Development of selective, potent RabGGTase inhibitors

Article abstract of DOI:10.1021/jm300624s

Members of the Ras superfamily of small GTPases are frequently mutated in cancer. Therefore, inhibitors have been developed to address the acitivity of these GTPases by inhibiting their prenylating enzymes FTase, GGTase I, and RabGGTase. In contrast to FTase and GGTase I, only a handful of RabGGTase inhibitors have been developed. The most active RabGGTase inhibitor known until recently was an FTase inhibitor which hit RabGGTase as an off-target. We recently reported our efforts to tune the selectivity of these inhibitors toward RabGGTase. Here we describe an extended set of selective inhibitors. The requirements for selective RabGGTase inhibitors are described in detail, guided by multiple crystal structures. In order to relate in vitro and cellular activity, a high-throughput assay system to detect the attachment of [ ³H]geranylgeranyl groups to Rab was used. Selective RabGGTase inhibition allows the establishment of novel drug discovery programs aimed at the development of anticancer therapeutics.

Full text of DOI:10.1021/jm300624s

Article
pubs.acs.org/jmc
Development of Selective, Potent RabGGTase Inhibitors
E. Anouk Stigter,†,§,⊥ Zhong Guo,‡,⊥ Robin S. Bon,†,⊥ Yao-Wen Wu,‡ Axel Choidas,∥ Alexander Wolf,∥
Sascha Menninger,‡ Herbert Waldmann,†,§ Wulf Blankenfeldt,*,‡ and Roger S. Goody*,‡
‡
†Department of Chemical Biology and Department of Physical Biochemistry, Max Planck Institute of Molecular Physiology,
Otto-Hahn-Strasse 11, 44227 Dortmund, Germany
§Fakultat Chemie, Fachbereich Chemische Biologie, Technische Universitat Dortmund, Otto-Hahn-Strasse 6, 44227 Dortmund,
̈
̈
Germany
∥Lead Discovery Center GmbH, Otto-Hahn-Strasse 15, 44227 Dortmund, Germany
S
* Supporting Information
ABSTRACT: Members of the Ras superfamily of small GTPases are frequently mutated in cancer.
Therefore, inhibitors have been developed to address the acitivity of these GTPases by inhibiting their
prenylating enzymes FTase, GGTase I, and RabGGTase. In contrast to FTase and GGTase I, only a
handful of RabGGTase inhibitors have been developed. The most active RabGGTase inhibitor known
until recently was an FTase inhibitor which hit RabGGTase as an off-target. We recently reported our
efforts to tune the selectivity of these inhibitors toward RabGGTase. Here we describe an extended set
of selective inhibitors. The requirements for selective RabGGTase inhibitors are described in detail,
guided by multiple crystal structures. In order to relate in vitro and cellular activity, a high-throughput
assay system to detect the attachment of [3H]geranylgeranyl groups to Rab was used. Selective
RabGGTase inhibition allows the establishment of novel drug discovery programs aimed at the
development of anticancer therapeutics.
proteins, FTase and GGTase I have become popular drug
INTRODUCTION
■
targets.9,10 FTase and GGTase I inhibitors showed promising
results in cancer cell lines and mouse models but were not as
successful in clinical trials as expected.11 Since some FTase
inhibitors were even active in Ras-independent tumor cells,12
it has been suggested that FTase inhibition is not their only
disease modifying mode of action and that they might target
additional proteins. Indeed, Lackner et al. found that some
imidazole substituted tetrahydrobenzodiazepines (THBs), such
as BMS3 (1; see Figure 1) developed by Bristol-Myers Squibb
Prenylation is one of the crucial post-translational modifications
that allow certain proteins to function properly after their
synthesis in the cytosol.1−3 The largest protein family that
undergoes prenylation is the Ras superfamily of small GTPases,
in particular the Ras, Rho, Cdc42, Rac, Rap, and Rab proteins.
Most members of this family need to undergo post-translational
prenylation for their biological role.4
Different groups of the Ras superfamily are involved in
diverse processes. Rab GTPases, the largest subfamily with
more than 60 human members in humans, are key regulators of
intracellular membrane trafficking processes.5,6 Covalent attach-
ment of prenyl moieties to their C-terminal cysteines is crucial
for membrane localization and hence function of these
small GTPases. Ras and Rho proteins are farnesylated and
geranylgeranylated at a C-terminal CAAX peptide sequence
(C = cysteine, A = aliphatic amino acid, X = any residue) by the
enzymes farnesyltransferase (FTase) and geranylgeranyltrans-
ferase I (GGTase I), respectively.7,8 In contrast, Rab GTPases
are mono- or digeranylgeranylated at one or two C-terminal
cysteines. Rab prenylation is mediated by Rab geranylgeranyl-
transferase (RabGGTase), which catalyzes two sequential
geranylgeranylations for most Rabs. Because Rab GTPases do
not contain a specific recognition sequence for prenylation,
they are recruited by the Rab escort protein (REP); REP is
responsible for recognizing Rab GTPases, presenting them to
RabGGTase and delivering prenylated Rab GTPases to their
target membranes.2
Figure 1. Dual FTase/RabGGTase inhibitor 1 (BMS3).
as FTase inhibitors, also inhibited RabGGTase. In addition, they
found several genes associated with RabGTPases as strong
inducers of apoptosis by using an RNAi knockdown screen in
Ras proteins are often mutated in cancer cells, and since
Received: May 4, 2012
prenylation is essential for the function of these oncogenic
© XXXX American Chemical Society
A
dx.doi.org/10.1021/jm300624s | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
Figure 2. Comparison of active sites of RabGGTase, FTase, and GGTase I. (a) Surface representation of the active site of 1−RabGGTase−GGPP
(PDB access code 3PZ2). The black dashed lines indicate ligand−protein interactions. Ring A (THB core) π-stacks with Phe289. Ring B (imidazole)
binds to the zinc ion. Ring C (3-benzyl) T-stacks with Trp52 and Phe289. The sulfonyl interacts with Tyr44, and its corresponding ring D π-stacks
internally with ring A. (b) Surface representation of the active site of 1−FTase−FPP (PDB access code 3PZ4). The black dashed lines indicate
ligand−protein interactions. Ring A (THB core) π-stacks with Tyr361. Ring B (imidazole) binds to the zinc ion. Ring C (3-benzyl) T-stacks with
Trp102, Trp106, and Tyr 289. Ring D π-stacks internally with ring A and is mainly solvent exposed. (c) Superposed 1 (gray sticks) and docked 1
(green wire) in the active site of GGTase I (PDB access code 1S64). GGTase I has less aromatic residues in the binding site; therefore, stabilizing
π- and T-stacking interactions with rings A and C are not realized (red dashed lines). Docking studies show no consensus for 1 binding (best
solution is shown in green sticks). The lack of interaction partners for 1 seems to ensure selectivity over GGTase I toward RabGGTase and FTase.
Crystal structures are shown in gray, and docked structures are shown in yellow.
a
Scheme 1. Synthesis of General Building Blocks for THB-Based Libraries21
a
Reagents and conditions: (a) D-tyrosine methyl ester hydrochloride, DMAP, pyridine, reflux, 3 days, 68%; (b) BH3 in THF, reflux, 16 h, 67% (or
60% over two steps when 3 was not purified); (c) CuCN, DMF, 210 °C (microwave), 70%; (d) 4-methoxybenzenesulfonyl chloride, pyridine, 0 °C
to rt, 16 h, 71% (R = Br) or 61% (R = CN); (e) 1-methylimidazole-5-carboxaldehyde, TFA, TFAA, triethylsilane, rt, 16 h, 80% (R = Br) or 83%
(R = CN).
C. elegans. These include Rab5, Rab7, and the HOPS complex
that interacts with Rab7. In addition RNAi against the
RabGGTase-α and RabGGTase-β subunits was effective in
inducing apoptosis. In contrast, RNAi against FTase-β or GGTase
I-β had no effect on induction of apoptosis in C. elegans.13
Rab GTPases and their interacting proteins are involved in
cancer, genetic diseases, and pathogen uptake mechanisms.6,14
However, only a few RabGGTase inhibitors have been
developed.15−18 Phosphonocarboxylate inhibitors led to the
development of a low micromolar, specific RabGGTase inhib-
itor, inhibiting only the first prenylation step.19,20 Recently, we
reported a potent highly selective RabGGTase inhibitor, which
inhibits cellular Rab prenylation and proliferation of several
cancer cell lines without displaying general cytotoxicity to
human peripheral blood cells (PMBCs).21
Here we describe in detail the requirements for selective
RabGGTase inhibitors, using the tetrahydrobenzodiazepine (THB)
as a guiding scaffold. While previous studies gave first insights
into the requirements for selectivity, we have now designed selec-
tive inhibitors based on the exploration of different substituents
around the THB core and crystallization studies with both FTase
and RabGGTase. In order to relate in vitro and cellular activity, a
high-throughput assay system, using scintillation proximity
beads to detect the transfer of [3H]geranylgeranyl groups from
[3H]GGPP to Rab, was established.
RESULTS AND DISCUSSION
■
Comparison of the Active Sites: FTase, RabGGTase,
and GGTase I. The THB scaffold already ensures selectivity
for RabGGTase and FTase with respect to GGTase I.21
Comparison of the active sites for the 1−RabGGTase−GGPP
complex and the 1−FTase−FPP complex from our previously
published work with the active site of GGTase I allowed us to
delineate initial insights into activity and selectivity require-
ments.21 An overlay of the active site in 1−RabGGTase with
the active site of GGTase I shows that the THB core cannot
form similarly favorable van der Waals interactions because of
the absence of aromatic residues in the GGTase I binding site.
In contrast, the presence of multiple aromatic residues in both
the FTase and RabGGTase active sites allows efficient binding
of THBs with similar potency (Figure 2). These observations
are supported by the solutions obtained by virtual screening
carried out for FTase, GGTase I, and RabGGTase. In contrast
to the solutions obtained for 1 in FTase and RabGGTase,
no binding consensus was found for 1 in GGTase I (for details
see the Supporting Information). Small molecules comprising
B
dx.doi.org/10.1021/jm300624s | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
a
Scheme 2. Introduction of Substituents To Address the LBS
a
Reagents and conditions: (a) alkyl bromide, NaH, DMF, rt, 85−94%; (b) carbamoyl chloride or isocyanate (for R2 = H), Et3N, DCM, rt, 72−98%;
(c) TfN(Ph)Tf, Et3N, DCM, 0 °C to rt, 94%; (d) pyridine-3-boronic acid, Pd(PPh3)4, K2CO3, DME/H2O, 80 °C, 16 h, 94%.
π-stacking and T-stacking elements close to the zinc binding
group thus appear to distinguish GGTase I from RabGGTase
and FTase.
introduced by reductive amination to obtain building blocks 6a
and 6b.
Synthesis and Analysis of THBs Targeting the LBS. To
study the influence of different substitutions of 6b directed
toward the GGPP binding site on inhibitory activity and se-
lectivity, the phenol group of 6b was derivatized using a range of
reactions (Scheme 2). Alkylation was achieved in high yields by
addition of an alkyl, allyl, or benzyl bromide with sodium hydride
as a base (Table 1). Aminoacylation was performed by addition
of either an isocyanate or a carbamoyl chloride in the presence of
base, leading to the corresponding products in excellent yields.
Building block 14 was obtained in high yield using the triflating
reagent N-phenyl-bis(trifluoromethanesulfonimide), whereas
standard conditions using triflic anhydride gave 14 in only 55%.
The triflate was subjected to coupling with pyridine-3-boronic
acid under standard Suzuki coupling conditions. This variety of
modifications resulted in the first generation of inhibitors based
on 1 and emphasizes the phenolic hydroxyl group as a versatile
handle for modification.
To achieve further selectivity, there are two major differences
between RabGGTase and FTase that can be exploited. One
unique feature of RabGGTase is a tunnel adjacent to the GGPP
binding site, the so-called TAG tunnel, identified during the
evaluation of peptide-based selective RabGGTase inhibitors.17
The second opportunity to create selective inhibitors is presented
by the lipid binding site (LBS). Extensions into or near the
LBS were expected to be accommodated in RabGGTase while
generating a clash with Trp102 of FTase, an aromatic residue in
FTase that is the main determinant of FPP over GGPP selectivity
(Figure 2).21 However, inhibitors extended to either one of these
sites, unexpectedly, were not sufficient to obtain high selectivity.
To obtain a deeper insight into the requirements for selectivity
toward RabGGTase with respect to FTase at the molecular level,
we developed a larger subset of THBs and, together with pre-
viously reported THBs,21 subjected them to a range of prenyla-
tion assays and crystallization experiments.
Because these extensions approach the LBS, we reasoned
that the obtained inhibitors could result in different inhibition
modes with respect to GGPP. Therefore, we decided to analyze
the compounds using a radiometric [3H]GGPP assay that was
expected to resemble the native situation more closely than
the fluorometric NBD-FPP assay used previously.21 The radio-
metric assay allows a double geranylgeranylation reaction, and
[3H]GGPP obviously has a Kd equal to that of GGPP. The
latter feature could be important to predict cellular activity
for inhibitors that are competitive with GGPP or for inhibitors
that interact with the isoprenoid group. In order to screen a
large set of substances, the radiometric assay was adapted to a
high-throughput format using scintillation proximity beads (see
Experimental Section for details).
Synthesis of the General Building Block. The necessary
building blocks for the THB series were synthesized according
to previously published procedures (Scheme 1).21 In short,
condensation of 5-bromoisatoic anhydride 2 with D-tyrosine-
methyl ester led to 7-bromo-1,4-benzodiazepine-2,5-dione 3,
which was reduced to the corresponding tetrahydrobenzodiaze-
pine (THB) 4a using borane in THF. Subsequent copper
catalyzed cyanation led to the formation of 4b, the building
block for the first generation of inhibitors, meant to target the
LBS. Compound 4a, containing the 7-bromo substituent, served
as a precursor for the later generations of THBs, meant to target
the TAG tunnel. Introduction of the sulfonamide moiety via
selective N-sulfonylation by use of a weak base such as pyridine
led to 5a and 5b. The imidazolylmethyl group was subsequently
C
dx.doi.org/10.1021/jm300624s | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
a
Table 1. Synthesis and Biological Evaluation of RabGGTase Inhibitors Approaching the Lipid Binding Site
a
FL= fluorometric assay. RA= radiometric assay. ($) Lower detection limit. (#) single point measurement. ger = geranyl. IF = improvement factor.
Figure 3. Cocrystal structures of 12 and 13. (a) Surface representation of active site of the 12−FTase−FPP complex (PDB access code 4GTQ).
Similar to BMS3, the imidazole coordinates to the zinc ion. The THB interacts with Tyr361 and is further involved in internal π-stacking with the
anisylsulfonyl group. The 3-benzyl moiety extends to the back, disrupting the interactions with Trp102 and Trp106. The black dashed lines indicate
interactions between ligand and enzyme. (b) Surface representation of the active site of 12−RabGGTase complex (PDB access code 4GTT). Similar
to 1, the imidazole coordinates to the zinc ion, whereas the sulfonamide forms hydrogen bonds with Tyr44. The 3-benzyl moiety interacts with Trp52
and Phe289 by edge−face hydrophobic stacking, whereas the THB moiety stacks face−face with Phe289. The conformation is further stabilized by
internal π-stacking of the THB with the anisylsulfonyl group. The additional benzylcarbamate contributes with an edge−face hydrophobic stacking
with Phe147. The black dashed lines indicate interactions between ligand and enzyme. (c) Surface representation of active site of 13−FTase complex
(PDB access code 4GTR). The imidazole coordinates to the zinc ion. The THB interacts with Tyr361 and is further involved in internal π-stacking
with the anisylsulfonyl group. The additional diethylcarbamate functionality is adopted via a reorientation of Trp102 by 180° and forms hydrogen
bonds with Trp106. The black dashed lines indicate interactions between ligand and enzyme. (d) Surface representation of the active site of 13−
RabGGTase complex (PDB access code 4GTV). Similar to 1, the imidazole coordinates to the zinc ion, whereas the sulfonamide forms hydrogen
bonds with Tyr44. The 3-benzyl moiety interacts with Trp52 and Phe289 by edge−face hydrophobic stacking, whereas the THB moiety stacks face−
face with Phe289. The conformation is further stabilized by internal π-stacking of the THB with the anisylsulfonyl group. The additional
diethylcarbamate contributes with cation−π-stacking with Arg144. The black dashed lines indicate interactions between ligand and enzyme.
D
dx.doi.org/10.1021/jm300624s | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
a
Table 2. Synthesis and Screening of THB-Based Inhibitor Meant To Target the TAG Tunnel
a
($) Lowest detection limit. IF = improvement factor. (∗) Prepared from 17 in two steps: (a) NH2OH·H2O, Et3N, DCM/MeOH; (b) TFAA,
DCM, rt.
The assay results are summarized in Table 1. The previously
reported activity in the NBD-FPP assay is also included for
comparison.21 The IC50 values measured in the radiometric
assay for 1 resemble the previously reported activity determined
in a [3H]GGPP filter binding assay (IC50 = 21 nM).13 The data
clearly suggest a different trend for this compound set in the
[3H]GGPP and NBD-FPP assays, respectively, which in most
cases results in higher IC50 values compared to 1 (e.g., com-
pounds 7, 8, and 13). The difference in IC50 values between
the two assay systems suggests different interactions of the
different lipid substrates with RabGGTase. For this generation
of inhibitors only the incorporation of a benzylcarbamate or
3-pyridine resulted in compounds as active as 1 in the radio-
metric assay. Competition experiments with NBD-FPP indeed
showed different binding modes due to additional extensions
(for experimental details, see the Supporting Information).
Compound 10 was found to be competitive with regard to the
lipid substrate, whereas 1 was noncompetitive. Most extensions
had less effect on FTase than expected based on the predicted
clash with Trp102, thus still showing significant FTase inhibiton.
In order to find an explanation for the observed trends, crystalli-
zation studies were carried out. Inhibitors 12 and 13 could
be successfully soaked into both FTase and RabGGTase
(Figure 3). The cocrystal structures revealed that the binding
E
dx.doi.org/10.1021/jm300624s | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
additional hydrogen bond interaction is formed between
Trp106 and the diethylcarbamate group. In total, these changes
result in an approximately similar IC50 for FTase. In the complex
13−RabGGTase (Figure 3d), additional interactions are realized
by a cation−π interaction between Arg144 and the 3-benzyl
moiety (ring C).
Scheme 3. Introduction of Substituents To Address the TAG
Tunnel
a
The crystal structure analysis shows that both the inhibitors
and the FTase binding site adopt different conformations due
to the extensions of the THBs, resulting in selectivity profiles
similar to that of 1. This accentuates the need for extensions in
other or additional directions to obtain selective inhibitors.
Design, Synthesis, and Assaying of THBs Approach-
ing the TAG Tunnel. Since the new insights obtained from
the cocrystal structures indicated the need for different and/or
additional extensions to the candidate substances, we decided to
introduce additional substituents toward the RabGGTase
specific TAG tunnel. Incorporation of substituted aromatic
rings would most probably result in a clash with the RabGGTase
surface, which would disrupt the important π-stacking with
the THB moiety (ring A) and imidazole zinc binding.21 In order
to identify additional promising candidates to the previously
identified furan moiety (Table 2, entries 2 and 3), we carried out
a virtual high throughput screening (VHTS) with conforma-
tional constraints. A virtual library was created using Pipeline
Pilot by assembly of the THB core structure with various
substituents. The compounds in the library were minimized in
energy to obtain their local-minimum 3D-conformations, which
were used as input for the virtual screening. The enzymes were
prepared for docking by removal of the inhibitors and prenyl
pyrophosphates followed by the addition of hydrogens using
the Protonate3D function in MOE. All members in the virtual
library were docked in both RabGGTase and FTase using
GOLD. In RabGGTase, the conformation of the THB was fixed
between Phe289 and the zinc ion in order to mimic the
π-stacking and imidazole zinc interaction, which seems essential
for activity (see Supporting Information). Compounds that
scored high for RabGGTase and low for FTase were expected
to represent selective RabGGTase inhibitors that correctly
approach the TAG tunnel. Several groups in addition to the
furanaldehyde (16) and furannitrile (17) could be identified.
Heteroaromatic groups such as pyridine point toward the
TAG tunnel without distorting the general binding character.
a
Reagents and conditions: (a) arylboronic acid, Pd(PPh3)4, K2CO3,
DME/H2O 2/1, 40−82%; (b) amine, Mo(CO)6, Fu’s salt, Hermann’s
palladacycle, DBU, THF, 160 °C (microwave), 13−16%; (c) amine,
NaO-t-Bu, Pd2(dba)3, JohnPhos, THF, reflux or 100 °C (microwave),
30−33%.
mode of the THB moiety is highly conserved in the two
enzymes. The π-stacking of ring A with the aromatic enzyme
residues and imidazole zinc binding locks the seven-membered
ring connected to the imidazole moiety in an L-shaped fashion.
The orientation of the phenolic portion (ring C), either
connected to a phenylcarbamate or diethylaminecarbamate,
appears to be more flexible; 12 shows a conformational change
in this area, while 13 adopts the same conformation as 1.
In the complex 12−FTase−FPP structure (Figure 3a) this
conformational change, which also leads to loss of some
hydrophobic interactions, is the main difference between 12
and parent compound 1. However, this difference does not
result in a decrease in potency of inhibition. In the complex
12−RabGGTase structure (Figure 3b) an additional hydrogen
bond between the sulfonyl group of the ligand and Tyr44 is
formed. There also appears to be an additional hydrophobic
interaction between the introduced benzylcarbamate and the
lipid binding site of the enzyme. The cocrystal structure of
13−FTase (Figure 3c) revealed that the Trp102 residue,
originally contributing with an edge−face T-stacking interaction,
is flipped by 180°. This opens up extra space in the hydro-
phobic area, allowing the extended THBs to bind. However, an
Figure 4. Cocrystal structures of 16 in FTase and RabGGTase. (a) Surface representation of active site of 16−FTase−FPP complex (PDB access
code 4GTP). The imidazole coordinates to the zinc ion. The THB interacts with Tyr361 and is further involved in internal π-stacking with the
anisylsulfonyl group. The additional furanal barely fits in the FTase pockets and induces a conformational change of the 3-benzyl moiety, disrupting
the favorable hydrophobic interactions with Trp106. The black dashed lines indicate interactions between ligand and enzyme. (b) Surface
representation of the active site of 16−RabGGTase complex (PDP access code 4GTS). Similar to 1, the imidazole coordinates to the zinc ion,
whereas the sulfonamide forms hydrogen bonds with Tyr44. The 3-benzyl moiety interacts with Trp52 and Phe289 by edge−face hydrophobic
stacking, whereas the THB moiety stacks face−face with Phe289. The conformation is further stabilized by internal π-stacking of the THB with the
anisolylsulfonyl group. The additional furanal occupies the entrance of the TAG tunnel, but the expected hydrogen bond interaction between the
aldehyde and Tyr30 cannot be observed. The black dashed lines indicate interactions between ligand and enzyme.
F
dx.doi.org/10.1021/jm300624s | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
a
Scheme 4. Synthesis of THBs Targeting Both the TAG Tunnel and LBS
a
Reagents and conditions: (a) alkyl bromide, NaH, DMF, rt, 69−81%; (b) benzyl isocyanate, Et3N, DCM, rt, 71−86%; (c) TfN(Ph)Tf, Et3N, DCM,
0 °C to rt, 86−99%; (d) arylboronic acid, Pd(PPh3)4, K2CO3, DME/H2O, 80 °C, 68−84%; (e) aniline, NaO-t-Bu, Pd2(dba)3, JohnPhos, toluene,
reflux, 43−75%.
Other suggestions from VHTS were flexible groups such as N-
acetylpiperazine, which could adjust toward the TAG tunnel.
This group would have the additional advantage of increasing
the solubility of the compounds.
crystal structures of 16−FTase−FPP and 16−RabGGTase
(Figure 4).
The cocrystal structure of 16−FTase−FPP revealed that the
addition of the furanal at R1 results in a compressed binding
mode. The furanal moiety barely fits into the small hydrophobic
space and, as expected, orients toward the exit groove of the
active site, which seems to result in a conformational change at
the other end of the ligand. In order to keep the favorable
conformation for π-stacking and zinc binding, the phenolic
portion of the molecule turns by 180°, a similar phenomenon as
seen for ligand 12. This causes the loss of favorable hydrophobic
interactions with the binding site Trp residues (Figure 3a).
In complex 16−RabGGTase, the furanaldehyde moiety orients
toward the TAG tunnel. Further, the structure closely resembles
the 1−RabGGTase complex, showing the same binding mode
and conformation (Figure 3b). The inhibitor does not occupy
the lipid binding site, which was expected from both VHTS and
competition experiments.
The crystal structures of the complexes with compounds
targeting either the LBS or the TAG tunnel complement the in
vitro data. Obviously, extension of the THB in only one direction
is not sufficient for selectivity. The predicted clashes with the
binding site of FTase were circumvented via reorientation of
Trp102 in the enzyme or via adoptive behavior of the ligand.
Design, Synthesis, and Assaying of Inhibitors
Approaching Both the LBS and TAG Tunnels. Interest-
ingly, in FTase, the introduction of the furanal moiety induces a
conformational change of the 3-benzyl moiety, which
approaches the surface region of FTase composed of residues
Ala151, Arg202, and Trp102. This could also explain the
selectivity of previously reported 28, bearing extensions toward
both sides.21 This new binding pose thus shows an increased
likelihood to obtain selective RabGGTase inhibitors by combining
The suggestions from VHTS could be verified by synthesis
using transition metal based cross-coupling procedures on
building block 6a (Scheme 3). The Suzuki coupling with
heteroaromatic rings resulted in low to moderate yields (e.g., 19,
40%), most probably related to the instability of the boronic
acid counterpart. Aminocarbonylation using Herrmann’s pallada-
cycle, air stable Fu’s salt ([(t-Bu)3PH]BF4), and molybdenum
hexacarbonyl under microwave irradiation afforded 24 and 25.
Buchwald−Hartwig couplings of 6a and N-acetylpiperazine or
aniline gave THBs 26 and 27. The new compound set was
screened against the three prenyltransferases (Table 2).
Table 2 reveals that the introduction of five-membered
rings as well as the introduction of pyridine derivatives and
amine linkers led to enhanced selectivity for RabGGTase (1- to
4.4-fold enhancement). The introduction of larger groups
such as amides (entries 8 and 9) results in a slight preference
for FTase. Interestingly, the previously reported para-chloro-
substituted THB (18) still showed some remaining activity in
the radioactive assay, although it is 100-fold less potent than 1.
The binding mode of this second generation of inhibitors,
determined for 18 and 20, was, as expected, noncompetitive
with respect to NBD-FPP (see Supporting Information).
However, also for this second generation of inhibitors,
complete selectivity was not reached for RabGGTase over
FTase. One plausible explanation could be that the additional
groups could orient toward the exit groove of FTase, thereby
preventing a clash with the surface. In order to test this
hypothesis, we subjected the inhibitors to crystallization studies
in both FTase and RabGGTase. We successfully solved the
G
dx.doi.org/10.1021/jm300624s | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
a
Table 3. Synthesis and Evaluation of Inhibitors Targeting Both the RabGGTase TAG Tunnel and the LBS
a
(§) Method used. ($) Lower detection limit. nc = not calculated. nd = not determined. IF = improvement factor.
substituents that approach both the TAG tunnel and the
LBS. In order to obtain a representative, larger set of selective
substances, we combined the most active groups of generations
1 and 2 to obtain additional active and selective RabGGTase inhib-
itors. In addition, we carried out an additional virtual screening
with GGPP in the binding site of RabGGTase to identify sub-
stituents that were predicted to be compatible with GGPP (see
Supporting Information). These GGPP-compatible groups would
give the opportunity to compare the difference between compet-
itive and uncompetitive inhibitors.
To obtain a set of inhibitors approaching both the LBS and
the TAG tunnel, we followed a similar strategy as described
above. First R1 was decorated by means of transition-metal-
catalyzed coupling reactions, after which R2 was substituted using
various strategies (Scheme 4). The results for this generation are
listed in Table 3. Introduction of substituents pointing toward
both the TAG tunnel and the LBS resulted in a set of selective
RabGGTase inhibitors (28, 29, 30, 31, 32). In contrast to the
data previously reported using the fluorometric NBD-FPP
assay21 all combinations resulted in a 10- to 1000-fold decrease
in inhibition of RabGGTase, which would suggest competitive
binding modes for most compounds with respect to GGPP.
Competition experiments with 28 and NBD-FPP showed,
however, that 28 was noncompetitive with respect to NBD-
FPP. The GGPP-compatible design principle, represented by
compounds 34, 35, 39, and 40, showed mixed success. The
THBs decorated with aniline (35, 40), which were predicted to
give stabilizing hydrophobic interactions with the two terminal
isoprene units of GGPP, were practically inactive for both en-
zymes. In contrast, the THBs extended with 3-pyridine (34, 39)
that should give similar stabilizing hydrophobic interac-
tions yielded the most active inhibitors targeting the LBS.
H
dx.doi.org/10.1021/jm300624s | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
Figure 5. (a) Surface representation of the active site of the 28−RabGGTase complex (PDB access code 3PZ3). Similar to 1, the imidazole
coordinates to the zinc ion, whereas the sulfonamide forms hydrogen bonds with Tyr44. The 3-benzyl moiety interacts with Trp52 and Phe289 by
edge−face hydrophobic stacking, whereas the THB moiety stacks face to-face with Phe289. The conformation is further stabilized by internal
π-stacking of the THB with the anisolylsulfonyl group. The additional furanal occupies the entrance of the TAG tunnel, but the expected hydrogen
bond interaction between the aldehyde and Tyr30 cannot be observed. The additional benzylcarbamate contributes with an edge−face hydrophobic
stacking with Phe147. The black dashed lines indicate interactions between ligand and enzyme (adapted from Bon et al.21). (b) Surface
representation of the active site of 13−FTase (PDB access code 4GTR) with 34 docked in the active site. The docking solution shows that R2 would
fit in the enlarged pocket caused by Trp102 rotation. An overlay of the original Trp102 position exemplifies the clash.
Unfortunately, this was the case for both RabGGTase and FTase.
This finding suggests that the introduction of this moiety leads
to a similar behavior in FTase as observed with compound 13
(Figure 3).
In order to find an explanation for the observed dual activity
of compound 34, we docked the structure to the FTase surface
of 13−FTase−FPP and compared it to the solution obtained
for the docking in 1−FTase−FPP. Indeed, the docking score
observed in 13−FTase−FPP was significantly higher than the
original score in 1−FTase−FPP (Figure 5b). This supports
the hypothesis that the flexibility of Trp102 leads to more tolerance
toward enlarged THBs.
inhibition. In this case, the inhibition of proliferation of HeLa
cells was retained whereas the activity against HCT116 and
A2780 slightly dropped. These data combined suggest that
HCT116 and A2780 cells are more sensitive to FTase inhibition
than HeLa cells.
Comparison of inhibitors 31 and 34 shows that the GGPP
compatible design is not as crucial for inhibition of cellular
prenylation as has been observed in the in vitro assays. One
explanation could be that the GGPP concentration in the cells is
different from the GGPP concentration in the radiometric assay
and hence has less impact on the inhibition. The concentration of
GGPP in NIH/3T3 cells has been determined at 0.145 pmol/106
cells,23 which would correspond to approximately 65 nM GGPP
and is 30-fold less than that used in the in vitro assay.
Inhibitor 17, which shows a remarkable cellular inhibition of
RabGGTase and moderate in vitro inhibition of FTase, shows a
significant effect on cancer cell line proliferation, giving IC50
values in the range of 2−21 nM. This demonstrates the potential
of low nanomolar inhibitors of RabGGTase to inhibit cancer cell
proliferation.
Attempts to crystallize the third generation of inhibitors
with FTase and RabGGTase were successful for 28 with
RabGGTase, as reported previously (Figure 5a).21
Cellular Assays. The complete set of inhibitors was evaluated
for their potential to inhibit cellular RabGGTase prenylation
using a previously described method (reprenylation assay)22 and
to inhibit proliferation of several cancer cell lines. The results are
summarized in Table 4. It is important to note that most tested
inhibitors are inactive up to 10 μM in the PBMC assay, providing
first indications that this compound class is not generally
cytotoxic.
CONCLUSION AND OUTLOOK
■
Diversification of the THB scaffold by an iterative cycle of
synthesis, screening, and design guided by crystal structure
determination led to the conversion of a dual FTase/RabGGTase
inhibitor into potent, selective, not generally cytotoxic
RabGGTase inhibitors. Selective inhibitor 31 (49 nM, cellular
IC50) inhibits cancer cell proliferation as potently as dual inhibitor
1 and shows the potential of selective RabGGTase inhibition
for novel drug discovery programs aimed at the development of
anticancer therapeutics.
The compounds all have different selectivity profiles.
Compound 14 is a selective FTase inhibitor, whereas 28 and
31 are selective RabGGTase inhibitors. The other compounds
are all dual inhibitors, with different potencies for RabGGTase
and FTase. It could be shown that the selective RabGGTase
inhibitors (28, 31) both resulted in inhibition of cancer cell
proliferation. Compound 31, a selective inhibitor of RabGGTase
with a similar cellular potency against RabGGTase as 1, inhibited
proliferation of all cancer cell lines as potently as the dual
inhibitor 1. Interestingly, the selective FTase inhibitor 14 with a
similar FTase potency as 1 only retained its activity against
A2780 cells. These results indicate that inhibition of RabGGTase
alone is sufficient to inhibit cancer cell proliferation, whereas the
efficiency of inhibition of FTase seems to be more cancer cell
line specific. Close inspection of the data revealed similar trends.
Dual inhibitor 13, which showed potent FTase and moderate
RabGGTase activity, effectively inhibited proliferation of the
HCT116 and A2780 cells but showed a clear drop in activity for
HeLa cells. These data were complemented by the activity profile
of dual inhibitor 12, showing potent RabGGTase and FTase
The selectivity prerequisites for RabGGTase, FTase, and
GGTase I uncovered here could be used while exploring
additional scaffolds. An interesting approach that could be
pursued would be a fragment-based design. Crystal structures
with (low affinity) fragments could reveal additional RabGG-
Tase specific interaction partners. In addition, it would increase
the chance to identify motifs that would accommodate to the
TAG tunnel.
Additional studies to confirm the therapeutic relevance of
RabGGTase will give extra insight into the effect of RabGGTase
inhibition and hence Rab GTPase disturbance.
I
dx.doi.org/10.1021/jm300624s | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
EXPERIMENTAL SECTION
■
For full details, refer to the Supporting Information.
Virtual High Throughput Screening (VHTS). The virtual library
was prepared by decoration of the tetrahydrobenzodiazepine (THB)
core with 62 different R1 and 62 different R2 groups (62 × 62). The
library was prepared in Pipeline Pilot. The structures were assembled,
and stereoisomers and tautomers were enumerated. This database was
minimized in MOE (version 2009.10) using database minimize, with
MMFF94x force field and an rms gradient of 0.1. The crystal structures
of RabGGTase, FTase, and GGTase I were prepared for docking
by removal of the inhibitor and the prenylpyrophosphate, followed by
the addition of hydrogens using the Protonate3D function in MOE.
Ligands were docked into the proteins using GOLD (version 4.1.1).
Since the imidazole is known to bind to the zinc atom, we set a
binding constraint by defining an imidazole substructure with defined
distance to the zinc atom (min 1.5 Å, max 3.5 Å, spring constant = 5).
Further, the binding site was defined by a sphere of radius 15 Å around
the phenylalanine/tyrosine residue. The Chemscore scoring function
was used in combination with most accurate automatic genetic
algorithm settings (autoscale = 1). Ten solutions were generated for all
inhibitors. The scores for RabGGTase and FTase were compared for
each inhibitor, assuming that an inhibitor with a high score for
RabGGTase and a low score for FTase would be most selective. The
compounds that satisfied these conditions and showed a preserved
binding mode in RabGGTase were further evaluated for their synthetic
feasibility and synthesized.
In Vitro Radioactive RabGGTase Assay. RabGGTase activity
was measured employing the scintillation proximity assay (SPA),24 and
the experiment was performed in a 384-well plate format. A typical
assay (20 μL scale) contained variable amounts of tested compounds
in DMSO (0.4%), 4 nM RabGGTase, 0.5 μM geranylgeranyl
pyrophosphate [3H], 1.5 μM geranylgeranyl pyrophosphate, 2 μM
SUMO-HIS-Rab7, and 0.5 μM REP-1 in prenylation buffer (50 mM
HEPES, pH 7.5, 50 mM NaCl, 2 mM MgCl2, 0.5 mM TCEP, 20 μM
GDP, 0.01% Triton X-100, 0.6% DMSO). Enzymatic reaction was
conducted at room temperature for 1 h and was stopped by addition of
50 μL of stopper-bead mix (0.625 mg/mL copper chelate yittrium
silicate SPA scintillation beads, 36% (w/v) ethanol, 50% (w/v) 2×
PBS buffer with 0.4% BSA, 11% (w/v) H2O). Beads were allowed to
settle in the dark for 8 h. Plates were then measured employing
a Wallac 1450 MicroBeta Trilux scintillation counter (Perkin-Elmer
LAS GmbH). The readout of probes without inhibitor was defined as
100% activity, whereas the data of reactions without RabGGTase were
defined as 0%. IC50 determinations were calculated employing Quattro
“Workflow” software (Quattro Research GmbH).
ASSOCIATED CONTENT
■
S
* Supporting Information
Full experimental details, virtual screening procedures, crystal
structure determination, and assay descriptions. This material is
available free of charge via the Internet at http://pubs.acs.org.
AUTHOR INFORMATION
■
Corresponding Author
*For W.B.: (phone) (49) 921 55 2427; (e-mail) wulf.blankenfeldt@
uni-bayreuth.de; (present address) Lehrstuhl fur Biochemie,
̈
Universitat Bayreuth, Universitatsstraße 30, 95447 Bayreuth,
̈
̈
Germany. For R.S.G.: (phone) (49) 231 133 2300; (e-mail)
goody@mpi-dortmund.mpg.de.
Author Contributions
⊥These authors contributed equally.
Notes
The authors declare no competing financial interest.
J
dx.doi.org/10.1021/jm300624s | J. Med. Chem. XXXX, XXX, XXX−XXX

Journal of Medicinal Chemistry
Article
(13) Lackner, M. R.; Kindt, R. M.; Carroll, P. M.; Brown, K.;
Cancilla, M. R.; Chen, C.; de, S. H.; Franke, Y.; Guan, B.; Heuer, T.;
Hung, T.; Keegan, K.; Lee, J. M.; Manne, V.; O’Brien, C.; Parry, D.;
Perez-Villar, J. J.; Reddy, R. K.; Xiao, H.; Zhan, H.; Cockett, M.;
Plowman, G.; Fitzgerald, K.; Costa, M.; Ross-Macdonald, P. Chemical
genetics identifies Rab geranylgeranyl transferase as an apoptotic target
of farnesyl transferase inhibitors. Cancer Cell 2005, 7, 325−336.
(14) Hutagalung, A. H.; Novick, P. J. Role of Rab GTPases in
membrane traffic and cell physiology. Physiol. Rev. 2011, 91, 119−149.
(15) Deraeve, C.; Guo, Z.; Bon, R. S.; Blankenfeldt, W.; Dilucrezia,
R.; Wolf, A.; Menninger, S.; Stigter, E. A.; Wetzel, S.; Choidas, A.;
Alexandrov, K.; Waldmann, H.; Goody, R. S.; Wu, Y. W. Psoromic acid
is a selective and covalent Rab-prenylation inhibitor targeting
autoinhibited RabGGTase. J. Am. Chem. Soc. 2012, 134, 7384−7391.
(16) Guo, Z.; Wu, Y. W.; Tan, K. T.; Bon, R. S.; Guiu-Rozas, E.;
Delon, C.; Nguyen, T. U.; Wetzel, S.; Arndt, S.; Goody, R. S.;
Blankenfeldt, W.; Alexandrov, K.; Waldmann, H. Development of
selective RabGGTase inhibitors and crystal structure of a RabGGTase-
inhibitor complex. Angew. Chem., Int. Ed. 2008, 47, 3747−3750.
(17) Tan, K. T.; Guiu-Rozas, E.; Bon, R. S.; Guo, Z.; Delon, C.;
Wetzel, S.; Arndt, S.; Alexandrov, K.; Waldmann, H.; Goody, R. S.;
Wu, Y. W.; Blankenfeldt, W. Design, synthesis, and characterization of
peptide-based rab geranylgeranyl transferase inhibitors. J. Med. Chem.
2009, 52, 8025−8037.
(18) Watanabe, M.; Fiji, H. D.; Guo, L.; Chan, L.; Kinderman, S. S.;
Slamon, D. J.; Kwon, O.; Tamanoi, F. Inhibitors of protein
geranylgeranyltransferase I and Rab geranylgeranyltransferase identi-
fied from a library of allenoate-derived compounds. J. Biol. Chem.
2008, 283, 9571−9579.
(19) Baron, R. A.; Tavare, R.; Figueiredo, A. C.; Blazewska, K. M.;
Kashemirov, B. A.; McKenna, C. E.; Ebetino, F. H.; Taylor, A.; Rogers,
M. J.; Coxon, F. P.; Seabra, M. C. Phosphonocarboxylates inhibit the
second geranylgeranyl addition by Rab geranylgeranyl transferase. J.
Biol. Chem. 2009, 284, 6861−6868.
ACKNOWLEDGMENTS
■
We thank the X-ray communities of the Max-Planck-Institut fur
Molekulare Physiologie (Dortmund, Germany) and the Max-
̈
Planck-Institut fur Medizinische Forschung (Heidelberg,
̈
Germany) for collecting diffraction data at the Swiss Light
Source of the Paul Scherrer Institute (Villigen, Switzerland) and
for giving us generous access to and support for station X10SA.
The research leading to these results has received funding from
the European Research Council under the European Union’s
Seventh Framework Programme (FP7/2007-2013)/ERC Grant
Agreement No. 268309. This work was supported in part by
DFG grants to H.W. and R.S.G. (Grant No. SFB642) and by
the Zentrum fur Angewandte Chemische Genomik. R.S.B.
̈
thanks the Alexander von Humboldt Stiftung for a fellowship.
E.A.S. thanks the IMPRS-CB for a Ph.D. scholarship.
ABBREVIATIONS USED
■
DBU, 1,8-diazabicycloundec-7-ene; DCM, dichloromethane;
DMAP, N,N-dimethylaminopyridine; DME, dimethoxyethane;
DMF, N,N-dimethylformamide; FPP:, farnesyl pyrophosphate;
FTase, farnesyltransferase; GGPP, geranylgeranyl pyrophos-
phate; GGTase I, geranylgeranyltransferase; LBS, lipid binding
site; NBD-FPP, 3,7,11-trimethyl-12-(7-nitrobenzo[1,2,5]-
oxadizao-4-ylamino)dodeca-2,6,10-trienyl pyrophosphate;
RabGGTase:, rab geranylgeranyltransferase; REP, rab escort
protein; TAG, tunnel adjacent to GGPP binding site; Tf, tri-
fluoromethylsulfonyl; TFA, trifluoroacetic acid; TFAA, tri-
fluoroacetic anhydride; THB, tetrahydrobenzodiazepine; THF,
tetrahydrofuran; VHTS, virtual high throughput screening
(20) McKenna, C. E.; Kashemirov, B. A.; Blazewska, K. M.; Mallard-
Favier, I.; Stewart, C. A.; Rojas, J.; Lundy, M. W.; Ebetino, F. H.;
Baron, R. A.; Dunford, J. E.; Kirsten, M. L.; Seabra, M. C.; Bala, J. L.;
Marma, M. S.; Rogers, M. J.; Coxon, F. P. Synthesis, chiral high
performance liquid chromatographic resolution and enantiospecific
activity of a potent new geranylgeranyl transferase inhibitor, 2-
hydroxy-3-imidazo[1,2-a]pyridin-3-yl-2-phosphonopropionic acid. J.
Med. Chem. 2010, 53, 3454−3464.
(21) Bon, R. S.; Guo, Z.; Stigter, E. A.; Wetzel, S.; Menninger, S.;
Wolf, A.; Choidas, A.; Alexandrov, K.; Blankenfeldt, W.; Goody, R. S.;
Waldmann, H. Structure-guided development of selective RabGGTase
inhibitors. Angew. Chem., Int. Ed. 2011, 50, 4957−4961.
(22) Nguyen, U. T.; Guo, Z.; Delon, C.; Wu, Y.; Deraeve, C.;
Franzel, B.; Bon, R. S.; Blankenfeldt, W.; Goody, R. S.; Waldmann, H.;
Wolters, D.; Alexandrov, K. Analysis of the eukaryotic prenylome by
isoprenoid affinity tagging. Nat. Chem. Biol. 2009, 5, 227−235.
(23) Tong, H.; Holstein, S. A.; Hohl, R. J. Simultaneous
determination of farnesyl and geranylgeranyl pyrophosphate levels in
cultured cells. Anal. Biochem. 2005, 336, 51−59.
(24) Glickman, J. F.; Schmid, A.; Ferrand, S. Scintillation proximity
assays in high-throughput screening. Assay Drug Dev. Technol. 2008, 6,
433−455.
REFERENCES
■
(1) Lane, K. T.; Beese, L. S. Structural biology of protein
farnesyltransferase and geranylgeranyltransferase type I. J. Lipid Res.
2006, 47, 681−699.
(2) Leung, K. F.; Baron, R.; Seabra, M. C. Thematic review series:
lipid posttranslational modifications. Geranylgeranylation of Rab
GTPases. J. Lipid Res. 2006, 47, 467−475.
(3) Nguyen, U. T.; Goody, R. S.; Alexandrov, K. Understanding and
exploiting protein prenyltransferases. ChemBioChem 2010, 11, 1194−
1201.
(4) Colicelli, J. Human RAS superfamily proteins and related
GTPases. Sci. STKE 2004, 2004, RE13.
(5) Schwartz, S. L.; Cao, C.; Pylypenko, O.; Rak, A.; Wandinger-
Ness, A. Rab GTPases at a glance. J. Cell Sci. 2007, 120, 3905−3910.
(6) Stenmark, H. Rab GTPases as coordinators of vesicle traffic. Nat.
Rev. Mol. Cell Biol. 2009, 10, 513−525.
(7) Reiss, Y.; Goldstein, J. L.; Seabra, M. C.; Casey, P. J.; Brown, M.
S. Inhibition of purified p21ras farnesyl:protein transferase by Cys-
AAX tetrapeptides. Cell 1990, 62, 81−88.
(8) Seabra, M. C.; Reiss, Y.; Casey, P. J.; Brown, M. S.; Goldstein, J.
L. Protein farnesyltransferase and geranylgeranyltransferase share a
common alpha subunit. Cell 1991, 65, 429−434.
(9) Appels, N. M.; Beijnen, J. H.; Schellens, J. H. Development of
farnesyl transferase inhibitors: a review. Oncologist 2005, 10, 565−578.
(10) Konstantinopoulos, P. A.; Karamouzis, M. V.; Papavassiliou, A.
G. Post-translational modifications and regulation of the RAS
superfamily of GTPases as anticancer targets. Nat. Rev. Drug Discovery
2007, 6, 541−555.
(11) Sousa, S. F.; Fernandes, P. A.; Ramos, M. J. Farnesyltransferase
inhibitors: a detailed chemical view on an elusive biological problem.
Curr. Med. Chem. 2008, 15, 1478−1492.
(12) Rose, W. C.; Lee, F. Y.; Fairchild, C. R.; Lynch, M.; Monticello,
T.; Kramer, R. A.; Manne, V. Preclinical antitumor activity of BMS-
214662, a highly apoptotic and novel farnesyltransferase inhibitor.
Cancer Res. 2001, 61, 7507−7517.
K
dx.doi.org/10.1021/jm300624s | J. Med. Chem. XXXX, XXX, XXX−XXX