Investigating the mechanisms of aromatic amine-induced protein free radical formation by quantitative structure-activity relationships: Implications for drug-induced agranulocytosis

Article abstract of DOI:10.1021/tx900432d

Aromatic amine drugs have been associated with agranulocytosis (neutrophil depletion) for which the mechanism is unknown. We have previously shown that the metabolism of two aromatic amine drugs by human myeloperoxidase (MPO) results in phenyl radical metabolite formation and also in protein free radical formation on MPO. Because the concentration of drug required to produce a maximum signal for MPO protein free radical (MPO^?) detection was different for each drug, this prompted us to consider that other aromatic amines may also show varying degrees of ability to induce MPO^? formation. Immunoassay experiments using the immuno-spin-trapping technique were performed, which evaluated the potency of different aromatic amines containing the aniline substructure to generate the MPO^?. Each reaction contained equal amounts of H₂O₂, 5,5-dimethyl-1-pyrroline- N-oxide, MPO, and variable concentrations of aniline derivatives. Several physicochemical parameters for aniline derivatives were used to derive quantitative structure-activity relationship equations, which showed that the Hammett constant (-) best correlated with the MPO^? formation for all aniline derivatives. More statistically robust equations were derived if the anilines were separated into mono- and disubstituted groups. However, some aniline derivatives did not induce MPO^? formation. Using electron spin resonance spectroscopy, we evaluated the ability of all aniline derivatives tested to produce phenyl radical metabolites, as previously shown by spin trapping for the aromatic amine drugs. Interestingly, we found that only those aniline derivatives that produced a phenyl radical also formed MPO ^?. We propose that the phenyl radical is the reactive free radical metabolite responsible for generating the MPO^?+.

Full text of DOI:10.1021/tx900432d

880
Chem. Res. Toxicol. 2010, 23, 880–887
Investigating the Mechanisms of Aromatic Amine-Induced Protein
Free Radical Formation by Quantitative Structure-Activity
Relationships: Implications for Drug-Induced Agranulocytosis
Arno G. Siraki,*,† JinJie Jiang,‡ and Ronald P. Mason‡
Faculty of Pharmacy and Pharmaceutical Sciences, UniVersity of Alberta, 3133 Dentistry/Pharmacy Centre,
Edmonton, Alberta, Canada T6G 2N8, and Laboratory of Pharmacology, National Institute of EnVironmental
Health Sciences/National Institutes of Health, Research Triangle Park, North Carolina 27709
ReceiVed December 1, 2009
Aromatic amine drugs have been associated with agranulocytosis (neutrophil depletion) for which the
mechanism is unknown. We have previously shown that the metabolism of two aromatic amine drugs by
human myeloperoxidase (MPO) results in phenyl radical metabolite formation and also in protein free
radical formation on MPO. Because the concentration of drug required to produce a maximum signal for
MPO protein free radical (MPO•) detection was different for each drug, this prompted us to consider that
other aromatic amines may also show varying degrees of ability to induce MPO• formation. Immunoassay
experiments using the immuno-spin-trapping technique were performed, which evaluated the potency of
different aromatic amines containing the aniline substructure to generate the MPO•. Each reaction contained
equal amounts of H2O2, 5,5-dimethyl-1-pyrroline-N-oxide, MPO, and variable concentrations of aniline
derivatives. Several physicochemical parameters for aniline derivatives were used to derive quantitative
structure-activity relationship equations, which showed that the Hammett constant (σ) best correlated
with the MPO• formation for all aniline derivatives. More statistically robust equations were derived if
the anilines were separated into mono- and disubstituted groups. However, some aniline derivatives did
not induce MPO• formation. Using electron spin resonance spectroscopy, we evaluated the ability of all
aniline derivatives tested to produce phenyl radical metabolites, as previously shown by spin trapping
for the aromatic amine drugs. Interestingly, we found that only those aniline derivatives that produced a
phenyl radical also formed MPO•. We propose that the phenyl radical is the reactive free radical metabolite
responsible for generating the MPO•.
induced agranulocytosis, which is due in part to the low
frequency of this side-effect as well as the lack of suitable
models of the condition.
Introduction
Aromatic amine xenobiotics are associated with numerous
acute toxicities, including methemoglobinemia and hemolytic
anemia, whereas chronic exposure to certain aromatic amines
(such as 4-aminobiphenyl) is associated with cancer. Oc-
cupational exposure has been a focal point for the determination
of biomarkers to evaluate human exposure. Drug molecules
containing the aromatic amine or aniline substructure are also
associated with toxic side-effects, including agranulocytosis, the
depletion of neutrophils (<500/ µL), which increases the risk
of sepsis. The mechanism of drug-induced agranulocytosis is
not clear; however, it is believed that reactions resulting from
the formation of reactive drug metabolites are involved (3, 4).
The reaction center for aromatic amines is the aniline moiety.
In the instance of procainamide-induced lupus, N-acetyl-
procainamide, which also has pharmacological activity, was
shown to significantly delay the onset of antinuclear antibody
formation (5). The N-acetyl group is electron withdrawing and,
therefore, deactivating with respect to oxidation, suggesting that
oxidation of the aromatic amine of procainamide to a reactive
metabolite may play an etiological role. Furthermore, procaina-
mide-induced lupus developed more rapidly in slow acetylator
phenotype individuals (6). No similar findings exist for drug-
The possibility of the involvement of myeloperoxidase
(MPO)-generated free radical metabolites in drug-induced
agranulocytosis has been proposed for some time (7). Peroxidase
enzymes are capable of oxidizing numerous substrates to free
radical metabolites (8). Since MPO is located in neutrophils, it
has been suggested as a catalyst for reactive metabolite
generation that could lead to neutrophil death (9). It was recently
found that in a case of dapsone-induced agranulocytosis, the
patient presented an absence of neutrophils, eosinophils, and
CD34+ progenitor cells, but basophil levels were unchanged
(10); this finding was attributed to reactive metabolite formation
by peroxidases (MPO and eosinophil peroxidases) since baso-
phils are typically low in peroxidase activity.
We previously found that myeloperoxidase catalyzes the
oxidation of two aromatic amine drugs, which leads to the
formation of protein free radicals on MPO in the HL-60 cell
line as well as in purified enzyme preparations (1, 2). In this
study, we build upon our previous findings to investigate the
factors responsible for these observations. We focus on the
physicochemical properties of aniline derivatives (representative
structures shown in Figure 1a) that lead them to generate protein
free radicals, and developed quantitative structure activity
relationships (QSARs, Hansch analysis) that relate their
potency in generating protein free radicals to their physico-
chemical properties. In addition, we examine the relationship
* To whom correspondence should be addressed. Fax: + 1-780-492-
1534. E-mail: siraki@ualberta.ca.
† University of Alberta.
‡ National Institute of Environmental Health Sciences/NIH.
10.1021/tx900432d  2010 American Chemical Society
Published on Web 03/03/2010

QSARs for Protein Free Radical Formation
Chem. Res. Toxicol., Vol. 23, No. 5, 2010 881
the plate was washed and blocked overnight with 4% cold fish
gelatin, and anti-DMPO detection of MPO-DMPO was carried
out as described previously (1). The concentration of aniline
derivative necessary to induce a 2-fold increase above back-
ground for detection of the myeloperoxidase protein free radical
(MPO•) was designated the EC2 value.
Electron Spin Resonance Spectroscopy Detection of
Aniline Free Radical Metabolites. Free radical metabolites
were detected by electron spin resonance (ESR) spectroscopy
using the spin-trapping technique (13). The spin trap we used
was MNP, which we previously found provided significant detail
for revealing the structure of the free radical that was formed
(1, 2). ESR spectra were recorded using an Elexsys E500
spectrometer (Bruker Biospin Ltd., Billerica, MA) using an ER
4122 SHQ cavity. The settings for most spectra were recorded
at 9.78 GHz, 100 kHz modulation field, 20 mW power, 0.4 G
modulation amplitude, 1024 points, and 6.32 × 105 receiver
gain. In certain cases where the spectra were not well resolved,
we used the following modifications to these settings: 1 mW
power, 0.07 G modulation amplitude, conversion time and time
constant of 2621.44 ms, 2048 data points, and 5368 s sweep
time. In order to generate free radical metabolites, we added
10 mM MNP, 5 mM aniline, 1.7 µM HRP, and 1 mM H2O2 in
200 µL buffer (described above), briefly vortexed the reaction,
and added it to a quartz flat cell for recording spectra. We chose
these concentrations based on preliminary experiments.
Physicochemical Parameters and QSAR Derivation. The
Hammett constant (σ) values were obtained from (14); lipo-
philicity (experimental values of the log octanol/water partition
coefficient, or log P) were retrieved from the ALOGPS 2.1
program (15); and ionization potential of the parent aniline (IP-
P), electron affinity of the parent aniline (EA-P), and ionization
potential and electron affinity of the corresponding phenyl
radical metabolites (IP-R, EA-R, respectively) were calculated
using MOPAC with PM3 parameters after molecular mechanics
geometry optimization (MM3/PM3) in Sybyl 7.1. Free radical
enthalpy of formation values has been previously calculated with
PM3 parameters (16). QSAR equations were derived using
SigmaPlot 11.0. Best subset regression was used to identify the
physicochemical parameter that best correlated with the EC2
values, and the inclusion of additional parameters was evaluated
to see whether or not they would improve the statistical
robustness of the equation.
Figure 1. a. Representative structures of compounds tested in the present
study. b. Phenyl radical metabolite involvement in MPO free radical
formation (MPO•). The dashed line from the cation radical indicates
possible involvement, but based on the results of this study, we propose
that the phenyl radical metabolite is more significantly involved in MPO•
formation.
between the detection of phenyl radicals formed from
aromatic amines and their ability to form protein free radicals
(shown in Figure 1b).
Experimental Procedures
Reagents. Human MPO was purchased from Athens Re-
search & Technology (Athens, GA) and was dissolved in 0.1
M phosphate buffer, pH 7.4, and dialyzed twice for 12 h each
time. MPO concentration was determined by its extinction
coefficient (178 mM-1 cm-1 at 429 nm) (11). Horseradish
peroxidase type VI (HRP, Sigma Chemical Co., St. Louis, MO)
was desalted in a PD-10 desalting column (GE Healthcare,
Waukesha, WI) prior to use and its concentration was deter-
mined by 102 mM-1 cm-1 at 402 nm. 5,5-Dimethyl-1-pyrroline
N-oxide (DMPO) was purchased from Alexis Biochemicals (San
Diego, CA), purified twice by vacuum distillation at room
temperature, and stored under argon atmosphere at -80 °C until
use. Hydrogen peroxide (30% v/v, H2O2, Fisher Scientific Co.,
Fair Lawn, NJ) was assayed by its extinction coefficient of 43.6
M-1cm-1 at 240 nm (12). All aniline compounds, 2-methyl-2-
nitrosopropane (MNP), and diethylenetriaminepentaacetic acid
were purchased from Sigma Chemical Co. (St. Louis, MO) at
the highest grade available. MNP was dissolved in buffer (2
mg/mL), protected from light, sealed, and shaken overnight at
32 °C. Chelex-100 resin (Bio-Rad Laboratories, Hercules, CA)
was mixed with purified water overnight, and filtered before
use. Phosphate buffer used in experiments was mixed with
Chelex-100 resin overnight to remove metal contaminants, and
the resin was removed by vacuum filtration. Diethylenetriamine-
pentaacetic acid (100 mM) was included in the buffer to prevent
metal-catalyzed reactions.
Enzyme-Linked Immunosorbent Assay (ELISA) and
MPO Protein Free Radical Detection. Ninety-six-well ELISA
plates (Greiner Labortechnik, Germany) were used to carry out
reactions in order to detect the DMPO-MPO adducts. These
reactions were carried out in buffer and contained 100 mM
DMPO, 100 µM H2O2, 50 nM MPO, and varying concentrations
of anilines. The reactants were mixed in 1.5 mL centrifuge tubes,
then 100 µL of this reaction was transferred to each of three
wells containing 200 µL water. This reaction was carried out
for 2 h in an Eppendorf Thermomixer R (Germany) at 37 °C
with mixing at 500 rpm. At the end of the incubation period,
Results
Aniline-induced MPO Free Radical Formation. The aniline
derivatives used in this study were assessed for their ability to
induce anti-DMPO detection (formation of MPO•) that was
2-fold greater than control levels (defined as the EC2 value).
As shown in Table 1, we were able to derive EC2 values for
most aniline compounds used in this study. The EC2 values
ranged from 2.4 µM to 733 µM. The most potent inducers of
protein free radical formation were 2,4-dimethylaniline and
4-ethylaniline, and the least potent was 2,6-dichloroaniline. We
were not able to derive EC2 values for p-anisidine, 3,4-
dimethoxyaniline, 2,4-dimethoxyaniline, and 3-nitro- or 2-nitro-
aniline because, for these compounds, there was no increase in
anti-DMPO detection significantly greater than the control found
in the concentration range between 0.1 µM - 500 µM of aniline
derivative. The physicochemical parameters were obtained as
described in the legend and Experimental Procedures. A
selection of representative structures from the current test set
in Table 1 is shown in Figure 1a.

882 Chem. Res. Toxicol., Vol. 23, No. 5, 2010
Siraki et al.
Table 1. EC2 Values for Aniline-induced MPO Free Radical Formation and Physicochemical Parameters for Anilines
CASRN
NAME
p-anisidine
aminoglutethimide
4-ethylaniline
p-toluidine
m-toluidine
aniline
o-anisidine
4-fluoroaniline
o-toluidine
m-anisidine
4-chloroaniline
procainamide
3-chloroaniline
2-chloroaniline
4-aminobenzonitrile
3-nitroaniline
Class
EC2 (µM)
log EC2
o´1
log P2
EA-R3
IP-R3
IP-P3
EA-P3
104-94-9
125-84-8
589-16-2
106-49-0
108-44-1
62-53-3
90-04-0
371-40-4
95-53-4
536-90-3
106-47-8
51-06-9
108-42-9
95-51-2
873-74-5
99-09-2
100-01-6
88-74-4
2735-04-8
95-64-7
6315-89-5
95-68-1
87-62-7
95-69-2
133-15-3
95-76-1
554-00-7
608-31-1
mono
mono
mono
mono
mono
mono
mono
mono
mono
mono
mono
mono
mono
mono
mono
mono
mono
mono
di
di
di
di
di
di
di
di
di
-*
1.5
2.4
3.1
8.9
20.5
3.6
16.3
3.6
6.8
19.3
3.2
248
61.2
279
--
345
--
--
2.7
--
2.4
7.3
3.64
29.8
455
67.6
733
-
-0.28
-0.154
-0.15
-0.14
-0.06
0
0
0.06
0.1
0.11
0.24
0.365
0.37
0.67
0.7
0.74
0.78
1.72
-0.28
-0.2
-0.17
-0.04
0.2
0.95
1.1
1.96
1.39
1.4
5.01
4.90
4.92
4.92
4.9
10.04
9.93
9.94
9.95
9.89
8.21
8.3
-0.55
-0.67
-0.65
-0.61
-0.59
-0.64
-0.57
-0.29
-0.6
-0.62
-0.28
0.34
-0.26
-0.29
0.19
0.75
0.78
0.176
0.38
0.49
0.95
1.31
0.56
1.21
0.56
0.83
1.29
0.505
2.39
1.79
2.45
--
2.54
--
--
0.43
--
0.38
0.86
0.56
1.47
2.66
1.83
2.87
8.33
8.35
8.48
8.53
8.26
8.55
8.44
8.5
0.9
4.96
5.04
5.27
4.91
5.03
5.24
5.35
5.26
5.32
5.49
5.82
5.89
5.94
5.10
4.88
5.07
4.89
4.89
5.2
9.97
1.18
1.15
1.32
0.93
1.83
0.88
1.88
1.9
10.06
10.34
9.93
10.01
10.26
10.40
10.22
10.31
10.53
10.88
10.93
10.96
10.07
9.88
10.04
9.92
9.92
10.23
11.1
10.46
10.58
10.64
8.6
8.43
8.73
8.62
8.87
9.39
9.27
9.25
8.35
8.32
8.33
8.28
8.35
8.51
9.47
8.74
8.68
8.73
0.9
1.37
1.39
1.85
1.056
1.84
1.086
1.68
1.84
2.1
0.89
2.69
2.78
2.76
4-nitroaniline
2-nitroaniline
0.79
2,4-dimethoxyaniline
3,4-dimethylaniline
3,4-dimethoxyaniline
2,4-dimethylaniline
2,6-dimethylaniline
4-chloro-2-methylaniline
4-aminosalicylic acid
3,4-dichloroaniline
2,4-dichloroaniline
2,6-dichloroaniline
-0.65
-0.61
-0.63
-0.61
-0.6
-0.29
0.52
0.34
0.57
0.61
0.91
1.34
6
5.49
5.58
5.67
0
-0.01
0.01
di
1 From Ref 14. 2 From Ref 15 (http://www.vcclab.org). 3 Values were determined by MOPAC calculations as described in Experimental Procedures.
EA-R, electron affinity of the radical metabolite; IP-R, ionization potential of the radical metabolite; IP-P, ionization potential of the parent aniline;
EA-P, electron affinity of the parent aniline. Values are in eV. 4 Value for -C(CH3)3. 5 Value for -CONHCH3. 6 Calculated logP (AlogP).
Table 2. Pearson Correlation (correlation matrix) for the
Physicochemical Parameters used to Derive QSAR
Equations in This Study
small test set for final QSAR evaluation, we excluded from all
model training sets the two monosubstituted aniline drugs,
aminoglutethimide and procainamide.
log P
EA-R
IP-R
0.795
IP-P
EA-P
Table 3 shows the best 1- and 2-parameter QSAR equations
derived from the aniline activities used in this study. Equations
1, 2, and 3 were derived from the combined set of 21 mono
and disubstituted anilines for which EC2 values were experi-
mentally determined. Interestingly, the best overall 1-parameter
fit, by a significant margin, was obtained using σ, and not log
P or IP-P (Figure 2). Equations 4-6 and 7-9 were derived for
the subsets of 13 mono- and 8 disubstituted anilines, respec-
tively. In both cases, a 1-parameter QSAR based on σ provided
reasonable statistics with improved r2 when the mono- and
disubstituted anilines were modeled as separate subsets (eqs 4
and 7; Figures 3 and 4) compared to eq 1. Interestingly, in the
case of the monosubstituted anilines, a 1-parameter QSAR based
on IP-P (eq 7) gave slightly better overall statistics, whereas
this parameter alone gave poor statistics in modeling either the
full data set or the subset of disubstituted anilines. The inclusion
of 4-aminosalicylic acid in the data set may have led to less
robust equations since the latter was the only aniline containing
a strong acid group which could have different values for its
physicochemical parameters based on its ionization.
σ
0.47a 0.83
0.648
0.754
0.033 3.1 × 10-7 1.7 × 10-6
0.00150
-0.109
0.638
7.8 × 10-5
0.0736
0.751
log P
EA-R
IP-R
IP-P
0.16
0.48
0.107
0.643
0.996
0.918
0.962
3.7 × 10-21 4.8 × 10-9 4.0 × 10-12
0.925
0.965
2.1 × 10-9 1.5 × 10-12
0.93
6.4 × 10-10
a Each pair of values corresponds to the correlation (r) and
significance (p) between the physicochemical parameters that are
compared. Values in bold typeface show highly significant correlation.
Derivation of QSAR Equations. Before derivation of
QSARs it is important to determine the relationship between
the physicochemical parameters to be used in QSAR derivation
(intercorrelation) to select the fewest number of parameters that
could potentially account for the most variance in the data. In
Table 2 we show the pairwise Pearson correlations between the
different physicochemical parameters computed for the 21
compounds in Table 1 having measured EC2 values. These
values indicate that the calculated quantum chemical parameters
are highly collinear (>92%), whereas log P is not significantly
correlated with any other parameters (<16%). The σ parameter,
on the other hand, shows variable correlation with quantum
mechanical parameters, ranging from 65% (IP-P) to 83% (EA-
R). Hence, only 3 parameters were considered in subsequent
QSAR development: σ, IP-P (which had the least colinearity
with σ), and log P. QSAR derivation is intended to reveal
whether one or more parameters correlate significantly with
activity; hence, only compounds with experimental EC2 values
are included in model development. In addition, to provide a
In addition, we performed a best subset regression that
revealed whether more than one parameter would contribute to
improving the statistics of the QSAR expression. In the case of
the full set of 21 anilines, addition of IP-P with σ (eq 2, Table
3) provided slight improvement in overall statistics. However,
replacing σ with EA (eq 3, Table 3) produced the best two-
parameter QSAR equation; EA was not a significant parameter
when the data set was separated into mono- and disubsituted
anilines. In the case of the 13 monosubstituted anilines, however,
inclusion of IP-P with σ provided significant overall improve-
ment (eq 6), consistent with the finding that IP-P alone gave
the best 1-parameter correlation for this subset. For the 8

QSARs for Protein Free Radical Formation
Chem. Res. Toxicol., Vol. 23, No. 5, 2010 883
Table 3. QSAR Equations for Aniline-induced MPO Free Radical Formation
Eqn Compound Sets
n
r2
F-ratio
SE
All anilines
log EC2 ) 0.794((0.126) + 1.74 σ((0.249)
(1)
(2)
21
21
21
0.719
0.747
0.772
48.6
26.6
30.5
0.458
0.447
0.423
log EC2 ) -4.21((3.56) + 1.452 σ((0.319) + 0.593 IP-P((0.426)
log EC2 ) -21.351((2.94) + 0.877 EA-R((0.154) + 2.117 IP-P((0.295) (3)
Monosubstituted anilines
log EC2 ) 0.851((0.126) + 2.12 σ((0.338)
(4)
(5)
(6)
13
13
13
0.781
0.795
0.852
39.3
42.5
28.8
0.379
0.368
0.327
log EC2 ) -21.2((3.46) + 2.63 IP-P((0.403)
log EC2 ) -11.7((5.81) + 1.08 σ((0.563) + 1.48 IP-P((0.639)
Disubstituted anilines
log EC2 ) 0.578((0.256) + 1.73 σ((0.386)
(7)
(8)
(9)
8
8
8
0.769
0.770
0.844
20.0
8.36
13.6
0.515
0.563
0.464
log EC2 ) -0.404((5.41) + 1.68 σ((0.489) + 0.639 IP-P((0.639)
log EC2 ) -17.5((4.35) + 1.89 IP-P((0.478) + 1.25 LogP((0.281)
disubstituted anilines, although the most significant 1-parameter
correlation was also based on σ, the best 2-parameter equation
was based on IP-P and log P. Given the small number of cases
(8) in this subset, however, use of a 2-parameter correlation is
not recommended due to high probability of chance correlation.
However, these results do seem to support the observation of
differences in physicochemical determinants of activity between
the mono- and disubstituted anilines.
For each of the QSAR equations reported in Table 3, we
considered cases of statistical outliers as well as compounds
whose activities had large leverage on the resulting statistics
(not necessarily the same). In the case of outliers (all having
Studentized Residuals >2.4), no consistent or generalizable
pattern could be identified to justify their exclusion, which could
artificially inflate the statistics in relation to prospective pre-
dictivity of the equation. In addition, we examined cases where
individual compound activities exerted large leverage on the
statistics. The importance of this distinction is that excluding
outliers almost always improves a statistical correlation (r2),
whereas if a compound is not an outlier but exerts large leverage
on the statistics, its exclusion generally degrades the overall
statistics but provides a more realistic estimate of the prospective
predictivity of the equation. For example, 2,6-dichloroaniline
was identified as exerting large leverage on both eqs 1 and 2 in
Table 3 (e.g., its exclusion from eq 1 decreased the r2 value
from 0.720 to 0.678; results not shown). However, this same
compound did not exert large leverage on eqs 7-9. Given the
relatively small size of the present data set and the relatively
Figure 3. Experimental vs Calculated log EC2 values for monosubsti-
tuted anilines. Each data point was derived by applying eq 4, Table 3.
Circles represent anilines that were used in derivation of the QSAR
equation. The EC2 values predicted by this equation for aminoglute-
thimide and procainamide are plotted as squares.
Figure 2. Experimental vs Calculated log EC2 values for anilines. Each
data point was derived by applying eq 1 (Table 3), which contained
both mono- and disubstituted anilines. Circles represent data points
that were used in the regression.

884 Chem. Res. Toxicol., Vol. 23, No. 5, 2010
Siraki et al.
Discussion
The mechanism of drug-induced agranulocytosis is unknown,
but is believed to have many components, with reactive
metabolites being a major one. We previously determined that
aminoglutethimide and procainamide, both of which contain
aniline substructures, produce protein free radicals upon oxida-
tion by myeloperoxidase. In addition, it was found that both
drugs generate a phenyl radical metabolite (1, 2). The relation-
ship between the phenyl radical metabolite and protein radical
formation is unknown. Moreover, the capacity for aniline
compounds to induce MPO• formation was not reported previ-
ously. We thus undertook this study to identify the physico-
chemical parameters that may be involved in protein free radical
formation and to determine whether phenyl radical metabolites
could be implicated.
We have derived QSAR equations that demonstrate a
relationship between two electronic parameters (σ, IP-P) of
aniline compounds and the log of their EC2 values for protein
free radical formation. This suggests that the ease of oxidation
of aniline derivatives was a significant determinant of activity.
Aniline lipophilicity, as approximated by log P, was not found
to be a significant determinant of activity, suggesting that
hydrophobic binding differences are less important in accounting
for protein free radical formation among the studied compounds.
It has been shown previously that σ was inversely related to
the rate constants for HRP Compound I reduction (20). We
derived an equation from these published data, but instead of
using the ratio of substituted aniline to aniline, we used the log
of the rate constant (kX) and correlated this with σ values:
Figure 4. Experimental vs Calculated log EC2 values for disubstituted
anilines. Circles represent the data points derived using eq 7 from Table
3.
modest effect of excluding compounds with high leverage, no
compounds were excluded from the present analysis and
equations in Table 3.
As a limited prediction challenge, we applied the QSAR
equations in Table 3 for the full set of anilines (eqs 1-3) and
for the monosubstituted anilines (eqs 4-6) to predict the log
EC2 values for the previously studied monosubstituted aniline
drugs, aminoglutethimide and procainamide (1, 2). As shown
in Table 4, absolute values of the log EC2 predictions in each
case differ significantly from the experimental values. It should
be noted, however, that the experimental log EC2 value for
aminoglutethimide (0.176) was a factor of 2 lower than the
lowest log EC2 value in the entire QSAR training set (0.38).
Hence, this compound represented a true extrapolation of the
QSAR equations outside of the training set range of activities.
Given this consideration, the QSAR predictions, which all fall
in the lowest third of the activity range, are reasonable. In
addition, the relative order of predicted activities is correct for
each QSAR equation, i.e., log EC2 (procainamide) > logEC2
(aminoglutethimide), and the ratio of experimental values (2.8)
is reproduced quite accurately (2.3-3.0) in all but eq 5 (1.5).
This last result further argues against use of IP-P alone in
modeling the disubstituted anilines.
Detection of HRP-catalyzed Phenyl Radical Metabolite
Formation from Anilines. HRP/H2O2-catalyzed aniline free
radical metabolites were spin-trapped using MNP, and the
coupling constants obtained from their spectra were compared
with the ESR coupling constants for tertiary butyl aryl nitroxide
radicals (obtained from reference tables in the Landolt-Bornstein
book series) (17-19). As shown in Table 5, we were able to
detect free radical metabolites of most anilines; however, the
complete structure (i.e., what atom of the free radical reacted
with MNP) could not be determined for many of the anilines
(indicated by “unresolved H-splittings”). In addition, some
anilines did not produce any detectable adducts. The compounds
that produced well-defined ESR spectra were aniline, p-toluidine,
4-chloroaniline, 4-fluoroaniline, 4-ethylaniline, 2-chloroaniline,
and 4-aminobenzonitrile. These compounds all had substitutions
on the benzene ring para to the amine, except for 2-chloroa-
niline. The ESR spectra from 4-, 3-, and 2-chloroaniline
including their simulated spectra are shown in Figure 5.
log kX ) -7.147σ + 6.739, r2 ) 0.94, s ) 0.563, F ) 92.4
This equation indicates that anilines that reduce HRP
Compound I the fastest (i.e., the best HRP substrates) have more
negative σ values. In the present study, the derived QSAR
equations showed a similar relationship, with lower EC2 values
(more potent MPO• inducers) corresponding to anilines with
smaller or more negative σ values. This implies that better
substrates, with higher rates of reactivity, are more effective
MPO free radical generators.
In our experiments, however, we found that there were two
groups of substrates that did not induce MPO free radical
formation. One group, p-anisidine and 3,4-, and 2,4-dimethoxy-
aniline, would appear on the basis of relatively low σ values to
be highly efficient donor substrates for peroxidases; p-anisidine
has a σ ) -0.28, 3,4-dimethoxyaniline σ ) -0.17, and 2,4-
dimethoxyaniline σ ) -0.28. Interestingly, these compounds
fall at the extreme of the σ range for the tested compounds,
implying that there is an optimal range of σ values beyond which
effectiveness drops off. Interestingly, p-toluidine (σ ) -0.14)
and aminoglutethimide (σ ) -0.15) both induced MPO free
radical formation with great efficacy. On the other hand,
relatively inefficient peroxidase substrates also appeared to be
capable of generating MPO•, e.g., dichloroanilines and nitroa-
nilines. However, they did so with much less efficacy as
indicated by their relatively high EC2 values. Further evidence
that an optimal range of σ values, or electronic factors, is
necessary for MPO• generation is provided by the inability to
measure EC2 values for 2- and 3-nitroaniline (σ ) 1.72, 0.74,
respectively), which both fall at the other extreme of large σ
values (see Table 1). 4-Nitroaniline and 3,4-dichloroaniline were
the third and second least effective in generating MPO•,
respectively, and neither produced a detectable phenyl radical.
Presumably, the rate of production of the precursor (nitrogen-

QSARs for Protein Free Radical Formation
Chem. Res. Toxicol., Vol. 23, No. 5, 2010 885
Table 4. Predicted log EC2 Values for Aminoglutethimide and Procainamide
Method1
Experimental
n
logEC2-A aminoglutethimide
logEC2-P procainamide
logEC2-P/logEC2-A
0.176
0.526
0.488
0.517
0.525
0.629
0.418
0.505
1.43
1.32
1.18
1.62
0.971
1.17
2.8
2.7
2.7
2.3
3.0
1.5
2.8
QSAR eq 1 (σ)
21
21
21
8
8
8
QSAR eq 2 (σ, IP-P)
QSAR eq 3 (EA-R, IP-P)
QSAR eq 4 (σ)
QSAR eq 5 (IP-P)
QSAR eq 6 (σ, IP-P)
1 QSAR Equations are listed in Table 3 based on the properties listed here, where n ) total number of compounds used to derive the equation.
Table 5. Spin Trapping of Aniline-derived Free Radicals Formed by HRP/H2O2 Using MNP
Coupling constants (aryl nitroxides) from references (G)*
aN ) 11.86, aH ) 2.16, aH ) 0.92, aH ) 2.21 (22)
Coupling constants using HRP/H2O2 (G)
aniline
aN ) 15, aH ) 1.6, aH ) 0.92
o
m
p
o,p
m
p-toluidine
m-toluidine
o-toluidine
aN ) 13.6, aH ) aH (CH3) ) 1.86, aH ) 0.93, (23)
aN ) 15, aH ) 0.99, aH
) 1.7
o
m
m
o, p-CH3
aN ) 13.45, aH ) 1.8, aH ) aH
) 0.6 (24)
p
aN ) 15.5, unresolved H-splittings
o,p
m
(CH3)
aN ) 13.5, aH ) 0.8, aH ) 0.6, 0.7, aH ) 0.39,
aN ) 15.6, unresolved H-splittings
o
aH
) 0.24, aH
)m0.23# (25-29)
(CH3)
(tbu)
p-anisidine
m-anisidine
o-anisidine
aN ) 13.9, aH ) 1.86, aH ) 0.93 (23, 28, 30)
Not detected
o
m
aN ) 12.8, aH ) 1.8, aH ) 1.8, aH ) 0.8 (24)
aN ) 15.6, unresolved H-splittings
aN ) 15.6, unresolved H-splittings
o
p
m
aN ) 14.5, aH ) 0.99, aH ) 0.61, aH ) 0.41,
o
(3)
(4)
aH ) 0.87, aH
) 0.063, aH
) 0.26# (28)
(5)
(OCH3)
(tbu)
3,4-dimethylaniline
2,4-dimethylaniline
aN ) 15.6, unresolved H-splittings
aN ) 15.6, unresolved H-splittings
aN ) 13.5, aH
) 0.24, aH ) 0.685, aH
) 0.41,
(2, CH3)
3
(4,CH3)
aH ) 0.685, aH ) 0.8# (26, 27)
5
6
2,6-dimethylaniline
aN ) 13.4, aH
)
) 0.17, aH
) 0.68,
aN ) 15.6, unresolved H-splittings
(6,CH3)
(3,5)
aH ) 0.15, (a2H, CH3) ) 0.34# (25-28, 31)
4
(tbu)
3,4-dimethoxyaniline
2,4-dimethoxyaniline
4-chloroaniline
3-chloroaniline
2-chloroaniline
aN ) 14.5 (32)
Not detected
Not detected
aN ) 13.0, aH ) 2.01, aH ) 0.98 (23)
aN ) 14.6, aH ) 1.8, aH ) 0.94
o
m
o
m
aN ) 12.45, aH ) 1.9, aH ) 1.9, aH ) 0.8 (24)
aN ) 15.6, unresolved H-splittings
o
p
m
aN ) 13.9, aH ) 0.74, aH ) 0.47,0.82, aH ) 0.29,
aN ) 15.3, aH ) 0.83, aH ) 0.50,0.87,
o
m
p
o
m
aH
) 0.25# (28, 33)
aH ) 0.37, aH ) 0.23
(tbu)
p
tbu
4-fluoroaniline
aN ) 13.7, aH ) 1.72, aH ) 0.86, aF ) 3.5 (23)
aN ) 15.1, aH ) 1.68, aH ) 0.89, aF ) 3.1
o
m
o
m
2,4-dichloroaniline
3,4-dichloroaniline
2,6-dichloroaniline
4-nitroaniline
aN ) 13.8# (28, 29)
aN ) 15.2, unresolved H-splittings
Not detected
aN ) 13.1, aH ) 0.69, aH ) 0.1, aH
) 0.35 (28)
aN ) 14.8, unresolved H-splittings
Not detected
m
p
(tbu)
3-nitroaniline
Not detected
2-nitroaniline
4-ethylaniline
4-aminosalicylic acid
4-chloro-2-methylaniline
4-aminobenzonitrile
Not detected
aN ) 13, aH ) 1.9, aH ) 0.9# (25, 34-36)
aN ) 15.0, aH ) 1.87, aH ) 1.1
o
m
o
m
aN ) 15.2, unresolved H-splittings
aN ) 15.4, unresolved H-splittings
aN ) 13.3, aH ) 2.1, aH ) 0.97,
aN ) 10.06, aH ) 2.46, aH ) 0.94, aN
) 0.41 (22)
o
m
(CN)
o
m
aN
) 0.27
(CN)
* The aN is typically lower in organic solvents. # Value for 2-chloro-4-bromoaniline.
centered radical) and formation of the phenyl radical were low
and below the limits of detection. 2,6-Dichloroaniline, which
was the least effective substrate (EC2 ) 733 µM), had a
detectable (albeit weak) phenyl radical spectrum which was
likely due to stabilization of the radical metabolite by the
o-chloro groups.
Identification of the presumed free radical metabolite that is
responsible for generating the MPO• is technically challenging
because it is difficult to differentiate the effects of the nitrogen-
centered cation radical from those of the carbon-centered phenyl
radical. In order to generate the phenyl radical metabolites, the
nitrogen-centered radical has to be sufficiently thermodynami-
cally unstable in order to break down into the phenyl radical.
Formation of the phenyl radical requires cleavage of the C-N
bond, which requires activation energy. This appears to be a
two-step process; the first step, where an initial radical
metabolite is formed (nitrogen-centered radical), is necessary
but not sufficient for MPO• formation; the second step is the
spontaneous formation of the phenyl radical (cleavage of the
C-N bond) which appears to react with MPO, forming MPO•.
Since we previously determined that aminoglutethimide and
procainamide formed phenyl radicals, we wished to investigate
whether the anilines used in this study also formed phenyl
radicals. In addition, we wished to determine the relationship
between the formation of phenyl radical metabolites and MPO•.
With two exceptions (4-nitroaniline and 3,4-dichloroaniline) out
of 28 aniline derivatives, it appeared that there was a qualitative
relationship between phenyl radical metabolite formation and
MPO•. More importantly, the aniline compounds for which we
Our initial studies on drug-induced MPO• formation involved
the peroxidase metabolism of aminoglutethimide and procaina-
mide, which are both aniline-based drugs. Having derived QSAR
equations to describe the physicochemical relationship between
aniline compounds and MPO formation, we wished to determine
whether these equations would reasonably predict the EC2 values
of the drugs. The drugs were not used to derive the QSAR
equations and therefore represented a small test set. Interestingly,
equations using σ did not accurately predict the absolute values
of EC2, but did correctly and accurately predict their relative
values. One reason for this disparity could be that the σ values
for the drugs are not known (we used σ values from structurally
similar para substituents). Another reason may be that the group
in the para position for procainamide is electron withdrawing
(-CONH2CH2R, σ ) 0.36), and would be predicted to be a
weak protein radical inducer and a relatively poor peroxidase
substrate.

886 Chem. Res. Toxicol., Vol. 23, No. 5, 2010
Siraki et al.
were unable to detect phenyl radicals did not form MPO•. This
suggests that the phenyl radical is the metabolite responsible
for MPO• formation.
The mechanisms of drug-induced agranulocytosis are still
unresolved and are considered to be multifactoral, i.e., a
particular alignment of multiple parameters must be present for
the adverse drug reaction to occur. In the absence of suitable
models, investigations of such parameters have been hindered.
It has been proposed for some time that leukocyte (MPO)-
generated reactive metabolites may be involved in the etiology
of drug-induced agranulocytosis (21); a role for MPO-generated
free radical metabolites has also been proposed (7). We have
focused on the physicochemical side of these reactions by
evaluating the tendency of aromatic amines that generate phenyl
radical metabolites to induce MPO•. This study established that
the electronic withdrawing nature of the substitution on the
aniline ring can determine the outcome of protein free radical
formation on MPO. In the future, this research needs to establish
the significance of MPO• in the etiology of agranulocytosis and
identify potential in ViVo pathways of inducing and altering this
reaction.
Acknowledgment. The authors thank Ms. Jean Corbett
(NIEHS) for purification of DMPO, Ms. Mary Jane Mason for
her careful review of the manuscript, and the internal reviewer
for her valuable critique. This research was supported by the
Intramural Research Program of the NIH, and NIH/NIEHS, as
well as the University of Alberta Startup Fund.
References
(1) Siraki, A. G., Bonini, M. G., Jiang, J., Ehrenshaft, M., and Mason,
R. P. (2007) Aminoglutethimide-induced protein free radical formation
on myeloperoxidase: a potential mechanism of agranulocytosis. Chem.
Res. Toxicol. 20, 1038–1045.
(2) Siraki, A. G., Deterding, L. J., Bonini, M. G., Jiang, J., Ehrenshaft,
M., Tomer, K. B., and Mason, R. P. (2008) Procainamide, but not
N-acetylprocainamide, induces protein free radical formation on
myeloperoxidase: a potential mechanism of agranulocytosis. Chem.
Res. Toxicol. 21, 1143–1153.
(3) Uetrecht, J. P. (1989) Idiosyncratic drug reactions: Possible role of
reactive metabolites generated by leukocytes. Pharm. Res. 6, 265–
273.
(4) Christie, G., Breckenridge, A. M., and Park, B. K. (1989) Drug-protein
conjugates-XVIII. Detection of antibodies towards the antimalarial
amodiaquine and its quinone imine metabolite in man and the rat.
Biochem. Pharmacol. 38, 1451–1458.
(5) Lahita, R., Kluger, J., Drayer, D. E., Koffler, D., and Reidenberg,
M. M. (1979) Antibodies to nuclear antigens in patients treated with
procainamide or acetylprocainamide. N. Engl. J. Med. 301, 1382–
1385.
(6) Reidenberg, M. M. (1983) Aromatic amines and the pathogenesis of
Figure 5. a. ESR spectrum of 2-chloroaniline when oxidized by HRP/
H2O2. 2-chloroaniline, MNP, and H2O2 were mixed in Chelex-100-
treated phosphate buffer, and the reaction was initiated by adding HRP.
The reaction was transferred to a flat cell and placed in the ESR cavity
for recording. The spectrum for 2-chloroaniline was not well resolved,
which required specific parameters to be used (see Experimental
Procedures). The simulation of this spectrum (correlation: r ) 0.99) is
shown below the experimental spectrum (top), and the center peak is
zoomed to show the overlay. The hyperfine splitting constants are shown
in Table 5. Figure 5b ESR spectrum of 3-chloroaniline oxidized by
HRP/H2O2. The reaction was carried out as described in Figure 5a. In
this case, the signal:noise was not as high as it was for 2-chloroaniline,
which required 10 mW power to be used in order to obtain as resolved
a spectrum. The other parameters were as described in the Experimental
Procedures. This spectrum could not be simulated, presumably because
of mixed product formation. Figure 5c. ESR spectrum obtained upon
4-chloroaniline oxidation by HRP/H2O2. The reaction was carried out
as described in Figure 5a; however, spectrum resolution was attained
using the following instrument settings: 20 mW power, modulation
amplitude of 0.4 G, and conversion and time constant at 655.36 ms.
The simulation of the experimental spectrum (correlation: r ) 0.97) is
shown below. The hyperfine splitting constants are shown in Table 5.
lupus erythematosus. Am. J. Med. 75, 1037–1042.
(7) Fischer, V., Haar, J. A., Greiner, L., Lloyd, R. V., and Mason, R. P.
(1991) Possible role of free radical formation in clozapine (clozaril)-
induced agranulocytosis. Mol. Pharmacol. 40, 846–853.
(8) Dunford, H. B. (1999) Heme peroxidases, John Wiley, New York.
(9) Uetrecht, J. P. (1992) The role of leukocyte-generated reactive
metabolites in the pathogenesis of idiosyncratic drug reactions. Drug
Metab. ReV. 24, 299–366.
(10) Besser, M., Vera, J., Clark, J., Chitnavis, D., Beatty, C., and Vassiliou,
G. (2009) Preservation of basophils in dapsone-induced agranulocytosis
suggests a possible pathogenetic role for leucocyte peroxidases. Int.
J. Lab. Hematol. 31, 245–247.
(11) Hsuanyu, Y., and Dunford, H. B. (1999) Oxidation of Clozapine and
Ascorbate by Myeloperoxidase. Arch. Biochem. Biophys. 368, 413–
420.
(12) Beers, R. F., Jr., and and Sizer, I. W. (1952) A spectrophotometric
method for measuring the breakdown of hydrogen peroxide by catalase.
J. Biol. Chem. 195, 133–140.
(13) Lagercrantz, C. (1971) Spin trapping of some short-lived radicals by
the nitroxide method. J. Phys. Chem. 75, 3466–3475.
(14) Perrin, D. D., Dempsey, B. Serjeant, E. P. (1981) pKa prediction for
organic acids and bases, Chapman and Hall, London; New York.

QSARs for Protein Free Radical Formation
Chem. Res. Toxicol., Vol. 23, No. 5, 2010 887
(15) Tetko, I. V., Gasteiger, J., Todeschini, R., Mauri, A., Livingstone,
D., Ertl, P., Palyulin, V. A., Radchenko, E. V., Zefirov, N. S.,
Makarenko, A. S., Tanchuk, V. Y., and Prokopenko, V. V. (2005)
Virtual computational chemistry laboratory--design and description.
J. Comp. Aid. Mol. Des. 19, 453–463.
(16) Burcat, A., Khachatryan, L., and Dellinger, B. L. (2003) Thermody-
namics of Chlorinated Phenols, Polychlorinated Dibenzo- p -Dioxins,
Polychlorinated Dibenzofurans, Derived Radicals, and Intermediate
Species. J. Phys. Chem. Ref. Data 32, 443–517.
(17) Forrester, A. (1989) Aryl Radicals. In Nitroxide Radicals (Series:
Landolt-Bo¨rnstein - Group II Molecules and Radicals Numerical Data
and Functional Relationships in Science and Technology) pp 168-
175, Springer-Verlag, Berlin.
(18) Forrester, A. (1979)Tert. alkyl-aryl nitroxides. In Organic N-Centered
Radicals and Nitroxide Radicals (Series: Landolt-Bo¨rnstein - Group
II Molecules and Radicals Part 2) pp 651-693, Springer-Verlag,
Berlin.
(19) Forrester, A. (1989) Aryl radicals, Nitroxides formed by trapping with
nitrosoalkanes. In Nitroxide Radicals. (Series: Landolt-Bo¨rnstein -
Group II Molecules and Radicals Part 2) pp 168-175, Springer-
Verlag, Berlin.
(20) Job, D., and Dunford, H. B. (1976) Substituent effect on the oxidation
of phenols and aromatic amines by horseradish peroxidase compound
I. Eur. J. Biochem. 66, 607–614.
(21) Uetrecht, J. (1989) Drug metabolism by leukocytes and its role in
drug-induced lupus and other idiosyncratic drug reactions. Crit. ReV.
Toxicol. 20, 213–235.
(26) Calder, A., Forrester, A. R., Emsley, J. W., Luckhurst, G. R., and
Storey, R. A. (1970) Nitroxide radicals 0.7. Nuclear and electron
resonance spectra of some aromatic tert-butyl nitroxides. Mol. Phys.
18, 481–489.
(27) Forrester, A. R., and Hepburn, S. P. (1970) Nitroxide radicals. VIII.
Stability of ortho-alkyl-substituted phenyl tert-butyl nitroxides.
J. Chem. Soc. C 9, 1277–1280.
(28) Pedersen, J. A., and Torssell, K. (1971) Electron spin resonance and
nuclear magnetic resonance spectra of sterically hindered aromatic
nitroxide radicals - synthesis of stable nitroxide radicals. Acta Chem.
Scand. 25, 3151–3162.
(29) Torssell, K. (1970) Investigation of radical intermediates in organic
reactions by use of nitroso compounds as scavengers the nitroxide
method. Tetrahedron 26, 2759–2773.
(30) Razuvaev, G. A., Abakumov, G. A., Sanaeva, E. P., and Abakumova,
L. G. (1973) 1. Synthesis and oxidative decomposition. IzVestiya
Akademii Nauk SSSR, Seriya Khimicheskaya 1, 68–71.
(31) Lemaire, H., and Rassat, A. (1964) Hyperfine structure of nitrogen in
nitroxide radicals. J. Chim. Phys. Physico-Chimie Biologique 61, 1580–
1586.
(32) Frangopol, P. T., Frangopol, M., and Baikan, R. (1975) New aryl tert-
alkyl nitroxyls. IzVestiya Akademii Nauk SSSR, Seriya Khimicheskaya
11, 2506–2511.
(33) Calder, A., Forrester, A. R., and Hepburn, S. P. (1973) Nitroxide
radicals. XII. Decomposition of o-, m-, and p-halophenyl tert-butyl
nitroxides. J. Chem. Soc., Perkin Trans. 1: Organic and Bio-Organic
Chemistry 5 456–465.
(34) Calder, A., Forrester, A. R., and Thomson, R. H. (1969) Nitroxide
radicals. IV. Oxidation of cyclic N-hydroxyimides. J. Chem. Soc. C
3, 512–516.
(22) Nelsen, S. F., Landis, R. T., Kiehle, L. H., and Leung, T. H. (1972)
N-tert-Butylanilino Radicals. J. Am. Chem. Soc. 94, 1610–1614.
(23) Lemaire, H., Marechal, Y., Ramasseul, R., and Rassat, A. (1965)
Nitroxides. IX. Electron paramagnetic resonance of parasubstituted
tert-butylphenylnitroxide. Bull. Soc. Chim. Fr. 2, 372–378.
(24) Barbarella, G., and Rassat, A. (1969) Nitroxides. XXXIV. Substituted
tert-butyl phenyl nitroxide derivatives electron paramagnetic resonance.
Ultrafine deviations due to nitrogen. Bull. Soc. Chim. Fr. 7, 2378–
2385.
(35) Forrester, A. R., Hepburn, S. P., and McConnachie, G. (1974) Nitroxide
radicals. XVI. Unpaired electron distribution in para substituted aryl
tert-butyl nitroxides and 2-naphthyl phenyl nitroxides. J. Chem. Soc.,
Perkin Trans. 1: Org. Bio-Org. Chem. 19, 2213–2219.
(36) Torssell, K., Goldman, J., and Petersen, T. E. (1973) Spin delocal-
ization of Group IV B elements in organic compounds. Justus Liebigs
Annalen der Chemie 2, 231–240.
(25) Calder, A., and Forrester, A. R. (1967) Stability of aryl-tert-
butylnitroxides. Chem. Commun. (Camb) 14, 682–684.
TX900432D