Doxazolidine, a proposed active metabolite of doxorubicin that cross-links DNA

Article abstract of DOI:10.1021/jm050678v

A crystal structure establishes doxoform as a dimeric formaldehyde conjugate of the oxazolidine of doxorubicin. Doxoform is a prodrug of doxazolidine, a monomeric doxorubicin formaldehyde-oxazolidine. Both doxoform and doxazolidine inhibit the growth of cancer cells at 1-4 orders of magnitude lower concentration than doxorubicin. They also inhibit the growth of cancer cells better than doxsaliform, a prodrug for an acyclic doxorubicin-formaldehyde conjugate. Doxoform rapidly hydrolyzes to doxazolidine, which then hydrolyzes to doxorubicin with a half-life of 3 min in human serum at 37 °C. Both doxoform and doxazolidine are taken up by multidrug-resistant MCF-7/Adr cells 3- to 4-fold better than doxorubicin. A molecular model suggests that doxazolidine can cross-link DNA by direct reaction with a G-base in a tautomeric form with synchronous ring opening of the oxazolidine. These results point to doxoform being a prodrug for doxazolidine that is the reactive species that directly cross-links DNA.

Full text of DOI:10.1021/jm050678v

7
648
J. Med. Chem. 2005, 48, 7648-7657
Doxazolidine, a Proposed Active Metabolite of Doxorubicin That Cross-links
DNA
‡
‡
‡
‡
,‡,§
Glen C. Post, Benjamin L. Barthel, David J. Burkhart, John R. Hagadorn, and Tad H. Koch*
Department of Chemistry and Biochemistry, University of Colorado at Boulder, Boulder, Colorado 80309-0215, and
Colorado Cancer Center, Aurora, Colorado 80010
Received July 15, 2005
A crystal structure establishes doxoform as a dimeric formaldehyde conjugate of the oxazolidine
of doxorubicin. Doxoform is a prodrug of doxazolidine, a monomeric doxorubicin formaldehyde-
oxazolidine. Both doxoform and doxazolidine inhibit the growth of cancer cells at 1-4 orders
of magnitude lower concentration than doxorubicin. They also inhibit the growth of cancer
cells better than doxsaliform, a prodrug for an acyclic doxorubicin-formaldehyde conjugate.
Doxoform rapidly hydrolyzes to doxazolidine, which then hydrolyzes to doxorubicin with a half-
life of 3 min in human serum at 37 °C. Both doxoform and doxazolidine are taken up by
multidrug-resistant MCF-7/Adr cells 3- to 4-fold better than doxorubicin. A molecular model
suggests that doxazolidine can cross-link DNA by direct reaction with a G-base in a tautomeric
form with synchronous ring opening of the oxazolidine. These results point to doxoform being
a prodrug for doxazolidine that is the reactive species that directly cross-links DNA.
Introduction
Chart 1. Structures for Doxorubicin and Its Clinical
Congeners: Epidoxorubicin, Daunorubicin, and
Idarubicin
Doxorubicin (Dox) is a broad-spectrum antitumor
drug of the anthracycline class that has been used in
the clinic for more than 30 years for the treatment of
leukemias, lymphomas, sarcomas, and solid tumors.1
Treatment is limited by chronic cardiotoxicity, and
extensive research has focused on the mechanism of
action toward discovery of methodology to improve the
therapeutic index.2,3 Many synthetic and natural de-
rivatives have been explored as well as improved drug
delivery systems. A stealth liposomal formulation is
4
emerging as a significant improvement. Shortly after
the discovery of Dox, DNA was identified as an impor-
tant target into which Dox and its clinical congeners,
daunorubicin, idarubicin, and epidoxorubicin (Chart 1),
of anthracycline-DNA adducts and virtual cross-
links, consult a recent paper by Phillips and co-
5
,6
reversibly intercalate. Circumstantial evidence ac-
cumulating over a period of years from laboratories of
Phillips, Konopa, Wang, and Koch, among others, now
points to covalent bonding of Dox and its congeners to
DNA within the intercalation complex, mediated by
21
workers.
DoxF, synthesized by reaction of Dox with formalin,
has a dimeric structure with two molecules of Dox
condensed with three molecules of formaldehyde (Scheme
1
). DoxF is unstable with respect to hydrolysis back to
7
-14
formaldehyde.
The result is a virtual cross-linking
Dox, with a half-life of only a few minutes at 25 °C in
RPMI 1640 cell culture medium, but despite this
of the two strands of DNA by drug, illustrated by the
crystal structure in Figure 1. The source of the form-
aldehyde in vitro and in vivo remains uncertain;15
however, treatment of MCF-7 breast cancer cells with
doxorubicin or daunorubicin results in an intracellular
elevation of formaldehyde concentration.16 These dis-
coveries prompted the syntheses of the doxorubicin-
17
instability, it inhibits the growth of both sensitive and
multidrug-resistant tumor cells about equally, with IC50
values ranging from 1 to 4 orders of magnitude lower
than those with Dox (Table 1). The magnitude of the
difference in efficacy appears to parallel the magnitude
of the multidrug resistance. Reaction of DNA with DoxF
yields the same cross-links as treatment with Dox and
1
7
formaldehyde conjugates doxoform (DoxF) and dox-
1
8
saliform (DoxSF) as drug candidates that carry their
own formaldehyde (Scheme 1) and exploration of coad-
ministration of doxorubicin with formaldehyde-releasing
17
formaldehyde, suggesting that DoxF serves as a pro-
drug to a monomeric species that cross-links DNA
(
Scheme 1).
1
9,20
prodrugs.
For an excellent review of the discovery
DoxSF is the salicylamide N-Mannich base of Dox
with formaldehyde. It releases the acyclic doxorubicin-
*
To whom correspondence should be addressed at University of
formaldehyde conjugate with a half-life of 1 h under
physiological conditions. The mechanism for cleavage
Colorado at Boulder. Phone: 303-492-6193. Fax: 303-492-5894.
E-mail: tad.koch@colorado.edu.
18
‡
of the N-Mannich base is proposed on the basis of
University of Colorado at Boulder.
Colorado Cancer Center.
§
22
mechanistic studies by London and co-workers and our
1
0.1021/jm050678v CCC: $30.25 © 2005 American Chemical Society
Published on Web 11/04/2005

Doxazolidine Cross-links DNA
Journal of Medicinal Chemistry, 2005, Vol. 48, No. 24 7649
Figure 1. Crystal structure of daunorubicin virtually cross-linking double-stranded 5′-CGCGC-3′/3′-GCGCG-5′ DNA at 5′-CGC-
′. The daunosamine sugar of daunorubicin is in a chair conformation. The virtual cross-linking of the two DNA strands is achieved
3
through a covalent bond to one strand and hydrogen bonding to the other strand plus hydrophobic interactions with both strands.
The image was created from data provided by Wang and co-workers10 at the Rutgers Protein Data Bank (PDB code 1D33) using
Chem 3D and PyMOL software.
1
observation that the cleavage is inhibited at low pH
polymer of formaldehyde, with monitoring by H NMR
showed formation of Doxaz followed by formation of
DoxF. Doxaz was isolated 90% pure (73% yield) by
stopping the reaction at an intermediate time with the
only impurities being traces of Dox and DoxF. Cor-
respondingly, DoxF was isolated in greater than 90%
purity (79% yield) by allowing the reaction to continue,
again with the only impurity being Doxaz. The structure
of Doxaz was established from an intense, doubly
charged molecular ion at m/z 278.8 in the electrospray
because of protonation at the 3′-amino substituent of
1
8
DoxSF. The lifetime of the acyclic doxorubicin-
formaldehyde conjugate with respect to hydrolysis to
Dox is unknown but presumably is very short. DoxSF
also inhibits the growth of sensitive and resistant cancer
cells at lower concentrations than Dox; however, it is
substantially less active than DoxF (Table 1).
We now present evidence that DoxF is a prodrug for
monomeric doxorubicin oxazolidine, doxazolidine (Dox-
az), and that Doxaz is more reactive than acyclic
conjugates in cross-linking DNA. Further, we report the
synthesis, isolation, and characterization of Doxaz as
well as kinetics for hydrolysis of DoxF to Dox via Doxaz,
in vitro activity of Doxaz against sensitive and resistant
cancer cells, and a molecular model for reaction of Doxaz
with DNA. During the course of these studies we also
obtained a crystal structure of DoxF.
1
mass spectrum and from the high-resolution H NMR
data reported in Table 2 with all of the J couplings
assigned in comparison with data for DoxF. Of particu-
lar note in the NMR spectrum of Doxaz is the absence
of the singlet peak for the methylene connecting the two
oxazolidine rings of DoxF and the characteristic small
geminal coupling constant for the methylene protons of
the oxazolidine.
The NMR data also partially establish the conforma-
tion of the daunosamine sugar of Doxaz and DoxF in
chloroform solution as a chair. The proton at the
1′-position is coupled approximately equally to the two
protons at the 2′-position in both structures, indicating
similar dihedral angles and a chair conformation.
Although this is consistent with what others have
Results and Discussion
Synthesis of Doxaz and DoxF. The original syn-
theses of DoxF and its congeners daunoform and epi-
doxoform (EpiF) were all performed by reaction of the
respective clinical drug as its hydrochloride salt with
formalin, an aqueous methanolic solution of formalde-
hyde, in acetate buffer at pH 6. The dimeric conjugates
were extracted into chloroform as they were formed.
With this method, the presumed intermediate, Doxaz,
was not observed. In contrast, reaction of Dox free base
in chloroform-d solvent with paraformaldehyde, the
observed in the crystal structures of various anthracy-
cline antitumor drugs2
3,24
and in the crystal structures
of daunorubicin (Figure 1) and epidoxorubicin cross-
linking DNA,1
0,14
it is not consistent with what we now
observe in the crystal structure of DoxF.

7
650 Journal of Medicinal Chemistry, 2005, Vol. 48, No. 24
Post et al.
Scheme 1. Synthesis of Doxoform (DoxF) and Doxsaliform (DoxSF) from Dox and Their Partial Hydrolysis to
Doxazolidine (Doxaz) and Doxorubicin-Formaldehyde Conjugate (Schiff Base or Aminol), Respectively, the Presumed
Intermediates for Cross-linking DNA
Table 1. Comparison of Growth Inhibition of Breast and
Prostate Cancer Cells by Dox, DoxSF, DoxF, Doxaz, Epi, and
EpiFa
little, from 3.37 to 3.44 Å. The crystal structure shows
an assortment of intramolecular and four nota-
ble intermolecular hydrogen bonds: O6-H6‚‚‚O12. 2.82
Å; O7-H7‚‚‚O17, 2.94 Å; O17-H17‚‚‚O1, 2.84 Å;
O18-H18‚‚‚O6, 3.05 Å. Loss in stability from the twist
boat sugar conformations in the solid state with little
or no anomeric effect is clearly compensated, partially
through π-stacking but probably more importantly
through favorable intermolecular interactions including
the hydrogen bonds noted.
IC50
compd
Dox
MCF-7
MCF-7/Adr
MDA-MB-435
DU-145
240c
b
b
b
200 ( 26 10000 ( 1300
150 ( 14
b,d
b,d,e
b
DoxSF 70-80
800-2000
50 ( 9
f
f
3c
DoxF
2
1
Doxaz 3 ( 0.2
3 ( 0.2 g
7 ( 0.3
4 ( 0.6
380 ( 52
g
c
Epi
EpiF
200
>10000
g
g
c
65
70
26 ( 5
a
Cell Experiments. Inhibition of growth of three
breast and one prostate cancer cell lines by Dox, DoxSF,
DoxF, and Doxaz is compared in Table 1. DoxF and
Doxaz inhibit 50% growth at approximately the same
concentration for each cell line and inhibit growth at 1
to greater than 3 orders of magnitude lower concentra-
tion than Dox. The more dramatic difference occurs with
the multidrug-resistant MCF-7/Adr cells that overex-
press P-170 glycoprotein efflux pump among other
resistance mechanisms.25 The ability of Doxaz to inhibit
the growth of MCF-7/Adr cells as well as DoxF is
illustrated in Figure 3A, which shows cell growth as a
function of drug treatment within a single experiment.
All data points fall on the same growth inhibition curve.
Growth inhibition parallels drug uptake as measured
by flow cytometry measuring drug fluorescence as a
function of time after drug treatment. MCF-7/Adr cells
take up significantly more DoxF and Doxaz than Dox
as shown in Figure 3B. DoxF and Doxaz may be able to
overcome P-170 glycoprotein drug efflux pump because
unlike Dox they are not cations at physiological pH and
Units for IC50 values with Dox, DoxSF, Doxaz, and Epi are
nM, and units with DoxF and EpiF are nM equiv to correct for
DoxF and EpiF having two active compounds per molecule. IC50
values for Doxaz were determined as described in the Experimen-
tal Section. All determinations were done at least in duplicate,
with average data shown above. Errors represent 1 standard
deviation about the mean for the six wells per lane measured for
b
c
d
each drug concentration. Reference 36. Reference 28. Refer-
ence 18. e Reference 38. f Reference 17. g Reference 29.
Crystal Structure of DoxF. X-ray quality crystals
were grown at the interface of a chloroform solution of
DoxF and a mixture of ethyl acetate mixed with hexane.
Crystal data and collection parameters appear in Table
3
, and a crystallographic information file (CIF format)
is provided in Supporting Information. The crystal
structure shows a compact structure with the daun-
osamine sugars in a twist boat conformation and the
anthraquinone rings in a π-stacking arrangement shown
in Figure 2 together with the crystallographic number-
ing system. Two least-squares planes were calcu-
lated for the groups of six carbon atoms C47-C52 and
C19-C24. The two planes are nearly coplanar with a
dihedral angle of 0.81(9)°. The atoms from each plane
are on average 3.402 Å apart; the individual values vary
2
6
consequently are more lipophilic than Dox. Further,
they are fast-acting drugs in that they can rapidly form

Doxazolidine Cross-links DNA
Journal of Medicinal Chemistry, 2005, Vol. 48, No. 24 7651
Table 2. 1H NMR Spectral Assignments for DoxF and Doxaza
structure position number
1
2
3
4
7
8
9
10
14 Ar-OH
1′
2′
3′
4′
5′
2 2
NCH O NCH N
Doxf
7
.85 7.69
7.23
dd
3.90 5.10 1.92
OMe dd dd
(2, 3) (3, 15)
.50
dt
15, 2)
4.86 2.86
4.76 12.92 5.48 1.77
3.40
dt
3.98 4.04
dd dq
4.21
d
3.52
s
dd dd
s
d
s
s
dd
ddd
(
8, 1) (8, 8.5) (1, 8.5)
s
(OH) (19)
3.04
3.04 13.75 (5, 6) (5, 6, 15) (7, 5, 5) (2, 7) (2, 6.5)
bs
OH
(4)
4.73
d
2
s
2.19
dt
dd
(2, 19)
1.36
Me
q
(
(15, 5, 5)
(4)
(
6.5)
Doxaz
4.78 13.30 5.39 1.70
dd ddd
8
.06 7.80
7.41
d
(8)
4.10 5.36 2.16
4.90 3.07
3.52
ddd
3.75 3.96
dd dq
4.31
d
d
t
OMe
s
t
dd
s
d
s
s
(
8)
(8)
(3)
(3, 15)
(OH) (19)
3.30
3.02 13.97 (6, 7) (5, 7,
(4, 5, 8) (2, 8) (2, 6.5)
(6.5)
4.68
d
2
.49
ddd
1.5, 3, 15)
bs
OH
s
s
15)
1.34
q
(6.5)
dd
(1.5, 19)
2.30
ddd
(
(6.5)
(
4, 6, 15)
a
Spectra were taken at 500 MHz with compounds in DCCl3. Chemical shifts are in ppm on the δ scale. Coupling constants are in Hz
and appear in parentheses. See Scheme 1 for the structures and the number system.
Table 3. Crystallographic Data and Collection Parameters
Table 4. Rate Constant for the Hydrolysis of Doxaz to Dox at
4 °C as a Function of pH
1
DoxF
pH
k (min-1)
t1/2 (min)
formula
formula wt
space group
temp (°C)
a (Å)
C57H58N2O22
1123.05
5.0
6.1
0.17 ( 0.01
4
9
16
20
18
P31 (No. 144)
0.074 ( 0.002
0.0424 ( 0.0002
0.035 ( 0.001
0.038 ( 0.002
-119
7.5
15.6681(3)
9.0
10.4
b (Å)
15.6681(3)
c (Å)
17.4559(7)
R (deg)
â (deg)
γ (deg)
Z
90
formaldehyde molecules; however, because of the trans
stereochemistry of the vicinal amino alcohol, the struc-
ture is bicyclic with seven-membered rings (Scheme 2).
It also virtually cross-links DNA, and the crystal
structure of the cross-link is very similar to that formed
with daunorubicin shown in Figure 1.14 Upon hydroly-
sis, EpiF slowly forms a monomeric species with one
90
120
3
3
V (Å )
3711.12(18)
3
dcalc (g/cm )
1.508
θ range (deg)
1.90-27.87
-
1
µ (mm
)
0.117
crystal size
30 0. 43 3m 2m × 0.3 mm × 0.2 mm
reflections collected
data/restraints/params
R1 (for Fo > 4σFo)
R1, wR2 (all data)
GOF
28
formaldehyde attached as an aminol (Scheme 2). An
11780/1/730
0.1071
0.1732, 0.3144
1.025
0.48, -0.43
aminol structure is also proposed for the intermediate
from partial hydrolysis of DoxSF (Scheme 1). On this
basis, our working hypothesis is that an anthracycline-
formaldehyde conjugate that has an oxazolidine ring or
releases a derivative with an oxazolidine ring inhibits
the growth of tumor cells much better than an anthra-
cycline-formaldehyde conjugate that releases a deriva-
tive with the formaldehyde incorporated as an aminol.
Hydrolytic Stability of DoxF and Doxaz. Both
DoxF and Doxaz are relatively stable in dry chloroform
and dry DMSO over a period of days. In dry DMSO,
Doxaz hydrolyzes to Dox at less than 2% per day at
ambient temperature. In aqueous medium, DoxF is very
hydrolytically unstable with respect to formation of
Doxaz, and Doxaz is hydrolytically unstable with respect
to formation of Dox. Injection of DoxF on reverse-phase
HPLC only shows peaks for Doxaz and Dox. Conse-
quently, HPLC was used to monitor the kinetics of
hydrolysis of Doxaz to Dox at 14 °C as a function of pH.
Nonlinear least-squares fitting of the data to a first-
order rate law gave the rate constants reported in Table
4. The rate constant for hydrolysis at pH 7.5 corresponds
to a half-life of 16 min. The hydrolysis was too rapid to
measure the rate at 37 °C. Extrapolation using the
rough rule that the rate will double for every 10 °C
increase in temperature gives an estimation of the half-
life at pH 7.4 and 37 °C of 3-4 min. The rate constant
increases with decreasing pH except for the transition
3
largest peak, hole (e/Å )
virtual cross-links to DNA such that they are not
accessible to the efflux pump. In this respect Figure 3B
may underestimate uptake of Doxaz and DoxF because
DNA quenches the fluorescence of the Dox chromophore
when it cross-links or intercalates DNA.27 Hence, in cells
treated with DoxF or Doxaz, a higher percentage of the
drug is bound to DNA, which quenches the fluorescence.
DoxSF shows intermediate growth inhibition; of note
is the variability in the measurement of the IC50 with
MCF-7/Adr cells over numerous measurements by sev-
eral different co-investigators. Careful scrutiny of the
experiments now indicates that DoxF is a byproduct of
the synthesis of DoxSF and that higher IC50 values were
determined with more highly purified samples of Dox-
SF. Hence, we conclude that incorporation of formalde-
hyde in an acyclic structure as in DoxSF yields a much
less active drug conjugate.
Interestingly, the formaldehyde conjugate of epidoxo-
rubicin (Epi), epidoxoform (EpiF), is an order of mag-
nitude less active at inhibition of tumor cell growth than
is DoxF, even though Epi is only slightly less active than
2
8,29
Dox (Table 1).
Like DoxF, EpiF has a dimeric
structure from reaction of two Epi molecules with three

7
652 Journal of Medicinal Chemistry, 2005, Vol. 48, No. 24
Post et al.
Figure 2. Crystallographic numbering system and three-dimensional structure of DoxF showing twist boat daunosamine sugars
and π-stacking arrangement of anthraquinone rings. The image was created in PyMOL from the crystallographic data reported
here.
Figure 3. (A) Inhibition of growth of multidrug-resistant MCF-7/Adr breast cancer cells with Doxaz and DoxF as a function of
concentration. Cells were treated with drug for 3 h in RPMI 1640 medium containing 10% fetal bovine serum (FBS), and cell
growth in the same medium was measured at 5 days. The concentration of DoxF is in nM equiv to correct for DoxF functioning
as a prodrug for 2 equiv of Doxaz. (B) Relative fluorescence of the Dox chromophore in MCF-7/Adr cells treated with 500 nM
Doxaz, 500 nM equiv of DoxF, or 500 nM Dox as a function of time after inoculation of the cell culture, measured by flow cytometry.
Scheme 2. Synthesis of Epidoxoform (EpiF) from Epidoxrubicin (Epi) and Partial Hydrolysis of EpiF to
Epidoxorubicin Aminol, the Presumed Intermediate for Cross-linking of DNA
from pH 10.4 to pH 9.0 where the rate constant
decreases slightly. At pH 10.5 the solution appears
purple, indicative of deprotonation at one of the hydro-
quinone functional groups. Hence, at this pH, at lease
some of the Doxaz has a different charge state. The
higher rate at low pH is consistent with acid catalysis,
probably assisting hydrolytic ring opening as the slow
step. Which bond of the oxazolidine ring is broken first,
the CH2-O bond or the CH2-NH bond, is unknown but
may be relevant to a small portion of the biological
activity. An intermediate from CH2-NH bond cleavage
would be an unlikely candidate for cross-linking DNA
based on the NMR and crystal structures of the virtual
cross-link, which all show a diaminomethane linkage

Doxazolidine Cross-links DNA
Journal of Medicinal Chemistry, 2005, Vol. 48, No. 24 7653
Figure 4. Inhibition of the growth of MCF-7/Adr multidrug-resistant breast cancer cells as a function of time for hydrolysis of
Doxaz or DoxF at 37 °C. Cells were treated with drug for 3 h, and cell growth in RPMI 1640 medium containing 10% FBS was
measured at 5 days. (A) Starting concentration was 1000 nM equiv of Doxaz or DoxF in RPMI 1640 growth medium. (B) Starting
concentration was 100 nM equiv of DoxF in RPMI 1640 growth medium or in 100% pooled human serum.
(
Figure 1).10,13,14 An intermediate from CH2-O bond
ments all point to DoxF serving as a prodrug to Doxaz
and Doxaz being the reactive intermediate for virtually
cross-linking DNA. Of primary significance is the dra-
matic difference in cell growth inhibition by Doxaz in
comparison with DoxSF. DoxSF releases doxorubicin
conjugated to formaldehyde as a Schiff base or aminol
(Scheme 1). In principle, Schiff base or aminol could
directly cross-link DNA, could cyclize to Doxaz which
could cross-link DNA, or could lose formaldehyde to
yield Dox. Doxaz could react directly with DNA to create
the cross-link or partially hydrolyze to the aminol which
could cross-link DNA or lose formaldehyde. In contrast,
EpiF can only react with DNA as the aminol because of
steric constraints. Further, crystal structures of DNA
cleavage should be the same reactive intermediate
produced by hydrolysis of DoxSF (Scheme 1). With an
acid-catalyzed mechanism, CH2-O bond cleavage would
be favored by protonation on oxygen and CH2-NH bond
cleavage would be favored by protonation on nitrogen.
This analysis of the ring opening reaction of Doxaz
under hydrolytic conditions is relevant to a subsequent
discussion of the direct reaction of Doxaz with DNA.
A functional measure of the rate of hydrolysis of
Doxaz to Dox (and DoxF to Dox) is the effect of
prehydrolysis as a function of time on the growth of
MCF-7/Adr cells. This technique provides kinetic infor-
mation because the concentration of Doxaz at time zero
is 1 µM, and at this concentration the product of
hydrolysis, Dox, has little activity (Table 1). The result
is shown in Figure 4A starting with either Doxaz or
DoxF at 1000 nM equiv at 37 °C in RPMI 1640 media
and in Figure 4B starting with 100 nM equiv of DoxF
in either RPMI 1640 media or in 100% human serum.
The media during cell growth was RPMI 1640 contain-
ing 10% fetal bovine serum in all the experiments in
Figure 4. The data in Figure 4A give an estimate for
the half-lives of Doxaz and DoxF at 37 °C in cell culture
media. This estimate comes from the IC50 value for cell
growth inhibition by Doxaz and DoxF of approximately
virtually cross-linked by daunorubicin and epidoxoru-
bicin show very little structural difference,1
0,14
suggest-
ing that the difference in cytotoxicity arises from some
factor other than the structure of the cross-link. The
data in Table 1 indicate collectively that the oxazolidine
inhibits the growth of cancer cells better than the
aminol. Is the transition state for direct reaction of the
oxazolidine with DNA reasonable? Figure 5 proposes a
possibility both schematically and with a molecular
model. Whether the reactive intermediate is the oxazo-
lidine or the aminol (or Schiff base), the G-base of the
DNA as a HN(2)dC-N(3)H tautomer has the best
location of nonbonding electrons and protons for cross-
linking without major disruption of Watson-Crick
hydrogen bonding. Although formally this tautomeriza-
tion interconverts double and single bonds, X-ray data
2
nM (Table 1) and the requirement of nine hydrolytic
half-lives to reach 2 nM Doxaz starting at 1000 nM
equiv. Note that after 9 half-lives no DoxF will be
present because the rate of hydrolysis of DoxF to Doxaz
is much faster than the rate of hydrolysis of Doxaz to
Dox. Figure 4A shows 50% growth inhibition by Doxaz
after about 8 min for hydrolysis and by DoxF after about
on nucleosides bases place the H N(2)-C(2) bond only
2
0.01 Å longer than the C(2)dN(3) bond, which is within
experimental error.30 In this tautomeric form, bond
breaking and bond forming can occur in concert without
development of formal charge. The model suggests that
for direct reaction of the oxazolidine, the Doxaz must
approach the DNA with the long axis of its anthraquino-
ne approximately parallel to the long axis of the stacked
base pairs at the intercalation site. This orientation
locates the lone pair of the nucleophilic nitrogen of the
N
1
3 min. From these times we estimate the half-life of
Doxaz at 8/9 or approximately 1 min and of DoxF at
1
3/9 or approximately 1.5 min. From the data in Figure
B, we estimate that human serum extends the half-
4
life of DoxF to about 3 min. For this calculation, 6 half-
lives are required to reach 2 nM Doxaz from the starting
concentration of 100 nM equiv of DoxF. Human serum
may double the life of DoxF through hydrophobic
interactions with proteins.
cross-linking G-base properly for S 2 cleavage of the
CH -O bond of the oxazolidine and locates the oxygen
2
Model for Reaction of Doxaz with DNA. Cell
of the oxazolidine near the N-H at the 3-position of the
growth inhibition experiments and hydrolysis experi-
G-base tautomer for proton transfer. After bond forma-

7
654 Journal of Medicinal Chemistry, 2005, Vol. 48, No. 24
Post et al.
Chart 2. Structures for Mitomycin C and FR-900482
Despite the apparent favorability of this tautomer of a
G-base for reaction with drugs in the minor groove, it
has received little or no attention in the literature
probably because detection at low levels is problematic
for crystallography or NMR spectroscopy. Further,
drugs might induce the tautomerization in a prelude to
covalent bonding.
The higher cytoxicity of DoxF and correspondingly
Doxaz might also be explained by DoxF simply serving
as a prodrug for Dox that achieves higher Dox levels in
cancer cells with virtual cross-linking being only an
artifact observed with extracellular DNA. Evidence in
support of DNA virtual cross-linking in cancer cells
treated with DoxF comes from a labeling experiment
reported earlier.33 MCF-7 and MCF-7/Adr cells were
treated with DoxF synthesized with tritiated formalde-
hyde. After 1 h the cells were lysed and separated into
DNA, RNA, and protein fractions, and the fractions
were counted for tritium. The ratio of counts in DNA to
RNA per milligram of macromolecule was greater than
2
0 with counts in protein similar to counts in RNA.
Control experiments with just tritiated formaldehyde
or doxorubicin and tritiated formaldehyde showed a
more even distribution of counts in DNA and RNA with
a ratio of approximately 2. In control experiments,
counts in protein were similar to the experiment with
Figure 5. Schematic and molecular model of the reaction of
Doxaz with 5′-CGCGC-3′/3′-GCGCG-5′ dsDNA to form virtu-
ally cross-linked DNA. The model maintains the integrity of
electrons in the σ- and π-regions of space, and proceeding
through the transition state to product yields no separation
of formal charge. The starting point for the molecular model
was the structure shown in Figure 1. The daunorubicin was
disconnected from the DNA and modified in Chem 3D to create
Doxaz. The Doxaz structure was then partially rotated out of
the DNA from its fully intercalated position and translated to
locate the reactive oxazolidine carbon about 1.6 Å from the
reactive nitrogen of the G-base, also in Chem 3D. The final
picture is displayed in PyMOL.
3
[ H]-DoxF. From this series of experiments, we conclude
that DoxF, and now also Doxaz, delivers formaldehyde
to DNA in cancer cells. Whether the formaldehyde
delivered to cancer cells is actually used with Dox to
virtually cross-link DNA is difficult to prove unambigu-
ously, but circumstantial evidence presented here and
in earlier papers by our group and other groups points
in this direction.
Conclusions
The monooxazolidine derivative of Dox, Doxaz, is
synthesized in good yield by the reaction of doxorubicin
free base with paraformaldehyde in dry chloroform.
Further reaction yields the dimeric compound DoxF,
also in good yield. In the solid state, DoxF has π-stacked
anthraquinones with daunosamine sugars in a twist
boat conformation to achieve a compact structure. In
chloroform solution, both DoxF and Doxaz have daun-
osamine sugars in a relaxed chair conformation. In
aqueous medium, DoxF very rapidly hydrolyzes to
Doxaz, which relatively rapidly hydrolyzes to Dox. DoxF
and Doxaz inhibit the growth of tumor cells approxi-
mately equally and significantly better than DoxSF,
EpiF, Epi, or Dox despite their rapid hydrolysis to Dox.
Of particular significance is the higher activity of Doxaz
relative to DoxSF, which is a prodrug for an acyclic
Dox-formaldehyde conjugate such as the Schiff base or
aminol. Doxaz contains a formaldehyde equivalent in a
more stable oxazolidine ring than does Dox aminol or
Schiff base. The higher cytotoxicity of Doxaz relative to
DoxSF might be explained by the oxazolidine directly
tion, bond rotations will allow the anthraquinone to
intercalate into the DNA to achieve the virtual cross-
link shown in Figure 1.
The proposed mechanism for creation of the covalent
bond between Dox-formaldehyde conjugate and the
2
-amino group of a G-base (Figure 4) involves a prior
tautomerization of the purine at the 2-3 position. This
simple in plane proton shift transforms the nitrogen
substituent at the 2-position into a nucleophile akin to
the nitrogen at the 7-position. The 7-position is the most
common site for alklyation in the major groove by the
nitrogen mustard, diaziridinylquinone, and cis-platin
31
antitumor drugs. Other antitumor drugs active in the
minor groove that form covalent bonds to the amino
group at the 2-position of a G-base include mitomycin
C and antibiotic FR-900482 among related structures
3
1
(
Chart 2). These drugs might also react via the same
G-base tautomer. Mitomycin C forms two covalent bonds
at the 2-amino groups of the G-bases at 5′-CG-3′ sites.32

Doxazolidine Cross-links DNA
Journal of Medicinal Chemistry, 2005, Vol. 48, No. 24 7655
cross-linking DNA without initial ring opening to the
Schiff base or aminol. A model for such a direct reaction
has the 2-amino of a G-base tautomer attack the
oxazolidine ring in an SN2 reaction. Several strategies
for selective delivery of DoxSF to tumor cells show
(Norcross, GA). All cells were maintained at 37 °C in a
2
humidified atmosphere of 5% CO and 95% air.
2
. Synthesis. 2.1. Doxorubicin Oxazolidine, Doxazoli-
dine (Doxaz). Dox hydrochloride (40 mg, 69 µmol) formulated
with lactose (clinical sample) was dissolved in 100 mL of pH
.5 saturated sodium carbonate/sodium bicarbonate buffer.
8
34-38
promise in cell experiments.
Further, two strategies
The aqueous solution was then extracted three times with 250
mL of chloroform. The chloroform extracts were combined,
dried over sodium sulfate, and filtered, and the chloroform was
removed by rotary evaporation yielding doxorubicin as the free
base. Dox free base (30 mg, 55 µmol) was dissolved in 17 mL
of deuteriochloroform (3.7 mM) that had been dried over 4 Å
molecular sieves for at least 24 h, and the solution was
degassed with argon. To this solution, 10 mg of paraformal-
dehyde (30 wt % of doxorubicin) was added, and the solution
was allowed to stir in the dark at ambient temperature (25-
for delivery of formaldehyde in conjunction with an
anthracycline show promise in mouse experiments, one
1
9,20,39
using formaldehyde prodrugs
and the other using
40
EpiF. A drug design strategy that protects Doxaz from
premature hydrolysis to Dox and delivers Doxaz selec-
tively to tumor cells should dramatically exceed the
therapeutic potential of these earlier strategies.
1
Experimental Section
28 °C). Progress of the reaction was followed by H NMR, and
additional paraformaldehyde (10 mg) was added at 2 and 4
days if further progress was not observed. After 7 days, the
reaction was complete as determined by observation of the
appearance of oxazolidine doublets at 4.31 and 4.68 ppm and
shift of the peak for the 5′-methyl from 1.36 to 1.34 ppm. The
reaction mixture was filtered to remove excess paraformalde-
hyde, and solvent was removed by rotary evaporation to
dryness followed by evacuation (∼0.05 Torr) for 30 min to give
1
. General Remarks. Dox hydrochloride clinical samples
(
formulated with lactose) were received as a gift from FeRx,
Inc. (Aurora, CO). NMR solvents were purchased from Cam-
bridge Isotope Laboratories, Inc. (Andover, MA). All other
chemicals were purchased from Aldrich (Milwaukee, WI).
Analytical HPLC was performed on a Hewlett-Packard 1090
chromatograph equipped with a diode array UV-vis detector
interfaced to an Agilent ChemStation data system (Palo Alto,
CA). Analytical HPLC injections were onto an Agilent Zorbax
22 mg of Doxaz (40 µmol, 73% from doxorubicin free base)
isolated as a red film. Product was characterized and analyzed
5
µm reverse-phase octadecylsilyl (ODS) microbore column, 4.6
1
for purity by 500 MHz H NMR in chloroform-d (>90% pure,
mm i.d. × 150 mm, eluting at 1.0 mL/min, and the eluent was
monitored at 280 and 480 nm. Analytical separation was
achieved using method 1 parameters: flow rate, 1.0 mL/min;
eluent A ) HPLC grade acetonitrile and eluent B ) 20 mM
triethylammonium acetate, pH 7.4; gradient, 25:75 A/B at 0
min to 70:30 A/B at 10 min, isocratic to 11 min, back to 25:75
A/B at 13 min. HPLC method 2 parameters were used for
monitoring hydrolysis of drugs: flow rate, 1.0 mL/min; eluent
A ) HPLC grade acetonitrile and eluent B ) 20 mM triethy-
lammonium acetate, pH 7.4; gradient, 25:75 A/B at 0 min to
Table 2). Positive ion electrospray mass spectrometry of a
solution in THF showed a doubly charged ion at m/z 278.8 ((M
+
+
2H )/2, 100% relative intensity, calcd 278.8). HPLC using
method 1 shows a peak for Doxaz at 6.6 min (Dox elutes at
.7 min). HPLC was not reliable for product purity because of
4
some hydrolysis to Dox during elution. The NMR spectrum is
provided in Supporting Information.
2.2. Doxoform (DoxF). Dox free base (35 mg 64 µmol) was
dissolved in 17 mL of dry chloroform-d (3.7 mM), and the
solution was degassed with argon. To this solution, 70 mg of
paraformaldehyde (200 wt % of doxorubicin) was added, and
the solution was allowed to stir in the dark at ambient
temperature (25-28 °C). Progress of the reaction was followed
5
8
6:44 A/B at 7 min, isocratic to 7.5 min, back to 25:75 A/B at
.5 min, isocratic to 9 min. H NMR spectra were acquired
1
with a Varian Unity INOVA 500 MHz spectrometer (Palo Alto,
CA). Electrospray mass spectra were measured with a Perkin-
Elmer Sciex API III instrument (Norwalk, CT), equipped with
an ion-spray source, at atmospheric pressure. UV-vis spec-
trometry was performed with a Hewlett-Packard Agilent
1
by H NMR, which showed the reaction to proceed through
the Doxaz intermediate. Additional paraformaldehyde (70 mg)
was added after 48 h. After 5 days, reaction was 90% complete
as determined by observation of the shift of the peak for the
methoxy group from 4.11 to 3.90 ppm. The reaction mixture
was filtered to remove excess paraformaldehyde, and the
solvent was removed by rotary evaporation to dryness followed
by evacuation (∼0.05 Torr) for 30 min to give 28 mg (56 µmol,
8
452A diode array spectrophotometer interfaced to an Agilent
ChemStation data system (Palo Alto, CA). The 96-well cell
culture plates were read using a PowerWave X plate reader
from BIO-TEK Instruments Inc. (Winooski, VT) using their
Keniticalc software. Flow cytometry was performed on a
Becton Dickinson Biosciences FACScan (San Jose, CA) flow
cytometer using Becton Dickinson Biosciences CellQuest
Software. All tissue culture materials were obtained from
Gibco Life Technologies (Grand Island, NY) unless otherwise
noted. MCF-7 human breast adenocarcinoma cells were ob-
tained from American Type Culture Collection (Rockville, MD).
MCF-7/Adr, a doxorubicin-resistant subline, was a gift from
Dr. William W. Wells (Michigan State University; East
Lansing, MI). MDA-MB-435 human breast adenocarcinoma
cells were a gift from Drs. Renata Pasqualini and Janet Price
7
9%) of DoxF as a red film. Product was analyzed for purity
1
by 500 MHz H NMR in chloroform-d (>90% pure, Table 2).
The spectrum is provided in Supporting Information. DoxF is
not stable to HPLC, and injections of DoxF only yield peaks
for Doxaz and Dox.
3. Crystal Structure of DoxF. 3.1. Growth of DoxF
Single Crystals. To synthesize DoxF for growing crystals, 25
mg of Dox hydrochloride (43 mmol) was dissolved in 30 mL of
triethylammonium acetate (20 mM)/acetic acid buffer at pH
6.0. Formaldehyde (formalin, 34.9 mL of 37% formaldehyde,
10 M equiv, in aqueous methanol) was added, and the mixture
was stirred for 30 min at ambient temperature. Chloroform
(100 mL) was added, and the mixture was stirred overnight.
The chloroform was then removed with a separatory funnel,
and 100 mL of fresh chloroform was added to the aqueous
layer. This was allowed to stir until all the color had trans-
ferred from the aqueous buffer to the chloroform. The chloro-
form fractions were combined and dried over sodium sulfate,
and the solvent was removed by rotary evaporation. The
product was redissolved in a small amount of chloroform and
again subjected to rotary evaporation. Three washes with 50
mL aliquots of water were done by swirling the water inside
the flask for 1 min. Residual water was removed by evacuation
at high vacuum (0.5 Torr). The washed product was dissolved
once more in chloroform and collected by removing the solvent.
(
MD Anderson Cancer Center, Houston, TX). DU-145, human,
metastatic, prostate, and adenocarcinoma cells were provided
by Dr. Andrew Kraft (University of Colorado Health Sciences
Center, Denver, CO). MCF-7 and MCF-7/Adr cells were
maintained in vitro by serial culture in RPMI 1640 medium
supplemented with 10% fetal bovine serum, L-glutamine (2
mM), HEPES buffer (10 mM), penicillin (100 units/mL), and
streptomycin (100 µg/mL). DU-145 cells were maintained in
vitro by serial culture in DMEM media with 10% fetal bovine
serum. MDA-MB-435 cells were maintained in vitro by special
culture in DMEM medium supplemented with 5% fetal bovine
serum, L-glutamine (2 mM), sodium pyruvate (1 mM), nones-
sential amino acids and vitamins for minimum essential
media, penicillin (100 units/mL), and streptomycin (100 µg/
mL). Pooled human serum was obtained from Serologicals

7
656 Journal of Medicinal Chemistry, 2005, Vol. 48, No. 24
Post et al.
This product was collected into a clean, dry Eppendorf tube
concentration was then corrected by measuring the solution
1
-1
-1
and determined to be sufficiently pure by H NMR. Crystal-
absorbance at 480 nm (ꢀ ) 11 500 M cm ). Serial dilutions
(1:3 and 1:10) were made in sterile DMSO to yield seven
solutions of decreasing drug concentration at 100× the respec-
tive working concentrations. The resulting solutions were
individually diluted 1:10 in RPMI 1640 medium; 10 µL of the
resulting 10× solution was immediately added to the ap-
propriate lane of cells. Additionally, two lanes were treated
with 10 µL of growth medium containing 10% sterile DMSO
and one lane was treated with 90 µL of 1.5 M Tris-HCl. The
cells were incubated at 37 °C for 3 h, at which time the drug
solutions were replaced with 200 µL of fresh growth medium.
The cells were then incubated for 5 days, and the extent of
colony formation was determined using a crystal violet staining
lization of the DoxF was accomplished by placing 10 mg of
the crude material inside a 4 mm i.d. glass tube and dissolving
it in 0.5 mL of chloroform. This was gently overlaid with 2
mL of 3:1 ethyl acetate/hexane, and the glass tube was sealed
to prevent solvent loss. Orange-red block-shaped crystals were
formed at the interface between the two solvent systems and
were harvested after approximately 1 week.
3
.2. X-ray Crystal Structure of DoxF. Crystals of DoxF
were suspended in Paratone-N hydrocarbon oil. A single
crystal was cut from a cluster, and it was mounted on the tip
of a glass fiber and transferred to a Siemens SMART diffrac-
tometer/CCD area detector fitted with a low-temperature
nitrogen-flow device. The crystal was centered in the beam
1
7
assay measuring optical density at 588 and 770 nm.
(Mo KR; λ ) 0.710 73 Å; graphite monochromator). A prelimi-
Drug uptake was measured by flow cytometery as follows.
Cells were dissociated with trypsin/EDTA, counted, and
nary orientation matrix and unit cell constants were deter-
mined by the collection of 60 10-s frames, followed by spot
integration and least-squares refinement. A sphere of data
were collected at -119 °C using 0.3° ω scans. The raw data
were integrated and the unit cell parameters refined using
SAINT. Data analysis was performed using XPREP. Absorp-
tion correction was applied using SADABS. The data were
corrected for Lorentz and polarization effects, but no correction
for crystal decay was applied. Structure solutions and refine-
5
suspended in growth media to 1 × 10
cells/mL. This cell
suspension was dispensed in 2.5 mL aliquots into 6-well tissue
culture plates. Plates were then incubated for 24 h at 37 °C
in a humidified atmosphere of 5% CO and 95% air. The
2
growth media was replaced with 2 mL of fresh growth media
prior to addition of the drug. Doxaz, DoxF, and Dox HCl were
dissolved in DMSO each at 50 µM. The concentration was then
corrected by measuring the solution absorbance at 480 nm (ꢀ
2
ments were performed (SHELXTL-Plus V5.0) on F . Table 3
-1
-1
)
11 500 M cm ). Drug solution (20 µL) was added to an
individual well and was incubated for 5 min, 15 min, 30 min,
h, 2 h, or 3 h at 37 °C in a humidified atmosphere of 5% CO
lists a summary of crystal data and collection parameters for
DoxF. A crystallographic information file (CIF format) appears
in the Supporting Information.
1
2
and 95% air. All drug treatments were performed, so all
treatment times would end all at once. After treatment, culture
media was removed, cells were washed once with 0.5 mL of
HBSS and washed once with 0.5 mL of trypsin/EDTA, 0.5 mL
of trypsin/EDTA was added, and cells were incubated for 5
min. After all cells were trypsinized, cells were aspirated with
1.5 mL of cold D-PBS (Dulbecco’s phosphate buffered saline,
no calcium or magnesium), and this solution was added to 3
mL of cold D-PBS (4.5 mL total D-PBS) in a conical vial. Cells
were centrifuged at 200g for 5 min, and D-PBS was decanted
off. Cells were washed once more in 5 mL of cold D-PBS and
centrifuged at 200g for 5 min, and D-PBS was decanted off.
The cells were then aspirated with 1 mL of D-PBS, placed in
a sample tube, and kept on ice until needed with FACScan.
Cells were analyzed with excitation at 488 nm (15 mW Ar ion
laser), with emission monitored between 570 and 600 nm.
Instrument settings were optimized for the cell line and held
constant for all experiments. For the anthracycline fluores-
cence analysis, 10 000 cells were analyzed for each sample.
The data are presented as the mean fluorescence for each
condition.
Preliminary data indicated a primitive trigonal cell. Sys-
tematic absences and intensity statistics suggested space
groups P3
1
(No. 144) and P3
2
(No. 145). Acceptable solution
was achieved with stere-
and refinement in space group P3
1
ochemistry consistent with that known for DoxF. Since DoxF
is a weak anomalous scatterer, the Flack parameter was poorly
defined, but it is near zero (-0.24(18)). The inverted structure
1
gave unsatisfactory refinement in P3 and a Flack parameter
approaching unity. All non-H atoms in the model were refined
anisotropically. Hydrogens were placed in idealized positions
and were included in structure factor calculations but were
not refined.
4
. DoxF and Doxaz Stability Studies. 4.1. Doxaz
Stability in DMSO. Doxaz (22 mg, 40 µmol) was dissolve in
mL of DMSO-d (stored over activated 4 Å molecular sieves)
1
6
1
and analyzed for purity by 500 MHz H NMR. After a 10×
dilution with DMSO to 4 mM, the stability of Doxaz was then
follow by HPLC using method 1, observing the relative peak
area of the Doxaz peak at 6.6 min with constant injection
volume. Doxaz hydrolyzed at a rate of <2% per day in DMSO.
4
.2. Doxaz Stability in Buffers. HPLC buffer, pH 7.4, 20
Hydrolysis of Doxaz or DoxF in RPMI medium was per-
formed as follows. Cells were dissociated with trypsin/EDTA,
mM TEAA, was used for the measurement of hydrolysis of
Doxaz. The buffer’s pH was adjusted as needed with either
triethylamine or acetic acid to pH 10.4, 9.0, 6.0, and 5.0 with
one sample left unchanged at pH 7.4. A volume of 1950 µL of
each of these solutions was added to a conical vial and then
cooled to 14 °C. An amount of 50 µL of a 4 mM solution of
Doxaz was added to each tube, and hydrolysis was followed
by HPLC using method 2 with injections every 10 min. The
pH of the buffers was measure before the addition of drug and
3
counted, and suspended in growth media to 5 × 10 cells/mL.
This cell suspension was dispensed in 200 µL aliquots (1000
cells/well) into the inside wells of 96-well tissue culture plates.
Plates were then incubated for 24 h at 37 °C in a humidified
atmosphere of 5% CO and 95% air. The medium was replaced
2
with 180 µL of growth medium prior to addition of the drug.
RPMI media (no serum added) was divided into 5 mL aliquots
in conical vials and heated to 37 °C in a constant-temperature
bath. Doxaz and DoxF were dissolved in DMSO at 1 mM equiv.
The concentration was then corrected by measuring the
1
.5 h after the start of hydrolysis. The data were fit to a first-
order rate law using regression software from Blackwell
Scientific Software (Oxford, U.K.); correction for hydrolysis on
the HPLC column was incorporated into the preexponential
term of the rate law.
-
1
-1
solution absorbance at 480 nm (ꢀ ) 11 500 M cm ). Drug
solution (50 µL) was added to an individual conical vial, and
the drug was allowed to hydrolyze (at 10 µM) for 0, 10, 20, 30,
45, 60, 75, or 90 min. Experiments were performed such that
all hydrolysis times would end simultaneously. After hydroly-
sis, 20 µL of the drug solution was added to a lane of wells on
the 96-well plate to give a final concentration of 1 µM equiv
(sum of Dox and Doxaz and/or DoxF) for treatment of cells.
Additionally, two lanes were treated with 20 µL of growth
medium containing 1% sterile DMSO for a control. The cells
were incubated at 37 °C for 3 h, at which time the drug
solutions were replaced with 200 µL of fresh growth media.
The cells were then incubated for 5 days, and the extent of
5
. Cell Experiments. IC50 measurements were performed
as follows. Cells were dissociated with trypsin/EDTA, counted,
3
and suspended in growth media to 5 × 10 cells/mL. This cell
suspension was dispensed in 200 µL aliquots (1000 cells/well)
into the inside wells of 96-well tissue culture plates. Outside
wells contained 200 µL of media. Plates were then incubated
2
for 24 h at 37 °C in a humidified atmosphere of 5% CO and
9
5% air. The medium was replaced with 90 µL of growth
medium prior to addition of the drug. Doxaz was dissolved in
DMSO at concentrations ranging from 50 µM to 1 mM. The

Doxazolidine Cross-links DNA
Journal of Medicinal Chemistry, 2005, Vol. 48, No. 24 7657
colony formation was determined using a crystal violet staining
assay measuring optical density at 588 and 770 nm.
Hydrolysis of DoxF in RPMI medium or 100% human serum
was performed the same as above except drug hydrolysis was
performed at 1 µM equiv and drug treatment was at 100 nM
equiv. During drug treatment, the medium was either 90%
RPMI/10% FBS or 90% RPMI/10% human serum. During cell
growth, the medium was 90% RPMI/10% FBS.
(18) Cogan, P. S.; Fowler, C. R.; Post, G. C.; Koch, T. H. Doxsali-
form: A novel N-Mannich base prodrug of a doxorubicin
formaldehyde conjugate. Lett. Drug Des. Dis. 2004, 1, 247-255.
19) Swift, L. P. C., S. M.; Rephaeli, A.; Nudelman, A.; Phillips, D.
R. Activation of adriamycin by the pH-dependent formaldehyde-
releasing prodrug hexamethylenetetramine. Mol. Cancer Ther.
2003, 2, 189-198.
(20) Nudelman, A.; Levovich, I.; Cutts, S. M.; Phillips, D. R. The role
of intracellularly released formaldehyde and butyric acid in the
anticancer activity of acyloxyalkyl esters. J. Med. Chem. 2005,
(
4
8, 1042-1054.
Acknowledgment. The authors thank the U.S.
Army Prostate Cancer Research Program (Grant No.
DAMD17-01-1-0046) and the National Cancer Institute
of the NIH (Grant No. CA-92107) for financial support,
the National Science Foundation for help with the
purchase of NMR equipment (Grant No. CHE-0131003),
Dr. William Wells for MCF-7/Adr cells, Drs. Renata
Pasqualini and Janet Price for MDA-MB-435 cells, and
Dr. Andrew Kraft for DU-145 cells.
(
21) Cutts, S. M.; Nudelman, A.; Rephaeli, A.; Phillips, D. R. The
power and potential of doxorubicin-DNA adducts. IUBMB Life
2005, 5, 73-81.
(
22) Loudon, G. M.; Almond, M. R.; Jacob, J. N. Mechanism of
hydrolysis of N-(1-aminoalkyl) amides. J. Am. Chem. Soc. 1981,
1
03, 4508-4515.
(23) Wani, M. C.; Taylor, D. J.; Wall, M. E.; McPhail, A. T.; Onan,
K. D. Antitumor agents. XIII. Isolation and absolute configura-
tion of Carminomycin I from Streptosporangium sp. J. Am.
Chem. Soc. 1975, 97, 5955-5957.
(24) Pettit, G. R.; Einck, J. J.; Herald, C. L.; Ode, R. H.; Von Dreele,
R. B.; et al. The structure of carminomycin I. J. Am. Chem. Soc.
1975, 97, 7387-7389.
Supporting Information Available: Crystallographic
(
25) Mimnaugh, E. G.; Fairchild, C. R.; Fruehauf, J. P.; Sinha, B. K.
Biochemical and pharmacological characterization of MCF-7
drug-sensitive and Adr multidrug-resistant human breast tumor
xenografts in athymic mice. Biochem. Pharmacol. 1991, 42, 391-
information file for DoxF (in CIF format), HPLC data of Doxaz
1
monitored at 280 and 480 nm, and H NMR spectra for Doxaz
and DoxF establishing the state of purity. This material is
available free of charge via the Internet at http://pubs.acs.org.
4
02.
(
(
26) Lampidis, T. J.; Kolonias, D.; Podona, T.; Isreal, M.; Safa, A. R.;
et al. Circumvention of P-GP MDR as a function of anthracycline
lipophilicity and charge. Biochemistry 1997, 36, 2679-2685.
27) Crooke, S. T.; DuVernay, V. H. Fluorescence Quenching of
Anthracyclines by Subcellular Fractions. Anthracyclines Current
Status and New Developments; Academic Press: New York,
References
(
1) DeVita, V. T.; Hellman, S.; Rosenberg, S. A. Cancer: Principles
and Practice of Oncology, 6th ed.; Lippincott, Williams and
Wilkins: Philadelphia, PA, 2001.
(
2) Gewirtz, D. A. A critical evaluation of the mechanisms of action
proposed for the antitumor effects of the anthracycline antibiot-
ics adriamycin and daunorubicin. Biochem. Pharmacol. 1999,
1
980; pp 151-155.
(
(
28) Taatjes, D. J.; Fenick, D. J.; Koch, T. H. Epidoxoform:
a
hydrolytically more stable anthracycline-formaldehyde conju-
gate, cytotoxic to resistant tumor cells. J. Med. Chem. 1998, 41,
57, 727-741.
(
3) Minotti, G.; Menna, P.; Salvatorelli, E.; Cairo, G.; Gianni, L.
Anthracyclines: molecular advances and pharmacologic develop-
ments in antitumor activity and cardiotoxicity. Pharmacol. Rev.
1
306-1314.
29) Taatjes, D. J.; Koch, T. H. Growth inhibition, nuclear uptake,
and retention of anthracycline-formaldehyde conjugates in
prostate cancer cells relative to clinical anthracyclines. Anti-
cancer Res. 1999, 19, 1201-1208.
2
004, 56, 185-229.
(
(
(
4) Thigpen, J. T. Innovations in anthracycline therapy: overview.
Community Oncol. 2005, 2 (S1), 3-7.
5) Arcamone, F. Doxorubicin Anticancer Antibiotics; Academic
Press: New York, 1981; pp 246-254.
(30) Taylor, R.; Kennard, O. The molecular structures of nucleosides
and nucleotides. 1. The influence of protonation on the geom-
etries of nucleic constituents. J. Mol. Struct. 1982, 78, 1-28.
(31) Rajski, S. R.; Williams, R. M. DNA interstrand cross-linking
agents as antitumor drugs. Chem. Rev. 1998, 98, 2723-2796.
(32) Norman, D.; Live, D.; Sastry, M.; Lipman, R.; Hingerty, B. E.;
et al. NMR and computational characterization of mitomycin
cross-linked to adjacent deoxyguanosines in the minor groove
of the d(TACGTA)-d(TACGTA) duplex. Biochemistry 1990, 29,
2861-2875.
6) Chaires, J. B.; Satyanarayana, S.; Suh, D.; Fokt, I.; Przewloka,
T.; et al. Parsing the free energy of anthracycline antibiotic
binding to DNA. Biochemistry 1996, 35, 2047-2053.
7) Cullinane, C.; van Rosmalen, A.; Phillips, D. R. Does adriamycin
induce interstrand cross-links in DNA? Biochemistry 1994, 33,
(
(
(
4632-4638.
8) Cullinane, C.; Cutts, S. M.; van Rosmalen, A.; Phillips, D. R.
Formation of adriamycin-DNA adducts in vitro. Nucleic Acids
Res. 1994, 22, 2296-2303.
(33) Taatjes, D. J.; Fenick, D. J.; Koch, T. H. Nuclear targeting and
nuclear retention of anthracycline-formaldehyde conjugates
implicates DNA covalent binding in the cytotoxic mechanism of
anthracyclines. Chem. Res. Toxicol. 1999, 12, 588-596.
9) Skladanowski, A.; Konopa, J. Interstrand DNA crosslinking
induced by anthracyclines in tumour cells. Biochem. Pharmacol.
1
994, 47, 2269-2278.
(
(
(
10) Wang, A. H. J.; Gao, Y. G.; Liaw, Y. C.; Li, Y. K. Formaldehyde
cross-links daunorubicin and DNA efficiently: HPLC and X-ray
diffraction studies. Biochemistry 1991, 30, 3812-3815.
(
34) Cogan, P. S.; Koch, T. H. Rational design and synthesis of
androgen receptor targeted non-steroidal anti-androgen ligands
for the tumor specific delivery of a doxorubicin-formaldehyde
conjugate. J. Med. Chem. 2003, 46, 5258-5270.
35) Cogan, P. S.; Koch, T. H. Studies of targeting and intracellular
trafficking of an anti-androgen-doxorubicin-formaldehyde con-
jugate in PC-3 prostate cancer cells bearing androgen receptor-
GFP chimera. J. Med. Chem., submitted.
11) Leng, F.; Savkur, R.; Fokt, I.; Przewloka, T.; Priebe, W.; et al.
Base specific and regiospecific chemical cross-linking of dauno-
rubicin to DNA. J. Am. Chem. Soc. 1996, 118, 4731-4738.
12) Taatjes, D. J.; Gaudiano, G.; Resing, K.; Koch, T. H. A redox
pathway leading to the alkylation of DNA by the anthracycline,
anti-tumor drugs, adriamycin and daunomycin. J. Med. Chem.
(
(
36) Burke, P. J.; Koch, T. H. Design, synthesis, and biological
evaluation of doxorubicin-formaldehyde conjugates targeted to
breast cancer cells. J. Med. Chem. 2004, 47, 1193-1206.
37) Burke, P. J.; Kalet, B. T.; Koch, T. H. Antiestrogen binding site
1
997, 40, 1276-1286.
(
13) Zeman, S. M.; Phillips, D. R.; Crothers, D. M. Characterization
of covalent adriamycin-DNA adducts. Proc. Natl. Acad. Sci.
U.S.A. 1998, 95, 11561-11565.
(
(
AEBS) and estrogen receptor (ER) mediate uptake and distri-
(
14) Podell, E. R.; Harrington, D. J.; Taatjes, D. J.; Koch, T. H. Crystal
structure of epidoxorubicin-formaldehyde virtual crosslink of
DNA and evidence for its formation in human breast-cancer
cells. Acta Crystallogr. 1999, D55, 1516-1523.
bution of 4-hydroxytamoxifen-targeted doxorubicin-formalde-
hyde conjugate in breast cancer cells. J. Med. Chem., submitted.
38) Burkhart, D. J.; Kalet, B. T.; Koch, T. H. Doxorubicin-
formaldehyde conjugates targeting alpha-v beta-3 integrin. Mol.
Cancer Ther. 2004, 3, 1593-1604.
(
(
15) Taatjes, D. J.; Gaudiano, G.; Koch, T. H. Production of formal-
dehyde and DNA-adriamycin or -daunomycin adducts, initi-
ated through redox chemistry of DTT/iron, xanthine oxidase/
NADH/iron, or glutathione/iron. Chem. Res. Toxicol. 1997, 10,
(39) Cutts, S. M.; Rephaeli, A.; Nudelman, A.; Hmelnitsky, I.; Phillips,
D. R. Molecular basis for the synergistic interaction of adria-
mycin with the formaldehyde-releasing prodrug pivaloyloxym-
ethyl butyrate (AN-9). Cancer Res. 2001, 61, 8194-8202.
(40) Dernell, W. S.; Powers, B. E.; Taatjes, D. J.; Cogan, P.; Gaudiano,
G.; et al. Evaluation of the epidoxorubicin-formaldehyde con-
jugate, epidoxoform, in a mouse mammary carcinoma model.
Cancer Invest. 2002, 20, 712-723.
953-961.
(
16) Kato, S.; Burke, P. J.; Fenick, D. J.; Taatjes, D. J.; Bierbaum,
V. M.; et al. Mass spectrometric measurement of formaldehyde
generated in breast cancer cells upon treatment with anthra-
cycline antitumor drugs. Chem. Res. Toxicol. 2000, 13, 509-516.
17) Fenick, D. J.; Taatjes, D. J.; Koch, T. H. Doxoform and dauno-
form: Anthracycline-formaldehyde conjugates toxic to resistant
tumor cells. J. Med. Chem. 1997, 40, 2452-2461.
(
JM050678V