Improving the oral bioavailability of the iron chelator HBED by breaking the symmetry of the intramolecular H-bond network

Article abstract of DOI:10.1021/jm990261n

Physicochemical analysis and Monte Carlo simulations were used to identify structural features which prevent oral absorption of HBED, a potent iron chelator. In water the dominant conformations of HBED involve the hydrophobic collapse of the two aromatic

Full text of DOI:10.1021/jm990261n

J . Med. Chem. 2000, 43, 1467-1475
1467
Im p r ovin g th e Or a l Bioa va ila bility of th e Ir on Ch ela tor HBED by Br ea k in g th e
Sym m etr y of th e In tr a m olecu la r H-Bon d Netw or k †
Bernard Faller,* Carsten Spanka, Thomas Sergejew, and Vincenzo Tschinke
Research Department, Novartis Pharma AG, WKL-122.P.33, CH-4002 Basel, Switzerland
Received May 25, 1999
Physicochemical analysis and Monte Carlo simulations were used to identify structural features
which prevent oral absorption of HBED, a potent iron chelator. In water the dominant
conformations of HBED involve the hydrophobic collapse of the two aromatic rings. These
conformations are favored in polar media because they expose the polar phenolic hydroxy groups
to the solvent and partially shield the nonpolar aromatic rings. In a less polar solvent such as
chloroform, a symmetrical H-bond network between the carboxylates and the amines dominates
the conformational space. This leads to the exposure of the phenolic hydroxy groups to the
solvent, which is unfavorable for solvation. The low solubility of HBED in nonpolar solvents
was confirmed experimentally by determination of the partition coefficients in octanol,
chloroform, and cyclohexane and may explain the poor membrane permeability of this
compound. The high conformational stability which disfavors partitioning into phospholipids
is mainly due to the symmetrical H-bond network. Potentiometric titrations of a monoester of
HBED in MeOH/water indicate that the protonation sequence was changed compared to that
of the parent compound, suggesting that the symmetrical H-bond network was disrupted.
Conformational analysis in chloroform confirmed that, in contrast to HBED, no symmetric
interaction between the carboxylate and the nitrogen amines is possible in the half-ester and
a variety of conformations which allow partial shielding of the polar phenolic OH groups are
energetically possible. This theoretical model predicting a better solubility of the half-esters
in nonpolar solvents was supported by the large increase in the partition coefficients in octanol,
chloroform, and cyclohexane measured experimentally. The high absorbability predicted by
physicochemical and computer simulation methods was corroborated by in vivo experiments
in marmoset monkeys where the monoethyl ester derivative of HBED was well-absorbed orally
while the parent compound was nearly ineffective in the same model.
In tr od u ction
in rodents where the compound turned out to be
effective and safe.5 However, the good oral bioavailabil-
ity originally observed in rats could not be reproduced
in a primate model6 (Cebus monkey), a more relevant
animal model for human iron metabolism and drug
absorption. This disappointing result observed in pri-
mates was confirmed by a phase I clinical trial7,8 with
â-thalassemia patients where oral HBED was almost
ineffective compared to sc DFO, due to poor absorption
of the drug. Because of the otherwise attractive proper-
ties of HBED, many attempts to improve its oral
bioavailability have been undertaken. Pitt et al.9 made
a series of HBED esters by reacting the charged
carboxylate groups with simple alcohols. The dimethyl
ester derivative was extensively studied and was found
very potent in rodents but, again, failed to be effective
in Cebus monkeys6 apparently because the Cebus is not
able to hydrolyze HBED esters once taken up from the
gut.
Desferal (desferrioxamine B) has been used for the
treatment of iron overload for over 30 years1 and
remains the only drug generally available for this
purpose. However, despite its proven efficacy Desferal
suffers from its impractical mode of application due to
its short plasma half-life and poor oral bioavailability.
The poor compliance to the drug of many patients,
especially children, has underlined the necessity to
develop alternative chelators which are orally active and
safe.2,3 HBED is a potent, hexadentate ligand which was
first synthesized over 30 years ago.4 It strongly binds
ferric iron with an overall formation constant (log KML
)
of 40, 9 log units higher than that of desferrioxamine B
and several orders of magnitude higher than that of
serum transferrin, making this molecule a very potent
ligand for the chelation of ferric iron in vivo. Encourag-
ing results initially came from pharmacological studies
† Abbreviations: HBED (1), N,N′-bis(2-hydroxybenzyl)ethylenedi-
amine-N,N′-diacetic acid hydrochloride dihydrate; dm-HBED (2a ),
dimethyl N,N′-bis(2-hydroxybenzyl)ethylenediamine-N,N′-diacetate;
de-HBED (2b), diethyl N,N′-bis(2-hydroxybenzyl)ethylenediamine-
N,N′-diacetate; mm-HBED (3a ), sodium N,N′-bis(2-hydroxybenzyl)-
ethylenediamine-N-methoxycarbonylmethyl-N′-acetate; me-HBED (3b),
sodium N,N′-bis(2-hydroxybenzyl)ethylenediamine-N-ethoxycarbonyl-
methyl-N′-acetate; EDDA, ethylenediaminediacetic acid; EHPG, N,N′-
ethylenebis(o-hydroxyphenyl)glycine; po, per os; sc, subcutaneous.
* To whom correspondence should be addressed. Tel: +41 61
696 5406. Fax: +41 61 696 8663. E-mail: bernard.faller@
pharma.novartis.com.
This manuscript reports on the physicochemical prop-
erties of new HBED derivatives, which do not neces-
sitate enzymatic cleavage and which retain oral activity
in primates.
Resu lts
Ch em istr y. HBED monoalkyl esters 3 were prepared
by partial hydrolysis of the dialkyl esters 2 with 1 equiv
of sodium hydroxide in the corresponding alcohol. The
10.1021/jm990261n CCC: $19.00 © 2000 American Chemical Society
Published on Web 03/28/2000

1468 J ournal of Medicinal Chemistry, 2000, Vol. 43, No. 8
Faller et al.
Sch em e 1a
a
Reagents and conditions: (a) 10% HCl(g) in ROH, 50 °C, 18 h; (b) 1.0 equiv 2 M NaOH, ROH, 40 °C, 18 h.
Sch em e 2a
cosolvent of dielectric constant lower than water, equi-
libria with overall decreased charges are favored. Thus,
in the case of a weak acid the pKa will increase as the
dielectric constant of the titration medium is decreased.
Results are presented in Table 1.
With HBED, the aqueous pKa’s determined by po-
tentiometric titration compare very well with the values
reported earlier by L’Eplattenier4 (Table 2). The analysis
of the slope of the pKa vs % cosolvent plot indicates that
pKa3 and pKa4 correspond to the association of protons
with the basic amino groups. The same conclusion was
drawn by L’Eplattenier using another method (com-
parison of the pKa’s of HBED with related compounds,
EDDA and EHPG). The sequence of protonation of
HBED involves the two phenols first, followed by the
two amines and the two carboxylates. This sequence of
events strongly suggests the existence of zwitterions in
solution (Table 3).
With th e tw o d iester s 2a a n d 2b, the aqueous pKa’s
were obtained using a cosolvent approach. The presence
of the cosolvent (i) maintains the compound in solution
allowing potentiometric titration to be performed and
(ii) protects the esters from being hydrolyzed at alkaline
pH by shifting the equilibrium toward diester formation.
For this reason, MeOH was used with 2a and EtOH
with 2b. Comparison of the protonation constants with
those of HBED showed little difference for the phenols,
whereas the amino group pKa’s are significantly lower
in the diesters than in HBED itself.
With th e tw o m on oester s 3a a n d 3b, the aqueous
pKa’s could be determined directly by potentiometric
titration in 0.15 M KCl. Even at alkaline pH no ester
cleavage was observed during the titration indicating
that the ester bonds are relatively stable. Determination
of the pKa’s in media with different dielectric constants
was also performed to identify microspecies. This analy-
sis shows a different sequence of protonation compared
to HBED, with an inversion between the amino group
and the carboxylates (Table 3). This is the consequence
of the disrupted symmetry of the molecule as compared
to the diesters or to HBED itself.
a
Reagents and conditions: (a) 1 equiv FeCl3‚6H2O, 2 M HCl,
30 °C, 30 min, then 4 M NaOH, pH 10, 0 °C, 1 h; (b) 1 equiv
FeCl3‚6H2O, EtOH, 0 °C, 30 min, then 1 M NaOH pH 3, 2 h, rt.
desired compounds precipitate directly from the reaction
mixture and were isolated in the form of their mono-
sodium salts which are stable, nonhygroscopic solids.
The required HBED dialkyl esters 29 were synthesized
by hydrogen chloride-catalyzed esterification of HBED
(1) with alcohols.The iron complexes 4 and 5 were
prepared by treatment of the free ligands 1 and 3b with
1 equiv of iron(III) chloride hexahydrate followed by
neutralization with sodium hydroxide.
Ion iza tion Con sta n ts. HBED (1), mm-HBED (3a ),
and me-HBED (3b) are soluble in water within a large
pH range which allows direct measurement of the
majority of their aqueous ionization constants. The
water solubility of the diesters (2a and 2b) is so small
that direct potentiometric titration in water is not
possible. In these cases, titrations were performed in
the presence of different amounts of cosolvent (MeOH
and EtOH, respectively) and the aqueous pKa’s were
extrapolated using the method of Yasuda and Shed-
lowsky.10,11 The low pKa’s were determined by spectro-
photometric titration because of the poor performance
of the pH electrode under very acidic conditions (pH < 2).
Titrations in mixed-solvent media were also used to
identify microspecies present in solution. Although pH
metric titration cannot identify protonation sites in the
molecule, the variation of the ionization constants with
the dielectric constant can be used to distinguish
between acid and base functions.21 In the presence of a
P a r tition Coefficien ts. Ca lcu la ted log P o/w : Theo-
retical ocanol/water partition coefficients were gener-
ated using two different software packages. The ACD
log P (from Advanced Chemistry Development Inc.,
Canada) uses a structure-fragment approach. The
second program was developped in-house and calculates
partition coefficients using an atom-based approach
following the method proposed by Wakita et al.12 Data
are summarized in Table 4a.

Oral Bioavailability Improvement of HBED
J ournal of Medicinal Chemistry, 2000, Vol. 43, No. 8 1469
Ta ble 1. Ionization Constants for HBED and Its Methyl and Ethyl Ester Derivativesa
compd
cosolvent
dielectric constant
78
pKa1
pKa2
pKa3
pKa4
pKa5
pKa6
1
none
YS
12.90 (0.80)
11.10 (0.03)
nd
8.34 (0.02)
8.36
4.61 (0.01)
4.58
nd
nd
nd
1.62
0.92
MeOH
MeOH
MeOH
none
62
56
49
78
nd
nd
nd
nd
8.36 (0.01)
8.36 (0.01)
8.37 (0.01)
7.67 (0.01)
7.68
7.56 (0.01)
7.48 (0.01)
7.41 (0.01)
5.70
4.58 (0.02)
4.57 (0.01)
4.71 (0.01)
2.61 (0.01)
2.49
2.68 (0.01)
2.77 (0.01)
2.94 (0.01)
1.40
2.25 (0.05)
2.34 (0.02)
2.72 (0.02)
nd
1.05 (0.02)
1.11 (0.02)
1.18 (0.02)
12.53 (0.10)
nd
10.93 (0.01)
11.50
11.81 (0.02)
12.05 (0.02)
12.22 (0.02)
10.90
3a
2a
12.05 (0.05)
nd
nd
nd
nd
12.2
YS
1.30
MeOH
MeOH
MeOH
YS
62
56
49
1.15 (0.02)
0.97 (0.02)
0.90 (0.02)
MeOH
MeOH
MeOH
none
EtOH
YS
EtOH
EtOH
EtOH
62
56
49
78
63
12.60 (0.15)
12.91 (0.25)
nd
12.31 (0.30)
nd
12.20*
nd
nd
11.34 (0.02)
11.55 (0.02)
11.87 (0.06)
10.91 (0.05)
11.44 (0.05)
10.90*
5.58 (0.01)
5.46 (0.01)
5.40 (0.02)
7.64 (0.03)
7.50 (0.03)
5.70
5.45 (0.02)
5.25 (0.01)
5.09 (0.01)
1.10 (0.01)
0.82 (0.02)
0.69 (0.02)
2.91 (0.03)
2.75 (0.05)
1.80
1.42 (0.25)
1.21 (0.13)
1.05 (0.25)
3b
2b
0.90 (0.04)
nd
63
57
52
nd
11.55 (0.02)
11.80 (0.02)
nd
a
nd, not determined; *estimated by analogy with 2a ; YS, Yasuda-Shedlovsky extrapolation to zero cosolvent from mixed solvent
titrations. The best values for the aqueous pKa are indicated in bold. Standard deviations are indicated in parentheses.
Ta ble 2. Comparison of the Aqueous Ionization Constants for
HBED with Literature Values
Ta ble 4.
a: Partition Coefficients (in log units) in Different Solventsa
pKa1 pKa2 pKa3 pKa4 pKa5 pKa6
12.90 11.10 8.34 4.61 1.62. 0.92 0.15 M KCl this work
12.46 11.00 8.32 4.64 nd nd 0.1 M KNO3
remarks
ref
molecular fragment- atom-
species
based based
compd partition
calcd
calcd octanol
chloroform cyclohexane
1
1
3a
2a
(LH4)0
(LH3)0
(LH2)0
3.4
3.8
4.1
4.3
5.2
2.2 0.16 (0.06) <-2
2.2 0.50 (0.05) -0.29 (0.05) <-2
2.3 2.86 (0.03) nd
2.6 1.15 (0.05) 0.18 (0.03)
3.0 3.35 (0.05) 3.35 (0.05)
<-2
Ta ble 3. Assignment of Chemical Functions to Ionization
2.11 (0.06)
-1.30 (0.10)
3.22 (0.09)
Constantsa
3b (LH3)0
2b (LH2)0
compd
pKa1
pKa2
pKa3
pKa4
pKa5
1.62
pKa6
0.92
b: Hydrogen-Bonding Potentials
1
12.90 11.10 8.34 4.61
chem group P-OH P-OH
3a 12.05 10.93 7.67 2.61
chem group P-OH P-OH COOH
2a 12.20 10.90 5.70 1.40
chem group P-OH P-OH
3b 12.31 10.91 7.64 2.91
chem group P-OH P-OH COOH
2b 12.20 10.90 5.70 1.80
N
N
COOH COOH
1.30
N
H-bond
H-bond
H-bond
oral
compd
acidityb
basicityc
total
activity
N
1
>2.2
0.8
nd
1.0
0
nd
>1.7
nd
1.5
0.1
.2.2
>2.5
0.7
2.4
0.1
-
+/-
+
+
+/-
3a
2a
3b
2b
N
N
0.90
N
N
a
b
Standard deviations are indicated in parentheses. A rough
chem group P-OH P-OH
N
N
estimate of H-bond acidity was obtained using log Poct - log PCHCl
.
c A rough estimate of H-bond basicity potential was obtained usin3g
a
P-OH, phenolic OH; N, nitrogen amine; COOH, carboxylate.
log PCHCl - log PC H
.
3
6
12
Exp er im en ta l p a r tition coefficien ts: Dual-phase
titrations were used to measure the partition coefficients
in octanol, chloroform, and cyclohexane. Partition coef-
ficients reported in this work refer to macroconstants13
which means that the values refer to overall tautomeric
equilibria of the given species. Data are summarized in
Table 4a.
In 1-octa n ol, the two diesters 2a and 2b are much
more soluble than the three other compounds. The dm-
HBED (2a ) has a log P of 2.86, and the addition of two
carbons further increases the log P to 3.35 for the diethyl
ester (2b). In contrast to what was predicted in silico,
compounds 3a and 3b are significantly more soluble in
octanol than 1a (0.5 and 1.15 vs 0.16, respectively).
In ch lor ofor m , the solubility of compounds 3a and
3b is again higher than that of 1a but the difference
between the half-esters 3a and 3b and free HBED (1a )
is higher than in octanol.
Con for m a tion a l An a lysis. Monte Carlo calculations
were performed with solvent effects taken into account
for water and chloroform. The protonation constants
determined experimentally strongly suggest that HBED
and its esters exist as zwitterions in solution. We have
therefore chosen the shorter separation of the charges
between the carboxylate and amine groups on the two
esters (Figure 1). This study was designed to provide
insight on the changes undergone by the species as they
cross a water/lipid barrier. Since the species dominant
in the nonpolar phase are neutral and since these
species also exist in water, we restricted our analysis
to the neutral species, as the most likely to cross the
water/lipid barrier. Computer simulations were carried
out for HBED and its monomethyl and monoethyl
esters, as well as for its dimethyl ester.
Con for m a tion s in w a ter : In water all four species
(1-3) are predicted to be very flexible. The dominant
conformations present a hydrophobic collapse of the two
aromatic rings (Figure 2).
Con for m a tion s in ch lor ofor m : Monte Carlo simu-
lations in a less polar solvent suggest a different
behavior of HBED as opposed to its esters. HBED itself
In cycloh exa n e, only the two diesters have positive
log P values. The partitioning of both HBED and mm-
HBED (3a ) was very small (log P < -2) and could not
be measured. In this aprotic solvent me-HBED (3b) is
significantly more soluble than mm-HBED (3a ).

1470 J ournal of Medicinal Chemistry, 2000, Vol. 43, No. 8
Faller et al.
F igu r e 1. Two-dimensional structures with assigned proto-
nation patterns: (a) HBED (1), (b) mm-HBED (3a ), (c) me-
HBED (3b), (d) dm-HBED (2a ).
F igu r e 3. Low-energy conformations in chloroform: (a)
HBED (1), (b) mm-HBED (3a ), (c) me-HBED (3b), (d) dm-
HBED (2a ). In panels b and c, the thin lines indicate hydrogen
bonds between the phenolic OH groups and the carboxylate
group; in panel d, the thin lines indicate hydrogen bonds
between the phenolic OH groups and the ester CO groups.
Ta ble 5. Minimum and Maximum Computed Distancesa (Å)
between Aromatic Ringsb
compd
water
chloroform
1
4.0-10.4
3.8-9.6
3.8-9.9
3.8-9.3
8.5-10.3
4.4-9.5
4.1-8.9
4.1-9.9
3a
3b
2a
a
Obtained from the sets of Monte Carlo conformations within
4 kcal/mol of and including the lowest-energy conformation.
b Distance between the two centroids of the aromatic rings, defined
as the center of mass of the six ring carbon atoms.
F igu r e 2. Low-energy conformations in water: (a) HBED (1),
(b) me-HBED (3a ), (c) me-HBED (3b), (d) dm-HBED (2a ).
Kin etic Sta bility of m e-HBED-F e Ch ela te. me-
HBED-Fe (5) converts into HBED-Fe (4) in a process
induced by iron. The decay of 5 into 4 can be followed
easily since the two chelates can be separated by
extraction with dichloromethane. 4 is much more polar
than 5 and stays in the aqueous phase, while more than
99% of 5 is extracted in the organic solvent. The me-
HBED-Fe chelate was dissolved in pure DMSO at an
initial concentration of 5 mM. As a control, the stability
of 5 in DMSO was checked. In pure DMSO the chelate
was stable for at least 4 h at room temperature. At time
0, a small aliquot of the DMSO stock solution of Fe
chelate was diluted into a pH 7.40 HEPES buffer. After
various incubation times aliquots were taken and the
Fe chelates of 3b and 1 were separated by the addition
of an equal volume of dichloromethane. The time-
dependent decay of 5 and the concomitant formation of
4 were both well-described by first-order kinetics (Figure
strongly prefers conformations with the two aromatic
rings separated by a large distance (Figure 3a). These
conformations are stabilized by an internal H-bond
network between the protonated nitrogen atoms and the
carboxylate groups on opposite sites of the molecule.
By contrast, both 3a and 3b are flexible in chloroform.
They assume different conformations with shorter dis-
tances between the aromatic rings compared to HBED
(Figure 3b,c). The dimethyl ester can also assume a
variety of conformations with short distances between
the two aromatic rings. As expected, some conformations
possess a binary axis of rotation (Figure 3d). The shapes
of the mono- and diesters in chloroform are less compact
than in water but more compact than HBED itself in
water. A critical factor influencing the shape of the
molecules is the distance between the two aromatic
rings (Table 5).

Oral Bioavailability Improvement of HBED
J ournal of Medicinal Chemistry, 2000, Vol. 43, No. 8 1471
Ta ble 6. Induction of Fe Excretion in Marmoset Monkeysa
compd
sc
po
1
1586
nd
968
1782
nd
167
519
1127
1412
739
3a
2a
3b
2b
a
The values indicate the amount of induced Fe excretion in µg/
kg of body weight after po or sc administration of 150 µmol/kg of
test compound; nd, not determined.
illustrated in Table 6. Although HBED is effective when
delivered by the sc route, when given orally it is only
marginally active, as in humans,7,8 while the half-esters
were highly active after po application. A comparison
of the iron excretion values after po and sc application
(Table 6) indicates a high oral availability for 3b in our
primate model.
F igu r e 4. J ob plot at pH 7.40 for HBED (1) and me-HBED
(3b). The total ligand (chelator + metal ion) concentration was
fixed at 0.5 mM. Filled and open symbols indicate compounds
1 and 3b, respectively.
4). Nonlinear regression analysis gave rate constants
of 0.014 and 0.015 min-1 for the formation of 4 (filled
symbols) and the decay of 5 (open symbols), respectively.
The first-order conversion of 5 into 4 indicates that
the reaction is not a process in which Fe acts as a
catalyst but that one atom of Fe induces the conversion
of one molecule of 3b into one molecule of 1. This
conclusion was further substantiated in a second set of
experiments where different initial concentrations of 5
(0.05, 0.10, 0.15, and 0.2 mM) were converted into 4 with
similar rate constants (k ) 0.015 ( 0.003 min-1). In
conclusion, at pH 7.40 the half-time for the conversion
of 5 into 4 is about 47 min (t1/2 ) 0.7/k), this value being
independent of the initial concentration of 5.
Ir on -Bin d in g Stoich iom etr y a n d Sta bility of m e-
HBED-F e. The iron-binding stoichiometry was deter-
mined at pH 7.40 using the mole ratio variation method
also known as the J ob plot.28 For both 1 and 3b the peak
concentration of Fe chelate is reached at a mole ratio
of 0.50 (Figure 4), indicating that the binding metal ion/
ligand stoichiometry is 1:1 in both cases. A sharp peak
is indicative of a high affinity of the chelator for the
metal ion. Figure 4 shows that the absorbance vs mole
ratio curve for 3b is slightly more bell-shaped compared
to that of 1; a weaker binding affinity is therefore
anticipated with 3b.
The iron-binding affinity of 3b was measured by
spectrophotometric pH titration in KCl (0.15 M). Ligand
and ferric iron were reacted in the presence of different
H+ concentrations (10-100 mM) in 1-mL aliquots. The
spectra were recorded immediately after steady-state
was reached, usually within 1-2 min. The best theo-
retical absorbance vs pH curve was generated using a
simple model with only three species: ML and free M
and L, respectively (Supporting Information). The ti-
tration was repeated with two other ligand/metal ion
initial concentrations: 1.0/0.15 and 0.25/0.15 mM, re-
spectively. Similar log KML values were found (30.6 and
30.7).
In Vivo Activity. The oral availability of the chela-
tors was indirectly assessed by their pharmacodynamic
effect (induction of iron excretion) in iron-overloaded
animals. It is well-known that rodents differ from
primates in drug absorption; for example HBED is
orally effective in rats6 but not in humans.7,8 For these
reasons, the compounds were evaluated in a primate
model (Callithrix jacchus, commonly known as marmo-
set).
Discu ssion
In water, the dominant conformations of HBED show
a hydrophobic collapse of the two aromatic rings (Figure
2a). These conformations are favored in polar media
because they result in the exposure of the polar groups
to the solvent and partial shielding of the nonpolar
hydrophobic aromatic rings.
In a less polar medium such as chloroform, HBED
exhibits very different conformations. According to
Monte Carlo simulations in chloroform, HBED has the
tendency to assume conformations with the two aro-
matic rings separated by a large distance between the
centroids of the aromatic rings (Table 5 and Figure 3a).
These conformations are stabilized by internal hydrogen
bonds between the two amines and the two carboxylates.
The formation of hydrogen bonds across the two sides
of a flexible molecule result in a conformation with a
large distance between the two aromatic rings which
also shields the polar groups from the solvent. Our
conformational analysis predicts little conformational
variation in a nonpolar solvent mainly because any
conformational change disrupting the symmetrical H-
bond network between the carboxylates and the nitro-
gen amines is strongly penalized energetically. For the
same reason, the phenolic OH groups are not engaged
in hydrogen bonds and remain exposed to the solvent.
The exposure of these polar groups to the solvent results
in a very low solubility of HBED in nonpolar solvents,
as confirmed experimentally by the determination of the
partition coefficients in chloroform/water and cyclohex-
ane/water (both less than -2 in log units).
With the monomethyl (3a ) and monoethyl (3b) esters,
several opportunities to form internal hydrogen bonds
are possible among the polar groups. In low polarity
media such as chloroform these interactions result in a
high number of conformations which allow partial
shielding of the polar groups from the solvent. In
contrast to HBED no symmetric interaction between the
carboxylates and the nitrogen amines is possible. In
addition, the polar phenolic OH groups participate in
the H-bond network with other polar groups, particu-
larly with the charged carboxylate (see Figure 3b,c and
Table 6). As a result, a variety of conformations with
intermediate distances between the aromatic rings are
observed. The partial shielding through internal hydro-
gen bonds of all polar groups including the phenolic
The effects of the HBED half-esters (3a and 3b),
compared with the parent compound (HBED), are

1472 J ournal of Medicinal Chemistry, 2000, Vol. 43, No. 8
Faller et al.
OH’s should contribute to an increased solubility of the
monoesters (3) in nonpolar solvents. This theoretical
model is supported by the experimental data showing
a large increase in solubility of the monoesters in
chloroform and cyclohexane (see Table 4).
Computer simulations with 2a in chloroform gave
results similar to those obtained with 3a and 3b, the
polar phenolic OH groups being effectively shielded from
the solvent by intramolecular hydrogen bonds between
the tertiary amine and the ester groups.
In the second part of the discussion we want to show
which physicochemical parameters are responsible for
the different behavior of the monoesters compared to
HBED (1) with respect to their oral bioavailability in
marmoset monkeys. Lipophilicity and hydrogen bonds
are known to be important parameters influencing
membrane permeability and thus oral absorption.14,15
The lipophilicity of molecules is usually estimated via
determination of their octanol/water partition coef-
ficient. On the basis of calculated octanol/water partition
coefficients, one did not expect major absorption differ-
ences between 1a and compounds 3a and 3b as their
values differ by less than 1 log unit with both programs
used. When one switches to experimental octanol/water
partition coefficients where the conformational aspect
should be better modeled, the difference between the
values of 1a and 3b is about 1 log unit. Strong differ-
ences between the free HBED (1a ) and the half-esters
(3a and 3b) were only observed in chloroform and
cyclohexane. Therefore octanol was less predictive of
different po activities than chloroform or cyclohexane,
presumably because of its inadequate H-bond acidity
component.16
One way to estimate the hydrogen-bonding capability
of a molecule experimentally is to measure the partition
coefficients in solvents with different properties. In this
study we used octanol (amphiprotic), chloroform (proton
donor), and cyclohexane (aprotic). Total hydrogen-bond-
ing potential and hydrogen-bonding acidity and basicity
of the different compounds investigated are summarized
in Table 4b. Several investigators have shown that, to
cross a lipid bilayer, the compound needs to proceed
through a desolvation step in order to accommodate the
nonpolar environment of the inner layer of the phos-
pholipid membranes. The ability of a compound to
diffuse into a low-polarity environment can be estimated
by measuring its relative solubility in an aprotic solvent
like cyclohexane.
The data presented in Table 4b show that although
the total hydrogen-bonding potential is very high for free
HBED, esterification of one carboxylate translates into
a reduction of the H-bond acidity and basicity. One can
also notice a shift in ionization constants of 3a vs 1
(increase of the carboxylate pKa and decrease of the
amine pKa’s). The decrease in hydrogen bonds is ampli-
fied in me-HBED (3a ) with a total hydrogen-bonding
potential of only 1.62. One can also observe a further
increase in carboxylate pKa and a decrease in amine pKa
values compared to 3a . In contrast to HBED and 3a ,
3b falls in a ∆log Po/h range where good oral absorption
is expected.
than the monoester 3a following oral application. Two
possible explanations may account for this finding: (i)
decreased oral absoption possibly due to the low water
solubility which then becomes the limiting factor for
absorption17 and (ii) the diesters are not chelating
agents by themselves and require enzymatic cleavage
to be activated.
Con clu sion
We have shown that the me-HBED (3b) is able to
form 1:1 chelates with ferric iron and that its binding
affinity is sufficiently high to compete successfully with
endogenous ligands such as transferrin or citrate. In
addition, the resulting iron chelate is converted into
HBED-Fe in a nonenzymatic way.
Furthermore, Monte Carlo simulations could show
that disrupting the symmetry of the parent compound
(HBED) results in a major change in the H-bond
network within the molecule. In low-polarity media (i.e.
membrane phospholipid tails) the polar groups of the
mono- and diesters are neutralized since they are
engaged in internal hydrogen bonds. Measurement of
the relative solubility of the compounds in chloroform
and cylohexane gave experimental evidence in support
of conformations predicted by computer simulation and
allowed quantification of the H-bond networks. The
partition coefficients summarized in Table 4a could be
used to quantify the hydrogen-bonding capability of the
different molecules. Experimental data showed that
both H-bond acidity and basicity were decreased in me-
HBED (3b) compared to mm-HBED (3a ), the former
compound falling within a range where good oral
availability is expected. The data presented here do
correlate with the observations made in monkeys.
HBED half-esters represent an elegant way to gain oral
activity over the parent compound, and it is reasonable
to believe that it will be effective in humans as well, as
no enzymatic cleavage is required to release the active
compound.
Exp er im en ta l Section
Ma ter ia ls. HBED (1) was purchased as dihydrochloride
dihydrate from Strem Chemicals (Newburyport, MA). Larger
quantities can be prepared according to the procedure pub-
lished by L’Eplattenier et al.4
Meth od s. Syn th esis. Melting points were determined on
a Bu¨chi B-540 melting point apparatus and are uncorrected.
Thinlayer chromatography (TLC) was done on silica gel 60 F254
coated glass plates (Merck, Art. 5719), product spots were
visualized by illumination with UV light (254 nm) or by
staining with 0.1 M FeCl3 in 1 M HCl. Flash column chroma-
tography18 was performed using silica gel 60 (230-400 mesh;
Merck, Art. 9385). Purity of compounds was evaluated on a
Kontron HPLC system equipped with a 465 autosampler, a
322 pump, and a 430A detector using a NUCLEOSIL 100-5
C18 analytical column (250 mm × 4.6 mm; Macherey-Nagel,
Du¨ren, FRG): linear gradient of 0.1% TFA in MeCN/0.1% TFA
in H2O from 20:80 to 100:0 over 11 min and 100:0 for 5 min;
flow rate 1 mL/min, detection at 215 nm. Fast atom bombard-
ment mass spectra (FAB) were taken on a ZAB-HF instrument
(VG Analytical). Electrospray ionization mass spectra (ESI)
were measured on a VG platform spectrometer (Fisons Instru-
ments). 1H NMR (300 MHz) spectra were recorded on a Varian
GEMINI-300 spectrometer, samples were dissolved in either
DMSO-d6 or CDCl3. Chemical shifts are reported in ppm (δ)
using the residual solvent signal as reference (δ 7.26 for CDCl3
and δ 2.49 for DMSO-d6). The following abbreviations are used
to describe the peak patterns observed: br ) broad, s ) singlet,
The two diesters 2a and 2b are highly soluble in
cyclohexane. However, in vivo data (Table 6) show that
they are not more effective inducers of iron excretion

Oral Bioavailability Improvement of HBED
J ournal of Medicinal Chemistry, 2000, Vol. 43, No. 8 1473
d ) doublet, t ) triplet, q ) quartet, m ) multiplet; quotation
marks (e.g. ‘d’) are being used for higher-order signals.
Coupling constants (J ) are given in hertz (Hz). Elemental
analyses were performed by Organische Elemente Analytik,
Novartis Services AG, Switzerland.
N-CH2CO2Me) 3.38 (s, 2H, N-CH2Ar) 3.44 (s, 2H, N-CH2Ar),
3.58 (s, 3H, OMe), 6.60-6.75 (m, 5H, ArH), 6.92 (‘dd′, 1H, J )
1.1, 7.5, ArH), 7.07 (‘dq′, 2H, J ) 1.5, 7.7, ArH), 11.20 (br s,
2H, ArOH). Anal. (C21H25N2NaO6 [424.43]) C, H, N, Na+.
Sod iu m N,N ′-bis(2-h yd r oxyben zyl)eth ylen ed ia m in e-
N-eth oxyca r bon ylm eth yl-N ′-a ceta te (3b): colorless micro-
crystalline powder (3.77 g, 86%): mp 192-193 °C dec; TLC Rf
) 0.21 (DCM/MeOH/H2O/AcOH 90:10:1:0.5); HPLC tR ) 9.50
min; MS (+FAB) m/z 439 (MH+); 1H NMR (300 MHz, DMSO-
d6) δ 1.17 (t, 3H, J ) 7.20, OCH2CH3), 2.37 (t, 2H, J ) 6.4,
NCH2CH2N), 2.51 (t, 2H, J ) 6.4, NCH2CH2N), 2.79 (s, 2H,
NCH2CO2H), 3.33 (s, 2H, N-CH2CO2Et) 3.36 (s, 2H, N-CH2-
Ar) 3.45 (s, 2H, N-CH2Ar), 4.06 (q, 2H, OCH2CH3), 6.58-6.76
(m, 5H, ArH), 6.90 (‘dd′, 1H, J ) 1.0, 7.2, ArH), 7.05 (‘dq′, 2H,
J ) 1.4, 7.4, ArH), 11.25 (br s, 2H, ArOH). Anal. (C22H27N2-
NaO6 [438.46]) C, H, N, Na.
Sod iu m [[N,N ′-bis(2-h yd r oxyben zyl)eth ylen ed ia m in e-
N,N ′-d ia ceta to](3-)-N,N ′,O,O′,ON,ON ′]-fer r a te(1-), Na -
[F e(h bed )] (4). To a solution of iron(III) chloride hexahydrate
(5.41 g, 20 mmol) in 2 M hydrochloric acid (250 mL, 0.5 mol)
was added HBED hydrochloride dihydrate (1; 9.22 g, 20 mmol).
The suspension was stirred at 30 °C until all material went
in to solution. Then 4 M sodium hydroxide (150 mL, 0.6 mol)
was added with vigorous stirring and cooling in an ice bath.
After 1 h at 0 °C the precipitated product was filtered off,
washed with ice water (250 mL) and ice-cold methanol
(approximately 400 mL) until the filtrate became almost
colorless. Drying in a stream of air gave a reddish brown
powder (7.22 g, 59%): mp > 300 °C; TLC Rf ) 0.29 (DCM/
MeOH/H2O/AcOH 75:27:5:0.5), reddish spot; HPLC tR ) 5.92
min (very broad peak); MS (-ESI) m/z 440 ([56Fe(HBED)]-);
1H NMR (300 MHz, DMSO-d6) no signals, material is para-
magnetic. Anal. (C20H20FeN2NaO6‚2NaCl‚2H2O [616.15]) C, H,
N, Fe, Na, H2O.
[[N,N ′-Bis(2-h yd r oxyben zyl)eth ylen ed ia m in e-N-m eth -
oxyca r bon ylm eth yl-N ′-a ceta to](3-)-N,N ′,O,O′,ON,ON ′]-
fer r a te(0), [F e(m e-h bed )] (5). To an ice-cold suspension of
3b (877 mg, 2 mmol) in ethanol (100 mL) was added a solution
of iron(III) chloride hexahydrate (595 mg, 2.2 mmol) in ethanol
(50 mL). After completion of the addition the now deeply purple
colored solution was stirred for 30 min at 0 °C. By addition of
1 M NaOH the solution was adjusted to pH 3 and was stirred
for a further 2 h at room temperature. The solvent was distilled
off in a vacuum at room temperature. The black partly
crystalline residue was taken up in dichloromethane, washed
(1× buffer pH 7, 2× H2O, 1× brine), and dried (Na2SO4).
Precipitation of the iron complex by addition of dry ether gave
black fine crystals (600 mg, 64%): mp > 300 °C; TLC Rf )
0.20 (DCM/MeOH/H2O/AcOH 90:10:1:0.5), purple spot; HPLC
tR ) 9.78 min (complex partly dissociates: 25% free ligand
detectable); MS (+ESI) m/z 470 ([56FeH(me-HBED)]+); 1H
NMR (300 MHz, CDCl3) no signals, material is paramagnetic.
Anal. (C22H25FeN2O6 [469.30]) C, H, N, Fe.
Ion iza tion Con sta n ts. Potentiometric titrations were per-
formed in 0.15 M KCl at 25 °C with a PCA 101 automatic
titrator and the data analyzed with pKaLOGP for Windows
software 4.02 (Sirius Analytical Instruments, Forest Row,
U.K.). Spectrophotometric titrations were performed with a
Beckman DU-7400 diode array spectrophotometer. Before each
titration the pH electrode was first calibrated using a pH 7.00
phosphate buffer. Since all constants given in this report are
true thermodynamic constants, an additional step was used
to convert the operational pH scale (obtained from the mV
readings) to a concentration scale where pcH ) -log [H+].
Before each set of titrations, a blank titration was performed
and the pH electrode was standardized using a nonlinear
procedure based on the following semiempirical equation:19
N,N ′-Bis(2-h yd r oxyben zyl)eth ylen ed ia m in e-N,N ′-d i-
a cetic Acid Hyd r och lor id e Dih yd r a te, HBED (1). The
commercially available material was purified by the following
method: HBED dihydrochloride dihydrate (20 g, 40.2 mmol)
was dissolved in 2 M NaOH (250 mL), and insoluble matter
was removed by suction filtration. The slightly turbid pink
colored filtrate was acidified by careful addition of 2 M HCl
(pH 1.5). The desired product precipitated in form of fine
crystals. The suspension obtained was stirred for 3 h in an
ice bath and then stored at 4 °C overnight. The crystals
deposited were collected on a Buchner funnel, washed with
ice water (3 × 100 mL), and vacuum-dried at 40 °C for 18 h to
give 1 (15.95 g, 86%) as a light- and air-sensitive colorless
powder: mp 136-141 °C dec; TLC Rf ) 0.03 (DCM/MeOH/
1
H2O/AcOH 90:10:1:0.5); HPLC tR ) 7.50 min; H NMR (300
MHz, DMSO-d6) δ 3.21 (s, 4H, N-CH2-CH2-N), 3.66 (s, 4H,
N-CH2-CO2H), 4.05 (s, 4H, N-CH2-Ar), 6.78 (‘t’, 2H, J ) 7.50,
ArH), 6.94 (‘d’, 2H, J ) 8.25, ArH), 7.12-7.30 (m, 4H, ArH).
Anal. (C20H24N2O6‚HCl‚2H2O [460.92]) C, H, N, Cl-, H2O.
Gen er a l P r oced u r e for Acid -Ca ta lyzed Ester ifica tion
of HBED (1). The preparation of HBED dialkyl esters 2 was
performed according to the procedure published by Pitt et al.9
The yields could be significantly improved by utilization of
gaseous hydrogen chloride instead of thionyl chloride: HBED
(4.61 g, 10 mmol) was suspended in the selected alcohol (50
mL). Then HCl gas (approximately 5 g) was introduced in a
moderate stream with cooling in an ice bath. After this the
reaction mixture was heated to 50 °C for 18 h with stirring.
The solvent was stripped off in vacuo, the residue was taken
up in ethyl acetate (100 mL), washed (1× NaHCO3, 2× H2O,
1× brine), dried (MgSO4), and evaporated in vacuo. The crude
material was purified by flash chromatography (ethyl acetate/
hexanes 1:4) and recrystallization.
Dim eth yl N,N ′-bis(2-h yd r oxyben zyl)eth ylen ed ia m in e-
N,N ′-d ia ceta te (2a ): colorless prisms from hot methanol
(3.91 g, 94%); mp 93-96 °C; TLC Rf ) 0.26 (ethyl acetate/
hexanes 1:2); HPLC tR ) 9.87 min; 1H NMR (300 MHz, CDCl3)
δ 2.74 (s, 4H, N-CH2-CH2-N), 3.30 (s, 4H, N-CH2-CO2Me), 3.74
(s, 3H, CO2Me), 3.76 (s, 4H, N-CH2-Ar), 6.79 (‘dt′, 2H, J )
1.13, 7.36, ArH), 6.84 (‘dd′, 2H, J ) 0.95, 8.03, ArH), 6.92 (‘dd′,
2H, J ) 1.65, 7.48, ArH), 7.19 (‘dt′, 2H, J ) 1.71, 7.88, ArH),
9.50 (br s, 2H, ArOH). Anal. (C22H28N2O6 [416.48]) C, H, N.
Dieth yl N,N ′-bis(2-h yd r oxyben zyl)eth ylen ed ia m in e-
N,N ′-d ia ceta te (2b): colorless needles from hot ethanol (3.51
g, 79%); mp 77-78 °C; TLC Rf ) 0.40 (ethyl acetate/hexanes
1:2); HPLC tR ) 10.79 min; 1H NMR (300 MHz, CDCl3) δ 1.28
(t, 6H, J ) 7.14, OCH2CH3), 2.73 (s, 4H, N-CH2-CH2-N), 3.28
(s, 4H, N-CH2-CO2Et), 3.75 (s, 4H, N-CH2-Ar), 4.19 (q, 4H,
OCH2CH3), 6.77 (‘dt′, 2H, J ) 1.16, 7.36, ArH), 6.85 (‘dd′, 2H,
J ) 1.01, 8.06, ArH), 6.91 (‘dd′, 2H, J ) 1.65, 7.51, ArH), 7.19
(‘dt′, 2H, J ) 1.71, 7.97, ArH), 9.54 (br s, 2H, ArOH). Anal.
(C24H32N2O6 [444.53]) C, H, N.
Gen er a l P r oced u r e for P a r tia l Hyd r olysis of HBED
Diester s 2. To a solution of the HBED diester (10 mmol) in
the corresponding alcohol (100 mL) was added 2 M NaOH (5
mL, 10 mmol). The clear solution obtained was stirred at 40
°C for 18 h. During this time the sodium salt of the HBED
monoester precipitated in the form of very fine colorless
crystals. The reaction flask was placed in an ice bath and
stirred for further 4 h at 0 °C. The crystals formed were filtered
off, washed with ice-cold alcohol (3 × 50 mL), and dried in a
high vacuum at 50 °C for 24 h.
Sod iu m N,N ′-bis(2-h yd r oxyben zyl)eth ylen ed ia m in e-
N-m eth oxyca r bon ylm eth yl-N ′-a ceta te (3a ): colorless mi-
crocrystalline powder (2.16 g, 51%); mp 211-213 °C dec; TLC
pH ) R + SpcH + jH[H+] + jOHKw/[H+]
Rf ) 0.18 (DCM/MeOH/H2O/AcOH 90:10:1:0.5); HPLC tR
)
where Kw is the ionization constant for water, S is the Nernst
slope, jH and jOH are the asymmetry potentials (or junction
potentials) in the acid and alkaline directions, and R is a
constant. When a cosolvent was used, the pH electrode was
9.17 min; MS (+FAB) m/z 425 (MH+); 1H NMR (300 MHz,
DMSO-d6) δ 2.37 (t, 2H, J ) 5.6, NCH2CH2N), 2.52 (t, 2H, J
) 5.6, NCH2CH2N), 2.82 (s, 2H, NCH2CO2H), 3.36 (s, 2H,

1474 J ournal of Medicinal Chemistry, 2000, Vol. 43, No. 8
Faller et al.
(3) Olivieri, N. F.; Brittenham, G. Iron-chelating therapy and the
treatment of thalassemia. Blood 1997, 89(3), 739-761.
(4) L’Eplattenier, F.; Murase, I.; Martell, A. E. New multidentate
ligands. IV. Chelating tendencies of N, N′-di(2-hydroxybenzyl)-
ethylenediamine-N,N′-diacetic acid. J . Am. Chem. Soc. 1967, 89,
837-843.
standardized using the same procedure in the presence of a
defined amount of cosolvent and the corresponding Kw value
was taken from the literature. Argon was used to prevent
carbon dioxide uptake.
P a r tition Coefficien ts. Partition coefficients of the free
ligands were measured with the pH metric technique20,21 which
basically consists of two linked titrations: a normal titration
followed by a two-phase titration in the presence of the
partition solvent.
(5) Grady, R. W.; Hersko, C. HBED: a potential oral iron chelator.
Ann. N. Y. Acad. Sci. 1990, 612, 361-368.
(6) Peter, H. H.; Bergeron, R. J .; Streiff, R. R.; Wiegand, J . A
comparative evaluation of iron chelators in a primate model. In
The Development of Iron Chelators for Clinical Use; Bergeron,
R. J ., Brittenham, G. M., Eds.; CRC Press Inc.: Boca Raton,
1993; pp 373-394.
(7) Grady, R. W.; Salbe, A. D.; Hilgartner, M. W.; Giardina, P. J .
Preliminary results from a phase I clinical trial of HBED. In
The Development of Iron Chelators for Clinical Use; Bergeron,
R. J ., Brittenham, G. M., Eds.; CRC Press Inc.: Boca Raton,
1993; pp 395-406.
(8) Grady, R. W.; Salbe, A. D.; Hilgartner, M. W.; Giardina, P. J .
Results from phase I clinical trial of HBED. Adv. Exp. Med. Biol.
1994, 356, 351-359.
(9) Pitt, C. G.; Bao, Y.; Thompson, J .; Wani, M. C.; Rosenkrantz,
H.; Metterville, J . Esters and Lactones of Phenolic Amino
Carboxylic Acids: Prodrugs for Iron Chelation. J . Med. Chem.
1986, 29, 1231-1237.
(10) Yasuda, M. Dissociation constants of some carboxylic acids in
mixed aqueous solvents. Bull. Chem. Soc. J pn. 1959, 32, 429-
432.
(11) Shedlovsky, T.; Kay, R. L. The ionization constant of acetic acid
in water-methanol mixtures at 25 °C from conductance mea-
surements. J . Am. Chem. Soc. 1956, 60, 151-155.
(12) Wakita, K.; Yoshimoto, M.; Miyamato, H.; Watanabe, H. A
method for the calculation of the aqueous solubility of organic
compounds by using new fragment solubility constants. Chem.
Pharm. Bull. 1986, 34, 4663.
(13) Pagliara, A.; Carrupt, P.-A.; Caron, G.; Gaillard, P.; Testa, B.
Lipophilicity Profiles of Ampholytes. Chem. Rev. 1997, 97, 3385-
3400.
(14) Von Geldern, T. W.; Hoffman, D. J .; Kester, J . A.; Nellans, H.
N.; Dayton, B. D.; Calzadilla, S. V.; Marsh, K. C.; Hernandez,
L.; Chiou, W.; Dixon, D. B.; Wu-Wong, J . R.; Opgenorth, T. J .
Azole endothelin antagonists. 3. Using ∆logP as a tool to improve
absorption. J . Med. Chem. 1996, 39, 982-991.
(15) Young, R. C.; Mitchell, R. C.; Brown, T. H.; Ganellin, C. R.;
Griffiths, R.; J ones, M.; Rana, K. K.; Saunders, D.; Smith, I. R.;
Sore, N. E. Development of a new physicochemical model for
brain penetration and its application to the design of centrally
acting H2 receptor histamine antagonists. J . Med. Chem. 1988,
31, 657-671.
(16) Steyaert, G.; Lisa, G.; Gaillard, P.; Boss, G.; Reymond, F.;
Girault, H. H.; Carrupt, P.-A.; Testa, B. Intermolecular forces
expressed in 1,2-dichloroethane-water partition coefficients. J .
Chem. Soc., Faraday Trans. 1997, 93(3), 401-406.
(17) Amidon, G. L.; Lennernas, H.; Shah, V. P.; Crison, J . R. A
theoretical basis for a biopharmaceutic drug classification: the
correlation of in vitro drug product dissolution and in vivo
bioavailability. Pharm. Res. 1995, 12, 413-420.
(18) Still, W. C.; Kahn, M.; Mitra, A. J . Org. Chem. 1978, 43, 2923.
(19) Avdeef, A. Weighting scheme for regression analysis using pH
data from acid-base titration. Anal. Chim. Acta 1983, 148, 237-
244.
(20) Clarke, F. H.; Cahoon, N. M. Ionization constants by curve
fitting: determination of partition and distribution coefficients
of acids and bases and their ions. J . Pharm. Sci. 1987, 76, 611-
620.
The lipophilicity of the iron chelates was determined by
measuring their distribution coefficient at pH 7.40 in (50 mM
HEPES, 100 mM NaCl/n-octanol) using the shake-flask method.
Con for m a tion a l An a lyses. The conformational analysis
studies were performed using the Monte Carlo module of the
modeling package Macromodel22 version 4.0. Water and chlo-
roform solutions were simulated according to the method by
Hasel and Still.23,24 The conformations were generated by
random variations of all rotatable bonds in the different
molecules, followed by energy minimization with the AMBER25
force field. The protonation patterns were assigned based on
our analysis of the titration curves recorded in media with
different dielectric constants. All conformations within 4 kcal/
mol of the respective computed global minima were analyzed
within the module “Analysis” of the package InsightII26
(software to convert Macromodel multiconformational files to
MSI’s archive files by V.T., unpublished results). The analysis
was focused on a visual assessment of the shapes of the
different molecules and their H-bond patterns, as well as on
the graphic plots of the distances between the centroids of the
two phenoxy rings. Solvent-accessible surfaces and molecular
volumes were computed using the method by Huron and
Claverie.27 The representative conformations chosen for dis-
play in the figures of the present work are within 1 kcal/mol
of the respective computed global minima.
Ir on -Bin d in g Stoich iom etr y a n d Com p lex Sta bility.
The iron-binding stoichiometry was determined at pH 7.40
using the mole ratio variation method also known as the J ob
plot.28 The affinity of 3b (me-HBED) for ferric iron was
determined by spectrophotometric titration in 0.15 M KCl.
Computation procedures were done as described by Nagano
and Metzler29 using Excel solver.
Kin etic An a lyses. The conversion of 5 into 4 was followed
spectrophotometrically following extraction with dichloro-
methane. The organic phase contains me-HBED-Fe while
HBED-Fe stays in the aqueous phase. Kinetic analyses were
performed by nonlinear regression analysis with Enzfitter
(Biosoft, Cambridge, U.K.).
In Vivo Effica cy. The in vivo efficacy of the test compounds
was determined in primates using the marmoset (C. jacchus)
with an adaptation of the Cebus apella model of Bergeron et
al.30 Briefly, marmosets were iron-overloaded by injections of
iron dextran. For iron-balance studies, animals were kept in
metabolic cages an maintained on low-iron diet in order to
reduce fecal background. After compound administration the
excretion of iron in urine and feces was followed for 2 days.
Urinary iron was measured colorimetrically (bathophen-
antroline method) and fecal iron was determined by atomic
aborption spectroscopy.
(21) Avdeef, A.; Comer, J . E. A.; Thomson, S. J . pH-metric logP. 3.
Glass electrode calibration in methanol-water, applied to pKa
determination of water insoluble substances. Anal. Chem. 1993,
65, 42-49.
(22) Mohamadi, F.; Richards, N. G. J .; Guida W. C.; Liskamp, R.;
Lipton, M.; Caufield, C.; Chang, G.; Hendrickson, T.; Still, W.
C. MacroModel An Integrated Software System for Modeling
Organic and Bioorganic Molecules using Molecular Mechanics.
J . Comput. Chem. 1990, 11, 440-467.
(23) Hasel, W.; Hendrickson, T.; Still, W. C. A Rapid Approximation
to the Solvent Accessible Surface Areas of Atoms. Tetrahedron
Comp. Methodol. 1988, 1, 103-116.
Su p p or tin g In for m a tion Ava ila ble: Figures showing the
spectrophotometric titration of me-HBED and kinetics of
conversion of 5 into 4 at pH 7.4 and table showing the solvent-
accessible surface area in water of compounds 1, 3a , 3b, and
2a . This material is available free of charge via the Internet
at http://pubs.acs.org.
(24) Still, W. C.; Tempczyk, A.; Hawley, R. C.; Hendrickson, T.
Semianalytical Treatment of Solvation for Molecular Mechanics
and Dynamics. J . Am. Chem. Soc. 1990, 112, 6127-6129.
(25) Weiner, S. J .; Kollman, P. A.; Nguyen, D. T.; Case, D. A. An All
Atom Force Field for Simulations of Proteins and Nucleic Acids.
J . Comput. Chem. 1986, 7, 230-252.
Refer en ces
(1) Porter, J . B.; Huhens, E. R.; Hider, R. C. The development of
iron chelating drugs. Clin. Hematol. 1989, 2, 257-292.
(2) Hershko, C.; Link, G.; Tzahor, M.; Pinson, A. In The Development
of Iron Chelators for Clinical Use; Bergeron, R. J ., Brittenham,
G. M., Eds.; CRC Press: Boca Raton, Ann Arbor, London, Tokyo,
1994; 75pp.
(26) MSI. InsightII Molecular Modeling System (v. 95.0); MSI: San
Diego, CA, 1995.
(27) Huron, M.; Claverie, P. Calculation of Molecular Surfaces and
Volumes. J . Phys. Chem. 1972, 76, 2123-2133.

Oral Bioavailability Improvement of HBED
J ournal of Medicinal Chemistry, 2000, Vol. 43, No. 8 1475
(28) Huang, C. Y. Determination of binding stoichiometry by the
continuous variation method: the J ob plot. Methods Enzymol.
87, 509-525.
(29) Nagano, K.; Metzler, D. E. Machine computation of equilibrium
constants and plotting of spectra of individual ionic species in
the pyridoxal-alanine system. J . Am. Chem. Soc. 1967, 89, 2891,
2900.
(30) Bergeron, R. J .; Streiff, R. R.; Weigand, J .; Vinson, J . R. T.;
Luchetta, G.; Evans, K. M.; Peter, H. H.; J enny, H. B. A.
Comparative Evaluation of Iron Clearance Models. Ann. N. Y.
Acad. Sci. 1990, 612, 378-393.
J M990261N