Dipeptidyl peptidase-4 greatly contributes to the hydrolysis of vildagliptin in human liver

Article abstract of DOI:10.1124/dmd.114.062331

The major metabolic pathway of vildagliptin in mice, rats, dogs, and humans is hydrolysis at the cyano group to produce a carboxylic acid metabolite M20.7 (LAY151), whereas the major metabolic enzyme of vildagliptin has not been identified. In the present study, we determined the contribution rate of dipeptidyl peptidase-4 (DPP-4) to the hydrolysis of vildagliptin in the liver. We performed hydrolysis assay of the cyano group of vildagliptin using mouse, rat, and human liver samples. Additionally, DPP-4 activities in each liver sample were assessed by DPP-4 activity assay using the synthetic substrate H-glycyl-prolyl-7-amino-4-methylcoumarin (Gly-Pro-AMC). M20.7 formation rates in liver microsomes were higher than those in liver cytosol. M20.7 formation rate was significantly positively correlated with the DPP-4 activity using Gly-Pro-AMC in liver samples (r = 0.917, P < 0.01). The formation of M20.7 in mouse, rat, and human liver S9 fraction was inhibited by sitagliptin, a selective DPP-4 inhibitor. These findings indicate that DPP-4 is greatly involved in vildagliptin hydrolysis in the liver. Additionally, we established stable single expression systems of human DPP-4 and its R623Q mutant, which is the nonsynonymous single-nucleotide polymorphism of human DPP-4, in human embryonic kidney 293 (HEK293) cells to investigate the effect of R623Q mutant on vildagliptin-hydrolyzing activity. M20.7 formation rate in HEK293 cells expressing human DPP-4 was significantly higher than that in control HEK293 cells. Interestingly, R623Q mutation resulted in a decrease of the vildagliptin-hydrolyzing activity. Our findings might be useful for the prediction of interin-dividual variability in vildagliptin pharmacokinetics.

Full text of DOI:10.1124/dmd.114.062331

DMD Fast Forward. Published on January 16, 2015 as DOI: 10.1124/dmd.114.062331
This article has not been copyedited and formatted. The final version may differ from this version.
Dipeptidyl peptidase-4 greatly contributes to the hydrolysis of
vildagliptin in human liver
Mitsutoshi Asakura, Hideaki Fujii, Koichiro Atsuda, Tomoo Itoh, and Ryoichi Fujiwara
Graduate School of Pharmaceutical Sciences (M.A.) and School of Pharmacy, Kitasato
University, Tokyo, Japan (H.F., K.A., T.I., R.F.)

DMD Fast Forward. Published on January 16, 2015 as DOI: 10.1124/dmd.114.062331
This article has not been copyedited and formatted. The final version may differ from this version.
Running title: DPP-4 greatly contributes to the vildagliptin hydrolysis
Corresponding author:
Ryoichi Fujiwara, Ph.D., School of Pharmacy, Kitasato University, 5-9-1 Shirokane, Minato-ku,
Tokyo 108-8641, JAPAN, fujiwarar@pharm.kitasato-u.ac.jp, +81-3-5791-6249
Number of text pages:
Number of tables:
Number of figures:
Number of references:
Number of words:
Abstract:
45
2
6
29
242
Introduction: 782
Discussion:
1,120
Abbreviations: DPP-4, dipeptidyl peptidase-4; CYP, cytochrome P450; NADPH, nicotinamide
adenine dinucleotide phosphate; UGT, uridine diphosphate glucuronosyltransferase; NIT,
nitrilase-like protein; HEK293, human embryonic kidney 293; Gly-Pro-AMC,
H-glycyl-prolyl-7-amino-4-methylcoumarin;
GAPDH,
glyceraldehyde-3-phosphate
dehydrogenase; BNPP, bis(p-nitrophenyl) phosphate; SNP, single-nucleotide polymorphism

DMD Fast Forward. Published on January 16, 2015 as DOI: 10.1124/dmd.114.062331
This article has not been copyedited and formatted. The final version may differ from this version.
Abstract
The major metabolic pathway of vildagliptin in mice, rats, dogs, and humans is hydrolysis at
the cyano group to produce a carboxylic acid metabolite M20.7 (LAY151), while the major
metabolic enzyme of vildagliptin has not been identified. In the present study, we determined the
contribution rate of dipeptidyl peptidase-4 (DPP-4) to the hydrolysis of vildagliptin in the liver.
We performed hydrolysis assay of the cyano group of vildagliptin using mouse, rat, and human
liver samples. Additionally, DPP-4 activities in each liver sample were assessed by DPP-4
activity assay using the synthetic substrate Gly-Pro-AMC. M20.7 formation rates in liver
microsomes were higher than those in liver cytosol. M20.7 formation rate was significantly
positively correlated with the DPP-4 activity using Gly-Pro-AMC in liver samples (r = 0.917, P <
0.01). The formation of M20.7 in mouse, rat, and human liver S9 fraction was inhibited by
sitagliptin, a selective DPP-4 inhibitor. These findings indicate that DPP-4 is greatly involved in
vildagliptin hydrolysis in the liver. Additionally, we established stable single expression systems
of human DPP-4 and its R623Q mutant, which is the non-synonymous single nucleotide
polymorphism of human DPP-4, in HEK293 cells to investigate the effect of R623Q mutant on
vildagliptin-hydrolyzing activity. M20.7 formation rate in HEK293 cells expressing human

DMD Fast Forward. Published on January 16, 2015 as DOI: 10.1124/dmd.114.062331
This article has not been copyedited and formatted. The final version may differ from this version.
DPP-4 was significantly higher than that in control HEK293 cells. Interestingly, R623Q mutation
resulted in a decrease of the vildagliptin-hydrolyzing activity. Our findings might be useful for
the prediction of interindividual variability in vildagliptin pharmacokinetics.

DMD Fast Forward. Published on January 16, 2015 as DOI: 10.1124/dmd.114.062331
This article has not been copyedited and formatted. The final version may differ from this version.
Introduction
Dipeptidyl peptidase-4 (DPP-4, CD26, EC 3.4.14.5), a serine protease belonging to type II
transmembrane glycoproteins, is widely expressed on the surface of epithelial cells of diverse
tissues including liver, kidney, and intestine, on endothelial cells of blood vessels, and on
lymphoid cells (Mentlein, 1999; Gorrell et al., 2001). In addition to the integral membrane form,
a soluble form of DPP-4 presents in serum (Durinx et al., 2000). By cleaving dipeptides from the
N-terminal end of peptides and polypeptides with proline or alanine in the second position, DPP-4
controls the activity of many bioactive molecules, including incretins, cytokines, chemokines,
and neuropeptides (Boonacker and Van Noorden, 2003).
Vildagliptin (LAF237) is a potent, orally active inhibitor of DPP-4 for the treatment of type 2
diabetes mellitus (Villhauer et al., 2003). DPP-4 inhibitors, so-called incretin enhancers, are
attracting attention among therapeutic agents for type 2 diabetes mellitus, because they improve
glucose control with a low risk of hypoglycemia (Scheen, 2010a; Deacon, 2011). While most
DPP-4 inhibitors allow one single oral administration per day for management of type 2 diabetes
mellitus, twice-daily administration is recommended for vildagliptin because of its shorter
half-life (Deacon, 2011). Vildagliptin is extensively metabolized via at least four pathways before

DMD Fast Forward. Published on January 16, 2015 as DOI: 10.1124/dmd.114.062331
This article has not been copyedited and formatted. The final version may differ from this version.
excretion, with the major metabolite M20.7 (LAY151) resulting from cyano group hydrolysis (He
et al., 2009a). The parent compound and the major metabolite M20.7, which is pharmacologically
inactive, account for the majority of vildagliptin-related materials in human plasma
(approximately 25.7 and 55%, respectively) (He et al., 2009a). Minor metabolites, which resulted
from amide bond hydrolysis (M15.3), glucuronidation (M20.2), and oxidation on the pyrrolidine
moiety of vildagliptin (M20.9 and M21.6), have been identified (He et al., 2009a). While 85% of
an oral dose of vildagliptin is ultimately excreted by the kidney, the liver is the major site of
vildagliptin metabolism in humans (He et al., 2007). Although cytochrome P450 (CYP), which is
a superfamily of microsomal hemoproteins, catalyzes the reduced nicotinamide adenine
dinucleotide phosphate (NADPH)-dependent oxidative metabolism of numerous drugs (Meunier
et al., 2004), M20.7 was not formed by incubating vildagliptin with recombinant human CYP in
the presence of NADPH (He et al., 2009a). These findings suggest that clinically relevant
drug-drug interactions with vildagliptin via CYP are very unlikely (He et al., 2009a; Scheen,
2010b; He, 2012). In contrast, non-CYP enzymes, such as uridine diphosphate
glucuronosyltransferase (UGT) and esterase, have attracted recent attention in pharmacokinetics,
including metabolism and drug-drug interaction (Nagai, 2010; Akabane et al., 2012; Fukami and

DMD Fast Forward. Published on January 16, 2015 as DOI: 10.1124/dmd.114.062331
This article has not been copyedited and formatted. The final version may differ from this version.
Yokoi, 2012). Identification of the major metabolic enzyme responsible for vildagliptin
hydrolysis is important in predicting the individual differences in pharmacokinetics and is useful
for the prediction of drug-drug interactions with vildagliptin via non-CYP.
In mice, rats, dogs, and humans, the main metabolite of vildagliptin is M20.7 (He et al.,
2009b). In our previous study, we investigated whether human nitrilase-like protein (NIT) 1 and
NIT2, which are members of the nitrilase superfamily, are involved in the hydrolysis of
vildagliptin, because the hydrolysis of the cyano group of vildagliptin is similar to nitrilase
reaction (Asakura et al., 2014). However, the result of measurement of M20.7 formation rate in
human embryonic kidney 293 (HEK293) cells expressing human NIT1 or NIT2 suggested that
human NIT1 or NIT2 was not involved in the metabolism of vildagliptin. A previous report has
shown that based on findings from the in vivo metabolism data in DPP-4-deficient and
DPP-4-normal rats, approximately 20% of the cyano group hydrolysis reaction, which is the
formation of M20.7, may be attributable to DPP-4 (He et al., 2009b). Furthermore, M20.7 was
formed by incubation of vildagliptin with the recombinant human DPP-4 (He et al., 2009a).
Although these data raise the possibility that DPP-4 is partially involved in formation of M20.7 in
rats, contribution rate of DPP-4 to the vildagliptin hydrolysis in humans has not been clarified.

DMD Fast Forward. Published on January 16, 2015 as DOI: 10.1124/dmd.114.062331
This article has not been copyedited and formatted. The final version may differ from this version.
In the present study, we determined the contribution rate of DPP-4 to the hydrolysis of
vildagliptin in the liver, which is the major site of vildagliptin hydrolysis. First, we performed
hydrolysis assays of the cyano group of vildagliptin using mouse, rat, and human liver samples.
Additionally, the protein expression levels of DPP-4 in each liver sample were assessed by
western blot analysis. DPP-4 activities in each liver sample were also assessed by DPP-4 activity
assay using the synthetic substrate H-glycyl-prolyl-7-amino-4-methylcoumarin (Gly-Pro-AMC).
Second, to clarify the contribution rate of DPP-4 to the metabolism of vildagliptin, we assessed
inhibitory effect of sitagliptin, a selective DPP-4 inhibitor, on vildagliptin-hydrolyzing activity of
mouse, rat, and human liver S9 fractions. Finally, we established stable single expression systems
of human DPP-4 and its mutants (R623Q and S630A) in HEK293 cells to assess the
vildagliptin-hydrolyzing activity of human DPP-4 and to investigate the effect of the mutations
on vildagliptin-hydrolyzing activity.
Materials and methods
Chemicals and reagents

DMD Fast Forward. Published on January 16, 2015 as DOI: 10.1124/dmd.114.062331
This article has not been copyedited and formatted. The final version may differ from this version.
Pooled human liver S9 fraction (50 donors, mixed gender), pooled human liver cytosol (50
donors, mixed gender), pooled rat liver S9 fraction (105 male Fischer 344 rats), pooled rat liver
cytosol (105 male Fischer 344 rats), and pooled rat liver microsomes (105 male Fischer 344 rats)
were obtained from XenoTech, LLC (Lenexa, KS, USA). Pooled human liver microsomes (50
donors, mixed gender) was obtained from BD Gentest (Woburn, MA, USA). AMC was
purchased from Setareh Biotech, LLC (Eugene, OR, USA). Gly-Pro-AMC was obtained from
Bachem (Bubendorf, Switzerland). G418 was purchased from Nacalai Tesque (Kyoto, Japan).
Goat polyclonal anti-human DPP-4 antibody (AF1180) was purchased from R&D Systems
(Minneapolis, MN, USA). Mouse monoclonal anti-glyceraldehyde-3-phosphate dehydrogenase
(GAPDH) antibody (6C5) was purchased from American Research Products, Inc. (Belmont, MA,
USA). Rabbit polyclonal anti-actin antibody (A2103), bis(p-nitrophenyl) phosphate (BNPP), and
ethopropazine were purchased from Sigma-Aldrich (St. Louis, MO, USA). Vildagliptin was
synthesized in our laboratory using the standard technique (Asakura et al., 2014). Vildagliptin
carboxylic acid metabolite (M20.7) was purchased from Santa Cruz Biotechnology (Delaware
Avenue, CA, USA). Sitagliptin was obtained from LKT Laboratories (St. Paul, MN, USA). All
other chemicals were of the highest grade available.

DMD Fast Forward. Published on January 16, 2015 as DOI: 10.1124/dmd.114.062331
This article has not been copyedited and formatted. The final version may differ from this version.
Hydrolysis assays of the cyano group of vildagliptin
A typical incubation mixture (200 µL of total volume) contained 50 mM Tris-HCl buffer,
pH 7.4, vildagliptin (1 or 10 µM), and various enzyme sources. The final concentration of mouse,
rat, and human liver S9 fraction and cytosol was 1.0 mg/mL, respectively. The final concentration
of mouse, rat, and human liver microsomes and S9 fraction of HEK293 cells expressing human
DPP-4 or its mutants was 0.5 mg/mL, respectively. Mouse liver S9 fraction, cytosol, and
microsomes were prepared previously (Asakura et al., 2014). The reaction was initiated by adding
vildagliptin after a 5 min preincubation at 37 °C. After incubation at 37 °C for 2 h, the reaction
was terminated by an addition of 200 µL of cold acetonitrile. After removal of protein by
centrifugation at 15,000 g for 5 min, a 50-µL portion of the sample was subjected to liquid
chromatography/tandem mass spectrometry (LC-MS/MS) assay to measure the concentration of
M20.7. The non-reaction control samples were kept on ice for the full period before
centrifugation. The concentration of M20.7 obtained with the non-reaction control was subtracted
from that obtained with each enzyme source. The vildagliptin-hydrolyzing activity was expressed
as an M20.7 formation rate (pmol/h/mg protein).
LC-MS/MS conditions

DMD Fast Forward. Published on January 16, 2015 as DOI: 10.1124/dmd.114.062331
This article has not been copyedited and formatted. The final version may differ from this version.
A Waters Micromass tandem quadrupole Quattro micro mass spectrometer (Waters, Milford,
MA, USA) was interfaced with a Waters Alliance 2695 Separation Module (Waters) via an
electrospray ionization probe in the positive ion mode. M20.7 was separated on a Polaris 5 µm
C18-A50 × 2.0-mm column (25 °C) (Agilent Technologies, Amstelveen, The Netherlands) with a
MetaGuard 2.0 mm Polaris 5 µm C18-Aguard column (Agilent Technologies). The mobile phase
of A/B (1:3, v/v) was used, where A was methanol/10 mM ammonium acetate, pH 8.0 (5:95, v/v),
and B was acetonitrile/methanol (10:90, v/v). The flow rate was adjusted to 0.2 mL/min, and an
eluent between 0 and 5 min was introduced into the mass spectrometer. The ionization conditions
were as follows: capillary voltage, 3.6 kV; cone voltage, 32 V; collision energy, 20 V;
sourcetemperature, 120 °C; desolvationtemperature, 400 °C; collision gas, argon. The sample
was analyzed during the multiple reaction monitoring mode of the mass spectrometer at a dwell
time of 0.5 s per channel using m/z 323.1 > 173.3 as the transition.
DPP-4 activity assays
DPP-4 activity was determined by cleavage rate of AMC from the synthetic substrate
Gly-Pro-AMC (Leiting et al., 2003; Lee et al., 2009). These assays were performed in 96-well
flat-bottom plates in a total assay volume of 100 µL. The diluted enzyme sources (0.5 to 40

DMD Fast Forward. Published on January 16, 2015 as DOI: 10.1124/dmd.114.062331
This article has not been copyedited and formatted. The final version may differ from this version.
µg/mL) were incubated for 15 min at room temperature in assay buffer (50 mM glycine, pH 8.7, 1
mM EDTA). After 15 min incubation, Gly-Pro-AMC was added. The final concentration of
Gly-Pro-AMC was 50 µM. The plates were incubated at 25 °C for 10 min. After incubation,
fluorescence was measured using a SpectraMax M5 96-well plate spectrophotometer (excitation,
360 nm; emission, 460 nm, Molecular Devices, Sunnyvale, CA, USA). The standard curve of free
AMC was constructed using 0 to 10 µM solutions of AMC. The DPP-4 activity was expressed as
the amount of cleaved AMC per minute per mg protein (nmol/min/mg protein).
SDS-PAGE and western blot analysis
Mouse liver samples (100 µg of protein), rat liver samples (100 µg of protein), human liver
samples (100 µg of protein), and cell S9 fractions (3 µg of protein) were subjected to NuPAGE
4-12% Bis-Tris Gel (Lifetechnologies, Carlsbad, CA, USA) and transferred to a PVDF
membrane Immobilon-P (Millipore, Bedford, MA, USA) following the manufacurer’s protocol.
The membrane was blocked for 1 h with 50 mg/mL skimmed milk in phosphate buffered saline
(PBS) and then incubated with anti-human DPP-4 antibody (AF1180) diluted with PBS (1:1,000)
as a primary antibody overnight. The membrane was washed with PBS three times and incubated
with HRP-labeled secondary antibody diluted with PBS (1:10,000) for 1 h. The bands were

DMD Fast Forward. Published on January 16, 2015 as DOI: 10.1124/dmd.114.062331
This article has not been copyedited and formatted. The final version may differ from this version.
detected using Chemi-Lumi One L Western-blotting detection reagents (Nacalai Tesque).
Anti-GAPDH antibody (1:5,000) or anti-Actin antibody (1:5,000) was used as the protein loading
control.
Inhibition study of vildagliptin-hydrolyzing activity
The inhibitory effect of sitagliptin, a selective DPP-4 inhibitor, on vildagliptin-hydrolyzing
activity was investigated using mouse, rat, and human liver S9 fraction as an enzyme source. The
inhibitory effect of bis(p-nitrophenyl) phosphate and ethopropazine, a carboxylesterase and
butyrylcholinesterase inhibitor, respectively, on vildagliptin-hydrolyzing activity of human liver
S9 fractions was also investigated. Vildagliptin (10 µM) and inhibitors (10, 100, and 1000 µM)
were incubated with the each liver S9 fraction at 1.0 mg/mL. In the presence of the inhibitor,
vildagliptin-hydrolyzing activity was measured as described above.
Expression of human DPP-4 in HEK293 cells and mutant construction
Human DPP-4 (National Center for Biotechnology Information (NCBI) accession number
NM_001935.3) cDNA were prepared by a reverse transcription-polymerase chain reaction
technique from human liver total RNA with sense and antisense oligonucleotide primers (5’-TGT

DMD Fast Forward. Published on January 16, 2015 as DOI: 10.1124/dmd.114.062331
This article has not been copyedited and formatted. The final version may differ from this version.
TTA ACT CGG GGC CGA AA-3’ and 5’-CAG ACC AGG ACC GGA ACA TC-3’ for DPP-4
vector 1; 5’-GTT GGA AGA TTT AGG CCT TCA GA-3’ and 5’-CCC TAG TGA CAT CAC
TGC CC-3’ for DPP-4 vector 2). The PCR products were subcloned into pTARGET Mammalian
Expression Vector (Promega, Madison, WI, USA) and the DNA sequences of the inserts were
determined by a BigDye Terminator v3.1 cycle sequencing kit (Applied Biosystems, Foster City,
CA, USA) using a 3130 Genetic Analyzer (Applied Biosystems). The two vectors, DPP-4 vector
1 and DPP-4 vector 2, have two Stu sites, respectively (Supplemental Fig. 1). To construct the
expression vector of full-length human DPP-4, we subcloned the 2.95-kb Stu fragment of DPP-4
vector 2 into the 5.56-kb Stu fragment of DPP-4 vector 1. Additionally, we constructed the two
expression vectors of human DPP-4 mutant. The expression vector of DPP-4 vector 2 was used as
the template for constructing mutants. A KOD-plus-Neo DNA polymerase (Toyobo, Osaka,
Japan) was used to create two single mutants of human DPP-4 (R623Q or S630A) in this study.
The forward and reverse primers used for the mutagenesis were shown as follows: 5’-CAA CAA
ACA AAT TGC AAT TTG GGG CTG GTC-3’ and 5’-CCC AAA TTG CAA TTT GTT TGT
TGT CCA CAA ATC CC-3’ for the R623Q mutant, and 5’-GCA ATT TGG GGC TGG GCA
TAT GGA GGG TAC GTA ACC-3’ and 5’-GGT TAC GTA CCC TCC ATA TGC CCA GCC

DMD Fast Forward. Published on January 16, 2015 as DOI: 10.1124/dmd.114.062331
This article has not been copyedited and formatted. The final version may differ from this version.
CCA AAT TGC-3’ for the S630A mutant. To construct the expression vector of full-length
R623Q or S630A, we subcloned the 2.95-kb Stu fragment of DPP-4 vector 2 mutated R623Q or
S630A into the 5.56-kb Stu fragment of DPP-4 vector 1, respectively. To confirm the desired
mutation and verify the absence of unintended mutations, the constructs were sequenced.
HEK293 cells were obtained from the Cell Resource Center for Biomedical Research,
Tohoku University (Sendai, Japan). The cells were grown in Dulbecco’s modified Eagle’s
medium containing 100 U/mL penicillin, 100 µg/mL streptomycin, and 10% fetal bovine serum
with 5% CO2 at 37 °C. The wild-type human DPP-4-, R623Q-, or S630A-expressing HEK293
cells and control cells (Mock) were constructed by transfecting the expression vector and control
pTARGET vector, respectively, into cells using a Lipofectamine LTX & Plus Reagent
(Invitrogen, Carlsbad, CA, USA). Stable transfectants were selected in medium containing 600
µg/mL G418 and several clones were isolated.
Preparation of cell S9 fraction
HEK293 cells expressing single wild-type human DPP-4, R623Q, or S630A was suspended
in cold PBS and homogenized with a teflon-glass homogenizer for 40 strokes. The total cell

DMD Fast Forward. Published on January 16, 2015 as DOI: 10.1124/dmd.114.062331
This article has not been copyedited and formatted. The final version may differ from this version.
homogenate was centrifuged at 600 g for 10 min at 4 °C. The pellet, which contained nuclear, was
discarded. The supernatant, which contained mostly cell protein, was collected as the cell
homogenate. Additionally, the cell homogenate was centrifuged at 9,000 g for 20 min at 4 °C. The
supernatant was used as the S9 fraction.
Kinetic analysis of DPP-4 activity and vildagliptin-hydrolyzing activity
The kinetic analysis of DPP-4 activity using the synthetic substrate Gly-Pro-AMC was
performed using human liver samples, human DPP-4-expressing HEK293 cells, and DPP-4
mutant-expressing HEK293 cells with substrate concentrations from 10 to 200 µM. The kinetic
analysis of vildagliptin-hydrolyzing activity was performed using human liver samples with
substrate concentrations from 1 to 1000 µM and using HEK293 cells expressing human DPP-4
and its mutants with substrate concentrations 1 to 500 µM, respectively. Kinetic parameters were
estimated from the fitted curves using a KaleidaGraph computer program (Synergy Software,
Reading, PA, USA) designed for nonlinear regression analysis. The following equations were
applied for Michaelis-Menten kinetics (eq. 1) or biphasic kinetics (eq. 2):
V = Vmax × S/ (Km + S)
(1)

DMD Fast Forward. Published on January 16, 2015 as DOI: 10.1124/dmd.114.062331
This article has not been copyedited and formatted. The final version may differ from this version.
V = Vmax1 × S/ (Km1 + S) + Vmax2 × S/ (Km2 + S)
(2)
where V is the initial velocity of the reaction, Vmax is the maximum velocity, S is the substrate
concentration, and Km is the Michaelis constant, which is defined as the substrate concentration at
one-half the maximum velocity. Km1 and Km2 are affinity constants for high and low-affinity sites,
respectively, and Vmax1 and Vmax2 are maximum velocities for high and low-affinity sites,
respectively. Kinetic data were also analyzed using Eadie-Hofstee plots.
Statistical analysis
Data were presented as means ± S.D. and were assessed for statistical significance using the
unpaired t-test. The correlation analysis was preformed using the Spearman rank method. A value
of P < 0.05 was considered statistically significant.
Results
Vildagliptin-hydrolyzing activity of mouse, rat, and human liver samples

DMD Fast Forward. Published on January 16, 2015 as DOI: 10.1124/dmd.114.062331
This article has not been copyedited and formatted. The final version may differ from this version.
M20.7 formation rates in mouse, rat, and human liver samples were measured. Time and
protein linearity studies were performed using human liver S9 fraction. It was demonstrated that
formation of M20.7 in human liver S9 fraction increased linearly for at least 2 h and 2.0 mg/mL,
respectively (data not shown). M20.7 formation rate in human liver S9 fraction, cytosol, and
microsomes was 0.21 ± 0.03, 0.13 ± 0.01, and 0.64 ± 0.05 pmol/h/mg protein, respectively, at a
substrate concentration of 10 µM (Fig. 1A). M20.7 formation rate in human liver microsomes was
statistically higher than that in the liver cytosol (P < 0.01). M20.7 formation rate in mouse liver
S9 fraction, cytosol, and microsomes was 0.41 ± 0.04, 0.36 ± 0.01, and 0.51 ± 0.02 pmol/h/mg
protein, respectively. M20.7 formation rate in rat liver S9 fraction, cytosol, and microsomes was
0.50 ± 0.01, 0.32 ± 0.02, and 0.95 ± 0.03 pmol/h/mg protein, respectively. M20.7 formation rate
in mouse and rat liver microsomes was also statistically higher than that in the liver cytosol,
respectively (P < 0.01). These results indicate that the cyano group hydrolysis reaction of
vildagliptin can be catalyzed by enzymes present in the liver, particularly in liver microsomes. A
kinetic analysis of vildagliptin-hydrolyzing activity was performed using human liver samples
(Fig. 1B). The Eadie-Hofstee plots of vildagliptin-hydrolyzing activities in human liver S9
fraction, cytosol, and microsomes were not linear at the low substrate concentration, indicating

DMD Fast Forward. Published on January 16, 2015 as DOI: 10.1124/dmd.114.062331
This article has not been copyedited and formatted. The final version may differ from this version.
that multiple enzymes might be responsible for the hydrolysis of vildagliptin (Supplemental Fig.
2A). Because the kinetic parameters of human liver samples could not been calculated using the
KaleidaGraph computer program, these parameters were calculated from the Eadie-Hofstee plots
(Supplemental Fig. 2A). The Km value for high-affinity sites of the vildagliptin-hydrolyzing
activity in human liver S9 fraction, cytosol, and microsomes was 1.95 ± 0.6, 2.34 ± 1.0, and 0.75
± 0.4 µM, respectively (Table 1). The Km value for low-affinity sites of the
vildagliptin-hydrolyzing activity in human liver S9 fraction, cytosol, and microsomes was 335 ±
10, 189 ± 5.0, and 448 ± 104 µM, respectively (Table 1).
DPP-4 activities and protein expression levels of DPP-4 in mouse, rat, and human liver
samples
We performed DPP-4 activity assay using Gly-Pro-AMC, which is a synthetic
DPP-4-specific substrate (Leiting et al., 2003; Lee et al., 2009), and western blot analysis using
mouse, rat, and human liver samples. The DPP-4 activity in human liver S9 fraction, cytosol, and
microsomes was 1.9 ± 0.2, 1.4 ± 0.2, and 3.3 ± 0.1 nmol/min/mg protein, respectively, at a
substrate concentration of 50 µM (Fig. 2A). The DPP-4 activity in human liver microsomes was
statistically higher than that in the liver cytosol (P < 0.01). Additionally, DPP-4 activity in mouse

DMD Fast Forward. Published on January 16, 2015 as DOI: 10.1124/dmd.114.062331
This article has not been copyedited and formatted. The final version may differ from this version.
and rat liver microsomes was also statistically higher than that in the liver cytosol, respectively.
These results suggest that DPP-4 is abundantly present in liver microsomes. Indeed, the protein of
DPP-4 was highly detected in liver microsomes (Fig. 2B). The protein expression level of DPP-4
in liver samples correlated well with the DPP-4 activity in each liver sample (Fig. 2A and 2B). A
kinetic analysis of the DPP-4 activity was performed using human liver samples. The DPP-4
activity by human liver S9 fraction, cytosol, and microsomes followed the Michaelis-Menten
equation (Fig. 2C), as the Eadie-Hofstee plots were almost linear, respectively (Supplemental Fig.
3A). The Km value of the DPP-4 activity in human liver S9 fraction, cytosol, and microsomes was
26.2 ± 2.5, 21.3 ± 1.5, and 24.6 ± 1.1 µM, respectively (Table 2).
We further preformed correlation analyses between M20.7 formation rates and DPP-4
activities in the liver samples (Fig. 3). M20.7 formation rate was significantly positively
correlated with the DPP-4 activity using Gly-Pro-AMC in liver samples (r = 0.917, P < 0.01).
This finding suggests that DPP-4 is primarily responsible for vildagliptin hydrolysis in the liver.
Inhibition study of vildagliptin-hydrolyzing activity
The inhibitory effect of sitagliptin, a selective DPP-4 inhibitor, on vildagliptin-hydrolyzing
activity of mouse, rat, and human liver S9 fractions was assessed. The formation of M20.7 was

DMD Fast Forward. Published on January 16, 2015 as DOI: 10.1124/dmd.114.062331
This article has not been copyedited and formatted. The final version may differ from this version.
inhibited by sitagliptin in a concentration-dependent manner (Fig. 4A). The formation of M20.7
in human liver S9 fraction was approximately 60% inhibited by 1000 µM sitagliptin. The
formation of M20.7 in mouse and rat liver S9 fraction was approximately 50 and 90% inhibited
by 1000 µM sitagliptin, respectively.
We also assessed inhibitory effect of BNPP (a carboxylesterase inhibitor) and ethopropazine
(a butyrylcholinesterase inhibitor) on vildagliptin-hydrolyzing activity of human liver S9
fractions, because it has been reported that anagliptin, which has the same cyanopyrrolidine motif
as vildagliptin, is metabolism by DPP-4, carboxylesterase, and butyrylcholinesterase (Furuta et
al., 2013). However, the formation of M20.7 in human liver S9 fraction was not inhibited by
BNPP and ethopropazine (Fig. 4B). These results indicate that DPP-4 is greatly involved in
vildagliptin hydrolysis in mouse, rat, and human liver, whereas carboxylesterase and
butyrylcholinesterase is not involved in vildagliptin hydrolysis in human liver.
Establishment of single expression systems of wild-type human DPP-4, R623Q, and S630A
in HEK293 cells
To assess the vildagliptin-hydrolyzing activity of human DPP-4, we constructed the stable
single expression systems of wild-type human DPP-4 in HEK293 cells (HEK/DPP-4). To assess

DMD Fast Forward. Published on January 16, 2015 as DOI: 10.1124/dmd.114.062331
This article has not been copyedited and formatted. The final version may differ from this version.
the effects of two DPP-4 mutants on DPP-4 activity and vildagliptin-hydrolyzing activity, stable
single expression systems of R623Q and S630A in HEK293 cells (HEK/R623Q and
HEK/S630A) were also constructed. R623Q (rs147614497) is the non-synonymous human
DPP-4 single-nucleotide polymorphism (SNP) with the highest minor allele frequency (MAF =
0.0248) in December 2013, which has been reported in the NCBI dbSNP (build 141;
http://www.ncbi.nlm.nih.gov/projects/SNP/). S630A is an active site mutant that lacked the
catalytic serine 630 residue required for DPP-4 activity (Tanaka et al., 1993; Metzler et al., 2008;
Greene et al., 2011). The protein expression levels of DPP-4 were assessed by western blot
analysis. We selected the clones with the highest DPP-4 protein levels for the subsequent analyses.
Significant DPP-4 protein expression was observed in the cell S9 fraction of the wild-type human
DPP-4-transfectant compared with that of Mock cells (Fig. 5A). The R623Q- or
S630A-transfectant also showed significant DPP-4 protein expression compared with Mock,
respectively. In addition, the protein expression levels of DPP-4 in HEK/R623Q and HEK/S630A
were similar and higher than that in HEK/DPP-4, respectively. The two bands reacting with the
anti-human DPP-4 antibody were observed in HEK/DPP-4 and HEK/R623Q. To elucidate
whether the two bands were different glycosylation forms of DPP-4, the S9 fraction of

DMD Fast Forward. Published on January 16, 2015 as DOI: 10.1124/dmd.114.062331
This article has not been copyedited and formatted. The final version may differ from this version.
HEK/DPP-4 was deglycosylated under denaturing conditions and analyzed by western blotting
with the anti-human DPP-4 antibody. Glycosidase treatment reduced the size of two bands to a
single 110-kDa band in HEK/DPP-4 (Supplemental Fig. 4). This result suggested that the two
bands reacting with the anti-human DPP-4 antibody in HEK/DPP-4 and HEK/R623Q were
different glycosylation forms of DPP-4.
Effects of DPP-4 mutants on DPP-4 activity
To assess the DPP-4 activity of wild-type human DPP-4, we performed DPP-4 activity
assays using the synthetic substrate Gly-Pro-AMC. To assess the effects of R623Q and S630A on
DPP-4 activity, we also performed DPP-4 activity assays using HEK/R623Q and HEK/S630A.
The DPP-4 activity in Mock, HEK/DPP-4, HEK/R623Q, and HEK/S630A was 1.2 ± 0.1, 134 ±
5.3, 132 ± 6.2, and 1.0 ± 0.1 nmol/min/mg protein, respectively, at a substrate concentration of 50
µM (Fig. 5B). The DPP-4 activity in HEK/DPP-4 was approximately 100- fold higher than that in
Mock. The DPP-4 activity in HEK/R623Q was comparable with that in HEK/DPP-4. Although
S630A-transfectant showed significant DPP-4 protein expression compared with Mock, the
DPP-4 activity in HEK/S630A was comparable with that in Mock. A kinetic analysis of DPP-4
activity was performed using Mock, HEK/DPP-4, and HEK/R623Q. The DPP-4 activity in

DMD Fast Forward. Published on January 16, 2015 as DOI: 10.1124/dmd.114.062331
This article has not been copyedited and formatted. The final version may differ from this version.
recombinant human DPP-4 (hDPP-4) and R623Q (hR623Q) was expressed as the DPP-4 activity
(nmol/min/mg protein) in which the DPP-4 activity obtained with Mock was subtracted from that
obtained with HEK/DPP-4 and HEK/R623Q, respectively. The DPP-4 activity by wild-type
hDPP-4 and hR623Q followed the Michaelis-Menten equation (Fig. 5C), as the Eadie-Hofstee
plots were almost linear, respectively (Supplemental Fig. 3B). The Km value of the DPP-4 activity
in wild-type hDPP-4 and hR623Q was 54.0 ± 2.6 and 45.6 ± 8.0 µM, respectively (Table 2). The
Km value of the DPP-4 activity in hR623Q was close to that in wild-type hDPP-4.
Effects of DPP-4 mutants on vildagliptin-hydrolyzing activity
To assess the vildagliptin-hydrolyzing activity of wild-type human DPP-4, we performed
hydrolysis assays of the cyano group of vildagliptin using HEK/DPP-4. To assess the effects of
R623Q and S630A on vildagliptin-hydrolyzing activity, we also performed hydrolysis assays of
the cyano group of vildagliptin using HEK/R623Q and HEK/S630A. The M20.7 formation rate in
Mock, HEK/DPP-4, HEK/R623Q, and HEK/S630A was 0.18 ± 0.01, 7.4 ± 0.2, 5.2 ± 0.1, and
0.13 ± 0.01 pmol/h/mg protein, respectively, at a substrate concentration of 1 µM (Fig. 5D). The
M20.7 formation rate in HEK/DPP-4 was statistically higher than that in Mock (P < 0.01).
Interestingly, although DPP-4 activity in HEK/R623Q was comparable with that in HEK/DPP-4,

DMD Fast Forward. Published on January 16, 2015 as DOI: 10.1124/dmd.114.062331
This article has not been copyedited and formatted. The final version may differ from this version.
M20.7 formation rate in HEK/R623Q was statistically lower than that in HEK/DPP-4 (P < 0.01).
Furthermore, M20.7 formation rate in HEK/S630A was comparable with that in Mock. These
results suggest that the catalytic serine 630 residue is directly involved in the vildagliptin
hydrolysis in human DPP-4 and that R623Q mutation results in a decrease of the
vildagliptin-hydrolyzing activity. A kinetic analysis of vildagliptin-hydrolyzing activity was
performed using Mock and HEK/DPP-4. The vildagliptin-hydrolyzing activity in hDPP-4 was
expressed as an M20.7 formation rate (pmol/h/mg protein) in which the vildagliptin-hydrolyzing
activity obtained with Mock was subtracted from that obtained with HEK/DPP-4. Because the
vildagliptin-hydrolyzing activity in Mock at the high substrate concentration was higher than that
in HEK/R623Q, the parameters of hR623Q could not been calculated. The
vildagliptin-hydrolyzing activity by hDPP-4 followed the biphasic equation (Fig. 5E), as the
Eadie-Hofstee plots were biphasic (Supplemental Fig. 2B). The Km value for high and
low-affinity sites of vildagliptin-hydrolyzing activity in hDPP-4 was 0.62 ± 0.1 and 151 ± 71 µM,
respectively (Table 1). The Km value for high-affinity site of vildagliptin-hydrolyzing activity in
hDPP-4 was similar to that in human liver samples.

DMD Fast Forward. Published on January 16, 2015 as DOI: 10.1124/dmd.114.062331
This article has not been copyedited and formatted. The final version may differ from this version.
Discussion
Identification of major enzyme responsible for the metabolism of drugs is important to
understand interindividual variability in pharmacokinetics. In the present study, we clarified the
contribution rate of DPP-4 to the cyano group hydrolysis reaction of vildagliptin in mouse, rat,
and human liver. The results of correlation analysis (Fig. 3) and inhibition study (Fig. 4A)
indicated that DPP-4 was greatly involved in vildagliptin hydrolysis in mouse, rat, and human
liver (Fig. 6). The formation of M20.7 in HEK/DPP-4 was approximately 90% inhibited by 1000
µM sitagliptin (Supplemental Fig. 5), indicating that the concentration of sitagliptin was adequate
for inhibition of vildagliptin-hydrolyzing activity in human DPP-4. Therefore, the results of
inhibition study suggest that the contribution rate of DPP-4 to vildagliptin hydrolysis in human
liver is approximately 60% (Fig. 4A). To predict the in vivo clearance of vildagliptin, plasma
protein binding and affinity of vildagliptin for DPP-4 are also important factors. While the
affinity of vildagliptin for DPP-4 is high, binding of vildagliptin to plasma proteins is very low
(9.3%) (He et al., 2009a). Therefore, it has been suggested that protein binding of vildagliptin is
clinically negligible (He et al., 2009a; Golightly et al., 2012). Additionally, to understand the
DPP-4 protein expression in various tissues, we performed western blot analysis using 8 mouse

DMD Fast Forward. Published on January 16, 2015 as DOI: 10.1124/dmd.114.062331
This article has not been copyedited and formatted. The final version may differ from this version.
tissues (heart, lung, stomach, liver, kidney, spleen, small intestine, and large intestine). Protein of
DPP-4 was expressed not only in liver but also in other tissues, such as lung, kidney, and small
intestine, in mice (Supplemental Fig. 6), which is in agreement with previous reports that human
DPP-4 is localized not only in liver but also in other tissues (Mentlein, 1999; Gorrell et al., 2001).
It has also been reported that human liver, kidney, and intestinal microsomes are all capable of
hydrolyzing vildagliptin to M20.7 in in vitro (He et al., 2007). These findings indicate that the
hydrolysis of vildagliptin by DPP-4 may occur in whole body in in vivo and that the contribution
rate of DPP-4 to vildagliptin hydrolysis in humans could be more than 60%.
The DPP-4 activity was determined using the synthetic substrate Gly-Pro-AMC. In our study,
the Km value of the DPP-4 activity in hDPP-4 was approximately 50 µM (Table 2). This Km value
was almost consistent with previous reports (Leiting et al., 2003; Aertgeerts et al., 2005; Lee et al.,
2009). Furthermore, the Km values of the DPP-4 activity in human liver samples were similar to
that in hDPP-4 (Table 2). The DPP-4 activity and the protein expression level of DPP-4 in human
liver microsomes were higher than those in the human liver cytosol, respectively (Fig. 2Aand 2B).
DPP-4 has the integral membrane form and soluble form (Durinx et al., 2000). Our findings
indicate that liver microsomal fraction abundantly contains a functional DPP-4, which is probably

DMD Fast Forward. Published on January 16, 2015 as DOI: 10.1124/dmd.114.062331
This article has not been copyedited and formatted. The final version may differ from this version.
the membrane form of DPP-4.
The result of inhibition study suggested that the enzyme responsible for the hydrolysis of
vildagliptin in human liver was not only DPP-4 (Fig. 4A). Indeed, although the Km values for
high-affinity sites of vildagliptin-hydrolyzing activity in human liver samples were close to that
in hDPP-4, the Km values for low-affinity sites of vildagliptin-hydrolyzing activity in human liver
samples were not close to that in hDPP-4 (Table 1). Furthermore, although the protein expression
of DPP-4 and the DPP-4 activity in Mock was hardly observed, the vildagliptin-hydrolyzing
activity in Mock was observed in a concentration-dependent manner (Supplemental Fig. 7), and
the Eadie-Hofstee plots were linear (Supplemental Fig. 2B). The enzyme responsible for the
hydrolysis of vildagliptin at the high vildagliptin concentration may be present in human liver
samples and the S9 fraction of Mock cells (Table1 and Supplemental Fig. 7). Involvement of
multiple enzymes in the cyano group hydrolysis reaction of vildagliptin was first demonstrated in
the present study. The Cmax of vildagliptin after the therapeutic dose of 50 mg twice daily is
approximately 1 µM (He, 2012). Human DPP-4 was greatly involved in vildagliptin hydrolysis at
1 µM vildagliptin, whereas the enzymes, which have expressed in Mock cells, were involved in
vildagliptin hydrolysis at 50 to 500 µM vildagliptin (Fig. 5D and Supplemental Fig. 7). These

DMD Fast Forward. Published on January 16, 2015 as DOI: 10.1124/dmd.114.062331
This article has not been copyedited and formatted. The final version may differ from this version.
findings also indicate that human DPP-4 could be greatly involved in vildagliptin hydrolysis in in
vivo.
The result of inhibition study using rat liver S9 fraction (Fig. 4A), which suggested that
DPP-4 mainly contributes to the hydrolysis of vildagliptin in rats, was not consistent with the
previous report, which suggested that approximately 20% of the formation of M20.7 in rats is
likely attributable to DPP-4 (He et al., 2009b). This inconsistency might be due to the fact that the
two sets of studies are very different experimental designs. To understand the pharmacokinetics
of vildagliptin in humans and other animals, the exploration of multiple enzymes responsible for
hydrolysis of vildagliptin needs to be conducted in the future.
We examined the effects of two DPP-4 mutants, R623Q and S630A, on DPP-4 activity and
vildagliptin-hydrolyzing activity. It has been reported that vildagliptin forms a reversible covalent
bounds with DPP-4 to inhibit DPP-4 enzymatic activity (Ahrén et al., 2011). The covalent
binding between vildagliptin and DPP-4 is a consequence of the formation of an unstable
complex by reaction of the cyano group of vildagliptin with amino acid residue serine 630 within
the catalytic domain of DPP-4 (Potashman et al., 2009; Ahrén et al., 2011). It has been suggested
that the amino acid residue serine 630 of DPP-4 is involved in formation of M20.7 (He et al.,

DMD Fast Forward. Published on January 16, 2015 as DOI: 10.1124/dmd.114.062331
This article has not been copyedited and formatted. The final version may differ from this version.
2009a; Ahrén et al., 2011). In the present study, we demonstrated that the catalytic serine 630
residue is directly involved in the vildagliptin hydrolysis in human DPP-4 using
S630A-expressing HEK293 cells (Fig. 5D). We also assessed the effects of R623Q, which is the
non-synonymous human DPP-4 SNP, on DPP-4 activity and vildagliptin-hydrolyzing activity.
Interestingly, although DPP-4 activity in HEK/R623Q was comparable with that in HEK/DPP-4,
M20.7 formation rate in HEK/R623Q was statistically lower than that in HEK/DPP-4 (P < 0.01)
(Fig. 5B and 5D). This reduction of M20.7 formation rate in HEK/R623Q may be due to the
influence that the position of its mutation is near to the catalytic serine 630 residue of DPP-4.
These results suggest that the DPP-4 genetic polymorphism leads to a decrease of the
vildagliptin-hydrolyzing activity.
In conclusion, DPP-4 was greatly involved in vildagliptin hydrolysis in mouse, rat, and
human liver. The contribution rate of DPP-4 to vildagliptin hydrolysis in human liver was
approximately 60%. Additionally, R623Q mutant, which is the non-synonymous SNP of human
DPP-4, led to a decrease of the vildagliptin-hydrolyzing activity. There have been rare cases of
hepatic dysfunction in patients treated with vildagliptin (Deacon, 2011; Karagiannis et al., 2014;
Kurita et al., 2014). Our findings might be useful for the prediction of liver injury induced by