Activity of sulfonium bisphosphonates on tumor cell lines

Article abstract of DOI:10.1021/jm700991k

We investigated three series of sulfonium bisphosphonates for their activity in inhibiting the growth of three human tumor cell lines. The first series consisted of 6 cyclic sulfonium bisphosphonates, the most active species having an (average) IC₅₀ of 89 μM. The second consisted of 10 phenylalkyl and phenylalkoxy bisphosphonates, the most active species having an IC₅₀ of 18 μM. The third series consisted of 17 n-alkyl sulfonium bisphosphonates, the most active species having an IC₅₀ of ～240 nM. Three QSAR models showed that the experimental cell growth inhibition results could be well predicted. We also determined the structures of one sulfonium bisphosphonate bound to farnesyl diphosphate synthase, finding that it binds exclusively to the dimethylallyl diphosphate binding site. These results are of interest since they show that sulfonium bisphosphonates can have potent activity against a variety of tumor cell lines, the most active species having IC₅₀ values much lower than conventional nitrogen-containing bisphosphonates.

Full text of DOI:10.1021/jm700991k

J. Med. Chem. 2007, 50, 6067-6079
6067
Activity of Sulfonium Bisphosphonates on Tumor Cell Lines
Yonghui Zhang,† Michael P. Hudock,§ Kilannin Krysiak,† Rong Cao,§ Kyle Bergan,† Fenglin Yin,§ Annette Leon,§ and
Eric Oldfield*,†,§
Department of Chemistry, UniVersity of Illinois at Urbana-Champaign, 600 South Mathews AVenue, Urbana, Illinois 61801, and Center for
Biophysics and Computational Biology, UniVersity of Illinois at Urbana-Champaign, 607 South Mathews AVenue, Urbana, Illinois 61801
ReceiVed August 9, 2007
We investigated three series of sulfonium bisphosphonates for their activity in inhibiting the growth of
three human tumor cell lines. The first series consisted of 6 cyclic sulfonium bisphosphonates, the most
active species having an (average) IC50 of 89 µM. The second consisted of 10 phenylalkyl and phenylalkoxy
bisphosphonates, the most active species having an IC50 of 18 µM. The third series consisted of 17 n-alkyl
sulfonium bisphosphonates, the most active species having an IC50 of ∼240 nM. Three QSAR models showed
that the experimental cell growth inhibition results could be well predicted. We also determined the structures
of one sulfonium bisphosphonate bound to farnesyl diphosphate synthase, finding that it binds exclusively
to the dimethylallyl diphosphate binding site. These results are of interest since they show that sulfonium
bisphosphonates can have potent activity against a variety of tumor cell lines, the most active species having
IC50 values much lower than conventional nitrogen-containing bisphosphonates.
Introduction
of ATP,17 such as 5: which act (in osteoclasts) by inhibiting
Bisphosphonates are an important class of drug molecules,
used to treat a variety of bone resorption diseases, such as
osteoporosis, Paget’s disease, and hypercalcemia due to malig-
nancy. In addition, some bisphosphonates have direct activity
against cancer cells, both in Vitro1-5 and in ViVo,4,6-8 while
others have antiparasitic9 and antibacterial activity.10 Bisphos-
phonates can also activate γδ T cells (containing the Vγ2Vδ2
T cell receptor) of the immune system, and such activated T
cells have both antitumor11-13 as well as antibacterial activity.14
There is, therefore, interest in the further development of
bisphosphonates, for a variety of indications. The early develop-
ment of bisphosphonates has been somewhat unusual in that it
did not follow the current paradigm of target identification f
drug discovery. Indeed, as noted by Reszka and Rodan,15
bisphosphonates were used to treat bone resorption diseases for
over 20 years without understanding their mechanism of action.
Their discovery came about in early work,16 in which it was
shown that the simple molecule (or ion) diphosphate (1) had
an effect on hydroxyapatite formation and stability, but the
hydrolytic instability of diphosphate prevents its use as a bone
antiresorption drug.
the mitochondrial ADP/ATP translocase,18 leading to osteo-
clastic inactivity and, potentially, apoptosis. The doses of each
of these drugs required for efficacy are, however, quite high,
which led to the development of “second generation” bispho-
sphonates, such as pamidronate (6), alendronate (7), and
ibandronate (8). These species all contain nitrogen and are
generally referred to as NBPsa or nitrogen-containing bispho-
sphonates. Their mechanism of action differs from that of the
non-nitrogen containing bisphosphonates (such as 2, 3) and
involves inhibition of the enzyme farnesyl diphosphate
synthase15,19-22 (FPPS, EC 2.5.1.10). This enzyme is involved
in the formation of farnesyl diphosphate (FPP), which is used
by cells in protein prenylation, heme a, dolichol, and ubiquinone
biosynthesis. In more recent work, aromatic-containing NBPs
such as risedronate (9), zoledronate (10), and minodronate (11)
have been developed, although minodronate is not currently
marketed.
Consequently, nonhydrolyzable analogues of diphosphate,
such as the methylene bisphosphonates clodronate (2), etidronate
(3), and tiludronate (4), were developed, and these represent
the so-called “first generation” bisphosphonates, used primarily
to treat osteoporosis, although their molecular targets (which
are different to those of modern bisphosphonates) were not
known at the time of their introduction. Much later, these
compounds were shown to be converted into cytotoxic analogues
These aromatic NBPs are the most active bisphosphonates
in bone resorption, have IC50 values of ∼2 nM versus the human
FPPS target, and are generally referred to as “third generation”
* Corresponding author. Phone: 217-333-3374. Fax: 217-244-0997.
E-mail: eo@chad.scs.uiuc.edu.
a Abbreviations: FPP, farnesyl diphosphate; FPPS, farnesyl diphosphate
synthase; IPP, isopentenyl diphosphate; NBP, nitrogen-containing bispho-
sphonate; NNBP, non-nitrogen-containing bisphosphonate; QSAR, quantita-
tive structure-activity relationship.
† Department of Chemistry.
§ Center for Biophysics and Computational Biology.
10.1021/jm700991k CCC: $37.00 © 2007 American Chemical Society
Published on Web 10/27/2007

6068 Journal of Medicinal Chemistry, 2007, Vol. 50, No. 24
Zhang et al.
phate synthase, binding to the allylic (dimethylallyl diphosphate
or geranyl diphosphate) substrate binding site.25-29 The bis-
phosphonate moieties undergo strong electrostatic interactions
with three Mg2+ (which bind to a highly conserved DDXXD
motif in the active site), while the cationic center is stabilized
by the more diffuse electrostatic potential provided by a network
of electronegative groups, with additional stabilization in some
cases being provided via hydrophobic/van der Waals attractive
interactions with hydrophobic amino acid side chains in the
distal part of the allylic binding site. The most potent NBPs in
tumor cell growth inhibition have been shown to be zoledronate
(10) and minodronate (11), which we find have mean IC50 values
of 15 ((7.3, N ) 63) and 7.9 ((5.1, N ) 8) µM, respectively,
against three human tumor cell lines (NCI-H460 lung, MCF-7
breast, and SF-268 CNS), in good accord with the results
obtained by other workers on a broad array of cell lines.
bisphosphonates. As with the other NBPs, their primary
mechanism of action is thought to be in inhibiting FPPS,
resulting in decreased protein (Rho, Ras, Rap 1A) prenyla-
tion,23,24 and in most cases, enzyme inhibition also leads to
formation of a strongly apoptotic species, the isopentenyl ester
of ATP (12, ApppI),18 produced from the isopentenyl diphos-
phate which accumulates.
From a structure/function perspective, we proposed early on
that NBPs might be efficacious simply because, in their
protonated-nitrogen forms, they would closely resemble the
likely cationic transition state/reactive intermediate20 formed
during the FPPS catalytic cycle. For example, the protonated
form of ibandronate (8) might mimic the carbocation formed
from geranyl diphosphate (13):
The question then arises as to how it might be possible to
enhance activity in cell-based assays. There would seem to be
at least four possibilities. First, it might be possible to “mask”
the highly hydrophilic phosphonate groups by esterification,32
and indeed in some cases this is possible, although in general
this approach is synthetically challenging and also requires that
the protecting group be cleaved in the cell. A second strategy
is to enhance the lipophilicity of the cation by varying the nature
of the cationic center, e.g., N+H2 f N+HMe f S+Me f P+-
Me2. A third approach might be to vary the actual locus of the
cationic center. As can be seen in the Introduction, the earlier
generation NBPs (pamidronate, alendronate) have primary
ammonium charge centers at γ and δ positions in the side chain,
but these compounds have relatively poor activity (IC50 values
of ∼200-300 µM) in tumor cell growth inhibition assays. The
more potent species, zoledronate and minodronate, have more
delocalized charges, and in the case of zoledronate, one of the
likely cationic centers is at the â-position. In previous quantita-
tive structure-activity relationship (QSAR) investigations,33,34
our results suggested the desirability of increasing charge at
this position in order to enhance activity. It might therefore be
desirable to introduce a fixed charge at this site, although
investigating the effects of varying the number of CH2 group
spacers attached to the bisphosphonate backbone would also
seem prudent, since the above discussion does not absolutely
rule out optimum activity with a more distal charge center. Of
course, the introduction of a fixed charge might further limit
the already low bioavailability of a bisphosphonate (as, most
likely, with pamidronate and alendronate), although this effect
might well be offset by enhanced lipophilicity. And finally, it
might be possible to improve activity by varying the actual
nature of the side-chain substitution pattern.
a view now supported by a wide variety of X-ray crystal-
lographic studies,25-27 as well as by NMR spectroscopy,28 the
latter technique providing direct evidence for the presence of
charged side chains for NBPs bound to FPPS. This then leads
to the idea that it might be worth investigating other cationic
species, such as sulfonium, phosphonium, pyridinium, and
guanidinium bisphosphonates, since some might have improved
activity or other beneficial properties (e.g., improved lipophi-
licity). Here, we report the results of an initial investigation of
sulfonium bisphosphonates, chosen for investigation at least in
part because the S-Me sulfonium fragment occurs in common
metabolites such as S-adenosylmethionine,29 as well as in drugs
such as suplatast.30,31 We synthesized and tested a structurally
diverse range of 33 sulfonium bisphosphonates, determined their
activities in three tumor cell lines, and used QSAR methods to
analyze the results obtained. We also determined the X-ray
crystallographic structure of one sulfonium bisphosphonate,
bound to FPPS, and used computational docking to see how
others might bind.
We thus decided to synthesize and test three series of
compounds containing (1) cyclic sulfonium side chains, (2) aryl-
substituted side chains analogous to those reported by Widler
et al.,35 and (3) n-alkyl side chains. The general synthetic routes
to produce the sulfonium bisphosphonates were as follows, and
full details are given in the Experimental Section.
Dialkyl sulfides reacted smoothly with bromoacetic acid at
room-temperature resulting in the corresponding thiobetaine;
phosphorylation using PCl3-H3PO3-pyridine afforded the
â-sulfonium bisphosphonates (Scheme 1). γ-Sulfonium bispho-
sphonates were produced in a similar manner, except that the
carboxylic acid precursors were prepared via addition of alkyl
sulfides to acrylic acid (Scheme 2).
Results and Discussion
General Strategies. There is now strong evidence that all
of the commercially available nitrogen-containing bisphospho-
nates act, primarily, by inhibiting the enzyme farnesyl diphos-
The synthesis of deoxy-sulfonium bisphosphonates (lacking
the 1-OH group) was achieved in high yield via addition of

Sulfonium Bisphosphonates
Journal of Medicinal Chemistry, 2007, Vol. 50, No. 24 6069
Scheme 1a
although the differences between the five and six ring-atom-
containing species were generally small. The presence of an H
as opposed to an OH group at C1 (on the bisphosphonate
backbone) appeared to make little difference to activity. The
relative lack of activity of these compounds was on the one
hand disappointing; however, on the other hand, the observation
that the most active compound was about twice as active as
pamidronate and alendronate was encouraging since it clearly
demonstrates that sulfonium bisphosphonates, containing a
cationic center at the â-position, can be quite active and indeed
the activity of 14 is some three times greater than that of the
analogous NBP 47, which has an IC50 of 294 µM. As can be
seen in Table 1 and Figure 1, optimum activity is obtained with
a single CH2-group spacer between the cationic charge center
and the bisphosphonate C1 carbon, with activity falling off upon
the addition of more CH2 groups. For example, in the 1-OH
series: n ) 1, 15, IC50 ) 127 µM; n ) 2, 16, IC50 ) 158 µM;
n ) 3, 19 (>400 µM), no detectable activity. This result is
interesting in that with the nitrogen-containing bisphosphonates,
alendronate has n ) 3 CH2 groups and is more active (in bone
resorption and FPPS inhibition) than is pamidronate (n ) 2)
but is consistent with our previous observation of the importance
of a cationic feature at the â-position.33,34
a Reagents and conditions: (i) BrCH2COOH, acetone, rt; (ii) PCl3, H3PO3,
pyridine-HCl.
Scheme 2a
a Reagents and conditions: (i) acrylic acid, HCl, acetone, rt; (ii) PCl3,
H3PO3, pyridine-HCl.
Scheme 3
(b) Phenylalkyl and Phenyloxyalkyl Bisphosphonates. To
try to improve activity, we reasoned that it might be desirable
to remove the structural constraints imposed by the ring
structure, since it might be that chair/boat or skew boat
conformers would be a poor steric match to the FPPS binding
site, since they are unlike the planar features found in the most
potent bisphosphonates, zoledronate, and minodronate, which
are particularly potent FPPS/bone resorption/tumor cell growth
inhibitors. However, another series of NBP compounds devel-
oped previously by Novartis35 have almost equipotent activity
to that of zoledronate. For example, compound 48 has about
alkyl sulfides to vinylidene-1,1-diphosphonic acid in trifluoro-
acetic acid (Scheme 3).
Qualitative Structure-Activity Relationship Results. We
synthesized 6 thiacycloalkyl, 10 aryl (or aryloxy) alkyl, and 17
n-alkyl bisphosphonates: structures are shown in Figures 1-3,
and elemental analysis results are in Supporting Table S1. All
compounds were then tested for their ability to inhibit the growth
of three tumor cell lines: NCI-H460, MCF-7, and SF-268, and
representative growth inhibition results for the most active
species in each category are shown in Figure 4A-C (for NCI-
H460 cells), together with, for comparison, results for zoledr-
onate (Figure 4D) as a control. On average, the fitted r2 value
(over all compounds and all cell lines) was r2 ) 0.94. All IC50
values (as well as pIC50 values, pIC50 ) -log[IC50 (M)]) are
shown in Table 1. We find a good correlation between all three
data sets with r2 values of 0.90, 0.94, and 0.98 (shown in the
correlation matrix in Figure S1 in the Supporting Information)
for MCF-7, NCI-H460, and SF-268. For simplicity, in what
follows we discuss, in detail, solely data for NCI-H460, based
on the results of the QSAR investigations discussed below,
which show that the NCI-H460 predictions are slightly better
than those of MCF-7, but worse than SF-268, consistent with
the pattern of r2 values in the experimental results noted above.
(a) Thiacycloalkyl Bisphosphonates. In the first set of
compounds, Figure 1, we investigated a series of 6 thiacycloalkyl
sulfonium (and one NBP, 47) bisphosphonates, as exemplified
by the most active species, 14, shown below: The logic here
30% the activity of zoledronate,35 so we reasoned that sulfonium
analogues such as 26 (similar in overall length to 15) might
also have good activity. Interestingly, as can be seen in Table
1 (and Figure 2), 26 is actually not the most active species;
rather, the compound (20) with a single bridging CH2 group is
the most active species. This is the same trend as seen with the
thiacycloalkyl species in which 15, containing a single bridging
group, is more active than homologous compounds containing
two and three CH2 backbone spacer groups, consistent again
with our previous QSAR observations.33 The IC50 for 20 is ∼14
µM, essentially the same as that found for zoledronate (15 µM),
an encouraging improvement. More systematically, the activity
of 20 decreases from 14 µM to 66 µM on addition of a second
CH2 group (25), and the activity of 21 also decreases from 17
µM to 151 µM (26) on addition of the second bridging CH2
group. The presence of an OH group (24 f 20) also appears to
confer activity in all three tumor cell lines, by a factor of ∼3-
4. The least active compounds generally contain shorter side
chains, with activities in the ∼100 µM f ∼300 µM range. An
was to try to mimic to some degree the five-membered ring
and â-position charge center found in zoledronate. The results
obtained with the thiacycloalkyl bisphosphonates showed activ-
ity generally similar to that obtained with the second generation,
nitrogen-containing bisphosphonates pamidronate and alendr-
onate, with IC50 values in the range ∼100 to >400 µM, Table
1. The most active compounds contained five-membered rings,

6070 Journal of Medicinal Chemistry, 2007, Vol. 50, No. 24
Zhang et al.
Figure 1. Structures of thiacycloalkyl bisphosphonates, activity decreasing left to right.
Figure 2. Structures of phenylalkyl and phenyloxyalkyl bisphosphonates, rank ordered by activity.
Figure 3. Structures of n-alkylsulfonium bisphosphonates, rank ordered by activity.
exception is that the addition of a polar moiety to the aryl group
(an attempt to design a compound which might bind to His93
in the active site of T. brucei FPPS) also abrogated activity,
presumably because of unfavorable interactions with more
hydrophobic active site residues.
The activities of the compounds having the cationic center
in the â-position and with longer side chains are all in the 50
µM range, and the two most active species have activities of
∼14 µM (20) and ∼17 µM (21), about the same as zoledronate.
This encouraged us to further investigate the effects of chain
length on activity, focusing on n-alkyl sulfonium species, since
these are very readily prepared from commercially available
materials and provide a simple method with which to deduce
the optimum overall chain length.
(c) n-Alkylsulfonium Bisphosphonates. We prepared a
series of n-alkylsulfonium bisphosphonates (Figure 3) and
determined their activities in tumor cell growth inhibition (Table
1). The most active compound (30) (NCI-H460) had an IC50 of
210 nM, a factor of ∼70× more active than zoledronate. All of
the most active n-alkyl sulfonium bisphosphonates (IC50 ∼200
nM to 8 µM) contained a single CH2 group spacer, the most
active dimethylene group-containing species (42) having an IC50
of ∼95 µM (Table 1), basically the same pattern of activity
observed for the phenyl-containing species described above.
Quantitative Structure-Activity Relationships. On the
basis of the results discussed above, it seemed clear that there
is a relationship between the activity and the overall length of
the inhibitor, shown graphically in Figure 4E in which we plot
pIC50 as a function of the molecular diameter36 for all molecules
investigated (Figures 1-3). Tumor cell growth inhibition is
maximal with a ∼15 Å chain length. We see the same trend
when plotting the FPPS pIC50 against molecular diameter
(Figure 4F, Supporting Table S2). This is not unexpected, since
FPPS is a known target for bisphosphonates and likely a major
target for sulfonium bisphosphonates. To obtain a more pictorial
view of the results discussed above, we produced a common
feature pharmacophore model using the Molecular Operating
Environment (MOE)37 program. The pharmacophore (Figure 5)
shows the importance of distal hydrophobic (green) and
aromatic (pink) features, together with a cationic (blue) feature,
centered as anticipated on the â position, in addition to the
presence of two acceptor/anionic/metal binding sites, the bis-
phosphonate backbone (yellow, Figure 5). These observations
correlate well with the features noted previously, but based on

Sulfonium Bisphosphonates
Journal of Medicinal Chemistry, 2007, Vol. 50, No. 24 6071
Figure 4. Representative dose-response curves and molecular diameter-activity scatter plots. (A-D) Dose response curve fitting for most active
compounds from each group: (A) 14, 80 µM; (B) 20, 9.4 µM; (C) 30, 262 nM; (D) zoledronate control, 15 µM. (E) Tumor cell growth inhibition
activity as a function of molecular diameter; most active compounds are ∼15 Å long. (F) FPPS inhibition as a function of molecular diameter; most
active compounds again have a ∼15 Å length (dotted lines are to “guide the eye”).
Table 1. Tumor Cell Growth Inhibition by Sulfonium Bisphosphonates
NCI-H460 activity MCF-7
SF-268
IC50 (µM)
class
compda IC50 (µM)
IC50 (µM)
thiacycloalkyl
14
82.7
127
158
207
96
89.1
197
304
517
87.8
101
129
240
431
15b
16b
17b
18b
19
359
>400 mM
>400 mM
>400 mM
phenylalkyl and
phenyloxyalkyl-
sulfonium
20
13.7
17.7
10.7
Figure 5. Common feature pharmacophore. Pharmacophore features
are superimposed on the alignment of all sulfonium species. Yellow
sphere, combined feature of acceptor, anion, and metal ligator; blue
sphere, cation; green spheres, hydrophobic; magenta sphere, aromatic.
21
22
23
24
25
26
27
28b
29
16.8
23.5
48.1
54.4
65.8
151
94.2
21.2
45.5
58.1
106
184
251
337
>400 mM
0.24
0.62
10.9
2.8
14.2
0.96
1.43
1.12
21.3
32
161
241
258
199
34.5
17.9
43.9
48.6
65.9
171
In the first, we constructed a molecular descriptor-based
model using the program MOE37 in combination with the
AutoQuaSAR38 module. To provide sufficient granularity,
subdivided van der Waals surface area descriptors39 (as imple-
mented in MOE) were utilized in the molar refractivity and
positive charge terms, as opposed to whole molecule-based
descriptors (see Supporting Methods for full details). The final
cross-validated (NCI-H460) model yielded r2 ) 0.84, q2 ) 0.77,
F ) 74.3, n ) 25. Training and test set predictions are shown
in Figure 6A and Table 2, with similar results for SF-268 and
MCF-7 shown in the Supporting Figure S2 and Tables S3-S7.
A second analysis was performed using the HQSAR40 module
in Sybyl 7.3. In this method, molecular holograms (fingerprints)
were used to encode structural information about each molecule.
These holograms were then utilized by a PLS algorithm within
Sybyl to produce a linear model, yielding r2 ) 0.76 and q2 )
0.635, Figure 6B and Supporting Figure S3 and Table S8.
Of course, while these statistics are not wholly unreasonable,
it is obvious from the results shown in Figure 6A,B that these
linear models are oversimplified. More specifically, they predict
that the longer chain species (30 and 31) are less active than
they actually are; plus they overestimate the activity of the
shorter chain and phenylalkyl species (24, 23, 46). We therefore
next used the comparative molecular similarity analysis (CoM-
183
243
198
264
>400 mM
0.21
0.32
0.48
3.4
>400 mM
0.29
0.37
0.88
3.1
n-alkylsulfonium 30
31
32
33
34
35
36
37
38
39
40
3.7
3.7
1.15
1.23
1.41
12.4
25.5
74.6
94.6
95.1
189
1.49
1.71
1.08
14.7
20.3
96.1
223
100
171
189
196
41b
42b
43
44
128
197
759
224
386
815
45b
46b
714
a Structures shown in Figures 1-3. b Highest concentration tested, 5 mM.
our previous work, was not expected to provide a highly
predictive model. We thus next consider three QSAR modeling
approaches.

6072 Journal of Medicinal Chemistry, 2007, Vol. 50, No. 24
Zhang et al.
Figure 6. Molecular descriptor, HQSAR, and CoMSIA results. (A) Molecular descriptor QSAR based on 2D and 3D alignment independent
descriptors showing q2 ) 0.77, r2 ) 0.76. Test and training set shown, training ) b, test ) O. (B) Hologram QSAR results based on 2D connectivity
information, q2 ) 0.63, r2 ) 0.75. (C) CoMSIA modeling results for NCI-H460 cell line and feature-based alignment, q2 ) 0.86, r2 ) 0.95, test
and training shown, training ) b, test ) O.
Table 2. Molecular Descriptor QSAR Results for NCI-H460 Tumor Cell Line.
contiguous
rotatable bonds
molar refractivity
(SMR VSA7)
Gasteiger charge
NCI-H460
pIC50 (M)
predicted
pIC50 (M)
CV predictedb
pIC50 (M)
compda
Weiner path
(VSA+2)
residual
30
31
32
35
37
36
20
38
21
22
39
23
24
25
14
40
41
15
16
43
17
45
28
18
46
14
12
16
22
18
20
6
10
7
9
10
8
7
7
3
4
5
2
3
2
2
7
3
1818
1353
2379
4718
3044
3821
1154
976
1360
1464
1261
1249
1056
1362
351
413
366
397
499
270
478
756
1129
595
110.91148
110.91148
110.91148
110.91148
110.91148
110.91148
77.585464
110.91148
77.585464
77.585464
110.91148
77.585464
77.585464
77.585464
45.894287
110.91148
110.91148
45.894287
45.894287
109.27663
45.894287
110.91148
58.534134
45.894287
110.91148
12.37335
12.37335
12.37335
12.37335
12.37335
12.37335
10.677375
12.37335
10.677375
12.37335
30.056063
12.37335
12.37335
30.056063
23.050724
10.677375
12.37335
21.35475
40.733437
0
6.7
6.5
6.3
5.9
5.9
5.9
4.9
4.9
4.8
4.6
4.6
4.3
4.3
4.2
4.1
4.1
4
3.9
3.8
3.7
3.7
3.7
3.6
3.4
3.1
5.9
5.6
6.1
6
6.1
6.1
4.4
5.3
4.5
5
4.6
4.9
4.7
4
4.2
3.9
4.2
3.9
3.7
3.6
3.8
4.1
3
5.7
5.4
6
1
1.1
0.3
-0.2
-0.4
-0.3
0.6
-0.5
0.3
-0.4
0
-0.6
-0.5
0.2
-0.2
0.3
-0.2
0
0.2
6.2
6.2
6.2
4.3
5.4
4.5
5.1
4.6
4.9
4.8
4
4.2
3.8
4.2
3.9
3.6
3.6
3.9
4.2
2.6
3.6
3.8
0.2
21.35475
30.056063
40.733437
40.733437
30.056063
-0.2
-0.5
1
-0.2
-0.7
3
5
3.6
3.6
516
a Structures shown in Figures 1-3. b Predicted values obtained from leave-one-out cross validation.
SIA) method to more explicitly model three-dimensional
structural information (Table 3). For CoMSIA, we used two
sets of alignments, and partial least-squares (PLS) to regress
against cell activity and field data. In the first alignment, we
used the maximum common substructure (MCS) alignment, and
in the second, we used an alignment constructed in the flexible
alignment module41 of MOE, which perceives common features
within the ligands (e.g., similar charge, hydrophobic area,
volume, etc.). In both cases, we used fully deprotonated
bisphosphonate groups, based on the results of 31P NMR
chemical shift and quantum chemical calculations reported
previously,28 and the observation that bisphosphonates bind three
Mg2+ in the FPPS active site.25-29 For alignment 1 (MCS, Figure
7A), we obtained: q2 (no. of components), r2, F-test, pIC50 error
values of 0.819 (4), 0.94, 80.18, 0.27 for NCI-H460; 0.61 (4),
0.91, 51.7, 0.33 for MCF-7; and 0.80 (4), 0.94, 79.0, 0.27 for
SF-268. Interestingly, we observed better performance from the
feature-based alignment (Figure 7B), with q2 (no. of compo-
nents), r2, F-test, error values of: 0.86 (4), 0.96, 131, 0.216 for
NCI-H460; 0.73 (4), 0.93, 61.8, 0.30 for MCF-7; and 0.85 (4),
0.96, 118, 0.22 for SF-268. We then performed five iterations
of a leave-two-out test/training set approach, observing (on
average for the feature alignment) a factor of 1.7× error between
predicted and experimental IC50 values in three cell lines.
Training and test set predictions are shown in Figure 6C and
Supporting Figure S4 with full details of the CoMSIA descrip-
tors given in the Supporting Information (Tables S9-S13).
Selected CoMSIA fields (for the NCI-H460 models, both
alignments) are shown in Figure 7C-F. Both sets of fields were
generally consistent, particularly in the steric (green, Figure
7C,D) and hydrophobic-favorable regions (yellow, Figure 7E,F).
In the MCS alignment, Figure 7C, a steric penalty feature
(yellow) near the location of the thiacycloalkyl ring, and
additionally, a hydrophobic penalty near the OH group at the
CR position were observed, consistent with the experimental
activities seen: reduced activity beyond an optimum chain
length (∼15 Å), reduced activity for thiacycloalkyl species
versus n-alkyl, and the overall most active compounds lack the
backbone OH group. A comparison of the experimental results
and computational predictions for each alignment and dataset
are shown graphically in Supporting Figure S1.
For further validation, a stability test was performed using
the progressive scrambling module42 in Sybyl 7.11, which
applies small, random perturbations to the activity data of
structurally similar compounds, allowing determination of the
sensitivity of a PLS model to these small systematic perturba-

Sulfonium Bisphosphonates
Journal of Medicinal Chemistry, 2007, Vol. 50, No. 24 6073
Table 3. CoMSIA Results for NCI-H460 Tumor Cell Line Using the Feature Alignment
CoMSIA pIC50 predictions
NCI-H460
experimental activity
test setsb
compda
IC50 (µM)
pIC50 (M)
training set
1
2
3
4
5
pred
residual
30
31
32
35
36
37
38
20
21
22
39
23
24
25
40
14
41
15
16
43
45
17
28
18
46
0.21
0.32
0.48
1.15
1.23
6.7
6.5
6.3
5.9
5.9
5.9
4.9
4.9
4.8
4.6
4.6
4.3
4.3
4.2
4.1
4.1
4.0
3.9
3.8
3.7
3.7
3.7
3.6
3.5
3.1
6.6
6.2
6.5
5.7
5.8
6.1
5.2
4.9
4.7
4.9
4.7
4.4
4.4
4.1
3.9
3.8
3.8
3.8
3.6
3.8
3.7
3.8
3.7
3.5
3.6
6.7
6.2
6.4
5.7
5.8
6.1
5.2
4.9
4.7
4.8
4.7
4.3
4.3
4.1
3.9
3.7
3.8
3.7
3.6
3.8
3.6
3.8
3.7
3.5
3.6
6.7
6.2
6.5
5.6
5.8
6.1
5.2
4.9
4.6
4.8
4.4
4.4
4.4
4.1
3.8
3.8
3.9
3.8
3.6
3.7
3.6
3.8
3.7
3.5
3.5
6.4
5.8
6.4
5.7
5.8
6.1
5.1
4.9
4.5
4.8
4.6
4.4
4.5
4.1
3.9
3.8
3.9
3.8
3.5
3.7
3.6
3.9
3.6
3.5
3.5
6.4
6.1
6.3
5.7
5.8
6.1
5.3
4.9
4.7
5.2
4.8
4.4
4.5
4.1
3.9
3.8
3.9
3.7
3.6
3.8
3.6
3.8
3.6
3.5
3.6
6.7
6.2
6.6
5.7
5.9
6.2
5.2
4.9
4.7
4.8
4.6
4.4
4.4
4.1
3.9
3.8
3.9
3.8
3.6
3.8
3.6
3.9
3.6
3.5
3.5
6.4
5.8
6.4
5.7
5.8
6.1
5.2
4.9
4.7
4.8
4.6
4.4
4.5
4.1
3.9
3.8
3.9
3.8
3.6
3.8
3.6
3.8
3.6
3.5
3.5
-0.28
-0.70
0.13
-0.24
-0.07
0.25
0.27
0.05
-0.08
0.17
0.03
0.09
0.24
-0.08
-0.26
-0.26
-0.12
-0.14
-0.25
0.08
1.41
12.30
13.80
16.60
23.44
25.70
47.86
54.95
66.07
74.13
83.18
95.50
125.89
158.49
190.55
199.53
208.93
245.47
354.81
758.58
-0.08
0.16
-0.01
0.06
0.42
q2
0.86
0.96
131.4
4
0.83
0.96
117.6
0.86
0.96
140.9
0.83
0.96
113.1
0.79
0.95
91.74
4
0.84
0.96
120.9
R2
F test
N
4
4
4
4
n
25
23
23
23
23
23
a Structures shown in Figures 1-3. b Bold values represent predicted activities of compounds that were not included in the training set.
Figure 7. Representative CoMSIA fields for MCS and feature-based alignments (NCI-H460 model). (A, C, E) MCS alignment and fields: (A)
MCS alignment, only heavy atoms displayed for clarity; (C) steric fields, green favorable, yellow disfavored; (E) hydrophobic fields, yellow favored,
gray disfavored. (B, D, F) feature-based alignment and fields; (B) feature-based alignment, only heavy atoms displayed for clarity; (D) steric fields,
green favorable, yellow disfavored; (F) hydrophobic fields, yellow favored, gray disfavored.
tions.43 The analysis yielded an effective slope of 1.1 for
NCI-H460 (MCF-7 ) 0.91, SF-268 ) 1.1), close to the ideal
value of 1, indicating good stability. Data randomization was
also performed to check that a model could not be the result
of a chance correlation. The randomization tests produced no
usable models (q2 < 0.1) among 10 iterations for each
alignment.
Sulfonium Bisphosphonate-FPPS Interactions. The results
discussed above are of general interest since we have produced
numerous bisphosphonates that, at least in cell growth inhibition
assays, are very active. This, of course, then leads to the question
as to how they might bind to their FPPS target. To answer this
question, we elected to determine the single-crystal X-ray
crystallographic structure of the simplest bisphosphonate, 43,

6074 Journal of Medicinal Chemistry, 2007, Vol. 50, No. 24
Zhang et al.
Table 4. Data Collection and Refinement Statistics for 15a Bound to Human FPPS and T. Brucei FPPS (PDB: 2OPN, 2OGD)
crystals
15a + human FPPS
15a + T. brucei FPPS
Data Collection
space group
P41212
C2
unit cell dimension (Å)
a, b, c
R, â, γ (°)
X-ray source
wavelength (Å)
resolution (Å)c
no. of reflections obsd
unique
completeness (%)
R-merge
I/σI
multiplicity
111.65, 111.65, 66.84
90, 90, 90
134.61, 118.37, 62.76
90, 112.158, 90
BNL-X8Cb
1.1
30-2.00 (2.07-2.00)
461,159
61,155 (6,046)
99.8 (99.5)
0.082 (0.538)
33.6
BNL-X29b
0.9791
30-2.70 (2.80-2.70)
135,612
12,031 (1,154)
99.0 (97.7)
0.107 (0.384)
30.0
11.3 (8.4)
7.5 (7.4)
Refinement Statistics
resolution range (Å)
R-work/R-free (%)
RMSD
30.0-2.70
27.24/29.79
30.0-2.00
20.70/24.12
bond lengths
bond angles
0.004
1.414
0.008
1.21
no. of atoms
protein
ligand
2,664
14
5,715
28
magnesium ion
3
6
solvent (water)
98
563
28.4
36.1
22.9
B average (Å2) of protein
B average (Å2) of solvents
B average (Å2) of ligands
(bisphosphonates, Mg2+, and PO4
48.9
55.2
67.6
3-
)
a 15 is 2-hydroxy-2,2-bis-phosphonoethyl)dimethylsulfonium. b Brookhaven National Laboratory. c Values in the parentheses are for the highest resolution
shell.
bound to human and T. brucei FPPS, the latter being of interest
(14, 20, and 30) to the 43 human FPPS structure (PDB: 2OPN,
minus the bisphosphonate ligand). The results obtained are
shown in Figures 8D-F and indicate that, in each case, the
bisphosphonate moiety binds to the 3Mg2+ cluster, while the
large side chains bind in the hydrophobic central channel in
the protein. For the phenylalkyl species, this binding pattern is
very similar to that seen crystallographically with 15 bound to
T. brucei FPPS (PDB: 2PlC), giving further confidence in the
predictions for the other compounds. Of course, on the basis of
these results alone, we cannot definitely state that FPPS is the
sole target for, in particular, the n-alkyl bisphosphonates, where
geranylgeranyl diphosphate synthase might also be targeted.
However, the most active n-alkyl bisphosphonate (30) does have
an IC50 of 0.26 µM against human FPPS (Table S2), supporting
FPPS as a “primary” target for these types of compounds.
as a target for antiinfective therapy. We anticipated, based on
the transition state/reactive intermediate model described above,
that this bisphosphonate would bind to the allylic binding site
and chelate to Mg2+, although given its small size, binding to
the IPP binding site would be hard to exclude. Bisphosphonate
43 has a 1 µM IC50 versus human FPPS (Table S2) and an IC50
in cell growth inhibition of ∼190 µM, to be compared (in the
same assay) with a 0.12 µM FPPS IC50 for zoledronate, which
as noted above, has a 15 µM value in cell growth inhibition.
The activity of 43 is, therefore, basically similar to that of
second-generation bisphosphonates, such as pamidronate. Full
details of protein expression, purification, and crystallization
were as described previously,28,44 and X-ray crystallographic
data collection and refinement statistics are given in Table 4.
43 binds to the allylic or DMAPP site in both human and T.
brucei FPPS, the latter structure having improved resolution
over that of the human FPPS, 2.07 Å versus 2.80 Å. The two
phosphonate groups interact with three Mg2+, Figure 8A,B, as
reported previously for many nitrogen-containing bisphosphon-
ates.25-28 Also of interest is the observation that the (cationic)
sulfonium charge center is in essentially the same location as
is N1 (the â-position) in zoledronate, bound to human FPPS,
shown superimposed on 43 in Figure 8B. To see if this binding
pose might be predicted computationally, we used the Glide45
program to dock 43 to the human and T. brucei structures
determined here (PDB: 2OPN, 2OGD) and to one other
structure, of 48 bound to T. brucei FPPS (PDB: 2PlC), in each
case, minus the bisphosphonate ligand. For the six pairwise
experimental/predicted structure comparisons (two experimental
structures, three predicted poses) there was a 1.0 ( 0.19 Å rmsd
between the experimental and predicted ligand heavy atom
positions, and one such comparison is shown in Figure 8C. This
gave some confidence in the method, so we next docked the
most active species in each of the three classes of compounds
Experimental Section
Synthetic Aspects. All reagents used were purchased from
Aldrich (Milwaukee, WI). The purities of all compounds produced
were routinely monitored by using 1H and 31P NMR spectroscopy
at 400 or 500 MHz on Varian (Palo Alto, CA) Unity spectrometers.
In some instances, absolute spin-count quantitative analyses using
an internal imidazole standard were performed. The elemental
analysis results for all new compounds are provided in the
Supporting Information (Table S1).
General Method A: Synthesis of 2-(Sulfonium-1-yl)acetic
Acid. A mixture of alkyl sulfide (1 mmol) and bromoacetic acid
(1.2 mmol) was stirred in anhydrous acetone (2 mL) under N2 at
room temperature overnight. The precipitate was washed with ether
and dried under vacuum to afford 2-(sulfonium-1-yl)acetic acid as
a white powder.
General Method B: Synthesis of 3-(Sulfonium-1-yl)propanoic
Acid. Concentrated HCl (2 mL) was added to a mixture of alkyl
sulfide (10 mmol) and acrylic acid (10 mmol) in acetone (5 mL) at
0 °C. The mixture was allowed to rise to room temperature and
stirred overnight. Upon evaporation of solvent, the residue was

Sulfonium Bisphosphonates
Journal of Medicinal Chemistry, 2007, Vol. 50, No. 24 6075
Figure 8. X-ray and docking results. (A) 43 bound to the allylic or DMAPP binding site of human FPPS (PDB: 2OPN); (B) 43 (from human
FPPS, PDB: 2OPN) structure superimposed on the zoledronate-FPPS structure (PDB: 1ZW5); (C) Docked pose of 43 obtained using Glide
superimposed onto 43 in human FPPS (PDB: 2OPN); (D-F) Docked poses of most active compounds from each category, neighboring Asp
residues labeled for clarity; (D) 14; (E) 20; (F) 30.
washed with ethyl acetate (2 × 5 mL) and then dried under vacuum
to afford the 3-(sulfonium-1-yl)propanoic acid as a syrup or white
powder.
tetrahydrothiopyran (1.02 g, 10 mmol) following General Methods
A and C (673 mg, 2.2 mmol, 22%). Anal. (C7H16O7P2S) C, H. 1H
NMR (400 MHz, D2O): δ 3.40-3.48 (m, 2H), 3.27-3.38 (m, 2H),
1.92-2.05 (m, 2H), 1.60-1.80 (m, 2H), 1.4-1.6 (dm, 2H, J )
6.5 Hz). 31P NMR (D2O): δ 15.50.
General Method C: Synthesis of Sulfonium Bisphosphonate
from a Carboxylic Acid. A mixture of sulfonium carboxylic acid
(10 mmol), H3PO3 (40 mmol), and pyridine-HCl (30 mmol) was
heated to 70 °C and melted. PCl3 (40 mmol) was added slowly
and the reaction mixture stirred at 70 °C for 4 h. H2O (5 mL) was
then cautiously added and the mixture refluxed for 1 h. Upon
filtration and evaporation, the residue was precipitated with
2-propanol, washed with acetone (5 × 5 mL), dried, and then further
purified by recrystallization from H2O/i-PrOH. In some cases, the
bisphosphonate was neutralized with NaOH and crystallized as the
sodium salt from H2O/EtOH.
General Method D: Syntheis of Deoxy-sulfonium Bisphos-
phonates (Scheme 3). A substituted sulfide (1.2 mmol) and
vinylidene-1,1-bisphosphonic acid (1 mmol) was refluxed overnight
in trifluoroacetic acid (3 mL). Upon removal of solvent, the residue
was triturated with acetone (3 mL), and the resulting white
suspension filtered and washed with acetone (2 × 2 mL), affording
pure 2-(sulfonium-1-yl)ethylidene-1,1-bisphosphonic acid as a white
powder.
1-Hydroxy-3-(tetrahydrothiopyranium-1-yl)propylidene-1,1-
bisphosphonic Acid (18). Compound 18 was prepared from
tetrahydrothiopyran (1.02 g, 10 mmol) following General Methods
B and C (560 mg, 1.7 mmol, 17%). Anal. (C8H18O7P2S·0.5H2O)
1
C, H. H NMR (400 MHz, D2O): δ 3.40-3.48 (m, 2H), 3.27-
3.38 (m, 2H), 2.1-2.22 (m, 2H), 1.92-2.05 (m, 2H), 1.60-1.75
(m, 2H), 1.4-1.6 (dm, 2H, J ) 6.5 Hz). 31P NMR (D2O): δ 17.80.
1-Hydroxy-4-(tetrahydrothiophenyium-1-yl)butylidene-1,1-
bisphosphonic Acid (19). Compound 19 was prepared from 1-(3-
carboxypropyl)tetrahydrothiophenium bromide (1.0 g, 3.9 mmol)
following General Method C (357 mg, 1.1 mmol, 28%). Anal.
1
(C8H18O7P2S·0.4H2O) C, H. H NMR (400 MHz, D2O): δ 3.30-
3.42 (m, 2H), 3.18-3.28 (m, 2H), 3.00-3.10 (m, 2H), 2.0-2.2
(m, 4H), 1.75-1.98 (m, 4H). 31P NMR (D2O): δ 19.88.
1-Hydroxy-2-(phenylpropyl, methyl sulfonium-1-yl)propy-
lidene-1,1-bisphosphonic Acid (20). Compound 20 was prepared
from methyl phenylpropyl sulfide (500 mg, 3.01 mmol) following
General Methods A and C (136 mg, 0.36 mmol, 12%). Anal.
2-(Tetrahydrothiophenyium-1-yl)ethylidene-1,1-bisphospho-
nic Acid (14). Compound 14 was prepared from tetrahy-
drothiophene (210 mg, 2.4 mmol) and vinylidene-1,1-diphosphonic
acid (380 mg, 2 mmol) following General Method D (420 mg, 1.50
1
(C12H20O7P2S·0.4H2O). H NMR (400 MHz, D2O): δ 7.05-7.20
(m, 5H), 3.41-3.65 (m, 2H), 3.0-3.11 (m, 1H), 3.15-3.24 (m,
1H), 2.75 (s, 3H), 2.60 (t, J ) 7.5 Hz), (1.85-2.00) (m, 2H). 31P
NMR (D2O): δ 18.00.
1
mmol, 75%). Anal. (C6H14O6P2S·0.25H2O) C, H. H NMR (400
MHz, D2O): δ 3.20-3.42 (m, 4H), 2.00-2.35 (m, 5H). 31P NMR
(D2O): δ 14.73.
1-Hydroxy-2-(methyl, 3-phenyloxypropylsulfonium-1-yl)ethyl-
idene-1,1-bisphosphonic Acid (21). Compound 21 was prepared
from methyl 3-phenyloxypropyl sulfide (1.1 g, 6 mmol) following
General Methods A and C (358 mg, 9 mmol, 15%). Anal.
1-Hydroxy-2-(tetrahydrothiophenyium-1-yl)ethylidene-1,1-
bisphosphonic Acid (15). Compound 15 was prepared from
tetrahydrothiophene (880 mg, 10 mmol) following General Methods
A and C (780 mg, 2.7 mmol, 27%). Anal. (C6H14O7P2S) C, H. 1H
NMR (400 MHz, D2O): δ 3.20-3.42 (m, 4H), 2.00-2.35 (m, 5H).
31P NMR (D2O): δ 15.01.
1
(C12H20O8P2S·0.5H2O) C, H. H NMR (400 MHz, D2O): δ 7.18
(t, J ) 7.5 Hz, 2H), 6.85-6.90 (m, 3H), 4.04 (t, J ) 6 Hz, 2H),
3.50-3.70 (m, 2H), 3.18-3.40 (m, 2H), 2.75 (s, 3H), 2.02-2.15
(m, 2H). 31P NMR (D2O): δ 15.30.
1-Hydroxy-3-(tetrahydrothiophenyium-1-yl)propylidene-1,1-
bisphosphonic Acid (16). Compound 16 was prepared from
tetrahydrothiophene (880 mg, 10 mmol) following General Methods
B and C (460 mg, 1.5 mmol, 15%). Anal. (C7H16O7P2S·0.25H2O)
2-(Methyl, phenypentylsulfonium-1-yl)ethylidene-1,1-bispho-
sphonic Acid (22). Compound 22 was prepared from methyl
phenylpentyl sulfide (466 mg, 2.4 mmol) and vinylidene-1,1-
diphosphonic acid (380 mg, 2 mmol) following General Method
D (453 mg, 1.16 mmol, 58%). Anal. (C14H24O6P2S·0.5H2O) C, H.
1H NMR (400 MHz, D2O): δ 7.00-7.21 (m, 5H), 3.20-3.35 (m,
2H), 3.10-3.20 (m, 1H), 2.95-3.00 (m, 1H), 2.90-3.00 (m, 1H),
2.63 (s, 3H), 2.5-2.60 (t, J ) 7.5 Hz, 2H), 1.90-2.05 (m, 1H),
1
C, H. H NMR (400 MHz, D2O): δ 3.00-3.48 (m, 4H), 1.88-
2.15 (m, 4H). 31P NMR (D2O): δ 18.01.
1-Hydroxy-3-(tetrahydrothiopyranium-1-yl)ethylidene-1,1-
bisphosphonic Acid (17). Compound 17 was prepared from

6076 Journal of Medicinal Chemistry, 2007, Vol. 50, No. 24
Zhang et al.
1.60-1.70 (m, 2H), 1.41-1.55 (m, 2H), 1.20-1.35 (m, 2H). 31P
NMR (D2O): δ 15.01 (d, J ) 17.4 Hz).
Anal. (C13H30O6P2S•0.3H2O) C, H. 1H NMR (400 MHz, D2O): δ
3.30-3.50 (m, 2H), 3.10-3.20 (m, 1H), 2.90-3.00 (m, 1H), 2.63
(s, 3H), 1.90-2.05 (m, 1H), 1.62-1.70 (m, 2H), 1.22-1.34 (m,
2H), 1.00-1.20 (m, 14H), 0.69 (t, J ) 6.5 Hz, 3H). 31P NMR
(D2O): δ 15.01 (dd, J ) 4.4, 19.7 Hz).
2-(Methyl, tetradecylsulfonium-1-yl)ethylidene-1,1-bisphos-
phonic Acid (32). Compound 32 was prepared from methyl
tetradecyl sulfide (586 mg, 2.4 mmol) and vinylidene-1,1-diphos-
phonic acid (380 mg, 2 mmol) following General Method D (590
mg, 1.34 mmol, 67%). Anal. (C17H38O6P2S‚0.5H2O) C, H. 1H NMR
(400 MHz, D2O): δ 3.30-3.50 (m, 2H), 3.10-3.20 (m, 1H), 2.90-
3.00 (m, 1H), 2.66 (s, 3H), 1.90-2.05 (m, 1H), 1.60-1.70 (m,
2H), 1.22-1.34 (m, 2H), 1.02-1.20 (m, 20H), 0.82 (t, J ) 6.5
Hz, 3H). 31P NMR (D2O): δ 15.51 (d, J 15.7 Hz).
2-(Methyl, phenylbutylsulfonium-1-yl)ethylidene-1,1-bispho-
sphonic Acid (23). Compound 23 was prepared from methyl
phenylbutyl sulfide (466 mg, 2.4 mmol) and vinylidene-1,1-
diphosphonic acid (380 mg, 2 mmol) following General Method
D (524 mg, 1.4 mmol, 70%). Anal. (C13H22O6P2S·0.3H2O) C, H.
1H NMR (400 MHz, D2O): δ 7.00-7.22 (m, 5H), 3.20-3.40 (m,
2H), 3.10-3.20 (m, 1H), 2.95-3.00 (m, 1H), 2.90-3.00 (m, 1H),
2.63 (s, 3H), 2.5-2.60 (t, J ) 7.5 Hz, 2H), 1.90-2.05 (m, 1H),
1.60-1.70 (m, 4H). 31P NMR (D2O): δ 15.01 (d, J ) 16.8 Hz).
2-(Methyl, phenylpropylsulfonium-1-yl)ethylidene-1,1-bispho-
sphonic Acid (24). Compound 24 was prepared from methyl
phenylpropyl methyl sulfide (400 mg, 2.4 mmol) and vinylidene-
1,1-diphosphonic acid (380 mg, 2 mmol) following General Method
D (414 mg, 1.14 mmol, 57%). Anal. (C12H20O6P2S·0.5H2O) C, H.
1H NMR (400 MHz, D2O): δ 7.01-7.25 (m, 5H), 3.27-3.53 (m,
2H), 3.13-5-3.20 (m, 1H), 2.90-3.03 (m, 1H), 2.61 (s, 3H), 1.95
(t, J ) 7.5 Hz, 2H). 31P NMR (D2O): δ 15.18.
2-(Methyl, tridecylsulfonium-1-yl)ethylidene-1,1-bisphospho-
nic Acid (33). Compound 33 was prepared from methyl tridecyl
sulfide (550 mg, 2.4 mmol) and vinylidene-1,1-diphosphonic acid
(380 mg, 2 mmol) following General Method D (517 mg, 1.18
1
mmol, 59%). Anal. (C16H36O6P2S‚1.2H2O) C, H. H NMR (400
MHz, D2O): δ 3.30-3.51 (m, 2H), 3.10-3.20 (m, 1H), 2.91-
3.00 (m, 1H), 2.75 (s, 3H), 1.91-2.00 (m, 1H), 1.60-1.70 (m,
2H), 1.25-1.35 (m, 2H), 1.02-1.20 (m, 18H), 0.84 (t, J ) 6.5
Hz, 3H). 31P NMR (D2O): δ 15.41 (d, J 15.7 Hz).
2-(Methyl, undecylsulfonium-1-yl)ethylidene-1,1-bisphospho-
nic Acid (34). Compound 34 was prepared from methyl undecyl
sulfide (485 mg, 2.4 mmol) and vinylidene-1,1-diphosphonic acid
(380 mg, 2 mmol) following General Method D (579 mg, 1.43
mmol, 71%). Anal. (C14H32O6P2S‚H2O) C, H. 1H NMR (400 MHz,
D2O): δ 3.30-3.51 (m, 2H), 3.10-3.20 (m, 1H), 2.91-3.00 (m,
1H), 2.75 (s, 3H), 1.91-2.00 (m, 1H), 1.60-1.70 (m, 2H), 1.25-
1.35 (m, 2H), 1.02-1.20 (m, 14H), 0.80 (t, J ) 6.5 Hz, 3H). 31P
NMR (D2O): δ 15.41 (d, J 15.7 Hz).
2-(Methyl, licosylsulfonium-1-yl)ethylidene-1,1-bisphosphonic
Acid (35). Compound 35 was prepared from methyl icosyl sulfide
(780 mg, 2.4 mmol) and vinylidene-1,1-diphosphonic acid (380 mg,
2 mmol) following General Method D (892 mg, 1.62 mmol, 81%).
Anal. (C23H49NaO6P2S‚0.7H2O) C, H. 1H NMR (400 MHz, D2O):
δ 3.31-3.53 (m, 2H), 3.12-3.20 (m, 1H), 2.90-3.00 (m, 1H), 2.69
(s, 3H), 1.90-2.00 (m, 1H), 1.60-1.70 (m, 2H), 1.22-1.34 (m,
2H), 1.02-1.20 (m, 32H), 0.78 (t, J ) 6.5 Hz, 3H). 31P NMR
(D2O): δ 15.41 (d, J 15.7 Hz).
2-(Methyl, octadecylsulfonium-1-yl)ethylidene-1,1-bisphos-
phonic Acid (36). Compound 36 was prepared from methyl
octadecyl sulfide (720 mg, 2.4 mmol) and vinylidene-1,1-diphos-
phonic acid (380 mg, 2 mmol) following General Method D (645
mg, 1.18 mmol, 59%). Anal. (C21H44Na2O6P2S‚H2O) C, H. 1H NMR
(400 MHz, D2O): δ 3.31-3.53 (m, 2H), 3.12-3.20 (m, 1H), 2.90-
3.00 (m, 1H), 2.69 (s, 3H), 1.90-2.00 (m, 1H), 1.60-1.70 (m,
2H), 1.22-1.34 (m, 2H), 1.02-1.20 (m, 28H), 0.82 (t, J ) 6.5
Hz, 3H). 31P NMR (D2O): δ 15.41 (d, J 15.7 Hz).
1-Hydroxy-3-(phenylpropyl, methylsulfonium-1-yl)propylidene-
1,1-bisphosphonic Acid (25). Compound 25 was prepared from
phenylpropyl sulfide (1.0 g, 6 mmol) following General Methods
B and C (283 mg, 0.72 mmol, 12%). Anal. (C13H22O7P2S‚0.5H2O).
1H NMR (400 MHz, D2O): δ 7.10-7.25 (m, 5H), 2.95-3.51 (m,
4H), 2.7-2.90 (m, 5H), 2.0-2.22 (m, 2H), 1.70-1.95 (m, 2H).
31P NMR (D2O): δ 18.00.
1-Hydroxy-3-(methyl, phenyloxypropylsulfonium-1-yl)pro-
pylidene-1,1-bisphosphonic Acid (26). Compound 26 was prepared
from methyl phenyloxypropyl sulfide (0.6 g, 3.3 mmol) following
General Methods B and C. (189 mg, 0.43 mmol, 13%). Anal.
(C13H13NaO9P2S) C, H, N. 1H NMR (400 MHz, D2O): δ 6.8-7.2
(m, 5H), 4.04 (t, J ) 6 Hz, 2H), 3.50-3.70 (m, 2H), 3.18-3.40
(m, 2H), 2.72 (s, 3H), 2.2-2.3 (m, 2H). 31P NMR (D2O): δ 17.89.
1-Hydroxy-3-(phenylethyl, methylsulfonium-1-yl)propylidene-
1,1-bisphosphonic Acid (27). Compound 27 was prepared from
methyl phenylethyl sulfide (0.69 g, 4.5 mmol) following General
Methods B and C (166 mg, 0.405 mmol, 9%). Anal. (C12H19-
1
NaO7P2S‚1.3H2O). H NMR (400 MHz, D2O): δ 7.15-7.27 (m,
5H), 3.30-3.61 (m, 4H), 3.0 (t, J ) 8.5 Hz, 2H), 2.65 (s, 3H),
2.2-2.3 (m, 2H). 31P NMR (D2O): δ 17.75.
1-Hydroxy-3-(3-phenyltetrahydrothiophenium-1-yl)propy-
lidene-1,1-bisphosphonic Acid (28). Compound 28 was prepared
from 3-phenyltetrahydrothiophene (0.75, 4.57 mmol) following
General Methods B and C (205 mg, 4.6 mmol, 10%). Anal.
1
(C13H19O6NaP2S‚2.5H2O). H NMR (400 MHz, D2O): δ 7.12-
7.3 (m, 5H), 3.45 (t, J ) 7.8 Hz, 2H), 2.7 (s, 6H), 3.00-3.10 (m,
2H), 2.0-2.2 (m, 4H), 1.75-1.98 (m, 4H). 31P NMR (D2O): δ
18.05.
2-[Methyl 4-(3-carboxypropyl)phenybutylsulfonium-1-yl]eth-
ylidene-1,1-bisphosphonic Acid (29). Compound 29 was prepared
from methyl 4-(3-carboxypropyl)phenybutyl sulfide (0.75 g, 2.8
mmol) and vinylidene-1,1-diphosphonic acid (0.44 g, 2.3 mmol)
following General Method D (0.75 g, 1.65 mmol, 72%). Anal.
(C17H28O8P2S‚0.5H2O) C, H. 1H NMR (400 MHz, D2O): δ 77.21
(s, 4H), 3.20-3.40 (m, 2H), 3.10-3.20 (m, 1H), 2.95-3.00 (m,
1H), 2.63 (s, 3H), 2.45-2.60 (m, 4H), 2.06 (t, J ) 6.9 Hz), 1.95-
2.0 (m, 1H), 1.55-1.75 (m, 6H). 31P NMR (D2O): δ 15.01 (d, J
) 17.4 Hz).
2-(Methyl, hexadecylsulfonium-1-yl)ethylidene-1,1-bisphos-
phonic Acid (37). Compound 37 was prepared from methyl
hexadecyl sulfide (650 mg, 2.4 mmol) and vinylidene-1,1-diphos-
phonic acid (380 mg, 2 mmol) following General Method D (630
mg, 1.31 mmol, 65%). Anal. (C19H42O6P2S‚1.2H2O) C, H. 1H NMR
(400 MHz, D2O): δ 3.31-3.53 (m, 2H), 3.12-3.20 (m, 1H), 2.90-
3.00 (m, 1H), 2.69 (s, 3H), 1.90-2.00 (m, 1H), 1.60-1.70 (m,
2H), 1.22-1.34 (m, 2H), 1.02-1.20 (m, 24H), 0.82 (t, J ) 6.5
Hz, 3H). 31P NMR (D2O): δ 15.41 (d, J 15.7 Hz).
2-(Methyl, dodecylsulfonium-1-yl)ethylidene-1,1-bisphospho-
nic Acid (30). Compound 30 was prepared from methyl dodecyl
sulfide (0.52 g, 2.4 mmol)and vinylidene-1,1-diphosphonic acid
(380 mg, 2 mmol) following General Method D (554 mg, 1.17
2-(Methyl, octylsulfonium-1-yl)ethylidene-1,1-bisphosphonic
Acid (38). Compound 38 was prepared from methyl octyl sulfide
(384 mg, 2.4 mmol) and vinylidene-1,1-diphosphonic acid (380 mg,
2 mmol) following General Method D (480 mg, 1.38 mmol, 69%).
1
mmol, 58%). Anal. (C15H34O6P2S‚0.3H2O) C, H. H NMR (400
1
MHz, D2O): δ 3.30-3.50 (m, 2H), 3.10-3.20 (m, 1H), 2.90-
3.00 (m, 1H), 2.66 (s, 3H), 1.90-2.05 (m, 1H), 1.60-1.70 (m,
2H), 1.22-1.34 (m, 2H), 1.02-1.20 (m, 16H), 0.73 (t, J ) 6.5
Hz, 3H). 31P NMR (D2O): δ 15.02 (dd, J ) 4.4, 19.7 Hz).
2-(Methyl, decylsulfonium-1-yl)ethylidene-1,1-bisphosphonic
Acid (31). Compound 31 was prepared from methyl decyl sulfide
(450 mg, 2.4 mmol) and vinylidene-1,1-diphosphonic acid (380 mg,
2 mmol) following General Method D (535 mg, 1.4 mmol, 70%).
Anal. (C11H26O6P2S) C, H. H NMR (400 MHz, D2O): δ 3.3-
3.47 (m, 2H), 3.13-3.21(m, 1H), 2.9-3.0 (m, 1H), 2.66 (s, 3H),
1.90-2.00 (m, 1H), 1.52-1.57 (m, 2H), 1.27-1.37 (m, 2H), 1.02-
1.25 (m, 8H), 0.72 (t, J ) 6.5 Hz, 3H). 31P NMR (D2O): δ 15.23.
1-Hydroxy-3-(methyl, octylsulfonium-1-yl)propylidene-1,1-
bisphosphonic Acid (39). Compound 39 was prepared from methyl
octyl sulfide (480 mg, 3 mmol) following General Methods B and
C (152 mg, 0.36 mmol, 12%). 1H NMR (400 MHz, D2O): 1H NMR

Sulfonium Bisphosphonates
Journal of Medicinal Chemistry, 2007, Vol. 50, No. 24 6077
(400 MHz, D2O): δ 3.00-3.41 (m, 4H), 2.69 (s, 3H), 2.2-2.3
(m, 2H), 1.60-1.75 (m, 2H), 1.15-1.37 (m, 10H), 0.77 (t, J ) 7.5
Hz, 3H). 31P NMR (D2O): δ 17.57.
(ATCC, Manassas, VA) was performed to obtain dose response
curves. Briefly, cells were incubated for 2 h under culture conditions
with the tetrazolium salt, lysed with detergent, incubated overnight,
protected from light at room temperature. The absorbance at 600
nm was read the following day using a SpectraMax Plus 384
spectrophotometer (Molecular Devices, Sunnyvale, CA). The
compound concentration for 50% growth inhibition values (IC50)
were obtained by fitting absorbance data to a rectangular hyperbolic
function: I ) (Imax )(C)/IC50 + C where I is the percent inhibition,
Imax ) 100% inhibition and C is the concentration of the inhibitor,
using GraphPad PRISM 3.0 software for windows (Graphpad
Software Inc., San Diego, CA).
Computational Methods. To relate the activity of the sulfonium
bisphosphonates to their chemical structure, we used four compu-
tational methods: 2D-QSAR, hologram QSAR (HQSAR),40 com-
parative molecular similarity indices analysis (CoMSIA),46 and
pharmacophore modeling. For molecular descriptor QSAR, chemi-
cal structures were imported into the Molecular Operating Environ-
ment (MOE)37 program, 3D structures were generated, and >200
2D and 3D molecular descriptors were computed. Alignment of
molecules is not required for this method. The AutoQuaSAR38
module was then used to systematically evaluate molecular descrip-
tor space in a stepwise fashion, building a series of cross-validated
partial least-squares (PLS) models using combinations of the
descriptors while monitoring the r2 and cross-validated r2 (q2)
trajectory. Models with the fewest number of components (descrip-
tors) and maximum predictive ability (r2, q2) where then evaluated
manually, for chemical significance. The final model selected
included terms accounting for the number of contiguous rotatable
bonds, molar refractivity (SMR VSA7),39,47 partial positive charge
(PEOE VSA+2),39,48 and molecular compactness (Weiner Path).49
1-Hydroxy-2-(methyl, propylsulfonium-1-yl)ethylidene-1,1-
bisphosphonic Acid (40). Compound 40 was prepared from methyl
propyl sulfide (220 mg, 2.4 mmol) following General Methods A
and C (82 mg, 0.21 mmol, 9%). Anal. (C6H16O7P2S‚1.3H2O) C,
1
H. H NMR (400 MHz, D2O): δ 3.30-3.50 (m, 2H), 2.90-3.20
(m, 2H), 2.71 (s, 3H), 1.62-1.73 (m, 2H), 0.93 (t, J ) 6.5 Hz,
3H). 31P NMR (D2O): δ 15.24.
2-(Methyl, propylsulfonium-1-yl)ethylidene-1,1-bisphosphonic
Acid (41) Compound 41 was prepared from methyl propyl sulfide
(360 mg, 4 mmol) and vinylidene-1,1-diphosphonic acid (380 mg,
2 mmol) following General Method D (272 mg, 0.95 mmol, 49%).
1
Anal. (C6H16O6P2S) C, H. H NMR (400 MHz, D2O): δ 3.25-
3.50 (m, 2H), 3.15-3.20 (m, 1H), 2.90-3.03 (m, 1H), 2.67 (s,
3H), 1.92-2.05 (m, 1H), 1.62-1.71 (m, 2H), 0.91 (t, J ) 6.5 Hz,
3H). 31P NMR (D2O): δ 15.01.
1-Hydroxy-3-(dimethylsulfonium-1-yl)propylidene-1,1-bis-
phosphonic Acid (42). Compound 42 was prepared from dimethyl
sulfide (0.4 g, 6.4 mmol) following General Methods B and C (172
mg, 0.58 mmol, 9%). Anal. (C5H4O7P2S‚0.5CH3OH). 1H NMR (400
MHz, D2O): δ 3.45 (t, J ) 7.8 Hz, 2H), 2.7 (s, 6H), 3.00-3.10
(m, 2H), 2.0-2.2 (m, 4H), 1.75-1.98 (m, 4H). 31P NMR (D2O):
δ 19.88.
1-Hydroxy-2-(dimethylsulfonium-1-yl)ethylidene-1,1-bispho-
sphonic Acid (43). Compound 43 was prepared from dimethyl
sulfide (1.5 g, 2.4 mmol) following General Methods A and C (447
mg, 1.7 mmol, 7%). Anal. (C4H12O7P2S) C, H. 1H NMR (400 MHz,
D2O): δ 3.50-3.61 (m, 4H), 2.7 (s, 6H). 31P NMR (D2O): δ 15.6.
2-(Methyl, pentylsulfonium-1-yl)ethylidene-1,1-bisphosphonic
Acid (44). Compound 44 was prepared from methyl pentyl sulfide
(283 mg, 2.4 mmmol) and vinylidene-1,1-diphosphonic acid (380
mg, 2 mmol) following General Method D (427 mg, 1.34 mmol,
67%). Anal. (C8H20O6P2S‚0.7H2O) C, H. 1H NMR (400 MHz,
D2O): δ 3.23-3.43 (m, 2H), 3.15-3.21(m, 1H), 2.90-3.03 (m,
1H), 2.77 (s, 3H), 1.90-2.00 (m, 1H), 1.62-1.77 (m, 2H), 1.27-
1.37 (m, 2H), 1.12-1.25 (m, 2H), 0.74 (t, J ) 6.5 Hz, 3H). 31P
NMR (D2O): δ 14.93.
1-Hydroxy-3-(methyl, pentenylsulfonium-1-yl)propylidene-
1,1-bisphosphonic Acid (45). Compound 45 was prepared from
methyl pentyl sulfide (0.7 g, 5.9 mmol) following General Methods
B and C (158 mg, 0.41 mmol, 7%). Anal. (C13H13NaO7P2S) C, H,
N. 1H NMR (400 MHz, D2O): δ 3.06-3.49 (m, 4H), 2.72 (s, 3H),
2.2-2.3 (m, 2H), 1.60-1.75 (m, 2H), 1.15-1.25 (m, 2H), 1.30-
1.37 (m, 2H), 0.72 (t, J ) 7.5 Hz, 3H). 31P NMR (D2O): δ 17.57.
1-Hydroxy-3-(methyl, propylsulfonium-1-yl)propylidene-1,1-
bisphosphonic Acid (46). Compound 46 was prepared from methyl
propyl sulfide (0.5 g, 5.6 mmol) following General Methods B and
C (148 mg, 0.42 mmol, 8%). Anal. (C7H17NaO7P2S) C, H. 1H NMR
(400 MHz, D2O): δ 3.00-3.52 (m, 4H), 2.75 (s, 3H), 2.1-2.3
(m, 2H), 1.53-1.70 (m, 2H), 1.23-1.35 (m, 2H), 0.78 (t, J ) 7.5
Hz, 3H). 31P NMR (D2O): δ 17.79.
Cell Growth Inhibition Assays. Human tumor cell lines MCF-7
(breast adenocarcinoma), NCI-H460 (lung large cell), and SF-268
(central nervous system glioblastoma) were obtained from the
National Cancer Institute and maintained at 100% humidity and
5% CO2 at 37 °C. Cell lines were cultured in RPMI-1640 medium
supplemented with 2 mM L-glutamine and 10% fetal bovine serum
(Gibco, Grand Island, NY) at 37 °C in a 5% CO2 atmosphere with
100% humidity. Compound stock solutions were typically prepared
in water at a concentration of 0.01 M. A broth microdilution method
was used to determine the bisphosphonate growth inhibition IC50
values. Compounds were half log serial diluted using cell culture
media into 96-well TC-treated flat bottom plates (Corning Inc.,
Corning, NY) typically from 316 µM to 10nM, but in some cases
compounds were run over a larger concentration range to enable
accurate IC50 determinations. Cells were plated at a density of 5000
cells/well. Cells were then incubated under the same culture
conditions for 4 days at which time an MTT ((3-(4,5-dimethylthi-
azole-2-yl)-2,5-diphenyltetrazolium bromide) cell proliferation assay
Hologram QSAR (HQSAR).40 Chemical structures were imported
into Sybyl 7.11 (Tripos Inc., St. Louis, MO) along with activity
data. Molecular holograms were computed using hologram lengths
of 97, 151, 199, 257, 307, and 353 and fragments of four to seven
atoms, with atom, bond, and connection information. The best
models were chosen based on the highest cross-validated r2.
CoMSIA46 analysis was performed in the Sybyl 7.350 program
using default settings. All alignments for the compounds were
prepared in the program MOE.37 Three-dimensional energy mini-
mized structures were generated and two alignments constructed,
one using the maximum common substructure (MCS) approach,
the second utilizing the flexible alignment module within MOE.
The MCS approach was carried out manually on the energy-
minimized structures, with the energy-minimized structure of the
most active molecule serving as the template. For the flexible
alignment method, molecules were added sequentially to the
alignment, performing up to 1000 flexible refinement iterations
using a gradient test of 0.01 to 1.0 with hydrophobe, logP, and
partial charge similarity features used in addition to the default
options (H-bond donor, acceptor, aromaticity, polar hydrogens, and
volume). The alignments were then exported into the Sybyl 7.350
program, where atomic charges were determined by using the
Gasteiger-Marsili method.48 CoMSIA indices were calculated on
a rectangular grid containing each of the sets of aligned molecules
using steric, electrostatic, hydrophobic, donor, and acceptor fields.
Default grid spacing and probe atoms were used. To relate the
structural features to activity, a partial least-squares analysis was
applied. The optimal number of components for each model was
determined by using SAMPLS51 cross-validation. Test-set valida-
tions and scrambling stability analyses were then performed using
Sybyl 7.3.
A common feature pharmacophore model was constructed in
MOE37 using the consensus pharmacophore module and the
polarity-charge-hydrophobicity (PCH) scheme. The final pharma-
cophore was generated from energy-minimized structures of the
top most active compounds using consensus features suggested by
the algorithm: hydrophobic, aromatic, cationic, and a combination
of features representing the phosphonate groups (acceptor, anion,
and metal ligator). These features representing the phosphonate

6078 Journal of Medicinal Chemistry, 2007, Vol. 50, No. 24
Zhang et al.
(3) Morgan, C.; Lewis, P. D.; Jones, R. M.; Bertelli, G.; Thomas, G.
A.; Leonard, R. C. The in vitro anti-tumour activity of zoledronic
acid and docetaxel at clinically achievable concentrations in prostate
cancer. Acta Oncol. 2007, 46 (5), 669-77.
(4) Issat, T.; Nowis, D.; Legat, M.; Makowski, M.; Klejman, M. P.;
Urbanski, J.; Skierski, J.; Koronkiewicz, M.; Stoklosa, T.; Brzezinska,
A.; Bil, J.; Gietka, J.; Jakobisiak, M.; Golab, J. Potentiated antitumor
effects of the combination treatment with statins and pamidronate in
vitro and in vivo. Int. J. Oncol. 2007, 30 (6), 1413-25.
(5) Koshimune, R.; Aoe, M.; Toyooka, S.; Hara, F.; Ouchida, M.;
Tokumo, M.; Sano, Y.; Date, H.; Shimizu, N., Anti-tumor effect of
bisphosphonate (YM529) on non-small cell lung cancer cell lines.
BMC Cancer 2007, 7, 8.
groups were combined using logical expressions within MOE to
produce two master features, each mapping to a phosphonate group.
Tolerances (sizes) of the spheres encompassing the features were
uniformly enlarged slightly to relax the model.
Molecular docking was performed using the Glide 4.545 program
within the Maestro 8.052 interface. The target proteins were prepared
using the protein preparation wizard. Hydrogens were added, Mg
ions present in the active site were assigned a +2 charge, and waters
beyond 5 Å of the heterogroups were deleted. Hydrogen bonds were
optimized and the hydrogen positions minimized to the default
value. The receptor grid was generated from the prepared target
structures using the default parameters, with additional constraints
added for each Mg2+, and two hydrophobic constraints at the end
of the hydrophobic tunnel present in FPPS. Any, all, or none of
these constraints can be activated during the docking calculation.
Ligands were imported as SD files, with atom types and bond orders
verified. Ligands were then minimized using the MacroModel
multiple minimization module in Maestro 8.0.52 Docking calcula-
tions were performed specifying standard precision, together with
constraints enabled requiring interactions with at least two Mg2+
ions. The final results were visualized using the Glide pose viewer.
(6) Kubo, T.; Shimose, S.; Matsuo, T.; Tanaka, K.; Yasunaga, Y.; Sakai,
A.; Ochi, M. Inhibitory effects of a new bisphosphonate, minodronate,
on proliferation and invasion of a variety of malignant bone tumor
cells. J. Orthop. Res. 2006, 24 (6), 1138-44.
(7) Dieli, F.; Vermijlen, D.; Fulfaro, F.; Caccamo, N.; Meraviglia, S.;
Cicero, G.; Roberts, A.; Buccheri, S.; D’Asaro, M.; Gebbia, N.;
Salerno, A.; Eberl, M.; Hayday, A. C. Targeting human gd T cells
with zoledronate and interleukin-2 for immunotherapy of hormone-
refractory prostate cancer. Cancer Res. 2007, 67 (15), 7450-7.
(8) Sato, K.; Yuasa, T.; Nogawa, M.; Kimura, S.; Segawa, H.; Yokota,
A.; Maekawa, T., A third-generation bisphosphonate, minodronic acid
(YM529), successfully prevented the growth of bladder cancer in
vitro and in vivo. Br. J. Cancer 2006, 95 (10), 1354-61.
(9) Rodriguez, N.; Bailey, B. N.; Martin, M. B.; Oldfield, E.; Urbina, J.
A.; Docampo, R. Radical Cure of Experimental Cutaneous Leish-
maniasis by the Bisphosphonate Pamidronate. J. Infect. Dis. 2002,
186 (1), 138-140.
(10) Leon, A.; Liu, L.; Yang, Y.; Hudock, M. P.; Hall, P.; Yin, F.; Studer,
D.; Puan, K. J.; Morita, C. T.; Oldfield, E. Isoprenoid biosynthesis
as a drug target: bisphosphonate inhibition of Escherichia coli K12
growth and synergistic effects of fosmidomycin. J. Med. Chem. 2006,
49 (25), 7331-41.
(11) Wilhelm, M.; Kunzmann, V.; Eckstein, S.; Reimer, P.; Weissinger,
F.; Ruediger, T.; Tony, H. P., γδ T Cells for Immune Therapy of
Patients with Lymphoid Malignancies. Blood 2003, 102 (1), 200-6.
(12) Sato, K.; Kimura, S.; Segawa, H.; Yokota, A.; Matsumoto, S.; Kuroda,
J.; Nogawa, M.; Yuasa, T.; Kiyono, Y.; Wada, H.; Maekawa, T.
Cytotoxic effects of gammadelta T cells expanded ex vivo by a third
generation bisphosphonate for cancer immunotherapy. Int. J. Cancer
2005, 116 (1), 94-9.
(13) Das, H.; Wang, L.; Kamath, A.; Bukowski, J. F., Vgamma2Vdelta2
T-cell receptor-mediated recognition of aminobisphosphonates. Blood
2001, 98 (5), 1616-8.
(14) Wang, L.; Kamath, A.; Das, H.; Li, L.; Bukowski, J. F., Antibacterial
Effect of Human Vγ2Vδ2 T Cells In ViVo. J. Clin. InVest. 2001,
108 (9), 1349-57.
(15) Reszka, A. A.; Rodan, G. A., Nitrogen-containing bisphosphonate
mechanism of action. Mini ReV. Med. Chem. 2004, 4 (7), 711-9.
(16) Fleisch, H.; Russell, R. G.; Francis, M. D. Diphosphonates Inhibit
Hydroxyapatite Dissolution in vitro and Bone Resorption in Tissue
Culture and in vivo. Science 1969, 165, 1262-1264.
(17) Lehenkari, P. P.; Kellinsalmi, M.; Napankangas, J. P.; Ylitalo, K.
V.; Monkkonen, J.; Rogers, M. J.; Azhayev, A.; Vaananen, H. K.;
Hassinen, I. E. Further Insight Into Mechanism of Action of
Clodronate: Inhibition of Mitochondrial ADP/ATP Translocase by
a Nonhydrolyzable, Adenine-Containing Metabolite. Mol. Pharmacol.
2002, 61 (5), 1255-62.
Conclusions
The results we have described above are of interest for a
number of reasons. First, we have synthesized a broad range of
novel sulfonium bisphosphonates. Second, we have tested these
bisphosphonates as growth inhibitors of three human tumor cell
lines (MCF-7, NCI-H460, and SF-268). The most active species
were found to have IC50 values in the 200 nM range and are
considerably more active than any other bisphosphonates, such
as minodronate, zoledronate, risedronate, ibandronate, alendr-
onate, and pamidronate. Third, we carried out a series of QSAR
investigations of these compounds and showed that their
activities can be predicted within a factor of ∼2, using CoMSIA
methods, obtaining similar results from both maximum common
substructure and feature-based alignments. Fourth, we report
the first X-ray crystallographic structures of sulfonium bispho-
phonate-FPPS complexes, together with additional (docking)
results, which indicate how other bisphosphonates bind to FPPS.
Overall, the observation that these novel cationic bisphospho-
nates have high activity against three tumor cell lines is of
general interest, and suggests that these and other, related
species, may be of interest in the context of new anticancer
therapies and potentially as antiinfectives.
Acknowledgment. This work was supported by the United
States Public Health Service (NIH grant GM-073216). YZ was
supported by a Postdoctoral Fellowship from the American
Heart Association, Midwest Affiliate. A.L. was supported by
an NIH Institutional NRSA in Molecular Biophysics (NIH
Training Grant GM-08276). We thank Howard Robinson of
Brookhaven National Laboratories for obtaining the X-ray
diffraction data. We also thank Schro¨dinger, LLC for providing
evaluation licenses of Glide.
(18) Monkkonen, H.; Auriola, S.; Lehenkari, P.; Kellinsalmi, M.; Hassinen,
I. E.; Vepsalainen, J.; Monkkonen, J. A new endogenous ATP analog
(ApppI) inhibits the mitochondrial adenine nucleotide translocase
(ANT) and is responsible for the apoptosis induced by nitrogen-
containing bisphosphonates. Br. J. Pharmacol. 2006, 147 (4), 437-
45.
(19) Cromartie, T. H.; Fisher, K. J.; Grossman, J. N. The discovery of a
novel site of action for herbicidal bisphosphonates. Pestic. Biochem.
Physiol. 1999, 63 (2), 114-126.
(20) Martin, M. B.; Arnold, W.; Heath, H. T., 3rd; Urbina, J. A.; Oldfield,
E., Nitrogen-Containing Bisphosphonates as Carbocation Transition
State Analogs for Isoprenoid Biosynthesis. Biochem. Biophys. Res.
Commun. 1999, 263 (3), 754-8.
Supporting Information Available: Supporting methods,
figures, and tables, including microanalysis data, human FPPS and
molecular diameter, 2D-QSAR output, hologram QSAR results, and
CoMSIA results. This material is available free of charge via the
Internet at http://pubs.acs.org.
References
(21) 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 (1), 231-
41.
(22) Grove, J. E.; Brown, R. J.; Watts, D. J. The intracellular target for
the antiresorptive aminobisphosphonate drugs in Dictyostelium
discoideum is the enzyme farnesyl diphosphate synthase. J. Bone
Miner. Res. 2000, 15 (5), 971-81.
(1) Shipman, C. M.; Croucher, P. I.; Russell, R. G.; Helfrich, M. H.;
Rogers, M. J. The bisphosphonate incadronate (YM175) causes
apoptosis of human myeloma cells in vitro by inhibiting the
mevalonate pathway. Cancer Res. 1998, 58 (23), 5294-7.
(2) Shipman, C. M.; Rogers, M. J.; Apperley, J. F.; Russell, R. G.;
Croucher, P. I. Bisphosphonates induce apoptosis in human myeloma
cell lines: a novel anti-tumour activity. Br. J. Haematol. 1997, 98
(3), 665-72.

Sulfonium Bisphosphonates
Journal of Medicinal Chemistry, 2007, Vol. 50, No. 24 6079
(23) Luckman, S. P.; Hughes, D. E.; Coxon, F. P.; Graham, R.; Russell,
G.; Rogers, M. J., Nitrogen-containing bisphosphonates inhibit the
mevalonate pathway and prevent post-translational prenylation of
GTP-binding proteins, including Ras. J. Bone Miner. Res. 1998, 13
(4), 581-9.
(24) Dunford, J. E.; Rogers, M. J.; Ebetino, F. H.; Phipps, R. J.; Coxon,
F. P. Inhibition of protein prenylation by bisphosphonates causes
sustained activation of Rac, Cdc42, and Rho GTPases. J. Bone Miner.
Res. 2006, 21 (5), 684-94.
(25) Kavanagh, K. L.; Guo, K.; Dunford, J. E.; Wu, X.; Knapp, S.; Ebetino,
F. H.; Rogers, M. J.; Russell, R. G.; Oppermann, U. The molecular
mechanism of nitrogen-containing bisphosphonates as antiosteoporo-
sis drugs. Proc. Natl. Acad. Sci. U.S.A. 2006, 103 (20), 7829-34.
(26) Rondeau, J. M.; Bitsch, F.; Bourgier, E.; Geiser, M.; Hemmig, R.;
Kroemer, M.; Lehmann, S.; Ramage, P.; Rieffel, S.; Strauss, A.;
Green, J. R.; Jahnke, W. Structural basis for the exceptional in vivo
efficacy of bisphosphonate drugs. ChemMedChem 2006, 1 (2), 267-
73.
(27) Hosfield, D. J.; Zhang, Y.; Dougan, D. R.; Broun, A.; Tari, L. W.;
Swanson, R. V.; Finn, J. Structural Basis for Bisphosphonate-
Mediated Inhibition of Isoprenoid Biosynthesis. J. Biol. Chem. 2004,
279 (10), 8526-9.
(35) Widler, L.; Jaeggi, K. A.; Glatt, M.; Muller, K.; Bachmann, R.;
Bisping, M.; Born, A. R.; Cortesi, R.; Guiglia, G.; Jeker, H.; Klein,
R.; Ramseier, U.; Schmid, J.; Schreiber, G.; Seltenmeyer, Y.; Green,
J. R. Highly Potent Feminal Bisphosphonates. From Pamidronate
disodium (Aredia) to Zoledronic Acid (Zometa). J. Med. Chem. 2002,
45 (17), 3721-38.
(36) Petitjean, M. Applications of the Radius Diameter Diagram to the
Classification of Topological and Geometrical Shapes of Chemical-
Compounds. J. Chem. Inf. Comput. Sci. 1992, 32 (4), 331-337.
(37) MOE, 2006.08; Chemical Computing Group, Inc.: Montreal, Quebec,
2006.
(38) Goto, J. AutoQuaSAR 2006.08, Ryoka Systems, Inc.: Tokyo, Japan,
2006.
(39) Labute, P. A widely applicable set of descriptors. J. Mol. Graph.
Modelling 2000, 18 (4-5), 464-477.
(40) Hologram QSAR (HQSAR) module part of Sybyl 7.11, Tripos Inc.:
St. Louis, MO.
(41) Labute, P.; Williams, C.; Feher, M.; Sourial, E.; Schmidt, J. M.
Flexible alignment of small molecules. J. Med. Chem. 2001, 44 (10),
1483-1490.
(28) Mao, J.; Mukherjee, S.; Zhang, Y.; Cao, R.; Sanders, J. M.; Song,
Y.; Zhang, Y.; Meints, G. A.; Gao, Y. G.; Mukkamala, D.; Hudock,
M. P.; Oldfield, E., Solid-state NMR, crystallographic, and compu-
tational investigation of bisphosphonates and farnesyl diphosphate
synthase-bisphosphonate complexes. J. Am. Chem. Soc. 2006, 128
(45), 14485-97.
(29) Sheng, L. S.; Horning, E. C.; Horning, M. G. Synthesis and
Elimination-Reactions of Methylsulfonium Ions Formed from Styrene
Oxide and Methylthio Compounds Related to Methionine and
Cysteine. Drug Metab. Dispos. 1984, 12 (3), 297-303.
(30) Sano, Y.; Yamada, H. Progress in suplatast tosilate research. Clin.
Exp. Allergy 2007, 37 (7), 970-2.
(31) Yanagihara, Y.; Kiniwa, M.; Ikizawa, K.; Yamaya, H.; Shida, T.;
Matsuura, N.; Koda, A.; Suppression, of Ige, Production, by Ipd-
1151t (Suplatast, Tosilate), a New, Dimethylsulfonium, Agent, .1.
Regulation of Murine Ige Response. Jpn. J. Pharmacol. 1993, 61
(1), 23-30.
(32) Zhang, Y.; Leon, A.; Song, Y.; Studer, D.; Haase, C.; Koscielski, L.
A.; Oldfield, E. Activity of nitrogen-containing and non-nitrogen-
containing bisphosphonates on tumor cell lines. J. Med. Chem. 2006,
49 (19), 5804-14.
(33) Sanders, J. M.; Song, Y.; Chan, J. M.; Zhang, Y.; Jennings, S.;
Kosztowski, T.; Odeh, S.; Flessner, R.; Schwerdtfeger, C.; Kotsiko-
rou, E.; Meints, G. A.; Gomez, A. O.; Gonzalez-Pacanowska, D.;
Raker, A. M.; Wang, H.; van Beek, E. R.; Papapoulos, S. E.; Morita,
C. T.; Oldfield, E., Pyridinium-1-yl Bisphosphonates are Potent
Inhibitors of Farnesyl Diphosphate Synthase and Bone Resorption.
J. Med. Chem. 2005, 48 (8), 2957-63.
(34) Sanders, J. M.; Ghosh, S.; Chan, J. M.; Meints, G.; Wang, H.; Raker,
A. M.; Song, Y.; Colantino, A.; Burzynska, A.; Kafarski, P.; Morita,
C. T.; Oldfield, E., Quantitative Structure-Activity Relationships for
©TM T Cell Activation by Bisphosphonates. J. Med. Chem. 2004,
47 (2), 375-84.
(42) Clark, R. D.; Sprous, D. G., Validating models based on large
datasets. In Rational Approaches to Drug Design; Holtje, H.-D.;
Sippl, W., Eds.; Prous Science SA: Barcelona, 2001; pp 475-485.
(43) Tripos Bookshelf 7.3, Tripos, Inc.: St. Louis, MO.
(44) Mao, J.; Gao, Y. G.; Odeh, S.; Robinson, H.; Montalvetti, A.;
Docampo, R.; Oldfield, E., Crystallization and Preliminary X-ray
Diffraction Study of the Farnesyl Diphosphate Synthase from
Trypanosoma brucei. Acta Crystallogr. Ser. D: Biol. Crystallogr.
2004, 60 (Pt 10), 1863-6.
(45) Glide 4.5, Schrodinger, Inc.: Portland, OR.
(46) Klebe, G.; Abraham, U.; Mietzner, T. Molecular Similarity Indices
in a Comparative Analysis (CoMSIA) of Drug Molecules to Correlate
and Predict their Biological Activity. J. Med. Chem. 1994, 37 (24),
4130-46.
(47) Wildman, S. A.; Crippen, G. M. Prediction of physicochemical
parameters by atomic contributions. J. Chem. Inf. Comput. Sci. 1999,
39 (5), 868-873.
(48) Gasteiger, J.; Marsili, M. Iterative Partial Equalization of Orbital
Electronegativity-a Rapid Access to Atomic Charges. Tetrahedron
1980, 36 (22), 3219-3228.
(49) Balaban, A. T.; Chemical, Graphs, .34. 5 New Topological Indexes
for the Branching of Tree-Like Graphs. Theor. Chim. Acta 1979, 53
(4), 355-375.
(50) Sybyl 7.3, Tripos, Inc.: St. Louis, MO.
(51) Bush, B. L.; Nachbar, R. B., Sample-Distance Partial Least-Squares-
Pls Optimized for Many Variables, with Application to Comfa. J.
Comput.-Aided Mol. Des. 1993, 7 (5), 587-619.
(52) Maestro 8.0, Schrodinger Inc.: Portland, OR.
JM700991K