B

w
doi:10.1016/S0022-2836(02)00428-X available online at http://www.idealibrary.com on J. Mol. Biol. (2002) 320, 189–200

Crystal Structure of the Potassium Form of an

Oxytricha nova G-quadruplex
Shozeb Haider, Gary N. Parkinson and Stephen Neidle*
The Cancer Research UK
Biomolecular Structure Unit
Chester Beatty Laboratories
The Institute of Cancer
Research, 237 Fulham Road
London SW3 6JB, UK
*Corresponding author
The crystal structures of the potassium-containing quadruplex formed
from the Oxytricha nova sequence d(GGGGTTTTGGGG) are reported, in
two space groups, the orthorhombic P212121 and the trigonal P3221,
which diffract to 2.0 Å and 1.49 Å, respectively. The orthorhombic form
contains two independent quadruplexes in the asymmetric unit, and
the trigonal form contains one. All three of these quadruplexes adopt an
identical fold, with two strands forming an antiparallel diagonal arrange-
ment. This is identical with that observed previously in NMR studies of
the native sodium and potassium forms, and a crystallographic analysis
of it complexed with an O. nova protein. The present analysis demon-
strates that the native structure is the same in solution and in the crystal-
line state and, moreover, that the nature of the counter-ion does not
affect the overall fold of this quadruplex. The analysis corrects an earlier
crystallographic study of this quadruplex. The conformation of the tetra-
thymine loop is described in detail, which involves the third thymine
base folding back to interact with the first thymine base. The water net-
works in the grooves and loops are described and, in particular, the ability
of water molecules to form a continuous spine of hydration in the narrow
groove is detailed. Each quadruplex has five potassium ions organised in
a linear channel, with square antiprismatic coordination to each ion from
oxygen atoms.
q 2002 Elsevier Science Ltd. All rights reserved
Keywords: G-quadruplex; crystal structures; Oxytricha telomere;
potassium ions

Introduction in a few promoter sequences, such as that of the

Telomeres are the specialised ends of linear
chromosomes.1 – 4 They contain tandem repeats
of short guanine-rich DNA sequences such as
d(TTAGGG) in vertebrates, d(TTGGGG) in Tetra-
hymena and d(TTTTGGGG) in Oxytricha, and are
associated with a number of telomere-specific pro-
teins. Telomeres protect chromosome ends from
recombination, aberrant fusion events and from
being recognised as damaged DNA. They have
been implicated in the formation of nuclear
domains that may be important for transcriptional
regulation, for sister chromatid pairing at mitosis
and for homologous meiotic synapsis. Analogous
guanine-rich tracts are present in some immuno-
globulin switch regions of higher organisms and
Abbreviation used: MAD, multiple wavelength
anomalous dispersion.
E-mail address of the corresponding author:
s.neidle@icr.ac.uk
c-myc oncogene.5
The length of telomeric DNA can range from
tens of bases to over 100 kb in mice. Telomeres
are typically 5– 10 kb long in human cells. The
100– 200 bases at the extreme 30 end of vertebrate
telomeres are single-stranded, and can fold back
into a double-stranded region to form the so-called
t-loop structure,6 although this arrangement has
not as yet been described in atomic-level detail.
Telomeric guanine-rich oligonucleotides can adopt
a variety of unusual DNA structures of which
four-stranded quadruplex DNAs7,8 are the best
characterised. There are various types of four-
stranded arrangements built upon the motif of
the guanine quartet (Figure 1), which has four
in-plane guanine bases hydrogen-bonding together
by means of Hoogsteen and Watson – Crick base-
pairing. G-quadruplexes, can be formed from one,
two or four separate DNA strands, with varying
topologies.7 They require cations for the main-
tenance of structural integrity, with sodium and
especially potassium ions providing optimal

0022-2836/02/$ - see front matter q 2002 Elsevier Science Ltd. All rights reserved
190
Figure 1. The guanine quartet, coloured by atom types.
effect.8 A number of NMR structures have
been reported for DNA quadruplexes,9 but only
two crystal structure determinations of native
quadruplexes to date. One is for the intermo-
lecular quadruplex formed by four strands of
d(TGGGGT).10,11 All four strands in this structure
are parallel, with the quadruplex consisting of
four stacked planes of guanine-quartets, in com-
plete accord with NMR studies on this sequence.12
The second crystal structure is of the potassium
form of the ciliate Oxytricha nova sequence d(GGG-
GTTTTGGGG), at 2.5 Å resolution.13 The structure,
in space group P212121, has been reported to be a
dimer of two strands forming a stack of four gua-
nine-quartets held together with a folding topology
of two lateral thymine loops in each quadruplex
(Figure 2(a)). Adjacent strands are antiparallel in
this hairpin dimer arrangement. By contrast, the
NMR structure of the sodium form in solution
shows (Figure 2(b)) a quadruplex with distinct
diagonal loops.14,15 This conclusion was confirmed
Figure 2. Schematic of the folding of d(GGGGTTTTGGGG): (a) the crystal structure;13 (b) the NMR structures.14 – 16
Guanosine nucleosides with an anti conformation are indicated in blue, and those in a syn conformation are in yellow.
Crystal Structure of Oxytricha nova Quadruplex
and extended in a subsequent NMR study, which
also concluded that the potassium form maintains
the same fold as the sodium-bound one.16 The
same topology has been found in the solution
structures of the quadruplexes formed from
d(GGGTTTTGGG).17 – 19 The crystal structure of the
single-stranded O. nova telomere-binding protein
has been reported recently,20 with the d(GGGGTT-
TTGGGG) sequence bound. Unexpectedly, a dis-
crete quadruplex structure, surrounded by protein,
has been observed in this structure (which contains
sodium ions), with the same topology as that
found in the NMR studies on this sequence. The
discrepancy in folds between the native crystal
and NMR structures of d(GGGGTTTTGGGG) has
been ascribed to the differing stabilising abilities
of potassium versus sodium ions.
As part of a programme to examine the struc-
tures and small-molecule interactions of quadru-
plex DNAs,21 – 24 we have re-determined the crystal
structure of the original sequence d(GGGGTTTT-
GGGG) in a potassium environment, both in the
original space group and in a new crystal form.
These structures allow us now to unequivocally
determine the fold of this quadruplex in the
crystalline state, and to assess in detail the struc-
tural roles of potassium ions and water molecules.
Results
Overall features of the structure
We have solved the crystal structures of the
sequence d(GGGGTTTTGGGG) in two space
groups, the orthorhombic P212121 and the trigonal
P3221. These two crystal structures have two and
one independent quadruplex molecules, respec-
tively, in the crystallographic asymmetric unit.
Thus we have a total of three independent quadru-
plexes to compare (Figure 3). The analyses show
that all have the same diagonal fold topology,
which is identical with that reported from the
Crystal Structure of Oxytricha nova Quadruplex 191

Table 1. The rms deviations (Å) between the various orthorhombic structure, even though both are in
O. nova quadruplex structures the same space group and have very similar cell
Trigonal Orth1 Orth2 1JB7 156D 1K4X
dimensions. The rmsd between the various struc-
tures having the diagonal fold topology are given
Trigonal – 0.35 0.38 0.65 1.28 2.46 in Table 1. The rmsd values between the present
Orth1 0.35 – 0.21 0.65 1.24 2.43 structures, especially between the two independent
Orth2 0.38 0.21 – 0.66 1.24 2.43
1JB7 0.65 0.65 0.66 – 1.18 – molecules in the orthorhombic crystal, are almost
156D 1.28 1.24 1.24 1.18 – 2.41 within experimental error, indicating no significant
1K4X 2.46 2.43 2.43 – 2.41 – deviation from equivalence for all three of the
The trigonal, Orth1 and Orth2 are the three structures
quadruplex molecules presented here. They are
reported here, and the others are indicated by their PBD codes. also closely similar to the Naþ crystal structure in
1JB7 is that from the crystal structure of the Naþ form in the the O. nova protein complex,20 indicating that both
O. nova protein complex,20 156D is that of the refined Naþ NMR the overall fold and the detailed structure does
model,15 and 1K4X is that of the Kþ NMR model.16 not change significantly with crystallographic
environment. The rmsd values between these and
the Naþ-containing NMR model15 are all . 1 Å,
NMR studies14 – 16 on both sodium and potassium- and . 2 Å compared to the Kþ-containing NMR
bound forms, as well as that in the O. nova protein model.16 Again, these reflect the overall identities
complex.20 This equivalence extends to almost all of overall fold between the crystallographic and
details of the structures. The structure of the NMR structures; the larger rmsd values mostly
orthorhombic form reported here is thus quite dis- reflect the differences between the loop geometries
tinct from (and corrects) the earlier report13 of an in the crystal structures and NMR models.

Figure 3. (a) Two sets of orthogonal views of the crystal structures reported here. In each view the two quadruplexes
in the orthorhombic structure are at the top, and the trigonal structure is below. Thymine bases are coloured red and
guanine bases are green. (b) Stereo view of the structure of the trigonal quadruplex, showing the five potassium ions.
192 Crystal Structure of Oxytricha nova Quadruplex

Figure 4. Plot of groove widths for the d(GGGGTTTTGGGG) quadruplex, comparing them in the present structures,
in the sodium-containing NMR15 and previous crystal structures,13 and in the O. nova protein complex.20

The G-quartets pared to the other structures. Figure 4 shows that

one consequence of the incorrect fold assignment
All four quartets in each crystal structure have
in the earlier crystal structure13 is to abolish the
all Hoogsteen N1· · ·O6 and N2· · ·N7 hydrogen
wide groove found in all the other structures.
bonds intact, with average distances of 2.88(8) and
2.88(8) Å, respectively. The glycosidic torsion
angles for guanine bases are alternating syn –anti The potassium ions
along each strand and syn –syn – anti – anti within
An almost perfectly linear row of five equi-
each quartet (Figure 2). All guanine bases in the
spaced potassium ions lies along the helical axis
central two quartets in each quadruplex are
within the central core of all three quadruplexes,
approximately coplanar. One guanine base in each
terminal quartet is slightly more tilted than
the others and stacks effectively with the adjacent
30 -thymine base. The average rise between adjacent
quartets is consistently 3.3 Å.

The grooves
As a result of the diagonal topology, a total of
one wide, two medium and one narrow groove is
present in each quadruplex. The distance between
the adjacent strands generates inverse tapered
groove-widths for the narrow groove in particular
(and rather less so for the medium grooves), so
that it is broad at the ends and relatively narrower
in the centre. Figure 4 shows that each groove
type is remarkably similar in width across all
three of the independent quadruplex molecules
reported here. The grooves of the Naþ-containing
NMR14 and protein-complexed20 ones are also
very close in width to these, although some small
differences are apparent for the wide groove. All
the wide and medium grooves in the Kþ-contain-
ing NMR model16 are consistently narrower than Figure 5. The linear array of potassium ions in the
in any of these other structures, by a maximum of structures, showing the first coordination shell of each
several ångström units, and, conversely, the ion. Solvent oxygen atoms are indicated by S and
narrow groove is widened in this, by .1 Å com- thymine O2 atoms by T.
Crystal Structure of Oxytricha nova Quadruplex
Figure 6. Plot of the loop structure in the trigonal
quadruplex, showing the potassium ion and directly
coordinated water molecules, and thymine nucleotides
1, 2 and 4 in red. Guanine bases are coloured blue. The
thymine bases are coloured brown, except for the third
thymine base (labelled T7, coloured blue), which is
flipped out of the loop and oriented in the reverse direc-
tion so that it stacks on a guanine base (in the 50 to 30
direction) at the same time as hydrogen bonding to the
first thymine base. The water molecules are coloured
red and the potassium ion in the loop is coloured green.
Dotted lines show the various co-ordinations and the
hydrogen bonding interactions between the water
molecules and the thymine bases, and between the two
thymine bases T5 and T7.
in both crystal forms (Figure 5). The average
Kþ· · ·Kþ distance is 3.38(9) Å. Each quadruplex has
three potassium ions embedded within the planes
of the quartets and one each within the thymine
loops, i.e. outside the stack of G-quartets. All the
potassium ions have full occupancy and low-
temperature factors, with the least mobility being
for those in the trigonal structure (Table 3). This
reflects the overall greater rigidity and consequent
higher resolution of the trigonal structure
compared to the orthorhombic one. All the quadru-
plexes have three octahedrally coordinated potas-
sium cations, each sandwiched symmetrically
between two planes of the four G-quartets, and
193
each forming a square antiprismatic arrangement.
By contrast, sodium ions have been suggested to
lie close to each quartet plane.11,15 The outer potas-
sium ions are located within the loops, where
they also achieve octahedral coordination with the
outer G-quartet carbonyl oxygen atoms, and the
O2 atoms from the adjacent thymine bases,
together with two oxygen atoms donated by water
molecules. These potassium ions have only slightly
increased mobilities compared to the three central
ones. Binding sites, within the G-quartets, have
been suggested for at least three K ions by the
NMR study, although it was not possible to gener-
ate models consistent with the NMR data that
have Kþ (or NHþ4 ions) in loop regions.16
The loops
All six loops in the three quadruplexes analysed
here, each consisting of four consecutive thymine
bases, have closely similar conformations. They
are also very similar to those found in the quadru-
plex within the O. nova protein complex, though
less so to those in the NMR structures. In each of
the present loops, the first thymine base in the 30
to 50 direction is stacked on a guanine base in a
terminal quartet (Figure 6). Its adjacent thymine
base stacks in turn on it. The third thymine loops
back to stack on the adjacent guanine base, from
the other strand. It forms an O2· · ·N3 hydrogen
bond (of length 3.0(1) Å) with the first thymine
base. The fourth thymine base is swung back in
the other direction so that it stacks on the second
thymine base.
Each loop interacts with an adjacent loop in the
two crystal structures (Figure 7). In the orthor-
hombic form, the two crystallographically-inde-
pendent quadruplexes interact with each other
via these loops. Thymine bases 2 and 4 from one
quadruplex hydrogen-bond with the equivalent
thymine bases 2 and 4 on the other quadruplex,
to form a pair of pseudo 2-fold related thymine· · ·
thymine base-pairs. Both hydrogen bonding are
O4· · ·N2, and are short: 2.6 – 2.7 Å. The exocyclic
methyl groups are in close non-bonded contact.
This pattern of thymine interactions is essentially
identical with that occurring in the crystal lattice
Figure 7. Stereo view of the symmetry-
related interactions of the loops, showing two
molecules in the trigonal space group.
194
Figure 8. The residue numbering scheme used here.
between symmetry-related molecules, involving
the thymine loops at the other ends of the pair of
orthorhombic quadruplexes, and both loops in the
trigonal crystal structure. In these instances, there
is exact 2-fold symmetry relating the crystallo-
graphically equivalent loops. We cannot discount
the possibility that the particular loop confor-
mations that we observe are the consequence
of crystal-packing interactions, although their
similarity to those in the quadruplex within
the O. nova protein complex,20 suggests other-
wise. There is clearly considerable flexibility in
the loops, as seen in the differences between them
in the sodium versus potassium forms in
solution.14 – 16
Water networks
All three quadruplexes are heavily hydrated. It is
apparent that no one groove has precisely the same
pattern of observed hydration in each of the three
quadruplexes reported here. This reflects, to some
extent, the non-atomic resolution (and hence the
limitations in refinement) of these three structures,
which hinders us from locating all possible asso-
ciated water molecules. However, the differences
in the detail of first-shell hydration patterns for a
particular groove suggest that water molecules are
more mobile than in the very constant spine
observed in A/T duplex minor grooves. However,
it is possible to discern several consistent patterns
in the hydration networks. In general, water
molecules tend to cluster around phosphate ions,
and there are many water· · ·water bridges across
the grooves, linking one phosphate group with
another. These and other water bridges are a reflec-
tion of the width of each groove. There is a marked
preference for water to hydrogen bond to the N2
Crystal Structure of Oxytricha nova Quadruplex
exocyclic amino group of guanine, whereas
contacts with N3 are not made. The sugar ring
atom O40 is also commonly observed as a hydro-
gen-bond donor to water molecules. The residue
numbering scheme is shown in Figure 8.
All the three narrow grooves show strong simi-
larities in their hydration and a complete spine is
observed each of the three narrow grooves (Figure
9(a) and (b)). The spine of hydration in the narrow
groove of one orthorhombic quadruplex begins at
the O2 of Thy-B8. Due to the alternating syn– anti
arrangement of the bases and the sugar puckering,
the phosphate groups face inwards into the groove.
The spine runs down the groove, making contacts
with phosphate groups on both backbone strands
forming the narrow groove. The water molecules
also form hydrogen bonding with the N2 group of
the guanine residues. The prominent interactions
in the groove are the two-water molecule bridge
between the N2 atoms of GuaB9 and GuaA11 on
the opposite strand and a single water molecule
that bridges between the phosphate groups of the
two Gua10 nucleotides. At the wide end of this
groove, where the phosphate groups are . 8.6 Å
apart, a bridge of three water molecules links the
phosphate groups of Gua-B/D12 and Gua-A/C9,
the second of which also hydrogen bonds to N2 of
this Gua-A/C9. A single water molecule bridges
the phosphate groups of Gua-A11 and Gua-A12 of
the same strand. Furthermore, the spine continues
in the narrow groove of the second orthorhombic
quadruplex, with two water molecules diagonally
bridging the phosphate groups of Gua-B12 of one
quadruplex with the Gua-D12 of the other (this
extended spine is not observed in the other
grooves of the orthorhombic form). The spine
ends with the terminal water molecule hydrogen
bonding with the phosphate group of Thy-D8. The
overall pattern of contacts is maintained in the
Crystal Structure of Oxytricha nova Quadruplex
narrow groove of this second quadruplex. A
complete spine of hydration is also seen in the
narrow groove of the trigonal form. The first
water molecule hydrogen bonds with the O2 of
Thy-A8. The pattern observed is similar as seen
in the orthorhombic form, in which a two-water
molecule bridge is observed between N2 atoms of
Gua-B9 and Gua-A11. However, in the trigonal
form, two water molecules bridge the phosphate
groups of Gua-A/B10, unlike the orthorhombic
form, where only one water molecule bridges the
two phosphate groups, even though the widths
are closely similar. The interactions observed at the
end of the groove are analogous but not identical.
In the trigonal form, three water molecules link
the phosphate groups of Gua-B12 and Gua-A9, the
second of which also hydrogen bonds with the N2
of Gua-A9.
Each quadruplex contains two medium-sized
grooves. In the first of the two (Figure 9(c)– (e)),
two water molecules bridge the N2 of Gua-A/C10
to O50 of GuaB/D1 in two of the three quadru-
plexes, orthorhombic 1 and trigonal. Both trigonal
and orthorhombic 2 quadruplexes have two-water
molecule bridges hydrogen bonding between the
N2 of Gua12 to the phosphate group of Gua3 on
the other strand. This bridging is incomplete in
the analogous orthorhombic 1 quadruplex medium
groove (Figure 9(c)). Other hydration and contact
differences are apparent between these three
grooves, indicating that water molecules have
considerable mobility in this groove.
The second medium grooves show incomplete
spines of water molecules, in both the orthor-
hombic and trigonal forms (Figure 9(f) – (h)), with
characteristic patterns of short runs of linked
water molecules. There is a consistent motif
observed in this groove for all the three quadru-
plexes of a two-water molecule bridge between
strands, often between a phosphate group on one
strand and a guanine N2 on the opposite strand.
Sometimes a two or three-water molecule bridge
links a phosphate group and a furanose O40 atom.
The bridging of guanine N2 atoms on opposite
strands is quite rare, being achieved by one water
molecule in the second medium groove of the
orthorhomic 1 form.
A large number of connected water molecules are
observed in the wide groove of the high-resolution
trigonal form (Figure 9(k)). These are clustered into
two groups. One forms a linear onnected zig-zag
of nine water molecules that make only one contact
with base edges. The other spine, of 11 water
molecules, is less regular, and involves 11 water
molecules, five of which make contact with DNA
atoms, alternating between N2 and phosphate
atoms, ending in a pair of water molecules contact-
ing the O40 atom of Gua-A3. The hydration in the
wide grooves of the two orthorhombic quadru-
plexes is far less complete, undoubtedly due to
the lack of high-resolution diffraction data. The
only interactions analogous to those observed in
the trigonal quadruplex is a single water molecule
195
bridge between the N2 group of Gua2 and the
furanose O40 atom of Gua3 of the same strand in
the orthorhombic-1 form (Figure 9(i)).
A pronounced pattern of hydration is observed
in the loops, especially of the trigonal form quad-
ruplex. Two water molecules bridge the phosphate
groups of Thy-A6 and Thy-A7 and again of resi-
dues Thy-A7 and Thy-A8. Two water molecules
link the O2 atom of Thy-A5 to O2 of Thy-A8, and
coordinate to the N3 atom of Thy-A5, as well as
with the potassium ion at the centre of the loop
(Figure 6). These are probably the key interactions
helping to maintain the integrity of the loop
conformation. Four water molecules bridge the
phosphate group of Thy-A6 with O4 of Thy-A8. In
general, there is a consistent pattern of the struc-
tured networks of hydration in the grooves linking
to the water molecules in the loops. The first water
molecule in the narrow groove spine is hydrogen
bonded to the O2 of Thy-A8 on one end and O2 of
Thy-B6 on the other. This forms a part of the octa-
hedral co-ordination of the potassium ion found
in the loop. The spine in the wide groove begins
at the O4 of Thy-A7 on one end and O4 of Thy-B7
in the other loop. The spines in the two medium
grooves trace their origins to O4 of Thy-B6 and
the phosphate group of Thy-B8, respectively. It
was hard to find the water networks in the loops
for the orthorhombic form, again probably as a
result of the lack of high-resolution diffraction
data. The spine in the wide groove originates from
the O4 of Thy-B7 and the spine in the narrow
groove begins from the O2 of Thy-B8. Other promi-
nent interactions involving water molecules was a
three-water molecule link between O4 of Thy-B/
D5 and O2 of Thy-B/D6 and a four-water molecule
bridge between the phosphate groups of Thy-D7
and Thy-D8.
The other important network between the
two orthorhombic quadruplexes is a four-water
molecule bridge that links the phosphate groups
of Thy-A5 on one quadruplex to the phosphate
group of Thy-C5 on the other. There is also another
two-water molecule link that hydrogen bonds
the phosphate group of residue Gua-B12 of one
quadruplex with the phosphate group of Gua-D12
on the other. This link is a part of the extension
that joins the spine of hydration in the two narrow
grooves in the two quadruplexes.
Discussion
We have shown here that the overall fold and
pattern of strand polarities/glycosidic angles in
the structure of the potassium-containing quad-
ruplex formed by the sequence d(GGGGTTTT-
GGGG), is the same as reported by NMR
studies14,16 for this quadruplex in solution, asso-
ciated with either Naþ or Kþ. It is identical
with those features reported in the sodium-
containing quadruplex complex with the telo-
mere-binding protein from O. nova.20 We conclude
196 Crystal Structure of Oxytricha nova Quadruplex

Figure 9 (legend opposite)

that d(GGGGTTTTGGGG) has a unique fold, that ive and quantitative patterns of groove widths is
is independent of whether sodium or potassium retained across all the NMR and crystal structures,
ions are present, and that is retained in both sol- apart from the initial report13 on this sequence.
ution and crystalline states. Similarly the qualitat- The total of six independent loop conformations
Crystal Structure of Oxytricha nova Quadruplex 197

Figure 9. Diagrams of the hydration in the grooves.

reported here are closely similar in all three of the The location of five firmly bound potassium
quadruplexes and the O. nova protein-bound one. ions, two of which are in the T4 loops, is at first
They resemble, but are not identical with, the loop sight surprising, in view of the conclusion from
conformations found in the NMR analyses.14,16 previous NMR studies16 on the same structure
198
with potassium ions, that modelled only three
sites, within the G-quartets. NMR analysis25 of the
potassium form of the closely related quadruplex
formed from d(GGGCTTTTGGGC) has modelled
a near-linear array of five potassium ions, closely
analogous to the continuous, equi-spaced linear
array of five potassium ions observed here, that
extend into the thymine loops. Those potassium
ions in the loops are playing a key role in stabili-
sing loop conformation, via water molecules that
hydrogen-bond bridge to thymine base edges.
Molecular dynamics simulations26,27 on these quad-
ruplexes have reported that the five-potassium sys-
tem is highly unstable, and that three are expelled
during a simulation, in particular those within the
loops. The present results are clearly at variance
with these findings; we suggest that there may be
specific features in these simulations that may
need be modified, such as the Kþ parameters, in
order to produce physically realistic simulations
for these type of quadruplexes.
This is the first visualisation of an extensive
potassium channel in a G-quadruplex structure.
It is analogous to that recently described in the
structure of a Kþ channel-Fab complex,28 which
also has a linear array of potassium ions in the
channel. The square anti-prismatic coordination of
the potassium ions is common to both types of
structure. We see in both that the potassium ions
that are embedded within the channel are coordi-
nated to amino acid side-chains, analogous to the
coordination to guanine bases here. By contrast,
the ions at the rim of the channel (the loops in the
case of the quadruplexes) are coordinated to water
molecules, albeit with the involvement of thymine
bases in the quadruplexes. The channel structure
provides insight into potassium ion transport; in
view of the equivalence of coordination to a potas-
sium ion in the interior and at the mouth of the
channel it has been proposed that it is energetically
cheap for an ion to move from one position to
another. This analogy between potassium channels
has been suggested,29 and a lipophilic G-quadru-
plex has recently been devised as an artificial ion
channel.30 In the environment of the cell, the
primary function of the quadruplex channel might
be to ensure a rapid influx of potassium ions and
ejection of sodium ions, thus ensuring maximum
quadruplex stability. It is known29 that G-quadru-
plexes and potassium ion channels both show the
same order of Eisenman ion stability:31 Kþ . Rbþ .
Naþ . Csþ . Liþ.
Materials and Methods
Crystallisation and data collection
The DNA sequences d(GGGGTTTTGGGG),
d(GGGGUBrTTTGGGG) and d(GGGG(S)TTTTGGGG)
(with a phosphothioate linkage between the fourth
guanosine base and the first thymidine base) were
purchased from Oswel Research Products Ltd,
Southampton, UK, having been purified by HPLC.
Crystal Structure of Oxytricha nova Quadruplex
Table 2. Crystallisation conditions
Components Orthorhombic Trigonal
MgCl2 (mM) 10.0 10.0
KCl (mM) 40.0 40.0
Spermine (mM) 4.10 3.3
DNA (mM) 1.0 1.0
MPD (%, v/v) 5.0 5.0
MPD (%, v/v) in reservoir 35 25
Oligonucleotide solutions were made up in potassium
cacodylate buffer and then were heated to 80 8C for 15
minutes. They were then allowed to cool gradually to
room temperature. Crystals were grown using the
hanging-drop method at 12 8C as colourless, rod-like
crystals, with the conditions given in Table 2. The orthor-
hombic form was obtained with the first two sequences,
and the trigonal with the brominated one. In spite of
the small difference in conditions, it was possible to
consistently obtain both crystal forms. The crystals grew
over a period of two to four weeks and were flash-frozen
in liquid nitrogen. The majority of orthorhombic crystals
examined showed morphological signs of twinning
and/or high mosaic spread, and data collected on them
could not be indexed readily.
Data for the trigonal form were collected at the ESRF,
Grenoble using an ADSC CCD detector and 0.98 Å wave-
length radiation, whereas the dataset for the orthorhom-
bic form was collected in-house using an RAXIS-IV
detector using 1.542 Å wavelength radiation. Data
reduction and scaling were carried out using the HKL
package.32 Unit cell dimensions and data statistics are
summarised in Table 3. The higher Rmerge value for the tri-
gonal form data set reflects the presence of the bromine
atom, which is commonly found to increase mosaic
spread and hence adversely affect diffraction data quality.
Structure solution and refinement
Numerous attempts were made at structure solution
using either the published orthorhombic crystal struc-
ture (NDB ID number UDL018) or a set of coordinates
from the NMR structure file (PDB ID number
PDB1K4X), using a variety of molecular replacement
programs. No solution refined to an acceptable structure,
as judged by Rfree and R values, as well as by visual
inspection. Heavy-atom and multiple wavelength amor-
phous dispersion (MAD) phasing methods were unsuc-
cessful in the case of the trigonal form, as were attempts
to soak mercury or platinum salts in the instance of the
phosphothiolated orthorhombic form. Molecular replace-
ment methods were successful for both structures using
the program EPMR33 only when the search model was
the quadruplex unit from the PDB submission 1JB7 of
the O. nova protein complex.20
For the trigonal structure, a single outstanding solu-
tion was obtained, with a correlation coefficient of 0.668.
There are two possible space groups that fit the system-
atic absences and had equivalent merging statistics. A
successful translation search was found in space group
P3221, giving an R factor of 40.3%, suggestive of a correct
solution. Examination of 2Fo 2 Fc and Fo 2 Fc difference
Fourier maps showed that the DNA fitted well in the
electron density and that there were no regions of
extraneous density. The XPLOR version 3.134 program
was used to carry out several cycles of rigid body
and positional refinement, followed by the CNS
program35 for simulated annealing and temperature
Crystal Structure of Oxytricha nova Quadruplex 199

Table 3. Crystallographic data

Orthorhombic Trigonal

Space group P212121 P3221

Cell dimensions a, b, c (Å) 26.50, 47.43, 96.46 27.54, 27.54, 145.81
Wavelength (Å) 1.542 0.98
VM (Å3/Da) 1.73 1.97
Resolution range (Å) 8–2.0 (2.07–2.0) 8–1.6 (1.66– 1.60)
Max. resolution (Å) 2.0 1.49
Completeness (%) 98 (98.2) 96.8 (98.4)
Mosaicity 1.0 1.1
Rmerge on I 0.052 (0.19) 0.094 (0.22)
Net I/s(I ) 31.5 (5.84) 18.6 (6.32)
No. unique reflections 8565 8862
Redundancy (over 2I/s) 2.08 1.12
R-factor (%) 21.5 21.8
Rfree (%) 28.2 27.5
rmsd bond distances (Å) 0.005 0.007
rmsd bond angles (deg.) 0.014 0.019
No. DNA strands/asymmetric unit 4 2
No. DNA atoms 1012 506
No. water molecules 213 137
No. K ions 10 5
Average B-factor (Å2)
Quartet 23.2 18.2
Loops 27.4 23.9
Water 43.5 40.6
K ions 22.8 16.0
PDB ID 1JRN 1JPQ
NDB ID UD0014 UD0013

factor refinement in the resolution range 8 – 1.6 Å. This
resulted in a large reduction in vales for the R-factor, to
28.2%, and for Rfree, to 29.8%.
Five strong peaks of electron density were apparent
between the bases. These were assigned as potassium
ions on the basis of their behaviour during subsequent
refinement. Alternative assignments as sodium ions
resulted in acceptably low B-factors for them, and a
significant increase in R and Rfree. The automated
solvent-searching program SHELXWAT† was then
employed, in conjunction with manual peak-picking to
locate solvent molecules. These were accepted on the
basis of standard distance and B-factor criteria. Refine-
ment was continued with the programme SHELX97.†
The final R value was 21.8% (Rfree 27.5%).
Similar procedures were employed to obtain a structure
solution in the P212121 space group. We assumed that there
are two independent quadruplex molecules in the crystal-
lographic asymmetric unit, on the basis of the cell volume.
The starting model was the same as before, again using the
program EPMR.33 A single high value for the correlation
coefficient (0.765) was obtained, but now with two quadru-
plexes. The program determined the relationship between
them. Initial translation searches resulted in an R value of
41.8%. The XPLOR and CNS programs were used as
before for further refinement in the 8–2.0 Å range, yield-
ing an R value of 31.3%. Automated solvent searching
was employed as before, and potassium ions were
located by visual inspection. The structure was further
refined, giving a final R value of 21.5% (Rfree of 28.2%).
† Sheldrick, G. M. (1997). SHELX-97, A Crystallographic
Refinement Program. The University of Göttingen,
Göttingen, Germany.
Data Bank accession numbers
Coordinates and structure factors for both structures
have been deposited in the RCSB Protein and Nucleic
Acid Data Banks with accession numbers IJPQ, IJRN,
UD0013 and UD0014.
Acknowledgements
The Cancer Research UK (programme grant SP1384
to S.N.), and the Institute of Cancer Research (research
studentship to S.H.) are thanked for support of these
studies. We are grateful to Martin Read of these labora-
tories, as well as Julie Feigon and Alex Rich for useful
discussions.
References
1. Blackburn, E. H. (1991). Structure and function of
telomeres. Nature, 350, 569 – 573.
2. McEachern, M. J., Krauskopf, A. & Blackburn, E. H.
(2000). Telomeres and their control. Annu. Rev.
Genet. 34, 331 – 358.
3. Blackburn, E. H. (2001). Switching and signalling at
the telomere. Cell, 106, 661 –673.
4. Cech, T. R. (2000). Life at the end of chromosomes:
telomeres and telomerase. Angew Chem., Int. Ed. 39,
34 – 43.
5. Simonsson, T., Pecinka, P. & Kubista, M. (1998). DNA
tetraplex formation in the control region of c-myc.
Nucl. Acids Res. 26, 1167 – 1172.
6. Griffith, J. D., Corneau, L., Rosenfeld, S., Stansel,
R. M., Bianchi, A., Moss, H. & de Lange, T. (1999).
Mammalian telomeres end in a large duplex loop.
Cell, 97, 503 – 514.
200
7. Simonsson, T. (2001). G-quadruplex DNA structures—
variations on a theme. Biol. Chem. 382, 621– 628.
8. Williamson, J. R. (1994). G-quartet structures in telo-
meric DNA. Annu. Rev. Biophys. Biomol. Struct. 23,
703 – 730.
9. Patel, D. J., Bouaziz, S., Kattani, A. & Wang, I. (1999).
Structures of guanine-rich and cytosine-rich quadru-
plexes formed in vitro by telomeric, centromeric and
triplet repeat disease DNA sequences. In Oxford
Handbook of Nucleic Acid Structure (Neidle, S., ed.),
pp. 389 – 453, Oxford University Press, Oxford.
10. Laughlan, G., Murchie, A. I., Norman, D. G., Moore,
M. H., Moody, P. C., Lilley, D. M. J. & Luisi, B. (1994).
The high-resolution crystal structure of a parallel-
stranded guanine tetraplex. Science, 265, 520– 524.
11. Phillips, K., Dauter, Z., Murchie, A. I., Lilley, D. M. J.
& Luisi, B. (1997). The crystal structure of a parallel-
stranded guanine tetraplex at 0.95 Å resolution.
J. Mol. Biol. 273, 171– 182.
12. Aboul-ela, F., Murchie, A. I. & Lilley, D. M. J. (1992).
NMR study of parallel-stranded tetraplex formation
by the hexadeoxynucleotide d(TG4T). Nature, 360,
280 – 282.
13. Kang, C., Zhang, X., Ratliff, R., Moyzis, R. & Rich, A.
(1992). Crystal structure of four-stranded Oxytricha
telomeric DNA. Nature, 356, 126– 131.
14. Smith, F. W. & Feigon, J. (1992). Quadruplex struc-
ture of Oxytricha telomeric DNA oligonucleotides.
Nature, 356, 164– 168.
15. Schultze, P., Smith, F. W. & Feigon, J. (1994). Refined
solution structure of the dimeric quadruplex formed
from the Oxytricha telomeric oligonucleotide
d(GGGGTTTTGGGG). Structure, 2, 221– 233.
16. Schultze, P., Hud, N. V., Smith, F. W. & Feigon, J. (1999).
The effect of sodium, potassium and ammonium ions
on the conformation of the dimeric quadruplex formed
by the Oxytricha nova telomere repeat oligonucleotide
d(G4T4G4). Nucl. Acids Res. 27, 3018– 3028.
17. Strahan, G. D., Shafer, R. H. & Keniry, M. A. (1994).
Structural properties of the [d(G3T4G3)]2 quadruplex:
evidence for sequential syn – syn deoxyguanosines.
Nucl. Acids Res. 22, 5447– 5455.
18. Smith, F. W., Lau, F. W. & Feigon, J. (1994). d(G3T4G3)
forms an asymmetric diagonally looped dimeric
quadruplex with guanosine 50 -syn– syn – anti and
50 -syn– anti – anti N-glycosidic conformations. Proc.
Natl Acad. Sci. USA, 91, 10546– 10550.
19. Keniry, M. A., Strahan, G. D., Owen, E. A. & Shafer,
R. H. (1995). Solution structure of the Naþ form of
the dimeric guanine quadruplex [d(G3T4G3)]2. Eur.
J. Biochem. 233, 631– 643.
20. Horvath, M. P. & Schultz, S. C. (2001). DNA
G-quartets in a 1.86 Å resolution structure of an
Oxytricha nova telomeric protein – DNA complex.
J. Mol. Biol. 310, 367– 377.
21. Neidle, S. & Read, M. A. (2001). G-quadruplexes as
therapeutic targets. Biopolymers, 56, 195– 208.
22. Read, M. A., Wood, A. A., Harrison, J. R., Gowan,
S. M., Kelland, L. R., Dosanjh, H. S. & Neidle, S.
(Received 23 January 2002; received in revised form 25 April 2002; accepted 2 May 2002)
Crystal Structure of Oxytricha nova Quadruplex
(1999). Molecular modelling studies on G-quadru-
plex complexes of telomerase inhibitors: structure–
activity relationships. J. Med. Chem. 42, 4538– 4546.
23. Neidle, S., Harrison, R. J., Reszka, A. P. & Read, M. A.
(2000). Structure– activity relationships among
guanine-quadruplex telomerase inhibitors. Pharma-
col. Therapeut. 85, 133 – 139.
24. Read, M. A., Harrison, R. J., Romagnoli, B., Tanious,
F. A., Gowan, S. H., Reszka, A. P. et al. (2001). Struc-
ture-based design of selective and potent G quadru-
plex-mediated telomerase inhibitors. Proc. Natl Acad.
Sci. USA, 98, 4844– 4849.
25. Bouaziz, S., Kettani, A. & Patel, D. J. (1998). A Kþ
cation-induced conformational switch within a loop
spanning segment of a DNA quadruplex containing
G – G –G –C repeats. J. Mol. Biol. 282, 637 – 652.
26. Špaéková, N., Berger, I. & Šponer, J. (1999). Nano-
second molecular dynamics simulations of parallel
and antiparallel guanine quadruplex DNA molecules.
J. Am. Chem. Soc. 121, 5519– 5534.
27. Špaéková, N., Berger, I. & Šponer, J. (2001). Structural
dynamics and cation interactions of quadruplex
DNA molecules containing mixed guanine/cytosine
quartets revealed by large-scale MD simulations.
J. Am. Chem. Soc. 123, 3294– 3307.
28. Zhou, Y., Morais-Cabral, J. H., Kaufman, A. &
MacKinnon, R. (2001). Chemistry of ion coordination
and hydration revealed by a Kþ channel-Fab com-
plex at 2.0 Å resolution. Nature, 414, 43 – 48.
29. Hud, N. V., Smith, F. W., Anet, F. A. L. & Feigon, J.
(1996). The selectivity for Kþ versus Naþ in DNA
quadruplexes is dominated by relative free energies
of hydration: a thermodynamic analysis by 1H
NMR. Biochemistry, 35, 15383– 15390.
30. Forman, S. L., Fettinger, J. C., Pieraccini, S., Gottarelli,
G. & Davis, J. T. (2000). Toward artificial ion
channels: a lipophilic G-quadruplex. J. Am. Chem.
Soc. 122, 4060– 4067.
31. Eisenman, G. & Horn, R. (1983). Ionic selectivity
revisited: the role of kinetic and equilibrium pro-
cesses in ion permeation through channels. J. Membr.
Biol. 76, 197 – 225.
32. Otwinowski, Z. (1993). Oscillation data reduction
program. In Data Collection and Processing (Sawyer,
L., Isaacs, N. W. & Bailey, S., eds), pp. 55 – 62, SERC
Daresbury Laboratory, Warrington, UK.
33. Kissinger, C. R., Gehlhaar, K. & Fogel, D. B. (1999).
Rapid automated molecular replacement by evolu-
tionary search. Acta Crystallog. sect. D, 55, 484 – 491.
34. Brünger, A. T., Kuriyan, J. & Karplus, M. (1987).
Crystallographic R-factor refinement by molecular
dynamics. Science, 235, 458 – 460.
35. Brünger, A. T., Adams, P. D., Clore, G. M., DeLano,
W. L., Gros, P., Grosse-Kunstleve, R. W. et al. (1998).
Crystallography and NMR system: a new software
suite for macromolecular structure determination.
Acta Crystallog. sect. D, 54, 905 – 921.
Edited by A. Klug