|Table of Contents|
Table of Contents
List of Tables
List of Figures
Chapter 1. Introduction
Chapter 2. Materials and methods
Chapter 3. Results and discussions
Chapter 4. Conclusions
PHYSICOCHEMICAL CHARACTERIZATION AND
STABILITY OF DOXORUBICIN IN
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
To my parents- who have patiently endured my excuses and
explanations for the delay in preparing this manuscript.
I am extremely grateful to Dr. Hans Schreier, Associate
Professor of Pharmaceutics, for his continuous academic and
financial support, encouragement, patience and tolerance
during the time it took to complete the work and generate this
document. His infallible confidence in my abilities and his
friendly disposition made working for him an enjoyable
I am deeply indebted to Dr. Edward R. Garrett, Professor
Emeritus of Pharmaceutics, for giving me the opportunity to
work with him. During my years of association with him, he has
at various times been a 'father figure', persevering teacher
and an advisor. I attribute my scientific background and
knowledge to his unique and effective style of graduate
instruction. I am also grateful to Mrs. Irene B. Garrett, for
her warm and caring disposition and her constant support
especially during some rough times.
I am also in the debt of Dr. Shangxian Chen, Visiting
Professor, University of Florida. His extensive research
experience and fundamental sciences background helped me in
understanding and solving many problems. I am also extremely
grateful to Mrs. Chen for helping with the experiments.
I am also indebted to all my committee members, Drs.
Kenneth B. Sloan, Mark Longer and Kathryn Williams, for their
scientific assistance at various times during my research.
Of the many colleagues that worked with me, Robert
Townsend has become a good friend. His infinite patience
during periods when I would bounce ideas off him and his
intelligent feedback were extremely helpful throughout my
graduate education. Other colleagues who have established long
lasting friendships and who at various times have helped me
unselfishly and unconditionally in preparing and compiling
documents such as these are Jaimini Patel, Janice Cacace,
Victoria Saldajeno, Cary Mobley, Vivek Shenoy, Patricia Khan
and Sharon Fussell-Carter.
I am extremely thankful to my parents NandLal and Shiela
and my brothers Aroop and Ajoy for being a patient and
supportive family at all times.
TABLE OF CONTENTS
ACKNOWLEDGEMENTS . . . . . . . . . iii
LIST OF TABLES . . . . . . . . .. viii
LIST OF FIGURES . . . . . . . . . . x
ABSTRACT . . . . . . . . . . .. xv
CHAPTER 1 INTRODUCTION . . . . . . . 1
History . . . . . . . . . . 3
Structural Configuration of Doxorubicin . . 4
Pharmacology and Mechanism of Action . . . . 6
Interaction with DNA . . . . . . 6
Interaction with Cell Membrane . . . . 7
Pharmacokinetics . . . . . . . . .. 10
Metabolism and Excretion . . . . .. .11
Physical Properties of Doxorubicin . . . .. .12
Partition Coefficient . . . . . .. 13
Metal Ion Complexation . . . . . .. .15
Molecular Spectroscopy . . . . . . .. .17
UV/Vis spectra . . . . . . . .. 17
Fluorescence spectra ..... .............. .18
Dissociation Equilibria of Doxorubicin . . .. .19
Self Association . . . . . . . . .. 22
Adsorption to Glass . . . . . . . .. 26
Analytical Methods ......... ............... . 28
Stability and Kinetics of Degradation . . .. .32
Stability Studies on Doxorubicin . . .. .33
Stability of (I) in Pharmaceutical Formulations . 43
Stability in Biological Media . . . . .. .46
Farmitalia and U.S. Patent # 4946831 . . . .. .48
CHAPTER 2 MATERIALS AND METHODS . . . . .. 52
Materials . . . . . . . . . .. .52
Test Compound . . . . . . . .. 52
Reagents/Solvents . . . . . . .. 53
Instrumentation .... . . . . . .. . 53
Melting Point determination . . . .. 53
Differential Scanning Calorimetry . . .. 53
pH measurements . . . . . . .. 54
Ultraviolet and Visible Spectroscopy . . .. 54
Self association . . . . . . .. 55
HPLC . . . . . . . . . . 55
Hydrolysis . . . . . . . .. 57
Fluorescence Spectroscopy . . . . . .. 57
HPLC . . . . . . . . . .. 57
Self association . . . . . . .. 59
Chromatography . ........ ..................... .61
Mobile Phase(s) . . . . . . .. 61
Solvent Delivery . . . . . . .. 62
Column . . . . . . . . . .. 62
Injector . . . . . . . . .. 63
Detector . . . . . . . . .. 63
Recorder/Integrator . . . . . . .. 63
Temperature controlled studies . . . . .. .64
Water bath(s) . . . . . . . .. 64
Methods . . . . . . . . . . .. 64
Silylating Glassware . . . . . .. 64
Self Association . . . . . . .. 65
High Performance Liquid Chromatography . . .. .69
Assay validation . . . . . . .. 70
Glass Binding . . . . . . . . .. 70
Effect of Temperature . . . . . .. 70
Effect of pH . . . . . . . .. 71
Effect of Concentration . . . . . .. 71
Experimental . . . . . . . .. 71
Sampling . . . . . . . . .. 72
Data Treatment . . . . . . . .. 73
Chemical Kinetics . . . . . . . .. 73
Effect of acids . . . . . . . .. 73
Effect of concentration . . . . . .. 74
Effect of Temperature . . . . . .. 74
Effect of pH . . . . . . . .. 75
Effect of buffers . . . . . . .. 75
Experimental . . . . . . . .. 76
Data Treatment . . . . . . . .. 76
Identification of degradation products . . .. .77
Analysis of the solution and precipitate . 78
Effect of oxygen on the hydrolysis . . .. .79
Detection of product X-COOH by potentiometry 80
Detection of product X-COOH by HPLC . . .. .81
Detection of product X-COOH by
derivatization . . . . . . . .. 82
Precipitate analysis by LC-MS . . . .. .82
CHAPTER 3 RESULTS AND DISCUSSIONS . . . . .. 84
Purity Determination . . . . . . .. 84
Melting Point . . . . . . . .. 84
Differential Scanning Calorimetry . . .. 84
Chromatography . . . . . . .. 86
HPLC Assay . . . . . . . . . .. 89
Internal Standard . . . . . . .. 89
Capacity factor (k') . . . . . .. 89
Calibration curves and statistics . . .. 92
Validation of the Assay .
Self Association . . . . . .
Fluorescence Measurements . . .
Normalization Experiments . . .
Selected Excitation Study . . .
Estimation of Association Constants
Estimation of E[MC-DOX] . . . .
Estimation of OpN-DOXi / MC-DOXI . .
Discussion . . . . . . . . .
Structure of the Dimer . . . . . .
Glass Binding . . . . . . . . .
Data Analysis . . . . . . . .
Types of Binding Sites . . . . . .
Capacity of the Binding Sites . . . .
Effect of Temperature . . . . . .
Effect of Concentration . . . . . .
Amounts Bound to Glass . . . . . .
Effect of pH . . . . . . . .
Effect of Silylation . . . . . .
Kinetics of Hydrolysis . . . . . . .
Rate Constants . . . . . . .
Statistics of the Kinetic Data . . . .
Effect of Concentration . . . . . .
Effect of Mineral Acids . . . . . .
Effect of pH . . . . . . . .
pH of Maximum Stability . . . . . .
Effect of Buffers and General Acid-Base
Catalysis . . . . . . . . .
Determination of the Catalytic Species of the
Buffer . . . . . . . . . .
Effect of Temperature . . . . . .
Degradation of (I) in the pH range 0.8 to
3 .0 . . . . . . . . . . .
Degradation of (I) in the pH range 4.0 to 8.0
~~~~~~ . . . .
Identification of the Degradation Products
Mechanism of degradation of (I) . . . .
Effect of oxygen and nitrogen on the
degradation of (I) . . . . . .
Identification of Glycollic Acid and
Glycolaldehyde . . . . . . . .
CHAPTER 4 CONCLUSIONS . . . . . . . .
APPENDIX . . . .
REFERENCE LIST . ..
BIOGRAPHICAL SKETCH .
. . 196
. . . . 92
LIST OF TABLES
Arrhenius parameters for the
hydrolysis of (I) . . .
Farmitalia patent data fitted
to the Arrhenius function . .
Capacity factors (k') and
Retention times (tr) of various
analytes studied by HPLC . .
Statistics of a typical
calibration curve for all
analytes in the concentration
range 25-700 ng/ml . . . .
Analysis of Variance of the
calibration curves for the
various analytes (I) to (V) .
Molar absorptivity (Mc-DOx) of
the monocation form of (I) at
ambient temperature . . .
Association constants estimated
at different ionic strengths .
Hybrid constants obtained from
the fitting of the glass
binding data . . . .
Derived microscopic rate
constants obtained from the
hybrid constants . . . .
. . . 90
. . . 95
. . 124
. . 126
Equilibrium constants derived
from the microscopic rate
constants . . . . .
Binding capacities of the type
1 (A/a) and type 2 (B/9) sites
and the predicted amounts bound
to the type 1 site per unit
area . . . . . . .
Predicted amounts bound
(ig/cm2) to the type 1 and type
2 binding sites . . . .
Validation of y, . . .
Effect of mineral acids on the
specific acid catalysis of (I)
at 50C . . . . . .
Second order rate constants
obtained from log kobs vs pH
profiles corrected for buffer
effects at various
temperatures . . . . .
pH of maximum stability as a
function of temperature . .
Values of kobs and kcat for the
general acid-base catalysis of
(I ) . . . . . . .
Second order catalytic rate
constants for the hydrolysis of
(I) and their corresponding
Arrhenius parameters . . .
Activation Energy Ea + Standard
Error and Frequency Factor (Ln
A) for the hydrolysis of (I) at
different pH . . . . .
. . 129
. . 130
. . 150
. . 154
. . 156
. . 165
. . 168
LIST OF FIGURES
Chemical structures of
Anthracyclines (I) to (VI) and
Rhodomycin . . . . . .
Twisted boat conformation of
the cyclohexyl ring of (I) . .
UV spectra for analytes (I) -
(VI) dissolved in mobile phase
C . . . .
S . 2
. . . . 56
Fluorescence spectra for
analytes (I) (VI) in mobile
phase C . . . . . .
Front surface and right angled
illumination setup . . . .
Typical fluorescence spectra of
(I) in aqueous solution . .
A typical DSC endotherm of the
melting of (I) for purity
evaluation . . . . . .
Plot of Ts vs 1/F for (I),
based on equation 3.1 in text
. . . . . . . . .
A typical chromatogram for
analytes (I) (VI) . . .
Calibration curves (25-700
ng/ml) for analytes (I) (VI)
prepared in mobile phase C . .
Relative Fluorescence Intensity
vs Concentration at 590 nm . .
Representative example of a
normalization spectra . . .
Diffrential spectra of (I) . .
Emission spectra of (I),
selectively excited at 580 nm
Fluorescence spectra used to
determine E[N-DOX] /([MC-DOX] at
boundary conditions of pH as
per equation 3.15 in text . .
Non-Linear fitted plots for the
self-association of (I) at pH
5 .72 . . . . . . .
Non-Linear fitted plots for the
self-association of (I) at pH
6 .72 . . . . . . .
Non-Linear fitted plots for the
self-association of (I) at pH
7 .69 . . . . . . .
Effect of ionic strength on the
association constant . . .
Representative plot of percent
Remaining vs time for glass
binding of (I) at pH 3.18,
temp: 5C and Conc: 1 p.g/ml .
. . . 91
. . . 98
. . 100
. . 102
. . 103
. . 111
. . 113
. . 114
. . 115
. . 116
. . 121
Representative plot of percent
Remaining vs time for glass
binding of (I) at pH 5.01,
temp: 5C and Conc: 10 gg/ml .
Representative plot of percent
Remaining vs time for glass
binding of (I) at pH 7.00,
temp: 5C and Conc: 100 ig/ml
Temperature dependence of glass
binding; pH 3.18; Conc: 10
jig/ml . . . . . . .
Effect of concentration on
glass binding; pH 5.01 . . .
Amount bound to glass at pH
5 .01 . . . . . . .
Typical first order plot for
the degradation of (I) in HCl
at 60C . . . . . .
Effect of Mineral Acids on the
hydrolysis of (I) at 50C . .
Log kobs vs -Log a[H+] plot
showing no influence of the
acid type on the activity of
[H'] ions . . . . . .
Log kobs vs pH plot for the
hydrolysis of (I) at zero
buffer concentration . . .
General Acid-Base catalysis-
Effect of Formate buffer at
60C . . . . . . .
General Acid-Base catalysis-
Effect of Acetate buffer at
60C . . . . . . .
General Acid-Base catalysis-
Effect of Phosphate buffer at
60C . . . . . . .
. . 135
. . 136
. . 142
. . 145
. . 149
. . 158
. . 159
General Base catalysis-Effect
of CH3COO on the hydrolysis of
(I ) . . . . . . .
General Base catalysis-Effect
of HP042 on the hydrolysis of
(I ) . . . . . .
Arrhenius plot for the
hydrolysis of (I) in HC1 . .
Arrhenius plot for the
hydrolysis of (I) at pH 3 & 4
Arrhenius plot for the
hydrolysis of (I) at pH 4.5, 5
& 5 .5 . . . . . .
Arrhenius plot for the
hydrolysis of (I) at pH 6, 7 &
8 . . . . . .
Degradation of (I) and the
formation of (IV) in the pH
range 0.8 to 3 . . . .
Sigma minus plot showing first
order formation of (IV) in the
pH range 0.8 to 3 . . .
Representative chromatogram of
a degraded solution of (I) at
pH values from 4 to 8 . .
Degradation of (I)
presence of oxy
nitrogen . . .
. . 183
Zero order formation of [H]
when (I) is degraded in the
presence of oxygen . . .
. . 163
. . 166
. . . 172
. . . 174
Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
PHYSICOCHEMICAL CHARACTERIZATION AND
STABILITY OF DOXORUBICIN IN
Chairperson: Hans Schreier Ph.D.
Major Department: Pharmaceutics
Doxorubicin (I), an anthracycline antitumor antibiotic,
is unstable in aqueous solution, resulting in loss of
activity. A sensitive, reproducible HPLC assay to quantify (I)
and its major metabolites/degradation products in the low
ng/ml range was developed.
Literature reports on the value of the equilibrium
constant for the self-association of (I) are conflicting.
Using a 'front surface illumination' fluorescence technique,
a mathematical model was developed which uniquely estimates
the influence of pH and ionic strength on the association
process and recognizes the ability of the various ionic
species of (I) to self-associate.
Nonspecific glass binding, contributes to losses of (I)
from aqueous solutions. This phenomenon, studied as a function
of concentration, temperature and pH, found that lowering the
temperature or concentration or increasing the pH increased
the amount of (I) bound to glass. The percent loss of (I) from
solution was fitted to a biexponential equation, suggesting
the possibility of two independent binding sites, each of
which could be saturated at higher concentrations.
The effect of pH, temperature and buffers on the
hydrolysis of (I) was studied. Specific acid catalysis of (I)
was observed from pH 0.8 to 3.0. The pH of maximum stability
was between 4.5 and 5.0. At pH>3 the hydrolysis was influenced
by general base catalysis. The contribution of each of these
buffer species was calculated. Hydrolysis was carried out
under accelerated conditions at 70, 65, 60 and 55C (50C for
the mineral acids). Arrhenius parameters showed the activation
energy to depend on pH, indicating different mechanisms of
degradation. At pH<3, the aglycone (V) was formed as a 1:1
degradation product. At pH>3, the chromatograms showed
unidentified peaks which appeared irrespective of the buffer
species, indicating that degradation might be pH specific.
Aqueous solutions of (I), degraded in the presence of oxygen,
showed a lowering of pH, indicating the formation of an acidic
product. A structure for this product is proposed.
Cancer (1) is a generic term for a variety of malignant
neoplasms, arising in all human and animal tissues composed
of dividing cells, which have adverse effects on the host
tissue by invasive growth and metastases. The most rapidly
developing therapy in cancer treatment uses chemical agents
with activity against human neoplastic disease.
Doxorubicin (I) (Adriamycine, Adriblastina) of the
anthracycline antitumor antibiotics has been shown to be
active against a wide spectrum of human malignancies
including soft tissue, osteogenic and other sarcomas like
Hodgkin's disease; non-Hodgkin's lymphomas; breast,
genitourinary, thyroid, lung, and stomach cancers;
neuroblastoma and acute leukeamias (acute granulocytic and
acute lymphocytic) (2).
The anthracyclines used as part of the studies
conducted in this thesis are (Figure 1.1): doxorubicin (I);
doxorubicinol (II), doxorubicinolone (III), doxorubicinone
(IV), 7-deoxy doxorubicinone (V) and daunorubicin (VI).
Chemical structures of Anthracyclines (I) to (VI) (see text) and
COMPOUND R1 R2 R3 R4 R'
RHODOMYCIN B OH S H CH3 (NCH3)2
DOXORUBICIN (I) OCH3 S -0 CH2OH NH2
DOXORUBICINOL (10) OCH3 S OH CH2OH NH2
DOXORUBICINOLONE (111) OCH3 OH OH CH2OH
DOXORUBICINONE (IV) OCH3 OH = 0 CH2OH -
7-DEOXY DOXORUBICINONE (V) OCH3 H = O CH2OH -
DAUNORUBICIN (VI) OCH3 S O CH3 NH2
(INTERNAL STANDARD) 3
In the late 1950's a pigmented compound, rhodomycin B
(Figure 1.1), originating from a strain of the Streptomyces
species found in soil samples in India, was characterized
and found to exhibit antitumor properties (3,4). In the
early sixties (VI), a precursor to (I), was isolated from a
strain of Streptomyces peucetius (5). Daunomycin was shown
to have rapid and excellent activity, with no cross
resistance against acute leukeamias (6). Major side effects
included severe aplasia, immunodepression and cardiac
complications, the latter requiring a limitation of the
total dose to 25 mg/kg with consequent impossibility of use
for maintenance treatment. In light of the pharmacological
results with (VI), subsequent research was developed along
a) elucidation of the structure and stereochemistry of the
antibiotic, and b) search for new biosynthetic analogs in
cultures derived from the (VI) producing micro organisms
Streptomyces peucetius. The latter approach, which resulted
in the isolation and characterization of (I), was based on
the premise that variations in the anthracycline structure
could induce a remarkable improvement of pharmacological
Doxorubicin was isolated from the cultures of one of
the strain varieties derived from a mutant of the original
Streptomyces peucetius, namely Streptomyces peucetius var.
caesius (7). Doxorubicin had better activity against tumors
than (VI) (8), but as with all anthracycline antibiotics its
use is restricted due to a dose limiting cardiac toxicity,
resulting in an acute cardiac myopathy (8). In clinical
practice the cumulative dose of (I) should not exceed 450-
550 mg/m2. Attempts to circumvent the problem of
cardiotoxicity have dealt with alternative routes of
administration (9-11), alternative dosage schedules (12-18),
pre- and co-administration of free radical scavengers (19-
26) and encapsulation in liposomes (27-35).
Structural Configuration of Doxorubicin
Most of the work in this field was first conducted on
(VI) and carminomycin using X-ray crystallography. Since
(VI) and (I) are similar in structure, these studies also
apply to (I). X-ray crystallographic studies (36,37,38) on
(VI) and carminomycin hydrochloride (37,38) showed that the
anthracycline part of the molecule was flat. This was also
indicated by the resonance stabilization that led to an
extended conjugation and a colored compound with a strong
fluorescence. The NMR spectra of (VI) confirmed that the
cyclohexyl ring is in the half chair conformation (or
twisted boat conformation) (Figure 1.2) and the sugar moiety
is nearly perpendicular to the plane of the chromophore.
Studies (39) on the solution conformation of (I) indicated
the same structure described by X-ray crystallography. The
distance between the 0-5 and 0-6 atoms (2.45 A) was found to
be much lower than between the 0-11 and 0-12 (2.67 A)
indicating a different charge distribution within the
quinonic chromophore. The relevance of this structural
property for antitumor activity has not been established.
The half-chair conformation of the cyclohexyl ring
stabilizes the molecule due to intramolecular hydrogen
bonding (Figure 1.2).
The 0-9 and 0-7 oxygen atoms can hydrogen bond via the
-OH at 0-9. Alternatively there could be a binding
interaction between the OH at C-9 and the oxygen of the
The positively charged NH2 group of the molecule in
(VI) has been shown to be hydrogen bonded with the C-4' OH
Another important finding indicated the axial H-8 and
H-10 protons had a long range coupling in all the
derivatives of (VI) studied (40) which suggested the
preferred half chair conformation of the cyclohexyl ring.
Figure 1.2: Twisted boat conformation of the cyclohexyl
ring of (I)
Pharmacology and Mechanism of Action
Interaction with DNA.
Most of the (I) in the cell accumulates in the nucleus
(41,42) up to a maximum of 1 drug molecule per 9 base pairs
(43). Doxorubicin binds to DNA by intercalation between
successive base pairs of the helix. The affinity constant,
determined by Scatchard analysis, is about 2 X 106 M-i and
the number of binding sites is 0.2 per DNA phosphate.
The biological significance of this interaction is not
clear. Both DNA and RNA syntheses are inhibited, with the
effect on initiation being greater than on elongation (44)
and the effect on repair being less than on replication
(45,46). Additionally, DNAase (47,48), reverse transcriptase
(49,50) and DNA polymerase II (51) are also inhibited. Drug
binds to chromatin with the same affinity as DNA but with
lower number of binding sites (52,53). The drug binds to the
internucleosomal regions, leading to compaction of the
chromatin (54). This is probably the lethal event, rather
than the inhibition of nucleic acid syntheses. Cytotoxicity
does correlate with nuclear drug content (55).
Additionally, covalent binding to DNA can occur (56,57,58)
through reductive activation, as well as photoactivation
(59,60) resulting in the degradation of the DNA (61). This
damage was thought to occur due to the generation of
reactive oxygen species like the OH radical (56,57).
However, breakage in the DNA also occurred in the absence of
oxygen, implicating an enzymic mechanism (62).
Interaction with Cell Membrane.
Doxorubicin undergoes a physical interaction with
membranes (63,64,65). An initial ionic attraction is
followed by the insertion of the hydrophobic region of (I)
into the phospholipid region of the membrane (64,66). The
affinity of the binding is equivalent to that for DNA (64).
The interaction is noncovalent unless the lipid is
peroxidized, in which case Schiff's base formation can occur
(65). The effect of the insertion into the membrane is an
increase in membrane fluidity (67,68,69) which may be the
cause of experimentally noted effects such as stimulation of
membrane NADH oxidase (70), inhibition of membrane ascorbic
oxidase (71), histamine release from mast cells (72),
increased glycosylation of the cell surface (73) and
deacylation of phospholipid (74). Whilst (I) will insert
into membranes of any phospholipid composition, there is a
specific binding interaction with cardiolipin, a
phospholipid generally only found in the mitochondria
(65,67,75). The specialized cardiolipin organization within
the membrane is destroyed, the (I)-cardiolipin complex
segregates, (76) there is an increase in the rigidity of the
membrane, (77) and the complex can act as an electron
transport system (77,78). These effects, along with lipid
peroxidation, could be the cause of damage to heart cells
Another membrane effect associated with (I) is the
effect on the Ca"2 uptake and transport. This effect could
be related to the lipid peroxidation, (I)-cardiolipin or
(I)-membrane interactions (79,80). Calcium transport is
affected (81,82), leading to a reduction of fast exchanging
calcium in heart cell mitochondria and sarcolemmal vesicles
(83,84). Calcium linked processes such as the Na'/K pump
are impaired (82,85,86), giving a prolongation of the action
potential (87,88). Other consequences include inhibition of
Ca2 dependent protein kinases (89), inhibition of actin
polymerization (90) and leukotriene formation via
phospholipase A2 activation (91). Support for the hypothesis
that Ca2 control impairment is related to the cardiotoxic
effects is given by the reversal of the cardiotoxicity by
the chelating agent ICRF-159 and its levo isomer ICRF-187
(92,93) and by verapamil, propranolol and hydralazine (94).
It is apparent that (I) has multiple effects and the
effect on cardiac mitochondria is probably the major cause
of the cardiotoxicity. This may be mediated by lipid
peroxidation (due to free radical formation) and/or physical
interaction with cardiolipin coupled with a direct or
indirect action on the control of Ca2 flux. A recent
hypothesis seems to implicate (II), a metabolite of (I), in
the cardiac toxicity. This hypothesis is supported by the
fact that cardiotoxicity is time delayed and is unrelated to
the plasma or cardiac levels of (I). The terminal half life
of (I) in plasma is 24 hours, and thus (I) is virtually
eliminated from plasma and the heart in 4 to 5 days (95,96).
The pharmacokinetics are similar in multiple as well as
single dose studies in rats (97). In contrast, the
cardiotoxicity takes weeks to develop and seems to be
correlated with the appearance of (II), which appears in
detectable levels after 24 hours (95,97,98). Doxorubicinol
accumulates in cardiac tissue and appears much more toxic
and more potent in isolated cardiac preparations than (I)
(98,99). However, direct comparisons of cardiotoxicity
between (I) and (II) are confounded by other factors.
Cardiac dysfunctions correlate to time dependency with (I)
whereas the cardiotoxicity of (II) is time independent.
Thus, critical evaluation of the metabolite hypothesis and
the use of in vitro vs in vivo models to assess the
cardiotoxicity should help to resolve and elucidate the
mechanism of cardiotoxicity (100).
Some of the other toxic manifestations (common to most
anticancer agents) are immunosuppression, myelosuppression,
reversible alopecia, diarrhoea, nausea and vomiting (101). A
serious side effect is severe dermatological toxicity if (I)
is extravasated during infusion. Severe pain is noted
immediately upon extravasation, followed in a few hours by
swelling and reddening. Skin lesions can become severe over
a period of several months resulting in necrosis of the skin
and the underlying tissues (102-104) down to the bone.
Spontaneous healing of these ulcers is rare, and surgical
excision of the involved tissues has become the recommended
It is generally accepted that plasma pharmacokinetics
of (I) following intravenous administration is best
described by a sum of three exponentials in both man and
animals (105-114). Earlier studies (115-119) showed fits to
a sum of two exponentials, probably due to a lack of
analytical sensitivity. The terminal half life is 14 to 30
hours in human patients suffering from cancers other than
hepatic tumors (110,114,120), 21-30 hours in the dog (113),
40 hours in the guinea pig (121), and 24-61 hours in the rat
(116,117,122,123). Plasma protein binding is about 59% in
rabbits, 90% in humans, 66% in rats and 48% in guinea pigs
(121,122). Hepatic enzyme induction and enzyme inhibition in
rats (122,125) pretreated with phenobarbital (an enzyme
inducer) reduced the plasma half life of (I) from 26 to 16
hours (125) and increased the excretion of (I) by 27% (122).
There was also an increase in the plasma levels of the
metabolite (II). Pretreatment with carbon tetrachloride (an
enzyme inhibitor) resulted in significantly higher levels of
(I) in plasma by reducing the total elimination by about 30%
(122). A similar observation made in humans being treated
for various cancers with impaired hepatic functions resulted
in a corresponding reduction in dose (126,127). In order to
expose tumors to prolonged levels of (I), and to reduce the
accompanying cardiotoxcity, a general consensus seems to be
to deliver the drug as a slow i.v. infusion in periods
ranging from 10 to 96 hours (110,120,128-131).
Metabolism and Excretion.
The principal metabolite in all species appears to be
(II) and the 7-deoxyaglycones of (I) and (II) (132). The
conversion to the alcohol is catalyzed by widely distributed
cytoplasmic aldo-keto reductases (133,134) and requires the
use of NADPH as a cofactor. Doxorubicin was reduced about 20
times more slowly than its analog (VI) (135), suggesting a
possible reason for its higher efficacy. The reductive
splitting of the amino sugar moiety of (I) and its principal
metabolite (II) to their corresponding deoxyaglycones was
demonstrated to occur in the microsomal fraction of the rat
liver (136). This catalysis was inhibited by oxygen,
required NADPH as a cofactor, was induced by
phenobarbital,and was not affected by enzyme inhibitors such
as carbon monoxide, j-diethylaminoethyldiphenylpropyl
acetate or Mg2, Mn+2, Fe2 and Ni+2, although it was inhibited
by Cu2 and Zn2. In the dog, i.v. administered (II) had a
half life of 3.7 hours but demonstrated a half-life similar
to the parent compound when analyzed after administration of
(I). This indicated the rate-determining nature of the
metabolic step (113). Over a 24-hour period only 5.2% of the
unchanged drug and 0.13% of the derived (II) was excreted in
the urine of dogs (113). In rats, between 6-8% of the dose
was recovered in the urine. About 35% was excreted in the
bile after 10 hours, of which 70% was as unchanged drug and
30% was as (II), with small amounts of the corresponding
deoxyaglycones (137). A mass balance study of radiolabelled
(I) in rats showed that after 96 hours 7.5% of the
radioactivity was in the urine, 65% in the feces and 6.75%
in expired CO2, with the remaining 21% in other tissues,
where a higher concentration in cardiac muscle than skeletal
muscle, liver, lymphatic and glandular tissues was found
(138). In man, the metabolism and excretion was
qualitatively similar to that of the rat. However, only 60%
of the dose could be accounted for, the rest probably being
present as nonfluorescent metabolites. Of this, only 10%
appeared in the urine over a period of 5 days and about 50%
in the bile (118,139). The principal products excreted in
the bile and the urine were about 40% of (I) and 29% of the
major metabolite (II) with 9% of the combined aglycones, 10%
of the 4-demethoxysulfate and about 12% of the 4-
Physical Properties of Doxorubicin(141)
The chemical name for (I) is (8S,10S)-10-(3 amino-
napthacenedione (CAS # 23214-92-8).
Appearance and Color.
The hydrochloride salt is a free flowing crystalline
powder, and the freeze dried formulation containing lactose
is a red cake.
The empirical formula for (I) is C27 H29 01 N.HCl
The molecular weight of (I) is 579.98 (of HCl salt).
The melting point of (I) has been determined as 205C
The hydrochloride salt is readily soluble in water,
physiological saline and methanol. It is slightly soluble or
practically insoluble in organic non polar solvents.
Strongly dextrorotatory [a], = +255 at 589 nm in 0.1%
The determination of the partition coefficient of any
drug between an aqueous and organic phase(s) is an important
parameter. In addition to revealing important
physicochemical information like ionization behavior,
lipophilicity and solubility, one can also extrapolate this
information to indicate binding to tissues, cells and other
macromolecules, uptake into cells and transference across
biological membranes. In addition, partition behavior can be
utilized to establish sample handling strategies such as
extraction of the drug from interfering biological or
chemical matrices prior to detection or quantification by
various analytical techniques. The apparent partition
coefficient of (I) between 1-octanol and Tris buffer at pH
7.0 with constant ionic strength (1:0.1) at room temperature
after shaking for 15 hours was determined to be 0.52 (141).
Eksborg (142) studied the partitioning behavior of a series
of anthracyclines as a function of pH in order to study the
self association of (I). Using a chloroform:l-pentanol (9:1)
system as the organic phase he found that the distribution
coefficient was strongly dependent on the pH of the aqueous
phase with optimal extraction occurring between pH 8.0 to
8.6. The apparent dissociation constant (pKj) of (I) was
determined to be 7.20, resulting in the isoelectric point
(isoelectric point is that pH where 50% or more of the
species is in the zwitterion form) of the molecule lying
between 8.0 and 8.6 (pK2 = 9.3). It can be argued that the
value of the pK2 has been underestimated from the
spectroscopic data (see discussion on dissociation
equilibria) and that the isoelectric point and the pH range
of maximum extraction lies between 9.0 and 9.2. The
extraction efficiency decreased as concentration increased,
presumably due to the formation of dimers and tetramers of
doxorubicin in the aqueous phase. Nakazawa et al.(143)
employed a countercurrent extraction method to study the
effect of various inorganic salts on the partition
coefficients of (I) and (VI) using three different organic
phases. The highest extraction was obtained between a 1-
butanol:1.0 M Na2HPO4 (1:1) system. However, in this study
the effect of pH on the extraction efficiency was not
reported. In studies dealing with the analytical development
of (I) from biological matrices, extraction with
chloroform:isopropanol (4:1 v/v) (144) and
chloroform:methanol (9:1 v/v) (145) have been considered
satisfactory (>90% extraction efficiency). The pH of the
aqueous phase in most cases is adjusted between 8.0 and 9.5
prior to extraction.
Metal Ion Complexation.
Aluminum, copper, magnesium, iron, calcium, gallium and
a host of other metal ions bind to (I) in solution (146).
The metal binding seems to affect the activity of (I) with
respect to its interaction with DNA. Thus, physiological
levels of Cu2 and Fe"3 ions interact with (I) resulting in
complexes that intercalate to a larger degree with DNA (147,
148-153). The Fe"3 complex intercalates with the DNA and the
metal ion is released (154). Both Cu2 and Fe"3 form ternary
complexes with (I), and these complexes are capable of
producing the superoxide free radical, which in turn is
capable of inducing DNA damage (147,151,155,156). Studies
dealing with the interaction of these ternary complexes with
phospholipids, erythrocyte ghost membranes and ADP show that
all of these systems initiate the destruction of these
biomembranes (157-159). Other metal ions like Ca2, Cd2,
Mg2, Pb+2, and Zn+2 do not interact with the (I)-DNA complex.
Complexation studies using UV-Vis, circular dichroism, NMR,
Raman, Mossbauer and IR spectroscopy indicate the
involvement of the deprotonated hydroxyanthraquinone moiety,
which binds with the metal ion forming a six membered
chelate (154,147,160,161,162,163-165). Initially the C-lI
hydroxy and C-12 carbonyl functions are the preferred sites
for chelation. The second metal ion is introduced at the C-5
ketone and C-6 hydroxy site. Complexation occurs in the pH
range from 2 to 8 and the optimum pH largely depends on the
metal ion involved (161). Binding does not take place at pH
values below 2 and above 8 (where metal hydroxide formation
predominates). Metal binding is accompanied by a change in
spectral characteristic and a drop in pH (147,160-165)
indicating the displacement of the proton from the C-ll or
C-6 hydroxy groups on the chromophore.
A polymeric form of Fe1-(I) complex, called
Quelamycin, is undergoing clinical trials. The complex was
first described by Gosalvez et al.(166), and is believed to
be less cardiotoxic than (I). It exists only in concentrated
solutions and hydrolyses to Fe-(I)-OH upon dilution (167).
Studies have shown that Quelamycin is in fact a mixture of
Fe(I)3 and polymeric ferric hydroxide (154).
Absorption maxima occur at 233, 253, 290, 477, 495 and
530 nm. The UV spectra shows the characteristic peaks of
extended conjugation of an aromatic nucleus (141). The broad
hump starting at 420 nm is indicative of the highly
conjugated anthraquinone moiety. This gives the compound its
red color. Even though the molar absorptivity between 420
and 540 nm is low, the lack of interference from other
components makes this region of the spectrum an attractive
possibility for chromatographic analysis. On adding alkali
(pH>9), the UV-Vis spectrum shifts towards longer wavelength
due to the characteristic indicator-like properties of
quinones. The color change associated with this spectral
shift is from orange-red to violet-blue. The absorption
spectrum of (I) is also dependent on the presence of metal
ions (147,149,157,158), solvent (168,169), drug
concentration (168,161-174) and ionic strength (146,169).
The red shift associated with the deprotonation of the
phenolic proton from the chromophore is analogous to the one
observed when (I) interacts with metal ions. The spectra of
the monomer and the dimer of (I) can be established by
differential measurements (172). Increase in the ionic
strength has a similar effect as the concentration
indicating the possibility that increased ionic strength
promotes aggregation of (I) in solution (169). UV-Vis
spectroscopy has been extensively used to study the
interaction of (I) with nucleic acids, as this interaction
is accompanied by changes in the spectra (146,150,175-178).
Excitation at the longest wavelength absorption maximum
at 485 un gives an emission band between 520-620 nm. Since
the absorption and emission bands overlap, certain inner-
cell effects, such as self-absorption and concentration
quenching, can affect the spectral behavior. The
fluorescence emission has a high quantum yield, and is used
as a sensitive method for the analysis of (I) from solutions
or biological matrices (179-186). However, to ensure
specificity, a technique like chromatography is necessary to
separate the principal metabolites and degradation products
of (I) which have identical spectral characteristics.
In some cases of fluorescence spectroscopy the apparent
fluorescence intensity and spectral distribution can be
dependent on the optical density of the sample and the
precise geometry of sample illumination by the excitation
light (187). The most commonly used geometry is the right
angle observation of the center of a centrally illuminated
cuvette. Frequently, however, off-center illumination
techniques are used, especially in samples having a high
optical density or turbidity, in order to reduce the path
length of illumination of the sample and thereby reduce the
possibility of attenuation of the emitted signal due to
self-absorption and inner-cell effects (187).
In one case of the off-center illumination technique,
front face illumination is performed using triangular
cuvettes or square cuvettes oriented at 45 to the incident
beam of light. A drawback of this system is that a large
portion of the incident light is reflected off the front
surface of the cuvette into the emission monochromator
increasing the chances of interference from stray light. A
better position (187) is to orient the cuvette at 30 to the
incident beam, thereby reducing interference from stray
Dissociation Equilibria of Doxorubicin.
Knowledge of the protolytic behavior of (I) and related
compounds in aqueous solution is of obvious importance for
the evaluation of the biochemical interactions with
biologically significant macromolecules and receptors, and
for the interpretation of their degradation kinetics in
biological fluids and aqueous solutions.
Sturgeon and Schulman (188) investigated the protolytic
equilibria of (I) using UV-Vis and fluorescence spectroscopy
(Scheme 1.1). As shown, they concluded that (I) was present
in aqueous solutions at pH<7.0 as the monocation MC-DOX.
However, at higher pH values MC-DOX could lose a proton to
form a neutral species N-DOX or a zwitterion Z-DOX. Further
removal of a proton from N-DOX or Z-DOX will give rise to
the anion MA-DOX. The negative charge on Z-DOX or MA-DOX
could be placed on either the C-6 or C-lI structurally
equivalent oxygen atoms. The doubly charged cation can be
observed when the drug is placed in concentrated sulfuric
acid. A doubly charged anion can be seen at pH>13. The
values for the microscopic constants were pK1=8.22,
pK2=10.10, pK3=9.01l, pK4=9.36. The macroscopic constants were
pK1=8.15 and pK2=10.16. In the above study self-association
was avoided by maintaining concentrations at less than 10-6
In another study using isoelectric focusing techniques,
Righetti et al.(189) reported that the isoelectric point of
the compound was 8.76 Spectrophotometric titration of the
phenolic OH group gave a pKa of 9.6. On this basis, the pKa
of the NH3' group was estimated at 7.92. This estimate did
not take into account the presence of the zwitterion and the
apparent equilibria that could exist between the zwitterion
and neutral species. Direct titration of (I) with 0.05 M
NaOH (141) gave a PKa value of 8.22. The pKi and pK2
estimated by Eksborg (142) using solvent extraction and
spectroscopic techniques described before were 7.20 and
The ionization of (I) has been shown to be dependent on
the degree of self-association in aqueous solutions
(132,190), configurational changes in the daunosamine moeity
pK1 = 8.1
K=[Z = 0.16
Kt =[Zl/[N] =0.16
pK2 = 10.16
i T OH- CH2OH
p 0OHl NEUrH2 L [N-OX
pe(=1 0.10 NEUTRAL [N.Doxi
Protolytic equilibria of (I) in aqueous solution
(191,192) and the number of hydroxyl groups on the sugar
moeity (191,193) (lowering the number of hydroxy groups on
the sugar increases the pKa of the amino group). In
addition, the interaction of the amino sugar with the
quinoidal moeity (161) can influence the pKas. There is a
suggestion that the side chain of (I) might be involved in
some stabilizing interactions with the amino sugar via
hydrogen bonds. This could also conceivably influence the
ionization of the NH3 group (142).
Many dyes, including compounds endowed with biological
activity such as acridines (194-196), purines (197), and
actinomycin D (198), can form molecular aggregates by
vertical stacking, even in dilute solutions. The tendency of
anthracycline molecules to self-associate to dimeric or even
polymeric aggregates was first revealed by X-ray diffraction
(36) of freshly crystallized (VI) from acidified methanol.
Studies conducted using circular dichroism and NMR
demonstrated that for low concentrations (<5 mM) the
association process could be represented by a dimerization
model with an association constant of 570-700 M-'1, whereas
at higher concentrations further association to polymers was
possible. Spectroscopic studies (199) reported somewhat
higher values of 3.0 X 103 M-1, 1.83 X 103 M-1 and 1.16 X 103
M1 at 25C, 35C and 45C, respectively, at pH 7.0 (0.01 M
phosphate buffer). Thermodynamic parameters estimated by
microcalorimetric measurements (199) gave the association
constant of (I) as 1.1 X 103 M-1, with a A Ha = -9.6
kcal/mole and A Sa = -18.3 eu (at 25C). These values are
similar to enthalpy values of other dyes undergoing
association (199). An enthalpy decrease of 8-9 kcal/mole
suggests that the dimerization process is based on the
stacking interactions of the planar chromophores. Eksborg
(142) studied the dimerization process by absorption
spectroscopy and solvent extraction techniques using a
chloroform:l-pentanol (9:1) system as the organic phase. He
estimated association constants of 15,848 and 19,952 M-'
with the former technique and 30,199 and 30,902 M-1 from the
latter for (VI) and (I), respectively. Spectroscopic studies
carried out in the pH range 9.8-10.5, where the effect of
MCDOX could be ignored, yielded values that were attributed
to the neutral species. He further concluded that the
interaction of the molecules was due to their ring systems
since no dimerization of the negatively charged phenolate
ion was found. He also reported tetramerization constants of
2 X 1012 and 1 X 1012 M-1 for 'I) and (VI), respectively.
Martin (173), while working on the interaction of (VI) with
calf thymus DNA, studied the phenomenon with respect to
temperature, ionic strength and solvent composition, using
visible absorption, fluorescence and circular dichroism
spectroscopy. He estimated an association constant of 6,400
M-1 and found that increasing the ionic strength increased
the aggregation of (VI) almost twofold. This was similar to
behavior reported by Blauer (200), and was thought to be a
modification of the solvent structure by the salt which had
been observed in the case of actinomycin D (198). Chaires et
al.(171) studied the self association of (VI) using
absorbance, sedimentation and NMR measurements and used an
indefinite association model to postulate a dimerization
constant of 1,500 M-1. Thermodynamic values from NMR data
indicated stacking of the planar aromatic structures.
However, similar studies on actinomycin D (201) did not
favor the indefinite association model. NMR data showed that
the ring protons and the 0-CH3 on Ring D (Figure 1.1) was
the major site of interaction. The side chain COCH3 and
sugar moieties did not affect the aggregation process.
In an attempt to resolve the discrepancies in the
literature, Arcamone et al.(172) studied the self-
association of (I), (VI) and related compounds in aqueous
solutions using absorption spectroscopy. Their method
involved the measuring of the difference spectra of these
compounds by using cells of different path lengths in the
sample and reference beams. They were able to differentiate
between the monomer and the dimer spectra. They estimated
the association constant for (I) to be 1.29 X 104 M-' which
was dependent on buffer composition and ionic strength. An
increase in the ionic strength resulted in an increase in
the association constant. Since the association constant of
(VI) was more affected by increase in ionic strength than
that of (I), it was concluded that the OH group at the C-14
position of (I) was involved in the association process
either directly via hydrogen bond formation, or indirectly
via dipole effects. McLennan et al. (202) used NMR
spectroscopy, and particularly the shift of the OCH3 protons
to estimate the self-association. Using the rotational
correlation time of the associated species as compared to
the monomer species, they were able to estimate a value of
4.0 X 10 M-'1 at 22C for the association constant.
The effect of formulation additives like NaCl and
methyl p-hydroxybenzoate on the self-association process
have been studied (203) using viscosity and NMR
measurements. The addition of higher concentrations of NaCi
increases the viscosity of (I) solutions indicating that an
increase in the ionic strength facilitates self-association.
By contrast, the addition of various amounts of methyl p-
hydroxybenzoate reverses the effect of the NaCl. Most
studies (142,171,199,202,203) conclude that association
takes place by the vertical stacking of (I) molecules at the
chromophoric part.The forces responsible for this
association are not completely understood. A R-n interaction
between the quinone chromophores is thought to occur, but
the contribution of hydrophobic or electrostatic
interactions cannot be ruled out (132). In any case, the
aggregation of (I) molecules should not be important at the
biologically active concentration of 106 M and less. This
may be due partly to the high protein binding of the drug
(90%) resulting in very low free drug concentrations.
The self-association literature discussed above shows
conflicting values of the association constant and is
inconclusive about the effect of pH and ionic strength on
the association process. To this end, a systematic study of
the association process of (I) in aqueous solutions was
carried out, using fluorescence with a 'front surface
illumination' technique. A mathematical model allowed for
the prediction of the association constants of each
individual ionic species generated due to pH changes and the
influence of ionic strength on these processes.
Adsorption to Glass.
During studies involving uptake of (I) into tumor cells
in vitro, Tomlinson and Malspeis found significant loss of
(I) from Hank's culture medium (204) to the walls of the
glass container. A detailed study carried out in different
containers at 37C showed that about 7.3% of (I) was bound
to glass, 4.6% to polyethylene Petri dishes, and 45% to
polytetrafluoroethylene (PTFE). However, there was no
binding to polypropylene and to siliconized glass. In order
to minimize losses due to adsorption, they recommended that
a) the number of wall surface contacts should be reduced, b)
an appropriate container material like polypropylene or
siliconized glass should be chosen, c) the walls of the
container should be presaturated with the drugs and d)
cosolvents should be used. Adsorption of hydrophobic drugs
to the walls of a container have been reported by various
authors (205-212). Invariably the use of cosolvents
(containing a more nonpolar solvent than water) or
protecting the glass surface by siliconization helps to
alleviate the problem. Similar effects were observed for (I)
during analysis (213) and when sampling from solution (214).
Conventional approaches like cosolvent extraction
techniques, have been used to resolve the difficulty
For HPLC analysis desipramine has been added to mobile
phases (217,227) and to extraction media (218-221) to
prevent the adsorption of (I) to HPLC hardware and to glass.
Desipramine appears to occupy the same adsorptive sites as
(I) on these surfaces.
In complete contrast to the study of Tomlinson and
Malspeis (204), Schutz et al.(174) found that (VI) binds to
siliconized glass below concentrations of 2 x 10-' M.
Andrews et al. (222) found that silylation of glassware had
no effect on the extraction of (I) from aqueous solutions.
In studies (223) carried out to compare the stability of
antitumor agents in glass and polyvinylchloride bags (I) was
found to be more stable in the plastic bags than in glass
bottles. Binding to end line (0.2 pm) infusion filters was
found to be negligible (224), as greater than 96% of the
drug was recovered when a saline solution of the drug was
passed through the filters at 80-100 ml/hour for 2.5 hours.
The apparent complexity of the nonspecific adsorption
of (I) to glass prompted the design of a study to quantify
and explain the phenomenon as a function of concentration,
temperature and pH of aqueous solutions of (I).
A fluorometric method of anthracycline analysis
developed by Schwartz (225) involved the extraction of
tissue and plasma samples with isoamyl alcohol and
measurement of the solution's fluorescence with excitation
at 490 nm and emission at 560 nm. Calibration curves were
linear in the range 0.05-10 Lg/ml. Prior to the extraction,
0.2 ml of a 33% w/v solution of silver nitrate was added to
precipitate the proteins, nucleotides, etc. and to aid in
the subsequent release and extraction of the anthracyclines.
This method was nonspecific for (I) and its major
metabolites. The oxidative effects of the silver ion were
Cummings et al.(226) determined the concentrations of
(I), (II) and their 7-deoxy aglycones in human serum by
HPLC. A reverse phase C-18 column with fluorescence
detection (excitation 480 nm and emission 560 nm) was used.
The mobile phase was 62.5% orthophosphoric acid, pH 3.2 (5mM
final concentration) and 37.5% of a mixture of methanol,
acetonitrile and 2-propanol (either 12.5:12.5:12.5 or
15:15:7.5). Elution was isocratic with a flow rate of 2.5
ml/min. The limits of detection of (I), (II) and (IV) were
4.4 ng/ml, 2.1 ng/ml and 4.2 ng/ml, respectively. The
internal standard was (VI). Serum samples were extracted
with chloroform:2-propanol (2:1). Unfortunately, the binding
of (I) to PTFE (206) tubes used in the study was not
evaluated. Thus, their report that 9.1% of (I) degraded per
hour at 25C cannot be accepted.
HPLC was also used by Oosterbaan et al.(144) to analyze
(I), (II), carminomycin and the 4'-epi and 4-demethoxy
derivatives of (I). The analytical system consisted of a 15
cm reverse phase C-8 column (8 im) with a dual pump solvent
delivery system. The first pump delivered a 10 Lg/ml
solution of desipramine in demineralized water to a 25 cm
silica column used to concentrate the sample. The
desipramine prevented the adsorption of the anthracyclines
to column and tubing materials. The second pump delivered
the mobile phase of acetonitrile:citric acid buffer, pH
2.2:water (35:10:55% v/v) and transferred the concentrated
anthracyclines from the silica column onto the analytical
column. Calibration curves were established in the range 5-
500 ng/ml. The limit of detection for (I) was 0.5 ng/ml and
for (II) was 0.4 ng/ml. Detection was by fluorescence
(excitation at 474 nm and emission at 590 nm).
In an attempt to prevent the adsorption of (I) and its
major metabolite (II) on to the chromatographic system,
Ichiba et al.(227) used normal phase HPLC with a
fluorescence detector (excitation at 480 nm and emission at
590 nm). The stationary phase was a 25 cm silica gel column
and the mobile phase was a mixture of methylene chloride,
methanol, glacial acetic acid and 0.01 M MgCl2 solution
(200:50:7:5 v/v). The addition of Mg2 ions was purported to
reduce adsorption and give sharper peaks. The calibration
curves for (I) and (II) were linear down to 1.5 ng/ml, when
(VI) was used as an internal standard.
Watson et al.(228) studied the effects of different
surfactants on the HPLC capacity factors of (I) and its
major metabolites. They used a reverse phase 10 cm C-18
column (5 Jm), and a mobile phase of acetonitrile:phosphoric
acid (0.01 M) (35:65% v/v). Detection was by fluorescence
(excitation 450 nm; emission 550 nm). They found that 6mM
Brij-35, polyoxyethylene lauryl ether, a non ionic
surfactant, in the mobile phase reduced the analysis time
from 18 to 12 minutes. The limit of detection for
doxorubicin was 1.5 ng/ml.
Weenan et al.(229) successfully separated and
quantified doxorubicin, 4'-epidoxorubicin and their
metabolites using a 25 cm reverse phase C-18 column (3 pim).
The mobile phase consisted of 70 parts of a 0.06 M potassium
phosphate monobasic buffer (pH 4) and 32.5 parts (v/v) of
acetonitrile. A fluorescence detector (excitation at 470 nm;
emission at 580 nm) was employed. Plasma samples were
treated with chloroform:isopropanol (4:1 v/v) and the
organic phase evaporated to dryness. The total analysis time
was about 30 minutes. Calibration curves were linear from 3
ng/ml to 3 jg/ml.
Riley et al.(230) used electrochemical detection to
quantify a series of closely related anthracycline
antibiotics, including (I). Taking advantage of the
oxidation-reduction behavior of the phenolic groups and
quinone moieties of (I), they were able to effectively
quantify (I) to 1 ng/ml in plasma. The chromatographic
conditions consisted of a reverse phase C-18 column with an
optimum mobile phase of acetonitrile:isopropanol:0.1M
phosphate buffer, pH 4.5 (25:3:72 v/v). The electrochemical
detector was set at +0.8 V for the best signal-to-noise
In similar studies (231-233) (I) and its major
metabolites were analyzed in biological samples using
reverse phase HPLC systems with a C-18 column (232), a
phenyl column (231) or a cyano column (233). In all cases,
mobile phases had varying concentrations of acetonitrile and
aqueous format or phosphate buffers, and fluorescence
detection was used (excitation 470-475 nm; emission 565-580
All authors using HPLC claimed a high sensitivity,
specificity and reproducibility, whereas in prior assays of
(I) specificity was a major problem, since fluorescence and
absorbance of metabolites and degradation products had
identical spectral characteristics as the parent compound.
Most studies did not mention whether the assayed (I)
solutions were protected from light, oxygen or glass
binding, processes which would affect the concentration of
The assay presented herein for (I) and its major
metabolites and/or degradation products (e.g.(II),(III),
(IV) and (V)) were developed in preparation for kinetic
studies on the degradation of (I).
Stability and Kinetics of Degradation
In order to establish appropriate storage and handling
conditions, all drugs are evaluated for their stability
characteristics as a function of temperature, solution
composition (pH, ionic strength, solvents, effects of
buffers and additives), light and oxygen. These studies are
usually carried out under accelerated, conditions wherein
the drug is subjected to extreme levels of the conditions
described above. The results thus obtained are
mathematically interpreted using simple laws of
thermodynamics, chemical kinetics and statistics, and are
extrapolated to yield information about the stability
(chemical and consequently biological) at room or ambient
temperatures, refrigerator and freezer conditions, in the
presence or absence of light, oxygen and moisture and in the
solid or solution form. These studies are carried out on the
unformulated as well as formulated drug. A logical extension
to these studies is the degradation behavior of the drug,
the rate and extent of formation of the degradation
products, the chemical pathways by which this degradation
occurs, the pharmacological usefulness or uselessness of
such products and the conditions under which one could
inhibit or accelerate the formation of these products. This
information helps all branches of the scientific community
who may handle the drug, and would therefore need this
knowledge, to design better studies.
Stability Studies on Doxorubicin.
The anthracyclines, especially (I) and (VI), have been
extensively studied for their stability in the solid and
solution state. However as will be detailed below, most
studies did not attempt to evaluate the whole spectrum of
conditions that can affect the stability of (I). Some
studies indicated that the conditions were considered, but
for some reason the authors failed to report the results in
Stability in the solid state. In the crystalline state
(I).HC1 was stable (<10% degradation) for more than five
years at room temperature, without chemical modification and
without loss of activity (141,234). The lyophilized
formulation containing lactose as a diluent was stable for
more than two years under the same conditions.
Stability in solutions. The solvolysis of (I) is
catalyzed by hydrogen and hydroxide ions (specific acid-base
catalysis). This catalysis is enhanced in the presence of
other additive ions (ionic strength), buffer components
(general acid-base catalysis) and an increase in
temperature. In addition, (I) is photolabile in solution and
the photolysis is catalyzed by oxygen. The latter also plays
a role in the absence of light. Metal ions that can interact
with (I) may also catalyze the oxidative degradation of (I)
Effect of pH. The hydrolysis of (I) increases with
higher acidity, temperature and ionic strength (235,236).
The log kobs vs pH (236,237) (kobs is the pseudo-first-order
degradation rate constant) profile shows a slope of -1 up to
a pH of 3 which is indicative of a specific hydrogen ion
catalyzed reaction. In this region hydrolysis is solely
catalyzed by hydrogen ions with no influence of the
conjugate base (237). Different authors report that the pH
rate profile approaches a minimum (pH of maximum stability)
between pH 3 and 4 (238) or between 4 and 5 (237). An
earlier study (239) indicated considerably greater stability
at a pH of 2.6 compared to pH 4.8 or 5.5. In this case the
authors failed to do a detailed pH study of (I) and relied
on studies at two pH values to claim their result. At pH
values greater than 5 the overall rate of degradation
increases with increasing pH (234,237,240,241). The slope of
the alkaline phase (pH>6) of the pH rate profile gives a
value of + 0.5 (237). This slope holds true up to pH 10,
where there is an indication of a slight inflection in the
curve, from ionization of the phenolic proton of the
molecule resulting in the differential rates of hydrolysis
of the two species formed. The authors (237) considered all
four species of (I) formed from the protolytic equilibria of
the amino group (of the amino sugar), and the two phenolic
protons on the chromophore. Each one of these four species
could have its C-9 side chain in the keto or the enol form
especially at pH ranges greater than 10 (the pKa of C-17 a-
ketol side chain in hydrocortisone is 11.05 at 25C) (242).
This factor was determined to be inconsequential in the pH
range studied: 1-11. The overall observed rate constant was
fitted to the following equation:
kObs= (kO+koH- [OH-] +kH.H] ) .[H (1.1)
Here kobs was the overall pseudo first order rate constant;
ko was the first order rate constant for the solvent
catalyzed degradation; koH- and k, represented the specific
second order rate constant for the hydroxide and hydrogen
ion catalyzed degradation; and fi was the fraction of
species involved at each pH. At pH values less than 4, ko
was determined to be negligible (not different from zero).
Effect of ionic strength. In the range from 0.1 to 0.4
M, the ionic strength does not seem to play any significant
role in the hydrolysis of (I) at pH values above 4 (237).
However, there does not seem to be any agreement from
observations made below pH 4. Studies carried out in acidic
conditions (236) using HC1 as the catalyzing acid showed no
influence of the ionic strength at values up to 0.5 M. In
contrast, studies carried out in the same pH range, but
using perchloric acid as the source of protons, revealed a
linear dependance of the log of the overall pseudo first
order rate constant on the square root of the ionic strength
(slope: 0.365; intercept: -4.53) in the range from 0.058 to
0.56 M (237). At ionic strength values greater than 0.56 M
no linearity existed.
In the present studies of the hydrolysis of (I) the
ionic strength was maintained at 0.2 M throughout by adding
appropriate amounts of KC1.
Effect of solvent. There is no detailed study of the
effect of organic cosolvents (methanol, ethanol etc) on the
degradation of (I). Methanolic solutions of (I) degrade
rapidly (246). In water a 2 mg/ml solution of (I) has a pH
of around 5.5 at room temperature and is stable (10%
degradation) for at least a month (141). However, the
general consensus is that water is an inappropriate solvent
for (I) (243,244,245,246). In sterile water for injection
(USP), (I) can be stored at -20C for at least a month
(247). Eksborg (248) recommended storage of (I) solutions in
0.01-0.1 M phosphoric acid, which is ludicrous in light of
the results of other studies (236,237).
Effect of light. In solutions (I) is photolabile
(59,249-252). This decomposition is dependent on pH (251),
the nature of the solvent and the concentration of (I). The
rate of degradation was shown to be inversely proportional
to the concentration (250), but the data were fitted to a
first-order kinetic equation (250) in obvious contradiction
to its concentration dependency. Explanations given to
rationalize this anomaly were related to the intensely
strong color of concentrated solutions of (I), which in turn
self-protect themselves from photolytic degradation (250).
The fact, that there was negligible photolytic degradation
of (I) in bile supported the hypothesis, since the strong
color of bile protected the sample from photolysis. The free
radical scavenger butylated hydroxy toluene (BHT), was found
to reduce photodegradation (249), and its presence was given
as an alternate explanation for the protection properties of
bile, since animal feed is a source of BHT which gets
excreted primarily through the bile and might be present in
it (253). Williams and Tritton (252) found that (I) was
photolytically (with 366 nm light) converted to a product
that did not exhibit any cytotoxicity against Sarcoma 180
cells, probably because these products (probably polymeric
in nature) were not taken up by the cells. In photolytic
studies of (VI) (58) under anaerobic conditions only the
aglycones were observed as photoproducts, including a
product with an aromatized A-ring (see Figure 1.1) and with
the C-9 acetyl group replaced by the C-9 hydroxyl group.
Spin trapping and direct electron spin resonance studies
(254) concluded that, upon irradiation at 310 nm, (I) and
(VI) generate the superoxide anion radical and at least two
carbon centered radicals which could degrade (I) to yield
the A-ring aromatized deacetylated aglycone (58). This was
not observed when the samples were irradiated at 490 nm,
indicating that the formation of the superoxide free radical
was not the photolytic pathway in the visible region of the
spectrum. Additional studies (255) confirmed the formation
of the superoxide free radical ion via a one electron
transfer mechanism in both aprotic and protic media. The
semiquinone formed as a consequence yields a diamagnetic
dimer which further decomposes to yield the 7-deoxyaglycone
(V) of (I). Glutathione (GSH) added to buffered solutions of
(I) in clear glass vials appreciably enhanced the stability
of (I) (t1,2 with GSH = 462 hours vs 38 hours without GSH)
from photolytic degradation (256).
In the present case, care was taken to ensure that the
solutions of (I) were protected from light (wrapped in
aluminum foil) at all times and each solution was purged
with nitrogen before and during sampling to exclude oxygen
from the reaction mixture.
Effect of temperature. In general an increase in the
temperature increases the overall rate of hydrolysis of (I).
This increase follows the Arrhenius equation (257).
kobs = Ae RT (1.2)
where kobs is the overall degradation rate constant, A is the
frequency factor, Ea is the activation energy, R is the gas
constant, and T is the absolute temperature. Using elevated
temperatures that result in accelerated conditions of
hydrolysis and the Arrhenius equation, one can predict the
half-life and shelf-life at room temperature and under
refrigerated conditions (4-5C).
The Arrhenius parameters have been estimated for the
hydrolysis of (I) by various authors (236,237,238) and are
summarized in Table 1.1. At pH values less than 3.0 one can
see that the estimate of the activation energy is not
consistent among the various authors (even though specific
acid catalysis was shown) indicating that the mechanism of
hydrolysis depends on the mineral acid used.
At pH values greater than 4.0 the activation energy
varies inconsistently with pH. This suggests a different
hydrolytic pathway (and a different rate limiting step) for
the degradation of (I). The formation of more than one
degradation product above pH 4.0 accounts for the varying
estimates of Ea. It is also possible that the different pH
dependence of the Ea may be due to the varying degrees of
protonation of (I) in this pH region (237).
Janssen et al.(238) found that the Arrhenius plot
deviated from linearity between 61 and 72C and concluded
that there may be a change in the rate-controlling mechanism
in that temperature region. This appears to be unlikely, and
experimental error seems to be a more plausible cause since
the authors used a kinetic study at 4C as one of the points
to establish linearity in the Arrhenius plots. It can be
shown that the half life of degradation of (I) is very large
and prone to error at 4C. Moreover, at 4C glass binding
accounts for about 3-8% of loss of (I) from solution in the
concentration range 1-100 ig/ml (see present study) and
should be considered in the interpretation of the results.
Table 1.1: Arrhenius parameters for the hydrolysis of (I).
Ea A CONDITIONS REFERENCE
8.2 x i07
3.3 x 107
4.0 x 108
2.6 x 1011
HC1 (0.01-0.5 M)
pH = 0.43-2.13
. = 0.2 or 0.5
Ho/pH = 0.43-3.0
Acetate (0.01 M)
pH = 4.0
Acetate (0.01 M)
pH = 5.0
Phosphate (0.01 M)
pH = 6.0
Phosphate (0.01 M)
pH = 7.0
Tris (0.01 M)
pH = 7.4
1.2 x 1012
8.5 x 1010
2.3 x 10"
5.0 x 107
pH = 7.4
pH = 7.4
pH = 8.0
pH = 9.0
pH = 10.0
* In these studies g = 0.3 and [EDTA] = 5 x 10-4 M
Effect of concentration. In studies on the photolytic
degradation of (I), there was no evidence of a concentration
dependent hydrolysis of (I), even though there was a
concentration-dependent photolytic degradation (250). This
was confirmed by Poochkian et al.(258) for 10 and 20 jig/ml
samples and by Beijnen et al.(237) in the concentration
range 1-20 jig/mi. Janssen et al.(238) claimed that (I)
degrades faster in solutions containing 500 jig/ml as
compared to 50 jig/ml at pH 7.4. A possible explanation for
this discrepancy was provided by Beijnen et al.(237), who
noticed the formation of precipitates when (I) was degraded
at high concentrations. Analysis of the precipitates
revealed the presence of considerable amounts of undegraded
(I). This could disrupt the kinetics and make it difficult
to interpret the results.
The present studies on the concentration dependency
were conducted at 10 and 100 jig/ml in 0.2 M HCl.
Effect of buffers. The overall rate of hydrolysis
depended on the concentration of the buffer used to maintain
a certain pH (141,237,238). The buffer species could be a
general acid and/or base and catalyze the degradation of
(I). At pH values less than 10, catalysis by acetate,
phosphate and carbonate were demonstrated (237). There was
no buffer catalysis at pH values greater than 10. The kobs vs
buffer concentration plots deviated from linearity at higher
buffer concentrations. The authors failed to use their data
to estimate the specific second-order rate constants for
each of the conjugate acids and bases of the buffers used in
In order to evaluate the kinetics of hydrolysis of (I)
in solution, a carefully designed study was established to
consider the effect of pH, temperature, and to determine the
catalytic contribution of each buffer species on the
hydrolysis of (I). Ionic strength was maintained constant
throughout these studies to avoid interference from this
variable. Each hydrolysis study was conducted at four
different temperatures to give statistically (t-values < 3
at a = 0.05; where a is the type 1 statistical error) better
and more reliable estimates of the Arrhenius parameters
rather than the fewer than 4 conducted to justify the claims
of a Farmitalia patent (U.S. Patent # 4946831, August 7,
1990). The pH of maximum stability (from the log kobs vs pH
profile) is evaluated. Possible mechanisms and degradation
pathways are proposed.
Stability of (I) in Pharmaceutical Formulations.
The commercial preparation of (I) in its hydrochloride
form (Adriamycin, Adriablastina) is a lyophilized powder
containing 5 parts by weight of lactose as a diluent. The
compound in this form is quite stable when its moisture
content is less than 1% (234,259). These preparations might
contain trace quantities of the aglycones as impurities
which are inherent to the mode of preparation (260). In a
recent study (261) it was shown that dissolution
(reconstitution with an appropriate infusion fluid) prior to
administration was facilitated when the freeze-dried mixture
contained sub-preservative amounts of hydroxy benzoates
(methyl and propyl parabens). Once dissolved, the
anthracyclines were susceptible to hydrolytic and photolytic
The manufacturer (Adria labs, Cincinnati, Ohio) states
that the stability (>10% degradation) of the prepared
infusion solution is 24 hours at room temperature and 48
hours in the refrigerator (4-10C). The manufacturer advises
against freezing (262), as this might degrade the drug. In
contrast, some of the literature (247) suggests that, when
reconstituted and filtered, (I) is stable for 6 months in a
refrigerator (4C) and 30 days when frozen (247). Repeated
freezing and thawing does not have any effect on the
stability (247). In 0.9% sodium chloride and 5% dextrose
solution (co-formulated with methotrexate) (I) was stable
(10% degradation) for at least 1 month at -20C (263). The
stability of (I) in four different infusion fluids was
studied by HPLC at ambient temperature (21C) (258). The to.9
(shelf life: defined as the time it takes for 10%
degradation) values for (I) in 5% dextrose injection (USP)
at pH 4.5 was estimated at 100 hours; about 63 hours in 0.9%
sodium chloride injection (USP), pH 6.2; about 28 hours in
lactated Ringer's injection (USP), pH 6.3; and 24 hours in
Normosol-R, pH 7.4. This indicated a pH dependence on the
stability of (I) in the various infusion fluids with a 5%
dextrose solution being the most stable. In contrast,
Ketchum et al. (264) reported that (I) formulations
reconstituted in 0.9% sodium chloride showed no loss in
potency (determined spectrophotometrically by a non-specific
method) during storage for 28 days at 5 or 25C. Beijnen et
al.(265) studied the degradation of (I) in four different
infusion fluids in the absence of light. Using a sensitive
and specific HPLC assay they found that (I) had a shelf-life
(10% degradation) at 25C of at least 4 weeks in 5%
dextrose, pH 4.7 and a mixture of 3.3% dextrose and 0.3%
NaCI, pH 4.4. In lactated Ringer's solution, pH 6.8 the
shelf life was 1.7 days and in 0.9% sodium chloride, pH 7
the shelf life was 6 days. The stability of (I) in the
presence of vincristine (266) was investigated in three
different infusion solutions at three different
temperatures. In 0.9% sodium chloride injection solution and
0.45% sodium chloride and 2.5% dextrose injection solution,
(I) was stable (<10% degradation) in the presence of the
vinca alkaloid vincristine for at least 7 days at 15 and 4C
(<5% loss). However, when placed in a 0.45% NaCI and
Ringer's acetate injection solution at the same
temperatures, (I) was not as stable (>20% loss after 7
days). When delivered by an implanted battery operated
infusion device, (I) degraded by more than 10% in 14 days at
37C (267). In portable pump reservoirs (268) (I) at a
concentration of 2 mg/ml was stable (<10% degradation) for
up to 14 days at 3 or 23C and for an additional 28 days at
30C. In a hydrophilic ointment (I) was stable (<10%
degradation) for four weeks at 5C and at room temperature
(269). There was no significant difference in the stability
of (I) when encapsulated in liposomes (238).
Stability in Biological Media.
The analysis and stability evaluation of (I) and its
metabolites and degradation products in biological
preparations is crucial for pharmacokinetic studies, tissue
binding and uptake studies, and for binding studies to
macromolecules like DNA, plasma proteins, etc.
Heparin an anticoagulant used in blood collecting tubes
binds (I) to form a complex (132). The formation of the
complex interfered with the bioanalysis of (I) from blood
(226), and hence heparin was avoided. In studies carried out
on implantable devices in dogs, heparin and (I) could be
mixed provided the concentration of heparin was below 1.3
units/ml (270). Doxorubicin was also incompatible with 5-
fluorouracil, dexamethasone, sodium phosphate (271) and
NaHCO3 (272). With the latter, the unknown degradation
product retained its anti-proliferative property but lost
its lethality to Colo 320 (colon cancer) and CCL2 (cervical
cancer) tumor cell lines. Red blood cells degrade (I)
rapidly by a cytoplasmic aldo-keto reductase enzyme
(248,273,274). Thus, blood should be rapidly centrifuged and
the RBC removed from the collected blood (275).
In plasma (I) is stable (<10% degradation) for long
periods of time at -80C (248). However the reported
stability at -20C in plasma is conflicting (144,273).
Oosterbaan et al. (144) found (I) to be stable (<10%
degradation) for 14 days at -20C even if repeatedly frozen
and thawed, while Eksborg et al.(273) came to the opposite
conclusions. At 4C plasma samples of (I) were found to have
unchanged concentrations for at least 24 hours (273,276).
Urine samples can be preserved at -20C (244) or, after
acidification, at 4C in the dark for at least a month
(277). The biological activity of (I) (tumor cell kill) is
a) reduced by 50% in vitro in the presence of 25 mg/mL of
human albumin (278); b) affected by cell density in a
monolayer system (279); and c) dependent on temperature
(280-283). Pavlik et al.(272) showed that (I) degraded in
the dark in a tissue culture medium (not containing serum)
to a product that was cytostatic but not cytotoxic to Colo
320 and CCL2 tumor cell lines. Doxorubicin was stable under
the conditions of a clonogenic assay (284) and under storage
conditions of -40 and -196C with a t,/2 greater than 6
weeks. Under conditions of the assay at 37C (I) had a
degradation t1/2 of 29 hours.
In the conventional organic solvents that extract drugs
from biological matrices, no detailed stability study of (I)
has been reported.
Farmitalia and U.S.Patent # 4946831 (August 7, 1990).
The Italian pharmaceutical company Farmitalia Carlo
Erba, Milan, Italy filed a patent for a new ready-to-use
injectable solution of (I). It was contended that, prior to
administration (I) did not have to be reconstituted from
lyophilized powders, where spillage and loss during the
reconstitution process could occur. The formulation was a
hydrochloride salt of (I) dissolved in nitrogen-purged
sterile water for injection with the pH of the resulting
solution adjusted to 3.0 using HC1 and stored at 4C.
Kinetic and statistical analysis of the data presented
does not appear to support the claims of the patent. The
'successful' formulation presented (example 2 in the patent)
was analyzed by fitting the Arrhenius function (equation
1.2), at temperatures of 55, 45 and 35C (Table 2 in the
patent). The results of the analysis are presented in the
Table 1.2. The R2 values of the least squares plots of in
concentration vs time were 0.9976, 0.9166 and 0.9699 at 55,
45 and 35C, respectively. However, when the obtained rate
constants kobs were plotted in accordance with the Arrhenius
equation (in kobs vs l/T, where kobs are slopes of the in
concentration vs time plots and T is the absolute
temperature) an R2 value of 0.999995 was obtained. This
suggests an almost perfect (R2=1.0000) Arrhenius fit of the
data and hence an error-free estimate of the rate constants
for the corresponding half-lives and shelf-life (to.9) at 4
and 8C given in their table. It is suspicious that kinetic
studies carried out at only three higher temperatures (55,
45 and 35C) yielded statistically poor correlation
coefficients (notwithstanding that errors in rate constants
obtained at elevated temperatures are smaller than those
obtained at lower temperatures) and yet the values of the
shelf life (to.9), estimated from the Arrhenius equation were
predicted with such precision. Estimates of rate constants
at room or refrigerated temperatures extrapolated from only
three elevated temperatures have high standard errors since
they have only two degrees of freedom, in which case
statistically significant (a = 0.05) results are difficult
The claims of Farmitalia that the stability maximum was
observed at pH 3, based on their Arrhenius and kinetic data,
disagreed with the pH profile of Beijnen et al. (237), whose
log kobs vs pH profile, corrected for buffer effects, had a
pH of maximum stability between 4 and 5.
Table 1.2: Farmitalia patent data fitted to the
Arrhenius function (equation 1.2 in text).
Temp. kobs In 1/T R2* Ea#
K (wk-1) kobs K-1 kJ/mole
328 0.33769 -1.08563 0.003049
318 0.09192 -2.38684 0.003145 0.999995 112.19
308 0.02335 -3.75716 0.003247
Numbers in parentheses are the R2 values of the
regressions in the estimates of the rate constants.
Absolute temperature (273.15 + tC)
* Correlation coefficient
# Activation energy
Farmitalia's scientific rationale for developing this
formulation was flawed due to a misinterpretation of the log
kobs-pH rate profile. The pH of maximum stability in the
presence of buffers that catalyze degradation (general acid-
base catalysis) shifts towards the acid region as the buffer
concentration increases. Since studies in buffer solutions
showed the pH of maximum stability to be about 3, Farmitalia
assumed that, if pH 3 was attained by a non-buffering
species such as HC1, one would have a 'ready-to-use' product
with a better shelf life. Additionally, as demonstrated in
this thesis, (I) degrades in solution in the presence of
oxygen. This in turn results in a pH drop from 4.95 to 2.97.
Thus a solution of (I) would inherently come to a pH of
about 3 after degradation.
This dissertation attempts to: 1) eliminate some of the
discrepancies present in the literature concerning self-
association of (I) in aqueous solution, 2) evaluate the
glass binding of (I) and the effects of pH, temperature and
silylation on the binding, and 3) establish the kinetics of
hydrolysis of (I) in aqueous solution to explain unresolved
issues concerning pH of maximum stability, degradation
behavior of (I) and possible mechanisms of degradation.
MATERIALS & METHODS
Doxorubicin hydrochloride (I) was initially purchased
from Sigma Chemical Co. St.Louis, MO. It was subsequently
received as a gift from Adria Laboratories, Columbus, OH and
from Farmitalia Carlo Erba, Milan, Italy. Metabolic and
degradation products like doxorubicinol (II), doxorubicinone
(III), doxorubicinolone (13, dihydroxydoxorubicinone) (IV),
and 7-deoxydoxorubicinone (V) were received as a gift from
Farmitalia Carlo Erba, Milan, Italy, through the auspices of
Dr.Federico Arcamone. Daunomycin (VI), used as an internal
standard during chromatographic experiments, was obtained
from Sigma Chemical Co. St. Louis, MO. All samples received
were assumed to be pure and used as received, excepting (I),
which was subjected to multiple tests to evaluate its
purity. These included melting point determination,
chromatographic analysis and differential scanning
Each of the samples received was stored in glass tubes
placed in larger tubes containing an excess of anhydrous
calcium sulfate as a desicant. These tubes were protected
from light and stored in a freezer maintained from 0-5C.
All chemical reagents, pH standards and solvents used
were of HPLC or A.C.S. analytical grade and were used as
supplied (Fisher Scientific, Fairlawn, NJ).
Hexamethyldisilazane in hexane (PCR Inc., Gainesville,
FL) was used as a glass silylating agent. Specific reagents
used for specific purposes will be described in the
Water used for preparing mobile phases, buffers,
reaction media etc. was deionized, distilled water which was
boiled, cooled and filtered through a 0.2 pinm nylon membrane
filter (Rainin). Water used for any other purpose was
deionized, distilled water.
Melting Point determination.
Approximately 10 mg of (I) was powdered in a dry glass
mortar. A portion of this sample was packed into a melting
point capillary tube (0.8 X 1.10 X 90 mm; KIMAX 51, Kimble
products, USA). The tube was heated in a melting point
apparatus (Thomas Hoover Capillary melting point apparatus,
A.H. Thomas Co., Philadelphia, PA) at an average rate of
22C/min from room temperature to 220C.
Differential Scanning Calorimetry (DSC).
All samples of (I) received had their purity evaluated
using DSC. In order to reduce the degradation of (I) at
temperatures lower than its melting point, samples were
heated at a rate of 20C/min rather than the 2-5C/min
recommended for thermal equilibration. Using a micro balance
(Cahn electrobalance, Ventron Corpn., Cerritos, CA),
approximately 1 mg of (I) was accurately weighed in an
aluminum pan and crimped. This pan was placed in the DSC (PE
DSC 7, Perkin Elmer Corpn, Norwalk, CT) against a blank pan
and heated at a rate of 20C/min up to 220C. The resulting
endotherm was recorded on a graphics plotter (Perkin Elmer
2, Perkin Elmer Corp., Norwalk, CT) and the data processed
with the supplied software on the interfaced microprocessor.
To measure pH (Corning 190 pH meter, Corning Scientific
Products, England) at elevated hydrolysiss and glass binding
studies) or low temperatures (glass binding studies), the pH
meter was calibrated using standard pH buffers (pH 4,7 and
10) maintained at the temperature of the studies. These
calibrating buffers were brought to the appropriate
temperature in a water bath and using the electrode the pH
was measured and calibrated using the mV control. An
"apparent" pH was recorded in mixed aqueous-organic solvents
for use in chromatography.
Ultraviolet and Visible Spectroscopy (UV-Vis).
All UV-Vis measurements were made at ambient
temperature on a Cary 219 spectrophotometer (Varian
Associates, Palo Alto, CA). The instrument was calibrated
absorbancee at different wavelengths) using a 14.2 g/l
solution of potassium chromate in 0.1 M potassium hydroxide.
Self association. In order to assess the non-linearity
of the plots of absorbance vs concentration of (I) in
solutions at various pH's, the UV-Vis spectra of (I) were
measured in different buffers (Appendix) from pH 5 to 10.
Concentrations ranging from 5 X 10-8M to 1 X 10-4M were
recorded at each pH using longer path length quartz cuvettes
(10 and 5 cm) for the dilute solutions (<5 X 10-6M) and
regular path length (1 cm) for the more concentrated
solutions. The reference compartment holder for the longer
path length cell was designed in-house.
The molar absorptivities (E) of the monomer species of
(I) at various pH's, were determined from the slopes of the
absorbance vs concentration plots obtained in the dilute
concentration range of (I) (<5 X 10-6M).
HPLC. UV-Vis spectroscopy was used to establish the
longest wavelength absorption band for compounds (I to VI)
to be analyzed by HPLC. Appropriate amounts of the analytes
were dissolved in mobile phase 'C' (see page 61 for
composition) and the resulting solutions were scanned from
650-200 nm versus a reference cell containing blank mobile
phase (Figure 2.1). The longest wavelength absorption band
with the highest molar absorptivity was used as the
excitation wavelength in the fluorescence detector.
400 450 500 550 600
Figure 2.1: UV spectra for analytes (I)-(VI) dissolved in mobile phase C
(see page 61 of text).
Hydrolysis. UV-Vis scans were obtained from 750-200 nm,
for solutions of (I) that were degraded under various
conditions of pH, in order to detect the formation of highly
conjugated degradation products appearing red shifted
compared to (I).
The fluorescence spectrophotometer was a Perkin Elmer
MPF-44 A (Hitachi Ltd., Tokyo, Japan). Typical settings for
fluorescence measurements were: excitation monochromator at
495 nm (as determined from absorption studies); emission
monochromator scanned from 500-650 nm at 120 nm/min. Chart
speed (recorder) was 60 mm/min. When necessary, the
sensitivity setting on the fluorimeter was altered during
the self-association study to record the spectra of dilute
solutions of (I). Repeat spectra on both scales were run to
estimate the conversion from one scale to the other. The
full scale setting on the chart recorder was 10 mV.
Excitation and emission bandwidths were set at 10 nm. The
instrument was calibrated absorbancee at different
wavelengths) using a 1% quinine bisulfate solution in 0.1 M
sulfuric acid. All measurements were made at ambient
HPLC. Spectral scans of compounds (I to VI) in mobile
phase 'C' (see page 61 for composition) were obtained from
500-650 nm with excitation at 495 nm (established through UV
spectral studies). Based on the maximum emission wavelength
525 550 575 600
Fluorescence spectra for analytes (I)-(VI) in mobile phase C
(see page 61 of text).
for all compounds (Figure 2.2) an optimum wavelength of 590
nm was selected for the HPLC fluorescence detector.
Self-association. The self-association of (I) was
studied using fluorescence spectroscopy. In order to reduce
the effects due to 'inner cell effects' (concentration
quenching) the cuvette holder in the cell compartment was
modified from the conventional right angle arrangement
(where the source of incident light and the emission light
detector are at right angles to each other) to a 'front
surface illumination' set up (where the emitted light is
detected from the same surface as the incident light).
An unused cuvette holder was detached from its base
plate and then glued back on to another plate in a manner
where the incident beam of light from the excitation
monochromator struck the front surface of the cuvette at an
angle of 30 to the normal at the surface (Figure 2.3). This
orientation gave the maximum signal response for any
solution ( e.g. 1 X 10'6 M) of (I), at a fixed setting. This
also allowed for the reflected beam (also at 30) from the
front surface to be dissipated into the back of the sample
compartment and away from the emission monochromator,
avoiding interference with the emission signal (Figure 2.3).
This orientation was used throughout the self-
association study and permitted the measurement of dilute (1
X 10-M) as well as high concentrations of (I) (1 X 10-'M)
without a significant loss of signal due to inner cell
/ \-,* \
1 -< -~ FRONT SURFACE
R E B \ ILLUMINATION
i I RIGHT ANGLED
Front surface and right angled illumination
setup. Shaded area represents incomplete
penetration by excitation light.
A High Performance Liquid Chromatography (HPLC) method
using a fluorescence detection system was developed that was
specific for (I), its possible degradation products (II to
V) and the internal standard (VI).
Mobile Phase(s). Three different mobile phases were
used during chromatographic studies. Mobile phases (A) and
(B) were used for purity assessment and mobile phase (C) was
used for routine analysis of compounds (I to VI). Mobile
phase (A) consisted of 70 volumes acetate buffer (0.05 M, pH
3.98) + 30 volumes acetonitrile containing 5% v/v
tetrahydrofuran (THF). Mobile phase (B) consisted of 70
volumes of phosphate buffer (0.05 M, pH 3.86 adjusted with
orthophosphoric acid) + 30 volumes of acetonitrile
containing 3% v/v THF. Mobile phase (C) consisted of 70
volumes of 0.05 M phosphate buffer (7.8 g/l of sodium
phosphate monobasic dihydrate in water), 30 volumes of
acetonitrile and 0.4 ml/l of tetrabutyl ammonium hydroxide
(1 M solution in water). The "apparent" pH of the resulting
solution was adjusted to 3.55 with orthophosphoric acid. All
mobile phases were filtered through a 0.2 gm nylon membrane
filter (Rainin) under vacuum using a Millipore filtration
unit (Millipore Corp., Milford, MA). Each mobile phase was
recycled (except during evaluation of the reproducibility of
the calibration curve) and replaced if chromatographic
properties of the analytes changed or after about 2-3 days.
Fresh mobile phase was equilibrated with the column
overnight. Between mobile phases the HPLC system was rinsed
overnight with a "wash solvent" system consisting of
water:methanol:isopropanol in the ratio 2:1:1. This washed
out potentially damaging buffer and analyte residues in the
HPLC system, thereby increasing system lifespan.
Solvent Delivery. The mobile phase was isocratically
delivered initially by a Waters' pump (Model M-6000 A,
Waters Associates, Millipore Corp., Milford, MA) and later
by a LDC/Milton Roy pump (Model CM 4000, LDC/Milton Roy,
Riviera Beach, FL). No significant difference was observed
in the retention times or capacity factors of the various
analytes on replacing the pumps. Flow rate for mobile phase
(A) and (B) was optimized at 1.0 ml/min and for (C) at 1.2
Column. The stationary phase consisted of a Whatman
Partisil (Whatman Inc., Clifton, NJ) ODS (C-18) column (25
cm X 0.45 cm i.d.; 5 pm particle size) for mobile phases (A)
and (B). The column for mobile phase (C) was a Whatman
Partisil (Whatman Inc., Clifton, NJ) ODS (C-18) (10 cm X
0.45 cm i.d.; 5 pm particle size). Both analytical columns
were protected by a 1 cm guard column packed in-house with a
Permaphase ODS 10 pm material (Dupont Instruments,
Wilmington, DE) using a traditional tap and fill technique.
Injector. All injections were made with a 50 pl syringe
(Hamilton Co., Reno, NV) through a Rheodyne 7125 high
pressure injector (Rheodyne Corp., Cotati, CA) fitted with a
20 p.1 loop.
Detector. Analytes eluting from the column were
detected by a Perkin Elmer 650S fluorescence detector
(Hitachi Ltd., Tokyo, Japan) with excitation and emission
wavelengths set at 470 nm and 553 nm, respectively for
mobile phases (A) and (B). The excitation and emission
wavelengths were optimized to 495 nm and 590 nm,
respectively, in the case of mobile phase (C). Excitation
and emission bandwidths were set at 10 nm. Power to the lamp
was supplied by a Perkin Elmer 150 B Xenon Power supply unit
(Hitachi Ltd., Tokyo, Japan).
Recorder/Integrator. Data was recorded on a HP 3394 A
integrator (Hewlett Packard Cc-p., San Fernando, CA) and a
Fisher recordall series 5000 strip chart recorder (Bausch
and Lomb, Houston Instrument division, Austin, TX). Peak
heights recorded on the strip chart recorder were measured
for data analysis. The integrator data was maintained as a
means of peak identification (printed retention times),
column performance (Area/Peak height ratios) and for
estimation of the peak heights that were "off" scale on the
strip chart recorder (using peak area ratios).
In studies involving glass binding and hydrolysis, the
temperature was rigorously controlled ( 0.1C) throughout
Water bath(s). Conditions designed to mimic those in a
refrigerator were maintained by a Lauda K-2/R refrigeration
unit (Brinkmann Instruments, Germany). Water at 5.0 + 0.1C
was delivered by gravity assist to a reaction bath and was
recycled back to the refrigeration unit by a pump. Ambient
and high temperature settings ( 0.1C) were maintained by a
Haake water bath and alternately an Isotemp Immersion
Circulator (Model 70; Fisher Scientific, Pittsburgh, PA).
All glassware in contact with (I) was pre-silylated.
Each item was thoroughly washed with detergent, rinsed with
water and methanol and dried in a hot air oven at 160C for
30 minutes. The dried glassware was allowed to cool and
about 8-20 ml of hexamethyldisilazane (silylating agent) in
hexane was added. The liquid was swirled in the glassware so
that all exposed surfaces of the glass would be coated. The
silylating liquid was used sequentially for up to four
times. Silylated glassware were allowed to drain and dry in
air, rinsed with methanol and dried in a hot air oven for 2-
3 hours. This glassware was used for handling of (I)
solutions. Care was taken to avoid scratching the inside
surface of the treated glassware in order that the
unprotected surface would not be reexposed.
Stock solutions. All stock solutions of (I) (10-s, 10-4,
10-3 M) used for the self association study were made in
deionized, distilled, deoxygenated (with nitrogen) and
filtered water. These solutions were protected from light
and stored in silylated glass volumetric flasks in the
refrigerator at 5C without freezing, and allowed to warm to
ambient temperature before use. All stock solutions were
discarded after one week.
Sample preparation. In order to minimize losses that
might occur due to hydrolysis, precalculated aliquots of (I)
from stock solutions were pipetted directly into 1 cm square
quartz cuvettes (fluorescence studies) and 5 or 10 cm
cylindrical quartz cuvettes absorbancee measurements).
Immediately prior to spectral recording, sufficient volume
of the appropriate buffer was pipetted into the cuvettes to
yield the desired final concentrations of (I). Between
successive samples the cuvette was thoroughly rinsed with
acetone and distilled water.
Normalization and Selective Excitation Measurements. In
order to determine the presence of a longer wavelength
emission band (if the dimer fluoresces) than a 'monomer
spectrum', the latter was recorded at a concentration of 5 X
10'7 M (it was assumed that dimerization was negligible at
this concentration, since concentrations of dimer formed
will be < 10% of the total concentration of (I) used if the
association constant is assumed to be 4 X 10' M-1 the
largest value cited in the literature (202)). This spectrum
had two maxima, at 560 and 590 nm (cf. Figure 3.5). Each of
these maxima was selected as a 'Normalizing Wavelength'. The
concentration of (I) solution was progressively increased in
the cuvette and at each increase the emission spectra of (I)
was scanned till 560 nm (or 590 nm) when the recorder pen
would show a higher fluorescence intensity than the 'monomer
spectrum' for the 5 X 10'7 M solution. Using the 'coarse'
and 'fine' sensitivity selectors on the spectrophotometer,
the emission signal at 560 nm (or 590 nm) was attenuated
such that the recorder pen overlay the spectral recording
for the 'monomer spectrum'. Using this 'new' instrument
setting the spectrum of the concentrated solution was
recorded and overlayed on the 'monomer spectrum'. If the
dimer, fluoresces, the overlayed spectra at higher
concentrations would be shifted towards longer wavelengths
than the 'monomer spectrum' (cf. Chapter 3 for explanation).
Using the technique of El-Sayed et al. (285), and the
results published by Arcamone et al. (172), the excitation
monochromator was set at 580 nm (a wavelength at which the
dimer spectra does not overlap the monomer spectra) and the
emission scanned from 580 to 710 nm. Since the absorbance of
Typical fluorescence spectra of (I) in
aqueous solution, a: 1 X 10-4M; b: 1 X 10-6M.
the dimer at 580 nm is low, the sensitivity settings of the
instrument were maximized to allow for the weakest signal to
Sample Spectral Measurements. Using the modified cell
holder, spectral recordings of (I) from 1 X 10-7 to 1 X 10-4
M were obtained at three different pH's (5.72, 6.72, 7.69)
with each maintained at three different ionic strengths
(0.01, 0.05, 0.1 M). For estimating the ratios of the
quantum efficiencies of the monocation (MC-DOX) and neutral
(N-DOX) species of (I) (ON-DOX/MC-DOX) fluorescence
measurements were recorded for 1 X 10-6 and 1 X 10-4M
solutions, each maintained at pH's of 5 and 10 to bracket
the pH range of the first ionization (pKI) constant of (I)
in solution (cf. Chapter 3: Results and Discussion for
rationale). To determine the molar absorptivity of the
monomer species of (I), spectral scans of (I) solutions in
concentration ranges from 8 X 10' to 5 X 10-M were obtained
in buffers of pH's 6, 7, 8 and 9 at ionic strengths of 0.01,
0.05, 0.1 and 0.2 M each. Since the molar absorptivity of
the monomeric monocation (MC-DOX) and neutral (N-DOX)
species of (I) were assumed to be identical (cf. Chapter 3:
Results and Discussion), they were averaged across the pH
range studied. Compositions of buffers for the above studies
are detailed in the Appendix.
Data Treatment. Typical fluorescence spectra are
shown in Figure 2.4. Peak heights were measured at 590 nm in
terms of relative fluorescence intensity (relative to a
blank) and plotted versus concentration. The absorbance at
495 nm was recorded versus the concentration and used to
calculate the molar absorptivity. The fluorescence data was
used to develop an equilibrium model to explain self-
association. The model development is outlined in Chapter 3:
Results and Discussion.
High Performance Liqcruid Chromatoqraphy.
Samples used to generate calibration curves for HPLC
studies were prepared in mobile phase (C) for all analytes.
Appropriate volumes from stock solutions (= 10 .g/ml) of the
various analytes (I-V) were placed in silylated 3 ml glass
tubes. An appropriate volume of (VI) as the internal
standard was added to yield a final concentration of 500
ng/ml. Calibration curves were generated in the
concentration range from 25-700 ng/ml for all analytes. To
establish the ruggedness and the reproducibility of the
chromatographic system, calibration curves were generated on
various days using the same and/or different mobile phase.
Additional calibration curves were generated prior to and
after each study to ensure consistency of the
chromatographic system. In cases where a study proceeded
over a period of days, calibration samples were run at the
beginning of each day and also during the study. The slopes
and intercepts obtained from the calibration curves
generated for a study were averaged over the number of days
of the study and used for the estimation of concentration.
Assay validation. Validation was performed by first
measuring peak heights of the HPLC calibration samples using
a standard ruler (with 1 mm rulings). The peak height ratio
was calculated by dividing the measured peak height of the
analyte by that of the internal standard. Plots of the peak
height ratio vs the formal concentration of the analytes
gave rise to calibration curves. Simple linear regression of
the calibration curves gave the slope, intercept,
correlation coefficient, and the standard errors of the
slope and regression. Analysis of Variance (ANOVA) of the
slope and the intercept between days and with different
mobile phases gave interday variability. Based on the
signal-to-noise ratio (S/N), and the regression parameters
of the calibration curves, information on Limits of
Detection (LOD), Sensitivity and Minimum Quantifiable Limit
(MQL) were obtained. The results and interpretation of these
analysis are in Chapter 3: Results and Discussion.
The non-specific binding of (I) to glass was studied as
a function of concentration, pH and temperature.
Effect of Temperature. Glass binding was studied at
three different temperatures: 5C, 25C and 50C. The
temperatures were selected to mimic the binding of (I) to
glass in the refrigerator, at ambient temperature (room),
and at elevated temperature (e.g., in a hydrolysis
experiment), respectively. All buffer solutions used were
pre-equilibrated at the appropriate temperature before start
of the study.
Effect of pH. The influence of pH and the ionization
of (I) was achieved using buffers at pH 3.18 (acetate 0.05
M), 5.01 (acetate 0.05 M) and 7.00 (phosphate 0.05 M) (see
Appendix). Each buffer was thoroughly deoxygenated with
nitrogen and equilibrated at the temperature of the study.
Effect of Concentration. The percent loss of (I) due to
glass binding from dilute solutions was expected to be
greater than from concentrated ones. To evaluate the extent
of binding as a function of concentration 1, 10 and 100
pg/ml solutions of (I) were studied at each of the pH's and
temperatures mentioned above. Increasing concentrations
should lead to saturation of the binding sites that would
show as a negligible loss due to glass binding.
Experimental. Corning 50 ml (15 ml for 100 gJg/ml
solutions) test tubes with ground glass stoppers and a
tapered bottom were used for the study. Appropriate aliquots
of (I) from stock solutions (2 mg/ml) were pipetted into
temperature equilibrated deoxygenated buffer solutions to
yield 30 mL (10 mL for 100 gg/mL) of the appropriate
concentrations of (I). This volume allowed for a limited
head space to be maintained above the liquid level and thus
reduced the diffusion of air into (and subsequent oxidation
of) the sample. To further prevent oxidation the head space
was constantly flushed with nitrogen before, during, and
after every sample was taken. The nitrogen was humidified at
the temperature of the study. Each study was conducted in
silylated and unsilylated tubes as a means of comparison and
to study the effectiveness of the silylation process. Due to
the large surface area of contact/unit volume of the tubes,
there was negligible change (<5%) in the contact area at the
end of a study (after multiple sampling). All tubes were
wrapped with aluminum foil to protect samples from
photolytic oxidation. Each study was run in triplicate.
Sampling. Sample volumes for analysis depended on the
concentration of (I) used in the study. In a clean
previously silylated glass tube 200, 50 and 10 gl samples
were withdrawn from the 1, 10 and 100 gg/ml reaction
solutions respectively and plunged into an ice bath. To each
solution an appropriate volume of (VI) was added as an
internal standard to yield a final concentration of 500
ng/ml. The mobile phase was used to make up volumes to 1 ml
for the 1 and 10 ug/ml solutions and 2 ml for the 100 gg/ml
solution. Solutions were sometimes stored on ice for a
period not exceeding 24 hours prior to analysis. Twenty
microliters of each solution was injected into the HPLC for
Data Treatment. The Peak Height Ratio (PHR) of (I)
was calculated. Using the most recent, (generated before the
study) calibration curve the PHR were converted to the
corresponding concentrations. The percent of the initial
concentration remaining and the amount of (I) bound to the
glass were determined as a function of time. The kinetic
data curves generated were fitted to a sum of exponentials
which in turn yielded important binding parameters.
Chromatograms for studies undertaken at 50C showed the
appearance of the aglycone (IV), making it difficult to
quantitate the extent of binding at this temperature.
The kinetics of hydrolysis of (I) was studied under
various conditions of pH, temperature and buffer
concentrations in order to establish log kobs-PH profiles, pH
of maximum stability, the Arrhenius activation energy (Ea)
and frequency factor (A), the existence of general and
specific acid-base catalysis, the possible mechanisms,
pathways and structures of degradation products.
Effect of acids. In order to evaluate the effect of the
conjugate bases of different mineral acids, hydrochloric,
perchloric and sulfuric acids were used to study the
hydrolysis of (I) at 50C. Further, to compare the activity
of the hydrogen ion obtained from these acids, their
concentrations (0.01, 0.05, 0.1, 0.2 M) and ionic strengths
(0.2 M with KC1) were kept constant. The activity of the
hydrogen ion (as determined by its activity coefficient)
varied as a function of the temperature, ionic strength and
the conjugate species present in solution (289). In the
present case measuring the pH at the temperature of the
study (i.e., 50C) for concentrated acid solutions was not a
reliable estimate of the activity of the hydrogen ions in
solution, since significant errors develop in glass
electrodes under these conditions. Estimates of the activity
coefficients of the hydrogen ions under the influence of the
above mentioned combination of conditions were obtained from
Harned and Owens (289). In cases where the literature did
not provide the relevant information, estimates of the
activities of the hydrogen ions were extrapolated from the
available data using the extended Debye Huckel equation
(257). These extrapolations and the rationale for their use
are presented in detail in Chapter 3: Results and
Effect of concentration. The concentration dependence
of the hydrolysis, was studied by evaluating the reaction
rate of 10 and 100 gg/ml solutions of (I) in 0.2 M HC1 at
Effect of Temperature. To obtain Arrhenius parameters
and to predict rate constants of degradation at ambient
storage conditions (room temperature or refrigerated
temperature), accelerated studies were carried out at four
elevated temperatures in each of the buffer systems. Four
different concentrations of each buffer system at a given
temperature and pH were studied. The temperature studies
were carried out at 70C, 65C, 60C and 55C for all buffers
and 50C for HC1 studies instead of 55C.
Effect of pH. The hydrolysis study focused
predominantly on the acid-catalyzed hydrolysis of (I) in
solution. The pH from 0.70 to 2.90 was generated by using
different concentrations of HC1. The pH range from 3.00 to
8.00 was maintained by various buffers starting with format
buffer for pH 3.00 and 4.00; acetate for pH 4.50, 5.00 and
5.50 and phosphate for pH 6.00, 7.00 and 8.00 (see
Appendix). Each pH study was conducted at the temperatures
mentioned above. The pH of each solution was measured at the
temperature of the study using an electrode that had been
calibrated using standard buffers at the same temperature.
Effect of buffers. The general acid/base catalysis of
(I) was studied using each buffer at four different
concentrations, taking care to maintain the pH and the ionic
strength constant. The format and acetate buffers were made
at concentrations of 0.01, 0.05, 0.1 and 0.2 M at a constant
ionic strength of 0.2 M (with KCl). In case of the phosphate
buffers the concentrations were reduced to 0.01, 0.05, 0.1
and 0.12 M for pH 6.00; 0.01, 0.025, 0.05 and 0.075 M for pH
7.00 and 0.01, 0.025, 0.05 and 0.065 M for pH 8.00 (see
Appendix). This compensated for the presence of the
monobasic and dibasic phosphate ion, each of which
contributes to the ionic strength. Each buffer solution was
deoxygenated with nitrogen before the pH was adjusted.
Experimental. A fixed volume of each of the buffers
was equilibrated to the desired temperature in 50 ml
silylated test tubes protected from light. Each buffer was
thoroughly deoxygenated with nitrogen. An appropriate
aliquot of a stock solution of (I) was added to each tube
(time, t=0) to yield a final concentration of 10 pg/ml (100
gg/ml for concentration dependence studies). At various time
intervals during the study (selected on the basis of the
predicted or previously reported half-lives at that
temperature, pH and buffer concentration) 50 gl aliquots of
the test sample were transferred to previously silylated
test tubes and cooled in an ice bucket (to arrest any
further reaction). Appropriate amounts of (VI) were added as
an internal standard to yield a final concentration of 500
ng/ml. The remaining volume was made up to 1 ml with the
mobile phase. Twenty microliters of this solution was
injected into the HPLC. Some solutions (that could not be
analyzed on the day of sampling) were stored in an ice
bucket without freezing for no longer than 24 hours prior to
analysis without any detrimental effects. Glass binding did
not interfere with this phase, probably due to the presence
of the organic component acetonitrilee) in the mobile phase.
Data Treatment. As described for the glass binding
studies, the peak heights of (I) and its degradation
products, as well as the internal standard (VI), were
measured and the corresponding peak height ratios were
calculated. Using the most recently generated calibration
curves or the mean parameters of a series of calibration
curves generated prior to, during and at the end of each
study, the peak height ratios were converted to
concentrations. These data were converted to percent of (I)
remaining at the corresponding times and fitted to a
monoexponential decay equation (in the ln C vs t form). The
resulting slope gave the overall pseudo first order observed
rate constant kobs for hydrolysis. This rate constant was
kinetically processed to yield Arrhenius parameters (plots
of ln kobs vs 1/T), second order rate constants (when plotted
versus [H']) for the various catalytic species ([H], [OH-]
and [H20]) and second order rate constants (when plotted
versus buffer concentration) for the general acid base
catalysis by the various species that constituted the
buffers. The respective kinetic treatments are presented in
detail in Chapter 3: Results and Discussion.
Identification of degradation products.
During buffer catalyzed hydrolysis of (I), sample
chromatograms showed various peaks which did not match
(based on retention times) the chromatogram of the standards
(I-VI). Therefore an attempt was made to isolate and, if
possible, to elucidate the structure of these degradation
products. In addition, all solutions undergoing hydrolysis
(from pH 0.79 to 8) showed the appearance of a precipitate.
This precipitate was also collected and an attempt was made
to analyze it.
Analysis of the solution and precipitate (obtained
after hydrolysis) in the pH range from 0.8 to 3. The
hydrolysis study in the pH range from about 0.8 to 3 showed
the appearance of one additional peak on the chromatograms,
having the same retention time as the aglycone (IV). The
aglycone was quantitatively generated from (I) in this range
as shown in Chapter 3: Results and Discussion. The
precipitate was collected, washed with cold water and then
dissolved in the mobile phase. HPLC analysis of the
resultant solution showed the presence of only one peak
having the same retention time as the aglycone (IV). Thus,
the aglycone was the only degradation product formed in this
Analysis of the solution and precipitate (obtained
after hydrolysis) in the pH range from 4 to 8. In all
solutions at pH>3, the formation of the aglycone was
negligible. Instead unknown peaks were observed (see Chapter
3 for details), that did not fit the chromatograms of the
known analytes (I-VI). At the end of a study the test
solution acquired a deep pink color (instead of the orange
red color of (I)) indicating the formation of a product that
appeared to be more conjugated than (I).
The solutions and precipitates obtained at the end of
the hydrolysis were subjected to UV-Vis spectroscopy as
described in 'Ultra violet and Visible Spectroscopy:
Hydrolysis'. The results are discussed in Chapter 3: Results
Solutions of the precipitate prepared in mobile phase
and injected into the HPLC showed the same chromatograms as
the solutions themselves. Isolation of the various peaks by
manual fraction collection did not yield sufficient
quantities for characterization and could not be separated
from the buffer components of the mobile phase.
Effect of oxygen on the hydrolysis. In order to study
the effect of oxygen on the hydrolysis of (I), three 50 ml
silylated tubes A, B and C each protected from light were
filled with approximately 10 ml of deionized, distilled,
boiled and filtered (2pm filter) water. Each tube was
preequilibrated to a temperature of 70C in a water bath.
Nitrogen was continuously bubbled through the water in tubes
A and B, and tube C was equilibrated with oxygen. Just prior
to the start of the study, 8 mg/ml stock solution of (I) was
prepared in deionized, distilled and deoxygenated water. An
aliquot of the stock solution was pipetted into each of the
equilibrated tubes to yield a final concentration of 2 mg/ml
(concentration of the reconstituted formulation of (I) used
in clinical practice), and the tubes were stoppered. The pH
of the solutions were measured throughout the study using an
electrode calibrated at 70C. The appropriate gas was
bubbled through each tube for 1 minute to displace any air
that might have entered the tubes during pH measurements. At
a predetermined time (based on a similarly run pilot study)
the nitrogen gas that was bubbled through tube B was
replaced with oxygen until the conclusion of the experiment
(no change in pH).
Due to a drop in the pH noted (from about 4.8 to 2.9)
in the tube equilibrated with oxygen, the formation of an
acidic product was thought to have formed. Kinetic treatment
of the resulting pH vs time data suggested the formation of
a product having a pKa around 3 (see Chapter 3: Results and
Discussion for kinetic development). Since this pH drop was
not observed in solutions equilibrated with nitrogen, the
formation of the acidic product appeared to be due to an
oxidation process. Beijnen et al. (237) had suggested that
the side chain of (I) could break apart and tautomerize to
form glycolaldehyde (hydroxy-acetaldehyde). The logical
explanation to the formation of an acidic product was the
oxidation of the aldehyde to the corresponding hydroxy
acetic acid (X-COOH), which could be responsible for
lowering the pH of the oxygenated solution.
Detection of product X-COOH by potentiometry. An
experiment similar in design to the one above was set up to
generate the acidic product and titrate it with dilute NaOH
to determine its pKa. Consequently about 20 ml of a 2 mg/ml
solution of (I) was prepared in deionized, distilled and
filtered water which had been equilibrated at 70C. The
solution was allowed to react with oxygen bubbled
continuously through the solution. A pH electrode calibrated
at 70C was placed in the solution to record the pH of the
solution. At the end of the experiment (no further change in
pH took place), the flow of oxygen was stopped and a
magnetic stirrer bar was dropped into the solution. Aliquots
of 10, 50 and 100 gil of a 0.01 M or 0.1 M solution of NaOH
were added to the solution at 70C with continuous stirring.
The pH after each addition was recorded until no further
change in pH could be noticed (pH>12). A blank solution was
titrated in a similar manner and the difference between the
sample solution and the blank was plotted using the Parke-
Davis method (290) as a function of the measured pH.
Detection of product X-COOH by ion-pair extraction and
HPLC. Aliquots of the hydrolyzedd' solutions from each tube
(A, B and C) were adjusted to a pH between 6 and 7 with
NaOH. This aqueous mixture was saturated with NaCI and
extracted repeatedly with small volumes of
chloroform:isopropanol (4:1 v/v) until colorless. The NaCI
helped to salt out the colored component into the organic
phase. To the remaining aqueous phase, 0.5 ml
tetrabutylammonium hydroxide (1 M) was added as an ion pair
agent. The resulting solution was extracted three times
using small volumes of diethyl ether. The ether extracts
were pooled and evaporated to dryness. The residue obtained
was dried under vacuum.
A HPLC system was set up consisting of a 25 cm reverse
phase C-18 column. The mobile phase was a 85:15% v/v of 0.05
M tetrabutylammonium hydroxide in water: acetonitrile.
Detection was by UV with a X. set at 210 nm. Flow rate was
1.0 ml/min. The residue obtained from the ether extracts was
reconstituted in the mobile phase and 20 gl of the sample
was injected into the HPLC. Standard solutions of known two
carbon carboxylic acids glycollicc, glyoxylic, oxalic etc.)
were prepared and injected into the HPLC as a means of
Detection of product X-COOH by derivatization of the
aldehyde. Since the acidic product X-COOH was theorized to
have formed from the oxidation of an aldehyde, another
method of confirming the existence of X-COOH would be to
detect its precursor. Consequently a 2 mg/ml solution of (I)
in water was heated to 70C. Nitrogen gas was continuously
bubbled through this solution and the gas outflow was led
into a methanolic solution of 2,4-dinitrophenylhydrazine.
The appearance of an 'osazone' precipitate would confirm the
presence of an aldehyde.
Precipitate analysis by LC-MS: The precipitate
generated in the oxygen study showed the same
chromatographic profile as that obtained in the hydrolysis
study. This precipitate was collected, washed with cold
water and dried in a vacuum desiccator. A Kratos LC-MS with
a thermospray attachment (vaporizes the LC column effluent
prior to mass spectral analysis) was set up. In order to
maximize the efficiency of the thermospray assembly, the
mobile phase of the LC was devoid of non-volatile buffers
like phosphate, carbonate, etc. In addition sodium and
potassium salts and organic modifiers like
tetrabutylammonium hydroxide that can create occlusion of
the exit pore of the thermospray assembly, were avoided. A
fresh mobile phase was prepared consisting of a 70% solution
of 0.05 M ammonium acetate (pH 3.5, adjusted with glacial
acetic acid) and 30% acetonitrile. This mobile phase also
separated and resolved all the degradation peaks as mobile
phase C (see 'Methods: Mobile Phase(s)'). The only
difference in the chromatograms was that the aglycone (IV)
peak appeared later than the internal standard (VI). The
precipitate obtained was dissolved in 1 ml of the mobile
phase and about 100 gl of this solution was injected into
the LC-MS. The high concentration of the acetate buffer
interfered with the analysis and no satisfactory results
could be obtained.
RESULTS AND DISCUSSION
The melting point measured for each of the lots of (I)
hydrochloride received, showed similar characteristics. In
all cases (I) melted at 205C with decomposition (sample
decolorized from a reddish orange to a charred ashy
residue). If the samples melted and discolored at a
temperature of 205C, as mentioned in the literature (141),
they were considered pure. All the samples received passed
the melting point test.
Differential Scanning Calorimetry (DSC).
A typical DSC endotherm of the melting of (I) is shown
in Figure 3.la. The sample was rapidly heated to its melting
point, the endotherm recorded and the purity estimated using
a derived form of the van't Hoff equation (3.1).
T = T T02 .1 (3.1)
where Ts (K) is the sample temperature, To (K) is the
melting point of a 100% pure sample, R is the gas constant
(8.314 J/mole.K), AHf0 is the molar heat of fusion (J/mole),
X2 is the mole fraction of the impurity and F is the
0 a a
190 200 210 220 230
Figure 3.1 a: A typical DSC endotherm of the melting of (I) for purity evaluation.