What is the conformation of physiologically-active dinucleoside polyphosphates in solution? Conformational analysis of free dinucleoside polyphosphates by NMR and molecular dynamics simulations

Article abstract of DOI:10.1039/c005122e

Dinucleoside polyphosphates, or dinucleotides (Np_nN′; N, N′ = A, U, G, C; n = 2-7), are naturally occurring ubiquitous physiologically active compounds. Despite the interest in dinucleotides, and the relevance of their conformation to their biological function, the conformation of dinucleotides has been insufficiently studied. Therefore, here we performed conformational analysis of a series of Np_nN′ Na⁺ salts (N = A, G, U, C; N′ = A, G, U, C; n = 2-5) by various NMR techniques. All studied dinucleotides, except for Up_4/5U, formed intramolecular base stacking interactions in aqueous solutions as indicated by NMR. The conformation around the glycosidic angle in Np_nN′s was found to be anti/high anti and the preferred conformation around the C4′-C5′, C5′-O5′ bonds was found to be gauche-gauche (gg). The ribose moiety in Np_nN′s showed a small preference for the S conformation, but when attached to cytosine the ribose ring preferred the N conformation. However, no predominant conformation was observed for the ribose moiety in any of the dinucleotides. Molecular dynamics simulations of Ap₂A and Ap₄A Na⁺ salts supported the experimental results. In addition, three modes of base-stacking were found for Ap_2/4A: α-α, β-β and α-β, which exist in equilibrium, while none is dominant. We conclude that natural, free Np_nN′s (n = 2-5) at physiological pH exist mostly in a folded (stacked), rather than extended conformation, in several interconverting stacking modes. Intramolecular base stacking of Np_nN′s does not alter the conformation of each of the nucleotide moieties, which remains the same as that of the mononucleotides in solution.

Full text of DOI:10.1039/c005122e

View Online / Journal Homepage / Table of Contents for this issue
PAPER
www.rsc.org/obc | Organic & Biomolecular Chemistry
What is the conformation of physiologically-active dinucleoside
polyphosphates in solution? Conformational analysis of free dinucleoside
polyphosphates by NMR and molecular dynamics simulations†
Noa Stern, Dan Thomas Major, Hugo Emilio Gottlieb, Daniel Weizman and Bilha Fischer*
Received 15th April 2010, Accepted 24th June 2010
DOI: 10.1039/c005122e
Dinucleoside polyphosphates, or dinucleotides (NpnN¢; N, N¢ = A, U, G, C; n = 2–7), are naturally
occurring ubiquitous physiologically active compounds. Despite the interest in dinucleotides, and the
relevance of their conformation to their biological function, the conformation of dinucleotides has been
insufficiently studied. Therefore, here we performed conformational analysis of a series of NpnN¢ Na+
salts (N = A, G, U, C; N¢ = A, G, U, C; n = 2–5) by various NMR techniques. All studied dinucleotides,
except for Up4/5U, formed intramolecular base stacking interactions in aqueous solutions as indicated
by NMR. The conformation around the glycosidic angle in NpnN¢s was found to be anti/high anti and
the preferred conformation around the C4¢–C5¢, C5¢–O5¢ bonds was found to be gauche–gauche (gg).
The ribose moiety in NpnN¢s showed a small preference for the S conformation, but when attached to
cytosine the ribose ring preferred the N conformation. However, no predominant conformation was
observed for the ribose moiety in any of the dinucleotides. Molecular dynamics simulations of Ap2A
and Ap4A Na+ salts supported the experimental results. In addition, three modes of base-stacking were
found for Ap2/4A: a–a, b–b and a–b, which exist in equilibrium, while none is dominant. We conclude
that natural, free NpnN¢s (n = 2–5) at physiological pH exist mostly in a folded (stacked), rather than
extended conformation, in several interconverting stacking modes. Intramolecular base stacking of
NpnN¢s does not alter the conformation of each of the nucleotide moieties, which remains the same as
that of the mononucleotides in solution.
processes. Thus, the Ap5A–Zn2+ complex was used to reveal
preferred pathways that create the configuration capable of
Introduction
proficient chemistry in the catalysis of adenylate kinase.8
One of the physiological roles of both purine and pyrimidine
dinucleotides is extracellular signaling mediated by the ubiqui-
tous nucleotide P2-receptors.4,6 Dincleotides may also bind to
adenosine receptors, P1 receptors.1,2 Furthermore, dinucleotides
are attractive as therapeutic agents due to their increased chem-
ical and metabolic stability as compared to the corresponding
mononucleotide P2-receptor ligands.9 For example, several NpnN¢
analogues have been administered in human clinical trials, in-
cluding Ap4A, for lowering blood pressure during anesthesia,
and Up4U and Up4dC (P1-(uridine 5¢-)-P4-(2¢-deoxycytidine 5¢-)
tetraphosphate) for the treatment of dry eye disease and cystic
fibrosis, respectively.5,9
Although there is a significant volume of data on the phar-
macology and biochemistry of ApnAs (GpnGs)2,3 and despite the
increasing interest in the members of the dinucleotide family
as therapeutic agents, the current information on their physico-
chemical properties is insufficient. For instance, the actual con-
formation of sodium or potassium salts of dinucleotides, in a
physiological medium, was scarcely determined experimentally.
Such data are required for the understanding of the structure–
activity relationship of either natural or synthetic dinucleotides
and the design of potent therapeutic agents based on a dinucleotide
scaffold.
Dinucleoside polyphosphates, or dinucleotides (NpnN¢; N, N¢ =
A, U, G, C; n = 2–7) Fig. 1, are naturally occurring ubiquitous
compounds. They consist of a polyphosphate chain linked to a
nucleoside, at the 5¢-position, on both edges. Many important
functions, both intra- and extracellular, are suggested for ApnAs,
ApnGs and GpnGs, including involvement in proliferation pro-
cesses, in conditions of cellular stress, regulation of enzymes,
neurotransmission, platelet disaggregation and modulation of
vascular tone.1–6 NpnN¢s may also function as inhibitors of
enzymatic reactions, acting as competitive or noncompetitive
inhibitors or as alternative substrates. For example, Ap5A is
an inhibitor of adenylate kinase (Ki = 2.9 ¥ 10-8 M).7 Various
dinucleotides primarily inhibit enzymes from the kinase family,
but also IMP dehydrogenase, endoribonuclease, adenylsuccinate
synthetase, poly(ADP-ribose) polymerase, etc.7 In some cases,
dinucleotides may act as positive effectors, wherein an NpnN¢
binds to an enzyme (such as GMP reductase or AMP deaminase),
thus increasing the activity rate of the enzyme.7 In addition,
dinucleotides, such as Ap4A, may be used as primers for DNA
synthesis by DNA polymerase or for RNA polymerase.7 Moreover,
dinucleotides were also used as probes for analyzing biochemical
Department of Chemistry and the Lise Meitner-Minerva Center of Com-
putational Quantum Chemistry, Bar Ilan University, Ramat-Gan, 52900,
Israel; Fax: +972-3-6354907; Tel: +972-3-5318303. E-mail: bfischer@
mail.biu.ac.il
† Electronic supplementary information (ESI) available: Experimental
details. See DOI: 10.1039/c005122e
Previous conformational studies on dinucleotides in solution,
mainly on ApnAs, focused on the base stacking interactions using
CD, UV and 1H NMR.10–14 Yet the mode of p-stacking in ApnAs
This journal is
The Royal Society of Chemistry 2010
Org. Biomol. Chem., 2010, 8, 4637–4652 | 4637
©

View Online
with potential of mean force (PMF) calculations. The conforma-
tional analysis by both NMR and theoretical calculations included
parameters such as ribose puckering (North↔South equilibrium),
glycosidic bond angle (c), a, b and g angles (e.g., Fig. 3) and the
mode of intramolecular p-stacking interactions (i.e. a–a/b–b/a–
b mode) between the nucleobases. Investigating the conformations
of the dinucleotides in solution is a prerequisite for understanding
their activity as enzyme inhibitors and receptor ligands.
Results
Selection of experimental and computational techniques and data
analysis
To study the effect of both the nature of the nucleobase and
the length of the polyphosphate linker on the conformation of
dinucleotide sodium salts in aqueous solution, we selected a series
of 17 dinucleotide analogues, Fig. 4. Among those 17 analogues
we investigated more extensively the ApnAs (analogues 1–4), since
these are the most abundant dinucleotides in organisms.9
Our methods of choice for the investigation of the NpnN¢
conformation were NMR techniques including 1H-, 13C-, and
31P-1D NMR and COSY, HMQC, HMBC and NOESY 2D-
experiments. Chemical shifts (dH, dC, dP) and coupling constants
(JHH, JCH, JPH, JCP, JPP) of NpnN¢s spectra were measured. Data
are tabulated in the Supplementary Material (Table S1†). The
NMR data were analyzed by eqn (1)–(10) (in Experimental) and
the following conformational features of NpnN¢ analogues were
determined: value of c, dominant conformers around C4¢–C5¢
and C5¢–O5¢ bonds, and dominant ribose conformer, in addition
to identification of intramolecular p–p stacking interactions.
Furthermore, we used MD simulations combined with PMF
calculations to study the conformation of two representative
dinucleotides from the NpnN¢ series, Ap2A and Ap4A, where
additional information on torsion angles a, b and g was obtained.
The analysis of the MD data yields the conformational parameters
of ApnAs as well as the occurrence and the mode of intramolecular
p-stacking (i.e. a–a, b–b or a–b stacking). The PMF simulations
provide free energy associated with the stacking process.
Fig. 1 Structures of dinucleoside 5¢,5¢- polyphosphate analogues.
(i.e. a–a, a–b or b–b stacking, Fig. 2) is still inconclusive.15,16
Internal stacking interactions in other NpnN¢s (N π A) were
studied but no conformational parameters (e.g. sugar puckering,
a, b, g , c angles, Fig. 3) were provided.17,18 Likewise, detailed data
on the conformation of free dinucleotides in solution is limited
mainly to Ap2A.18
Crystal structures of several kinases or hydrolases-bound dinu-
cleotides in enzymes were determined.19–22 In protein-bound dinu-
cleotides, the polyphosphate chain adopts various conformations:
Synthesis of the studied dinucleotides
21
extended,19–22 S-shaped or folded.22 In some cases the protein-
bound dinucleotides coordinate with a Mg2+ ion. Nevertheless,
crystal structures of free NpnN¢s were hardly determined and a
crystal structure is available only for Ap4A sodium salt.23
Briefly, nucleoside-5¢-monophosphate (NMP) or nucleoside-5¢-
diphosphate (NDP) tributylammonium or trioctylammonium
salts were first activated as the corresponding phosphoroimida-
zolide analogues, using CDI in DMF and then treated by NMP or
NDP trialkylammonium salts.9,33 Synthesis of Up5U was achieved
by activating UTP tributylammonium salt with DCC followed
by the addition of UDP tributylammonium salt in DMF.9,33 The
dinucleotides were obtained in 10–60% yield. Dinucleotides with
long phosphate linker (with 4–5 phosphates e.g., Ap4U, Up5U) or
some of the mixed dinucleotides (with 2–3 phosphates e.g., Gp2C)
were obtained in yields lower than 10%.
Likewise, the conformation of NpnN¢s was scarcely studied
computationally. Specifically, only Ap2A was studied and its
stable conformations were calculated by energy minimizations.24
Computational-based conformational studies were performed
for related compounds, such as 3¢,5¢-dinucleotides,25–28 sugar-
nucleotides (UDP-glucose),29,30 ATP-Mg2+,31 and nucleosides.32
The absence of a large body of comprehensive data on the
conformation of a large and diverse series of free dinucleotides
prompted us to conduct a systematic conformational study
employing NMR and computational techniques. Specifically, we
analyzed the conformation of a diverse series of NpnN¢ (N¢ = A, U,
G, C; n = 2–5) analogues 1–17 (Fig. 4) in a neutral aqueous solution
by 1H and 31P-NMR. To augment the experimental data, we
studied the conformation of Ap2A and Ap4A in an explicit water
environment by molecular dynamics (MD) simulations combined
NMR Results
Selection of NMR measurements conditions
We first established the dinucleotide concentration range suitable
for the NMR experiments. The concentration range should be
4638 | Org. Biomol. Chem., 2010, 8, 4637–4652
This journal is
The Royal Society of Chemistry 2010
©

View Online
Fig. 2 Different stacking modes of Ap2A. One base was colored purple for clarity. (a) b–b base-stacked mode (b) a–a stacked mode, stabilized by two
intramolecular hydrogen bonds (black arrows) (c) a–b stacked mode. Side view. (d) a–b stacked mode. Top view.
Fig. 3 Torsional angles of Ap2/4A.
kept low enough to avoid inter-molecular base stacking (self asso-
ciation). Nucleotides tend to aggregate via p-stacking interactions
in aqueous solutions.34 However, at relatively low concentrations
they remain in their monomeric form. Specifically, 97% of 9 mM
ADP3-, 16 mM GDP3-, 22 mM CDP3- and 26 mM UDP3- exist
in their monomeric form.35 According to Mayo et al.,18 only at
dinucleotide concentrations above 20 mM are there concentration
effects on the chemical shifts due to inter-molecular interactions.
Therefore, we used 7.2–7.6 mM dinucleotide NMR samples.
Several of our dinucleotides were measured at two concentration
This journal is
The Royal Society of Chemistry 2010
Org. Biomol. Chem., 2010, 8, 4637–4652 | 4639
©

View Online
and NOESY spectra. In several cases, assignment was achieved
by comparison to mononucleotides.36,37 For purine analogues, the
ribose proton signals are always ordered as follows: d H1¢ > H2¢
> H3¢ > H4¢ > H5¢, H5¢¢. However, for pyrimidine analogues,
the ribose protons, except for H1¢, appear in no specific order. The
H5¢ signal was assumed to be more downfield than that of the H5¢¢
signal following Remin and Shugar’s observations.38 In most cases,
parts of the spectra measured are of second order, especially for
pyrimidine containing dinucleotides, and the obtained coupling
constants are either not accurate or can not be obtained at all.
Assignment of 3JH4¢H5¢/H5¢¢ and 3JPH5¢/H5¢¢ was based on the J-resolved
spectrum of Ap4A and 31P NMR of Ap3A (see Supplementary
Material†). The J-resolved spectrum for Ap4A was used to analyze
the spectra of other NpnN¢s. The error of 1H NMR coupling
constants is J 0.2 Hz. In cases where the spectra were of second
order due to overlapping signals, the error increased to 1 or 2 Hz,
as for H4¢, H5¢ and H5¢¢ in Ap2A, Up2U and Gp2G. Strongly
coupled protons showed broadened signals and the error was ca.
0.5 Hz, as for H3¢ in Ap2A and Gp2G. The error for 13C NMR
and 31P NMR coupling constants was J 0.5 Hz. The error in
the percentage of N/S conformers is 2–3% for J 0.2 Hz, which
is true for most of the examined dinucleotides. However, for J
0.5 Hz the error increases to 5–6%, and for J 1 Hz the error is
ca. 11%. Thus, Up2U has a larger error as compared to the other
dinucleotides.
Analysis of NMR data of the Np2N¢ series
p-stacking interactions. To study intramolecular p-stacking
interactions in dinucleotides we compared the chemical shifts of
the base protons of dinucleotides to those of the corresponding
mononucleotides at the same concentration. An upfield shift of a
dinucleotide nucleobase proton reflects that this proton is located
above the other aromatic ring plane and is affected by the ring
current.18,39,40
Base stacking was reported to be most dominant in Ap2A,
estimated as 60% of the population.24 Here, we found that base
stacking interactions are significant not only for Ap2A, but also
for all tested dinucleoside diphosphates (Np2N¢) as compared
to their homologues with longer phosphate linkers (Table 1).
Specifically, upfield shifts with Dd values of ca. 0.1–0.4 ppm for
H2/H8 or H5/H6 for A/G or C/U, respectively, were observed
as compared to the corresponding mononucleotides. We observed
more significant upfield shifts of the aromatic base protons
in Np2N¢ as compared to dinucleotides with longer phosphate
linker (for example, dH8: ATP>Ap5A>Ap4A>Ap3A>Ap2A). The
effect of the polyphosphate chain length on base stacking was
reported before for ApnAs and GpnGs, based on NMR and CD
experiments.10,13,18 Here, we show that p-stacking occurs also in
ApnU and GpnC analogues (Table 1). However, for the longer
UpnU analogues, Up4U and Up5U, there is no clear evidence from
NMR of base stacking, since the Dd values are within the error
range. Dd values for Up2U are larger than the error and indicate
the existence of stacking interactions.
Fig. 4 Dinucleoside polyphosphate (NpnN¢) analogues investigated in
this study.
ranges, 7.2–7.6 mM and 1.5–1.8 mM and their 1H NMR spectra
were compared. There were no significant differences in the
coupling constants and chemical shifts (Dd = 0.01–0.02 ppm)
between both concentration ranges. For example, for Ap4A DdH2
dH2,(1.7 mM) - dH2,(7.4 mM) = 0.02 ppm.
=
Ap2A exhibited the largest Dd values as compared to other
dinucleoside diphosphates, whereas Up2U exhibited the smallest
Dd values. For mixed dinucleotides (e.g. Ap2U), a large effect of
purine base on the pyrimidine protons was observed. However,
only a small effect of the pyrimidine on purine protons was
Assignment of the NMR signals and coupling constants of
dinucleotides 1–17
Assignment of H1¢, H2¢, H3¢, H4¢, H5¢, H5¢¢ and 13C NMR signals
of the analogues 1–17 was based on 2D COSY, HMQC, HMBC
4640 | Org. Biomol. Chem., 2010, 8, 4637–4652
This journal is
The Royal Society of Chemistry 2010
©

View Online
Table 1 Differences in chemical shifts (in ppm) between dinucleoside polyphosphates and the corresponding mononucleotide. (Dd = dmononucleotide
-
ddinucleotide
)
No.
Dinucleotide
DdH8
DdH2
DdH6
DdH5
DdH1¢(pu)
DdH1¢(py)
1
2
3
4
5
6
7
8
Ap2A
Ap3A
Ap4A
Ap5A
Ap2U
Ap3U
Ap4U
Up2U
Up4U
Up5U
Gp2G
Gp3G
Gp4G
Gp5G
Gp2C
Gp3C
Gp2C¢
0.37
0.26
0.17
0.11
0.10
0.03
0.01
—
0.24
0.18
0.14
0.12
0.06
0.04
0.03
—
—
—
—
—
—
—
—
—
0.34
0.21
0.14
0.07
0.02
0.02
—
—
—
—
0.35
0.27
-0.01
—
—
—
—
0.31
0.25
0.14
0.04
-0.01
0
—
—
—
—
0.23
0.20
-0.81
0.20
0.15
0.13
0.10
0.07
0.04
0.03
—
—
—
—
—
0.19
0.15
0.08
0.05
0.01
-0.01
—
—
—
—
9
—
—
—
—
10
11
12
13
14
15
16
17
0.20
0.14
0.08
0.06
0.09
0.03
0.15
0.15
0.11
0.08
0.06
0.06
0.04
-0.07
—
—
—
—
0.16
0.11
0.16
—
observed in Ap2U and Gp2C. The reason for those observations
is the much smaller ring current of the pyrimidines vs. purines.41,42
In the carbamate substituted cytosine in Gp2C¢, 17, downfield
shifts were observed, especially in cytosine H5, as compared
to CMP. This phenomenon is probably due to the electron
withdrawing effect of the carbamate group. Nevertheless, there
is an upfield shift for H8 in the guanine base which may be due to
base stacking effect.
C5¢ bond and Jt = 20.9 Hz and Jg = 1.8 Hz for C5¢–O5¢.44 Hence,
for Ap2A, Gp2G and Up2U the conformation around C4¢–C5¢
is gg (3J4¢5¢/5¢¢ = 2.1–2.7 1 Hz). For the ribose attached to U in
Ap2U, only one J value, J4¢5¢¢ = 2.5 1 Hz, was extracted from
the spectrum. Therefore, we conclude that H5¢¢ is gauche to H4¢
3
(Fig. 5).
The ring current effect in dinucleotides influences not only the
base protons, but also the anomeric H1¢.11 Indeed, we observed
an upfield shift in the H1¢ signal of most studied dinucleotides. In
addition, there were also upfield shifts in the H2¢ and H3¢ signals
(Supplementary Material, Table S2†).
Ribose puckering. We found that in most cases, for dinu-
cleotide ribose attached to any of the bases (A,G,U,C) there was
no pre-dominant conformation (i.e. Northern, N, or Southern, S,
conformation), Table 2. For adenine and guanine dinucleotides
there is a slight preference for the S conformer (ca. 55%),
whereas, for the unsymmetrical dinucleotides (Ap2U, Gp2C) ca.
65% preference was observed.
Uracil base attached to ribose induces only a slight preference
for the S conformer in the ribose (54% S in Ap2U). Up2U exhibited
almost no N↔S preference. Ribose attached to cytosine prefers
the N conformer (60%). N4-Carbamate substituted C induced a
greater preference for the N conformer (71%). This preference
of the S conformer in purine linked ribose and N conformer in
pyrimidine linked ribose is similar to that of the corresponding
mononucleotides.43
Fig. 5 Staggered rotamers around (a) C4¢–C5¢ and (b) C5¢–O5¢ bonds.
3
Moreover, some information may be obtained from JC4¢P
coupling constants about the conformation around the C5¢–O5¢
bond.43 We used ATP as a reference for determining rotamers
around the C5¢–O5¢ bond in dinucleotides. Thus, for ATP a
population of 63% gg around C5¢–O5¢ was calculated applying eqn
(6)–(8) using 3J4¢5¢ and 3J4¢5¢¢. In addition, a coupling constant 3JC4¢P
of 9.2 Hz was extracted for ATP from the 13C NMR spectrum.
3
For Ap2A, Up2U and Gp2G we measured the following JC4¢P
Conformation around the ribose –CH2O– exocyclic group. The
percentage of gg, gt and tg conformer populations around the
ribose –CH2O– exocyclic group in dinucleotides 1, 5–12, 15–17
could not be obtained from eqn (3)–(8), since the required coupling
constants could not be determined accurately from 1H NMR data
due to second order spectra in most cases. However, the dominant
conformer could sometimes be determined by examining the J4¢5¢/5¢¢
values and comparing them to the values of the pure rotamers.
The coupling constants for pure rotamers were estimated as Jg =
2.04 Hz (gauche rotamer) and Jt = 11.72 (trans rotamer) for C4¢–
values ( 0.5 Hz): 9.0, 9.4, 9.5 Hz, respectively. Therefore, based on
comparison to 3JC4¢P in ATP, we concluded that the gg rotamer is
the predominant one also for dinucleotides. For Ap2U there was a
slight difference in 3JC4¢P: 8.2 Hz for the adenine moiety and 7.6 Hz
for the uracil moiety. This may imply that the rotamer population
percentages for Ap2U are somewhat different than those of the
other mentioned Np2N¢s, namely, Ap2A, Up2U, Gp2G and Gp2C.
Glycosidic angle. We calculated the glycosidic angle for
3
3
Np2N¢s based on JC8/6-H1¢ and JC4/2-H1¢ (eqn (9)–(10)). Each
This journal is
The Royal Society of Chemistry 2010
Org. Biomol. Chem., 2010, 8, 4637–4652 | 4641
©

View Online
Table 2 Conformational data obtained for the studied dinucleoside polyphosphates
C4¢–C5¢
% gg
C5¢–O5¢
% gg
No.
Compound
c degrees (3JC8/6-H1¢
)
c degrees (3JC4/2-H1¢
)
% Puckering
% tg
% gt
% tg
% gt
1
2
3
4
5
5
6
6
7
7
8
9
10
11
12
13
14
15
15
16
16
17
17
Ap2A
Ap3A
Ap4A
Ap5A
Ap2U
Ap2U
Ap3U
Ap3U
Ap4U
Ap4U
Up2U
Up4U
Up5U
Gp2G
Gp3G
Gp4G
Gp5G
Gp2C
Gp2C
Gp3C
Gp3C
Gp2C¢
Gp2C¢
ATP
213/267
206/274
218/262
219/261
214/266
210/270
214/266
211/269
211/269
211/269
210/270
210/270
210/270
211/269
211/269
211/269
214/266
210/270
205/275
211/269
204/276
202/278
192/288
210/270
210/270
221/259
210/270
213/267
n.d.b
n.d.
55 (S)
51 (N)
62 (S)
65 (S)
63 (S)
54 (S)
64 (S)
52 (N)
67 (S)
55 (S)
52 (N)
59 (S)
60 (S)
57 (S)
57 (S)
67 (S)
69 (S)
65 (S)
57 (N)
69 (S)
62 (N)
61 (S)
71 (N)
64 (S)
55 (S)
67 (S)
61 (N)
n.d.
84
79
n.d.
5
8
n.d.
11
13
n.d.
64
63
n.d.
23
22
n.d.
13
15
213/267
226/254
218/262
226/254
213/267
226/254
218/262
218/262
218/262
220/260
226/254
230/250
79
8
13
64
23
13
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
67
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
15
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
18
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
61
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
22
n.d.
n.d.
n.d.
n.d
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
17
230/250
a
70
15
15
60
23
17
a
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
82
n.d.
70
84
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
8
n.d.
15
7
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
10
n.d.
15
9
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
63
n.d.
61
73
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
23
n.d.
25
10
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
14
n.d.
14
17
n.d.
240
204/276
n.d.
n.d.
198/282
218/262
240
UTP
GTP
CMP
213/267
3
3
a Solutions of the glycosidic angle are out of the expected range, possibly due to error of measurement ( 0.5 Hz of JC8/6-H1¢ and JC4/2-H1¢). b n.d. not
determined due to second order spectrum, The percentage of preferred conformer refers to the ribose the base of which appears in bold italics. (e.g. Ap2U
refers to the ribose attached to adenine).
equation gives four (sometimes two) solutions. One solution in
the anti region, one in the high anti region, one is in the syn region,
and the last one is syn or anti. We selected the solutions that are
in the anti and high anti regions as the most reasonable possibility,
considering the NOESY and the MD results, mentioned below
(Table 2). The range of glycosidic angles for analogues 1, 5, 8,
11, 15, 17 was 192–226◦ (anti)/254–288◦ (high-anti). The error in
angle values ranges from 6◦ to 22◦.
Though the c angles reported for mononucleotides were 200
20◦ 43 and one of the solutions obtained here for c of dinucleotides
was in the range of 210 20◦, our MD results showed larger
c angle values, in the high anti region, consistent with another
solution resulting from NMR data. Therefore, we concluded that
the c angle range in dinucleotides is in the anti/high anti range.
NOESY spectra provided additional evidence for the anti
conformation of Np2N¢ analogues (Fig. 6–7). Clear cross peaks
between H8–H1¢, H8–H3¢, H8–H2¢ for the purine base moiety
supported the anti conformation in Ap2A and Gp2G. Likewise,
one cross peak between H8–H3¢ for the adenine base in Ap2U
and for the guanine base in Gp2C, and H6–H2¢/3¢, H6–H1¢ for
Fig. 6 NOESY spectrum of 7.4 mM Ap2A in D2O, pD 7.32, 300 K at
the pyrimidine base moiety, supported the anti conformation in
Up2U, Ap2U, Gp2C and Gp2C¢.
600 MHz.
In Ap3U and Gp3C there is a large effect of purine base on the
pyrimidine protons, but only a small effect of the pyrimidine on
purine protons, as observed for Ap2U and Gp2C.
Analysis of NMR data of the Np3N¢ series
The upfield shifts of the nucleobase protons in Np3N¢ series
indicate base stacking interactions in this series, yet, to a smaller
extent than those in the corresponding Np2N¢ series (Table 1).
Among the ApnAs (n = 2–5), Ap3A has the largest percentage
of N conformer. The dominant conformer around both C4¢–C5¢
and C5¢–O5¢ is gg, 84% and 64%, respectively (Table 2). In the
4642 | Org. Biomol. Chem., 2010, 8, 4637–4652
This journal is
The Royal Society of Chemistry 2010
©

View Online
Analysis of NMR data of the Np4N¢ series
For the Np4N¢ series base stacking interactions were still observed,
although weaker than those of the corresponding Np3N¢ series,
except for Up4U for which stacking interactions could not be
determined by NMR due to a small ring current effect (Table 1).
Unlike the X-ray crystal structure of Ap4A which adopts the
N conformer,23 Ap4A in solution prefers the S conformer (62%),
Table 2. Up4U in solution prefers the S conformer as well (59%).
However, the dominant Ap4A conformer in solution is gg around
both C4¢–C5¢ (79%) and C5¢–O5¢ (63%), as found in the crystal
structure.23 Gp4G has, as Ap4A, a larger preference for the S
conformer (67%) as compared to Gp2/3G, Table 2. Likewise, the
preferred conformer is gg around both C4¢–C5¢ (67%) and C5¢–
O5¢ (61%). Both ribose moieties of Ap4U are preferentially S (67%
in adenylated ribose, and 55% in the uracilated ribose). Uracilated
ribose in Ap4U has about the same conformation as Ap2U (54%
S), unlike Ap3U (52% N).
Conformation around the exocyclic group (C5¢–O5¢) was not
1
determined for most Np4N¢s due to the second order H NMR
3
spectrum. However, JC4¢P data show that for adenylated ribose
in Ap4A and Ap4Us the C5¢–O5¢ conformation is gg as in
ATP. Uracilyated ribose in Up4U and Ap4U conformation is also
gg based on 3JC4¢P (Supplementary Material, Table S1†).
The conformation around glycosidic angle in the Np4N¢ series
is anti or high anti around both glycosidic bonds, according to eqn
(9)–(10) (Table 2) and NOESY data (showing H8–H3¢ cross peak
for Ap4A and Ap4U, and H6–H2¢, H6–H3¢ cross peaks for Ap4U
and Up4U).
Fig. 7 NOE interactions in dinucleotides in the anti conformation.
GpnG series, Gp3G prefers the S conformer (57%), as observed
for Gp2G.
The predominant conformation of the adenylated ribose in
Ap3U is S (64%), as observed for Ap2U (64%). Yet, the uracilated
ribose in Ap3U slightly prefers the N conformer (53%). Among
ApnUs, n = 2–4, Ap3U has the highest percentage of N conformer
in the uracilated ribose. There is a significant difference in the
conformation of the adenylated ribose in Ap3U compared to
Ap3A: 36% S vs. 51% N, respectively. Yet, the origin of the effect
of the neighboring U vs. A on the adenylated ribose is not clear.
We used 3JC4¢P for deducing the conformation around C5¢–O5¢,
since 3JP5¢/5¢¢ could not be extracted for most of the Np3N¢ series. We
found that the conformation around C5¢–O5¢ in Ap3A, Gp3G and
Ap3U is gg for both ribose moieties based on comparison with
Analysis of NMR data of the Np5N¢ series
In the Np5N¢ series, the base stacking interactions are least
significant of all dinucleotides investigated here. Generally, the
chemical shifts of Np5N base signals are highly similar to those of
free mononucleotides (e.g. ATP), indicating a negligible percentage
of stacked species. For Ap5A the upfield shifts (Dd) are 0.12 ppm
for H2, and 0.11 ppm for H8, as compared to ATP. For Gp5G
the upfield shift is negligible, i.e., DdH8 = 0.06 ppm as compared
to GTP H8 signal. Likewise, for Up5U there is no upfield shift of
H5, and Dd of only 0.02 ppm was observed for H6 as compared to
UTP H6 signal. Therefore, we assume that stacking interactions
probably still occur in Ap5A and Gp5G, but there is no clear
evidence for their occurrence in Up5U.
3
the UTP, GTP and ATP JC4¢P values (Supplementary Material,
Table S1†).
The conformation around the glycosidic bond in the Np3N¢
3
3
series is anti/high anti according to JC8/6-H1¢ and JC4/2-H1¢ values,
and NOESY data (H8–H3¢, H8–H2¢ and H8–H3¢ cross peaks were
observed for the adenine moiety in Ap3A and Ap3U, respectively)
(Table 2).
In Ap5A and Gp5G the dominant conformer is gg around
both C4¢–C5¢ (79% and 70% respectively) and C5¢–O5¢ (64%
and 60% respectively) bonds, as in Ap4A and Gp4G. Population
analysis for Up5U could not be done due to missing coupling
The Gp3C ribose conformation is similar to that of Gp2C. The
G/C bases are in anti/high anti conformation (Table 2). Compari-
son of 3JC4¢P values for mononucleotides and Gp3C (8.7 and 9.3 Hz
for guanosine and cytidine moieties, respectively), shows that the
conformation around C5¢–O5¢ in Gp3C is gg.
It is noteworthy that the chemical shifts of the a phosphates of
Np3N¢s are shifted upfield as compared to the other Pa chemical
shifts (e.g. dPa (Ap2A) = -10.39, dPa (Ap4A) = -10.42, dPa
(Ap5A) = -10.44, vs. dPa (Ap3A) = -10.75 ppm). This shift of
Pa signal in Ap3A may be due to a change in the phosphate chain
conformation.45
3
constants. However, the JC4¢P constant of 9.2 Hz in Up5U
provides evidence for a gg conformation of C5¢–O5¢, as in mono-
nucleotides.
Np5N¢s are similar to the corresponding mononucleotide and
the Np4N¢ analogues also in the sugar puckering. Ap5A, Gp5G
and Up5U prefer the S conformer (60–69%).
The conformation around the glycosidic bonds in Gp5G, Ap5A,
Up5U is anti or high anti, according to the Eq. 9–10 and NOESY
spectra. Thus, for instance, we observed cross peaks between H8–
H3¢, H8–H2¢ for Ap5A, and H6–H3¢ for Up5U.
This journal is
The Royal Society of Chemistry 2010
Org. Biomol. Chem., 2010, 8, 4637–4652 | 4643
©

View Online
Molecular dynamics results
We chose Ap2A and Ap4A for our computational investigation
since they both belong to the most abundant and studied series of
ApnAs. These two analogues helped us evaluate the dependence
of the dinucleotide conformation in solution on the length of the
phosphate linker.
To verify the appropriateness of the adenine nucleotide
CHARMM27 force field parameters in the case of adenine
dinucleotides, we performed crystal simulations, starting with the
experimental structure as input.23 According to the simulation
results, there is good agreement between the experimental and
simulated molecular structure and cell unit parameters of Ap4A
(Fig. 8). Ap4A experimental crystal p◦aramete◦rs (a, b, c, a◦, b, g )
˚
˚
˚
are 12.75 A, 20.27 A, 8.56 A, 90.00 , 90.38 , and 90.00 , and
˚
˚
the para◦meters after minimi◦zation are 12.46 A, 19.59 A, 8.52
◦
˚
A, 90.00 , 91.98 , and 90.00 . The root mean square deviation
(RMSD) between the crystal structure and the structure after
˚
minimization was 0.293 A for all the dinucleotide atoms, including
hydrogens (excluding water and ions). The average unit cell values
˚
obtained during 1 ns of MD simulation are a = 12.09 0.17 A, b =
19.38 0.23 A, c = 9.14 0.18 A, angle b = 88.23 2.05◦. The graph
showing the changes of a, b, c and b during the 1 ns simulation is
shown in the Supplementary Material, Fig. S1–S2.† The average
˚
˚
Fig. 9 The reaction coordinate, N9–N92 distance, used for examination
of Ap2/4A base-stacking interactions.
˚
RMSD during simulation is 1.293 A. The structural integrity
observed during the crystal minimization and MD simulation
serves as an indication of the appropriateness of the nucleotide
force field parameters for dinucleotides.
apart. This was indeed observed in numerous snapshots. Thus,
Ap2A was found to be stacked 77% of the simulation time, while
Ap4A was stacked 70% of the simulation time (see Supplementary
Material, Fig. S3†). During the course of the simulations, several
occurrences of stacked conformations which opened up and
subsequently re-stacked were observed.
Previous CD studies of base-stacking in dinucleotides (ApnA n =
2–6 as well as GpnGs n = 2–6 and mixed dinucleotides) proposed
that there are two stereo-configurations: a–a and b–b (Fig. 2).16
Specifically, Ap2A was suggested to be preferentially stacked in the
b–b mode, unlike other ApnAs which were assumed to be stacked
in the a–a mode.10,11,16 Previous reports suggested that Ap2A can
have several stacking modes, including a–a, b–b, and a–b modes.24
It was also suggested that the a–a mode possibly enables hydrogen
bonding between the amine nitrogen of each adenine ring and the
C2¢–OH of the opposite ribose residue for Ap3/4A.16
To obtain qualitative information regarding the base stacking
and its mode in the adenine dinucleotides, we performed PMF
simulations for Ap2A and Ap4A in the a–a, b–b and a–b modes. In
Fig. 2a–d we show representative stacked conformations of Ap2A
in the a–a, b–b, and a–b modes. We differentiate between a–a,
b–b, and a–b stacking by visual inspection of frequent snapshots
from the MD trajectory.
Fig. 8 Conformations of Ap4A. A. Crystal structure (green). B. After
1 ns molecular dynamics at 293 K (other colours).
For Ap2A and Ap4A there is a preference for the stacked
conformation over the extended one, regardless of the stacking
mode (Fig. 10). Moreover, there is no free energy barrier separating
the extended and folded forms. For Ap2A, the a–a, b–b and a–b
modes have stacking free energies of 3.1, 1.5 and 3.2 kcal mol-1,
respectively. This finding of several equilibrating conformers of
Ap2A is in agreement with previous studies by Thornton et al.24
and Pettegrew et al.15 Yet, our finding is not consistent with CD
data, which suggested that Ap2A is only b–b stacked.10,16 This
point will be addressed further in the Discussion. Ap4A has also
several stacked conformers in equilibrium. Specifically, the a–a,
Intramolecular base stacking and hydrogen bonding in Ap2/4
sodium salts
A
Initially, to obtain qualitative information regarding the p-
stacking in the adenine dinucleotides, we performed 10 ns of un-
biased MD simulations (termed free MD below). To differentiate
between open and p-stacked conformations, we defined a N9–
˚
N92 distance <6.5 A (Fig. 9) as a stacked conformation. The
˚
somewhat large stacking limit of 6.5 A was chosen since the base
planes could be close although the N9–N92 atoms relatively far
4644 | Org. Biomol. Chem., 2010, 8, 4637–4652
This journal is
The Royal Society of Chemistry 2010
©

View Online
Fig. 10 Potential mean force (kcal mol-1) for the base stacking in ApnA: (A) Ap2A (B) Ap4A in a–a mode (red line) the b–b mode (blue line) and the
a–b mode (green line) as a function of the distance N9–N92 between the two adenine moieties.
b–b and a–b modes have stacking free energies of 3.3, 2.3 and 2.4
kcal mol-1, respectively.
other simulations the Ap4A a–a form was not always hydrogen
bonded.
Based on previous suggestions16 we examined the existence of
intramolecular hydrogen bonds between the N6 amine and the
ribose O2¢ groups in Ap2A and Ap4A. The N6–O22¢ and N62–O2¢
distances (Fig. 11) were measured for the adenine dinucleotides
in a–a, b–b, and a–b stacking modes based on PMF simulation
windows in which the bases were stacked. The distances were
averaged over ca. 2000 structures collected during 1 ns of MD
simulations. For Ap2A in the a–a mode the N6–O22¢ and N62–
From the PMF for the adenine dinucleotides we can also obtain
the average minimum N9–N92 stacking distance. For Ap2A the
˚
˚
˚
stacking distances are 4.8 A, 5.1 A and 4.5 A for a–a, b–b and
a–b, respectively, while for Ap4A the stacking distances are 5.0
˚
˚
˚
A, 5.3 A and 6.6 A for a–a b–b and a–b , respectively. In the
Ap4A a–b structure the average N9–N92 distance is large due to
the relative approach of the two bases in this case, although the
base planes are close and stacking clearly occurs.
˚
˚
O2¢ distances were 3.40 0.54 A and 3.51 0.61 A, whereas
˚
˚
for b–b the distances were 6.90 1.25 A and 8.58 1.76 A. For
˚
the a–b mode these distances were 7.24 0.65 A and 7.37
0.84 A. This indicates that in the a–a stacked mode, Ap2A forms
Study of conformations in stacked vs. extended forms of Ap2/4A in
the various stacking modes during PMF simulations
˚
intramolecular hydrogen bonds, while in the b–b and a–b modes
no hydrogen bonds are formed. For Ap4A the distances were
The average c values are in the anti or high anti range for all forms
(Tables 3–4). Specifically, there is a tendency towards the high anti
for the extended conformations of Ap2/4A. In addition, both c
angles for Ap4A b–b are in the anti range, while the c angles of
Ap4A a–a and a–b are closer to the high anti range . The largest c
values are for the a–b mode, with one angle in the high anti range.
˚
˚
3.72 0.89 A and 3.16 0.40 A for a–a, whereas for b–b 5.91
˚
˚
˚
0.54 A and 5.58 0.61 A, and for a–b 11.54 0.98 A and
4.74
˚
0.63 A. As for Ap2A, there was a preference for the
hydrogen bonded form in the a–a stacked Ap4A (Supplemen-
tary Material, Fig. S4–S9 for all Ap2/4A species†). However, in
This journal is
The Royal Society of Chemistry 2010
Org. Biomol. Chem., 2010, 8, 4637–4652 | 4645
©

View Online
˚
Table 3 Torsion angles of Ap2A during 1 ns simulation at N9–N92 distances of 4.5 (in 3 stacking modes) and 9.0 A (extended form)
Ap2A a–a (◦)
Ap2A b–b (◦)
Ap2A a–b (◦)
Ap2A ex. (◦)
a1
a2
b1
b2
g 1
g 2
c1
c2
180.4 21.9
184.6 22.4
190.8 31.2
189.7 25.9
64.6 13.3
62.3 15.1
230.7 14.9
229.9 13.7
281.7 44.7
141.2 90.0
151.0 26.3
208.1 27.2
54.2 13.4
62.9 23.5
241.0 30.2
253.1 22.6
77.9 16.0
292.2 12.6
241.1 10.5
168.0 13.5
83.4 11.9
58.6 14.4
267.3 11.9
304.0 15.4
210.6 77.2
152.3 110.0
185.8 26.2
192.4 25.7
61.0 15.0
60.0 16.2
263.3 28.7
267.5 40.5
˚
Table 4 Torsion angles of Ap4A during 1 ns simulation at N9–N92 distances of 4.5 (in 3 stacking modes) and 9.0 A (extended form)
Ap4A a–a (◦)
Ap4A b–b (◦)
Ap4A a–b (◦)
Ap4A ex. (◦)
a1
a2
b1
b2
g 1
g 2
c1
c2
144.4 31.4
108.3 51.1
213.1 31.3
276.4 53.9
65.5 14.8
72.4 17.8
255.2 19.7
261.2 18.3
282.0 30.3
278.4 19.0
149.8 19.5
112.8 15.0
48.9 15.7
47.0 16.0
215.3 17.5
222.4 17.3
72.8 18.1
292.4 84.7
226.9 10.7
111.7 13.2
69.0 11.0
45.6 15.7
257.5 28.7
258.4 22.3
294.4 26.4
-34.4 25.2
139.8 20.4
164.2 18.0
41.9 14.6
62.4 16.0
256.8 15.3
275.0 43.4
Fig. 11 Atom type definition for Ap2A.
The average angle g is in the gg range (60.0 30.0◦, Tables 3–
4). The average angle b is in the gg range in most cases (180.0
30.0◦). Additionally, in the 10 ns simulation the average b fell
mostly within the gg range (Supplementary Material, Table S3†).
The a angle varies many times during the simulations, and it has a
broad set of values with larger standard deviations. The a values
may be similar for both ribose moieties (e.g. in the Ap2A a–a
mode the a values remain ca. 180◦) or different (as in the Ap2A
a–b mode) (Tables 3–4).
Inspection of the results from the two 10 ns free MD simulations
revealed that both N and S conformations are present for Ap2A
and Ap4A. In agreement with the NMR data we find there is a
preference for the S conformer over the course of 20 ns of MD
simulations: The average values for the two ribose moieties in
Ap2A and Ap4A are 84% and 92%, respectively. The preference
for the S conformer is also found in the restrained 1 ns PMF
simulations for both Ap2A and Ap4A (80% and 88% respectively)
as in the free MD simulations. However, during 1 ns simulation in
the Ap2A a–b structure, one ribose preferred the S conformer
and the other preferred the N conformer. These findings are
in agreement with NMR data showing that the preference for
the S conformer is larger in Ap4A than in Ap2A, although the
difference is small. The dominance of the S conformer remained
even when the simulations were extended for additional 40 ns
(not shown). We note that there is a difference between the
NMR and simulation S/N populations which probably results
from insufficient simulation time and inaccuracies in the potential
energy surface.
The effect of sodium ions was also examined by measuring
the distance between the sodium ions and phosphorous atoms
of Ap2/4A during free simulations of lasting 50 ns. The average
˚
distances were all larger than 11 A and therefore we conclude that
the sodium ion’s role is neutralization of charge and it does not
have an effect on the conformation of dinucleotides.
4646 | Org. Biomol. Chem., 2010, 8, 4637–4652
This journal is
The Royal Society of Chemistry 2010
©

View Online
Crystal structure of Ap4A vs. structure in solution
stacking mode, while Ap3/4/5A prefer the a–a mode, in contrast
to our results suggesting no preferred stacking modes. However,
the CD study was based on several unsettled assumptions about
the transition moments in adenine.16 In addition, in that study,
the existence of only one stacking mode for each dinucleotide was
assumed. Our findings are consistent with a CD study on Ap2A15
suggesting that Ap2A exists in solution in a multi-conformation
equilibrium, as opposed to a simple stacked-unstacked equilib-
rium, yet, none of the previously suggested stacked modes (a–a,
b–b or a–b) was attributed to Ap2A. On the other hand, all three
stacking modes were suggested for Ap2A by computational studies
in the gas phase.24 We found that Ap4A equilibrates between those
three stacking modes, the same as we (and others)15,24 found for
Ap2A. The multi-conformation equilibrium found for Ap2/4A is
likely for all other NpnN¢s as well.
Comparison to the crystal structure of Ap4A shows that there
are differences compared to the structure in solution. The modes
of stacking and the puckering are different. Specifically, in the
crystal structure the stacking mode is b–b and the ribose prefers
the N conformer while in solution the stacking mode is a–a
and the preferred ribose conformer is S. A similar difference in
puckering exists for ATP.46 The torsion angles b and g in the
crystal structure are in the gg range, as was observed by NMR
and in the PMF simulations. The glycosidic angle c is in the anti
range for the crystal structure, and is in the high anti (a–a)/anti
(b–b) range according to MD simulations. The torsion angles a,
b, g , c resulting from the PMF simulation for Ap4A b–b have
very similar values to those of the crystal structure (See Table 4,
a angles: 293◦, 295◦, b angles: 166◦, 158◦, g angles: 47◦, 47◦, c
angles: 210◦, 208◦).
Dinucleotides with longer phosphate chain have weaker stacking
interactions
Discussion
Our NMR data show chemical shift differences (Dd) for the dinu-
cleotide nucleobase protons vs. the corresponding mononucleotide
protons that may be interpreted as a different degree of base-
stacking for the different NpnN¢s. In each dinucleotide series, e.g.,
ApnUs n = 2–4, the longer the dinucleotide phosphate chain, the
smaller were the Dd values, suggesting greater stacking for the
shorter dinucleotides in each series.
Though according to NMR results we may suggest that Ap2A
and Ap4A have different populations of stacked conformation, we
do not obtain a significant difference between the populations of
Ap2A and Ap4A in the PMF simulations. The stacking free energy
difference between Ap2A and Ap4A is expected to be within the
statistical error of the simulations. Moreover, the difference in Dd
in the NMR spectra does not provide quantitative information on
the populations of stacked structures.
The preference for stacking in shorter dinucleotides may be a
result of an entropic effect. The stacking in Ap4A involves the
loss of more degrees of freedom (of the phosphate linker) as
compared to Ap2A and therefore it may be less favorable. Another
possible explanation is the repulsion of more negative charges in
a longer phosphate chain that may be reduced in an extended
conformation.
Dinucleotides in an aqueous solution exist as a mixture of equili-
brating species most of which are in the folded conformation.
In this study, we addressed the question of the preferred confor-
mations of dinucleoside polyphosphate (NpnN¢; N, N¢ = A, U, G,
C; n = 2–7) sodium salts in an aqueous solution. In particular, we
studied whether the dinucleotides exist in the extended or folded
(stacked) conformations.
Nucleic acids, nucleosides and nucleotides form p–p stack-
ing interactions through their bases in water.41–43,47 The base
stacking that results from p–p interactions is assumed to be
stabilized mainly by London dispersion forces (momentary dipole-
induced dipole) and hydrophobic effects.25 Several reports sug-
gest that London dispersion forces are the most significant
interaction,48 while others propose that hydrophobic effects make
the most significant contribution.49 Additional minor factors
that influence stacking are electrostatic interactions and the
de-solvation effect.49,50 In addition, the polyphosphate linker
may influence the stacking of dinucleotide bases. The stacking
interactions may occur inter- and/or intra-molecularly. While
in concentrated solutions we expect intermolecular stacking,
as in mononucleotides,18,42,51 in diluted solutions, such as the
physiological medium, intramolecular stacking is assumed to
occur.18
Based on extensive NMR studies for dinucleotides 1–17 we
conclude that the intramolecularly base-stacked conformation
exists for all dinucleotides studied here in neutral aqueous solution,
except for Up4/5U, for which base stacking interactions could not
be determined by NMR.41 Intramolecular base stacking is sup-
ported by MD simulations for the Ap2A and Ap4A dinucleotides,
which showed a preference of 1.5–3 kcal mol-1 for the p-stacked
configurations (i.e., equilibrating a–a, b–b, and a–b, stacked
species). Indeed, intramolecular p-stacking interactions were also
observed by similar PMF simulations for related dinucleoside
monophosphate analogues, e.g. adenylyl-3¢-5¢-adenosine (3¢-5¢-
ApA).27 Specifically, dinucleotides with at least one purine base
showed favorable stacking interactions with a DG difference of ≥2
kcal mol-1 between the stacked and unstacked state.52
Purine dinucleotides show greater stacking interactions than
pyrimidine dinucleotides
The base type, namely purine or pyrimidine, affects the interaction
between the bases. The self association tendency of bases in
mononucleotides was rated as: adenine > guanine > cytosine ~
uracil.42 Similar data were obtained for oligonucleotides,.49 Purine
bases in nucleosides interact with each other more favorably
than purine and pyrimidine bases, while pyrimidine–pyrimidine
interactions could not be determined by NMR methods, though
there was evidence for their association.41 A similar tendency for
base–base interactions was found in a computational work on
dinucleoside monophosphate analogues.27
Here, the larger Dd shifts were observed for purine dinucleotides,
or mixed purine–pyrimidine dinucleotides, where the purine base
affects the pyrimidine base (i.e. Dd of H5, H6 of U or C in dinu-
cleotides vs. mononucleotides UTP and CMP), while the effect of
A previous study on base-stacking interactions in ApnAs and
GpnGs, applying CD,10,11,16 proposed that Ap2A adopts the b–b
This journal is
The Royal Society of Chemistry 2010
Org. Biomol. Chem., 2010, 8, 4637–4652 | 4647
©

View Online
pyrimidine on purine (i.e. Dd of H2, H8 of A or G in dinucleotides
vs. mononucleotides ATP and GTP), is small. Though the Dd
values in UpnUs (vs. UTP) are small, it does not necessarily mean
that there are no stacking interactions, only that they cannot
be monitored by NMR. Indeed, computational data for related
ribodinucleotides monophosphates (e.g. 3¢-5¢-ApA, ApG, GpA
etc.) showed the following stacking preference: purine–purine
> purine–pyrimidine > pyrimidine–pyrimidine.27 Therefore, we
suggest that ApnAs and GpnGs undergo intramolecular stacking to
a greater extent as compared to UpnUs. The stacking preference of
purines over pyrimidines could be explained by the larger surface
area of purines which increases dispersion interactions and also
affects hydrophobic effects. Moreover, the aromaticity of purines
is greater than that of pyrimidines,53 thus possibly resulting in
more significant dispersion forces. Among purines, we observed
greater stacking for the adenine vs. guanine analogues, based on
the respective Dd values, consistent with literature data.39,54
of dinucleotides at a physiological pH will be reported in due
course.
Conclusions
Based on our MD and NMR data, obtained at physiological
pH, we propose that under physiological conditions dinucleotide
sodium salts adopt preferentially the folded structure, which
exists as an equilibrating mixture of a–a, b–b and a–b stacking
modes. However, the conformation of each nucleotide moiety
in a dinucleotide is similar to the standard mononucleotide
conformation.
Investigating the conformations of the dinucleotides in solution
is a prerequisite for understanding their activity as enzyme
inhibitors and receptor ligands, since the DG of dinucleotide
binding to proteins depends on the change of the dinucleotide
conformation during the transition from solution to protein envi-
ronment, in addition to de-solvation effects. Therefore, we expect
our findings will be valuable to the community of biochemists and
medicinal chemists.
Each nucleoside moiety in a folded dinucleotide retains a standard
mononucleotide conformation
Experimental
We systematically investigated the conformation of dinucleotide
analogues and the corresponding mononucleotides (ATP, UTP,
GTP and CMP) as reference compounds (Table 2). We found
that the preferred conformation around the glycosidic-; C4¢–C5¢-
; C5¢–O5¢-bonds are: anti/high anti, gg, and gg, respectively, for
the investigated mononucleotides. Both purine and pyrimidine
mononucleotides prefer the S conformer, except for CMP that
prefers the N conformer. The reported sugar puckering for purine
nucleoside monophosphates is preferentially S (60%), while for
pyrimidine mononucleotides it is N (60%). Thus, in both cases
there is no predominant sugar puckering, as we found here.43 The
reported conformation for mononucleotides around the glycosidic
angle is anti and around C4¢–C5¢– and C5¢–O5¢– bonds it is gg.43
All dinucleotides studied here adopted preferentially the high
anti/anti range of the c angle according to C–H coupling con-
stants, NOESY spectra, and MD simulation for Ap2/4A. The b and
g torsion angles in ApnAs and Gp4/5Gs are in the gg conformation
range, yet the gg preference is greater for g - than for b-angles
according to NMR (~70–80% and ~60%, respectively), and MD
data. For the other dinucleotides, due to second order NMR
spectra, the only information about the exocyclic group (C4¢–
C5¢ and C5¢–O5¢) is derived from JC4¢P, which upon comparison
to JC4¢P of mononucleotides, provides information about the C5¢–
O5¢ bond. The b angle in those cases adopts the gg conformation
similar to mononucleotides.43
Most of the dinucleotides prefer the S conformer (~50–70%),
except for cytosinylated ribose that prefers the N conformer (~60–
70%). MD simulation for Ap2/4A also supported the preference
for the S conformer. A slight preference for the N conformer
was found for: Ap3A (51%), Ap3U (52%), Up2U (52%) dinu-
cleotides. These sugar puckering preferences are similar to those
of mononucleotides.43
Here, we studied the conformation of various dinucleotides
sodium salts. Yet, in a physiological medium, dinucleotides may
exist either as sodium (or potassium) salts or coordinate to
Mg2+/Ca2+ ions. The preference of dinucleotides to Mg2+/Ca2+
coordination and the possible conformations of these complexes
General. Ap4A and Ap5A, 3 and 4, were purchased from
Sigma (Steinheim, Germany) and Fluka (Steinheim, Germany).
Dinucleotides 1–2 and 5–17 were synthesized according to
literature.9,33 The air- and moisture-sensitive reactions were per-
formed in flame dried, nitrogen or argon flushed two necked
flasks sealed with rubber septa. The reagents were introduced
with a syringe. The progress of the reactions was monitored
by TLC on precoated Merck silica-gel plates (60F-254) and
visualization was accomplished by UV light. Primary purification
of the dinucleoside polysphates was achieved on a LC (Isco
UA-6) system using a Sephadex DEAE-A25 column, which was
swelled in 1 M NaHCO3. Final purification was achieved on
a HPLC (Merck-Hitachi) system using semipreparative reverse-
phase Gemini 5u C18 110A 250-10 (Phenomenex, Torrance
CA, USA) or LichroCART 250-10 column (Merck, Darmstadt,
Germany) in the following solvent system: MeOH/0.1 M TEAA
(triethylammonium acetate, pH = 7.5). Each dinucleotide was
purified using a different gradient. The gradients ranged from
1 : 99 to 13 : 87 MeOH–TEAA in 20 min, flow rate 5 mL min-1. The
purity of the dinucleotides was evaluated on an analytical reverse-
phase column system (Gemini 5u, C-18, 110A, 150x4.60 mm,
5 mm, Phenomenex, Torrance, CA), in two solvent systems as
described below. To obtain the dinucleotides as sodium salts
they were passed through a column of Sephadex-CM C25 (Na+-
form) column. Some of the dinucleotides, e.g. Gp2/3C, were passed
instead througha columnofDowex 50WX8-200(H+ form that was
washed with NaOH to obtain the Na+ form of this resin), since
they degraded on Sephadex- CM C25 (Na+ form). Compounds
were characterized by NMR spectroscopy using Bruker DPX-
300, DMX-600, AC-200, Avance III -500 and Avance III - 700
spectrometers. Chemical shifts are reported in ppm relative to
glycerol and D2O as internal standard (1H and 13C NMR).
The chemical shifts of 31P NMR are reported in ppm relative
to 85% H3PO4 as an external standard. The pD values of the
NMR samples were adjusted with diluted NH4HCO3–TFA D2O
solutions or NaOD–DCl D2O solutions. Measurements of pD
4648 | Org. Biomol. Chem., 2010, 8, 4637–4652
This journal is
The Royal Society of Chemistry 2010
©

View Online
were performed with a Knick pH-meter (Knick, Berlin, Germany)
equipped with biotrode 238140 or tiptrode 238080 Hamilton
electrode (Hamilton, Bonaduz, Switzerland). Mass spectra were
recorded on Micromass QToF low resolution mass spectrometer
in ESI negative mode.
154.37 (C2G), 152.22 (C4G), 145.07 (C6C), 138.14 (C8G), 116.81
(C5G), 97.37 (C5C), 91.15 (C1¢C), 87.75 (C1¢G), 84.21 (d, JCP
=
8.7, C4¢G), 83.01 (d, JCP = 8.7, C4¢C), 75.50 (C2¢C), 74.22 (C2¢G),
70.97 (C3¢G), 68.98 (C3¢C), 66.05 (d, JCP = 5.0, C5¢G), 64.76 (d,
JCP = 3.7, C5¢C), 53.99 (CH3) ppm. MS (MALDI): m/z 725((M -
H)-, 69%) 515(33), 548(100), 699(51), 705(15).
Typical procedure of dinucleotide synthesis: Guanosine-5¢-
diphospho-5¢-cytidine (15) and guanosine-5¢-diphospho -5¢-N4-
methyl carbamate-cytidine (17). CMP(Bu3NH+)2 was prepared
from the CMP free acid (119.7 mg, 0.37 mmol) and Bu3N (176 mL,
2 eq) in 95% EtOH. GMP(Bu3NH+)2 was prepared from GMP
disodium salt by passing it through a column of activated Dowex
50WX-8 200 mesh, H+ form. The column eluate was collected
in an ice-cooled flask containing Bu3N (2 eq) and EtOH. The
resulting solution was freeze-dried to yield GMP(Bu3NH+)2. Bis-
tributylammonium CMP and CDI (348 mg, 2.15 mmol, 5.8
eq) were stirred in dry DMF under argon atmosphere at room
temperature for 2 h. Dry MeOH (140 mL, 3.45 mmol, 9.3 eq) was
added. After 20 min GMP(Bu3NH+)2 (311 mg, 0.42 mmol) in dry
DMF (6 mL) was added. The resulting solution was stirred at room
temperature for 24 h. Then DMF was evaporated and a viscous
oil was obtained. The resulting oil was separated on Sephadex
DEAE-A25, eluting the product with a buffer gradient of 0–0.3 M
NH4HCO3 (total volume 1000 mL). The product was obtained
in ca. 10% yield (26 mg) after LC separation. N4-Me-carbamate-
Gp2C, 17, was obtained in ca. 54% yield (152 mg). Final separation
of 15 was achieved by applying isocratic elution of TEAA–MeOH
94 : 6 in 30 min (5 mL min-1): tR 8.07 min. Final separation of 17
was achieved by applying isocratic elution of TEAA–MeOH 93 : 7
in 20 min (5 mL min-1): tR 16.73 min. Purity data obtained on
an analytical column for 15 and 17: tR 1.60 min; 3.73 min (99%
and 98% purity, respectively) using solvent system I (99 : 1 to 97 : 3
ammonium acetate–MeOH over 23 min, 1 mL min-1); tR 2.43 min;
5.72 min (93% and 99% purity respectively) using solvent system
II (99 : 1 to 97 : 3 of 0.01M KH2PO4 (pH=4.6)–MeOH over 23 min,
1 mL min-1).
1H NMR for 15 (600 MHz, D2O) d: 8.01 (s, 1H, H-8), 7.72
(d, J = 7.6, 1H, H-6), 5.88 (d, J = 7.6, 1H, H-5), 5.84 (d, J =
6.0, 1H, H-1¢G), 5.82 (d, J = 4.0, 1H, H-1¢C), 4.78 (dd, J = 6.0,
5.3,1H, H-2¢G), 4.46 (dd, J = 5.3, 3.3, 1H, H-3¢G), 4.31 (m, 1H,
H-4¢G), 4.24 (m, 3H, H-5¢G), 4.23 (m, 3H, H-5¢C), 4.22 (m, 3H,
H-3¢C), 4.18 (m, 3H, H-4¢C), 4.16 (m, 3H, H-2¢C), 4.15 (m, 3H,
H-5¢¢G), 4.12 (m, 1H, H-5¢¢C) ppm. 31P NMR (243 MHz, D2O) d:
-10.72 (s, Pa) ppm. 13C NMR (151 MHz, D2O) d :165.27 (C4C),
158.24 (C6G), 156.81 (C2C), 153.24 (C2G), 151.13 (C4G), 140.27
(C6C), 137.00 (C8G), 115.68 (C5G), 95.55 (C5C), 88.81 (C1¢C),
86.40 (C1¢G), 83.21 (d, JCP = 9.0, C4¢G), 81.74 (d, JCP = 9.0, C4¢C),
73.99 (C2¢C), 72.87 (C2¢G), 69.88 (C3¢G), 68.31 (C3¢C), 64.83 (d,
JCP = 4.0, C5¢G), 63.84 (d, JCP = 4.0, C5¢C) ppm. MS (MALDI):
m/z 667 ((M - H)-, 100%).
Guanosine-5¢-triphospho-5¢-cytidine (16). CMP(Bu3NH+ )2
was prepared from the CMP free acid (139 mg, 0.43 mmol) and
Bu3N (205 mL, 2 eq) in EtOH. GDP(Bu3NH+)2 was prepared
from GDP disodium salt (199.8 mg, 0.41 mmol) by passing it
through a column of activated Dowex 50WX-8 200 mesh, H+ form.
The column eluate was collected in an ice-cooled flask containing
Bu3N (2 eq) and EtOH. The resulting solution was freeze-dried
to yield GDP(Bu3NH+)2. Bis-tributylammonium CMP and CDI
(350.5 mg, 2.16 mmol, 5 eq) were stirred in dry DMF under argon
atmosphere at room temperature for 3.5 h. Dry MeOH (140 mL,
3.44 mmol, 8 eq) was added. After 5 min GDP(Bu3NH+)2 in dry
DMF (4 mL) was added. The resulting solution was stirred at room
temperature overnight. Then DMF was evaporated and a viscous
oil was obtained. The resulting oil was separated on Sephadex
DEAE-A25, eluting the product with a buffer gradient of 0–0.5 M
NH4HCO3 (total volume 1000 mL). The product was obtained in
ca. 3% yield (12 mg). Final purification of Gp3C was achieved by
HPLC isocratic elution with MeOH–0.1 M TEAA (pH = 7.5) 5 : 95
over 30 min, 5 mL min-1. Purity data obtained on an analytical
column for 16: tR 1.81 min (90% purity) using solvent system I
(100% ammonium acetate over 23 min, 1 mL min-1); tR 1.87 min
(99% purity) using solvent system II (100% of 0.01 M KH2PO4
(pH 4.6) over 23 min, 1 mL min-1).
1H NMR (600 MHz, D2O) d: 8.07 (s, 1H, H-8), 7.80 (d, J =
7.8, 1H, H-6), 5.91 (d, J = 7.8, 1H, H-5), 5.87 (d, J = 3.5, 1H,
H-1¢C), 5.86 (d, J = 6.5, 1H, H-1¢G), 4.49 (dd, J = 5.0, 3.0, 1H, H-
3¢G), 4.31 (quint, 1H, H-4¢G), 4.25 (m, 6H, H-3¢C), 4.24 (m, 6H,
H-5¢C), 4.24 (m, 6H, H-5¢¢C), 4.21 (m, 6H, H-2¢C), 4.20 (m, 6H,
H-5¢G), 4.20 (m, 6H, H-5¢¢G), 4.18 (m, 6H, H-4¢C) ppm. 31P NMR
(243 MHz, D2O) d: -10.78 (d, J = 18.7, Pa), -10.90 (d, J = 19.7,
Pa), -22.42 (t, J = 19.1, Pb) ppm. 13C NMR (151 MHz, D2O) d :
166.46 (C4C), 159.42 (C6G), 158.00 (C2C), 154.52 (C2G), 152.37
(C4G), 141.51 (C6C), 138.11 (C8G), 116.78 (C5G), 96.79 (C5C),
89.91 (C1¢C), 87.22 (C1¢G), 84.58 (d, JCP = 8.7, C4¢G), 82.82 (d,
JCP = 9.3, C4¢C), 75.08 (C2¢C), 74.50 (C2¢G), 71.32 (C3¢G), 69.33
(C3¢C), 65.95 (d, JCP = 4.5, C5¢G), 64.87 (d, JCP = 4.5, C5¢C) ppm.
MS (MALDI): m/z 747((M - H)-, 100%) 515(15), 548(48).
NMR Experiments
Preparation of NMR samples. dinucleotide sodium salts were
disssolved in 99.9% D2O and pD was adjusted to a physiological
pH (pD = 7.4
0.2, pD = pH + 0.4). NMR spectra were
measured at 600 MHz, 300 K 0.5, using glycerol as an internal
reference. The sample concentration was kept low enough to
avoid intermolecular base stacking yet, high enough to enable
the detection of clear and sharp NMR signals.42,51 Therefore, the
concentration of dinucleotides in the NMR samples was 7.5 mM.
1H NMR for 17 (600 MHz, D2O) d: 8.08 (d, J = 7.6, 1H, H-
6), 7.95 (s, 1H, H-8), 6.92 (d, J = 7.6, 1H, H-5), 5.97 (d, J =
5.7, 1H, H-1¢G), 5.82 (d, J = 2.7, 1H, H-1¢C), 4.71 (dd, J = 5.7,
5.3,1H, H-2¢G), 4.45 (dd, J = 5.3, 3.7, 1H, H-3¢G), 4.34 (m, 1H,
H-5¢C), 4.29 (m, 2H, H-4¢G), 4.28 (m, 2H, H-5¢G), 4.24 (m, 3H,
H-3¢C), 4.24 (m, 3H, H-4¢C), 4.21 (m, 3H, H-2¢C), 4.18 (m, 2H,
H-5¢¢G), 4.17 (m, 2H, H-5¢¢C), 3.77 (s, 3H, OMe) ppm. 31P NMR
(243 MHz, D2O) d: -10.48 (s, Pa) ppm. 13C NMR (151 MHz, D2O)
d : 163.53 (C4C), 159.29 (C6G), 157.24 (C2C), 154.80 (COOMe)
Analysis of 1H, 13C, 31P-NMR data for compounds 1–17.
The ribose conformation of the dinucleotides was analyzed in
terms of a dynamic equilibrium between two favored puckered
conformations North (N) conformer and South (S) conformer N
This journal is
The Royal Society of Chemistry 2010
Org. Biomol. Chem., 2010, 8, 4637–4652 | 4649
©

View Online
and S equilibrium populations were calculated from the observed
J
method was employed to treat electrostatics.61 The number of grid
point for the Ewald summation in each dimension was 32, while
the k parameter was set to 0.34 A . A spherical cutoff scheme
1¢2¢ and J3¢4¢ couplings as reported previously (eqn (1)–(2)).43,44,55,56
-1
˚
(1)
was employed for van der Waals and real-space electrostatic
J1¢2¢ = 9.3(1 - XN) = 9.3XS
˚
interactions with a value of 12.0 A. Heating dynamics were
performed from 50 to 300 K in 5 intervals of 5 psec with the velocity
Verlet algorithm for a total time of 25 psec (timestep 1 fsec).62
Equilibration of the system at 300 K was performed for 75 psec in
the constant particle–volume–temperature (NVT) ensemble using
the Nose thermostat with an effective mass of 100 kcal mol-1-ps2.
Production dynamics was performed at 300 K for Ap2A and Ap4A
in the particle–pressure–temperature (NPT)63 ensemble employing
the leapfrog algorithm.62 Specifically, the extended system pressure
algorithm of Andersen with an effective mass of 500.0 amu,64and
the Hoover thermostat with an effective mass of 1000.0 kcal
mol-1-ps2.65 For each dinucleotide a 10 ns production simulation
was performed. The trajectories were analyzed and structural
data were collected. All the minimizations and simulations were
performed with CHARMM program59,66 and the CHARMM27
force-field.67,68 All visualization employed Discovery Studio 2.1
(Accelrys Inc.).
The distance between the dinucleotide bases is defined as the
distance between the glycosidic nitrogen atoms of the bases (N9-
N92).26,27 The torsional angles were defined according to Saenger
(Fig. 3).46 c: O4¢–C1¢–N9–C4, a: O3P–P–O5¢–C5¢, b: P–O5¢–C5¢–
C4¢, g : O5¢–C5¢–C4¢–C3¢. Endocyclic torsion angles: n0: C4¢–O4¢–
C1¢–C2¢, n1: O4¢–C1¢–C2¢–C3¢, n2: C1¢–C2¢–C3¢–C4¢, n3: C2¢–
C3¢–C4¢–O4¢, n4: C3¢–C4¢–O4¢–C1¢.
(2)
J3¢4¢ = 9.3XN
Information on the torsional angles around C4¢–C5¢ (g -angle)
and C5¢–O5¢ (b-angle), Fig. 3, was obtained by measuring J4¢5¢ and
4¢5¢¢, JP5¢ and JP5¢¢ values and using previously reported equations
eqn (3)–(5) and (6)–(8), respectively44
J
(3)
(4)
(5)
(6)
(7)
(8)
rgg = [(Jt + Jg) – (J4¢5¢ + J4¢5¢¢)]/(Jt - Jg)
rtg = (J4¢5¢ - Jg)/(Jt - Jg)
rgt = (J4¢5¢¢ - Jg)/(Jt - Jg)
where Jg = 2.04 Hz, Jt = 11.72 Hz.
rgg = [(Jt + Jg) - (JP5¢ + JP5¢¢)]/(Jt - Jg)
rtg = (JP5¢ - Jg)/(Jt - Jg)
rgt = (JP5¢¢ - Jg)/(Jt - Jg)
The conformation around the glycosidic bond is defined as anti
for 180 90◦ and syn for 0 90◦.In the syn region, the range
between 270–330◦ is also referred to as high anti. The exocyclic
group torsion angles b and g are in the gg range when b = 180
30◦ and g = 60 30◦. Pseudorotation was calculated according to
Altona and Sundaralingam:55
where Jt = 20.9 Hz, Jg = 1.8 Hz.
The determination of the conformation around the glycosidic
bond (c-angle), Fig. 3., was obtained by Karplus-type equations
using the vicinal coupling constants JC8/6-H1¢ and JC4/2-H1¢, which
were extracted from 13C NMR spectra for each dinucleotide (eqn
(9)–(10)).57
3
3
3JC6/8-H1¢ = 4.5cos2(c - 60◦) - 0.6cos(c - 60◦) + 0.1
(9)
(11)
Where we add 180 to the calculated P if n2 < 0. Here P = 0 is for
the Northern conformer and P =180 is for the Southern conformer.
3JC2/4-H1¢ = 4.7cos2(c - 60◦) + 2.3cos(c - 60◦) + 0.1
(10)
p–p Stacking interactions were deduced based on Dd of H2/H8
of the adenine/guanine ring and H5/H6 of the uracil/cytosine
ring in each of the analogues as compared to their values in ATP
and GTP, UTP, and CMP, correspondingly.
Crystal simulation of Ap4A
Coordinates for Ap4A were obtained from the Cambridge
Structural Database (entry HIFKEM)23 for the tetrasodium-
dodecahydrate form of Ap4A. We tested the CHARMM27 force
field for the dinucletides by crystal minimization and simulation.69
The crystal unit cell of Ap4A with 4 sodium ions and 12 water
molecules was minimized by 120 cycles of the ABNR method.
A minimization of the lattice parameters was done as well by 60
cycles of the ABNR method. The system was heated stepwise from
50 to 293 K23 for 25 psec in 4 intervals of 50 K and one final interval
of 43 K, and further equilibrated for 25 psec in the NVT ensemble.
Finally, a simulation was run in the NPT ensemble for 1 ns.
MD Simulations
System preparation. Initially, the Ap2A molecule was built in
its extended conformation with the bases in the anti conformation
(as in mono-nucleotides43). The initial conformation of Ap4A was
generated from crystallographic data.23 Each dinucleotide was
3
˚
placed in a 40 ¥ 40 ¥ 40 A pre-equilibrated box of 2044 TIP3
water molecules.58 Sodium counter ions were added in order to
neutralize the system. Water molecules that overlapped with the
˚
solute or sodium ions were deleted (<2.5 A heavy atom distance).
PMF simulations
The initial systems were minimized by 150 cycles of the Adopted
Basis Newton-Raphson (ABNR) method.59 Periodic boundary
conditions were applied. The hydrogen atom–heavy atom bond
lengths were constrained by the SHAKE algorithm.60 A relative
dielectric constant of 1.0 was used. The Particle Mesh Ewald
The PMF was obtained from umbrella sampling MD
simulations.70 The PMF was defined as:
W(z) = -kBT lnr*(z) - Ubias(z)
(12)
4650 | Org. Biomol. Chem., 2010, 8, 4637–4652
This journal is The Royal Society of Chemistry 2010
©

View Online
where kB is the Boltzman constant, T is the temperature and r*(z)
is the probability distribution in the presence of a biasing potential.
To enable uniform sampling along the reaction coordinate, z, a
biasing potential is added to the potential energy of the system,
References
1 A. G. McLennan, Pharmacol. Ther., 2000, 87, 73–89.
2 C. H. V. Hoyle, R. H. Hilderman, J. J. Pintor, H. Schluter and B. F.
King, Drug Dev. Res., 2001, 52, 260–273.
3 V. Ralevic, J. Jankowski and H. Schluter, Br. J. Pharmacol., 2001, 134,
1073–1083.
U
bias(z) = Uumbr(z) + Uharm(z). The umbrella potential, Uumbr(z), is
defined as -W(z), which essentially removes the free energy barrier,
thereby enabling uniform sampling. Furthermore, the reaction
coordinate is divided into regions (windows) centered around z0,
and a harmonic potential, Uharm(z) = k(z - z0)2/2, is added.
Here k is a force constant, z is the reaction coordinate defined
as the distance between the N9 atoms in the two nucleotide bases
(Fig. 9)26 and z0 is the distance which the system is restrained to in a
specific simulation window. This allows the dinucleotide/solvent
system to relax along the reaction coordinate. Practically, since
W(z) is required for Uumbr (z) and the optimal force constant, k,
in Uharm (z) are not known at the outset, the PMF is obtained
by employing adaptive umbrella sampling molecular dynamics
simulations whereby the biasing potential is refined in an iterative
manner.
4 E. G. Delicado, M. T. Miras-Portugal, L. M. Carrasquero, D. Leon, R.
Perez-Sen and J. Gualix, Pflugers Arch.-Eur. J. Physiol., 2006, 563–572.
5 A. Guzman-Aranguez, A. Crooke, A. Peral, C. H. V. Hoyle and J.
Pintor, Prog. Retinal Eye Res., 2007, 26, 674–687.
6 J. G. Douglass, R. I. Patel, B. R. Yerxa, S. R. Shaver, P. S. Watson, K.
Bednarski, R. Plourde, C. C. Redick, K. Brubaker, A. C. Jones and J. L.
Boyer, J. Med. Chem., 2008, 51, 1007–1025.
7 A. G. McLennan, Ap4A and other dinucleoside polyphosphates, CRC
Press, Boca Raton, Florida, 1992.
8 K. A. Henzler-Wildman, V. Thai, M. Lei, M. Ott, M. Wolf-Watz, T.
Fenn, E. Pozharski, M. A. Wilson, G. A. Petsko, M. Karplus, C. G.
Hubner and D. Kern, Nature, 2007, 450, 838–844.
9 S. R. Shaver, J. L. Rideout, W. Pendergast, J. G. Douglass, E. G. Brown,
J. L. Boyer, R. I. Patel, C. C. Redick, A. C. Jones, M. Picher and B. R.
Yerxa, Purinergic Signalling, 2005, 1, 183–191.
10 E. Holler, B. Holmquist, B. L. Vallee, K. Taneja and P. Zamecnik,
Biochemistry, 1983, 22, 4924–4933.
11 M. Ikehara, S. Uesugi and K. Yoshida, Biochemistry, 1972, 11, 836–842.
12 N. H. Kolodny and L. J. Collins, J. Biol. Chem., 1986, 261, 14571–14575.
13 N. H. Kolodny, E. Kisteneff, C. Redfield and E. Rapaport, FEBS Lett.,
1979, 107, 121–124.
The PMF windows were connected using the weighed-
histogram analysis method.71 The reaction coordinate is expected
to be representative of the principal stacking features, although
there can be situations where bases are close but not parallel (i.e.
T-stacked as opposed to parallel stacking).26,27
14 R. B. Westkaemper, Biochem. Biophys. Res. Commun., 1987, 144, 922–
929.
15 J. W. Pettegrew, D. W. Miles and H. Eyring, Proc. Natl. Acad. Sci.
U. S. A., 1977, 74, 1785–1788.
16 J. F. Scott and P. C. Zamecnik, Proc. Natl. Acad. Sci. U. S. A., 1969,
64, 1308–1314.
17 J. Stepinski, M. Bretner, M. Jankowska, K. Felczak, R. Stolarski, Z.
Wieczorek, A.-L. Cai, R. E. Rhoads and A. Temeriusz, Nucleosides,
Nucleotides Nucleic Acids, 1995, 14, 717–721.
18 K. H. Mayo, O. M. Mvele and R. N. Puri, FEBS Lett., 1990, 265,
97–100.
For each energy profile there were 22 windows. The reaction
˚
coordinate range was 3.5–14.0 A for both dinucleotides. Each
window was equilibrated for ca. 50 psec and further sampled for
ca. 500 ps. Convergence was verified by building the PMF in stages
starting from 100 ps to 500 ps with increments of 100 ps. The PMFs
obtained from the different stages were identical to within 1 kcal
mol-1. Based on this analysis, we estimate the statistical error to
be ca. 1 kcal mol-1, (see Supplementary Material Fig. S12–S17†).
More PMF simulations were run for 1 ns for Ap2A and Ap4A in
19 K. Scheffzek, W. Kliche, L. Wiesmueller and J. Reinstein, Biochemistry,
1996, 35, 9716–9727.
20 M. B. Berry and G. N. Phillips, Proteins: Struct., Funct., Genet., 1998,
32, 276–288.
˚
all the stacking modes at 4.5 A, except for Ap4A a–b which was at
21 M. D. Baker, D. E. Holloway, G. J. Swaminathan and K. R. Acharya,
Biochemistry, 2006, 45, 416–426.
22 D. K. Simanshu, H. S. Savithri and M. R. N. Murthy, Proteins: Struct.,
Funct., Bioinf., 2008, 70, 1379–1388.
23 D. Watanabe, M. Ishikawa, M. Yamasaki, M. Ozaki, T. Katayama and
H. Nakajima, Acta Crystallogr., Sect. C: Cryst. Struct. Commun., 1996,
52, 338–340.
˚
˚
6.5 A, corresponding to the stacked configuration, and at 9.0 A,
corresponding to the open configuration. The torsion angles c, a,
b and g were monitored as well as pseudorotation angle and the
distances N6–O22¢, N62–O2¢.
24 J. M. Thornton and P. M. Bayley, Biopolymers, 1976, 15, 955–975.
25 J. Norberg and L. Nilsson, Biophys. J., 1994, 67, 812–824.
26 J. Norberg and L. Nilsson, Biophys. J., 1995, 69, 2277–2285.
27 J. Norberg and L. Nilsson, J. Am. Chem. Soc., 1995, 117, 10832–10840.
28 J. Norberg and L. Nilsson, J. Phys. Chem., 1996, 100, 2550–2554.
29 P. Petrova, J. Koca and A. Imberty, Eur. J. Biochem., 2001, 268, 5365–
5374.
Abbreviations
A
G
U
C
ApnAs
GpnGs
CD
adenine
guanine
uracil
cytosine
diadenosine polyphosphates
diguanosine polyphosphates
circular dichroism
30 P. Petrova, C. Monteiro, C. Herve du Penhoat, J. Koca and A. Imberty,
Biopolymers, 2001, 58, 617–635.
31 J. Liao, S. Sun, D. Chandler and G. Oster, Eur. Biophys. J., 2004, 33,
29–37.
32 N. Foloppe and L. Nilsson, J. Phys. Chem. B, 2005, 109, 9119–9131.
33 W. Pendergast, B. R. Yerxa, J. G. Douglass, S. R. Shaver, R. W.
Dougherty, C. C. Redick, I. F. Sims and J. L. Rideout, Bioorg. Med.
Chem. Lett., 2001, 11, 157–160.
34 H. Sigel and R. Griesser, Chem. Soc. Rev., 2005, 34, 875–900.
35 K. H. Scheller and H. Sigel, J. Am. Chem. Soc., 1983, 105, 5891–5900.
36 O. K. Kalinowski, S. Berger, and S. Braun, Carbon-13 NMR Spec-
troscopy, John Wiley and Sons, Chichester, 1994.
37 T. Son and C. Chachaty, Biochim. Biophys. Acta., 1977, 500, 405–418.
38 M. Remin and D. Shugar, Biochem. Biophys. Res. Commun., 1972, 48,
636–642.
39 C. Giessner-Prettre and B. Pullman, J. Theor. Biol., 1970, 27, 87–95.
40 P. O. Ts’o, N. S. Kondo, M. P. Schweizer and D. P. Hollis, Biochemistry,
1969, 8, 997–1029.
NMP
NDP
NTP
CDI
DCC
COSY
HMQC
HMBC
NOESY
MD
nucleoside monophosphate
nucleoside diphosphate
nucleoside triphosphate
carbonyl diimidazole
dicyclohexylcarbonyldiimide
correlation spectroscopy
heteronuclear multiple quantum coherence
heteronuclear multiple bond coherence
nuclear Overhauser effect spectroscopy
molecular dynamics
PMF
potential of mean force
This journal is
The Royal Society of Chemistry 2010
Org. Biomol. Chem., 2010, 8, 4637–4652 | 4651
©

View Online
41 M. P. Schweizer, S. I. Chan and P. O. P. Ts’o, J. Am. Chem. Soc., 1965,
87, 5241–5247.
59 B. R. Brooks, R. E. Bruccoleri, B. D. Olafson, D. J. States, S.
Swaminathan and M. Karplus, J. Comput. Chem., 1983, 4, 187–217.
60 J. P. Ryckaert, G. Ciccotti and H. J. C. Berendsen, J. Comput. Phys.,
1977, 23, 327–341.
42 K. H. Scheller and H. Sigel, J. Am. Chem. Soc., 1983, 105, 5891–5900.
43 D. B. Davies, Prog. Nucl. Magn. Reson. Spectrosc., 1978, 12, 135–225.
44 D. B. Davies and S. S. Danyluk, Biochemistry, 1974, 13, 4417–4434.
45 M. Cohn, Structural and Chemical properties of ATP and its Metal
Complexes in Solution, The New York Academy of Sciences, New York,
1990.
61 K. Nam, J. Gao and D. M. York, J. Chem. Theory Comput., 2005, 1,
2–13.
62 M. P. Allen, and D. J. Tildesley, Computer Simulation of Liquids, Oxford
University Press, Oxford, U.K., 1987.
46 W. Saenger, Principles of Nucleic Acid Structure, Springer-Verlag, New-
63 S. E. Feller, Y. Zhang, R. W. Pastor and B. R. Brooks, J. Chem. Phys.,
York, 1984.
1995, 103, 4613–4621.
47 K. H. Scheller, F. Hofstetter, P. R. Mitchell, B. Prijs and H. Sigel, J. Am.
Chem. Soc., 1981, 103, 247–260.
48 J. Sponer, K. E. Riley and P. Hobza, Phys. Chem. Chem. Phys., 2008,
10, 2595–2610.
49 K. M. Guckian, B. A. Schweitzer, R. X.-F. Ren, C. J. Sheils, D. C.
Tahmassebi and E. T. Kool, J. Am. Chem. Soc., 2000, 122, 2213–2222.
50 J. Sponer, P. Jurecka, I. Marchan, F. J. Luque, M. Orozco and P. Hobza,
Chem.–Eur. J., 2006, 12, 2854–2865.
51 H. Sigel, R. Tribolet, R. Malinibalakrishnan and R. B. Martin, Inorg.
Chem., 1987, 26, 2149–2157.
64 H. C. Andersen, J. Chem. Phys., 1980, 72, 2384–2393.
65 W. G. Hoover, Phys. Rev. A, 1985, 31, 1695–1697.
66 B. R. Brooks, C. L. Brooks, III, A. D. Mackerell, Jr., L. Nilsson, R. J.
Petrella, B. Roux, Y. Won, G. Archontis, C. Bartels, S. Boresch, A.
Caflisch, L. Caves, Q. Cui, A. R. Dinner, M. Feig, S. Fischer, J. Gao,
M. Hodoscek, W. Im, K. Kuczera, T. Lazaridis, J. Ma, V. Ovchinnikov,
E. Paci, R. W. Pastor, C. B. Post, J. Z. Pu, M. Schaefer, B. Tidor,
R. M. Venable, H. L. Woodcock, X. Wu, W. Yang, D. M. York and M.
Karplus, J. Comput. Chem., 2009, 30, 1545–1614.
67 N. Foloppe and J. MacKerell, A.D., J. Comput. Chem., 2000, 21, 86–
52 J. Norberg and L. Nilsson, Acc. Chem. Res., 2002, 35, 465–472.
53 A. F. Pozharskii, Khim. Geterotsikl. Soedin., 1985, 867–905.
54 B. W. K. Shum and D. M. Crothers, Biopolymers, 1983, 22, 919–
933.
104.
68 J. MacKerell, A. D. and N. Banavali, J. Comput. Chem., 2000, 21,
105–120.
69 A. D. MacKerell, Jr., D. Bashford, M. Bellott, R. L. Dunbrack, J. D.
Evanseck, M. J. Field, S. Fischer, J. G. Gao, H., S. Ha, D. Joseph-
McCarthy, L. Kuchnir, K. Kuczera, F. T. K. Lau, C. Mattos, S.
Michnick, T. Ngo, D. T. Nguyen, B. Prodhom, W. E. Reiher, III, B.
Roux, M. Schlenkrich, J. C. Smith, R. Stote, J. Straub, M. Watanabe, J.
Wiorkiewicz-Kuczera, D. Yin and M. Karplus, J. Phys. Chem. B, 1998,
102, 3586–3616.
70 G. M. Torrie and J. P. Valleau, J. Comput. Phys., 1977, 23, 187–199.
71 S. Kumar, D. Bouzida, R. H. Swendsen, P. A. Kollman and J. M.
Rosenberg, J. Comput. Chem., 1992, 13, 1011–1021.
55 C. Altona and M. Sundaralingam, J. Am. Chem. Soc., 1972, 94, 8205–
8212.
56 C. Altona and M. Sundaralingam, J. Am. Chem. Soc., 1973, 95, 2333–
2344.
57 J. H. Ippel, S. S. Wijmenga, R. De Jong, H. A. Heus, C. W. Hilbers, E.
De Vroom, G. A. Van der Marel and J. H. Van Boom, Magn. Reson.
Chem., 1996, 34, S156–S176.
58 W. L. Jorgensen, J. Chandrasekhar, J. D. Madura, R. W. Impey and
M. L. Klein, J. Chem. Phys., 1983, 79, 926–935.
4652 | Org. Biomol. Chem., 2010, 8, 4637–4652
This journal is
The Royal Society of Chemistry 2010
©