Sequence selective binding of bis-daunorubicin WP631 to DNA
Keith R. Fox
1
, Richard Webster
1
, Robin J. Phelps
1
, Izabela Fokt
2
and Waldemar Priebe
2
1
School of Biological Sciences, University of Southampton, Bassett Crescent East, Southampton, UK;
2
The University of Texas MD
Anderson Cancer Center, Houston, TX, USA
We have used footprinting techniques on a wide range of
natural a nd synthetic footprinting substrates to examine t he
sequence-selective i nteraction of the bis-daunorubicin anti-
biotic WP631 with DNA. The ligand produces clear
DNase I footprints that are very diff erent f rom t hose s een
with other anthracycline antibiotics such as daunorubicin
and nogalamycin. Footprints are found in a diverse range of
sequences, many of which are rich in G T ( AC) o r G A ( TC)
residues. As expected, the ligand binds well to the sequences
CGTACG and CGATCG, but clear footprints are also
found at h exanucleotide sequences s uch GCATGC and
GCTAGC. The various footprints do not contain any par-
ticular unique d i-, tri- or tetranucleotide s equences, but are
frequently contain the sequence (G/C)(A/T)(A/T)(G/C). All
sequences with this composition a re protected by the ligand,

restrictions imposed by the linkers between the two
intercalators. In addition, because the parent compounds
bind to almost all DNA sequences, t he oligomers show
little o r n o s equence selectivity.
The anthracycline a ntibiotics are well known antitumour
agents [8–11] a nd, although they d isplay a p leiotropic
mechanism of action, DNA is their p rimary cellular target.
The best characterized members of this group are dauno-
rubicin (daunomycin) and doxorubicin ( adriamycin). These
agents bind to DNA by intercalation, with the amino sugar
daunosamine positioned i n the DNA minor groove. Several
crystal s tructures h ave been reported f or the interaction of
these ligands with oligonucleotides, i ncluding CGTACG
[12,13], CGATCG [14,15] a nd TGTACA and T GATCA
[16]. They possess some sequence specificity and high
resolution footprinting has suggested that they bind best to
the sequences 5 ¢-(A/T)CG and 5¢-(A/T)GC [1 7–19].
There have been a number of attempts to produce bis-
intercalating dau norubicin derivatives, with increased affin-
ity for DNA. I n early studies these were linked through C 13
and C 14 as these are chemically accessible [20,21]. H owever
these positions are involved in DNA binding and the
modifications decreased the affinity of each mono mer.
More recently dimers of daunorubicin have been produced
by linking between t he C-4¢ or C-3¢ sugar positions [22,23].
These compounds were designed after examination of the
crystal structure of daunorubicin bound to CGTACG [13].
This s tructure c ontains two daunorubicin molecules which
are intercalated at the CpG steps with their amino sugars
facing each other at the centre o f the complex, with the

bound preferentially to GC-rich DNA regions [27]. B y
comparison with the crystal structures of daunorubicin
we would expect these ligands to bind to the s equence
CG(A/T)(A/T)CG. NMR [28] a nd cry stal structures [ 29]
have been derived f or the interaction of WP631 with
CGTACG and CGATCG, respectively, and as expected
show that the ligand binds by bisintercalation with each
chromophore inserted into the CpG steps, with four base
pairs sandwiched between them. In contrast, prolonged
incubation of WP652 with CGTACG resulted in precipi-
tation, and the NMR structure was d etermined for this
ligand bound to TGTACA [28]. In t his structu re the ligand
is bound a cross the sequence P yGTPu, with only t wo base
pairs between the i ntercalated chromophores.
These studies have demonstrated that WP631 binds
tightly to DNA by bisintercalation and assume that it
recognizes the sequence CG(T/A)(T/A)CG. How ever, there
have been no previous studies examining its sequence
binding preferen ces, t hough i t has been demonstrated that
WP631 inhibits Sp1-activated transcription in vitro [24,30 ].
In this paper we e xamine t he DNA sequence s pecificity o f
WP631 using a r ange of footprinting techniques o n several
different DNA fragments.
Materials and methods
Chemicals and enzymes
Oligonucleotides for preparing the v arious DN A f ragments
were purchased from Oswel DNA service (Southampton,
UK). These were stored in water at )20 °C,anddilutedto
working concentrations immediately before use. P lasmid
pUC19 was purchased from Pharmacia. DNase I was

.C
6
T
6
and G
6
A
6
.T
6
C
6
inserted
into the BamHI site of pUC18 [34]. Radiolabelling visualizes
the purine-rich strand of AG1, but the p yrimidine-contain-
ing s trand o f G A1. F ragments WPseq 1 and WPseq2 were
obtained by cloning appropriate oligonucleotides into the
BamHI site of pUC18. The sequences were confirmed by
manual sequencing with a T7 sequencing kit (Amersham
Pharmacia). Fragment W Pseq2 was found to contain a
dimer of the required i nsert. These fragments were obtained
by cutting the plasmids with HindIII and SacIandtheywere
labelled a t the 3¢-end of the Hin dIII site with [
32
P]dATP[aP]
using reverse transcriptase. Radiolabelled DNA was s epar-
ated from the remainder of the plasmid on 6–8% non-
denaturing polyacrylamide gels. The bands containing the
radiolabelled DNA were excised and eluted into 10 m
M

digesting with either DNase I or a hydroxyl radical
generating mixture as p reviously described [31]. DNase I
digestion was achieved by adding 2 lL enzyme (typically
0.01 UÆmL
)1
) dissolved in 20 m
M
NaCl, 2 m
M
MgCl
2
,and
2m
M
MnCl
2
. The digestion was stopped a fter 1 min by
adding 5 lL o f 80% formamide c ontaining 10 m
M
EDTA,
10 m
M
NaOH and 0 .1% (w/v) bromophenol blue.
Hydroxyl radical footprinting
Hydroxyl radical cleavage was performed by adding 6 lLof
a freshly prepared mixture containing 50 l
M
ferrous
ammonium sulf ate, 100 l
M

potassium
permanganate was added a nd the reaction s topped after
1 min by adding 2 lL o f mercaptoethanol. T he DNA was
then precipitated with ethano l in the presence of 0.3
M
sodium acetate. For both DEPC and permanganate the
dried DNA pellets were boiled i n 10% (v/v) piperidine f or
30 min, reduced to dryness in a Sp eedvac, and redissolved in
8 lL o f 80% form amide containing 10 m
M
EDTA, 10 m
M
NaOH and 0 .1% (w/v) bromophenol blue.
Fig. 2. DNase I, DEPC and KMnO
4
foot-
prints showing the inte raction of WP631 with
tyrT(43–59). WP631 concentrations ( lM) are
shownatthetopofeachgellane;ÔconÕ cor-
responds to cleavage in the absence of added
ligand. Tracks labelled ÔGAÕ are markers
specific for purines.
3558 K. R. Fox et al. (Eur. J. Biochem. 271) Ó FEBS 2004
Denaturing gel electrophoresis
The products o f footprintin g reactions were resolved on
6–10% polyacrylamide gels (depending on the location of
the t arget site) containing 8
M
urea. DNA samples were
boiled f or 3 min immediately before loading onto the gels.

M
), the ligand shows a general i nh ibition o f
cleavage at most positions in the fragment. However specific
regions of protection are evident with ligand c oncentrations
between 0.2 and 1 l
M
. Examples of bands that are protected
by the ligand include positions 34, 41, 53 and 61. In contrast,
cleavage at positions 31–32 and 47–50 is enhanced in the
presence of the ligand. These r esults are presented as a
differential cleavage plot in the top panel of F ig. 3 , showing
the intensity of each band in the drug-treated lanes
compared with that in the control. Examination o f the
patterns does not reveal any obvious sequence p reference,
though some o f the clearest footprints are located in regions
containing both G and A residues. The enhancements are
located in oligo(dA) tracts, as often noted with intercalating
agents. These footprints are of variable lengths. The
footprints around positions 40, 63 and 8 0 c over about six
bases, as might be e xpected for a bis-intercalator. However,
below position 60 t here are two smaller f ootprints of about
Fig. 3. Differential c leavage plots showing the i nteraction of WP631 with tyrT(43–59), AG1 and GA1. The plots were calculated from th e cleavage
patterns in the presence of 1 l
M
WP631 shown in F ig. 2 (tyrT)andFig.6(AG1andGA1).Onlyapartof each sequence is s hown and is written
reading 5 ¢)3¢ from left to right; the right-hand e nd corresponds to the bottom of the gels. The ordinate, w hich is plotted on a logarithmic scale,
shows the intensity of e ach band i n the drug-treate d lanes relative to that in the control. Values lessthanonecorrespondtoprotectionbytheligand,
while values above indicate enhanced cleavage. The black b ars highlight the regions that are p rotected from cleavage. For tyrT t he arrows indicate
the positions of WP631-induced c leavage by DEPC (grey arrows) and KMnO
4

DEPC are located at positions 18, 32, 48, 67, 83 and 84. In
some instances these a re located i n regions o f enhanced
DNase I cleavage (positions 32 and 48), while others are
adjacent to regions of DNa se I protection (18, 67, 83, 84) .
Enhanced reactivity to KMnO
4
can be seen at posit ions 29,
33, 60, 6 8, 81, 86, 88 and 9 1.
These results show that WP631 produces distinct foot-
printing patterns, w hich are different to those produced by
daunorubicin and nogalamycin [17,19,35,36]. The ligand
must therefore possess some sequence selectivity, though no
consensus b inding sites can be deduced from these patterns.
We have therefore examined the interaction of this ligand
with a r ange of DN A fragments, i n o rder to elucidate the
characteristics of t he preferred binding sites.
Fig. 4. DNase I and KMn O
4
footprints showing the interaction of WP631 with fragments MS1 and MS2. WP631 concentrations (l
M
)areshownat
the top of each gel lane; ÔconÕ corre sponds to cleavage in the absence of added liga nd. Tracks labelled ÔGAÕ are markers specific for purines. The
numbered black bars s how the positions of D Nase I footprints, while the aste risks indicate b ands that be come sensitive t o reaction with K MnO
4
in the presence of W P631.
3560 K. R. Fox et al. (Eur. J. Biochem. 271) Ó FEBS 2004
Fragments M S1 and MS2 were designed so as to contain
all 136 tetranucleotide sequences [32]. They contain i dentical
sequences, but are cloned i n opposite orientations thereb y
simplifying analysis of bands at the ends of the fragments.

.C
6
T
6
, r espectively [34]. DNase I footprinting patterns
for WP631 with these fragments are shown in Fig. 6 and
differential clea vage plots derived from these are present ed
in Fig. 3. As these oligop urine tracts were both cloned into
the polylinker s ite o f p UC18 the sequences surrounding the
inserts are common to both fragments and show similar
cleavage patterns in the presence of the ligand. For both
fragments there is a large footprint below the insert,
corresponding to the sequence TCCTCT. Similarly cleavage
is attenuated above the inserts in vicinity of the sequence
GGATC. H owever the ligand h as very different effects on
cleavage of the two inserts. WP631 protects from D Nase I
cleavage at the centre of AG1, but causes enhanced cleavage
at the centre of GA1. It therefore appears that A
n
G
n
is a
much better binding site than G
n
A
n
.
Fragment DMG60 a lso c ontains oligopurine tracts that
are interrupted by isolated thymine r esidues [ 33]. DNase I
digestion p atterns for the pyrimidine-rich strand of this

shown in Fig. 7 . It can be seen that there a re footprints at all
the potential sites, which a re most clearly seen in the
differential cleavage plot. T he strongest s ites are a t
GTGTTG and CTTCTC. There is little or no protection
in the junctions between t he various sites a nd there i s
enhanced cleavage between GTGGTG and CCACAC.
These r egions of protection a re located towards the 3¢-end
of each target site as normally observed w ith D Nase I
footprinting, as th is enzyme c uts across t he width of the
DNA minor groove.
Daunorubicin is thought to bind best to sequences of the
type 5¢-(A/T)CG and 5¢-(A/T)GC [17–19] and previous
NMR and crystallographic studies with bis-daunorubicins
have investigated their interaction with CGTACG and
TGTACA [28,29]. None of these sequences are r epresented
in any o f the footprinting substrates mentioned a bove. We
therefore prepared a novel fragment (WPseq1) containing
the sites CGATCG, CGTACG, GCATGC, GCTAGC and
TGTACA each separated b y the sequence A ATT t o which
the drug is not expected to bind. The r esults of footprinting
experiments with t his fragment a re presented i n Fig. 8 . T he
DNase I c leavage patterns show footprints at each of these
sites, some of which persist to between 0.1 a nd 0.2 l
M
.The
positions of these sites are confirmed i n the differential
cleavage plot shown i n Fig. 8(B). Although DNase I
footprinting cannot usually be used to determine ligand
binding sites t o single b ase r esolution, some interesting
features of WP631 binding can be deduced by comparing

)
are shown at the t op of each gel lane; ÔconÕ
corresponds to cle avage in the absence of
added ligand. Tracks labelled ÔGAÕ are m ark -
ers s pe cific for purines. The nu mbered black
bars show the positions of DNase I foo tprints
with DMG60Y.
3562 K. R. Fox et al. (Eur. J. Biochem. 271) Ó FEBS 2004
GCATGC and GCTAGC, while this is at the second
adenine (AATT) for CGATCG and TGTACA. T here is no
enhancement i n r eactivity t o DEPC after CGTACG. These
subtle differences suggest that WP631 does not have exactly
the same m ode of binding at each of these s ites.
Discussion
The footprinting results p resented in this paper demonstrate
that WP631 binds to DNA in a sequence s elective fashion
and that i ts preferred binding sites a re different from t hose
of daunorubicin and nogalamycin. By c omparison with
daunorubicin it was expected that WP631 should bind best
to sequences such as CGATCG and CGTACG, which have
been used in X-ray and NMR structural studies with this
ligand [28,29]. The experiments with fragment WPseq2
confirm that WP631 does indeed bind to this site at
concentrations as low as 0 .2 l
M
, but experiments with t his
and other fragments show that it also b inds equally well to
other sequences.
Another difference between these patterns a nd those
produced by daunorubicin is t heir temperature d ependence.

Onlyapartofeachsequenceisshownandiswrittenreading5¢)3¢ fro m left to right; the r ight-hand end co rrespo nds to the b ottom o f the gel. The
ordinate, which is plotted on a logarithmic s cale, shows t he intensity of eachbandinthedrug-treatedlanesrelative t o that in the c ontrol. Values less
thanonecorrespondtoprotectionbytheligand,while values above indicate enhanced cleavage. The vertical lines divide the f ragment into t he
various h exanucleo tide repeats.
Ó FEBS 2004 WP631 sequence selectivity (Eur. J. Biochem. 271) 3563
These different str uctures suggest that WP631 may b ind to
different sequences in different modes, sandwiching between
two a nd four base pairs b etween the c hromophores. These
different modes will depend on the local DNA structure and
flexibility as well a s any contacts between the ligand and its
binding site. A further complication is the possibility that
WP631 might bind to some sequences by mono-intercala-
tion, leaving the second chromophore i n free solution or
stacked within t he groove. The possibility of additional
sequence-specific groove binding may further complicate
the footprinting pattern. The coexistence of different
binding modes is suggested by the footprinting d ata
presented i n t his paper. Some binding sites are six to eight
base pairs long, as expected for a ligand that spans six base
pairs, while others are much shorter, and appear to cover
only t hree bases . Thes e r esults are consistent with a recent
study sugges ting that WP631 can bind in two d ifferent
modes with stoichiometries of 6 : 1 a nd 3 : 1 base pairs per
drug [41].
Sequence selectivity
The results with these DNA fragments show that
WP631binds t o DNA in a sequence s elective fashion, as
specific footprints are g enerated at moderate ligand c on-
centrations (about 0.3 l
M

(G/C)(A/T)(A/T)(G/C), and that there a re no occasions
when this is not part of a drug b inding site. For example, on
Fig. 8. Interaction of WP631 with fragment WPseq2. (A) DNase I and DEPC foo tprints. W P631 c oncentrat ions (l
M
)areshownatthetopofeach
gel lane; ÔconÕ co rresponds to cleavage in the absence of added ligand. Tracks labelled ÔGAÕ are m arkers specific for purines. The p otential
hexanucleotide binding sequences are indicated alongside the gel. (B) Differential cleavage plot s howing the interaction of WP631 with WPseq1. The
plot wa s calculated from the cleavage patterns in the pre sence of 0.2 l
M
WP631 shown i n Fig. 8A. Only a part o f each s equence is s how n and is
written r eading 5¢)3¢ from l eft to righ t; the r ight-hand end corresponds to the bottom o f the gel. The o rdinate, which is plotted on a logarithmic
scale, shows the intensity of each b and in the drug -treated lanes relative to that in the control. Valuesoflessthanonecorrespond to protection by
the ligand, while values above one indicate enhanced c leavage. The arrows indicate the positions of W P631-induced cleavage by DEPC.
3564 K. R. Fox et al. (Eur. J. Biochem. 271) Ó FEBS 2004
MS1 the foot prints are at s ite 1 (CATC), s ite 3 (GTAC) and
site 4 (GAAG), while on MS2 they are seen at site 7
(CATG), s ite 8 (GTTG), s ite 9 (CTTG and GATC). In
addition the weaker regions of protection between sites 8
and 9 contain the sequences CTAC and CTAG. This
consensus sequence is also found on the tyrT fragment at
positions 25 (CATC), 38 (GTTG), 43 (GAAC) and 5 7
(GAAG) each of wh ich corresponds to a r egion t hat is
protected by t h e ligand. The s equences GATC and C TGA
are a lso f ound in the polylinker r egions of pAG1 and
pGA1, and at site 2 in DMG60 (CTTC). We therefore
suggest that WP631 binds well to the sequence (G/C)(A/
T)(A/T)(G/C). However, t his s equence c annot be the only
good ligand b inding site. F or example, the f ootprint at t he
centre of pAG1 contains the sequence AGGG (in contrast
to GGGA, which does not produce a footprint with pAG2).

WP631 strongly self-associates and t he total ligand concen-
tration may overestimate the concentration of the free
momomer.
Acknowledgements
This work was supported by grants from the Cancer Research UK, the
Association for International C ancer R esearch and The Welch
Foundation, Houston , Texas, USA.
References
1. Thuong, N.T. & H e
´
le
`
ne, C. (1993) Sequen ce specific rec ognition
and modification of double helical D N A by o ligonuc leotides.
Angew. Chemie. Int. Ed. Eng. 32, 666–690.
2. Fox, K.R. (2000) Targeting DNA w ith triplexes. Curr. Med.
Chem. 7, 17–37.
3. Dervan, P.B. & Bu
¨
rli, R.W. (1999) Sequence-specific DNA
recognition b y polyamides. Curr. Opin. Chem. Biol. 3, 688–693.
4. Wemmer, D.E. ( 2000) D esigned se quence-specific m inor groove
ligands. Annu. R ev. Biophys. Biomol. Struc t. 29 , 439 –461.
5. Guelev, V.M., Cubberley, M.S., M urr, M.M., Lokey, R.S. &
Iverson, B.L. (2 001) Design, s ynthesis and characterization of
polyintercalating ligands. Methods Enzymol. 340, 556–570.
6. Chaires, J.B. (1998) D rug–DNA interactions. Curr. Opin . S truct.
Biol. 8, 314–320.
7. Wakelin, L.P.G. (1986) Polyfunctional DNA intercalating agents.
Med. Res. Rev. 6, 275–340.

J. Mol. Biol. 222, 167–177.
17. Chaires, J.B., Fox, K.R., Herrera, J.E., Britt, M. & Waring, M.J.
(1987) Sit e and sequence specificity of the da unomycin–DNA
interaction. Bioc hemistry 26 , 8 227–8236.
18. Skorabogaty, A., White, R.J., Phillips, D.R. & Reis, J.A. (1988)
The 5¢-CA DNA-sequence p refere nce of daunomycin. FE BS Le tt.
227, 103–106.
19. Chaires, J.B., Herrera, J.E. & Waring, M.J. (1990) Preferential
binding of daunomycin to 5 ¢A/TCG a nd 5¢A/T GC sequences
revealed by footprinting titration experiments. Biochemistry 29,
6145–6153.
20. Phillips, D.R., Brownlee, R.T.C., R eiss, J.A. & Scourides, P.A.
(1992) Bis–daunomycin hydrazones – interactions with DNA.
Invest. N ew Dru gs 10, 79–88.
21. Skorobogarty, A., Brownlee, R.T.C., Chandler, C .J., Kyratzis, I.,
Phillips, D.R., R iess, J.A. & Trist, H. (1988) The DNA association
and biological activity of a new bis(14-thiadaunomycin). Anti-
cancer Drug De s. 3, 41–56.
22. Priebe, W., Fokt, I., Przewloka, T., Chaires, J.B., Portugal, J. &
Trent, J.O. (2001) Exploiting anthracycline scaffold for designing
DNA-targeting agents. Meth ods Enz ymol. 340, 529–555.
23. Chaires,J.B.,Leng,F.,Przewloka,T.,Fokt,I.,Ling,Y H.,Perez-
Soler, R. & Priebe, W. (1997) Structure-based design of a new
bisintercalating an thracycline antibiotic. J. Med. Chem. 40, 261 –
266.
24. Martin, B., Vaquero, A., Priebe, W. & Portugal, J. ( 1999) Bisa n-
thracycline WP631 inhibits basal and Sp1-activated transcription
initiation in vitro. N ucleic Acids Res. 27 , 3402–3409.
25. Ashikawa, K., Shishodia, S., Fokt, I., Priebe, W. & A ggarwal,
B.B. (2004) Evidence that ac tivation of nuclear factor-kappa B is

33. Fox, K.R., Flashman, E. & G owers, D. (2000) Secondary
binding sites for triplex-forming oligonucleotides containing bul-
ges, lo ops, and mismatches i n the t hird strand. Bioc hemistry 39,
6714–6725.
34. Stonehouse, T.J. & Fox, K.R. (1994) DNase I footprinting of
triple helix formation at polypurine tracts by acridine-linked
oligopyrimidine. Bioc him. Biophys. Acta 1218, 322–330.
35. Fox, K.R. & Waring, M.J. (1 986) Nucleotide sequence b inding
preferences of nogalamycin investigated by DNase I footprinting.
Biochemistry 25, 4349 –4356.
36. Fox, K.R. (198 8) Footpr inting studies on the interactions of
nogalamycin, arugomycin, decilorubicin and viriplanin with
DNA. Anti-Cancer Drug D esign 3, 157–168.
37. Churchill, M.E.A., Hayes, J.J. & Tullius, T.D. (1990) Detection of
drug bi nding to DNA by h ydroxyl rad ical footprinting: relation-
ship of distamycin binding sites to DNA structure and p os itioned
nuclesomes on 5S RNA genes of Xenopus. Biochemistry 29, 6043–
6050.
38. Portugal, J ., Fox, K.R ., McLean, M.J., Richenberg, J.L. &
Waring, M.J. (1988) Diethyl pyrocarbonate can detect a modified
DNA structure induced by t he binding o f quinoxaline a ntibiotics.
Nucleic Acids Res . 16 , 3655–3670.
39. Jeppesen, C. & Nielsen, P.E. (1988) Detection of intercalation-
induced c hanges in DNA structure by reaction with die thylpyro-
carbonate o r potassium permanganate. Evidence against the
induction o f Hoogsteen b ase pairing b y echinomycin. FEBS Lett.
231, 172–176.
40. Fox, K.R. & Grigg, G.W. (1988) Diethylpyrocarbonate and per-
manganate provide evidence for an unusual DNA conform ation
induced by binding of t he antitumour a ntibiotics bleom ycin and