Chemical stability and fate of the cytostatic drug ifosfamide and its N- dechloroethylated metabolites in acidic aqueous solutions

Article abstract of DOI:10.1021/jm980587g

³¹P NMR spectroscopy was used to study the products of the decomposition of the antitumor drug ifosfamide (IF, 1d) and its N- dechloroethylated metabolites, namely, 2,3-didechloroethylIF (1a) and 2- (1b) and 3-dechloroethylIF (1c), in buffered solutions at acidic pH. The first stage of acid hydrolysis of these four oxazaphosphorines is a P-N bond cleavage of the six-membered ring leading to the phosphoramidic acid monoesters (2a-d) of type R'HN(CH₂)₃OP(O)(OH)-NHR, with R and/or R' = H or (CH₂)₂Cl. The electron-withdrawing chloroethyl group at the endocyclic and/or exocyclic nitrogens counteracts the endocyclic P-N bond hydrolysis. This effect is even more marked when the N-chloroethyl group is in the exocyclic position since the order of stability is 1d > 1c > 1b > 1a. In the second stage of hydrolysis, the remaining P-N bond is cleaved together with an intramolecular attack at the phosphorus atom by the non-P-linked nitrogen of the compounds 2a-d. This leads to the formation of a 2- hydroxyoxazaphosphorine ring with R = H (3a coming from compounds 2a,c) or (CH₂)₂Cl (3b coming from compounds 2b,d) and to the release of ammonia or chloroethylamine. The third step is the P-N ring opening of the oxazaphosphorines 3a,b leading to the phosphoric acid monoesters, H₂N(CH₂)₃OP(O)(OH)₂ (4a) and Cl(CH₂)₂HN(CH₂)₃OP(O)(OH)₂ (4b-1), respectively. For the latter compound, the chloroethyl group is partially (at pH 5.5) or totally (at pH 7.0) cyclized into aziridine (4b-2), which is then progressively hydrolyzed into an N-hydroxyethyl group (4b-3). Compounds 3a,b are transient intermediates, which in strongly acidic medium are not observed with ³¹P NMR. In this case, cleavage of the P-N bond of the type 2 phosphoramidic acid monoesters leads directly to the type 4 phosphoric acid monoesters. The phosphate anion, derived from P-O bond cleavage of these latter compounds, is only observed at low levels after a long period of hydrolysis. Compounds 1a-c and some of their hydrolytic degradation products (4b-1, 4b-2, diphosphoric diester [Cl(CH₂)₂NH(CH₂)₃OP(O)(OH)]₂O (5), and chloroethylamine) did not exhibit, as expected, any antitumor efficacy in vivo against P388 leukemia. ³¹P NMR determination of the N- dechloroethylated metabolites of IF or its structural isomer, cyclophosphamide (CP), and their degradation compounds could provide an indirect and accurate estimation of chloroacetaldehyde amounts formed from CP or IF.

Full text of DOI:10.1021/jm980587g

2542
J . Med. Chem. 1999, 42, 2542-2560
Ch em ica l Sta bility a n d F a te of th e Cytosta tic Dr u g Ifosfa m id e a n d Its
N-Dech lor oeth yla ted Meta bolites in Acid ic Aqu eou s Solu tion s†
Ve´ronique Gilard, Robert Martino,* Myriam Malet-Martino, Ulf Niemeyer,‡ and J o¨rg Pohl‡
Biomedical NMR Group, IMRCP Laboratory, Universite´ Paul Sabatier, 118, route de Narbonne, 31062 Toulouse, France, and
ASTA Medica AG, Weismu¨llerstrasse 45, 60314 Frankfurt am Main, Germany
Received February 5, 1999
31P NMR spectroscopy was used to study the products of the decomposition of the antitumor
drug ifosfamide (IF, 1d ) and its N-dechloroethylated metabolites, namely, 2,3-didechloroethylIF
(1a ) and 2- (1b) and 3-dechloroethylIF (1c), in buffered solutions at acidic pH. The first stage
of acid hydrolysis of these four oxazaphosphorines is a P-N bond cleavage of the six-membered
ring leading to the phosphoramidic acid monoesters (2a -d ) of type R′HN(CH2)3OP(O)(OH)-
NHR, with R and/or R′ ) H or (CH2)2Cl. The electron-withdrawing chloroethyl group at the
endocyclic and/or exocyclic nitrogens counteracts the endocyclic P-N bond hydrolysis. This
effect is even more marked when the N-chloroethyl group is in the exocyclic position since the
order of stability is 1d > 1c > 1b > 1a . In the second stage of hydrolysis, the remaining P-N
bond is cleaved together with an intramolecular attack at the phosphorus atom by the non-
P-linked nitrogen of the compounds 2a -d . This leads to the formation of a 2-hydroxyoxaza-
phosphorine ring with R ) H (3a coming from compounds 2a ,c) or (CH2)2Cl (3b coming from
compounds 2b,d ) and to the release of ammonia or chloroethylamine. The third step is the
P-N ring opening of the oxazaphosphorines 3a ,b leading to the phosphoric acid monoesters,
H2N(CH2)3OP(O)(OH)2 (4a ) and Cl(CH2)2HN(CH2)3OP(O)(OH)2 (4b -1), respectively. For the
latter compound, the chloroethyl group is partially (at pH 5.5) or totally (at pH 7.0) cyclized
into aziridine (4b-2), which is then progressively hydrolyzed into an N-hydroxyethyl group
(4b-3). Compounds 3a ,b are transient intermediates, which in strongly acidic medium are not
observed with 31P NMR. In this case, cleavage of the P-N bond of the type 2 phosphoramidic
acid monoesters leads directly to the type 4 phosphoric acid monoesters. The phosphate anion,
derived from P-O bond cleavage of these latter compounds, is only observed at low levels after
a long period of hydrolysis. Compounds 1a -c and some of their hydrolytic degradation products
(4b-1, 4b-2, diphosphoric diester [Cl(CH2)2NH(CH2)3OP(O)(OH)]2O (5), and chloroethylamine)
did not exhibit, as expected, any antitumor efficacy in vivo against P388 leukemia. 31P NMR
determination of the N-dechloroethylated metabolites of IF or its structural isomer, cyclophos-
phamide (CP), and their degradation compounds could provide an indirect and accurate
estimation of chloroacetaldehyde amounts formed from CP or IF.
In tr od u ction
The oxazaphosphorine ifosfamide (IF, 1d ) (Figure 1)
is an alkylating antitumor agent employed in the
chemotherapy of several tumors, including lung, cervi-
cal, and testicular cancers, Ewing’s sarcoma, neuroblas-
toma, and soft-tissue sarcomas. In common with its
structural isomer, cyclophosphamide (CP), IF is a pro-
drug that is entirely dependent on metabolism for its
activity.1-6 Apart from the activation pathway, which
F igu r e 1. Structures of ifosfamide and its N-dechloroethy-
lated metabolites.
leads among other metabolites to the formation of the
alkylating species isophosphoramide mustard (IPM) and
DDCIF) amounts of chloroacetaldehyde (CAA), a com-
the release of urotoxic and nephrotoxic acrolein, there
pound that may be responsible for the oxazaphospho-
is a competitive deactivation pathway via oxidation of
rine-induced neurotoxicity, urotoxicity, and cardio-
toxicity.11-13
the chloroethyl group.3,5,6 This reaction results in the
loss of one or both of the N-chloroethyl side chains of
IF leading to the formation of 2-dechloroethylifosfamide
(2DCIF, 1b), 3-dechloroethylifosfamide (3DCIF, 1c),3,5-7
or 2,3-didechloroethylifosfamide (DDCIF, 1a )8-10 (Fig-
ure 1) and liberation of equimolar (or 2-fold molar for
Knowledge of the amounts of CAA formed are thus
of clinical interest. Although specific methods for analy-
sis of CAA in blood or plasma have been validated,
formaldehyde must be added to the blood samples at
the bedside to prevent rapid degradation of CAA (half-
life in whole blood ranging from 6 to 30 min14,15). Due
to the rapid in vivo conversion of CAA into chloroacetate
which is extensively metabolized mainly to S-carboxy-
† This work has been presented in part at the XIIth International
Conference on Phosphorus Chemistry (Toulouse, France, J uly 6-10,
1992) and abstracted in Phosphorus, Sulfur, Silicon 1993, 77, 200.
‡ ASTA Medica AG.
10.1021/jm980587g CCC: $18.00 © 1999 American Chemical Society
Published on Web 06/25/1999

Chemical Stability and Fate of Ifosfamide
J ournal of Medicinal Chemistry, 1999, Vol. 42, No. 14 2543
Sch em e 1. Hydrolytic Pathways of Ifosfamide Reported
reaction sequence shown in Scheme 1 which is akin to
the Friedman mechanism for hydrolysis of CP.28 Under
strongly acidic conditions (pD ≈ 0) at 20 °C, they
obtained an equimolecular mixture of B and C (Scheme
1) (as evidenced by proton NMR) resulting from the
cleavage of the P-N bonds. In an unbuffered D2O
solution of IF maintained at 50 °C, Mun˜oz et al.24
observed the disappearance of IF after 21 days and the
formation of two final products, the major one having
the structure D from 1H and 13C NMR spectroscopy.
From hydrolysis of IF at 40 °C and pH 4, 7, or 10, Higley
et al.26 identified, using GC-MS, the methyl esters of
oxazaphosphorine E and phosphoric acid (G) after
derivatization by diazomethane and, after derivatization
by trifluoroacetic anhydride, the trifluoroacetylated
derivatives of chloroethylamine (CEA) (C) and 2-(chlo-
roethyl)-3-hydroxypropylamine (F). They proposed that
IF is hydrolyzed according to the sequential pathway
of Scheme 1. Several degradation products were de-
tected (up to nine after 48-h degradation at 75 °C in
distilled water) by micellar electrokinetic chromatogra-
phy, but only chloride ion was identified.27 In alkaline
solutions (pH 10-12), the degradation product detected
by GC or HPLC was not formally identified,22 but the
aziridine structure H (Scheme 1) was proposed from its
molecular weight of 224.18 This was consistent with the
well-documented cyclization of primary and secondary
â-haloamines into aziridines.29 Up to now, the behavior
of IF toward hydrolysis has not been fully determined.
In an attempt to determine the stability and the
behavior of IF and its dechloroethylated metabolites
2DCIF, 3DCIF, and DDCIF in urine, we investigated
the time course of their hydrolysis in buffered solutions
at neutral and various acidic pH by 31P NMR. The
degradation compounds were isolated and characterized
by NMR and mass spectrometry. The toxicity and
antitumor activity of IF, its dechloroethylated metabo-
lites, and some of their degradation compounds were
also evaluated.
in the Literaturea
a
All the compounds are represented in neutral forms.
methylcysteine and thiodiglycolic acid,5,7 low CAA con-
centrations relative to those of 3DCIF and 2DCIF have
been detected in plasma samples from patients treated
with IF.14,15 Assay of 2DCIF, 3DCIF, and DDCIF in
biofluids of patients would therefore give a better
estimate of amounts of CAA formed. Various methods
have been reported for the determination of the dechlo-
roethylated IF metabolites (up to 1992, they are re-
viewed in ref 16; see also refs 9, 10, 14, and 17).
However, before concluding that the amount of dechlor-
ethyl derivatives of IF in the urine of patients corre-
sponds to the amount of CAA liberated, we wanted to
determine their aqueous stability. If unstable in aque-
ous solution, we also needed to identify their degrada-
tion products.
The chemical stability in aqueous solutions of DDCIF
has not been reported, while a study by HPLC and GC
chromatography failed to detect any degradation prod-
ucts of 2DCIF or 3DCIF.18 On the other hand, numerous
studies have been devoted to the aqueous stability of
their parent compound IF.19-27 The hydrolytic cleavage
of IF was first studied by Zon et al.19 From an unbuf-
fered solution of IF (starting pD ≈ 6.0) refluxed for 12
days, only one compound was isolated in low yield (8%
as determined from data reported in the paper) and its
structure determined from elemental analysis and
proton NMR as that of compound A (Scheme 1). The
formation of this compound may be accounted for by the
Resu lts
For more clarity, the data supported by the experi-
ments reported below are summarized in Scheme 2.
Hyd r olysis of DDCIF . 1. 31P NMR-Der ived Tim e
Cou r ses for Hyd r olytic Br ea k d ow n . DDCIF is not
stable in aqueous solution, and its rate of degradation
was higher at acidic pH than at neutral pH. Indeed, its
half-life was 33 min at pH 5.4 and 1.6 h at pH 6.1 versus
6.8 h at pH 6.8 (Table 1).
Three phosphorylated hydrolysis products were de-
tected with 31P NMR (Figure 2A). The concentration-
time plots, which depend on pH, are shown in Figure
2B,C. Their chemical shifts, which are also a function
of pH in the range 5.5-7 for two of them,30 are listed in
Table 2. On the basis of the formation in time and on
the temporal patterns of the degradation compounds of
DDCIF (Figures 2), compound 2a at 10.58 ppm is the
first intermediate which is transformed into 3a (δ )
6.04-6.31 ppm) and then in turn into compound 4a , the
final hydrolysis product (δ ) 2.58-4.61 ppm). The
second intermediate 3a is much more stable at neutral
pH than at acidic pH: the ratio of concentrations of 3a
and 4a is 7 at pH 6.8 versus 0.6 at pH 6.1 and 0.3 at
pH 5.4 after hydrolysis of DDCIF for ≈2 days.

2544 J ournal of Medicinal Chemistry, 1999, Vol. 42, No. 14
Gilard et al.
Sch em e 2. Hydrolytic Pathways of Ifosfamide and Its Dechloroethylated Metabolites from Our Data
a
Observed at low levels at pH 5.5 after hydrolysis for g7 days (from 2DCIF) or g17 days (from 3DCIF) and at pH 6.8 after hydrolysis
b
for ≈40 days from 2DCIF. Observed from hydrolysis of IF at pH ≈0 and 2.0, from hydrolysis of 2DCIF and IF at pH 3.0 and 5.5, and
from hydrolysis of compounds 2b,d at pH 6.0. c Observed from hydrolysis of 2DCIF and IF at pH 5.5, from hydrolysis of compounds 2b,d
d
at pH 6.0, and from hydrolysis of 2DCIF at pH 6.8. Observed at pH 6.8 for >20 days of hydrolysis (from 2DCIF). All the compounds are
represented in neutral forms.
Ta ble 1. 31P NMR-Derived Kinetic Data for Hydrolysis Reactions of DDCIF, 2DCIF, 3DCIF, and IF as Well as Their First
Intermediates at Various pH and 25 °C (unless otherwise specified)
degradation rate constants and half-lifea
pH
6.8
compd
k1
t1/2 (1)
k2
t1/2 (2)
k3
t1/2 (3)
DDCIF (1a )
DDCIF (1a )
DDCIF (1a )
2DCIF (1b)
2DCIF (1b)
2DCIF (1b)
2b
3DCIF (1c)
3DCIF (1c)
2c
1.7 × 10-3 min-1
6.8 h
1.6 h
33 min
36 days
3.8 days
25 min
6.1
5.4
6.8
5.5
3.0
6.0
5.5
3.0
6.0
3.0
2.0
7.1 × 10-3 min-1
2.4 × 10-4 min-1
2.0 days
7.0 × 10-4 min-1
16.4 h
2.1 × 10-2 min-1
1.3 × 10-5 min-1 (≈20 °C)
1.3 × 10-4 min-1 (≈20 °C)
2.8 × 10-2 min-1
8.6 × 10-4 min-1
3.9 × 10-4 min-1
13.5 h
5.0 × 10-3 min-1
9.3 × 10-4 min-1
2.3 h
1.0 × 10-5 min-1 (≈20 °C)b
49 days
2.8 h
4.1 × 10-3 min-1b
1.2 days
12.4 h
IF (1d )
IF (1d )
IF (1d )
1.1 × 10-4 min-1 (≈20 °C)
2.2 × 10-3 min-1b
0.16 min-1
4.3 days
5.3 h
4.4 min
≈0
6.0
2d
2.7 × 10-3 min-1
4.3 h
7.0 × 10-3 min-1
1.6 h
a
k1, k2, and k3 are the rate constants for the disappearance of compounds 1, 2, and 3, respectively, and their derived half-lives are
b
t
1/2(1), t1/2(2), and t1/2(3). Since the pH of the solution drifted more than 0.2 pH unit (0.25-0.35 unit), this rate constant can only be
viewed as an estimate.
We did not observe any formation of phosphate anion
(Pi) over hydrolysis for 4 days (pH 5.4 and 6.1) or 5 days
(pH 6.8) as verified by spiking with authentic standard.
2. Id en tifica tion of Hyd r olysis P r od u cts. The
three hydrolysis products of DDCIF (2a -4a ) were
separately obtained in aqueous solutions in purities
ranging from 80% to 100% as determined by 31P NMR
by acting on pH conditions and duration of hydrolysis.
Their structures could thus be determined from
their mass spectrometric and NMR characteristics
(Table 3).
of C4 in compounds 2a and 4a is indicative of break-
down of the endocyclic P-N bond in DDCIF. So the first
degradation intermediate of DDCIF, compound 2a with
a molecular mass of 154 Da (18 Da higher than that of
DDCIF (136 Da)), comes from hydrolysis of the endocy-
clic P-N bond of DDCIF and corresponds to the phos-
phoramidic acid monoester shown in Scheme 2 and
Table 3.
Compound 3a , in which the C4 has a C-P coupling
constant, must have a cyclic structure. The molecular
weight of 137 Da of this compound is 1 Da above that
of DDCIF and 17 Da less than that of 2a . This indicates
that the amino group in DDCIF is replaced by a
hydroxyl group. Compound 3a is thus a 2-hydroxyox-
azaphosphorine (Scheme 2 and Table 3).
The carbon and proton signals from DDCIF were
readily attributed from their characteristic chemical
shifts. The three carbon atoms in DDCIF are coupled
to the phosphorus atom, and disappearance of coupling

Chemical Stability and Fate of Ifosfamide
J ournal of Medicinal Chemistry, 1999, Vol. 42, No. 14 2545
F igu r e 2. Stack plot of 31P NMR spectra recorded during DDCIF (1a ) hydrolysis in 0.1 M cacodylate buffer at pH 6.8, 25 °C (A).
31P NMR time course at pH 6.8 (B) and 6.1 (C).
Ta ble 2. 31P NMR Chemical Shifts of Hydrolysis Products of IF and Its N-Dechloroethylated Metabolites
a
31P chemical shifts
b
starting compounds
DDCIF (1a ) 2DCIF (1b) 3DCIF (1c) IF (1d )
17.63
hydrolysis products
pH
2a
2b
2c
2d
3a
3b
4a
4 b-1 4 b-2 4 b-3
Pi
7
10.58
10.58
10.58
6.31
6.24
6.04
4.61
3.46
2.58
6.0
5.5
7
17.68
10.64
10.64
10.64
8.91
4.78
3.66
2.68
4.85 2.70
1.10
6.0c
5.5
3
7.83
7.83
3.76
2.76
1.31
7
15.72
4.61
3.46
2.58
1.31
6.0c
5.5
3.0
6.0c
5.5
3.0
2.0
9.49
9.49
9.37
6.24
6.04
-6.28
1.10
9.54
7.83
3.76
2.76
1.31
1.27
0.53
3.66
2.68
15.75
9.37
8.25
5.68
≈0
a
b
Chemical shifts (δ) are expressed in ppm relative to external 85% H3PO4. Chemical shifts of starting compounds are insensitive to
pH. c Chemical shifts of compounds obtained on hydrolysis of first degradation intermediates of 2DCIF (2b), 3DCIF (2c), or IF (2d ).
The final hydrolysis product, 4a , has a molar mass
of 155 Da, 1 Da above that of 2a . This is indicative of a
substitution of an amino group of 2a by a hydroxyl
group. Compound 4a thus corresponds to the phosphoric
acid monoester (Scheme 2 and Table 3), the structure
of which was confirmed by independent synthesis of
compound.28
pounds than in the cyclic ones. Moreover, a strong
shielding of the 31P chemical shift was observed (≈11-
12 ppm for the oxazaphosphorine derivatives and ≈6-8
ppm for the linear compounds at pH 5.5-7) when the
2-amino group was replaced by a 2-hydroxyl group
(Table 2). These observations helped determine the
structures of the hydrolysis products of 2DCIF, 3DCIF,
and IF.
It can be seen from Table 3 that the 13C chemical
shifts of the cyclic compounds (DDCIF and 3a ) are
similar, as are those of the linear compounds 2a and
4a . On the other hand, the C4 and C6 are more shielded
(3-5 and 5-7 ppm, respectively) in the linear com-
Hyd r olysis of 2DCIF . 1. 31P NMR-Der ived Tim e
Cou r ses for Hyd r olytic Br ea k d ow n . The rate of
hydrolysis of 2DCIF was lower than that of DDCIF but
increased with decrease in pH. The half-life of 2DCIF

2546 J ournal of Medicinal Chemistry, 1999, Vol. 42, No. 14
Gilard et al.
Ta ble 3. Mass Spectrometry and NMR Data of DDCIF (1a ) and Its Phosphorus-Containing Hydrolysis Compounds
a
b
Observed pH readings, not corrected for kinetic isotope effect. Coupling constants determined from homo- or heteronuclear spin-
c 3
decoupling experiments.
J
.
H-P
F igu r e 3. Stack plot of 31P NMR spectra recorded during 2DCIF (1b) hydrolysis in 0.1 M cacodylate buffer at pH 5.5, 25 °C (A).
Corresponding 31P NMR time course (B).
fell from 36 days at pH 6.8, to 3.8 days at pH 5.5, to 25
min at pH 3.0 (Table 1).
detected between the 2nd and 11th days of hydrolysis.
Compound 3b (δ ) 7.83 ppm) was only observed in
small amounts (<3% of all 31P NMR signals) and only
after hydrolysis at pH 5.5. After hydrolysis for 6 h at
pH 5.5, compounds 4b-1 and 4b-2 giving signals at 2.76
and 2.68 ppm, respectively (Figure 3A), were detected,
and their signal intensities in an ≈60/40 ratio increased
with time (Figure 3B). Compound 4b-1 was the sole
detected one at pH 3.0 (δ ) 1.31 ppm) and compound
Six different phosphorylated degradation products
were observed depending on the pH of the medium and
the duration of hydrolysis. Compound 2b giving a signal
at 8.91-10.64 ppm30 (Table 2) is the first intermediate
in the hydrolytic pathway (Figure 3). It was detected
at all values of pH, but at pH 6.8 it never made up more
than 2% of all phosphorylated compounds and was only

Chemical Stability and Fate of Ifosfamide
J ournal of Medicinal Chemistry, 1999, Vol. 42, No. 14 2547
4b-2 the sole one at pH 6.8 (δ ) 4.78 ppm). The signal
from Pi (δ ) 1.10 ppm), identified by spiking with
standard, was detected after 7 days of hydrolysis at pH
5.5, and its intensity increased with time to make up
≈2% of the hydrolysis products of 2DCIF after 7 days
and ≈6% after 28 days of hydrolysis.
ance of signals highly shielded by the strain of the three-
membered ring: in H NMR, two doublets at δ ) 2.02
1
and 1.68 ppm corresponding to the two protons of the
aziridine ring in trans and cis positions with respect to
the third substituent on N331 and in 13C NMR, one
singlet whose intensity corresponds to two carbon
atoms.
The mass spectrum of compound 4b-3 showed that it
did not contain a chlorine atom and that relative to
compound 4b-1, this chlorine had been replaced by a
hydroxyl group (Table 4). The 13C NMR data showing
the presence of two CH2-O groups and the fact that this
compound was always found together with compound
4b-2 were evidence that compound 4b-3 was the phos-
phoric acid monoester resulting from hydrolysis of the
aziridine ring of compound 4b-2 (Scheme 2 and Table
4).
Hyd r olysis of 3DCIF . 1. 31P NMR-Der ived Tim e
Cou r ses for Hyd r olytic Br ea k d ow n . 3DCIF is more
stable than either DDCIF or 2DCIF. We did not observe
any degradation products after 4 days of hydrolysis at
pH 6.8 and only a single product (≈2% degradation)
after 28 days under the same conditions. But as for
DDCIF and 2DCIF, the rate of degradation of 3DCIF
was inversely proportional to pH: half-life of 49 days
at pH 5.5 and 2.8 h at pH 3.0 (Table 1). Since the pH
variations in these experiments respectively reached
0.25 and 0.3 pH unit, these values can only be consid-
ered as an estimate.
Three phosphorylated degradation compounds were
formed (Figure 4A). Compound 2c (δ ) 9.37-9.49
ppm)30 (Table 2) was the first intermediate in the 3DCIF
degradation pathway (Figure 4B), but it was not ob-
served at pH 6.8 and only detected in small amounts at
pH 5.5 (≈2% of all 31P NMR signals) due to the high
stability of 3DCIF at these values of pH. A second
compound whose chemical shift lays between -6.28 and
+6.04 ppm was detected at pH 3.0 and 5.5, but it never
exceeded 3% and 1%, respectively, of all the phospho-
rylated products. The third compound at δ ) 1.31-4.61
ppm was the final hydrolysis product as its concentra-
tion increased with time (Figure 4B). It was the only
degradation product observed at pH 6.8 and by far the
major (g80%) product detected after 17 days of hydroly-
sis at pH 5.5.
At pH 6.8, after hydrolysis for more than 20 days, we
observed another signal at 4.85 ppm (compound 4b-3)
whose intensity increased with time and to a greater
extent than that of compound 4b-2. At this pH, we
observed traces of Pi (δ ) 2.70 ppm) after hydrolysis
for ≈40 days.
2. Id en tifica tion of Hyd r olysis P r od u cts. Com-
pounds 2b, 4b-1, and 4b-2 were separately obtained
from 2DCIF in aqueous solutions at ≈100% purity as
determined by 31P NMR by acting on pH conditions and
duration of hydrolysis. Compound 4b-3 was analyzed
in a 64/20/16 mixture with compounds 4b-2 and starting
2DCIF (1b), respectively. Compound 3b could not be
isolated due to its instability, and its structure was only
established from its 13C NMR features on monitoring
the degradation of compound 2b at pH 6.0. The mass
spectrometry data and the NMR characteristics (13C and
in some cases 1H) of 2DCIF and all its degradation
compounds are listed in Table 4.
The proton and carbon-13 signals of 2DCIF were
attributed from 1H spectra with 31P decoupling and
COSY, HETCOR, and HMBC correlations. The endocy-
2
clic C4 did not exhibit an observable J C-P coupling
3
constant in contrast to the exocyclic C7. The J C8-P
2
coupling constant was higher than the J C7-P one.
The FAB mass spectra of compound 2b showed that
it contained a chlorine atom and had a molar mass 18
units (216, 218 Da) above that of 2DCIF (198, 200 Da).
Together with the disappearance of the C-P couplings
on the two carbons C7 and C8 of the chloroethyl chain,
this indicated that compound 2b was the phosphor-
amidic acid monoester formed by hydrolysis of the
endocyclic P-N bond of 2DCIF (Scheme 2 and Table
4).
The 13C NMR characteristics of compound 3b re-
sembled those of 2DCIF where the two carbons atoms
in the N-chloroethyl chain exhibit coupling with the
phosphorus atom. Compound 3b thus has the same
carbon skeleton as 2DCIF. The 31P chemical shift (Table
2) was close to that of compound 3a indicating that the
immediate neighborhood of the P atom was the same
as that in 3a . Moreover, the high shielding (≈10 ppm)
of its 31P chemical shift with respect to that of 2DCIF
is comparable to the 31P ∆δ between compound 3a and
DDCIF (Table 2). Compound 3b is thus a 2-hydroxyox-
azaphosphorine (Scheme 2 and Table 4).
The molar mass of compound 4b-1 indicated that it
contained a chlorine atom and, in comparison with
compound 2b, an amino group replaced by a hydroxyl
group (Table 4). The protons and carbons were at-
tributed from the COSY and HETCOR correlations,
showing that compound 4b-1 was the phosphoric acid
monoester shown in Scheme 2 and Table 4. Its structure
was confirmed by obtaining this compound from IF
hydrolysis for 24 h in 1 M HCl.
The signal of Pi was observed after 17 days of
hydrolysis at pH 5.5, with an intensity that increased
with time. However, after 31 days of hydrolysis, it only
made up <10% of the degradation products. It was not
detected after 28 days of hydrolysis at pH 6.8.
2. Id en tifica tion of Hyd r olysis P r od u cts. Com-
pounds 2c and 4a were separately obtained from 3DCIF
(1c) at ≈100% purity as determined by 31P NMR by
acting on pH conditions and duration of hydrolysis, and
after purification by HPLC for 2c. Compound 3a was
obtained with a “phosphorus purity” of ≈75% from
aqueous degradation of 2c at pH 8.
The proton and carbon-13 signals of 3DCIF were
1
attributed from the H spectra with phosphorus decou-
pling and the COSY, HETCOR, HETCOR-LR, and
HMBC correlations (Table 5). The 13C (δ and J ) char-
acteristics of the carbon atoms in the oxazaphosphorine
Compound 4b-2 harbored an aziridine ring as evi-
1
2
denced by the loss of the H and 13C signals from the
ring resembled those of DDCIF with a J C4-P coupling
N-chloroethyl chain of compound 4b-1 and the appear-
observable, contrarily to 2DCIF.

2548 J ournal of Medicinal Chemistry, 1999, Vol. 42, No. 14
Gilard et al.

Chemical Stability and Fate of Ifosfamide
J ournal of Medicinal Chemistry, 1999, Vol. 42, No. 14 2549
F igu r e 4. Stack plot of 31P NMR spectra recorded during 3DCIF (1c) hydrolysis in 0.2 M phthalate buffer at pH 3.0, 25 °C (A).
Corresponding 31P NMR time course (B).
The structure of compound 2c was established by
comparison of its 13C NMR and MS data with those of
3DCIF. The 13C characteristics of compound 2c were
close to those of 3DCIF for the carbon atoms in the
directly bonded R-carbon (-1.3 ppm). The CH2Cl group
is thus more shielded than the CH2N group at acidic
pH, the chemical shift of the two groups being the same
at pD 7.7.
3
Hyd r olysis of IF . 1. 31P NMR -Der ived Tim e
Cou r ses for Hyd r olytic Br ea k d ow n . IF is more
stable than its N-dechloroethylated metabolites. At pH
6.8 and 5.5, no degradation product was detected after
hydrolysis for 1 month. At pH 5.5, after 49 days of
hydrolysis, the two degradation products detected in
almost equal amounts at δ ) 2.76 and 2.68 ppm only
made up around 4% of all the phosphorylated com-
pounds. The rate of degradation of IF was inversely
related to the pH of the medium, with a half-life of 4.3
days at pH 3.0, 5.3 h at pH 2.0 (estimated value since
the pH of the solution drifted 0.35 pH unit during the
experiment), and 4.4 min in 1 M HCl (Table 1). At these
pH, hydrolysis gave rise to a degradation intermediate
(compound 2d ) at δ between 9.37 and 5.68 ppm30 (Table
2) whose rate of formation and level were inversely
related to pH: ≈3% after 21 h of hydrolysis at pH 3.0,
≈8% after 2 h of hydrolysis at pH 2.0, and 27% after
≈4 min in 1 M HCl. This intermediate was transformed
into the final hydrolysis product (δ ) 0.53-1.31 ppm)
identical to the one obtained at pH 5.5 and resonating
at 2.76 ppm. We failed to detect Pi after 2-3 days of
hydrolysis at pH ranging from ≈0 to 3.0 or 30-49 days
at pH 6.8 and 5.5.
chloroethyl chain with a J C10-P coupling and to those
2
of compound 2a for the other carbons with no J C4-P
coupling. Moreover, C4 and C6 were shielded by ≈3 and
≈7 ppm, respectively, with respect to those of 3DCIF,
as were those of compound 2a with respect to those of
DDCIF (Tables 5 and 3). This indicated that compound
2c derived from hydrolysis of the endocyclic P-N bond
of 3DCIF like compound 2a from DDCIF. The mass
spectral data of compound 2c confirmed that it con-
tained a chlorine atom ([MH]+ ions at m/z ) 217 and
219 in a 3/1 ratio) and had a molar mass 18 Da more
than that of 3DCIF (Table 5). Compound 2c was thus
the phosphoramidic acid monoester shown in Scheme
2 and Table 5.
The 31P chemical shifts of the second (δ ) -6.28-
+6.04 ppm) and final (δ ) 1.31-4.61 ppm) hydrolysis
products of 3DCIF were identical to those of compounds
3a and 4a resulting from hydrolysis of DDCIF. This was
demonstrated by spiking the hydrolytic mixture of
3DCIF with that of DDCIF. Their 13C NMR and mass
spectrometric characteristics confirmed their structures
as being those of compounds 3a and 4a (no chlorine
atom or N-chloroethyl chain). This indicated that at this
stage of hydrolysis of 3DCIF, CEA had been liberated.
No signal at ≈25 ppm characteristic of free aziridine
was detected. By spiking the samples with authentic
material, signals from CEA were detected in the 13C
NMR spectra of compounds 3a (from hydrolysis of
compound 2c at pH ≈8) and 4a (from hydrolysis of
3DCIF in 1 M HCl) and hydrolytic mixtures of 3DCIF
at pH 5.5 or 3.0 (Table 5). The 13C signals of CEA were
attributed from HETCOR correlations of CEA solutions
in D2O at pD in the range 6-13. The direction of the
titration shift is opposite from 1H and 13C nuclei:
downfield for 1H, upfield for 13C with decreasing pH.
Throughout the pH range studied, the 1H chemical shift
of the CH2Cl group is downfield from that of the CH2N
group. The same finding was observed for the 13C
chemical shifts at basic pH. Protonation of the amino
group causes a larger upfield shift in the resonance of
the â-carbon (-6.9 ppm for the CH2Cl group) than the
2. Id en tifica tion of Hyd r olysis P r od u cts. The
proton and carbon-13 signals of IF were attributed from
similar NMR analyses to those conducted on 3DCIF.
The 13C characteristics of IF resembled those of 3DCIF
for C9 and C10 and those of 2DCIF for the other carbon
atoms, although in contrast to 2DCIF, a C4-P coupling
was observed for IF (Tables 4 and 5).
Pure compound 2d was obtained after Sep-Pak chro-
matography of the hydrolytic mixture after 30 min of
IF hydrolysis in 1 M HCl. The FAB positive mass
spectrometry data showed that compound 2d contained
two chlorine atoms and had a molar mass 18 Da above
that of IF (Table 5), which indicated that, in common
with the other compounds 2a -c, the endocyclic P-N
bond of IF had been hydrolyzed. Compound 2d was thus
the phosphoramidic acid monoester shown in Scheme
2 and Table 5. The 13C NMR data showing the disap-

2550 J ournal of Medicinal Chemistry, 1999, Vol. 42, No. 14
Gilard et al.

Chemical Stability and Fate of Ifosfamide
J ournal of Medicinal Chemistry, 1999, Vol. 42, No. 14 2551
F igu r e 5. Stack plot of 31P NMR spectra recorded during hydrolysis of the phophoramidic acid monoester 2b in cacodylate
buffer at pH 6.0, 25 °C (A). Corresponding 31P NMR time course (B).
pearance of the couplings between P and C4, C7, and
C8 were consistent with this structure.
increased with increase in pH (0% at pH 3.0, ≈40% at
pH 5.5-6.0, and 100% at pH 7.0), a cyclization of the
N-chloroethyl chain of compound 4b-1 into an aziridine
may occur giving rise to compound 4b-2. We therefore
followed by 31P NMR the behavior of compound 4b-1 at
pH 5.5 and 7.0. At pH 5.5, no cyclization of the
N-chloroethyl chain into an aziridine was observed.
Compound 4b-1 was slowly hydrolyzed into Pi (≈20%
after 14 days and ≈40% after 44 days). At pH 7.0, no
transformation of compound 4b-1 into the aziridine 4b-2
was observed during hydrolysis for 3 days. On the other
hand, after 15 days, the aziridine represented ≈30% of
the initial 4b-1 and ≈40% after 23 days. We thus
concluded that compound 4b-2 resulted from hydrolysis
of compound 3b and not from cyclization into an
aziridine of the N-chloroethyl chain of compound 4b-1.
Similarly, compound 4b-1 was also derived from hy-
drolysis of the intermediate 3b and not from opening
of aziridine ring of 4b-2 by chloride ion since, at pH 5.8
(obtained by addition of HCl), 31P and 13C NMR failed
to detect any trace of degradation of compound 4b-2
after 18 h at room temperature.
The compounds at δ ) 0.53-2.76 ppm (pH ≈ 0-5.5)
and 2.68 ppm (pH 5.5) had the same 31P chemical shifts
as compounds 4b-1 and 4b-2, respectively, as demon-
strated by spiking hydrolytic mixtures of IF with
standards. The mass spectra and 13C NMR character-
istics of these two compounds obtained with a “phos-
phorus purity” of 100% from IF hydrolysis for 24 h in 1
M HCl for the first one followed with alkalinization at
pH ≈ 13 for 24 h for the second one confirmed that they
were compounds 4b-1 and 4b-2. This means that
hydrolysis liberated CEA, which was detected in the 13
C
spectrum of a reaction mixture from hydrolysis of IF at
pH ≈ 0.
Hyd r olysis of th e P h osp h or a m id ic Acid Mon o-
ester s 2a -d . The hydrolysis at pH 6 of the first
intermediates 2a -d confirmed the pathways shown in
Scheme 2. Compound 2a , the first intermediate in the
hydrolytic pathway of DDCIF, is relatively stable at pH
6.1, with a half-life of ≈2 days (Table 1). The breakdown
of compound 2a led to the formation of compounds 3a
(δ ) 6.24 ppm) and 4a (δ ) 3.46 ppm) (Table 2).
Compound 3a is an intermediate in the degradation
process, which makes up a maximum of 19% of all the
phosphorylated compounds between ≈34 and ≈46 h of
At pH 6.0, the hydrolysis rate of compound 2c (half-
life 1.2 days) lays between those of compounds 2a (half-
life 2.0 days) and 2b (half-life 0.6 day) (Table 1). The
products obtained were identical to those obtained from
hydrolysis of compound 2a , i.e., compound 3a , an
intermediate which peaked at ≈20% after 23-40 h
hydrolysis, and compound 4a , the final hydrolysis
product.
hydrolysis. It is less stable than compound 2a (t1/2
≈
0.7 day versus 2 days) (Table 1), and its hydrolysis leads
to compound 4a , the terminal degradation product
(Scheme 2).
The hydrolysis rate of compound 2b at pH 6.0 was
faster than that of compound 2a (half-life ≈ 0.6 day
versus 2 days) (Table 1). The products formed were the
same as those obtained during the hydrolysis of 2DCIF
at pH 5.5. Compound 3b appeared as an intermediate
in the degradation process. Its 31P signal reached a
maximum between ≈6 and 9 h of hydrolysis and never
exceeded ≈12% of all the phosphorylated compounds
(Figure 5). Its degradation rate constant was much
higher than its formation rate constant (5.0 × 10-3
min-1 versus 8.6 × 10-4 min-1) (Table 1). The final
products of hydrolysis were compounds 4b-1 and 4b-2,
whose signal intensities increased with time in a fairly
fixed ratio of 60/40 (ratio of their formation rate
constants ) 1.32) (Figure 5B). After 30 h of hydrolysis,
the Pi signal was detected, but it only made up 2% of
all phosphorylated products observed after 4 days of
hydrolysis. Since the amount of 4b-2 relative to 4b-1
At pH 6.0, the hydrolysis rate of compound 2d was
much faster than that of the three other phosphoramidic
acid esters 2a -c (t1/2 4.3 h) (Table 1). The degradation
pathway leads to the formation of a transient compound
at 7.83 ppm (Table 2), never observed during hydrolysis
of IF and which made up a maximum of ≈18% of the
phosphorylated compounds observed between ≈3 and
5 h of hydrolysis (Figure 6). The final hydrolysis
products were compounds 4b-1 and 4b-2, observed in a
60/40 ratio (Figure 6B) (ratio of their rate constants of
formation: 1.33), already obtained on hydrolysis of the
first intermediate (2b) in the 2DCIF hydrolytic pathway.
The transient compound whose 31P chemical shift was
identical to that of compound 3b could not be isolated.
Its 13C NMR characteristics, determined by monitoring
the degradation of compound 2d in the NMR probe,
were identical to those of compound 3b (Tables 4 and
5). Characteristic quasi-molecular ions [MH]+ at m/z

2552 J ournal of Medicinal Chemistry, 1999, Vol. 42, No. 14
Gilard et al.
F igu r e 6. Stack plot of 31P NMR spectra recorded during hydrolysis of the phophoramidic acid monoester 2d in cacodylate
buffer at pH 6.0, 25 °C (A). Corresponding 31P NMR time course (B).
Ta ble 6. Changes in 31P Chemical Shifts Due to Protonation of
Hydrolysis Products of IF and Its Dechloroethylated
Metabolites
range of pH ∆δ shifta
compd
ifosfamide (1d )
studied
(ppm)
pKa
12.1-1.2
≈0b
phosphoramidic acid esters
(2b)
(2c)
(2d )
8.5-0.15
9.1-0.1
9.1-0.1
9.1-1.15
-14.3
-4.2
2.3
1.6
1.6
4.1
-4.4
2-hydroxyoxazaphosphorine (3a )
phosphoric acid esters
(4a )
-14.3
9.1-0
9.9-0
-3.65
-0.8
-3.7
-0.8
5.9
0.8
5.9
0.8
(4b-1)
F igu r e 7. 31P chemical shifts of IF (1d ) (9), phosphoramidic
acid monoesters 2b (b) and 2d (O), and 2-hydroxy-1,3,2-
oxazaphosphorine (3a ) (2) as a function of pH at 25 °C.
a
b
Upfield shifts are negative with decreasing pH. A ∆δ shift
of 0.04 ppm was observed, but it was not correlated with pH.
200, 202 were observed in the FAB+ mass spectra of
degradation mixtures of compound 2d after different
hydrolysis times, along with those at m/z 80, 82 corre-
sponding to the CEA released on transformation of
compound 3b, at m/z 218, 220 corresponding to the final
hydrolysis products 4b-1, and at m/z 182 corresponding
to the other final hydrolysis product 4b-2.
31P Ch em ica l Sh ifts a s a F u n ction of p H a n d p Ka
of Hyd r olysis P r od u cts of IF a n d Its Dech lor o-
eth yla ted Meta bolites. To facilitate characterization
by 31P NMR of the hydrolysis products of DDCIF,
2DCIF, 3DCIF, and IF, the 31P chemical shifts as a
function of pH were determined for each type of struc-
ture: phosphoramidic acid monoester (compounds 2),
2-hydroxyoxazaphosphorine (compounds 3), and phos-
phoric acid monoester (compounds 4) along with the
starting compound (compounds 1). The 31P chemical
shift for IF (1d ) was insensitive to pH, as was that of
phosphoric triamide P(O)(NH2)3, which, in common with
IF, has no titratable oxygen bound to a phosphorus.32
For the other compounds, the plots of 31P chemical shifts
versus pH followed a single sigmoid curve (compounds
2 and 3) or a biphasic curve (compounds 4). Selected
data are presented in Figure 7.
These values are in agreement with those expected for
compounds with this type of structure. Indeed, the first
and the second pKa’s of compounds 4 (4a and 4b-1), 0.8
and 5.9, are in line with those (1.5 and 6.3) of CH3-
OPO3H2 (which has a similar structure around the
phosphorus atom to that of compounds 4).33 Moreover,
the pKa of 2b (2.3) is very similar to that (2.5) of
monomethylphosphoramidate.34 The ∆pKa of 0.7 be-
tween compounds 2b and 2c,d is due to the acid-
strengthening effect of the N-chloroethyl group. This is
in agreement with the difference observed (0.6) between
the pKa of phosphoramide mustard (protonation of
P-NH2 group) and IPM (protonation of P-NCH2CH2Cl
group).32,35
Apart from IF, the magnitude of the pH-induced shift
of the 31P resonance fell into three ranges: ≈-1 (i.e.,
upfield shift with decreasing pH), ≈-4, and ≈-14 ppm.
The two smaller shifts correspond to O-protonation for
the first and second P-OH ionization respectively, as
demonstrated by the 31P protonation shifts observed in
compounds of type 4. These values agree well with the
upfield 31P chemical shifts due to protonation of inor-
ganic phosphate: -0.2, -2.5, and -2.7 ppm for the first,
second, and third P-OH ionization, respectively.32 Gamc-
sik et al. showed that a large pH-induced titration shift
(-12 to -14 ppm) was indicative of an N-protonation
site.32,35 The ≈14 ppm upfield shift was thus assumed
to stem from N-protonation. The phosphoramidic acid
ester 2b (and probably also 2a ) and the 2-hydroxy-1,3,2-
Table 6 gives the values of pKa calculated from curve
fitting as well as the changes in 31P chemical shift due
to protonation. The pKa ranged from 0.8 to 5.9 and
corresponded to protonation of either oxygen or nitrogen
for compounds 2 and 3 and oxygens for compounds 4.

Chemical Stability and Fate of Ifosfamide
J ournal of Medicinal Chemistry, 1999, Vol. 42, No. 14 2553
Ta ble 7. Biological Activity of IF, Its Dechloroethylated
Metabolites, and Some of Its Degradation Compounds
Discu ssion
Our findings are consistent with the facile cleavage
of the P-N bond of phosphoramidates, phosphonami-
dates, or phosphinamidates in aqueous acidic solu-
tions36,37 since the P-N bond in IF and its dechloro-
ethylated metabolites was cleaved at pH between 0 and
7. No P-O bond cleavage was observed.
antitumor activityb
compd
species route LD50 (mg/kg)a in vitro
in vivo
DDCIF (1a )
2DCIF (1b) rat
3DCIF (1c) rat
nd
-
-
-
-
nd
-
iv
iv
iv
ip
iv
ip
ip
ip
>1000
≈800
300
-
IF (1d )
rat
mouse
rat
mouse
mouse
mouse
+++
+++
-
The detailed investigations carried out by Modro and
co-workers on the acidic cleavage of the P-N bond in
acyclic and cyclic phosphoramidates have shown that
it proceeds via the N-protonated species as a reactive
form and that the amino group departs directly from
the apical position of the trigonal-bipyramidal transition
state or intermediate.38 Indeed, in cyclic phosphorodi-
amidates for which a phosphorus atom and a nitrogen
atom are incorporated into a five-membered ring, the
endocyclic P-N bond is broken much more readily than
the exocyclic P-N bond. Our results are in accordance
with this behavior since the first compounds formed in
the hydrolytic pathway of the four oxazaphosphorinanes
studied are the phosphoramidic acid monoesters (2a -
d ) from cleavage of the P-N bond in the six-membered
ring (Scheme 2). These compounds were detected on
hydrolysis at pH 0-7 provided the degradation was not
too slow. If it was slow, only the final hydrolysis
products are observable by 31P NMR as was found for
hydrolysis of 3DCIF (1c) at pH 7.0 or IF (1d ) at pH 5.5
(Table 2).
The decreasing order of stability to acid hydrolysis of
the oxazaphosphorines studied (IF > 3DCIF > 2DCIF
> DDCIF) showed that the presence of the chloroethyl
electron-withdrawing group at the endocyclic and/or
exocyclic nitrogens counteracted cleavage of the P-N
bonds. Since only the endocyclic P-N bond is cleaved
and this cleavage involves the N-protonated form, the
key step governing the hydrolysis rate of 1a -d is the
protonation of the endocyclic nitrogen. The presence of
an electron-withdrawing chloroethyl group on this
nitrogen decreases its basicity, hindering N-protonation.
The phosphorus atom is therefore less electrophilic and
less reactive toward the weakly nucleophilic water.39
This explains the reduced rate of P-N bond hydrolysis
in 2DCIF (1b) relative to DDCIF (1a ) by a factor of ≈150
at pH 6.5 and 5.5 and of IF (1d ) relative to 3DCIF (1c)
by a factor of ≈40 at pH 3 (Table 1), but the presence of
a chloroethyl group on the exocyclic nitrogen decreases
the reactivity of the endocyclic P-N bond by a much
higher factor than on the endocyclic nitrogen. Indeed,
the hydrolytic rate of 3DCIF (1c) relative to DDCIF (1a )
is reduced by a factor of ≈2000 at pH 5.5 and that of IF
(1d ) relative to 2DCIF (1b) by a factor of ≈250 at pH 3
(Table 1). In other words, the hydrolysis rate of the P-N
endocyclic bond was reduced by a factor of ≈10 by the
presence of a chloroethyl group on the exocyclic rather
than endocyclic nitrogen: the ratios of hydrolytic rate
constants 2DCIF/3DCIF were 13 and 7 at pH 5.5 and
3.0, respectively (Table 1). These values are comparable
to that of 9 obtained by Kaijser et al. for the ratio of
2DCIF and 3DCIF degradation rate constants in acidic
aqueous medium at 70 °C.18 These results can only be
explained in terms of lower basicity of the endocyclic
NH or NCH2CH2Cl group relative to the exocyclic
nitrogen even when bearing a chloroethyl group. The
possible reason is that the protonation has a more
700
4b-1
4b-2
5
≈800
-
180
-
-
>1000
2200
-
-
CEA HCl
nd
nd
a
b
nd, not determined. +, effective; -, no activity.
oxazaphosphorine 3a which bear a P-NH2 or P-NH-alkyl
group give rise to a zwitterion on protonation:
The large effect on pKa produced by the N-alkyl
substitution on going from compound 2b (pKa ) 2.3) to
compound 3a (pKa ) 4.1) is also consistent with N-
protonation in view of the marked difference in pKa
between the ammonium ion and its N-alkyl derivatives
(9.2 versus 10.6-10.7).33 On the other hand, the ≈-4
ppm shift observed for the phosphoramidic acid esters
2c,d was much smaller than that observed for N-
protonation but greater than the shifts normally ob-
served for O-protonation of the first acidity of a phos-
phate group. This is consistent with the existence of
both N- and O-protonated forms as already stated for
IPM, a compound in which the phosphorus atom bears
an anionic oxygen (or a hydroxyl group) and an N-
chloroethyl group as in 2c,d :35
Biologica l Activity of Com p ou n d s 1a -d , 4b-1, 4b-
2, a n d 5. The biological activity of these compounds was
investigated in three models: (i) acute toxicity (approxi-
mative LD50) after a single ip or iv injection in mice or
rats, (ii) direct cytotoxic activity in vitro from the
inhibition of colony formation of murine leukemic L1210
cells, and (iii) antitumor efficacy in vivo against murine
P388 leukemia.
The most predictive parameter of biological activity
or toxic hazard seems to be the acute toxicity. The LD50
values and the antitumor activity are listed in Table 7.
Symptoms of toxicity were rather unspecific, i.e., re-
tracted flanks, reduced locomotion, and piloerection.
Death occurred 3-5 days after administration. No direct
cytotoxic activity in vitro, i.e., no inhibition of colony
formation of L1210 cells, was observed at concentrations
ranging from 0.1 to 100 µg/mL for all the compounds.
Similarly, no antitumor efficacy in vivo was observed
except for IF.

2554 J ournal of Medicinal Chemistry, 1999, Vol. 42, No. 14
Gilard et al.
unfavorable steric effect on the endocyclic than on the
exocyclic nitrogen for which the P-N rotation minimizes
the torsional strain at the tetrahedral protonated
moiety. The strongly electronegative protonated exocy-
clic +NH2CH2CH2Cl group increases the electrophilicity
of the phosphorus atom and thus, by a compensating
effect, decreases the basicity of the endocyclic nitrogen
(more than the presence of a directly bonded chloroethyl
group). The protonation is then much more difficult, and
the cleavage of the P-N bond is slowed.
stants of compounds 2a -d at pH 6.0 (Table 1) showed
that (i) compound 3b is formed faster than compound
3a (by a factor of ≈4 if there is liberation of ammonia
and by a factor of 7 if there is liberation of CEA),
demonstrating that the nonprotonated form of the non-
P-linked N-chloroethyl group is present at a higher
concentration than that of the non-P-linked NH2 moiety;
(ii) compounds 3a ,b are formed 2-3 times more rapidly
when the P-linked nitrogen bears the electron-with-
drawing chloroethyl group which gives a better leaving
ability.
Our observations on the stability of aqueous solutions
of IF (no degradation product detected after 1 month
at pH 6.8 and two degradation products representing
together only ≈4% of all the phosphorylated compounds
after 49 days of hydrolysis at pH 5.5) are in line with
its documented stability at room temperature close to
neutral pH. IF has been shown to be stable for at least
1 week in sterile water for injection (pH ≈ 6.5) at room
temperature and for 6 weeks under refrigeration (2-8
°C).20 No evidence of IF decay in normal saline solution
(0.9% NaCl solution) was observed after 9 days at room
temperature or 27 °C, and a loss of only 7% was reported
over a 9-day period at 37 °C in the same solutions.21
Less than 0.6% degradation was observed in water for
injection or 0.9% NaCl solution (pH ≈ 6.5) stored for 8
days in the disposable cassettes for portable iv pumps
at 4, 25, or 35 °C.23 A 3.2% loss of IF was found over a
7-day period at 37 °C in Ringer lactate buffer at pH
7.3.25 The only determined IF half-life at 37 °C at
natural pD (≈6) was 119 days.19 Various other values
for t1/2 have been reported at higher temperatures: ≈12
days at 50 °C and natural pH,24 20 h at 70 °C in the pH
range 5-922 (but, at the same temperature in normal
saline solution (pH ≈ 6.5), another study claimed that,
after 72 h, only 32-44% of IF was degraded21), and 6.5
h at 75 °C in water (pH ≈ 6.5).27 From the observed
rate constants for degradation of IF at various temper-
atures, Kaijser et al.22 calculated a half-life of 254 days
at 20 °C in the pH range 5-9 and Mun˜oz et al.24 619
days at 25 °C and pH 5.0. Since ≈4% of IF is degraded
after 49 days at pH 5.5, we estimated a half-life at room
temperature of 832 days, which is closer to the value
reported by Mun˜oz et al.24 Among the various studies
on the stability of IF, only Highley et al.26 reported a
nonnegligible degradation of IF in aqueous solution at
pH ranging from 4 to 10 at 37-40 °C. They detected
significant amounts of CEA in the medium after 5-12
h of hydrolysis. However, acid hydrolysis during the
derivatization procedure may have contributed to some
of this degradation.40 The rate constant we determined
for hydrolysis of IF in 1 M HCl (0.16 min-1, t1/2 ) 4.4
min) was close to that reported by Zon et al. at pD ≈0
and 20 °C (0.10 min-1, t1/2 ) 6.7 min).19
The following stage in hydrolysis leads to the forma-
tion of type 4 phosphoric acid monoesters, the final
hydrolysis products, by cleavage of the P-N bond of
compounds 3a ,b. Compounds 3a ,b are transient reac-
tion intermediates. Even at pH 6.0, their rate of
degradation is faster than their rate of formation (Table
1). The P-N bond in the oxazaphosphorine ring is made
even more labile by the chloroethyl group on the
nitrogen: t1/2 ≈ 2 h for 3b versus ≈12-16 h for
compound 3a (Table 1). Provided that the precursors
(2a -d ) make up more than 2% of all 31P NMR signals
in the reaction medium, compound 3a resulting from
hydrolysis of DDCIF (1a ) or 3DCIF (1c) is detected from
pH 3.0 to 7.0, whereas compound 3b resulting from the
hydrolysis of 2DCIF (1b) or IF (1d ) is only detected at
pH between 5.5 and 7.0, but not at 3.0 (Table 1). At
highly acidic pH, the non-P-linked nitrogen of com-
pounds 2a -d is totally protonated, so their P-N bonds
are directly cleaved leading to phosphoric acid mono-
esters 4 without any transformation into the type 3
oxazaphosphorines. The breakdown of the P-O bond of
compounds of type 4 leading to formation of Pi was only
observed after long periods of hydrolysis (footnotes to
Scheme 2).
It is likely that formation of compounds of type 3 or
4 from compounds 2a -d occurs via a unimolecular
mechanism (formation of a metaphosphate intermedi-
ate) with some nucleophilic participation. It should be
borne in mind that prerequisites for the fragmentation
of a substrate with the extrusion of a reactive meta-
phosphate derivative X-PO2 are the accumulation of a
negative charge on an atom directly bonded to phos-
phorus and a good leaving group.41,42 For the compounds
studied here, these criteria depend on the pH and the
presence of a chloroethyl substituent on a nitrogen atom.
For example, in weakly acidic medium, the P-NHR
group of compounds 2a -d is not protonated, and
assistance by an incoming nucleophile (here the other
nitrogen atom of the same molecule) is required for
breaking the P-NHR bond and the formation of com-
pounds 3 (Scheme 2). In strongly acidic medium,
compounds 2a -d will be in the P-NHR protonated form
(at least in part), and the conditions for fragmentation
via a transient metaphosphate are fulfilled. The pres-
ence, in the 31P NMR spectrum of the hydrolytic mixture
of 0.5 M IF (1d ) at pH ≈ 0, of a singlet resonating at δ
) -10.63 ppm (characteristic of a symmetrical diphos-
phate moiety43) is in favor of such a mechanistic
pathway. Formal identification of this compound as the
diphosphate 5 was obtained by spiking with authentic
standard, which led to an increase in signal intensity.
This results from the trapping of the metaphosphate
At weakly acidic pH, cleavage of the P-N bond of
compounds 2a -d is accompanied by an intramolecular
attack at the phosphorus atom by the non-P-linked
nitrogen of the compounds 2a -d . This leads to the
formation of a new oxazaphosphorine ring with a
2-hydroxy substituent (instead of a 2-amino or 2-(2-
chloroethyl)amino as in compounds 1a -d ) with the loss
of ammonia or CEA moiety (compounds 3a ,b) (Scheme
2). Substitution on the non-P- or P-linked nitrogen by a
chloroethyl group accelerates the formation of these
compounds. Comparison of the degradation rate con-

Chemical Stability and Fate of Ifosfamide
J ournal of Medicinal Chemistry, 1999, Vol. 42, No. 14 2555
intermediate by the hydrolysis product 4b-1 as reported
below:
aziridine ring of 4b-2 reacts slowly with water at neutral
pH to form an N-(2-hydroxyethyl) group giving rise to
compound 4b-3. This compound was observed after
hydrolysis for more than 20 days.
Our results on the hydrolysis of IF are consistent in
most respects with literature data. For example, the
formation of compound 4b-1 and 2-chloroethylammo-
nium ion on hydrolysis under highly acidic conditions
1
is in line with the H NMR study of Zon et al.19 These
authors detected an equimolecular mixture of these two
compounds under similar experimental conditions
(Scheme 1). The lack of detection of Pi by capillary
electrophoresis among the nine degradation products of
IF after hydrolysis for 48 h at 75 °C27 is in agreement
with our results and confirms the absence of P-O bond
cleavage of the oxazaphosphorine ring in acidic medium.
However, compound A identified by Zon et al.19 in the
degradation mixture obtained after refluxing an aque-
ous solution of IF (pH ≈ 6.5) for 12 days or its presumed
precursors (Scheme 1) was not formed during hydrolysis
of IF at room temperature at values of pH between 0
and 7.
The amount of diphosphate 5 formed rises with
increase in the molar ratio of the two available nucleo-
philes, 4b-1 and H2O. In 2 mL of 1 M HCl, 4% of 1 mmol
of IF (1d ) was transformed into diphosphate 5 and 96%
was transformed into compound 4b-1, while 17% of the
0.7 mmol of further IF added to this mixture was
transformed into diphosphate 5 and 83% into compound
4b-1. Similarly, although 0.035 mmol of 2DCIF (1b) in
acetonitrile in the presence of 0.1 mL of 12 M HCl was
exclusively hydrolyzed into compound 4b-1, 97% of the
further 0.035 mmol of 1b added to this mixture was also
transformed into compound 4b-1 but 3% into diphos-
phate 5 identified by spiking with authentic standard.
Similarly, the hydrolytic decomposition of compound
3b at neutral or weakly acidic pH leading to compound
4b-2 can proceed via the metaphosphate route. At such
pH, compound 4b-2 cannot be formed by cyclization of
the N-chloroethyl group of 4b-1 into an aziridine as
demonstrated in the Results section. The formation of
4b-2 can be accounted for by a unimolecular fragmenta-
tion of compound 3b akin to that of the N-phosphory-
lated nitrogen mustards described by Modro et al.41,42
This fragmentation leads to a metaphosphate species
which quickly reacts with water and to an aziridine
moiety which is stable at neutral or weakly acidic pH,
but an alternative mechanism involving an intermediate
aziridinium zwitterion with a subsequent P-N bond
hydrolysis might also be operative:19
We are not in agreement with the structure of
compound D (Scheme 1) attributed by Mun˜oz et al. to
the major of the two final hydrolysis products of IF in
unbuffered D2O at 50 °C after 21 days.24 Our results
indicate that the phosphorylated final product of IF
hydrolysis should have a phosphate structure rather
than the phosphoramidate structure claimed by Mun˜oz
et al.24 The 31P NMR characteristics of the hydrolytic
IF mixture should have allowed these authors to
propose a phosphate ester rather than a phosphorami-
date structure since the chemical shift of the phosphorus
atom in these two environments is different. Indeed, the
extent of the downfield 31P chemical shift was directly
related to the number of nitrogens bonded to phosphorus
atom (Table 2). The compounds with two P-N bonds of
structure 2-amino-1,3,2-oxazaphosphorine 2-oxide
(DDCIF (1a ) and 2DCIF (1b)) exhibited a chemical shift
of ≈18 ppm. The presence of a chloroethyl group on
the exocyclic nitrogen induced a further upfield shift of
≈2 ppm, producing a 31P chemical shift for 3DCIF (1c)
and IF (1d ) of ≈16 ppm. The first intermediates in the
process of hydrolysis, the phosphoramidic acid mono-
esters (2a -d ) (Scheme 2), only harbor a single P-N
bond, and their 31P chemical shifts were observed
between 9.5 and 10.5 ppm (Table 3) at weakly acidic
and neutral pH. Their chemical shifts were shifted
upfield at pH 2-3 as they have a pKa ≈ 2 (Table 6). For
this type of compounds, the presence of a chloroethyl
group on the nitrogen bound to the phosphorus leads
to an upfield shift of ≈1 ppm (≈9.5 ppm for 2c,d versus
≈10.5 ppm for 2a ,b). The second intermediates with a
single P-N bond in a 1,3,2-oxazaphosphorine ring (3a ,b)
at pH 5-7 exhibited a 31P chemical shift of ≈6-8 ppm,
which is shifted upfield at more acidic pH. The final
hydrolysis products (4a and 4b1-3) are phosphoric
monoesters with no P-N bond, and their 31P chemical
shifts are upfield (<5 ppm). Since they have a pKa ≈ 6
(Table 6), their 31P chemical shifts range from ≈1 to 5
ppm for values of pH ranging from ≈2 to 7 (Table 2).
With decrease in pH, cleavage of the P-N bond of the
oxazaphosphorine ring in compounds 3a ,b probably
proceeds via the N-protonated species, as for compounds
1a -d , leading to compound 4a or 4b-1. At moderately
acidic pH, compound 4b-1 cannot derive from opening
of the aziridine ring of compound 4b-2 by chloride ion
as 31P NMR demonstrated that 4b-2 was stable at such
pH (see Results section). On the other hand, the
In an attempt to identify the final hydrolysis products
of IF (1d ) at 50 °C, we replicated the experiment
described by Mun˜oz et al.24 After hydrolysis for 17 days

2556 J ournal of Medicinal Chemistry, 1999, Vol. 42, No. 14
Gilard et al.
in unbuffered H2O at 50 °C, 31P NMR detected the
presence of unchanged IF (1d ) (16%) and five phospho-
rylated hydrolysis products with chemical shifts close
to that of the external reference, H3PO4. This indicated
that they were phosphates. After extraction of the
residual IF with chloroform, 31P NMR showed the
presence of compound 4b-3 (55%) and four minor
components. Three of these four compounds were iden-
tified by spiking with authentic standards (4b-1 (15%),
4b-2 (4%), Pi (11%)). The unknown compound made up
the remaining 15%. The 13C NMR data allowed the
identification of the two major products (4b-3 and CEA)
and two minor compounds (4b-1 and 2-hydroxyethyl-
amine). The mistaken identification of Mun˜oz et al.24
for the major product of hydrolysis of IF at 50 °C from
13C NMR data was due to the fact that (i) there are two
major hydrolysis products (4b-3 and CEA) rather than
one and (ii) the N-chloroethyl group is no longer bound
to the phosphorus atom but is found in the medium as
CEA.
2c, 3a , and 4a , represented ≈22% of 3DCIF excreted
in urine samples at pH 5.8 from patients treated with
CP at a dose of 60 mg/kg/day as a 3-h infusion. This
percentage was markedly higher in urine at more acidic
pH: 57% at pH 5.1-5.3.44 31P NMR determination of
the N-dechloroethylated metabolites of IF or CP and
their degradation compounds provides an indirect and
accurate estimation of CAA amounts formed from CP
or IF.
Exp er im en ta l Section
An a lytica l Meth od s. NMR spectra were recorded on
Bruker WB-AM300 or Bruker ARX 400 spectrometers. Chemi-
cal shifts (δ) are reported in ppm, relative to 3-(trimethylsilyl)-
propanesulfonic acid sodium salt (TMPS) as an internal
1
standard for H and 13C NMR spectra and relative to 85% H3-
PO4 as an external standard for 31P NMR spectra. Fast atom
bombardment mass spectra in positive mode (FAB+) were
obtained with a VG ZAB HS spectrometer. A Waters (model
991) system equipped with a photodiode array detector was
used for the HPLC analyses.
Ch em ica ls. DDCIF (1a ), 2DCIF (1b), and 3DCIF (1c) were
generously supplied by ASTA Medica AG (Frankfurt, Ger-
many). IF (1d ) was obtained from ASTA Medica (Bordeaux,
France). All other chemicals used were of highest purity
obtainable.
Some IF degradation products were also identified by
GC-MS analysis of IF aqueous solutions at pH 4-10
heated at ≈37 °C for 5-12 h after treatment with either
diazomethane or trifluoroacetic anhydride.26 The methyl
ester of phosphoric acid or the trifluoroacetylated de-
rivative of 2-(chloroethyl)-3-hydroxypropylamine (com-
pound F, Scheme 1) can be obtained only after cleavage
of the P-O bond of IF, a mechanism that does not occur
in water at any value of pH (vide supra). These findings
were indicative of IF degradation during the derivati-
zation procedure rather than in the aqueous medium.
The other compounds detected, the methyl ester of
2-hydroxy-1,3,2-oxazaphosphorine 2-oxide (3b) and the
trifluoroacetylated derivative of CEA, may have been
produced by hydrolysis of IF at pH 4 but not at pH 7 or
10.40 They could not have derived from rupture of the
exocyclic P-N bond of IF as described by Highley et al.26
Indeed, this cleavage was consistently observed after
that of the endocyclic P-N bond (Scheme 2).
Evaluation of biological activity of IF, its dechloro-
ethylated metabolites, and some of their hydrolytic
degradation products showed that, except for IF, all
these compounds have no antitumor efficacy toward
murine P388 leukemia. Neutral or acidic hydrolysis of
IF led to compounds that are less toxic than IF. IF
dechloroethylated metabolites and their hydrolysis prod-
ucts are also less toxic than IF. The sole exception is
the aziridino compound 4b-2, but 2DCIF and IF which
are the sources of 4b-2 are slowly or not hydrolyzed
under physiological conditions.
Syn th esis of 1-P r op a n ol, 3-[(2-Ch lor oeth yl)a m in o]-,
Dih yd r ogen P h osp h a te (Ester ) (4b-1). IF (10 g, 38.3 mmol)
was dissolved in 500 mL of 1 M HCl. When the exothermal
reaction ended, the solution was evaporated under reduced
pressure (about 20 mbar). The residue was dissolved in 100
mL of chloroform:ethanol (1:1). After alkalinization with
triethylamine, the solution was stored for crystallization. The
crystalline product was filtered off, washed with ethanol, and
recrystallized in water:ethanol to give compound 4b-1 in 57%
yield (3.7 g): mp 190-191 °C. Anal. (C5H13NO4PCl) C; H:
calcd, 6.07; found, 6.23. N: calcd, 6.44; found, 6.40. NMR and
mass spectrometry data are identical to those reported in Table
4.
Syn th esis of Dip h osp h or ic Acid P ,P ′-Bis[3-[(2-ch lor o-
et h yl)a m in o]p r op yl] E st er ([Cl(CH 2)2NH (CH 2)3OP (O)-
(OH)]2O, 5). 3-(Chloroethyl)-2-chlorooxazaphosphorinane 2-ox-
ide (50 g, 230 mmol) synthesized as previously described45 was
dissolved in 400 mL of dry THF; 11.4 g (285 mmol) of crushed
sodium hydroxide was added and the suspension stirred at
room temperature and then refluxed for 10 h. The precipitate
of sodium chloride was filtered off and the solution concen-
trated. The oily residue was resuspended in 300 mL of THF:
water (1:1), stirred for 1 h after acidification with HCl, and
then concentrated in vacuo. The residue was dissolved in 100
mL of water and 350 mL of ethanol. After alkalinization with
triethylamine to pH 8, the solution was stirred for 2 h at 0 °C.
The next day, the crystalline product was filtered off and
washed with ethanol and diethyl ether. Recrystallization from
acetone:water at 4 °C gave compound 5 in 29% yield (14 g):
mp 193-194 °C. Anal. (C10H24N2O7P2Cl2) C, H; N: calcd, 6.72;
found, 6.56. FAB+ mass spectrometry: ions at m/z 417, 419,
421 [MH+] (2 Cl). NMR (D2O, pD 2.6): δ (1H) 4.00 (4H, m as
they are diastereotopic, CH2O), 3.82 (4H, m, CH2Cl), 3.40 (4H,
t, 5.6 Hz, NCH2 CH2Cl), 3.22 (4H, t, 7.1 Hz, NCH2CH2CH2O),
2.02 (4H, app quin, 6.3 Hz, NCH2CH2CH2O); δ (13C) 66.5 (broad
signal, CH2O), 51.6 (s, NCH2 CH2Cl), 48.2 (s, NCH2CH2CH2O),
41.8 (s, CH2Cl), 28.92 and 28.95 (NCH2CH2CH2OP, the carbon
of the two chains are magnetically nonequivalent, each giving
In conclusion, the present study showed the value of
31P NMR for determining the successive steps in the
acidic hydrolytic pathways of IF and its dechloroethy-
lated metabolites (DDCIF, 2DCIF, and 3DCIF). The
1
combination of data from MS and H, 13C, and 31P NMR
led to the identification of all their degradation products
in neutral or acidic aqueous solutions. Some of these
compounds, 2a -c, 3a , and 4a , were detected and
quantified in urine of patients treated with CP or IF.
Degradation compounds of DDCIF, 2a and 3a , and
2DCIF, 2b, represented ≈10% of 2DCIF (DDCIF itself
was not detected) excreted over 24 h in urine of patients
receiving 3 g/m2 IF as a 3-h infusion.9 In a more recent
and detailed study, degradation compounds of 3DCIF,
3
1
a doublet with a J CP 3.8 Hz); δ (31P) -10.0 (s). The H and 13C
signals were attributed from COSY and HETCOR correlations.
H yd r olysis P r od u ct s of DDCIF (1a ). 1-P r op a n ol, 3-
a m in o-, h yd r ogen p h osp h or a m id a te (ester ) (2a ): DDCIF
(1a ) (0.012 g, 0.09 mmol) was dissolved in 3 mL of water, the
pH of which was then adjusted to 2.3 with HCl. After 20 min
at room temperature, the solution was neutralized with NaOH.
31P NMR showed disappearance of the DDCIF (1a ) signal and
appearance of those of 2a (>95%) and traces of 3a and 4a .

Chemical Stability and Fate of Ifosfamide
J ournal of Medicinal Chemistry, 1999, Vol. 42, No. 14 2557
The NMR and mass spectrometric data for compound 2a are
listed in Table 3.
2-H yd r oxy-2-oxo-1,3,2-oxa za p h osp h or in e (3a ): This
compound was not isolated, but it represented ≈75% of the
reaction mixture (estimated by 31P NMR) after keeping an
aqueous solution of 2c for 3 days at room temperature after
adjusting the pH to 8 with 0.1 M NaOH. The structure of
compound 3a was characterized by positive-ion FAB mass
spectrometry: ions at m/z 138 [MH]+ and 160 [MNa]+. 13C
NMR: δ C4 44.9 (d, 4.4 Hz), C5 29.8 (d, 6.7 Hz), C6 70.7 (d,
5.4 Hz) in H2O at pH 7.9. Liberation of CEA in the medium
was evidenced by the presence of a 13C signal at 44.4 ppm
corresponding to the CH2Cl and CH2N carbons as demon-
strated by spiking the mixture with authentic standard.
2-Hyd r oxy-2-oxo-1,3,2-oxa za p h osp h or in e (3a ): A solu-
tion of compound 2a , whose pH was maintained at 8.5 by
periodic addition of NaOH, was sequentially analyzed (once a
day) by 31P NMR. After 7 days, the compound 3a represented
80% of all the 31P signals. The solution was then neutralized,
and the mass and 13C NMR characteristics of compound 3a
were determined (Table 3).
1-P r op a n ol, 3-a m in o-, d ih yd r ogen p h osp h a te (ester )
(4a ): DDCIF (1a ) (0.013 g, 0.10 mmol) was dissolved in 3 mL
of 1 M HCl. After 10 h at room temperature, the 31P NMR
spectrum showed the presence of a sole compound whose mass
and, after raising the pH to ≈5, NMR spectrum (Table 3) were
identical to that of compound 4a obtained by an independent
synthesis.28
1-P r op a n ol, 3-a m in o-, d ih yd r ogen p h osp h a te (ester )
(4a ): 3DCIF (0.014 g, 0.07 mmol) was dissolved in 2 mL of 1
M HCl. After 24 h at room temperature, 31P NMR showed the
presence of the sole compound 4a , the positive-ion FAB mass
data (glycerol matrix) showed that of 4a with ion at m/z 156
[MH]+ and CEA with ions at m/z 80, 82 [MH]+ and 172, 174
[MH + glycerol]+, and the 13C NMR spectrum showed also the
presence of 4a with signals at δ 40.5 (s, C4), 30.6 (d, 7.0 Hz,
C5), and 65.8 (d, 5.1 Hz, C6), as well as that of CEA with
signals at δ 43.8 (s, CH2Cl) and 44.3 (s, CH2N) in H2O at pH
4.1.
Hyd r olysis P r od u cts of 2DCIF (1b). 1-P r op a n ol, 3-[(2-
ch lor oeth yl)a m in o]-, h yd r ogen p h osp h or a m id a te (ester )
(2b): 2DCIF (1b) (0.008 g, 0.04 mmol) was dissolved in 2 mL
of water. The pH of the solution was then adjusted to 2.4 with
HCl. After 20 min at room temperature, the pH of the solution
was neutralized with NaOH. The 31P NMR spectrum, recorded
at 4 °C, showed the presence of 2b and traces of 4b-1 (≈3%).
The degradation kinetics of compound 2b were obtained after
dissolving the freeze-dried solution in cacodylate buffer. The
13C NMR and mass spectrometric data of compound 2b are
listed in Table 4.
Hyd r olysis P r od u cts of IF (1d ). 1-P r op a n ol, 3-[(2-
ch lor oeth yl)a m in o]-, N-(2-ch lor oeth yl)-, h yd r ogen p h os-
p h or a m id a te (ester ) (2d ): IF (0.1 g, 0.38 mmol) was dis-
solved in 10 mL of water, and the pH was adjusted to 0.9 with
1 M HCl. After stirring for 30 min at room temperature, the
solution was neutralized with 1 M KOH and then extracted
with CHCl3 to remove unreacted IF. The aqueous phase was
freeze-dried and the residue chromatographed on a SepPak
C18 column and eluted with distilled water at pH 7. Fractions
were collected and analyzed by HPLC under the following
conditions: column, Lichrosorb-RP Select B 5 µm (150 × 4
mm); UV detection at 195 nm; eluant, water/acetonitrile (95/
5) at a flow rate of 0.8 mL/min. The retention time of 2d was
5.8 min, whereas that of 4b-1 (which was also formed in acidic
medium) was 2.8 min. The fractions containing compound 2d
were immediately freeze-dried and stored at -80 °C. The mass
and 13C NMR data of 2d are listed in Table 5.
1-P r op a n ol, 3-[(2-ch lor oet h yl)a m in o]-, d ih yd r ogen
p h osp h a te (ester ) (4b-1): 0.033 g (0.17 mmol) of 2DCIF (1b)
was dissolved in 8 mL of 1 M HClO4. After 15 h at room
temperature, 31P NMR showed the disappearance of the 2DCIF
signal and the presence of a sole signal. Aliquots of this
solution (solution A) were analyzed by mass spectrometry and,
after raising the pH to 5.8 with 1 M KOH, centrifugation to
remove precipitated KClO4, freeze-drying, and redissolution
1
in D2O, by H and 13C NMR. The mass and NMR character-
istics of the compound obtained (4b-1) are listed in Table 4.
1-P r op a n ol, 3-a zir id in o-, d ih yd r ogen p h osp h a te (es-
ter ) (4b-2): The pH of the solution A remaining was adjusted
to 13 by addition of 1 M KOH. After 24 h at room temperature,
the solution was centrifuged and the supernatant, analyzed
by 31P NMR, showed the sole presence of compound 4b-2 whose
structure was characterized by 1H and 13C NMR (Table 4). The
high salt concentration of this solution excluded mass spec-
trometry, but the MS characteristics of compound 4b-2 (m/z
ion [MH]+ at 182) were obtained by analysis of the degradation
mixture of compound 2d .
2-Hyd r oxy-2-oxo-3-(2-ch lor oeth yl)-1,3,2-oxa za p h osp h o-
r in e (3b): Compound 2d in neutral and slightly acidic
aqueous solutions is degraded to compound 3b, which is
unstable and is transformed into 4b-1 and/or 4b-2 depending
on pH. Compound 3b could not, therefore, be isolated, and its
characteristics (mass and 13C NMR) had to be obtained by
monitoring the degradation of compound 2d .
1-P r op a n ol, 3-[(2-ch lor oet h yl)a m in o]-, d ih yd r ogen
p h osp h a te (ester ) (4b-1) a n d 1-P r op a n ol, 3-a zir id in o-,
d ih yd r ogen p h osp h a te (ester ) (4b-2): IF (0.04 g, 0.15 mmol)
was dissolved in 2 mL of 1 M HClO4. After 24 h at room
temperature, the solution containing only one phosphorylated
compound as evidenced by 31P NMR was split into two equal
parts.
1-P r op a n ol, 3-[(2-h yd r oxyet h yl)a m in o]-, d ih yd r ogen
p h osp h a te (ester ) (4b-3): 2DCIF (1b) (0.0085 g, 0.04 mmol)
was dissolved in 2 mL of 0.1 M sodium cacodylate buffer at
pH 6.9. After 58 days at room temperature, the 31P NMR
spectrum of the resulting mixture (pH 6.7) showed the pres-
ence of three strong signals at 17.68 (16%) (2DCIF, 1b), 4.64
(64%) (4b-3) and 4.56 ppm (20%) (4b-2), the latter being
identified by spiking with authentic standard. Mass spectrom-
etry and 13C NMR analysis of the reaction mixture also showed
the presence of the same three compounds. The mass and 13C
NMR characteristics of compound 4b-3 are listed in Table 4.
The first part was analyzed by MS and, after raising the
pH to 3.4 with 1 M NaOH, by 13C NMR. FAB-positive MS
showed the presence of compound 4b-1 with ions at m/z 218,
220 [MH]+. 13C NMR signals were characteristic of compound
4b-1 (δ 48.3 (s, C4), 29.3 (d, 6.9 Hz, C5), 65.6 (d, 5.2 Hz, C6),
51.8 (s, C9), 42.2 (s, C10)) and chloroethylamine (δ 43.8 (s,
CH2Cl), 44.3 (s, CH2N)).
H yd r olysis P r od u ct s of 3DCIF (1c). 1-P r op a n ol, 3-
a m in o-, N-(2-ch lor oeth yl)-, h yd r ogen p h osp h or a m id a te
(ester ) (2c): 3DCIF (0.015 g, 0.08 mmol) was dissolved in 2
mL of aqueous 0.1 M HCl. After a few minutes, the solution
was neutralized with 1 M KOH. 31P NMR showed that 3DCIF
had been completely hydrolyzed into 2c and a compound
previously identified as 4a . We analyzed the mixture by HPLC
with a 5-µm Lichrosorb-RP Select B column (150 × 4 mm)
under the following conditions: eluant, water/acetonitrile (98/
2) at a flow rate of 0.7 mL/min, with UV detection at 200 nm.
The retention time of 2c was 5.6 min, whereas that of 4a was
3.2 min. The fractions containing 2c were immediately frozen
in liquid nitrogen, then freeze-dried, and stored at -80 °C.
The compound isolated was 98% pure as estimated from 31P
NMR. Its mass and 13C NMR data are listed in Table 5.
The pH of the second part was raised to 13 with 1 M KOH.
After centrifugation to remove the precipitate of KClO4, the
solution was left for 24 h at room temperature. 31P NMR
showed the presence of the sole phosphorylated compound 4b-
2, confirmed by 13C NMR, which also indicated the presence
of the characteristic resonance of aziridine at δ ) 20.3 ppm.
The solution was freeze-fried to remove the highly volatile
aziridine. After the pellet was dissolved in H2O, the pH was
adjusted to 7.5 with 1 M HCl. 13C NMR showed the presence
of the sole compound 4b-2 with δ ) 55.5 (s, C4), 30.4 (d, 5.8
Hz, C5), 64.7 (d, 5.0 Hz, C6), 32.0 (s, C9/C10) of the aziridine
ring.

2558 J ournal of Medicinal Chemistry, 1999, Vol. 42, No. 14
Gilard et al.
P r olon ged Hyd r olysis of IF a t 50 °C. IF (0.023 g, 0.05
mmol) was dissolved in 3 mL of water. The solution (initial
pH 6.5) was maintained at 50 °C in a circulating water bath
for 17 days (final pH 4.6). At this time, 31P NMR showed the
presence of unreacted IF (1d , ≈16% of all the 31P NMR signals)
along with five other signals with chemical shifts ranging from
1.4 to 0.8 ppm, characteristic of phosphate structures. This
solution was then extracted with 2 × 1 mL of CHCl3 to remove
unreacted IF and the aqueous phase analyzed by 31P and 13C
NMR.
0.990. C and C0 refer to the percent of 31P NMR signal
intensities of the starting material relative to that of all
observable 31P signals. We checked for various degradation
kinetics that the rate constants for the disappearance of the
starting material were close to those obtained with values of
the concentrations (C and C0) of the starting materials
measured by comparison of the integrated intensities of their
31P NMR signals with that of the 31P NMR signal of meth-
ylphosphonic acid, a quantification standard placed in a sealed
coaxial capillary.
The 31P NMR spectrum recorded at pH 6.0 (pH giving the
best discrimination) contained five signals of which four were
attributed to 4b-1, 4b-2, and Pi by spiking with authentic
standards and 4b-3 by comparison of its chemical shift with
that of an authentic standard. The chemical shifts, relative
intensities, and attribution were as follows: 3.76 ppm, 15%,
4b-1; 3.71 ppm, 55%, 4b-3; 3.69 ppm, 4%, 4b-2; 1.76 ppm, 11%,
Pi; 1.73 ppm, 15%, unknown.
The 13C NMR spectrum recorded at pH 4.6 had major
signals from compounds 4b-3 and CEA attributed by compari-
son with those of known standards. 4b-3: δ 48.1 (s, C4), 29.3
(d, 7.0 Hz, C5), 65.4 (d, 5.2 Hz, C6), 52.1 (s, C7), 59.5 (s, C8).
CEA: δ 44.25 (s, CH2N), 43.8 (s, CH2Cl).
The rate constants for consecutive reactions were deter-
mined by a least-squares fit method using a computer program
developed in the laboratory.47 The program analyzed the
measured time dependence of the relative concentrations of
compounds involved (1-4 or 2-4) for kinetic equations
k2
k3
describing a consecutive-irreversible two 2 f 3 f 4 or three 1
k1
k2
k3
f 2 f 3 f 4 stage reaction. We assumed that all the
reactions obeyed a first-order rate law, with k1, k2, and k3 the
rate constants for the disappearance of compounds 1, 2, and
3, respectively. For 2b,d , k3 is the sum of the rate constants
for the disappearance of 3 leading to simultaneous formation
of 4b-1 and 4b-2. The k1 values differed little on fitting either
the first step, the first and second steps, or the three steps of
the hydrolytic reaction of 1a . Similar values for k2 were found
by fitting either the first step or the two steps of the hydrolysis
of 2b-d . Half-lives were derived from the rate constants (t1/2
) ln 2/k).
Among the numerous minor signals, we identified those of
4b-1 and 2-hydroxyethylamine by comparison with those of
authentic standards. 4b-1: δ 48.2 (s, C4), 29.2 (d, 7.2 Hz, C5),
65.3 (d, 5.2 Hz, C6), 51.7 (s, C7), 42.0 (s, C8). 2-Hydroxyethyl-
amine: δ 60.5 (s, CH2OH), 44.3 (s, CH2N).
p Ka Deter m in a tion s. For the determinations of pKa, the
compound of interest was added to aqueous 0.1 M NaCl to
avoid variations in ionic strength during titration. The pH was
measured before and after each 31P NMR spectrum recorded
at 25 °C and modified by the addition of NaOH or HCl. The
pKa values were calculated according to the equation reported
by Lachmann and Schnackerz.48
Biologica l Meth od s. An im a ls: Female SD rats and male
CD2F1 mice were obtained from the Zentralinstitut fu¨r
Versuchstierzucht, Berlin, Germany. They were kept under
specific-pathogen-free conditions, were fed with a standard
pellet diet (Altromin 1324) ad libitum, and had an unrestricted
supply of acidified water (pH 3).
Cell lin es: P388 and L1210 mouse leukemia were obtained
from Dr. Atassi, Institut J ules Bordet, Brussels, Belgium.
Ma ter ia ls: The materials used included compounds 1a -
d , 4b-1, 4b-2, and 5, CEA HCl, RPMI-1640 (Gibco, D-7500
Karlsruhe, Germany), fetal calf serum (Seromed, D-1000
Berlin, Germany), Bacto-Agar (Difco, Detroit, MI), and phos-
phate-buffered saline (Seromed, D-1000 Berlin, Germany).
Acu te toxicity: Two female SD rats or male CD2F1 mice
per dose group were injected iv or ip with the test compound.
The animals were observed for lethality and clinical signs of
toxicity for 28 days. An approximate LD50 value was thus
determined.
Colon y a ssa ys: In vitro assays for inhibition of colony
growth in soft agar (colony assay) were performed according
to the method of Hamburger and Salmon.98 Briefly, tumor cells
were incubated in RPMI-1640 medium containing 20% fetal
calf serum and solidified with 0.3% agar in the presence of
different concentrations of the cytostatic agent at 37 °C, 95%
relative humidity, and 7.5% CO2. The experiments were
performed in triplicate. The incubation time was 6 days.
Subsequently, the colonies of more than 50 cells were counted.
The concentration of the cytostatic agent resulting in an
inhibition of colony formation by 90% (EC90) was determined
graphically.
The ratio of the signal intensities of 4b-3/4b-1 was ≈4 in
common with that observed on 31P NMR.
Hyd r olysis Kin etics by 31P NMR. Compounds to be
hydrolyzed were dissolved in 2.5 mL of the appropriate buffer
(0.1 or 0.2 M sodium cacodylate at pH 6.8, 6.0, and 5.5, 0.2 M
sodium phthalate at pH 3.0, 0.5 M KCl-HCl at pH 2.0, and 1
M HCl at pH ≈ 0), and the pH of the resulting solutions was
measured. The initial concentration of all these compounds
was 2.8 × 10-2 M.
After recording initial “zero-time” 31P NMR spectra, the
samples of 2DCIF (1b) at pH 6.8 and 5.5, 3DCIF (1c) at pH
5.5, and IF (1d ) at pH 3.0 whose degradation kinetics are very
slow (t1/2 ≈ 4 days) were maintained at room temperature near
20 °C and their 31P NMR spectra then recorded at various
times. The degradation kinetics of all the other compounds
were continuously monitored in the NMR probe regulated at
25 °C, and the spectra were acquired at different intervals
depending of their rate of hydrolysis. Time points for each
spectrum were taken at the midpoint of data acquisition.
Spectra were run on a Bruker WB AM 300 or ARX 400
spectrometer using the inverse-gated decoupling technique
under the instrumental conditions described elsewhere,28,40,46
shortly a 35° pulse of 5 µs and a recycle time of 2.08 s.
Negligible reaction took place during the NMR recording
(≈10-30 min) of the samples analyzed over several days.
pH was measured before and after each NMR spectrum.
Only a slight change in pH (0.2 pH unit) was generally
observed during the hydrolysis, except for that of 3DCIF at
pH 5.5 (0.25 pH unit) and pH 3.0 (0.3 pH unit) and IF at pH
2.0 (0.35 pH unit). All runs were followed over reaction periods
>50%, except for degradation of 3DCIF at pH 5.5 and IF at
pH 3.0 for which only ≈35% of the starting compounds were
hydrolyzed. All kinetic runs were made in duplicate or
triplicate, except for the very slow degradation reactions of
2DCIF at pH 6.8 and 3DCIF at pH 5.5 where only one run
was made.
The relative concentrations of the phosphorylated com-
pounds observed in the 31P NMR spectra were determined from
their integrated signal intensities (peak heights). In a previous
study,28 we found that this procedure gave an accurate
estimate of the relative concentrations of CP (an isomer of IF)
and its hydrolysis products, which have similar structures to
the oxazaphosphorines 1 and their degradation products.
Linear least-squares fits of pseudo-first-order plots of ln-
[starting materials]t/[starting materials]t0 (ln C/C0) versus time
gave the values for the rate constants, with in all cases r g
Exp er im en ta l tu m or s in vivo: Six female SD rats or male
CD2F1 mice per group were inoculated ip with 106 freshly
harvested P388 cells in 0.2 mL of phosphate-buffered saline.
The animals were given the test substances iv or ip either once
on the following day (day 1) or daily from days 1 to 4. Control
animals were treated with solvent. Median survival times
(MST) were determined for the respective groups, and an
increase of life span (ILS) was determined as a percentage of
the control group.

Chemical Stability and Fate of Ifosfamide
J ournal of Medicinal Chemistry, 1999, Vol. 42, No. 14 2559
P a tien ts’ u r in e: 31P NMR analysis was carried out directly
on urine without any extraction. For patients treated with IF,
crude (i.e., nonconcentrated) urine samples were analyzed,
whereas they were concentrated from 20 to 6 mL for patients
treated with CP as these patients exhibited a marked diuresis
(2.5-7.9 L/day).
(21) Radford, J . A.; Margison, J . M.; Swindell, R.; Lind, M. J .;
Wilkinson, P. M.; Thatcher, N. The stability of ifosfamide in
aqueous solution and its suitability for continuous 7-day infusion
by ambulatory pump. Cancer Chemother. Pharmacol. 1990, 26,
144-146.
(22) Kaijser, G. P.; Beijnen, J . H.; Bult, A.; Hogeboom, M. H.;
Underberg, W. J . M. A systematic study on the chemical stability
of ifosfamide. J . Pharm. Biomed. Anal. 1991, 9, 1061-1067.
(23) Mun˜oz, M.; Girona, V.; Pujol, M.; Dura´n, S.; Vicente, P.; Sole´,
L. A. Stability of ifosfamide in 0.9% sodium chloride solution or
water for injection in a portable i.v. pump cassette. Am. J . Hosp.
Pharm. 1992, 49, 1137-1139.
(24) Mun˜oz, M.; Bonjoch, J .; Prat, J .; Pujol, M.; Girona, V.; de Bolos,
J . Degradation kinetics of ifosfamide in aqueous solution, Int.
J . Pharm. 1996, 139, 249-253.
(25) Leone, L.; Comandone, A.; Oliva, C.; Bussi, P.; Goffredo, F.;
Bretti, S.; Bumma, C. Stability of ifosfamide in solutions for
multiday infusion by external pump. Anti-Cancer Drugs 1995,
6, 604-607.
(26) Highley, M. S.; Momerency, G.; Van Cauwenberghe, K.; Van
Oosterom, A. T.; de Bruijn, E. A.; Maes, R. A. A.; Blake, P.;
Mansi, J .; Harper, P. G. Formation of chloroethylamine and 1,3-
oxazolidine-2-one following ifosfamide administration in hu-
mans. Drug Metab. Dispos. 1995, 23, 433-437.
(27) Hill, S.; Aexiou, J .; Kavanagh, P. E.; Livingstone, P. C.; Shalliker,
R. A.; Soscic, J . Separation of ifosfamide and its degradation
products using micellar electrokinetic chromatography. Anal.
Commun. 1996, 33, 235-237.
(28) Gilard, V.; Martino, R.; Malet-Martino, M. C.; Kutscher, B.;
Mu¨ller, A.; Niemeyer, U.; Pohl, J .; Polymeropoulos, E. E.
Chemical and biological evaluation of hydrolysis products of
cyclophosphamide. J . Med. Chem. 1994, 37, 3986-3993.
(29) Dermer, O. C.; Ham, G. E. Ethylenimine and other aziridines;
Academic Press: New York, 1969.
Ack n ow led gm en t. This work was supported by
grants from the Association pour la Recherche sur le
Cancer (Grant 6635) and the Ligue Nationale Franc¸aise
contre le Cancer. The authors wish to thank Drs. J ean-
Claude Micheau, Dominique Lavabre, and Ve´ronique
Pimienta for helpful discussions on the kinetics.
Refer en ces
(1) Connors, T. A.; Cox, P. J .; Farmer, P. B.; Foster, A. B.; J arman,
M. Some studies of the active intermediates formed in the
microsomal metabolism of cyclophosphamide and isophospha-
mide. Biochem. Pharmacol. 1974, 23, 115-129.
(2) Sladek, N. E. Metabolism of oxazaphosphorines. Pharmacol.
Ther. 1988, 37, 301-355.
(3) Lind, M. J .; Ardiet, C. Pharmacokinetics of alkylating agents.
Cancer Surveys 1993, 17, 157-188.
(4) Sladek, N. E. Metabolism and pharmacokinetics behavior of
cyclophosphamide and related oxazaphosphorines. In Anticancer
Drugs: Reactive metabolism and drug interactions; Powis, G.,
Ed.; Pergamon Press: Oxford, 1994; pp 79-156.
(5) Kaijser, G. P.; Beijnen, J . H.; Bult, A.; Underberg, W. J . M.
Ifosfamide metabolism and pharmacokinetics. Anticancer Res.
1994, 14, 517-532.
(6) Wagner, T. Ifosfamide clinical pharmacokinetics. Clin. Phar-
macokinet. 1994, 26, 439-456.
(30) The 31P chemical shifts for the degradation compounds of IF and
its dechloroethylated metabolites are pH-dependent since their
phosphorus atom is directly linked to protonation sites such as
amino and/or hydroxyl groups. To determine if the compounds
resulting from the hydrolysis of DDCIF, 2DCIF, 3DCIF, and IF
had identical structures, all the degradation mixtures obtained
were systemically spiked by aliquots of the other degradation
mixtures and their 31P NMR spectra recorded. The compounds
which exhibited a single 31P signal were assumed to have the
same structures. This assumption was then checked by compar-
ing their 13C NMR spectra.
(31) Yonezawa, T.; Morishima, I.; Fukuta, K. Solvent effects in NMR
spectroscopy. II. A repulsive interaction between the benzene
solvent molecule and the nitrogen lone-pair, and the effects
induced by the addition of protic substances. Bull. Chem. Soc.
J pn. 1968, 41, 2297-2305.
(7) Norpoth, K. Studies on the metabolism of isophosphamide in
man. Cancer Treat. Rep. 1976, 60, 437-443.
(8) Martino, R.; Crasnier, F.; Chouini-Lalanne, N.; Gilard, V.;
Niemeyer, U.; de Forni, M.; Malet-Martino, M. C. A new approch
to the study of ifosfamide metabolism by the analysis of human
body fluids with 31P nuclear magnetic resonance spectroscopy.
J . Pharmacol. Exp. Ther. 1992, 260, 1133-1144.
(9) Gilard, V.; Malet-Martino, M. C.; de Forni, M.; Niemeyer, U.;
Ader, J . C.; Martino, R. Determination of the urinary excretion
of ifosfamide and its phosphorated metabolites by phosphorus-
31 nuclear magnetic resonance spectroscopy. Cancer Chemother.
Pharmacol. 1993, 31, 387-394.
(10) Wang, J . J . H.; Chan, K. K. Identification of new metabolites of
ifosfamide in rat urine using ion cluster technique. J . Mass
Spectrom. 1995, 30, 675-683.
(11) Goren, M. P.; Wright, R. K.; Pratt, C. B.; Pell, F. E. Dechloro-
ethylation of ifosfamide and neurotoxicity. Lancet II 1986, 1219.
(12) Pohl, J .; Stekar, J .; Hilgard, P. Chloroacetaldehyde and its
contribution to urotoxicity during treatment with cyclophospha-
mide or ifosfamide. Arzneim. Forsch. 1989, 39, 704-705.
(13) J oqueviel, C.; Malet-Martino, M.; Martino, R. A 13C NMR study
of 2-13C-chloroacetaldehyde, a metabolite of ifosfamide and
cyclophosphamide, in the isolated perfused rabbit heart model.
Initial observations on its cardiotoxicity and cardiac metabolism.
Cell. Mol. Biol. 1997, 43, 773-782.
(14) Kurowski, V.; Wagner, T. Comparative pharmacokinetics of
ifosfamide, 4-hydroxyifosfamide, chloroacetaldehyde, and 2- and
3-dechloroethylifosfamide in patients on fractionated intrave-
nous ifosfamide therapy. Cancer Chemother. Pharmacol. 1993,
33, 36-42.
(15) Kaijser, G. P.; Beijnen, J . H.; J eunink, E. L.; Bult, A.; Keizer,
H. J .; de Kraker, J .; Underberg, W. J . M. Determination of
chloroacetaldehyde, a metabolite of oxazaphosphorine cytostatic
drugs, in plasma. J . Chromatogr. 1993, 614, 253-259.
(16) Kaijser, G. P.; Korst, A.; Beijnen, J . H.; Bult, A.; Underberg, W.
J . M. The analysis of ifosfamide and its metabolites. Anticancer
Res. 1993, 13, 1311-1324.
(17) Momerency, G.; Van Cauwenberghe, K.; de Bruijn, E. A.; Van
Oosterom, A. T.; Highley, M. S.; Harper, P. G. Determination of
iphosphamide and seven metabolites in blood plasma, as stable
trifluoroacetyl derivatives, by electron capture chemical ioniza-
tion GC-MS. J . High Resolut. Chromatogr. 1994, 17, 655-661.
(18) Kaijser, G. P.; Aalbers, T.; Beijnen, J . H.; Bult, A.; Underberg,
W. J . M. Chemical stability of cyclophosphamide, trofosfamide,
and 2- and 3-dechloroethylifosfamide in aqueous solutions. J .
Oncol. Pharm. Pract. 1996, 2, 15-21.
(19) Zon, G.; Ludeman, S. M.; Egan, W. High-resolution nuclear
magnetic resonance investigations of the chemical stability of
cyclophosphamide and related phosphoramidic compounds. J .
Am. Chem. Soc. 1977, 99, 5785-5795.
(20) Trissel, L. A.; Kleinman, L. M.; Cradock, J . C.; Davignon, J . P.
Investigational drug information. Drug Intell. Clin. Pharm.
1979, 13, 340-343.
(32) Gamcsik, M.; Ludeman, S. M.; Shulman-Roskes, E. M.; McLen-
nan, I. J .; Colvin, M. E.; Colvin, O. M. Protonation of phosphor-
amide mustard and other phosphoramides. J . Med. Chem. 1993,
36, 3636-3645.
(33) Albert, A.; Serjeant, E. P. Ionization constants of acids and bases;
J ohn Wiley and Sons: New York, 1962.
(34) O¨ ney, I.; Caplow, M. Phosphoramidate Solvolysis. J . Am. Chem.
Soc. 1967, 89, 6972-6980.
(35) Millis, K. K.; Colvin, M. E.; Shulman-Roskes, E. M.; Ludeman,
S. M.; Colvin, O. M.; Gamcsik, M. P. Comparison of the
protonation of isophosphoramide mustard and phosphoramide
mustard. J . Med. Chem. 1995, 38, 2166-2175.
(36) Koizumi, T.; Haake, P. Acid-catalyzed and alkaline hydrolyses
of phosphinamides. The lability of phosphorus-nitrogen bonds
in acid and the mechanisms of reaction. J . Am. Chem. Soc. 1973,
95, 8073-8079.
(37) Rahil, J .; Haake, P. Reactivity and mechanism of hydrolysis of
phosphonamides. J . Am. Chem. Soc. 1981, 103, 1723-1734.
(38) Modro, T. A.; Graham, D. H. Phosphoric Amides. 3. Acidic
cleavage of the phosphorus-nitrogen bond in acyclic and cyclic
phosphoramidates. J . Org. Chem. 1981, 46, 1923-1925.
(39) Tyssee, D. A.; Bausher, L. P.; Haake, P. Displacement at
phosphorus by a mechanism with A1 character. Acid-catalyzed
hydrolysis of phosphinanilides. J . Am. Chem. Soc. 1973, 95,
8066-8072.
(40) Gilard, V.; Martino, R.; Malet-Martino, M.; Niemeyer, U. Stabil-
ity of commercial formulations and aqueous solutions of ifosfa-
mide (a reply). Drug Metab. Dispos. 1997, 25, 927-931.
(41) Le Roux, C.; Modro, A. M.; Modro, T. A. Decomposition of
N-phosphorylated nitrogen mustards: a mechanistic investiga-
tion. J . Org. Chem. 1995, 60, 3832-3839.
(42) Wan, H.; Modro, T. A. Chemistry of N-phosphorylated nitrogen
mustards: the effect of a second nitrogen substituent at phos-
phorus on the stability of the system. Phosphorus, Sulfur Silicon
1996, 108, 155-168.

2560 J ournal of Medicinal Chemistry, 1999, Vol. 42, No. 14
Gilard et al.
(43) Berger, S.; Braun, S.; Kalinowski, H. O. 31P NMR spectroscopy.
NMR spectroscopy of the non metallic elements; J ohn Wiley and
Sons: Chichester, England, 1997; pp 700-1019.
(44) J oqueviel, C.; Martino, R.; Gilard, V.; Malet-Martino, M.; Canal,
P.; Niemeyer, U. Urinary excretion of cyclophosphamide in
humans, determined by phosphorus-31 nuclear magnetic reso-
nance spectroscopy. Drug Metab. Dispos. 1998, 26, 418-428.
(45) Sociedad Espanola Especialidades Farmaco-Therapeuticas SA.
Preparation of ifosfamide. Spanish Pat. ES 526194, 1985; Chem.
Abstr. 1987, 107, 7385r.
cyclophosphamide and ifosfamide. Cancer Chemother. Pharma-
col. 1997, 40, 391-399.
(47) Pimienta, V.; Lavabre, D.; Levy, G.; Micheau, J . C. Kinetic
modeling of the KMnO4/H2C2O4/H2SO4 reaction: origin of the
bistability in a CSTR. J . Phys. Chem. 1995, 99, 14365-14371.
(48) Lachmann, H.; Schnackerz, K. D. 31P nuclear magnetic resonance
titrations. Simultaneous evaluation of all pH-dependent reso-
nance signals. Org. Magn. Reson. 1984, 22, 101-105.
(49) Hamburger, A. W.; Salmon, S. E. Primary bioassay of human
myeloma stem cells. J . Clin. Invest. 1977, 60, 846-854.
(46) J oqueviel, C.; Gilard, V.; Martino, R.; Malet-Martino, M.; Ni-
emeyer, U. Urinary stability of carboxycyclophosphamide and
carboxyifosfamide, two major metabolites of the anticancer drugs
J M980587G