Probing the molecular and structural elements of ligands binding to the active site versus an allosteric pocket of the human farnesyl pyrophosphate synthase

Article abstract of DOI:10.1016/j.bmcl.2014.12.089

In order to explore the interactions of bisphosphonate ligands with the active site and an allosteric pocket of the human farnesyl pyrophosphate synthase (hFPPS), substituted indole and azabenzimidazole bisphosphonates were designed as chameleon ligands. NMR and crystallographic studies revealed that these compounds can occupy both sub-pockets of the active site cavity, as well as the allosteric pocket of hFPPS in the presence of the enzyme's Mg²⁺ ion cofactor. These results are consistent with the previously proposed hypothesis that the allosteric pocket of hFPPS, located near the active site, plays a feed-back regulatory role for this enzyme.

Full text of DOI:10.1016/j.bmcl.2014.12.089

Bioorganic & Medicinal Chemistry Letters 25 (2015) 1117–1123
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry Letters
journal homepage: www.elsevier.com/locate/bmcl
Probing the molecular and structural elements of ligands binding to
the active site versus an allosteric pocket of the human farnesyl
pyrophosphate synthase
Dimitrios Gritzalis a, Jaeok Park b, Wei Chiu a, , Hyungjun Cho a, , Yih-Shyan Lin a, Joris W. De Schutter a,
Cyrus M. Lacbay a, Michal Zielinski b, , Albert M. Berghuis b,c,d, Youla S. Tsantrizos a,b,d,
⇑
a Department of Chemistry, McGill University, 801 Sherbrooke Street West, Montreal, QC H3A 0B8, Canada
b Department of Biochemistry, McGill University, 3655 Promenade Sir William Osler, Montreal, QC H3G 1Y6, Canada
c Department of Microbiology and Immunology, McGill University, 3775 Rue University, Montreal, QC H3A 2B4, Canada
d Groupe de Recherche Axé sur la Structure des Protéines, McGill University, 3649 Promenade Sir William Osler, Montréal, QC H3G 0B1, Canada
a r t i c l e i n f o
a b s t r a c t
Article history:
In order to explore the interactions of bisphosphonate ligands with the active site and an allosteric pocket
of the human farnesyl pyrophosphate synthase (hFPPS), substituted indole and azabenzimidazole bis-
phosphonates were designed as chameleon ligands. NMR and crystallographic studies revealed that these
compounds can occupy both sub-pockets of the active site cavity, as well as the allosteric pocket of hFPPS
in the presence of the enzyme’s Mg2+ ion cofactor. These results are consistent with the previously pro-
posed hypothesis that the allosteric pocket of hFPPS, located near the active site, plays a feed-back reg-
ulatory role for this enzyme.
Received 17 October 2014
Revised 26 December 2014
Accepted 29 December 2014
Available online 13 January 2015
Keywords:
Bisphosphonates
Human FPPS
Ó 2015 Elsevier Ltd. All rights reserved.
Allosteric inhibitors
Bisphosphonates (BPs) are chemically stable bioisosteres of
inorganic pyrophosphate that were initially developed as antiscal-
ing, anticorrosion and water softening agents.1,2 The subsequent
discovery that BPs can also prevent bone loss in vivo opened the
door to the development of therapeutic agents for the prevention
of osteoclast-mediated bone diseases. The most effective of the
clinically validated compounds are the nitrogen-containing BPs
(N-BPs) having an aliphatic [e.g., pamidronic acid (1a), alendronic
acid (1b), ibandronic acid (1c)] or a heterocyclic side chain, such
as risedronic acid (2), zoledronic acid (3) and minodronic acid
(5). The key biological target of N-BPs is the human farnesyl pyro-
phosphate synthase (hFPPS),3 the enzyme which catalyzes the
sequential condensation of isopentenyl pyrophosphate (IPP) with
dimethylallyl pyrophosphate (DMAPP) to give geranyl pyrophos-
phate (GPP), and then catalyzes a second condensation between
GPP and IPP to form farnesyl pyrophosphate (FPP). Given the stra-
tegic location of hFPPS in the mevalonate pathway, this enzyme
also plays a pivotal role in controlling the intracellular levels of
all human isoprenoids, as well as numerous other essential metab-
PO(OH)2
PO(OH)2
R' =
OH
R1
N
R2
R'
NH
R'
N
N
N
n
N
N
N
R'
R1
R'
R'
1a, n = 1, R1=R2=H
2
3
4a, R1=H
5
1b, n = 2, R1=R2=H
4b, R1=Me
1c, n = 1, R1=H, R2=n-pentyl
The key pharmacophore of the current N-BP drugs is their
a
-hydroxyl bisphosphonate moiety (R0). This moiety acts as a tri-
C
dentate ligand for Mg2+ ions in the active site of hFPPS, as well as
for Ca2+ ions and hydroxyapatite in bone. Since removal of the
Ca-hydroxyl usually leads to decreased affinity for bone, this sub-
stituent is commonly referred to as the ‘bone hook’.5 The nitrogen
on the side chain of N-BPs also plays a role in the activity of these
compounds. It is presumed to be protonated and participating in
bifurcated hydrogen bond interactions with the side chain hydro-
xyl of Thr 201 and the carbonyl oxygen of Lys 200 within the hFPPS
active site.3 These interactions mimic those of the putative transi-
tion-state allylic carbocation formed during the hFPPS catalytic
cycle.6 N-BPs with an imidazole ring as part of their side chain
4
olites.
⇑
Corresponding author. Tel.: +1 514 398 3638; fax: +1 514 398 3797.
E-mail address: youla.tsantrizos@mcgill.ca (Y.S. Tsantrizos).
Undergraduate Research Participants.
http://dx.doi.org/10.1016/j.bmcl.2014.12.089
0960-894X/Ó 2015 Elsevier Ltd. All rights reserved.

1118
D. Gritzalis et al. / Bioorg. Med. Chem. Lett. 25 (2015) 1117–1123
(i.e., 3, 4, 5) are amongst the most potent inhibitors of hFPPS, as
well as the most effective therapeutics in blocking osteoclastic
bone resorption. Earlier structure–activity relationship (SAR) stud-
ies at Ciba-Geigy (now Novartis) explored the direction of attach-
ment from the imidazole to the bisphosphonate (e.g., 3 vs 4), as
well as the influence of substitution on the potency of these com-
pounds.7 A linker of one methylene unit between the imidazole
biologically relevant conditions were estimated by ITC.10 However,
since in vitro evaluation of hFPPS activity requires Mg2+ as the
cofactor (thus biasing the binding of bisphosphonates to the allylic
sub-pocket), the ability of bisphosphonates to bind in part or exclu-
sively in the allosteric pocket under biologically relevant conditions
remained difficult to prove. In this report, we describe the design
and synthesis of novel bisphosphonate inhibitors of hFPPS that
are structurally related to the clinical drug 5. We provide data
which strongly suggests that some of these inhibitors can occupy
simultaneously both the allylic sub-pocket and IPP sub-pocket of
the active site cavity. In addition, we obtained conclusive evidence
that bisphosphonates inhibitors of hFPPS can occupy the allosteric
pocket of the enzyme, even in the presence of high concentrations
of Mg2+ ions. These findings, together with the fact that bisphos-
phonates are structural mimics of isoprenyl metabolites (i.e., mim-
ics of DMAPP and GPP), support the previously proposed theory
that the allosteric pocket of hFPPS plays a regulatory role in product
feedback inhibition.
Based on the previously reported binary and ternary structures
of hFPPS/N-BP and hFPPS/N-BP/IPP complexes, it is well estab-
lished that the initial occupancy of the allylic sub-pocket by an
N-BP inhibitor leads to a structural change of the active site cavity
from the fully ‘open’ to the ‘half-closed’ conformation.3 This con-
formational transition of the protein re-defines the shape of the
IPP binding sub-pocket, allowing high-affinity binding of the
homoallylic substrate (IPP). Co-occupancy of the allylic sub-pocket
and the IPP sub-pocket sets in motion a second conformational
change, which involves folding of the C-terminal residues
350KRRK353 over the IPP sub-pocket and complete ‘closing’ of the
active site cavity, thus sequestering both bound ligands from bulk
water. In contrast, binding of an allosteric inhibitor near the IPP
sub-pocket freezes the active site cavity in the catalytically incom-
petent open (or half closed) conformation, even in the presence of a
co-bound N-BP (PDB code: 3N46) and blocks binding of IPP; more
specifically, it blocks the binding of the pyrophosphate moiety of
IPP. Based on all these data, the role of the lipophilic tail of IPP in
the binding contributions and the biological relevance of the allo-
steric pocket remained unclear.
and the Ca of the bisphosphonate, together with direct attachment
of the imidazole nitrogen to the methylene linker was shown to
provide optimum potency in reducing hypercalcemia (i.e., blood
levels of Ca2+) in rat. For example, the drug ZometaÒ (3) is approx-
imately 4-fold more potent in vivo than analogs 4a (i.e., EC50 differ-
ence reflects the dose of compound administered subcutaneously,
which results in a 50% reduction of hypercalcemia in rat). Interest-
ingly, although the methyl derivative 4b is 50-fold less potent than
4a and >200-fold less potent than 3, the fused imidazole derivative
5 is nearly equipotent to 3 both in vitro and in vivo. The IC50 values
in inhibiting hFPPS of 3 and 5 are 4.1 nM and 1.9 nM, respectively,
with a minimum effective dose (MED) of 0.003 mg/kg (s.c. admin-
istration) in reducing hypercalcemia in rat.8 A number of structur-
ally related analogs of 5 and their corresponding azaindoles are
also potent inhibitors of hFPPS.9 Consequently, we selected the
structurally related fused bicyclic imidazole (8c), indole (9) and
azabenzimidazole (10) scaffolds for the design of compounds used
in the structural investigations, described in this study.
R
Y
X
PO(OH)2
PO(OH)2
CO2H
CO2H
S
N
N
N
N
PO(OH)2
R1
N
N
H
PO(OH)2
1
8a
, X=Y=CH, R =H
7
1
8b
, X=Y=N, R =H
6a, R=Me
6b, R =cyclopropyl
1
8c
, X=Y=CH, R =alkyl
R3
Y
X
We noted that in the enzyme bound state, the side chains of
N-BP inhibitors are within van der Waals radius from the IPP
hydrophobic tail (Fig. 1a–c). We also recently showed that
bisphosphonate 6b binds mainly in the allylic sub-pocket, but its
pyrimidine ring protrudes into the IPP sub-pocket, thus occupying
part of the region that is normally occupied by the IPP isoprenyl
tail (PDB code: 4L2X; Fig. 1c). Consequently, we reasoned that
N-BPs with an appropriate molecular design could possibly occupy
both the allylic and IPP sub-pockets simultaneously and potentially
exhibit much higher affinity for the active site cavity. Structural
characterization of the interactions between such inhibitors and
the enzyme could lead to the designs of novel hFPPS inhibitors
with lower dependency on the bisphosphonate anchor for binding
to the active site cavity.
Initially, we used an in silico model to dock benzimidazole bis-
phosphonates of general structure 8c to the fully closed active site
of hFPPS using GLIDE (version 5.5, Shrödinger, LLC, New York, NY
2009; standard parameters of XP-mode were used). Favourable
outputs for the binding of the R1 alkyl moiety into the IPP
sub-pocket were obtained; an example of a docked molecule,
where the R1 substituent of 8c is –C(CH3)cyclopropyl, is shown in
Figure 1d. However, mindful of the significant protein plasticity
of hFPPS and the large conformational changes previously
observed upon ligand binding to this target, we were cautious
about the validity of these data.
PO(OH)2
PO(OH)2
N
N
R1
PO(OH)2
PO(OH)2
N
R1
HN
R2
9
1
10a
10b
, X=CH, Y=N, R =Me
, X=N, Y=CH, R =Me
1
The interactions between N-BPs and their primary biological tar-
get, hFPPS, have been extensively characterized by X-ray crystallog-
raphy.3 Numerous structures of hFPPS/ligand complexes have been
reported, including the binary complex of hFPPS/2 (PDB codes:
1YQ7, 1YV5) and the ternary complex of hFPPS/3/IPP (PDB code:
1ZW5). These studies have revealed that in the enzyme-bound
state, the bisphosphonate of N-BPs is fully ionized to the tetra anion
and interacts with three Mg2+ ions. These metal-mediated interac-
tions allow binding of N-BPs to two highly conserved aspartate-rich
(DDXXD) motifs that define the allylic sub-pocket (DMAPP/GPP-
binding site) of the hFPPS active site. Until very recently, all known
N-BPs were found to bind exclusively in this manner and only in the
allylic sub-pocket of the enzyme. In a recent report, we identified
for the first time thienopyrimidine-based bisphosphonates (e.g.,
6) that exhibit a mixed binding mode.10 We provided DSF, NMR
and crystallographic data which strongly suggested that inhibitors
of general structure 6 can bind mainly in the allylic sub-pocket of
hFPPS in the presence of Mg2+ ions and in the allosteric pocket of
hFPPS in the absence of Mg2+ ions (PDB IDs: 4JVJ vs 4LPG, respec-
tively). The relative contributions of the two binding modes under
Benzimidazole-based derivatives (e.g., 8a,b) have been previously
reported as potent inhibitors of hFPPS.11 We initiated the synthesis of
analogs with general structure 8c, assuming that the R1 alkyl group

D. Gritzalis et al. / Bioorg. Med. Chem. Lett. 25 (2015) 1117–1123
1119
Figure 1. Ligands in the active site of hFPPS. (a) Structure of the hFPPS/3/IPP complex (PDB ID: 1ZW5).3a The surface of the ligands is represented in mesh. (b) The protein is
removed to clearly show the close contact between the two ligands. (c) Superimposition of the hFPPS/3/IPP and hFPPS/6b/PPi ternary structures (PDB codes: 1ZW5 and 4L2X,
respectively); 3 and PPi are omitted for clarity. The potential steric clash between the thienopyrimidine core of 6b and the prenyl tail of IPP is evident (closest carbon-to-
carbon contact is less than 2.7 Å). (d) Compound 8c [where the R1 is –C(CH3)cyclopropyl] docked in the active site of the fully closed protein (PDB codes: 1ZW5) using GLIDE
and superimposed with 3 and IPP; the protein was removed to clearly show the binding proximity between the R1 alkyl moiety of 8c and the prenyl tail of IPP. The carbon
backbone of 3, 6b, 8c and IPP are highlighted in cyan, green, blue and magenta, respectively. Oxygen, nitrogen, sulfur and phosphorus atoms are colored in red, blue, yellow
and orange, respectively. Yellow spheres represent Mg2+ ions.
could mimic the prenyl tail of IPP. However, in our hands these com-
Table 1
pounds were found to be chemically unstable and further profiling in
biological or structural assays were not pursued. We subsequently
explored structurally related indoles 9 (Table 1). Inhibitors 9a–9f
were prepared via the classical Madelung indole synthesis from com-
mercially available 2-methyl aniline (11). The aniline was reacted
with various acid chlorides and the amide product was cyclized to
the indole 12 in the presence of n-BuLi (Scheme 1). Alkylation of
the indole nitrogenwas easily achieved with alkyl halides in the pres-
ence ofNaH. The indolescaffold was thenreactedwithtetraethyl eth-
ene-1,1-diylbis(phosphonate) under acidic conditions to introduce
the C-3 bisphosphonate side chain, which was then deprotected with
TMSBr/MeOH as previously reported (Scheme 1).12 Analogs with an
aryl or heteroaryl substituent at R3 were prepared starting with p-
bromoaniline (13) via a modified Fischer indole synthesis, followed
by Suzuki–Miyaura cross-coupling to obtain the key fragment 14
(Scheme 1). Reduction of the ethyl ester 14 followed by oxidation
provided the aldehyde 15, which was reacted with various Grignard
reagents to give the secondary alcohol 16. Re-oxidation of 16–17, fol-
lowed by a Wittig reaction provided intermediates with an olefinic
branched alkyl substituent at C-2 (18). Hydrogenation and alkylation
at the nitrogen (19), followed by introduction of the bisphosphonate
gave analogs with a branched R2 moiety, such as inhibitor 9i
(Scheme 1, Table 1).
Indole-(9) and azabenzimidazole-based (10) inhibitors of the human FPPS; (a)
structures of compounds, (b) % inhibition of hFPPS at 10
l
M of inhibitors (numbers
indicate the average of three independent determinations)
PO(OH)2
(a)
PO(OH)2
R3
Y
X
PO(OH)2
R2
N
PO(OH)2
NH
N
R1
N
R1
1
2
3
1
9a
, R =Me, R =R =H
9b
10a
10b
, X=CH, Y=N, R =Me
1
2
3
1
, R =Et, R =R =H
, X=N, Y=CH, R =Me
9c, R1=i-Pr, R2=R3=H
9d, R1=R2=Me, R3=H
9e, R1=R3=H, R2=Me
9f, R1=R3=H, R2=t-Bu
1
3
2
9g
9h
9i
, R =R =H, R =CH(CH3)CH(CH3)2
1
3
2
, R =R =H, R =CH(CH3)cyclopentyl
1
2
3
, R =H, R =CH(CH3)CH(CH3)2, R =Ph
(b)
Azabenzimidazoles of general structure 10 were also synthe-
sized following the reactions shown in Scheme 2. The aminopyri-
dines 21 and 23 were converted to the diamines 22 and 24,
respectively, with an alkyl amine (Scheme 2).13 Carbonylation of
these diamines with 1,10-cabonyldiimidazole (CDI), provided imi-
dazolone intermediates 25, which were converted to the chloride
26 with POCl3 as previously reported.14 Subsequently, displace-
ment of the chloride with ammonia (27), followed by coupling
with tetraethyl ethene-1,1-diylbis(phosphonate) and hydrolysis
of the bisphosphonate tetraesters (28) gave compounds 10 (e.g.,
10a,b; Table 1) in good overall yields.

1120
D. Gritzalis et al. / Bioorg. Med. Chem. Lett. 25 (2015) 1117–1123
R3
R3
Br
O
H
a, b
f, g, h, i
j, k
R2
CO2Et
N
H
N
H
N
H
NH2
NH2
15
12
11
13
14
l
c, d
PO(OR")2
PO(OR")2
R2
R3
R3
R3
R3
OH
R'
X
d
n, c
k
N
N
R1
N
R1
N
H
R'
R'
H
20
, R" = Et
19
17
, X = O
18, X = CH2
16
m
e
9, R" = H
Scheme 1. Synthesis of indole-based bisphosphonates of general structure 9. Reagents and conditions: (a) R2COCl, CH2Cl2, ꢀ78 to 23 °C, 15 h, 85%; (b) nBuLi, THF, 0–23 °C,
90%; (c) R1Br or R1I, NaH, 0–23 °C, 40–90% yield; (d) tetraethyl ethene-1,1-diylbis(phosphonate), AcOH, reflux, 5d, 57%; (e) TMSBr, CH2Cl2, 23 °C, 5d, 98%; (f) (i) HCl, NaNO2,
0 °C, 15 min, (ii) SnCl2/HCl, 0 °C, 4 h, 60%; (g) ethyl pyruvate, EtOH, reflux, 15 h 85%; (h) polyphosphoric acid, 130 °C, 24 h, 45%; (i) ArB(OH)2, 5% Pd(PPh3)4, K2CO3, dioxane/
H2O, 110 °C, 3 h, 50–70%; (j) LiAlH4, THF, 0–23 °C, 15 h, 40%; (k) IBX, EtOAc, 80 °C, 3 h, 45–98%; (l) R0MgCl, THF, 0–23 °C,2 h, 80%; (m) Ph3PMeBr, THF, 0–23 °C, 3 h, 90%; (n) H2,
Pd/C, MeOH, 23 °C, 4 h, 90%.
NH2
Cl
NH2
a
a
N
21
N
22
NHR1
H
N
Y
X
Y
X
N
N
b
c
O
Cl
N
NHR1
NH2
R1
R1
Br
25a
, X=CH, Y=N
26a
26b
, X=CH, Y=N
25b, X=N, Y=CH
, X=N, Y=CH
N
NH2
N
23
24
d
PO(OR')2
Y
X
Y
X
N
N
N
N
e
NH
PO(OR')2
NH2
R1
R1
28a, X=CH, Y=N, R'=Et
28b, X=N, Y=CH, R'=Et
27a
, X=CH, Y=N
, X=N, Y=CH
27b
f
1
10a
10b
, X=CH, Y=N, R'=H, R =Me
1
, X=N, Y=CH, R'=H, R =Me
Scheme 2. Synthesis of indole-based bisphosphonates of general structure 10. Reagents conditions: (a) R1NH2, CuSO4ꢁH2O, mw 160 °C, 1 h, 65–70%; (b) CDI, THF, 23 °C, 15 h
(50–70%); (c) POCl3, mw, 125 °C, 30 min (45–50%); d) 28% NH4OH(aq), 100 °C, 15 h (75–80%); (e) tetraethyl ethene-1,1-diylbis(phosphonate), CH2Cl2/DMF, 23 °C, 15 h, (60–
65%); (f) (i) TMSBr, CH2Cl2, (ii) MeOH (80–90%).
Representative analogs of indole (9) and azabenzimidazole (10)
bisphosphonates were initially screened in our hFPPS inhibition
assay at a fixed concentration of 10 lM using compound 2 and 7
resulted in some loss of potency, possibly due to the decrease in
the basicity of that nitrogen (in silico docking suggested that this
nitrogen could be involved in H-bonding with Thr 201 and Lys
200). Interestingly, the azabenzimidazole 10a was found to be
equivalent in potency to the indole 9a, whereas 10b was inactive.
Although the exact mode of binding of these analogs is not yet con-
firmed crystallographically, binding of 10a in the allylic sub-pocket
was confirmed by competition studies using 1H line broadening
NMR. Broadening of the 1H resonances of inhibitor 10a were
observed in the presence of Mg2+ ions and hFPPS (Fig. 2), suggest-
ing fast exchange between the free and enzyme-bound state. Addi-
tion of inhibitor 2 to the same NMR sample restored the 1H
resonances of 10a, suggesting that the two compounds compete
for binding in the allylic sub-pocket of the hFPPS active site
(Fig. 2). However, addition of inhibitor 7 to the hFPPS/10a NMR
solution did not produce any change in the 1H resonances of 10a,
suggesting that this compound binds exclusively, or predomi-
nantly, in the active site.
as the positive controls. The inhibition data observed for some of
the most important analogs are summarized in Table 1 (% inhibi-
tion of hFPPS; average values of triplicate measurements with
standard deviation <10%). The main objective of this study was
to probe the ability of a single bisphosphonate inhibitor to simul-
taneously occupy both the allylic and IPP sub-pocket of the active
site cavity and also possibly bind in the allosteric pocket under bio-
logically relevant conditions (i.e., in the presence of Mg2+). Thus a
detailed structure–activity relationship study was not pursued
and a full dose response inhibition curve (IC50) was determined
only for the most potent analogs 9i (1.1 lM); the IC50 values of
N-BP drug 2 (11 nM) and the allosteric inhibitor 7 (0.92 lM) were
also determined under the same assay conditions.
The inhibition observed with analogs 9a,b,c suggested that
small alkyl groups at R1 (i.e., methyl, ethyl, isopropyl) did not
adversely affect potency (Table 1). Similarly, a small alkyl substitu-
ent at R2 was also well tolerated (e.g., 9d,e,f). However, removal of
the R1 alkyl group from the indole nitrogen (Table 1; 9d vs 9e),
We subsequently investigated the indoles of general structure 9
with alkyl groups at C-2 of increasing size and steric bulk, which
produced an apparent (albeit very modest) improvement in

D. Gritzalis et al. / Bioorg. Med. Chem. Lett. 25 (2015) 1117–1123
1121
Figure 2. 1H line broadening NMR studies exploring competitive binding of inhibitors 10a and 2 in the presence of Mg2+ ions. Stacked spectra of inhibitor free in solution and
in the presence of hFPPS and inhibitor 2; for clarity, only a section the aromatic region is shown; the proton resonances of 10a are labeled at the bottom of the stacked spectra.
The molar ratio of ligands and protein are indicated in brackets next to each spectrum; the concentrations of 10a and MgCl2 in the sample were 100 lM and 2 mM,
respectively.
potency (Table 1; 9e–9h). Competitive binding of analogs 9h and
N-BP 2, or the allosteric inhibitor 7, was monitored by 1H line
broadening NMR. These studies revealed that even in the presence
of excess Mg2+ ions, compound 9h (i.e., the analog with the largest
R2 substituent) exhibited a dual binding mode, competing for bind-
ing mainly with 7 and to a smaller extent with the active site inhib-
itor 2 (Fig. 3a and b). Similarly, analog 9i, having a smaller R2
moiety than 9h and more similar to the prenyl tail of IPP, exhibited
clear competitive binding with both 2 and 7 in the presence of
Mg2+ ions (Fig. 3c and d).
interacts with the protein in a significantly different manner. One
of the phosphonates forms a direct polar interaction with the side
chain nitrogen of Asn59 and an indirect electrostatic interaction
with the guanidinium ion of Arg60 via a co-bound phosphate ion
and a water molecule (Fig. 4a). Interestingly, the other phospho-
nate group interacts directly with the side chain carboxylate of
Asp243 (Fig. 4a), presumably forming H-bonds upon protonation
(inter-oxygen distances = 2.5 and 2.6 Å). Although such an interac-
tion is highly unlikely, especially at pH ꢂ8 (refer to SI for crystalli-
zation protocol), the electron density data provide irrefutable
evidence (Fig. 4a). This interaction may represent an atypical pro-
tonation state, which has been observed previously to occur in the
bound state, depending on the local environment.15 In this light,
Lys57 is of special interest as its side chain stabilizes the negative
charge on the bisphosphonate moiety (together with Arg60), form-
ing a salt bridge (Fig. 4a). As a result of this interaction, the side
chain of Lys57 is fully ordered in the hFPPS/9iallo complex, unlike
in the hFPPS/6aallo complex (Fig. 4b). Another notable difference
between these two complexes is that the indole nitrogen of 9i
recruits a water molecule, which in turn forms H-bonds with the
side chain oxygen of Gln242 and the main chain carbonyl of
Lys347 (Fig. 4a). Although inhibitor 7 is somewhat structurally
related to the indole 9i, their binding interactions with the alloste-
ric pocket are also clearly different (Fig. 4c). It is noteworthy that
the R2 group of 9i does not make any close contact with the protein
surface. This observation, in addition to the accumulated
We characterized the binding interactions of 9i at the allosteric
pocket of hFPPS by determining the crystal structure of the hFPPS/
9iallo complex at 1.9 Å resolution (Fig. 4a). Since 9i also competes
for binding with the active site inhibitor 2 (Fig. 3c) hFPPS/9iallo
indicates specifically the binding of 9i to the allosteric pocket.
The protein is in the open conformation, and its overall structure
superposes well with those of the apoenzyme (PDB code: 2F7 M)
and the hFPPS/6aallo complex (PDB code: 4LPG), with core (C
a
atoms) r.m.s.d. values of 0.50 and 0.56 Å, respectively. Interest-
ingly, however, the bound inhibitors 9i and 6a, as well as the pro-
tein residues within the allosteric pocket do not overlap (Fig. 4b).
Similarly, the residues that form the opening of the allosteric
pocket (i.e., Asn59, Arg60, Phe239, Lys347, and Ile348) show
slightly different conformations in the two complexes. While the
side chain of 9i still forms
p-stacking and hydrophobic interactions
analogous to those observed with 6a, its bisphosphonate moiety

1122
D. Gritzalis et al. / Bioorg. Med. Chem. Lett. 25 (2015) 1117–1123
Figure 3. 1H line broadening NMR studies exploring competitive binding of inhibitors 9h, 9i, 2 and 7 in the presence of 20-fold molar excess of Mg2+ ions (the molar ratio of
ligands and protein are indicated in brackets next to each spectrum; the concentrations of 9h and 9i in each sample were 100 M, whereas that of MgCl2 was 2 mM). Stacked
l
spectra of inhibitors free in solution and in the presence of hFPPS and either 2 or 7; for clarity, only sections the aromatic regions are shown; the proton resonances of 9h and
9i are labeled at the bottom of the stacked spectra.
knowledge on the mode of binding of bisphosphonates in the
allylic sub-pocket, suggests that the observed difference in potency
between analog 9e and 9h is likely due to favourable binding inter-
actions of the R2 substituent with the IPP sub-pocket.
In summary, the human FPPS has been an important therapeu-
tic target for skeletal diseases for many years. However, recent
clinical data provide evidence that hFPPS inhibitors may also be
useful for the treatment of cancer and neurodegeneration.4 In an
effort to understand the molecular recognition elements that gov-
ern the binding of ligands to the active site and/or an allosteric
pocket of this enzyme near the active site, we have explored the
binding mode of inhibitors 9 and 10 using NMR and X-ray crystal-
lography. Inhibitors such as 9g–9i were designed to possibly
simultaneously occupy both the allylic and the IPP sub-pocket of
the active site. The results of our studies suggest that this objective
was accomplished and that the R2 alkyl moiety of these com-
pounds may be binding in the region normally occupied by the
prenyl tail of IPP during catalysis. Since these analogs 9g–9i are
less potent in inhibiting hFPPS than the simpler N-BPs (e.g., 5),
which were used as the molecular blueprints for their design, fur-
ther structural optimization is required to compensate for the
induced-fit ligand binding observed with this target. Interestingly,
inhibitor 9i appears to have a higher affinity for binding in the
allosteric pocket of hFPPS than the active site, confirming that

D. Gritzalis et al. / Bioorg. Med. Chem. Lett. 25 (2015) 1117–1123
1123
These findings are consistent with the previously proposed theory
that the allosteric pocket adjacent to the IPP sub-pocket plays a
regulatory role in hFPPS catalysis and may be sensitive to the intra-
cellular levels of isoprenyl pyrophosphates. Similar to the active
site cavity, the structure of this allosteric pocket is flexible and
can accommodate various scaffolds binding in slightly different
orientations; such biological targets are not uncommon, but they
make the design of high affinity ligands more challenging.
Acknowledgments
We are greatful to C.-Y. Leung for his assistance with the in vitro
inhibition assays.
We gratefully acknowledge the financial support from the Nat-
ural Sciences and Engineering Research Council of Canada (NSERC)
and the Canadian Institute of Health Research (CIHR) for grants to
A.M.B. and Y.S.T.
Supplementary data
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.bmcl.2014.12.
089.
References and notes
1. Russell, R. G. G. Bone 2011, 49, 2.
2. Ebetino, F. H.; Hogan, A.-M.; Sun, S.; Tsoumpra, M. K.; Duan, X.; Triffitt, J. T.;
Kwaasi, A. A.; Dunford, J. E.; Barnett, B. L.; Oppermann, U.; Lundy, M. W.; Boyde,
A.; Kashemirov, B. A.; McKenna, C. E.; Russell, R. G. Bone 2011, 49, 20.
3. (a) 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.
ChemMedChem 2006, 1, 267; (b) Kavanagh, K. L.; Guo, K.; Dunford, J. E.; Wu, X.;
Knapp, S.; Ebetino, F. H.; Rogers, M. J.; Russell, R. G. G.; Oppermann, U. Proc.
Natl. Acad. Sci. U.S.A. 2006, 103, 7829.
4. Park, J.; Matralis, A.N.; Berghuis, A.M.; Tsantrizos, Y.S. Fron. Chem. 2014, article
50, http://dx.doi.org/10.3389/fchem.2014.00050.
5. Marma, M. S.; Xia, Z.; Stewart, C.; Coxon, F.; Dunford, J. E.; Baron, R.;
Kashemirov, B. A.; Ebetino, F. H.; Triffitt, J. T.; Russell, G. G.; McKenna, C. E. J.
Med. Chem. 2007, 50, 5967.
6. Martin, M. B.; Arnold, W.; Heath, H. T., III; Urbina, J. A.; Oldfield, E. Biochem.
Biophys. Res. Commun. 1999, 263, 754.
7. Widler, L.; Jaeggi, K. A.; Glatt, M.; Müller, 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. J. Med. Chem. 2002, 45, 3721.
8. Takeuchi, M.; Sakamoto, S.; Kawamuki, K.; Kurihara, H.; Nakahara, H.; Isomura,
Y. Chem. Pharm. Bull. 1998, 46, 1703.
9. (a) Ebetino, F. H.; Mazur, A.; Lundy, M. W.; Russell, R. G. G. WO 2010/033978
A2.; (b) Ebetino, F. H.; Mazur, A.; Lundy, M. W.; Russell, R. G. G. WO 2010/
033980 A2.; (c) Ebetino, F. H.; Mazur, A.; Lundy, M. W.; Russell, R. G. G. WO
2010/033981 A2.
10. De Schutter, J. W.; Park, J.; Leung, C.-Y.; Gormley, P.; Lin, Y.-S.; Hu, Z.; Berghuis,
A. M.; Poirier, J.; Tsantrizos, Y. S. J. Med. Chem. 2014, 57, 5764.
11. Simoni, D.; Gebbia, N.; Invidiata, F. P.; Eleopra, M.; Marchetti, P.; Rondanin, R.;
Baruchello, R.; Provera, S.; Marchioro, C.; Tolomeo, M.; Marinelli, L.; Limongelli,
V.; NovellinKwaasi, A.; Dunford, J.; Buccheri, S.; Caccamo, N.; Dieli, F. J. Med.
Chem. 2008, 51, 6800.
Figure 4. Bisphosphonate binding at the hFPPS allosteric site. (a) Structure of the
hFPPS/9iallo complex (PDB code: 4QXS). The green mesh represents a simulated-
annealing omit map (FO ꢀ FC, contoured at 3 sigma) for 9i, and the blue mesh, a
2FO ꢀ FC map (1 sigma) for Asp243. (b) Previously determined structure of the
hFPPS/6aallo complex (PDB code: 4LPG) shown in the same orientation. The
structure of the hFPPS/9iallo complex is superposed, with only the inhibitor
displayed semitransparently. (c) The structure of the hFPPS/7 complex (PDB code:
3N6 K) with 9i superposed. The carbon backbone of 9i, 6a and 7 are highlighted in
blue, pink and yellow, respectively. Oxygen, nitrogen, sulfur and phosphorus atoms
are colored in red, blue, yellow and orange, respectively. Red spheres represent
water molecules.
12. Lacbay, C. M.; Mancuso, J.; Lin, Y.-S.; Bennette, N.; Götte, M.; Tsantrizos, Y. S. J.
Med. Chem. 2014, 57, 7435.
13. Chrisman, W.; Knize, M. G.; Tanga, M. J. J. Heterocycl. Chem. 2008, 45, 1641.
14. Ognyanov, V. I.; Balan, C.; Bannon, A. W.; Bo, Y.; Dominguez, C.; Fotsch, C.;
Gore, V. K.; Klionsky, L.; Ma, V. V.; Qian, Y.-X.; Tamir, R.; Wang, X.; Xi, N.; Xu, S.;
Zhu, D.; Gavva, N. R.; Treanor, J. J. S.; Norman, M. H. J. Med. Chem. 2006, 49,
3719.
15. (a) Steuber, H.; Czodrowski, P.; Sotriffer, C. A.; Klebe, G. J. Mol. Biol. 2007, 373,
1305; (b) Czodrowski, P.; Sotriffer, C. A.; Klebe, G. J. Chem. Inf. Model. 2007, 47,
1590.
bisphosphonates can bind to this allosteric pocket under physio-
logically relevant conditions (i.e., in the presence of Mg2+ ions).