Oxidative deamination of emixustat by human vascular adhesion protein-1/semicarbazide-sensitive amine oxidase

Article abstract of DOI:10.1124/dmd.118.085811

Emixustat potently inhibits the visual cycle isomerase retinal pigment epithelium protein 65 (RPE65) to reduce the accumulation of toxic bisretinoid by-products that lead to various retinopathies. Orally administered emixustat is cleared rapidly from the

Full text of DOI:10.1124/dmd.118.085811

DMD Fast Forward. Published on February 20, 2019 as DOI: 10.1124/dmd.118.085811
This article has not been copyedited and formatted. The final version may differ from this version.
Oxidative Deamination of Emixustat by Human Vascular
Adhesion Protein-1/Semicarbazide-sensitive Amine Oxidase
Michael J Reid, Russell Eyre, and Terry Podoll
Acucela Inc., Seattle, WA (MJR)
MavuPharma, Kirkland, WA (RE)
IV-PO, LLC, Seattle, WA (TP)

DMD Fast Forward. Published on February 20, 2019 as DOI: 10.1124/dmd.118.085811
This article has not been copyedited and formatted. The final version may differ from this version.
DMD # 85811
a) Emixustat oxidative deamination by human VAP-1 (SSAO).
b) Michael Reid, PhD, 818 Stewart Street, Seattle, WA, 98101, 425-308-1456,
mjr33_33@yahoo.com.
c) Number of text pages: 27
Number of tables: 4
Number of figures: 7
Number of references: 44
Number of words in the Abstract: 221
Number of words in the Introduction: 718
Number of words in the Discussion: 1737
d) List of nonstandard abbreviations (alphabetical order): AMD = age-related macular
degeneration, AOC = amine oxidase copper-containing, A2E = N-retinylidene-N-
retinylethanolamine, BA = benzylamine, BAL = benzaldehyde, BALDNP =
benzaldehyde dinitrophenylhydrazone, CPQ = carboxyprimaquine, DEABAL = 4-
diethylamino-benzaldehyde, DNPH = 2,4-dinitrophenylhydrazine, hAM = human aorta
membranes, hLM = human liver microsomes, hP = human plasma, hUCM = human
umbilical cord membranes, LOX = lysyl oxidase, MPA = mobile phase A, PQ =
primaquine, pTALDNP = p-tolualdehyde dinitrophenylhydrazone, RPE65 = retinal
pigmented epithelium protein 65, SSAO = semicarbazide-sensitive amine oxidase, VAP-
1
= vascular adhesion protein-1
-
2 -

DMD Fast Forward. Published on February 20, 2019 as DOI: 10.1124/dmd.118.085811
This article has not been copyedited and formatted. The final version may differ from this version.
DMD # 85811
Abstract
Emixustat potently inhibits the visual cycle isomerase RPE65 to reduce the accumulation
of toxic bisretinoid by-products that lead to various retinopathies. Orally administered
emixustat is cleared rapidly from the plasma, with little excreted unchanged. The
hydroxypropylamine moeity that is critical in emixustat’s inhibition of RPE65 is
oxidatively deaminated to three major carboxylic acid metabolites that appear rapidly in
plasma. These metabolites greatly exceed the plasma concentrations of emixustat and
demonstrate formation-rate limited metabolite kinetics. This study investigated in vitro
deamination of emixustat in human vascular membrane fractions, plasma and
recombinant human vascular adhesion protein-1 (VAP-1); demonstrating single enzyme
kinetics for the formation of a stable aldehyde intermediate (ACU-5201) in all in vitro
systems. The in vitro systems employed herein established sequential formation of the
major metabolites with addition of assay components for aldehyde dehydrogenase
(
ALDH) and cytochrome P450 (CYP). Reaction phenotyping experiments using selective
chemical inhibitors and recombinant enzymes of monoamine oxidase (MAO), VAP-1
and lysyl oxidase (LOX), showed that only VAP-1 deaminated emixustat. In
individually-derived human vascular membranes from umbilical cord and aorta, rates of
emixustat deamination were highly correlated to VAP-1 marker substrate activity
(
benzylamine) and VAP-1 levels measured by ELISA. In donor-matched plasma samples,
soluble VAP-1 activity and levels were lower than in aorta membranes. A variety of
potential co-medications did not strongly inhibit emixustat deamination in vitro.
-
3 -

DMD Fast Forward. Published on February 20, 2019 as DOI: 10.1124/dmd.118.085811
This article has not been copyedited and formatted. The final version may differ from this version.
DMD # 85811
Introduction
Emixustat (CAS number 1141777-14-1, ACU-4429) is a potent orally administered
small-molecule inhibitor of retinal pigment epithelium 65 (RPE65) under clinical
development by Acucela Inc. for the potential treatment of various retinopathies
including Stargardt disease and proliferative diabetic retinopathy. RPE65 is a key
isomerohydrolase enzyme in the pathway that regenerates the visual chromophore (11-cis
retinal) in vertebrate retinal tissue. Emixustat decreases the levels of vitamin A
derivatives (11-cis- and all-trans-retinal) and the accumulation of toxic lipid-retinoid by-
products (e.g. A2E and lipofuscin) that result in dysfunction and eventual death of the
specialized photoreceptor-containing cells of the retina (Bavik et al., 2015). While
important advances have been achieved in treatment of the wet form of age-related
macular degeneration (AMD; Gower et al., 2016) using injectable vascular endothelial
growth factor inhibitors, oral administration for retinal diseases remains an attractive
objective given the infection risks and frequent office visits associated with intravitreal
injection (Awh, 2016).
Emixustat was designed to treat the retina following oral administration by
focusing on potent and durable pharmacologic activity against RPE65 while displaying
rapid systemic elimination. Preclinical demonstration of these traits translated well to
clinical application in a Phase 1 study wherein dose-dependent effects were measured in
the retina more than 24 hours after dose, at a time when emixustat levels were very low
due to its rapid elimination (4 – 6 hour half-life in plasma; Kubota et al., 2012).
Emixustat absorption was also rapid with Tmax values ranging from 3 – 5 hours while no
accumulation was observed upon repeated-dose administration (Kubota et al., 2014). A
1
4
human ADME study characterized three major (ie, ≥10% of total systemic C exposure)
circulating metabolites, ACU-5116, ACU 5124, and ACU-5149, accounting for 11.5%,
29.0%, and 10.6% of the total area under the plasma drug concentration-time curve
-
4 -

DMD Fast Forward. Published on February 20, 2019 as DOI: 10.1124/dmd.118.085811
This article has not been copyedited and formatted. The final version may differ from this version.
DMD # 85811
(
AUC) of circulating radioactivity (Fitzsimmons et al., 2018). In this study >90%
emixustat dose-related radioactive material was excreted via the urine as metabolites,
similar to nonclinical species (data on file).
Preclinical in vivo metabolism of emixustat was very similar across rat, dog, and
monkey (Podoll et al., 2013). The three major metabolites are carboxylic acids that are
products of oxidative deamination of the primary amine of emixustat (Fitzsimmons et al.,
2018). Two of the three metabolites, ACU-5124 and ACU-5149, were secondary
metabolites that had also been differentially hydroxylated in the cyclohexyl ring. In
human hepatocytes, primary hydroxylation to the cyclohexanols dominated the product
formation observed; averaging 26% of radioactivity, and only 4 – 9% observed as the
oxidatively deaminated carboxylic acids (Fitzsimmons et al., 2018). In human liver
microsomes no oxidative deamination was observed and NADPH-dependent turnover of
[14
C]emixustat was low, yet CYP2D6 and CYP3A4 were identified as capable of
catalyzing the hydroxylation of emixustat in vitro when using the cDNA-expressed CYP
isoforms (data on file). Other extra-hepatic, human in vitro systems were then used to
characterize the predominant oxidative deamination reaction observed in the human
ADME clinical trial.
Initial studies focused on monoamine oxidases (MAO), yet emixustat was not a
substrate for MAO (data on file), so alternative methods were sought to characterize its
oxidative deamination. The first evidence of vascular adhesion protein-1/semicarbazide-
sensitive amine oxidase (VAP-1/SSAO from here forward referred to as VAP-1)
involvement in the metabolism of emixustat was revealed during cross-species plasma
protein binding study (Fitzsimmons et al., 2018; with additional nonclinical species data
on file). Plasma stability assessments demonstrated inter-species differences that were
consistent with the highest soluble plasma deamination activity observed in pig and dog
(
Schwelberger 2007). Prior studies have used human umbilical artery microsomes
-
5 -

DMD Fast Forward. Published on February 20, 2019 as DOI: 10.1124/dmd.118.085811
This article has not been copyedited and formatted. The final version may differ from this version.
DMD # 85811
(
(
(
hUCM) to demonstrate the involvement of VAP-1 in the metabolism of tresperimus
Claud et al., 2002). Since then potent, selective VAP-1 inhibitors have been developed
Salter-Cid et al., 2005) and recombinant VAP-1 has become commercially available.
The results of human in vitro reaction phenotyping for the oxidative deamination of
emixustat are described here. Since emixustat was targeting retinopathies that are more
prevalent in adults aged 65 years and above (e.g. AMD) and this population can have
inflammatory diseases impacting VAP-1 (Salmi et. al., 1993; Pannekoek et al., 2015),
human aorta tissue was obtained from elderly donors to characterize emixustat oxidative
deamination in this patient population. Once established, the in vitro model was used to
assess the impact of potential VAP-1 inhibitors on emixustat and the impact of emixustat
on the VAP-1 catalyzed oxidative deamination of model substrate, benzylamine (BA).
-
6 -

DMD Fast Forward. Published on February 20, 2019 as DOI: 10.1124/dmd.118.085811
This article has not been copyedited and formatted. The final version may differ from this version.
DMD # 85811
Materials and Methods
Materials. Emixustat hydrochloride (99.7% chemical purity) was manufactured by IRIX
Pharmaceuticals (Florence, SC). ACU-4861 [(1R,4r)-4-((3-((R)-3-amino-1-
hydroxypropyl)phenoxy)methyl)cyclohexan-1-ol], ACU-5116 [(R)-3-(3-
(
(
cyclohexylmethoxy)phenyl)-3-hydroxypropanoic acid], ACU-5124 [(R)-3-hydroxy-3-
3-(((1r,4R)-4-hydroxycyclohexyl)methoxy)phenyl)propanoic acid], ACU-5141 (d11-
ACU-5116), ACU-5142 (d3-ACU-5124) and ACU-5201 [(E)-3-(3-
cyclohexylmethoxy)phenyl)acrylaldehyde] were prepared by Acucela (Bothell, WA) as
(
described previously (Fitzsimmons et al., 2018, supplemental). LJP-1207 [N′-(2-phenyl-
allyl)-hydrazine hydrochloride] was also prepared and characterised by Acucela
according to previously published methods (Salter-Cid et al., 2005). Acetic acid,
methanol (HPLC grade), phosphate buffered saline, sodium bicarbonate and water
(
HPLC grade) were obtained from Fisher Scientific USA (Pittsburg, PA). All other
chemicals and reagents were obtained from Sigma-Aldrich (St. Louis, MO).
Recombinant human VAP-1, the ELISA kit for human VAP-1, anti-human VAP-
1
antibody as well as recombinant human Lysyl Oxidase Homolog 2 (LOXL2) were
obtained from R&D Systems (Minneapolis, MN). Anti-mouse HRP antibody was
obtained from Promega (Madison, WI). Human liver microsomes (hLM; UltraPool™
HLM 150 Mixed Gender; lot # 38290) were obtained from BD Biosciences (Woburn,
MA). Human umbilical cord membranes (hUCM), and elderly human aorta membranes
(
hAM) with matching heparinized plasma samples from the elderly humans were
prepared under contract by BioIVT (Baltimore, MD). Twenty individual lots each of
hUCM, and hAM and 18 donor-matched (to hAM) individual plasma samples were
provided to assess inter-individual variability of VAP-1 activity. An 8-donor pool of
hUCM and a 20-donor pool of hAM were also provided to determine the kinetics of
VAP-1 activity. Aorta tissues were obtained from individuals that were at least 65 years
-
7 -

DMD Fast Forward. Published on February 20, 2019 as DOI: 10.1124/dmd.118.085811
This article has not been copyedited and formatted. The final version may differ from this version.
DMD # 85811
of age with the following demographics:11 male (M) and 9 female (F), 3 African-
America, 4 non-white Hispanic and 13 Caucasian with body mass index ranging from
16.9 to 43. Eight individuals were identified as having diabetes and 17 had cardiovascular
complications and/or hypertension.
Sequential Oxidative Metabolism of Emixustat in Human Tissue. Pooled hUCM,
NAD+, emixustat and its hydroxycyclohexyl metabolite (ACU-4861) were prepared in
10 mM sodium bicarbonate buffer, pH 7.4, and kept on wet ice until added to
incubations. Final incubation volumes of hUCM reactions were 0.2 ml at a final protein
concentration 0.1 mg/ml with final concentrations of substrates (100 μM emixustat or
+
ACU-4861), and the cofactor NAD (1 mM). Reactions were started by the addition of
+
hUCM immediately followed by the addition of NAD and stopped after 20 minutes with
the addition of 0.2 ml of acetonitrile (ACN), vortexed briefly and centrifuged at 13,000 x
g for 10 minutes to precipitate proteins. Supernatants were filtered through 0.2 µm
polypropylene filter membranes (Captiva, Agilent) and loaded onto 96-well plates.
Human liver microsomes (hLM), NADPH, and the deaminated carboxylic acid
metabolite (ACU-5116) were prepared in 10 mM potassium phosphate buffer, pH 7.4,
and kept on wet ice until added to incubations. Final incubation volumes of hLM
reactions were 0.2 ml, final microsomal protein concentrations were 1.0 mg/ml, final
ACU-5116 concentrations were 100 µM and final concentrations of NADPH were 1 mM.
Reactions were started by the addition of NADPH to hLM incubations preincubated at
37°C for 3 minutes. Reactions were stopped after 20 minutes with the addition of 0.1 ml
of ACN, vortexed briefly and centrifuged at 13,000 x g for 10 minutes to precipitate
proteins. Supernatants were filtered through 0.2 µm polypropylene filter membranes
(
Captiva, Agilent) and loaded onto 96-well plates.
In Vitro Metabolism of Emixustat in Human Sources of VAP-1. In vitro metabolism
of emixustat by human sources of VAP-1 was determined by measuring the amount of its
-
8 -

DMD Fast Forward. Published on February 20, 2019 as DOI: 10.1124/dmd.118.085811
This article has not been copyedited and formatted. The final version may differ from this version.
DMD # 85811
stable aldehyde product (ACU-5201) formed per minute per mg protein (or ml plasma) in
triplicate. Final incubation volumes were 0.1 ml and the amount of VAP-1 source and
incubation time were linearly optimized. Pooled and individual lots of hUCM or hAM
were prepared at a final protein concentration 0.1 mg/ml and recombinant human VAP-1
was prepared at a final protein concentration 0.01 mg/ml while human plasma (hP) was
assayed volumetrically (0.05 ml) in 10 mM sodium bicarbonate buffer, pH 7.4 (NaHCO
3
buffer). Stock solutions of emixustat and inhibitors were also prepared in NaHCO buffer
3
and were pre-incubated at 37°C together (when inhibition was tested) prior to adding the
enzyme source. Human sources of VAP-1 were kept on dry ice until thawed at 37°C prior
to diluting with NaHCO buffer. When possible, timed incubations (10 minutes for
3
hUCM, hAM and rhVAP-1 and 30 minutes for plasma) were started immediately by the
addition of the enzyme source and stopped by the addition of 0.2 ml of ACN fortified
with 10 ng/ml of 4-diethylamino-benzaldehyde (DEABAL as an internal standard),
vortexed briefly and centrifuged at 13,000 x g for 10 minutes to precipitate proteins.
During simultaneous incubations of tissue from several individuals the membrane
preparations were kept on wet ice until aliquoted and brought to 37°C before combination
with substrate. Emixustat concentrations ranged from 5.0 – 500 µM when determining
enzyme kinetic parameters in the various human sources of VAP-1 and between 60-160
µM when assessing interindividual activity and inhibition. Supernatants were filtered
through 0.2 µm PVDF membranes (Captiva, Agilent) and loaded onto 96-well plates for
quantification by LC-MS/MS.
LC-MS/MS Quantitation of Emixustat’s Primary In Vitro Aldehyde Metabolite
(
ACU-5201). Although ACU-5201 has only been observed in vitro (Fitzsimmons et al.,
018), the proposed intermediate aldol (Figure 1) has not been identified nor could it be
2
synthesized due to apparent instability during isolation and purification. The amount of
ACU-5201 formed in VAP-1 incubations was quantitated by LC-MS/MS. ACU-5201
-
9 -

DMD Fast Forward. Published on February 20, 2019 as DOI: 10.1124/dmd.118.085811
This article has not been copyedited and formatted. The final version may differ from this version.
DMD # 85811
standard was diluted in ACN and used for preparation of calibration standard stock
solutions that were added at 10% (v:v) to appropriate volumes of incubation reaction
media and acetonitrile to match the sample incubations.
Liquid chromatography (LC20AD; Shimadzu, Pleasanton, CA) separation was
achieved on a Poroshell 120 Phenyl Hexyl, 2.1 x 50 mm, 2.7 µm column (Agilent) using
a step gradient of 10 mM ammonium formate (mobile phase A: MPA) and ACN at 0.45
ml/min as follows: 0.25 min 60% MPA, 0.25 min linear gradient to 17% MPA, 0.75 min
17% MPA, 0.25 min linear gradient to 8% MPA, then 0.5 min at 8% MPA before
flushing with 100% ACN and re-equilibrating to 60% MPA. Column temperature was set
to 30°C while autosampler temperature was set to 10°C. Samples (20 µl) were injected
with an external rinse program using 100% ACN.
The mass spectrometer (API5000; SCIEX, Redwood City, CA) was operated in
the positive ion mode and MRM scan type. The first and third quadrupoles were set to
+
detect the positively charged molecular ions [M+H] and fragments of ACU-5201 (m/z
245.2→103.1 Da) and DEABAL (m/z 178.2→106.1 Da) with unit resolution in both
quadrupoles. Compound-dependent MS/MS parameters for quantitating ACU-5201 and
DEABAL were: Declustering Potential = 60.0 and 175.0 V; Entrance Potential = 10.0 V;
Collision Energy = 37.0 and 40.0 V; Exit Potential = 15.0 and 17.0 V, respectively.
Source and gas parameters were: Collision Gas = 8 psi, Curtain Gas = 11 psi, Nebulizer
Gas = 50 psi, Turbo Gas = 60 psi, ESI Voltage = 5500 V and Temperature = 675°C.
In Vitro Deamination of Benzylamine in Human Sources of VAP-1. Oxidative
deamination of benzylamine (BA) was employed as a biomarker of VAP-1/SSAO
enzymatic activity in human sources of VAP-1 in vitro. The rate of BA deamination was
determined by measuring the amount of benzaldehyde (BAL) formed per minute per mg
protein (or ml plasma) in triplicate (only duplicates for some plasma incubations). Final
incubation volumes were 0.1 ml and the amount of VAP-1 source and incubation time
-
10 -

DMD Fast Forward. Published on February 20, 2019 as DOI: 10.1124/dmd.118.085811
This article has not been copyedited and formatted. The final version may differ from this version.
DMD # 85811
were linearly optimized. Pooled and individual lots of hUCM or hAM were prepared at a
final protein concentration 0.1 mg/ml and recombinant human (rh) VAP-1 was prepared
at a final protein concentration 0.0025 mg/ml while plasma was assayed volumetrically
(
0.05 ml) in NaHCO buffer. Final reaction volumes were 0.1 ml and timed incubations
3
were started by the addition of the enzyme source at 37°C. When inhibition and
interindividual activity was tested, BA concentrations were 80 µM for membrane and
recombinant sources and 200 µM for plasma. Reactions were stopped after 10 minutes
for membrane and recombinant sources and after 30 minutes for plasma by the addition
of 0.2 ml of 60/40 ACN/MeOH fortified with 5 µM of the internal standard p-
tolualdehyde dinitrophenylhydrazone (pTALDNP), vortexed briefly and centrifuged at
13,000 x g for 10 minutes to precipitate proteins. Supernatants (0.25 ml) were filtered
through 0.2 µm PVDF membranes (Captiva, Agilent) and loaded onto 96-well plates with
each well containing 50 µl of 2,4-dinitrophenylhydrazine (DNPH; 3 mg/ml) solution.
Derivatizations were incubated at 40°C for 20 minutes and stopped by the addition of 10
µl of 25% formic acid, and then vortexed briefly prior to loading into the autosampler.
Initial incubations in hUCM did not incorporate the internal standard in a reaction stop
solution of acetonitrile with 2% formic acid and these derivatizations were not stopped
with formic acid. These method optimizations resulted in improved recovery of the
benzaldehyde dinitrophenylhydrazone (BALDNP) by > 50% across the calibration range
when compared to authentic BALDNP standard spiked into reaction mixtures (data not
shown), affording better sensitivity to assess inhibition.
LC-MS/MS Quantitation of Derivatized Benzaldehyde. The amount of derivatized
benzaldehyde (BALDNP) formed in VAP-1 incubations was quantitated by LC-MS/MS.
Benzaldehyde standard was diluted in methanol to 4 mM and used for preparation of
calibration standard and QC stock solutions. To match the sample incubations, stock
solutions of BAL were added at 10% (v:v) to appropriate volumes of incubation reaction
-
11 -

DMD Fast Forward. Published on February 20, 2019 as DOI: 10.1124/dmd.118.085811
This article has not been copyedited and formatted. The final version may differ from this version.
DMD # 85811
media and internal standard solution. Calibration standards and QCs were derivatized
along with incubation samples following centrifugation according to the procedure
outlined above.
Liquid chromatography separation was achieved on a Kinetix Biphenyl 2.5 µm, 2
x 30 mm column (Phenomenex) using a step gradient of 0.15 formic acid (MPA) and
6
1
3
0% ACN/40% MeOH (mobile phase B) at 0.8 ml/min as follows: 0.8 min 34% MPA,
.0 min linear gradient to 2% MPA, then 1.0 min 2% MPA before re-equilibrating to
4% MPA. Column temperature was set to 30 °C while autosampler temperature was set
to 6°C. Samples (20 µl) were injected with an external rinse program using 100% ACN.
The mass spectrometer was operated in the positive ion mode and MRM scan
type. The first and third quadrupoles were set to detect the positively charged molecular
ions [M+H]+ and fragments of BALDNP (m/z 287.2→104.1 Da) and pTALDNP (m/z
301.1→118.3 Da) with unit resolution in both quadrupoles. Compound-dependent
MS/MS parameters for quantitating BALDNP and pTALDNP were: Declustering
Potential = 70.0 V, Entrance Potential = 10.0 V, Collision Energy = 24.0 V and Exit
Potential = 15.0 V. Source and gas parameters were: Collision Gas = 10 psi, Curtain Gas
=
14 psi, Nebulizer Gas = 50 psi, Turbo Gas = 55 psi, ESI Voltage = 5200 V and
Temperature = 700°C.
Quantification of VAP-1 by ELISA. Vascular adhesion protein-1 levels in membranes
and plasma were quantified using the R&D Systems Quantikine ELISA for human VAP-
1. The ELISA was performed according to the instructions in the manual. The samples
were diluted 1:100 with Calibrator Diluent RD6X. The optical density for the plate was
read at 450 nm and 570 nm. To account for wavelength correction, the 570 nm value was
subtracted from the 450 nm value. The background value (value from well without any
sample, standard or control) was subtracted from all wells. VAP-1 concentrations (ng/ml)
for all samples determined from the calibration standards were normalized to total protein
-
12 -

DMD Fast Forward. Published on February 20, 2019 as DOI: 10.1124/dmd.118.085811
This article has not been copyedited and formatted. The final version may differ from this version.
DMD # 85811
concentrations determined by standard methods (Smith et al., 1985) into μg VAP-1/mg
total protein.
Clinical Study Analyses. Plasma concentrations of emixustat and its three major
metabolites were determined in a clinical pharmacology study titled: Pharmacokinetic
and Pharmacodynamic Study of Emixustat Hydrochloride in Subjects With Geographic
Atrophy Associated With Dry Age-Related Macular Degeneration (Dugel et al., 2015)
using a validated LC-MS/MS method (Podoll et al., 2018). The study was performed in
accordance with the ethical principles stated in the Declaration of Helsinki, FDA
regulation 21 CFR, Parts 50, 56, and 312, and International Conference on
Harmonization guidelines for good clinical practice. The protocol, advertisement, and
Informed Consent Form (ICF) were reviewed and approved by the Institutional Review
Board. Written informed consent for the study was obtained from all subjects before
protocol-specific procedures were carried out. Emixustat was administered orally at doses
ranging from 2 to 10 mg once daily with plasma samples collected to meet study
objectives.
Data Analysis. Linear regression analysis of calibration curve data and resulting
unknown sample concentration determination was performed with Analyst software
(
versions 1.5.1 through 1.6.2; AB Sciex, Redwood City, CA) using the ratio of product to
internal standard as the dependent variable. Mathematical conversion of product
formation data and basic statistical calculations were performed using Microsoft Excel
2
010. Michaelis-Menten and other enzyme kinetic data analysis was performed using
nonlinear hyperbolic one-site binding or competition fits on GraphPad Prism software
versions 4-7; La Jolla, CA). Log values from the ELISA standards (log ng/ml VAP-1)
(
were plotted against the corrected OD570 also using GraphPad Prism with the 4-
parameter logistic curve fit. Pearson’s and Spearman’s rank correlation analysis and
ANOVA of VAP-1 metabolic activities and VAP-1 levels was also performed using
-
13 -

DMD Fast Forward. Published on February 20, 2019 as DOI: 10.1124/dmd.118.085811
This article has not been copyedited and formatted. The final version may differ from this version.
DMD # 85811
GraphPad Prism software. Semi-log concentration-time plots for the clinical
pharmacology study were also constructed using GraphPad Prism software. All
noncompartmental pharmacokinetic (PK) analyses were performed using WinNonlin™
(
version 6.3, Certara L.P., Princeton, NJ).
-
14 -

DMD Fast Forward. Published on February 20, 2019 as DOI: 10.1124/dmd.118.085811
This article has not been copyedited and formatted. The final version may differ from this version.
DMD # 85811
Results
Oxidative Deamination of Emixustat. The three major circulating metabolites
1
4
observed following the dosing of [ C]emixustat during a human ADME clinical trial,
ACU-5116, ACU-5124, and ACU-5149, were all carboxylic acid products of oxidative
deamination of the primary amine. While the intermediate aldehydes were not observed
in vivo in human plasma, ACU-5201 was identified during in vitro human plasma
14
stability tests using [ C]emixustat (Fitzsimmons, et al., 2018). Similar to a previous
tresperimus report (Claud et al., 2001), exposure of emixustat was increased
approximately 2-fold in rats following prior administration of hydralazine and
semicarbazide, suggesting the rate-determining role of VAP-1 in the elimination of
emixustat (data on file). The metabolic scheme shown in Figure 1 was proposed to
initiate in vitro studies to characterize the enzymology relevant to the human in vivo
metabolism of emixustat. Range-finding incubations of emixustat in human umbilical
cord membranes (hUCM), an established source of VAP-1 activity, suggested the ACU-
5201 aldehyde was the predominant in vitro product. Further experiments were
conducted with emixustat to confirm that ACU-5201 formation kinetic parameters could
be accurately determined.
To monitor formation of the carboxylic acid metabolite, a quantitative method for
ACU-5116 using synthetic reference standard and deuterium-labeled internal standard
was established. Emixustat (50 µM; 80% maximal rate) was incubated at 37ºC with
hUCM at 0, 0.1, and 0.25 mg/ml total protein up to 16 minutes. While levels of ACU-
5116 were BLQ, ACU-5201 was formed linearly with both time and protein at
concentrations ranging from approximately 100 to 1000 ng/ml. Although several
oxidases may contribute to the rapid oxidation of the aldehyde intermediate in vivo, the
role of aldehyde dehydrogenase (ALDH) in the conversion to the carboxylic acid was
-
15 -

DMD Fast Forward. Published on February 20, 2019 as DOI: 10.1124/dmd.118.085811
This article has not been copyedited and formatted. The final version may differ from this version.
DMD # 85811
tested. Emixustat (16 µM; at K , see below) was added to pooled hUCM (1 mg/ml
M
+
protein) in the presence and absence of NAD . High levels of ACU-5116 were observed
+
in the presence of NAD and were nearly the molar equivalent to the decrease in the
+
formation of ACU-5201 when compared to the incubations without NAD (Figure 2A).
Experiments exploring the formation route(s) of the secondary metabolite, ACU-5124,
resulting from both oxidative deamination of the primary amine and hydroxylation of the
cyclohexyl ring of emixustat (Figure 1), were conducted with two primary metabolites of
emixustat. To maximize potential product formation, 100 µM ACU-5116 was incubated
with human liver microsomes (1 mg/ml protein) in the presence and absence of NADPH.
NADPH-dependent formation of ACU-5124 was observed, confirming that ACU-5116 is
a substrate for CYP enzymes resulting in the formation of ACU-5124. ACU-4861 (100
µM), a monohydroxylated metabolite previously shown to be formed by CYP enzymes in
human hepatocytes (Fitzsimmons et al., 2018) was also incubated in order to maximize
product formation with pooled hUCM (1 mg/ml protein) for 20 minutes in the presence
+
+
and absence of NAD . NAD -dependent formation of the carboxylic acid metabolite,
ACU-5124, was observed demonstrating that its formation could also proceed from initial
hydroxylation of emixustat, followed by oxidative deamination and aldehyde oxidation of
the monohydroxylated amine metabolite (Figure 2B). Given these results, vascular
membranes were used to generate kinetic parameters for the oxidative deamination of
emixustat via ACU-5201 formation rates as an in vitro model for assessing the variability
of VAP-1 reaction rates across a population of individually-derived vascular tissues and
for the assessment of emixustat as a potential victim of in vitro drug interactions driven
by inhibition of VAP-1.
Enzyme kinetics were assessed for emixustat deamination activity with the
linearly optimized protein concentration (or volume of plasma) and incubation time for
-
16 -

DMD Fast Forward. Published on February 20, 2019 as DOI: 10.1124/dmd.118.085811
This article has not been copyedited and formatted. The final version may differ from this version.
DMD # 85811
rhVAP-1, hUCM, human aorta membranes (hAM) and human plasma (hP). Michaelis-
Menten kinetic constants were determined by fitting the rate equation for a single-enzyme
using nonlinear regression; substrate saturation data are presented graphically in Figure 3.
Mean VAP-1 enzyme kinetic constants in rhVAP-1, and pooled hUCM, hAM, and hP are
presented in Table 1. The apparent Michaelis constant (K ) values for the formation of
M
ACU-5201 of 13, 14.7 and 15.6 µM in rhVAP-1, hUCM and hAM, respectively, are in
very close agreement indicating that the different human sources of the VAP-1 enzyme
had the same binding affinity for emixustat. These data confirmed that in vitro product
formation of the aldehyde metabolite was adequate to assess the oxidative deamination of
emixustat. Limited turnover of emixustat was observed in hP. The reaction rates did not
reach saturating conditions resulting in an overestimation of the apparent K
compared to other VAP-1 sources.
M
in hP when
In Vitro VAP-1 Oxidative Deamination Activity. To correlate the amine oxidase
activity in human vascular membrane preparations (hUCM and hAM) and plasma with
emixustat oxidative deamination rates, we adapted an assay for the commonly used
amine oxidase substrate, benzylamine (BA), which yields benzaldehyde (BAL). Although
BA is not a physiologic substrate for VAP-1, there are numerous assays based on the
oxidative deamination of BA to form BAL as a marker substrate for quantifying various
1
4
amine oxidase activities including: radiometric detection of extracted C-BAL (Lizcano
et al., 1998), HPLC with fluorometric detection of derivatized BAL (van Dijk et al.,
1995), UV monitoring of BAL (Heuts et al., 2011) and measurement of the H
2
O
2
produced using Amplex red (Holt et al., 1997). A common method used for the
measurement of aldehydes is an acid catalyzed derivatization with DNPH followed by
HPLC with ultraviolet or fluorometric detection (Claud et. al., 2002.). A direct injection
LC-MS/MS method of the DNPH-derivatized BAL (BALDNP) was developed to
-
17 -

DMD Fast Forward. Published on February 20, 2019 as DOI: 10.1124/dmd.118.085811
This article has not been copyedited and formatted. The final version may differ from this version.
DMD # 85811
measure the amount of BAL formed per minute per mg protein in in vitro incubations.
Following the initial experiments in hUCM, the LC-MS/MS method for BALDNP was
modified slightly to improve the sensitivity of the assay in order to better characterize
inhibition in latter experiments with hAM and hP. Efficiency and recovery of the 2,4,
DNPH derivatization was improved by increasing the pH of the reaction stop solvent
closer to 4.5 with 0.1 M ammonium acetate added to the organic solvent. These changes
improved the recovery BALDNP from ~30-40% across the calibration range to >85%
when compared to neat BALDNP spiked samples (data on file).
Enzyme kinetics were assessed for BA deamination activity with the linearly
optimized protein concentration and incubation time for hUCM, hAM and hP (plasma
volume). The Michaelis-Menten kinetic constants were determined by fitting the rate
equation for a single-enzyme using nonlinear regression. Mean VAP-1 enzyme kinetic
constants in rhVAP-1, and pooled hUCM, hAM and hP are presented in Table 2. The
apparent BA K
rhVAP-1 and pooled hUCM, hAM and hP, respectively. These K
3.5 ± 9.5 µM previously reported for BA deamination in human umbilical artery
M
values for the formation of BAL of 56.6, 78.7, 78.9 and 75.2 µM, in
M
values are similar to
6
microsomes (Claud et al., 2002) indicating that the different human sources of the VAP-1
enzyme had the same binding affinity for this marker substrate. Although the rates of
product formation in plasma are much lower than in membranes or recombinant sources,
the Vmax in pooled hP was 17.6 nmol/ml/hr which is comparable to other reported clinical
values for: diabetic vs normal subjects with means of 20.6 and 14.3 nmol/ml/hr,
respectively (Garpenstrand et al., 1999), and chronic liver disease vs healthy normal
subjects with means of 18.8 and 10.7 nmol/ml/hr, respectively (Kurkijärvi et al., 2000).
Reaction Phenotyping of VAP-1 Oxidative Deamination.
Correlation Analysis. The interindividual variability in VAP-1 activity observed across a
population of tissue donors was assessed in hUCM, hAM and hP acquired from
-
18 -

DMD Fast Forward. Published on February 20, 2019 as DOI: 10.1124/dmd.118.085811
This article has not been copyedited and formatted. The final version may differ from this version.
DMD # 85811
commercial sources. Saturating concentrations of emixustat and BA were incubated in
triplicate for the individual sources of VAP-1 at the linearly optimized conditions. Mean
rates of ACU-5201 formation ranged from 36.4 to 170, 5.82 to 100 and 0.748 to 7.51
pmol/min/mg protein in hUCM, hAM and hP, respectively. The 10-fold difference in
emixustat deamination in hP and 17-fold difference in hAM had more variability than the
4.7-fold difference in hUCM, however the mean rate of 20 individuals in both hUCM and
hAM were very similar at 44.7 and 49.4 pmol/min/mg protein, respectively. Mean rates
of BAL product formation ranged from 13.7 to 40.4, 4.16 to 36.6 and 0.0637 to 1.17
nmol/min/mg protein in hUCM, hAM and hP, respectively. The 18-fold difference in BA
deamination in hP and 9-fold difference in hAM also exhibited more variability than the
3-fold difference in hUCM. The interindividual variability in VAP-1 protein levels
measured by ELISA was also assessed across a population of tissue donors in hUCM,
hAM and hP. Mean levels of VAP-1 protein measured ranged from 267 to 1690, 168 to
506 and 0.329 to 1.10 ng/mg protein in hUCM, hAM and hP, respectively. The
variability of VAP-1 measured by ELISA normalized to total protein varied less (3- to 6-
fold) than the oxidative deamination activities.
Correlation constants for comparisons of individual BA deamination rates and
VAP-1 protein levels to each other and to emixustat deamination activity using both
Gausian (Pearson’s) and nonparametric (Spearman’s) analyses are presented in Table 3.
Correlations of individual BA deamination rates and VAP-1 protein levels to emixustat
deamination activity are graphically presented in Figure 4. VAP-1 BA deamination
activity was highly correlated with the VAP-1 levels in plasma and significantly
correlated in hUCM and hAM supporting the use of BA deamination as a marker for
VAP-1 activity. Benzylamine deamination activity and VAP-1 protein levels were well
correlated with emixustat deamination in hUCM, hAM and hP. These results indicate that
interindividual variability in VAP-1 contributes to variability emixustat deamination.
-
19 -

DMD Fast Forward. Published on February 20, 2019 as DOI: 10.1124/dmd.118.085811
This article has not been copyedited and formatted. The final version may differ from this version.
DMD # 85811
Also of interest, the plasma activity or protein levels were not predictive of the oxidative
deamination activity in the elderly human aorta membranes.
Recombinant Human Enzymes. Recombinant forms of human VAP-1, lysyl oxidase-like
protein-2 (LOXL2; another amine oxidase copper containing), MAO-A and MAO-B (BD
Gentest, Franklin Lakes, NJ) were also used to assess the specificity of this amine
oxidase activity in deaminating emixustat to ACU-5201. As noted earlier, rhVAP-1 had
similar affinity to emixustat as the other human tissue sources of VAP-1 with high rates
of maximum reaction velocity (see Table 1 and Figure 3) affirming the deaminase
activity of this enzyme for emixustat. No apparent metabolites of radiolabeled emixustat
were evident after incubation with rhMAO-A and rhMAO-B for 30 minutes, while
activities for the positive control substrates kynuramine and benzylamine confirmed
active monoamine oxidases (data on file). In addition, no ACU-5201 was observed when
incubating emixustat with rhLOXL2 for 30 minutes.
Selective Chemical Inhibitors. Selective inhibition of amine oxidase activity was also
used to assess the contribution of VAP-1 and MAO-A and -B in both hUCM and hAM.
Semicarbazide and LJP-1207 were co-incubated with emixustat as specific inhibitors of
VAP-1, clorgyline and deprenyl were co-incubated to selectively inhibit MAO-A and
MAO-B, respectively. Quinidine was tested because it is commonly employed in clinical
drug-interaction studies as a CYP2D6 inhibitor (FDA DDI guidance) and its potential
effects on amine oxidase activity have not been established. Figure 5 shows the effects of
high and low concentrations (10-fold difference) of each inhibitor to assess dose
dependence; however, the high and low concentrations were not the same for the
different compounds since low concentrations were chosen based on available maximum
plasma concentrations or in vitro IC50 values. LJP-1207 had the most pronounced effect
on emixustat deamination with complete inhibition at 1.0 µM. Semicarbazide had
moderate effects at 10 µM, but more complete inhibition at 100 µM. Clorgyline and
-
20 -

DMD Fast Forward. Published on February 20, 2019 as DOI: 10.1124/dmd.118.085811
This article has not been copyedited and formatted. The final version may differ from this version.
DMD # 85811
deprenyl did not inhibit emixustat deamination in hUCM at the low concentration of 50
µM, but both inhibited at 500 µM. Neither clorgyline nor deprenyl inhibited at 10 and
100 µM in hAM. Quinidine did not inhibit oxidative deamination of emixustat in vitro,
nor did β-aminoproprionitrile (BAPN; a LOX specific inhibitor; Mercier et al., 2009) in
hAM even at concentrations as high as 500 µM.
In Vitro Drug-Drug Interaction Potential.
Emixustat as the victim. The reaction phenotyping for oxidative deamination of
emixustat by VAP-1 infers its predominant role in the biotransformation to the three
major metabolites observed in human plasma. Since metabolic clearance by VAP-1 is an
uncommon pathway, its sensitivity to inhibition by other drugs as an underlying
mechanism for drug-drug interactions (DDI) has been seldomly scrutinized. To assess the
potential for emixustat exposure to increase in response to being victimized by an
inhibitor of VAP-1, in vitro screening for VAP-1 inhibition was conducted following
review of therapeutic agents with various functional moieties (e.g. hydrazine, guanidine,
phthalazine) that have been demonstrated to inhibit VAP-1 activity. Additionally,
inhibitors of MAO isoforms were included because they often have overlapping
inhibition profiles with VAP-1/SSAO (Kinemuchi et al., 2004; Dunkel et al., 2008).
Various potential inhibitors (15 compounds in total) were screened to assess their
potential to inhibit the formation of ACU-5201 in pooled hAM at 60 µM concentration of
emixustat which was less than maximal saturation but high enough to remain sensitive to
inhibition in pooled hAM, pooled hP and rhVAP-1. Inhibition of benzylamine
deamination was similarly screened at 80 µM. Most compounds were screened at a low
(
2 or 10 µM) and high (20 or 100 µM) concentration. Low concentrations were chosen
based on available maximum plasma concentrations or published in vitro IC50 values for
their targets and high concentrations were ten times the lower concentration. If a
-
21 -

DMD Fast Forward. Published on February 20, 2019 as DOI: 10.1124/dmd.118.085811
This article has not been copyedited and formatted. The final version may differ from this version.
DMD # 85811
compound exhibited > 50% inhibition at either of the screening concentrations, an IC50
was subsequently determined. Results are presented in Supplemental Table 1.
Guanabenz (20 µM), hydralazine (20 µM) and semicarbazide (100 µM) inhibited
the formation of ACU-5201 by 69, 87 and 70%, respectively. The specific VAP-1
inhibitor, LJP-1207, also inhibited oxidative deamination of emixustat by more than
50%; its IC50 as determined in earlier hUCM experiments was 75 nM. When tested for
inhibition of emixustat deamination in pooled hAM, IC50 values for guanbenz,
hydralazine and isoniazid were 10.2, 7.4 and 55.4 µM, respectively. IC50 values for
guanbenz and hydralazine were 46.9 and 6.4 µM when tested for inhibition of BA
deamination in pooled hAM, and 9.46 and 26.5 µM in rhVAP-1, respectively.
Anticipated results were observed as reported above for MAO-specific (clorgyline and
deprenyl) and LOX-specific (BAPN) inhibitors. The CYP2D6 inhibitors, paroxetine and
quinidine, also did not inhibit. Although benserazide (Boomsma et al., 1995) and caffeine
(
Olivieri and Tipton, 2011) are reported to be inhibitors of plasma VAP-1 and
antidepressants have been observed to inhibit BA deamination, these compounds were
not potent in vitro inhibitors of emixustat or BA deamination in vascular membranes.
Emixustat as a perpetrator. Inhibitory potential of emixustat as a perpetrator against
VAP-1-mediated BA (80 µM) deamination activity in hUCM and hAM yielded IC50
values of 161 and 109 µM, respectively. During in vitro incubations of human plasma,
1
00 µM emixustat inhibited BA deamination by < 30%. More thorough kinetic analyses
to determine the K value were performed in hUCM with emixustat concentrations of
00, 200 and 400 µM as shown in Figure 6 yielding a K value of 186 µM using a mixed
inhibition model.
i
1
i
Since VAP-1 appears to be upregulated in several human pathologies
(
Pannecoeck et al., 2015), there are numerous published reports of VAP-1 inhibition as a
potential anti-inflammatory therapy (Kinemuchi et al., 2004; Salter-Cid et al., 2005;
-
22 -

DMD Fast Forward. Published on February 20, 2019 as DOI: 10.1124/dmd.118.085811
This article has not been copyedited and formatted. The final version may differ from this version.
DMD # 85811
Dunkel et al., 2008). This study did not establish an in vitro system to examine the
potential to chemically induce VAP-1, but observations were made regarding VAP-1
variability across the donor tissues. Greater variability in VAP-1 activity in the elderly
aorta tissue compared to umbilical cords suggests that emixustat clearance may be more
variable in an elderly population. Mean rates of emixustat deamination via vascular tissue
(
membrane-bound) and soluble (plasma) VAP-1 in elderly individuals were evaluated
using two-way ANOVAs to determine if in vitro emixustat VAP-1 metabolism
differences were discernable between select sub-populations of the demographic data in
this small population of donors. There was no statistical significance between males (M)
and females (F), nor between diabetics (d) and non-diabetics (n). However, Caucasians
did have significantly higher emixustat metabolic rates than non-Caucasians in hAM
(
p<0.005, membrane-bound), but not in hP (soluble).
Clinical Pharmacokinetics. The mean plasma concentration–time profiles and
noncompartmental PK parameters of emixustat and its 3 major metabolites, ACU-5116,
ACU-5124 and ACU-5149, following daily emixustat doses of 2.5, 5 and 10 mg/kg for 7
days are shown in Figure 7 and Table 4, respectively. Mean plasma concentrations at
each sample collection time point for all 3 metabolites were generally ten-fold higher
than emixustat and the metabolite plasma concentration profiles declined in parallel with
emixustat on both days 1 and 7 (Figure 7). Maximum plasma concentrations (Cmax) of
emixustat (~11 nM at 10 mg QD) and its three deaminated metabolites increased dose
proportionally over the dose range and the average time to reach Cmax (Tmax) values
ranged between 2.1 and 4.1 hours for all 4 analytes at all 3 dose levels indicating rapid
metabolism with ACU-5116 having the earliest peak concentrations. Comparable to Cmax
values, AUC values for emixustat and its 3 metabolites ranged approximately 4- to 6-fold
across the 18 individuals in each cohort, within each dose level, and their mean values
increased in an apparent dose proportional manner on both days 1 and 7 consistent with
-
23 -

DMD Fast Forward. Published on February 20, 2019 as DOI: 10.1124/dmd.118.085811
This article has not been copyedited and formatted. The final version may differ from this version.
DMD # 85811
linear kinetics over the 2.5 to 10 mg dose range. AUC-based metabolite-to-parent ratios
indicated plasma exposure to each inactive metabolite were 6- to 11-fold greater than to
the pharmacologically active parent. Accumulation ratios of exposure (based on AUC) on
day 7 ranged from 1.09 to 1.31 for emixustat and its 3 metabolites indicating minimal
accumulation for any of the analytes as predicted since their elimination half-lifes (t1/2
)
were a fraction of the dosing interval (24 hr). Mean elimination half-lifes of emixustat
and its three metabolites were between 5.9 and 7.1 hr.
-
24 -

DMD Fast Forward. Published on February 20, 2019 as DOI: 10.1124/dmd.118.085811
This article has not been copyedited and formatted. The final version may differ from this version.
DMD # 85811
Discussion. In a previous human ADME study, three major deaminated metabolites of
emixustat were identified in the plasma following a single 40 mg dose: ACU-5116, ACU-
5124 and ACU-5149 (Fitzsimmons et al., 2018). The clinical PK results in subjects with
geographic atrophy presented here confirm that these three deaminated metabolites
predominate in plasma following single and multiple daily oral doses and reach
maximum concentrations rapidly; demonstrating formation-rate limited elimination with
similar t1/2 to the parent (see Table 2 and Figure 7).
Inconsistent with these in vivo results, in vitro metabolism of emixustat in human
hepatocytes resulted in a large fraction of emixustat (~34%) being unchanged with the
oxidatively deaminated metabolites each accounting for a smaller fraction (4–9%) of the
14C after 2 hours. The remaining metabolites observed in human hepatocytes were mainly
cyclohexanol metabolites and a cyclohexanone metabolite of the hydroxypropylamine,
yet none of these metabolites were ≥10% of the circulating plasma radioactivity.
Although the secondary hydroxylated metabolites, ACU-5124 and ACU-5149, both
achieve >10% of the total circulating plasma radioactivity, the primary metabolite, ACU-
5116, appeared in human plasma more quickly than both the major secondary
metabolites, emixustat (see Figure 7) or any of the other minor metabolites (Fitzsimmons
et al., 2018). This suggests that formation of the major secondary metabolites occurs
primarily via monohydroxylation of ACU-5116 by CYP isoform(s) (Figures 1 and 2).
Although in vivo formation of carboxylic acid metabolites from a primary amine
is not a common xenobiotic biotransformation, one example is primaquine (PQ), whose
major human plasma metabolite is the carboxylic acid, carboxyprimaquine (CPQ).
Plasma levels of CPQ rapidly exceeded PQ concentrations with Cmax of CPQ ten-fold
1
4
higher than the parent. Even though 64% of the radioactivity of a single C-PQ oral dose
was recovered in the urine, less than 4% was due to PQ (Mihaly et al., 1984). In vivo
-
25 -

DMD Fast Forward. Published on February 20, 2019 as DOI: 10.1124/dmd.118.085811
This article has not been copyedited and formatted. The final version may differ from this version.
DMD # 85811
biotransformation of PQ to a carboxylic acid had similar conversion kinetics to that
observed for emixustat and its primary carboxylic acid metabolite, ACU-5116. The high
rates of emixustat deamination motivated an early study that collected portal vein blood
from rats. High levels of ACU-5116, ACU-5124, and ACU-5149 were observed
presystemically within 15 minutes of an emixustat oral dose (data on file).
In vitro investigations into the sequential metabolism of PQ determined that
initial conversion of PQ to an aldehyde via MAO-A (Constantino et al., 1999; Pybus et
al., 2012) was followed by further oxidation to CPQ by aldehyde dehydrogenase
(
Frischer et al., 1991). This in vitro sequential metabolism of PQ is similar to the
conversion of emixustat to ACU-5116 where the initial aldol product of VAP-1 (see
+
Figure 1) is converted to the carboxylic acid only when NAD is added (Figure 2A). A
similar pattern was described for tresperimus wherein aldehyde metabolites were
predominantly observed following its incubation in plasma in vitro, while acid
metabolites were also observed in plasma collected from humans and rats dosed with
tresperimus (Claud et al., 2001). Because high levels of emixustat’s acid metabolites
were observed in the portal circulation of rats (as noted above), gastric mucosal and small
intestinal mucosal homogenate human tissue preparations were tested alongside the
1
4
rhMAO enzymes. The stability of [ C]emixustat in these in vitro systems drove the
decision to seek a different extra-hepatic model for emixustat oxidative deamination.
The present work indicates the biotransformation of emixustat is predominantly
driven by the copper-containing amine oxidase, VAP-1. Vascular adhesion protein-1 was
found to be identical in gene sequence and oxidative deamination function to SSAO
(
Smith et al., 1998), a member of AOC gene family that are ubiquitous across nature and
distinct from FAD cofactor-containing monoamine and polyamine oxidases (Salmi and
Jalkanen, 1992). As a homodimeric 180 – 200 kDa ectoenzyme with a topaquinone
cofactor (Holt et al., 1998), VAP-1 is found mainly on plasma membranes of
-
26 -

DMD Fast Forward. Published on February 20, 2019 as DOI: 10.1124/dmd.118.085811
This article has not been copyedited and formatted. The final version may differ from this version.
DMD # 85811
endothelium, adipose, and smooth muscle of mammals (Barrand and Callingham, 1985,
Precious et al., 1988, Wibo et al., 1980). Currently, amlodipine is the only marketed drug
with oxidative deamination metabolism by a copper-containing amine oxidase (Stopher et
al., 1998), yet this is a relatively minor metabolic clearance pathway. Vascular adhesion
protein-1 was also established as the major metabolic clearance enzyme of the
unmarketed immunosuppressive compound tresperimus (Claud et al., 2001; Claud et al.,
2002).
Low to no yields of carboxylic acid metabolites in hepatocytes, liver
microsomes, liver mitochondria, recombinant CYP and MAO enzymes was congruent
with constitutive VAP-1 expression in sinusoidal endothelium within the liver (Lalor et
al., 2002). The data presented here expands on the utility of endothelial membrane
preparations for investigating VAP-1 oxidative deamination metabolism of xenobiotics
that Claud et al (2002) first introduced. Similar enzyme kinetics for the oxidative
deamination of emixustat to ACU-5201 were observed in hUCM, hAM and rhVAP-1
(
see Table 1). The oxidative deamination activities in these sources of VAP-1 were also
confirmed using BA as a marker substrate in an assay where the aldehyde product, BAL,
was measured directly. The affinity of BA for VAP-1 in rhVAP-1, hUCM, hAM and hP
were very similar to the previously reported values (Claud et al., 2002) and its rates of
deamination were highly correlated with both the oxidative deamination of emixustat and
the VAP-1 levels measured by ELISA in individual human tissue sources of VAP-1
(
(
Figure 4 and Table 3). The variability of in vitro oxidative deamination in elderly hAM
8.8- to 17-fold) is comparable to the in vivo variability of emixustat plasma AUCs in
AMD patients within each dose cohort (4- to 6-fold). This suggests that clinical
variability in emixustat exposure is explained in part by interindividual variability in the
expression of VAP-1 and reinforces the use of these membranes as an in vitro model for
assessing the potential impact of inhibitors of VAP-1 on emixustat metabolic clearance.
-
27 -

DMD Fast Forward. Published on February 20, 2019 as DOI: 10.1124/dmd.118.085811
This article has not been copyedited and formatted. The final version may differ from this version.
DMD # 85811
In several mammalian species, including humans, VAP-1 has been found both on
vascular cell membranes and in a soluble form in plasma, with both forms determined to
be the same enzyme from protein sequencing and enzymology characterization (Smith et
al., 1998). Soluble VAP-1 in human plasma may originate from proteolytic cleavage of
the membrane-spanning N-terminal domains of membrane-bound homodimers, resulting
in a circulating enzyme that retains deamination activity (Salmi and Jalkanen, 1992) or
from circulating inflammatory endothelial cells that detach from sites of injury (Holmén
et al., 2005). VAP-1 deamination in human plasma and tissues has been compared to
other mammals yielding extreme variation in Vmax values (10,000-fold), with less
variation in K . Plasma BA deamination in humans has been reported to be
M
comparatively low, with BA deamination in human tissues being higher than in tissues
from rat and pig (Boomsma et al., 2000). Markedly higher BA deamination in human
vascular membranes than in plasma are consistent with the findings in the current study.
In addition, we previously reported that ACU-5201 was detected after 1 hr incubation in
1
4
human plasma (Fitzsimmons et al., 2018) with 85 - 96% of [ C]emixustat remaining
after 1 hr in human plasma and blood. Collectively this suggests that the membrane-
bound form of VAP-1 has a larger contribution to the rapid in vivo clearance of emixustat
than the soluble plasma form.
Initially selective chemical inhibition was used in reaction phenotyping to
confirm that ACU-5201 formation was inhibited by known inhibitors of VAP-1, but not
those selective for MAO-A, MAO-B, or LOX (see Figure 5). Although inhibition of
VAP-1 has been investigated extensively as an anti-inflammatory therapy over the last
few decades (Kinemuchi et al., 2004; Salter-Cid et al., 2005; Dunkel et al., 2008), clinical
drug interactions have yet to be observed since few drugs are eliminated by this pathway.
In the absence of established clinical observations, an in vitro screening approach was
undertaken for emixustat oxidative deamination. Therapeutic agents were surveyed in the
-
28 -

DMD Fast Forward. Published on February 20, 2019 as DOI: 10.1124/dmd.118.085811
This article has not been copyedited and formatted. The final version may differ from this version.
DMD # 85811
literature and screened in pooled sources of human VAP-1 to investigate the potential for
inhibitory DDIs. Inhibitors and each substrate (emixustat or BA) were co-incubated prior
to adding the appropriately diluted human VAP-1 sources to start the reactions. The
sensitivity of the two substrates to inhibition differed by as much as 5-fold in IC50 values
for guanabenz in hAM. A similar difference was seen for hydralazine against BA
deamination in hAM relative to rhVAP-1. The homodimeric structure of VAP-1 may
contribute to these differences since both cooperativity and half-site reactivity have been
observed for VAP-1 (Klema and Wilmot, 2012). Additionally, formation of imine Schiff
bases with primary amine substrates in the active site of VAP-1 competes with the
hydrazine inhibitors as the analogous hydrazone adducts. Given the greater intrinsic
hydrolytic stability of the hydrazones, time-dependent processes that were not accounted
for in the coincubation design may also contribute to the differential sensitivity.
Hydralazine is a hydrazine-containing drug that was a relatively potent inhibitor
of emixustat and BA deamination in pooled hAM with IC50 values of 7.4 and 6.4 µM,
respectively. At this time the ratio of the plasma levels of VAP-1 inhibitors relative to
their in vitro inhibition parameters has yet to be explored as quantitatively meaningful in
assessing the likelihood of a clinically relevant drug interaction. The potential for
hydralazine to have an impact on emixustat plasma exposures via VAP-1 inhibition
remains valuable in identifying potential extrinsic factors that may impact therapeutic
response to emixustat. Summation of the in vitro inhibition results presented here
suggests that VAP-1 is relatively less sensitive to semicarbazide inhibition.
In summary, the data presented here shows that VAP-1 is predominantly
responsible for emixustat oxidative deamination in human vascular membranes (hUCM
and hAM) and human plasma (hP) with much higher rates in the membranes. Chemical
inhibition and recombinant human enzyme strategies were effective in establishing the
-
29 -

DMD Fast Forward. Published on February 20, 2019 as DOI: 10.1124/dmd.118.085811
This article has not been copyedited and formatted. The final version may differ from this version.
DMD # 85811
role of VAP-1 in emixustat metabolism and offers a simple path for drug discovery
candidates comprising primary amine to carboxylic acid biotransformations. The three
major metabolites found in plasma are formed rapidly and eliminated in a formation-rate
limited fashion and can be formed sequentially in vitro with additional assay components
for ALDH and CYP. The rates of VAP-1 marker substrate activity and emixustat
deamination were highly variable in a population of 20 elderly individuals, yet they were
highly correlated with each other and VAP-1 levels measured by ELISA. Variability
observed in the in vitro systems tested were similar to that observed in a clinical PK
study. Most potential co-medications tested did not inhibit emixustat oxidative
deamination in vitro with the exception of the vasodilators hydralazine and guanabenz.
Lastly, clinically relevant inhibiton of VAP-1 by emixustat seems unlikely since its
plasma concentrations are over ten thousand-fold less than the calculated K .
i
-
30 -