Discovery of novel allosteric site and covalent inhibitors of FBPase with potent hypoglycemic effects

Article abstract of DOI:10.1016/j.ejmech.2019.111749

Fructose-1,6-bisphosphatase (FBPase) is an essential enzyme of GNG pathway. Significant advances demonstrate the FBPase plays a critical role in treatment of diabetes. Numerous FBPase inhibitors were developed by targeting AMP site, nevertheless, none of these inhibitors has exhibited suitable potency and druggability. Herein, a new allosteric site (C128) on FBPase was discovered, and several nitrostyrene compounds exhibiting potent FBPase inhibitions were found covalently bind to C128 site on FBPase. Mutagenesis suggest that C128 is the only cysteine that can influence FBPase inhibition, the N125–S124–S123 pathway was most likely involved in allosteric signaling transmission between C128 and active site. However, these nitrostyrenes may bind with multiple cysteine besides C128 in FBPase. To improve pocket selectivity, a series of novel compounds (14a-14n) were re-designed rationally by integrating fragment-based covalent virtual screening and machine-learning-based synthetic complexity evaluation. As expected, the mass spectrometry validated that the proportion of title compounds binding to the C128 in FBPase was significantly higher than that of nitrostyrenes. Notably, under physiological and pathological conditions, the treatment of compounds 14b, 14c, 14i or 14n led to potent inhibition of glucose production, as well as decreased triglyceride and total cholesterol levels in mouse primary hepatocytes. We highlight a novel paradigm that molecular targeting C128 site on FBPase can have potent hypoglycemic effect.

Full text of DOI:10.1016/j.ejmech.2019.111749

Journal Pre-proof
Discovery of novel allosteric site and covalent inhibitors of FBPase with potent
hypoglycemic effects
Yunyuan Huang, Lin Wei, Xinya Han, Haifeng Chen, Yanliang Ren, Yanhong Xu,
Rongrong Song, Li Rao, Chen Su, Chao Peng, Lingling Feng, Jian Wan
PII:
S0223-5234(19)30901-8
DOI:
https://doi.org/10.1016/j.ejmech.2019.111749
EJMECH 111749
Reference:
To appear in: European Journal of Medicinal Chemistry
Received Date: 25 May 2019
Revised Date: 20 September 2019
Accepted Date: 28 September 2019
Please cite this article as: Y. Huang, L. Wei, X. Han, H. Chen, Y. Ren, Y. Xu, R. Song, L. Rao, C. Su, C.
Peng, L. Feng, J. Wan, Discovery of novel allosteric site and covalent inhibitors of FBPase with potent
hypoglycemic effects, European Journal of Medicinal Chemistry (2019), doi: https://doi.org/10.1016/
j.ejmech.2019.111749.
This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition
of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of
record. This version will undergo additional copyediting, typesetting and review before it is published
in its final form, but we are providing this version to give early visibility of the article. Please note that,
during the production process, errors may be discovered which could affect the content, and all legal
disclaimers that apply to the journal pertain.
© 2019 Published by Elsevier Masson SAS.

Graphical Abstract.

Discovery of novel allosteric site and covalent
inhibitors of FBPase with potent
hypoglycemic effects
1,#
Yunyuan Huang,1,# Lin Wei,
Xinya Han,1,2,# Haifeng Chen,1,3,# Yanliang Ren,1,*
Yanhong Xu,1 Rongrong Song, 1 Li Rao,1 Chen Su,4 Chao Peng4, Lingling Feng,1 Jian
Wan,1,*
1International Cooperation Base of Pesticide and Green Synthesis (Hubei), Key
Laboratory of Pesticide & Chemical Biology (CCNU), Ministry of Education,
Department of Chemistry, Central China Normal University, Wuhan 430079, China
2School of Chemistry and Chemical Engineering, Anhui University of Technology,
Maanshan 243002, China
3Ocean College, Beibu Gulf University, Qinzhou 535011, Guangxi, P. R. China
4
National Facility for Protein Science in Shanghai, Zhangjiang Lab, Shanghai,
201210, China

Highlights
•
•
A new cryptic allosteric site (C128) in FBPase was discovered.
Several nitrostyrene compounds (12 and 13) exhibiting potent FBPase inhibitions
were found covalently bind to C128 site on FBPase.
•
The N125-S124-S123 pathway was involved in allosteric signaling transmission
between C128 and active site.
• The treatments of compounds 14a, 14c, 14i or 14n led to potent inhibition of
glucose production, as well as decreased triglyceride and total cholesterol levels in
mouse primary hepatocytes.
KEYWORDS Fructose-1,6-bisphosphatase (FBPase), Novel allosteric site, Covalent
inhibitors, Fragment-based drug design

Abstract
Fructose-1,6-bisphosphatase (FBPase) is an essential enzyme of GNG pathway.
Significant advances demonstrate the FBPase plays a critical role in treatment of
diabetes. Numerous FBPase inhibitors were developed by targeting AMP site,
nevertheless, none of these inhibitors has exhibited suitable potency and druggability.
Herein, a new allosteric site (C128) on FBPase was discovered, and several
nitrostyrene compounds exhibiting potent FBPase inhibitions were found covalently
bind to C128 site on FBPase. Mutagenesis suggest that C128 is the only cysteine that
can influence FBPase inhibition, the N125-S124-S123 pathway was most likely
involved in allosteric signaling transmission between C128 and active site. However,
these nitrostyrenes may bind with multiple cysteine besides C128 in FBPase. To
improve pocket selectivity, a series of novel compounds (14a-14n) were re-designed
rationally by integrating fragment-based covalent virtual screening and
machine-learning-based synthetic complexity evaluation. As expected, the mass
spectrometry validated that the proportion of title compounds binding to the C128 in
FBPase was significantly higher than that of nitrostyrenes. Notably, under
physiological and pathological conditions, the treatment of compounds 14b, 14c, 14i
or 14n led to potent inhibition of glucose production, as well as decreased triglyceride
and total cholesterol levels in mouse primary hepatocytes. We highlight a novel
paradigm that molecular targeting C128 site on FBPase can have potent
hypoglycemic effect.

1. Introduction
Fructose-1,6-bisphosphatase (FBPase) is an essential enzyme of the
gluconeogenesis (GNG) pathway, and a predominant factor involved with increased
hepatic glucose output[1, 2]. Numerous FBPase inhibitors have been explored as
potential agent for the treatment of diabetes, especially type 2 diabetes mellitus
(T2D),[3, 4] thus FBPase also represents an attractive target for diabetes
treatment.[5-9] To the best of our knowledge, most FBPase inhibitors reported
previously[10] bind to the AMP site on FBPase via non-covalent interactions (e.g.,
hydrogen-bonds, salt bridges, van der Waals interactions). Nevertheless, none of these
inhibitors exhibit suitable potency, specificity or biological activity, except compound
MB06322(CS-917).[11, 12] Although MB06322 has been undergoing clinical
evaluation,[13, 14] phase I clinical studies have indicated that MB06322 can cause
lactic acidosis in human hepatocytes[15].
On the other hand, covalent drugs have been profoundly successful in the
treatment of a wide array of human diseases. In particular, the clinical success of
ibrutinib[16] and afatinib[17] has prompted a resurgence of interest in covalent
inhibitor discovery.[18, 19] Compared with traditional non-covalent drugs, covalent
inhibitors possess exceptionally high potency, ligand efficiency and exhibit
long-lasting effects, because covalent bonds are significantly stronger than
non-covalent interactions.[20, 21] In fact, previous covalent drugs are usually
designed by attaching a “covalent warhead” to non-covalent inhibitors. For instance,

the covalent drug afatinib, targeting a Bruton’s tyrosine kinase-associated cysteine
for treatment of non-small-cell lung carcinomas (NSCLCs), stem from the
non-covalent inhibitor gefitinib and an acrylamide “warhead”[22, 23]. In other words,
the non-catalytic residues that might covalently react with the “warhead” are
indispensable[20, 24] for the discovery of covalent inhibitors. As an uncommon
residue, non-catalytic cysteine is not highly conserved in the enzyme sequence[23, 25]
and is often observed in the binding pocket of covalent inhibitors[20, 26] due to the
electrostatic properties of residues. Targeting of non-catalytic cysteine residues has
emerged as an indispensable strategy for the design of drugs and inhibitors against
protein kinases[22, 27], such as G-protein-coupled receptors (GPCRs)[28],
proteases[29] and GTPases[30]. Many covalent inhibitors that bind to non-catalytic
cysteines of kinases, e.g., ibrutinib, acalabrutinib, afatinib and osimertinib, have
been approved for commercial use by US FDA[20].
Inspired by the deficiency of traditional FBPase inhibitors[15] and the benefit of
covalent drugs[16], [17], our approach reported here will focus on the exploration of
novel cryptic allosteric site and covalent inhibitors of FBPase. Herein, a cryptic site
(cysteine 128) on FBPase was identified and verified using an electrophilic
fragment-based screen strategy[31, 32] for the first time. Furthermore, nitrostyrene
derivatives that covalently bind to the C128 site of FBPase via a nitrovinyl covalent
warhead, which is also included in ranitidine,[33, 34] were screened in this work.
However, the Mass Spectrometry (MS) showed that nitrostyrene might bind with
more than one cysteine residue besides C128 in FBPase. To improve the pocket

selectivity, a drug moiety was attached to nitrostyrene for the generation of new
compounds by integrating the fragment-based covalent virtual screening (FBCVS)
protocol and machine-learning-based synthetic complexity evaluation based on the
C128 site of FBPase. In addition, T2D has become a prevalent public health concern
around the world, and is associated with other cardiovascular complications, such as
hypertension[35], hyperlipidemia[36, 37], and atherosclerosis[38]. Diabetes mellitus
is associated with an increased risk of dyslipidemia and coronary artery disease.
Patients with T2D increased the incidence of coronary artery disease and dyslipidemia
that is 2- to 6-fold compared to persons without diabetes, and this epidemic has
remained unabated in recent years.[36][39] Therefore, the strategy of attaching a
fragment with anti-dyslipidemic function to nitrostyrene may contribute to improve
the selectivity and dual-functionalization of resulting compounds. As expected, the
proportion of title compounds covalently bound to the C128 site of FBPase increased
significantly compared with that of nitrostyrene. The enzymatic experiments indicated
that compound 14b exhibited the potential inhibitory activity against FBPase with an
IC50 of 0.61 µM. Importantly, compounds 14b, 14c, 14i or 14n showed excellent
inhibition of glucose production, as well as the ability to decrease triglycerides (TG)
and total cholesterol (TC) levels.[40]
2. Results and discussions
2.1 Screening of covalent warheads.
Cysteine-reactive screening is the most popular strategy for discovery of cryptic
site. For instance, the C290 site of caspase-7 was identified by using the library

compound FICA[41]. Discovery of the KRASG12C allosteric site also benefited from
this screening method[42]. According to crystal structure (PDB ID: 1FTA), there
were seven cysteines (Fig. S1) near the substrate (FBP) or AMP binding pocket[43]
on FBPase, but no remarkable interaction between these cysteines and the substrate
FBP/AMP[44] could be observed. To our knowledge, cysteine reacts easily with
reactive electrophilic species[45]. Therefore, reactive electrophilic species that are
commonly present in covalent drugs[46] or endogenous species, were selected as
probes to identify a suitable covalent warhead that reacted with the cysteine in
FBPase. As shown in Fig. S2, compound 10 (9-nitro-oleic acid) exhibited higher
FBPase inhibition (IC50 = 58 µM) than the other compounds (1–9). In fact, the lipid
containing nitrovinyl moiety is involved in some post-translational modification of
proteins in vivo[47-49]. Ranitidine[33, 34] is also a well-known drug with a nitrovinyl
group. Dimethyl fumarate (2), derived from the endogenous electrophilic species 6
(fumarate)[50], is used for the treatment of multiple sclerosis. These evidences
suggest that nitrovinyl (10) moiety is most likely a suitable covalent warhead for
reaction with the cysteine in FBPase.
Fig. 1. The endogenous nitrostyrene electrophilic 10 (9-nitro-oleic acid) and its

derivatives (11, 12 and 13) structure, as well as their inhibition activities against
FBPase.
2.2 Covalent binding of nitrostyrene to multiple cysteines in FBPase.
Although compound 10 could inhibit FBPase, it is difficult to further synthesize
and optimize because its structure is too complex. To identify the binding pocket of
nitrovinyl covalent warhead on FBPase, three simple nitrovinyl derivatives (11, 12
and 13) were further synthesized. Fig. 1 shows that cyano- and fluorine-substituted
nitrostyrene (12 and 13) exhibited a 150-fold increase in inhibitory activities
compared with compound 10 (IC50 = 58 µM). To verify that these nitrostyrene
derivatives may react with the thiol group, 100 µM compound 12 was further reacted
with β-mercaptoethanol (βME). The UV-visible absorption (Fig. S3A) gradually
decreased as the βME concentration increased, this suggests that compound 12 can
actually react with the thiol group. The Kd (54 µM) was determined by plotting the
proportion of the βME adduct vs βME concentration (Fig. S3B). In fact, nitrovinyl has
also been shown to be highly biologically active[51, 52] via reactions with the thiol
group of cysteine. For instance, dipeptidyl nitroalkene might covalently bind to the
CYS25 of the proteases rhodesain and cruzain via the nitrovinyl moiety[53].
3,4-Methylenedioxy-β-nitrostyrene has been demonstrated to inhibit NLRP3 ATPase
activity by covalently binding to the cysteine of NLRP3 for the therapy
inflammasome[54].
To determine the number of cysteine residues in FBPase that could react with
nitrostyrene, compounds 12 and 13 were incubated with FBPase for 2 h. Then, the

covalent adducts were detected by Liquid chromatography−mass spectrometry
(LC−MS)[55]. The MS results (Fig. S4) showed that more than two nitrostyrene
derivatives (12 and 13) could covalently bind with FBPase.
2.3 Identifying the cryptic allosteric site (C128) on FBPase.
To improve the pocket selectivity of nitrostyrene, we first identified a functional
cysteine pocket that can regulate the catalytic activities of FBPase. For this purpose,
seven cysteines in FBPase were mutated to serine. Table 1 shows that only the C128S
mutation could result in significant decrease in the inhibitory activity of 12 and 13
(IC50 > 350 µM), decreasing the FBPase activity by more than 1000-fold compared
with the activity of wild-type (WT) FBPase. This evidence suggests that C128 is the
only cysteine that can influence FBPase inhibition by 12 and 13, although both
compounds may covalently bind with three cysteine residues on FBPase.
Table 1. Kinetic parameters of seven cysteine mutations of FBPase, and the IC50 of 12
/ 13 against mutations and WT.
WT
4.0±0.2
2.5±0.1
0.020±0.002
0.14±0.02
2.3±0.1
0.36±0.08
1
C38S
1.3±0.1
0.79±0.04
0.043±0.005
0.50±0.06
0.37±0.02
0.49±0.08
1.4
C92S
1.6±0.1
C116S
1.6±0.2
1.0±0.1
0.093±0.009
0.13±0.01
0.82±0.05
0.33±0.04
0.94
C128S
2.7±0.2
1.7±0.1
0.10±0.02
1.1±0.1
5.2±1.2
> 350
C179S
2.3±0.1
1.4±0.1
0.083±0.006
0.11±0.01
1.0±0.1
0.43±0.06
1.2
C183S
3.8±0.2
2.4±0.2
0.12 ± 0.02
0.16±0.02
6.3±1.0
0.44±0.02
1.2
C281S
3.6±0.1
2.2±0.1
0.12±0.01
0.23±0.02
1.1±0.1
0.61±0.03
1.7
Vmax (U/mg)
kcat (1/s)
0.99±0.07
0.051±0.009
0.11±0.01
1.3±0.2
Km (mM）
Ka (mM)
AMP IC50 (µM)
12 IC50 (µM)
0.34±0.02
0.97
12(ICM50 / ICW
)
> 1000
>350
50
13 IC50 (µM)
0.32±0.03
1
1.2±0.3
3.8
0.62±0.10
1.9
1.1±0.2
3.4
0.89±0.10
2.8
0.92±0.21
2.8
2.3±0.2
7.2
13(ICM50 / ICW
)
> 1010
50
The crystal structure of FBPase (R state, PDB ID: 1EYI) shows that C128 is
located on the interface of two subunits, and is distant (12 Å) from the substrate

catalysis site (Fig. S5); thus, direct interaction of C128 and the catalytic site of
FBPase is difficult to achieve. In order to verify the allosteric regulation mechanism
of C128, the possible binding model of FBPase and compound 12 was predicted using
Autodock 4[56] and the molecular dynamic (MD) simulation was performed. To our
knowledge, there are two famous mechanisms to explain the reaction of cysteine and
nitroethylene warheads: One was slow-binding mechanism[57] that sulfur group of
cysteine attack the nitro group of nitroethylene warhead. The other was Michael
addition mechanism [54, 58, 59] that nitroethylene warheads was recognized as
Michael acceptor and undergo conjugate addition with thiol group. Hence, two
possible binding modes (Michael acceptor or reaction between thiolate and nitro
group) of nitrostyrene compounds and C128 of FBPase were used for the further
simulations. As shown in Fig. 2A, the MD conformation of N125-S124-S123 (white)
on FBPase almost overlaps with the crystal structure (PDB: 1EYI, green). However, if
nitrostyrene acts as the Michael acceptor, C7 of 12 was attacked by the C128 thiolate.
The MD of FBPase+12 (Fig. 2B, purple) result in the remarkable conformation
change of N125-S124-S123 compared with that of Apo-FBPase (Fig. 2A, white). The
nitro group of 12 and -CONH2 group of N125 would repel each other, which cause
the movement of N125 away from C128, in turn result in the movement (2.3 Å) of the
S123 far from the phosphate of FBP. Similarly, if the reaction occurring between
C128 thiolate and nitro group of 12, the MD result (Fig. 2C) of FBPase+12 (N9 of 12
was attacked) also shows the remarkable conformation change of N125-S124-S123,
and the movement of the S123 far from the phosphate of FBP (Fig. 2C, blue)

compared with that of Apo-FBPase (Fig. 2A, white). On the other hand, one hydrogen
bond between S123 and the phosphate of FBP was observed in the crystal FBPase
structure (Fig. S5). Thus, the addition of nitrostyrene 12 onto C128 could actually
weaken the hydrogen-bond of S123--FBP (Fig. 2B and Fig. 2C) and inhibit FBPase
catalysis activity though N125-S124-S123 allosteric signaling pathway, N125 is a
most important residue on this pathway. Notably, our simulated results indicate that
both covalent binding modes result in the same allosteric regulation mechanism of
C128. Additionally, the addition of nitrostyrene 12 onto C128 could also result in the
movement of Mg2+ (Fig. 2B) far from the phosphate of FBP about 2.5 Å. This result
partly demonstrated that compound 12 may weaken the binding abilities of Mg2+.
Fig. 2. Proposed conformation of compound 12 covalent binding to C128 simulated
by molecular dynamic (MD). (A) Crystal FBPase structure (PDB:1EYI) was
superposed with that (FBPase, white) of MD simulation, (B) The MD structures

superposition of Apo-FBPase (white) and FBPase+12 (purple), with the binding
model that the C7 of 12 was attacked. (C) The MD structures superposition of
Apo-FBPase (white) and FBPase+12 (blue), with the binding model that N9 of 12
was attacked. (D) Plots of the MD simulation root-mean-square deviation (RMSD, in
Å) MD simulation vs time of FBPase and FBPase+12.
To verify the importance of this helix H4 pathway, the residues surrounding this
network were further mutated to Ala. As listed in Table S1, the effects of D127A,
R243A, R254A and Y258A mutations on the Ka (Mg2+ binding affinity) and IC50 (12
and 13) values were almost negligible. According to the crystal structure of R state
FBPase (Fig. S5) and the discussions mentioned-above, N125 is the residue (in helix
4) closest to C128 (3.3 Å), thus the addition of nitrostyrene (12 or 13) onto C128 will
cause the remarkable movement of N125 away from C128, which in turn significantly
decrease the FBPase activity. However, when N125 was mutated to Ala,
N125-S124-S123 will move close to C128, which in turn result in the increase of
FBPase activity. As expected, the N125A lead to more than 500-fold IC50 increase
(Table S1) of compounds 12/13 compared with WT (0.36 µM). In comparison, S124
is far from the C128 (6.8 Å), S124A could not remarkably affect the movement of
N125-S124-S123. Correspondingly, S124A caused 3.8- and 37.8- fold increase of the
IC50 of compounds 12/13. Furthermore, as shown in Table S1, the Kcat value of
S123A (0.041 s-1) decreased approximately 61-fold compared to that of the WT (2.5
s-1), and S123A lacked detectable activity. This suggests that S123 is critical for the
catalytic activity of FBP. These mutant results further confirm our speculation that

C128 could affect the catalytic activity of FBPase via H4 allosteric signaling pathway
(N125-S124-S123), and N125 is a most important residue on H4 pathway.
Furthermore, the effect of Mg2+ on the inhibitory activities of 12 and 13 was also
investigated. As illustrated in Fig. S6, at a fixed Mg2+ concentration, the relative
activity of FBPase may gradually decrease with increasing concentration of
compound 12 or 13, however, the FBPase activities may increase as Mg2+
concentrations increase from 1 mM to 10 mM. This evidence suggests that both 12
and 13 may weaken the binding abilities of Mg2+ via the N125-S124-S123 pathway,
thus competitively inhibiting Mg2+.
Furthermore, 10× the IC50 concentration of compounds 12 and 13 (Fig. S7) were
incubated with FBPase for complete inhibition. Then, the mixture was diluted 20
folds and then detected enzyme catalysis activity. These assays reveal that compounds
12 and 13 covalently likely binds FBPase in a reversible manner. As documented
previously, whether the reaction of nitroethylene warhead and target protein is
slow-binding mechanism[57] or Michael addition mechanism[57, 60, 61], the
corresponding reaction show reversible behavior. Thus, we believe that compound 12
most likely react with C128 of FBPase in a reversible behavior.
Taken together, these findings suggest that, in addition to the FBP and AMP sites,
the C128 pocket of FBPase is a new non-catalytic site, which was identified by our
group for the first time, and could be used for the discovery of novel covalent
inhibitors against FBPase. Based on the document by Beglov et al.[32], such a
binding site is not easily detectable, and is thus called a “cryptic site”.

2.4 Redesign of nitrostyrene-based covalent inhibitors (14a-14n).
Based on the MS experiments in Fig. S4, nitrostyrene could bind with more than
one cysteine residue in FBPase, likely because nitrostyrene is very small, and can
easily bind to multiple cysteine pockets. To improve the selectivity of these
nitrostyrenes, rational optimization by attachment of drug moieties to nitrostyrene
based on the cryptic allosteric pocket (C128) of FBPase is necessary. To achieve this
purpose, a new FBCVS protocol (Fig. S8) based on the C128 pocket of FBPase was
used to generate new hits. In fact, fragment-based approach[23] has been successfully
used for identification of covalent inhibitors for RAS G12C from a library of
disulfides, followed by conversion to an irreversible warhead.[62] Kathman et al.
utilized a fragment-based screening method to test the ability to inhibit papain, a
model cysteine protease.[63] The fragment-based approach has also been validated by
a large number of fragment-to-lead studies reported in recent literature,[64] as well as
by successful translation of more than 30 compounds to clinical studies, and by the
introduction of two FDA approved drugs (vemurafenib and venetoclax) into the
market. Nonetheless, fragment-based covalent inhibitors screening is rare. Moreover,
ensuring the synthetic tractability of the resulting molecule generated by the FBCVS
protocol remains challenging.[65] To overcome this challenge, the synthetic
complexity score (SCscore) developed by Green et al.[66] and learned from Reaxys
reaction database was used to evaluate the compounds designed by this FBCVS
protocol for the first time.

In addition, patients with T2D increased the incidence of coronary artery disease
and dyslipidemia that is 2- to 6-fold compared to persons without diabetes, and this
epidemic has remained unabated in recent years.[36],[39] Attaching a fragment with
anti-dyslipidemic function to nitrostyrene may contribute to improvement of the
selectivity of the resulting compounds, as well as to dual-functionalization of these
compounds. Currently available anti-dyslipidemic drugs mainly include strains,[67,
68] fibrates,[69, 70] niacin[71] and ezetimibe[72], thus, functional moieties on these
drugs were selected as fragment parameters (Fig. S9).
Fig. 3. Designed nitrostyrene derivatives (Top 10) by using FBCVS protocol;
PM7_score is the binding energy calculated by PM7@MOPAC; SCScore is the

synthetic complexity score developed by Green[66] et al.
Combining PM7@MOPAC score and machine-learning-based synthetic
complexity evaluation (ScScore), a second generation of nitrostyrene-based covalent
inhibitors (14a, Table S2 and Fig. 3), which were de novo designed based on the
cryptic allosteric site (C128) on FBPase.
Table 2. The inhibitory activities of synthesized compounds (14a-14n) against
FBPase in vitro
Compds
R1
R2
R3
R4
IC50 (µM)
H
Nitrovinyl
H
H
1.0±0.1
0.61±0.11
2.9±0.2
0.79±0.12
0.83±0.14
1.6±0.2
2.7±0.1
7.0±0.3
8.2±0.3
3.1±0.4
1.6±0.3
1.8±0.2
5.4±0.5
0.84±0.07
2.3±0.1
14a
14b
14c
14d
14e
14f
H
Nitrovinyl
F
H
H
Nitrovinyl
Cl
H
H
Nitrovinyl
Br
H
H
Nitrovinyl
H
Cl
H
Nitrovinyl
H
OCHF2
Nitrovinyl
H
H
H
H
H
H
H
H
H
H
-
14g
14h
14i
Nitrovinyl
H
F
Nitrovinyl
H
Cl
H
Cl
H
H
H
-
H
Nitrovinyl
Nitrovinyl
Nitrovinyl
Nitrovinyl
Cl
14j
H
14k
14l
Cl
CH3
14m
14n
AMP
2Br-2-Nitrovinyl
-
-
FBPase inhibition by synthesized compounds. Although the anti-dyslipidemic
fragment of clofibrate was attached to compound 11, the inhibitory activities of 14a
compounds may be retained at approximately 1.0 µM (Table 2) compared with that of
our previous nitrostyrene compound 11 (3.5 µM). To discover potent hit compounds, a

series of nitrostyrene derivatives were further synthesized (Table 2) and characterized
(Supplementary Methods) to explore the structure-activity relationships (SARs).
When a nitrovinyl group was introduced at R2 of these compounds, the IC50 values of
compounds 14a-14f were 0.61~2.9 µM. When the R1 position was replaced with a
nitrovinyl group, the inhibitory activities of the title compounds (14g-14i) were
decreased slightly, with IC50 values of 2.7~8.2 µM. In comparison, compounds with
nitrovinyl substituents at the R3 position (14j-14m) had IC50 values of 1.6-5.4 µM. To
determine whether these compounds could also bind to the C128 site, we exploited
compound 14b as a probe molecule to explore the binding mode. As shown in Table
S3, only the C128S mutations could remarkably affect the IC50 value of 14b, causing
an 820-fold increase compared to the wild type. These consequences demonstrated
that C128 is extremely essential for the binding of 14b and FBPase, and a covalent
reaction likely occurs between 14b and C128, similar to that observed for compound
11.

Fig. 4. Mass spectrometry of covalent binding of FBPase (A) and compounds,
(B)FBPase+14b, (C) FBPase+14c, (D)FBPase+14i, (E) FBPase+14n
To further determine the number of cysteines in FBPase that may react with
these compounds, MS analysis of FBPase+compound was also performed (Fig. 4).
Fig. 4B shows that the main peak (38,009 Da) belongs to FBPase, and the peak at
38,306 Da belongs to FBPase+14b. These results suggest that 14b might covalently
bind with one cysteine on FBPase. Combining previous mutation experiments (Table
S3) and this result (Fig. 4B), one could infer that 14b predominately bind with C128
in FBPase. Similar to 14b, 14i could also specifically bind with C128 on FBPase as
illustrated in Fig. 4D. Additionally, the mass spectra of FBPase and 14c were also
determined in Fig. 4C. Fig.4C clearly shows that, although the covalent interactions
FBPase+2˟14c and FBPase+3˟14c occurred (based on the peaks at 38,637 Da and
38,951 Da), the predominant component was FBPase+14c (38,323 Da). Similar to 14c,

14n could also bind to FBPase with three forms as illustrated in Fig. 4E
Fig. 5. The fraction adduction ratio of compounds (12, 13, 14b, 14c, 14i and 14n)
and FBPase complex (FBPase +1×comp, FBPase +2×comp, FBPase +3×comp).
In conclusion, the proportion (Fig. 5) of title compounds (14a, 14c, 14i or 14n)
that bound to the C128 site on FBPase increased compared with that of compounds
12/13. These positive results indicate that anti-dyslipidemic fragments are
predominantly accommodated in the C128 pocket of FBPase. The optimization
strategy adopted in this study could remarkably improve the pocket selectivity of
FBPase inhibitors.
2.5 Binding model of compound 14b and FBPase.
As shown in Fig. S10A and Fig. S10B, the possible binding model for FBPase
and 14b act as Michael acceptor (C7 of 14b was attacked) was predicted using
Autodock 4[56]. In Fig. S10A, the -NH2 of R243 can form two hydrogen bonds with
-NO2 and –C=O of 14b. The NH2 moiety of R254 can form hydrogen bonds with
-NO2 of 14b. Two hydrogen-bonds between S124 and 14b could also be observed in
Fig. S10A. In addition, another possible binding model for FBPase and 14b (N9 of

14b was attacked) was also predicted in Fig. S10C and Fig. S10D. As illustrated in
Fig. S10C, the -OH group of S124 can form hydrogen bond with ether oxygen group
of 14b. A π-π staking interaction between the R243 and 14b could be observed in Fig.
S10C. The NH2 moiety of R254 likely forms halogen bond with F of 14b. To verify
these binding modes, mutations of the critical residues mentioned-above were
performed (Table S3). The IC50 values of R243A (17 µM) and R254A (12 µM)
increased 28- and 20-fold compared with that of WT FBPase (0.61 µM). In addition,
S124A result in 39-fold IC50 increase of compounds 14b compared with that of WT
(0.61 µM). These mutant results demonstrate the importance of residues S124, R254
and R243 for both binding modes of 14b and FBPase. In addition, a weakly π-π
interaction (3.6 Å) between benzene (Y258) and benzene (14b) looks likely to be
existed. However, the π-π stacking distance of Y258 and 14b is longer (3.5Å), this π-π
interaction could be negligible (Fig S10A). Thus, the IC50 of Y258A (2.9 µM)
increased only 4.8-fold compared to that of WT (0.36 µM). As discussed for
compounds 12/13, N125 is the residue (in helix 4) closest to C128 (Fig. S5, 3.3 Å),
the addition of 14b onto C128 will also cause the remarkable movement of N125
away from C128, which in turn significantly decrease the FBPase activity. However,
when N125 was mutated to Ala, N125-S124-S123 will move close to C128, which in
turn result in the increase of FBPase activity. As expected, N125A lead to the
significant IC50 increase (492-fold, Table S3) of 14b compared with WT (0.36 µM).
These results further indicate that, N125-S124-S123 pathway is very important for the
14b inhibition against FBPase, N125 is a key residue that affect the movement of

N125-S124-S123.
2.6 Inhibition of glucose output in primary rat hepatocytes.
To the best of our knowledge, FBPase is a rate-limiting enzyme in the GNG
pathway, which converts carbon substrates to glucose, and is a main factor of
endogenous glucose production[14, 73]. Furthermore, primary rat hepatocytes were
used to explore the effect of title compounds on glucose production under
physiological conditions.
Fig. 6. Effects of Clofibrate and synthesized compounds on glucose production in
primary rat hepatocytes in physical condition. 14b(A), 14c(B), 14i(C), 14n(D). *P <
0.05, **P < 0.01, ***P < 0.001 vs DMSO. Data shown are means ± SEM of four
independent experiments. Metformin (Met)
As shown in Fig. 6A, B, C, D, representative compounds 14b, 14c, 14i and 14n
resulted in dose-dependent inhibition of glucose production in primary rat hepatocytes.

Both 14c (Fig. 6B) and 14i (Fig. 6C) begin to inhibit glucose production at 10 µM,
with a significant difference of P < 0.05 and P<0.01 compared to the DMSO group.
Maximal inhibition of glucose production (≈ 90%) was achieved with 400 µM title
compounds, and a significant difference (P < 0.001) was observed compared to the
DMSO group. These results suggest that title compounds actually affect glucose
production of the GNG pathway and that -Cl at the R3 position is beneficial for
pharmaceutical effects under physiological conditions.
In fact, the FBPase or GNG pathway is overexpressed in T2D patients[1, 2], thus,
the effects of title compounds on the GNG pathway under pathological conditions
were further tested (Fig. 7). It has been reported[74] that cyclic AMP (cAMP) and
dexamethasone (Dex) can enhance hepatic GNG levels, therefore, this strategy was
used as model for evaluation of the pharmaceutical effects. Primary rat hepatocytes
were treated with 1 µM dexamethasone and 200 µM cAMP for 3 h [74, 75]. As shown
in Fig. 7 vehicle, glucose production via GNG (vehicle) dramatically increased
5.8-fold compared to the DMSO group after these treatments. Consistent with the
results obtained under physical conditions, the glucose lowering that occurs under
pathological conditions could be observed with 1 µM 14c (Fig. 7B) and 14i (Fig. 7C),
with a significant difference (P < 0.001) vs vehicle. In addition, title compounds (14b,
14c, 14i, 14n) also caused dose-dependent (10~100 µM) inhibition of glucose
production under pathological conditions (P <0.001 vs vehicle). Maximal glucose
lowering under pathological conditions (≈ 95%) was achieved at 100 µM 14b. In
conclusion, these title compounds also exhibit potential hypoglycemic effects under

pathological conditions, even more so than under physiological conditions.
Fig. 7. Effects of title compounds (14b, 14c, 14i and 14n) on glucose production in
primary rat hepatocytes in pathology condition causing by 1 µM dexamethasone (Dex)
and 200 µM cyclic AMP (cAMP). 14b (A), 14c (B), 14i (C), 14n (D). ***P < 0.001
versus DMSO. †P < 0.05, †† P < 0.01, ††† P < 0.001 vs vehicle. Data shown are
means ± SEM of four independent experiments.
2.7 Reducing lipid accumulation in primary rat hepatocytes.
It has been reported that high glucose and insulin levels can induce
hyperglycemia and insulin resistance[76]. In particular, the title compounds were
derived by combining the nitrostyrene fragment and anti-dyslipidemic functional
fragments, thus, it is possible for title compounds to exhibit anti-dyslipidemic effects.
Thus, this method was also used out to establish a model of lipid accumulation to

evaluate the pharmaceutical effect[77]. High concentrations of glucose (25 mM) and
insulin (500 nM) were used to treat primary rat hepatocytes for 24 h resulting in lipid
accumulation.
Fig. 8. Effects of title compound on lipid accumulation of TG (A, B) and TC (C, D) in
primary rat hepatocytes by 25 mM high glucose and 500 nM insulin, Clo (Clofibrate).
***P < 0.001 vs DMSO. †† P < 0.01, ††† P < 0.001 vs vehicle. Data shown are
means ± SEM of four independent experiments.
As shown in Fig. 8, the TG and TC quality in the vehicle group increased sharply
compared with that in the DMSO group, suggesting that this treatment actually led to
lipid accumulation. Both 14c and 14i begin to show a lipid–lowering (TG and TC)
effects at 25 µM, while the minimum lipid-lowering concentration of 14b and 14n is
10 µM (Fig. 8A, C). 14c and 14i exhibit stronger lipid-lowering effects than 14b and

14n and, also, potentially, Clo (50 µM), as illustrated in Fig. 8A, C. Upon increasing
the concentration of these compounds to 100 µM, inhibitions of TG production
increased slightly (Fig. 8B), but TC production decreased markedly (Fig. 8D).
Therefore, these title compounds can dose-dependently inhibit TC production, and
exhibits anti-dyslipidemic activity.
2.8 Evaluation of cytotoxicity and acute toxicity.
To the best of our knowledge, a protein assay is the general experiment for the
measurement of toxic substances cytotoxicity[78]. For the control of these compounds
cytotoxicity, the glucose, triglyceride and total cholesterol content were normalized to
the cellular protein concentration (cell viability) of primary rat hepatocytes (Fig. S11,
Fig. S12, Fig. S13) as described in previous documents.[78, 79] The protein
concentration of each group has been determined by using Bicinchoninic Acid (BCA)
method.[79] As shown in Fig. S11, Fig. S12 and Fig. S13, most of title compounds
exhibit no cytotoxicity except for 14n in pathology condition (Fig. S14D).
In addition, Sulforhodamine B (SRB) cytotoxicity assay of primary rat
hepatocytes in physical condition was further performed[80]. As shown in Fig. 9, cell
viability exceeded 80% even if the concentration of compounds 14b, 14c, 14i or 14n
reached 300 µM. These results further suggest that title compounds pharmaceutical
effect were independent of their influences on the cell viability. Combining positive
results of the protein concentration (Fig. S11, S12, S13) and SRB assay (Fig. 9), we
could conclude that the hypoglycemic and hypolipidemic effects are predominantly
due to the inhibition of title compounds.

Fig. 9. The cell viability (Sulforhodamine B cytotoxicity assay, SRB) of primary rat
hepatocytes in physical condition after treated by title compounds. The cell viability
of 14b(A), 14c(B), 14i(C), 14n(D). Data shown are means ± SEM of four
independent experiments
Furthermore, acute toxicity of 14b in the mice was evaluated. Sixteen
8-week-old male C57BL/6 mice were randomly divided into two groups to receive 0
(vehicle), 600 mg/kg of 14b oral on first day. All treated animals represented no
mortality, significant weight loss or allergic action and were as well health as the
control animals for 14 days, indicating that 14b has no acute toxicity at a dose of 600
mg/kg of oral (Fig. S14). These results also indicate to some extent that compounds
14b, 14c, 14i and 14n exhibit better target selectivity in primary hepatocyte.
3. Conclusion

We first discovered a novel cryptic cysteine allosteric site (C128) in FBPase,
several nitrostyrene derivatives exhibiting potent FBPase inhibition (IC50=0.32 µM).
The N125-S124-S123 pathway was most likely involved in the allosteric regulation
between C128 and active site of FBPase. However, the MS analysis revealed that
these nitrostyrenes might bind with more than one cysteine besides C128 in FBPase.
To improve pocket selectivity, a series of novel compounds were redesigned
rationally by attaching a drug moiety with anti-dyslipidemic function to a nitrostyrene
by integrating the FBCVS protocol and machine-learning-based synthetic complexity
evaluation. As expected, the MS results validated the proportion of title compounds
(14b, 14c, 14i or 14n) that bind to C128 site in FBPase was higher than that of 12/13.
In addition, these title compounds not only exhibited dose-dependent (10~100 µM)
inhibition of glucose production under pathological conditions (P <0.001 vs vehicle)
and pathological conditions but also might dose-dependently reduce the production of
TG and TC. These positive results provide further evidence that molecular targeting
of the C128 site of FBPase can have anti-T2D effects. This strategy could provide an
alternative general approach for improvement of the selectivity and
dual-functionalization of covalent drugs in the future.
4. Experimental section
4.1 Chemicals.
The following chemicals were commercially obtained and used without further
purification: compounds 1~9, 99% pure, Sigma-Aldrich; compound 10 98% pure,
Cayman; Nitrostyrene compounds 11~13, and compounds 14a-14n were synthesized

as described in Scheme S1, S2 and S3, and characterized as described in the
Supplementary Methods, and the characteristics and purity matched those of
published standards.
4.2 FBPase inhibition assays and mutagenesis.
Positive transformants containing FBPase genes were transformed into E.coli
BL21 (DE3) for protein expression. Cloning, expression, and purification were
performed as described in our previous work[81]. The final purified proteins were
validated by SDS-PAGE and preserved in 50% (v/v) glycerol at -20 .
FBPase activity was measured by the method of malachite green[82]. 30 µL
reaction solution contained 300 nM Hu-FBPase, 50 mM Tris-HCl (pH 7.4), 0.8 mM
Mg2+, 0.4 mM FBP and 0.5 µL DMSO. All compounds were dissolved by DMSO.
The reaction time is 5 min, and the reaction temperature is 37 . Then, the reaction
was stopped by 15 µL 1M perchloric acid. The released phosphate was further to form
a colored complex that was detected spectrophotometrically at 620 nM. Thus, the
reaction solution was incubated with 450 µL malachite green solution (0.035%
malachite green, 0.35% polyvinyl alcohol) at 25
for 30 min and the enzymatic
activity was monitored at 620 nm. The half maximal inhibitory concentration (IC50)
values or kinetic constants were identified using logistic equations (nonlinear analysis)
or Hill kinetic equations in the Origin 7.5 program package. All of the FBPase
mutants were also expressed, purified and analyzed.
4.3 Fragment-based covalent virtual screen