Preparation and physicochemical characterization of a novel water-soluble prodrug of carbamazepine

Article abstract of DOI:10.1002/jps.21952

N-Acyl-urea derivatives of carbamazepine (CBZ) were synthesized through the reactions of iminostilbene with acyl-isocyanates to form N-glycyl-carbamazepine (N-Gly-CBZ, after a deprotection step) or N-acetyl-carbamazepine (N-acetyl-CBZ). N-Gly-CBZ was isolated as its water-soluble HCl salt and was designed to act as a prodrug and convert to CBZ and glycine in vivo by enzymatic cleavage of the acyl-urea bond. The stability pH-rate profiles for N-Gly-CBZ and N-acetyl-CBZ were determined. The stability of N-Gly-CBZ was found to range over four orders of magnitude with its greatest stability at pH 3-4 and a t ₉₀ value of 5.9 day at pH 4 at 25°C. From the fit of the pH rate profile two pK_a values were estimated to be 7.2 (terminal amine) and 10.0 (imide), which were independently verified using UV-visible spectroscopic analysis. The solubility of N-Gly-CBZ in aqueous solution was determined in the range of pH 5.5-7.5. The intrinsic solubility of the neutral form of the prodrug was found to be 4.4 mg/mL, and the solubility of the prodrug increased exponentially (log linear) as pH was decreased below its pK_a1 value. N-Gly-CBZ was found to have an aqueous solubility in excess of 50 mg/mL at pH 4. The presence of N-Gly-CBZ was found to increase the aqueous solubility of CBZ, a degradation product. CBZ showed an 8.6-fold greater solubility in an aqueous solution containing 23 mg/mL of N-Gly-CBZ than in water alone. The solubilization of CBZ by N-Gly-CBZ was investigated by examining the diffusion coefficients of the predominant species in D₂O and was found to be more consistent with stacking complex formation than micelle formation. The stability ofN-Gly-CBZ makes a ready-to-use parenteral formulation impractical, but a freeze-dried preparation for reconstitution appears to be feasible.

Full text of DOI:10.1002/jps.21952

Preparation and Physicochemical Characterization of a
Novel Water-Soluble Prodrug of Carbamazepine
JEFFREY N. HEMENWAY,1 PEKKA JARHO,1 JOHN T. HENRI,2 SAJIV K. NAIR,2 DAVID VANDERVELDE,1
GUNDA I. GEORG,2 VALENTINO J. STELLA1
1Department of Pharmaceutical Chemistry, University of Kansas, 2095 Constant Avenue, Lawrence, Kansas 66047
2Department of Medicinal Chemistry, University of Kansas, Lawrence, Kansas 66045
Received 6 February 2009; revised 13 August 2009; accepted 20 August 2009
Published online 22 September 2009 in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/jps.21952
ABSTRACT: N-Acyl-urea derivatives of carbamazepine (CBZ) were synthesized
through the reactions of iminostilbene with acyl-isocyanates to form N-glycyl-carba-
mazepine (N-Gly-CBZ, after a deprotection step) or N-acetyl-carbamazepine (N-acetyl-
CBZ). N-Gly-CBZ was isolated as its water-soluble HCl salt and was designed to act as a
prodrug and convert to CBZ and glycine in vivo by enzymatic cleavage of the acyl-urea
bond. The stability pH-rate profiles for N-Gly-CBZ and N-acetyl-CBZ were determined.
The stability of N-Gly-CBZ was found to range over four orders of magnitude with its
greatest stability at pH 3–4 and a t90 value of 5.9 day at pH 4 at 258C. From the fit of the
pH rate profile two pKa values were estimated to be 7.2 (terminal amine) and 10.0
(imide), which were independently verified using UV–visible spectroscopic analysis. The
solubility of N-Gly-CBZ in aqueous solution was determined in the range of pH 5.5–7.5.
The intrinsic solubility of the neutral form of the prodrug was found to be 4.4 mg/mL, and
the solubility of the prodrug increased exponentially (log linear) as pH was decreased
below its pKa1 value. N-Gly-CBZ was found to have an aqueous solubility in excess of
50 mg/mL at pH 4. The presence of N-Gly-CBZ was found to increase the aqueous
solubility of CBZ, a degradation product. CBZ showed an 8.6-fold greater solubility in an
aqueous solution containing 23 mg/mL of N-Gly-CBZ than in water alone. The solubi-
lization of CBZ by N-Gly-CBZ was investigated by examining the diffusion coefficients of
the predominant species in D2O and was found to be more consistent with stacking
complex formation than micelle formation. The stability of N-Gly-CBZ makes a ready-to-
use parenteral formulation impractical, but a freeze-dried preparation for reconstitution
appears to be feasible. ß 2009 Wiley-Liss, Inc. and the American Pharmacists Association
J Pharm Sci 99:1810–1825, 2010
Keywords: stability; solubility; carbamazepine; prodrug; pH-rate profile; glycine
INTRODUCTION
the treatment of partial and generalized tonic–
clonic epileptic seizures.1,2 It is also commonly
prescribed for trigeminal neuralgia and bipolar
depression.3–5 CBZ has a low aqueous solubility
The anticonvulsant carbamazepine (CBZ) is one of
the most frequently prescribed first-line drugs for
Jeffrey N. Hemenway’s present address is Pharmaceutics
R&D, Bristol-Myers Squibb, 1 Squibb Drive, New Brunswick,
NJ 08901.
David VanderVelde’s present address is Division of
Chemistry and Chemical Engineering, California Institute of
Technology, Pasadena, CA.
Pekka Jarho’s present address is Department of Pharma-
ceutical Chemistry, University of Kuopio, Kuopio, Finland.
John T. Henri’s present address is 3549 Larkspur Drive,
Longmont, CO 80503.
Gunda I. Georg’s present address is Department of
Medicinal Chemistry, University of Minnesota, MN 55414.
Correspondence to: Valentino J. Stella (Telephone: 785-864-
3755; Fax: 785-864-5736; E-mail: stella@ku.edu)
Sajiv K. Nair’s present address is Medicinal Chemistry,
Pfizer Global R & D, San Diego, CA 92121.
Journal of Pharmaceutical Sciences, Vol. 99, 1810–1825 (2010)
ß 2009 Wiley-Liss, Inc. and the American Pharmacists Association
1810
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 99, NO. 4, APRIL 2010

NOVEL WATER-SOLUBLE PRODRUG OF CARBAMAZEPINE
1811
(120 mg/mL),6 which seriously hampers parent-
eral administration of the drug. The drug
substance is a polymorphic compound with at
least four known forms and a dihydrate,7–10 which
show different aqueous solubilities and dissolu-
tion rates, of which the dihydrate being the least
soluble and slowest to dissolve.8 Anhydrous CBZ
rapidly transforms into the dihydrate form when
dispersed in water,11,12 and the rate and extent of
oral absorption of CBZ in humans has been shown
to be dissolution rate limited.13 The poor solubility
of CBZ can be overcome by the use of high
concentrations of cosolvents and/or surface active
agents,14–18 but these are not recommended for IV
formulations due to the risk of severe allergic
reaction. It has been shown that formulation-
related toxicities can occur from these types of
preparations.19–24 A water-soluble prodrug of CBZ
could be administered intravenously in a safe
and effective formulation and oral administration
could give rise to faster, less erratic and more
complete oral absorption.
Scheme 1. Acyl-urea prodrug strategy for water-
soluble prodrugs of CBZ. N-Gly-CBZ is hypothesized
to be a water-soluble prodrug of CBZ, which will convert
to CBZ and glycine in vivo.
lysis in vivo. Amino acid promoieties are also
favorable because they release a naturally occur-
ring potentially nontoxic entity upon bioreversion.
MATERIALS AND METHODS
Chemicals
Prodrugs are bioreversibly modified derivatives
of drug molecules, which have more desirable
physicochemical properties and/or improved
delivery characteristics, which undergo rapid
enzymatic or chemical cleavage in vivo to release
the parent drug. In the case of CBZ, the property
to be overcome is poor aqueous solubility. The
work presented here includes the synthesis and
evaluation of the physicochemical properties of
an N-acyl a-amino acid derivative of CBZ as a
possible water-soluble prodrug for parenteral and
oral use. This prodrug strategy involves the
attachment of an amino acid to the urea nitrogen
of the parent drug forming a peptide like acyl-urea
functionality. The amine group of the attached
promoiety can be ionized to form a water-soluble
salt. It was initially hypothesized that an N-amino
acid derivative of a urea-containing drug will
rapidly convert to the parent urea drug and the
respective amino acid in vivo by enzymatic
cleavage of the acyl-urea bond as illustrated in
Scheme 1. Amino acids have been extensively
investigated for their use as promoieties to
improve the aqueous solubility of various amine
and alcohol containing drugs.25–32 Amino acid
amides have shown good chemical stability and
enzymatic lability in vivo.30,33–35 N-amino acid
ureas have chemical structures that are suffi-
ciently similar to amino acid amides that it is
reasonable to expect that peptidase and other
nonspecific enzymes will recognize this peptide
like acyl-urea prodrug as a substrate for hydro-
Acetamide, acetyl chloride, carbamazepine, N,N0-
dicyclohexyl carbodiimide, di-tert-butyl dicarbo-
nate, glycine, hydrochloric acid (1.0 M in diethyl
ether), iminostilbene, methyl hydrazine, 4-nitro-
phthalic anhydride, oxalyl chloride, p-toluene
sulfonic acid, and triethyl orthoformate, were
purchased from Aldrich Chemical Company
(Milwaukee, WI). Methyl sulfoxide and sodium
sulfide nonahydrate were purchased from Sigma-
Aldrich (St. Louis, MO). The solvents, 1,2-
dichloroethane and anhydrous tetrahydrofuran
(THF) were purchased from Fisher Scientific
(Pittsburgh, PA). Dry THF refers to tetrahydro-
furan (from anhydrous) that was freshly distilled
under argon atmosphere over sodium metal and
benzophenone. All other chemicals were reagent
and/or analytical grade commercially available
products.
Synthesis and Characterization/Chemistry
Synthetic reactions were monitored where possi-
ble by template liquid chromatography. Flash
chromatography was performed using silica gel
(40–63 mm particle size) purchased from Fisher
1
Scientific. H NMR and 13C NMR analysis was
performed on a Bruker Avance 400 instrument
and the chemical shifts are reported in parts per
million. Mass spectroscopy (ESþ) was performed
by The University of Kansas Mass Spectroscopy
Laboratory.
DOI 10.1002/jps
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 99, NO. 4, APRIL 2010

1812
HEMENWAY ET AL.
N-Glycyl-Carbamazepine (N-Gly-CBZ)
44.5 mmol), ammonium bicarbonate (3.518 g,
44.5 mmol), and pyridine (0.498 g, 6.18 mmol)
were added to the reaction vessel and stirred for
12 h. The reaction mixture was added to 200 mL of
5% HCl in water and extracted with three 100 mL
portions of ethyl acetate. The organic phases were
combined, dried over sodium sulfate, and con-
centrated under reduced pressure. Hexane was
added to the organic layer to precipitate com-
pound 2 as a white solid (95%): 1H NMR (DMSO-
d6) d 4.20 (s, 2H), 7.32 (br s, 1H), 7.74 (br s, 1H),
8.16 (d, 1H, J ¼ 8.13 Hz), 8.53 (d, 1H, J ¼ 1.84),
The synthetic preparation of N-Gly-CBZ is illu-
strated in Scheme 2. 4-Nitrophthalic anhydride
(10.0 g, 51.8 mmol) was added to glycine (3.89 g,
51.8 mmol) without solvent and heated to 1908C
in an oil bath for 1 h. The brown residue was
dissolved in methanol, filtered through a medium
glass filter frit, and crystallized from methanol/
water to yield compound 1 as a white solid (87%):
1H NMR (acetone-d6) d 4.52 (s, 2H), 8.22 (d, 1H,
J ¼ 8.17 Hz), 8.63 (d, 1H, J ¼ 1.98 Hz), 8.77 (dd,
1
1
1H, J ¼ 2.01 Hz, J ¼ 8.17 Hz). H NMR (DMSO) d
8.65 (dd, 1H, J ¼ 8.14 Hz, J ¼ 1.95 Hz). H NMR
4.39 (s, 2H), 8.19 (d, 1H, J ¼ 8.16 Hz) 8.56 (d, 1H,
J ¼ 1.94Hz) 8.67 (dd, 1H, J ¼ 2.02, J ¼ 8.15 Hz).
Compound 1 (8.00 g, 34.2 mmol) was dissolved in
ethyl acetate. Di-tert-butyl dicarbonate (9.71 g,
(acetone-d6) d 4.41 (s, 2H), 6.69 (br s, 1H), 7.31
(br s, 1H), 8.19 (d, 1H), 8.60 (s, 1H), 8.73 (d, 1H).
13C NMR (acetone-d6) d 40.71, 118.48, 125.00,
129.99, 134.04, 137.15, 152.40, 166.02, 166.27,
Scheme 2. Synthesis of N-Gly-CBZ. Reagents and conditions: (i) 1908C, fusion.
(ii) 1. BOC2O, acetone, RT. 2. NH4HCO3, cat. amount of pyridine. (iii) (COCl)2 DCE
reflux, Ar. (iv) iminostilbene, THF, RT. (v) Na2SÁ9H2O, THF/dioxane/H2O, 08C. (vi) DCC,
THF, 08C, Ar. (vii) CH3NHNH2, THF, À808C, Ar. (viii) HCl, ether/acetone.
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 99, NO. 4, APRIL 2010
DOI 10.1002/jps

NOVEL WATER-SOLUBLE PRODRUG OF CARBAMAZEPINE
1813
167.78. Compound 2 (3.00 g, 12.9 mmol) was
refluxed in 1,2-dichloroethane with oxalyl chlor-
ide (16.3 g 129 mmol) under argon to form an acyl
isocyanate,36 which was not isolated. The solvent
and excess oxalyl chloride were distilled off under
argon and further removed under vacuum. The
residue was dissolved in dry THF, and a solution
of iminostilbene (4.97 g, 25.8 mmol) in dry THF
was added in situ under argon and stirred at RT
for 4 h. The reaction mixture was concentrated
under reduced pressure and purified by flash
chromatography to yield 3 as a white solid (81%):
1H NMR (CDCl3) d 5.08 (s, 2H), 7.04, (brm, 3H,
NH þ 2CH), 7.38-7.57 (brm, 8H), 8.06 (d, 1H,
J ¼ 8.40), 8.60 (dd, 1H, J ¼ 8.40, J ¼ 1.40), 8.69
(d, 1H, J ¼ 1.40). 1H NMR (DMSO-d6) d 4.72
(s, 2H), 7.07 (s, 2H), 7.41 (t, 2H, J ¼ 7.65), 7.49–
7.55 (m, 4H), 7.64 (d, 2H, J ¼ 7.65), 8.19 (d, 1H,
J ¼ 8.82), 8.58 (d, 1H, J ¼ 1.47), 8.66 (dd, 1H, J ¼
8.28, J ¼ 1.47). 13C NMR (DCCl3) d 43.31, 119.36,
125.14, 129.71, 130–131 (8 Ar C), 133.98, 136.98,
151.67, 151.97, 166.04, 168.40. MS (ESþ) m/z
469.1 (M þ H)þ.
from the first step was taken up in dry THF under
argon and cooled to 08C. N,N0-dicyclohexyl carbo-
diimide (DCC, 0.45 g, 2.18 mmol) was dissolved in
dichloromethane under argon and cannulated
into the reaction vessel and allowed to stir at
08C for 6 h, presumably forming the phthalisoi-
mide. The reaction was cooled to À808C under
argon and methyl hydrazine (0.098 g, 2.14 mmol)
was added in situ to give the free amine, which
was not isolated. The reaction was filtered under
argon to remove N,N0-dicyclohexylurea, a spar-
ingly soluble side product of the recyclization
reaction and the solvent was removed under
reduced pressure. The free amine and 4-nitro-N-
methylphthalylhydrazide, the side product of the
methyl hydrazine reaction, were taken up in
anhydrous acetone. One molar HCl in ether
(2.78 mL, 2.78 mmol) was added to the acetone
solution. Upon standing at room temperature for
24 h, the product N-Gly-CBZ crystallized as the
hydrochloride salt. The salt was then filtered,
washed with acetone and dried under vacuum.
The overall yield of this deprotection sequence
1
Deprotection of the 4-nitrophthaloyl protecting
group using methyl hydrazine resulted in the
formation of carbamazepine rather than the
desired product. Deprotection of phthaloyl pro-
tecting groups from base sensitive compounds
reported by Kukolja and Lammert37 involved a
three-step process in which the reactivity of the
imide functionality was enhanced by conversion to
an isoimide, which was then removed by methyl
hydrazine. In the first step, compound 3 (1.00 g,
2.14 mmol) was dissolved in 5 mL of THF, 3 mL
dioxane, and 2 mL H2O. The solution was cooled
to 08C, sodium sulfide nonahydrate (0.512 g,
2.14 mmol) was added and the reaction mixture
was stirred at 08C for 2 h. The reaction mixture
was concentrated under reduced pressure and
the remaining residue was dissolved in 20 mL
of H2O. Twenty-two milliliters of 0.1 N HCl was
added to the aqueous layer, which caused a cloudy
white precipitation. The aqueous layer was
extracted with three 10 mL portions of ethyl
acetate, and the disappearance of the cloudy
precipitation was observed. The organic layer was
dried over NaSO4, filtered through a medium
glass filter frit, and the solvent was removed
under vacuum. What is presumed to be the
phthalamic acid shown in Scheme 2 was not
further purified but carried directly to the next
step. The remaining steps of the deprotection
sequence were carried out under argon and not
exposed to atmospheric moisture. The dry residue
was 46% as a white solid. H NMR (DMSO-d6) d
4.00 (s, 2H), 5.05 (s, 2H), 7.38 (td m, 2H,
Jt ¼ 7.36 Hz, Jd ¼ 0.84 Hz), 7.49 (br m, 4H), 7.57
(d, 2H, J ¼ 8.04 Hz), 8.23 (br s, 3H), 9.69 (br s, 1H).
Addition of one drop of D2O to the DMSO-d6
sample results in disappearance of peaks at 8.23
and 9.69 ppm. 1H NMR (D2O) d 4.00 (q, 2H,
J ¼ 4.80 Hz), 6.99 (s, 2H), 7.46 (m, 8H). HRMS
(TOF ESþ) m/z calc. C17H16N3O2 (M þ H)þ
294.1243, found 294.1233.
N-Acetyl-Carbamazepine (N-Acetyl-CBZ)
Acetamide (0.100 g, 2.32 mmol) was refluxed in
1,2-dichloroethane with oxalyl chloride (0.295 g,
2.32 mmol) under argon for 9 h to form an acyl
isocyanate (Scheme 3), which was not isolated. A
Scheme 3. Synthesis of N-Acetyl-CBZ. Reagents and
conditions: (i) (COCl)2 THF reflux, Ar. (ii) iminostilbene,
THF, RT.
DOI 10.1002/jps
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 99, NO. 4, APRIL 2010

1814
HEMENWAY ET AL.
solution of iminostilbene (0.448 g, 2.32 mmol) in
THF was added in situ under argon and stirred at
RT for 4 h. The reaction mixture was concentrated
under reduced pressure and purified by flash
chromatography (ethyl acetate/methylene chlor-
ide) to give N-acetyl-CBZ as a white solid (68%):
1H NMR (CDCl3) d 2.46 (s, 3H), 6.98 (s, 2H), 7.18
(s, 1H), 7.45 (m, 8H). The addition of one drop
of D2O to the CDCl3 sample resulted in disap-
pearance of the peak at 7.18 ppm.
quadrupole mass spectrometer interfaced with
electrospray ionization (ESI). A C18 ODS Hypercil
(150 Â 4.6 mm2 i.d., 5 mm) HPLC-column was used
for separation with an injection volume of 20 mL
and isocratic flow rate of 1.0 mL/min. The mobile
phase consisted of 42.5% pure water and 57.5%
acetonitrile with the eluate divided between the
UV detector monitoring absorption at 280 nm and
the mass spectrometer. Retention time (tR) values
were 3.5 min for CBZ and 2.5 min for the unknown
degradation product.
pKa Determinations by UV–Visible Spectroscopy
Solutions of N-Gly-CBZ (50 mM) were prepared in
25 mM phosphate buffers with pH values in the
range of 5–12 and ionic strength adjusted to 0.15
with NaCl. UV–visible spectroscopy analysis was
performed using a Varian Cary 50 Bio UV–Visible
Spectrophotometer in the range of wavelengths
between 200 and 400 nm to determine the pKa
values of N-Gly-CBZ by a method similar to that
reported by Zalipsky et al.38 The greatest change
in the spectra with respect to pH was observed at
the wavelength of 250 nm and the spectral change
at this wavelength was used to determine pKa
values.
Aqueous Stability of Acyl-Urea Prodrugs
The stability characteristics of N-Gly-CBZ and
N-acetyl-CBZ were studied in buffered solutions
of various pH values at temperatures of 25, 50,
and/or 708C. Aqueous buffer for the pH value of 3
was prepared using HCl. Acetate buffers (25 mM)
were used for pH range of 4–5, and phosphate
buffers (25 or 50 mM) were used for the pH range
of pH 6–11. Samples for pH value 12 were made
using NaOH. The ionic strength of all buffers used
for stability was adjusted to 0.15 with NaCl.
Stability experiments were initiated by adding
approximately 0.10 mg of the experimental com-
pound to 10 mL buffered solutions. Reaction
vessels were placed in a shaking water bath
for storage conditions of 25.08C or a constant
temperature cabinet for storage conditions of 50
and 708C. Samples were drawn from stability
vessels at appropriate intervals and directly
injected into the HPLC instrument without
dilution or any other preparation. The observed
pseudo-first-order rate constants (kobs) for the
degradation reactions were determined from the
linear semi logarithmic plots of remaining pro-
drug versus time.
HPLC and LC/MS/MS Analysis
All separations were conducted on a C18 ODS
Hypercil (150 Â 4.6 mm2 i.d., 5 mm) HPLC-column
using an injection volume of 20 mL and isocratic
flow rate of 1.0 mL/min with eluate monitored by a
UV absorption detector at 280 nm (lmax of CBZ).
Mobile phase for the stability indicating HPLC
assay for N-Gly-CBZ consisted of 42.5% 20 mM
phosphate buffer, pH 5 and 57.5% acetonitrile
and retention time (tR) values were 5.2 min for
N-Gly-CBZ, 2.9 min for CBZ and 11.0 min for
iminostilbene. Mobile phase for the stability
indicating HPLC assay for N-acetyl-CBZ con-
sisted of 45% 20 mM phosphate buffer, pH 6 and
55% acetonitrile and retention time (tR) values
were 3.8 min for N-acetyl-CBZ, 2.9 min for CBZ
and 14.0 min for iminostilbene. External stan-
dards were used to create calibration curves to
quantify N-Gly-CBZ, N-acetyl-CBZ, and CBZ
for all stability and solubility assays. LC/MS/MS
was used for the identification of an unknown
degradation product of N-Gly-CBZ. Analysis was
performed using a Waters Alliance 2690 HPLC
system and a Micromass Quattro Micro tandem
Solubility Studies
An excess amount of the compound of interest
was added to 25 mM phosphate buffer (m ¼ 0.15)
of known pH. Suspensions were shaken in a
water bath at 258C for 24 h to achieve equili-
bration. Samples were then centrifuged at
12,400 RPM for 10 min, and 100 mL aliquots of
supernatant were diluted and analyzed by HPLC.
The pH of the undiluted supernatant was
measured to obtain the final pH of the solubility
samples.
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 99, NO. 4, APRIL 2010
DOI 10.1002/jps

NOVEL WATER-SOLUBLE PRODRUG OF CARBAMAZEPINE
1815
Pulse Gradient NMR Determination of
Diffusion Coefficients
25.08C. The degradation of N-Gly-CBZ exhibited
pseudo-first-order kinetics and the observed rate
constants (kobs) for the loss of N-Gly-CBZ were
determined from the plot of the natural logarithm
of concentration of N-Gly-CBZ remaining versus
time, where kobs is equal to the negative slope of
the linear fit. Buffer catalysis for the aqueous
degradation of N-Gly-CBZ was examined by
comparing the observed first order rate constants
for the loss of N-Gly-CBZ in aqueous solutions
with pH values of 4, 7, and 10 with buffer
concentrations of 25 and 50 mM and ionic strength
adjusted to 0.15. The kobs values for these buffer
concentrations are listed in Table 1. Buffer
concentration effects appear to become more
significant at higher pH values but an extensive
analysis of buffer effects were not performed. The
prodrug was most stable in lower pH range where
the buffer effects seem to be least significant.
Scheme 4 shows a proposed model for potential
hydrolytic reactions of the various ionic species of
N-Gly-CBZ deemed necessary to describe the
pH-rate profile. The reactions shown in Scheme 4
include water hydrolysis of the protonated form of
N-Gly-CBZ (k0), water hydrolysis of the neutral
form of N-Gly-CBZ ðk00Þ, and base catalyzed
hydrolysis of the neutral species ðk0OHÞ. Kinetically
equivalent reaction pathways could be invoked.
The equation for observed rate constant (kobs) in
terms of rate and equilibrium constants as a
function of hydrogen ion concentration is given by
Eq. (2).
The 1H NMR signal attenuation for a single
diffusing species in the presence of a single pair of
pulsed field gradients is given by Eq. (1).39–41
2
I ¼ I0exp½ÀDðggdÞ ðD À d0=3ފ
(1)
The parameters in Eq. (1) are defined as follows:
I is the signal intensity, I0, the intensity at zero
gradient, D, the diffusion coefficient, D, the delay
time (time for diffusion), d0, the correction for
finite gradient, g, the gyromagnetic ratio, g, the
amplitude of applied signal, and d, the duration of
signal. Diffusion coefficients can be determined
by a semi-log plot of signal intensity (I) versus
gradient field strength [(ggd)2(D–d0/3)], which
yields a linear plot with a slope of ÀD.40
NMR diffusion coefficients were determined
for concentrated (10 mg/mL) and dilute (2 mg/mL)
samples of N-Gly-CBZ in D2O. Diffusion co-
efficients were also determined for the same
concentrations of N-Gly-CBZ saturated with
CBZ. Diffusion experiments were performed on
a Bruker Avance 400 instrument using a 4-nuc.
probe. The bipolar longitudinal eddy delay se-
quence from the Johnson Laboratory was applied.42
The diffusion delay was fixed at 200 ms. Sine-
shaped z-gradient pulses ranging in amplitude
from 5% to 50% and 2.5 ms in length were
employed in the experiments. Ten experimental
points were obtained and the diffusion coefficient
was extracted from the slope of the semi-log plot of
signal intensity versus gradient field strength.
2
k0½HþŠ þ k00 ½HþŠKa1 þ k0OHKa1Kw
kobs
¼
(2)
2
½HþŠ þ ½HþŠKa1 þ Ka1Ka2
RESULTS
The pH-rate profile for the aqueous degradation
of N-Gly-CBZ at 258C is shown in Figure 1. The
stability of N-Gly-CBZ was found to range over
four orders of magnitude, with its greatest stab-
ility at pH 3 and decreasing stability as the pH
Synthesis of N-Gly-CBZ and N-Acetyl-CBZ
The synthetic procedures to prepare N-Gly-CBZ
and N-acetyl-CBZ were dictated by earlier fail-
ures to directly couple various activated acyl
derivatives to CBZ. In the case of N-Gly-CBZ,
other N-protected glycine isocyanate derivatives
were attempted without success. Although the
simple phthalimide (vs. the 4-nitrophthalamide)
derivative was successfully prepared, one could
not successfully deprotect this derivative to give
N-Gly-CBZ.
Table 1. Effects of 25 and 50 mM Buffer
Concentration on the Observed Pseudo-First-Order
Rate Constants for the Loss of N-Gly-CBZ in Aqueous
Solutions With pH Values of 4, 7, and 10
pH
kobs (sÀ1),
kobs (sÀ1),
Value
25 mM buffer
50 mM buffer
2.1 Â 10À7
5.6 Â 10À6
1.5 Â 10À4
2.6 Â 10À7
7.2 Â 10À6
2.5 Â 10À4
Stability Characteristics of N-Gly-CBZ
4
7
10
The aqueous stability of N-Gly-CBZ was deter-
mined for pH values in the range of 3–12 at
DOI 10.1002/jps
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 99, NO. 4, APRIL 2010

1816
HEMENWAY ET AL.
Scheme 4. Model for potential hydrolytic reactions of the various ionic species of
N-Gly-CBZ. The reactions include water hydrolysis of the protonated form (k0), water
hydrolysis of the neutral form ðk00Þ, and base catalyzed hydrolysis of the neutral species
ðk0OHÞ.
increases. The solid line in Figure 1 represents the
fit of the pH-rate profile to Eq. (2). From the fit of
the pH-rate profile, the two apparent pKa values
of N-Gly-CBZ were calculated to be 7.2 and 10.0.
The values for the rate constants obtained from
the fit of the pH-rate profile to Eq. (2) were: k0 ¼
in the ranges of 3–4. An initially unknown minor
degradation product was formed in stability
samples with pH values of 8, 9, 11, and 12 and
was identified as the major degradation product
at pH 10. LC/MS/MS was used to identify this
product. The LC retention time of the earlier
eluting unknown degradation product is sharp in
nature compared to the amine containing N-Gly-
CBZ peak, which is broad and exhibits tailing. The
MS scan of the chromatographically separated
unknown LC peak revealed a base peak of m/z ¼
316 along with m/z peaks of 332, 294, 227, 249,
220, and 194. The 294 ESþ peak has the same m/z
ratio as the M þ H of N-Gly-CBZ, indicating that it
is likely to result from a rearrangement reaction.
The 316 ESþ peak is expected to result from the
sodium adduct of the rearrangement product and
the 332 ESþ peak is expected to result from the
potassium adduct. The sodium and potassium
1.50 Â 10À7 sÀ1, k00 ¼ 1:42 Â 10À5sÀ1, and kO0 H
¼
1:15 Â 10À2sÀ1MÀ1
.
Degradation Products of N-Gly-CBZ
Products from the aqueous degradation of N-Gly-
CBZ are shown in Scheme 5. The major degrada-
tion product from N-Gly-CBZ in aqueous solution
was found to be CBZ for all pH values except for
around pH 10. Iminostilbene was formed to a
lesser extent in stability samples with pH values
Figure 1. pH rate profile (uncorrected for buffer
effects) of N-Gly-CBZ at 258C. The solid line represents
the fit of the pH rate profile to Eq. (1). The values for
the rate constants obtained from the fit of the pH rate
profile are: k0 ¼ 1.50 Â 10À7 sÀ1, k00 ¼ 1:42 Â 10À5sÀ1
,
Scheme 5. Degradation products for N-Gly-CBZ in
the pH range of 3–12. CBZ is the major degradation
product at all pH values except pH 10.
and k0OH ¼ 1:15 Â 10À2sÀ1MÀ1 and the pKa values
obtained from the fit are 7.2 and 10.0.
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 99, NO. 4, APRIL 2010
DOI 10.1002/jps

NOVEL WATER-SOLUBLE PRODRUG OF CARBAMAZEPINE
1817
adduct phenomenon is common with ESI meth-
ods.43–45 The MS scan of the chromatographically
separated CBZ peak showed similar sodium and
potassium adducts. The MS/MS daughter peaks of
294 ESþ were found to be m/z ¼ 277, 249, 220, 194,
and 192, which are consistent with the m/z ratios
of expected ESþ fragments of the proposed
rearrangement product shown in Scheme 5.
Rate constants and half-lives for the aqueous
degradation of N-Gly-CBZ are listed in Table 2,
where kobs represents the rate and t1/2 the half-life
for of loss of N-Gly-CBZ for each pH value.
Formation rate constants for the appearance of
CBZ and the rearrangement product are also
listed. kCBZ represents the calculated rate of
formation of CBZ, kRAP is the formation rate
constant of the rearrangement product, and kIMN
is the formation constant for iminostilbene. The
rate constants for the formation of products from
parallel degradation reactions of N-Gly-CBZ were
calculated from the product peak areas at the
completion of the individual stability experi-
ments. For parallel first order reactions, the
rate law is expressed by Eq. (3) and the ratio of
the concentration of products formed is equal to
the ratio of the rate constants k1 and k2 for parallel
reactions according to Eq. (4).46
HPLC-UV peak area can be used in place of
concentration values in with Eq. (4). It is reason-
able to assume that a neutral organic compound
with a very similar chemical structure to CBZ
and iminostilbene would have a nearly identical
extinction coefficient. This assumption is further
supported by the difference in the response factors
for N-Gly-CBZ and CBZ (1.2%) and that the total
peak area was conserved throughout the stability
experiments within the variance of the assay
(2.7%).
Temperature Dependence of N-Gly-CBZ Stability
The pH of maximum stability for N-Gly-CBZ was
found to be in the acidic pH range. This also
corresponds to the pH range of greatest solubility
for primary amines such as N-Gly-CBZ. The pH
value of 4.0 appears to be the most suitable pH for
parenteral formulation. The half-life of N-Gly-
CBZ in pH 4 solution at 258C is 35 days, which
corresponds to a t90 value of 5.9 days, although the
actual product shelf-life is expected to be limited
by precipitation of CBZ from degradation of
the prodrug (see later discussion). The rate of
degradation of N-Gly-CBZ in aqueous solution
with a pH value of 4.0 at 508C was 7.14 Â 10À6 sÀ1
.
½AŠ ¼ ½AŠ0exp À ½ðk1 þ k2ÞtŠ
(3)
The activation energy (Ea) value calculated from
Arrhenius behavior of the hydrolysis of N-Gly-
CBZ at 25 and 508C in aqueous solution with a pH
value of 4.0 is 123 kJ/mol (29.3 kcal/mol). This Ea
value can be used to calculate the predicted rate of
degradation at refrigerated temperature (48C).
Under refrigerated conditions, the kobs value is
predicted to be 4.54 Â 10À9 sÀ1, and the corre-
sponding t90 value is 249 days.
½BŠ k1
¼
(4)
½CŠ k2
If the assumption is made that the extinction
coefficients of the rearrangement product and
iminostilbene are not significantly different,
Table 2. Various Parameters Describing the Aqueous Degradation of N-Gly-CBZ at 258C
pH Value
kobs (sÀ1
)
t1/2 (h)
kCBZ (sÀ1
)
kRAP (sÀ1
)
kIMN (sÀ1
)
3
4
5
6
7
8
9
10
11
12
1.16 Â 10À7
2.07 Â 10À7
4.30 Â 10À7
1.28 Â 10À6
5.62 Â 10À6
1.99 Â 10À5
3.34 Â 10À5
1.512 Â 10À4
2.19 Â 10À4
2.16 Â 10À4
1510
847
408
9.81 Â 10À8
1.93 Â 10À7
4.30 Â 10À7
1.28 Â 10À6
5.62 Â 10À6
1.62 Â 10À5
2.08 Â 10À5
6.86 Â 10À5
1.54 Â 10À4
1.58 Â 10À4
—
—
—
—
—
1.78 Â 10À8
1.34 Â 10À8
—
—
—
—
—
—
—
—
137
31.2
8.81
5.25
1.16
0.80
0.81
3.76 Â 10À6
1.26 Â 10À5
8.25 Â 10À5
6.42 Â 10À5
5.81 Â 10À5
kobs and and t1/2 represent the rate of loss and the half-live of N-Gly-CBZ respectively. kCBZ represents the calculated rate of
formation of CBZ, kRAP is the formation rate constant of the rearrangement product, and kIMN is the formation constant for
iminostilbene.
DOI 10.1002/jps
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 99, NO. 4, APRIL 2010

1818
HEMENWAY ET AL.
Stability of N-Acetyl-CBZ
To help understand the relatively rapid degrada-
tion kinetics of N-Gly-CBZ, the degradation of
the model compound N-acetyl-CBZ was examined.
Clearly, the intramolecular reaction seen with
N-Gly-CBZ at around pH 10 is not possible for
N-acetyl-CBZ and the effect of the protonated
amine group in N-Gly-CBZ can be determined.
The stability characteristics of N-acetyl-CBZ were
determined for pH values 4, 5, 6, 7, 8, 9, 10, 11, and
12 at 708C. The degradation reactions exhibited
pseudo-first-order kinetics and the observed rate
constants (kobs) for the loss of N-acetyl-CBZ were
determined from the plot of the natural logarithm
of concentration of N-acetyl-CBZ remaining ver-
sus time. Iminostilbene is the major product
formed at pH values 4, 5, and 6, and CBZ is the
major product formed at pH values 7–12.
Scheme 6 shows a model for potential hydrolytic
reactions with respect to the ionic equilibrium of
N-acetyl-CBZ. The reactions shown in Scheme 6
include water hydrolysis (k0) and base catalyzed
hydrolysis (kOH) of the neutral species. The
equation for the observed rate constant (kobs) in
terms of individual rate and the ionic equilibrium
constant as a function of hydrogen ion concentra-
tion, or pH where [Hþ] ¼ 10ÀpH is described by
Eq. (5).
Figure 2. pH rate profile (uncorrected for buffer
effects) of N-acetyl-CBZ at 708C. The solid line repre-
sents the fit of the pH rate profile to Eq. (5). The values
for the rate constants obtained from the fit of the
pH rate profile are: k0 ¼ 1.53 Â 10À7 sÀ1, kOH ¼ 4.20 Â
10À2 sÀ1 MÀ1, and the pKa value is 10.0.
CBZ. The pH of maximum stability of N-acetyl-
CBZ is in the acidic range near pH 4–5 and the
rate of loss of N-acetyl-CBZ was determined in
aqueous solution at 508C and pH 5 to give a kobs
value of 5.05 Â 10À7 sÀ1. The Ea value calculated
from Arrhenius behavior of the hydrolysis of
N-acetyl-CBZ in aqueous solution with a pH value
of 5.0 at 50 and 708C is 129 kJ/mol (30.9 kcal/mol).
The predicted kobs value for the degradation of
N-acetyl-CBZ in aqueous solution with a pH value
of 5.0 at 258C is 8.92 Â 10À8 sÀ1, which corresponds
to a half life (t1/2) value of 899 days compared to
19 days for N-Gly-CBZ under similar conditions.
k0½HþŠ þ kOHKw
kobs
¼
(5)
½HþŠ þ Ka
Figure 2 shows the pH-rate profile of N-acetyl-
CBZ at 708C, in which the solid line represents
the fit of the pH-rate profile to Eq. (5). The values
for the rate constants obtained from the fit of the
pKa Values of N-Gly-CBZ
pH-rate profile were: k0 ¼ 1.53 Â 10À7 sÀ1, kOH
¼
4.20 Â 10À2 sÀ1 MÀ1, and the pKa value was 10.0, a
value very similar to that estimated for N-Gly-
The pKa values of N-Gly-CBZ were determined
independent of the kinetics by UV–visible spectro-
scopy. Absorbance values were determined for
solutions of N-Gly-CBZ (50 mM) in 25 mM phos-
phate buffers with pH values in the range of 5–12
and ionic strength adjusted to 0.15 with NaCl.
The most significant change in the spectra with
respect to pH was observed at the wavelength of
250 nm. The spectral change at this wavelength
was used to determine the pKa values of the
prodrug by fitting the UV absorbance at 250 nm
to Eq. (6).
Scheme 6. Model for potential hydrolytic reactions
with respect to the ionic equilibrium of N-acetyl-CBZ.
The reactions include water hydrolysis (k0) and base
catalyzed hydrolysis ðk0OHÞ of the neutral species.
2
A½HþŠ þ BKa1½HþŠ þ CKa1Ka2
Abs ¼
(6)
2
½HþŠ þ Ka1½HþŠ þ Ka1Ka2
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 99, NO. 4, APRIL 2010
DOI 10.1002/jps

NOVEL WATER-SOLUBLE PRODRUG OF CARBAMAZEPINE
1819
The plot of UV absorbance at 250 nm versus pH
is shown in Figure 3. The solid line represents
the fit of the data points to Eq. (6), and the values
obtained for pKa1 and pKa2 were 7.42 and 10.2,
respectively.
Aqueous Solubility of N-Gly-CBZ at 25-C
The aqueous solubility of N-Gly-CBZ was deter-
mined for the pH range of 5.5–7.5. Figure 4 shows
the plot of the log S versus pH of N-Gly-CBZ where
S is the total solubility of N-Gly-CBZ. The
intrinsic solubility of the neutral form of N-Gly-
CBZ was found to be 4.4 mg/mL and the solubility
of the prodrug increases, as expected, as pH
decreases below the lowest pKa value of N-Gly-
CBZ. The aqueous solubility of N-Gly-CBZ was
projected to exceed 50 mg/mL at pH 4, which
appears to be the most suitable pH for a potential
parenteral formulation.
Figure 4. Plot of log Stot N-Gly-CBZ versus pH where
Stot N-Gly-CBZ is the total solubility of N-Gly-CBZ. The
intrinsic solubility of the neutral form of N-Gly-CBZ
was found to be 4.4 mg/mL, and the solubility increases
log linearly with decrease in pH below the pKa.
CBZ and concentration of N-Gly-CBZ appears to
be linear over the range studied and CBZ showed
an 8.6-fold greater solubility in an aqueous
solution containing 23 mg/mL of N-Gly-CBZ than
in water alone. To account for this increased
solubility, an effort was made to see if N-Gly-CBZ
formed associated species that could account for
this solubility enhancement.
Aqueous Solubility of CBZ in Concentrated
Solutions of N-Gly-CBZ
It was found that the solubility of CBZ itself
increases with increased concentration of N-Gly-
CBZ. The solubility experiments were equili-
brated for 24 h and since anhydrous CBZ is known
to rapidly transforms into the dihydrate form
when dispersed in water,11,12 it is assumed that
the major solid state CBZ entity in suspension
was the dihydrate. Figure 5 shows the solubility
of CBZ in increasingly concentrated solutions of
N-Gly-CBZ. The relationship between solubility of
Pulse Gradient NMR Determination of
Diffusion Coefficients
Diffusion coefficients of species in solutions of
N-Gly-CBZ in D2O (and in the presence of CBZ)
Figure 3. UV–visible absorption at wavelength of
250 nm for solutions of 50 mM N-Gly-CBZ in 25 mM
phosphate buffers in the pH range of 5–12. The solid
line represents the fit of the data points to Eq. (6).
Figure 5. Solubility of CBZ (SCBZ) versus concentra-
tion of N-Gly-CBZ (CN-Gly-CBZ). CBZ showed an 8.6-fold
greater solubility in an aqueous solution containing
23 mg/mL of N-Gly-CBZ than in water alone.
DOI 10.1002/jps
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 99, NO. 4, APRIL 2010

1820
HEMENWAY ET AL.
were determined by pulse gradient NMR studies.
The diffusion coefficients were determined by a
semi-log plot of signal intensity (I) versus gradient
field strength [(ggd)2(D–d0/3)], which yields a linear
plot with a slope of ÀD.40 Figure 6 shows the plots
obtained for concentrated (10 mg/mL) and a more
dilute (2 mg/mL) samples of N-Gly-CBZ in D2O,
and Figure 7 shows plots for the same concentra-
tions of N-Gly-CBZ in D2O and saturated with
CBZ prior to analysis.
DISCUSSION
Figure 7. Pulse gradient NMR studies of solutions of
N-Gly-CBZ saturated with CBZ in D2O. Plots of Inten-
sity (I) versus gradient field strength [(ggd)2(D–d0/3)] for
concentrated (&, 10 mg/mL) and dilute (&, 2 mg/mL)
samples of N-Gly-CBZ in D2O saturated with CBZ.
N-Gly-CBZ was shown to be extremely soluble in
the lower pH range. The pKa values of N-Gly-CBZ
determined by UV–visible absorption were 7.42
and 10.0. These values agree well with the pKa
values estimated from the fit of the pH-rate profile
(7.21 and 10.0). It is interesting to note that the
larger change in the UV absorbance at 250 nm
(Fig. 3) is associated with the pKa value of 10.0
and is likely the pKa value associated with the
ionization of the acyl-urea N–H group, which is
closest to the chromophore. This data provides
support that the predominant form at pH values
between the two pKa values is the neutral specie
and not the zwitterion.
mL for the free base form of N-Gly-CBZ seems
excessive based on its structural elements but
it may be that the base is in an amorphous,
noncrystalline, state. A significant increase in the
solubility of CBZ itself was found in the presence
of increasing concentration of N-Gly-CBZ (Fig. 5).
The solubility-pH profile was not easily fit to
the expected pH-solubility equation for a simple
amine. If N-Gly-CBZ were to be studied further,
refinements to the solubility curve and better
solid-state characterization of all solid species
will be needed. Assuming the solubility increase
effect remains linear at higher concentrations of
prodrug, a CBZ solubility increase of more than
23.6-fold (3.67 mg/mL compared to 0.155 mg/mL)
would be predicted from extrapolation of the
available data to an N-Gly-CBZ concentration of
69.7 mg/mL (0.2116 M, 50 mg/mL CBZ equivalent)
prodrug solution. Unfortunately, this could not be
experimentally verified due to limited availability
of the experimental compound. The two most
likely mechanisms by which the solubility of
species in aqueous solution can be increased as
a function of concentration are micellar forma-
tion47,48 and other forms of self-association.49,50
The solubilization phenomenon of CBZ by N-Gly-
CBZ was investigated by examining the diffusion
coefficients of the predominant species in solu-
tions containing CBZ and/or N-Gly-CBZ. The
diffusion coefficients determined from the slope
values from Figures 6 and 7 are shown in Table 3.
There is little change in the diffusion coefficients
of species in concentrated and a more dilute
The apparent intrinsic aqueous solubility of
N-Gly-CBZ was about 4 mg/mL and solubility
increases log linearly below the 7.4 pKa value
resulting in greatest aqueous solubility in the
acidic pH range. The intrinsic solubility of 4.4 mg/
Figure 6. Pulse gradient NMR studies of solutions of
N-Gly-CBZ in D2O. Plots of intensity (I) versus gradient
field strength [(ggd)2(D–d0/3)] for concentrated (*,
10 mg/mL) and dilute (*, 2 mg/mL) samples of N-Gly-
CBZ in D2O.
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 99, NO. 4, APRIL 2010
DOI 10.1002/jps

NOVEL WATER-SOLUBLE PRODRUG OF CARBAMAZEPINE
1821
Table 3. Diffusion Coefficients (D), Hydrodynamic Radiuses (rs), and Mass Ratios of Concentrated (10 mg/mL, Mc)
and More Dilute (2.0 mg/mL, Md) of Species in Solutions of N-Gly-CBZ and CBZ in D2O
D (cm2 sÀ1
)
rs (A)
Mc/Md
˚
Sample ID
Concentrated (10 mg/mL) N-Gly-CBZ
Dilute (2 mg/mL) N-Gly-CBZ
Concentrated (10 mg/mL) N-Gly-CBZ saturated w/CBZ
Dilute (2 mg/mL) N-Gly-CBZ saturated w/CBZ
4.61 Â 10À6
5.04 Â 10À6
3.97 Â 10À6
4.73 Â 10À6
5.3
4.9
6.2
5.8
1.30
1.69
solution of N-Gly-CBZ with and without CBZ.
Eq. (7) shows the relationship between diffusion
(D) and temperature (T), viscosity (h), and
hydrodynamic radius (rs). k is the Boltzmann
constant.
The solutions saturated with CBZ also show
greater hydrodynamic radii, which suggest the
possibility of sandwich complex formation
through p–p stacking interactions of the conju-
gated planar ring systems of N-Gly-CBZ and CBZ.
N-Gly-CBZ exhibits a maximum aqueous sta-
bility in the pH range where injectable solutions of
>50 mg/mL CBZ equivalent (projected, 69.7 mg/
mL) are possible. The pH-rate profile for N-Gly-
CBZ is consistent with the pH-rate profiles for
most activated amide type compounds, with
the pH of maximum stability in the acidic range
and undergoing base catalyzed hydrolysis. The
maximum stability of N-Gly-CBZ, however, was
significantly less than expected. The N-amino acid
urea functionality of N-Gly-CBZ is very similar to
N-amino acid amide prodrugs of aromatic amines,
which have shown very good chemical stability,
with predicted t90 values of 6–7 years.30,33
Comparison of the N-glycine derivative of CBZ
to the N-glycine prodrug of the aromatic amine
dapsone,31 which exhibits very good aqueous
chemical stability (predicted t90 >2 years at pH
4 and 258C compared to 5.9 days for N-Gly-CBZ),
reveals the significance of the structural differ-
ences of the parent compounds on the stability of
their N-amino acid derivatives. The urea of N-Gly-
CBZ possesses a much greater electron with-
drawing capability than the aromatic amine of
dapsone, making CBZ a better leaving group and
there is a potential for intramolecular catalysis to
contribute significantly to the aqueous instability
of N-Gly-CBZ. N-Gly-CBZ has the potential to
undergo keto-enol tautomerization as illustrated
in Scheme 7. The enol/enolate form of acyl-urea
compounds such as N-Gly-CBZ could make a
significant contribution to the reactivity of the
acyl-urea bond toward hydrolysis. The stability
of the model compound N-acetyl-CBZ was eval-
uated to determine the effect of the electron
withdrawing amine group associated with N-Gly-
CBZ apart from the potential keto-enol tautomer-
ization shared by the two structurally similar
acyl-urea compounds. Accelerated stability stu-
kT
D ¼
(7)
6phrs
Values for the hydrodynamic radius for the
species in each solution are listed in Table 3,
using the viscosity of water and the assumption
that viscosity is constant in all solutions. These
calculations show hydrodynamic radii on the order
of angstroms, which is consistent with self-
association and not indicative of supramolecular
interactions such as micelle formation. Eq. (8)
reported by Yao et al.40 show the relationship of
molecular mass (M) of species in solution to their
diffusion coefficient.
ꢀ
ꢁ
3
kT
4pNA
M ¼
(8)
6phFD 3ðn2 þ d1n2Þ
F is the shape factor (Perrin factor), and the y
constants are the partial specific volumes (mole-
cule1, water2), NA is Avogadro’s number, and d1
represents the hydration number (bound water).
The absolute values of these constants need not be
known to determine the ratio of molecular masses
of dilute and concentrated solutions of N-Gly-CBZ
in D2O. The molecular mass of the species in
solution is inversely proportional to the cubed
diffusion coefficient (Eq. 9), allowing the calcula-
tion of a ratio of the molecular mass values for
species in the dilute and concentrated solutions of
N-Gly-CBZ (Tab. 3).
1
D3
M /
(9)
Interestingly, the ratio of molecular masses for
dilute and concentrated solutions of N-Gly-CBZ
in D2O show a significantly greater difference in
the presence of CBZ (1.69) than without (1.30).
DOI 10.1002/jps
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 99, NO. 4, APRIL 2010

1822
HEMENWAY ET AL.
The LC/MS/MS data strongly suggest that the
additional degradation product seen around pH
10 and alluded to earlier is the rearrangement
product shown in Scheme 5, which presumably
results from intramolecular aminolysis of the
urea functionality of N-Gly-CBZ (Scheme 8). This
result also provides further evidence that the
intermediate ionic species of N-Gly-CBZ shown in
Scheme 4 exists predominantly as the neutral
species over the zwitterion. It would be unlikely
that the protonated amine group of a zwitterionic
species would be able to participate in the
nucleophilic attack associated with an intramo-
lecular aminolysis reaction. Similarly, the anionic
species shown in Scheme 4 would not be expected
to undergo the cyclization reaction to form the
rearrangement product because of decreased
electrophilicity of both carbonyl carbons from
the electron donation effect of the negative charge
associated with the imide. The plateau of the
observed rate constant for loss of N-Gly-CBZ
above pH 10 results from the parallel reactions of
the base catalyzed hydrolysis and the rearrange-
ment reaction of the neutral species (also appears
to be base catalyzed and therefore kinetically
equivalent) as shown by the product formation
rate constants for pH values 11 and 12 in Table 2.
Under refrigerated conditions, the kobs value for
the loss of N-Gly-CBZ is predicted to be 4.54 Â
10À9 sÀ1, and the corresponding t90 value is
249 days. This predicted stability limitation
is sufficient to prevent a ready-to-use aqueous
injectable product for N-Gly-CBZ. Furthermore,
the actual shelf-life of a concentrated solution of
N-Gly-CBZ is likely not to be limited by a loss of
the prodrug, but rather by the possible precipita-
tion of CBZ formed from the aqueous degradation
of N-Gly-CBZ. Based on the predicted solubility
increase observed for CBZ in concentrated solu-
tions of N-Gly-CBZ, a series of successive approx-
imations show the expected solubility limit of
CBZ produced from a 50 mg/mL CBZ equivalent
Scheme 7. Potential keto-enol tautomerization of
acyl-urea amino acid compounds.
dies of N-acetyl-CBZ predict a half-life of 899 days
for N-acetyl-CBZ compared to 19 days for N-Gly-
CBZ at pH 5 and 258C. The much greater
hydrolytic stability of N-acetyl-CBZ compared to
N-Gly-CBZ indicates that the keto-enol tautomer-
ization does not appear to play a significant role
and that the electron withdrawing effects of the
amine group of N-Gly-CBZ contribute signifi-
cantly to the destabilization of the acyl-urea bond.
It is also possible that the protonated amine group
of N-Gly-CBZ could participate in intramolecular
H-bonding with the acyl-urea carbonyl, which
could further activate the carbonyl carbon
towards nucleophilic attack. In the absence of
the electron withdrawing and potentially catalytic
amine group of N-Gly-CBZ, the reactivity of the
acyl-urea bond of N-acetyl-CBZ towards hydro-
lysis is greatly reduced and becomes slower than
the hydrolysis of the tertiary nitrogen urea bond
to release iminostilbene. Furthermore, iminostil-
bene is formed as a major product from the
degradation of N-acetyl-CBZ in the acidic pH
range where hydrolysis by water becomes the
dominant reaction and the rate of loss of N-acetyl-
CBZ and the distribution of hydrolysis products
appear to become more or less pH independent.
Scheme 8. Suggested mechanism for formation of the rearrangement product result-
ing from the degradation of N-Gly-CBZ in the pH range of 8–12.
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 99, NO. 4, APRIL 2010
DOI 10.1002/jps

NOVEL WATER-SOLUBLE PRODRUG OF CARBAMAZEPINE
1823
(69.7 mg/mL, 0.2116 M) solution of N-Gly-CBZ.
When the concentration of N-Gly-CBZ reaches
64.9 mg/mL (0.1971 M), 3.43 mg (0.0145 M) of CBZ
is produced, which is the predicted solubility limit
of CBZ in a 0.1971 M solution of N-Gly-CBZ. Using
this information along with the rate constant
above, a modified version of the half-life Eq. (10)
can be used to predict this time.
bioavailability. The enzymatic stability and
in vivo performance of N-Gly-CBZ in rats after
oral and IV administration is the subject of a
companion paper. If successful, the acyl-urea
prodrug approach could provide a viable strategy
to improve the solubility of urea containing drug
compounds with poor aqueous solubility.
lnðC0=CÞ
tprecip
¼
(10)
ACKNOWLEDGMENTS
k
The authors would like to gratefully acknowledge
Dr. Serena Tongiani for her assistance in the
NMR diffusion experiments and Dr. Brian Carl-
son for his assistance with the LC/MS/MS analysis
used to identify the rearrangement product.
This work was supported in part by The American
Foundation for Pharmaceutical Education Pre-
Doctoral Fellowship and a Tuition Fellowship
from Perdue Pharma.
The constant C0 is the initial concentration of
N-Gly-CBZ (0.2116 M), C is the concentration of
N-Gly-CBZ when the solubility limit of CBZ is
reached (0.1971 M) and the value of k can be taken
from the observed rate constant for the loss
of N-Gly-CBZ in an aqueous solution with a
pH value of 4.0 at 258C (k ¼ 2.07 Â 10À7 sÀ1).
Although the associated t90 value for N-Gly-CBZ
under these conditions would be 5.9 days, the
predicted time for the precipitation of CBZ from a
50 mg/mL CBZ equivalent solution of N-Gly-CBZ
is 4.0 days, which would appear to be the shelf-life
limiting factor in this case.
REFERENCES
1. Levy RH, Mattson RH, Meldrum BS. 1995. Anti-
epileptic drugs, 4th edition. New York: Raven
Press. 1120. p.
2. Loiseau P, Duche B, Pedespan JM. 1985. Absence
epilepsies. Epilepsia 36:1182–1186.
CONCLUSION
3. Bazil CW, Pedley TA. 1998. Advances in the medical
treatment of epilepsy. Annu Rev Med 49:135–162.
4. Birkhimer LJ, Curtis JL, Jann MW. 1985. Use
of carbamazepine in psychiatric disorders. Clin
Pharm 4:425–434.
5. Williams Robin SB, Cheng L, Mudge Anne W, Har-
wood Adrian J. 2002. A common mechanism of
action for three mood-stabilizing drugs. Nature
417:292–295.
6. Brewster ME, Anderson WR, Estes KS, Bodor N.
1991. Development of aqueous parenteral formula-
tions for carbamazepine through the use of modified
cyclodextrins. J Pharm Sci 80:380–383.
7. McMahon LE, Timmins P, Williams AC, York P.
1996. Characterization of dihydrates prepared from
carbamazepine polymorphs. J Pharm Sci 85:1064–
1069.
8. Kobayashi Y, Ito S, Itai S, Yamamoto K. 2000.
Physicochemical properties and bioavailability of
carbamazepine polymorphs and dihydrate. Int J
Pharm 193:137–146.
N-acyl-urea derivatives of CBZ were successfully
synthesized through the reactions of iminostil-
bene with acyl-isocyanates to form N-Gly-CBZ
and N-acetyl-CBZ. The stability of N-Gly-CBZ
was found to range over four orders of magnitude
with its greatest stability in the pH range of
3–4 where injectable solutions of >50 mg/mL CBZ
equivalent (69.7 mg/mL) appear to be possible.
The presence of N-Gly-CBZ was found to increase
the aqueous solubility of CBZ, a degradation
product. The solubilization of CBZ by N-Gly-CBZ
was investigated by examining the diffusion
coefficients of the predominant species in D2O
and was found to be more consistent with stacking
complex formation than micelle formation and
solubilization. The stability of N-Gly-CBZ makes a
ready-to-use parenteral formulation impractical,
but a freeze-dried preparation for reconstitution
prior to administration appears to be a viable
option.
9. Kaneniwa N, Ichikawa J, Yamaguchi T, Hayashi K,
Watari N, Sumi M. 1987. Dissolution behavior of
carbamazepine polymorphs. Yakugaku Zasshi 107:
808–813.
10. Lang M, Kampf Jeff W, Matzger Adam J. 2002.
Form IV of carbamazepine. J Pharm Sci 91:1186–
1190.
It is likely that the improved solubility char-
acteristics of N-Gly-CBZ could also be beneficial
in a solid oral dosage formulation of the prodrug.
This could result in less erratic, and more comp-
lete oral absorption of CBZ with overall better
DOI 10.1002/jps
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 99, NO. 4, APRIL 2010

1824
HEMENWAY ET AL.
11. Laine E, Tuominen V, Ilvessalo P, Kahela P. 1984.
Formation of dihydrate from carbamazepine anhy-
drate in aqueous conditions. Int J Pharm 20:307–
314.
12. Kahela P, Aaltonen R, Lewing E, Anttila M, Kris-
toffersson E. 1983. Pharmacokinetics and dissolu-
tion of two crystalline forms of carbamazepine. Int J
Pharm 14:103–112.
13. Lake OA, Olling M, Barends DM. 1999. In vitro/in
vivo correlations of dissolution data of carbamaze-
pine immediate release tablets with pharmacoki-
netic data obtained in healthy volunteers. Eur J
Pharm Biopharm 48:13–19.
14. Chen X-Q, Venkatesh S. 2004. Miniature device for
aqueous and non-aqueous solubility measurements
during drug discovery. Pharm Res 21:1758–1761.
15. Brewster M, Anderson W, Pop E, Sawchuk R,
Cheung B, Webb A. 1995. Improved oral bioavail-
ability of carbamazepine through the use of chemi-
cally modified cyclodextrins. World Meeting on
Pharmaceutics, Biopharmaceutics and Pharmaceu-
tical Technology, 1st, Budapest, May 9–11, 1995:
627–628.
16. Loescher W, Hoenack D, Richter A, Schulz H-U,
Schuerer M, Duesing R, Brewster ME. 1995. New
injectable aqueous carbamazepine solution through
complexing with 2-hydroxypropyl-b-cyclodextrin:
Tolerability and pharmacokinetics after intrave-
nous injection in comparison to a glycofurol-based
formulation. Epilepsia 36:255–261.
17. Loescher W, Hoenack D. 1997. Intravenous carba-
mazepine: Comparison of different parenteral for-
mulations in a mouse model of convulsive status
epilepticus. Epilepsia 38:106–113.
18. Steulet AF, Schmutz M, Maitre L, Bernasconi R,
Stahl PH. 1991. Intravenous solutions of antiepi-
leptics. Eur. Pat. Appl. 435826 ed.: (Ciba-Geigy
A.-G., Switz.) 12 pp.
19. Yalkowsky SH, Krzyzaniak JF, Ward GH. 1998.
Formulation-related problems associated with
intravenous drug delivery. J Pharm Sci 87:787–
795.
24. Yalkowsky SH, Valvani SC. 1977. Precipitation of
solubilized drugs due to injection or dilution. Drug
Intell Clin Pharm 11:417–419.
25. Trivellato E, Malesani L, Concilio C. 1958. Some
new esters of chloramphenicol. IV. Pharmacological
research on chloramphenicol glycinate. Farmaco
[Prat] 13:399–405.
26. Bundgaard H, Larsen C, Thorbek P. 1984. Prodrugs
as drug delivery systems. XXVI. Preparation and
enzymic hydrolysis of various water-soluble amino
acid esters of metronidazole. Int J Pharm 18:67–77.
27. Stella VJ, Kennedy PE. 1987. Water-soluble rapa-
mycin prodrugs. Us Patent 4650803. ed.: (University
of Kansas, USA). p 6.
28. Pop E, Soti F, Brewster ME, Barenholz Y, Kora-
blyov V, Mechoulam R, Nadler V, Biegon A. 1996.
Derivatives of dexanabinol. II. Salts of amino acid
esters containing tertiary and quaternary hetero-
cyclic nitrogen with increased water-solubility.
Pharm Res 13:469–475.
29. Hirai K, Koike H, Ishiba T, Ueda S, Makino I,
Yamada H, Ichihashi T, Mizushima Y, Ishikawa
M, Ishihara Y, Hara Y, Hirose H, Shima N, Doteu-
chi M. 1991. Amino acid amides of 2-[(2-aminoben-
zyl)sulfinyl]benzimidazole as acid-stable prodrugs
of potential inhibitors of Hþ/Kþ ATPase. Eur J Med
Chem 26:143–158.
30. Hutchinson I, Jennings SA, Vishnuvajjala BR,
Westwell AD, Stevens MFG. 2002. Antitumor benzo-
thiazoles. 16. Synthesis and pharmaceutical prop-
erties of antitumor 2-(4-aminophenyl)benzothiazole
amino acid prodrugs. J Med Chem 45:744–747.
31. Pochopin NL, Charman WN, Stella VJ. 1995.
Amino acid derivatives of dapsone as water-soluble
prodrugs. Int J Pharm 121:157–167.
32. Pochopin NL, Charman WN, Stella VJ. 1994.
Kinetics of rearrangement and hydrolysis of amino
acid derivatives of prazosin. Int J Pharm 105:169–
176.
33. Pochopin NL, Charman WN, Stella VJ. 1994. Phar-
macokinetics of dapsone and amino acid prodrugs of
dapsone. Drug Metab Dispos 22:770–775.
20. Akers MJ. 2002. Excipient—Drug interactions
in parenteral formulations. J Pharm Sci 91:2283–
2300.
21. Johnson JLH, He Y, Yalkowsky SH. 2003. Predic-
tion of precipitation-induced phlebitis: A statistical
validation of an in vitro model. J Pharm Sci 92:
1574–1581.
22. Yalkowsky SH, Valvani SC, Johnson BW. 1983.
In vitro method for detecting precipitation of par-
enteral formulations after injection. J Pharm Sci
72:1014–1017.
23. Oshida S, Degawa K, Takahashi Y, Akaishi S.
1979. Physico-chemical properties and local toxic
effects of injectables. Tohoku J Exp Med 127:301–
316.
34. Bradshaw TD, Bibby MC, Double JA, Fichtner I,
Cooper PA, Alley MC, Donohue S, Stinson SF,
Tomaszewjski JE, Sausville EA, Stevens MFG.
2002. Preclinical evaluation of amino acid prodrugs
of novel antitumor 2-(4-amino-3-methylphenyl)-
benzothiazoles. Mol Cancer Ther 1:239–246.
35. Bradshaw TD, Westwell AD. 2004. The develop-
ment of the antitumour benzothiazole prodrug,
phortress, as a clinical candidate. Curr Med Chem
11:1009–1021.
36. Speziale AJ, Smith LR. 1963. The reaction of oxalyl
chloride with amides. II. Oxazolidinediones and
acyl isocyanates. J Org Chem 28:1805–1811.
37. Kukolja S, Lammert SR. 1975. Azetidinone anti-
biotics. XIV. Removal of a phthaloyl protective
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 99, NO. 4, APRIL 2010
DOI 10.1002/jps

NOVEL WATER-SOLUBLE PRODRUG OF CARBAMAZEPINE
1825
group from acid and base sensitive compounds.
J Am Chem Soc 97:5582–5583.
38. Zalipsky JJ, Patel DM, Darnowski RJ, Reavey-
Cantwell NH. 1976. The pKa determination of
methaqualone. J Pharm Sci 65:460–461.
39. Stejskal EO, Tanner JE. 1965. Spin diffusion mea-
surements: Spin echoes in the presence of a time-
dependent field gradient. J Chem Phys 42:288–292.
40. Yao S, Howlett GJ, Norton RS. 2000. Peptide self-
association in aqueous trifluoroethanol monitored
by pulsed field gradient NMR diffusion measure-
ments. J Biomol NMR 16:109–119.
41. Momot KI, Kuchel PW. 2003. Pulsed field gradient
nuclear magnetic resonance as a tool for studying
drug delivery systems. Concepts Magn Reson A
19A:51–64.
metric studies of ginkgolides and bilobalide using
simultaneous monitoring of proton, ammonium and
sodium adducts. Analyst 127:641–646.
45. Jonkers N, Govers H, De Voogt P. 2005. Adduct
formation in LC-ESI-MS of nonylphenol ethoxy-
lates: Mass spectrometrical, theoretical and quan-
titative analytical aspects. Anal Chim Acta 531:
217–228.
46. Espenson JH. 1981. Chemical kinetics and reaction
mechanisms, 1st edition. New York: McGraw-Hill.
p 56.
47. Anderson BD, Conradi RA, Johnson K. 1983. Influ-
ence of premicellar and micellar association on the
reactivity of methylprednisolone 21-hemiesters in
aqueous solution. J Pharm Sci 72:448–454.
48. Anderson BD, Conradi RA, Knuth KE, Nail SL.
1995. Strategies in the design of solution-stable,
water-soluble prodrugs II: Properties of micellar
prodrugs of methylprednisolone. J Pharm Sci 74:
375–381.
42. Wu D, Chen A, Johnson CS, Jr. 1995. An improved
diffusion-ordered spectroscopy experiment incor-
porating bipolar-gradient pulses. J Magn Reson A
115:260–264.
43. Jemal M, Almond RB, Teitz DS. 1997. Quantitative
bioanalysis utilizing high-performance liquid chro-
matography/electrospray mass spectrometry via
selected-ion monitoring of the sodium ion adduct
[M þ Na]þ. Rapid Commun Mass Spectrom 11:
1083–1088.
49. Surakitbanharn Y, McCandless R, Krzyzaniak JF,
Dannenfelser RM, Yalkowsky SH. 1995. Self-
association of dexverapamil in aqueous solution.
J Pharm Sci 84:720–723.
50. Abdul Rasool A, Hussain AA, Dittert LW. 1991.
Solubility enhancement of some water-insoluble
drugs in the presence of nicotinamide and related
compounds. J Pharm Sci 80:387–393.
44. Li X-F, Ma M, Scherban K, Tam Yun K. 2002.
Liquid chromatography-electrospray mass spectro-
DOI 10.1002/jps
JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 99, NO. 4, APRIL 2010