Download Mechanistic Pharmacokinetic-Pharmacodynamic Modeling of

Supplemental material to this article can be found at:
http://dmd.aspetjournals.org/content/suppl/2013/04/12/dmd.112.050864.DC1
1521-009X/41/7/1319–1328$25.00
DRUG METABOLISM AND DISPOSITION
Copyright ª 2013 by The American Society for Pharmacology and Experimental Therapeutics
http://dx.doi.org/10.1124/dmd.112.050864
Drug Metab Dispos 41:1319–1328, July 2013
Mechanistic Pharmacokinetic-Pharmacodynamic Modeling of
BACE1 Inhibition in Monkeys: Development of a Predictive Model for
Amyloid Precursor Protein Processing s
Xingrong Liu, Harvey Wong, Kimberly Scearce-Levie, Ryan J. Watts, Melis Coraggio,
Young G. Shin, Kun Peng, Kristin R. Wildsmith, Jasvinder K. Atwal, Jason Mango,
Stephen P. Schauer, Kelly Regal,1 Kevin W. Hunt, Allen A. Thomas, Michael Siu,
Joseph Lyssikatos, Gauri Deshmukh, and Cornelis E. C. A. Hop
Genentech Inc., South San Francisco, California (X.L., H.W., K.S.-L., R.J.W., M.C., Y.G.S., K.P., K.R.W., J.K.A., J.M., S.P.S., M.S.,
J.L., G.D., C.E.C.A.H.); and Array BioPharma Inc., Boulder, Colorado (K.R., K.W.H., A.A.T.)
Received December 21, 2012; accepted April 12, 2013
This study was conducted to determine the pharmacokinetics (PK)
and pharmacodynamics (PD) of two novel inhibitors of b-site
amyloid precursor protein (APP)–cleaving enzyme (BACE1), GNE-629
[(4S,4a9S,10a9S)-2-amino-89-(2-fluoropyridin-3-yl)-1-methyl-39,49,4a9,10a9tetrahydro-19H-spiro[imidazole-4,109-pyrano[4,3-b]chromen]5(1H)-one] and GNE-892 [(R)-2-amino-1,39,39-trimethyl-79-(pyrimidin5-yl)-39,49-dihydro-29H-spiro[imidazole-4,19-naphthalen]-5(1H)-one], and
to develop a PK-PD model to predict in vivo effects based solely
on in vitro activity and PK. GNE-629 and GNE-892 concentrations
and PD biomarkers including amyloid b (Ab) in the plasma and
cerebrospinal fluid (CSF), and secreted APPb (sAPPb) and secreted
APPa (sAPPa) in the CSF were measured after a single oral administration of GNE-629 (100 mg/kg) or GNE-892 (30 or 100 mg/kg)
in cynomolgus monkeys. A mechanistic PK-PD model was developed
to simultaneously characterize the plasma Ab and CSF Ab, sAPPa,
and sAPPb using GNE-629 in vivo data. This model was used to
predict the in vivo effects of GNE-892 after adjustments based on
differences in in vitro cellular activity and PK. The PK-PD model
estimated GNE-629 CSF and free plasma IC50 of 0.0033 mM and
0.065 mM, respectively. These differences in CSF and free plasma IC50
suggest that different mechanisms are involved in Ab formation in
these two compartments. The predicted in vivo effects for GNE-892
using the PK-PD model were consistent with the observed data. In
conclusion, a PK-PD model was developed to mechanistically
describe the effects of BACE1 inhibition on Ab, sAPPb, and sAPPa
in the CSF, and Ab in the plasma. This model can be used to
prospectively predict in vivo effects of new BACE1 inhibitors using
just their in vitro activity and PK data.
Introduction
APP is processed via two distinct pathways: namely, the amyloidogenic
and nonamyloidogenic. The amyloidogenic pathway includes cleavage of APP by b-secretase, also known as b-site APP cleavage
enzyme 1 (BACE1), producing a secreted ectodomain of APP (sAPPb)
and a membrane-tethered C-terminal fragment with 99 amino acids,
which is further cleaved by g-secretase and results in multiple Ab
peptides of 38 to 42 amino acids long (Sinha and Lieberburg, 1999). The
most abundantly produced variants are Ab1-40 and Ab1-42. Although
Ab1-40 is more abundant, Ab1-42 is considered the most toxic peptide
(Younkin, 1995; Lesné et al., 2006). In contrast, the nonamyloidogenic
pathway includes cleavage of APP by a-secretase and formation of
secreted APPa (sAPPa) and a C-terminal fragment with 83 amino acids.
These peptides do not contribute to the formation of plaques in the brain
(Kandalepas and Vassar, 2012).
One approach to reduce the levels of the toxic amyloidogenic
peptides is to inhibit BACE1 and thereby direct the APP cleavage
Alzheimer’s disease (AD) is characterized pathologically by the
accumulation of extracellular amyloid b (Ab) plaques and intracellular
neurofibrillary tangles in the brain, subsequently leading to neuronal
loss in brain regions important for cognition and memory. These
amyloid plaques are primarily composed of Ab peptides generated
from amyloid precursor protein (APP) (Glenner and Wong, 1984;
Masters et al., 1985). AD caused by the accumulation of Ab in the
brain through either overproduction or reduced clearance is the leading
hypothesis in AD research and is known as “the amyloid hypothesis”
(Hardy and Selkoe, 2002).
This research was supported by Genentech Inc.
1
Current affiliation: ProPharma Services, Superior, Colorado.
dx.doi.org/10.1124/dmd.112.050864.
s This article has supplemental material available at dmd.aspetjournals.org.
ABBREVIATIONS: Ab, amyloid b peptide; AD, Alzheimer’s disease; APP, amyloid precursor protein; AUC, area under concentration-time curve;
AZD3839, (S)-1-(2-(difluoromethyl)pyridin-4-yl)-4-fluoro-1-(3-(pyrimidin-5-yl)phenyl)-1H-isoindol-3-amine hemifumarate; BACE, b-site APP cleavage enzyme; BBB, blood-brain barrier; BSA, bovine serum albumin; CNS, central nervous system; CSF, cerebral spinal fluid; FBS, fetal bovine
serum; fu, unbound fraction; GNE-629, 4S,4a’S,10a’S)-2-amino-8’-(2-fluoropyridin-3-yl)-1-methyl-39,4’,4a’,10a’-tetrahydro-1’H-spiro[imidazole-4,10’pyrano[4,3-b]chromen]-5(1H)-one; GNE-892, (R)-2-amino-1,39,39-trimethyl-7’-(pyrimidin-5-yl)-39,4’-dihydro-2’H-spiro[imidazole-4,1’-naphthalen]5(1H)-one; LC-MS/MS, liquid chromatography with tandem mass spectrometry; LY2811376, (4S)-4-(2,4-difluoro-5-pyrimidin-5-ylphenyl)4-methyl-5,6-dihydro-4H-1,3-thiazin-2-amine; MDCK, Madin-Darby canine kidney cell line; MDR1, human multidrug resistance gene; P-gp,
P-glycoprotein; PD, pharmacodynamics; PK, pharmacokinetics; sAPPa, secreted APPa peptide; sAPPb, secreted APPb peptide.
1319
Downloaded from dmd.aspetjournals.org at ASPET Journals on May 7, 2017
ABSTRACT
1320
Liu et al.
Fig. 1. Structures of GNE-629 and GNE-892.
Materials and Methods
Chemicals. GNE-629 and GNE-892 were synthesized at Array BioPharma
Inc. (Boulder, CO). All chemicals used in the experiments were of the highest
available commercial grade.
BACE1 Enzyme Activity Assay. Recombinant human BACE1 expressed
and isolated from Chinese hamster ovary cells was incubated with 0.15 mM
peptide substrate (Biotin-KTEEISEVNLDAEFRHDSGYEVHHQKL) (American
Peptide Company, Inc., Sunnyvale, CA) in 6 ml 50 mM sodium acetate, pH 4.4,
and 0.1% 3-[(3-cholamidopropyl)dimethylammonio]-1-propanesulfonate for 6
hours at room temperature. The reaction with various concentrations of BACE1
inhibitors was quenched with 5 ml of 200 mM Tris pH 8.0 buffer containing
20 mM EDTA, 0.1% bovine serum albumin (BSA), 0.8 M KF, the time-resolved
fluorescence resonance energy transfer acceptor strepavidin-D2 (final concentration 0.043 mM; CisBio Bioassays, Codolet, France), and the time-resolved
fluorescence resonance energy transfer donor Europium cryptate–labeled antipeptide antibody 6E10 (final concentration 0.0011 mM; CisBio). After 2 hours
at room temperature, the ratio of the emissions at 665 and 615 nm was determined and the IC50 was calculated using a 4-parameter logistic fit.
BACE1 Cellular Inhibition Assay. HEK-293 cells stably transfected with
a DNA construct containing the coding sequence for the wild-type human
APP695 sequence (293-hAPP cells) were grown in Dulbecco’s modified
Eagle’s medium supplemented with 10% fetal bovine serum (FBS), penicillin/
streptomycin, and 150 mg/ml G418 [O-2-amino-2,7-didesoxy-D-glycero-aD-gluco-heptopyranosyl-(1→4)-O-(3-desoxy-4-C-methyl-3-(methylamino)-b-Larabinopyranosyl-(1→6))-D-streptamin]. Cells were plated in 96-well plates at
35,000 cells per well and allowed to attach for 8–12 hours. Media were changed
to Dulbecco’s modified Eagle’s medium supplemented with 10% FBS and
penicillin/streptomycin 15 minutes prior to compound addition. Compound was
then added at a final concentration of 0.5% dimethylsulfoxide. After 48 hours,
4 ml media from each well was added to a corresponding well of a 384-well plate
(Cat# 6008280; PerkinElmer, Waltham, MA) containing the homogeneous timeresolved fluorescence reagents. Homogeneous time-resolved fluorescence
analysis was performed using the CisBio amyloid b1-40 peptide assay kit
(Cat# 62B40PEC) following the manufacturer’s instructions.
In Vitro Transporter Assay. Madin-Darby canine kidney cell line 1
(MDCKI) cells transfected with the human multidrug resistance gene 1
Downloaded from dmd.aspetjournals.org at ASPET Journals on May 7, 2017
through the nonamyloidogenic pathway (McConlogue et al., 2007;
Citron, 2010). However, it has been particularly challenging to
develop orally available and brain-penetrable small-molecule BACE1
inhibitors. Great progress has been made in the last few years to design potent BACE1 inhibitors that can reduce brain Ab after oral
administration. There are numerous reports studying the effects of
BACE1 inhibition on cerebral spinal fluid (CSF) and/or brain Ab and
other biomarkers in mice, rats, dogs, monkeys, and humans (Hussain
et al., 2007; Elvang et al., 2009; Sankaranarayanan et al., 2009; Zhu
et al., 2009; Fukumoto et al., 2010; May et al., 2011; Dineen et al.,
2012; Weiss et al., 2012). Sankaranarayanan et al. (2009) were the
first to show in vivo reduction of plasma and CSF Ab after single
or multiple oral administration of a BACE1 inhibitor, N-[4-{5-[(1R)1-amino-1-methyl-2-phenylethyl]-1,3,4-oxadiazol-2-yl}-3-chloro-6-((2methoxyethyl){[(1S,2S)-2-methylcyclopropyl]methyl}amino)pyridin2-yl]-Nmethylmethanesulfonamide, in nonhuman primates (rhesus
monkeys). However, due to its extensive first-pass metabolism,
coadministration of the cytochrome P450 inhibitor ritonavir was
necessary to boost the oral exposure of N-[4-{5-[(1R)-1-amino-1methyl-2-phenylethyl]-1,3,4-oxadiazol-2-yl}-3-chloro-6-((2-methoxyethyl){[(1S,2S)-2-methylcyclopropyl]methyl}amino)pyridin-2-yl]-Nmethylmethanesulfonamide to achieve efficacious levels in the reported
preclinical studies. May et al. (2011) reported that a potent BACE1
inhibitor LY2811376 [(4S)-4-(2,4-difluoro-5-pyrimidin-5-ylphenyl)4-methyl-5,6-dihydro-4H-1,3-thiazin-2-amine] reduced Ab levels in
human CSF. The development of this compound was discontinued due
to ocular toxicity in rats. Recently, Jeppsson et al. (2012) disclosed that
a potent BACE1 inhibitor AZD3839 [(S)-1-(2-(difluoromethyl)pyridin4-yl)-4-fluoro-1-(3-(pyrimidin-5-yl)phenyl)-1H-isoindol-3-amine hemifumarate] with an IC50 of 0.0048 mM can reduce CSF Ab in monkeys.
However, the reduction was not robust as AZD3839 has a very short
half-life (18 minutes) in monkeys. Therefore, more potent and safe
BACE1 inhibitors are needed to test the amyloid hypothesis in humans
(Probst and Xu, 2012). In this article, we report the in vivo characterization
of BACE1 inhibition in monkeys for the following two novel BACE1
inhibitors: GNE-629 [4S,4a’S,10a’S)-2-amino-8’-(2-fluoropyridin-3-yl)-1methyl-39,4’,4a’,10a’-tetrahydro-1’H-spiro[imidazole-4,10’-pyrano[4,3b]chromen]-5(1H)-one] and GNE-892 [(R)-2-amino-1,39,39-trimethyl7’-(pyrimidin-5-yl)-39,4’-dihydro-2’H-spiro[imidazole-4,1’-naphthalen]5(1H)-one] (Fig. 1).
Mechanistic pharmacokinetic-pharmacodynamic (PK-PD) modeling can play an important role in the drug discovery and development
process by providing an integrated understanding of relationships
between compound plasma concentrations and biomarker responses
(Gabrielsson and Green, 2009; Gibbs, 2010; van der Graaf and
Benson, 2011; Wong et al., 2012a,b). In addition, PK-PD modeling
is a useful tool that can facilitate the translation of preclinical data to
humans. Recently, Lu et al. (2012) reported the use of an indirect
response PK-PD model for Ab lowering in CSF and brain (Jusko and
Ko, 1994). The reported PK-PD modeling efforts on in vivo BACE1
inhibition studies characterized the effects of BACE1 inhibition on
Ab, and no models have been reported that examine the effects
of BACE1 inhibition on overall APP processing. The development
of such a model would aid in better understanding the interactions of
BACE1 inhibitors with various products of APP processing and allow
for predictions of how these potential PD markers would behave after
BACE1 inhibition. Therefore, the objectives of the present study were
to examine BACE1 inhibition in cisterna magna cannulated cynomolgus monkeys for two novel BACE1 inhibitors GNE-629 and GNE-892
and to use a modeling approach to study BACE1 inhibition in the
monkeys.
PK-PD Modeling of BACE1 Inhibition
(MDR1) gene (MDCKI–MDR1) were maintained at 37°C, 95% humidity, and
5% CO2 in culture with Eagle’s minimum essential medium (0.1% nonessential
amino acids, 1 mM sodium pyruvate, 2 mM L-glutamine, 1.5 g/l sodium
bicarbonate) supplemented with 10% FBS. The MDCKI–MDR1 cell medium
was further supplemented with 0.2 mM colchicine to sustain MDR1 expression.
The monolayers were equilibrated for 30 minutes in transport buffer (Hanks’
balanced salt solution with 10 mM HEPES, pH 7.4) at 37°C with 5% CO2 and
95% humidity prior to the experiment. Drug concentrations were determined by
liquid chromatography with tandem mass spectrometry (LC-MS/MS). The
transport of GNE-629 or GNE-892 (5 mM) was examined in the apical-tobasolateral and basolateral-to-apical directions. The apparent permeability
(Papp) was calculated using eq. 1:
Papp ¼
dQ 1 1
• •
dt C0 A
ð1Þ
luminescence was measured with a Sector Imager (Meso Scale Discovery). The
assay had lower limit of quantification of 0.05 ng/ml for sAPPa and 0.03 ng/ml
for sAPPb.
Plasma Protein Binding. The unbound fractions in plasma for GNE-629
and GNE-892 were determined in a 48-well rapid equilibrium dialysis device
using a dialysis membrane with a molecular mass cut-off value of 8000 (Pierce
Biotechnology, Rockford, IL). Blank plasma was spiked with GNE-629 or
GNE-892 at a final concentration of 5 mM. Three hundred microliters of plasma
containing compound was added to the donor side and 500 ml of buffer was
added to the receiver side. The plate was placed on a rocking platform at 400
rpm with 1 mm radius agitation for 4 hours at 37°C. Twenty microliters of
buffer and plasma samples was transferred to a 96-well plate and 20 ml of blank
plasma and buffer was added to the buffer and plasma samples, respectively.
Plasma proteins were precipitated using 65% acetonitrile containing internal
standard (0.1 mM propranolol), drug concentrations were quantitated by
LC-MS/MS, and the unbound fraction (fu) was calculated as the ratio of the
buffer concentration versus the plasma concentration.
LC-MS/MS Analysis of GNE-629 and GNE-892 in Plasma and CSF. The
plasma samples were prepared for analysis by adding 25 ml GNE-629 or GNE892 plasma into a 96-well plate followed by the addition of 25 ml internal
standard solution and 200 ml acetonitrile to each sample for protein precipitation.
The samples were vortexed and centrifuged for 5 minutes at 1500g. Fifty
microliters of the resulting supernatant was mixed with 300 ml water and 10 ml
was injected onto an LC-MS/MS for analysis. For the CSF samples, 25 ml of
GNE-629 or GNE-892 CSF samples was mixed with 25 ml blank plasma and
then 25 ml of each CSF-blank plasma mixture was extracted using the same
protein precipitation procedure as described above for the plasma samples.
Concentrations of GNE-629 and GNE-892 in plasma and CSF extracts were
determined using an AB Sciex QTrap4000 mass spectrometer (AB Sciex, Foster
City, CA) connected to a CTC PAL autosampler (Leap Technologies, Chapel
Hill, NC). GNE-629 and GNE-892 were separated using an ACE 5 phenyl
reverse phase (100 2.1 mm, 5 mm) column (Advanced Chromatography
Technologies, Aberdeen, UK) at room temperature using a LC-10AD pump with
a Shimadzu SCL-10A controller (Shimadzu, Columbia, MD). The mobile phase
consisted of mobile phase A (water with 0.1% formic acid) and mobile phase B
(acetonitrile with 0.1% formic acid). GNE-629 and GNE-892 were eluted using
an initial condition of 20% mobile phase B followed by a linear gradient to 80%
mobile phase B over 5 minutes with a flow rate of 0.5 ml/min. The mass
spectrometer was used in the positive-ion mode using an electrospray interface at
400°C with nitrogen as both the nebulizing and heating gases. GNE-629 and
GNE-892 were analyzed in the multiple reaction monitoring using the following
transitions: m/z 383.2→299.1 and 336.2→252.2, respectively. The calibration
curves for GNE-629 and GNE-892 were prepared by plotting the appropriate peak
area ratios of analyte versus internal standard against the known concentrations of
GNE-629 and GNE-892 in plasma or CSF using a quadratic regression with 1/x2
weighing. Concentrations of GNE-629 and GNE-892 in samples were determined
by interpolation from their respective standard curves. The dynamic ranges for
both compounds were 0.0027 to 52 mM in plasma and 0.00014 to 0.52 mM in
CSF. A run was considered as acceptable when quality control samples as well as
calibration curve samples were within 625% of the nominal concentration, except
for the lowest quantification sample where 630% was acceptable.
Data Analysis. The concentrations of all biomarkers are expressed as the
percentage of the average baseline from 216 to 24 hours prior to the oral dose
for each group. The plasma and CSF area under concentration-time curve
(AUC) values were calculated using linear trapezoidal rule. The unpaired t test
was used to determine the significance of the biomarker changes at each time
point using Microsoft Excel (Redmond, WA).
Mechanistic PK-PD Modeling. A mechanistic PK-PD model describing the
pharmacokinetics and pharmacodynamics of GNE-629 in monkey was constructed using SAAM II v1.2.1 software (Saam Institute, University of Washington,
Seattle, WA). All parameters were estimated by fitting the PK and PD models
to the average of the observed data at each time point. Estimates of all parameters are presented as the parameter estimate followed by the percent standard
error in parentheses.
The configuration of the full mechanistic PK-PD model is shown in Fig. 2.
An oral two compartment model with an additional compartment describing
GNE-629 CSF concentrations (Fig. 2A) was fit to GNE-629 plasma and CSF
concentrations as described by eqs. 2–5.
Downloaded from dmd.aspetjournals.org at ASPET Journals on May 7, 2017
where dQ/dt is the rate of compound appearance in the receiver compartment,
C0 is the concentration in the donor compartment at time = 0, and A is the
surface area of the insert.
In Vivo Experiments. The PK and PD studies were conducted in male
cynomolgus macaques monkeys (3–5 years, 2–3 kg) implanted with catheters
in the cisterna magna that allowed for repeated sampling of CSF in conscious
animals (Maccine Pte Ltd, Singapore). After the catheters were implanted,
animals were maintained for at least 2 weeks on a 12-hour light/dark cycle in
a temperature- and humidity-controlled environment with free access to food
and water prior to dosing the inhibitors. Dosing solutions of each drug were
prepared as a suspension in 1% carboxymethylcellulose sodium, 0.5% Tween
80, and 5 mM citrate buffer at pH 5.0. The GNE-629 PK-PD study was
conducted in a two-way crossover design with vehicle and the 100 mg/kg dose
being given orally with an 18-day washout between each dosing period (n = 11
animals). The GNE-892 PK-PD study was conducted in a three-way crossover
design with the vehicle, 30 mg/kg, and 100 mg/kg doses being given orally
with 18- to 20-day washout periods between each arm of the study (n = 13
animals). Whole blood was collected from the femoral vein. Baseline blood and
CSF samples were collected from 216 to 24 hours prior to dosing. Serial blood
and CSF samples were collected postdose from 0 to 72 hours for GNE-629 and
from 0 to 48 hours for GNE-892. CSF samples were diluted 1:2 with 5% BSA
solution in polypropylene tubes. Whole blood samples were stored on ice until
centrifuged for the preparation of plasma. The plasma and CSF samples were
immediately frozen on dry ice and then stored at 270°C. All studies were
conducted in accordance with Genentech Animal Care and Use Procedures.
Determination of Ab1-40 in Plasma. The concentrations of Ab1-40 in
plasma were determined using an assay modified based on the MA6000 Human
(6E10) Ab Kit (Cat# K111BVE-2; Meso Scale Discovery, Gaithersburg, MD).
Assay procedures followed the manufacturer’s instructions. The capture antibody,
specific for the C terminus of Ab1-40, was precoated on the 96-well plates (Meso
Scale Discovery), and Sulfo-Tag anti-Ab monoclonal antibody 6E10 was used for
detection. A detection substrate was added and electrochemical luminescence was
measured with a Sector Imager (Meso Scale Discovery). The assay had a lower
limit of quantification of 0.060 ng/ml in plasma.
Determination of Ab 1-40 in CSF. The concentrations of total Ab1-40 in
CSF samples were determined using a sandwich enzyme-linked immunosorbent assay. Briefly, rabbit polyclonal antibody specific for the C terminus of
Ab1-40 (Millipore, Bedford, MA) was coated onto 96-well plates, and
biotinylated anti-Ab1-40 monoclonal antibody 6E10 (Covance, Dedham, MA)
was used for detection. The assay had a lower limit of quantification of 0.14 ng/ml
in CSF.
Determination of sAPPa and sAPPb in CSF. CSF concentrations of
sAPPa and sAPPb were determined using the sAPPa/sAPPb multi-spot assay
from Meso Scale Discovery. CSF was thawed on ice, and then diluted 1:20
into 1% BSA in Tris-buffered saline/Tween 20 (TBST) (10 mM Tris buffer,
pH 8.0, 150 mM NaCl, 0.1% Tween-20). The assay was performed per the
manufacturer’s protocol. Briefly, the multi-spot plate (Meso Scale Discovery)
was blocked for 1 hour at room temperature with 3% BSA in TBST. Plates
were rinsed three times with TBST, and the diluted CSF was added to the wells
in duplicate for 1 hour. After analyte capture, plates were rinsed with TBST and
detection antibody was added for 1 hour. After a final TBST rinse, a detection
substrate (Sulfo-Tag; Meso Scale Discovery) was added and electrochemical
1321
1322
Liu et al.
Fig. 2. Schematic representation of the mechanistic PK-PD model for BACE1 inhibition.
(A) PK model for GNE-629. (B) PK model
for GNE-892. (C) PD model for CSF Ab,
sAPPa, and sAPPb. (D) PD model for plasma
Ab. Unbound plasma concentrations were
linked to the plasma PD model and CSF
concentrations were linked to the CSF PD
model.
ð2Þ
dX1
¼ ka X0 þ k21 X2 þ kcsf 21 Xcsf 2 ðke þ k12 þ k12csf fu ÞX1
dt
ð3Þ
dX2
¼ k12 X1 2 k21 X2
dt
ð4Þ
dXcsf
¼ k12csf fu X1 2 kcsf 21 Xcsf
dt
ð5Þ
X0, X1, X2, and Xcsf are the amounts of compound (i.e., GNE-629) (mmol/kg)
in the gastrointestinal, central, peripheral, and CSF compartments, respectively.
ka and ke are the absorption and elimination rate constants (h21), respectively.
k12 and k21 are the rate constants from the central to peripheral and from the
peripheral to central compartment, respectively. k12csf and kcsf21 are the rate
constants from the central to CSF and from the CSF to central compartment,
respectively. The unbound fraction (fu) of GNE-629 in plasma was fixed to 0.39
as determined by equilibrium dialysis (Table 1).
V1 is the apparent volume of the central compartment after oral dosing (l/kg)
was fixed to 12.3 l/kg (calculated as Vc/F) based on a volume of the central
compartment of 2.5 l/kg (Vc) estimated from an independent intravenous
cynomolgus monkey study and a bioavailability (F) of 0.21 at the 100 mg/kg oral
dose (unpublished data). Vcsf is the volume of the CSF compartment (l/kg) and
was fixed to a value of 2 ml/kg (Poplack et al., 1977; Muraszko et al., 1993;
Rubenstein et al., 2003). Amounts in the central (or plasma) compartment and
CSF compartment were scaled to concentrations by dividing by V1 and Vcsf,
respectively. CL1-csf is the intercompartmental clearance between the central and
the CSF compartment (l/h per kilogram) and CLcsf21 is the intercompartmental
clearance between the CSF and central compartment (l/h per kilogram). They are
secondary calculated parameters and were calculated using methods listed in
Table 3. As noted in eqs. 3 and 5 above, unbound GNE-629 in the central
compartment is assumed to have access to the CSF compartment. Parameter
estimates from fitting the plasma and CSF concentrations were fixed prior to
modeling the CSF and plasma PD models described below.
The effects of GNE-629 on CSF APP processing are depicted by the CSF
PD model in Fig. 2C and described by eqs. 6–9. The proposed PK-PD model
assumes that the rate-limiting step for Ab formation is the cleavage of APP via
b-secretase. This assumption is supported by the findings that the Ab profiles
are similar to the sAPPb-time profiles for both BACE1 inhibitors in monkeys.
Furthermore, it has been demonstrated in rats that brain Ab and C99 were also
changed similarly after BACE1 inhibition (May et al., 2011). In addition, we
assume the maximal inhibition is 100%.
TABLE 1
Molecular mass and in vitro parameters for GNE-629, GNE-892, and LY2811376
Parameter
GNE-629
GNE-892
LY2811376
Molecular mass
BACE1 enzyme IC50 (mM)
BACE1 cell IC50 (mM)
MDR1 Pab (1026 cm/s)
MDR1 Pba (1026 cm/s)
MDR1 Pba/Pab ratio
fu (monkey)
fu (human)
382.4
0.010 6 0.003
0.0080 6 0.002
1.0 6 0.2
45 6 5
44
0.39 6 0.02
NA
335.4
0.048 6 0.041
0.027 6 0.014
1.2 6 0.1
31 6 3
25
0.55 6 0.02
NA
320.4
0.21 6 0.09
0.023 6 0.009
21 6 2
56 6 2
2.6
NA
0.44 6 0.01
NA, not available.
Downloaded from dmd.aspetjournals.org at ASPET Journals on May 7, 2017
dX0
¼ 2 k a X0
dt
1323
PK-PD Modeling of BACE1 Inhibition
dAPP
Ccsf
¼ kinAPP 2 kAPPð1 2
Þ 2 ka APP
dt
IC50 þ Ccsf
ð6Þ
dAb
Ccsf
¼ kAPPð1 2
Þ 2 kAb
dt
IC50 þ Ccsf
ð7Þ
dsAPPb
Ccsf
¼ kAPPð1 2
Þ 2 ksAPPb
dt
IC50 þ Ccsf
ð8Þ
dsAPPa
¼ ka APP 2 ka sAPPa
dt
ð9Þ
APP, sAPPa, sAPPb, and Ab are APP, sAPPa, sAPPb, and Ab1-40 levels
in CSF (%), kinAPP is the formation rate of APP (%/h), k is a rate constant (h21)
describing Ab and sAPPb formation from APP and their elimination, IC50
(mM) is the CSF concentration of a BACE1 inhibitor (i.e., GNE-629) that
inhibits the formation rate of Ab and sAPPb from APP by 50%, and ka is the
rate constant (h21) describing the formation and elimination of sAPPa. The
initial values of APP, sAPPa, sAPPb, and Ab were set at 100%.
Changes in plasma Ab concentrations are characterized using an indirect
response model in Fig. 2D and described by eq. 10.
ð10Þ
Abplasma is the plasma concentrations of Ab (%), kinplasma (%/h) is the
formation rate of plasma Ab, IC50(plasma) (mM) is the unbound GNE-629
plasma concentration required for inhibiting the plasma Ab formation by 50%,
and koutplasma (h21) is the first-order plasma Ab elimination rate constant. The
initial value of Abplasma was set to 100%. The value of koutplasma was fixed to
equal kinplasma/(initial value of Abplasma).
Simulations of Ab, sAPPb, and sAPPa CSF Levels and Plasma Ab
after Oral Administration of GNE-892 to Monkey. To test the performance
of the mechanistic PD models, simulations of CSF levels of Ab, sAPPb, and
sAPPa and plasma Ab levels were performed for the second BACE inhibitor,
GNE-892. Prior to the simulation of the PD, the GNE-892 plasma and CSF
concentrations were fit to PK model shown in Fig. 2B to provide GNE-892
concentrations to drive the PD responses. A CSF peripheral compartment was
added to the GNE-629 PK model (Fig. 2B) to more accurately describe the CSF
concentrations at the terminal phase. The model is described by eqs. 2–4 and
eqs. 11 and 12 describe the additional CSF peripheral compartment.
dXcsf
¼ k12csf fu X1 2 ðkcsf 21 þ kcsf 2 csfp ÞXcsf þ kcsfp 2 csf Xcsfp
dt
ð11Þ
dXcsfp
¼ kcsf 2 csfp Xcsf 2 kcsfp 2 csf Xcsfp
dt
ð12Þ
Xcsfp is the amount of GNE-892 in the CSF peripheral compartment (mmol/kg).
kcsf2csfp and kcsfp2csf are the rate constants from the CSF to CSF peripheral
and CSF peripheral to CSF compartments, respectively (h21). The fu of GNE-892
in plasma was fixed to 0.55 as determined by equilibrium dialysis (Table 1).
Simulations of PD response in monkey were performed using the PK model
for GNE-892 to provide GNE-892 plasma and CSF concentrations to drive the
PD models. The in vivo plasma and CSF IC50 values for GNE-892 were
calculated using the in vivo IC50 of GNE-629 (Table 3) after adjustment for the
3.4-fold decrease in potency for GNE-892 compared with GNE-629 in the
BACE1 cell assay (Table 1). The in vivo CSF IC50 was set to 0.0111 mM and
the plasma in vivo IC50(plasma) was set to 0.221 mM.
Simulations of Ab, sAPPb, and sAPPa CSF Levels after Oral
Administration of LY2811376 to Humans. CSF levels of Ab, sAPPb, and
sAPPa after oral administration of 30 mg and 90 mg LY2811376 were simulated
using the CSF PD model for GNE-629. LY2811376 CSF concentrations after
a 90-mg oral dose of LY2811376 were simulated based on the following reported
human values (May et al., 2011): CSF AUC of 4.74 mM*h. The CSF half-life of
LY2811376 was assumed to be identical to the plasma half-life of 40 hours. CSF
Tmax was assumed to either match reported plasma Tmax (2 hours) or to be
slightly delayed (i.e., 5 hours). No substantial differences were observed in the
CSF Ab, sAPPb, and sAPPa profiles when a CSF Tmax of 2 hours versus 5 hours
was used; therefore, simulations were performed with a CSF Tmax of 2 hours.
Results
In Vitro Data. GNE-629 and GNE-892 exhibited potent BACE1
inhibition in the recombinant human BACE1 enzymatic assay with
IC50 values of 0.010 and 0.048 mM, and in an APP cellular assay with
IC50 values of 0.0080 and 0.027 mM, respectively (Table 1). Both
compounds showed relatively low protein binding with unbound
fraction of 0.39 for GNE-629 and 0.55 for GNE-892. Their permeability efflux ratios (permeability from basolateral side to apical
side versus permeability from apical to basolateral side, Pba/Pab) in
P-glycoprotein (P-gp) transfected cells (MDCKI-MDR1) were 44 for
GNE-629 and 25 for GNE-892, indicating that they are both P-gp
substrates (Table 1).
GNE-629 Plasma and CSF PK in Monkeys: Experimental
Observation and PK Modeling. Similar concentration-time profiles
were observed in monkey plasma and CSF after a single dose of GNE629 at 100 mg/kg (Fig. 3, A and B). The plasma and CSF Tmax values
were approximately 2 hours. CSF concentrations of GNE-629 were
notably lower relative to the corresponding unbound plasma concentrations, resulting in a CSF to unbound plasma AUC ratio of 0.0348 6
0.0017, indicating the blood-brain barrier (BBB) limits GNE-629 brain
penetration as predicted by the in vitro P-gp assay (Table 2).
As shown in Fig. 3, A and B, GNE-629 concentrations were well
characterized by the PK model described in Fig. 2A. The predicted
plasma and CSF concentrations matched the observed concentrations
as demonstrated by a correlation coefficient of 0.99 (Supplemental Fig.
1A). The PK parameters associated with the GNE-629 PK modeling
were estimated with good precision, generating a percent standard error
for all parameters of less than 30% (Table 3). An approximately 30-fold
higher distribution clearance from the CSF to central compartment
relative to the reverse distribution clearance from the central compartment to CSF was observed. This result is consistent GNE-629 being
effluxed by P-gp at the BBB.
GNE-629 Plasma and CSF PD in Monkeys: Experimental
Observation and Mechanistic PK-PD Modeling. The baseline for
plasma Ab was 0.238 6 0.037 ng/ml and the baseline for CSF Ab,
sAPPa, and sAPPb was 4.21 6 1.78 ng/ml, 32.0 6 14.5 ng/ml, and
34.5 6 16.8 ng/ml, respectively. GNE-629 significantly reduced the
Ab in the plasma and CSF after a single oral dose at 100 mg/kg from
4 to 72 postdosing (P , 0.001) (Fig. 3, C and D). The maximal APPrelated biomarker changes in monkeys are listed in Table 2. Similar
maximal Ab reduction was observed in the plasma and the CSF
(31.7–35.1% of the baseline or 64.9–68.3% reduction from the
baseline). The extent of reductions in CSF sAPPb and Ab were similar
as well. The maximal sAPPa increased to 143% of the baseline. The
average Tmax for biomarker changes ranged from 8 to 24 hours, which
was notably longer than the observed Tmax for plasma and CSF
concentrations of GNE-629 (2 hours) (Fig. 3).
Inhibition of plasma Ab formation by GNE-629 was well characterized by an indirect response PD model (Figs. 2D and 3C) (Jusko and Ko,
1994). The mechanistic CSF PD model (Fig. 2C) was able to describe
both the decrease in CSF Ab and sAPPb levels as well as the increase
in CSF sAPPa levels (Fig. 3, D–F). PD parameters were estimated
with good precision, all showing a percent standard error of less than
35% (Table 3). The estimated in vivo CSF IC50 for GNE-629 (0.0033
Downloaded from dmd.aspetjournals.org at ASPET Journals on May 7, 2017
dAbplasma
fu C 1
¼ kinplasma ð1 2
Þ 2 koutplasma Abplasma
dt
IC50ðplasmaÞ þ fu C1
Concentrations of LY2811376 in CSF after a 30-mg dose were extrapolated
from the simulated CSF concentrations for the 90-mg dose, assuming linear
pharmacokinetics. The simulated CSF concentrations of LY2811376 at the two
doses were used to simulate the PD effects of LY2811376 using the GNE-629
PD model after making adjustments for the in vivo GNE-629 IC50 value to
account for the 3-fold lower potency of LY2811376 relative to GNE-629
(Table 1). The adjusted in vivo CSF IC50 for LY2811376 was set to 0.010 mM.
1324
Liu et al.
mM) was in the same range as the IC50 values from the cellular assay
(0.0080 mM). The estimated unbound plasma IC50(plasma) (0.065 mM)
was approximately 20-fold higher than the CSF IC50. Overall, the
proposed mechanistic PK-PD model adequately described the plasma
Ab, and CSF Ab, sAPPa, and sAPPb levels in cynomolgus monkeys.
The predicted PD biomarkers in the plasma and CSF correlated well
with the observed values showing a correlation coefficient of 0.86
(Supplemental Fig. 1B).
GNE-892 Plasma and CSF PK in Monkeys: Experimental
Observation and PK Modeling. Plasma and CSF concentration-time
profiles for GNE-892 were similar after a single oral dose of 30 mg/kg
or 100 mg/kg (Fig. 4, A and B). The plasma and CSF Tmax values
were similar and ranged from 2 to 4 hours at 30 and 100 mg/kg. The
CSF concentrations again were notably lower than the unbound
plasma concentrations as demonstrated by the CSF to unbound plasma
AUC ratios of 0.0408 6 0.0036 for 30 mg/kg and 0.0423 6 0.032 for
100 mg/kg, indicating that the BBB limits GNE-892 brain penetration
as predicted by in vitro P-gp assay (Table 2).
GNE-892 plasma and CSF concentrations (Fig. 4, A and B) were
well described by its PK model as shown in Fig. 2B. The addition of
a central nervous system (CNS) depot compartment was necessary to
accurately capture the CSF concentrations of GNE-892 at the later
time points as they were substantially underpredicted when the GNE629 PK model (Fig. 2A) was used (unpublished data). PK parameters
for GNE-892 were estimated with good precision, generating a percent
standard error for all parameters of less than 20% (Table 4). The
predicted plasma and CSF concentrations correlated well with the
experimentally measured results, with a correlation coefficient of 0.96
(Supplemental Fig. 1C). Distribution clearance from CSF to the
central compartment was approximately 21-fold higher compared with
the distribution clearance from the central compartment to CSF for
GNE-892, similar to what was observed for GNE-629 (Table 4).
Downloaded from dmd.aspetjournals.org at ASPET Journals on May 7, 2017
Fig. 3. GNE-629 significantly inhibited BACE1 in cynomolgus monkeys. Monkey plasma (A) and CSF (B) concentrations of GNE-629, and levels of plasma Ab (C), CSF
Ab (D), CSF sAPPb (E), and CSF sAPPa (F) after a single 100 mg/kg oral dose of GNE-629 are presented. The symbols represent the observations (mean 6 S.D.). n = 9–10
from 0–48 hours and n = 4–5 at 72 hours. *P , 0.05; **P , 0.01; and ***P , 0.001 in comparison with the corresponding vehicle controls. The lines in (A) and (B)
represent the best fit of the PK model in Fig. 2A to the observed plasma and CSF concentrations, and lines in (C–F) represent the best fit of the PD model in Fig. 2, C and D,
to the observed PD markers.
1325
PK-PD Modeling of BACE1 Inhibition
TABLE 2
Summary of concentration and activity from monkey studies with GNE-629 and GNE-892
Data are presented as the mean 6 S.E.
GNE-629
GNE-892
Parameter
100 mg/kg
Plasma AUC (mM*h)
CSF AUC (mM*h)
CSF/unbound plasma AUC ratio
Minimal plasma Ab (% baseline)
Minimal CSF Ab (% baseline)
Maximal CSF sAPPa (% baseline)
Minimal CSF sAPPb (% baseline)
63.5
0.870
0.0348
31.7
35.1
143
29.4
6
6
6
6
6
6
6
30 mg/kg
11.5
0.164
0.0017
2.46
3.5
7
4.0
Discussion
In the present study, we report that two novel and potent BACE1
inhibitors, GNE-629 and GNE-892, significantly reduced Ab levels in
6
6
6
6
6
6
6
2.1
0.075
0.0036
4.4
5.0
25
6.3
100 mg/kg
138
2.95
0.0423
24.6
29.2
148
31.2
6
6
6
6
6
6
6
21
0.38
0.0032
5.4
2.5
14
3.6
plasma and CSF, and modulate other biomarkers of APP processing
such as sAPPa and sAPPb after a single dose oral administration in
monkeys. The observed changes in these biomarkers are consistent
with selective BACE1 inhibition in the brain. Therefore, these two
compounds can be used as tool compounds to study BACE1 inhibition
in monkeys. For the first time, a mechanism-based PK-PD model has
been developed to characterize plasma Ab and CSF Ab, sAPPa, and
sAPPb simultaneously. The modeling revealed that the free plasma
IC50 was approximately 20-fold of the CSF IC50, suggesting that Ab
in the plasma and brain may be formed by different mechanisms.
CSF Ab has been used as a biomarker for Ab in the brain. For
GNE-629, we observed that reduction of Ab1-40 in CSF was
associated with the reduction of Ab1-40 in the brain in guinea pigs but
the extent of reduction was generally greater in the CSF than in the
brain (unpublished data). Similar results were reported in mouse and
rats after other BACE1 inhibitors or g-secretase inhibitors (Lu et al.,
2011, 2012; Tai et al., 2012; Weiss et al., 2012). Recently, Lu et al.
(2012) demonstrated that the PK/PD relationship for the Ab lowering
in CSF was predictive of Ab lowering in the brain in mice, rats, and
guinea pigs for BACE1 or g-secretase inhibitors and modulators and
the extent of Ab lowering was greater in the CSF than in the brain.
There are several Ab peptides resulting from the sequential cleavage
TABLE 3
Summary of PK-PD model parameters estimated from cynomolgus monkey studies
with GNE-629
Parameter
PK Parameter
ka (h21)
ke (h21)
k12 (h21)
k21 (h21)
k1-csf (h21)
kcsf-1 (h21)
Vcsf (l/kg)
V1 (l/kg)
fu
CL1-csf (l/h per kilogram)b
CLcsf-1 (l/h per kilogram)b
Ratio CLcsf-1/CL1-csfb
PD Parameter for CSF PD Model
kinAPP (%/h)
k (h21)
ka (h21)
IC50 (mM)
PD Parameter for Plasma PD Model
kinplasma (%/h)
IC50(plasma) (mM)
Estimate (% S.E.)a
1.24
2.00
1.52
4.53
0.812
0.352
0.164
0.0827
1025
2.27
1023
12.3
0.39
1024
1023
29.7
(7.0)
(3.4)
(8.2)
(28)
(16)
(16)
(fixed)
(fixed)
(fixed)
55.7
0.140
0.411
0.00328
(26)
(6.8)
(34)
(13)
17.7 (7.8)
0.0654 (13)
a
Fitted parameters are expressed as the estimate followed by the % standard error in
parentheses.
b
Parameters are calculated as follows: CL1-csf = V1 k1-csf; CLcsf-1 = Vcsf kcsf-1.
Downloaded from dmd.aspetjournals.org at ASPET Journals on May 7, 2017
GNE-892 Plasma and CSF PD in Monkeys: Experimental
Observation and PK-PD Simulation. GNE-892 significantly reduced
Ab in plasma and CSF after a single oral dose at 30 mg/kg or 100 mg/kg
from 4 to 24 hours (Fig. 4, C and D) (P , 0.01). It also reduced
sAPPb and increased sAPPa in CSF (Fig. 4, E and F) (P , 0.05).
Similar maximal Ab reduction was observed in the plasma for the
30 and 100 mg/kg doses (23.4–24.6% of the baseline or 75.4–76.6%
reduction from the baseline) (Table 2). The biomarker changes in the
CSF were generally dose dependent. The maximal Ab reduction in CSF
was 48.6% (51.4% reduction from the baseline) of the baseline at 30
mg/kg and 29.2% of the baseline (71.8% reduction from the baseline) at
100 mg/kg. The levels of CSF sAPPb and Ab after administration of
GNE-892 were reduced to a similar extent. The maximal increases in
sAPPa levels in the CSF were 119% of the baseline at 30 mg/kg and
148% of the baseline at 100 mg/kg. The Tmax values for the biomarker
changes (8–24 hours) were notably longer than the Tmax values for the
plasma and CSF drug concentrations (2–4 hours) (Fig. 4).
Simulations of APP-related biomarker changes in plasma and CSF for
GNE-892–treated monkeys were made using the mechanistic PK-PD
model developed with GNE-629 data to validate its performance. As
detailed in the Materials and Methods, the in vivo CSF and plasma IC50
values for GNE-892 were adjusted from the corresponding in vivo GNE629 IC50 values (0.00328 mM in CSF and 0.0654 mM in plasma;
Table 3) to account for the decrease in cell potency for GNE-892 (0.0080
versus 0.027 mM; Table 1). The in vivo CSF IC50 for GNE-892 was set
to 0.0111 mM and the plasma IC50 was set to 0.221 mM. The remaining
system-dependent PD parameters were fixed to those obtained from the
GNE-629 PD model parameters (Table 3). Simulations of monkey Ab,
sAPPb, and sAPPa in CSF and Ab in plasma at doses of 30 and 100
mg/kg GNE-629 are shown in Fig. 4, C–F. The simulations captured the
dose dependence of the PD responses for all of the PD markers; the
decreases in CSF Ab, CSF sAPPb, and plasma Ab; and the increases in
CSF sAPPa. Overall, the CSF and plasma PD levels from simulation
were in good agreement with the observed concentrations in the monkey
showing a correlation of coefficient of 0.85 (Supplemental Fig. 1D).
LY2811376 CSF PD in Humans: Experimental Observation
and Mechanistic PK-PD Simulation. The utility of the mechanistic
PK-PD model developed for GNE-629 was further investigated by
testing its ability to predict the PD biomarker changes observed in
human CSF for LY2811376 after oral administration of 30 and 90 mg.
The predicted maximum reductions of CSF Ab for the 30-mg and 90mg doses were approximately 61 and 81%, respectively, and were
comparable with the observed average maximum changes of 53–59%
at 30 mg and 82–91% at 90 mg (May et al., 2011).
25.2
0.571
0.0408
23.4
48.6
119
47.3
1326
Liu et al.
of APP by b-secretase and g-secretase. A good concordance between
CSF Ab1-40 and Ab1-42 has been demonstrated in the literature for
other rodents, monkeys, and humans with BACE1 inhibitors (Hussain
et al., 2007; Sankaranarayanan et al., 2009; Fukumoto et al., 2010;
May et al., 2011). Therefore, we used Ab1-40 to represent the overall
amyloid peptides based on its relative abundance.
GNE-629 and GNE-892 Significantly Inhibit Brain BACE1 in
Monkeys after a Single Oral Dose. The present study demonstrated
that GNE-629 at 100 mg/kg inhibited monkey Ab by 64.9–68.3% in
the plasma and CSF and GNE-892 at 30 or 100 mg/kg inhibited
monkey Ab by 75.4–76.6% in the plasma and 51.4–71.8% in the CSF.
Since sAPPb is a direct product of b-secretase cleavage for APP, it
would be expected that BACE1 inhibition also would lead to
a reduction of sAPPb levels and accumulation of full-length APP in
brain, potentially an increase of sAPPa via a-secretase cleavage. Our
data support this prediction because both GNE-629 and GNE-892
led to increased sAPPa levels, correlating with decreased sAPPb in
monkey CSF. The observed changes in multiple APP-related biomarkers
demonstrate both inhibitors inhibit CNS BACE1 activity and can be
used as tool compounds to study BACE1 inhibition in nonhuman
primates.
BACE1 Inhibition can be Characterized by a Mechanistic PKPD model. The development of new treatments for AD is challenging
for many reasons, including the time lag between the disease onset and
clinical symptoms, unreliability of outcomes measured in clinical
trials, and the lack of understanding of relationships between drug
effects and disease in preclinical models and in humans (Becker and
Greig, 2008). Therefore, the discovery and development of new
Alzheimer’s medicines will benefit from an improved understanding
of PK-PD relationships through the development of mechanistic PKPD models. Several of the PK-PD models for APP processing largely
rely on the effects of g-secretase inhibition on Ab levels (Craft et al.,
2002; Lu et al., 2011; Tai et al., 2012). Recently, Lu et al. (2012)
reported a PK-PD model for Ab lowering in CSF and brain in rodents.
Downloaded from dmd.aspetjournals.org at ASPET Journals on May 7, 2017
Fig. 4. GNE-892 significantly inhibited BACE1 in cynomolgus monkeys. Monkey plasma (A) and CSF (B) concentrations of GNE-892, as well as levels of plasma Ab (C),
CSF Ab (D), CSF sAPPb (E), and CSF sAPPa (F) after a 30 or 100 mg/kg oral dose of GNE-892 are presented. The symbols represent the observations (mean 6 S.D.) n =
12–13 for plasma Ab and CSF Ab, n = 7 for CSF sAPPb and sAPPa. *P , 0.05; **P , 0.01; ***P , 0.001 in comparison with the corresponding vehicle controls. The
lines in (A) and (B) represent the best fit of the PK model in Fig. 2B to the plasma and CSF concentrations. The lines in (C–F) are the simulated levels of plasma Ab and CSF
Ab, sAPPa, and sAPPb using PD models in Fig. 2, C and D.
PK-PD Modeling of BACE1 Inhibition
TABLE 4
Summary of pharmacokinetic parameters estimated from cynomolgus monkey
studies with GNE-892
PK Parameter
ka (h21)
ke (h21)
k12 (h21)
k21 (h21)
k1-csf (h21)
kcsf-1 (h21)
kcsf-csfp (h21)
kcsfp-csf (h21)
Vcsf (l/kg)
V1 (l/kg)
fu
CL1-csf (l/h per kilogram)b
CLcsf-1 (l/h per kilogram)b
Ratio CLcsf-1/CL1-csfb
Estimate (% S.E.)a
1.87
2.00
1.17
2.49
1.00
0.608
0.0284
0.0783
1025
1.25
0.692
0.100
1023
6.23
0.55
1024
1023
21.3
(15)
(8.1)
(11)
(3.6)
(18)
(12)
(16)
(5.5)
(fixed)
(8.6)
(fixed)
To our knowledge, no models have been reported that detail the effects
of BACE1 inhibition on overall APP processing including Ab,
sAPPa, and sAPPb. Although the PK-PD modeling was based on the
GNE-629 dataset, an exploratory modeling exercise showed that use
of the GNE-892 dataset generated similar PK/PD model parameters.
Our PK-PD model was able to characterize the responses of the APPrelated PD markers in both plasma and CSF. Our modeling work
revealed the kinetics of Ab in the monkeys. The estimated elimination
rate constant for the CSF and plasma Ab was 0.140 and 0.177 h21,
respectively, resulting in an Ab half-life of 5 and 4 hours, respectively.
These values are consistent with the reported Ab elimination rate
constant (0.08 h21) and half-life (8.3 hours) in monkeys (Cook et al.,
2010). A similar elimination rate constant was also observed in humans
(Bateman et al., 2006).
The model estimated an in vivo CSF IC50 of 0.0033 mM which is
similar to GNE-629 cell-based BACE1 inhibition IC50 of 0.0080 mM.
Interestingly, the free plasma IC50 for GNE-629 was 0.065 mM, a near
20-fold of the in vivo CSF IC50. In the PD model, it was assumed that
100% Ab in the plasma and brain can be inhibited by BACE1
inhibition. If we assume that only 60% of Ab in the plasma can be
reduced via BACE1 inhibition, the free plasma IC50 for GNE-629 is
0.0056 mM, which is more in line with in vivo CSF and in vitro cellbased IC50s. These results suggest that Ab in the brain may be mainly
generated by BACE1 but Ab in plasma may be generated by BACE1
and other mechanisms. This hypothesis is supported by recent
observations that Ab in the brain of BACE1/2 double knockout mice
was reduced by more than 90% relative to wild-type littermates, but
Ab in the plasma was only reduced to 50–60% of wild-type levels
(William J. Meilandt et al., submitted manuscript). Previous observations
that only approximately 50% plasma Ab can be inhibited by high
concentrations of an anti-BACE1 antibody in monkeys also support
this hypothesis (Atwal et al., 2011).
The general utility of this PD model built with GNE-629 data were
tested by performing simulations of the in vivo PD markers for
another BACE1 inhibitor, GNE-892, following simple adjustments to
account for differences in compound in vitro potency and PK. For
GNE-892, a CSF peripheral compartment (Fig. 2B) was added to the
CSF compartment to accurately describe the CSF concentrations in the
terminal phase. Potentially slower distribution between plasma, brain,
and CSF compartments may contribute to this observation. A limitation
of the GNE-892 PK-PD data set was that the last sampling time
point was 48 hours where the PD biomarkers had not yet returned to
the baseline. Despite this, the simulated biomarker changes in the
plasma and CSF compartments matched the observed values reasonably
well. We also evaluated whether the monkey PD model can predict the
CSF biomarkers for LY2811376 in humans after potency adjustments
using its cell IC50 generated under the same condition as GNE-629 in
the authors’ laboratory, and its CSF concentrations reported by May
et al. (2011). Our model predicted maximal CSF Ab reductions were
61% at 30 mg and 81% at 90 mg. These values are similar to the clinical
observed maximal CSF Ab reductions of 53–59% at 30 mg and 82–91%
at 90 mg.
GNE-892 and GNE-629 Significantly Inhibit CNS BACE1
Despite P-gp Efflux in Monkeys. Both GNE-629 and GNE-892 are
P-gp substrates as indicated by their high efflux ratios in P-gp
transfected cells. P-gp at the BBB likely impaired brain penetration of
these two compounds as demonstrated by the low CSF to free plasma
concentration ratios in monkeys as well as approximately 21- to 30fold higher distribution clearance from the CSF compartment to the
central compartment than that from the central compartment to
the CSF compartment. This is supported by the observations that the
brain-to-plasma AUC(0–1 h) ratios increased approximately 11-fold for
GNE-629 and 7-fold for GNE-892, in the Mdr1a/1b knockout or
Mdr1a/1b/Bcrp knockout mice compared with the wild-type mice.
There were not significant changes observed in the Bcrp knockout
mice (Supplemental Figs. 2 and 3). These observations are also in
agreement with previous findings that P-gp substrates have low CSF
to unbound plasma concentration ratios (Liu et al., 2006, 2008, 2009).
Although P-gp at the BBB limits the brain penetration of GNE-629
and GNE-892, sufficient concentrations were achieved to inhibit
BACE1 in the brain. Thus, P-gp is just one of many hurdles to overcome
in designing CNS drugs. If a P-gp substrate is sufficiently potent and
selective against its target, it still may be useful as a CNS drug (de Leon
et al., 2010).
In summary, two novel and potent BACE1 inhibitors, GNE-629 and
GNE-892, are capable of eliciting significant inhibition of Ab formation
in the monkey plasma and CSF after oral administration and can be used
as tool compounds to study BACE1 inhibition. A novel mechanismbased PK-PD model was developed to provide a better understanding of
the relationships between drug concentrations and BACE1 biomarkers
in plasma and CSF. This modeling suggests the potential of different
mechanisms being responsible for Ab formation in the plasma and brain.
This work illustrates the utility of a mechanistic PK-PD model in
quantitatively understanding the biologic processes as well as in the
translation of in vitro activities to in vivo activities in drug discovery.
Acknowledgments
The authors thank Emile Plise, Jonathan Cheong, Quynh Ho, and Mary Geck
Do for their support of in vitro studies, and Sonali Medakkar and Daniel
Hutcheson at Maccine Pte Ltd and Eric Bourhis for their support of in vivo studies.
Authorship Contributions
Participated in research design: Liu, Scearce-Levie, Watts, Regal, Hunt,
Siu, Lyssikatos, Hop.
Conducted experiments: Liu, Coraggio, Shin, Peng, Wildsmith, Atwal,
Mango, Schauer, Deshmukh.
Contributed new reagents or analytic tools: Hunt, Thomas, Siu, Lyssikatos.
Performed data analysis: Liu, Wong, Atwal, Deshmukh.
Wrote or contributed to the writing of the manuscript: Liu, Wong, ScearceLevie, Watts, Regal, Hop.
References
Atwal JK, Chen Y, Chiu C, Mortensen DL, Meilandt WJ, Liu Y, Heise CE, Hoyte K, Luk W,
and Lu Y, et al. (2011) A therapeutic antibody targeting BACE1 inhibits amyloid-b production
in vivo. Sci Transl Med 3:84ra43.
Downloaded from dmd.aspetjournals.org at ASPET Journals on May 7, 2017
a
Fitted parameters are expressed as the estimate followed by the % standard error in
parentheses.
b
Parameters are calculated as follows: CL1-csf = V1 k1-csf; CLcsf-1 = Vcsf kcsf-1.
1327
1328
Liu et al.
biomarker for brain Ab lowering verified by quantitative preclinical analyses. J Pharmacol Exp
Ther 342:366–375
Lu Y, Zhang L, Nolan CE, Becker SL, Atchison K, Robshaw AE, Pustilnik LR, Osgood SM,
Miller EH, and Stepan AF, et al. (2011) Quantitative pharmacokinetic/pharmacodynamic
analyses suggest that the 129/SVE mouse is a suitable preclinical pharmacology model for
identifying small-molecule g-secretase inhibitors. J Pharmacol Exp Ther 339:922–934.
Masters CL, Simms G, Weinman NA, Multhaup G, McDonald BL, and Beyreuther K (1985)
Amyloid plaque core protein in Alzheimer disease and Down syndrome. Proc Natl Acad Sci
USA 82:4245–4249.
May PC, Dean RA, Lowe SL, Martenyi F, Sheehan SM, Boggs LN, Monk SA, Mathes BM,
Mergott DJ, and Watson BM, et al. (2011) Robust central reduction of amyloid-b in humans
with an orally available, non-peptidic b-secretase inhibitor. J Neurosci 31:16507–16516.
McConlogue L, Buttini M, Anderson JP, Brigham EF, Chen KS, Freedman SB, Games D,
Johnson-Wood K, Lee M, and Zeller M, et al. (2007) Partial reduction of BACE1 has dramatic
effects on Alzheimer plaque and synaptic pathology in APP Transgenic Mice. J Biol Chem
282:26326–26334.
Muraszko K, Sung C, Walbridge S, Greenfield L, Dedrick RL, Oldfield EH, and Youle RJ (1993)
Pharmacokinetics and toxicology of immunotoxins administered into the subarachnoid space in
nonhuman primates and rodents. Cancer Res 53:3752–3757.
Poplack DG, Bleyer WA, Wood JH, Kostolich M, Savitch JL, and Ommaya AK (1977) A primate
model for study of methotrexate pharmacokinetics in the central nervous system. Cancer Res
37:1982–1985.
Probst G and Xu YZ (2012) Small-molecule BACE1 inhibitors: a patent literature review (2006 2011). Expert Opin Ther Pat 22:511–540.
Rubenstein JL, Combs D, Rosenberg J, Levy A, McDermott M, Damon L, Ignoffo R, Aldape K,
Shen A, and Lee D, et al. (2003) Rituximab therapy for CNS lymphomas: targeting the
leptomeningeal compartment. Blood 101:466–468.
Sankaranarayanan S, Holahan MA, Colussi D, Crouthamel MC, Devanarayan V, Ellis J, Espeseth
A, Gates AT, Graham SL, and Gregro AR, et al. (2009) First demonstration of cerebrospinal
fluid and plasma A beta lowering with oral administration of a beta-site amyloid precursor
protein-cleaving enzyme 1 inhibitor in nonhuman primates. J Pharmacol Exp Ther 328:
131–140.
Sinha S and Lieberburg I (1999) Cellular mechanisms of beta-amyloid production and secretion.
Proc Natl Acad Sci USA 96:11049–11053.
Tai LM, Jacobsen H, Ozmen L, Flohr A, Jakob-Roetne R, Caruso A, and Grimm HP (2012) The
dynamics of Ab distribution after g-secretase inhibitor treatment, as determined by experimental and modelling approaches in a wild type rat. J Pharmacokinet Pharmacodyn 39:
227–237.
van der Graaf PH and Benson N (2011) Systems pharmacology: bridging systems biology and
pharmacokinetics-pharmacodynamics (PKPD) in drug discovery and development. Pharm Res
28:1460–1464.
Weiss MM, Williamson T, Babu-Khan S, Bartberger MD, Brown J, Chen K, Cheng Y, Citron M,
Croghan MD, and Dineen TA, et al. (2012) Design and preparation of a potent series of
hydroxyethylamine containing b-secretase inhibitors that demonstrate robust reduction of
central b-amyloid. J Med Chem 55:9009–9024.
Wong H, Choo EF, Alicke B, Ding X, La H, McNamara E, Theil FP, Tibbitts J, Friedman LS,
and Hop CE, et al. (2012a) Antitumor activity of targeted and cytotoxic agents in murine
subcutaneous tumor models correlates with clinical response. Clin Cancer Res 18:3846–3855.
Wong H, Vernillet L, Peterson A, Ware JA, Lee L, Martini JF, Yu P, Li C, Del Rosario G,
and Choo EF, et al. (2012b) Bridging the gap between preclinical and clinical studies using
pharmacokinetic-pharmacodynamic modeling: an analysis of GDC-0973, a MEK inhibitor.
Clin Cancer Res 18:3090–3099.
Younkin SG (1995) Evidence that A beta 42 is the real culprit in Alzheimer’s disease. Ann Neurol
37:287–288.
Zhu Y, Xiao K, Ma L, Xiong B, Fu Y, Yu H, Wang W, Wang X, Hu D, and Peng H, et al. (2009)
Design, synthesis and biological evaluation of novel dual inhibitors of acetylcholinesterase and
beta-secretase. Bioorg Med Chem 17:1600–1613.
Address correspondence to: Dr. Xingrong Liu, Genentech Inc., MS 41-2A, 1
DNA Way, South San Francisco, CA 94080. E-mail: [email protected]
Downloaded from dmd.aspetjournals.org at ASPET Journals on May 7, 2017
Bateman RJ, Munsell LY, Morris JC, Swarm R, Yarasheski KE, and Holtzman DM (2006)
Human amyloid-beta synthesis and clearance rates as measured in cerebrospinal fluid in vivo.
Nat Med 12:856–861.
Becker RE and Greig NH (2008) Alzheimer’s disease drug development: old problems require
new priorities. CNS Neurol Disord Drug Targets 7:499–511.
Citron M (2010) Alzheimer’s disease: strategies for disease modification. Nat Rev Drug Discov 9:
387–398.
Cook JJ, Wildsmith KR, Gilberto DB, Holahan MA, Kinney GG, Mathers PD, Michener MS,
Price EA, Shearman MS, and Simon AJ, et al. (2010) Acute gamma-secretase inhibition of
nonhuman primate CNS shifts amyloid precursor protein (APP) metabolism from amyloid-beta
production to alternative APP fragments without amyloid-beta rebound. J Neurosci 30:
6743–6750.
Craft DL, Wein LM, and Selkoe DJ (2002) A mathematical model of the impact of novel
treatments on the A beta burden in the Alzheimer’s brain, CSF and plasma. Bull Math Biol 64:
1011–1031.
de Leon J, Wynn G, and Sandson NB (2010) The pharmacokinetics of paliperidone versus
risperidone. Psychosomatics 51:80–88.
Dineen TA, Weiss MM, Williamson T, Acton P, Babu-Khan S, Bartberger MD, Brown J, Chen
K, Cheng Y, and Citron M, et al. (2012) Design and synthesis of potent, orally efficacious
hydroxyethylamine derived b-site amyloid precursor protein cleaving enzyme (BACE1)
inhibitors. J Med Chem 55:9025–9044.
Elvang AB, Volbracht C, Pedersen LO, Jensen KG, Karlsson JJ, Larsen SA, Mørk A, Stensbøl
TB, and Bastlund JF (2009) Differential effects of gamma-secretase and BACE1 inhibition on
brain Abeta levels in vitro and in vivo. J Neurochem 110:1377–1387.
Fukumoto H, Takahashi H, Tarui N, Matsui J, Tomita T, Hirode M, Sagayama M, Maeda R,
Kawamoto M, and Hirai K, et al. (2010) A noncompetitive BACE1 inhibitor TAK-070
ameliorates Abeta pathology and behavioral deficits in a mouse model of Alzheimer’s disease.
J Neurosci 30:11157–11166.
Gabrielsson J and Green AR (2009) Quantitative pharmacology or pharmacokinetic pharmacodynamic integration should be a vital component in integrative pharmacology. J Pharmacol
Exp Ther 331:767–774.
Gibbs JP (2010) Prediction of exposure-response relationships to support first-in-human study
design. AAPS J 12:750–758.
Glenner GG and Wong CW (1984) Alzheimer’s disease: initial report of the purification and
characterization of a novel cerebrovascular amyloid protein. Biochem Biophys Res Commun
120:885–890.
Hardy J and Selkoe DJ (2002) The amyloid hypothesis of Alzheimer’s disease: progress and
problems on the road to therapeutics. Science 297:353–356.
Hussain I, Hawkins J, Harrison D, Hille C, Wayne G, Cutler L, Buck T, Walter D, Demont E,
and Howes C, et al. (2007) Oral administration of a potent and selective non-peptidic BACE-1
inhibitor decreases beta-cleavage of amyloid precursor protein and amyloid-beta production in
vivo. J Neurochem 100:802–809.
Jeppsson F, Eketjäll S, Janson J, Karlström S, Gustavsson S, Olsson LL, Radesäter AC, Ploeger
B, Cebers G, and Kolmodin K, et al. (2012) Discovery of AZD3839, a potent and selective
BACE1 inhibitor clinical candidate for the treatment of Alzheimer disease. J Biol Chem 287:
41245–41257.
Jusko WJ and Ko HC (1994) Physiologic indirect response models characterize diverse types of
pharmacodynamic effects. Clin Pharmacol Ther 56:406–419.
Kandalepas PC and Vassar R (2012) Identification and biology of b-secretase. J Neurochem 120
(Suppl 1):55–61.
Lesné S, Koh MT, Kotilinek L, Kayed R, Glabe CG, Yang A, Gallagher M, and Ashe KH (2006)
A specific amyloid-beta protein assembly in the brain impairs memory. Nature 440:352–357.
Liu X, Chen C, and Smith BJ (2008) Progress in brain penetration evaluation in drug discovery
and development. J Pharmacol Exp Ther 325:349–356.
Liu X, Smith BJ, Chen C, Callegari E, Becker SL, Chen X, Cianfrogna J, Doran AC, Doran SD,
and Gibbs JP, et al. (2006) Evaluation of cerebrospinal fluid concentration and plasma free
concentration as a surrogate measurement for brain free concentration. Drug Metab Dispos 34:
1443–1447.
Liu X, Van Natta K, Yeo H, Vilenski O, Weller PE, Worboys PD, and Monshouwer M (2009)
Unbound drug concentration in brain homogenate and cerebral spinal fluid at steady state as
a surrogate for unbound concentration in brain interstitial fluid. Drug Metab Dispos 37:787–793.
Lu Y, Riddell D, Hajos-Korcsok E, Bales K, Wood KM, Nolan CE, Robshaw AE, Zhang L,
Leung L, and Becker SL, et al. (2012) Cerebrospinal fluid amyloid-b (Ab) as an effect