Inhibition of geranylgeranyl diphosphate synthase by bisphosphonates and diphosphates: A potential route to new bone antiresorption and antiparasitic agents

Article abstract of DOI:10.1021/jm010412y

We report the inhibition of a human recombinant geranylgeranyl diphosphate synthase (GGPPSase) by 23 bisphosphonates and six azaprenyl diphosphates. The IC₅₀ values range from 140 nM to 690 μM. None of the nitrogen-containing bisphosphonates that inhibit farnesyl diphosphate synthase were effective in inhibiting the GGPPSase enzyme. Using threedimensional quantitative structure-activity relationship/comparative molecular field analysis (CoMFA) methods, we find a good correlation between experimental and predicted activity: R² = 0.938, R_cv² = 0.900, R_bs² = 0.938, and F-test = 86.8. To test the predictive utility of the CoMFA approach, we used three training sets of 25 compounds each to generate models to predict three test sets of three compounds. The rms pIC₅₀ error for the nine predictions was 0.39. We also investigated the pharmacophore of these GGPPSase inhibitors using the Catalyst method. The results demonstrated that Catalyst predicted the pIC₅₀ values for the nine test set compounds with an rms error of 0.28 (R² between experimental and predicted activity of 0.948).

Full text of DOI:10.1021/jm010412y

J . Med. Chem. 2002, 45, 2185-2196
2185
In h ibition of Ger a n ylger a n yl Dip h osp h a te Syn th a se by Bisp h osp h on a tes a n d
Dip h osp h a tes: A P oten tia l Rou te to New Bon e An tir esor p tion a n d An tip a r a sitic
Agen ts
Christina M. Szabo,† Yoshihiro Matsumura,‡ Sayaka Fukura,‡ Michael B. Martin,† J ohn M. Sanders,†
Suraj Sengupta,† J ohn A. Cieslak,† Timothy C. Loftus,† Christopher R. Lea,† Hyung-J ae Lee,† Ali Koohang,†
Robert M. Coates,† Hiroshi Sagami,‡ and Eric Oldfield*,†
Department of Chemistry, University of Illinois at Urbana-Champaign, 600 South Mathews Avenue, Urbana, Illinois 61801,
and Institute of Multidisciplinary Research for Advanced Materials, Tohoku University, Katahira 2-1-1, Aoba-ku,
Sendai 980-8577, J apan
Received J anuary 1, 2002
We report the inhibition of a human recombinant geranylgeranyl diphosphate synthase
(GGPPSase) by 23 bisphosphonates and six azaprenyl diphosphates. The IC50 values range
from 140 nM to 690 µM. None of the nitrogen-containing bisphosphonates that inhibit farnesyl
diphosphate synthase were effective in inhibiting the GGPPSase enzyme. Using three-
dimensional quantitative structure-activity relationship/comparative molecular field analysis
(CoMFA) methods, we find a good correlation between experimental and predicted activity:
2
2
R2 ) 0.938, Rcv ) 0.900, Rbs ) 0.938, and F-test ) 86.8. To test the predictive utility of the
CoMFA approach, we used three training sets of 25 compounds each to generate models to
predict three test sets of three compounds. The rms pIC50 error for the nine predictions was
0.39. We also investigated the pharmacophore of these GGPPSase inhibitors using the Catalyst
method. The results demonstrated that Catalyst predicted the pIC50 values for the nine test
set compounds with an rms error of 0.28 (R2 between experimental and predicted activity of
0.948).
In tr od u ction
prene (terpene) biosynthesis inhibitors act further down-
stream of FPPSase and directly block sterol formation.
Examples are squalene synthase inhibitors, squalene
monooxygenase inhibitors (e.g., terbinafine), and the
azole inhibitors (e.g., fluconazole, a cytochrome P450 14-R
demethylase inhibitor). Compounds such as terbinafine
and the azoles are important antifungal agents, which
inhibit the formation of the essential sterol, ergosterol.
Moreover, they have activity as antiparasitic agents
since in parasitic protozoa, such as Trypanosoma cruzi
and Leishmania spp., ergosterol is required for cell
growth.
The isoprenoid biosynthetic pathway is a key target
for chemotherapeutic intervention in a number of
diseases. For example, the statins are an important
class of inhibitors of the enzyme hydroxymethylglutaryl
coenzyme A reductase (HMGCoA-R) and drugs such as
Lipitor, Zocor, and Pravachol are used extensively as
cholesterol-lowering agents.1 A second class of inhibitors
consists of the nitrogen-containing bisphosphonates,
such as pamidronate (Aredia, 36), alendronate (Fosa-
max, 35), and risedronate (Actonel, 27), used in treating
bone resorption diseases. These compounds are thought
to act as aza-carbocation reactive intermediate ana-
logues of prenyl carbocation intermediates,2 and they
have recently been shown to inhibit farnesyl diphos-
phate synthase (FPPSase).3-7 A reduction in farnesyl
diphosphate (FPP) levels results in decreases in gera-
nylgeranyl diphosphate (GGPP) concentrations since
GGPP is formed from FPP by geranylgeranyl diphos-
phate synthase (GGPPSase). This effectively blocks
protein geranylgeranylation and appears to lead to
apoptotic programmed cell death (PCD) at high concen-
trations8 and to ultrastructural changes at lower con-
centrations.9 The IC50 (Ki) values for some of the
bisphosphonates, such as pamidronate, are quite high
(∼0.5 µM), but elevated local levels in the target cells,
osteoclasts, are thought to be present due to the specific
adsorption of the bisphosphonates onto the calcium
hydroxyapatite surfaces of bone. A third class of iso-
Our interest in developing novel approaches to treat
diseases caused by parasitic protozoa recently led to
the discovery that bisphosphonates of the type used to
treat bone resorption diseases are in vitro inhibitors
of the growth of T. cruzi, Trypanosoma brucei rhod-
esiense, Leishmania donovani, Toxoplasma gondii, Plas-
modium falciparum, and Cryptosporidium parvum, the
causative agents of Chagas’ disease, human East Afri-
can trypanosomiasis, visceral leishmaniasis, toxoplas-
mosis, malaria, and cryptosporidiosis.10,11 We also found
that the bisphosphonate, risedronate, effects an in
vivo cure of L. donovani infection in a BALB/c mouse
model12 and that pamidronate effects a parasitological
cure of cutaneous leishmaniasis, again in a BALB/c
mouse model.13 These bisphosphonates are, however,
only poorly absorbed orally (a few percent). Oral ad-
ministration is clearly a desirable feature of any drug
for use in less-developed nations, and indeed even in
developed areas, drugs that are more orally available
would be very desirable for treating bone resorption
diseases.
* To whom correspondence should be addressed. Tel.: (217)333-
3374. Fax: (217)244-0997. E-mail: eo@chad.scs.uiuc.edu.
† University of Illinois at Urbana-Champaign.
‡ Tohoku University.
10.1021/jm010412y CCC: $22.00 © 2002 American Chemical Society
Published on Web 04/20/2002

2186 J ournal of Medicinal Chemistry, 2002, Vol. 45, No. 11
Szabo et al.
F igu r e 1. Isoprene chain extension reaction catalyzed by geranylgeranyl PP synthase: Ionization of farnesyl PP generates the
first of two carbocation intermediates. Condensation with isopentenyl PP gives rise to a second carbocation that undergoes proton
elimination to form geranylgeranyl PP. FPP can also be produced by the enzyme from GPP and IPP.
Because protein geranylgeranylation is of key impor-
tance in bone resorption,14 we reasoned that a more
direct approach to developing new chemotherapeutic
agents would be to investigate GGPPSase inhibitors. In
this context, it was reported some time ago that native
GGPPSases from rat liver and in rat brain homogenates
are inhibited by the aza-carbocation analogue 3-aza-2,3-
dihydrogeranylgeranyl diphosphate (3-azaGGPP, 5).15
This inhibitor has a considerably longer carbon chain
than the conventional nitrogen-containing bisphospho-
nates currently in use in bone resorption therapy, which
also have substantial activity against a broad spectrum
of parasitic protozoa, so the activity of this compound
suggested that GGPPSase might represent a useful
new target. In addition, the more lipophilic nature of
typical GGPPSase inhibitors such as 3-azaGGPP (5)
might have the added advantage of improved absorption
relative to nitrogen-containing bisphosphonates in cur-
rent use.
Syn th etic Meth od s
Procedures developed previously for a six step syn-
thesis of 3-azaGGPP (5) from farnesol15 were adapted
to the preparation of the shorter chain polyene amino
diphosphates, 3-azaFPP (6) and 3-azaGPP (7), as shown
in eq 1. The principal difference was the use of N-
methyl-N′-tert-butylformamidine in place of the corre-
sponding N′-cyclohexyl derivative in the nucleophilic
aminomethylation step by Meyers’ lithioformamidine
method.19 This change was made to facilitate separation
of the relatively volatile amine 2c from the primary
amine byproduct generated concurrently in the alkaline
hydrolysis of the formamidine intermediate. N-Alkyla-
tion of secondary amines 2b,c with tert-butyl bromoac-
etate (THF, Et3N), LiAlH4 reductions (ether, 25 °C),
conversion to the methanesulfonates (MsCl, CH2Cl2,
Et3N, -15 °C), and displacements with tetrabutylam-
monium pyrophosphate (2.5 equiv, CH3CN, 25 °C) were
conducted according to the published procedures.15,20
The Tohoku group recently reported the cloning,
expression, and partial characterization of a human
GGPPSase, which catalyzes the condensation of geranyl
diphosphate (GPP) or FPP with isopentenyl diphosphate
(IPP) to give the C20 polyprenyl PP homologue.16 Al-
though FPPSase and GGPPSase have only a 16%
sequence identity, the five conserved amino acid motifs
characteristic of trans-prenyl transferases are present
in both proteins. Consequently, it appears reasonable
to suppose that the mechanisms of the reactions they
catalyze are similar and proceed through analogous
carbocation+/OPP- ion pairs, as illustrated in Figure
1.17 Therefore, it seemed likely that bisphosphonates
and related compounds that mimic these or related
carbocation intermediates might have similar inhibitory
effects on GGPPSase.
In this article, we report the first results on the
inhibition of a human recombinant GGPPSase by bis-
phosphonates and azaprenyl diphosphates. The IC50
values range from 140 nM to ∼690 µM. We also report
the results of a three-dimensional quantitative struc-
ture-activity relationship (3D-QSAR)/comparative mo-
lecular field analysis (CoMFA) of the inhibitory activity
of these compounds, together with predictions of the
inhibitory activities of three sets of three compound test
sets against GGPPSase using 3D-QSAR/CoMFA and
Catalyst18 pharmacophore modeling approaches.
After ion exchange, the hygroscopic ammonium salts,
6 and 7, were freed from contaminating diphosphate by
selective precipitations of the inorganic salt in am-
monium bicarbonate buffer. The purities with respect
to contamination by inorganic diphosphate and mono-
phosphate as well as other phosphorus-containing im-
purities were estimated by integration of 31P nuclear
magnetic resonance (NMR) spectra. The homologous
azaprenyl diphosphates, 3-azahomoGGPP (11) and 3-aza-
homoFPP (12), were prepared in similar fashion from
amines 2a ,b by N-alkylation with tert-butyl acrylate to

Bone Antiresorption and Antiparasitic Agents
J ournal of Medicinal Chemistry, 2002, Vol. 45, No. 11 2187
F igu r e 2. Structures of the six azaprenyl diphosphates investigated.
F igu r e 3. Structures of the 23 bisphosphonates investigated.
the corresponding â-amino propionates 8 and 9 (MeOH,
reflux; 70-72%), hydride reductions to tertiary amino
alcohols 10a ,b (82-91%), and conversion to the diphos-
phate monoesters by SN2 displacements of the derived
mesylates and/or chlorides (80-81%) (eq 2).
Resu lts a n d Discu ssion
We investigated the activities of the 29 compounds
shown in Figures 2 and 3 in inhibiting the formation of
[14C]GGPP from either GPP or FPP, using [14C]IPP
substrate. To confirm the formation of radioactive
products, one-dimensional thin-layer chromatography
(TLC), with autoradiographic detection of [14C]FPP/
[14C]GGPP, was used and representative results are
shown in Figure 4. The inhibitor concentrations re-
quired for 50% enzyme inhibition (the IC50 values) were
obtained from a graphical fit of the data using a
rectangular hyperbolic function. Representative results
are shown in Figure 5. The IC50 values so obtained are
compiled in Table 1 together with the trivial names of
several compounds and the code numbers used by the
Procter and Gamble Company in their publications.23
The compound numbers in the text refer to Figures 2
and 3.
The range of activity of these compounds is quite
large, 140 nM to ∼690 µM, with the most active com-
pounds having similar IC50 values to FPPSase inhibitors
such as pamidronate and alendronate.24,25 However, the
N-containing bisphosphonates have almost no activity
against GGPPSase, Table 2. It is thus of interest to
develop more QSARs for GGPPSase (and FPPSase)
inhibition to help clarify the origin of these rather large
differences (e.g., risedronate, 27: IC50 ∼10 nM, FPP-
Sase; ∼350 µM, GGPPSase) and potentially to develop
more effective drugs.
Most of the 1-hydroxymethane-1,1-bisphosphonates
shown in Figure 3 were prepared previously by conden-
sation of the corresponding carboxylic acids with phos-
phorous acid and PCl3.10,21 Ibandronate (32) was ob-
tained for the present study by phosphonylation of
N-methyl-N-pentyl-3-aminopropionic acid according to
this method. The aminomethanebisphosphonates were
synthesized from the corresponding primary amines by
condensation with ethyl orthoformate and diethyl phos-
phite followed by acid hydrolysis.10,22

2188 J ournal of Medicinal Chemistry, 2002, Vol. 45, No. 11
Szabo et al.
F igu r e 4. (A) Schematic flowchart for assay of GGPPSase inhibition. (B-F) TLC results for compounds 5, 30, 15, 19, and 32,
respectively, for [14C]IPP incorporation into FPP and GGPP with GPP substrate (left half of TLC lanes) and [14C]IPP incorporation
into GGPP with FPP substrate (right half of TLC lanes). Inhibitor levels used are as indicated.
To begin this process, we first carried out a 3D-QSAR/
CoMFA investigation of SARs for the 23 bisphospho-
nates together with the 5 azaprenyl diphosphates that
had measurable activities (Table 1, Figures 2 and 3) to
try to deduce which structural factors are most impor-
tant in GGPPSase inhibition. Each of the bisphospho-
nate compounds was aligned to the most active bispho-
sphonate (16) acting as a template by performing an
rms fitting of the pharmacophoric atoms of each con-
former to those of the template by using the shape
reference alignment function of the QSAR module of
Cerius2 4.6.26 The alignments of each of the molecular
mechanics geometry-optimized structures, obtained
through pairwise superpositioning using the maximum
common subgroup (MCSG) method, placed all 23 bis-
phosphonates in the same reference frame as the shape
reference. We used the geometry of GPP bound to the
active site of FPPSase27 as a model with which to build
the energy-minimized structure of 3-azaGGPP (5). The
structures of 3-azaFPP (6), 3-azaGPP (7), 3-azahomoG-

Bone Antiresorption and Antiparasitic Agents
J ournal of Medicinal Chemistry, 2002, Vol. 45, No. 11 2189
F igu r e 5. GGPPSase enzyme inhibition by (A) 5; (B) 12; (C) 15; (D) 14; (E) 17; and (F) 19 together with least-squares fits to the
equation I ) (ImaxC)/(IC50 + C) where I is the inhibition fraction, Imax ) 1, C is the concentration of the inhibitor (µM), and IC50
is the concentration for 50% inhibition (µM).
GPP (11), and 3-azahomoFPP (12) were all then built
by suitable modifications of 5. All of the azaprenyl
compounds were aligned to the bisphosphonates by
performing a least-squares fit in Cerius2 between four
atom centers. Specifically, the two phosphorus atoms,
the oxygen adjacent to the prenyl side chain, and C-1
in the azaprenyl diphosphates were aligned to the
corresponding phosphorus atoms, C-2, and C-3 in com-
pound 16. All ring nitrogens in the aromatic species
were assumed to be protonated, as were the amino-alkyl
side chains in the nonaromatic bisphosphonates (such
as ibandronate, 32), while the phosphonate groups each
had a -1 charge. The diphosphate charges were taken
to be -2. In the case of the protonated form of 32, the
molecule is chiral: the enantiomer that best fit the
overall alignment, (S)-ibandronate, was selected for
incorporation into the QSAR model. The S stereroiso-
mers of the protonated forms of 5-7 and 33 and the R
stereoisomers of the protonated forms of 11 and 12 were
also used. The results of the alignment are shown in
Figure 6, which includes the superimposition of all
training set molecules in order to illustrate the overall
alignment of the conformers used. Charges were calcu-
lated by using the Gasteiger-Marsili method.28 We then
used 3D-QSAR/CoMFA methodologies, as embodied in
the Cerius2 4.6 suite of programs, to analyze the
inhibition data shown in Table 1. We performed a
regression analysis of the inhibition results using the
relationship pIC50 ) -log (IC50, M) and the genetic
function approximation (GFA) algorithm29 to obtain the
following QSAR equation
pIC50 ) 3.55 + 0.44 × “CH3/473” - 0.37 ×
“CH3/330” + 0.04 × “CH3/461” - 0.01 ×
“H+/318” (3)
The descriptors H+, HO-, and CH3 represent the cor-
responding probe interaction energies (electrostatic,
hydrogen bond donor/acceptor, and steric, respectively)
at defined grid points. The locations of the four grid
points appearing in this particular CoMFA equation are
shown in Figure 6. As expected, there are no points

2190 J ournal of Medicinal Chemistry, 2002, Vol. 45, No. 11
Szabo et al.
Ta ble 1. Experimental IC50, pIC50, and Predicted pIC50 Values for GGPPSase Inhibitors and Statistical Results of 3D-QSAR CoMFA
and Catalyst Models (See Figures 2 and 3 for Compound Structures)
a
a
3D-QSAR/CoMFA predicted pIC50
28 compd
Catalyst predicted pIC50
experimental activity
28 compd training set
3 compd test sets
compd
5 (3-azaGGPP)
12 (3-azahomoFPP)
11 (3-azahomoGGPP)
IC50 (µM)
pIC50 training set
3 compd test sets
0.14 ( 0.03 6.84
0.31 ( 0.02 6.51
0.37 ( 0.03 6.43
0.72 ( 0.02 6.14
0.74 ( 0.04 6.13
0.92 ( 0.03 6.04
6.50
6.15
6.32
6.15
6.51
6.15
6.15
5.87
5.06
4.17
4.86
3.91
3.72
3.65
3.33
3.91
3.92
3.60
3.61
3.73
3.62
3.63
3.47
3.33
3.54
3.58
3.92
3.32
6.55
6.20
6.37
6.20
6.55
6.20
6.20
5.91
5.10
4.17
4.91
3.90
3.72
3.64
3.33
3.90
3.90
3.60
3.61
3.78
3.61
3.63
3.47
3.33
3.54
3.57
3.90
3.33
6.46
6.08
6.26
6.08
6.46
6.08
6.08
5.75
4.97
4.18
4.76
3.95
3.74
3.67
3.33
3.96
3.96
3.62
3.63
3.64
3.64
3.65
3.48
3.32
3.55
3.59
3.96
3.29
6.55
6.50
6.09
6.09
6.56
6.09
6.09
5.84
4.99
4.07
4.79
3.80
3.68
3.57
3.50
3.81
3.81
3.57
3.56
3.73
3.56
3.57
3.52
3.39
3.50
3.45
3.81
3.40
6.25
6.46
6.43
5.85
6.22
6.03
6.01
5.92
4.85
4.85
5.00
3.82
4.48
3.44
3.43
3.43
3.43
3.44
3.44
4.89
3.85
3.43
3.43
3.43
3.44
3.44
3.43
3.43
6.32
6.48
6.46
6.25
6.30
6.27
6.25
6.23
4.80
4.80
4.89
4.08
4.42
3.43
3.44
3.43
3.43
3.44
3.43
4.96
4.20
3.43
3.43
3.43
3.44
3.43
3.43
3.43
6.59
6.51
6.57
5.74
6.55
5.89
5.74
5.64
4.77
4.77
4.89
4.00
4.27
3.41
3.41
3.41
3.41
3.41
3.41
4.89
4.11
3.41
3.41
3.41
3.41
3.41
3.41
3.41
6.54
6.54
6.68
5.85
6.54
5.89
5.85
5.85
4.85
4.80
4.96
3.82
4.47
3.42
3.43
3.42
3.42
3.43
3.42
4.89
3.92
3.42
3.42
3.42
3.43
3.42
3.42
3.42
16
6 (3-azaFPP)
14
15
30
17
18
31
19
32 (ibandronate)
36 (pamidronate)
22
20
24
23
25
7 (3-azaGPP)
26
33
27 (NE58095, risedronate)
28 (NE58051, homorisedronate)
35 (alendronate)
29
1.4 ( 0.2
2.2 ( 0.2
4.3 ( 0.4
11 ( 1
5.84
5.66
5.37
4.95
4.72
4.28
4.08
3.75
3.73
3.70
3.66
3.66
3.66
3.62
3.58
3.49
3.45
3.39
3.36
3.26
3.21
3.16
19 ( 1
53 ( 6
83 ( 13
180 ( 20
180 ( 60
200 ( 40
220 ( 50
220 ( 40
220 ( 50
240 ( 20
260 ( 20
330 ( 90
350 ( 50
410 ( 130
440 ( 60
550 ( 200
620 ( 270
690 ( 370
21
34
13 (15-azaGGPP)
. 100
R2 b
0.938
86.8
0.900
0.938
5
0.936
73.7
0.845
0.937
5
0.938
75.1
0.907
0.938
5
0.956
108.7
0.921
0.956
5
0.914
0.899
0.913
0.911
c
Ftest
2d
Rcv
2e
Rbs
Nf
Ng
28
25
25
25
28
25
25
25
a
b
Bold values represent predicted activities of compounds that were not included in the training set. Correlation coefficient. c Ratio of
R2 explained to unexplained ) R2/(1 - R2). Cross-validated correlation coefficient after leave-one-out procedure. e Average squared
d
correlation coefficient calculated during the validation procedure. f Optimal number of principal components. Number of compounds.
g
Ta ble 2. Comparison of Bisphosphonate Inhibitory Activities
against Human Recombinant FPPSase and GGPPSase (See
Figure 3 for Compound Structures)
To test the predictive ability, we removed three
compounds from the training set prior to GFA analysis.
The resulting model based on the reduced training set
was then used to predict the activities of the three
excluded compounds. This procedure was repeated three
times using different test sets, and the predicted IC50
values are shown in bold in Table 1. The three com-
pounds in each test set were chosen at random while
maintaining a distribution of activities. In Figure 7B,
we show graphical results for the three training sets
(O) together with the three test set points (9). The QSAR
equations for the three training sets were
experimental activity, IC50 (µM)
compd
FPPSasea
GGPPSaseb
32 (ibandronate)
36 (pamidronate)
27 (risedronate)
28 (homorisedronate)
35 (alendronate)
0.02
0.20
0.01
2.93
0.05
83 ( 13
180 ( 20
350 ( 50
410 ( 130
440 ( 55
a
b
From ref 24. From Table 1.
associated with the bisphosphonate/diphosphate head-
group, since these highly polar motifs are common to
all of the compounds studied and consequently do not
contribute significantly to the range of activity seen in
GGPPSase inhibition. As shown in Table 1 and Figure
7A, the GFA method provided a good QSAR correla-
tion between experimental and predicted pIC50 values
pIC50 ) 3.54 + 0.44 × “CH3/473” - 0.38 ×
“CH3/330” + 0.04 × “CH3/461” - 0.01 × “H+/318”
(4)
pIC50 ) 3.56 + 0.43 × “CH3/473” - 0.36 ×
with an R2 of 0.938 and a cross-validation Rcv of
2
“CH3/330” + 0.03 × “CH3/461” - 0.01 × “H+/318”
0.900. The optimal number of components in the final
GFA model ()5) was determined using cross-validated
R2 and standard error prediction analyses, as obtained
from the leave-one-out cross-validation technique.30 We
decided to proceed further and investigate the actual
predictive utility of the QSAR model, which if successful
should facilitate the design of additional GGPPSase
inhibitors.
(5)
pIC50 ) 3.46 + 0.45 × “CH3/473” - 0.39 ×
“CH3/330” + 0.04 × “CH3/461” - 0.01 × “H+/344”
(6)
Statistical data for all QSAR equations are given in

Bone Antiresorption and Antiparasitic Agents
J ournal of Medicinal Chemistry, 2002, Vol. 45, No. 11 2191
Table 1. The rms error in predicted pIC50 of the test set
compounds was 0.39, which corresponds to a factor of
∼2.5 in predicting the experimental IC50 values over the
∼700 µM range in activity tested in Table 1. Using solely
the nine predicted test set points, the correlation
between experiment and prediction is R2 ) 0.882.
SARs. The library of compounds investigated con-
tains many diverse structural features: ionic bisphos-
phonate and diphosphate groups; alkyl, alkenyl (prenyl),
aryl, and heteroaryl side chains; 1-OH- and 1-H-bearing
bisphosphonates; and nitrogen-containing or nitrogen-
free side chains, together with different locations of the
side chain nitrogens. With so many diverse structural
features, QSAR methods are clearly desirable in order
to begin to predict in a quantitative manner inhibitory
activity over a wide range of values, but of course, it is
also helpful to draw at least some more qualitative
structural conclusions from an inspection of all of the
results, to guide future inhibitor design. Here, we briefly
outline some of the trends that are apparent.
First, it is evident that all eight aromatic bisphos-
phonates are extremely poor inhibitors of GGPPSase,
even though several of them have low nanomolar IC50
values vs FPPSase. Second, the short alkyl chain
nitrogen-containing bisphosphonates are also poor in-
hibitors. As shown in Table 1, ibandronate (32) has an
83 µM IC50 vs GGPPSase but only a 20 nM IC50 vs
FPPSase.24 These results might imply a rather different
local structure in the GGPPSase and FPPSase E‚I
complexes or even different modes of binding. Alterna-
tively, perhaps the side chains in the known potent
FPPSase inhibitors are simply too short to act as
competitive inhibitors since the substrate (FPP) has a
15 carbon side chain (or at least a continuous 12 car-
bon fragment) that can be expected to play an im-
portant role in binding to the active site, due to van der
Waals dispersion forces or hydrophobic bonding. In the
GGPPSase 3D-QSAR CoMFA, CH3-probe (van der
Waals) interactions are clearly of importance, as may
be seen in eq 3-6 above, while in FPPSase 3D-QSAR
CoMFA models,31 electrostatic interactions appear to be
more important. Third, our results clearly confirm the
importance of hydrophobic interactions in the case of
the n-alkyl 1-hydroxy-1,1-bisphosphonates. As the n-
alkyl chain length increases, the IC50 values decrease
essentially monotonically, from ∼620 µM for C3 to ∼1
µM for C9, C10, and C11. Fourth, the strong inhibition
by 3-azaGGPP (IC50 ) 140 nM) is similar to that
previously reported for the effect of this amino diphos-
phate on native GGPPSase from rat liver (90% inhibi-
tion at 0.9 µM).15a This potency is perhaps not surprising
in view of its close resemblance to GGPP and the
previous observation of product inhibition with bovine
brain GGPPSase (Ki ) 1.2 µM).32
F igu r e 6. MCSG alignment of the 28 molecules that had
quantitatively measurable activities together with the loca-
tions of the four descriptors given in eq 3. The alignments used
16 as the shape reference compound, as discussed in detail in
the text.
second carbocation would be formed in the following
condensation step with IPP (Figure 1). The sequence of
reactions is clearly complex, but we have synthesized
several homologues of azaprenyl FPP and GGPP, which
allow us to test the hypothesis that these compounds
are acting as carbocation intermediate analogues. First,
the additional methylene group in 3-azahomoFPP (12)
would increase the charge separation between am-
monium and diphosphate moieties so that the product
would more closely resemble the putative farnesyl+/
OPP- ion pair intermediate, as illustrated in Figure 8A.
The experimental values of 740 nM for 6 but 310 nM
for 12 are consistent with this idea. Second, in the case
of the GGPP analogues (Figure 8B), we anticipated that
3-azaGGPP (5) might mimic the geranylgeranyl pyro-
phosphate carbocation (E‚I) prior to release from the
enzyme, while 3-azahomoGGPP (11) would be a worse
inhibitor, since it would mimic a C25 synthase inhibitor.
The experimental results support this idea since the
IC50 values for 3-azaGGPP (5) and 3-azahomoGGPP (11)
are 140 and 370 nM, respectively. These trends are not
seen in our QSAR modeling; however, this is most likely
because the CoMFA (or Catalyst) IC50 predictions have
about a factor of 2-2.5 error (see below). The chain
terminus analogue 15-aza GGPP (13) showed no inhibi-
tion at 100 µM, which evidently reflects the incompat-
ibility of its charged terminus with the interior of the
hydrophobic polyene binding pocket.
The various isoprenoid amino diphosphates 5, 6, 11,
and 12 might interact with the GGPPSase active site
in one of several ways: as analogues of the FPP
substrate (S1), as analogues of either of the two different
carbocation diphosphate ion pair intermediates, or as
analogues of the GGPP product. If the mechanism
follows the ionization-condensation-elimination se-
quence proposed by Poulter and co-workers for FPP-
Sase,17,33 the prenyl extension reaction would be initi-
ated by formation of a farnesyl+/OPP- ion pair, and the
Ca t a lyst P h a r m a cop h or e In vest iga t ion s. The
results presented above indicate that 3D-QSAR/CoMFA
methodologies permit relatively accurate predictions of
the activity of the bisphosphonate and diphosphate
GGPPSase inhibitors investigated. However, we only
investigated the single alignment described above, and
clearly, other alignments are possible. A convenient
approach to investigating multiple alignments is to use
the Catalyst18,34 approach in which large numbers of
inhibitor conformations are considered and numerous

2192 J ournal of Medicinal Chemistry, 2002, Vol. 45, No. 11
Szabo et al.
F igu r e 7. Graphical comparisons between experimental and
predicted pIC50 values for GGPPSase inhibition. (A) A 28
compound training set. (B) Superposition of three sets of
predictions for test set (9) and training set (O) compounds.
For the test set only, R2 ) 0.882 and rms error ) 0.39.
F igu r e 8. Schematic showing a comparison between (A)
3-azaFPP (6) and 3-azahomoFPP (12) structures with FPP and
the putative carbocation-pyrophosphate ion pair. (B) 3-azaG-
GPP (5) and 3-azahomoGGPP (11) with the putative (GG-
PPH)+ carbocation formed from FPP+ and IPP. The 3azaGGPP
species is thought to mimic the (GGPPH)+ carbocation.
pharmacophore hypotheses are generated. This ap-
proach has not only predictive utility but also provides
a potentially useful graphical representation of the key
structural features of a pharmacophore, suitable for
virtual screening.
spheres represent the two negative ionizable groups.
This hypothesis is shown superimposed on (A) 3-azaG-
GPP (5), (B) 17, and (C) risedronate (27) in Figure 9.
This figure clearly illustrates the key importance of the
hydrophobic interactions in controlling activity. Com-
pounds 5 and 17 have two and one hydrophobic
“matches”, respectively, while risedronate (27) has none.
This agrees with the increasing IC50 values of 140 nM,
4.3, and 350 µM, respectively, and similar low activities
are predicted for other potent FPPSase inhibitors; see
Table 2.
To further judge the predictive ability of this phar-
macophore model, we again used the three training sets
described above (results given in Table 1) to generate
hypotheses containing two negative ionizable and two
hydrophobic features to predict the activities of three
sets of three test compounds, this time employing the
Catalyst program. We found an R2 value of 0.948
between experiment and prediction (see Figure 10),
together with a 0.28 pIC50 rms error. We did not
investigate numerous other alignments, conformations,
or hypotheses in either the Catalyst or the CoMFA
We therefore generated a pharmacophoric model for
all compounds using the Hypogen module in the Cata-
lyst software package.18 To facilitate feature recognition,
we used the structures from the 3D-QSAR/CoMFA
model with explicitly defined single and double P-O
bonds. Activity uncertainities of 3.0 (Catalyst default)
were used for all compounds. Hydrophobic, negative
ionizable, and positive charge features were considered
for hypothesis generation, since these features are
common among the most active compounds. We re-
quired hypotheses to have two negative ionizable groups
in order to account for the bidentate interactions of
diphosphates observed in the crystal structures of
FPPS.27 While numerous hypotheses were generated,
they all typically had quite similar features, with two
negative ionizable groups and two or three hydrophobic
features. One representative hypothesis is illustrated
in Figure 9 in which the light blue spheres represent
the two hydrophobic features while the dark blue

Bone Antiresorption and Antiparasitic Agents
J ournal of Medicinal Chemistry, 2002, Vol. 45, No. 11 2193
F igu r e 9. Catalyst pharmacophore hypothesis for GGPPSase
inhibition by bisphosphonates and diphosphates. The light blue
spheres represent hydrophobic components of the hypothesis,
and the dark blue spheres represent negative ionizable
components. Pharmacophore superimposed on (A) 3-azaGGPP
(5); (B) 17; and (C) risedronate (27).
methods, since the results already appeared quite
respectable.
When taken together, these results clearly indicate
that long chain alkyl bisphosphonates and related
compounds can act as inhibitors of GGPPSase. While
the IC50 values of most of the compounds investigated
in this initial study are higher than those of the most
potent FPPSase inhibitors, the IC50 values for the
isoprenoid amino diphosphates are equal to or lower
than those of pamidronate and alendronate (IC50 ≈ 0.2-
0.5 µM), drugs widely used in bone resorption therapy.
Because the longer chain inhibitors we have investi-
gated are quite lipophilic, they can be expected to have
improved absorption characteristics and may thus
represent a novel approach to the design of bone
antiresorption and antiparasitic agents.
F igu r e 10. Graph showing correlation between experimental
and Catalyst predicted pIC50 values. (A) A 28 compound
training set. (B) Superposition of three sets of predictions for
test set (9) and training set (O) compounds. For the test set
only, R2 ) 0.948 and rms error ) 0.28.
pharmacophore model (hypothesis) for GGPPSase in-
hibition that also had promising predictive ability and
confirmed the importance of hydrophobic interactions
in GGPPSase inhibition together with, as expected, the
importance of negative ionizable groups. Overall, given
the important role of geranylgeranylation in osteoclast
function (bone resorption) as well as in the growth of
parasitic protozoa, these results imply that GGPPSase
inhibitors appear to be attractive candidates for the
development of future drugs, which can be expected to
have improved absorption characteristics due to their
more lipophilic character.
Con clu sion s
The results described above are of interest for a
number of reasons. First, we have carried out a detailed
study of the inhibition of an enzyme in the isoprenoid
biosynthetic pathway, GGPPSase, by a broad range of
bisphosphonate inhibitors and azaprenyl diphosphate
analogues. The range of IC50 values varies from 140 nM
to ∼690 µM. Second, we have carried out the first QSAR
study of the inhibition of GGPP synthase using 3D-
QSAR/CoMFA techniques. The results show good agree-
ment between experimental and predicted pIC50 values
Exp er im en ta l Section
In h ibitor s. Experimental procedures and characterization
data are provided below for 3-azaFPP (6) (eq 1) and 3-azaho-
moFPP (12) (eq 2). Purities of the 3-azaprenyl and 3-azaho-
moprenyl PPs were judged to be g90-95% from inspection
and integration of their 1H and 31P NMR spectra (see Sup-
porting Information). 15-azaGGPP (13) was available from
previous work.35
Samples of bisphosphonates 14-29, 31, and 33-36 were
available from previous work.10 The purity of all bisphospho-
nates was verified by microchemical analysis (H/C/N/P) and
via 13C, 31P, and 1H NMR spectroscopy; the 1H NMR experi-
2
2
(R2 ) 0.938, Rcv ) 0.900, Rbs ) 0.938, and F-test )
86.8), and the method has predictive utility as evidenced
by prediction of the activity of nine test set compounds
with an rms pIC50 error of 0.39. Third, an analysis of
the CoMFA inhibition results clearly indicates the
importance of van der Waals interactions in GGPPSase
inhibition. Fourth, using Catalyst, we developed a

2194 J ournal of Medicinal Chemistry, 2002, Vol. 45, No. 11
Szabo et al.
ments were performed in triplicate using an internal maleic
acid quantitation standard.
extractions were repeated g5 times until a white solid was
completely formed. The extraction sequence was then repeated
with 4:5:5 (v/v/v) 25 mM aqueous NH4HCO3/CH3CN/2-pro-
panol. The combined supernant solutions were concentrated
to near dryness by rotary evaporation. The solid was dissolved
in 25 mM NH4HCO3 (10 mL), and the solution was lyophilized
to dryness to give 3-aza FPP 6 (308 mg, 60% yield, 91% purity
by 31P NMR) as a white solid. 1H NMR (500 MHz, ND4OD/
D2O): δ 5.01 (br, 2 H), 4.18 (br, 2 H), 3.35 (br, 2 H), 3.03 (br,
2 H), 2.83 (s, 3 H), 2.39 (br, 2 H), 1.97 (q, 2 H, J ) 6.2), 1.90
(br, 2 H), 1.57 (s, 3 H), 1.55 (s, 3 H), 1.49 (s, 3 H). 13C NMR
(125 MHz, ND4OD/D2O): δ 139.3, 131.2, 124.2 (d, J ) 7.4),
118.0 (d, J ) 5.5), 59.5, 55.4 (d, J ) 17.5), 39.5 (d, J ) 8.3),
39.1, 38.4 (d, J ) 9.2), 26.0, 25.1 (d, J ) 6.4), 22.1, 17.0 (d, J
) 6.4), 15.4 (d, J ) 6.4). 31P NMR (202 MHz, ND4OD/D2O): δ
-6.34 (d, 1 P, J ) 19.8), -10.02 (d, 1 P, J ) 19.8). LRFABMS
N-Met h yl-4,8-d im et h yl-3E,7E-n on a d ien -1-a m in e (2b).
The procedure was based on the literature.15b,36 Lithiation of
N′-tert-butyl-N,N-dimethylformamidine (2.4 g, 18.8 mmol) in
dry THF (60 mL) with tert-butyllithium (13 mL, 1.7 M, 22.1
mmol) in pentane at -25 °C for 1 h was followed by alkylation
with geranyl chloride (2.9 g, 17 mmol) in THF (10 mL) at -78
°C to room temperature (2 h). Hydrolysis of the crude forma-
midine in 4:1 methanol-water (70 mL) containing KOH (4.8
g, 85 mmol) at reflux (ca. 12 h) afforded amine 2b (2.7 g, 95%)
as a colorless oil, which was used without further purification.
1
TLC Rf 0.12 (33% MeOH/CH2Cl2). H NMR, 13C NMR, and IR
data agreed with the literature values.15b
1,1-Dim eth yleth yl 2-[N-Meth yl-N-[4,8-d im eth yl-3E,7E-
n on a d ien yl]a m in o]eth a n oa te (3b). Alkylation of amine (2.7
g, 16.2 mmol) with tert-butyl bromoacetate (3.3 g, 17.0 mmol)
in the presence of triethylamine (1.8 g, 17.8 mmol) in dry THF
(18 mL) under N2 was carried out by a literature procedure.15b
Purification of the crude product by column chromatography
with 50% ether/pentane gave 4.1 g (90%) of the known amino
(H2O): m/z 386.2. HRFABMS (positive ion) calcd for C14H30
-
NO7P2, 386.1498; found, 386.1498. The 31P NMR spectrum
showed an extra peak attributed to inorganic pyrophosphate
[δP -7.49 (s, 0.09 P)].
1,1-Dim eth yleth yl 3-[N-Meth yl-N-[4,8-d im eth yl-3E,7E-
n on a d ien yl]a m in o]p r op a n oa te (9). The procedure was
1
ester 3b as a colorless oil. TLC Rf 0.63 (33% EA/hexane). H
based on a literature procedure for a different compound.38
A
NMR, 13C NMR, and IR data agreed with the literature
values.15b
solution of the N-methylamine 2b (715 mg, 4.3 mmol) in dry
MeOH (15 mL) under N2 was stirred at room temperature as
tert-butyl acrylate (1.1 mL, 8.5 mmol) was added. The solution
was refluxed for 1 day. After it was concentrated under
reduced pressure, the residue was subjected to column chro-
matography with 33% EA/hexane to give 913 mg (72%) of the
â-amino ester 9 as a colorless oil. TLC Rf 0.40 (33% EA/
hexane). 1H NMR (500 MHz, CDCl3): δ 5.09-5.03 (m, 2 H),
2.65 (t, 2 H, J ) 7.3), 2.35 (t, 2 H, J ) 7.7), 2.32 (t, 2 H, J )
7.5), 2.21 (s, 3 H), 2.12 (q, 2 H, J ) 7.9), 2.03 (q, 2 H, J ) 7.9),
1.94 (t, 2 H, J ) 8.1), 1.64 (d, 3 H, J ) 1.1), 1.58 (s, 3 H), 1.56
(s, 3 H), 1.41 (s, 9 H). 13C NMR (125 MHz, CDCl3): δ 172.0,
136.2, 131.3, 124.2, 121.7, 80.2, 57.2, 52.8, 41.8, 39.6, 33.6, 28.0,
26.6, 26.0, 25.6, 17.6, 16.0. IR (thin film): νmax 2975, 2927,
2-[N-Meth yl-N-[4,8-d im eth yl-3E,7E-n on a d ien yl]a m in o]-
eth a n ol (4b). The procedure was based on the literature for
a related compound.15b A suspension of LiAlH4 (742 mg, 19.5
mmol) in dry ether (50 mL) under N2 was stirred and cooled
at 0 °C as amino ester 3b (1.7 g, 5.9 mmol) in ether (15 mL)
was added dropwise. The mixture was stirred at room tem-
perature for 2 h and cooled to 0 °C. Water (0.8 mL), 15%
aqueous KOH (0.8 mL), and water (2.2 mL) were added in
succession.37 After it was stirred at room temperature for 1 h,
the insoluble salts were removed by filtration, and the filtrate
was concentrated. Purification by column chromatography
with 50% ether/pentane gave 1.3 g (94%) of amino alcohol 4b
1
as a colorless oil. TLC Rf 0.35 (33% MeOH/CH2Cl2). H NMR
2852, 2795, 1731, 1454, 1367, 1255, 1155, 1039, 848 cm-1
.
(500 MHz, CDCl3): δ 5.08-5.01 (m, 2 H), 3.53 (t, 2 H, J )
5.4), 3.22 (br. s, 1 H), 2.48 (t, 2 H, J ) 5.4), 2.37 (t, 2 H, J )
7.1), 2.22 (s, 3 H), 2.12 (q, 2 H, J ) 7.5), 2.01 (q, 2 H, J ) 7.5),
1.93 (t, 2 H, J ) 8.1), 1.62 (s, 3 H), 1.57 (s, 3 H), 1.55 (s, 3 H).
13C NMR (125 MHz, CDCl3): δ 136.3, 131.2, 124.1, 121.6, 58.6,
58.3, 57.4, 41.5, 39.6, 26.5, 25.8, 25.5, 17.5, 15.9. IR (thin
film): νmax 3421, 2966, 2923, 2854, 2798, 1668, 1452, 1376,
1041 cm-1. EIMS: m/z 226.2 [M + H]+ (1), 194.2 (2), 124.1
(1), 88.1 (100), 69.1 (9), 57.0 (5). HREIMS: m/z 224.2018 (calcd
for C14H26NO, [M - H]+ 224.2014).
EIMS: m/z 310.5 [M + H]+ (1), 254.2 (1), 236.2 (3), 194.2 (13),
172.1 (40), 116.1 (100), 88.1 (1), 74.1 (9), 57.1 (5). HREIMS:
m/z 308.2586 (calcd for C19H34NO2 [M - H]+ 308.2590).
3-[N-Meth yl-N-[4,8-d im eth yl-3E,7E-n on a d ien yl]a m in o]-
p r op a n ol (10b). The LiAlH4 reduction of 9 was carried out
as described above for 4b. Yield, 389 mg (91%). TLC Rf 0.35
(33% MeOH/CH2Cl2). 1H NMR (500 MHz, CDCl3): δ 5.09-
5.03 (m, 2 H), 3.76 (t, 2 H, J ) 5.1), 2.57 (t, 2 H, J ) 5.6), 2.35
(t, 2 H, J ) 7.3), 2.23 (s, 3 H), 2.15 (q, 2 H, J ) 7.7), 2.04 (q,
2 H, J ) 7.7), 1.95 (t, 2 H, J ) 7.9), 1.66 (quintet, 2 H, J )
5.1), 1.64 (s, 3 H), 1.59 (s, 3 H), 1.57 (s, 9 H). 13C NMR (125
MHz, CDCl3): δ 136.5, 131.2, 124.2, 121.3, 64.6, 58.3, 57.8,
41.8, 39.6, 27.6, 26.5, 25.8, 25.6, 17.6, 15.9. IR (thin film): νmax
3378 (-OH), 2925, 2852, 2800, 1668, 1452, 1376, 1128, 1051,
835 cm-1. EIMS: m/z 240.2 [M + H]+ (2), 194.2 (3), 172.1 (2),
116.1 (4), 102.1 (100), 88.1 (5), 69.1 (8), 58.1 (61). HREIMS:
m/z 238.2172 (calcd for C15H28NO [M - H]+ 238.2171).
3-[N-Meth yl-N-[4,8-d im eth yl-3E,7E-n on a d ien yl]a m in o]-
p r op yl Dip h osp h a te (3-a za h om oF P P , 12). The diphosphate
was prepared from amino alcohol 10b as described above for
4b except that the extraction buffer was 1:2:2 (v/v/v) 25 mM
aqueous NH4HCO3/CH3CN/2-propanol. Yield, 353 mg (80%,
93% purity by 31P NMR). 1H NMR (500 MHz, ND4OD/D2O):
δ 4.93 (br d, 2 H, J ) 6.9), 3.84 (t, 2 H, J ) 5.8, 5.6), 3.09 (t,
2 H, J ) 7.1), 2.90 (t, 2 H, J ) 7.5), 2.65 (s, 3 H), 2.26 (q, 2 H,
J ) 7.9), 1.92-1.86 (m, 4 H), 1.84 (q, 2 H, J ) 7.5), 1.47 (s, 6
H), 1.40 (s, 3 H). 13C NMR (125 MHz, ND4OD/D2O): δ 139.9,
132.6, 124.2, 118.1, 62.9, 54.7, 53.1, 39.7, 39.0, 38.5 (d, J )
3.7), 25.9, 25.1, 24.6 (d, J ) 8.3), 22.1, 17.1, 15.4. 31P NMR
(202 MHz, ND4OD/D2O): δ -7.66 (d, 1 P, J ) 19.8), -9.70 (d,
1 P, J ) 21.4). LRFABMS (H2O): m/z 400.2. HRFABMS
(positive ion) Calcd for C15H32NO7P2, 400.1654; found, 400.1652.
The 31P NMR spectrum showed an extra peak attributed to
inorganic monophosphate [δP 1.44 (s, 0.07 P)].
2-[N-Meth yl-N-[4,8-d im eth yl-3E,7E-n on a d ien yl]a m in o]-
eth yl Dip h osp h a te (3-a za F P P , 6). Literature methods20
were used with some modifications including omission of the
cellulose column chromatography step. A solution of amino
alcohol 4b (257 mg, 1.1 mmol) and triethylamine (173 mg, 173
mmol) in dry CH2Cl2 (11 mL) under N2 was stirred and cooled
at -15 °C as methanesulfonyl chloride (0.1 mL, 1.3 mmol) was
added dropwise. After 0.5 h, the product was extracted with
ice-cold CH2Cl2 (10 mL). The organic layer was washed with
ice-cold saturated NaHCO3 (2 × 10 mL), dried (MgSO4), and
concentrated to 5 mL. MeCN (15 mL) was added to the solution
of mesylate/chloride in CH2Cl2 (5 mL). The solution was
concentrated to 2 mL and stirred in an ice bath as tris(tetra-
n-butyl)ammonium hydrogen pyrophosphate (2.7 g, 2.8 mmol)
in MeCN (3 mL) was added. The suspension was stirred at
room temperature for 6 h. After it was washed with pentane
(3 × 10 mL) and concentrated, the residue was dissolved in 2
mL of ion exchange buffer (25 mM NH4HCO3 containing 2%
(v/v) 2-propanol). The resulting solution was passed through
a column (2 cm × 10 cm) of Dowex AG 50W-X8 (100-200
mesh) cation-exchange resin (ammonium form). The column
was eluted with 2 column volumes of ion exchange buffer, and
the eluant was lyophilized to dryness. The residual white solid
was partially dissolved in 20 mL of 3:5:5 (v/v/v) 25 mM aqueous
NH4HCO3-MeCN-2-propanol. The gellike mixture was vor-
texed, the suspension was centrifuged, and the supernatant
solution was decanted into another round-bottomed flask. The
N-Octyla m in om eth a n e-1,1-bisp h osp h on ic Acid (30).
This compound was prepared by reaction of n-octylamine with

Bone Antiresorption and Antiparasitic Agents
J ournal of Medicinal Chemistry, 2002, Vol. 45, No. 11 2195
ethyl orthoformate and diethyl phosphite followed by hydroly-
sis in 6 M HCl.10,22
Ack n ow led gm en t. This work was supported by the
United States Public Health Organization (NIH Grants
GM-50698 (E.O.) and GM-13956 (R.M.C.)), by the
UNDP/World Bank/WHO Special Program for Research
and Training in Tropical Diseases (TDR) (E.O.), and by
the National Computational Science Alliance (NSF
Grants MCB-000016N (E.O.), MCB-000018N (E.O.), and
MCB-000020N (E.O.)). C.M.S. was supported by NIH
Training Grant GM-08276. M.B.M. was an American
Heart Association, Midwest Affiliate, Predoctoral Fel-
low.
N-Meth yl-N-p en tyl-3-a m in o-1-h yd r oxyp r op a n e-1,1-bis-
p h osp h on ic Acid (Iba n d r on a te, 32). The bisphosphonate
was prepared by reaction of N-methyl-N-pentylamine and
acrylonitrile followed by hydrolysis in 75% H2SO4 and reduc-
tive phosphonylation with H3PO3/PCl3.10,21 Absolute compound
purities as determined from these experiments were 98.2 and
99.4% for 30 and 32, respectively.
In h ibition of h r GGP P Syn th a se. The following hexahis-
tagged human recombinant (hr) geranyl-geranyl diphosphate
synthase (320AA)16sMGSSHHHHHHSSGLVPRGSHMEKT-
QE TVQRILLE P YKYLLQLP GKQVRTKLSQAF NH WLK-
VP E DKLQIIIE VTE MLH NASLLIDDIE DNSKLRRGF P V-
AH SIYGIP SVINSANYVYF LGLE KVLTLDH P DAVKLF T-
RQLLE LH QGQGLDIYWRDNYTCP TE E E YKAMVLQKT-
GGLFGLAVGLMQLFSDYKEDLKPLLNTLGLFFQIRDD-
YAN L H S K E YS E N K S F C E D L T E G K F S F P T I H AI W -
S R P E S T Q VQ N I L R Q R T E N I D I K K YC VH YL E D VG S -
F E YTRN TLKE LE AKAYKQI DARGGN P E LVALVKH L-
SKMFKEENEswas used in this study. A cDNA coding for this
enzyme was cloned into the NdeI and BamHI sites of an
expression vector pET-15b (Novagen). Escherichia coli BL21
(DE3) cells transformed with the plasmid were grown to an
A600 of 0.6. Isopropyl 1-thio-â-D-galactopyranoside was added
to a final concentration of 1 mM, and after 3 h culture, the
cells were collected and suspended in 20 mM Tris-HCl buffer
(pH 7.9) containing 5 mM imidazole and 500 mM NaCl. The
suspensions were sonicated, and the expressed hexahis-tagged
proteins were purified by nickel affinity column chromatog-
raphy. The purified enzyme fractions were combined, dialyzed
against Tris-HCl buffer (pH 7.9) containing 500 mM NaCl, and
stored at -80 °C before use. Protein (100 ng) in 50 mM
potassium phosphate buffer (pH 7.0) containing 5 mM MgCl2,
2 mM DTT, 1 mg/mL BSA, and either 25 µM GPP or FPP in
a total volume of 23.5 µL was incubated at 0 °C for 15 min.
Then, 1.5 µL of a 0.35 mM solution of [14C]IPP was added, the
mixture was incubated at 37 °C for 15 min, and the pH was
lowered by addition of 75 µL of HCl/MeOH to effect hydrolysis
of the allylic PPs present. Following a second 15 min incuba-
tion at 37 °C, the reaction mixture was neutralized by addition
of 37.5 µL of 6 N NaOH. Prenyl alcohols were then extracted
with 200 µL of hexane, and a 100-µL aliquot was transferred
to a scintillation vial for radioactivity counting. The quantita-
tive activity results used in the QSAR analysis and reported
in Table 1 were derived only using FPP as the substrate. The
distribution of 14C before acidification and hydrolysis was
determined by separating IPP, GPP, FPP, and GGPP using
normal phase TLC (Merck 11845 silica gel plates) using a
2-propanol/ammonia/H2O 6:3:1 (v/v/v) solvent system (Figure
4). Spots were visualized with a Bio-image analyzer BAS1000
(Fuji Film).
Refer en ces
(1) J acobson, T. A.; Schein, J . R.; Williamson, A.; Ballantyne, C. M.
Maximizing the Cost-Effectiveness of Lipid-Lowering Therapy.
Arch. Intern. Med. 1998, 158, 1977-1989.
(2) Martin, M. B.; Arnold, W.; Heath, H. T., III; Urbina, J . A.;
Oldfield, E. Nitrogen-Containing Bisphosphonates as Carbo-
cation Transition State Analogues for Isoprenoid Biosynthesis.
Biochem. Biophys. Res. Commun. 1999, 263, 754-758.
(3) Cromartie, T. H.; Fisher, K. J .; Grossman, J . N. The Discovery
of a Novel Site of Action of Herbicidal Bisphosphonates. Pestic.
Biochem. Physiol. 1999, 63, 114-126.
(4) Keller, R. K.; Fliesler, S. J . Mechanism of Aminobisphosphonate
Action: Characterization of Alendronate Inhibition of the Iso-
prenoid Pathway. Biochem. Biophys. Res. Commun. 1999, 266,
560-563.
(5) Bergstrom, J . D.; Bostedor, R. G.; Masarachia, P. J .; Reszka, A.
A.; Rodan, G. Alendronate is a Specific, Nanomolar Inhibitor of
Farnesyl Diphosphate Synthase. Arch. Biochem. Biophys. 2000,
373, 231-241.
(6) Grove, J . E.; Brown, R. J .; Watts, D. J . The Intracellular Target
for the Antiresorptive Aminobisphosphonate Drugs in Dictyo-
stelium discoideum is the Enzyme Farnesyl Diphosphate Syn-
thase. J . Bone Miner. Res. 2000, 15, 971-981.
(7) van Beek, E.; Pieterman, E.; Cohen, L.; Lowik, C.; Papapoulos,
S. Farnesyl Pyrophosphate Synthase is the Molecular Target of
Nitrogen-Containing Bisphosphonates. Biochem. Biophys. Res.
Commun. 1999, 264, 108-111.
(8) Benford, H. L.; McGowan, N. W.; Helfrich, M. H.; Nuttall, M.
E.; Rogers, M. J . Visualization of Bisphosphonate-Induced
Caspase-3 Activity in Apoptotic Osteoclasts in vitro. Bone 2001,
28, 465-473.
(9) Halasy-Nagy, J . M.; Rodan, G. A.; Reszka, A. A. Inhibition of
Bone Resorption by Alendronate and Risedronate does not
Require Osteoclast Apotosis. Bone 2001, 29, 553-559.
(10) Martin, M. B.; Grimley, J . S.; Lewis, J . C.; Heath, H. T., III;
Bailey, B. N.; Kendrick, H.; Yardley, V.; Caldera, A.; Lira, R.;
Urbina, J . A.; Moreno, S. N.; Docampo, R.; Croft, S. L.; Oldfield,
E. Bisphosphonates Inhibit the Growth of Trypanosoma brucei,
Trypanosoma cruzi, Leishmania donovani, Toxoplasma gondii,
and Plasmodium falciparum:
a Potential Route to Chemo-
therapy. J . Med. Chem. 2001, 44, 909-916.
(11) Moreno, B.; Bailey, B. N.; Luo, S.; Martin, M. B.; Kuhlenschmidt,
M.; Moreno, S. N. J .; Docampo, R.; Oldfield, E. 31P NMR of
Apicomplexans and the Effects of Risedronate on Cryptospo-
ridium parvum Growth. Biochem. Biophys. Res. Commun. 2001,
284, 632-637.
3D-QSAR/CoMF A. CoMFA was performed within the
QSAR module of Cerius2 4.6 using default settings. Molecular
mechanics calculations were carried out with the Cerius2
standard universal force field, with a convergence criterion
requiring a minimum energy change of 0.001 kcal/mol. Charge
calculations were performed by using the Gasteiger-Marsili
method.28 CoMFA was used to evaluate the interaction energy
between a probe (H+, OH- (donor/acceptor), and CH3), and the
molecule of interest at a series of points defined by a rectan-
gular grid around the surface of the aligned molecules. The
atomic coordinates of the models were used to compute field
values on each point of the 3D grid. Grid size was adjusted to
2.00 Å. Each energy associated with an MFA grid point served
as input for calculation of the QSAR results. To obtain a
quantitative analysis of the dependence of biological activity
on MFA parameters, regression analysis was applied, leading
to the equations given in the text.
(12) Yardley, V.; Khan, A. A.; Martin, M. B.; Slifer, T. R.; Araujo, F.
G.; Moreno, S. N. J .; Docampo, R.; Croft, S. L.; Oldfield, E. In
Vivo Activity of the Farnesyl Pyrophosphate Synthase Inhibitors
Alendronate, Pamidronate and Risedronate Against Leishmanai
donovani and Toxoplasma gondii. Antimicrob. Agents Chemoth-
er. 2002, 46, 929-931.
(13) Rodriguez, N.; Bailey, B. N.; Martin, M. B.; Oldfield, E.; Urbina,
J . A.; Docampo, R. Unpublished results.
(14) Coxon, F. P.; Helfrich, M. H.; Van’t Hof, R.; Sebti, S.; Ralston,
S. H.; Hamilton, A.; Rogers, M. J . Protein Geranylgeranylation
is Required for Osteoclast Formation, Function, and Survival:
Inhibition by Bisphosphonates and GGTI-298. J . Bone Miner.
Res. 2000, 15, 1467-1476.
(15) (a) Sagami, H.; Korenaga, T.; Ogura, K.; Steiger, A.; Pyun, H.
J .; Coates, R. M. Studies on Geranylgeranyl Diphosphate Syn-
thase from Rat Liver: Specific Inhibition by 3-Azageranylgeranyl
Diphosphate. Arch. Biochem. Biophys. 1992, 297, 314-320. (b)
Steiger, A.; Pyun, H.-J .; Coates, R. M. Synthesis and Charac-
terization of Aza Analogue Inhibitors of Squalene and Gera-
nylgeranyl Diphosphate Synthases. J . Org. Chem. 1992, 57,
3444-3449.
Su p p or tin g In for m a tion Ava ila ble: Schemes 1 and 2
showing synthetic routes to azaprenyl diphosphates 5-7, 11,
and 12 and general experimental aspects. This material is
available free of charge via the Internet at http://pubs.acs.org.
(16) Kuzuguchi, T.; Morita, Y.; Sagami, I.; Sagami, H.; Ogura, K.
Human Geranylgeranyl Diphosphate Synthase: cDNA Cloning
and Expression. J . Biol. Chem. 1999, 274, 5888-5894.

2196 J ournal of Medicinal Chemistry, 2002, Vol. 45, No. 11
Szabo et al.
(17) (a) Koyama, T.; Ogura, K. Isopentenyl Diphosphate Isomerase
and Prenyl Transferases. In Isoprenoids Including Carotenoids
and Steroids; Cane, D. E., Ed.; Comprehensive Natural Products
Chemistry; 1999; Vol. 2, Chapter 4, pp 86-96. (b) Ogura, K.;
Koyama, T. Enzymatic Aspects of Isoprenoid Chain Elongation.
Chem. Rev. 1998, 98, 1263-1276.
(29) Hasegawa, K.; Funatsu, K. Partial Least Squares Modeling and
Genetic Algorithm Optimization in Quantitative Structure-
Activity Relationships. SAR QSAR Environ. Res. 2000, 11, 189-
209.
(30) Cramer, R. D.; Bunce, J . D.; Patterson, D. E.; Frank, I. E.
Crossvalidation, Bootstrapping, and Partial Least Squares
Compared with Multiple Regression in Conventional QSAR
Studies. Quant. Struct.-Act. Relat. 1988, 7, 18-25.
(31) (a) Martin, M. B.; Kendrick, H.; de Luca-Fradley, K.; Lewis, J .
C.; Grimley, J . S.; van Brussel, E. M.; Olsen, J . R.; Croft, S. L.;
Oldfield, E. A Study of the Activity of Biosphosphonates Against
Trypanosoma brucei rhodesiense: Farnesyl Pyrophosphate Syn-
thase as a Drug Target and an Analysis of Drug Toxicity.
Unpublished results. (b) Szabo, C. M.; Martin, M. B.; Oldfield,
E. An Investigation of Bone Resorption and Dictyostelium
discoidium Growth Inhibition by Bisphosphonate Drugs. Un-
published results.
(18) Catalyst, version 4.6; Molecular Simulations Inc.: Burlington,
MA, 2000.
(19) Meyers, A. I.; Ten Hoeve, W. R-Substitution of Amines via Dipole
Stabilized Carbanions from Formamidines. J . Am. Chem. Soc.
1980, 102, 7125-7126.
(20) (a) Davisson, V. J .; Woodside, A. B.; Neal, T. R.; Stremler, K.
E.; Muehlbacher, M.; Poulter, C. D. Phosphorylation and Iso-
prenoid Alcohols. J . Org. Chem. 1986, 51, 4768-4779. (b)
Woodside, A. B.; Huang, Z.; Poulter, C. D. Trisammonium
Geranyl Diphosphate. Organic Syntheses; Wiley and Sons: New
York, 1993; Collect. Vol. VIII, p 616. (c) Davission, V. J .;
Woodside, A. B.; Poulter, C. D. Synthesis of Allylic and Homoal-
lylic Isoprenoid Pyrophosphates. Methods Enzymol. 1985, 110A,
130.
(32) Sagami, H.; Morita, Y.; Ogura, K. Purification and Properties
of Geranylgeranyl-Diphosphate Synthase from Bovine Brain. J .
Biol. Chem. 1994, 269, 20561-20566.
(21) Kieczykowski, G. R.; J obson, R. B.; Melillo, D. G.; Reinhold: D.
F.; Grenda, V. J .; Shinkai, I. Preparation of (4-Amino-1-Hy-
droxybutylidene) Bisphosphonic Acid Sodium Salt, MK-217
(Alendronate Sodium). An Improved Procedure for the Prepara-
tion of 1-Hydroxy-1,1-Bisphosphonic Acids. J . Org. Chem. 1995,
60, 8310-8312.
(33) (a) Laskovics, F. M.; Poulter, C. D. Prenyltransferase. Determi-
nation of the Binding Mechanism and Individual Kinetic Con-
stants for Farnesylpyrophosphate Synthase by Rapid Quench
and Isotope Partitioning Experiments. Biochemistry 1981, 20,
1893-1901. (b) Mathis, J . R.; Poulter, C. D. Yeast Protein
Farnesyltransferase. A Presteady-State Kinetic Analysis. Bio-
chemistry 1997, 36, 6367-6376. (c) Poulter, C. D.; Rilling, H. C.
Prenyltransferase and Isomerase. In Biosynthesis of Isoprenoid
Compounds; Porter, J . W., Spurgenon, S. L., Eds; J . Wiley: New
York, 1981; Vol. I, Chapter 4.
(34) Kaminski, J . J .; Rane, D. F.; Snow, M. E.; Weber, L.; Rothofsky,
M. L.; Anderson, S. D.; Lin, S. L. Identification of Novel Farnesyl
Protein Transferase Inhibitors Using Three-Dimensional Data-
base Searching Methods. J . Med. Chem. 1997, 40, 4103-4112.
(35) Ravn, J . J .; J in, A. Q.; Coates, R. M. Synthesis of Allylic
Isoprenoid Diphosphates by SN2 Displacement of Diethyl Phos-
phate. Eur. J . Org. Chem. 2000, 8, 1401-1410.
(22) Soloducho, J .; Gancarz, R.; Wieczorek, P.; Korf, J .; Hafner, J .;
Lejczak, B.; Kafarski, P. Patent PL93-298436 93408, 1993;
Chem. Abstr. No. 128:61631.
(23) Benedict, J . J .; Perkins, C. B. European Patent Application
#853091403, 1995.
(24) Dunford, J . E.; Thompson, K.; Coxon, F. P.; Luckman, S. P.;
Hahn, F. M.; Poulter, C. D.; Ebetino, F. H.; Rogers, M. J .
Structure-Activity Relationships for Inhibition of Farnesyl
Diphosphate Synthase in vitro and Inhibition of Bone Resorption
in vivo by Nitrogen-Containing Bisphosphonates. J . Pharmacol.
Exp. Ther. 2001, 296, 235-242.
(25) Montalvetti, A.; Bailey, B. N.; Martin, M. B.; Severin, G. W.;
Oldfield, E.; Docampo, R. Bisphosphonates are Potent Inhibitors
of Trypanosoma cruzi Farnesyl Pyrophosphate Synthase. J . Biol.
Chem., 2001, 276, 33930-33937.
(36) Fung, A. K. L.; Baker, W. R.; Fakhoury, S.; Stein, H. H.; Cohen,
J .; Donner, B. G.; Garvey, D. S.; Spina, K. P.; Rosenberg, (1R,-
2â,3â,4R-1,2-Bis[N-propyl-N-(4-phenoxybenzyl)amino]carbon-
yl] cyclobutane-3,4-dicarboxylic Acid(A-87049): A Novel Potent
Squalene Synthase Inhibitor. J . Med. Chem. 1997, 40, 2123-
2125.
(26) Cerius2 Modeling Environment, version 4.6; Accelrys Inc.: San
Diego, CA, 2001.
(27) Tarshis, L. C.; Proteau, P. J .; Kellogg, B. A.; Sacchettini, J . C.;
Poulter, C. D. Regulation of Product Chain Length by Isoprenyl
Diphosphate Synthases. Proc. Natl. Acad. Sci. U.S.A. 1996, 93,
15018-15023.
(28) Gasteiger, J .; Marsili, M. Iterative Partial Equalization of Orbital
Electronegativity: a Rapid Access to Atomic Charges. Tetrahe-
dron 1980, 36, 3219-3222.
(37) Fieser, L. F.; Fieser, M. Reagents for Organic Synthesis; J ohn
Wiley & Sons: New York, 1967; pp 581-595.
(38) Ganesan, A.; Heathcock, C. H. A Stereochemical Test of the
Mechanism of Electrophilic Substitution in 3-Substituted In-
doles. Tetrahedron Lett. 1993, 34, 439-440.
J M010412Y