)) Sequences should be entered in column B. The top part of the
i,j εi,j (260 nm) ij εn-n
sheet calculates the peak wavelength and the extinction
ij (260 nm)
−1 −1
coefficients at 260 nm. The bottom part of the sheet predicts the
(5′ to 3′) (10 M 3
cm ) (5′ to 3′) (103 M− 1 cm− 1)
entire UV spectrum for a single sequence entered in the cell B27.
A 15.4 EA, AE 7.70 Sheets “SSData” and “DuplexData” contain tables of nearest-
C 7.4 EC, CE 3.70
neighbor parameters for single stranded and double stranded
G 11.5 EG, GE 5.75
T 8.7 ET, TE 4.35 DNA oligomers, respectively. Available custom functions are
AA 27.4 AA 12.00 described in the sheet “Help”.
AC 21.2 AC 9.80
AG 25.0 AG 11.55 4. Conclusion
AT 22.8 AT 10.75
CA 21.2 CA 9.80
CC 14.6 CC 7.20 This paper present methods, experimentally determined pa-
CG 18.0 CG 8.55 rameters and easy-to-use software that calculates the ultraviolet
CT 15.2 CT 7.15 spectrum of deoxyribonucleic acids. Spectra of both single
GA 25.2 GA 11.75 stranded and double stranded DNA oligomers were shown to be
GC 17.6 GC 8.15
accurately predicted.
GG 21.6 GG 10.10
GT 20.0 GT 9.90
TA 23.4 TA 11.35 Appendix A. Supplementary data
TC 16.2 TC 8.15
TG 19.0 TG 8.90 Supplementary data associated with this article can be found,
TT 16.8 TT 8.10
in the online version, at doi:10.1016/j.bpc.2007.12.004.
70 A.V. Tataurov et al. / Biophysical Chemistry 133 (2008) 66–70

References [10] I. Tinoco Jr., Hypochromism in polynucleotides, J. Am. Chem. Soc. 82 (1960)
4785–4790.
[1] C.R. Cantor, M.M. Warshaw, H. Shapiro, Oligonucleotide interactions. III. [11] D.M. Gray, Derivation of nearest-neighbor properties from data on nucleic
Circular dichroism studies of the conformation of deoxyoligonucleotides, acid oligomers. I. Simple sets of independent sequences and the influence
Biopolymers 9 (1970) 1059–1077. of absent nearest neighbors, Biopolymers 42 (1997) 783–793.
[2] G. Kallansrud, B. Ward, A comparison of measured and calculated single- [12] R.F. Goldstein, A.S. Benight, How many numbers are required to specify
and double-stranded oligodeoxynucleotide extinction coefficients, Anal. sequence-dependent properties of polynucleotides? Biopolymers 32 (1992)
Biochem. 236 (1996) 134–138. 1679–1693.
[3] M.J. Cavaluzzi, P.N. Borer, Revised UV extinction coefficients for nucleo- [13] W.H. Press, S.A. Teukolsky, W.T. Vetterling, B.P. Flannery, Numerical
side-5′-monophosphates and unpaired DNA and RNA, Nucleic Acids Res. Recipes in Fortran 77: The Art of Scientific Computing, Cambridge
32 (2004) e13. University Press, Cambridge, UK, 1992.
[4] G.D. Fasman, Handbook of Biochemistry and Molecular Biology, vol. I, [14] P.R. Bevington, D.K. Robinson, Data Reduction and Error Analysis for the
CRC Press, Cleveland, 1975, p. 589. Physical Sciences, WCB/McGraw-Hill, Boston, 1992.
[5] J.T. Henderson, A.S. Benight, S. Hanlon, A semi-micromethod for the [15] R. Owczarzy, Y. You, B.G. Moreira, J.A. Manthey, L. Huang, M.A.
determination of the extinction coefficients of duplex and single-stranded Behlke, J.A. Walder, Effects of sodium ions on DNA duplex oligomers:
DNA, Anal. Biochem. 201 (1992) 17–29. improved predictions of melting temperatures, Biochemistry 43 (2004)
[6] R.D. Blake, T.G. Hydorn, Spectral analysis for base composition of DNA 3537–3554.
undergoing melting, J. Biochem. Biophys. Methods 11 (1985) 307–316. [16] V.I. Danilov, On the origin of the hypochromic effect in double-stranded
[7] J.-L. Mergny, J. Li, L. Lacroix, S. Amrane, J.B. Chaires, Thermal polynucleotides, FEBS Lett. 47 (1974) 155–157.
difference spectra: a specific signature for nucleic acid structures, Nucleic [17] K.J. Breslauer, J.M. Sturtevant, I. Tinoco Jr., Calorimetric and spectro-
Acids Res. 33 (2005) e138. scopic investigation of the helix-to-coil transition of a ribo-oligonucleotide:
[8] J. Jaumot, N. Escaja, R. Gargallo, C. Gonzales, E. Pedroso, R. Tauler, rA7U7, J. Mol. Biol. 99 (1975) 549–565.
Multivariate curve resolution: a powerful tool for the analysis of conforma- [18] C.A. Sprecher, W.C. Johnson Jr., Circular dichroism of the nucleic acid
tional transitions in nucleic acids, Nucleic Acids Res. 30 (2002) e92. monomers, Biopolymers 16 (1977) 2243–2264.
[9] G. Felsenfeld, S.Z. Hirschman, A neighbor-interaction analysis of the [19] C.R. Cantor, I. Tinoco Jr., Absorption and optical rotary dispersion of
hypochromism and spectra of DNA, J. Mol. Biol. 13 (1965) 407–427. seven trinucleoside diphosphates, J. Mol. Biol. 13 (1965) 65–77.