Inhibition studies with rationally designed inhibitors of the human low molecular weight protein tyrosine phosphatase

Article abstract of DOI:10.1016/j.bmc.2004.01.042

The human low molecular weight protein tyrosine phosphatase (HCPTP) is ubiquitously expressed as two isoforms in a wide range of human cells and may be involved in regulating the metastatic nature of epithelial tumors. A homology model is presented for th

Full text of DOI:10.1016/j.bmc.2004.01.042

Bioorganic & Medicinal Chemistry 12 (2004) 1867–1880
Inhibition studies with rationally designed inhibitors of the
human low molecular weight protein tyrosine phosphatase
Adam P. R. Zabell,a,b Steven Corden,b Paul Helquist,b Cynthia V. Stauffachera
and Olaf Wiestb,*
aDepartment of Biological Sciences and the Purdue Cancer Center, Purdue University,
West Lafayette, Indiana 47907-1392, USA
bDepartment of Chemistry and Biochemistry and the Walther Cancer Research Center, University of Notre Dame,
Notre Dame, Indiana 46556-5670, USA
Received 27 January 2003; revised 19 January 2004; accepted 27 January 2004
Abstract—The human low molecular weight protein tyrosine phosphatase (HCPTP) is ubiquitously expressed as two isoforms in a
wide range of human cells and may be involved in regulating the metastatic nature of epithelial tumors. A homology model is
presented for the HCPTP-B isoform based on known X-ray crystal structures of other low molecular weight PTPs. A comparison
of the two isoform structures indicates the possibility of developing isoform-specific inhibitors of HCPTP. Molecular dynamics
simulations with CHARMM have been used to study the binding modes of the known adenine effector and phosphate in the active
site of both isoforms. This analysis led to the design of the initial lead compound, based on an azaindole ring moiety, which was
then also evaluated computationally. A comparison of these simulations indicates the need for a phosphonate group on the indole
and provides insight into inhibitor binding modes. Compounds with varying degrees of structural similarity to the azaindole have
been synthesized and tested for inhibition with each isoform. These molecular systems were examined with the program AutoDock,
and comparisons made with the kinetics and the explicit simulations to validate AutoDock as a screening tool for potential inhi-
bitors. Two compounds were experimentally found to have sub-millimolar inhibition, but the greater solubility of one reinforces the
need for experimental testing alongside computational analysis.
# 2004 Elsevier Ltd. All rights reserved.
1. Introduction
their signaling pathway, only PTP1B and the dual spe-
cificity phosphatase Cdc25 have been studied as drug
targets.10ꢀ15 That work proposed a set of compounds
with remarkable diversity, many of which achieve spe-
cificity through bidentate binding.16
Tyrosine phosphorylation and dephosphorylation are
crucial elements in eukaryotic signal transduction.1ꢀ3
Aberrant cell behavior is correlated with persistent ele-
vated levels of tyrosine phosphorylation in a variety of
proteins, which is considered one of the primary indica-
tors for oncogenic transformation.4,5 Several studies6ꢀ8
have indicated phosphatases are directly involved in cel-
lular transformation, where the action of a phosphatase
may increase or decrease the enzymatic activity of an
oncogene. This increased understanding of the role
dephosphorylation plays in regulation of cell behavior
has fostered an interest in using the family of protein
tyrosine phosphatases (PTPs) as drug targets.9
Although many PTPs have been studied with regard to
The Eph family of receptor kinases are involved in a
variety of epithelial intracellular contact systems, prin-
cipally during embryonic development.17ꢀ19 Elevated
concentrations of Eph receptors have been found in a
broad range of metastatic tumors, but at a reduced
phosphorylation state relative to nontransformed
cells.20,21 Regulation of EphB122 and EphA26 has been
shown to be associated with the binding of the low
molecular weight protein tyrosine phosphatase (LMW-
PTP) to these signaling kinases. In the latter study,
Kikawa et al. demonstrated that LMW-PTP over-
expression is sufficient to transform non-transformed
epithelial cells. This in turn suggests that inhibition of
LMW-PTP could provide a means to pharmaceutically
Keywords: EphA2; Human LMW-PTP; AutoDock; Inhibitor design.
* Corresponding author. Tel.: +1-574-631-5876; fax: +1-574-631-
6652; e-mail: owiest@nd.edu
0968-0896/$ - see front matter # 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmc.2004.01.042

1868
A. P. R. Zabell et al. / Bioorg. Med. Chem. 12 (2004) 1867–1880
regulate EphA2 phosphorylation levels and thus control
oncogenic potential.
isoform41 and show an average root mean squared dif-
ference of less than 0.5 A in backbone coordinates. The
variable region in the human isoform makes up one
edge of the active site and extends down the side of
the presumed specificity cleft seen in all known
LMW-PTP structures41 (Fig. 1a). Variable loop residues
of particular interest in the A isoform are Glu50 and
Asn53, which are changed to Asn50 and Arg53 in the
B isoform. These residues are close to or at the lip of
the active site and appear to be well suited to provide
substrate or inhibitor specificity.
The LMW-PTPs have been characterized in a wide
range of organisms23ꢀ36 as 18-kDa enzymes with the
conserved CX5R active site motif found in all PTPs.37
The enzyme exists as two distinct isoforms in humans,
HCPTP-A and HCPTP-B,38 that are identical in
sequence outside of the variable spliced region, which
extends from residue 40 to residue 73. Crystal structures
have been solved for bovine,39 yeast,40 and the human A
Figure 1. (a) Ribbon drawing of HCPTP-A with a superimposed transparent molecular surface. The Active site loop is colored cyan and the vari-
able region is colored magenta. Active site residue Cys12 and the aromatic residues Tyr49, Tyr131, and Tyr132 at the mouth of the active site are
explicitly labeled. An asterisk indicates the location of the Y131 dish. The specificity cleft extends down along the right-hand side of the Y131 dish.
This figure and subsequent ribbon figures were drawn with GRASP,45 Molscript46 and Raster3-D.47 (b) Sequence alignment of the two human
isoforms of the LMW-PTP (humA and humB) with bovine (BPTP) and yeast (LTP1) as performed by Modeller.59 The active site loop residues are
shown in bold and the variable region for the human isoforms is boxed. Sequence numbering and secondary structure elements come from the
HCPTP-A structure. Figure drawn with Alscript.48

A. P. R. Zabell et al. / Bioorg. Med. Chem. 12 (2004) 1867–1880
1869
Previous studies have shown that adenine activates
HCPTP-B and the yeast LMW-PTP, but inhibits
HCPTP-A.38,42 Several authors have suggested reasons
for the differences in this modulation,38,42,43 but no
complete test of these hypotheses via mutant studies or
structural experiments has been reported. The yeast
LMW-PTP cocrystalized with adenine42 provides the
only structural evidence to help explain this modulating
effect. Bound in the complex with a configuration
shown schematically in 1, the adenine is at the position
of the substrate leaving group with a water molecule
between the adenine and the phosphate cradled in the
active site loop. This water is positioned to act as a
nucleophile to release the phosphoenzyme intermediate,
hydrolysis of which has been shown to be the rate
determining step.44 While this helps to explain the acti-
vation for the yeast and HCPTP-B forms, inhibition of
HCPTP-A upon formation of this complex remains
unclear.
these compounds indicates that the relative ranking
predicted by AutoDock is valid provided there are at
least three functional groups on the molecule.
2. Computational methods
2.1. Molecular dynamics with compound 2
In order to investigate the interactions of adenine and
the proposed inhibitors with both human isoforms, it
was necessary to create a homology model for HCPTP-
B because there was no experimentally determined
structure of this enzyme available. The human A iso-
form,41 along with the bovine (BPTP),57 and yeast
(LTP1)42 enzymes were obtained (PDB58 codes 5PNT,
1DG9, and 1D2A respectively) and aligned according to
their amino acid sequence (Fig. 1b) by the program
Modeller 4.59 The homology model for HCPTP-B
as created by Modeller was based on this sequence
alignment and visually examined to ensure appropriate
placement of the catalytic residues.
The emerging importance of PTPs in a variety of disease
states and the variety of kinetic and structural data for
LMW-PTPs has led us to pursue inhibitor design for the
human LMW-PTPs with the eventual goal of designing,
synthesizing, and evaluating isoform-specific inhibitors.
As a guide to our experimental efforts to synthesize
HCPTP inhibitors, we use in silico screening of poten-
tial ligands. Molecular modeling to design specific inhi-
bitors provides a fast and inexpensive alternative to
traditional trial and error drug design approaches.49ꢀ53
Although highly dependent upon the system under
investigation, success rates when attempting to repro-
duce known complexes are reported in the range of 70–
80% if a suitable level of approximation is chosen.53 To
validate our rapid screening methodology, we compared
results generated from the rigorously developed energy
function in CHARMM54,55 to those from the faster and
empirically based energy function in AutoDock.56
CHARMM can provide detailed information about the
structure and dynamics of an enzyme–ligand complex,
while AutoDock can quickly generate multiple plausible
ligand binding orientations. Molecular dynamics in
CHARMM using an explicit water box was performed
on both HCPTP isoforms with the bound adenine,
inorganic phosphate, and several well-defined water
molecules in their initial orientation from the crystal
structure for the yeast LMW-PTP. The same calcula-
tions were then performed with a phosphonated 5-
azaindole (compound 2), which was designed based on
the adenine crystal structure and was expected to mimic
those interactions. These systems were then examined
using AutoDock, as were several molecules structurally
similar to 2 but more synthetically accessible (com-
pounds 3–6). The experimental inhibition of HCPTP by
2.2. Initial structures
The relative alignment of adenine, inorganic phosphate,
and four structurally relevant water molecules in the
active site was taken from the crystal structure of ade-
nine bound to LTP142 and reproduced with these mole-
cules in the same relative position within the HCPTP-A
and HCPTP-B isoforms. Superposition of the enzyme
structures was performed with CHARMM2854,55 using
residues 4-39 (HCPTP numbering). This sequence con-
tains the first strand of the b-sheet, the conserved active
site loop, and the first a-helix.
The geometry of the phosphonated 5-azaindole was
optimized in the neutral charge state with the program
Gaussian9860 using the HF/6-31G(d) method. Hydro-
gen atoms were removed to model the phosphonate as a
dianion, and the charges on the molecule were recalcu-
lated in a single-point calculation. The alignment of 2 in
the active site was performed in CHARMM from a
minimum RMSD overlay of the adenine ring coordi-
nates and corresponding atoms from the azaindole.
Energy minimization was then performed in the absence
of protein with restraints on the azaindole atoms to

1870
A. P. R. Zabell et al. / Bioorg. Med. Chem. 12 (2004) 1867–1880
maintain their position. Distance restraints were applied
to each atom in the –PO3 group to force them into the
corresponding phosphate atom locations.
and partial atomic charges were determined by the
Gasteiger method66 with modification to ensure unit
charge on each residue. Charge modifications were
empirically applied to backbone nitrogens. These chan-
ges were rarely more than ꢂ0.002 charge units except
for proline residues, which regularly deviated from unit
charge by ꢀ0.005. Ligands were allowed full internal
flexibility around rotatable (non-ring system) bonds as
well as motion within the simulation space. A total of 50
independent simulations with a population size of 50
members were run for each ligand using AutoDock with
default parameters. Simulations were ranked according
to the docked energy between the protein and the
ligand, a summation of internal ligand energy and
intermolecular energy terms.
2.3. Molecular dynamics simulations
Molecular dynamics (MD) simulations were performed
in CHARMM using the all-atom parameter set.55 Pro-
tonation states for the catalytic Cys12, the conserved
Asp129, and all histidine residues were based on model
and pKa studies.61ꢀ63 The catalytic Cys is the nucleophile
that attacks the –PO3 group and becomes the modified
residue in the phospho-enzyme intermediate. The con-
served Asp donates a hydrogen to the substrate phenolic
oxygen, protonating the tyrosyl leaving group. The
initial orientation of the Asp129 side-chain was adjusted
to permit a hydrogen bond from N9–H9 of adenine to
the side-chain carbonyl of the aspartic acid. The partial
charges for 2, the inorganic phosphate, and the anionic
Cys12 were constructed based on the values calculated
from Gaussian98 and on empirical adjustment to match
the values already present in the CHARMM parameter
set. Optimized coordinates and partial charges will be
made available upon request. Bond, angle, and dihedral
forces were similarly determined where necessary.
Hydrogen atoms were built using CHARMM onto
molecules whose coordinates were determined by
X-ray crystallography and the resulting complex mini-
mized while keeping all non-hydrogen atoms at fixed
positions.
3. Results and discussion
3.1. Molecular dynamics of HCPTP with adenine
The goal of this research is to develop inhibitors that
can distinguish between the two HCPTP isoforms. For
our initial models, we chose the crystal structure of
HCPTP-A which was solved with the buffer molecule 2-
(N-morpholino)-ethanesulfonic acid in the active site.41
The buffer molecule roughly approximates the active
site position of the phosphate and adenine from the
LTP1-adenine structure and allows an initial build of 1
and 2 with the aromatic amino acid side chains of the
active site positioned to interact with the effector.
Although there is no X-ray or NMR structure of the
HCPTP-B isoform currently available, the high struc-
tural similarity between the known structures with an
average backbone RMSD of 0.5 A across the secondary
structure elements and the extensive sequence identity of
87% between the human isoforms (Fig. 1b) indicated
that a homology model of this second isoform should
adequately represent the structure. The bovine LMW-
PTP, with a sequence identity of 94% to the HCPTP-B
isoform, was used to validate plausible side chain
orientations in the homology model with satisfactory
results. No additional adjustments of the model were
performed after checking the structure, and the homol-
ogy model of HCPTP-B was considered to be a valid
structure for all ensuing calculations.
A 1 ns MD simulation was performed for each isoform
with either the adenine, water, and phosphate from the
yeast crystal structure (structure 1) or compound 2
bound in the active site. A truncated octahedron water
box with a 9 A minimal distance between protein and
box edge was built for each complex and minimized
prior to MD. Dynamics were performed using Vertlet’s
algorithm64 with a 2 fs timestep at 298 K. A constant
pressure of 1 atm was maintained for the simulation and
box volume was monitored to ensure reasonable con-
sistency. For the first 20 ps of MD, a 5 kcal/mol
restraint was applied to all backbone non-hydrogen
atoms. These initial time-points were not included in the
analysis of the simulation.
2.4. Docking simulations
After aligning the active site residues of the yeast com-
plex with those in each human isoform, the adenine,
phosphate, and bridging water were placed in the
corresponding human enzymes, as shown in Figure 2a.
In addition, the three water molecules near the mouth of
the active site that form hydrogen bonds between the
yeast LMW-PTP and adenine were placed in each
human isoform. After hydrating this ‘instantly docked’
complex and performing energy minimization to alle-
viate possible bad atom contacts, a 1 ns MD simulation
was performed. The total potential energy of the system
(Fig. 3a) and the interaction energy between the ade-
nine, phosphate, and bridging water (APW) and the
enzyme (Fig. 3b) were examined to determine stability
of the system and the complex, respectively. Further
measures of complex stability were made by following
Docking was performed with the program Auto-
Dock3.0.5,56 applying the Lamarckian genetic algo-
rithm protocol.65 The simulation space was defined as a
box 18.0ꢁ20.8ꢁ20.8 A, which includes the active site
and all residues of the variable region that line the
proposed specificity cleft. The interaction energy
between ligand and enzyme was calculated on a 0.2 A
grid using atom affinity potentials, greatly increasing the
calculation speed relative to CHARMM. For simula-
tions with inorganic phosphate, the phosphate ion was
retained at the base of the active site in the same relative
location as seen in the yeast crystal structure and treated
as a part of the inflexible protein for the purposes of
calculating the grid potential. Polar hydrogen atoms
were retained, ring carbons were defined as aromatic,

A. P. R. Zabell et al. / Bioorg. Med. Chem. 12 (2004) 1867–1880
1871
characteristic distances between enzyme and APW
atoms, or movement of portions of APWduring the
simulation (Fig. 3c and d). The potential energy of both
isoenzyme-inhibitor systems is stable, although the
interaction energy between the enzymes and APWis
not, and instead shows an increase in energy of 50–75
kcal/mol over the course of the simulation. The distance
between the phosphate and the catalytic cysteine
remains constant at either 4.0 or 5.5 A, but the dis-
placement of adenine in both isoforms increases sig-
nificantly during the simulation to over 20 A for
HCPTP-A and to about 5 A for HCPTP-B. None of the
explicitly placed water molecules remained in their
intended hydrogen bonding position (data not shown).
This is consistent with a disruption of the interaction
between the adenine and the active site residues. A
snapshot of the complex taken at 0, 100, 500, and 800 ns
during the simulation (Fig. 2b) confirms that both the
adenine and the bridging water leave the active site. It
is important to note that the phosphate stays in the
original location, suggesting that a suitable inhibitor
would require covalent linkage between phosphate and
adenine mimic.
3.2. Molecular dynamics with inhibitor 2
The results described in the last section, together with
the well-ordered position of the APWwithin the LTP1-
adenine complex, suggest the design principle for the
lead compound of a new class of inhibitors of HCPTP.
The distance between the phosphorus atom and N3 of
adenine is 4.0 A, roughly equal to that of an ethyl
group. The use of such a linker would also replace the
water molecule thought to be essential for adenine
modulation of the enzyme.44 Retaining the two ade-
nine nitrogens directly involved in hydrogen bonding
to the yeast enzyme—the hydrogen bond acceptor
atom N1 and the hydrogen bond donor atom N9—and
Figure 2. (a) Ribbon drawing of HCPTP-A expanded around the active site with a transparent molecular surface. Residues Cys12, Tyr49, Tyr131,
and Tyr132 are explicitly shown. The adenine, phosphate, and bridging water molecules are in the same relative positions in this structure as in the
LTP1-adenine crystal structure.42 The hydrophobic packing between the adenine and the aromatic residues can be easily recognized. (b) Snapshots
of the MD with HCPTP-A and the APWcomplex at 0 (standard atom colors), 100 (red), 500 (yellow), and 800 (cyan) ps. The catalytic Cys12 is
shown as a reference for the phosphate motion. Adenine at 800 ps and the water at 500 and 800 ps have left the vicinity of the active site and are not
visible from this perspective. (c) Snapshots of the MD with HCPTP-A and 2 at 0 (standard atom colors), 100 (red), 500 (yellow), and 800 (cyan) ps.
(d) Top 10 AutoDock structures of 4 docked with HCPTP-A. The structure with the best docked energy is shown as a blue ball and stick model,
while the remaining nine structures are drawn in magenta lines.

1872
A. P. R. Zabell et al. / Bioorg. Med. Chem. 12 (2004) 1867–1880
Figure 3. Monitored values for the CHARMM dynamics simulations
of HCPTP-A (black) and HCPTP-B (grey) with the adenine, phos-
phate, and bridging water (APW). (a) Total potential energy of the
system, in kcal/mol. (b) Interaction energy between enzyme and
the APW, in kcal/mol. (c) Distance in Angstroms between the
phosphorus atom of the phosphate and Sg of Cys12. (d) Distance
RMSD between the adenine position at each timepoint and the
adenine position at the start of the MD, in Angstroms.
Figure 4. Monitored values for the CHARMM dynamics simulations
of HCPTP-A (black) and HCPTP-B (grey) with compound 2. (a)
Total potential energy of the system, in kcal/mol. (b) Interaction
energy between enzyme and 2, in kcal/mol. (c) Distance in Angstroms
between the phosphorus atom of the phosphonate and Sg of Cys12.
(d) Distance RMSD between the azaindole position at each timepoint
and the azaindole position at the start of the MD, in Angstroms.
(e) Relative rotation of the azaindole ring as part of the RMSD
calculation of (d).
modifying the phosphate to a phosphonate to improve
hydrolytic stability and reduce biological degradation,
we proposed that the 5-azaindole 2 would function as
an effective inhibitor for LMW-PTPs.
The validity of this design concept was tested by two
MD simulations, which differ only in that the APWis
replaced with 2. The total potential energy and interac-
tion energy between the compound and each of the two
isoenzymes were monitored as before, with slightly dif-
ferent distance measurements being taken to correspond
with the use of a single molecule in the active site. The
potential energy of the system is again stable (Fig. 4a),
but it is gratifying to note that in contrast to the APW
system, the interaction energy is constant in the complex
with HCPTP-A and slightly improves with HCPTP-B
(Fig. 4b), indicating that 2 is a reasonable inhibitor. The
distance measurements lend additional support to the
hypothesis of effective inhibition, with small fluctua-
tions between the phosphorus in the phosphonate group
and Cys12Sg of approximately 0.5 A (Fig. 4c). This is
the same degree of motion seen with the phosphate in
the simulations with 1. It is interesting to note that the
azaindole ring is not involved in close hydrogen bond-
ing either as an acceptor to residue 50 or a donor to
residue 129 (data not shown). Without specific atomic
interactions between 2 and HCPTP, the azaindole
undergoes a 1–3 A displacement over the course of the
simulation (Fig. 4d). At the same time, the azaindole is
limited to a rotation between 30ꢃ and 90ꢃ (Fig. 4e). This
indicates that the shape of 2 is more important than the
distribution of hydrogen bond donors and acceptors on
the aromatic ring. Consequently, a further simplifica-
tion of the system for the purpose of an easy experi-
mental proof-of-principle of the design concept should
be possible.
Analysis of the geometry in the simulations also pro-
vides insight into the nature of inhibitor binding. Figure
2c shows snapshots of 2 in the active site at the same
time-points used in Figure 2b. As expected from the
trajectories shown in Figure 4, 2 remains tightly bound
in the active site. While the phosphonate and the ethyl
linker show very similar positioning, the aromatic
azaindole ring exhibits torsional flexibility. Along with
the distance and energy graphs, these snapshots high-
light the three conclusions from our computer-based
inhibitor design for HCPTP. Most importantly, the
covalent linkage of a phosphate or phosphate mimic
with an aromatic system should ensure stronger binding
than adenine. Second, hydrogen bond donors or accep-
tors appeared to make no difference in binding affinity
for the inhibitor atoms at the top of the active site.
Third, the inhibitor moiety at the top of the active site
should have a planar shape and preferably be aromatic.
Additional support for these points is the aromatic nat-
ure of the natural substrate as well as the roughly planar
or aromatic nature of many inhibitors, among them
buffer molecules such as 2-(N-morpholino)-ethane-
sulfonic acid (MES) or N-2-Hydroxyethylpiperazine-N0-
2-ethanesulfonic acid (HEPES), and many of the inhi-
bitors designed for the similarly shaped active site of
PTP1B.10,16
3.3. Docking of adenine and azaindole to each isoform
The results of the MD simulations show that 2 should
effectively bind to either isoform, but a variety of similar

A. P. R. Zabell et al. / Bioorg. Med. Chem. 12 (2004) 1867–1880
1873
inhibitor structures are possible. The computational
requirements of the full 1 ns MD simulation are prohi-
bitive for use as a screening tool. If it could be validated
by comparison with the results from the explicit MD
simulation, a simplified energy function could be used
to increase the speed of analysis for a larger series of
compounds. We selected AutoDock for its ability to
determine ligand structures with internal flexibility and
the improved speed of calculation afforded by pre-cal-
culating a grid of interaction between each atom type
and the target molecule. Several recent publications73ꢀ76
have used this program to investigate problems of simi-
lar size and scope to ours, with good success.
Table 1. Relative ranking of AutoDock simulations with adenine
and HCPTP-A or HCPTP-B
HCPTP-A and Adenine, with PO4
Rank a
# in
cluster
Binding
siteb
Docked energy,
c
Ki, mM d
kcal/mol
1
2
3
4
5
6
7
8
5
16
2
1
12
7
Y131 dish
Active site
Y131 dish
Active site
Active site
Active site
Active site
binding cleft
ꢀ6.55
ꢀ5.36
ꢀ5.01
ꢀ4.94
ꢀ4.81
ꢀ4.71
ꢀ4.71
ꢀ4.09
15.9
119.
212.
238.
299.
352.
356.
997.
2
5
HCPTP-A and Adenine, without PO4
Potential energy grids were calculated for each HCPTP
isoform to represent the enzyme with or without phos-
phate in the active site. Although phosphate is a weak
inhibitor of LMW-PTPs, the yeast crystal structure with
adenine clearly indicated the presence of a phosphate
coordinated at the base of the active site, so phosphate
may play a role in modulating the orientation or activity
of adenine. We proposed that AutoDock calculations
with and without phosphate should indicate the degree
to which it affects binding of adenine. Two major bind-
ing sites for adenine were found, one in the active site
and the other in a small depression near the active site
on the opposite side of aromatic residues Tyr131 and
Tyr132 (asterisk, Fig. 1a). This depression, termed the
Y131 dish, is formed by aromatic residues and contains
a negative charge from a partially exposed Asp56 in the
HCPTP-A isoform and both a positive and a negative
charge from Asp56 and Arg58 in the HCPTP-B iso-
form. The Y131 dish can be considered a wide extension
of the specificity cleft that extends from the binding site
and may act as an initial binding location for adenine
from bulk solvent. The relative ranking of each cluster
by docked energy, position of binding for that cluster,
number of members, and estimated Ki is given in Table 1.
It is noteworthy that the lowest energy structure for
each isoform when phosphate is present was with ade-
nine bound in the Y131 dish. Of the structures that
show adenine bound in the active site, the presence of
phosphate promoted a binding in the same relative
position as that seen in the LTP1-adenine structure,
although the adenine would often be rotated around
either the N1-C2 or the C5-C6 bond. The absence of
phosphate permitted the adenine to bind deeper in the
active site with either the N9 or the amine group posi-
tioned in the location where the phosphate was found.
One other site with poor interaction energy was infre-
quently found, where the adenine binds in the specificity
cleft, well away from the active site. Taken together,
these results suggest that activity modulation of the
HCPTP isoforms by adenine may involve multiple
binding sites.
1
2
3
4
5
6
7
8
9
19
6
Active site
Active site
Active site
Active site
Active site
Active site
Active site
Active site
Y131 dish
ꢀ7.24
ꢀ7.06
ꢀ6.96
ꢀ6.88
ꢀ6.83
ꢀ6.79
ꢀ6.67
ꢀ6.64
ꢀ5.89
4.95
6.65
7.96
9.03
9.87
10.5
12.8
13.7
48.3
1
1
15
3
1
3
1
HCPTP-B and Adenine, with PO4
1
2
3
4
5
6
7
8
2
8
3
5
4
Y131 dish
Active site
Y131 dish
Y131 dish
Active site
Y131 dish
Active site
Y131 dish
Y131 dish
Y131 dish
Active site
binding cleft
ꢀ5.39
ꢀ5.26
ꢀ5.25
ꢀ5.24
ꢀ4.98
ꢀ4.92
ꢀ4.92
ꢀ4.85
ꢀ4.74
ꢀ4.61
ꢀ4.36
ꢀ3.84
113.0
139.0
142.0
143.0
223.0
246.0
249.0
277.0
338.0
414.0
642.0
N.B.e
2
21
1
9
1
1
1
1
10
11
12
HCPTP-B and Adenine, without PO4
1
2
3
4
5
6
7
27
2
Active site
Active site
Active site
Active site
Active site
Active site
Active site
ꢀ7.84
ꢀ7.68
ꢀ7.43
ꢀ7.39
ꢀ6.99
ꢀ6.96
ꢀ6.74
1.79
2.36
3.56
3.80
7.50
7.97
11.4
2
16
1
1
1
a Rank determined by final docked energy.
b Placement of cluster in the active site, the Y131 dish, or along the
binding cleft.
c Docked potential energy of the ligand, from the best complex in the
cluster.
d Estimated inhibition constant at 298.15 K.
e Not bound. The KI was calculated to be 0.0.
sume that when the phosphate is present, it is tightly
bound to the enzyme. The Ki for phosphate has been
experimentally determined to be in the low mM
range57,77 and indicates the phosphate is neither tightly
nor constantly bound. Furthermore, experimental data
has shown that while adenine inhibits the HCPTP-A
isoform, it acts as an activator for the HCPTP-B iso-
form. Presumably this activation involves a mechanism
other than simple binding in the active site of the
apoenzyme.42 Assuming that the calculated inhibition
constant, Ki, is equivalent to the dissociation constant,
KD, the results of the AutoDock calculations con-
sistently suggest a stronger binding of adenine to either
The calculated inhibition constants for the adenine
complexes are in the mM range and generally 1 to 2
orders of magnitude weaker when phosphate is present.
The calculations also predict that adenine is a somewhat
better inhibitor for HCPTP-A in the presence of phos-
phate, but is a better inhibitor for HCPTP-B when
phosphate is absent. However, these calculations pre-

1874
A. P. R. Zabell et al. / Bioorg. Med. Chem. 12 (2004) 1867–1880
enzyme by 2 or 3 orders of magnitude compared to
experimental data. The lowest energy docked structure
of adenine in the absence of phosphate was calculated
to be 4.95 mM and 1.79 mM, respectively, for HCPTP-A
and HCPTP-B. The reported dissociation constant is
2.1 mM and 0.3 mM, respectively.38 More important
than the absolute binding constants is the finding that
the relative order of dissociation remains consistent,
with adenine acting as a weaker effector of HCPTP-A.
We are therefore confident that this methodology can be
used as a rapid screening tool, indicating binding con-
stants relative to adenine, and is able to distinguish any
preference for a given isoform.
while the remaining structures rotate to position N9
near either Tyr131 or Tyr49. As we might predict, the
azaindole positioned to form the hydrogen bond to
Asp129 does exhibit the best docking energy.
These 10 structures of 2 demonstrated a much stronger
binding to the HCPTP isoforms when compared with
adenine (Table 2), with a predicted binding energy of
approximately ꢀ10 kcal/mol. The estimated inhibition
constant for the best inhibitor complex is also sig-
nificantly lower, with a Ki of 0.15 mM for HCPTP-A
and 0.015 mM for HCPTP-B. It should be recalled that
the estimated Ki for adenine is some 2 or 3 orders of
magnitude stronger than experimentally determined, so
these values may also be exaggerated.
The calculations with 2 were performed only in the
absence of phosphate, and the flexibility of the ethyl-
phosphonate chain led to a dramatic increase in the
number of independent structures. This apparent diver-
sity in structures, however, comes from a variety of
nearly identical structures where the indole is rotated
around the N1-C2 bond in a fashion similar to that seen
with the adenine complexes. The best 20–30 structures
of 2 in the HCPTP isoforms have a docking energy
within 2 kcal/mol of each other. The 10 structures with
the lowest energy showed each docked complex bound
with the phosphonate superimposed on the phosphate
position in the LTP1-adenine complex, and the azain-
dole located almost exactly in the position of the ade-
nine. Consistent with the results from the explicit MD
simulations, the predicted hydrogen bonding with the
acceptor atom N2 or donor atom N9 was obtained in a
limited number of structures. In the HCPTP-A isoform,
for example, only one structure has N9 positioned in
rough alignment to permit hydrogen bonding to Asp129
3.4. Docking of molecular probes for binding interactions
Our docking calculations indicate that 2 can be further
simplified while maintaining the key elements of our
lead concept. In order to partition out the role of the
binding interactions for individual optimization, it
would be of interest to study the interactions one at a
time. Although the smaller number of interactions in
the probes will of course lead to lower inhibition con-
stants, the information provided by this study is very
valuable for the longer-term development of more active
and selective HCPTP inhibitors. This has for example
been demonstrated by the popular ‘SAR by NMR’
technique, where binding of small molecular probes is
used to elucidate binding interactions in the active site,
which are then exploited by more complex molecules
incorporating these interactions. Four potential inhibi-
tors 3–6 were designed to probe the individual impor-
tance of those characteristics the simulations predicted
to be important for a potent inhibitor: size of the aro-
matic group, necessity for secondary hydrogen bonding
interactions, and flexibility of the phosphonate linker.
For example, the heterocycle of 2 is replaced by a
naphthalene group in 3, retaining the size and aromati-
city of the initial target while removing the hydrogen
bonding donor and acceptor. The relative importance of
the hydrogen bond donor and acceptor atoms is measured
by reducing the size of the heterocycle to a pyridine in
4 and a benzene in 5, while an amino group in the ortho
position to the phosphonate retains the proposed
Table 2. Relative ranking of AutoDock simulations with 2 HCPTP-A
and HCPTP-B a
Rank
Docked Energy,
kcal/mol
Binding Energy,
kcal/mol b
Ki, mM
HCPTP-A
1
2
3
4
5
6
7
8
9
10
ꢀ10.49
ꢀ10.45
ꢀ10.41
ꢀ10.35
ꢀ10.30
ꢀ10.04
ꢀ10.00
ꢀ10.00
ꢀ9.99
ꢀ9.3
0.153
0.128
0.154
0.194
0.193
0.313
0.294
0.357
0.242
0.393
ꢀ9.4
ꢀ9.29
ꢀ9.16
ꢀ9.16
ꢀ8.87
ꢀ8.91
ꢀ8.80
ꢀ9.03
ꢀ8.74
Table 3. Calculated and Experimental results with compounds 3–6 a
Compd Docked Energy, Binding Energy, Calculated Experimental
kcal/mol
ꢀ9.94
kcal/mol
Ki, mM
Ki, mM
HCPTP-B
HCPTP-A
1
2
3
4
5
6
7
8
9
10
ꢀ11.87
ꢀ11.78
ꢀ11.72
ꢀ11.60
ꢀ11.57
ꢀ11.52
ꢀ11.45
ꢀ11.36
ꢀ11.34
ꢀ11.21
ꢀ10.69
ꢀ10.69
ꢀ10.55
ꢀ10.40
ꢀ10.54
ꢀ10.57
ꢀ10.37
ꢀ10.19
ꢀ10.21
ꢀ10.18
0.0147
0.0146
0.0185
0.0238
0.0187
0.0178
0.0252
0.0337
0.0329
0.0346
3
4
5
6
ꢀ9.88
ꢀ10.03
ꢀ9.86
ꢀ9.28
ꢀ8.81
ꢀ8.51
ꢀ8.43
ꢀ8.13
0.346
0.581
0.662
1.10
14.0
0.30
0.25
5.30
HCPTP-B
3
4
5
6
ꢀ11.60
ꢀ11.45
ꢀ11.46
ꢀ10.61
ꢀ10.48
ꢀ9.86
ꢀ9.91
ꢀ9.55
0.0209
0.0594
0.0541
0.100
10.0
2.21
1.40
13.0
a Terms are the same as in Table 1.
b Free energy of ligand binding, which incorporates ligand flexibility.
a Terms are the same as in Table 2.

A. P. R. Zabell et al. / Bioorg. Med. Chem. 12 (2004) 1867–1880
1875
hydrogen bonding to Asp129. In 6, the ethyl group is
modified to an ethylene in order to examine the impor-
tance of flexibility in the phosphonated linker. Compu-
tational and experimental studies of 3–6 will therefore
provide insights into the effect of the different elements
of the lead concept.
The estimated docking energy, binding energy, and Ki
for the best complex of 3–6 with each HCPTP isoform
are shown in Table 3. All four compounds are predicted
to have lower binding energies and to be poorer inhibi-
tors than 2. Compound 3 should be a better inhibitor
than the systems with a single aromatic ring, with a
predicted Ki of 0.346 mM and 0.021 mM for HCPTP-A
and HCPTP-B, respectively, reemphasizing the impor-
tance of shape. The acceptor nitrogen of 4 is predicted
to have a negligible effect on inhibition, with Ki values
essentially equal to those 5 of 0.581 mM and 0.059 mM
for either isoform. A further two-fold loss of inhibition
is predicted with the loss of flexibility in the phospho-
nate chain of 6 as the ethyl group is modified to an eth-
ylene. Figure 2d shows the relative orientation of the ten
lowest energy structures of 4 bound to HCPTP-A, and
is representative of all eight enzyme-inhibitor calcula-
tions. In all cases, the ten lowest energy complexes have
the phosphonate at the base of the active site and the
ring system positioned roughly between the aromatic
residues at the mouth of the active site. As was the case
with atom N9 of the azaindole, the amino group in 4–6
is typically not aligned to interact with Asp129. Other
trends remain the same as those seen in the simulations
with 2, where the predominant difference between
docked positions involves the rotation of the ring sys-
tem relative to that seen in the crystal structure of the
yeast form with adenine.
Scheme 2. (a) Pivaloylchloride, TEA, DCM (63%); (b) BuLi, THF,
ꢀ78 ꢃC to 0 ꢃC, then DMF at ꢀ78 ꢃC, (61%); (c) tetraethyl methyl-
enediphosphonate, 50% NaOH, DCM, (95%); (d) H2, Pd/C, MeOH
(97%); (e) 6N HCl, reflux (54%).
Scheme 3. (a) NaH, THF, [t-BuOC(O)]2O, reflux (93%); (b)
Pd(PPh3)Cl2, diethyl vinylphosphate, TEA, DMF, 100 ꢃC (41%); (c)
TMSBr, CH3CN, rt (72%); (d) H2, Pd/C, MeOH (18%).
myl-4-pyridyl)-2,2-dimethylpropionamide
7
(Scheme
2).68 The Horner–Wadsworth–Emmons–Wittig reaction
of tetraethyl methylenebisphosphonate with this alde-
hyde carried out in an aqueous two-phase system affor-
ded solely the E-isomer 8.79 Hydrogenation of the
double bond over palladium on carbon provided com-
pound 9. Simultaneous hydrolysis of the amide and
phosphonate esters was achieved by refluxing for 18 h in
6 N hydrochloric acid providing 4.
3.5. Synthetic studies
Based on the results for adenine, it can be expected that
the real absolute binding constants are 2–3 orders of
magnitude lower than the ones predicted by AutoDock.
To determine the absolute and relative binding con-
stants for the molecular probes 3–6, these compounds
were synthesized and the inhibition constants for
HCPTP-A and HCPTP-B determined. 2-(1-Naphthyl)-
ethylphosphonic acid 3 was obtained in two steps as
shown in Scheme 1. The Arbusov reaction of commer-
cially available 1-(2-bromoethyl)naphthylene provides
the diethyl phosphonate ester, and subsequent depro-
tection using trimethylsilyl bromide gives the free phos-
phonic acid 3.67
The anilides 5 and 6 were synthesized from commer-
cially available 2-bromoaniline (Scheme 3). Protection
of the aniline with the t-butoxycarbonyl (Boc) group,69
followed by palladium catalyzed olefination of the aryl
bromide (Heck reaction)78 with diethyl vinylphosphate
gave the corresponding E-2-(2-aminophenyl)vinyl phos-
phonate 10. Simultaneous deprotection of the Boc
group and hydrolysis of the phosphonate esters was
accomplished with trimethylsilyl bromide. Following
each hydrolysis, the fully deprotected material was fur-
ther purified on Dowex 50 ion exchange resin giving 6 as
the ammonium salt. Subsequent hydrogenation with
palladium on carbon afforded saturated side chain
derivative 5.
Compound 4 was prepared through regiospecific ortho-
metalation of 4-(pivaloylamino)-pyridine and subsequent
reaction with dimethylformamide to provide N-(3-for-
3.5.1. Enzyme inhibition of 3–6. The results from the
AutoDock calculations with compounds 3–6 were tested
by direct comparison with experimentally determined Ki
values from a standard phosphatase assay using pNPP
Scheme 1. (a) P(OEt)3, 50 ꢃC (90%); (b) TMSBr, CH3CN, rt (30%).

1876
A. P. R. Zabell et al. / Bioorg. Med. Chem. 12 (2004) 1867–1880
as the substrate. Results from inhibitors 4–6 follow the
calculated results relatively well. Compound 5 was
found to be a slightly better overall inhibitor than 4,
with a Ki of 0.25 mM and 1.4 mM for HCPTP-A and
HCPTP-B, respectively, compared to 0.30 mM and 2.2
mM with the pyridine compound. As predicted, the
addition of a hydrogen bonding donor atom in the aro-
matic ring of 4 does not have a significant effect on
enzyme inhibition. The predicted mM to nM inhibition
of these compounds would have made them excellent
inhibitors, but the observed mM inhibition is more
consistent with results for a lead compound. As expec-
ted based on the calculated and experimental Ki for
adenine, the experimentally determined inhibition con-
stants for 3–6 were orders of magnitude poorer than
predicted by AutoDock (Table 3). Compound 4 is much
more soluble than 5, a necessary consideration for
effective drug design that AutoDock cannot take into
account. One surprise in these experiments involved
compound 6, designed with the expectation that a
rigidly positioned distance between the phosphonate
and the amine would improve the binding at both the
phosphate binding loop and the Asp129 side chain.
However, 6 was roughly 10-fold poorer than either 4 or
5, with a Ki of 5.3 and 13.0 mM for HCPTP-A and
HCPTP-B. It is gratifying to note that the relative scale
of experimental Ki values between the pyridine and ani-
lide molecules is essentially the same as the prediction,
demonstrating that the ethyl group flexibility in 4 and 5
is a critical component of inhibition. It is also interest-
ing to note that even these very simple compounds
begin to show selectivity of up to a factor of 7.5 for
differential binding to one isoform over the other. A
detailed analysis of the structural origin of this selectiv-
ity through MD simulation will help in the design of
more specific inhibitors of HCPTP.
4. Conclusions
The CHARMM molecular dynamics and AutoDock
docking simulations for 1 and 2 revealed a number of
elements necessary for the successful design of inhibitors
for the HCPTP isoforms. First, the MD shows that the
phosphate ion remains essentially unchanged in its
bound location in the active site loop, while adenine is
unable to maintain its position in the active site with
either isoform of HCPTP. By covalently linking the
features of phosphate and adenine in the azaindole
compound 2, the molecule remained bound to the
enzyme, supporting this scaffold as a basis for inhibitor
design. Second, the MD results with 2 suggested three
points of design strategy of a lead compound: a planar
shape for the inhibitor at the mouth of the active site,
indifference for hydrogen bonds between that moiety
and enzyme, and the requirement of the phosphonate to
provide the principal binding interaction between inhi-
bitor and enzyme. Third, the correlation of the results
from AutoDock and CHARMM suggested that a faster
virtual screening of potential compounds with the sim-
ple AutoDock scoring function is possible before com-
mitting to detailed analysis using extended MD
simulations. It is interesting to note that MD showed
the azaindole ring of 2 moving away from the initial
adenine position while all of the AutoDock solutions
show a greater superposition of azaindole on the ade-
nine. This discrepancy is possibly an artifact of the dif-
ferences between the explicit energy calculation and the
implicit, parameterized scoring function. The MD
simulation was performed with explicit solvent and
permitted free motion of the enzyme, allowing both the
indole and the aromatic residues around the active site
to move to favorable positions. The docking calcula-
tions, on the other hand, maintain a rigid enzyme
structure and use an energy function empirically para-
meterized for favorable binding modes.
The unexpected result of the experimental studies was
the relatively low activity of 3, with a Ki of 14.0 mM and
10.0 mM for HCPTP-A and HCPTP-B, respectively,
which is in contradiction to the docking results.
Although the relatively crude energy function of Auto-
Dock could of course be one reason, it is noteworthy
Separation of the design elements led to the even further
simplified compounds 3–6, which were studied compu-
tationally using AutoDock and experimentally through
synthesis and enzyme kinetics. Three of these four
compounds show good correlation between computed
and experimental binding values with the computed
binding energies being consistently overestimated by ꢄ4
kcal/mol, in agreement with the correlation obtained for
the well known effector adenine. Of the compounds stud-
ied, we found that compound 3 showed poor correlation,
most likely because of a different binding mode. This
demonstrates that shape complementarity and hydro-
phobic interactions alone are not sufficient for efficient
inhibition and that appropriately positioned functional
groups capable of hydrogen bonding are necessary. The
initial results for compounds 4 and 5 are also promising
in that they start to show a small specificity for HCPTP-
A, which will be further explored in future studies.
that the secondary plots of 1/Vmax,app and KM,app
/
Vmax,app indicate that 3 is a non-competitive inhibitor
(not shown). As a non-competitive inhibitor, 3 must be
able to bind to the enzyme in the presence and absence
of substrate. This implies that a secondary binding site,
which was not part of the original design strategy, can
bind 3. It is therefore an instructive example of the
breakdown of the model when alternative binding sites
interfere with inhibition. This is always a possibility that
needs to be evaluated by experimental methods,
emphasizing the strengths of the combined computa-
tional and experimental approach used here. Finally,
the results also establish the importance of hydrogen
bonding with the amino function at the mouth of the
active site. This can be accomplished by the inclusion of
the amino group in compounds 4–6, which adds a
functional group not present in compound 3, suggesting
that a minimum of three interactions are needed for
AutoDock to successfully predict the relative levels of
inhibition.
In summary, we have used computer aided rational
design to develop new and structurally simple inhibitors
of biologically relevant low molecular weight phospha-
tases. The relative importance of shape, hydrogen
bonding, and charge interactions in the active site were

A. P. R. Zabell et al. / Bioorg. Med. Chem. 12 (2004) 1867–1880
1877
elucidated and demonstrated separately in a series of
simple inhibitors. Although the binding constants of
these simplified compounds with only two or three
interactions is of course low, the thorough understanding
obtained will enable us to combine the individual design
elements into more complex, more potent inhibitors.
This process is helped by the finding that the binding
constants of the AutoDock scoring function correlates
reasonably well with the relative binding abilities of the
majority of inhibitors. These studies are currently in
progress and the results will be reported in due course.
after which a suspension formed. Water (4 mL) was
then added, and the organics were extracted with ethyl
acetate. The combined organic layers were dried over
magnesium sulfate, filtered and the solvent evaporated
under vacuum to give the title compound as a yellow oil
which slowly solidifies (785 mg, 95%) and was used
without further purification: mp: 75–78 ꢃC; 1H NMR
(DMSO) d 1.17–1.25 (15H, m, Piv+CH2CH3), 3.94–
4.05 (4H, m, J=7.4 Hz, CH2CH3), 6.55–6.72 (1H,
apparent t, J=18.0 Hz, PCH2), 7.21–7.40 (2H, m, Ar-
CH+ArH), 8.43–8.48 (1H, d, J=5.4, ArH), 8.87 (1H, s,
ArH), 9.51–9.58 (1H, bs, NH); 13C NMR (MeOH) d
15.55+15.63 (d), 26.53, 39.56, 62.49+62.57 (d),
115.34+117.85 (d), 120.67, 126.56+126.86 (d),
142.16+142.25 (d), 144.73, 148.15, 150.29, 178.88; MS
(+veFAB) (M+H) 341; HRMS calcd for
C16H25N2O4P (M+H), 341.1630, found 341.1630.
5. Experimental
All reagents were purchased from commercial suppliers
and were used without further purification. Melting
points were measured on a MEL-TEMP II device with-
out correction. 1H NMR (300 MHz) and 13C NMR
(75 MHz) were recorded on a Varian Inova-300 spec-
trometer. All spectra were recorded in CDCl3, DMSO-
d6 or CD3OD, and chemical shifts are given relative to
TMS. Mass spectrometry was performed with a JEOL
JMS-AX505HA double sector mass spectrometer. THF
was purified by an Innovative Technologies solvent
purification system. CH2Cl2 and CH3CN were both
distilled from CaH. Column chromatography was per-
formed with EM Science 60 A silica gel (230–400 mesh).
The DOWEX1 50WX4-400 ion-exchange resin was
purchased from Aldrich. The Superdex 75 prep grade
gel filtration media and HiPrep 16/10 SP cation
exchange column were purchased from Pharmacia.
5.1.3. Diethyl 2-{4-[(2,2-dimethylpropanoyl)amino]pyri-
din-3-yl}ethanephosphonate (9). A solution of 8 (600
mg, 1.8 mmol) and palladium on carbon (60 mg) in
methanol (30 mL) was stirred under an atmosphere of
hydrogen for 18 h. The solution was filtered through
Celite which was washed further with methanol, and the
solvent was evaporated under vacuum to give a grey oil
(585mg, 95%) that was used without further purifica-
tion: 1H NMR (CDCl3) d 1.15–1.22 (6H, t, J=7.1,
CH2CH3), 1.37 (9H, s, Piv), 2.03–2.15 (2H, dt, J=17.6,
7.2, CH2P), 2.85–2.99 (2H, dt, J=20.1, 7.1, ArCH2),
3.89–4.01 (4H, m, CH2CH3), 7.85–7.89 (1H, d, J=5.4,
ArH), 8.35–8.40 (2H, m, ArH), 8.82 (1H, br s, exchanges
with D2O, NH); 13C NMR (CDCl3) d 16.41, 21.67+25.60
(d), 27.46+27.63 (d), 27.80, 40.27, 62.06+62.15 (d),
118.23, 126.47+126.57 (d), 143.79, 148.85, 151.52, 178.17;
MS (+veFAB) (M+H) 343; HRMS calcd for
C16H27N2O4P (M+H) 343.1787, found 343.1777.
5.1. Chemical synthesis
5.1.1. 2-(1-Napthyl)ethanephosphonic acid (3). A mixture
of 1-(2-bromoethyl)naphthalene (900 mg, 3.8 mmol;
Aldrich) and triethyl phosphite (2.6 mL, 15 mmol) was
heated to 145–150 ꢃC for 3 h. Excess triethyl phosphite
was removed by evaporation under vacuum to give a
brown oil (1.0 g, 3.4 mmol, 90%). To this crude oil was
added anhydrous acetonitrile (30 mL) and trimethlysilyl
bromide (9.0 mL, 68 mmol), and the mixture heated to
60–65 ꢃC for 3 h. All volatiles were removed by eva-
poration under vacuum, and dichloromethane (5 mL)
was added to the resulting oil. The formed precipitate
was filtered and recrystallized from ethyl acetate to give
white crystals (271 mg, 30%): mp 165–167 ꢃC (lit mp67
5.1.4. 4-Amino-3-(2-phosphonoethyl)pyridinium chloride
(4). A mixture of 9 (300 mg, 0.88 mmol) in hydrochlo-
ric acid (4 mL, 6 N solution in water) was refluxed for
18 h after which the water was evaporated under
vacuum to give a tan colored solid. The solid was heated
in ethanol (1 mL) and the resulting suspension was
centrifuged for 10 min. Ethanol was decanted off leav-
ing a fine white powder (113 mg, 54%): 1H NMR
(DMSO) d 1.78–1.90 (2H, dt, J=18.0, 7.4, P-CH2),
2.64–2.76 (2H, m, ArCH2), 6.80–6.85 (1H, d, J=6.9,
ArH), 8.02–8.07 (2H, m, ArH); 13C NMR (CD3OD) d
21.63, 23.48+25.33 (d), 109.09, 121.08, 137.98, 138.47,
159.06; MS (+veFAB) (M+H) 203; HRMS calcd for
C7H11N2O3P (M+H) 203.0586, found 203.0587.
ꢃ
1
166–167 C); H NMR (DMSO) d 1.80–1.94 (2H, m,
CH2P), 3.15–3.25 (2H, m, ArCH2), 7.34–7.59 (4H, m,
ArH), 7.73–8.01 (3H, m, ArH); 13C NMR (DMSO) d
26.92, 28.95+30.72 (d), 123.88–131.66 (multiple peaks-
undeterminable), 134.18, 138.43+138.67 (d); MS
(+veFAB) (M+H) 237; HRMS calcd for C12H13O3P
(M+H) 237.0681, found 237.0711.
5.1.5. Diethyl (E)-2-{2-[(tert-butoxycarbonyl)amino]phe-
nyl}ethenephosphonate (10). A solution of diethyl vinyl-
phosphate (2.11 mL, 13.5 mmol), 2-bromo-N-(tert-
butoxycarbonyl)aniline (1.84 g, 6.8 mmol, prepared by
the lit.69 method), bis(triphenylphosphine)palladium II
chloride (0.487 g, 0.68 mmol), in dimethylformamide (5
mL, degassed) and triethylamine (1.93 mL, 14 mmol)
was heated to 90–100 ꢃC for 18–20 h. The solvent was
removed by evaporation under vacuum, and the residue
was taken up in ethylacetate and water. The precipitated
triphenylphosphine was removed by filtration, and the
5.1.2. Diethyl (E)-2-{4-[(2,2-dimethylpropanoyl)amino]-
pyridin-3-yl}ethenephosphonate (8). To a solution of N-
(3-formylpyridin-4-yl)-2,2-dimethylpropionamide 7 (485
mg, 2.4 mmol, prepared by the lit.68 method) and tetra-
ethyl methylenebisphosphonate (0.58 mL, 2.4 mmol) in
dichloromethane (3.5 mL) was added sodium hydroxide
solution (3.5 mL, 50% in water) and stirred for 1 min,

1878
A. P. R. Zabell et al. / Bioorg. Med. Chem. 12 (2004) 1867–1880
organic layer was separated from the filtrate solution
and extracted further with ethylacetate. The combined
organic extracts were dried over magnesium sulfate, and
the solvent was removed by evaporation under vacuum.
The residue was then subjected to bulb-to-bulb distilla-
tion at 90–100 ꢃC at 1mmHg to remove excess diethyl
vinylphosphate. The title compound was then further
purified by flash chromatography to give a brown solid
(992 mg, 41%): mp: 96–98 ꢃC; 1H NMR (CDCl3) d
1.32–1.39 (6H, t, J=7.3, CH2CH3), 1.51 (9H, s, tBu),
4.08–4.20 (4H, m, CH2CH3), 6.15–6.29 (1H, apparent t,
J=17.7 Hz, PCH), 6.50–6.64 (1H, br s, NH), 7.08–7.15
(1H, dt, J=0.6, 7.5, ArH), 7.32–7.39 (1H, dt, J=1.5,
7.8, ArH), 7.46–7.51 (1H, dd, J=1.5, 8.1, ArH), 7.58–
7.79 (2H, m, ArH+Ar-CH); 13C NMR (CDCl3) d
16.62+16.70 (d), 28.53, 26.17+26.24 (d), 81.22,
115.40+117.92 (d), 123.43, 124.66, 127.07, 131.03,
135.70, 136.59, 143.83+143.92 (d), 153.35; MS
(+veFAB) (M+H) 356; HRMS calcd for C17H26NO5P
(M+H) 356.1627, found 356.1626.
7.00 (2H, m, ArH); 13C NMR (DMSO) d 29.93,
31.71+33.47 (d), 120.35, 122.16, 130.92+131.13 (d),
132.13, 133.95, 150.86; MS (-veFAB) 200; HRMS calcd
for C8H11NO3P 200.0477, found 200.0483.
5.2. Protein expression and purification
Cloning, expression, and purification of each isoform
was performed as previously described70 with slight
modification. Briefly, the pET-11d vector containing the
wild-type sequence for either HCPTP-A or HCPTP-B
was transformed into fresh E. coli BL21(DE3) cells.
Cells were plated, selected, and grown in M9ZB media
at 37 ꢃC in the presence of 100 mg/mL ampicillin. When
the culture reached an optical density of 0.8–1 at 600
nm, cells were induced for 4 h with the addition of
0.4 mM isopropyl-1-thio-b-d-galactopyranoside. Cells
were harvested by centrifugation at 3,000ꢁg for 15 min
and stored at ꢀ80 ꢃC until purification.
The cell pellet was resuspended in the cation exchange
loading buffer consisting of 10 mM sodium acetate, pH
4.8, 10 mM sodium phosphate, 50 mM NaCl and 1 mM
EDTA. Cells were lysed using a French press and the
lysate centrifuged at 27,000ꢁg for 30 min. The super-
natant was then loaded directly onto a HiPrep 16/10 SP
cation exchange column. Elution was done with a buffer
of 300 mM sodium phosphate, pH 5.1, and 1 mM
EDTA. Fractions with phosphatase activity were con-
centrated and loaded onto a Superdex 75 gel filtration
column using the cation exchange loading buffer.
5.1.6. 3(E)-2 - (2 - Aminophenyl)ethenephosphonic acid
monoammonium salt (6). To 10 (500 mg, 0.14 mmol) in
acetonitrile (35 mL) was added trimethylsilyl bromide
(5.5 mL, 42 mmol) and the mixture was stirred for 15 h.
After quenching the reaction with water (5mL), the sol-
vent was removed by evaporation under vacuum. The
resulting residue was made slightly acidic (pH 5-6) and
purified on ion exchange resin (DOWEX1 50WX4-
400), eluting first with water and then ammonium
hydroxide solution (5% in water). All fractions that
turned positive to ninhydrin solution were combined
and water evaporated under vacuum to afford the title
5.2.1. Enzyme activity assay. All enzymatic assays were
performed at 37 ꢃC in 100 mM sodium acetate, pH 5.0,
adjusted to an ionic strength of 150 mM with NaCl, as
described previously,71 and the program DNRPEasy72
was used to perform curve fitting. Activity was monitored
with the substrate analogue p-nitrophenyl phosphate
(pNPP), with production of p-nitrophenol measured at
405nm via continuous or discrete assay. In either case,
enzyme concentration did not exceed 0.1% of substrate.
The discrete assay was quenched with 1N NaOH after a
4 min reaction and used an extinction coefficient of
18,000 Mꢀ1cmꢀ1, with fitting of the standard Michaelis-
Menten equation to provide Vmax and KM. The con-
tinuous assay is not quenched and is consequently
monitored at a pH roughly two units below the pKa of
the product. As such, a standard curve was performed
to determine the extinction coefficient of 185 Mꢀ1 cmꢀ1
at pH 5. The Michaelis-Menten parameters were fit to a
four variable equation in the continuous assay. Enzyme
inhibition was measured at inhibitor concentrations
from 0.5 to 5 mM. Compounds 5 and 6 have limited
solubility, so the stock solution was prepared as a satu-
rated solution then vortexed to a uniform suspension.
The dilutions in the kinetic assay provided homogeneous
solutions in all cases and gave reproducible results.
1
compound as a tan colored solid (219 mg, 72%): H
NMR (DMSO) d 6.08–6.22 (1H, dd, J=16.2, 15.3,
PCH), 6.45–6.52 (1H, t, J=7.5, ArH), 6.60–6.65 (1H, d,
J=8.4, ArH), 6.90–6.96 (1H, t, J=7.7, ArH), 7.03–7.17
(1H, dd, J=18.8, 20.1, ArCH), 7.17–7.22 (1H, d,
J=7.5, ArH); 13C NMR (DMSO) d 118.06, 120.44,
121.40+123.77 (d), 122.99, 127.23, 130.47, 137.41 (d,
ArCH), 143.88; MS (-veFAB) (M-H) HRMS calcd for
C8H9NO3P 198.0320, found 198.0316.
5.1.7. 2-(Aminophenyl)ethane phosphonic acid monoam-
monium salt (5). To 6 (200 mg, 0.93 mmol) in methanol
(10 mL) was added palladium on carbon (20 mg, 5% on
carbon). The atmosphere was purged with hydrogen,
and the solution was stirred overnight. After filtration
of the mixture through Celite, the solvent was removed
by evaporation under vacuum to give a white solid (205
mg). The solid was taken up in hot methanol and fil-
tered through a sintered glass funnel which was washed
further with hot methanol. Solvent was evaporated
under vacuum from the filtrate, and the residue was
further purified using ion exchange resin (DOWEX1
50WX4-400) eluting with water and then ammonium
hydroxide solution (5% sol in water). All ninhydrin
positive fractions were combined, and the water was
removed by evaporation to give a gray solid. The solid
was finally recrystallized from methanol to give a tan
solid (36 mg, 18%): 1H NMR (DMSO) d 1.68–1.89 (2H,
m, PCH2), 2.74–2.72 (2H, m, ArCH2), 6.50–6.62 (1H, d,
J=7.2 Hz, ArH), 6.59–6.66 (1H, d, J=8.1, ArH), 8.86–
Acknowledgements
The authors gratefully acknowledge the support of the
National Cancer Institute (CA82673 to CVS) and col-

A. P. R. Zabell et al. / Bioorg. Med. Chem. 12 (2004) 1867–1880
1879
laboration with the Walther Cancer Institutes at Purdue
University and the University of Notre Dame. We
thank Dr. Isamu Katsuyama for preliminary studies on
the synthesis of PTP inhibitors. Computing resources
were provided by the National Science Foundation
(DMR0079647 to OW) and the Office of Information
Technology at the University of Notre Dame.
29. Chernoff, J.; Schievella, A. R.; Jost, C. A.; Erikson, R. L.;
Neel, B. G. Proc. Natl. Acad. Sci. U.S.A. 1990, 87, 2735.
30. Lawrence, G. L.; Van Etten, R. L. Arch. Biochem.
Biophys. 1981, 206, 122.
31. Zhang, Z. Y.; Van Etten, R. L. Arch. Biochem. Biophys.
1990, 282, 39.
32. Modesti, A.; Cirri, P.; Raugei, G.; Carraresi, L.; Magherini,
F.; Manao, G.; Camici, G.; Ramponi, G. FEBS Lett.
1995, 375, 235.
33. Mondesert, O.; Moreno, S.; Russell, P. J. Biol. Chem.
1994, 269, 27996.
References and notes
34. Ostanin, K.; Pokalsky, C.; Wang, S.; Van Etten, R. L. J.
Biol. Chem. 1995, 270, 18491.
35. Park, E. K.; Warner, N.; Mood, K.; Pawson, T.; Daar,
I. O. Mol. Cell. Biol. 2002, 22, 3404.
1. Stauffacher, C. V.; Charbonneau, H. In Principles of
Molecular Regulation; Conn, P. M., Means, A. R. (Eds.).
Humana Press: Totowa, N.J., 2000, p 323.
2. Tonks, N. K.; Neel, B. G. Curr. Opin. Cell Biol. 2001, 13,
182.
36. Miller, D. T.; Read, R.; Rusconi, J.; Cagan, R. L. Gene
2000, 243, 1.
3. Li, L. W.; Dixon, J. E. Semin. Immunol. 2000, 12, 75.
4. Hunter, T. Cell 2000, 100, 113.
37. Raugei, G.; Ramponi, G.; Chiarugi, P. Cell. Mol. Life Sci.
2002, 59, 941.
5. Li, J.; Yen, C.; Liaw, D.; Podsypanina, K.; Bose, S.;
Wang, S. I.; Puc, J.; Miliaresis, C.; Rodgers, L.; McCombie,
R.; Bigner, S. H.; Giovanella, B. C.; Ittmann, M.; Tycko,
B.; Hibshoosh, H.; Wigler, M. H.; Parsons, R. Science
1997, 275, 1943.
38. Dissing, J.; Rangaard, B.; Christensen, U. Biochim. Bio-
phys. Acta 1993, 1162, 275.
39. Zhang, M.; Van Etten, R. L.; Stauffacher, C. V. Bio-
chemistry 1994, 33, 11097.
40. Wang, S.; Tabernero, L.; Zhang, M.; Harms, E.; Van
Etten, R. L.; Stauffacher, C. V. Biochemistry 2000, 39,
1903.
6. Kikawa, K. D.; Vidale, D. R.; Van Etten, R. L.; Kinch,
M. S. J. Biol. Chem. 2002, 277, 39274.
7. Wang, Q.; Holmes, D. I. R.; Powell, S. M.; Lu, Q. L.;
Waxman, J. Cancer Lett. 2002, 175, 63.
41. Zhang, M.; Stauffacher, C. V.; Lin, D.; Van Etten, R. L.
J. Biol. Chem. 1998, 273, 21714.
8. Cobb, B. S.; Parsons, J. T. Oncogene 1993, 8, 2897.
9. Burke, T. R.; Zhang, Z. Y. Biopolymers 1998, 47, 225.
10. Murthy, V. S.; Kulkarni, V. M. Bioorg. Med. Chem. 2002,
10, 897.
42. Wang, S.; Stauffacher, C. V.; Van Etten, R. L. Biochem-
istry 2000, 39, 1234.
43. Tanizaki, M. M.; Bittencourt, H. M.; Chaimovich, H.
Biochim. Biophys. Acta 1977, 485, 116.
11. Fu, H.; Park, J.; Pei, D. Biochemistry 2002, 41, 10700.
12. Shen, K.; Keng, Y. F.; Wu, L.; Guo, X. L.; Lawrence,
D. S.; Zhang, Z. Y. J. Biol. Chem. 2001, 276, 47311.
13. Iversen, L. F.; Andersen, H. S.; Branner, S.; Mortensen,
S. B.; Peters, G. H.; Norris, K.; Olsen, O. H.; Jeppesen,
C. B.; Lundt, B. F.; Ripka, W.; Moller, K. B.; Moller,
N. P. H. J. Biol. Chem. 2000, 275, 10300.
14. Ducruet, A. P.; Rice, R. L.; Tamura, K.; Yokokawa, F.;
Yokokawa, S.; Wipf, P.; Lazo, J. S. Bioorg. Med. Chem.
2000, 8, 1451.
44. Zhang, Z. Y.; Van Etten, R. L. J. Biol. Chem. 1991, 266,
1516.
45. Nicholls, A.; Sharp, K.; Honig, B. Proteins 1991, 11, 281.
46. Kraulis, P. J. J. Appl. Crystallogr. 1991, 24, 946.
47. Merritt, E. A.; Bacon, D. J. In Macromolecular Crystal-
lography, Pt B, 1997; Vol. 277, p 505.
48. Barton, G. J. Protein Eng. 1993, 6, 37.
49. Sippl, W. Bioorg. Med. Chem. 2002, 10, 3741.
50. Walters, W. P.; Murcko, M. A. Adv. Drug Deliv. Rev.
2002, 54, 255.
15. Peng, H. R.; Xie, W. G.; Otterness, D. M.; Cogswell, J. P.;
McConnell, R. T.; Carter, H. L.; Powis, G.; Abraham,
R. T.; Zalkow, L. H. J. Med. Chem. 2001, 44, 834.
16. Zhang, Z.-Y. Annu. Rev. Pharmacol. Toxicol. 2002, 42, 209.
51. Abagyan, R.; Totrov, M. Curr. Opin. Chem. Biol. 2001, 5,
375.
52. Kuntz, I. D. Science 1992, 257, 1078.
53. Taylor, R. D.; Jewsbury, P. J.; Essex, J. W. J. Comput.
Aid. Mol. Des 2002, 16, 151.
54. Brooks, B. R.; Bruccoleri, R. E.; Olafson, B. D.; States,
D. J.; Swaminathan, S.; Karplus, M. J. Comp. Chem.
1983, 4, 187.
55. MacKerell, A. D.; Bashford, D.; Bellott, M.; Dunbrack,
R. L.; Evanseck, J. D.; Field, M. J.; Fischer, S.; Gao, J.;
Guo, H.; Ha, S.; Joseph-McCarthy, D.; Kuchnir, L.;
Kuczera, K.; Lau, F. T. K.; Mattos, C.; Michnick, S.;
Ngo, T.; Nguyen, D. T.; Prodhom, B.; Reiher, W. E.;
Roux, B.; Schlenkrich, M.; Smith, J. C.; Stote, R.; Straub,
J.; Watanabe, M.; Wiorkiewicz-Kuczera, J.; Yin, D.;
Karplus, M. J. Phys. Chem. B 1998, 102, 3586.
56. Morris, G. M.; Goodsell, D. S.; Halliday, R. S.; Huey, R.;
Hart, W. E.; Belew, R. K.; Olson, A. J. J. Comp. Chem.
1998, 19, 1639.
57. Zhang, M.; Zhou, M.; Van Etten, R. L.; Stauffacher, C. V.
Biochemistry 1997, 36, 15.
58. 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.
17. Holmberg, J.; Frisen, J. Trends Neurosci. 2002, 25, 239.
´
18. Knoll, B.; Drescher, U. Trends Neurosci. 2002, 25, 145.
¨
19. Dodelet, V. C.; Pasquale, E. B. Oncogene 2000, 19, 5614.
20. Zantek, N. D.; Azimi, M.; Fedor-Chaiken, M.; Wang, B.;
Brackenbury, R.; Kinch, M. S. Cell Growth & Differ.
1999, 10, 629.
21. Zelinski, D. P.; Zantek, N. D.; Stewart, J. C.; Irizarry,
A. R.; Kinch, M. S. Cancer Res. 2001, 61, 2301.
22. Stein, E.; Lane, A. A.; Cerretti, D. P.; Schoecklmann,
H. O.; Schroff, A. D.; Van Etten, R. L.; Daniel, T. O.
Genes Dev. 1998, 12, 667.
23. Heinrikson, R. L. J. Biol. Chem. 1969, 244, 299.
24. Thomas, C. L.; McKinnon, E.; Granger, B. L.; Harms,
E.; Van Etten, R. L. Biochemistry 2002, 41, 15601.
25. Taga, E. M.; Van Etten, R. L. Arch. Biochem. Biophys.
1982, 214, 505.
26. Waheed, A.; Laidler, P. M.; Wo, Y. Y. P.; Van Etten,
R. L. Biochemistry 1988, 27, 4265.
27. Manao, G.; Pazzagli, L.; Cirri, P.; Caselli, A.; Camici, G.;
Cappugi, G.; Saeed, A.; Ramponi, G. J. Protein Chem.
1992, 11, 333.
59. Sali, A.; Blundell, T. L. J. Mol. Biol. 1993, 234, 779.
60. Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G.
E.; Robb, M. A.; Cheeseman, J. R.; Zakrzewski, V. G.;
28. Okada, M.; Owada, K.; Nakagawa, H. Biochem. J. 1986,
239, 155.

1880
A. P. R. Zabell et al. / Bioorg. Med. Chem. 12 (2004) 1867–1880
Montgomery, J., J. A.; Stratmann, R. E.; Burant, J. C.;
68. Janiak, C.; Deblon, S.; Uehlin, L. Synthesis 1999, 6, 959.
69. Lamba, J. J. S.; Tour, J. M. J. Am. Chem. Soc. 1994, 116,
11723.
70. Wo, Y.-Y. P.; McCormack, A. L.; Shabanowitz, J.; Hunt,
D. F.; Davis, J. P.; Mitchell, G. L.; Van Etten, R. L. J.
Biol. Chem. 1992, 267, 10856.
Dapprich, S.; Millam, J. M.; Daniels, A. D.; Kudin, K.
N.; Strain, M. C.; Farkas, O.; Tomasi, J.; Barone, V.;
Cossi, M.; Cammi, R.; Mennucci, B.; Pomelli, C.; Adamo,
C.; Clifford, S.; Ochterski, J.; Petersson, G. A.; Ayala,
P. Y.; Cui, Q.; Morokuma, K.; Malick, D. K.; Rabuck,
A. D.; Raghavachari, K.; Stefanov, B. B.; Liu, G.;
Liashenko, A.; Piskorz, P.; Komaromi, I.; Gomperts, R.;
Martin, R. L.; Fox, D. J.; Keith, T.; Al-Laham, M. A.;
Peng, C. Y.; Nanayakkara, A.; Challacombe, M.; Gill,
P. M. W.; Johnson, B.; Chen, W.; Wong, M. W.; Andres,
J. L.; Gonzalez, C.; Head-Gordon, M.; Replogle, E. S.;
Pople, J. A.; A.9 ed.; Gaussian, Inc.: Pittsburgh, PA,
1998.
71. Davis, J. P.; Zhou, M.-M.; Van Etten, R. L. Biochemistry
1994, 33, 1278.
72. Duggleby, R. G. Comput. Biol. Med. 1984, 14, 447.
73. Liu, H.; Huang, X. Q.; Shen, J. H.; Luo, X. M.; Li, M. H.;
Xiong, B.; Chen, G.; Shen, J. K.; Yang, Y. M.; Jiang,
H. L.; Chen, K. X. J. Med. Chem. 2002, 45, 4816.
74. Rockey, W. M.; Elcock, A. H. Proteins 2002, 48, 664.
75. Primozic, I.; Hrenar, T.; Tomic, S.; Meic, Z. J. Phys. Org.
Chem. 2002, 15, 608.
76. Buzko, O. V.; Bishop, A. C.; Shokat, K. M. J. Comput.
Aid. Mol. Des 2002, 16, 113.
77. Evans, B. In Department of Chemistry; Purdue University:
West Lafayette, 1997, p 218.
61. Dillet, V.; Van Etten, R. L.; Bashford, D. J. Phys. Chem.
B 2000, 104, 11321.
62. Zhou, M. M.; Davis, J. P.; Van Etten, R. L. Biochemistry
1993, 32, 8479.
63. Tishmack, P. A.; Bashford, D.; Harms, E.; Van Etten,
R. L. Biochemistry 1997, 36, 11984.
64. Verlet, L. Phys. Rev. 1967, 159, 98.
78. Cristau, H. J.; Pirat, J. L.; Drag, M.; Kafarski, P. Tetra-
hedron Lett. 2000, 41, 9781.
65. Solis, F. J.; Wets, R. J.-B. Math. Oper. Res. 1981, 6, 19.
66. Gasteiger, J.; Marsili, M. Tetrahedron 1980, 36, 3219.
67. Robinson, C. N.; Pettit, W. A.; Walker, T. O.; Shearon,
E.; Mokashi, A. M. J. Heterocyl. Chem. 1972, 9, 735.
79. Bigge, C. F.; Johnson, G.; Ortwine, D. F.; Drummond,
J. T.; Retz, D. M.; Brahce, L. J.; Coughenour, L. L.;
Marcoux, F. W.; Probert, A. W. J. Med. Chem. 1992, 35,
1371.