CYP26C1 is a hydroxylase of multiple active retinoids and interacts with cellular retinoic acid binding proteins

Article abstract of DOI:10.1124/mol.117.111039

The clearance of retinoic acid (RA) and its metabolites is believed to be regulated by the CYP26 enzymes, but the specific roles of CYP26A1, CYP26B1, and CYP26C1 in clearing active vitamin A metabolites have not been defined. The goal of this study was to

Full text of DOI:10.1124/mol.117.111039

Molecular Pharmacology Fast Forward. Published on February 23, 2018 as DOI: 10.1124/mol.117.111039
This article has not been copyedited and formatted. The final version may differ from this version.
MOL #111039
ꢀ
CYP26C1 is a hydroxylase of multiple active retinoids and interacts with cellular retinoic
acid binding proteins
Guo Zhong, David Ortiz, Alex Zelter, Abhinav Nath, Nina Isoherranen
Departments of Pharmaceutics (GZ, NI) and Medicinal Chemistry (DO, AN), School of Pharmacy,
University of Washington, Seattle, WA, USA and Department of Biochemistry, School of Medicine,
University of Washington, Seattle, WA, USA
1
ꢀ
ꢀ

Molecular Pharmacology Fast Forward. Published on February 23, 2018 as DOI: 10.1124/mol.117.111039
This article has not been copyedited and formatted. The final version may differ from this version.
MOL #111039
ꢀ
Running Title: CYP26C1 as a retinoid hydroxylase
Corresponding Author: Nina Isoherranen, Department of Pharmaceutics, School of Pharmacy,
University of Washington, Health Science Building, Room H-272M, Box 357610, Seattle,
Washington 98195-7610 USA, Phone: 206-543-2517; Fax: 206-543-3204; Email: ni2@uw.edu
Text Pages: 24ꢀꢁ
Figures: 12ꢀꢁ
References: 69ꢀꢁ
Words in Abstract: 246ꢀꢁ
Words in Introduction: 744 ꢁ
Words in Discussion: 1486ꢀ
Abbreviations:
ACN, acetonitrile; ALDH, aldehyde dehydrogenase; ADH, alcohol dehydrogenase; atRA, all-trans retinoic
acid; atRAL, all-trans-retinaldehyde; CAD, collision associated dissociation; CE, collision energy;
CRABP, cellular retinoic acid binding protein; CRBP, cellular retinol binding protein; CXP, collision exit
potential; CYP, cytochrome P450; DP, declustering potential; EP, entrance potential; EPI, enhanced
m
product ion; FA, formic acid; IS, internal standard; K , apparent Michaelis-Menten constant in biological
matrix; LC-MS/MS, liquid chromatography tandem mass spectrometry; LC-UV, liquid chromatography
ultraviolet; LRAT, lecithin retinol acyltransferase; MD, molecular dynamics; MRM, multiple reaction
monitoring; RAR, retinoic acid receptor; RDH, retinol dehydrogenase; REH, retinyl ester hydrolase;
RMSD, root-mean-square-deviation; ROL, retinol; TEM, temperature
2
ꢀ
ꢀ

Molecular Pharmacology Fast Forward. Published on February 23, 2018 as DOI: 10.1124/mol.117.111039
This article has not been copyedited and formatted. The final version may differ from this version.
MOL #111039
ꢀ
Abstract
The clearance of retinoic acid (RA) and its metabolites is believed to be regulated by the CYP26 enzymes,
but the specific roles of CYP26A1, CYP26B1 and CYP26C1 in clearing active vitamin A metabolites have
not been defined. The goal of this study was to establish the substrate specificity of CYP26C1, and
determine whether CYP26C1 interacts with cellular RA binding proteins (CRABPs). CYP26C1 was found
to effectively metabolize all-trans- retinoic acid (atRA), 9-cis-retinoic acid (9-cis-RA), 13-cis-retinoic acid
and 4-oxo-atRA with the highest intrinsic clearance towards 9-cis-RA. In comparison to CYP26A1 and
CYP26B1, CYP26C1 resulted in a different metabolite profile for retinoids, suggesting differences in the
active site structure of CYP26C1 when compared to other CYP26s. Homology modeling of CYP26C1
suggested that this is due to distinct binding orientation of retinoids within CYP26C1 active site. In
comparison to other CYP26 family members, CYP26C1 was up to 10-fold more efficient in clearing 4-
oxo-atRA (intrinsic clearance 153 µl/min/pmol) than CYP26A1 and CYP26B1 suggesting that CYP26C1
may be important in clearing this active retinoid. In support of this, CRABPs delivered 4-oxo-atRA and
atRA for metabolism by CYP26C1. Despite the tight binding of 4-oxo-atRA and atRA with CRABPs, the
m
K value of these substrates with CYP26C1 was not increased when the substrates were bound with
CRABPs, in contrast to what is predicted by free drug hypothesis. Together these findings suggest that
CYP26C1 is a 4-oxo-atRA hydroxylase and may be important in regulating the concentrations of this active
retinoid in human tissues.
3
ꢀ
ꢀ

Molecular Pharmacology Fast Forward. Published on February 23, 2018 as DOI: 10.1124/mol.117.111039
This article has not been copyedited and formatted. The final version may differ from this version.
MOL #111039
ꢀ
Introduction
Vitamin A (retinol; ROL), through its active metabolite retinoic acid (RA), plays a critical role in
regulation of gene transcription, cell proliferation, differentiation and apoptosis (Noy, 2010; Tang and
Gudas, 2011). Retinol is believed to be devoid of biological activity, and is metabolized by alcohol and
aldehyde dehydrogenases to retinoic acid (RA) (Fig.1; Napoli, 2012; Kedishvili, 2016). The endogenous
vitamin A metabolites detected in human serum include all-trans- (atRA), 9-cis-, 13-cis- and 9,13-di-cis-
RA, 4-oxo-13-cis-RA and 4-oxo-atRA (Fig. 1 and 2; Arnold et al., 2012). Of these retinoids, atRA activates
retinoic acid receptors (RARs) and is believed to be the most important and biologically active endogenous
RA isomer (Allenby et al., 1994; Chambon, 1996; Stevison et al., 2015). However, 13-cis-RA, 9-cis-RA
and 4-oxo-atRA can also activate RARs (Idres et al., 2002; Topletz et al., 2014). Due to this
pharmacological activity and their favorable in vivo pharmacokinetics, 13-cis-RA and 9-cis-RA are used to
treat acne, high risk neuroblastoma (Veal et al., 2007) and chronic hand eczema (Schmitt-Hoffmann et al.,
2
011). Similarly, both 13-cis-RA and 9-cis-RA are teratogenic demonstrating classic retinoid effects
Willhite et al., 1986; Kraft and Juchau, 1993). 9-cis-RA is also a ligand of retinoid X receptors (RXRs)
with significantly higher binding affinity than atRA and 13-cis-RA (Åström et al., 1990; Heyman et al.,
992; Allenby et al., 1993), but the in vivo significance of 9-cis-RA is unclear. 4-oxo-atRA exhibits higher
(
1
affinity to RARα and similar affinity to RARβ as atRA, (Pijnappel et al., 1993; Idres et al., 2002; Topletz
et al., 2014) consistent with its teratogenicity (Herrmann, 1995), and ability to modulate positional
specification in early embryo (Pijnappel et al., 1993). However, the role of the different RA isomers and 4-
oxo-atRA in human physiology is unclear.
Biological effects of RA isomers and metabolites are dependent on the RAR binding affinity and the
cellular concentrations of the retinoids. Therefore, strict regulation of the physiological concentrations of
endogenous retinoids via tissue specific expression of retinoid synthesizing and metabolizing enzymes and
retinoid binding proteins (Fig.1) is critical. The two enzymes of the CYP26 family, CYP26A1 and
CYP26B1, have been identified as key enzymes responsible for clearing atRA and in controlling atRA
concentrations (Fig.1). CYP26A1 appears to be the liver atRA hydroxylase eliminating biologically active
4
ꢀ
ꢀ

Molecular Pharmacology Fast Forward. Published on February 23, 2018 as DOI: 10.1124/mol.117.111039
This article has not been copyedited and formatted. The final version may differ from this version.
MOL #111039
ꢀ
retinoids, while CYP26B1 appears to be responsible for atRA clearance in extrahepatic tissues (Thatcher
et al., 2010; Topletz et al., 2014). CYP26A1 and CYP26B1 have also been shown to metabolize 4-oxo-
atRA and 9-cis-RA, but the catalytic efficiency is much lower than that towards atRA (Thatcher et al.,
2
011; Topletz et al., 2012; Diaz et al., 2016), suggesting that other enzymes may be responsible for the
elimination of 4-oxo-atRA and RA isomers in humans. The sequence similarity of human CYP26C1 is only
3% with CYP26A1 and 51% with CYP26B1 (Taimi et al., 2004), suggesting structural differences
4
between these CYPs. Hence, CYP26C1 may have different substrate specificity than CYP26A1 and
CYP26B1. Indeed, in CYP26C1 transfected COS-1 cells, both 9-cis-RA and atRA were identified as
potential substrates of CYP26C1 (Taimi et al., 2004). However, the role of CYP26C1 in retinoid clearance
or in formation of active metabolites such as 4-oxo-atRA has not been defined and biochemical
characterization of CYP26C1 in retinoid metabolism is lacking.
A distinct characteristic of biologically important enzymes contributing to retinoid metabolism is that they
interact with and obtain their substrates from cellular retinoid binding proteins (Fig.1; Napoli, 2017). atRA
is bound inside cells by cellular retinoic acid binding proteins (CRABPs) which appear to deliver atRA
either to nuclear RARs to regulate gene transcription, or to CYP26A1 and CYP26B1 for catabolism
(
Napoli, 2017). Hence, whether metabolic enzymes interact with CRABPs can define the biological
importance of the enzyme and the substrates. For example CRABPI-bound atRA was metabolized in rat
testes microsomes at a similar rate as free atRA, while CRABPI-binding of 4-oxo-atRA abolished 4-oxo-
atRA metabolism, suggesting substrate specific delivery to metabolism (Fiorella and Napoli, 1991, 1994).
Recently, atRA was shown to be channeled to recombinant CYP26B1 by CRABPs (Nelson et al., 2016),
but other substrates that bind to CRABPs were not studied, and the specificity of CRABP interactions with
CYP26 enzymes was not established. To better understand the role of CYP26 enzymes in retinoid
metabolism, the CRABP and retinoid specific delivery of substrates to metabolism by CYPs need to be
defined. The goal of this study was to establish the ligand specificity of CYP26C1 and characterize the
CRABP-CYP26C1 interactions to delineate the biological significance and the role of CYP26C1 in
regulating retinoid homeostasis.
5
ꢀ
ꢀ

Molecular Pharmacology Fast Forward. Published on February 23, 2018 as DOI: 10.1124/mol.117.111039
This article has not been copyedited and formatted. The final version may differ from this version.
MOL #111039
ꢀ
Materials and Methods
Chemicals and reagents. atRA, NADPH, and ketoconazole were purchased from Sigma-Aldrich (St.
Louis, MO, USA). Talarozole was purchased from MedChem Express (Princeton, NJ, USA). 4(S)-OH-
atRA and 4(R)-OH-atRA were synthesized as described previously (Shimshoni et al., 2012). All other
retinoids were from Toronto Research Chemicals (North York, ON, Canada). All solvents were HPLC or
Optima grade and were purchased from Thermo Fisher Scientific (Waltham, MA, USA).
Cloning, expression and characterization of recombinant CYP26C1. Human CYP26C1 cDNA was
purchased from OriGene Technologies (Rockville, MD, USA) (SKU: SC307567). Primers were designed
to amplify the CYP26C1 coding sequence minus the stop codon and the resulting PCR product was cloned
into Invitrogen pFastBac-CT TOPO following the manufacturer's instructions. The resulting coding
sequence consisted of CYP26C1 plus a C-terminal TEV cleavage site followed by a 6xHis tag. Full length
C-terminal 6xHis tagged CYP26C1 protein was then produced in Sf9 cells using Invitrogen’s Bac-to-Bac
baculovirus expression system according to the manufacturer's instructions. Sf-900 II SFM liquid media
(
Invitrogen; Carlsbad, CA, USA) supplemented with 2.5% fetal bovine serum was used for cell growth.
During protein expression, media was supplemented with δ-aminolevulinic acid (0.3 mM) and ferric citrate
0.2 mM) 24 h post-infection to enable heme synthesis. Infected cells were harvested after 48 h, washed in
(
PBS with 1 mM PMSF, and stored at −80 °C. The microsomal fractions of insect cells were prepared by
ultracentrifugation based on published methods (Topletz et al., 2012). The content of active P450 was
measured by a CO-difference spectrum (Omura and Sato, 1964), using an Aminco DW2 dual-beam
spectrophotometer (Olis Instruments; Bogart, GA, USA). Supersomes containing human cytochrome P450
oxidoreductase and b5 prepared from insect cells (Corning Inc.; Corning, NY, USA) were used as negative
control for CO-difference spectra. Protein concentration of the CYP26C1 microsomes was measured using
BCA protein assay kit according to manufacturer’s recommendations (Thermo Fisher Scientific; Waltham,
MA, USA). The CYP26C1 microsomes were used in all catalytic experiments. Rat P450 reductase was
expressed in E.coli and purified as described previously (Woods et al., 2011).
6
ꢀ
ꢀ

Molecular Pharmacology Fast Forward. Published on February 23, 2018 as DOI: 10.1124/mol.117.111039
This article has not been copyedited and formatted. The final version may differ from this version.
MOL #111039
ꢀ
Qualitative western blotting was conducted based on published method for CYP26A1 with minor
modifications (Lutz et al., 2009). CYP26C1 microsomes containing 2 µg total protein were loaded on the
gel. Mouse anti-6Xhis antibody (1:2000; Qiagen, Valencia, CA, USA) and IRDye® 680RD anti-mouse
antibody (1: 10000; LI-COR Biosciences, Lincoln, NE, USA) were used as primary and secondary
antibodies, respectively, to detect the his-tagged CYP26C1.
General protocol for incubations. Unless otherwise described, incubations and retinoid extractions were
performed following published procedures used for analysis of CYP26A1 and CYP26B1 activity (Thatcher
et al., 2010; Topletz et al., 2012). Similar incubation conditions and reductase to CYP ratios were used for
consistency. The rat reductase was added into CYP26C1 microsomes and allowed to incorporate into the
membranes at room temperature for 5 min (Thatcher et al., 2010; Topletz et al., 2012), before adding the
membranes into the incubation mixture containing substrate and 100 mM potassium phosphate buffer in a
total volume of 1 ml. After another 5 min pre-incubation at 37°C with the substrate, the reaction was
initiated with the addition of 1 mM NADPH. Incubations were quenched and extracted with 3 ml ethyl
acetate as previously described (Topletz et al., 2012). Ethyl acetate layer was collected, evaporated to
dryness under N
content and rat reductase concentration used for incubations are specified for each experiment below. 4-
oxo-atRA-d (2 µl of 5 µM solution) was added as an internal standard for all the incubations except those
2
flow and reconstituted with 100 µl methanol. Incubation time, microsomal CYP26C1
3
using UV-Vis detection. All product formation was confirmed to be NADPH dependent by comparison to
identical incubations in the absence of NADPH. Incubations were conducted in duplicates or triplicates and
all experiments repeated three times.
Evaluation of potential CYP26C1 substrates. To identify substrates of CYP26C1, individual retinoids at
1
µM concentration were incubated with 18 nM microsomal CYP26C1 and 36 nM rat reductase for 30 min.
Tested retinoids included RA isomers (atRA, 13-cis-RA and 9-cis-RA), 4-oxo-RA isomers (4-oxo-atRA,
-oxo-13-cis-RA and 4-oxo-9-cis-RA), 4-OH-RA isomers (4-OH-atRA, 4-OH-13-cis-RA and 4-OH-9-cis-
4
RA), ROL isomers (atROL, 13-cis-ROL and 9-cis-ROL) and 3-dehydro-atROL. For incubations with RA
isomers, 4-oxo-RA isomers or 4-OH-RA isomers, samples were extracted and analyzed by LC-UV as
7
ꢀ
ꢀ

Molecular Pharmacology Fast Forward. Published on February 23, 2018 as DOI: 10.1124/mol.117.111039
This article has not been copyedited and formatted. The final version may differ from this version.
MOL #111039
ꢀ
described previously (Thatcher et al., 2010) using an Agilent 1200 series HPLC system (Agilent
Technologies; Santa Clara, CA, USA) equipped with a Zorbax C18 column (3.5 µm, 2.1 X 100 mm; Agilent
Technologies; Santa Clara, CA, USA) and by using water instead of 50 mM ammonium acetate (pH 4.5)
in the mobile phase. The column temperature was maintained at room temperature. Analytes were
monitored at UV wavelength of 360 nm. For incubations with ROL isomers and 3-dehydro-atROL, hexanes
(
4 ml) were used to quench the reaction and extract analytes. After hexane layer was collected and
evaporated, samples were reconstituted with 100 µl acetonitrile (ACN). Retinyl acetate (5 µl of 5 µM
solution) was added to the samples as internal standard. Samples were analyzed by LC-UV using the
Agilent 1200 series HPLC system coupled with an Ascentis® RP-amide column (2.7µm, 15cm x 2.1 mm;
Sigma-Aldrich; St. Louis, MO, USA). Column temperature was maintained at 40°C. The flow rate was 0.5
ml/min. The gradient elution started with initial condition of 40% aqueous with 0.1% formic acid (FA) and
6
0% acetonitrile (ACN) with 0.1% FA, which was maintained for 2 min, followed by a gradient to 100%
ACN with 0.1% FA over 11 min, and held at that for 8 min before returning to the initial conditions before
next injection. Analytes were monitored at UV wavelength of 325 nm.
Stereoselective formation of 4-OH-RA enantiomers from RA isomers by CYP26C1. The
stereoselectivity in the formation of 4-OH-RA by CYP26C1 was determined for individual RA isomers by
incubating atRA, 13-cis-RA or 9-cis-RA at 0.75 µM with 15 nM CYP26C1, and 30 nM reductase for 15
min. For comparison, CYP26A1 was incubated under identical conditions with the same substrates.
Samples were extracted and analyzed by LC-UV using an Agilent 1200 series HPLC system coupled with
a Chiracel OD-RH column (5 µm, 2.1 X 150 mm; Chiral Technologies Inc.; Westchester, PA, USA) as
described previously to detect 4(S)- and 4(R)-OH-RA formation (Fig.2; Shimshoni et al., 2012). The peak
area ratio of 4(S)- and 4(R)-OH-RA was calculated and used as an indicator of the stereoselectivity of 4-
OH-RA formation.
Identification of metabolites formed by CYP26C1. To identify metabolites formed by CYP26C1, the
incubations described above for substrate identification were further analyzed by LC-MS/MS using an AB
Sciex API5500 Q/LIT mass spectrometer (AB Sciex LLC; Framingham, MA, USA) equipped with Agilent
8
ꢀ
ꢀ

Molecular Pharmacology Fast Forward. Published on February 23, 2018 as DOI: 10.1124/mol.117.111039
This article has not been copyedited and formatted. The final version may differ from this version.
MOL #111039
ꢀ
1
290 Infinity UHPLC (Agilent Technologies; Santa Clara, CA, USA). All analytes were monitored with
electrospray ionization under negative ion mode. To evaluate the identity of metabolites of RA isomers,
samples were first separated using a Zorbax C18 column (3.5 µm, 2.1 X 100 mm) and a mobile phase flow
rate of 0.35 ml/min, with a linear gradient from 90% aqueous with 0.1% FA and 10% ACN to 5% aqueous
with 0.1% FA and 95% ACN over 10 min. The column temperature was maintained at 40°C. The multiple
reaction monitoring (MRM) transitions included m/z 315 > 253 (4-OH- and 18-OH-RA), m/z 315 > 241
(
16-OH-RA) and m/z 313 > 269 (4-oxo-RA) as described previously (Topletz et al., 2012). All of the MS
parameters set for MRM scans are listed in Supplemental Table 1. Together with monitoring MRM
transitions, parallel information of [M-H] ions, including m/z 315 and m/z 313 Da, were acquired via
enhanced product ion (EPI) scan. The MS/MS spectra were collected from m/z 50 to 330 Da. EPI scan rate
1 3 0
was set at 10,000 Da/s, LIT fill time at 100 msec, Q at low resolution and Q entry barrier at 8 V. Q trap
was turned on. Other mass spectrometer detection settings were the same as those used in MRM scan,
except that declustering potential (DP) of -80 V and collision energy (CE) of -35 V with a spread of ± 10
V for both [M-H] ions were used. For further separation of the metabolites, the MS/MS spectra were also
collected after chiral separation of the products using the Chiracel OD-RH column and chromatography as
described above. Mass spectrometry settings for MRM and EPI scans were as described above.
To determine the identity of 4-oxo-atRA metabolites, a Zorbax C18 column was used for analyte
separation and the incubations were analyzed by LC-MS/MS. The flow rate was 0.2 ml/min, with a linear
gradient from 90% aqueous with 0.1% FA and 10% ACN to 10% aqueous with 0.1% FA and 90% ACN
over 15 min, held for 8 min before returning to initial conditions. The MRM transitions monitored included
m/z 329 > 255 (4-oxo-16-OH-atRA) and m/z 329 > 267 Da (4-oxo-18-OH-atRA) as described previously
(
Topletz et al., 2012). Parameters set for MRM scans are provided in Supplemental Table 1. The EPI scan
with the spectrum range of 50 to 350 Da was acquired for the [M-H] ion of m/z 329 Da. EPI settings were
similar to those described for metabolites of RA isomers except that the scan rate was 1,000 Da/S and Q
was set at unit resolution.
1
9
ꢀ
ꢀ

Molecular Pharmacology Fast Forward. Published on February 23, 2018 as DOI: 10.1124/mol.117.111039
This article has not been copyedited and formatted. The final version may differ from this version.
MOL #111039
ꢀ
Enzyme kinetics of RA isomers and 4-oxo-atRA with CYP26C1. To determine enzyme kinetic
parameters for RA isomers, varying concentrations of atRA (10-250 nM), 13-cis-RA (5-250 nM) or 9-cis-
RA (3-100 nM) were incubated with 2 nM CYP26C1 and 4 nM reductase for 30 seconds. Standard curves
of RA isomers, 4-OH-RA isomers and 4-oxo-RA isomers were constructed for quantification. Samples
were extracted and analyzed by LC-MS/MS using AB Sciex QTRAP 4500 mass spectrometer (AB Sciex
LLC; Framingham, MA, USA) coupled with Shimadzu UFLC XR DGU-20A5 (Shimadzu Corporation;
Kyoto, Japan) and a Zorbax C18 column as described above for metabolite identification. 4-OH-, 18-OH-
3
and 16-OH-RA isomers, 4-oxo-RA isomers, 4-oxo-atRA-d , and RA isomers were monitored by MRM,
and the MRM transitions and MS parameters used are listed in Supplemental Table 2. Based on the peak
area ratio of metabolites to the internal standard, the formation of major metabolites of RA isomers was
quantified using Analyst software (AB Sciex LLC; Framingham, MA, USA). Due to the sequential
metabolism of 4-OH-RA to 4-oxo-RA, to quantify the 4-OH-RA formation from atRA and 13-cis-RA, the
formation of 4-OH-RA and 4-oxo-RA were summed as previously described for CYP26A1 and CYP26B1
(
Topletz et al., 2014) (Supplemental Fig.1A). For 9-cis-RA, only the formation of 4-OH-9-cis-RA was
quantified because the formation of 4-oxo-9-cis-RA was quantitatively insignificant. The velocity of
metabolite formation was plotted as a function of RA concentration and substrate concentration was
corrected for depletion by calculating the average of the initial added concentration and final measured
concentration as described previously (Lutz et al., 2009). Enzyme kinetic parameters, including kcat and K
were obtained from GraphPad Prism 5.0 (GraphPad Software Inc.; San Diego, CA, USA) by fitting the
Michaelis-Menten equation to the data. Intrinsic clearance (Clint) was calculated as kcat/K
m
,
m
.
To obtain enzyme kinetic parameters for 4-oxo-atRA, both product formation and substrate depletion of
-oxo-atRA by CYP26C1 were measured. For product formation, varying concentrations of 4-oxo-atRA
4
(
3-100 nM) were incubated with 2 nM CYP26C1 and 4 nM reductase for 2 min. Samples were extracted
and analyzed by LC-MS/MS using ABsciex 4500 QTRAP mass spectrometer and the Zorbax C18 column
with the same LC conditions as described above. Monitored analytes included 4-oxo-16-OH-atRA, 4-oxo-
atRA and 4-oxo-atRA-d
3
. MRM transitions and MS parameters are listed in Supplemental Table 2. Due to
0ꢀ
1
ꢀ

Molecular Pharmacology Fast Forward. Published on February 23, 2018 as DOI: 10.1124/mol.117.111039
This article has not been copyedited and formatted. The final version may differ from this version.
MOL #111039
ꢀ
the tight binding of 4-oxo-atRA to CYP26C1, enzyme kinetic parameters, including kcat and K
obtained from GraphPad Prism 5.0 by fitting the Morrison equation to the data as described previously
Shimshoni et al., 2012) and assuming a CYP26C1 concentration of 2 nM in the incubations. Because of
the lack of synthetic 4-oxo-16-OH-atRA standard, the product formation was estimated as the ratio of peak
area of 4-oxo-16-OH-atRA to IS to allow K determination. For substrate depletion, experiments were
m
, were
(
m
conducted following published procedures (Shimshoni et al., 2012) with varying concentrations of 4-oxo-
atRA (5-150 nM) incubated with 2 nM CYP26C1 and 4 nM reductase in a total volume of 3 ml. At 0.5, 1,
2
3
and 4 min, 0.5 ml-aliquots were taken and quenched with 3 mL ethyl acetate. 4-oxo-atRA-d was added
as an internal standard and 4-oxo-atRA quantified as described for product formation assays. K
were obtained using GraphPad Prism 5.0 as described previously (Shimshoni et al., 2012).
m
and Clint
Differences in the kinetic parameters between substrates were evaluated using Graphpad Prism 5.0. The
differences for each kinetic parameter (K , kcat and Clint) were tested via ANOVA including all different
substrates in the analysis, and Tukey’s test was used as a post hoc test to identify differences between
substrates. Due to the multiple ANOVA comparisons done (comparison of K
m
m
, kcat and Clint values) a
Bonferroni adjustment for the p-value was done (0.05/3) and a p-value less than 0.01 was considered
significant.
Expression and purification of CRABPs. CRABP expression vectors were a gift from Dr. Noa Noy (Case
Western University). His-tagged CRABP-I and CRABP-II vectors were used to transform E.Coli.
Transformed cells were plated into LB agar plates containing 30 µg/ml kanamycin. Overnight cultures were
prepared from freshly streaked plates. Kanamycin added LB media of 500 ml was inoculated with overnight
cultures and incubated at 37°C with shaking until optical density at the wavelength of 600 nm reached 0.6.
IPTG was added at a final concentration of 1 mM and the culture was incubated at 37°C for another 2 hours.
Induced cells were collected by centrifugation at 6,000 X g for 20 min. Cell pellets were stored at -80°C
until protein purification. For purification of CRABPs, frozen cell pellets were thawed on ice, resuspended
with 20 ml lysis buffer (1 mg/ml lysozyme, 500 mM NaCl, 20 mM Tris, 5 mM imidazole, pH7.4) and
incubated at room temperature for 20 min with gentle shaking. Resuspended mixture was placed on ice and
1
1ꢀ
ꢀ

Molecular Pharmacology Fast Forward. Published on February 23, 2018 as DOI: 10.1124/mol.117.111039
This article has not been copyedited and formatted. The final version may differ from this version.
MOL #111039
ꢀ
sonicated for 30 seconds twice with a 30-sec interval followed by centrifugation at 12,000 X g for 30 min.
Supernatant was applied to a HisTrap HP column (GE Healthcare Bio-Sciences; Marlborough, MA, USA)
using Biologic DuoFlow chromatography system (Bio-Rad Laboratories; Hercules, CA, USA) following
manufacturer’s instructions. Loaded samples were first washed with 5X column volume of 20 mM Tris-
HCl, 500 mM NaCl, 20 mM imidazole, pH 7.4) and then CRABPs were eluted with 5X column volume of
2
0 mM Tris-HCl, 500 mM NaCl, 300 mM imidazole, pH 7.4). Fractions containing CRABPs were
concentrated and exchanged into HEDK buffer (10 mM Hepes, 0.1 mM EDTA, 0.5 mM DTT, 100 mM
KCl, pH 8.0) using Amicon 10kD molecular weight cutoff filters (EMD Millipore; Billerica, MA, USA).
Purified CRABPs were stored at -20°C in HEDK buffer with 50% glycerol.
Effect of CRABPs on retinoid metabolism by CYP26C1. To investigate how CRABPs affect atRA,
9
-cis-RA and 4-oxo-atRA metabolism by CYP26C1, three experiments were conducted as previously
described with minor modifications (Nelson et al., 2016). For all the CRABP-related incubations,
microsomal CYP26C1 (2 nM) and rat reductase (4 nM) were first pre-incubated with 1 mM NADPH in a
total volume of 1 ml at 37°C for 3 min. The reaction was initiated by adding free substrate (atRA, 4-oxo-
atRA, or 9-cis-RA) or substrates premixed with CRABP-I or CRABP-II in 1:1 concentration ratio. One to
one binding of substrates (atRA or 4-oxo-atRA) with CRABPs was confirmed with fluorescence titration
and via isolation of the holo-CRABP from CRABP-ligand mixture using the method adapted from Fiorella
et al (Fiorella and Napoli, 1991; Fiorella et al., 1993) in which the ligand is bound with CRABP and then
holo-CRABP is separated from any free ligand using a mini desalting column (Thermo Fisher Scientific).
After the incubation time specified for each experiment, the reaction was quenched with ACN (1 ml)
3
followed by the addition of ethyl acetate (3 ml) and 4-oxo-atRA-d (2 µl of 5 µM solution) as previously
described (Nelson et al., 2016). The rest of the procedures were as described above and metabolite
formation was measured as described above using LC-MS/MS. To characterize the effect of CRABPs on
CYP26C1 activity, first, CYP26C1 was incubated with 100 nM free substrate or 100 nM CRABP-ligand
mixture (1:1) for 2 min. The formation rates of 4-OH-atRA, 4-OH-9-cis-RA and 4-oxo-16-OH-atRA were
compared in the absence and presence of CRABPs. Second, the kinetics of 4-OH-atRA and 4-oxo-16-OH-
1
2ꢀ
ꢀ

Molecular Pharmacology Fast Forward. Published on February 23, 2018 as DOI: 10.1124/mol.117.111039
This article has not been copyedited and formatted. The final version may differ from this version.
MOL #111039
ꢀ
atRA formation as metabolites of atRA and 4-oxo-atRA were characterized, respectively. For this, varying
concentrations of CRABP-I and CRABP-II bound atRA- or 4-oxo-atRA (3-200 nM; 1:1 CRABP-ligand
ratio) were incubated with CYP26C1 for 1 (atRA) or 2 min (4-oxo-atRA) and extracted as described above.
m
The K and kcat of atRA and 4-oxo-atRA bound to CRABP were obtained by fitting the Michaelis-Menten
equation and the Morrison equation to the data, respectively. Third, the effect of excess apo-CRABPs on
CYP26C1 activity was evaluated as described previously for CRABPs and CYP26B1 (Nelson et al., 2016).
Individual substrate at 100 nM was pre-mixed with 0-400 nM CRABP-I or CRABP-II and incubated with
CYP26C1 in the presence of NADPH for 2 min. Metabolite formation rates were measured and compared
to the control which contained no CRABP in the incubation. Differences in kinetic values were compared
together with all substrates by ANOVA followed by Tukey’s test as a post hoc test as described above for
enzyme kinetic analysis.
Ligand-induced binding spectra and determination of IC50 values of talarozole and ketoconazole with
CYP26C1. To explore the active site characteristics of CYP26C1 and test if talarozole and ketoconazole
bind to the active site of CYP26C1 and interact with the heme, CYP26C1 microsomes at 0.7 µM in 100
mM potassium phosphate buffer (pH 7.4) were added into the sample and reference cuvettes, with 0.45 ml
in each. Microsomal CYP26C1 was titrated with ketoconazole (100, 149 and 198 µM) or talarozole (20, 79
and 157 µM) in methanol in the sample cuvette, while an equal volume of methanol was added into the
reference cuvette. The binding spectra were obtained using an Aminco DW2 dual-beam spectrophotometer
(
Olis Instruments). The spectra were normalized to 415 or 420 nm for comparison.
To determine if talarozole and ketoconazole are also inhibitors of CYP26C1 and to obtain IC50 values,
talarozole (0.01-25 µM dissolved in DMSO) and ketoconazole (25-200 µM dissolved in methanol) were
incubated with 2 nM CYP26C1, 4 nM reductase and 50 nM atRA for 2 min. Because the K values for 9-
m
cis-RA were much lower than atRA with CYP26C1, instead of using 9-cis-RA as a substrate in the
inhibition assays as previously described (Thatcher et al., 2011), assays were performed using atRA as a
substrate at 50 nM, a concentration similar to the K
volume of DMSO or methanol as those with talarozole or ketoconazole. Samples were analyzed by LC-
3ꢀ
m
value of atRA. Control incubation contained an equal
1
ꢀ

Molecular Pharmacology Fast Forward. Published on February 23, 2018 as DOI: 10.1124/mol.117.111039
This article has not been copyedited and formatted. The final version may differ from this version.
MOL #111039
ꢀ
MS/MS as described above for enzyme kinetic experiments. The peak area ratio of 4-OH-atRA to IS was
used as a measure of CYP26C1 activity. The percent remaining activity compared to control was calculated
and plotted as a function of inhibitor concentration. The IC50 values were determined by nonlinear
regression using GraphPad Prism 5.0 as described previously (Thatcher et al., 2011).
Generation of a CYP26C1 homology model and docking of retinoids. To explore the ligand binding
interactions of CYP26C1, a CYP26C1 homology model was created with I-TASSER (Roy et al., 2010)
using the sequence for human CYP26C1 (NCBI sequence number NP_899230.2). The following structures
were used as templates: the retinoic acid-bound cyanobacterial CYP120A1 (PDB code 2VE3; Kühnel et
al., 2008), the lanosterol- (PDB code 4LXJ), itraconazole- (PDB code 5EQB; Monk et al., 2014) and
fluconazole- (PDB code 4WMZ; Sagatova et al., 2015) bound structures of a CYP51 from S. cerevisiae,
and the thioperamide-bound human CYP46A1 (PDB code 3MDM; Mast et al., 2010). The heme group was
automatically docked in I-TASSER according to the docking position in structural analogs. To model a
structure in solution capable of binding CYP26 substrates, molecular dynamics simulations were performed
with GROMACS version 4.6.5 (Van Der Spoel et al., 2005). The topology files were modified to reflect
the cysteine thiolate linkage to the heme iron as previously described (Autenrieth et al., 2004; Oda et al.,
2
005). Atomic interactions were defined with the GROMOS 54a7 force field. The model was placed in a
dodecahedron periodic box of diameter 3.75 nm and then energy minimized by the steepest descent and
+
-
conjugate gradient methods. The system was solvated with ~89,000 SPC water molecules and Na and Cl
ions were added to neutralize the system with an effective 0.15 M ionic strength. Molecular dynamics (MD)
simulations reached equilibrium in ~7 ns and were allowed to proceed for 40 ns. Root-mean-square-
deviation (RMSD) and active site volume during the MD simulation is shown in Supplemental Fig. 2. To
select a conformation suitable for docking studies, the active site volume throughout the MD simulation
was measured with Pocket Volume Measurer 2.0 (Durrant et al., 2014). Candidate models were selected
from different time points representing conformations with maximum, minimum and intermediate active
site volumes. Each candidate was then subjected to in silico ligand docking screens in Autodock 4.0
(
Österberg et al., 2002). The heme charges were manually edited to match the high spin ferric state as
4ꢀ
1
ꢀ

Molecular Pharmacology Fast Forward. Published on February 23, 2018 as DOI: 10.1124/mol.117.111039
This article has not been copyedited and formatted. The final version may differ from this version.
MOL #111039
ꢀ
described previously (Shahrokh et al., 2012). The following ligands were obtained from PubChem and
docked to each candidate model: all-trans-RA (Compound Identifier CID 444795), 9-cis-RA (CID
4
49171), 13-cis-RA (CID 5282379) and 4-oxo-atRA (CID 6437063). Side chains from the following
residues were selected to be flexible during the docking simulations for all substrates: Trp-117, Phe-295,
Phe-299 and Ile-497. Additionally, Leu-131 was made flexible for the docking of 4-oxo-atRA, and Thr-
3
00 was made flexible for the docking of 9-cis-RA and 4-oxo-atRA. Initial docking screens were performed
5
with 10 simulations per ligand and 2.5 x 10 energy evaluations per docking simulation. The model derived
from the 8 ns time point has an active site volume that reaches a local maximum (Supplemental Fig.2B)
and permits the binding of each ligand with high affinity. This model was chosen as the most suitable
candidate. The final CYP26C1 homology model was subjected to more extensive docking studies of 200
simulations per ligand with the same parameters as above. The quality of the model before and after
refinement was validated by Ramachandran Plot analysis (RAMPAGE server; Lovell et al., 2003) and
Verify3D score (Eisenberg et al., 1997). Parameters of homology model quality assessment and the
secondary structure assignments are listed in Supplemental Tables 3 and 4.
Results
Expression and characterization of CYP26C1. Microsomal CYP26C1 prepared from baculovirus
infected Sf9 insect cells showed a typical CO-difference spectrum with a defined peak at 448 nm, which
was not observed in the negative control (Fig.3). The cytochrome P450 content, calculated from the CO-
difference spectrum and from the microsomal protein content, was 84.3 pmol/mg protein (2.2 µM) in the
membrane preparation. The western blot developed with anti-6XHis antibody indicated expression of a
protein with an approximate size of 50-60 kD, corresponding to the predicted molecular weight of
CYP26C1, 57 kD (Fig.3) demonstrating CYP26C1 expression in Sf9 insect cell microsomes.
Identification and characterization of CYP26C1 substrates. Among all tested retinoids, all three RA
isomers, 4-OH-RA isomers and 4-oxo-atRA were substrates of CYP26C1, showing significant metabolite
formation in the incubations with CYP26C1 based on UV absorbance (Fig.4-6). In contrast, no significant
1
5ꢀ
ꢀ

Molecular Pharmacology Fast Forward. Published on February 23, 2018 as DOI: 10.1124/mol.117.111039
This article has not been copyedited and formatted. The final version may differ from this version.
MOL #111039
ꢀ
metabolite formation was detected in incubations with retinol isomers, 3-dehydro-atROL (Supplemental
Fig.3), 4-oxo-13-cis-RA (Fig.6B) or 4-oxo-9-cis-RA (Fig.6C). With all RA isomers (atRA, 13-cis-RA or
9
-cis-RA), 4-hydroxylation pathway was the primary metabolic pathway resulting in formation of 4-OH-
RA isomers, and subsequent minor formation of 4-oxo-RA and other hydroxylation products (M1-M8).
The primary hydroxylation products at the C16 and C18 positions were observed from atRA and C16
hydroxylation was observed from 13-cis-RA and 9-cis-RA (Fig.4). The identity of these minor metabolites
was confirmed by LC-MS/MS (Fig.4G-4I). Notably, the polar metabolites (M1-M8) observed in the
incubations with atRA, 13-cis-RA and 9-cis-RA (Fig.4) corresponded exactly to those observed as products
from the incubations of the corresponding 4-OH-RA isomers after incubation with CYP26C1 (Fig.5),
suggesting sequential metabolism of 4-OH-RA isomers formed from RA by CYP26C1. For 4-oxo-RA
isomers, CYP26C1 metabolized 4-oxo-atRA to a single metabolite identified as 4-oxo-16-OH-atRA by
LC-MS/MS (Fig.6A and 6D), but metabolism of 4-oxo-13cisRA and 4-oxo-9-cis-RA was negligible
(
(
Fig.6). The identity of the metabolites formed by CYP26C1 was confirmed via collecting MS/MS spectra
Supplemental Fig.4-6). The fragmentation patterns of metabolites were consistent with those described in
previous studies with other CYP26s (Thatcher et al., 2011; Topletz et al., 2012).
To further define the metabolite pattern for CYP26C1, the stereo-specificity of the formation of 4-OH-
RA (Fig.2) from atRA, 13-cis-RA or 9-cis-RA by CYP26C1 was evaluated using chiral chromatography
(
Fig.7). In the incubation with atRA, 4(R)-OH-atRA was the main enantiomer produced by CYP26C1 with
a 4S/4R ratio of 0.2 (Fig.7G). In comparison, similar amounts of 4(S)- and 4(R)-OH-13-cis-RA were formed
by CYP26C1 with a 4S/4R ratio of 1.3 (Fig.7H). From 9-cis-RA only 4(S)-OH-9-cis-RA was detected,
demonstrating distinct stereo-specificity in 4-OH-RA formation (Fig.7I). The stereo-specificity of 4-OH-
RA formation was clearly different for CYP26A1 in comparison to CYP26C1. CYP26A1 produced mainly
4
(S)-OH-RA from all three RA isomers. The 4S/4R ratios were 5.5 and 7.6 with atRA and 13-cis-RA as
CYP26A1 substrates, respectively (Fig.7D and 7E). With 9-cis-RA as the substrate, only the peak of 4(S)-
OH-9-cis-RA was observed with CYP26A1 (Fig.7F).
1
6ꢀ
ꢀ

Molecular Pharmacology Fast Forward. Published on February 23, 2018 as DOI: 10.1124/mol.117.111039
This article has not been copyedited and formatted. The final version may differ from this version.
MOL #111039
ꢀ
In addition to the separation of the 4-OH-RA enantiomers, the separation of 16-OH-13-cis-RA from 4-
OH-13-cis-RA and 16-OH-9-cis-RA from 4-OH-9-cis-RA was much better using the chiral
chromatography and the Chiracel OD-RH Column (Supplemental Fig.4-6) than with a Zorbax C18 column
(
Fig. 4E and 4F), allowing confirmation of the formation of these metabolites and collection of good quality
MS/MS spectra of the metabolites of 13-cis-RA and 9-cis-RA (Supplemental Fig.5 and 6). The metabolites
identified included the hydroxylations at C4 and C16 positions and sequential formation of the 4-oxo-
metabolites.
Enzyme kinetic parameters of RA isomers and 4-oxo-atRA as substrates of CYP26C1. The kcat values
m
with CYP26C1 for all RA isomers were similar, while K values of atRA and 13-cis-RA were 5.1- and 1.6-
fold higher than that of 9-cis-RA, respectively (Table 1). The Clint value for 9-cis-RA was 4.2 and 2.4-fold
higher than that for atRA and 13-cis-RA, respectively. This suggests that CYP26C1 prefers 9-cis-RA as a
substrate over atRA and 13-cis-RA (Fig.4L -4N; Table 1). The relatively robust formation of the 16-OH-
metabolite from 9-cis-RA allowed the determination of the K
0.3 nM; mean ± S.D., n=3; Supplemental Fig.1B) as well as for 4-OH-9-cis-RA formation, and the K
values for the two metabolites were similar (Table 1). In the case of 4-oxo-atRA as a substrate of CYP26C1,
similar K values were obtained by measuring product formation and substrate depletion (p > 0.01; Table
; Fig.6E and 6F). These K values were similar to those of 9-cis-RA, but significantly lower than those of
m
value towards this metabolite formation (5.9
±
m
m
1
m
atRA and 13-cis-RA (p < 0.01; Table1), suggesting that 4-oxo-atRA has higher affinity to CYP26C1 than
atRA and 13-cis-RA, and CYP26C1 may prefer 4-oxo-atRA as a substrate. Because of the lack of synthetic
4
-oxo-16-OH-atRA standard, the kcat for product formation was not measured but the efficiency of 4-oxo-
atRA metabolism was determined via substrate depletion (Fig.6F). Of all the substrates, 9-cis-RA had the
highest Clint by CYP26C1 which was significantly (3.3- fold, p < 0.01; Table 1) greater than that of 4-oxo-
atRA, while the Clint of atRA, 13-cis-RA, and 4-oxo-atRA were not significantly different from each other
(
p > 0.01; Table 1).
Effect of CRABPs on metabolism by CYP26C1. Since retinoids are extensively bound to CRABPs in
cellular environment, the delivery of CRABP bound substrates to metabolic enzymes has often been used
1
7ꢀ
ꢀ

Molecular Pharmacology Fast Forward. Published on February 23, 2018 as DOI: 10.1124/mol.117.111039
This article has not been copyedited and formatted. The final version may differ from this version.
MOL #111039
ꢀ
as an indicator of the biological importance of a metabolic enzyme in retinoid homeostasis. In addition, the
delivery of atRA for metabolism by CYPs has been shown to be enzyme dependent (Nelson et al., 2016).
For example, binding of atRA to CRABP-I and CRABP-II abolished the metabolism of atRA by CYP3A4
and CYP2C8 (Nelson et al., 2016) as predicted by the free drug hypothesis, i.e. that only drug free in
solution and not protein bound is available for metabolism. However, CRABP-I and CRABP-II delivered
atRA for metabolism by CYP26B1 (Nelson et al., 2016). On the other hand, atRA binding to albumin
resulted in reduced metabolism of atRA by CYP26B1, as predicted by free drug hypothesis and by albumin
acting as a binding sink of atRA (Nelson et al., 2016). To investigate whether CYP26C1 takes any of its
substrates from CRABP-I or CRABP-II, and interacts with the binding proteins, CYP26C1 was incubated
with CRABP-bound atRA, 4-oxo-atRA and 9-cis-RA (CRABP-ligand ratio of 1:1; Fig.8A). In the case of
atRA, when compared to the no CRABP incubations, 4-OH-atRA formation was decreased by 60-70% in
the presence of CRABP-I or CRABP-II (Fig.8A). However, no change of 4-OH-9-cis-RA formation by
CYP26C1 was observed in the presence of CRABPs when 9-cis-RA was the substrate (Fig.8A). With 4-
oxo-atRA as a substrate, 4-oxo-atRA metabolism by CYP26C1 was about 75% and 45% lower in the
presence of CRABP-I and CRABP-II, respectively, in comparison to the incubations with 4-oxo-atRA as a
substrate in the absence of the CRABPs (Fig.8A).
To further investigate the interactions between CRABPs and CYP26C1 with different CYP26C1
substrates, the kinetics of metabolite formation were characterized using CRABP bound substrates. In
kinetic studies with CYP26C1 (Fig.8B; Table 1), the kcat of atRA was 64% lower in the presence of CRABP-
m
I and CRABP-II compared with free atRA but this difference was not statistically significant. The K value
of CRABPI-atRA and CRABPII-atRA remained unchanged compared with that of free atRA (p > 0.01;
Table 1). As atRA binds to CRABPs tightly, virtually no unbound atRA is expected to be present in the
incubations including CRABPs. Therefore, this data suggests that CRABPs deliver atRA to CYP26C1.
While there was a trend of decreased intrinsic clearance of atRA in the presence of CRABPs, the intrinsic
clearances were not significantly different (p > 0.01; Table 1) in the presence and absence of CRABPs.
CRABP-I and CRABP-II affected 4-oxo-atRA metabolism in a similar manner as atRA (Fig.8C; Table 1).
1
8ꢀ
ꢀ

Molecular Pharmacology Fast Forward. Published on February 23, 2018 as DOI: 10.1124/mol.117.111039
This article has not been copyedited and formatted. The final version may differ from this version.
MOL #111039
ꢀ
m
The K values for 4oxo-atRA obtained in the presence of CRABP-I and CRABP-II were similar to that
obtained in the absence of CRABPs (p > 0.01; Table 1). The estimated maximum peak area ratio of 4-oxo-
6OH-atRA to the internal standard obtained in the presence of CRABP-I and CRABP-II (Fig.8C), which
1
is an indicator of the kcat, was reduced about 80% and 60%, respectively, compared to that obtained in the
absence of CRABPs, suggesting a similar decrease in the intrinsic clearance of 4-oxo-16OH-atRA
formation. However, due to the lack of reference standard for this metabolite, the true kcat values could not
be quantified.
To study the effect of excess apo-CRABPs on product formation by CYP26C1, fixed concentration of
substrate was mixed with varying concentrations of CRABP-I and CRABP-II, with the ratio of CRABP to
substrate ranging from 0 to 4 (Fig.8D and 8E). In the case of atRA and 4-oxo-atRA, CYP26C1 activity
decreased as the ratio of CRABP to substrate increased, reaching a plateau after the ratio exceeded 1. This
suggests that apo-CRABPs do not compete with holo-CRABPs to interact with CYP26C1, but CRABPs
may decrease metabolism by CYP26C1 via allosteric modulation or via inhibition. Surprisingly, neither
CRABP-I nor CRABP-II had any effect on 9-cis-RA metabolism and there was no significant change in 4-
OH-9-cis-RA formation from 9-cis-RA by CYP26C1 as the concentration of CRABP-I or CRABP-II
increased from 0 to 400 nM. This further suggests that CRABP-CYP interactions are substrate dependent.
Binding spectra and IC50 values of talarozole and ketoconazole. The substrate specificity and
stereoselectivity of the product formation with CYP26C1 suggests that CYP26C1 has different ligand
interactions than other members of the CYP26 family. To explore the ligand specificity of CYP26C1, the
binding of the two azole inhibitors of CYP26A1 and CYP26B1, ketoconazole and talarozole was tested
with CYP26C1. The coordination of the triazole or imidazole nitrogen with the heme iron causes a high
spin to low spin shift of the iron, resulting in a type 2 binding spectrum. The binding of ketoconazole or
talarozole to microsomal CYP26C1 induced a type 2 binding spectrum with some atypical characteristics
with the minimum absorbance observed between 410 and 420 nm and the maximum absorbance between
4
20-430 nm (Fig.9A and 9B). This suggests that the azole nitrogen in these two inhibitors can coordinate
with the CYP26C1 heme. The binding was inhibitor concentration dependent and showed increased spectral
9ꢀ
1
ꢀ

Molecular Pharmacology Fast Forward. Published on February 23, 2018 as DOI: 10.1124/mol.117.111039
This article has not been copyedited and formatted. The final version may differ from this version.
MOL #111039
ꢀ
intensity with increasing inhibitor concentration. In agreement with the binding spectra, both ketoconazole
and talarozole inhibited CYP26C1 activity in an inhibitor concentration dependent manner (Fig.9C and
9
D). However, the IC50 values, 124 µM (26-594; 95% confidence intervals) with ketoconazole and 3.8 µM
(
1.8-8.1; 95% CI) with talarozole, were much higher than the IC50 values observed with CYP26A1 (<10nM
for talarozole and 0.55 µM for ketoconazole) and CYP26B1 (<10 nM for talarozole and 0.59 µM for
ketoconazole) (Thatcher et al., 2011; Diaz et al., 2016; Foti et al., 2016).
Homology Model of CYP26C1 and Molecular Docking of Retinoic Acid Isomers. The experimental
data of substrate selectivity, metabolic specificity, ligand binding and CRABP interactions suggest that
CYP26C1 has functional differences when compared to CYP26A1 and CYP26B1, and that substrates have
different orientations in the CYP26C1 active site when compared to other CYP26 enzymes. To investigate
the basis of substrate recognition by CYP26C1, a homology model of CYP26C1 was developed and refined
by energy minimization and MD simulations (Supplemental Tables 3 and 4 and Supplemental Figure 2).
RA isomers (atRA, 9-cis-RA, 13-cis-RA) and 4-oxo-atRA were then docked in silico to the CYP26C1
active site. Based on the docking simulations, atRA, 9-cis-RA and 13-cis-RA all positioned in a similar
fashion within the CYP26C1 active site, with the β–ionone ring closest to the heme iron. 4-oxo-atRA was
rotated about 90° relative to RA isomers in its preferred orientation within the active site (Fig.10). The
docking simulations could appropriately predict the sites of oxidation of all four CYP26C1 substrates.
According to the distance between the heme and carbons at position 4, 16 and 18 on the β–ionone ring, the
CYP26C1 model predicted that atRA, 9-cis-RA and 13-cis-RA are primarily oxidized at carbon 4 and 4-
oxo-atRA is mainly oxidized at carbon 16 (Table 2). At the pro-chiral carbon 4 position of atRA, the
hydrogen that would account for the formation of 4(R)-OH-atRA oriented towards the heme iron with a
closer distance (3.9 Å) compared with the hydrogen which, when abstracted, would lead to the formation
of 4(S)-OH-atRA (5.0 Å; Fig.10A). Thus, the model predicted that CYP26C1 preferentially forms 4(R)-
over 4(S)-OH-atRA from atRA. For 13-cis-RA, pro-S and pro-R hydrogen atoms at C4 position were 3.7
and 3.8 Å away from the heme iron, respectively, leading to the prediction of racemic 4-OH-13-cis-RA
formation by CYP26C1 (Fig.10B). Because the β–ionone rings of atRA and 9-cis-RA are flipped 180°
2
0ꢀ
ꢀ

Molecular Pharmacology Fast Forward. Published on February 23, 2018 as DOI: 10.1124/mol.117.111039
This article has not been copyedited and formatted. The final version may differ from this version.
MOL #111039
ꢀ
relative to each other, the formation of 4(S)-OH-9-cis-RA was more favorable than the formation of 4(R)-
OH-9-cis-RA (Fig.10C). The distance between the heme iron and the hydrogen atom at the pro-S and pro-
R position of 9-cis-RA were 3.1 and 4.6 Å, respectively.
Compared with the crystal structure of atRA-bound CYP120A1 (Kühnel et al., 2008), the CYP26C1
model revealed that a C-terminal loop (amino acid L484-L504) occludes the binding pocket occupied by
the nonatetraene chain of atRA bound to CYP120A1, forcing atRA to occupy a distinct binding pocket in
CYP26C1 (Fig.11). The amino acids proposed to interact with RA isomers in the active site are shown in
Figure 12 and Supplemental Table 5. Leu-221 and Phe-222 are predicted to interact with the carboxylic
acid tail of atRA, 9-cis-RA and 4-oxo-atRA, whereas Ser-120 is predicted to interact with that of 13-cis-
RA. Notably, based on the homology model, CYP26C1 active site lacks the positively charged amino acids
such as Arg within the active site to interact with the carboxylic acid moiety in atRA.
2
1ꢀ
ꢀ

Molecular Pharmacology Fast Forward. Published on February 23, 2018 as DOI: 10.1124/mol.117.111039
This article has not been copyedited and formatted. The final version may differ from this version.
MOL #111039
ꢀ
Discussion
To date, most studies related to the CYP26 family have focused on CYP26A1 and CYP26B1 which are
known to be important RA hydroxylases (Thatcher and Isoherranen, 2009; Topletz et al., 2012). The
-
/-
importance of CYP26C1 in retinoid clearance has been questioned as the Cyp26C1 mice do not show any
of the phenotypic malformations typical for knock-out models of retinoid metabolism (Uehara et al., 2007).
However, the tissue specific importance of the individual CYP26 enzymes in retinoid clearance has not
been defined, and depends on both the expression levels of the CYP26 enzymes and their catalytic activity
towards the specific substrates. As the sequence homology of CYP26C1 with CYP26A1 and CYP26B1 is
low (43-51%), we hypothesized that CYP26C1 differs functionally from CYP26A1 and CYP26B1. Our
data demonstrates that CYP26C1 has different ligand (inhibitor and substrate) specificity than CYP26A1
and CYP26B1, illustrated most strongly by the differences in the IC50 values for talarozole and ketoconazole
with CYP26 enzymes. While talarozole and ketoconazole binding to CYP26C1 active site induced a type
II binding spectrum with some atypical characteristics similar to CYP26A1 (Thatcher and Isoherranen,
2
009), talarozole and ketoconazole IC50 values were more than 100-fold higher with CYP26C1 than with
CYP26A1 and CYP26B1 (Thatcher et al., 2011; Diaz et al., 2016). Comparisons of IC50-values between
enzymes can be confounded by different substrates and by different substrate concentration/K ratios used,
and the inhibitory potency of competitive inhibitors can be underestimated if substrate is used at
concentrations approaching its K . Even still, the IC50 data obtained demonstrate clear differences in ligand
interactions within the CYP26 family, and CYP26C1 displays a clear lack of susceptibility to these
inhibitors. The different ligand interactions were also evident with retinoids. While atRA has a lower K
with CYP26A1 and CYP26B1 than 9-cis-RA, with CYP26C1 9-cis-RA exhibited the lowest K and highest
intrinsic clearance. In addition, the K of 9-cis-RA with CYP26C1 (K =9.7nM) was much lower than with
CYP26A1 (K =134nM; Thatcher et al., 2011) and CYP26B1 (K =555nM; Diaz et al., 2016). This is in
m
m
m
m
m
m
m
m
agreement with a previous study in which radio-labeled substrate competition assays showed that 9-cis-RA
was a better ligand for CYP26C1 than for the other CYP26s (Taimi et al., 2004; Helvig et al., 2011).
Interestingly, 9-cis-RA metabolism by CYP26C1 was unaffected by the CRABPs (Fig.8), possibly because
2
2ꢀ
ꢀ

Molecular Pharmacology Fast Forward. Published on February 23, 2018 as DOI: 10.1124/mol.117.111039
This article has not been copyedited and formatted. The final version may differ from this version.
MOL #111039
ꢀ
the binding affinity of 9-cis-RA to CRABPs is much weaker than that of atRA and 4-oxo-atRA (Fiorella et
al., 1993) resulting in higher free fraction of 9-cis-RA. These findings also indicate a high clearance of 9-
cis-RA in vivo which is unprotected by CRABPs. This may explain the lack of detection of 9-cis-RA in
common tissue samples. In human serum 9-cis-RA was detected at a very low level (0.08-0.1nM) only in
a subset of individuals (Arnold et al., 2012). Similarly, in mice 9-cis-RA was only detected in the pancreas
(
Kane et al., 2010; Kane, 2012). The results shown here suggest that the detection of 9-cis-RA may be
dependent on CYP26C1 expression, and further studies are needed to clarify the expression pattern of
CYP26C1 in humans, and the role of CYP26C1 in 9-cis-RA clearance. CYP26C1 mRNA has been detected
in human fetal liver and brain and in adult human brain, lung, liver, spleen and testis (Xi and Yang, 2008)
suggesting that CYP26C1 is expressed in these retinoid target tissues.
All the data presented here suggest that CYP26C1 plays a unique role in the clearance of the active atRA
metabolite 4-oxo-atRA as a 4-oxo-atRA hydroxylase, and that 4-oxo-atRA may be an important active
retinoid in humans. However, the microsomal enzyme(s) responsible for 4-oxo-atRA formation is still
unknown. Similar to CYP26A1 and CYP26B1 (Lutz et al., 2009; Topletz et al., 2012), CYP26C1 forms 4-
oxo-atRA from 4-OH-atRA but with minimal efficiency, supporting the prior report that CYP26 enzymes
are important in the elimination of bioactive retinoids, not in their formation (Lutz et al., 2009; Topletz et
al., 2012). Previously, 4-oxo-atRA has been shown to be a substrate of both CYP26A1 (K
Clint=91µl/min/pmol) and CYP26B1 (K =29nM; Clint=15µl/min/pmol) (Topletz et al., 2012). The K
values of 4-oxo-atRA with CYP26A1 and CYP26B1 were 4 and 1.8-fold higher than the K of 4-oxo-atRA
m
=63nM;
m
m
m
with CYP26C1, respectively, and the Clint of 4-oxo-atRA was 1.7 and 10-fold higher with CYP26C1 than
with CYP26A1 and CYP26B1, respectively. Most notably, CYP26C1 appeared to take CRABP-bound 4-
oxo-atRA as a substrate, similar to accepting CRABP-bound atRA as a substrate. This suggests that
CRABPs play a role in modulating 4-oxo-atRA clearance in the cellular environment. In contrast to our
findings, in rat testes microsomes CRABP binding eliminated 4-oxo-atRA metabolism completely (Fiorella
and Napoli, 1994), suggesting that the substrate delivery is CYP specific. This is consistent with the general
observation that retinoid exchange between cellular binding proteins and metabolic enzymes is unique to
2
3ꢀ
ꢀ

Molecular Pharmacology Fast Forward. Published on February 23, 2018 as DOI: 10.1124/mol.117.111039
This article has not been copyedited and formatted. The final version may differ from this version.
MOL #111039
ꢀ
specific binding protein-enzyme pairings (Napoli, 2017). Substrate delivery to CYP26C1 by CRABPs
together with the relative expression levels of CYP26 enzymes, likely determines the physiological
importance of CYP26C1.
The data shown here reveals important information of the CYP-CRABP interactions. Due to the high
d
binding affinity of 4-oxo-atRA and atRA with CRABPs (K <20nM) (Fiorella and Napoli, 1991; Fiorella et
al., 1993), 4-oxo-atRA and atRA are expected to be bound with CRABPs in the incubations and in cells
with virtually no free retinoids in solution. This lack of free atRA or 4-oxo-atRA in the holo-CRABP
containing incubations was confirmed via back-purifying the holo-CRABPs prior to incubations. If
CRABPs acted as a sink of retinoids and did not deliver the retinoids for metabolism (i.e. according to free
drug hypothesis), the K
the presence of albumin (Nelson et al., 2016). In contrast, the K
CRABP were similar to those of free atRA and 4-oxo-atRA, respectively, with a trend towards lower kcat
with CRABP bound substrates. These findings are similar to decreased kcat or Vmax and K for CRABP-
m
for retinoid metabolism should be increased in these incubations as observed in
m
values of atRA and 4-oxo-atRA bound to
m
bound atRA reported using rat testis microsomes and recombinant CYP26B1 (Fiorella and Napoli, 1991;
Nelson et al., 2016). The data with CYP26C1 supports the notion that CRABPs deliver active retinoids for
metabolism via substrate channeling, and strongly suggest that CRABPs deliver both atRA and 4-oxo-atRA
to CYP26C1 for metabolism. In contrast, the lack of effect of CRABPs on 9-cis-RA metabolism suggests
that CRABPs do not modulate 9-cis-RA clearance. The lack of effect of CRABPs on 9-cis-RA metabolism
m
can be explained by the low binding affinity of 9-cis-RA with CRABPs and the low K of 9-cis-RA with
CYP26C1.
The differences in substrate specificity, inhibition profiles and stereo-specificity of metabolite formation
by CYP26C1, all suggest that the retinoids have different binding orientations within CYP26C1 active site
when compared to CYP26A1 and CYP26B1. Specifically, CYP26C1 formed 4-OH-9-cis-RA as the major
metabolite of 9-cis-RA and the formation of 16-OH-9-cis-RA appeared very minor, while CYP26A1 made
these metabolites at equal levels (Thatcher et al., 2011). Similarly, CYP26C1 formed predominantly the
4
(R)-OH-atRA while CYP26A1 formed the 4(S)-OH-atRA, suggesting different orientation of atRA within
4ꢀ
2
ꢀ

Molecular Pharmacology Fast Forward. Published on February 23, 2018 as DOI: 10.1124/mol.117.111039
This article has not been copyedited and formatted. The final version may differ from this version.
MOL #111039
ꢀ
the CYP26A1 and CYP26C1 active sites. To explore the binding pocket of CYP26C1, a homology model
of CYP26C1 was built based on a complement of CYP crystal structures. Previously, CYP26A1 and
CYP26B1 homology models were constructed mainly based on the CYP120A1 crystal structure (Karlsson
et al., 2008; Kühnel et al., 2008; Shimshoni et al., 2012; Foti et al., 2016). The predicted orientation of
atRA in the active site in these homology models (Shimshoni et al., 2012; Foti et al., 2016) was similar to
that observed in the crystal structure of CYP120A1 (Kühnel et al., 2008), with the carboxylic acid group
of atRA interacting with Arg64, Arg86 and Arg90 of CYP26A1 or Trp65, Arg76, Tyr372, and Arg373 of
CYP26B1. However, the CYP26C1 homology model revealed that a hairpin loop located at the C terminus
protrudes ~5Å deeper into the active site than the homologous loop in CYP120A1, which forces atRA and
other docked substrates to orient differently, with carboxylic acid interacting with Leu221 and Phe222
(
atRA and 9-cis-RA) or Ser120 (13-cis-RA). This orientation was predicted to result in the different
stereospecificity of the C4-hydroxylation of atRA. The overall active site volume of CYP26C1 was
3
predicted to be approximately 200-250 Å , much smaller than the predicted active site volume of CYP26A1
3
3
(
918 Å ) and CYP26B1 (977 Å ) (Foti et al., 2016). This indicates that the CYP26C1 active site may offer
fewer potential substrate binding orientations than the other two CYP26 enzymes, an observation consistent
with the much less diverse metabolite profile observed with CYP26C1 than with CYP26A1 and CYP26B1
(
Thatcher et al., 2011; Topletz et al., 2012). The small active site volume may also explain the high IC50’s
of talarozole and ketoconazole with CYP26C1. Even bearing in mind the inherent limitations of homology
modeling, the distinct structural properties of the CYP26C1 active site predicted by the structural model
are consistent with the experimental findings presented here, and successfully captured the regio- and
stereo-specificity of the enzyme, providing insights to the CYP26C1 active site.
2
5ꢀ
ꢀ

Molecular Pharmacology Fast Forward. Published on February 23, 2018 as DOI: 10.1124/mol.117.111039
This article has not been copyedited and formatted. The final version may differ from this version.
MOL #111039
ꢀ
Acknowledgements
The authors wish to thank Dr. Jay Kirkwood for his skillful assistance on LC-MS/MS experiments and
Brian Buttrick for his help in cloning of the CYP26C1.
2
6ꢀ
ꢀ

Molecular Pharmacology Fast Forward. Published on February 23, 2018 as DOI: 10.1124/mol.117.111039
This article has not been copyedited and formatted. The final version may differ from this version.
MOL #111039
ꢀ
Authorship Contributions
Participated in research design: Zhong, Ortiz, Nath, Isoherranen
Conducted experiments: Zhong, Ortiz
Contributed new reagents or analytical tools: Zelter
Wrote or contributed to the writing of the manuscript: Zhong, Ortiz, Zelter, Nath, Isoherranen
2
7ꢀ
ꢀ

Molecular Pharmacology Fast Forward. Published on February 23, 2018 as DOI: 10.1124/mol.117.111039
This article has not been copyedited and formatted. The final version may differ from this version.
MOL #111039
ꢀ
References
Allenby G, Bocquel MT, Saunders M, Kazmer S, Speck J, Rosenberger M, Lovey A, Kastner P, Grippo
JF, and Chambon P (1993) Retinoic acid receptors and retinoid X receptors: interactions with
endogenous retinoic acids. Proc Natl Acad Sci U S A 90:30–4.
Allenby G, Janocha R, Kazmer S, Speck J, Grippo JF, and Levin AA (1994) Binding of 9-cis-retinoic
acid and all-trans-retinoic acid to retinoic acid receptors alpha, beta, and gamma. Retinoic acid
receptor gamma binds all-trans-retinoic acid preferentially over 9-cis-retinoic acid. J Biol Chem
2
69:16689–95.
Arnold SLM, Amory JK, Walsh TJ, and Isoherranen N (2012) A sensitive and specific method for
measurement of multiple retinoids in human serum with UHPLC-MS/MS. J Lipid Res 53:587–598.
Åström A, Pettersson U, Krust A, Chambon P, and Voorhees JJ (1990) Retinoic acid and synthetic
analogs differentially activate retinoic acid receptor dependent transcription. Biochem Biophys Res
Commun 173:339–345.
Autenrieth F, Tajkhorshid E, Baudry J, and Luthey-Schulten Z (2004) Classical force field parameters for
the heme prosthetic group of cytochrome c. J Comput Chem 25:1613–1622.
Bankson DD, Russell RM, and Sadowski JA (1986) Determination of retinyl esters and retinol in serum
or plasma by normal-phase liquid chromatography: Method and applications. Clin Chem 32:35–40.
Chambon P (1996) A decade of molecular biology of retinoic acid receptors. FASEB J 10:940–54.
Diaz P, Huang W, Keyari CM, Buttrick B, Price L, Guilloteau N, Tripathy S, Sperandio VG, Fronczek
FR, Astruc-Diaz F, and Isoherranen N (2016) Development and Characterization of Novel and
Selective Inhibitors of Cytochrome P450 CYP26A1, the Human Liver Retinoic Acid Hydroxylase. J
Med Chem 59:2579–2595.
Durrant JD, Votapka L, Sørensen J, and Amaro RE (2014) POVME 2.0: An enhanced tool for
determining pocket shape and volume characteristics. J Chem Theory Comput 10:5047–5056.
Eisenberg D, Lüthy R, and Bowie JU (1997) VERIFY3D: Assessment of protein models with three-
dimensional profiles. Methods Enzymol 277:396–406.
2
8ꢀ
ꢀ

Molecular Pharmacology Fast Forward. Published on February 23, 2018 as DOI: 10.1124/mol.117.111039
This article has not been copyedited and formatted. The final version may differ from this version.
MOL #111039
ꢀ
Fiorella PD, Giguere V, and Napoli JL (1993) Expression of Cellular Retinoic Acid-binding Protein
(
Type II) in Escherichia coli CHARACTERIZATION AND COMPARISON TO CELLULAR
RETINOIC ACID-BINDING PROTEIN (TYPE I)*. J Biol Chem 268:21545–21552.
Fiorella PD, and Napoli JL (1991) Expression of cellular retinoic acid binding protein (CRABP) in
Escherichia coli. Characterization and evidence that holo-CRABP is a substrate in retinoic acid
metabolism. J Biol Chem 266:16572–16579.
Fiorella PD, and Napoli JL (1994) Microsomal retinoic acid metabolism. Effects of cellular retinoic acid-
binding protein (type I) and C18-hydroxylation as an initial step. J Biol Chem 269:10538–44.
Foti RS, Diaz P, and Douguet D (2016) Comparison of the ligand binding site of CYP2C8 with
CYP26A1 and CYP26B1: a structural basis for the identification of new inhibitors of the retinoic
acid hydroxylases. J Enzyme Inhib Med Chem 31:148–161.
Foti RS, Isoherranen N, Zelter A, Dickmann LJ, Buttrick BR, Diaz P, and Douguet D (2016)
Identification of Tazarotenic Acid as the First Xenobiotic Substrate of Human Retinoic Acid
Hydroxylase CYP26A1 and CYP26B1. J Pharmacol Exp Ther 357:281–292.
Helvig C, Taimi M, Cameron D, Jones G, and Petkovich M (2011) Functional properties and substrate
characterization of human CYP26A1, CYP26B1, and CYP26C1 expressed by recombinant
baculovirus in insect cells. J Pharmacol Toxicol Methods 64:258–263, .
Herrmann K (1995) Teratogenic effects of retinoic acid and related substances on the early development
of the zebrafish (Brachydanio rerio) as assessed by a novel scoring system. Toxicol Vitr 9:267–283.
Heyman RA, Mangelsdorf DJ, Dyck JA, Stein RB, Eichele G, Evans RM, and Thaller C (1992) 9-cis
retinoic acid is a high affinity ligand for the retinoid X receptor. Cell 68:397–406.
Idres N, Marill J, Flexor MA, and Chabot GG (2002) Activation of retinoic acid receptor-dependent
transcription by all-trans-retinoic acid metabolites and isomers. J Biol Chem 277:31491–31498.
Kane MA (2012) Analysis, occurrence, and function of 9-cis-retinoic acid. Biochim Biophys Acta - Mol
Cell Biol Lipids 1821:10–20.
Kane MA, Folias AE, Pingitore A, Perri M, Obrochta KM, Krois CR, Cione E, Ryu JY, Napoli JL, and
2
9ꢀ
ꢀ

Molecular Pharmacology Fast Forward. Published on February 23, 2018 as DOI: 10.1124/mol.117.111039
This article has not been copyedited and formatted. The final version may differ from this version.
MOL #111039
ꢀ
Russell DW (2010) Identification of 9-cis-retinoic acid as a pancreas- specific autacoid that
attenuates glucose-stimulated insulin secretion. PNAS 107:21884–21889.
Karlsson M, Strid Å, Sirsjö A, and Eriksson LA (2008) Homology models and molecular modeling of
human retinoic acid metabolizing enzymes cytochrome P450 26A1 (CYP26A1) and P450 26B1
(
CYP26B1). J Chem Theory Comput 4:1021–1027.
Kedishvili NY (2016) Retinoic acid synthesis and degradation. Subcell Biochem 81:127–161.
Kraft JC, and Juchau MR (1993) 9-cis-retinoic acid: A direct-acting dysmorphogen. Biochem Pharmacol
4
6:709–716.
Kühnel K, Ke N, Cryle MJ, Sligar SG, Schuler MA, and Schlichting I (2008) Crystal structures of
substrate-free and retinoic acid-bound cyanobacterial cytochrome P450 CYP120A1. Biochemistry
4
7:6552–6559.
Lovell SC, Davis IW, Arendall WB, de Bakker PIW, Word JM, Prisant MG, Richardson JS, and
Richardson DC (2003) Structure validation by Calpha geometry: phi,psi and Cbeta deviation.
Proteins 50:437–450.
Lutz JD, Dixit V, Yeung CK, Dickmann LJ, Zelter A, Thatcher JE, Nelson WL, and Isoherranen N (2009)
Expression and functional characterization of cytochrome P450 26A1, a retinoic acid hydroxylase.
Biochem Pharmacol 77:258–68.
Maqbool A, Graham-Maar RC, Schall JI, Zemel BS, and Stallings VA (2008) Vitamin A intake and
elevated serum retinol levels in children and young adults with cystic fibrosis. J Cyst Fibros 7:137–
1
41.
Mast N, Charvet C, Pikuleva IA, and Stout CD (2010) Structural basis of drug binding to CYP46A1, an
enzyme that controls cholesterol turnover in the brain. J Biol Chem 285:31783–31795.
Monk BC, Tomasiak TM, Keniya M V., Huschmann FU, Tyndall JDA, O’Connell JD, Cannon RD,
McDonald JG, Rodriguez A, Finer-Moore JS, and Stroud RM (2014) Architecture of a single
membrane spanning cytochrome P450 suggests constraints that orient the catalytic domain relative
to a bilayer. Proc Natl Acad Sci 111:3865–3870.
3
0ꢀ
ꢀ