Download A comparison of the structure of echinomycin and triostin A

Volume 13 Number 7 1985
Nucleic Acids Research
A comparison of the structure of echinomycin and triostin A complexed to a DNA fragment
Giovanni Ughetto*, Andrew H.-J.Wang*§, Gary J.Quigley*, Gijs A.van der Marel"1", Jacques H.van
Boom4" and Alexander Rich*§
*Department of Biology, Massachusetts Institute of Technology, Cambridge, MA 02139, USA, and
+
Gorlaeus Laboratories, Leiden State University, Leiden, The Netherlands
Received 2 January 1985; Accepted 19 February 1985
ABSTRACT
Two members of the quinoxaline
antibiotic
family, echinomycin and
triostin A, form crystals complexed to a DNA fragment with the sequence
a(CpGpTpApCpG).
The crystal structure of both complexes was solved by X-ray
diffraction to near-atomic resolution. The two structures are similar to each
other with differences in some details due to the shorter cross bridge of
echinomycin.
Both molecules act as bis intercalators surrounding the d(CpG)
sequence at either end of the double helix.
Alanine forms sequence—specific
hydrogen bonds to guanines in the minor groove. The two central AT base pairs
are held
together
by Uoogsteen base pairing with
adenine
in the svn
conformation in both complexes.
An octahedrally hydrated magnesium ion is
found in the crystal lattice that plays an important role in organizing the
lattice as well as stabilizing the complex by hydrogen bonding both to base
pairs of UNA and to the quinoxaline ring nitrogen atoms in the major groove
side of the UNA double helix.
A functional description of the various amino
acids
in
quinoxaline
antibiotics
is
given,
together
with
possible
modifications that might affect biological activity.
INTKODUCTION
An important class of antibiotics are those that bind to UNA and thereby
modify
its biological
activities.
Streptomvces contain eight
The quinoxaline
antibiotics
amino acids in a cyclic depsipeptide
quinoxaline rings attached to them [1, 2 ] ,
These antibiotics
derived
from
and have two
are very active
against gram positive bacteria and also have cytotoxic effects on tumor cells.
Two members
of this
each
by
other
molecules
class are echinomycin and triostin A, which differ
the nature
of
the sulfur-containing
as shown in Figure 1.
They act by binding
replication as well as transcription [3, 4 ] .
action.
to
with
DNA is fundamental
the deoxyhexanucleoside
in
from
these
to DNA and inhibit DNA
[5].
to an understanding
Previously, we have described
bridge
Currently, echinomycin
used in clinical trials in human cancer chemotherapy
interaction
cross
is being
The nature of their
of
their
mode
of
the structure of triostin A complexed
pentaphosphate
d(CpGpTpApCpG)
[6],
We have now
solved the structure of echinomycin bound to the same oligonucleotide and here
we compare the two molecular
structures.
© IRL Press Limited, Oxford, England.
Both molecules appear to act as bis
2305
Nucleic Acids Research
intercalators
surrounding
intercalated
inside
the
sequence
the double
helical
over the terminal CG base pair.
d(CpG)
tragment
helix.
In
addition,
central AX base pairs
in
echinomycin.
interactions
the
unusual
together
Although
results
in
as Hoogsteen
because ot
ot
the
similar
shorter cross
the
many
two
base pairs [7]
the two structures are
to the
ring
stacks
ot the DNA
rearrangement
modifying
a major component
van
bridge
der
Waals
of the stabilization between the
The UNA fragment is considerably unwound and there are several
conformational modifications
to
quinoxaline
other ring
and the minor groove
in several details due
This
that provide
drugs and UNA.
is an
base pairs.
to each other, they differ
found
there
that are now held
instead of Watson-Crick
one
the
This binding is sequence-specific
hydrogen bonding between the peptide backbone
double
with
while
interaction
with
associated with alterations of base
the
antibiotic.
These
are
also
stacking due
found
to
differ
somewhat in the two structures due to the altered geometry ot the depsipeptide
ring.
Although
differences
these
both
activities [8] .
in
two
quinoxaline
their binding
antibiotics
constant
to
UNA
are
similar,
and
in
there
their
are
biological
Thus, a detailed comparison ot the two types of complexes may
be useful as it may be related to these differences.
MATERIALS ANU METHODS
The triostin A was a gift from Ur. T. Yoshida
The echinomycin
Cambridge.
The
earlier 19].
technique
hexanucleoside
described
pentaphosphate
before [6].
A
2 mM UNA hexamer, 20 mM sodium
triostin
A
or
echinomycin,
and
was
as
described
unit
cell
lattices with
dimensions
triostin A
asymmetric
were
&
the
synthesized
using the vapor phase
diffusion
typical
crystallization
cacodylate
space
= 30.71,
F222.
62.36,
c
For
unit
in both
The
crystals contained one
echinomycin,
= 61.70
the dimensions were a = 31.35, b = 62.38, and
antibiotic molecule.
(MPU).
The crystals of both complexes
group
b -
mixture
(pH 7 ) , 10 mM MgCl 2 . 1.5
51o 2-methyl-2,4-pentanediol
mixture was then equilibrated against 35% MPU.
had orthorhombic
(Shionogi, Osaka, Japan).
from Ur. M. J. Waring, University of
The complexes were crystallized
as
contained
mM
(Ciba-Geigy) was a gift
8,
while
the
tor
c = 61.26 ft. The
strand of UNA as well
The diffraction patterns produced by the two
as one
complexes
were very similar to each other.
Three-dimensional
diffraction
data
diffractometer in the omega scan mode.
10°C and
ad).
2,015
collected
on
a
Nicolet
P3
reflections were measured
with
an
intensity greater
than 1.0
The triostin A data collected in the same manner but at -16°C had 2,995
observed reflections.
2306
were
The echinomycin data were collected at
The structure of the triostin A complex was solved as
Nucleic Acids Research
NM,Vol
iN-MeCy
I-Ala
Ou.
Triostin A
Future 1^: The structure of triostin A and echinomycin axe indicated.
Triostin A has a OisulfiQe linkage between the cystine residues, whereas a
thioacetal linkage is found in echinomycin.
described
earlier
5-bromocytosine
by
the
method
ot
derivatives [6] .
patterns, trial coordinates
multiple
Due
to
isomorphous
the
replacement
similarity
tor the echinomycin
of
the
using
diffraction
structure were obtained
from
the triostin A structure
and then they were refined at 1.8 8 resolution using
the
constrained
were
Konnert-Henarickson
revealed
in the
asymmetric
refinement
unit
of the
the 137 found in the triostin A crystal.
residual
factor
was
19.9%
compared
to 18.6% for
tor
the
triostin
A
[10].
solvent
molecules
echinomycin crystal compared
to
After many cycles of refinement, the
echinomycin
at
88
1.67
%
complex
at 1.8
resolution.
In
8
resolution
the
refinment
procedure, a number of different constraints were imposed on the molecule and
the results were
studied.
In all of
placed on base pairing geometries.
distance
constraints
structure revealed
and
placed
echinomycin structure.
the
peptide
Accordingly,
and
the
similar
A similar
hydrogen
weak constraints
were
the
drug
and
the
nucleic
acid.
bonding interactions between the
That
antibiotic
type of refinement was carried out
for
the
This yielded two hydrogen bonding interactions between
nucleic
constraints
hydrogen bonds were
peptides.
between
three hydrogen
the nucleic acid.
the refinements,
In the triostin A structure there were no
imposed
bonding
acid
involving
but
the
to be close
interactions
one
of
Ml
of
them
was
alamne
implausibly
and
N3
of
to 2.8 8 and this resulted
between
Both crystals were nearly 50% solvent
the
nucleic
acid
and there were large
short.
guanine
in
and
three
the
solvent
channels found weaving along the £ axis.
The coordinates of both complexes are being published elsewhere.
2307
Nucleic Acids Research
Figure 2: Van Her Waals diagrams showing the interaction of triostin A
and echinomycin with the DNA duplex.
The antibiotics are shaded with dark
stippling.
At the top of the diagram the quinoxaline rings can be seen
protruding into the major groove of the helix.
In the lower part of the
diagram the cyclic peptide occupies virtually all of the minor groove.
Carbon
atoms are indicated by concentric circles, phosphorous atoms by concentric
circles with cross bars, nitrogen atoms by stippled circles.
Hydrogen atoms
are not shown.
RESULTS ANLI DISCUSSION
The
structure
of the echinomycin
complex
to the DNA oligomer is broadly
similar to that already reported for triostin A (Figure 2 ) .
like
triostin
A,
binds
to the minor
groove
quinoxaline rings stacking on either side
of
the
groove.
helix.
The
although the
The
intercalated
overall
different
similarity
of the
of the
rings
of
structures
contigurations
two
of the
the
two
two CG base pairs at the end
quinoxaline
the
Here echinomycin,
DNA duplex with
protrude
into
the
is readily
sultur-containing cross
major
apparent
bridges
can also oe seen in Figure 2.
A more
Both
2308
direct
antibiotics
comparison
have
a
of the
Oshaped
two antibiotics
form
in which
the
is
shown
cyclic
in Figure
3.
depsipeptide
is
Nucleic Acids Research
Ala
Ser
Figure 3:
A superposition ot both the echinomycin (E, solid bars) and
the triostin A molecules
(X, open bars) is shown.
In this view
the
quinoxal m e rings at the upper ana lower part of the diagram are pointing
toward the reader ana we are looking toward the side of the cyclic peptide
that is facing the nucleic acid.
The two pairs of quinoxaline rings are
virtually in the same position.
There are only small deviations in the
position of the various amino acids around the cyclic depsipeptide.
The
greatest deviation is found near the side chains of the cystine residues in
which the disulfiae linkage of triostin A pushes the two vertical peptide
units further apart than does the thioacetal cross bridge of echinomycin.
organized
each
in
other
toward
a rectangular
and
protruding
the face ot
form
on
with
the
the
same
the depsipeptide
two
quinoxaline
side.
In Figure
that
is exposed
quinoxaline rings coming toward the reader.
area of difference
is near the center
rings parallel
3, we
right
in
side
this
of
superposition
the
ring.
The
is
to the DNA with
of the molecule where
founa
distance
at
the
Ca
square
(rms) deviation from
the
the shorter
between the
atom
two
cross
The greatest
of cystine
two C a
across the Dridge of echinomycin is 0.22 8 less than that
root mean
to
looking
It can be seen that the principal
bridge in echinomycin brings the peptide chains closer together.
deviation
are
atoms of
on
the
cystine
in triostin A. The
the position of all of the atoms from
each other, excluding the side chains of the two cystines in both antibiotics,
is 0.39 8.
The
side
approximately
those
chain
into
projecting
directly
toward
orientation
two groups,
away.
the
The
nucleic
of
the
amino
those projecting
methyl
acid,
side
while
chains
the
acids
toward
ot
cystine
the
can
be
nucleic
alanine
are
side
chains
divided
acid
and
projecting
are
both
2309
Nucleic Acids Research
Figure 4.: A schematic diagram illustrating the
manner in which the quinoxaline antibiotics are
inserted into the DNA duplex.
The numbering system
of the nucleotides in the duplex is shown together with
the designation given to the different crosshatched
quinoxaline rings.
There is a two-fold axis in the
middle of the molecule, so that half of the complex
is the asymmetric unit.
projecting
while
the
away.
The
serine
side
valine
residues
chains are
have
a
somewhat
used directly
tangential
position,
the amide
bond
to form
that
links the quinoxaline ring to the lactone ester linkage.
ORGANIZATION OF THE COMPLEX
The
which
general
also
form
shows
of
the
the
complex
numbering
cross-hatched quinoxaline
is
shown diagramatically
system
rings surround
used
in
the
identical
the center
on the outer edge of the CG base pairs.
molecule lies on a crystallographic
is
polynucleotide.
to
the
lower halt.
and
4,
The
the two GC base pairs at either
with intercalation between AT and CG base pairs near
of quinoxaline
in Figure
end
stacking
In the crystal, the
two-fold rotation axis, so the upper half
A detailed
comparison
of the
interactions
between the antibiotic and the nucleic acid fragment is shown in Figure 5.
There
complex:
are three major interactions
hydrogen
bonding,
that
intercalation
are
and
involved
van
der
in stabilizing
Waals
the
interactions.
There is hydrogen bonding between both alanine residues and the guanine of the
two
CG
base
pairs
in
stippling in Figure 5.
the
minor
have a two-fold symmetry
tetrapeptide.
This deviation
the
axis
due
sulfur
to
in the hydrogen
side
the backbone
a two-told
of the
bonding
of the diagram
Nil of
are
alanine
axis
In echinomycin,
the asymmetry
on the
G12,
from
atoms.
also evident
right
These
shown
enhanced
with
from Figure 2, the antibiotic does not
axis even though it is built out of two repeats of a
positioning of
two—fold
groove.
As is evident
is most
cross bridge.
interactions.
forms
apparent
there could not
two hydrogen
is hydrogen bonded
This
asymmetry
In Figure 5, the
bonds with
to the
in
the
be a true
the
is
alanine
guanine
guanine N3
a ton,
while the carbonyl oxygen of alanine receives a hydrogen bond from the guanine
N2 atom.
2310
The NH-N hydrogen bond is 2.68 8 for echinomycin compared to 2.91 8
Nucleic Acids Research
A)
van der Waals
contacts
Echinomycin
Figure 5.: Skeletal diagrams showing the interaction of (A) triostin A or
(B) echinomycin with one half of the DNA duplex.
The quinoxaline antibiotic
is drawn with solid black lines while the DNA is drawn with open lines.
Oxygen atoms are shaded black, nitrogen atoms are stippled and sulfur atoms
are cross hatched.
The numbers represent the distance between atoms in
Angstroms illustrating van der Waals contacts of 3.5 8 or less between the
antibiotic and the nucleic acid. Nineteen such contacts are found in triostin
A compared to 21 in the echinomycin complex.
Hydrogen bonds between the
alanine residues and the guanine bases are emphasized by stippled shading
surrounding the hydrogen bonds.
2311
Nucleic Acids Research
tor the triostin A complex. The guanine N2 to the alanine carbonyl oxygen is a
long hydrogen
bo no in
both cases with 3.38
A tor
echinomycin
and
3.07 8 in
triostin A. The other alanine in Figure 5 has only one hydrogen bond trom
NH
of
alanine
to
the
N3
of
guanine
G2 which
compared to 3.01 8 in the triostin A complex.
is 2.99
8
in the
the
echinomycin
In both structures, there is no
hydrogen bond between the N2 amino group of guanine G2 and the alanine oxygen.
In the triostin A complex, the distance
between
the N2 of guanine G2 to
the
alanine carbonyl oxygen is 4.1 8 while it is 3.6 8 in the echinomycin complex.
Inspection of the two complexes in Figure S reveals that the nucleic acid
has
undergone
a
considerable
Alanine plays a central
chain points toward
and G2
on
results
in
one
of
its
role in this reorganization,
internal
side
of
the
ring
tilt
and
Cll
and G12
between
the
structure.
since the C
the polynucleotide wedging between
a considerable
cytosine Cll.
reorganization
on
plane
methyl side
the sugar residues Cl
the
of
other
guanine
side.
base
This
G12
and
In the echinomycin complex, these two bases have 11.3° dihedral
angle between them in contrast to the 17° found in the triostin A complex.
the other
side of the double helix, there
G2 and cytosine
Cl
is a tilt of 16.5° between
in the echinomycin complex, compared
the triostin A complex.
The differences between
On
guanine
to the 26° found
in
these tilts are due to their
high sensitivity to the exact position of the alanine methyl side chains.
Because
the
same
of this tilt, Cl is hydrogen bonded
plane.
These
are
tilted
14°
from
to G12 but
each
other
is not
in the
lying
on
echinomycin
complex compared to 20° in the triostin A. Likewise, G2 is hydrogen bonded to
Cll but
is tilted 14° in the echinomycin complex compared
triostin A complex.
found for
Despite
the hydrogen bonding interactions between
two quinoxaline
to 23° seen in the
this tilting of the bases, normal distances
rings, 013 and 014, are
the bases.
virtually parallel
However,
with each
are
the
other,
with only a 7° deviation in echinomycin and a 6° deviation in the triostin A.
Stacking of bases
is generally considered a stabilizing interaction with
interplanar spacings between unsaturated ring systems near 3.4 8.
However, in
these antibiotic—DNA complexes several interplanar distances have increased so
that
there
is a decrease
in
stabilization.
The
partly over cytosine Cll but the distance between
echinomycin
under
planes
the
is
complex.
2312
ana
3.97
pyrimidine
3.46
8
in
8
for
ring
the
the
of
triostin
adenine
echinomycin
A
quinoxaline
complex.
A10,
complex
but
and
ring 014
lies
the rings is 3.68 8 for the
the
Similarly,
distance
3.53
8
in
014
between
the
stacks
their
triostin
A
Nucleic Acids Research
VAN HER WAALS INTERACTIONS
A major
stabilizing interaction between the antibiotic ana DNA is due to
the large number of van der Waals interactions which are shown in Figure S for
both
complexes.
distances
These
der Waals
interactions.
echinomycin
acid
the
echinomycin
that are 3.S 8 or less.
van
subtle
In
and
triostin
in the
to
the
A
changes
in
the
alanine
side
chains
complex
chain has moved over
Cll.
serine
so that
changes
echinomycin relative
antibiotic ana
triostin
group
of
the
van
antibiotics
oxygen
01' of
sugar
G2.
binding.
The
total
represent
the
nucleic
with
the
clearly in association
In
in contact with the furanose
altered
ot
seen as the alanine
the
interactions
an oxygen
contact
atom
is
of
in
side
ring of
the sugar residue G12 have
positioning
of
the
with
found
the
valine
adenine
interactions
between
between
nucleic
and
All)
to those
seen
the
in
the
for
methyl
acid.
In
the
is 2.86 8 between an alanine methyl group and
In echinomycin,
the
closest
contact
and oxygen 02 of cytosine
shorter than normal van der Waals distances and they reflect
the
are 19.
shown for
which
to
Waals
stacking
associated
in echinomycin compared
between the N-methyl group of valine
addition of different
there
side of Figure S.
contacts are
der Waals
and
der
to ring
relative
to triostin A. There are fewer
closest
van
three van der Waals contacts; however,
number
and T3
ring
backbone
the right
it is now
the
21
of differences
This can be seen most
triostin A the closest distance
an
the
antibiotic
antibiotic
Likewise,
residues G2
A. The
of
between triostin A with
in echinomycin.
residues
the
two additional
Some of the contacts
lengthened
a number
shown on
triostin A complex, there are only
the echinomycin
are
of the van der Waals contacts
of
the
different sulfur cross bridges.
with
exclusive
reveals
positioning
there
In the triostin A complex,
are
A detailed comparison
shifts
due
distances
complex,
energy
of
the
binding
is
thus
is 2.90
Cl.
8
Both are
the tightness of
due
to
a
complex
stabilizing and destabilizing interactions.
STACKING OF THE BASES
The
stacking of the bases ana the quinoxaline
the approximate helix axis.
A complex;
rings is best viewed
This is illustrated in Figure 6 for the
the results for echinomycin are similar.
in which the base pairs G2-C11
down
triostin
Figure 6A shows the way
stack upon C1-G12 with considerable
unwinding
as the twist angles between the base pairs is only 10° compared to the normal
36° that
is found
in B-DNA.
In Figure 6B, the quinoxaline
rings Q13
and 014
(see Figure 4) have been added to Figure 6A so that their position relative to
the
base
pairs
can
be
seen.
The
quinoxaline
rings
stack
upon
both
bases
2313
Nucleic Acids Research
Future 6_: Stacking interactions in the triostin A-UNA complex.
All
views project along the helix axis of the complex.
(A) Base pair G2-C11 over
C1-G12.
There is an unwinding angle of 26° between base pairs.
(B) The
quinoxaline rings 013 and 014 are drawn on either side of the base pairs shown
in ( A ) . The solid quinoxaline ring is closer to the reader while the dashed
one is further away.
It can be seen that the quinoxaline rings span both
bases in the base pair with some overlap.
(C) The stacking interactions on
either side of quinoxaline ring 014 are shown. The base pair T3-A10 is closer
to the reader while the base pair C11-G2 is further away, behind
the
quinoxaline ring.
It can be seen that the quinoxaline ring stacks very well
with adenine A10 in the svn conformation and its attached carbonyl group sits
over T3.
If adenine A10 were in the anti conformation, there would be
virtually no stacking. (D) The intermolecular stacking of quinoxaline rings
between the different complexes. This stacking stabilizes the organization of
these complexes into long rods that pass through the lattice.
2314
Nucleic Acids Research
although there
is somewhat
better stacking upon the pyrimidine rings than the
purine ring of the base pair.
Figure 6C
shows the way in which the CG ana AT
base pairs surround the quinoxaline CU.4.
As was the case
in the triostin A complex, the central two AT base pairs
in the echinomycin complex are
pairs.
Instead,
the
thymine
residues
form
described
adenine
two
no longer
residues
hydrogen
by Hoogsteen [7] .
held together
are in
bonds
as Watson-Crick
the svn conformation,
with
the
adenine
in
These Uoogsteen base pairs involve
base
and
a
the
manner
the N6
amino
group of adenine hydrogen bonding to thymine 04, while the thymine N3
donates
a
in
hydrogen
in
echinomycin
distance
bonding
complex
Cl'
to
are
to Cl'
adenine
similar
from
the
N7.
to
The
those
sugar
hydrogen
in
atoms
the
bond
triostin
across
the
lengths
A
AT base
pair
shorter tor Hoogsteen base pairs than for Watson-Crick base pairs.
the effect
of bringing the polynucleotide
The
is 2
S
This has
closer to the quinoxaline
makes it possible for there to be a large number
the
complex.
rings and
of stabilizing van der Waals
contacts as shown in Figure S.
As
shown
in Figure
6C,
the
quinoxaline
ring
stacks
directly
upon
the
pyrimidine ring of adenine Alt) but does not stack at all with pyrimidine
T3.
The base pairs G2'C11
ring
ana A10'T3 have a twist angle of 26° and therefore
an unwinding angle of 10°.
Studies of DNA
angle of 45
in
solution with
echinomycin have
to 55° per echinomycin molecule
in the echinomycin complex to make
is an unwinding
[8].
an estimate
angle of 10° between the
and tinally
an unwinding
be another
comparable
angle of 26° between
of this in the complex.
may assume
that
the
the
1U°
unwinding
angle at
unwinding angle
pairs between the quinoxaline
angle
the
quinoxaline
is
residue
014,
there would
also
the other end of the molecule.
of 26°
associated with
residues will be found
that
There
the two CG base pairs bracketed
In a continuous DNA molecule,
unwinding
unwinding
information
two AT base pairs, also 10° between
the AT base pairs ana the CG base pairs around
by the two quinoxaline rings.
revealed an
We can use the
found
around
the two
in solution.
014
is
We
CG
base
Likewise,
likely
to
be
representative, and for two rings we come to a total of 46° of unwinding.
We
have ignored the unwinding found between the two Hoogsteen base pairs found in
the center of the complex.
However, as we do not know how many Hoogsteen base
pairs will fora in the vicinity of the intercalative
estimate their contribution to unwinding.
total unwinding would be close to 6 6 ° .
the
average
only
one
extra
site, it is difficult
to
If there were two on each side, the
However, if we assume that there is on
Hoogsteen
base
pair
on
either
side
of
the
2315
Nucleic Acids Research
2316
Nucleic Acids Research
quinoialine
close
to
ring,
then
the observed
the
estimate
value.
of
the unwinding
The unwinding
data of
angle
of 56° would
the complex
be
is thus in
general accord with the unwinding measurements observed in solution.
FACTORS STABILIZING THE LAITICE
There are two fundamental
of
both
the
stacking
6U).
echinomycin
interaction
Secondly,
forming
successive
between
these
sheets.
ana
the
linear
Successive
layers.
interactions used in building up the
triostin A
complexes.
two-fold
rod-like
sheets
related
elements
are
There
Q13
is an
quinoxalines
pack parallel
tilted
lattices
end-to-end
relative
(Figure
to each
to
each
Figure 7A shows a view down the c axis with the
A-DNA complexes organized along the diagonals of the unit cell.
other
other
in
triostin
Each isolated
figure represents the superposition of two complexes packed at an angle of 54°
to each other ana stacked on the quinoxaline residues Q13.
The interactions between
which
shows a view down
the diagonal
the a axis.
elements can be seen in Figure 7B,
The rods alternate
in their
with one rod tipping toward the reader at the top of the diagram
rod tipping away
from
the reader at the top of the, diagram.
In Figure 7B it
should be noted that there are two kinds of interaction between
rods.
One
set
of interactions
antibiotic-DNA residues
has a hydrated magnesium
(stippled regions) while
close lntermolecular sulfur—to-sulfur contact
A more
magnesium
detailed
ions
view
contribute
diagram of Figure 8.
of
to
the manner
the
lattice
orientation
and the next
the sheets of
complex between the
the other
interaction has a
(4.0 8 ) .
in which
the octahedrally
interaction
is
seen
in
hydrated
the
stereo
Two opposite faces of the octahedron form hydrogen bonds
with the base pair G2-C11 and quinoxaline QJ.4 from complexes that are oriented
at an angle to each other
in the diagonal array.
The six water molecules are
bound with hydrogen bond lengths of 2.80, 2.83, 2.VI 8 from water molecules to
quinoxaline N3, 06 of G2 and N7 of G2 respectively.
while
the magnesium
complex
plays
a key
role
It should
in organizing
be noted
the
that
lattice
by
Figure 2:
The lattice packing of the triostin A-DNA complex.
(A) The
view down the .c axis is shown ana the lattice is outlined.
Two different
levels of complexes are shown, which are organized as rod-like arrays that run
across the diagonals of the figure.
The t> axis is vertical and ji axis is
horizontal.
The rods along the diagonals are stabilized by the stacking
interactions between the quinoxaline rings at the outer edges of the complex.
(B) The outlined lattice is shown viewed down the a axis, b is vertical, c is
horizontal.
The long rod-like arrays of complexes are tipped so that
alternate ones are successively pointing toward and away from the reader.
There are two different types of interactions between the rods, one of which
is stabilized by
the octahedrally
coordinated magnesium
ions (some are
stippled) that are located between every other pair of rods.
2317
Nucleic Acids Research
Figure 8: A stereo diagram illustrating the manner in which the hydrated
magnesium
ion with octahedral coordination
torms hydrogen bonds to two
different antibiotic-DNA complexes.
In the unrefined structure, the Mg-ILO
complex was very close to a regular octahedron.
During the refinement, the
Mg-0 distances in the complex were all constrained to remain at 2.0 A in a
regular octahedral geometry. The magnesium ion lies on the b axis which is a
two-fold rotation axis.
Opposite octahedral faces which are related by the
two—fold axis form three hydrogen bonds to each DNA quinoxaline complex. The
water molecules are hydrogen bonded to both the intercalating quinoxaline
rings as well as the CG base pairs in the center of the complex.
bridging
7B,
it
the
successive
can also
serve
layers
to
of the DNA-drug
stabilize
the
complexes
complex
in
as shown in Figure
solution using
hydrogen bonding interactions from one face of the octahedron.
the
three
It is possible
that the role of the metal ion complex in stabilizing the complex by hydrogen
bonding both to the base pairs of UNA and to the intercalated quinoxaline ring
in the major
have
groove
observed
side ot the DNA double helix is a more
a similar
between
daunomycin
noticed
that
ana
other
interaction
d(CUTACG)
metal
ions
in the crystal
(unpublished
such
as
general one.
structure
data).
calcium
of the
Furthermore,
and
barium
ions
We
complex
we
have
in
the
crystallization mixture produced different crystal lattices in accordance with
these observations.
SPECIFICITY OF THE INTERACTION OF QUINOXALINE ANTIBIOTICS WITH DNA
Binding
preference
studies
tor
contents [8].
of both
triostin A
double-stranded
DNA,
and echinomycin
with
triostin
A
to DNA have
preferring
shown a
higher
GC
Footprinting studies have shown that echinomycin and triostin A
cover a binding site of tour to six Dase pairs [11, 12, 13] and the
tightest
binding was found when the sequence d(CpG) was located in the central two base
pairs.
The outer base pairs usually had AT or TA sequences.
in agreement with the details of this structure.
of
2318
the
central
two
base
pairs
is
determined
This result is
The CpG sequence
by
the
specificity
hydrogen
bonding
Nucleic Acids Research
interaction
different
guanine
of
the
sides
of
alanine
the
specificity
the
guanine
adenine
bond.
G2
with
helix.
for
the
has
not
a
It
necessarily
similar N3
the
is
alanine
The other alanine
does
residue
double
is high
bonds with guanine.
residues
guanine
bases
interesting
residue
to
position
it
that
note
the
two
that
the
that forms two hydrogen
that forms a single
define
on
for
hydrogen
guanine.
could
In
receive
bond
to
tact,
an
the
hydrogen
However, the footprinting experiments [11, 12, 13] show that all of the
strong binding sites have the sequence CG in the center of the four base pairs
ana
only
a tew
weak
binding
sites
have
a sequence
TG.
The
asymmetry
we
observe in the hydrogen bonding to guanine residues may not necessarily be an
inherent
feature
of
these
quinoxaline
antibiotics
that the UNA fragments we are using are
but
may
reflect
short and the quinoxaline
the
fact
ring 013
is
not intercalating the base pair that normally stacks on the other end.
The strongest
binding sites have a preference for AT base pairs flanking
the central CG sequence.
base pairs.
However,
Hoogsteen base pairs can be formed readily with AT
in order
to form
them with CG base pairs
must be protonated on N3 in order to hydrogen bond with
the
cytosine
the guanine N 7 .
This
apparently occurs less readily because flanking CG base pairs occur largely in
the weak binding sites rather
this
suggests
that
Hoogsteen base
strands
CG
pair geometry
approximately
Watson-Crick
than the
Hoogsteen
2
hydrogen
8
strong binding
base
brings
pairs
the
closer
to
bonding were
two
sugar
each
present
can
in
residues
other
and
sites [12] .
form
than
this
in
However,
solution.
on
the
they
turn
The
opposite
would
be
facilitates
if
a
large number of van der Waals interactions which has the effect of stabilizing
the binding.
The interaction of these quinoxaline antibiotics with the double
DNA fragment
make
are used to make
it possible
a bis
to
to the
triostin A. An exact
because
of
echinomycin
the
these particular
intercalating antibiotic.
pseudo-two-fold rotation axis due
This may be related
ask why
molecule,
symmetry
axis
the
actual
of
the
covalent
acids
The DNA double helix has a
strands.
in the molecular formula
is not found
dispositions
helical
amino
to the anti-parallel nature of the
two-fold
two-fold
skewed
eight
in the triostin A
cystine
bonding
of
side
the
of the
structure
chains.
In
the
thioacetal
bridge
on ribosomes but rather on special
enzyme
prevents a two-fold axis.
These antibiotics are not made
systems
that
can
accept
not
only
L-ami no
various other amino acid derivatives.
D—amino
acid
ana
two
N-methylated
acids
In the
amino
but
present
acids.
also
D— amino
instance
Only
acids
there
alanine
or
is one
has
no
2319
Nucleic Acids Research
Table I
Functional Descriptions o£ Amino Acids in Qninoxaline Antibiotics
Ami no Acid
Role £ l
Backbone
Role of
Side Chain
Role Qi
Modification
L-alanine
Cu methyl group
separates adjacent
sugars, wedges guanine
base out of plane; van
der Waals bonding to
sugars
NH groups form hydrogen
bond to both guanine
N3; carbonyl group
hydrogen bonds to one
guanine N2
L-cystine
Sulfur-containing cross
linkage provides rigidity
to cyclic peptide and is
positioned on the side
away from the base pairs
NH groups are methylated; N-methyl gronp prevents NH
carbonyl groups project
hydrogen bonding and allows
alanine carbonyl group in
away from base pairs
planar peptide linkage to
be oriented for hydrogen
bonding
L-valino
C methyl groups are in
der Waals contact with
sugar-phosphate linkage; other quinoxaline
antibiotics exist with
different side chains here,
including more aliphatic
groups
NH groups are methylated; carboxyl groups
form an ester linkage to
serine side chain; one
carbonyl group i s in
contact with adenine
base outside intercalative s i t e
N-methyl groups are in close
contact with cytosine rings,
changing base planes; methyl
groups prevent NH hydrogen
bonding and distortion of the
structure
D-sorine
Side chain provides ester
linkage to form cyclic
structure.
NH gronp i s used to form
peptide bond to quinoxaline, cannot be Nmethylated due to steric
hindrance; carbonyl
gronp projects away
from base pairs.
D-amino acid i s needed to
position quinoxaline ring
perpendicular to cyclic
depsipeptide structure.
modifications.
in
these
individual
central
Table 1 contains a functional
two quinoxaline
antibiotics.
side chains, the backbone
role
in both
hydrogen
description of the amino acids
It
summarizes
or thioacetal
the roles
and the modification.
bonding
with
the bases
interaction of its methyl group with the sugars.
forms a disulfide
Cannot have modifications
and maintain function
linkage
that
of the
Alanine plays a
residues
and in the
In L-cystine, the side chain
rigidifies
the cyclic
peptide.
Its Nil group is methylated, which prevents the intramolecular hydrogen bonding
that has been
seen
in the non-methylated
analog
of triostin A, TANDEM [14].
The NH group of cystine is linked to the carbonyl group of alanine
planar peptide linkage.
fixed
by hydrogen
carbonyl
bonding
to hydrogen
through a
Orientation of the alanine carbonyl group must not be
of the cystine
bond with
guanine*
NH group
This
in order
to allow the
is accomplished
by blocking
hydrogen bonding in the cystine Nil group through N-methylation.
The N—methylation of the L-valine groups plays a different role, since it
points
directly
contacts.
stabilizing
2320
toward
the cytosine
In the absence of a methyl
rings
group
and forms
there
tight
one would
van der Waals
lose
both the
interaction due to the methyl group and create a hole that
could
Nucleic Acids Research
not
accomodate
a
destabilization.
solvent
molecule
and
thus
would
represent
a
focus
of
Host of the carbon atoms of the valine side chain are not in
van der Waals contact with the sugar phosphate backbone, with the exception of
the C
Cl.
atom of triostin A, which is in van der Waals contact with the sugar
That
contact
echinomycin
are
is
both
lost
first
in
the
members
echinomycin
of
complex.
a family
of
Triostin
quinoxaline
A
The other members of the family differ by having variations on the side
of valine.
This involves not
and
antibiotics.
chain
only the normal amino acid but also side chains
in which there are additional carbon atoms that lengthen the side chains.
or
both
valine
side
chains
can
be
This is in agreement with the fact
substituted
that
this
for
alloisoleucine
is the
only side
One
residues.
chain of
the
peptide that can vary.
The serine residue has a side chain that plays a crucial role because its
hydroxyl group forms an ester linkage with
form the cyclic structure.
to position
structure.
to
The alpha amino group of serine is used to form an
amide bond to the quinoxaline ring.
in order
the carbonyl group of L valine
the
In this case, the D residue is essential
quinoxaline
rings
at
right
angles
to
the
cyclic
Use of the L residue would have the rings in the wrong orientation
for intercalation.
The
quinoxaline
occurring
Other
drugs
examples
that
antibiotics
interact
include
the
with
represent
DNA
by
another
binding
anthracyclines
class
onto
of
the
daunomycin [IS]
naturally
minor
and
groove.
adriamycin
(manuscript in preparation), actinomycin D [16, 17] and nonintercalators
as
netropsin
and
distamycin [IB].
Binding
to
the minor
groove
has
such
fewer
unique recognition features than does the major groove since, for instance, it
is difficult
However,
to distinguish
recent
they bind
analyses
largely
protein-UNA
of
AT from
TA base
bacterial
to the major groove
interactions,
pairs by hydrogen bonding [19] .
repressors
[20].
it is possible
that
bound
If that
to DNA
suggest
that
is a common feature
small molecule
binding may
of
be
biased in favor of minor groove binding.
The
structure
DNA have many
first
examples
of both the echinomycin and the
surprising features.
First
in which a peptide has
mechanism for the sequence
triostin A complexes with
of all, it illustrates
sequence-specific
one of
the
binding to DNA.
specificity involves hydrogen bonding
The
to bases and
also the participation of a conformational change involving the Hoogsteen base
pairs on the sequences
first
time
structure
on one
in which Hoogsteen
and
it suggests
that
side of
base
this
the intercalative
pairs
type
have been
of
site.
seen in
conf ormational
This is
the
oligonucleotide
lability
may
be
2321
Nucleic Acids Research
seen elsewhere
as well.
On the structural
level, the bis intercalative
mode
of action of the quinoxaline antibiotics that had been inferred from a variety
of
physical
studies [1, 2]
demonstrated here.
accomodate
itself
The crystal
It
was
well
as
model
to the more rigid peptide
structure
found
as
building [21, 22]
has
been
Finally, the ON A molecule is seen to be flexible enough to
that
of the
the
antibiotic
in a variety of ways.
triostin A alone has been solved recently [23].
dihedral
angles
of
the
peptide
backbone
in
that
structure are similar to those observed in the DNA-triostin A complex.
The structure of a complex of this type provides us with information that
can be used to direct the synthesis ot new antibiotics.
At the present time,
echinomycin
in the
human
what
is being
cancers [5] .
changes
might
we
biological
activity?
introduced,
one
exists
natural
used
in phase
If we were
introduce
There
II
clinical
going
in order
are
of which
is
in
variations
in
this
a
the
its
positions
can
be
tew
side chain could be decreased through
should
not
produce
a
side
of
where
chain
changes
where
antibiotics.
there
These
For example,
change
in
the
already
variations
the size of the
the introduction of alanine
significant
of
molecule,
of
valine
family
treatment
the echinomycin
to loox. for modifications
could be further modified in a number of ways.
This
trials
to redesign
or glycine.
interaction
of
the
antibiotic with the UNA but might modify the pharmacodynamics of the molecule.
Another example is to introduce positive charges by replacing a lysine residue
at that
position.
This may
increase
DNA due to charge attraction.
the binding of the antibiotic with the
A negatively
charged glutamic acid
that position might also change the biological effectiveness
Examination
of
intercalative
larger
the
quinoxaline
element
overlaps
intercalation.
with
with
three
the
bases
ring
suggests
fused
rings would
and
lead
biological
a
make
the
in
somewhat
for
larger
significantly
stabilization
due
to
Finally, if one or both of the alanine residue were changed to
glycine, there will be no methyl group
might
modify
that
residue
of the molecule.
to
a
significantly
activity
of
the
to wedge between the base pairs, which
modified
antibiotic.
interaction
These
that
ideas
would
affect
be
explored
can
the
by
chemical synthetic methods to search for more effective antitumor agents.
Acknowledgments:
Institutes
Office
the
of
of
Health,
Naval
Netherlands
acknowledges
This research was supported by grants from the National
National
Science
Research, National
Organization
support
from NATO
for
the
and
Advancement
American
Cancer
We
Society,
Space Administration
of
Pure
Research.
ano Istituto Strutturistica Chimica,
Nazionale Delle Richerche (Italy).
2322
Foundation,
Aeronautics
thank Dr. T. Yoshida
and
G.U.
Consiglio
of Shionogi Co.,
Nucleic Acids Research
Osaka, Japan, tor the t r i o s t i n A ana Dr. M.J. Waring of the University of
Cambridge, England, for the echinomycin.
§To whom correspondence should be addressed
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
HCES
Katagiri, K., Yoshida, T. and Sato, K. (1975)in Antibiotics III
Mechanism of Action of Antimicrobial and Antitumor Agents, J.W. Corcoran
ana F.E. Hahn, Eds., Ed., pp. 234-251, Springer-Verlag, Heidelberg,
Waring, H.J. (197!') in Antiobiotics V Mechanism of Action of
Antieukaryotic and Antiviral Compounds, F.E. Hahn, Ed., Ed., pp.
173-194, Springer-Verlag, Berlin, , part 2
Ward, D.C., Reich, E. and Goldberg, I.U. (1965) Science 149, 1259-1263.
Sato, K., Shiratori, 0. and Katagiri, K.J. (1967) J. Antibiotics 20,
270-276.
Foster, B.J., et al. (1985) , Submitted for publication
Wang, A.H.-J., Ughetto, G., Uuigley, G.J., Hakoshima, T., van der Harel,
G.A., van Boom, J.H. and Rich, A. (1984) Science 225, 1115-1121.
Hoogsteen, K. (1959) Acta Crystallography 12, 822-823.
Lee, J.S. and Waring, M.J. (1978) Biochem. J. 173, 115-128.
van der Marel, G., van Boeckel, C.A.A., Wille, G. and van Boom, J.H.
(1981) Tetrahedron Letters 22, 3887-3890.
R. Srinivasan Ed., (1979)Biomolecular Structure: Conformation, Function
and Evolution,, 1, Pergamon, Oxford, .
Low, C.M.L., Drew, H.R. and Waring, M.J. (1984) Nucleic Acids Research
12, 4865-4879.
van Dyke, M.M. and Dervan, P.B. (1984) Science 225, 1122-1127.
Low, C.M.L., Olsen, K.K. and Waring, M.J. (1984) FEBS Letters 176,
414-420.
Hossain, M.B. e£ al. (1982) J. Amer. Chem. Soc. 104, 3401-3408.
Uuigley, G.J., Wang, A.H.-J., Ughetto, G., van der Marel, G., van Boom,
J.H. ana Rich, A. (1980) Proc. Natl. Acad. USA 77, 7204-7208.
Sobell, H.M., Jain, S.C., Sakore, T.D. and Nordman, C.E. (1971) Nature
(London) New Biol. 231, 200-201.
Xakusagawa, F.M., Dabrow, S., Neidle, S. and Berman, U.M. (1982) Nature
(London) 290, 466-467.
Zimmer, C. (1975) Progr. Nucl. Acid. Res. Mol. Biol. 15, 285-318.
Seeman, N.C., Rosenberg, J.M. ana Rich, A. (1976) Proc. Natl. Acad. Sci.
USA 73, 8U4-808.
Pabo, C O . and Sauer, R.T. (1984) Ann. Rev. Biochem. 53, 293-321.
Ughetto, G. and Waring, M.J. (1977) Mol. Pharmacol. 13, 579-584.
Viswamitra, M.A. etal, (1981) Nature 289, 817-819.
Sheldrick, G.M., Guy, J.J., Kennard, 0., Rivera, V., Waring, M.J. (1984)
J. Chem. Soc. Perkin Trans. II 1601-1605.
2323