Structure-Activity Relationships for CYP4B1 Bioactivation of 4-Ipomeanol Congeners: Direct Correlation between Cytotoxicity and Trapped Reactive Intermediates

Article abstract of DOI:10.1021/acs.chemrestox.9b00330

Cytochrome P450 4B1 (CYP4B1) has been explored as a candidate enzyme in suicide gene systems for its ability to bioactivate the natural product 4-ipomeanol (IPO) to a reactive species that causes cytotoxicity. However, metabolic limitations of IPO necessitate discovery of new "pro-toxicant" substrates for CYP4B1. In the present study, we examined a series of synthetically facile N-alkyl-3-furancarboxamides for cytotoxicity in HepG2 cells expressing CYP4B1. This compound series maintains the furan warhead of IPO while replacing its alcohol group with alkyl chains of varying length (C1-C8). Compounds with C3-C6 carbon chain lengths showed similar potency to IPO (LD₅₀ ≈ 5 μM). Short chain analogs (<3 carbons) and long chain analogs (>6 carbons) exhibited reduced toxicity, resulting in a parabolic relationship between alkyl chain length and cytotoxicity. A similar parabolic relationship was observed between alkyl chain length and reactive intermediate formation upon trapping of the putative enedial as a stable pyrrole adduct in incubations with purified recombinant rabbit CYP4B1 and common physiological nucleophiles. These parabolic relationships reflect the lower affinity of shorter chain compounds for CYP4B1 and increased ω-hydroxylation of the longer chain compounds by the enzyme. Furthermore, modest time-dependent inhibition of CYP4B1 by N-pentyl-3-furancarboxamide was completely abolished when trapping agents were added, demonstrating escape of reactive intermediates from the enzyme after bioactivation. An insulated CYP4B1 active site may explain the rarely observed direct correlation between adduct formation and cell toxicity reported here.

Full text of DOI:10.1021/acs.chemrestox.9b00330

Article
pubs.acs.org/crt
Cite This: Chem. Res. Toxicol. XXXX, XXX, XXX−XXX
Structure−Activity Relationships for CYP4B1 Bioactivation of
4‑Ipomeanol Congeners: Direct Correlation between Cytotoxicity
and Trapped Reactive Intermediates
John P. Kowalski,† Matthew G. McDonald,† Dale Whittington,† Miklos Guttman,† Michele Scian,†
Marco Girhard,‡ Helmut Hanenberg,§ Constanze Wiek,⊥ and Allan E. Rettie*,†
†Department of Medicinal Chemistry, School of Pharmacy, University of Washington, Seattle, Washington 98105, United States
‡Institute of Biochemistry, Heinrich-Heine University, 40225 Du
̈
sseldorf, Germany
§Department of Pediatrics III, University Children’s Hospital Essen, University of Duisburg-Essen, 45122 Essen, Germany
⊥Department of Otorhinolaryngology and Head/Neck Surgery, Heinrich-Heine University, 40225 Du
sseldorf, Germany
̈
S
* Supporting Information
ABSTRACT: Cytochrome P450 4B1 (CYP4B1) has been
explored as a candidate enzyme in suicide gene systems for its
ability to bioactivate the natural product 4-ipomeanol (IPO)
to a reactive species that causes cytotoxicity. However,
metabolic limitations of IPO necessitate discovery of new
“pro-toxicant” substrates for CYP4B1. In the present study, we
examined a series of synthetically facile N-alkyl-3-furancarbox-
amides for cytotoxicity in HepG2 cells expressing CYP4B1.
This compound series maintains the furan warhead of IPO while replacing its alcohol group with alkyl chains of varying length
(C1−C8). Compounds with C3−C6 carbon chain lengths showed similar potency to IPO (LD50 ≈ 5 μM). Short chain analogs
(<3 carbons) and long chain analogs (>6 carbons) exhibited reduced toxicity, resulting in a parabolic relationship between alkyl
chain length and cytotoxicity. A similar parabolic relationship was observed between alkyl chain length and reactive intermediate
formation upon trapping of the putative enedial as a stable pyrrole adduct in incubations with purified recombinant rabbit
CYP4B1 and common physiological nucleophiles. These parabolic relationships reflect the lower affinity of shorter chain
compounds for CYP4B1 and increased ω-hydroxylation of the longer chain compounds by the enzyme. Furthermore, modest
time-dependent inhibition of CYP4B1 by N-pentyl-3-furancarboxamide was completely abolished when trapping agents were
added, demonstrating escape of reactive intermediates from the enzyme after bioactivation. An insulated CYP4B1 active site
may explain the rarely observed direct correlation between adduct formation and cell toxicity reported here.
relegated to where this bioactivating enzyme is newly
expressed. Specifically, this may have utility in an anticancer
setting as a molecular switch for T-cell inactivation6 since
adverse immunological responses pose serious safety issues
that can limit T-cell therapy.7−9 Use of mammalian CYP4B1 as
a suicide gene could potentially reduce the immunogenicity
that arises from other systems like the herpes simplex virus
thymidine kinase and ganciclovir tandem.6
However, IPO and PK possess structural features that could
negatively impact their metabolism, thereby limiting their
duration of action or effectiveness. Rodent studies suggest
rapid clearance of IPO by glucuronidation,10,11 and several
human uridine 5′-diphosphoglucuronosyl transferases (UGTs)
are known to conjugate IPO.12 While PK lacks this alcohol
moiety, replacement by a methyl group confers extensive
CYP4B1-mediated oxidation to nonreactive products.13 This
latter observation suggests that CYP4B1 can bind alkyl furans
INTRODUCTION
■
The cytochrome P450 4 (CYP4) family consists of 13 enzymes
in humans that are typically involved in fatty acid and
eicosanoid oxidation.1 CYP4B1 is considered an orphan
isozyme in the family, with unknown endogenous substrate
selectivity.2 In humans, CYP4B1 is catalytically deficient due to
low stability caused, in part, by a Pro-to-Ser change in the
meander region of the enzyme.3,4 However, other mammalian
orthologs, present in cattle and rabbit, exhibit good stability
and bioactivate several protoxins that induce toxicological
effects in expression-specific tissues.2
4-Ipomeanol (IPO), produced by the common sweet potato
in response to stress, is a 3′-substituted furan-containing
natural product, which is bioactivated by pulmonary CYP4B1
to an electrophilic species that binds tissue nucleophiles to
elicit its organ toxicity.5 Consequently, IPO and related furans,
like perilla ketone (PK), have been suggested as cytotoxic
prodrugs in a novel CYP4B1 suicide gene system in humans.
Introduction of exogenous and functional (e.g., rabbit)
CYP4B1 could theoretically allow for controlled cytotoxicity
Received: August 9, 2019
© XXXX American Chemical Society
A
DOI: 10.1021/acs.chemrestox.9b00330
Chem. Res. Toxicol. XXXX, XXX, XXX−XXX

Chemical Research in Toxicology
Article
amount of pyridine, was added dropwise. Reactions were stirred
overnight at room temperature, quenched with water, extracted with
ethyl acetate, and dried over magnesium sulfate. Compounds were
isolated via flash chromatography using a hexane/ethyl acetate
in distinct catalytic orientations that influence the bioactivation
versus inactivation processes.
The goal of this study is to clarify the relationship between
3′-substituted furan structure and CYP4B1-mediated metabo-
lism to either reactive intermediates (derived from oxidation of
the furan group) or nonreactive metabolites (derived from
oxidation of the alkyl group). In preliminary studies, we probed
the CYP4B1 permissible chemical space for furan-containing
substrates using a high-throughput cytotoxicity screen
performed with CYP4B1-transfected HepG2 cells and a
mini-library of >400 IPO-like compounds (data not shown).
Although no compounds proved as potent as IPO, N-alkyl-3-
furancarboxamides emerged as a common motif in the hits
from this library. Therefore, we adopted a simple synthetic
approach to generate a homologous series of eight N-alkyl-3-
furancarboxamides and evaluated their CYP4B1-mediated
cytotoxicity. Structure−metabolism relationships are discussed,
and the lack of mechanism-based inhibition of CYP4B1 by 3′-
substituted furans was identified as a key determinant of the
efficiency of the cytotoxic process.
1
gradient. Purity assessed by LC−MS/UV and H NMR were >95%
for all compounds. 1H NMR experiments were performed at 25 °C on
a 499.73 MHz Agilent DD2 spectrometer. For characterization and
spectral assignment purposes, the samples were ∼30 mM solutions in
CDCl3 (99.96% D, Cambridge Isotopes). Chemical shifts are reported
below relative to the solvent peak in CDCl3 at 7.26 ppm, and the
1
coupling constants (J) are noted in hertz (Hz). All annotated H
NMR spectra have been provided in the Supporting Information.
Accurate mass analyses of the compounds were determined via LC−
MS on a Thermo LTQ-Orbitrap mass spectrometer.
N-Ethyl-3-furancarboxamide, C2. Exact mass [M + H] calculated
140.0712, observed 140.0698, δ ppm 9.7. 1H NMR (CDCl3), 7.91 (s,
1H), 7.42 (s, 1H), 6.59 (s, 1H), 5.75 (bs, 1H), 3.46−3.42 (m, 2H),
1.22 (t, J = 7.5 Hz, 3H). Yield 58%.
N-Propyl-3-furancarboxamide, C3. Exact mass [M + H]
calculated 154.0868, observed 154.0871, δ ppm 1.9. 1H NMR
(CDCl3), 7.94 (s, 1H), 7.39 (s, 1H), 6.67 (s, 1H), 6.51 (bs, 1H), 3.31
(q, J = 7 Hz, 2H), 1.56 (sx, J = 7.5 Hz, 2H), 0.92 (t, J = 7 Hz, 3H).
Yield 56%.
N-Butyl-3-furancarboxamide, C4. Exact mass [M + H] calculated
168.1025, observed 168.1022, δ ppm 1.5. 1H NMR (CDCl3), 7.95 (s,
1H), 7.36 (s, 1H), 6.88 (bs, 1H), 6.70 (s, 1H), 3.31 (q, J = 7 Hz, 2H),
1.50 (p, J = 7.5 Hz, 2H), 1.31 (sx, J = 7.5 Hz, 2H), 0.86 (t, J = 7.5 Hz,
3H). Yield 65%.
N-Pentyl-3-furancarboxamide, C5. Exact mass [M + H]
calculated 182.1181, observed 182.1170, δ ppm 6.1. 1H NMR
(CDCl3), 7.94 (s, 1H), 7.46 (s, 1H), 6.62 (s, 1H), 5.76 (bs, 1H), 3.43
(q, J = 7 Hz, 2H), 1.62 (p, J = 7.5 Hz, 2H), 1.39 (m, 4H), 0.94 (t, J =
7 Hz, 3H). Yield 70%.
N-Hexyl-3-furancarboxamide, C6. Exact mass [M + H] calculated
196.1338, observed 196.1333, δ ppm 2.3. 1H NMR (CDCl3), 7.94 (s,
1H), 7.39 (s, 1H), 6.66 (s, 1H), 6.41 (bs, 1H), 3.35 (t, J = 7 Hz, 2H),
1.54 (p, J = 7.5 Hz, 2H), 1.29 (m, 6H), 0.86 (t, J = 7 Hz, 3H). Yield
64%.
N-Heptyl-3-furancarboxamide, C7. Exact mass [M + H]
calculated 210.1494, observed 210.1484, δ ppm 4.8. 1H NMR
(CDCl3), 7.94 (s, 1H), 7.46 (s, 1H), 6.62 (s, 1H), 5.76 (bs, 1H), 3.43
(q, J = 7 Hz, 2H), 1.61 (p, J = 7 Hz, 2H), 1.34 (m, 8H), 0.91 (t, J = 7
Hz, 3H). Yield 69%.
EXPERIMENTAL PROCEDURES
■
General Reagents. Recombinant rabbit CYP4B1 was expressed
in Escherichia coli and purified using an established protocol.14
Cytochrome P450 reductase (CPR) and cytochrome b5 coenzymes
used in CYP4B1 incubations were expressed and purified according to
previously published methods.15 Chloroform-d (CDCl3) and
deuterium oxide (D2O) used for NMR experiments were purchased
from Cambridge Isotope Laboratories, Inc. (Andover, MA). Reduced
β-nicotinamide-adenine dinucleotide phosphate (NADPH) for
CYP4B1 incubations was purchased from OYC Americas (Vista,
CA). 1,2-Dilauroyl-sn-glycero-3-phosphocholine (DLPC) for
CYP4B1 incubations was purchased from Avanti Polar Lipids
(Alabaster, AL). Tissue culture supplies (Gibco) and organic solvents
were purchased from Thermo-Fisher Scientific (Pittsburgh, PA), and
all other chemicals were obtained from Sigma-Aldrich (St. Louis,
MO).
Whole Cell Measurement of CYP4B1-Mediated Toxicity. A
cell proliferation assay was used to assess the cytotoxicity of
compounds C1−C8. HepG2 cells were modified by lentiviral
transduction to express rabbit CYP4B1, or a control vector, as
previously described.4,6,11,16 Western blot analysis of the cell lines
showed ∼130 pmol CYP4B1/mg of lysate protein and no detectable
enzyme in the control vector line. Enzyme levels were quantitated
using recombinantly expressed purified rabbit CYP4B1 as a standard
(Figure S1). Stably transduced cells were grown in tissue culture
media (DMEM, 10% FBS, 1% pen/strep, and 10 mM HEPES) and
selected with puromycin treatment (2 μg/mL) over a 7-day period.
Cells were seeded into white-walled 96-well plates at 10 000 cells/well
and grown for 24 h. Both cell lines were then treated with C1−C8 at
eight doses spanning the following concentrations: C1 0.55−1200
μM, C2 0.14−300 μM. C3−C6 0.4−50 μM, C7 0.14−300 μM, C8
1.6−150 μM (DMSO did not exceed 0.1%). Cells were incubated for
72 h, and the viability was determined with the CellTiter-Glo
Luminescent Assay (Promega, Madison, WI). Vehicle-treated wells
were set to 100% viability, and lethal dose 50% (LD50) values were
calculated from dose−response curves using nonlinear regression
curve fitting with GraphPad Prism 7.00.
N-Octyl-3-furancarboxamide, C8. Exact mass [M + H] calculated
1
224.1651, observed 224.1643, δ ppm 3.4. H NMR (CDCl3),7.94 (s,
1H), 7.38 (s, 1H), 6.67 (s, 1H), 6.52 (bs, 1H), 3.34 (q, J = 7 Hz, 2H),
1.54 (p, J = 7.5 Hz, 2H), 1.27 (m, 10H), 0.85 (t, J = 7 Hz, 3H). Yield
48%.
N-(2-Hydroxyethyl)-3-furancarboxamide, C2-OH. Exact mass [M
1
+ H] calculated 156.0655, observed 156.0652, δ ppm 1.9. H NMR
(CDCl3), 7.95 (s, 1H), 7.43 (s, 1H), 6.63 (s, 1H), 6.33 (bs, 1H), 3.82
(s, 2H), 3.59 (s, 2H). Yield 42%.
N-(2-Hydroxypentyl)-3-furancarboxamide, C5-OH. Exact mass
1
[M + H] calculated 198.1125, observed 198.1126, δ ppm 0.6. H
NMR (CDCl3), 7.93 (s, 1H), 7.43 (s, 1H), 6.61 (s, 1H), 5.94 (bs,
1H), 3.67 (t, J = 6.5 Hz, 2H), 3.43 (q, J = 7 Hz, 2H), 1.74 (bs, 1H),
1.64 (sx, J = 7 Hz, 4H), 1.47 (m, 2H). Yield 46%.
N-(2-Hydroxyoctyl)-3-furancarboxamide, C8-OH. Exact mass [M
1
+ H] calculated 240.1594, observed 240.1595, δ ppm 0.4. H NMR
(CDCl3), 7.92 (s, 1H), 7.43 (s, 1H), 6.61 (s, 1H), 5.84 (bs, 1H), 3.64
(t, J = 6.5 Hz, 2H), 3.4 (q, J = 7 Hz, 2H), 1.67 (bs, 1H), 1.58 (sx, J =
7.5 Hz, 4H), 1.35 (m, 8H). Yield 35%.
Chemical Synthesis. The general method for N-alkyl-3-
furancarboxamides was performed as follows. A solution containing
1 equiv (eq) of 3-furoic acid, 2 eq of thionyl chloride, and a catalytic
amount of DMF in anhydrous dichloromethane (DCM) was refluxed
for 3 h to generate a crude batch of 3-furoyl chloride. Solvent and
excess thionyl chloride were evaporated, and the 3-furoyl chloride was
used in the following step without further workup. A solution of 3-
furoyl chloride in DCM was next cooled over ice, and 2 eq of the
aliphatic amine (ethyl-, through octyl-amine), along with a catalytic
N-Methyl-3-furancarboxamide, C1. Synthesis of C1 was per-
formed as previously described.17 Briefly, 1 eq of 3-furoic acid, 1.2 eq
of EDC-HCl, 0.3 eq of HOBt, and 1.05 eq of MeNH3Cl were placed
in a sealed flask flushed with nitrogen. MeCN was added, and the
solution was stirred for 5 min, followed by addition of 1.05 eq of
Et3N. The reaction was stirred for 1.5 h, and water was added prior to
workup. The mixture was extracted thrice with ethyl acetate followed
by a brine wash, then dried over magnesium sulfate and purified via
B
DOI: 10.1021/acs.chemrestox.9b00330
Chem. Res. Toxicol. XXXX, XXX, XXX−XXX

Chemical Research in Toxicology
Article
flash chromatography with a hexane/ethyl acetate gradient. Exact
mass [M + H] calculated 126.0550, observed 126.0553, δ ppm 2.4. 1H
NMR, 7.92 (s, 1H), 7.43 (s, 1H), 6.60 (s, 1H), 2.96 (s, 3H). Yield
65%.
the following settings: capillary 3.50 kV, cone 30.0 V, collision 10.0 V,
source offset 60.0 V, source temperature 150 °C, desolvation
temperature 350 °C, cone gas flow 150 L/h, desolvation gas flow
800 L/h, and collision gas flow 0.15 mL/min. Using these conditions,
C5-OH, C5-Adduct, and C8-OH eluted at 3.7, 4.1, and 4.7 min and
were detected, respectively, in the following channels by MRM: m/z
198.1 > 95.1, 657.2 > 528.1, and 240.2 > 95.1. C5-Adduct was also
detected via NL129 with results practically identical to that obtained
with MRM. Separation of C2-OH was also achieved with a gradient of
0.1% formic acid in H2O (solvent A) and 0.1% formic acid in
acetonitrile (solvent B) at a flow rate of 0.3 mL/min. After holding at
2% for 3 min, B was increased linearly to 95% between 3 and 9 min
and held at 95% between 9 and 10 min. The total run time was 12
min following a 2 min equilibration period. C2-OH eluted at 4.9 min
and was detected by MRM at m/z 156.1 > 95.1. Calibration standards
for C2-OH, C5-OH, C5-Adduct, and C8-OH were prepared in blank
incubation matrix from DMSO stocks and metabolite analysis was
performed with MassLynx V4.1.
Characterization of C5-Adduct Isomers. Metabolic incubations
contained 50 pmol purified rabbit CYP4B1, 100 pmol CPR, 50 pmol
cytochrome b5, 5 μg of DLPC, 100 U catalase, 50 μg of superoxide
dismutase, 5 mM NAL, and 5 mM GSH in 50 mM KPi, pH 7.4, in a
final volume of 100 μL. Compound C5 was added to achieve a final
concentration of 200 μM, and then the same incubation and workup
procedure described above in the GSH/NAL trapping experiments
was used. A supernatant aliquot (10 μL) was analyzed by UPLC−
MS/MS on a Waters Acquity UPLC connected to an AB Sciex Triple
TOF 5600 (Framingham, MA) set to ESI+ mode with the following
settings. Ion source gas 1, 60 arbitrary units (au); ion source gas 2, 60
au; curtain gas, 35 au; temperature, 600 °C; ionspray voltage, 4500 V;
declustering potential, 100 V; collision energy, 35 V; collision energy
spread, 20 V; ion release delay, 67 au; and ion release width, 25 au.
Isomers were separated chromatographically using a Tosoh
Biosciences (Tokyo, Japan) TSKgel ODS-100 V column (2.0 × 150
mm2, 3 μm), starting with 100% mobile phase A (0.2% acetic acid in
H2O) at a flow rate of 0.2 mL/min. After a 2 min hold, mobile phase
B (0.2% acetic acid in methanol) was increased linearly from 0 to 60%
between 2 and 50 min, then ramped to 100% B rapidly and held at
100% B from 50.1 to 56 min. Total run time was 60 min following a 4
min equilibration period. Using these conditions, isomers numbered
1, 2, 3, and 4 eluted at retention times (RT) of 40.0, 40.8, 41.9, and
42.4 min, respectively. Analytes were detected by both TOF MS
positive mode using a 50−1000 Da window and positive mode
product ion scanning from the parent ion of m/z 657.29 Da.
CYP4B1 Substrate Docking. Substrate docking was performed
in AutoDock Vina19 using the crystal structure of rabbit CYP4B120
(PDB 5T6Q). Models of C2, C5, and C8 were generated in Open
Babel21 at pH 7.4, and the substrate and protein files were processed
using AutoDock tools. Docking was restrained to the active site area
of the enzyme above the heme moiety and used default settings.
Multiple simulations were performed for each substrate with
consistent results for the preferred pose directionality and docking
scores. Visualization of docked C2, C5, and C8 in the active site was
performed in Chimera.22
Glutathione (GSH)/N-α-Acetyl-L-lysine (NAL) C5 Adduct Stand-
ard, C5-GSH/NAL. A modified synthesis procedure that has been
published was utilized.18 To a mixture of 1 eq of compound C5 in a
1:1 solution of saturated aqueous NaHCO3/acetone was added 1.1 eq
of potassium peroxymonosulfate. The in situ generation of
dimethyldioxirane and oxidation of C5 was allowed to proceed for
30 min at room temperature with stirring. Subsequently, 1.1 eq of
GSH and 2 eq of NAL were added, and the reaction allowed to
proceed for 1 h at 70 °C. Upon cooling, the precipitates were filtered
out, and the crude mixture was subjected to semipreparative HPLC
purification on a Phenomenex (Torrance, CA) Luna C18 column (10
× 250 mm2, 5 μm, 100 Å), starting with 95% mobile phase A (0.1%
formic acid in H2O) and 5% mobile phase B (0.1% formic acid in
acetonitrile) at a flow rate of 5 mL/min. After holding for 5 min, B
was increased linearly from 5 to 30% between 5 and 30 min, then B
was increased linearly from 30 to 50% between 30 and 35 min. At
35.01 min, B was increased to 95% and held for 5 min; total run time
was 45 min following a 5 min equilibration period. The synthetic
standard was characterized by high resolution LC−MS/MS and 2D-
NMR techniques (Figure S3 and S4). Exact mass [M + H] calculated
657.2912, observed 657.290 (δ ppm 1.3). Yield 16%.
GSH/NAL Trapping and Adduct Comparison of CYP4B1-
Generated Reactive Intermediates from Compounds C1−C8.
Metabolic incubations contained 25 pmol purified rabbit CYP4B1, 50
pmol CPR, 25 pmol cytochrome b5, 2.5 μg of DLPC, 100 U catalase,
50 μg of superoxide dismutase, 5 mM NAL, and 5 mM GSH in 50
mM potassium phosphate buffer (KPi), pH 7.4, in a final volume of
100 μL, similar to as reported previously.13 Compounds C1−C8 were
added to achieve a final concentration of 100 μM. After 30 min
reconstitution on ice, metabolic reactions were initiated by the
addition of NADPH (1 mM final concentration) and allowed to
progress for 30 min at 37 °C. Reactions were quenched with an equal
volume of acetonitrile containing 2 μM furafylline as the internal
standard and centrifuged at 16 000g for 5 min. A supernatant aliquot
(10 μL) was analyzed by UPLC−MS/MS on a Waters Acquity UPLC
(Milford, MA) connected to a Xevo TQ-S instrument in ESI+ mode,
with the following settings: capillary 3.50 kV, cone 30.0 V, collision
10.0 V, source offset 60.0 V, source temperature 150 °C, desolvation
temperature 350 °C, cone gas flow 150 L/h, desolvation gas flow 800
L/hour, and collision gas flow 0.15 mL/min. Analytes were separated
chromatographically using a Waters HSS T3 column (2.1 × 100 mm2,
1.8 μm), starting with 95% mobile phase A (0.1% formic acid in H2O)
and 5% mobile phase B (0.1% formic acid in acetonitrile) at a flow
rate of 0.3 mL/min. After holding for 1 min, B was increased linearly
from 5 to 95% between 1 and 7 min and held at 95% between 7 and
10 min; total run time was 12 min following a 2 min equilibration
period. Under these conditions, the eight different GSH/NAL
adducts eluted at 3.2, 3.4, 3.6, 3.8, 4.0, 4.2, 4.5, and 4.7 min for
C1−C8, respectively, and were detected with neutral loss scanning of
129 Da (NL129). Windows spanning 5 Da were set for each
compound, surrounding the parent m/z for each adduct. The internal
standard, furafylline, was eluted at 4.1 min and was detected by
multiple reaction monitoring (MRM) at m/z 261.3 > 81.0. Peak area
ratios (PAR) for each adduct were determined by dividing the peak
areas of the analytes by that of furafylline, and metabolite analysis was
performed with MassLynx V4.1.
Time-Dependent Inhibition (TDI) Assays. Inhibition reactions
contained 100 pmol purified rabbit CYP4B1, 200 pmol CPR, 100
pmol cytochrome b5, 10 μg of DLPC, 100 U catalase, and 50 μg of
superoxide dismutase, in 50 mM KPi, pH 7.4, in a final volume of 125
μL. Compound C5 or 1-aminobenzotriazole (ABT) was added from
concentrated DMSO stocks to achieve a final concentration of 100
μM, and DMSO alone was also utilized as a vehicle control. DMSO
concentration did not exceed 0.2%. Reactions were set up with and
without 5 mM NAL and GSH to examine the effect of trapping agents
on time-dependent inhibition. After 30 min on ice for reconstitution,
the reactions were initiated by the addition of NADPH (1 mM final
concentration). Aliquots of the inhibition reaction (25 μL) were taken
at times ∼0 and 30 min, diluted 20-fold into a 0.5 mL probe substrate
reaction containing lauric acid, at 10× the reported Km for CYP4B1
(350 μM23), in 50 mM KPi buffer with 1 mM NADPH. The probe
substrate reactions progressed for 10 min at 37 °C and were quenched
Kinetics of Metabolite Formation by CYP4B1. Metabolic
incubations contained the same reconstituted enzyme mixture as
outlined above in the above. Concentrations of C2, C5, and C8
spanned 50−3000, 1−300, and 1−100 μM, respectively. The
reactions were initiated by the addition of NADPH (1 mM final
concentration) and allowed to progress for 15 min at 37 °C, followed
by the same workup described above in the GSH/NAL trapping
experiments. Chromatographic separation of C5-OH, C5-Adduct,
and C8-OH utilized the same instrument, column, and LC method as
described above. A Xevo TQ-S instrument was set to ESI+ mode with
C
DOI: 10.1021/acs.chemrestox.9b00330
Chem. Res. Toxicol. XXXX, XXX, XXX−XXX

Chemical Research in Toxicology
Article
and amine combination are added to oxidized furans5,18,25 and
(ii) the facile ability to track GSH-trapped reactive metabolites
via neutral loss 129 scanning from different parent furan-
containing compounds.18 Preliminary work, performed with
the same instrument settings as detailed in Experimental
Procedures, showed that all eight parent compounds had
ionization levels within ∼2-fold of each other. Presumably, the
neutral loss of 129 is not greatly influenced by the species with
which GSH reacts. Therefore, the combination of labile GSH
fragmentation and similar parent ionization efficiency should
enable semiquantitative comparisons between GSH/NAL-
trapped adducts for the different compounds C1−C8.
Although structurally more divergent homologues might be
trapped with different efficiencies, we reasoned that the simple
aliphatic modifications located away from the furan pharma-
cophore likely does not cause a serious quantitative
complication.
Therefore, we could monitor CYP4B1-dependent bioactiva-
tion of compounds C1−C8 by measuring the relative amounts
of GSH/NAL-trapped pyrrole adduct generated from their
respective reactive intermediates. C1−C8-Adducts were
detected by neutral loss scanning and had base peak m/z
values that matched a proposed pyrrole adduct structure with
one each of GSH and NAL, and LC retention times that
increased according to parent substrate chain length, as
expected (Figure 2A,B). The relative amount of adduct
formation was measured via PAR, with respect to the internal
standard furafylline, and then normalized such that PAR was
with 50 μL of 10% HCl. Subsequently, 10 nmol of 15-
hydroxypentadecanoic acid was added as internal standard, and the
solutions were extracted twice with 1 mL of ethyl acetate. The organic
solvent was evaporated under a nitrogen stream, and the residues
were redissolved in 50 μL of ethyl acetate. Then 50 μL of BSTFA
reagent (Sigma) was added, and the silylation reactions were heated
to 90 °C for 180 min. After cooling, the samples were analyzed by gas
chromatography−mass spectrometry (GC−MS) on a Shimadzu
QP2010 Gas Chromatograph Quadrupole mass spectrometer
(Columbia, MD) with electron impact (EI) ionization using a
previously reported method.24
RESULTS
■
CYP4B1-Mediated Cytotoxicity of N-Alkyl-3-furancar-
boxamide Analogs C1−C8. Previously, we demonstrated
the utility of HepG2 cells stably transfected with CYP4B1 as a
tool for investigating the potency of bioactivatable furan
substrates that elicit cytotoxicity.12,13 A series of N-alkyl-3-
furancarboxamides (C1−C8, Figure 1A) were chosen to
Figure 1. N-Alkylated 3-furancarboxamides (C1−C8) were synthe-
sized to establish optimal aliphatic tail length for CYP4B1-mediated
bioactivation, and compounds were utilized for cytotoxicity and
metabolism assays. (A) LD50 values from CYP4B1-HepG2 cells dosed
substrates C1−C8, and assayed for viability 72 h later, showed that
C3−C6 were the most potent congeners. Values presented are means
SD from three to six replicates. (B) Representative dose−response
curve for substrate C5 illustrates the 5 μM LD50 detected in the
CYP4B1-HepG2 cell line. No toxicity was observed in the control
vector HepG2 cell line from exposure to C5, nor from any of the
analogs up to doses outlined in the Experimental Procedures section.
establish the optimal aliphatic chain-length for CYP4B1 furan
substrate bioactivation. Dose-dependent toxicity was observed
in CYP4B1-HepG2 cells with all eight compounds. A
representative dose−response curve is shown for C5 in Figure
1B. Compounds C3, C4, C5, and C6 were the most potent
cytotoxins exhibiting LD50 values 6 1, 5 1, 5 1, and 5
1 μM, respectively. CYP4B1-mediated cytotoxicity was greatly
attenuated (one to two log-fold) when chain-length was
shortened to < C3 or lengthened beyond C6. LD50 values of
680 40 and 24 3 μM were measured for the short-chain
compounds C1 and C2, respectively, while LD50 values of 26
Figure 2. Assessment of GSH/NAL trapped reactive intermediates
derived from substrates C1−C8. (A) Separate CYP4B1 incubations
with C1−C8, in the presence of GSH (5 mM) and NAL (5 mM),
each produced an NADPH-dependent metabolite adduct with a
parent mass corresponding to a pyrrole structure containing one GSH
and one NAL moiety (termed C1−C8-Adduct). Relative amounts of
the adducts were evaluated via NL129 scanning and eluted according
to parent chain length. Peak sizes have been normalized for viewing.
Because of the short chromatographic separation, any different
isomers present could not be resolved. (B) Average PAR values of the
C1−C8-Adduct products with corresponding adduct m/z. Shown is
3 and 130
3 μM were measured for the long-chain
compounds C7 and C8 (Figure 1A). No toxicity was observed
in the control-HepG2 cells for any of the compounds, up to
doses allowed by aqueous solubility.
Relative CYP4B1-Mediated Reactive Intermediate
Formation from Compounds C1−C8. The nucleophile
tandem of GSH and NAL was originally chosen for our studies
due to (i) the stable pyrrole formation that occurs when a thiol
the normalized average
SD derived from three replicates.
Compounds C3−C6 produced substantially more trapped adduct
than either the short (C1, C2) or long (C7, C8) chain homologues.
D
DOI: 10.1021/acs.chemrestox.9b00330
Chem. Res. Toxicol. XXXX, XXX, XXX−XXX

Chemical Research in Toxicology
Article
set to 100 for the C5-Adduct. The respective amounts
generated from each substrate (average from three replicates
SD) are shown in Figure 2B. Adduct PAR showed an elevated
value for compounds C3 through C6 that inversely mirrored
the LD50 values reported in Figure 1A. Short-chain compounds
C1 and C2, as well as long-chain compounds C7 and C8,
generated ∼10- to >100-fold less GSH/NAL-adduct than their
midlength chain homologues.
MS/MS Characterization of CYP4B1-Mediated For-
mation of Isomeric C5-Adducts. For the simple com-
parative experiment above, we did not separate all isomeric
products that were formed and could be detected with much
longer chromatography runs. A detailed characterization was
carried out on the pyrrole adducts formed from a fortified
CYP4B1 incubation with compound C5. In the presence of
GSH, NAL, and NADPH, CYP4B1 produced four isobaric
metabolite peaks identified through an extended chromato-
graphic separation followed by MS/MS analysis. These
analytes eluted at retention times of 40.0, 40.8, 41.9, and
42.4 min and are numbered 1, 2, 3, and 4, respectively (Figure
3A). Each of these metabolites had an [M + H] peak at m/z
657.29 and likely represent different regio-isomers of a pyrrole
adduct structure containing one NAL and one GSH moiety.
Tandem MS of each analyte showed a fragment ion at m/z
528, indicative of a neutral loss of 129, as well as other
diagnostic ions that are common fragments observed with
GSH adducts (Figure 3B: isomer 1, Figure S2A,C: isomers 2
and 4). Isomer 3 lacked the characteristic GSH-adduct [M+H
− 75] fragment ion that was present in each of the other three
isomers at m/z 582. Therefore, 3 was suggested to be the result
of a cyclization reaction, perhaps resulting from the free acid of
the GSH-glycine moiety attacking a cyclic iminium inter-
mediate. A possible structure, along with assigned fragmenta-
tion data, is shown in Figure S2B and D.
Characterization of Synthesized C5-GSH/NAL Adduct
Standard. As a means of confirming our mass spectral
evidence for the C5-Adduct, we synthesized a chemical
standard of a C5-GSH/NAL adduct. Our synthesis produced a
dominant species that matched the CYP4B1 incubation-
derived C5-Adduct 1 by LC−MS at retention time 40.0 min
(Figure 3A). The synthesized compound had an [M + H] peak
at m/z 657.292 and exhibited the same MS/MS fragmentation
1
pattern as observed for isomer 1 (Figure 3B,C). H NMR
analysis of the C5-GSH/NAL standard showed the presence of
a pyrrole substructure and aliphatic protons consistent with
side chain NAL and GSH moieties. Additional structural
validation was performed via multiplicity-edited HSQC NMR
(Figure S3). ROESY was used to characterize the GSH thiol
attachment site to the pyrrole adduct from through-space
proton interactions (Figure S4). The assignments specified the
pyrrole protons as straddling the carboxamide with the GSH
moiety attached adjacent to the NAL side chain. When tested
with the aforementioned NL129 LC−MS/MS method, the
synthesized C5-GSH/NAL adduct displayed a linear response
from ∼1 nM to 10 μM. This broad range of linearity provided
additional confidence in comparing relative adduct amounts
deriving from compounds C1−C8. Of note, a very minor
second peak that matched the retention time of adduct 2 was
also present in the LC−MS chromatogram of the synthetic
standard, but insufficient material was available for further
NMR analysis.
Figure 3. Characterization of GSH/NAL-containing C5-Adducts.
(A) Extended chromatographic separation of the C5-Adduct from the
CYP4B1 incubation revealed four isobaric species at m/z 657.292
(upper chromatogram). Chemical synthesis of a C5-GSH/NAL
adduct produced one dominant species of the same mass, with the
same chromatographic profile as isomer 1. A trace amount of isomer 2
at RT 40.8 min was also observed (lower chromatogram). (B) MS/
MS analysis of isomer 1 was identical to that of the synthesized
standard and produced ions typical of a GSH adduct. MS/MS spectra
for isomers 2−4 is shown in the supplemental section. (C) Assigned
structure for isomer 1, as elucidated from 2D-NMR techniques
performed on the synthesized adduct (Figures S3 and S4), with
annotated fragmentation pattern corresponding to the MS/MS
spectrum.
Furancarboxamide Series. Preliminary experiments sug-
gested that the major hydroxylation (+16) metabolite of the N-
alkyl-3-furancarboxamides, produced by CYP4B1, occurred at
the ω-position of the alkyl tail, although metabolism of both
C5 and C8 produced additional mono-oxygenated products
Kinetics of CYP4B1-Mediated ω-Hydroxylation for
C2, C5, and C8 and Adduct Formation for C5 of 3-
E
DOI: 10.1021/acs.chemrestox.9b00330
Chem. Res. Toxicol. XXXX, XXX, XXX−XXX

Chemical Research in Toxicology
Article
that were tentatively assigned as the ω−1−OH analogs (Figure
S5). Therefore, compounds C2, C5, and C8 were chosen
(spanning the homologous series) as substrates for a kinetic
analysis of CYP4B1-dependent ω-hydroxylation, following
synthesis of the respective C2, C5, and C8 ω-OH metabolites
(Figure 4A−D). CYP4B1 exhibited kinetics consistent with the
Figure 5. CYP4B1 time-dependent inhibition (TDI) investigation for
compound C5. A 30 min (t30) incubation of CYP4B1 with 100 μM
C5 and NADPH resulted in a ∼44% reduction in enzyme activity, as
normalized to vehicle control. Addition of GSH and NAL to the
incubation abolished this observed TDI. The 0 min (t0) incubation,
prior to probe substrate turnover, shows C5 is not engaging in
reversible inhibition of CYP4B1. The pan-CYP inhibitor ABT was
used as a positive control for MBI and was unaffected by the addition
of the trapping agents GSH/NAL.
inhibitor 1-aminobenzotriazole26 (ABT) resulted in 78% loss
of activity for t30, with minimal inactivation observed at t0.
Addition of the trapping agent combination, GSH and NAL,
had no significant effect on t0 CYP4B1 activity; however, this
condition completely abrogated the C5 TDI for t30. Upon
bioactivation, a true mechanism-based inactivator reacts in the
active site of the catalyzing enzyme and is not significantly
affected by the addition of trapping agents. It is therefore
possible that bioactivated C5 causes its low amount of
CYP4B1 TDI through reaction with nucleophilic surface
amino acid residues belonging to CYP4B1 or CPR. TDI of
CYP4B1 by ABT was not affected by the addition of GSH and
NAL.
Docking of Compounds C2, C5, and C8 into the
CYP4B1 Crystal Structure. The rabbit CYP4B1 crystal
structure shows a narrow and hydrophobic active site, with
residues Phe309, Leu485, Val375, and Val378 acting to confine
aliphatic substrates, like n-octane.20 The Gln218 and Tyr379
residues at the mouth of the boot-like binding pocket may
participate in hydrogen-bonding to the carboxylate moiety of
fatty acids or other electron-rich functionalities such as the
furan in the 3-furancarboxamide series. As the carbon chain is
lengthened for compounds C2, C5, and C8, docking studies
predict increased affinity toward CYP4B1, generating scores
that are approximations of ΔG values of −5.7, −6.7, and −7.4
kcal/mol, respectively. The top poses for C2 and C5 (Figure 6;
colored yellow and green, respectively) position the furan ring
above the heme for epoxidation, with the 4′ and 5′ carbons of
the furan at 3.8 and 3.3 Å from the heme iron for both C2 and
C5. The top pose for C8 (Figure 6; colored plum) reoriented
the substrate with the aliphatic tail positioned over the heme
for ω-hydroxylation, situating the terminal carbon 3.1 Å away
from the heme iron. Additionally, the furan moiety oxygen of
C8 is located 3.3 and 3.6 Å from residues Gln218 and Tyr379,
respectively. This is consistent with hydrogen-bonding
contacts that may aid in terminal tail positioning for C8,
which might explain the observed CYP4B1 preference for ω-
hydroxylation of this substrate over epoxidation. The positions
of these substrates in the CYP4B1 active site should not be
considered as atomically exact, but as representative computa-
tional models consistent with our experimental data.
Figure 4. Kinetics for CYP4B1 metabolism of compounds C2, C5,
and C8. (A) Rate assessment for ω-hydroxylation of C2 (C2-OH),
(B) C5 (C5-OH), and (C) C8 (C8-OH). (B) Kinetics for C5
bioactivation and adduct (C5-Adduct) formation was also evaluated.
(D) Kinetic constants for ω-hydroxylation showed a decrease in Km
and an overall increase in catalytic efficiency as the aliphatic chain was
extended. The Vmax for C5-Adduct formation was ∼1.5-times that of
C5-OH. Shown is the mean SD from three replicates.
Michaelis−Menten model for a single-substrate reaction for
C2-, C5-, and C8-ω-OH formation, with Km values of 2.7 mM,
51 μM, and 15 μM along with Vmax values of 0.03, 7.4, and 10
pmol/pmol enzyme/min, respectively. Because of the fact that
we were unable to achieve full apparent saturation with C2, the
kinetic parameters calculated for this substrate should be
treated as an approximation. The catalytic efficiencies for ω-
hydroxylation of C2, C5, and C8 by CYP4B1 increased
substantially with increasing alkyl tail length, namely, 1.2e-5,
0.15, and 0.67 μM−1 min−1, respectively.
With a characterized C5-GSH/NAL adduct standard
(isomer 1) in hand, CYP4B1-mediated flux between reactive
intermediate and ω-hydroxylation could be directly compared
for compound C5. Metabolism of C5 by CYP4B1 to the
trapped C5-Adduct exhibited kinetics again consistent with the
Michaelis−Menten model of a single-substrate reaction and
apparent saturation by substrate was achieved. The Km, Vmax
,
and catalytic efficiency values for C5 were calculated to be 48
μM, 11 pmol/pmol enzyme/min, and 0.23 μM−1 min−1,
respectively.
Investigation of Time-Dependent Inhibition of
CYP4B1 by C5. A single concentration point, time-dependent
inhibition assay was used to determine if the reactive
electrophilic intermediates generated from compound C5
(chosen due to the high amount of trapped reactive
intermediate formation) were causing inactivation of
CYP4B1. A 30 min inhibition preincubation (t30) with
compound C5 showed a 44% reduction in CYP4B1 activity,
as measured by the rate of lauric acid metabolism to the 11-
and 12-OH metabolites (Figure 5). The 0 min preincubation
(t0) did not show any change relative to the vehicle control,
indicating lack of both reversible and time-dependent
inhibition after the dilution step for the probe substrate
turnover. Incubation with the pan-CYP mechanism-based
F
DOI: 10.1021/acs.chemrestox.9b00330
Chem. Res. Toxicol. XXXX, XXX, XXX−XXX

Chemical Research in Toxicology
Article
necessary for protoxicant effectiveness with CYP4B1, (ii) a 3′
positioned amide between the furan and tail substructure is an
effective linker for preserving the chemical reactivity of furan
metabolites, and (iii) an intermediary degree of aliphatic
character is ideal for bioactivation. Additionally, compounds
C2, C5, C8 were shown to be stable to hydrolysis in complete
tissue culture media because no significant decline in
concentration was observed after 72 h (Figure S6).
Consequently, conclusions drawn from studying this series is
not confounded by undesirable hydrolytic reactions.
Because lipophilicity undoubtedly plays an important role in
cellular permeability of the compounds, one might have
surmised that C8 would have been the most potent congener
based on accumulated intracellular concentration. Therefore,
the distinctive cytotoxicity profile we obtained prompted a
biochemical analysis of the CYP4B1-dependent metabolism of
C1−C8 homologues. Congeneric pyrrole adducts derived from
furan bioactivation were easily detected for C1−C8 utilizing
NL129 scanning. Upon investigating the relative amounts
generated, normalized PAR for the adduct species were
strikingly proportional to the cytotoxicity-dose response
curve observed in whole HepG2 cells. The parabolic
relationship seen between relative adduct amount and alkyl
chain length strongly mirrored the relationship observed
between cytotoxicity and alkyl chain length (Figure 7A),
thereby demonstrating a tight relationship between furan
bioactivation and cell death.
While clearly a useful tool in the analysis of xenobiotic
metabolic liability, attempts to demonstrate a direct link
between covalent binding and hepatotoxicity have been
elusive. Historically, there are convincing examples showing
the correlation between the covalent binding of reactive
metabolites and liver necrosis.28 However, as shown from
metabolic analysis of hepato- and nonhepato-toxins, in vitro
covalent binding studies, when analyzed alone, are not
predictive of liver toxicity at least.29 To our knowledge, the
data presented here are unique in demonstrating a highly
quantitative association between a covalent binding event (in
this case the trapped reactive species formed) and cytotoxicity.
This strong correlation facilitates construction of a simple
association model; a log−log plot of adduct PAR versus the
LD50 results in a linear relationship (Figure 7B). If a power
function does indeed model the dependency of cytotoxicity on
reactive intermediate formation in this HepG2 system, then
furan-containing congeners bioactivated by CYP4B1 that result
in even higher quantities of electrophilic species would only
show a nominal increase in potency. This is a unique finding
that illustrates the robust association between these two
variables that has not been observed to this degree before, and
thus could have implications for future toxicological studies. A
caveat is that the current research has not explored covalent
binding of radiolabeled versions of these compounds, although
we suggest a similar differential effect would be observed
between amount of covalent binding generated in vitro and
whole cell toxicity.
Figure 6. Docking of compounds C2 (yellow), C5 (green), and C8
(plum) in the CYP4B1 crystal structure. The CYP4B1 active site
contains hydrophobic residues Leu485, Val378, Val375, and Phe309
that aid in the enzyme’s preference for aliphatic substrates. As the
carbon tail chain length increases for C2 through C8, improved
affinity for these substrates is predicted. The preferred pose from
multiple simulations showed both C2 and C5 positioned with their
furan moiety located directly above the heme, with carbons at the 2′
and 3′ positions 3.3 and 3.8 Å away from the heme iron. The
preferred pose for C8 reversed this positioning, with the carbon tail
now centered over the heme, 3.1 Å away from the iron. Gln218 and
Tyr379 at the mouth of the active site may play a role in hydrogen-
bond stabilization of the C8 furan.
DISCUSSION
■
Of the known xenobiotics that undergo CYP4B1-mediated
bioactivation, the natural products IPO and PK have been
most extensively studied, both in vitro and in vivo, including the
use of IPO in humans.27 Given the potential liabilities of rapid
glucuronidation and metabolism to inactivated products for
IPO and PK, respectively, it seemed advantageous to
investigate other structurally related compounds to identify
new bioactivatable substrates for CYP4B1. The current study
was begun with the principal goal of elucidating alkyl furan
structure−activity relationships to better define elements of
protoxicant potency and metabolism by the enzyme. Rabbit
CYP4B1 is the most commonly studied form of the enzyme
and its use as a suicide gene was first suggested in the 1990s.33
In contrast to native human CYP4B1, rabbit CYP4B1 is both
highly stable and efficiently bioactivates furan protoxicants to
cytotoxic species. Although a re-engineered form of CYP4B1
with improved stability has been described recently,4 we used
rabbit CYP4B1 for all studies conducted here.
In general, trapping systems for reactive intermediates
typically produce multiple adduct species.25 Therefore, the
C5 furan was utilized for an in-depth interrogation of different
conceivable pyrrole isomers that could be trapped from a
CYP4B1 incubation. Previously, published research performed
with the GSH/NAL system had either not investigated the
possibility of multiple isomers formed or provided definitive
structural assignments of the bioactivated and trapped furan-
Substrates C1−C8 were readily synthesized and proved to
be an effective set of analogs to explore structure-metabolism
relationships for CYP4B1. Upon testing in our assay,
compounds C3−C6 were the most potent cytotoxins in the
CYP4B1-HepG2 cells, with LD50 values approximating that
seen with IPO and PK of ∼5 μM.12,13 From this we conclude
that clearly (i) an alcohol (i.e., such as present in IPO) is not
G
DOI: 10.1021/acs.chemrestox.9b00330
Chem. Res. Toxicol. XXXX, XXX, XXX−XXX

Chemical Research in Toxicology
Article
containing products.18 Using an extended chromatography
method with the GSH/NAL system, we detected two major
and two minor isobaric C5-Adducts formed in an NADPH-
dependent manner (Figure 3A, upper). MS/MS analysis for
three of the isomers provided a rationale for the generation of
the pyrrole adduct, while a fourth showed evidence for a
cyclized variant, akin to that reported for GSH trapping of
IPO.30 The fragmentation data could not definitively assign the
thiol attachment of GSH for these isomers. Therefore, we
attempted to synthesize C5-GSH/NAL adduct standards to
allow for their structural characterization. Interestingly, only
one pyrrole adduct isomer could be isolated in abundance
(potentially due to elevated temperature, differential stoichi-
ometry of trapping agents, or solvents in the synthetic reaction
compared to the biological incubation), which had the same
LC−MS/MS profile as the metabolically generated isomer 1
(Figure 3A, lower). 2D NMR techniques enabled unambig-
uous determination of GSH attachment to the 5′ position of
the pyrrole. We were thus able to assign a structure for the
CYP4B1-generated isomer 1, which shows that C5 is
bioactivated and trapped by GSH/NAL in the same manner
as IPO.5 However, for C5, it is likely that both 1,2-addition
and 1,4-addition by GSH, followed by NAL attack and
condensation, contribute to the adducts formed (Figure 8).
Note that all of these products are represented in the single
peaks analyzed in a shorter chromatography run performed
across all C1−C8 analogs (Figure 2A). While other trapped
species were also detected in CYP4B1 incubations with C5 and
GSH/NAL (a potential limitation with this nucleophile
tandem), these were not explored in depth for this study.
Given the proclivity of CYP4B1 to ω-hydroxylate fatty acid
substrates,31 and the recent publication on PK metabolism by
CYP4B1,13 we hypothesized that as the aliphatic tail is
lengthened for these compounds, substrate reorientation may
occur in the active site of CYP4B1. This could result in a
switching of metabolism from furan bioactivation to
hydroxylation at the terminal end of the carbon chain, and
thus lower the cytotoxic potency. A kinetic analysis of
CYP4B1-dependent metabolism to the ω-hydroxyl metabolites
of compounds C2, C5, and C8 (chosen to span the series)
indeed showed that catalytic efficiencies (Vmax/Km) for ω-
hydroxylation increased substantially with tail length, peaking
for C8 at 0.67 μM−1 min−1 (Figure 4D).
Figure 7. Correlation between CYP4B1-mediated reactive metabolite
formation and cytotoxicity. (A) Parabolic relationship between the
relative amount of C1−C8-Adduct formed from trapped reactive
intermediate and alkyl chain length directly reflects the association
between cytotoxic potency (as measured by LD50) and alkyl chain
length for C1−C8. Averages have been plotted, and standard
deviations (reported earlier) have been omitted for clarity. (B)
Linear dependency results from a log−log plot of normalized adduct
amount versus LD50 and can be modeled by the equation y = 2.6x−0.6
+ 90.3x−0.6, which has been fit to the data; this is indicative of a power
relationship between these two variables.
Figure 8. Proposed mechanism for the formation of GSH/NAL-trapped C5-Adducts. Characterized isomer 1 can be formed through 1,2-addition
of GSH to the putative enedial species and reaction with NAL. Both 1,2- and 1,4-addition of GSH to the enedial and reaction with NAL may
account for the additional C5-Adducts identified through LC−MS/MS; the structures for the potential isomers 2−4 are shown.
H
DOI: 10.1021/acs.chemrestox.9b00330
Chem. Res. Toxicol. XXXX, XXX, XXX−XXX

Chemical Research in Toxicology
Article
Additional kinetic studies were performed for C5 metabo-
lism by CYP4B1, to compare the relative rates of furan
bioactivation versus ω-hydroxylation. Although most likely an
underestimation of the total reactive metabolite amount
generated, the Vmax for the adduct was only 1.5-fold higher
than that for the ω-hydroxylation process for C5. This is a key
finding regarding future optimization of potency for CYP4B1
protoxicants since extensive (inactivating) hydroxylation may
be a pharmacokinetic liability. Nonproductive metabolism to
hydroxy metabolites could then render these compounds
substrates for glucuronidation. We suggest further that potent
protoxicants that undergo CYP4B1-mediated bioactivation
with optimal tail length may benefit from fluorine substitution
on the aliphatic moiety to block oxidation at terminal carbon
positions.
To investigate differences in substrate binding to CYP4B1,
compounds C2, C5, and C8 were docked into the recently
published CYP4B1 crystal structure.20 The results indicated
increased affinity as the carbon chain was lengthened, from a
comparatively loosely bound C2 to a much more tightly bound
C8. Interestingly, the increased binding affinity for C8
paralleled a preference for substrate reorientation, with the
C8 aliphatic tail, rather than its furan moiety, now oriented
above the heme iron (Figure 6). This modeling matched
extremely well to the trends we saw in the kinetic and
metabolic data we acquired and bolsters the use of the
CYP4B1 crystal structure for future substrate design work.
Given our metabolic data and docking models that describe
the interactions between CYP4B1 and N-alkyl-3-furancarbox-
amide analogs, the parabolic relationship between alkyl chain
length and cytotoxicity may be rationalized by poorer binding
of the shorter-chain compounds to CYP4B1 and increased
oxidative metabolism for the longer chain compounds at sites
distant from the furan warhead.
Numerous furan-containing compounds act as mechanism-
based inactivators of CYPs.25 The reactive intermediates that
are generated in the CYP4B1-HepG2 system could partition
between quenching by active site nucleophiles and escape from
the CYP4B1 active site to cause cytotoxicity. To evaluate this
potential flux, a TDI investigation was performed that showed
a 44% loss of activity from C5 treatment of CYP4B1 in the
presence of NADPH. However, the addition of trapping agents
completely nullified the small degree of enzyme inactivation
observed (kobs ≈ 0.02 min−1 at 100 μM). This indicates that
bioactivated C5 is not a true mechanism-based inhibitor of
CYP4B1 as assessed by the criteria given by Silverman.32 In the
absence of GSH and NAL, escaped reactive intermediate may
cause some minimal time-dependent loss of enzyme activity
through attachment to exterior amino acid residues on either
CYP4B1 or CPR in this reconstituted system. We posit this
would play a marginal role if the enzyme is expressed in target
cells dosed with similar furan-containing congeners when
excess biological nucleophiles, like GSH, are present in high
concentration. Therefore, the CYP4B1 active site architecture
likely insulates the enzyme from self-inactivation and
contributes to its efficient cytotoxic action.
carbons, they are highly toxic in HepG2 cells expressing
CYP4B1 and elicit no observed effect in control vector cells.
Reducing aliphatic character lowers the affinity of these
substrates for the enzyme, while extending the tail increases
ω-hydroxylation rates; both of which lower the potency for
CYP4B1-mediated cytotoxicity. Relative amounts of trapped
reactive intermediates formed from CYP4B1 metabolism of
these substrates with GSH and NAL are strikingly correlated to
cytotoxicity; to our knowledge, a comparison like this has
never been shown to this high degree. This effect may be
explained due in large part to the efficient nature by which the
reactive intermediates are able to escape the CYP4B1 active
site without causing measurable mechanism-based inactivation.
These studies further demonstrate the utility of CYP4B1 for
use in suicide gene systems, especially in settings requiring
highly controlled cytotoxicity. Future studies will investigate
the utility of fluorinated congeners and probe the specificity of
bioactivation by hepatic CYPs, particularly CYP1A2.
ASSOCIATED CONTENT
* Supporting Information
■
S
The Supporting Information is available free of charge at
https://pubs.acs.org/doi/10.1021/acs.chemrestox.9b00330.
Western blot assessment of CYP4B1 expression levels in
HepG2 cells; MS/MS fragmentation data for C5-
Adduct isomers 2−4; HSQC structural validation for
C5-GSH/NAL; ROESY characterization of C5-GSH/
NAL; C2-, C5-, C8-OH metabolite identification; N-
alkyl-3-furancarboxamide stability assessment in tissue
culture media; NMR spectroscopy; annotated 1H NMR:
C1−C8 and C2-OH, C5-OH, C8-OH (PDF)
AUTHOR INFORMATION
Corresponding Author
*E-mail: Rettie@uw.edu.
■
ORCID
John P. Kowalski: 0000-0002-0343-8425
Miklos Guttman: 0000-0003-2419-1334
Funding
Brady Fund for Natural Products, NIH T32 Pharmacological
Sciences Training Grant, Strategic Research Fund of the
̈
Heinrich-Heine University, Dusseldorf, Germany.
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
We would like to thank Dr. Tim Martins (Quellos, Seattle) for
assistance with preliminary screening efforts.
ABBREVIATIONS
■
CYP4B1, cytochrome P450 4B1; IPO, 4-ipomeanol; PK,
perilla ketone; UGT, uridine 5’-diphosphoglucuronosyl trans-
ferase; CPR, cytochrome P450 reductase; NAL, N-α-acetyl-L-
lysine; KPi, potassium phosphate buffer; NL129, neutral loss
scanning of 129 Da; MRM, multiple reaction monitoring;
PAR, peak area ratio; TDI, time-dependent inhibition; ABT, 1-
aminobenzotriazole; C1−C8, N-alkyl-3-furancarboxamide
compounds with aliphatic tails of one through eight carbons;
C5-Adduct, GSH/NAL trapped adduct(s) from CYP4B1
bioactivation of C5
In summary, we have undertaken an in-depth character-
ization of the structural elements required for potent IPO-like
congeners that undergo CYP4B1 bioactivation to elicit whole
cell toxicity. N-Alkyl-3-furancarboxamides are stable, easily
synthesized for analog and library generation, and possess a
similar cytotoxicity and reactivity profile as their ketone
counterparts. When these congeners are affixed with 3−6 tail
I
DOI: 10.1021/acs.chemrestox.9b00330
Chem. Res. Toxicol. XXXX, XXX, XXX−XXX

Chemical Research in Toxicology
Article
(16) Schmidt, E. M., Wiek, C., Parkinson, O. T., Roellecke, K.,
Freund, M., Gombert, M., Lottmann, N., Steward, C. A., Kramm, C.
M., Yarov-Yarovoy, V., Rettie, A. E., and Hanenberg, H. (2015)
Characterization of an Additional Splice Acceptor Site Introduced
into CYP4B1 in Hominoidae during Evolution. PLoS One 10,
No. e0137110.
(17) Hyster, T. K., and Rovis, T. (2010) Rhodium-catalyzed
oxidative cycloaddition of benzamides and alkynes via C-H/N-H
activation. J. Am. Chem. Soc. 132, 10565−10569.
(18) Li, C., Lin, D., Gao, H., Hua, H., Peng, Y., and Zheng, J. (2015)
N-Acetyl Lysine/Glutathione-Derived Pyrroles as Potential Ex Vivo
Biomarkers of Bioactivated Furan-Containing Compounds. Chem. Res.
Toxicol. 28, 384−393.
(19) Trott, O., and Olson, A. J. (2009) Auto Dock Vina: improving
the speed and accuracy of docking with a new scoring function,
efficient optimization and multithreading. J. Comput. Chem. 31, 455−
461.
(20) Hsu, M. H., Baer, B. R., Rettie, A. E., and Johnson, E. F. (2017)
The Crystal Structure of Cytochrome P450 4B1 (CYP4B1)
Monooxygenase Complexed with Octane Discloses Several Structural
Adaptations for omega-Hydroxylation. J. Biol. Chem. 292, 5610−5621.
(21) O’Boyle, N. M., Banck, M., James, C. A., Morley, C.,
Vandermeersch, T., and Hutchison, G. R. (2011) Open Babel: An
open chemical toolbox. J. Cheminf. 3, 33.
(22) Pettersen, E. F., Goddard, T. D., Huang, C. C., Couch, G. S.,
Greenblatt, D. M., Meng, E. C., and Ferrin, T. E. (2004) UCSF
Chimera–a visualization system for exploratory research and analysis.
J. Comput. Chem. 25, 1605−1612.
(23) Baer, B. R., Schuman, J. T., Campbell, A. P., Cheesman, M. J.,
Nakano, M., Moguilevsky, N., Kunze, K. L., and Rettie, A. E. (2005)
Sites of covalent attachment of CYP4 enzymes to heme: evidence for
microheterogeneity of P450 heme orientation. Biochemistry 44,
13914−13920.
(24) McDonald, M. G., Ray, S., Amorosi, C. J., Sitko, K. A.,
Kowalski, J. P., Paco, L., Nath, A., Gallis, B., Totah, R. A., Dunham, M.
J., Fowler, D. M., and Rettie, A. E. (2017) Expression and Functional
Characterization of Breast Cancer-Associated Cytochrome P450 4Z1
in Saccharomyces cerevisiae. Drug Metab. Dispos. 45, 1364−1371.
(25) Peterson, L. A. (2013) Reactive metabolites in the
biotransformation of molecules containing a furan ring. Chem. Res.
Toxicol. 26, 6−25.
(26) de Montellano, P. R. O. (2018) 1-Aminobenzotriazole: A
Mechanism-Based Cytochrome P450 Inhibitor and Probe of
Cytochrome P450 Biology. Med. Chem. (Los Angeles, CA, U. S.) 8, 38.
(27) Rowinsky, E. K., Noe, D. A., Ettinger, D. S., Christian, M. C.,
Lubejko, B. G., Fishman, E. K., Sartorius, S. E., Boyd, M. R., and
Donehower, R. C. (1993) Phase I and pharmacological study of the
pulmonary cytotoxin 4-ipomeanol on a single dose schedule in lung
cancer patients: hepatotoxicity is dose limiting in humans. Cancer Res.
53, 1794−1801.
(28) Gillette, J. R. (1974) Commentary. A perspective on the role of
chemically reactive metabolites of foreign compounds in toxicity. I.
Correlation of changes in covalent binding of reactivity metabolites
with changes in the incidence and severity of toxicity. Biochem.
Pharmacol. 23, 2785−2794.
(29) Obach, R. S., Kalgutkar, A. S., Soglia, J. R., and Zhao, S. X.
(2008) Can in vitro metabolism-dependent covalent binding data in
liver microsomes distinguish hepatotoxic from nonhepatotoxic drugs?
An analysis of 18 drugs with consideration of intrinsic clearance and
daily dose. Chem. Res. Toxicol. 21, 1814−1822.
(30) Chen, L. J., DeRose, E. F., and Burka, L. T. (2006) Metabolism
of furans in vitro: ipomeanine and 4-ipomeanol. Chem. Res. Toxicol.
19, 1320−1329.
REFERENCES
■
(1) Edson, K. Z., and Rettie, A. E. (2013) CYP4 enzymes as
potential drug targets: focus on enzyme multiplicity, inducers and
inhibitors, and therapeutic modulation of 20-hydroxyeicosatetraenoic
acid (20-HETE) synthase and fatty acid omega-hydroxylase activities.
Curr. Top. Med. Chem. 13, 1429−1440.
(2) Baer, B. R., and Rettie, A. E. (2006) CYP4B1: an enigmatic P450
at the interface between xenobiotic and endobiotic metabolism. Drug
Metab. Rev. 38, 451−476.
(3) Zheng, Y. M., Fisher, M. B., Yokotani, N., Fujii-Kuriyama, Y., and
Rettie, A. E. (1998) Identification of a meander region proline residue
critical for heme binding to cytochrome P450: implications for the
catalytic function of human CYP4B1. Biochemistry 37, 12847−12851.
(4) Wiek, C., Schmidt, E. M., Roellecke, K., Freund, M., Nakano, M.,
Kelly, E. J., Kaisers, W., Yarov-Yarovoy, V., Kramm, C. M., Rettie, A.
E., and Hanenberg, H. (2015) Identification of amino acid
determinants in CYP4B1 for optimal catalytic processing of 4-
ipomeanol. Biochem. J. 465, 103−114.
(5) Baer, B. R., Rettie, A. E., and Henne, K. R. (2005) Bioactivation
of 4-ipomeanol by CYP4B1: adduct characterization and evidence for
an enedial intermediate. Chem. Res. Toxicol. 18, 855−864.
(6) Roellecke, K., Virts, E. L., Einholz, R., Edson, K. Z., Altvater, B.,
Rossig, C., von Laer, D., Scheckenbach, K., Wagenmann, M.,
Reinhardt, D., Kramm, C. M., Rettie, A. E., Wiek, C., and
Hanenberg, H. (2016) Optimized human CYP4B1 in combination
with the alkylator prodrug 4-ipomeanol serves as a novel suicide gene
system for adoptive T-cell therapies. Gene Ther. 23, 615−626.
(7) Bonini, C., and Mondino, A. (2015) Adoptive T-cell therapy for
cancer: The era of engineered T cells. Eur. J. Immunol. 45, 2457−
2469.
(8) Ciceri, F., Bonini, C., Gallo-Stampino, C., and Bordignon, C.
(2005) Modulation of GvHD by suicide-gene transduced donor T
lymphocytes: clinical applications in mismatched transplantation.
Cytotherapy 7, 144−149.
(9) Turtle, C. J., Hanafi, L. A., Berger, C., Gooley, T. A., Cherian, S.,
Hudecek, M., Sommermeyer, D., Melville, K., Pender, B., Budiarto, T.
M., Robinson, E., Steevens, N. N., Chaney, C., Soma, L., Chen, X.,
Yeung, C., Wood, B., Li, D., Cao, J., Heimfeld, S., Jensen, M. C.,
Riddell, S. R., and Maloney, D. G. (2016) CD19 CAR-T cells of
defined CD4+:CD8+ composition in adult B cell ALL patients. J. Clin.
Invest. 126, 2123.
(10) Statham, C. N., Dutcher, J. S., Kim, S. H., and Boyd, M. R.
(1982) Ipomeanol 4-glucuronide, a major urinary metabolite of 4-
ipomeanol in the rat. Drug Metab. Dispos. 10, 264−267.
(11) Parkinson, O. T., Teitelbaum, A. M., Whittington, D., Kelly, E.
J., and Rettie, A. E. (2016) Species Differences in Microsomal
Oxidation and Glucuronidation of 4-Ipomeanol: Relationship to
Target Organ Toxicity. Drug Metab. Dispos. 44, 1598−1602.
(12) Teitelbaum, A. M., McDonald, M. G., Kowalski, J. P.,
Parkinson, O. T., Scian, M., Whittington, D., Roellecke, K.,
Hanenberg, H., Wiek, C., and Rettie, A. E. (2019) Influence of
Stereochemistry on the Bioactivation and Glucuronidation of 4-
Ipomeanol. J. Pharmacol. Exp. Ther. 368, 308−316.
(13) Roellecke, K., Jager, V. D., Gyurov, V. H., Kowalski, J. P.,
Mielke, S., Rettie, A. E., Hanenberg, H., Wiek, C., and Girhard, M.
(2017) Ligand characterization of CYP4B1 isoforms modified for
high-level expression in Escherichia coli and HepG2 cells. Protein Eng.,
Des. Sel. 30, 205−216.
(14) Cheesman, M. J., Baer, B. R., Zheng, Y. M., Gillam, E. M., and
Rettie, A. E. (2003) Rabbit CYP4B1 engineered for high-level
expression in Escherichia coli: ligand stabilization and processing of
the N-terminus and heme prosthetic group. Arch. Biochem. Biophys.
416, 17−24.
(15) Chen, W., Koenigs, L. L., Thompson, S. J., Peter, R. M., Rettie,
A. E., Trager, W. F., and Nelson, S. D. (1998) Oxidation of
acetaminophen to its toxic quinone imine and nontoxic catechol
metabolites by baculovirus-expressed and purified human cyto-
chromes P450 2E1 and 2A6. Chem. Res. Toxicol. 11, 295−301.
(31) Fisher, M. B., Zheng, Y. M., and Rettie, A. E. (1998) Positional
specificity of rabbit CYP4B1 for omega-hydroxylation1 of short-
medium chain fatty acids and hydrocarbons. Biochem. Biophys. Res.
Commun. 248, 352−355.
(32) Silverman, R. B. (1988) Mechanism-Based Enzyme Inactivation:
Chemistry and Enzymology; CRC Press.
J
DOI: 10.1021/acs.chemrestox.9b00330
Chem. Res. Toxicol. XXXX, XXX, XXX−XXX

Chemical Research in Toxicology
Article
(33) Rainov, N. G., Sena-Esteves, M., Fraefel, C., Dobberstein, K. U.,
Chiocca, E. A., and Breakefield, X. O. (1998) A chimeric fusion
protein of cytochrome CYP4B1 and green fluorescent protein for
detection of pro-drug activating gene delivery and for gene therapy in
malignant glioma. Advances in experimental medicine and biology 451,
393.
K
DOI: 10.1021/acs.chemrestox.9b00330
Chem. Res. Toxicol. XXXX, XXX, XXX−XXX