Novel fatty acid chain modified GLP-1 derivatives with prolonged in vivo glucose-lowering ability and balanced glucoregulatory activity

Article abstract of DOI:10.1016/j.bmc.2018.04.022

Glucagon-like peptide-1 is a potent hypoglycemic hormone with beneficial properties for the treatment of diabetes. However, its half-life is short because the rapid metabolic degradation. This study aims to prolong the half-life of glucagon-like peptide-1 through conjugation with the fatty acid side chain which helps the conjugates to interact with the albumin. Firstly, we chose two optimized polypeptide chains which have tremendous hypoglycemic effect named Cys₁₇-Gly₈-GLP-1(7-36)-NH₂ and Cys₃₇-Gly₈-GLP-1(7-37)-NH₂, and various fatty acid chains were modified. All conjugates preserved relatively strong GLP-1R activation and I-6 behaved best in glucose-lowering ability. The prolonged antidiabetic effects of I-6 were further confirmed by hypoglycemic efficacy test in vivo. Meanwhile, once daily injection of I-6 to diabetic mice achieved long-term beneficial effects on glucose tolerance, body weight and blood chemistry. It is concluded that I-6 is a promising agent for further investigation of its potential to treat obese patients with diabetes.

Full text of DOI:10.1016/j.bmc.2018.04.022

Bioorganic & Medicinal Chemistry xxx (2018) xxx–xxx
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry
journal homepage: www.elsevier.com/locate/bmc
Novel fatty acid chain modified GLP-1 derivatives with prolonged in vivo
glucose-lowering ability and balanced glucoregulatory activity
Xingguang Cai a,e, Lidan Sun a,b,e, Yuxuan Dai , Yosefa Avraham , Chunxia Liu , Jing Han , Yuan Liu ,
a
c
a
a
a
a
a,d,
⇑
a,d,
⇑
Dazhi Feng , Wenlong Huang
a
, Hai Qian
Center of Drug Discovery, State Key Laboratory of Natural Medicines, China Pharmaceutical University, Nanjing 210009, PR China
Department of Pharmaceutics, College of Medicine, Jiaxing University, Jiaxing 314001, PR China
Department of Human Nutrition and Metabolism, Braun School of Public Health, Faculty of Medicine, Hebrew University, Jerusalem, Israel
Jiangsu Key Laboratory of Drug Discovery for Metabolic Disease, China Pharmaceutical University, 24 Tongjiaxiang, Nanjing 210009, PR China
b
c
d
a r t i c l e i n f o
a b s t r a c t
Article history:
Glucagon-like peptide-1 is a potent hypoglycemic hormone with beneficial properties for the treatment
of diabetes. However, its half-life is short because the rapid metabolic degradation. This study aims to
prolong the half-life of glucagon-like peptide-1 through conjugation with the fatty acid side chain which
helps the conjugates to interact with the albumin. Firstly, we chose two optimized polypeptide chains
Received 8 March 2018
Revised 8 April 2018
Accepted 10 April 2018
Available online xxxx
8 2 8
which have tremendous hypoglycemic effect named Cys17-Gly -GLP-1(7-36)-NH and Cys37-Gly -GLP-
1
(7-37)-NH , and various fatty acid chains were modified. All conjugates preserved relatively strong
2
Keywords:
GLP-1
Diabetes
Fatty acid
Body weight
GLP-1R activation and I-6 behaved best in glucose-lowering ability. The prolonged antidiabetic effects
of I-6 were further confirmed by hypoglycemic efficacy test in vivo. Meanwhile, once daily injection of
I-6 to diabetic mice achieved long-term beneficial effects on glucose tolerance, body weight and blood
chemistry. It is concluded that I-6 is a promising agent for further investigation of its potential to treat
obese patients with diabetes.
Ó 2018 Elsevier Ltd. All rights reserved.
1
. Introduction
Glucagon-like peptide-1 (GLP-1) is a potent hypoglycemic hor-
showed a moderate improvement in binding affinity for GLP-1
receptor over parent Ex-4, PEGylation reduced its affinity. In expec-
tation, the conjugate possessed a far more better pharmacokinetic
profile than parent Ex-4 itself. Luginbuhl. et al. fused GLP-1 to an
elastin-like polypeptide biopolymer and injected it as a subcuta-
neous injection depot which had improved in vivo half-life and
mone, emphasized a great opportunity for its potential for the
treatment of diabetes.1 GLP-1 exhibits stimulation of pancreatic
2
insulin secretion in glucose-dependent way. Thus, the risk of
hypoglycemia is reduced.3 A short half-life of endogenous GLP-1
limits its clinical application, which accounts for the rapid meta-
bolic degradation and elimination by kidney filtration.4
controlled release features. It is reported that the depot formed
7
after a single injection of GLP-1-ELP conjugate results in zero-order
release kinetics. However, PEG is non-biodegradable and some
PEGylated proteins have been associated with generation of anti-
5
Up to now, several distinct strategies have been described.
8
Methods have been developed to reduce renal elimination through
enlarging the molecular size, such as, PEG or ELP conjugates. Lee
et al. linked two Ex-4 molecules together by using a commercially
available bimaleimide amine to which a PEG chain had already
been conjugated so as to prolong pharmacokinetic duration by
increasing the hydrodynamic radius.6 Although the Ex-4 polymer
PEG antibodies and cellular toxicity. Meanwhile, the industrializa-
tion of ELPylated peptide also takes a long time. Thus, the GLP-1
receptor agonist field still has a need for new sustained-release
therapies that can achieve safety hypoglycemic action and conve-
nience preparation of production. Such a formulation would be
likely to reduce gastrointestinal side effects, improve patient com-
fort, and further dissociate compliance from therapeutic outcome.
Towards this goal, considering the physicochemical characteris-
tics of fatty acid chain modified conjugates, like interacting with
⇑
Corresponding authors at: Jiangsu Key Laboratory of Drug Discovery for
Metabolic Disease, China Pharmaceutical University, 24 Tongjiaxiang, Nanjing
9
serum albumin. Our group has investigated the hypoglycemic
2
10009, PR China.
effect of many GLP-1 conjugates and supplied a series of site-speci-
fic cysteine substitution polypeptide chains which have tremen-
dous hypoglycemic effects and created an effective and
E-mail addresses: ydhuangwenlong@126.com (W. Huang), qianhai24@163.com
(
H. Qian).
e
These authors contributed equally to this work.
https://doi.org/10.1016/j.bmc.2018.04.022
968-0896/Ó 2018 Elsevier Ltd. All rights reserved.
0
Please cite this article in press as: Cai X., et al. Bioorg. Med. Chem. (2018), https://doi.org/10.1016/j.bmc.2018.04.022

2
X. Cai et al. / Bioorganic & Medicinal Chemistry xxx (2018) xxx–xxx
convenient strategy to screen desired peptides for further
Ex-4 and liraglutide. As shown in Fig. 2, 21.0 ± 3.6% of Ex-4 and
86.7 ± 2.5% of liraglutide were found to associate with albumin
resin respectively under physiological conditions (in PBS pH 7.4).
The majority fatty acid chain-modified conjugates presented con-
siderably increased albumin binding ability than Ex-4.
5
,10
investigation.
In this study, we firstly chose two optimized polypeptide chains
which have tremendous hypoglycemic effects named as Cys17
-
1
0
8 2 8 2
Gly -GLP-1(7-36)-NH (1) and Cys37-Gly -GLP-1(7-37)-NH (2).
In the previous work, these two polypeptide chains which acted
excellently in antidiabetic characteristics had superior competency
to activate GLP-1 receptor.11 Through the thiol group with high
chemo selectivity, maleimide modified fatty acid chain was then
attached to the cysteine. A series of fatty acid chain modified
GLP-1 conjugates were designed (Fig. 1), afterwards their struc-
tural, biological, and biochemical analysis were explored. The opti-
mized compound I-6 possessed the most long acting antidiabetic
properties. Further investigation revealed that I-6 could also pro-
mote INS-1 cell viability (Fig. 5). Compounds I-6 was identified
as long-acting and highly biologically active GLP-1 derivative,
Long-term beneficial effects in diabetic mice (Fig. 6) and diet
induced mice (Fig. 8) were also evaluated.
2
.3. Plasma stability
The in vitro stability of the 6 conjugates (I-1 ꢀ I-6), compared
with Ex-4 and liraglutide, was performed in the plasma stability
assay. As illustrated in Fig. 2, the half-life of Ex-4 was approximate
3
3
.6 h while that of long-acting agonist liraglutide was 15.4 h at
7 °C in rat plasma, nearly 4-fold longer. As expected, most fatty
acid chain modified conjugates were more stable than Ex-4. What’s
more, the longer fatty acid chain exhibited better in long-acting
effects than the shorter.
The albumin binding ability and in vitro stability of conjugates
I-1 ꢀ I-6 were correlated well each other. In this study, fatty acid
chain-modified conjugate I-6 showed noticeably highest plasma
stability and albumin binding ability, which was very important
for the accomplishment of long-acting in vivo activity. Conse-
quently, the in vivo biological activity of I-6 was performed in
the subsequent assay.
2
. Results
2.1. Synthesis and characterization of the fatty acid chain-modified
GLP-1 analogs
Applying standard solid-phase peptide synthesis protocol with
N-Fmoc/tBu chemistry and microwave irradiation, two cysteine
altered peptides Cys17-Gly
8
-GLP-1(7-36)-NH
2
8
(1), Cys37-Gly -GLP-
2.4. Glucoregulatory and insulin secretion assay
1
(7-36)-NH (2) were synthesized. Three different fatty acid chains
2
with maleimide were prepared and reacted with the cysteine
altered peptides to obtain 6 fatty acid chain-modified conjugates
The in vivo antidiabetic activity of fatty acid chain-modified
conjugate I-6 was examined by oral glucose tolerance test (OGTT)
in SD rats. As illustrated in Fig. 3, after the intraperitoneal admin-
istration of I-6, the mean blood glucose level rapidly increased to
13.8 ± 1.31 mmol/L in the saline treated group while that in the
Ex-4, liraglutide and I-6 treated group reduced to ꢀ7.6, ꢀ7.2 and
6.7 mmol/L at 30 min, respectively (Fig. 3A). Meanwhile, the glu-
cose AUC of Ex-4, liraglutide and I-6 was considerably lower than
that of the saline treated group (Fig. 3B). In this assay, the decrease
in plasma glucose was associated with the increase in plasma insu-
lin concentration which was coherent with the GLP-1-dependent
mechanism (Fig. 3C). The time courses for plasma insulin concen-
trations in Ex-4, liraglutide and I-6 observed was similar in where
all plasma insulin concentration from 15 to 60 min were notably
greater than those of the control group. Especially, I-6 behaved
comparable insulin secretion promoting ability than Ex-4 and
liraglutide (p < 0.05). The glucose-lowering abilities and insulin
tropic activities of I-6 were comparable with those of Ex-4 and
liraglutide.
(I-1 ꢀ I-6). These conjugates (Fig. 1) were purified and identified
by preparative PR-HPLC and Waters ACQUITY UPLC–MS System.
Using HEK293 cells over expressing human GLP-1R, we deter-
mined the potency of all 6 conjugates. As the data shown in Table 1,
all conjugates preserved relatively strong receptor activation while
slightly reduction was observed. Furthermore, the longer fatty acid
chain exhibited weaker receptor activation than the shorter fatty
acid chain, indicating that the length of the fatty acid chain had a
bearing on the GLP-1 receptor activation. The results suggested
that the conjugates of cysteine-substituted peptides and fatty acid
chain were strong GLP-1R agonists.
2.2. Albumin binding
The expected long-acting effects may be attributed to the
remarked albumin binding affinities of the fatty acid chain. Thus,
the albumin binding assay was achieved using conjugates I-1 ꢀ I-6,
Fig. 1. Structure of compounds I-1 ꢀ I-6.
Please cite this article in press as: Cai X., et al. Bioorg. Med. Chem. (2018), https://doi.org/10.1016/j.bmc.2018.04.022

X. Cai et al. / Bioorganic & Medicinal Chemistry xxx (2018) xxx–xxx
3
Table 1
MS data of conjugates.
Conjugates
EC50(pM)a
Purity
Formula
Molecular mass
Molecular mass calculated
Molecular mass found
[
M + 3H]3+
[M + 4H]4+
[M + 3H]3+
[M + 4H]4+
Gly
8
-GLP-1(7-36)-NH
2
5.6 ± 0.9
1.8 ± 0.8
5.6 ± 1.1*
14.3 ± 2.0*
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
Ex-4
I-1
I-2
I-3
I-4
I-5
I-6
95.7%
95.8%
97.2%
96.3%
96.8%
95.1%
C
C
C
C
C
C
155
158
164
158
161
167
H
231
H
237
H
249
H
236
H
242
H
254
N
N
N
N
N
N
41
O
41
O
41
O
42
O
42
O
42
O
48
S
48
S
48
S
50
S
50
S
50
S
3468.8
3510.9
3595.1
3555.9
3598.0
3682.2
1157.3
1171.3
1199.3
1186.3
1200.3
1228.4
868.2
878.7
899.7
890.0
900.5
921.6
1157.6
1171.1
1199.3
1186.8
1200.9
1228.8
868.9
878.7
899.2
890.5
900.4
921.3
*
26.3 ± 2.5
6.1 ± 1.9*
*
9.3 ± 1.1
19.8 ± 2.1*
a
*
The receptor potency data are given as mean ± SD. All experiments were performed in triplicate and repeated three times (n = 3).
p < 0.05, compared with Ex-4.
Fig. 2. The plasma stability and albumin binding affinity of the fatty chain-modified GLP-1 conjugates, Ex-4 and liraglutide.
Fig. 3. In vivo biological activity tests of I-6. Liraglutide, Ex-4 and I-6 (25 nmol/kg) were i.p. injected into SD rats; glucose was taken orally (10 g/kg). (A) The glucose-lowering
effects and calculated plasma glucose AUC0À180min (B) of liraglutide, Ex-4 and I-6. (C) Insulinotropic activities and calculated insulin AUC0À180min (D) of liraglutide, Ex-4 and I-
6. n = 6 per group. Data are given as mean ± SD. *p < 0.05, **p < 0.01 compared with saline (control).
Please cite this article in press as: Cai X., et al. Bioorg. Med. Chem. (2018), https://doi.org/10.1016/j.bmc.2018.04.022

4
X. Cai et al. / Bioorganic & Medicinal Chemistry xxx (2018) xxx–xxx
2.5. Hypoglycemic duration test
viability of control INS-1 cells was significantly decreased by
ꢀ
45%, while only ꢀ28.8% in 1 nM and ꢀ21.9% in ꢁ10 nM I-6 trea-
In the present study, normal blood glucose level was set as
ted INS-1 cells (p < 0.05 for 10 nM and 100 nM, compared with sal-
below 8.35 mmol/L, and englycemic durations below this magni-
tude were calculated and viewed as a practical indication for
ine) (Fig. 5A–C). After 24 h incubation with glucolipotoxicity media
or 50
analysis was also performed, 24 h treatment with glucolipotoxicity
media, 50 M H or 1 M STZ caused a drastic increase of apop-
2 2
lM H O , similar results were observed. Flow cytometric
1
0
antidiabetic treatment. The hypoglycemic effects of compound
I-6 were performed at two does (25 and 150 nmol/kg, i.p.) in
STZ-induced diabetic mice. As shown in Fig. 4A, a hyperglycemic
state (average > 20 mmol/L) was observed in saline treated control
mice but a normal blood glucose level was maintained in those
treated with Ex-4 and liraglutide for nearly 4.8 h and 17.7 h,
respectively. At the dose of 25 nmol/kg, the target durations in I-
l
2
O
2
l
tosis in INS-1 cells at saline-treated control group, however, less
apoptosis was found in Ex-4, liraglutide and I-6 treated INS-1 cells
(Fig. 5D), indicating that I-6 could protect b-cell viability and inhi-
bit apoptosis against glucolipotoxicity, oxidative stress and DNA
damage mediated stress.
6
treated mice (ꢀ9.7 h) were much longer than Ex-4 and were
shorter than that in liraglutide (Fig. 4A). When the dose was raise
to 150 nmol/kg, the AUC0–36h values of I-6 were much lower than
those of saline or Ex-4 treated mice (Fig. 4B). In this assay, the time
required to below a glucose level of 8.35 mmol/L was 11.5 h.
2.7. Chronic in vivo studies on STZ-induced diabetic mice
In vivo activity and potential therapeutic utility was probed
STZ-induced diabetic mice received chronic administration of I-6.
Ex-4 and liraglutide (25 nmol/kg) were injected (once daily) and
saline was used as control. Fasting blood glucose level in diabetic
mice prior to treatment was 17.5 ± 1.9 mmol/L and in saline trea-
ted group it was worsened somewhat to 20.5 ± 1.3 mmol/L after
2
.6. INS-1 cells protective effects against cytotoxicity induced by H
2 2
O ,
glucolipotoxicity or STZ
2
1 days. In contrast with saline treated control, administration of
I-6 lowered fasting glucose toward the 10.7 ꢀ 11.3 mmol/L levels
Fig. 6A). HbA1c was regarded as an indirect pointer of cumulative
To determine the protective effect of I-6 on cell viability and
apoptosis, Compared with saline controls, I-6 at 1–100 nM signifi-
cantly retained cell viability against glucolipotoxicity media, 50
(
blood glucose concentrations which was a more responsive indica-
tor of glycemic control than glucose concentration and resulted
2 2
lM H O or 1 lM STZ. After 24 h incubation with 1 lM STZ, cell
1
2
from non-enzymatic irreversible glycation. Initially HbA1c was
well matched in all diabetic groups which was considerably higher
than in nondiabetic littermates. As shown in Fig. 6B, after a 21-day
treatment, HbA1c in diabetic mice daily injected I-6 revealed a
gradual reduction while that in control group stayed to be high.
The medication of I-6 caused the HbA1c reduction, which was
indicative of promising role of I-6 in clinical studies.
IPGTT test was performed before and after treatment (day 0 and
day 24), to determine if long-term treatment of I-6 improved glu-
cose tolerance or not. At the beginning of the treatment, no differ-
ence was observed between each group. However, at the end of the
treatment, Ex-4, liraglutide and I-6 treated mice exhibited rapid
blood glucose reductions and the blood glucose of those returned
to base line at 120 min after injection of glucose, whereas saline
treated mice manifested a typical diabetic outline of slow reduc-
tion on the hyperglycemic state after 120 min of administration
(
Fig. 6C). The AUC of blood glucose I-6 treated was considerably
lower than that of the saline treated diabetic mice (p < 0.05)
Fig. 6D).
To estimate the protective and rehabilitation of I-6, pancreas of
(
mice was gathered, followed with fixation, section and HE stain. As
shown in Fig. 7A, the edge of islet b cells was such rough and the
nucleus was diminished along with the b cells narrowed. As stated
in Fig. 7B, islet area in the saline treat mice was significant smaller
than that in the normal mice while the islet area in the Ex-4,
liraglutide and I-6 treated mice was raised notably after 21 days
treatment.
2.8. Chronic in vivo studies on DIO mice
As shown in Fig. 8A, after the treatment in STZ-induced diabetic
mice, body-weight change was noticed while the body weight of I-
treated group was lower than that in saline treated group. As
6
shown in Fig. 8A, during the 4-week treatment progression, I-6
inhibited cumulative water and food intake by up to 33.5% and
24.9%, respectively. Meanwhile, nearly the same amount of water
was taken by I-6 treated mice as lean mice. To further verify the
appetite suppression effect, an in vivo acute study was performed
and I-6 inhibited food intake by up to 34.5% at 6 h after administra-
tion. Body weight gain of the saline treated DIO mice continually
Fig. 4. Glucose-lowering and stabilizing effect of Ex-4, liraglutide and I-6 as
determine by hypoglycemic duration test in nonfasted diabetic mice. (A) Time-
course average blood glucose levels of diabetic mice after an i.p. injection of Ex-4,
liraglutide or I-6 (25 nmol/kg or 150 nmol/kg). Times depict hypoglycemic duration
rebound to 8.35 mmol/L. (B) Hypoglycemic effects of Ex-4, liraglutide and I-6 based
on the calculated glucose AUC0-60h values. n = 6 per group. Data are given as mean ±
SD. *p < 0.05, **p < 0.01 compared with saline (control).
Please cite this article in press as: Cai X., et al. Bioorg. Med. Chem. (2018), https://doi.org/10.1016/j.bmc.2018.04.022

X. Cai et al. / Bioorganic & Medicinal Chemistry xxx (2018) xxx–xxx
5
Fig. 5. I-6 attenuates glucolipotoxic and oxidative stress and enhances viability and survival of INS-1 cells. I-6 preserved cell viability in response to STZ (A). I-6 preserved
viability and survival of INS-1 cells against oxidative stress (B). I-6 preserved viability and survival of INS-1 cells against glucolipotoxic stress (C). I-6 decreased cell apoptosis
in response to STZ (D), oxidative stress (E) and glucolipotoxic stress (F). All experiments were performed in triplicate and repeated three times (n = 3). Data are given as mean
±
SD. * p < 0.05, compared with saline.
Fig. 6. The effects of chronically administered I-6, Ex-4 and liraglutide in diabetic mice. (A) Fasted plasma glucose measured every five days. (B) HbA1C measured at day 24.
C) IPGTT test results in mice at day 24. (D) AUC of IPGTT test results in mice at day 24. n = 6 per group. Data are given as mean ± SD. *p < 0.05, **p < 0.01, ***p < 0.001
(
compared with saline (control).
increased over the 4-week treatment progression by ꢀ5 g whereas
it was suppressed after administration of either Ex-4, liraglutide
or I-6.
The serum levels of leptin and adiponectin were also amended
after I-6 treatment which were important adipocytokines
secreted from white adipose tissue. As shown in Fig. 8B, leptin
Please cite this article in press as: Cai X., et al. Bioorg. Med. Chem. (2018), https://doi.org/10.1016/j.bmc.2018.04.022

6
X. Cai et al. / Bioorganic & Medicinal Chemistry xxx (2018) xxx–xxx
Fig. 7. Effects of I-6 on islet size. Histological analysis was performed on the islet tissue from saline treated, normal mice, diabetic mice, Ex-4, liraglutide and I-6 treated
diabetic mice, respectively (A). Sections (5 m) were stained with H&E and the area of each islet. Islet areas were assessed by manually drawing around the islet perimeter
with the measuring tool contained in the software Image-Pro Plus 6.0. n = 6 per group (B). *p < 0.05, **p < 0.01 compared with saline (control).
l
concentration circulating in DIO mice was notably higher than that
in lean mice controls (p < 0.05). Meanwhile, I-6 administration
notably (p < 0.05) reduced leptin level circulating to a near normal
value of 2.5 ng/ml. Nearly 39% serum adiponectin level in DIO was
reduced after prolonged exposure to high diet induced compare
with that in lean mice (p < 0.05), and this level was significantly
increased (p < 0.05) by I-6. Recent evidence shows that overweight
represents high risk factors for steatohepatitis.1 In present study,
hepatocyte damage was also assessed by examining serum enzyme
activities of aspartate aminotransferase (AST) and alanine amino-
transferase (ALT). The result showed that I-6 treatment resulted
in a reduction in ALT (p < 0.05) and AST (p < 0.05) than that in sal-
ine treated LFD mice. Moreover, similar reduction trend was also
detected in serum total cholesterol (TC) and triglyceride (TG) (p
of cysteine residues (Cys17 and Cys37). The following in vitro
albumin binding test and stability test revealed that the plasma
stability and albumin binding ability of all conjugates I-1 ꢀ I-6
were improved to different degrees. Notably, the plasma stabilities
of the test conjugates associated well with their albumin binding
abilities. I-6 exhibited the longest plasma half-life and the highest
albumin binding ability, I-6 was further estimated for in vivo glu-
cose-lowering abilities and insulinotropic activities. It was shown
that I-6 showed comparable protection effects against glucolipo-
toxicity, oxidative stress and DNA damage mediated stress (Figs. 3
and 4), insulinotropic activities and glucose-lowering abilities with
those of liraglutide and Ex-4 (Fig. 5). Vitally, the in vivo biological
action of I-6 was observed to be exerted via the glucose-dependent
action mode. This action offered the advantage of increased drug
safety compared with drugs that increased insulin secretion via
glucose-independent mechanisms (e.g., sulfonylureas), indicating
that I-6 can be used clinically without the risk of hypoglycemia.
Then it was evaluated that the in vivo long-acting glucose-lower-
ing profiles of I-6. What’s more, the long-acting characteristic of
I-6 was notably more improved than Ex-4, most likely due to its
reduced renal clearance resulting from the strong albumin binding
ability and enzyme metabolism. The absorption of I-6 into sys-
temic circulation was delayed possibly due to the formation of
self-assembled colloidal structures in aqueous environment.14 It
is interesting to find of this study that concerned albumin binding
by I-6 and its long-acting profiles were harmonious with previous
successful reports regarding the protracted in vivo circulation half-
life values of fatty acid chain-modified GLP-1 analogs, attributed to
their physical interactions with serum albumin. The long-acting
glucose-lowering effects of I-6 were further corroborated by the
hypoglycemic duration test in STZ-induced diabetic mice. The
hypoglycemic durations of I-6 in diabetic mice were much longer
than those of Ex-4 and comparable with those of liraglutide,
regardless of the dose administered. Furthermore, once daily
administration of I-6 to STZ-induced diabetic mice achieved
long-term beneficial effects on HbA1c lowering and glucose toler-
3
<
0.05 and p < 0.05, respectively, compared to DIO mice). In addi-
tion, the ratio between ‘‘good” (HDL) and ‘‘bad” (LDL) cholesterol
values, LDL/HDL, were considered to analyze the degree of
increases in the two lipoproteins in different groups. As shown in
Fig. 8G, prolonged exposure to HFD increased this ratio by 20%
while I-6 treated DIO mice lead to a reduction trend but not signif-
icant following administration. Furthermore, the brown adipocytes
and the white adipocytes were gathered after treatment. As shown
in Fig. 9, the white or brown adipocytes cells in I-6 treated DIO
mice was significantly smaller than that in DIO mice. I-6 could
reduce body weight in obese mice.
3
. Discussion
In this study, a series of novel fatty chain modified GLP-1 ana-
logs were designed and prepared to search for long-acting GLP-1
conjugates with retained efficacy. The initial GLP-1 receptor activa-
tion assay granted direct evidence of the relevant structure-activity
relationships. As stated in Table 1, the receptor activation of com-
pounds I-1 ꢀ I-6 depended on the length of the fatty acid chain of
fatty acid chain-maleimide and were independent of the position
Please cite this article in press as: Cai X., et al. Bioorg. Med. Chem. (2018), https://doi.org/10.1016/j.bmc.2018.04.022

X. Cai et al. / Bioorganic & Medicinal Chemistry xxx (2018) xxx–xxx
7
Fig. 8. Effects of I-6 on food and water intake. Time course for chronic effect of I-6 administered for 4 weeks in DIO mice on cumulative food intake. Time course for chronic
effect of I-6 administered for 4 weeks in DIO mice on cumulative food (A) and water (B) intake. Time course of I-6 effect on changes in body weight. Time course of acute effect
on food intake in fasted mice with intraperitoneal doses of I-6. Effects of I-6 on serum biomarkers examined at the end of the 4-week treatment. Serum TC (E), TG (F), HDL/LDL
(
G), ALT (H), AST (I), leptin (J) and adiponectin (K) levels were assayed on a by Beckman Coulter Chemistry Analyzer. TC, total cholesterol; TG, triglyceride; HDL, high density
lipoprotein; LDL, low density lipoprotein; ALT, alanine aminotransferase; AST, aspartate aminotransferase. n = 6 per group. Data are given as mean ± SD. *p < 0.05, **p < 0.01,
**p < 0.001 compared with HDF (saline).
*
ance. The enhanced glucose tolerance of I-6 treated mice probably
resulted from the prolonged biological activity (b cell neogenesis
and/or proliferation). Including appetite suppression and weight
loss reduction, the results of chronic administration of I-6 to DIO
mice were also consistent with the chronic effects of administered
GLP-1 receptor agonists.
In summary, this present study demonstrates that the cysteine
substitution of GLP-1 residues and further fatty acid chain modifi-
cation offers a useful scientific approach to the development of
long-acting incretin-based antidiabetics. This study also shows
that I-6, preserved biological activity, long-acting antidiabetic
characteristics, and long-term beneficial effects well, is a promising
biological agent deserving further investigation to treat obesity
patients with diabetes.
purchased from Cisbio (Bedford, MA, USA). HbA1c kit was pur-
chased from Glycosal (Deeside, UK). Serum leptin and adiponectin
levels using EZML-82K and EZMADP-60K ELISA kits supplied by
Linco Research (St. Charles, MO, USA), respectively. Serum TG, TC,
HDL, LDL, ALT and AST were assayed on a by Beckman Coulter
Chemistry Analyzer (Beckman Coulter, CA, USA). Unless indicated,
all other reagents were purchased from Sigma (Saint Louis, MO,
USA). Peptides were synthesized in a Discover focused single mode
microwave synthesis system (CEM, NC, USA) using microwave irra-
diation procedures at 2450 MHz and the mass of obtained peptides
and target conjugates were confirmed by Waters ACQUITY UPLC
Systems (Waters, Milford, MA, USA). Sprague–Dawley rats (SD rats,
male, 200–250 g) and Kunming mice (male, 10 weeks old) were
purchased from the comparative medical center of Yangzhou
University (Jiangsu, China). C57BL/6J mice (male, 6–8 weeks old)
were obtained from Jiesijie Laboratory Animal (Shanghai, China).
Animals were housed in groups of three (rat) and six (mice) in
cages under controlled temperature (22 ± 2 °C) and relative air
humidity (set point 50%) with a 12 h light:12 h dark cycle. Tap
water and standard laboratory chow were provided ad libitum
throughout the study. C57BL/6J mice were fed with a DIO
(D12492; 60% fat, 20% protein and 20% carbohydrate; 5.24 kcal/g)
or a LDF (D12450B; 10% fat, 20% protein and 70% carbohydrate;
3.85 kcal/g) and watered ad libitum. Both diets were supplied by
1
5
4
. Experimental section
4.1. Materials and animals
Fmoc-protected amino acids, Fmoc Rink Amide-MBHA resin,
liraglutide and Ex-4 were purchased from GL biochem (Shanghai,
China). Acetonitrile and methanol (HPLC grade) were obtained
from Merck (Darmstadt, Germany). cAMP dynamic kit was
Please cite this article in press as: Cai X., et al. Bioorg. Med. Chem. (2018), https://doi.org/10.1016/j.bmc.2018.04.022

8
X. Cai et al. / Bioorganic & Medicinal Chemistry xxx (2018) xxx–xxx
Fig. 9. Effects of I-6 on white or brown adipocytes cells.
Research Diets (New Brunswick, NJ, USA). All animal experimental
protocols were approved by an ethical committee at China Phar-
maceutical University and conducted according to the Laboratory
Animal Management Regulations in China and adhered to the
Guide for the Care and Use of Laboratory Animals published by
the National Institutes of Health (revised 2011). The experiments
were conducted in such a way that the number of animals used
and their suffering was minimized. Prior to the blood sampling,
animals were anesthetized with diethyl ether.
Cys17-Gly
8
-GLP-1(7-36)-NH
2
8
(1) and Cys37-Gly -GLP-1(7-37)-
NH (2). Then the peptide was purified on preparative RP-HPLC
2
11
and was identified by electrospray mass spectrometry. Gener-
ally speaking, Fmoc Rink Amide-MBHA resin underwent repeated
procedures of deprotection and coupling with relevant Fmoc-
protected amino acids and then the final peptide was cleaved
from the resulting resin by reagent K (TFA/thioanisole/water/
phenol/EDT, 82.5:5:5:5:2.5) for 1.5 h at room temperature. The
crude peptides were purified on Shimadzu preparative RP-HPLC
with the following condition: Shimadzu C18 reversed-phase col-
4.2. General synthetic route and HPLC purification of peptides 1 and 2
umn (5
l
m, 340 mm  28 mm), a linear gradient of mobile phase
3
0–75% B in 30 min at a flow rate of 6.0 mL/min (mobile phase
Two optimized polypeptide chains were synthesized by the
method of the standard solid-phase peptide synthesis protocol,
A: water containing 0.1% TFA, mobile phase B: acetonitrile con-
taining 0.1% TFA).
Please cite this article in press as: Cai X., et al. Bioorg. Med. Chem. (2018), https://doi.org/10.1016/j.bmc.2018.04.022

X. Cai et al. / Bioorganic & Medicinal Chemistry xxx (2018) xxx–xxx
9
liraglutide) was 1000 ng/mL at 37 °C. At 0, 1, 2, 4, 6, 8, 12, 24, 36,
8 and 72 h time points, 100 L mixture was aliquoted and
4
l
extracted on an Oasis HLB 96-well plate (Waters, Milford, MA,
USA) and then analyzed by the LC-MS/MS system. The signal of test
articles was detected by multiple reaction monitoring with the use
of electrospray ionization mass spectrometry on a Sciex API-4000
and Turbo Ionspray (Applied Biosystems, Foster City, CA, USA).
The condition of reverse phase liquid chromatographic separation
was as aforementioned.
Scheme 1. Synthetic route of fatty acid chain-maleimide.
4.7. Albumin binding
4.3. General synthetic route of fatty acid chain-maleimide
The albumin binding properties of the fatty acid chain modified
conjugates were investigated in an albumin-binding assay with the
use of albumin-conjugated Sepharose resin according to the
As shown in Scheme 1, three fatty acid chain maleimide (3-(2,
-Dioxo-2H-pyrrol-1(5H)-yl)propanoic acid, 6-(2,5-Dioxo-2H-pyr-
5
1
0
reported literatures. Firstly, albumin-conjugated resin was pre-
pared by mixing NHS-activated Sepharose 4 fast flow (5 mL, wet
resin volume, Amersham Bioscience, Uppsala, Sweden) and human
serum albumin (HSA; 33 mg in 10 mM PBS, pH 7.4) to react at
room temperature by gentle shaking for 4.5 h. Then, the albu-
min-conjugated resin was recovered by centrifugation (1000 rpm,
rol-1(5H)-yl)hexanoic acid, 12-(2,5-Dioxo-2H-pyrrol-1(5H)-yl)do-
decanoic acid were synthesized as previously reported.10
4
.4. General synthetic route of conjugates (I-1, I-2, I-3, I-4, I-5, I-6)
The conjugates were synthesized as previously reported. Cys-
5
min). Meanwhile, the resin was inactivated by hydrolysis of the
teine altered peptide (5
mol) were reacted in 10 mL of 0.05 mol/L sodium phosphate buf-
fer (pH 7.0) at 20 °C under N till UPLC showed completion. The
analytical condition was as follows: Acquity UPLC HSS T3 column
1.8 m, 2.1 mm * 100 mm, Waters); a linear gradient of mobile
lmol) and fatty chain-maleimide (12
NHS-active ester to give an albumin-free control resin. The HSA
content of the wet resin was 6–7 mg/mL. Conjugates or liraglutide
l
2
(
100 lg/mL in PBS, 50 lL) were mixed with HSA resin or albumin-
free resin and incubated for 3 h at room temperature (25 °C). The
supernatant was separated by centrifugation (1000 rpm, 10 min)
to determine the unbound peptide contents using the aforemen-
tioned method by LC-MS/MS system.
(
l
phase 5–95% B (mobile phase A: water with 0.2% formic acid,
mobile phase B: acetonitrile with 0.2% formic acid) in 3.5 min at
a flow rate of 0.3 mL/min with ultraviolet (UV) detection at 214
nm. The crude conjugate was purified on Shimadzu preparative
RP-HPLC.
4.8. Cell viability assay
Cell viability assay was conducted as previously reported.16
INS-1 cells were maintained in RPMI 1640 medium (Gibco, Grand
Island, NY, USA) with 11.2 mM glucose supplemented with 10%
heat-inactivated fetal calf serum, 2 mM L-glutamine, 1 mM sodium
pyruvate, 10 mM N-2-hydroxyethylpi perazine-N-ethane-sulfonic
acid, 50 lM b-mercaptoethanol, 100 U/ml penicillin and 100 mg/ml
4
.5. GLP-1 receptor activation assay
The GLP-1 receptor activation assay was conducted as previ-
1
6
ously reported. To put it simply, our group constructed HEK293
cells over expressing human GLP-1 receptor stably, which was
used to assess the potency of peptides and conjugates toward the
GLP-1 receptor. Cells were grown in Dulbecco’s modified Eagle’s
medium-31053 (Invitrogen, Carlsbad, CA, USA) supplemented with
streptomycin at 37 °C in a humidified atmosphere (5% CO
INS-1 cells were seeded in 96- well plates (ꢀ5000 cells/well) for
4 h, then cells were treated with glucolipotoxicity media consist-
ing of RPMI media made to 25 mM glucose and 0.4 mM oleate
oleic acid-albumin from bovine serum, Sigma, Saint Louis, MO,
USA), or treated with H (50 M) to induce oxidative stress, or
treated with 1 M streptozotocin (STZ, b-cell-specific DNA-damag-
ing agent) to induce apoptosis followed by the addition of Ex-4
10 nM), liraglutide (10 nM) or I-3 (1, 10 or 100 nM). After incuba-
tion for 24 h, cell viability was measured by adding 200 g/ml
-(4,5-Dimethylthiazol-2-yl)-2, 5-diphenyltetrazolium bromide
MTT) (Dingguo, Beijing, China) and incubated for 3 h at 37 °C.
2
).
1
7
2
0
2
.5% FBS, 2 mmol/L
0 mmol/L HEPES (Sigma, Saint Louis, MO, USA), 50 units/ml peni-
g/ml streptomycin (Gibco, Grand Island, NY, USA) at
. Cells were plated in 96-well half area, solid black
L-glutamine (Sangon Biotech, Shanghai, China),
(
cillin, and 50
3
l
2
O
2
l
7 °C in 5% CO
2
l
microplates 2 h before the test started. Meanwhile, test articles
were solubilized in DMSO and further diluted in medium contain-
ing 0.1% BSA fraction V (Genview Scientific, Florida, USA). The
resulting solution was added to cells and incubated for 20 min,
then assayed for cAMP using the cAMP dynamic 2 kit with
homogenous time-resolved fluorescence technology (Cisbio, Bed-
ford, MA, USA) using an Envision 2104 Multilabel Reader according
to the manufacturer’s instructions. The potency of the conjugates
(
3
(
The reaction was stopped and the purple formazan precipitate
formed was dissolved in dimethyl sulfoxide and the color intensity
was measured at 570 nm with a multiwall spectrophotometer
(
EC50 values) was calculated by sigmoidal curve fitting using
(
Thermo Labsystems, Waltham, MA, USA).
GraphPad Prism version 7.0 (GraphPad, San Diego, CA, USA).
4.9. Detection of apoptotic cells
4.6. Plasma stability test
In flow cytometry analysis, INS-1 cell were seeded in 96-well
With a modification of a previously described method, Plasma
plates (ꢀ5000 cells/well) for 24 h, then treated with glucolipotox-
were obtained from adult male Sprague–Dawley (SD) rats (Certifi-
cate No.: 0011253) and compounds were then incubated with
plasma over 72 h. Samples from each time points underwent solid
phase extraction and the resulting extract was analyzed by LC-MS/
MS system to assess the profile of plasma degradation. In vitro sta-
bility test, the initial concentration in rat plasma of fatty chain
modified conjugates I-1 ꢀ I-6 and the positive controls (Ex-4 and
icity media, H (50 M) or 1 M STZ followed by the adding of
2
O
2
l
Ex-4 (10 nM), liraglutide (10 nM) or I-6 (10 nM) according to the
previously reported methods. After incubation for 24 h, cells were
collected by trypsinization, washed with cold PBS and incubated
with Annexin V-FITC (Sigma, Saint Louis, MO, USA) for 15 min, then
stained with propidium iodide (PI, Sigma, Saint Louis, MO, USA).
Cells in early apoptosis were Annexin V-FITC positive and PI nega-
Please cite this article in press as: Cai X., et al. Bioorg. Med. Chem. (2018), https://doi.org/10.1016/j.bmc.2018.04.022

1
0
X. Cai et al. / Bioorganic & Medicinal Chemistry xxx (2018) xxx–xxx
tive. The percentage of apoptotic cells was analyzed by the flow
cytometry (Beckman Coulter Accuri C6).
nmol/kg Ex-4, liraglutide or I-6, respectively with saline as control
after randomly assigned to treatment groups (n = 6) with matched
body weight. C57BL/6 mice fed with LFD (n = 6/group) were used
to index responses to normal values. Water consumption, food
intake, and body weight were measured daily. At the end of the
study, blood samples were collected and sera separated subse-
quently for further analyses.
4.10. Glucoregulatory and insulin secretion assay
The glucoregulatory and insulin secretion assays were carried
10
out in accordance with a previous method. Briefly, overnight
fasted (12 h) SD rats (n = 6/group, 200–250 g) were orally adminis-
trated with I-6, Ex-4 and liraglutide (25 nmol/kg) with saline used
as negative control. Male SD rats were administered saline, Ex-4,
liraglutide or I-3 half hour prior to oral glucose load (10 g/kg)
4.14. Data analysis and statistical assessment
Data were analyzed using Prism version 7 software (GraphPad,
San Diego, CA, USA). General effects were tested using a 1-way
ANOVA with Tukey’s multiple-comparison post hoc test. Data
throughout are stated as mean ± SD. p < 0.05 considered
significant.
(
the time point was set as 0 min). At 30, 0, 15, 30, 45, 60, 90, 120
and 180 min, blood sample was collected from the cut tip of the tail
vein to measure the blood glucose levels using a blood glucose
monitor. Meanwhile, blood samples (0.1 mL) were collected in
EDTA-containing microcentrifuge tubes from the lateral tail vein
at the same aforementioned time point. Plasma samples were then
obtained by centrifugation (1000 rpm, 15 min) and assayed for
insulin levels using a Rat Insulin ELISA kit (Millipore, Billerica,
MA, USA).
Acknowledgements
This study was supported by grants from National Natural
Science Foundation of China (81673299 and 81703346), National
Science
(
(
and
2018ZX09301034004), Science and Technology Projects of Jiaxing
2017AY33074) and National Found for Fostering Talents of Basic
Technology
Major
Project
of
China
4
.11. Multiple intraperitoneal glucose tolerance test
Multiple intraperitoneal glucose tolerance test (IPGTT) was car-
Science of China (J1310032). We thank Mr. Yanyong Yang and Dr.
Xin Su for giving assistant generously.
Notes
ried out on C57BL/6 mice to assess the ability of I-6 to reduce glu-
cose levels for longer time using a modification of a previously
1
0
described method. Briefly, the fasted overnight (12 h) male
C57BL/6 mice (n = 6/group, 6 weeks) were first intraperitoneally
administrated with saline, Ex-4, liraglutide, or I-6 (25 nmol/kg).
Half an hour later, they were intraperitoneally challenged with glu-
cose (2.0 g/kg) (0 h) at the interval of 6 h to simulate the condition
of three meals a day. The blood glucose levels were monitored at 0,
The authors declare no conflict of interest to disclose.
A. Supplementary data
Supplementary data associated with this article can be found, in
the online version, at https://doi.org/10.1016/j.bmc.2018.04.022.
0
.25, 0.5, 1, 2, 3 h using a blood glucose monitor. The following glu-
cose load time points were at 6 and 12 h after the first IPGTT.
Reference
4
.12. Hypoglycemic efficacies test
1
2
3
.
.
.
Ali S, Davies MJ, Brady EM, et al. Guidelines for managing diabetes in Ramadan.
Diabetic Med. 2016;33:1315–1329.
Bagger JI, Knop FK, Holst JJ, Vilsboll T. Glucagon antagonism as a potential
therapeutic target in type 2 diabetes. Diabetes Obes Metab. 2011;13:965–971.
Manandhar B, Ahn JM. Glucagon-like Peptide-1 (GLP-1) analogs: recent
Hypoglycemic efficacies of I-6 were evaluated in STZ-induced
diabetic mice using a modification of a previously described
1
0
method. Kunming mice received a single intraperitoneal injec-
tion of saline, Ex-4, liraglutide, I-6 (25 nmol/kg) with free access
to food and water. At 0, 1, 2, 3, 4, 6, 8, 12, 16, 20, 24, 30 and 36
h, the second drop of blood was drawn from a tail vein to deter-
mine the blood glucose levels by blood glucose monitor. Moreover,
we also calculated the hypoglycemic durations with the blood glu-
cose level <8.35 nmol/L (150 mg/dL).
advances, new possibilities, and therapeutic implications.
2015;58:1020–1037.
J Med Chem.
4
.
Emery EC, Diakogiannaki E, Gentry C, et al. Stimulation of GLP-1 Secretion
Downstream of the Ligand-Gated Ion Channel TRPA1. Diabetes.
2015;64:1202–1210.
5. Sun L, Huang X, Han J, et al. Site-specific fatty chain-modified exenatide analogs
with balanced glucoregulatory activity and prolonged in vivo activity. Biochem
Pharmacol. 2016;110:80–91.
6
.
Kim TH, Jiang HH, Lee S, et al. Mono-PEGylated dimeric exendin-4 as high
receptor binding and long-acting conjugates for type
2 anti-diabetes
therapeutics. Bioconjugate Chem. 2011;22:625–632.
4
4
.13. Chronic in vivo studies
7
.
Luginbuhl KM, Schaal JL, Umstead B, et al. One-week glucose control via zero-
order release kinetics from an injectable depot of glucagon-like peptide-1 fused
to a thermosensitive biopolymer. Nat. Biomed. Eng.. 2017;1(6).
Gaberc-Porekar V, Zore I, Podobnik B, Menart V. Obstacles and pitfalls in the
PEGylation of therapeutic proteins. Curr Opin Drug Discov Devel. 2008;11:242.
Jonassen I, Havelund S, Hoegjensen T, Steensgaard DB, Wahlund P, Ribel U.
Design of the novel protraction mechanism of insulin degludec, an ultra-long-
acting basal insulin. Pharm Res. 2012;29:2104–2114.
.13.1. Chronic in vivo studies on STZ-induced diabetic mice
Male KM mice, 6–8 weeks old, weighing 18–22 g, were injected
8
.
.
with STZ intraperitoneally (40 mg/kg/day) for 5 consecutive days
after fasted overnight. Five days after STZ injection, blood glucose
was measured to validate diabetic hyperglycemia. Mice with fast-
ing blood glucose level 11.1 mM or higher were assigned to groups
with matched body weight and were injected intraperitoneally
with 25 nmol/kg Ex-4, liraglutide, or 2c once daily for 3 weeks.
On days 0, 5, 10, 15, and 20, overnight fasting blood samples were
obtained from the tail vein to measure blood glucose levels. At the
end of the study, mice were sacrificed and blood samples were
collected to determine the levels of HbA1c by Beckman Coulter
Chemistry Analyzer (Tokyo, Japan).
9
1
0. Han J, Sun L, Chu Y, et al. Design, synthesis, and biological activity of novel
dicoumarol glucagon-like peptide conjugates. Med Chem.
2013;56:9955–9968.
1
J
1
1. Han J, Huang X, Sun L, Li Z, Qian H, Huang W. Novel fatty chain-modified
glucagon-like peptide-1 conjugates with enhanced stability and prolonged
in vivo activity. Biochem Pharmacol. 2013;86:297–308.
12. Young AA, Gedulin BR, Bhavsar S, et al. Glucose-lowering and insulin-
sensitizing actions of exendin-4: studies in obese diabetic (ob/ob, db/db)
mice, diabetic fatty Zucker rats, and diabetic rhesus monkeys (Macaca
mulatta). Diabetes. 1999;48:1026–1034.
3. Olleros ML, Martin ML, Vesin D, et al. Fat diet and alcohol-induced
steatohepatitis after LPS challenge in mice: role of bioactive TNF and Th1
type cytokines. Cytokine. 2008;44:118–125.
4. Kim K, Kwon S, Park JH, et al. Physicochemical characterizations of self-
assembled nanoparticles of glycol chitosan-deoxycholic acid conjugates.
Biomacromolecules. 2005;6:1154–1158.
1
4
.13.2. Chronic in vivo studies on DIO mice
C57BL/6 mice (6 weeks) were maintained on prescribed DIO for
2 weeks. They were injected intraperitoneally once daily with 25
1
1
Please cite this article in press as: Cai X., et al. Bioorg. Med. Chem. (2018), https://doi.org/10.1016/j.bmc.2018.04.022

X. Cai et al. / Bioorganic & Medicinal Chemistry xxx (2018) xxx–xxx
11
1
5. Rolin B, Larsen MO, Gotfredsen CF, et al. The long-acting GLP-1 derivative
NN2211 ameliorates glycemia and increases b-cell mass in diabetic mice. Am J
Physiol-Endoc M. 2002;283(4).
17. Sun L, Wang C, Dai Y, et al. Coumaglutide, a novel long-acting GLP-1 analog,
inhibits b-cell apoptosis in vitro and invokes sustained glycemic control in vivo.
Eur J Pharmacol. 2015;767:211–219.
1
6. Han J, Sun L, Huang X, et al. Novel coumarin modified GLP-1 derivatives with
enhanced plasma stability and prolonged in vivo glucose-lowering ability. Brit J
Pharmacol. 2014;171:5252–5264.
Please cite this article in press as: Cai X., et al. Bioorg. Med. Chem. (2018), https://doi.org/10.1016/j.bmc.2018.04.022