Pyrrolo[1,2-a]quinoxalines: Insulin Mimetics that Exhibit Potent and Selective Inhibition against Protein Tyrosine Phosphatase 1B

Article abstract of DOI:10.1002/cmdc.202000446

PTP1B dephosphorylates insulin receptor and substrates to modulate glucose metabolism. This enzyme is a validated therapeutic target for type 2 diabetes, but no current drug candidates have completed clinical trials. Pyrrolo[1,2-a]quinoxalines substituted at positions C1–C4 and/or C7–C8 were found to be nontoxic to cells and good inhibitors in the low- to sub-micromolar range, with the 4-benzyl derivative being the most potent inhibitor (0.24 μm). Some analogues bearing chlorine atoms at C7 and/or C8 kept potency and showed good selectivity compared to TCPTP (selectivity index '40). The most potent inhibitors behaved as insulin mimetics by increasing glucose uptake. The 4-benzyl derivative inhibited insulin receptor substrate 1 and AKT phosphorylation. Molecular docking and molecular dynamics simulations supported a putative binding mode for these compounds to the allosteric α3/α6/α7 pocket, but inconsistent results in enzyme inhibition kinetics were obtained due to the high tendency of these inhibitors to form stable aggregates. Computational calculations supported the druggability of inhibitors.

Full text of DOI:10.1002/cmdc.202000446

Full Papers
doi.org/10.1002/cmdc.202000446
ChemMedChem
1
2
3
4
5
6
7
8
9
Pyrrolo[1,2-a]quinoxalines: Insulin Mimetics that Exhibit
Potent and Selective Inhibition against Protein Tyrosine
Phosphatase 1B
Javier García-Marín,[a, b, f] Mercedes Griera,[b, c] Patricia Sánchez-Alonso,[a] Bruno Di Geronimo,[d]
Francisco Mendicuti,[e, f] Manuel Rodríguez-Puyol,*[b, c] Ramón Alajarín,*[a, b, f]
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
Beatriz de Pascual-Teresa,[d] Juan J. Vaquero,*[a, b, f] and Diego Rodríguez-Puyol*[b, c]
PTP1B dephosphorylates insulin receptor and substrates to
modulate glucose metabolism. This enzyme is a validated
therapeutic target for type 2 diabetes, but no current drug
candidates have completed clinical trials. Pyrrolo[1,2-a]quinoxa-
lines substituted at positions C1–C4 and/or C7–C8 were found
to be nontoxic to cells and good inhibitors in the low- to sub-
micromolar range, with the 4-benzyl derivative being the most
potent inhibitor (0.24 μm). Some analogues bearing chlorine
atoms at C7 and/or C8 kept potency and showed good
selectivity compared to TCPTP (selectivity index >40). The most
potent inhibitors behaved as insulin mimetics by increasing
glucose uptake. The 4-benzyl derivative inhibited insulin
receptor substrate 1 and AKT phosphorylation. Molecular
docking and molecular dynamics simulations supported a
putative binding mode for these compounds to the allosteric
α3/α6/α7 pocket, but inconsistent results in enzyme inhibition
kinetics were obtained due to the high tendency of these
inhibitors to form stable aggregates. Computational calculations
supported the druggability of inhibitors.
Introduction
insulin sensitivity and glucose tolerance.[3–5] Furthermore,
PTP1B-targeting antisense nucleotides have been shown to
exhibit anti-diabetic effects.[6,7]
The search for PTP1B inhibition-based anti-diabetes drug
candidates is still an unresolved issue despite almost two
decades of both industrial and academic research. Although
many different inhibitors have been reported (Figure 1), PTP1B
has remained an elusive target for many years. Currently, there
Protein tyrosine phosphatase 1B (PTP1B) dephosphorylates
insulin receptor (IR) and substrates (IRS), thereby modulating
glucose metabolism.[1,2] In addition, it is a validated therapeutic
target for type 2 diabetes mellitus (T2DM) as PTP1B knockout
mice and tissue-specific deletion of this enzyme lead to better
[a] J. García-Marín, Dr. P. Sánchez-Alonso, Dr. R. Alajarín, Prof. J. J. Vaquero
Departamento de Química Orgánica y Química Inorgánica
Universidad de Alcalá
28805 Alcalá de Henares (Spain)
E-mail: ramon.alajarin@uah.es
juanjose.vaquero@uah.es
[b] J. García-Marín, M. Griera, Prof. M. Rodríguez-Puyol, Dr. R. Alajarín,
Prof. J. J. Vaquero, Dr. D. Rodríguez-Puyol
Instituto Ramón y Cajal de Investigación Sanitaria (IRYCIS)
Ctra. Colmenar Viejo, km. 9100
28034 Madrid (Spain)
E-mail: manuel.rodriguez@uah.es
diego.rodriguez@uah.es
[c] M. Griera, Prof. M. Rodríguez-Puyol, Dr. D. Rodríguez-Puyol
Departamento de Biología de Sistemas
Universidad de Alcalá
28805 Alcalá de Henares (Spain)
[d] B. Di Geronimo, Prof. B. de Pascual-Teresa
Departamento de Química y Bioquímica
Facultad de Farmacia, Universidad San Pablo CEU
28925 Alcorcón (Spain)
[e] Prof. F. Mendicuti
Departamento de Química Analítica, Química Física e Ingeniería Química
Universidad de Alcalá
28805 Alcalá de Henares (Spain)
[f] J. García-Marín, Prof. F. Mendicuti, Dr. R. Alajarín, Prof. J. J. Vaquero
Instituto de Investigación Química Andrés M. del Río
Facultad de Farmacia, Universidad de Alcalá
28805 Alcalá de Henares (Spain)
Figure 1. Representative examples of PTP1B inhibitors and their IC50 data:
pTyr mimetic (top left) and bidentate (top right) catalytic-site inhibitors,
allosteric inhibitors at the α3/α6/α7 site (bottom left) and at the disordered
C-terminal site (bottom right).
Supporting information for this article is available on the WWW under
https://doi.org/10.1002/cmdc.202000446
ChemMedChem 2020, 15, 1–15
1
© 2020 Wiley-VCH GmbH
��
These are not the final page numbers!

Full Papers
doi.org/10.1002/cmdc.202000446
ChemMedChem
is no drug-like PTP1B inhibitor in post-phase II clinical trials.[8–11]
Since the discovery of the first PTP1B inhibitor,[12] many potent
competitive phosphotyrosine (pTyr)-mimetic PTP1B inhibitors
have been reported.[13,14] However, these ligands normally show
low selectivity due to the highly conserved catalytic site in the
phosphatase superfamily. They also frequently show poor oral
bioavailability since they need to be negatively charged in
order to bind the highly positively charged catalytic pocket.
Although druggability of phosphatases has been a controversial
issue,[15,16] the allosteric site of PTP1B exhibits better druggability
than the catalytic pocket.[17]
Binding sites other than the catalytic pocket (site A) have
been reported. Thus, in addition to the active site and its
immediate surroundings,[18] there are several binding pockets
(sites B, C, D and E) with two of them reported as secondary
pTyr binding pockets.[19–21] Somewhat farther from the catalytic
pocket, there is an allosteric binding pocket surrounded by α3,
α6 and α7 helixes,[22] and a further allosteric site has been
reported in the disordered C-terminal non-catalytic segment.[23]
Compounds 1 were prepared using the classical method
involving POCl3-mediated cyclization of N-(2-acylaminophenyl)
pyrroles (Scheme 1).[25]
1
2
3
4
5
6
7
8
9
The pyrrole ring was constructed initially using a Clauson–
Kaas synthesis with commercially available 2-nitroanilines 2 to
give 1H-1-(2-nitrophenyl)pyrroles 3 (60–99%). Compound 3e
was obtained in the lowest yield due to formation of the
corresponding indole derivative (10%) as a by-product. Com-
pounds 3 were then reduced to anilines 4 (52–91%) by catalytic
hydrogen-transfer reaction using hydrazine and Pd on carbon
or by reduction with SnCl2. Hydrazine only worked well with 3a
(91%), whereas the reaction with 3b was incomplete and
yielded the intermediate azoxy compound, as detected by
HPLC-MS analysis of the reaction mixture. Reduction of the nitro
group in 3b–j was performed with SnCl2. Anilines 4 were
acetylated with Ac2O or an acyl chloride to give amides 5 (70–
93%) together with small amounts of imides 6 (5–15%) as by-
product. Cyclisation of 5 in refluxing POCl3 led to 1a–i (77–
91%). Heating 6 for a longer reaction time also yielded 1a–i
(74–82%). For 1j, treatment of 4a with methyl orthoformate
resulted in acylation and cyclisation in a single step. The
substitution pattern at the pyrrole ring was explored by
bromination of 1b and 1g with NBS, which yielded mixtures of
mono- and dibromo derivatives 1k–m and 1n,o, respectively.
Compound 1b gave a mixture of 1k (52%), 1l (39%) and 1m
(8%), whereas 1g provided a mixture of 1n (69%) and 1o
(12%). Structural determination of the bromo derivatives was
based on chemical shifts and coupling constants reported for
the pyrrolo[1,2-a]quinoxaline core,[26] as well as nuclear Over-
hauser effect (NOE) and HMBC experiments. Compounds 1k,
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
During
a biological screening of in-house compound
libraries, 4-methylpyrrolo[1,2-a]quinoxaline (1a) was found to
be a hit for PTP1B inhibition (Figure 2). As such, we started a
hit-to-lead program based on this scaffold that involved
synthesizing a series of pyrrolo[1,2-a]quinoxalines 1 bearing
substituents at the benzene (Domain 1) and pyrrole (Domain 2)
rings and at R4 (Domain 3), as depicted in Figure 2. We also
performed the in vitro inhibition of PTP1B, enzyme kinetics
assays, spectroscopic measurements for aggregate formation,
cell experiments for glucose uptake and PTP1B-substrates (IRS1,
AKT, STAT3) phosphorylation, as well as molecular docking and
molecular dynamics. In this article we report the identification
of a new class of PTP1B inhibitors based on this pyrrolo[1,2-a]
quinoxaline scaffold.
Results and Discussion
Chemistry
The synthesis of pyrrolo[1,2-a]quinoxalines has recently at-
tracted renewed interest due to the biological activities shown
by compounds based on this scaffold.[24] We followed a classical
approach by using commercially available ortho-nitroanilines 2
as building blocks to provide the substituents at Domain 1.
Carboxylic acid anhydrides or chlorides, or methyl orthofor-
mate, allowed diversification at Domain 3, whereas N-bromo-
succinimide was the bromine source at Domain 2 (Figure 2).
Scheme 1. Synthesis of pyrrolo[1,2-a]quinoxalines 1. a) 2,5-dimeth-
oxytetrahydrofuran, AcOH, reflux; b) SnCl2·2H2O, EtOH, reflux; or N2H4, 10%
Pd/C, EtOH (for 3a); c) Ac2O, AcOH, reflux; or RCOCl, Et3N, CH2Cl2 (for 1h and
1i); d) HC(OMe)3, cat. AcOH, reflux (for 1j); e) POCl3, reflux; f) NBS, DMF,
Figure 2. Hit compound 1a with domains to explore and synthetic design
for compounds 1.
°
À 10 C to room temp.
ChemMedChem 2020, 15, 1–15
www.chemmedchem.org
2
© 2020 Wiley-VCH GmbH
��
These are not the final page numbers!

Full Papers
doi.org/10.1002/cmdc.202000446
ChemMedChem
1m and 1n showed a NOE between H3 and CH3 at C4, whereas
1l and 1o showed a NOE between H1 and H9.
an IC50 in the sub-micromolar range (Table 1 and Figure S1 in
the Supporting Information).
1
2
3
4
5
6
7
8
9
T-cell protein tyrosine phosphatase (TCPTP) is the human
tyrosine phosphatase with the highest degree of homology (ca.
75%) with PTP1B, thus making this enzyme a good control to
assess the selectivity of inhibition. As such, the IC50 was
determined for human TCPTP (63.2 kDa; 1.6 nm) with our best
inhibitors (IC50/PTP1B <2 μm) using conditions strictly identical to
PTP1B/ TCPTP in vitro inhibitory activity and selectivity
Compounds 1 were screened using single substrate and
inhibitor concentrations (Figure 3), with the colorimetric mala-
chite green assay being used as method of choice. Assays were
carried out using human recombinant PTP1B (residues 1–322;
37.4 kDa; 2.5 ng), inhibitor (1 μm) and insulin receptor substrate
5 (IRS-5, 75 μm, km=85 μm), a phosphopeptide containing
residue sequence 1142–1153 and the pTyr1146 residue from
the insulin receptor (IR) β-subunit domain. The compounds
tested showed inhibitory activity of between 22 and 53%
inhibition at 1 μm.
Compounds 1h–j were prepared after the activity screening
was performed and were later assayed for inhibitor activity. The
inhibitor concentration at 50% inhibition (IC50) was assayed for
compounds 1 using human recombinant PTP1B (residues 1–
322; MW=37.4 kDa; 1.6 nm), para-nitrophenyl phosphate
(pNPP, km=0.38 mm) and five inhibitor concentrations be-
tween 0.04 and 25 μm.[27] The best inhibitors 1d,g–j exhibited
those used for PTP1B. The selectivity indices (IC50/TCPTP/ IC50/PTP1B
)
obtained ranged from 0.5- to 44-fold (Table 1). Most of the
inhibitors assayed where selective for PTP1B, especially 1d and
1g, which showed selectivity indexes (SI) higher than 41-fold.
Compound 1h was fourfold selective, whereas 1i showed no
selectivity. Moreover, 1i and 1l were slightly selective (1.39-
and 2.1-fold, respectively) for TCPTP.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
It has been suggested that a Phe280 residue in the PTP1B
allosteric pocket, which is missing in TCPTP, bearing a Cys278
residue instead, could be the origin of the selective inhibition of
PTP1B versus TCPTP.[27] Recently, the selectivity of the benzbro-
marone derivative allosteric inhibitor BB2 reported by Wies-
mann (an analogue lacking of thiazole ring in inhibitor depicted
in Figure 1, bottom left) has been investigated by computa-
tional MD and end-point free energy methods.[28] García-Marín
found that the allosteric site of TCPTP presented higher mobility
(and higher fluctuations of the ligand BB2 inside the pocket)
and overall flexibility than that of PTP1B. This study illustrates
the crucial role that Phe280 plays in the allosteric recognition of
PTP1B, and highlights the importance of hydrophobic inter-
actions, especially sandwich π stacking, for the design of
effective allosteric inhibitors of PTP1B, in terms of not only free
energy but also stability of the final ligand-protein complex.
This Phe280 residue can stablish π-stacking interactions
with inhibitors that do not occur in TCPTP. Then, compounds
having two aromatic rings separated by a chain have been
reported to adopt a U-shape conformation able to doubly stack
up and down Phe280 by π-π stacking interaction through a
sandwich-type form (see Figure 1 bottom left). In this regard,
compound 1h was designed to stablish an additional π-
stacking interaction with Phe280 as it was observed in
Figure 3. Inhibition screening for compounds 1 (1 μm) with insulin receptor
substrate IRS5 as PTP1B substrate.
Table 1. [IC50] and selectivity data for compounds 1.
IC50 �SD [μM]
Compd.
R1
R2
R3
R4
R7
R8
PTP1B
TCPTP
SI [a]
1a
1b
1c
1d
1e
1f
1g
1h
1i
1j
1k
1l
1m
1n
1o
H
H
H
H
H
H
H
H
H
H
Br
H
Br
Br
H
H
H
H
H
H
H
H
H
H
H
H
H
Br
H
H
H
H
H
H
H
H
H
H
H
H
H
Br
H
H
Br
Me
Me
Me
Me
Me
Me
Me
Bn
Ph
H
Me
Me
Me
Me
Me
H
H
Me
1.00�0.01
>10
3.95�1.81
n.d.
n.
n.d.
Me
Me
Cl
CF3
OMe
H
H
H
H
Me
Me
Me
H
H
Cl
H
>10
n.d.
>25
n.d.
n.d.
>43.86
n.d.
0.57�0.22
>10
H
Cl
H
H
H
Me
Me
Me
Cl
Cl
>10
n.d.
>25
0.96�0.21
0.45�0.40
1.79�1.01
n.s.
n.d.
0.60�0.11
0.24�0.01
0.62�0.08
0.88�0.43
n.s.
>41.67
4.00
0.72
2.03
n.s.
0.47
n.s.
4.20�0.98
1.99�1.48
n.s.
n.s.
1.82�0.94
4.69�0.43
7.95�3.02
n.d.
4.37
n.d.
H
[a] Selectivity index, SI=IC50/TCPTP / IC50/PTP1B; n.d.: not determined; n.s.: not soluble; Experiments were carried out in duplicate (N=3).
ChemMedChem 2020, 15, 1–15
www.chemmedchem.org
3
© 2020 Wiley-VCH GmbH
��
These are not the final page numbers!

Full Papers
doi.org/10.1002/cmdc.202000446
ChemMedChem
molecular docking experiments (see Computational Approach
section). Surprisingly, inhibitors 1d and 1g, which lack the
benzyl group at C4 and bear chlorine atoms at C7/C8, showed
the highest selectivity with selectivity indexes (SI) above 40,
which is tenfold the SI for 1h. Then, the origin of the selectivity
of inhibitors 1d and 1g cannot be explained by π-stacking
interactions only.
performed by substrate, then obtaining low correlations plots
for each compound in every model.[29,30] Unfortunately, results
obtained were inconclusive. Then, we considered the possibility
of aggregate formation by compounds 1a, 1g and 1h for
shedding light upon the problems encountered during the
experiments of enzyme inhibition kinetics. Thus, to predict if
these compounds are prone to aggregate, we used Aggregate
Advisor tool. This tool helps to distinguish between true and
artefact screening hits based on Tanimoto structural similarity
index (compared to known aggregators) and on lipophilicity
criteria (based on calculated logP).[31] Compound 1a (clogP 2.9)
has 73% similarity to 1-methyl-β-carboline, a known aggrega-
tor. Although this tool gives no similarities for 1g or 1h to
known aggregators, their clogP values are high (3.5 and 4.8,
respectively) and we could not rule out aggregate formation by
these compounds due to their flatness and high lipophilicity.
We studied the possibility of aggregate formation by
compounds 1a, 1g and 1h in buffered solutions of TrisÀ HCl
containing ~1% of DMSO by performing measurements of
steady-state and time-resolved fluorescence, as well as dynamic
light scattering (DLS; Figures S5–S8 and Table S1). The
1
2
3
4
5
6
7
8
9
Structure-activity relationships
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
To better explain the IC50 data (Table 1), we considered three
sets of inhibitors: potent (IC50 <2 μm), good (2<IC50 <10 μm)
and poor (IC50 >10 μm). Although the IC50 dataset is small, some
qualitative structure–activity relationships can be extracted.
Thus, only hydrogen or chlorine atoms at R7 are allowed in
domain 1 (Figure 2), with a chlorine atom favouring inhibition
(1d vs 1g). In contrast, bulkier substituents such as methyl,
trifluoromethyl and methoxy are detrimental to the activity (1a
vs 1c, 1e, and 1f). A chlorine atom at R8 improves activity
slightly (1a vs 1g) but has no effect if R7 is substituted by
chlorine (1d vs 1g). In domain 2, the introduction of a bromine
atom pursued to mimic the bromine atoms as in benzbromar-
one-derived allosteric inhibitors (Figure 1, bottom left). How-
fluorescence emission spectra (Figure 4, left) exhibited
a
monomer band centred at ~306 nm and a second structureless
band shifted to the red at 399, 403 and 415 nm for 1a, 1g and
1h, respectively, which was attributed to the presence of
aggregates presumably formed by π-π stacking interaction. The
largest aggregation (which does not mean the largest aggre-
gate size) was observed for 1a, followed by 1h and much less
for 1g in the normalized spectra (Figure 4, right). Experiments
were performed in the same aqueous medium used in our
enzyme kinetics assays. As a result, these compounds formed
very stable aggregates in buffered solutions of Tris·HCl contain-
ing ~1% of DMSO at micromolar concentrations by presumable
π-π stacking interactions. The presence of stable aggregates for
inhibitors 1a, 1g and 1h, can probably be extended to all
compounds 1 due to their flatness and lipophilicity, thus
providing a strong reason for the unsuccessful results obtained
ever,
a bromine atom diminishes the activity and the
detrimental effect is stronger at R3 (1g vs 1o) than at R1 (1g vs
1n). In domain 3, alkyl and aryl substituents at R4 are allowed
and their activity follows the hierarchical order Bn (1h) > Ph
(1i) > H (1j)�Me (1a).
For selectivity against TCPTP, and despite the small dataset
available, we found that a chlorine atom at R8 is essential (1a vs
1g), whereas another chlorine atom at R7 has poor or little
effect (1g vs 1d). A hydrogen atom or an alkyl group (Me, Bn)
are allowed at R4 but have little effect on selectivity (1a vs 1h
vs 1j) and should not be conjugated to the ring like a phenyl
group (1i). A bromine atom at R1 is detrimental to selectivity
(1g vs 1n) by a factor of tenfold. Interestingly, two compounds
(1i and 1l) showed a slight preference for TCPTP, and a small
inversion in selectivity is observed when a phenyl group
occupies R4 (1i). This reversal in selectivity also takes place for
1l, although in this case it is not clear if the reason is the
presence of a bromine atom at R3, methyl groups at R7 and R8,
or both.
Enzyme inhibition kinetics and aggregate formation
In order to disclose the inhibition mechanism that operates
with these inhibitors, we performed enzyme inhibition kinetics
assays with compounds 1a, 1g and 1h, as representative
examples of potent inhibitors with no substitution (1a) and a
chlorine atom (1g) at Domain 1, and a benzyl group able to π-
stacking at Domain 3, respectively. Lineweaver–Burk plot for 1a
showed an apparent mixed inhibition mechanism, although for
1g and 1h plots did not fit into classic inhibition mechanisms
(Figures S2–S4). Experimental data could not fit to model kinetic
equations, neither those considering the activation/inhibition
Figure 4. Emission spectra for 1a, 1g and 1h derivatives in buffered
solutions of Tris·HCl containing ~1% of DMSO at a 25 μm concentration
(left); Emission spectra normalized to the maxima of monomer bands located
°
around 306 nm (right). Measurements were performed at 25 C.
ChemMedChem 2020, 15, 1–15
www.chemmedchem.org
4
© 2020 Wiley-VCH GmbH
��
These are not the final page numbers!

Full Papers
doi.org/10.1002/cmdc.202000446
ChemMedChem
in experiments on enzyme inhibition kinetics. Thus, the IC50 and
selectivity data reported herein (see former section) must be
taken with caution as correspond to mixtures aggregate-
aggregator where is not clear what is the inhibitory specie, the
aggregator, the aggregate, an enzyme-aggregate colloid or a
combination of them.
Shoichet lab and other researches as well, have largely
contributed to the study of aggregates in medicinal chemistry.
This is a matter of concern to the pharmaceutical community
because compounds physicochemical properties in aqueous
media influence drug behaviour.[32,33] However, aggregate
formation is also an opportunity to selectivity and to their use
in drug-delivery.[34–36] In particular, protein stability has implica-
tions for the binding onto aggregates and this could have an
effect on selectivity.[35] Moreover, Janssen lab hypothesized by
first time that aggregates are absorbed in the intestine tract as
opposed to individual molecules.[37] In fact, approved drugs like,
for instance, moxalactam, gefitnib, pranlukast, crizotinib and
clofazamine, among many others, or the antioxidant (E)-
resveratrol, are known aggregators.[35,38,39]
Aggregates have been reported to show selective protein
binding and this has been related to protein stability (the lower
is the stability, the higher is the loading onto the
aggregate).[34,35] As we are reporting here (see Computational
approach section), no potential PAINS (pan assay interference
compounds) were found, meaning that these compounds do
not show potential protein promiscuity.
On the other hand, it has been reported that aggregates
can have different IC50 values against different proteins. This has
been attributed to differences in protein concentration in the
pertinent assays.[40,41] As spectroscopic measurements showed
that 1a, 1g and 1h are aggregators that co-exist with
aggregates, then it is not possible to attribute the activity or
the selectivity of these compounds to one kind or another,
even performing IC50 assays using strictly identical conditions,
as we made. Thus, the activity and selectivity of compounds 1
needs to be addressed since it is still unclear what factors, other
than π-stacking with Phe280 in PTP1B, are operating for these
highly homologous enzymes, like protein conformational re-
straints/freedom, binding onto aggregates or both. Thus, the
observed selectivity for compounds 1 needs a deeper study
that is far from the aim of this work. This study is currently
ongoing and results will be soon reported elsewhere.
molecules.[42] Two main binding modes, namely, active-site
inhibition and allosteric inhibition, are found for these struc-
tures.
1
2
3
4
5
6
7
8
9
The active site has been reported to be an extremely
conserved domain among PTPs, and inhibition with small
molecules has been shown to be nonselective.[22,43] Enhanced
ligand selectivity can be achieved by allosteric binding. To
assess the selectivity of the most active compounds 1 between
two possible binding sites, in other words the catalytic and
allosteric site, docking experiments were performed using the
FlexX 4.0 module implemented in the LeadIT suite (Table S2).[44]
To model the interaction at the catalytic binding site, the crystal
structure with PDB ID: 2NTA was chosen. FlexX parameters were
set as default and protonation states were investigated at
pH 7.4�2 (Table S2). All compounds showed better docking
scores at the allosteric site rather than at the catalytic site.
Moreover, as implausible binding modes in the active site of
PTP1B were found, extensive molecular modelling studies were
performed at the allosteric site. In order to evaluate the
possibility that the compounds synthesized herein bind PTP1B
at the allosteric site, docking calculations were carried out using
the Schrödinger suite (http://www.Schrodinger.com). PTP1B has
been crystalized with an allosteric inhibitor BB3 and this
structure has been deposited in the PDB with code 1T48
(Figure 5a).[44] This structure was used as the target protein for
docking purposes. However, since it has not been completely
solved (there are some residues missing in the α7 helix),
homology modelling techniques were necessary to obtain a
complete target structure. The modelled structure was obtained
with SWISS-MODEL web server and 1T48 as template.[46]
Compounds 1 were directly docked inside this generated PTP1B
model, with the results showing that the tricyclic system
present in our ligands occupies the site of the dibromohydrox-
ibenzoyl moiety from BB3, filling the hydrophobic pocket
formed by helixes α3, α6, and α7. All compounds establish π–π
interactions with Phe196 (α3 helix) and Phe280 (α6 helix), as
well as a parallel displaced interaction with Trp291 (α7 helix)
(Figure 5b). Moreover, van der Waals interactions are estab-
lished with the side chains of Leu192 and Ile281. Atom N5 from
the tricyclic core hydrogen bonds to the side chain of Asn193 in
all cases. Interestingly, halogen-containing compounds such as
1d, 1g, 1m, 1n and 1o reproduce the halogen positions of the
reference compound (BB3), although in 1m, the molecule flips
to place bromine atoms at the pyrrole ring instead of chlorine
atoms at the benzene ring (see Figure S10). To test the stability
of the proposed binding modes and rationalize the biological
results described above, 20 ns molecular dynamics (MD)
simulations were carried out for all complexes. Analysis of the
MD trajectories highlights the stability of the docked com-
plexes, which is in complete agreement with the experimental
results, further supporting the putative allosteric site binding of
this series of compounds.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
Although the inhibition model could not be experimentally
demonstrated for these compounds, we performed a computa-
tional study for the binding of the aggregator form of these
inhibitors to the two major and best reported binding pockets
for PTP1B inhibitors: the active site and the allosteric pocket at
the α3/ α6/ α7 tunnel.
Computational approach
The ligand stability within the binding pocket can be
assessed from the low root mean square deviation (RMSD)
values for the ligand atoms (Figure 6 and Table 2), which range
from 0.81 to 4.49 Å. These values are, in general, in agreement
with a well-stabilized, low-energy predicted binding mode. The
PTP1B has been intensively studied as a druggable target.
Indeed, at the time of drafting this manuscript, a total of 144
crystal structures had been deposited in the Protein Data Bank
(PDB), with 86 of these involving complexation to small
ChemMedChem 2020, 15, 1–15
www.chemmedchem.org
5
© 2020 Wiley-VCH GmbH
��
These are not the final page numbers!

Full Papers
doi.org/10.1002/cmdc.202000446
ChemMedChem
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
Figure 5. a) PyMOL representation of crystal structure 1T48. The protein is coloured orange, and BB3 is shown as yellow sticks. b) Superimposition of docked
1a (pale blue) inside the PTP1B model with BB3 (yellow). The main interacting residues are represented as sticks.
Table 2. Predicted total binding energies (docking and MD simulations) and mean RMSD values for selected protein substructures and ligands along the
MD simulations.
Predicted binding energy
Mean RMSD values along 20 ns MD
Compd.
Docking
score
MM-
ISMSA
[kcal/mol]
Alpha 3
helix
[Å]
Alpha 6
helix
[Å]
Alpha 7
helix
[Å]
WPD loop
[Å]
Ligand
[Å]
[kcal/mol]
1a
1b
1c
1d
1e
1f
1g
1h
1i
1j
1k
1l
1m
1n
1o
À 6,40
À 6.52
À 6.51
À 6.55
À 6.55
À 5.88
À 6.55
À 9.02
À 7.43
À 6.16
À 6.94
À 7.26
À 6.58
À 6.95
À 6.98
À 27.19
À 31.01
À 29.70
À 32.60
À 29.46
À 23.86
À 29.41
À 33.31
À 34.51
À 24.60
À 30.62
À 33.88
À 36.15
À 29.55
À 34.10
–
0.397
0.468
0.423
0.411
0.444
0.530
0.453
0.421
0.402
0.483
0.468
0.367
0.507
0.413
0.695
0.442
0.3872
0.289
0.694
0.519
0.761
0.575
0.624
0.828
0.578
0.483
0.586
0.720
0.562
0.547
0.573
0.995
0.794
0.862
1.121
1.430
0.967
0.923
0.545
3.755
1.217
0.553
0.733
0.927
0.509
1.503
1.023
0.490
1.498
0.652
0.613
0.635
0.597
0.733
0.683
0.624
0.597
0.562
0.589
0.624
0.473
0.545
0.575
0.537
0.528
0.541
0.726
1.397
2.393
2.234
2.039
0.853
1.161
2.157
1.081
2.292
1.201
1.304
3.348
1.542
4.496
2.668
0.815
–
1T48 model
2HNP model
–
–
ligands remain within the allosteric binding pocket throughout
the entire simulation time and are mainly maintained by stable
π-stacking interactions with Phe196 (α3 helix) and Phe280 (α6
helix) and transient van der Waals and hydrophobic interactions
with Trp291, Leu192 and Ile281. However, a clear correlation
between complex stability and IC50 cannot be obtained due to
the minor structural differences in this series. It is interesting to
note that the higher RMSD values correspond to compounds
1k and 1m, for which IC50 values were not determined due to
their lack of solubility. In addition, to further support the
predicted results, the dynamic behaviour of the α3, α6 and α7
helixes and the WPD-loop of all complexes was compared to
that of the protein without the bound ligand. To that end, two
apo proteins were submitted to the same MD protocol
mentioned above. These proteins were structure 1T48, after
removal of the crystalized ligand BB3 and modelling of the
missing amino acids, and structure 2HNP, which corresponds to
the crystal structure of the protein with no bound ligand and
after modelling of the missing α7 helix. In the case of the WPD
loop, this analysis shows an increased stability for most of the
complexes when compared to the apo proteins (Table 2), with
mean RMSD values along the simulation time for this
substructure ranging between 0.47 and 0.73 Å for all com-
plexes. This same analysis for the apo structures leads to RMSD
values of 0.72 and 1.39 Å for the 1T48 and 2HNP models,
respectively. These results may account for the restricted
mobility of this WPD loop upon ligand binding, in contrast to
the apo structures, where this loop appears to be more flexible.
This observation further supports the already reported restric-
tion to WPD loop movement upon ligand binding to PTP1B via
the same allosteric pocket.[42] On the other hand, the mean
RMSD values throughout the MD simulation for α6 and α3 in all
complexes show a similar dynamic behaviour as for the apo
proteins.
In contrast, the α7 helix increases its mobility with respect
to the modelled apo structures (1T48 and 2HNP). Thus, the
RMSD values range from 0.49 to 3.75 Å in the complexes,
whereas this same analysis gives a value of 0.6 Å for the
ChemMedChem 2020, 15, 1–15
www.chemmedchem.org
6
© 2020 Wiley-VCH GmbH
��
These are not the final page numbers!

Full Papers
doi.org/10.1002/cmdc.202000446
ChemMedChem
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
Figure 6. Evolution of the RMSD for all PTP1B-1 complexes along the 20 ns MD simulations.
modelled apo forms. These increased values in most complexes
suggest a movement of the α7 helix which, interestingly, is in
agreement with the low structured characteristics of the C-
terminal end in experimentally solved allosteric complexes
(e.g., PDB IDs: 1T49 and 1T4 J).[22]
depth as the minor structural differences between them are not
sufficient to significantly affect the binding mode to the target
protein. However, our calculations strongly support the con-
formational rearrangement of secondary structural elements
observed experimentally for this type of inhibitor.[22,45]
As expected for an allosteric binding mode, the WPD loop is
further stabilized upon ligand binding and α7 exhibits high
mobility, as observed experimentally for the allosteric inhibitors
in crystal structures such as 1T48, 1T49 and 1T4 J.[22,47] Taken
together, these results further support a putative allosteric
binding mode for this series of compounds.
In summary, our simulations provide information regarding
the putative binding mode for this type of ligand but do not
allow us to scrutinize this series of compounds in any greater
To better rationalize the in vivo results (see section below:
Analysis of the effects in cells), the physicochemical properties
and ADME parameters were calculated using the versatile web
server SWISS–ADME.[48] All compounds showed similar physico-
chemical properties and ADME parameters (Table 3), with no
marked differences between them. Interestingly, no potential
PAINS (pan assay interference compounds) were found and the
iLOGP values obtained (close to 3) correspond to improved
glucose uptake (1h, 1l and 1o).[49,50] Ligand efficiency indices
ChemMedChem 2020, 15, 1–15
www.chemmedchem.org
7
© 2020 Wiley-VCH GmbH
��
These are not the final page numbers!

Full Papers
doi.org/10.1002/cmdc.202000446
ChemMedChem
Table 3. Calculated properties: molecular weight (MW), topological polar
surface area (TPSA), ilogP (inhouse logP), pIC50 values (calculated from
experimental IC50 values), binding efficiency index (BEI) and surface binding
efficiency index (SEI).
1
2
3
TPSA [Å2]
ilogP
pIC50
BEI
SEI
4
Compd.
MW [g/mol]
5
6
7
8
1a
1b
1c
1d
1e
1f
1g
1h
1i
1j
1k
1l
1m
1n
1o
182.22
210.27
196.25
251.11
250.22
212.25
216.67
258.32
244.29
168.19
289.17
289.17
289.17
295.56
295.56
0.17
0.17
0.17
0.17
0.17
0.26
0.17
0.17
0.17
0.17
0.17
0.17
0.17
0.17
0.17
2.39
2.84
2.63
2.82
2.67
2.64
2.67
2.97
3.02
2.11
3.09
3.14
3.39
2.91
2.96
6
5
5
6.24
5
32.93
23.78
25.48
24.85
19.98
23.56
28.71
25.59
25.38
35.97
–
34.68
28.90
28.90
36.07
28.90
18.85
35.95
38.21
35.84
34.97
–
9
5
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
6.22
6.61
6.2
6.05
–
5.37
–
18.57
–
19.39
18.00
31.04
–
33.12
30.75
5.73
5.32
were also calculated using the binding efficiency index (BEI)
and surface binding efficiency index (SEI) based on the
molecular weight (MW) and topological polar surface area
(TPSA).[51] As reported in the literature, the BEI and SEI values
should be around 27 and 18, respectively, for a compound to
be considered a good druggable lead. In general, the values for
our compounds are close to those reported previously,
although the SEI values are farther from the canonical ones.
However, according to the literature, the suitability of a series
of compounds to be considered as druggable leads should be
assessed from either BEI or SEI. Therefore, in this case, we can
further confirm that the reported series constitute a good
starting point for the development of allosteric PTP1B inhib-
itors. Taken together, our results point to future strategies for
lead optimization of this series by appropriate growth of the
lead structure via the pyrrole end in order to sandwich Phe280,
as demonstrated for BB3.
Figure 7. Viability of HepG2 (top) and C2C12 (bottom) cell lines, as
determined by using the MTT assay after incubation for 24 h with
compounds 1 at 50 μm. Values are expressed as percent of CT and as the
mean�SEM of three independent experiments in duplicate.
increased glucose uptake by cells, with this effect being more
relevant for compounds 1g, 1h, 1l and 1o. Compounds 1h, 1l
and 1o reached the level produced by insulin. Compound 1g
also increased glucose uptake, but to a slightly lesser degree
than former compounds. Compound 1d increased the level of
glucose uptake, although below the level reached by 1g.
An inspection of data provided by cell glucose uptake assay
(expressed as percent of control) and ilogP data of tested
compounds (see former section) allowed us to find a qualitative
relationship between both. Then, the hierarchical order found
for glucose uptake test was: 1h (205%) �1l (207%) �1o
(209%) >1g (184%) >1d (118%) >1a (73%), and for ilogP: 1h
(2.97) �1l (3.14) �1o (2.96) >1d (2.82) >1g (2.67) >1a (2.39).
Then, the higher is the lipophilicity, the biggest is the uptake of
glucose by C2C12 cells, with the most potent inhibitor 1h
(0.24 μm) among those producing the highest effect.
Additionally, 1h induced and increased phosphorylation of
the insulin receptor substrate 1 (IRS1; panel B) and AKT (panel
C), as was also the case for insulin. Combined treatment with
1h and insulin induced an additive effect on IRS1 (panel B) and
AKT (panel C) phosphorylation. This same compound did not
modify STAT3 phosphorylation (panel D). At peripheral nervous
system, the JAK/STAT pathway induces glucose uptake in
muscle via STAT3 phosphorylation by JAK, STAT3 translocation
and further gene transcription. Then, our results suggest that
Screening of cell viability
The determination of IC50 against PTP1B and TCPTP allowed us
to identify two potent and very selective inhibitors (1d and 1g)
and three good and slightly selective inhibitors (1a, 1h and
1n). After these promising results, the cell viability was
screened. Compounds 1 were tested using the MTT assay on
HepG2 and C2C12 (for the most potent inhibitors tested on cell
glucose metabolism) cell lines at 50 μM. Both cell lines showed
high viability (>90%) towards these compounds (Figure 7).
Analysis of the effects in cells
We tested the effect of some inhibitors in C2C12 cells to
validate the potential therapeutic use of these inhibitors. These
cells are derived from mouse muscle myoblasts and constitute
a good target for assessing the effects of insulin. As shown in
Figure 8 (panel A), all compounds tested, except 1a, induced an
ChemMedChem 2020, 15, 1–15
www.chemmedchem.org
8
© 2020 Wiley-VCH GmbH
��
These are not the final page numbers!

Full Papers
doi.org/10.1002/cmdc.202000446
ChemMedChem
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
Figure 8. Biological effects on C2C12 cells. A) Effect of inhibitors 1 on glucose uptake. Effect of 1h on the phosphorylation of B) insulin receptor substrate 1
(IRS1); C) AKT and D) STAT3. Keys: CT (control conditions, 1% DMSO); Ins (with 10 nm insulin final concentration, 30 min); 1h (with 2 μm inhibitor final
concentration, 90 min); FBS (with 10% foetal bovine serum, 30 min). For B, C and D, representative western blots for both the phosphorylated and total
proteins are shown in the upper part of the panels, and the ratios of the densitometric analysis for both signals are given in the lower part. Values are
expressed as percent of CT and as the mean � SEM of three independent experiments (in duplicate for glucose uptake). *p<0.05 vs. CT, **p<0.05 vs. the
other groups (CT and Ins for glucose uptake).
the increase of glucose uptake provided by inhibition of PTP1B
by 1h in C2C12 cells is not due to changes in the phosphory-
lated status of STAT3, but rather due to the inhibition of pAKT
dephosphorylation, then promoting the insulin–induced trans-
location of glucose transporter 4 (GLUT4) to the plasma
membrane.
hydrogen or chlorine atoms at C–7 and C–8, hydrogen atoms in
the pyrrole ring and benzyl or phenyl groups at C–4 to establish
π interactions with Phe196 and Phe280. Molecular docking and
molecular dynamics studies using the PTP1B homology model
support binding to the α3/α6/α7 pocket, thereby stabilizing the
WPD loop in an open, non–catalytic conformation. The
inhibition mode could not be revealed since enzyme inhibition
kinetics assays were unsuccessful due to aggregate formation,
as it has been proven by fluorescence and DLS methods applied
to 1a, 1g and 1h. The inhibitors tested showed enhanced
glucose uptake comparable to insulin when their ilogP value is
close to 3, as for 1h, 1l and 1o, and showed no effect when
ilogP is, at least, below 2.4. Compound 1h, which is one of best
inhibitors, also inhibits pIRS and pAKT dephosphorylation in
C2C12 cells, and enhances the insulin–dependent phosphoryla-
tion of these proteins, while STAT3 phosphorylation was not
modified by 1h. ADME properties and druggability indices were
also calculated, thereby further supporting the suitability of the
compounds as leads for inhibition of PTP1B by a putative
allosteric mechanism.
Conclusion
We have reported the discovery of pyrrolo[1,2–a]quinoxalines
as scaffolds that provide potent and selective PTP1B inhibitors
with insulin mimetic properties. This scaffold putatively binds to
the allosteric hydrophobic pocket formed by α3/α6/α7 helixes,
as supported by molecular docking and molecular dynamics
simulations. A set of 15 compounds bearing substituents at
three domains, namely the benzene, pyrrole and pyrazine rings,
have been prepared. The inhibitory activity of these compounds
ranges from 0.24 to 4.69 μm, their selectivity indices against
TCPTP are in the range >44–fold to 0.47–fold and they are
nontoxic to HepG2 or C2C12 cells. The best inhibitors bear
ChemMedChem 2020, 15, 1–15
www.chemmedchem.org
9
© 2020 Wiley-VCH GmbH
��
These are not the final page numbers!

Full Papers
doi.org/10.1002/cmdc.202000446
ChemMedChem
°
(91%); yellow solid; mp: 137–139 C; IR (νmax, KBr): 3435, 3116, 2919,
Experimental Section
1
2
3
4
5
6
7
8
9
1625, 1523, 1490, 1412, 1351, 1313, 1244, 1086, 1024, 888, 853, 748,
717, 603 cmÀ 1
;
1H NMR (300 MHz, CDCl3): δ=7.77 (dd, 1H, J=
Chemistry
2.6 Hz, J=1.2 Hz, H–1), 7.62 (s, 1H, H–6/H–9), 7.51 (s, 1H, H–9/H–6),
6.78 (m, 2H, H–2, H–3), 2.67 (s, 3H, CH3), 2.38 (s, 3H, CH3), 2.34 (s, 3H,
CH3) ppm; 13C NMR (75 MHz, CDCl3): δ=152.3, 136.1, 134.0, 133.7,
129.3, 126.1, 125.1, 113.9, 113.5, 112.9, 105.7, 21.9, 20.1, 19.6 ppm;
HRMS [ESI–TOF] m/z calcd for C14H15N2 [M+H]+ 211.1230, found
[M+H]+ 211.1200.
Solvents (HPLC quality, Scharlau) were dried in a Solvent Purifica-
tion System (MBraun) by passing through a pre–activated alumina
column or were purchased as anhydrous quality. Reagents were
purchased from Merck/Sigma-Aldrich or Acros and were used as
received. Reactions were monitored by thin–layer chromatography
(TLC) on silica–coated aluminium sheets (Alugram silica gel 60
F254). Spots were visualized by UV light (254 nm). Column
chromatography was carried out with Merck silica gel (0.030–
0.075 mm) and solvents were used as received (Scharlau). Infrared
spectra (IR, NaCl windows or KBr pellets) were recorded on a
Perkin-Elmer FTIR 1725X instrument. Frequencies (ν) of the more
intense bands are given in cmÀ 1. Nuclear magnetic resonance
spectra (1H and 13C NMR) were recorded using a Varian Gemini 200
(200 and 50 MHz, respectively), Varian UNITY-300 (300 and 75 MHZ,
respectively), Varian Mercury-VX-300 MHz and Varian-UNITYPLUS-
500 (500 and 125 MHz, respectively) instruments. Chemical shifts
(δ) are given in ppm and are referenced to the residual signal of
the non-deuterated solvent. Coupling constants (J) are given in Hz.
All compounds were analysed by tandem HPLC-MS. HPLC was
performed on a reverse-phase C18 column (Luna, Phenomenex,
3 μm, 3×100 mm) using a gradient of MeOH/water/4% formic acid
at a flow rate of 1.0 mL min–1. Products were detected at λ
=254 nm. Compounds were�95% pure. Mass spectra were
recorded on a Hewlett–Packard 5988 A mass spectrometer. High–
resolution mass spectra were recorded on an Agilent 6210 LC/MS
TOF mass spectrometer.
4,7-Dimethylpyrrolo[1,2–a]quinoxaline (1c). General procedure A.
Compound 5c (0.65 g, 3.05 mmol) was heated for 50 min. Yield:
0.46 g (77%); yellow solid; mp: 271–273 C: IR (νmax, KBr): 3423, 3095,
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
°
2624, 1885, 1624, 1598, 1550, 1507, 1405, 1377, 1285, 1260, 1111,
1046, 823, 753, 679, 601, 530 cm–1; 1H NMR (300 MHz, CDCl3) δ 8.38
(s, 1H, H–6), 8.21 (dd, 1H, J=2.6 Hz, J=1.6 Hz, H–1), 7.83 (d, 1H, J=
8.2 Hz, H–9), 7.47 (m, 2H, H–3, H–8), 7.13 (dd, 1H, J=4.6 Hz, J=
2.6 Hz, H–2), 3.16 (s, 3H, CH3), 2.52 (s, 3H, CH3) ppm; 13C NMR
(75 MHz, CDCl3) δ 152.5, 136.7, 129.7, 125.6, 124.9, 124.4, 117.9,
115.6, 113.8, 113.8, 111.7, 21.1, 19.5 ppm; HRMS (ESI+) m/z calcd
for C13H13N2 [M+H]+ 197.1079, found [M+H]+ 197.1074.
7,8-Dichloro-4-methylpyrrolo[1,2–a]quinoxaline (1d). General proce-
dure A. Compound 5d (2.69 g, 10 mmol) was heated for 45 min.
Yield: 2.49 g (99%); brown solid; mp: 197–199 C; IR (νmax, KBr): 3448,
°
3093, 2925, 2362, 1718, 1606, 1478, 1409, 1298, 1216, 1130, 1048,
879, 852, 745, 639, 604, 545 cmÀ 1
;
1H NMR (500 MHz, CDCl3): δ=
7.93 (s, 1H, H–6/H–9), 7.81 (s, 1H, H–9/H–6), 7.75 (dd, 1H, J=2.7 Hz,
J=1.3 Hz, H–1), 6.90 (dd, 1H, J=4.0 Hz, J=1.3 Hz, H–3), 6.84 (dd,
1H, J=4.0 Hz, J=2.7 Hz, H–2), 2.68 (s, 3H, CH3) ppm; 13C NMR
(75 MHz, CDCl3) δ 154.9, 134.7, 130.5, 129.8, 128.7, 126.2, 125.8,
115.2, 115.1, 114.5, 108.1, 21.7 ppm; HRMS (ESI+) m/z calcd for
C12H9Cl2N2 [M+H]+ 251.0137, found [M+H]+ 251.0158.
The synthesis of compounds 3, 4, 5 and 6 is reported in SI. The
synthesis of compounds 1 is reported below.
7-Trifluoromethyl-4-methylpyrrolo[1,2–a]quinoxaline (1e). General
procedure A. Compound 5e (0.96 g, 3.59 mmol) was heated for
40 min. Yield: 0.88 g (98%); brown solid; mp: 128–130 C; IR (νmax
Synthesis of pyrrolo[1,2–a]quinoxalines (1)
°
,
KBr): 3104, 1628, 1530, 1504, 1459, 1420, 1334, 1299, 1208, 1164,
1142, 1108, 1089, 1033, 899, 828, 766, 737, 696, 653, 606, 526 cm–1;
1H NMR (300 MHz, CDCl3): δ=8.17 (s, 1H, H–6), 7.92 (d, 1H, J=
2.6 Hz, H–1), 7.88 (d, 1H, J=8.6 Hz, H–9), 7.67 (d, 1H, J=8.6 Hz, H–
8), 6.94 (d, 1H, J=3.6 Hz, H–3), 6.89 (t, 1H, J=3.2 Hz, H–2), 2.73 (s,
3H, CH3) ppm; 13C NMR (75 MHz, CDCl3): δ=155.0, 135.3, 129.1,
General procedure A: A mixture of acetanilide 5 or N–acetyl–
acetanilide 6 (3–13 mmol) and POCl3 (5 mL/mmol) was heated at
reflux temperature. Next, the reaction mixture was cooled to room
temperature and evaporated to dryness. Iced water (20 mL/mmol)
was added over the resulting residue and the pH was adjusted to
7–8 with 5% NaHCO3. The mixture was extracted with CH2Cl2 (3 x
30 mL/mmol) and the organic extracts were pooled and dried
(MgSO4). The desiccant was filtered off and the solvent was
removed under reduced pressure to give a residue that was
chromatographed on silica gel using CH2Cl2/acetone 9:1 as eluent.
General procedure B: Compound 4 (0.312 mmol) was dissolved in
trimethyl orthoformate (0.8 mL/mmol). Some drops of AcOH were
added and the solution was stirred at reflux for 1.5 h. Next, water
(10 mL/mmol) was added and the mixture was extracted with
CH2Cl2 (3 x 16 mL/mmol). Organic extracts were washed with brine
(16 mL/mmol), dried over Na2SO4, filtered and evaporated under
reduced pressure to afford a brown solid. The solid was dissolved in
MeOH, water was added dropwise and the precipitate was filtered
and dried.
2
3
1
127.0 (q, JCF =32.8 Hz), 126.5 (q, JCF =3.7 Hz), 126.1. 123.9 (q, JCF
=
3
271.7 Hz), 123.1 (q, JCF =3.7 Hz), 114, 114.4, 114.1, 107.5, 21.8 ppm;
HRMS (ESI+) m/z calcd for C13H9F3N2 [M+H]+ 251.0791, found [M
+H]+ 251.0752.
4-Methyl-7-methoxypyrrolo[1,2–a]quinoxaline (1f). General procedure
A. Compound 5f (1.61 g, 7 mmol) was heated for 35 min. Yield:
1.38 g (93%); brown solid; mp: 49–50 C; IR (νmax, KBr): 3085, 1616,
°
1593, 1524, 1491, 1350, 1300, 1259, 1246, 1198, 1161, 1048, 1030,
932, 877, 847, 770, 717, 621 cm–1; 1H NMR (300 MHz, CDCl3): δ=7.81
(dd, 1H, J=2.3 Hz, J=1.3 Hz, H–1), 7.71 (d, 1H, J=8.9 Hz, H–9), 7.37
(d, 1H, J=2.6 Hz, H–6), 7.06 (dd, 1H, J=8.9 Hz, J=2.6 Hz, H–8), 6.85
(dd, 1H, J=3.9 Hz, J =1.3 Hz, H–3), 6.79 (dd, 1H, J=3.9 Hz, J=
2.6 Hz, H–2), 3.88 (s, 3H, CH3), 2.71 (s, 3H, CH3O) ppm; 13C NMR
(75 MHz, CDCl3): δ=157.0, 153.9, 136.8, 125.9, 121.5, 115.9, 114.5,
113.9, 113.2, 110.5, 106.3, 55.6, 21.9 ppm; HRMS [ESI–TOF] m/z calcd
for C13H13N2O [M+H]+ 213.1028, found [M+H]+ 213.1023.
4-Methylpyrrolo[1,2–a]quinoxaline (1a). General procedure A. Com-
pound 5a (2.18 g, 10.9 mmol) was heated for 1.25 h. Yield: 1.77 g
(89%); brown solid; mp: 138–140 C; IR (νmax, KBr): 3439, 3099, 1611,
°
1529, 1481, 1416, 1380, 1361, 1323, 1258, 1212, 1042, 947, 859, 760,
8-Chloro-4-methylpyrrolo[1,2–a]quinoxaline (1g). General procedure
A. Compound 5g (2.52 g, 10.7 mmol) was heated for 40 min. Yield:
2.03 g (87%); yellow solid; mp: 143–144 C; IR (νmax, KBr): 3099, 1610,
1530, 1476, 1458, 1418, 1381, 1351, 1312, 1253, 1210, 1130, 1118,
1085, 1036, 862, 846, 814, 793, 740, 677, 614, 570, 514, 463 cm–1; 1H
NMR (300 MHz, CDCl3): δ=7.79 (m, 3H, H–1, H–6, H–9), 7.34 (dd, 1H,
J=8.6 Hz, J=2.3 Hz, H–7), 6.88 (dd, 1H, J=3.9 Hz, J=1.3 Hz, H–3),
732, 690, 650, 609, 534, 470 cmÀ 1
;
1H NMR (200 MHz, CDCl3): δ=
7.89 (m, 2H), 7.82 (dd, 1H, J=7.4 Hz, J=2.3 Hz), 7.43 (m, 2H), 6.88
(dd, 1H, J=4.2 Hz, J=1.3 Hz), 6.83 (t, 1H, J=3.2 Hz), 2.72 (s, 3H)
ppm.
°
4,7,8-Trimethylpyrrolo[1,2–a]quinoxaline (1b). General procedure A.
Compound 5b (2.89 g, 12.8 mmol) was heated for 1 h. Yield: 2.43 g
ChemMedChem 2020, 15, 1–15
www.chemmedchem.org
10
© 2020 Wiley-VCH GmbH
��
These are not the final page numbers!

Full Papers
doi.org/10.1002/cmdc.202000446
ChemMedChem
6.84 (q, 1H, J=2.8 Hz, H–2), 2.69 (s, 3H, CH3) ppm; 13C NMR (75 MHz,
CDCl3): δ=153.8, 134.4, 132.2, 130.3, 127.8, 126.1, 125.4, 114.4,
114.0, 113.8, 107.0, 21.9 ppm; HRMS (ESI+) m/z calcd for C12H9ClN2
[M+H]+ 217.0527, found [M+H]+ 217.0551.
177–178 C. IR (νmax, KBr): 3434, 3121, 2914, 1574, 1486, 1475,
1436, 1403, 1341, 1192, 1160, 1013, 881, 855, 802, 676, 564 cm–1;
1H NMR (300 MHz, CDCl3): δ=8.93 (s, 1H, H–9/H–6), 7.62 (s, 1H, H–
3), 6.84 (s, 1H, H–6/H–9), 2.93 (s, 3H, CH3), 2.39 (s, 3H, CH3), 2.34 (s,
3H, CH3) ppm; 13 C NMR (75 MHz, CDCl3): δ=151.6, 135.6, 135.1,
129.3, 126.5, 125.1, 121.8, 117.6, 115.8, 115.5 (2 C), 24.6, 20.5,
19.4 ppm; HRMS [ESI–TOF] m/z calcd for C14H13BrN2 [M+H]+
368.9420, found [M+H]+ 368.9429.
°
1
2
3
4
5
6
7
8
9
4-Benzylpyrrolo[1,2-a]quinoxaline (1h). General procedure A. Com-
pound 5h (36 mg, 0.13 mmol) was heated for 3 h. Yield: 29 mg
1
°
(90%); White solid; mp, 90–93 C; H NMR (500 MHz, CDCl3): δ=7.99
(d, J=7.8 Hz, 1H), 7.88 (dd, J=2.7, 1.2 Hz, 1H), 7.82 (dd, J=8.1,
1.4 Hz, 1H), 7.50 (dd, J=7.3, 1.5 Hz, 1H), 7.46–7.41 (m, 3H), 7.31–
7.25 (m, 2H), 7.23–7.17 (m, 1H), 6.86 (dd, J=3.9, 0.9 Hz, 1H), 6.80
(dt, J=8.7, 4.3 Hz, 1H), 4.38 (s, 2H) ppm; 13C NMR (125 MHz, CDCl3):
δ=155.38, 138.12, 129.85, 128.99, 128.65, 128.64, 127.49, 127.37,
126.72, 125.99, 125.27, 114.38, 113.83, 113.76, 107.30, 42.72 ppm;
HRMS (ESI+) m/z calcd for C18H14N2 [M+H]+ 258.1238, found [M+
H]+ 258.1242.
1-Bromo-8-chloro-4-methylpyrrolo[1,2-a]quinoxaline (1n) and 3-
bromo–8-chloro-4-methylpyrrolo[1,2-a]quinoxaline (1o). From 1g
(0.51 g, 2.37 mmol). Compound 1n: Yield, 0.47 g (69%); yellow solid;
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
°
mp: 193–194 C. IR (νmax, KBr): 3127, 2919, 1607, 1532, 1468, 1418,
1
1375, 1095, 1049, 849, 819, 763, 755, 678, 577, 460 cmÀ 1; H NMR
(500 MHz, CDCl3): δ=9.25 (d, 1H, J=2.1 Hz, H–9), 7.80 (d, 1H, J=
8.7 Hz, H–6), 7.40 (dd, 1H, J=8.7 Hz, J=2.1 Hz, H–7), 6.88 (d, 1H, J=
4.2 Hz, H–3), 6.85 (d, 1H, J=4.2 Hz, H–2), 2.66 (s, 3H, CH3) ppm; 13C
NMR (75 MHz, CDCl3) δ 152.9, 135.5, 131.1, 130.3, 128.9, 127.9,
125.9, 119.0, 115.2, 107.6, 99.4, 21.5 ppm; HRMS (ESI+) m/z calcd
for C12H9BrClN2 [M+H]+ 294.9632, found [M+H]+ 294.9648;
4-Phenylpyrrolo[1,2-a]quinoxaline (1i). General procedure A. Com-
pound 5i (78 mg, 0.33 mmol) was heated for 3 h. Yield: 63 mg
1
°
(70%); White solid; mp, 85–87 C; H NMR (300 MHz, CDCl3): δ=8.07
(d, J=7.6 Hz, 1H), 8.04–7.97 (m, 2H), 7.87 (d, J=7.7 Hz, 1H), 7.63–
7.38 (m, 5H), 7.01 (d, J=3.7 Hz, 1H), 6.96–6.81 (m, 1H) ppm; 13C
NMR (125 MHz, CDCl3): δ=154.4, 238.4, 136.2, 130.2, 129.7, 129.1,
128.5, 128.3, 128.0, 127.1, 125.4, 114.5, 114.0, 113.6, 108.6 ppm;
HRMS (ESI+) m/z calcd for C17H12N2 [M+H]+ 244.1000, found [M+
H]+ 244.0983.
°
Compound 1o: Yield, 84.5 mg (12%); yellow solid; mp: 180–182 C.
IR (νmax, KBr): 3436, 3099, 2359, 1604, 1482, 1409, 1344, 1112, 1087,
1010, 987, 856, 819, 761, 729, 686, 568 cm–1; 1H NMR (500 MHz,
CDCl3): δ=7.75 (d, 1H, J=8.7 Hz, H–6), 7.72 (d, 1H, J=3.0 Hz, H–1),
7.71 (d, 1H, J=2.0 Hz, H–9), 7.34 (dd, 1H, J=8.7 Hz, J=2.1 Hz, H–7),
6.86 (d, 1H, J=3.0 Hz, H–2), 2.94 (s, 3H, CH3) ppm; 13C NMR (75 MHz,
CDCl3) δ 153.7, 132.6, 130.4, 130.3, 126.4, 125.9, 122.5, 117.8, 114.1,
113.2, 95.3, 24.4 ppm; HRMS (ESI+) m/z calcd for C12H9BrClN2 [M+
H]+ 294.9632, found [M+H]+294.9625.
Pyrrolo[1,2-a]quinoxaline (1j). General procedure B from 4a (50 mg,
0.312 mmol); Yield: 52 mg (89%); cream-coloured solid; mp: 129–
1
°
130 C; cream-coloured solid; H NMR (500 MHz, CDCl3): δ=8.81 (s,
1H), 8.00–7.91 (m, 2H), 7.87 (dd, J=8.1 Hz, J=1.2 Hz, 1H), 7.57–7.49
(m, 1H), 7.45 (td, J=7.6 Hz, J=1.4 Hz, 1H), 6.93–6.85 (m, 2H) ppm;
HRMS (ESI+) m/z calcd for C11H8N2 [M+H]+ 168.0687, found [M+
H]+ 168.0718.
Enzyme inhibition
PTP1B inhibition screening: To test the activity of compounds 1,
inhibition assays were performed using a commercial kit (“PTP1B
Tyrosine Phosphatase Drug Discovery Kit”, BML–AK822 Enzo Life
Sciences). This kit performs a colorimetric test that measures the
phosphatase activity of purified PTP1B, based on the malachite
green assay. This assay measures the inorganic phosphate released
into the aqueous solution that complexes with malachite green
molybdate. The absorbance of the complex at 620 nm is propor-
tional to the concentration of free phosphate.
Bromination of pyrrolo[1,2-a]quinoxalines 1. General procedure:
A solution of NBS (2–2.5 mmol) in DMF (5.6 mL/mmol) was added
over a solution of 1 (2–2.5 mmol) in DMF (7.2 mL/mmol) cooled at
°
À 10 C at a speed of 0.17 mL/min. Next, the reaction mixture was
allowed to reach room temperature and satd NaCl (15 mL/mmol)
was added. The mixture was then extracted with CH2Cl2 (5×15 mL/
mmol) and the organic extracts were pooled and dried (MgSO4).
The desiccant was filtered off and the solvent was removed under
reduced pressure to give a residue that was chromatographed on
silica gel using hexane/AcOEt 9:1 as eluent.
The inhibitor was dissolved in reagent–grade dimethyl sulfoxide
(DMSO) (Sigma-Aldrich, Saint Louis, MO) for assay. The assay kit
contained recombinant PTP1B and the phosphopeptide substrate
(IRS5). The enzyme has an estimated Km of 85 μm for the substrate
which is used to measure PTP1B activity. PTP1B was prepared for all
assays to allow for a final amount of 2.5 ng/well and IRS5 was
prepared to have a final assay concentration of 75 μm. Briefly, 35 μL
of assay buffer was added to each well and incubated for 15
1-Bromo-4,7,8-trimethylpyrrolo[1,2-a]quinoxaline (1k), 3-bromo-4,7,8-
trimethylpyrrolo[1,2-a]quinoxaline (1l) and 1,2-dibromo-4,7,8-trimeth-
ylpyrrolo[1,2–a]quinoxaline (1m). From 1b (0.40 g, 1.94 mmol).
Compound 1k: Yield, 0.29 g (52%); cream–coloured solid; mp: 139–
°
140 C; IR (νmax, KBr): 3435, 2970, 2915, 1682, 1623, 1579, 1530, 1478,
1411, 1380, 1352, 1218, 1154, 1042, 912, 883, 857, 764, 735, 684,
674 cm–1; 1H NMR (500 MHz, CDCl3): δ=9.00 (s, 1H, H–9), 7.64 (s, 1H,
H–6), 6.82 (d, 1H, J=4.2 Hz, H–3), 6.78 (d, 1H, J=4.2 Hz, H–2), 2.64
(s, 3H, CH3), 2.41 (s, 3H, CH3), 2.36 (s, 3H, CH3) ppm; 13C NMR
(75 MHz, CDCl3): δ=152.3, 136.6, 134.4, 133.6, 129.3, 125.0, 122.7,
116.8, 113.4, 113.3, 93.8, 24.3, 20.2, 19.5 ppm; HRMS (ESI+) m/z
calcd for C14H14N2Br [M+H]+ 289.0340, found [M+H]+ 289.0338;
°
minutes at 37 C. Inhibitors or DMSO (10 μL) as baseline control
(final concentration 1 μm) were added to each well followed by 5
μL of PTP1B (2.5 ng per well) solution. Reactions were initiated by
adding 50 μL of IRS5 (final concentration 75 μm) and the plate was
°
incubated for 30 min at 37 C. Reactions were then terminated by
adding 25 μL of the provided phosphate detection reagent and
wells were agitated gently to mix. Colour was allowed to develop
for 25–30 min and absorbance was read at 620 nm using a
spectrophotometer.
°
Compound 1l: Yield, 0.21 g (39%); yellow solid; mp: 190–192 C; IR
(νmax, KBr): 3434, 3108, 2966, 2918, 1708, 1621, 1576, 1512, 1485,
1408, 1376, 1342, 1218, 1137, 1098, 1007, 975, 917, 882, 853, 758,
1
735, 691, 673, 605 cm–1; H NMR (200 MHz, CDCl3): δ=7.70 (d, 1H,
PTP1B/TCPTP inhibitory activity (IC50): PTP1B was purchased from
Enzo Life Sciences and TCPTP from Merck–Sigma Aldrich. Inhibitor
concentration at 50% inhibition was measured as hydrolysis of
J=2.5 Hz, H–1), 7.57 (s, 1H, H–6), 7.45 (s, 1H, H–9), 6.78 (d, 1H, J=
2.5 Hz, H–2), 2.93 (s, 3H, CH3), 2.37 (s, 3H, CH3), 2.33 (s, 3H, CH3)
ppm; 13C NMR (75 MHz, CDCl3): δ=151.6, 135.1, 134.4, 129.4, 128.0,
126.5, 118.2, 115.6, 115.5, 106.7, 98.4, 21.5, 20.5, 19.4 ppm; HRMS
[ESI–TOF] m/z calcd for C14H14N2Br [M+H]+ 289.0340, found [M+
H]+ 289.0345; Compound 1m: Yield, 0.10 g (8%); brown solid; mp:
°
pNPP (400 mm) at 37 C and pH 7.5 in a 96–well plate with 100 μL/
well. Buffered solution consisted of Tris–HCl (25 mm, pH 7.5), β-
mercaptoethanol (2 mm), EDTA (ethylenediaminotetraacetic acid,
1 mm), and DTT (dithiothreitol, 1 mm). Increasing concentrations (0,
ChemMedChem 2020, 15, 1–15
www.chemmedchem.org
11
© 2020 Wiley-VCH GmbH
��
These are not the final page numbers!

Full Papers
doi.org/10.1002/cmdc.202000446
ChemMedChem
0.04, 0.2, 1, 5 and 10 μm) of compounds were used to determinate
the IC50. In each experiment, the hydrolysis of pNPP was measured
as an increment in absorbance at 405 nm by adding substrate
solution (50 μL) to a solution consisting of inhibitor (45 μL) and
enzyme (5 μL; 60 ng/mL). The nonenzymatic hydrolysis of pNPP
was corrected by measuring the increase in absorbance at 405 nm
obtained in the absence of enzyme. Each data point was measured
by triplicate and experiment was performed by duplicate. Parame-
ters were calculated using SigmaPlot® (SPCC Inc., Chicago, IL).
with anti-total anti-IRS, anti-AKT or anti-STAT3 antibodies. Specific
total or phosphoproteins were visualized after subsequent incuba-
tion with a 1:5,000 dilution of anti–mouse or rabbit IgG conjugated
to horseradish peroxidase and a SuperSignal Chemiluminescence
detection procedure and imaged using an ImageQuant LAS 500
chemiluminescent detection chamber (General Electric Healthcare,
Little Chalfont, UK). Densitometry was determined using ImageJ
software (NIH). Three independent experiments were performed for
each condition.
1
2
3
4
5
6
7
8
9
Biological methods
Molecular Modelling
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
MTT viability test: a) HepG2 cell line. Cells were seeded in 24-well
plate and were treated for 24 h with the compounds. Then DMEM
was supplemented with MTT (0.5 mg/mL final concentration) and
Homology model of the holo structure of PTP1B was built using
SWISS-MODEL web server, using as template the crystal structure of
1T48.[46] QMEAN value resulted in À 0.09.[57] Using the Protein
Preparation Wizard module of the Schrödinger Suite (http://
www.Schrodinger.com) the receptor geometry was optimized. The
protonation states of charged amino acids were calculated with the
PROPKA module.
°
the myotubes were incubated at 37 C. After 4 h, DMSO was added
to dissolve the formazan crystals. Absorbance was measured at
570 nm using a spectrophotometer;[52,53] b) C2C12 cell line. Cells
were seeded in 24-well, differentiated to myotubes and subse-
quently treated with compounds (50 μm) for 24 h. Cytotoxicity of
PTP inhibitors was determined by the MTT assay. For the MTT assay,
cells were incubated with 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl
tetrazolium bromide (MTT) (50 μL of 5 mg/mL in PBS, Sigma
The described ligands were built with the LigPrep module of
Maestro, generating all possible states at pH 7�2, resulting in no
protonated states for this series.
°
Aldrich) for 4 h at 37 C in CO2 incubator. The medium was removed
Docking experiments were performed using extra precision (XP)
mode of the GLIDE module and a van der Waals radii scale factor of
1.0/0.8. Receptor grid was calculated using compound BB3 from
1T48 as the centre of the 10 Å-size box that enclosed the cryptic
pocket. The lowest energy ligand poses were selected for further
visual inspection and analysis of the ligand–receptor interactions.
and the purple formazan crystals formed were then dissolved by
adding 500 μL of dimethyl sulfoxide (DMSO) and mixed effectively
by pipetting up and down. Spectrophotometric absorbance of the
purple blue formazan dye was measured using Multimode Plate
Reader (Perkin Elmer) at 570 nm. Optical density of each sample
was compared with control optical density.
Geometry optimization and charge distribution for each compound
was obtained by the quantum mechanically calculated (RHF/3-
21G*//RHF/6-31G**) as implemented in Gaussian 03 (Gaussian, Inc.,
Wallingford, CT).[58]
C2C12 cell differentiation and treatment. Mouse skeletal muscle cell
lines, C2C12 myoblasts (from ATCCH CRL-1772TM) were routinely
maintained in DMEM, supplemented with 10% heat-inactivated
foetal bovine serum (FBS), 1000 UI/mL penicillin, 1000 UI/mL
MD simulations for each complex in the lowest energy docking
pose were carried out with the general AMBER14 (http://amberm-
d.org/) force field and the GAFF force field for the parametrization
of the small molecules, using the same conditions as described
previously.[59] Systems were introduced in a truncated octahedron
box of approximately 4000 TIP3P water molecules with 13 Å cut–off
distance and sodium cations were added to achieve electrical
neutrality. Smooth particle mesh Ewald (PME) method with a
spacing grid of 1 Å was used for treating electrostatic interactions
and SHAKE algorithm was applied to all hydrogen bonds with 2.0 fs
integration step.[60,61] An initial energy minimization of the water
molecules and counter–ions was carried out on each system. The
systems where further heated from 100 to 300 K in 25 ps, and
solvent molecules were progressively allowed to move freely. To
explore the complex between ligand and macromolecule 10 ns MD
simulations were carried out without any restraints, generating
snapshots each 20 ps for further analysis. The trajectories were
collected and the cpptraj module from AMBER14 was used to
calculate the root mean square deviation (RMSD) of the atomic
positions of the ligands and the most populated conformers.[62]
Reference values were taken from Pro188 to Ser201 for α3-helix,
from Ala264 to Met282 for α6-helix and from Ser286 to Trp291for
α7-helix. The MM–ISMSA program was used to calculate the total
binding energies for each compound.[63] All the physicochemical
properties and ADME parameters were obtained from web server
SWISS-ADME (http://www.swissadme.ch/) and pIC50 values from
the web server tool (http://www.sanjeevslab.org/tools.html).[64]
°
streptomycin and 2 mm L-Glutamine at 37 C in humidified air
containing 5% CO2. Differentiation was induced by switching the
growth medium to DMEM supplemented with 2% horse serum (HS;
differentiation medium). The differentiation medium was changed
every 48 h.[54]
Glucose uptake. To determine glucose uptake, C2C12 cells were
seeded in 24–well plate. Differentiated C2C12 cells were cultured in
fresh DMEM for 2 h, and then treated with DMSO (1% final
concentration, 90 min), insulin (10 nm final concentration, 30 min),
°
or compounds 1 (2 μm final concentration, 90 min) at 37 C. After
incubation, the cells were washed with PBS and fresh glucose-free
DMEM containing 100 μm 2NBDG added for 30 min. Then, cells
were washed three times with ice-cold PBS to remove excess
fluorescence in the wells, and fluorescence was measured using a
microplate reader at excitation and emission wavelengths of 485
and 530 nm, respectively. Background was subtracted from all
values. The cells were then lysed, the protein content was
determined in each experiment with a BCA protein assay kit. For
each experiment, at least two assays of each condition were
performed, and each experiment was repeated at least three
times.[55]
Phosphorylation of PTP1B substrates. It was analysed by western
blot.[56] Antibodies were purchased from Cell Signaling Technology
Inc. In brief, cell lysates were subjected to 5 or 12% SDS–PAGE.
Proteins were transferred to a polyvinylidene difluoride membrane.
The membranes were then blocked for 1 hour at room temperature
with 3% BSA in Tris-buffered saline containing 0.1% Tween 20.
Immunostaining to detect each protein was achieved with O/N
incubation with a 1:1,000 dilution of anti-phosphospecific anti-IRS,
anti-AKT or anti-STAT3 antibodies. Membranes were also blotted
ChemMedChem 2020, 15, 1–15
www.chemmedchem.org
12
© 2020 Wiley-VCH GmbH
��
These are not the final page numbers!

Full Papers
doi.org/10.1002/cmdc.202000446
ChemMedChem
[18] D. Barford, A. J. Flint, N. K. Tonks, Science 1994, 263, 1397–1404.
[19] A. Salmeen, J. N. Andersen, M. P. Myers, N. K. Tonks, D. Barford, Mol. Cell
2000, 6, 1401–1412.
[20] Y. A. Puius, Y. Zhao, M. Sullivan, D. S. Lawrence, S. C. Almo, Z.–Y. Zhang,
Proc. Natl. Acad. Sci. USA 1997, 94, 13420–13425.
Acknowledgements
1
2
3
4
5
6
7
8
9
Authors thank the financial support provided by: Spanish Minister-
io de Economía y Competitividad (CTQ2017-85263-R MINECO/
FEDER UE to J.J.V., MINECO RTI2018-097609-B-C21 to F.M.);
Instituto de Salud Carlos III and FEDER funds (RD16/0009/0015,
RD16/0009/18, PI17/01513 and PI17/00625to J.J.V., D.R.-P. and
M.R.-P.); Comunidad de Madrid (NOVELREN, B2017/BMD-3751 to
J.J.V. and D.R.-P.); and Fundación Renal Íñigo Alvarez de Toledo
(FRIAT). Fellowships were provided by: REDinREN and University of
[21] P. J. Ala, L. Gonneville, M. Hillman, M. Becker–Pasha, E. W. Yue, B. Douty,
B. Wayland, P. Polam, M. L. Crawley, E. McLaughlin, R. B. Sparks, B. Glass,
A. Takvorian, A. P. Combs, T. C. Burn, G. F. Hollis, R. Wynn, J. Biol. Chem.
2006, 281, 38013–38021.
[22] C. Wiesmann, K. J. Barr, J. Kung, J. Zhu, D. A. Erlanson, W. Shen, B. J.
Fahr, M. Zhong, L. Taylor, M. Randal, R. S. McDowell, S. K. Hansen, Nat.
Struct. Mol. Biol. 2004, 11, 730–737.
[23] N. Krishnan, D. Koveal, D. H. Miller, B. Xue, S. D. Akshinthala, J. Kragelj,
M. R. Jensen, C.–M. Gauss, R. Page, M. Blackledge, S. K. Muthuswamy, W.
Peti, N. K. Tonks, Nat. Chem. Biol. 2014, 10, 558–566.
[24] A. Huang, A. Ma, C. Mini-Rev. Med. Chem. 2013, 13, 607–616.
[25] G. W. H. Cheeseman, B. Tuck, J. Chem. Soc. 1966, 852–855.
[26] M. L. Heffernan, G. M. Irvine, Aust. J. Chem. 1976, 837–845.
[27] J. Montalibet, K. I. Skorey, B. P. Kenedy, Methods 2005, 23, 2–8.
[28] J. García-Marín, J. Biomol. Struct. Dyn. 2020. DOI: 10.1080/
07391102.2020.1790421.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
Alcala (to P.S.-A. and J.G.–M.); Ministerio de Economía
y
Competitividad (FPU16/01647to J.G.-M); and Ministerio de Educa-
ción, Cultura y Deporte (FPU15/02857 to B.D.G.). We acknowledge
BioSolveIt for a free one–year licence for FlexX software to J.G.-M
(Scientific Challenge, Winter 2017) and to ChemAxon for a free
three–year license for Marvin Suite software. We also thank Dr.
Ryan Walsh (INRS, Québec, Canada) for his kind advice and
comments.
[29] R. A. Walsh, PeerJ 2018, 6, e6082.
[30] R. A. Walsh, Anal. Biochem. 2019, 572, 58–62.
[31] J. J. Irwin, D. Duan, H. Torosyan, A. K. Doak, K. T. Ziebart, T. Sterling, G.
Tumanian, B. K. Shoichet, J. Med. Chem. 2015, 58, 7076–7087.
[32] S. L. McGovern, E. Caselli, N. Grigorieff, B. K. Shoichet, J. Med. Chem.
2002, 45, 1712–1722.
Conflict of Interest
[33] S. Boulton, R. Selvaratnam, R. Ahmed, K. Van, X. Cheng, G. Melacini, J.
Med. Chem. 2019, 62, 5063–5079.
The authors declare no conflict of interest.
[34] D. Duan, H. Torosyan, D. Elnatan, C. K. McLaughlin, J. Logie, M. S.
Shoichet, D. A. Agard, B. K. Shoichet, ACS Chem. Biol. 2017, 12, 282–290.
[35] H. Torosyan, B. K. Shoichet, J. Med. Chem. 2019, 62, 9593–9599.
[36] A. N. Ganesh, E. N. Donders, B. K. Shoichet, M. S. Shoichet, Nano Today,
2018, 19, 188–200.
[37] P. Lewi, E. Arnold, K. Andries, H. Bohets, H. Borghys, A. Clark, F. Daeyaert,
K. Das, M. P. de Bethune, M. de Jonge, J. Heeres, L. Koymans, J.
Leempoels, J. Peeters, P. Timmerman, W. Van den Broeck, F. Vanhoutte,
G. Van’t Klooster, M. Vinkers, Y. Volovik, P. A. Janssen, Drugs R&D, 2004,
5, 245–257.
Keywords: biological activity · inhibitors · insulin mimetics ·
PTP1B · pyrrolo[1,2-a]quinoxaline
[1] B. J. Goldstein, A. Bittner–Kowalczyk, M. F. White, M. Harbeck, J. Biol.
Chem. 2000, 275, 4283–4289.
[2] S. Dadke, J. Kusari, J. Chernoff, J. Biol. Chem. 2000, 275, 23642–23647.
[3] M. Delibegovic, D. Zimmer, C. Kauffman, K. Rak, E. G. Hong, Y. R. Cho,
J. K. Kim, B. B. Kahn, B. G. Neel, K. K. Bence, Diabetes 2009, 58, 590–599.
[4] L. D. Klaman, O. Boss, O. D. Peroni, J. K. Kim, J. L. Martino, J. M.
Zabolotny, N. Moghal, M. Lubkin, Y. B. Kim, A. H. Sharpe, A. Stricker–
Krongrad, G. I. Shulman, B. G. Neel, B. B. Kahn, Mol. Cell. Biol. 2000, 20,
5479–5489.
[5] M. Elchebly, P. Payette, E. Michaliszyn, W. Cromlish, S. Collins, A. L. Loy,
D. Normandin, A. Cheng, J. Himms–Hagen, C. C. Chan, C. Ramachan-
dran, M. J. Gresser, M. L. Tremblay, B. P. Kennedy, Science 1999, 283,
1544–1548.
[6] C. M. Rondinone, J. M. Trevillyan, J. Clampit, R. J. Gum, C. Berg, P.
Kroeger, L. Frost, B. A. Zinker, R. Reilly, R. Ulrich, M. Butler, B. P. Monia,
M. R. Jirousek, J. F. Waring, Diabetes 2002, 51, 2405–2411.
[7] B. A. Zinker, C. M. Rondinone, J. M. Trevillyan, R. J. Gum, J. E. Clampit,
J. F. Waring, N. Xie, D. Wilcox, P. Jacobson, L. Frost, P. E. Kroeger, R. M.
Reilly, S. Koterski, T. J. Opgenorth, R. G. Ulrich, S. Crosby, M. Butler, S. F.
Murray, R. A. McKay, S. Bhanot, B. P. Monia, M. R. Jirousek, Proc. Natl.
Acad. Sci. USA 2002, 99, 11357–11362.
[38] S. R. LaPlante, R. Carson, J. Gillard, N. Aubry, R. Coulombe, S. Bordeleau,
P. Bonneau, M. Little, J. O’Meara, P. L. Beaulieu, J. Med. Chem. 2013, 56,
5142–5150.
[39] J. M. Lopez–Nicolas, M. Perez–Gilabert, F. Garcia–Carmona, J. Agric. Food
Chem. 2009, 57, 4630–4635.
[40] M. F. Sassano, A. K. Doak, B. L. Roth, B. K. Shoichet, J. Med. Chem. 2013,
56, 2406–2414.
[41] D. Duan, A. K. Doak, L. Nedyalkova, B. K. Shoichet, ACS Chem. Biol. 2015,
10, 978–988.
[42] H. M. Berman, J. Westbrook, Z. Feng, G. Gilliland, T. N. Bhat, H. Weissig,
I. N. Shindyalov, P. E. Bourne, Nucleic Acids Res. 2000, 28, 235–242.
[43] A. J. Barr, Future Med. Chem. 2010, 2, 1563–1576.
[44] FlexX version 4.0; BioSolveIT GmbH, Sankt Augustin, Germany, 2019,
www.biosolveit.de/FlexX.
[45] S. Li, J. Zhang, S. Lu, W. Huang, L. Geng, Q. Shen, J. Zhang, PLoS One
2014, 9, e97668/1–e97668/10.
[46] L. Bordoli, F. Kiefer, K. Arnold, P. L. Benkert, J. Battey, T. Schwede, Nat.
Protoc. 2009, 4, 1–13.
[47] W. Cui, Y.–H. Cheng, L.–L. Geng, D.–S. Liang, T.–J. Hou, M.–J. Ji, J. Chem.
[8] Trodusquemine (MSI–1436) completed phase I for obesity and T2DM
Inf. Model. 2013, 53, 1157–1167.
and currently is suspended at phase
(www.clinicaltrials.gov).
[9] Z. Y. Zhang, S. Y. Lee, Expert Opin. Invest. Drugs, 2003, 12, 223–33.
[10] K. A. Lantz, S. G. Hart, S. L. Planey, M. F. Roitman, I. A. Ruiz–White, H. R.
Wolfe, M. P. McLane, Obesity, 2010, 18, 1516–1523.
[11] S. Fukuda, T. Ohta, S. Sakata, H. Morinaga, M. Ito, Y. Nakagawa, M.
Tanaka, M. Matsushita, Diabetes Obes. Metab. 2010, 12, 299–306.
[12] E. J. Moran, S. Sarshar, J. F. Cargill, M. M. Shahbaz, A. Lio, A. M. M. Mjalli,
R. W. Armstrong, J. Am. Chem. Soc. 1995, 117, 10787–10788.
[13] L.–J. Wang, B. Jiang, N. Wu, S.–Y. Wang, D.–Y. Shi, RSC Adv. 2015, 5,
48822–48834.
I for breast cancer trials
[48] A. Daina, O. Michielin, V. Zoete, Sci. Rep. 2017, 7, 42717/1–42717/13.
[49] J. Baell, M. A. Walters, Nature 2014, 513, 481–483.
[50] A. Daina, O. Michielin, V. Zoete, J. Chem. Inf. Model. 2014, 54, 3284–
3301.
[51] C. Abad–Zapatero, J. T. Metz, Drug Discovery Today 2005, 10, 464–469.
[52] H. Yuan, Y. Hu, Y. Zhu, Y. Zhang, C. Luo, Z. Li, T. Wen, W. Zhuang, J. Zou,
L. Hong, X. Zhang, I. Hisatome, T. Yamamoto, J. Cheng, Mol. Cell.
Endocrinol. 2017, 443, 138–145.
[53] F. Angius, A. Floris, Toxicol. in Vitro 2015, 29, 314–319.
[54] H. Yoshitomi, R. Tsuru, L. Li, J. Zhou, M. Kudo, T. Liu, M. Gao, PLoS One,
2017, 8, e0183988/1– e0183988/13.
[55] M. Hatem-Vaquero, M. Griera, A. García-Jerez, A. Luengo, J. Álvarez, J. A.
Rubio, L. Calleros, D. Rodríguez-Puyol, M. Rodríguez-Puyol, S. De Frutos,
J. Endocrinol. 2017, 234, 115–128.
[14] A. P. Combs, J. Med. Chem. 2010, 53, 2333–2344.
[15] S. De Munter, M. Kohn, M. Bollen, ACS Chem. Biol. 2013, 8, 36–45.
[16] W. Peti, R. Page, Bioorg. Med. Chem. 2015, 23, 2781–2785.
[17] M. A. Ghattas, N. Raslan, A. Sadeq, M. Al Sorkhy, N. Atatreh, Drug Des.
Dev. Ther. 2016, 10, 3197–3209.
[56] N. Guex, M. C. Peitsch, T. Schwede, Electrophoresis, 2009, 30, S162–S173.
ChemMedChem 2020, 15, 1–15
www.chemmedchem.org
13
© 2020 Wiley-VCH GmbH
��
These are not the final page numbers!

Full Papers
doi.org/10.1002/cmdc.202000446
ChemMedChem
[57] P. Benkert, M. Kuenzli, T. Schwede, Nucleic Acids Res. 2009, 37, W510–
W514.
[58] M. J. Frisch, G. W. Trucks, H. B. Schlegel, G. E. Scuseria, M. A. Robb, J. R.
Cheeseman, J. A. Montgomery, Jr., T. Vreven, K. N. Kudin, J. C. Burant,
J. M. Millam, S. S. Iyengar, J. Tomasi et al., Gaussian 03, Revision C.02,
Gaussian, Inc.: Wallingford CT, 2004.
[63] J. Klett, A. Nunez-Salgado, H. G. Dos Santos, A. Cortes-Cabrera, A.
Perona, R. Gil-Redondo, D. Abia, F. Gago, A. Morreale, J. Chem. Theory
Comput. 2012, 8, 3395–3408.
[64] C. Selvaraj, S. K. Triphati, K. K. Reddy, S. K. Singh, Curr. Trends Biotechnol.
Pharm. 2011, 5, 1104–1109.
1
2
3
4
[59] M. Purwn, J. Hernandez-Toribio, C. Coderch, R. Panchuk, N. Skorokhyd,
K. Filipiak, B. de Pascual-Teresa, A. Ramos, RSC Adv. 2016, 6, 66595–
66608.
[60] T. Darden, D. York, L. Pedersen, J. Chem. Phys. 1993, 92, 10089–10092.
[61] M. A. Lill, Biochemistry, 2011, 50, 6157–6169.
5
6
7
8
Manuscript received: June 20, 2020
Revised manuscript received: July 29, 2020
Version of record online: ■■■, ■■■■
[62] D. R. Roe, T. E. Cheatham, J. Chem. Theory Comput. 2013, 9, 3084–3095.
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
ChemMedChem 2020, 15, 1–15
www.chemmedchem.org
14
© 2020 Wiley-VCH GmbH
��
These are not the final page numbers!

FULL PAPERS
1
2
3
4
5
6
7
8
9
Rather picky: Pyrrolo[1,2-a]quinoxa-
lines substituted at C1–C4 and C7–C8
behave as insulin mimetics that
potently and selectively inhibit
PTP1B rather than its closest
homologue TCPTP. They bind to the
α3/α6/α7 tunnel in MD simulations
and form stable aggregates in equili-
brium with aggregator in buffered
solutions. The 4-benzyl derivative
also inhibits p-AKT and p-IRS.
J. García-Marín, M. Griera, Dr. P.
Sánchez-Alonso, B. Di Geronimo,
Prof. F. Mendicuti, Prof. M. Rodríguez-
Puyol*, Dr. R. Alajarín*, Prof. B.
de Pascual-Teresa, Prof. J. J. Vaquero*,
Dr. D. Rodríguez-Puyol*
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
1 – 15
Pyrrolo[1,2-a]quinoxalines: Insulin
Mimetics that Exhibit Potent and
Selective Inhibition against Protein
Tyrosine Phosphatase 1B