Structure-based design of protein tyrosine phosphatase-1B inhibitors

Article abstract of DOI:10.1016/j.bmcl.2005.03.068

Using structure-based design, a new class of inhibitors of protein tyrosine phosphatase-1B (PTP1B) has been identified, which incorporate the 1,2,5-thiadiazolidin-3-one-1,1-dioxide template.

Full text of DOI:10.1016/j.bmcl.2005.03.068

Bioorganic & Medicinal Chemistry Letters 15 (2005) 2503–2507
Structure-based design of protein tyrosine
phosphatase-1B inhibitors
Emma Black,a Jason Breed,a Alexander L. Breeze,a Kevin Embrey,a Robert Garcia,a
Thomas W. Gero,b Linda Godfrey,a Peter W. Kenny,a,* Andrew D. Morley,a,*
Claire A. Minshull,a Andrew D. Pannifer,c Jon Read,a Amanda Rees,a Daniel J. Russell,b
Dorin Toaderb and Julie Tuckera
aAstraZeneca, Mereside, Alderley Park, Macclesfield, Cheshire SK10 4TG, UK
bAstraZeneca, 35 Gatehouse Drive, Waltham, MA 0245102451, USA
cPfizer Central Research, Sandwich, Kent CT13 9NJ, UK
Received 27 December 2004; revised 10 March 2005; accepted 17 March 2005
Available online 16 April 2005
Abstract—Using structure-based design, a new class of inhibitors of protein tyrosine phosphatase-1B (PTP1B) has been identified,
which incorporate the 1,2,5-thiadiazolidin-3-one-1,1-dioxide template.
Ó 2005 Elsevier Ltd. All rights reserved.
1. Introduction
has proved difficult to achieve potency without compro-
mising physical properties.
Reversible phosphorylation of tyrosine residues is a uni-
versal feature of eukaryotic intracellular signalling and
provides an important mechanism for enzymatic and
cellular regulation.1 Tyrosine phosphorylation is con-
trolled by protein tyrosine kinases (PTKs) and phospha-
tases2,3 (PTPs), both of which are attractive target
classes for pharmaceutical intervention. Although PTPs
have been less extensively exploited than PTKs as drug
targets, inhibition of PTP1B, the prototypical member
of this enzyme class, has emerged as a potential ap-
proach to treatment of type II diabetes and obesity.4–6
At the beginning of this programme, no suitable small
molecule inhibitors were available as starting points,
although the very acidic phenyldifluorophosphonate
group was well known as a phosphotyrosine bioiso-
stere.8 Public domain9 crystal structures were available
for complexes of PTP1B with both phosphotyrosine
peptides (in an inactive C215S mutant of the enzyme)25
and non-hydrolysable difluorophosphonates.10 The an-
ionic group sits at the N-terminus of a helix with phos-
phate oxygen atoms interacting with R221 and a
number of backbone amide hydrogen atoms. The ligand
aromatic system interacts with both F182 and Y46. Ra-
tional, structure-based ligand design approaches re-
sulted in the synthesis of a number of templates that
potentially mimic these interactions. Among the most
interesting of these was the 1,2,5-thiadiazolidin-3-one-
1,1-dioxide template 1.
Most approaches to PTP1B inhibition have focused on
competition with substrate,5,6 although allosteric inhibi-
tors have recently been characterised.7 A combination of
charge and shape makes the phosphomonoester group a
challenging target for mimicry. Such characteristics may
be beneficial in the context of signalling and could be a
factor in the evolution of protein phosphorylation as a
regulatory mechanism. However, these same character-
istics have frustrated the search for inhibitors, and it
2. Synthesis
The 1,2,5-thiadiazolidin-3-one-1,1-dioxide scaffold is
synthesised according to Scheme 1.11 Anilines 2–6 were
readily alkylated with bromoacetate in the presence of
diisopropylamine. The resultant secondary amines re-
acted cleanly with tert-butyl(chlorosulfonyl)carbamate
Keywords: Protein tyrosine phosphatase-1B; Structure-based design.
*
Corresponding authors. Tel.: +44 01625 512782; fax: +44 01625
510823; e-mail addresses: peter.kenny@astrazeneca.com; andy.
morley@astrazeneca.com
0960-894X/$ - see front matter Ó 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmcl.2005.03.068

2504
E. Black et al. / Bioorg. Med. Chem. Lett. 15 (2005) 2503–2507
O
S
H
N
O
N
O
1
H
N
H
O
H
N
CO2Me
O
S
H
R'
N
R'
a
H
CO2Me
R'
N
R
b
R
R
2 R = H
3 R = H
R' =H
R'= Br
4 R = Me R' = H
5 R = OMe R' = H
6 R = OMe R' = Ph
O
S
H
N
O
1
7
8
9
R = H
R = H
R = Me R' = H
R = OMe R' = H
R' =H
R'= Br
R'
N
O
c
10 R = OMe R' = Ph
R
7 R = H R' = Br
11 R = H R' = Ph
d
Scheme 1. Synthesis of 1,2,5-thiadiazolidin-3-one-1,1-dioxide template. Reagents and conditions: (a) methyl bromoacetate, 1 equiv, DIPEA
2.5 equiv, DMF, 60 °C, 14 h ; (70–90%); (b) (i) BocNHSO2Cl 0.5 M in CH2Cl2, Et3N, 0 °C, 3 h; (ii) TFA/H2O (9:1), rt, 2 h; (75–85%); (c) NaH, THF,
rt, 2 h; (70–90%); (d) PhB(OH)2, Pd(PPh3)4, Cs2CO3, DME/H2O, microwave, 30 min, 100 °C; (45–50%).
and treatment with aqueous TFA allowed removal of
the Boc protecting group, to give the phenylsulfamides
in 75–85% yield for the two steps.
Cyclisation with sodium hydride afforded the 1,2,5-thi-
adiazolidin-3-one-1,1-dioxide template in excellent yield.
Modification of the aryl ring substituents can also be
undertaken. Analogue 7 reacted smoothly with phen-
ylboronic acid under Suzuki conditions to afford 11 in
reasonable yield.
Figure 1. Overlay of docked anion of 1 (in cyan) with a difluoro-
phosphonate ligand (see Ref. 12). Also shown are Q266 side chain
amide and the water molecule that mediates the interaction of the
difluorophosphonate ligand with this residue. For clarity only a
3. Docking studies
portion of the difluorophosphonate ligand is displayed.
The anion of 1 was docked12,13 into a PTP1B crystal
structure. The predicted binding mode is shown over-
layed with the crystallographic difluorophosphonate
ligand in Figure 1.
in the docked form of 1 was especially significant.
Encouraged by these findings, we decided to explore
the scaffold in more detail.
Two of the phosphonate oxygen atoms are matched by
their sulfonyl counterparts, while the third overlays with
the anionic nitrogen of 1. The carbonyl oxygen (at posi-
tion 3 of the 1,2,5-thiadiazolidin-3-one-1,1-dioxide tem-
plate) occupies a similar position to the water molecule
(observed in a number of PTP1B X-ray structures) that
mediates the interaction between the Q266 side chain
and the ligand CF2 group. This carbonyl was included
originally with the intention of modulating pKa. Fortu-
itously, it also appears to be able to mimic the water-
mediated interaction with the protein. As discussed
below, the orthogonal orientation between the two rings
4. NMR and crystallographic studies
Protein NMR was used to determine if 1 bound to the
active site of PTP1B. NMR experiments were conducted
using samples (0.3 mM) of uniformly 15N-labelled re-
combinant human PTP1B (residues 1-321) in a 50 mM
sodium phosphate-based buffer, pH 7.0, at 600 MHz
and 25 °C. 15N–1H HSQC spectra acquired at increasing
concentrations of 1 showed chemical shift perturbations

E. Black et al. / Bioorg. Med. Chem. Lett. 15 (2005) 2503–2507
2505
Table 1. Computeda energy differences between orthogonal and
coplanar conformations of anion of 1
Basis set
DE(RHF)b
DE(B3LYP)c
DE(MP2)d
/kJ molÀ1
/kJ molÀ1
/kJ molÀ1
6-31G*
6-31+G*
12
12
24
23
19
17
a Calculations carried out with Gaussian 98 program (Ref. 15).
b Restricted Hartee Fock calculation (Ref. 16).
c Density functional theory using B3LYP functional (Ref. 17).
d Second order Møller-Plesset perturbation theory (Ref. 16).
Figure 2. Selected regions of the 15N–1H HSQC spectra of recombi-
nant human PTP1B(1-321) in the absence (black contours) and
presence (red contours) of 4.0 mM 1. Labelled cross-peaks experience
characteristic chemical shift perturbations on binding of ligands to the
active site of the enzyme.
Figure 4. Minimum energy conformations for anionic forms of 1, 8
and 9 from calculations at the RHF/6-31G* level.
ortho Substitution of the phenyl ring was predicted to
destabilise the lower energy conformer, thereby render-
ing the bound conformation more accessible. Energy
minimised structures of 8 and 9 in which the rings are
forced out of coplanarity by the introduction of an ortho
substituent are shown in Figure 4.
Figure 3. Crystal structure of 1 (coloured by atom type) bound to
PTP1B, showing molecular surface of binding site compared with the
result from docking (cyan).
6. Ligand binding and enzyme inhibition
in amide resonances of PTP1B known to be sensitive to
occupation of the enzymeÕs active site, confirming
that the 1,2,5-thiadiazolidin-3-one-1,1-dioxide motif
mimics the native substrate (Fig. 2). The Kd calculated
from the shift perturbations was 8 mM. Assuming
competitive inhibition and a Kd for inorganic phosphate
of around 20 mM,14 this equates to approximately
2 mM in the absence of buffer phosphate.
Following the observation of binding by NMR, motif 1
was tested in a biochemical assay for its ability to inhibit
a C-terminal truncated form of recombinant human
PTP1B using a para-nitrophenyl phosphate (pNPP) as-
say. Compound 1 showed an IC50 = 1.6 mM for the rate
of hydrolysis of the pNPP substrate. The results for
other analogues and their Kd from NMR experiments
are summarised in Table 2.
The crystal structure of the PTP1B complex with 1 was
determined in order to confirm the predicted binding
mode.18,20 Crystallographic and docked ligands are
shown in Figure 3.
There is a good correlation between activity and ability
to adopt an orthogonal conformation. Introduction of a
methyl or methoxy substituent at the ortho position
leads to an order of magnitude increase in potency. Fur-
ther kinetic analyses with these compounds indicated
that they were all competitive and reversible inhibitors
of PTP1B, and NMR chemical shift perturbation studies
corroborated the improvement in affinity and the preser-
vation of binding mode (data not shown).
5. Conformational lock
Protein crystallography confirmed the prediction from
docking12 that 1 would bind with rings in an orthogonal
orientation. This anionic system lies outside the para-
meterisation of empirical force fields so the energetic
consequences of adopting this conformation were
investigated with three commonly used quantum
mechanical models. Two stationary points were found
on the potential energy hypersurface and it is the higher
energy of these that most closely approximates the
bound conformation. Full details of the calculations
are provided in the Supplementary material (Table 1).
Analysis of literature data suggested how the com-
pounds might be elaborated to increase potency. While
phenyl difluorophosphonate shows no inhibition of
PTP1B at 200 lM,8 an IC50 of 15 lM is observed for
its 3-phenyl analogue.23 Following this approach it
was determined that the 5-position of the aromatic ring
was the preferred point for substitution. Subsequently,
compounds 10 and 11 were synthesised, with biological
results shown in Table 2.

2506
E. Black et al. / Bioorg. Med. Chem. Lett. 15 (2005) 2503–2507
Table 2. IC50 and NMR Kd values
References and notes
Compound
NMR Kd/lMa
PTP1B Inhibition
IC50/lMf
1. Hunter, T. Cell 2000, 100, 113–127.
2. Alonso, A.; Sasis, J.; Bottini, N.; Friedberg, I.; Osterman,
A.; Godzik, A.; Hunter, T.; Dixon, J.; Mustelin, T. Cell
2004, 117, 699–711.
3. Mustelin, T.; Feng, G.-S.; Bottini, N.; Alonso, A.;
Kholod, N.; Birk, D.; Merlo, J.; Huynh, H. Front. Biosci.
2002, 7, d85–d142.
4. Elchebly, M.; Payette, P.; Michaliszyn, E.; Cromlish, W.;
Collins, S.; Lee Loy, A.; Normandin, D.; Cheng, A.;
Himms-Hagen, J.; Chan, C.-C.; Ramachandran, C.;
Gresser, M. J.; Tremblay, M. L.; Kennedy, B. P. Science
1999, 283, 1544–1548.
1
8000b
1608 (425)
203 (43)
138 (37)
8
Not tested
183c
9
10
11
<10d
2.47 (0.19)
132 (43)
Not determinede
a Measured in presence of 50 mM inorganic phosphate; for true Kd
these numbers should be divided by a factor of 3–4.
b Fast exchange.
c Fast–intermediate exchange.
d Intermediate–slow exchange—assumes diffusion-limited kon
.
e Protein gel formation.
5. Johnson, T. O.; Ermolieff, J.; Jirousek, M. R. Nat. Rev.
Drug Discov. 2002, 1, 696–709.
f Compounds were pre-incubated with 27 nM recombinant human
PTP1B (1-314) in buffer (50 mM Bis–Tris, 2 mM EDTA, 5 mM
DTT, 0.001% Triton X-100) for 10 min at room temperature. pNPP
was then added to a final concentration of 400 lM and the reaction
allowed to run for a further 15 min at room temperature. The reac-
tion was stopped by the addition of 1 M NaOH. Absorbance at
405 nm was measured using a Fluoroskan. Values are means of three
experiments; standard deviation is given in parentheses.
[pNPP] = Km.
6. Hooft van Huijsduijnen, R.; Sauer, W. B.; Bombrun, A.;
Swinnen, D. J. Med. Chem . 2004, 47, 4142–4146.
7. Wiesmann, C.; Barr, K. J.; Kung, J.; Zhu, J.; Erlanson, D.
J.; Shen, W.; Fahr, B. J.; Zhong, M.; Taylor, L.; Randal,
M.; McDowell, R. S.; Hansen, S. K. Nat. Struct. Mol.
Biol. 2004, 11, 730–737.
8. Kole, H. K.; Smyth, M. S.; Russ, P. L.; Burke, T. R.
Biochem. J . 1995, 311, 1025–1031.
9. Berman, H. M.; Westbrook, J.; Feng, Z.; Gilliland, G.;
Bhat, T. N.; Weissig, H.; Shindyalov, I. N.; Bourne, P. E.
Nucleic Acids Res. 2000, 28, 235–242.
10. Groves, M. R.; Yao, Z. J.; Roller, P. P.; Burke, T. R.;
Barford, D. Biochemistry 1998, 37, 17773–17783.
11. Kenny, P. W.; Morley, A. D.; Russell, D. J.; Toader, D.
PCT Int. Appl. WO 2004050646, 2004.
12. Gold13 (Version 2.0, Cambridge Crystallographic Data
Centre, Cambridge CB2 1EZ, UK, www.ccdc.cam.ac.uk)
was used for docking using PDB9 structure 1BZJ10.
Difluorophosphonate ligand, water and Q266 residue in
Figure 1 are from 1BZC10 after alpha carbon alignment
with docking model.
13. Jones, G.; Willett, P.; Glen, R. C.; Leach, A. R.; Taylor,
R. J. Mol. Biol . 1997, 267, 727–748.
14. Zhang, Y.-L.; Zhang, Z. Y. Anal. Biochem. 1998, 261,
139–148.
Figure 5. Crystal structure of 10 bound to PTP1B showing molecular
surface of protein.
15. Gaussian 98, Revision A.11.3, Gaussian Inc., Carnegie,
PA 15106, USA, www.gaussian.com.
Encouragingly, these compounds show significant
improvements in inhibiting PTP1B activity compared
with their unsubstituted partners, SAR being additive.
The crystal structure of 10 bound into the active site
of PTP1B was obtained18,21 and is shown in Figure 5.
16. Hehre, W. J.; Radom, L.; Schleyer, P. v. R.; Pople, J. A.
Ab Initio Molecular Orbital Theory; Wiley, 1986.
17. Koch, W.; Holthausen, M. C. A ChemistÕs Guide to
Density Functional Theory, 2nd ed.; Wiley-VCH: Wein-
heim, 2001.
18. Protein and crystals were obtained according to established
procedures.19 For the co-crystallisation of 1, compound
was added to protein solution to a final concentration of
5 mM prior to crystallisation. Compound 10 was soaked
into apo crystals of PTP1B at a concentration of 0.5 mM.
Diffraction data for compound 1 were collected on beam-
line PX9.6 at SRS, Daresbury at 100 K.20 Diffraction data
for compound 10 were collected on beamline X13 at DESY,
Hamburg at 100 K.21 Data processing, data reduction and
structures solution by molecular replacement were carried
out using programs from the CCP4 suite.22 The protein
complex models were refined using CNX (version 2000.1,
Accelrys) and REFMAC522 and the final structures have
been deposited in the Protein Data Bank with deposition
codes 2BGE, and 2BGD together with structure factors
and detailed experimental conditions.
This work shows how joint application of protein NMR
and crystallography can allow a millimolar lead to be
exploited with a minimum amount of synthetic effort.
The approach highlighted above has identified low
micromolar PTP1B inhibitors, with molecular weights
of ꢀ300 Da and can be regarded as an example of frag-
ment based drug discovery.24 The 1,2,5-thiadiazolidin-3-
one-1,1-dioxide template has been shown to be an
excellent warhead for PTP1B inhibition. Compound 10
is a useful low molecular weight starting point for a drug
discovery programme.
Supplementary data
19. Barford, D.; Keller, J. C.; Flint, A. J.; Tonks, N. K. J.
Mol. Biol. 1994, 239, 762.
20. Crystallographic parameters and statistics for 1 are: space
Supplementary data associated with this article can be
found, in the online version, at doi:10.1016/
j.bmcl.2005.03.068.
˚
˚
group P3121, unit cell 89.0, 89.0, 104.7 A, resolution 1.8 A,
45,256 unique reflections from 160,793 observations give

E. Black et al. / Bioorg. Med. Chem. Lett. 15 (2005) 2503–2507
2507
99.7% completeness with Rmerge = 6.6% and I/r(I) of 11.0.
The final model containing 2424 protein, 378 water and 14
inhibitor atoms has an R-factor of 21.8% (Rfree using 5%
of data is 24.5%). Mean temperature factor for protein is
and 31 inhibitor atoms has an R-factor of 16.6% (Rfree
using 5% of data is 20.0%). Mean temperature factor for
protein is 25.7 A , and for ligand is 34.8 A .
22. CCP4 Acta Crystallogr. 1994, D50, 760.
23. Kotoris, C. C.; Chen, M.-J.; Taylor, S. D. Bioorg. Med.
Chem. Lett. 1998, 8, 3275–3280.
2
2
˚
˚
2
2
˚
˚
32.7 A , and for ligand is 31.3 A .
21. Crystallographic parameters and statistics for 10 are:
˚
space group P3121, unit cell 88.9, 88.9, 104.5 A, resolution
24. Erlansen, D. A.; McDowell, R. S.; OÕBrien, T. J. Med.
Chem. 2004, 47, 3463–3482.
˚
2.4 A, 17,876 unique reflections from 176,071 observations
give 98.6% completeness with Rmerge = 5.5% and I/r(I) of
11.2. The final model containing 2398 protein, 174 water
25. Jia, Z.; Barford, D.; Flint, A. J.; Tonks, N. K. Science
2004, 268, 1754–1758.