Crystal structure of human soluble adenylate cyclase reveals a distinct, highly flexible allosteric bicarbonate binding pocket

Article abstract of DOI:10.1002/cmdc.201300480

Soluble adenylate cyclases catalyse the synthesis of the second messenger cAMP through the cyclisation of ATP and are the only known enzymes to be directly activated by bicarbonate. Here, we report the first crystal structure of the human enzyme that reveals a pseudosymmetrical arrangement of two catalytic domains to produce a single competent active site and a novel discrete bicarbonate binding pocket. Crystal structures of the apo protein, the protein in complex with α,β-methylene adenosine 5′-triphosphate (AMPCPP) and calcium, with the allosteric activator bicarbonate, and also with a number of inhibitors identified using fragment screening, all show a flexible active site that undergoes significant conformational changes on binding of ligands. The resulting nanomolar-potent inhibitors that were developed bind at both the substrate binding pocket and the allosteric site, and can be used as chemical probes to further elucidate the function of this protein. Crystal clear design: The crystal structure of bicarbonate-bound human soluble adenylate cyclase shows how this protein discriminates between its activator and other anions. Low-affinity fragments binding at this and the nucleotide binding site induce significant conformational changes and highlight the challenges in successfully discovering new therapeutic agents.

Full text of DOI:10.1002/cmdc.201300480

CHEMMEDCHEM
FULL PAPERS
DOI: 10.1002/cmdc.201300480
Crystal Structure of Human Soluble Adenylate Cyclase
Reveals a Distinct, Highly Flexible Allosteric Bicarbonate
Binding Pocket
Susanne M. Saalau-Bethell, Valerio Berdini, Anne Cleasby, Miles Congreve, Joseph E. Coyle,
Victoria Lock, Christopher W. Murray, M. Alistair O’Brien, Sharna J. Rich, Tracey Sambrook,
Mladen Vinkovic, Jeff R. Yon, and Harren Jhoti*[a]
Soluble adenylate cyclases catalyse the synthesis of the second
messenger cAMP through the cyclisation of ATP and are the
only known enzymes to be directly activated by bicarbonate.
Here, we report the first crystal structure of the human
enzyme that reveals a pseudosymmetrical arrangement of two
catalytic domains to produce a single competent active site
and a novel discrete bicarbonate binding pocket. Crystal struc-
tures of the apo protein, the protein in complex with a,b-
methylene adenosine 5’-triphosphate (AMPCPP) and calcium,
with the allosteric activator bicarbonate, and also with
a number of inhibitors identified using fragment screening, all
show a flexible active site that undergoes significant conforma-
tional changes on binding of ligands. The resulting nanomolar-
potent inhibitors that were developed bind at both the sub-
strate binding pocket and the allosteric site, and can be used
as chemical probes to further elucidate the function of this
protein.
Introduction
Bicarbonate (hydrogen carbonate) is biochemically indispensa-
ble to most life forms. It is continually produced in cells as
a waste product of respiratory oxidation and plays an essential
role in diverse biological processes ranging from photosynthe-
sis, cellular and whole-body pH regulation, smooth and cardiac
muscle contraction and relaxation, to signal transduction.[1] In
mammals, bicarbonate regulates the concentration of the pro-
totypical second messenger 3’,5’cyclic adenosine monophos-
phate (cAMP) via the stimulation of soluble adenylate cyclases
(EC 4.6.1.1).[2–4] cAMP is solely produced by adenylate or ade-
nylyl cyclases (AC), and its intracellular levels are tightly regu-
lated. Mammalian cells express two different types of adeny-
late cyclases: the well-known transmembrane ACs (tmAC) and
the more recently discovered soluble adenylate cyclase
(solAC).[5] TmACs are a family of nine isoforms of plasma-mem-
brane-bound proteins stimulated by hormones or neurotrans-
mitters and activated by the a-subunit of the Gs protein. Each
member’s differential response to Gi, Go and Gz a-subunit, Gz
and Ca2+/calmodulin, allows strictly controlled intercellular sig-
naling to occur within the appropriate cellular context.[6,7] Solu-
ble AC (solAC) is found in the cytosol, is structurally and bio-
chemically distinct from the tmAC enzyme family, and is syn-
ergistically regulated by calcium and bicarbonate ions.[8] SolAC
works in conjunction with phosphodiesterases, protein kina-
se A (PKA) and PKA- anchoring proteins (AKAP), as well as
other effectors to establish and maintain signal transduction
pathways. The original hypothesis that cAMP could only be
produced at the cellular membrane by tmACs and would then
diffuse inside the cell to effect specific cellular responses has,
therefore, been superseded by a more comprehensive path-
[a] S. M. Saalau-Bethell, Dr. V. Berdini, Dr. A. Cleasby, Dr. M. Congreve,+
Dr. J. E. Coyle, V. Lock,++ Dr. C. W. Murray, Dr. M. A. O’Brien,+ S. J. Rich,
T. Sambrook,+++ Dr. M. Vinkovic, Dr. J. R. Yon, Dr. H. Jhoti
Astex Pharmaceuticals
Cambridge Science Park, Cambridge, CB4 0QA (UK)
E-mail: harren.jhoti@astx.com
[+] Current address: Heptares Therapeutics Ltd, BioPark, Broadwater Road,
way and a model that postulates cAMP-specific micro-
domains.[7,9,10]
Human solAC (hsolAC) is encoded by a single gene, ADCY10,
which produces two different-length splice variants: the mini-
mal functional 48 kD variant contains solely the two heterolo-
gous catalytic domains, C1 and C2, whilst the full-length
~190 kD variant comprises additional putative regulatory do-
mains at the C terminus.[11–13] Both forms are fully active and
synergistically stimulated by bicarbonate and calcium ions.
solAC was first detected in testes and spermatozoa, and its
role was thought to be confined to fertility and capacitation,
an idea that appeared to be supported by solAC knockout
mice, which appeared fully viable except for their inability to
Welwyn Garden City, AL7 3AX (UK)
++] Current address: Gregory Fryer Associates Ltd, 10–12 St Thomas’ Pl, Ely,
[
CB7 4EX (UK)
[
+++] Current address: StemCell Sciences UK, Babraham Research Campus,
Cambridge, CB22 3AT (UK)
Supporting information for this article is available on the WWW under
http://dx.doi.org/10.1002/cmdc.201300480.
ꢀ 2014 The Authors. Published by Wiley-VCH Verlag GmbH & Co. KGaA.
This is an open access article under the terms of the Creative Commons
Attribution-NonCommercial License, which permits use, distribution and
reproduction in any medium, provided the original work is properly
cited and is not used for commercial purposes.
ꢀ 2014 The Authors. Published by Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 2014, 9, 823 – 832 823

CHEMMEDCHEM
FULL PAPERS
www.chemmedchem.org
reproduce.[14–16] Consequently, solAC was considered a potential
target for male contraception and a candidate for drug discov-
ery efforts. Later, solAC was found to be present, albeit at
lower concentrations, in most organs[10] and to be indispensa-
ble for a wide range of intracellular signaling events. Here, we
describe the first crystal structures of a mammalian solAC, and
the application of fragment-based drug discovery methods to
identify small-molecule inhibitors of the human enzyme.
with the b,g phosphates of AMPCPP, the backbone carbonyl of
Ile48, carboxylate oxygen atoms from the strictly conserved
Asp47 and Asp99 amino acids and a water molecule, which in
turn mediates an interaction with the a phosphate. Adenylate
cyclases are known to require two metal ions to catalyse the
phosphoryl transfer reaction. The observed Ca2+ ion in our
structure occupies metal site B, but no discernible density was
observed for the second metal site (A), which is also expected
to be in close vicinity to the substrate.[19] Calcium is known to
activate hsolAC by lowering the Km value for ATP–Mg2+ from
about 10 mm to 0.9 mm.[8] The reduction in the Km value for
ATP suggests that the calcium-coordinated phosphates con-
tribute a large part of the binding energy for ATP. Comparison
of the complexes of hsolAC and S. platensis solAC with
AMPCPP/Ca2+ show that in both cases, the nucleotide ana-
logues assume very similar productive conformations and are
poised for the subsequent in-line reaction.[17] Binding of the
nucleotide analogue AMPCPP/Ca2+ to hsolAC causes two
major conformational changes in the active site pocket. Firstly,
the N terminus of helix a3 unwinds by one turn upon forma-
tion of the complex-extending strand b1; and second, the b-
hairpin loop from Ala97 to Ala100 moves by approximately
4 ꢁ, presumably to accommodate substrate binding (Fig-
ure 2b). The absence of a S. platensis solAC apo structure pre-
cludes such comparison; however, movements of the corre-
sponding a-helix (a1) and b-hairpin loop (b7-b8) were ascribed
Results
Overall structure and active site
We determined the crystal structure of the short-form hsolAC
at 1.7 ꢁ resolution and show that the two domains, C1 and C2,
adopt an approximate twofold pseudo-symmetrical conforma-
tion with the sole active site positioned off-center within the
interdomain cleft (Figure 1a; see also figure S1 in the Support-
ing Information). Comparison of the hsolAC structure with that
of the previously determined orthologue from the cyanobacte-
rium Spirulina platensis shows the secondary structural ele-
ments to be similar, though the external loops exhibit signifi-
cant divergence and are generally longer in hsolAC (Fig-
ure 1b).[17] These loops consist of N-terminal residues 1–28 and
the interdomain residues 219–285, which form a subdomain of
loops and helices that packs against the catalytic core of the
enzyme. The human enzyme also displays a significant degree
of mobility, and very weak main-chain density was seen for
amino acids 135–140 and 350–356, making it hard to interpret.
Furthermore, while in the cyanobacterial enzyme there are
two identical nucleotide binding sites, in the hsolAC structure
only one of these sites is accessible because an extension of
b strands 2 and 3 in C2 and the loop linking them precludes
the binding of a second substrate molecule (Figure 1c; see
also figure S2 in the Supporting Information).
À
to the binding of HCO3 to the binary-complex crystals during
flash soaking and freezing experiments.[17]
Bicarbonate binding site
As previously indicated, in contrast to the cyanobacterium
S. platensis protein, human solAC only binds one substrate
molecule of AMPCPP as the putative second substrate binding
site is significantly remodelled due to insertions. However,
after soaking apo crystals of hsolAC with bicarbonate, we did
observe additional electron density in this region, which is con-
sistent with a bound bicarbonate ion (Figure 2c). The bicar-
bonate anion appears to be tightly held in place by hydrogen
bonds with the side chain amino group of Lys95, the back-
bone amide and carbonyl groups of Val167, the backbone
amide of Met337 and both terminal nitrogen atoms of the
guanidino group of Arg176. The side chain of Arg176, which
in the apo structure forms a water-mediated interaction with
the backbone carbonyl of Ala97, undergoes an approximate
308 swing in order to bind the bicarbonate ion disrupting the
hydrogen bond to Ala97 (Figure 2c). Since this loop is located
very close to the adenine of the substrate AMPCPP and con-
tains one of the catalytically essential aspartic acid residues
(Asp99), the conformational flexibility observed by the binding
of the bicarbonate ion and propagated via the movement of
Arg176 could influence the mechanism by which the bicarbon-
ate ion exerts its allosteric regulatory role. Further evidence of
the functional importance of one of these residues involved in
the binding of bicarbonate, Lys95, is provided by its conserva-
tion across orthologous enzymes from species including cya-
nobacterial Anabaena sp. PCC7120, S. platensis and rat solAC.[20]
Indeed, the crystal structure of the ternary complex of
hsolAC with the substrate analogue a,b-methylene adenosine
5’-triphosphate (AMPCPP) and Ca2+ ion confirms this single
ATP binding site. The adenine group of AMPCPP sits close to
the bottom of a predominantly hydrophobic pocket formed by
the side chains of Ala97, Phe296, Leu345, Phe336 and Val411
(Figure 2a). The N6 amine forms hydrogen bonds to the back-
bone carbonyl of Val406 and to a water molecule located at
the very bottom of the pocket. The environment of the ribose
moiety is more amphipathic in nature, with the ring abutting
the hydrophobic side of the pocket and making van der Waals
contact with the side chains of Ala415 and Phe338. The guani-
dino group of Arg176 points into the hydrophobic interior of
the pocket, towards the 2’ and 3’ hydroxy groups of the ribose
ring. Further along the pocket, the highly conserved Asn412
and Arg416 of the NXXXR cyclase catalytic motif straddle the
ribose ring and the a phosphate (Figure 2b). Both of these res-
idues have been implicated in the stabilization of the transition
state.[18] In our structure, the guanidino group of Arg416 forms
a hydrogen bond with the oxygen atoms from the a-phos-
phate, while the 2 amino group of Asn412 hydrogen bonds to
the oxygen atoms on the g-phosphate. The Ca2+ ion interacts
ꢀ 2014 The Authors. Published by Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 2014, 9, 823 – 832 824

CHEMMEDCHEM
FULL PAPERS
www.chemmedchem.org
Figure 1. Crystal structure of the human adenylate cyclase. a) Ribbon dia-
gram of the human soluble adenylate cyclase enzyme in complex with the
substrate analogue AMPCPP (solid spheres). Domains C1 and C2 of the
single chain are coloured blue and grey, respectively, and show the twofold
pseudosymmetry of the protein. b) Overlay of the ribbon diagrams of the
hsolAC–AMPCPP complex (blue) and the cyanobacterial solAC homodimer
from Spirulina platensis (gold) showing the N-terminus and interdomain ex-
tensions of the human enzyme. The bound AMPCPP in hsolAC illustrates the
location of the single active site. c) View of the hsolAC active site with
bound AMPCPP (blue), with a molecule of AMPCPP (grey) from the S. platen-
sis complex (PDB: 1WC0[17]) superimposed on the putative second site. The
Fragment screening and inhibitor design
To identify inhibitors of human solAC, a library of 1600 frag-
ments was screened using thermal shift (Tm), functional bioas-
says, and X-ray crystallography. The crystal structures of the
protein–fragment complexes revealed high intrinsic flexibility
in the protein, as binding of these relatively small and low-af-
finity compounds was able to alter the shapes of both the
active and allosteric sites. The majority of the fragments bound
in the bicarbonate binding site and represented carboxylic
acids and acidic heterocycles binding in their anionic form.
However, three fragment hits were also found to bind in the
ATP binding site, and a further two fragments bound in an in-
duced pocket located between the ATP and bicarbonate bind-
ing sites (Figure 3a and Table 1). Initial fragment hits were fol-
lowed up with analogue searches from an in-house repository,
based on either substructure derivatives and/or iso-structural
heterocycles. These were ranked based on their affinity in
a functional bioassay and then soaked into hsolAC crystals to
obtain additional protein–fragment crystal structures. The
most potent inhibitors from this round of screening were
found to be amino-furazanes (e.g., 4; IC50 =19 mm) and tri-
fluoro-pyrazoles (e.g., 5; IC50 =28 mm), both of which bound in
the bicarbonate binding site.
To select which fragment hits to develop into inhibitors, we
evaluated the hits on the basis of binding interactions, novelty,
synthetic accessibility, as well as affinity. The amino-furazanes
(e.g., 4) were the most potent compounds identified from the
follow-up of the initial hits. Their unequivocal binding mode in
the bicarbonate pocket, noncharged (neutral) state, and well-
positioned vectors extending towards the ATP site (Figure 3b),
fulfilled most of the criteria required to develop these into
a lead series. The amino-furazane heterocycle, a constant scaf-
fold in the whole series, mimics the donor/acceptor hydrogen
bonding pattern of the bicarbonate with the Val167 backbone.
The molecule also forms two further hydrogen bonds to the
carbonyl (C=O) and backbone NH of Met337. The absence of
a negative charge, as in the bicarbonate anion, is likely respon-
sible for the rearrangement of the positively charged Arg176
side chain, which moves out of the pocket towards the sol-
vent-exposed entrance and is replaced by the side chain of
Phe338, creating a more hydrophobic environment. The
phenyl ring of the fragment sits in a narrow hydrophobic cleft
flanked by Phe45, Phe336 and Phe338, with the latter two res-
idues undergoing substantial conformational changes induced
by fragment binding.
ꢀ 2014 The Authors. Published by Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 2014, 9, 823 – 832 825

CHEMMEDCHEM
FULL PAPERS
www.chemmedchem.org
Figure 2. Close up views of the nucleotide and bicarbonate binding sites.
a) hsolAC active site in complex with AMPCPP. The adenine pocket is shaped
by hydrophobic residues Val411, Phe336, Leu345, Phe296 (Ala97 is towards
the back and not shown). The hydrogen bonds of the N6 amine to Val406
backbone carbonyl and a water molecule (red sphere) are represented by
dashed lines. Phe338 and Arg176 flank the ribose ring. b) Comparison be-
tween the apo (yellow) and AMPCPP–Ca2+ complex (blue) structures of
hsolAC. The highly conserved Asn412 and Arg416 of the NXXXR cyclase cat-
alytic motif hydrogen bond to the ribose and the phosphate oxygens. The
Ca2+ ion (silver sphere) exhibits an octahedral coordination with b,g oxygen
atoms of the AMPCPP phosphates, a carboxyl oxygen each from the invari-
ant Asp47 and Asp99 residues, the backbone carbonyl of Ile48 and a water
molecule (red sphere). The arrows highlight the movement of the b-hairpin
loop and the unwinding of helix a3. c) Comparison of the apo (yellow) and
bicarbonate-bound (cyan) hsolAC structures. Density for bicarbonate is
shown (blue mesh). The bicarbonate anion forms hydrogen bonds (black
dashed lines) with the backbone carbonyl and NH groups of Val167, the
guanidino group of Arg176 and amine side chain of Lys95. The side chain
of Arg176 undergoes an approximate 308 swing to hydrogen bond with
bicarbonate, altering the water-mediated interaction with Ala97.
The basic strategy in optimizing fragment 4 was to grow
from the bicarbonate pocket through the induced pocket
region that lies adjacent to Phe336 and into the ATP pocket.
Initially, a number of simple linker groups were attached onto
the phenyl of fragment 4, and the 3-methoxy analogue, 6,
proved to be the most promising with similar potency and ex-
cellent growth vectors (6; IC50 =11 mm). The next round of syn-
thesis focused on molecules capable of forming a hydrogen
bond with the side chain of Arg176; this was an attractive
design idea because the side chain was easily accessible and
sits at the mouth of the ATP pocket. A simple ethyl ester at-
tachment to the methoxy group resulted in a 10-fold improve-
ment in potency (7; IC50 =1.6 mm), and the crystal structure of
the co-complex is consistent with the formation of the target-
ed hydrogen bond (Figure 3c). A number of heterocycles were
then synthesized that could retain this hydrogen bond whilst
improving the p-stacking with Phe336. This culminated in the
discovery of the novel lead compound 8 (IC50 =0.36 mm). The
designed molecule induces a movement of the side chain of
Asp99 to form an additional long hydrogen bond with the
benzimidazole of compound 8 (Figure 3d). The molecule also
forms a hydrogen bond with the flexible side chain of Arg176,
although the geometry is not ideal. The fragment-to-lead cam-
paign outlined here illustrates two important points: firstly, the
binding mode of the starting fragment, 6, was recapitulated in
the final molecule (Figure 3d); and secondly, the fragment-
induced movement of the Phe338 and Arg176 side chains
could be exploited via careful structure-based optimisation.
Discussion
The hsolAC crystal structure provides the first description on
a molecular basis of an enzyme that is functionally regulated
by a bicarbonate ion. Moreover, the structure highlights how
the human protein has substantially diverged from its cyano-
bacterial orthologue. Most notably, there is a single binding
site for the substrate ATP, with the protein symmetry-related
site (equivalent to the second ATP site in the bacterial enzyme)
now able to bind the allosteric activator bicarbonate. It is
ꢀ 2014 The Authors. Published by Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 2014, 9, 823 – 832 826

CHEMMEDCHEM
FULL PAPERS
www.chemmedchem.org
Figure 3. Fragment binding and structure-guided compound optimisation.
a) Overlay of three different crystal structures showing fragments binding in
multiple regions of the hsolAC catalytic and allosteric pockets. The phenyl
pyrazole (cyan) binds in the ATP pocket, an aromatic carboxylic acid (2;
yellow) induces a new pocket and another pyrazole (3; magenta) binds in
the allosteric bicarbonate pocket. b) Overlay of the bicarbonate (cyan) and
amino-furazane fragment (4; orange) protein structures. Both molecules in-
teract with the Val167 backbone. Compound 4 triggers movement of the
side chains of Arg176 (close to the position the residue adopts in the apo
structure) and Phe338, altering the shape and electrostatic environment of
the pocket. The ketone group of compound 4 forms a hydrogen bond with
the backbone NH of Met337, whilst the carbonyl of Met337 forms an addi-
tional hydrogen bond with compound 4. The phenyl ring of the fragment
sits in a hydrophobic pocket between Phe45 and Phe336 (not labelled,
shown faded on the back). Position 3 on the phenyl ring of compound 4
presents a good vector to access the ATP pocket. c) Complex of hsolAC with
compound 7 (orange). The ethyl ester added to the 3-methoxy group of
compound 6 exploits the movement of Arg176 observed with compound 4
(panel b) forming an additional hydrogen bond with its side chain and
shows opportunity to further extend the inhibitors into the ATP pocket.
d) Overlay of the protein complex of hsolAC with compounds 4 (cyan) and 8
(orange) showing the growth of the molecule from the original fragment. In-
teractions with both Val167 and Met337 backbones are preserved along the
series. The molecule extends towards the ATP pocket forming two further
hydrogen bonds: one with Arg176 (also seen in the complex with com-
pound 7) and an additional one, accessible through the movement of the
Asp99 side chain.
therefore clear that the fusion event that created the single
chain C1/C2 mammalian enzyme[21] has allowed differences to
evolve in catalytic mechanism and regulatory control com-
pared with the homodimeric bacterial orthologues. Intriguing-
ly, the crystal structures of mammalian tmACs also show a simi-
lar C1/C2 architecture—a single active site located at the hete-
rodimer interface and a pseudo-symmetrical, catalytically in-
competent site, where the nonphysiological forskolin activator
is seen to bind.[19,22,23]
Bicarbonate acts on mammalian solACs both to relieve sub-
strate inhibition and to increase the maximum reaction veloci-
ty (Vmax).[8] The crystal structures presented here show how the
allosteric site is exquisitely designed to recognise the bicarbon-
ate ion, which is held by six hydrogen bonds including a hydro-
gen bond to the backbone carbonyl of Val167, which allows
À
the protein to discriminate between HCO3 and alternative
anions that are unable to act as hydrogen-bond donors. The
structures also suggest that bicarbonate exerts its regulatory
effect via an allosteric mechanism, in which conformational
changes are propagated to the substrate site via Arg176 (Fig-
ure 2c). One of the key residues involved in bicarbonate bind-
ing, Lys95, has also been shown to be important in the bacte-
rial orthologue from Anabaena; site-directed mutagenesis of
the analogous residue and subsequent enzyme kinetic studies
confirmed its crucial role for activation of the enzyme by bicar-
bonate.[18] However, given the fact that the other residues in-
volved in bicarbonate binding in the hsolAC structure are not
conserved in the bacterial orthologues, the precise mechanism
by which bicarbonate regulates these proteins remains
unclear.
The paper also describes the first structurally characterised
potent inhibitors of hsolAC. The inhibitors were discovered by
fragment-based drug discovery and occupy both the newly
ꢀ 2014 The Authors. Published by Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 2014, 9, 823 – 832 827

CHEMMEDCHEM
FULL PAPERS
www.chemmedchem.org
of binding potency, therefore, these approaches can be suc-
cessful at targeting highly flexible proteins. The resulting com-
pounds described here are not only potential drug candidates,
but also potent chemical probes and might be useful for the
further characterisation of the function of soluble adenylate
cyclase in higher organisms.
Table 1. In vitro activity profiles of compounds binding to hsolAC.
Compd Structure
CLogP IC50 [mm][a] LE[b]
1
2.3
2.4
>1000
>1000
–
–
2
3
Experimental Section
Protein expression and purification
2.6
2.5
3.4
>300
19
–
Cloning: An expression vector pCDNA3.1 encoding full-length
solAC (Nucleotide NM 018417 SwissProt) was used as a template
for polymerase chain reaction (PCR) amplification. The construct
M1V469 was amplified by PCR, using 3’ and 5’ primers.
4
5
0.46
0.41
5’ primer: 5’-GGGGACAAGTTTGTACAAAAAAGCAGGCTCCATGAA-
CACTCCAAAAGAAGAATTCCAGGACTGG-3’, which carried an attB1
recognition sequence, and a sequence corresponding to bases
1–11.
28
3’ primer: 5’-GGGGACCACTTTGTACAAGAAAGCTGGGTTCAGTGG-
TGGTGGTGGTGGTGGACTTTCTCAGTACGGCCC-3’, which introduced
a sequence corresponding to bases 463–469, a six his C-terminal
tag, a stop codon, and an attB2 recognition site.
6
7
2.4
2.6
11
0.42
0.38
1.6
PCR reagents: Thermopol buffer 10X (10 mL), dNTP mix (2 mL), DNA
template (1 mL), 5’ primer (1 mL), 3’ primer (1 mL), Vent (1 mL). Reac-
tion made up to 100 mL with H2O.
8
3.6
0.36 0.35
PCR cycling: 25 cycles of 948C for 30 s, 558C 1 min, 728C for 3 min,
followed by an extension of 10 min at 728C. The 1407 bp fragment
was recombined into the entry vector, pDONR, using the BP reac-
tion of the Gateway cloning technique. Product was transformed
into DH5a competent cells; plasmid was extracted from the bacte-
ria grown on the kanamycin selective plates. The sequence was
confirmed by DNA sequencing before transferring the clone to the
destination vector, pDEST8, via the LR reaction of the Gateway
cloning technique. An aliquot of the product was transformed into
DH5a competent cells. Plasmid was extracted from the bacteria
grown on the carbenicillin-selective plates. The sequence was con-
firmed by DNA sequencing before beginning expression of the
protein. Expression of the solAC constructs was done in High Five
cells (BTI-TN-5B1-4).
[a] The half maximal inhibitory concentration (IC50) of the compounds
was measured in the bioassay as described in the Experimental Section.
Assay conditions were optimized for maximum signal, and all compounds
except for compound 7 were independently tested at least twice in du-
plicate with SD in line with 5, the control compound. Compound 5 was
used as the assay standard (n=19 independent experiments; data is the
geometric mean; SD= Æ3.8 mm). [b] Ligand efficiency (LE) is defined as
the binding affinity per heavy atom and has units of kcalmolÀ1 NÀ1, where
N is the number of non-hydrogen atoms.
identified allosteric bicarbonate site of the protein as well as
the ATP pocket. The work provides further support for the po-
tential of fragment-based screening to identify and exploit
novel allosteric pockets for inhibitor design.[24–27] Alternative
protein conformations and loop movements are difficult to
predict, and the energetics of such movements could mask
the binding of low-affinity fragments. However, we observed
several fragment molecules binding at pre-existing sites, whilst
others caused significant loop movements and induced new
additional binding pockets. Major conformational movement
induced by the binding of low-affinity fragments has been re-
ported previously and indicates that, while the affinity of frag-
ments may be low, their binding efficiency can be signifi-
cant.[28] Indeed, we were able to elaborate the original low-af-
finity fragment hits into potent lead compounds, and we be-
lieve that the iterative use of experimentally determined bind-
ing modes was important in achieving this goal.
Expression of hsolAC: Production of recombinant virus of hsolAC
was performed in the following manner. Briefly, the pDEST8 vector
encoding the relevant gene was transformed into Escherichia coli
DH10BAC cells containing the baculovirus genome (bacmid DNA).
Via a transposition event in the cells, a region of the pDEST8
vector containing the gene and a gentamycin-resistance gene in-
cluding the baculovirus polyhedron promoter was transposed di-
rectly into the bacmid DNA. By selection on gentamycin, kanamy-
cin, tetracycline and Bluo-gal, resultant white colonies should con-
tain recombinant bacmid DNA encoding the relevant gene. Bacmid
DNA was extracted from a culture of white DH10BAC cells and
transfected into Spodoptera frugiperda sf9 cells grown in SF900 II
serum-free medium following the manufacturer’s instructions.
Virus particles were harvested 72 h post-infection. An aliquot
(1 mL) of harvested virus particles was used to infect sf9 cells
(100 mL) containing 1ꢂ106 cellsmLÀ1. Cell culture medium was har-
vested 72 h post-infection.
High Five insect cells were cultured in EX-Cell 405 (JRH) serum-free
medium to a density of 1ꢂ106 cellsmLÀ1. An aliquot (5 mL) of viral
stock was added to each litre of cells. The cultures were incubated
In summary, the active site flexibility of hsolAC did not pose
a barrier to fragment screening or to subsequent optimisation
ꢀ 2014 The Authors. Published by Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 2014, 9, 823 – 832 828

CHEMMEDCHEM
FULL PAPERS
www.chemmedchem.org
at 278C for 48–72 h. Cells were harvested by centrifugation at
4000 rpm for 8 min 108C. Pellets were frozen at À808C.
MLPHARE,[30] generated difference Fourier maps that were then in-
spected for additional sites. Map quality was improved using sol-
vent flattening (Solomon and cycles of ARP/WARP). The sequence
of hsolAC was built into the resulting electron density map. After
multiple rounds of rebuilding and refinement (using QUANTA,
REFMAC and CNX) at 1.7 ꢁ using the synchrotron data, the final
soluble adenylate cyclase model consisted of residues 1–468, with
two gaps (residues 135–140 and 350–356). The final refinement
statistics are shown in table S2 in the Supporting Information.
Purification of hsolAC: All procedures were performed at 48C unless
stated otherwise. Cell pellets were thawed on ice and re-suspend-
ed in lysis buffer [50 mm Tris (pH 7.5), 300 mm NaCl, 10% glycerol,
2 mm b-mercaptoethanol (BME), protease inhibitor cocktail (Calbio-
chem)]. Cells were lysed by sonication, and lysate was incubated
with DNase 1 for 1 h at 48C. Lysate was clarified by centrifugation
at 14000 or 25000 rpm for 1 h. The clarified lysate was further
clarified by centrifugation as in the previous step, then passed
through a 0.45 mm filter before being applied to metal chelating
matrix (GE Healthcare) pre-charged with Ni2+ in batch bind mode.
The resin or matrix was poured into a column, and the solAC pro-
tein was eluted by addition of lysis buffer containing 250 mm imi-
dazole. Fractions were analysed by sodium dodecyl sulfate–poly-
acrylamide gel electrophoresis (SDS–PAGE), and those containing
solAC protein were pooled. Fractions containing hsolAC were
tracked at each step by Western blotting using an hsolAC-specific
monoclonal antibody (kindly supplied by Schering AG). The pooled
protein was buffer-exchanged into low salt by application to
a G25-desalting column equilibrated in 50 mm Tris (pH 7.5), 30 mm
NaCl, 10% glycerol, 5 mm BME. The buffer-exchanged solAC pro-
tein was then applied to a 6 mL Resource Q cation exchange
column (GE Healthcare) and eluted using a gradient of 0–30% 1m
NaCl over 20 column volumes. Fractions were analysed by SDS–
PAGE and those containing solAC protein were pooled and applied
to a 26/60 superdex-75 gel filtration column pre-equilibrated using
50 mm Tris (pH 7.5), 330 mm NaCl, 10% glycerol, 5 mm BME. Frac-
tions were analysed by SDS–PAGE. SolAC fractions were pooled
and concentrated to a final concentration of ~10 mgmLÀ1 using
a vivaspin 2 centrifugal concentrator (HY membrane).
Preparation of hsolAC–AMPCPP co-crystals: 200 mm a,b-methyl-
ene adenosine 5’-triphosphate (AMPCPP), 40 mm CaCl2 and 40 mm
MnCl2 were dissolved in 0.1m sodium acetate (pH 4.8), 0.2m triso-
dium citrate, 16% PEG4K, and 10% glycerol. A previously grown
crystal of hsolAC was placed in 20 mL of this soaking solution and
allowed to equilibrate for 3 days. The crystal was then moved to
a solution of cryoprotectant, frozen in liquid nitrogen, and X-ray
data collected.
À
Preparation of solAC–HCO3 co-crystals: 50 mm sodium bicarbon-
ate was prepared in a solution consisting of 0.1m sodium acetate
(pH 4.8), 0.2m trisodium citrate, 16% PEG4000, and 10% glycerol.
A previously grown crystal of the solAC catalytic domain was
placed in 20 mL of the bicarbonate solution and allowed to equili-
brate for 3 h. The crystal was then moved to a solution of cryopro-
tectant and X-ray data collected.
Preparation of compound–solAC co-crystals: Compounds were
soaked into hsolAC crystals at a concentration of 25 mm, the crys-
tals subsequently being frozen, and data collected (tables 1a and
1b in the Supporting Information).
RCSB Protein Data Bank accession codes: The atomic coordinates
and structure factors for the apo and co-crystal complexes de-
scribed here have been deposited in the RCSB Protein Data Bank
with accession codes: 4OYW (apo hsolAC); 4OYX (hsolAC–
AMPCPP–Ca2+); 4OYZ (hsolAC–HCO3À); 4OZ2 (hsolAC–compd 1);
4OYP (hsolAC–compd 2); 4OYO (hsolAC–compd 3); 4OYI (hsolAC–
compd 4); 4OZ3 (hsolAC–compd 5); 4OYM (hsolAC–compd 6);
4OYB (hsolAC–compd 7); 4OYA (hsolAC–compd 8).
Crystallization
Crystals of hsolAC were grown by the hanging-drop vapour diffu-
sion method. Protein solution (1 mL) was mixed with 1 mL of reser-
voir solution (0.1m sodium acetate (pH 4.8), 0.2m trisodium citrate,
16–18% PEG4K and 10% glycerol) and left to equilibrate at 48C.
Crystals appeared in the drops after 3–6 days and reached a maxi-
mum size of 0.5ꢂ0.1ꢂ0.1 mm in 14 days. Consistency of crystal
size and quality was greatly improved by microseeding.
Biological evaluation
Bioassay: The enzymatic activity of hsolAC was measured in 96-
well (black) plate format using the Mediomics Bridge-It designer
cAMP kit (cat #122935, Mediomics, LLC, St. Louis, MO, USA). The
assay reactions contained 50 mm Tris-HCl (pH 7.4), 3 mm MnCl2,
0.1% BSA, 500 mm ATP, 62.5 pm hsolAC, and test compound, and
allowed to proceed for 20 min at RT. The reaction was stopped by
adding 5 mL of 15 mm EDTA. The Bridge-It assay reagents were pre-
pared as per the manufacturer’s instructions. The Bridge-It read
mix was incubated with the stopped reactions for 1 h at RT before
the fluorescence signal was read at 535 nm on excitation at
485 nm. The results from the hsolAC curves were calibrated against
a standard curve of cAMP in assay buffer. IC50 values were generat-
ed using Graphpad Prism software (version 3.02).
Data collection and processing: Crystals of solAC were briefly
transferred to a cryobuffer solution (0.1m sodium acetate (pH 4.8),
0.2m trisodium citrate, 30% PEG4K and 10% glycerol) and plunged
into liquid nitrogen prior to data collection at 100 K.
Datasets for hsolAC as well as for the heavy atom derivatives were
collected in house using either a Jupiter CCD detector or an RAXIS
HTC image plate detector. Both were mounted on Rigaku rotating
anode generators. The high-resolution data used to refine the
solAC structure at 1.7 ꢁ was collected on Beamline ID29–1 at the
European Synchrotron Radiation Facility (Grenoble, France), using
an ADSC Quantum4 CCD detector, with a wavelength of 0.934 ꢁ
and processed using MOSFLM version 6.2.3.[29] The dataset was
scaled using SCALA,[30] and the intensities converted to structure
factor amplitudes with TRUNCATE.[31] The crystals grew in space
group P63 with cell dimensions of a=b=99.15 ꢁ, c=97.51 ꢁ.
Thermal shift assay: Thermal shift assays were performed on an
Mx3005P quantitative PCR instrument (Stratagene, La Jolla, Califor-
nia) capable of temperature control and fluorescence detection.
Protein unfolding was monitored using the SYPRO orange dye (In-
vitrogen), which binds to hydrophobic regions on the protein
during the unfolding process. Compound stock solution (1 or 5 mL)
was aliquoted into 96-well PCR plates (Starlab) to produce final
nominal concentrations of 1 mm and 5 mm compound in 1 or 5%
DMSO, respectively. The hsolAC protein concentration was 0.25 mm
Structure determination and refinement: The structure of hsolAC
was solved by multiple isomorphous replacement using five
heavy-atom-containing derivatives (table S1 in the Supporting In-
formation). Difference Patterson methods were used to obtain ini-
tial positions for the heavy atoms. Preliminary phasing with
ꢀ 2014 The Authors. Published by Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 2014, 9, 823 – 832 829

CHEMMEDCHEM
FULL PAPERS
www.chemmedchem.org
in a buffer containing 50 mm HEPES (pH 7.5), 100 mm NaCl, 1 mm
MnCl2, 1 mm dithiothreitol, and 2.5X SYPRO orange. Control experi-
ments were performed on each 96-well plate, with six negative
and two positive control wells per plate. The plate was heated
over a temperature range of 25–708C, and the temperature was in-
creased by 18C per min. The fluorescence intensity was measured
every 0.58C. All data were processed and analysed using Origin 7.0
software.
while stirring. A small quantity of NaNO2 (a spatula full) was added,
and the mixture stirred at 608C for a further 2 h. The mixture was
allowed to cool to RT and then partitioned between 2m aq NaOH
and EtOAc. The organic layer was separated, dried over anhydrous
MgSO4, filtered and concentrated in vacuo to afford a viscous
orange oil. Purification by column chromatography on silica gel
(EtOAc/petroleum ether, 0–40%) afforded [4-(3-methoxybenzoyl)-2-
oxy-furazan-3-yl]-(3-methoxyphenyl)methanone as a viscous yellow
oil (12.2 g, 89.9%): 1H NMR (400 MHz, [D6]DMSO): d=7.74 (dm,
1H), 7.59 (m, 1H), 7.54 (m, 2H), 7.45 (m, 2H), 7.38 (dm, 1H), 7.30
(dm, 1H), 3.83 (s, 3H), 3.77 ppm (s, 3H); MS (ESI): m/z=208
[M+H]+.
Chemistry
Compound 2 was purchased from Lancaster Synthesis. Compounds
3 and 5 were purchased from Maybridge. Compound 4 was pur-
chased from Chemical Blocks. Compounds 1 and 6–8 were pre-
pared as shown in Schemes 1–4 and described below.
2m NH3 in MeOH (60 mL) was added to a solution of [4-(3-methox-
ybenzoyl)-2-oxyfurazan-3-yl]-(3-methoxyphenyl)methanone
(6 g,
13.3 mmol) in Et2O (30 mL) and MeOH (60 mL), and the mixture
was stirred at RT for 30 min then heated to 608C for 2 h. Upon
cooling to RT, the mixture was evaporated to dryness in vacuo,
and the residue was purified by column chromatography on silica
gel (EtOAc/petroleum ether, 0–30%) to give 6 as an off-white solid
(1.1 g, 37.8%): 1H NMR (400 MHz, [D6]DMSO): d=7.77 (dm, 1H),
7.65 (m, 1H), 7.55 (t, 1H), 7.35 (dm, 1H), 6.60 (br s, 2H), 3.85 ppm
(s, 3H).
Scheme 1. Reagents and conditions: a) DMF·DMA, toluene, reflux, 18 h;
b) N2H4·H2O, EtOH, N2, reflux, 72 h.
4-(4-Fluorophenyl)-3-methyl-1H-pyrazole (1): A mixture of 4-fluo-
rophenylacetone (2.67 mL, 20 mmol) and N,N-dimethylformamide–
dimethylacetal (4.0 mL, 30 mmol) in toluene (50 mL) was stirred at
reflux for 18 h and then allowed to cool to RT. The solvent was re-
moved in vacuo to afford (Z)-4-dimethylamino-3-(4-fluorophenyl)-
but-3-en-2-one as an orange solid, which was used without further
purification (4.15 g, >100%—likely due to solvent contamination):
1H NMR (400 MHz, [D6]DMSO): d=7.51 (s, 1H), 7.11 (m, 4H), 2.67
(br s, 6H), 1.97 ppm (s, 3H); MS (ESI): m/z=208 [M+H]+.
A mixture of (Z)-4-dimethylamino-3-(4-fluorophenyl)but-3-en-2-one
(4.1 g, 19.8 mmol) and NH2NH2·H2O (1.06 mL, 21.8 mmol) in EtOH
(50 mL) was stirred at reflux under a nitrogen atmosphere for 72 h
and then allowed to cool to RT. The solvent was removed in vacuo,
and the residue was purified by recrystallization from petroleum
ether and EtOAc. The resulting solid was collected by suction filtra-
tion, rinsed with petroleum ether, and dried in vacuo to afford 1 as
an off-white solid (2.78 g, 79.7%): 1H NMR (400 MHz, [D6]DMSO):
d=12.65 (br s, 1H), 7.71 (br s, 1H), 7.46 (m, 2H), 7.21 (t, 2H),
2.35 ppm (s, 3H); MS (ESI): m/z=177 [M+H]+.
Scheme 3. Reagents and conditions: a) HNO3, AcOH, NaNO2, 608C, 2 h;
b) NH3, Et2O, MeOH, 608C, 2 h; c) BCl3, CH2Cl2, RT, 30 min; d) NaH, DMF, 908C,
2 h.
[3-(4-Aminofurazan-3-carbonyl)phenoxy]acetic acid ethyl ester
(7): A mixture of 3-benzyloxyacetophenone (15 g, 66.4 mmol),
HNO3 (11.5 mL), AcOH (52 mL) and water (35 mL) was heated to
608C while stirring. A small quantity of NaNO2 (a spatula full) was
added, and the mixture stirred at 608C for a further 2 h. The mix-
ture was allowed to cool to RT and then partitioned between 2m
aq NaOH and EtOAc. The organic layer was separated, dried over
anhydrous MgSO4, filtered and concentrated in vacuo to afford
a viscous orange oil. Purification by column chromatography on
silica gel (EtOAc/petroleum ether, 0–35%) afforded [4-(3-benzyloxy-
benzoyl)-2-oxy-furazan-3-yl]-(3-benzyloxyphenyl)methanone
as
1
a yellow oil (9.3 g, 55.4%): H NMR (400 MHz, [D6]DMSO): d=7.76
(m, 2H), 7.58 (m, 3H), 7.50–7.30 (m, 13H), 5.16 (s, 2H), 5.08 ppm (s,
2H).
Scheme 2. Reagents and conditions: a) HNO3, AcOH, NaNO2, 608C, 2 h;
b) NH3, Et2O, MeOH, 608C, 2 h.
2m NH3 in MeOH (18 mL) was added to a solution of [4-(3-ben-
zyloxybenzoyl)-2-oxyfurazan-3-yl]-(3-benzyloxyphenyl)methanone
(9.3 g, 18.4 mmol) in Et2O (30 mL) and MeOH (60 mL), and the mix-
ture was stirred at RT for 30 min then heated to 608C for 2 h.
Upon cooling to RT, the mixture was evaporated to dryness in
(4-Amino-furazan-3-yl)-(3-methoxy-phenyl)-methanone (6):
mixture of 3-methoxyacetophenone (10 mL, 75.4 mmol), HNO3
(13 mL), AcOH (60 mL) and water (40 mL) was heated to 608C
A
ꢀ 2014 The Authors. Published by Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 2014, 9, 823 – 832 830

CHEMMEDCHEM
FULL PAPERS
www.chemmedchem.org
À
vacuo, and the residue was purified by column chromatography
on silica gel (EtOAc/petroleum ether, 0–50%) to give (4-aminofura-
zan-3-yl)-(3-benzyloxyphenyl)methanone as an off-white solid
(920 mg, 16.9%): 1H NMR (400 MHz, [D6]DMSO): d=7.78 (dt, 1H),
7.73 (t, 1H), 7.55 (t, 1H), 7.50 (dd, 2H), 7.43 (m, 3H), 7.35 (tt, 1H),
6.60 (br s, 2H), 5.20 ppm (s, 2H); MS (ESI): m/z=296 [M+H]+.
PDB for comparative HCO3 binding interactions. V.B. modelled
and designed the compounds and produced the pictures. A.C. per-
formed crystallization, X-ray data collection, structure solution by
multiple isomorphous replacement, and solution of the AMPCPP
and bicarbonate complexes. M.V. performed ligand soaking experi-
ments, X-ray data collection, solved and refined the complex struc-
tures, measured the pKa values of compounds, searched com-
pound databases for hit enrichment, and designed compounds.
M.C. and M.A.O’B designed and synthesized compounds. J.E.C per-
formed the Tm shift screen. H.J. and J.R.Y. helped define the alloste-
ric model and manage the project. S.M.S-B and J.E.C. were project
leaders during structure solution and pyramid screening phases,
respectively. S.M.S-B., C.W.M. and H.J. wrote the manuscript.
1m BCl3 in CH2Cl2 (5.22 mL, 5.22 mmol) was added dropwise to
a stirred solution of (4-aminofurazan-3-yl)-(3-benzyloxyphenyl)me-
thanone (770 mg, 2.61 mmol) in CH2Cl2 (20 mL), and the mixture
was stirred at RT for 30 min. MeOH (1 mL) was added, the mixture
was concentrated to dryness in vacuo, and the residue purified by
column chromatography on silica gel (EtOAc/petroleum ether, 0–
40%) to give (4-aminofurazan-3-yl)-(3-hydroxyphenyl)methanone
as a pale tan solid (500 mg, 93%): 1H NMR (400 MHz, [D6]DMSO):
d=9.97 (s, 1H), 7.62 (dm, 1H), 7.56 (m, 1H), 7.42 (t, 1H), 7.15 (dm,
1H), 6.57 ppm (br s, 2H).
Acknowledgements
NaH (60 wt% in mineral oil, 180 mg, 4.58 mmol) was added to
The authors acknowledge the contributions of Dr. Pamela Wil-
liams in producing the alignments, Kirsten Smith in assisting with
the baculo virus expressions, Dr. Deborah J. Davis in setting up
the bioassay, and Dr. Martyn Frederickson in assisting with the
chemistry write-up.
a
mixture of (4-aminofurazan-3-yl)-(3-hydroxyphenyl)methanone
(250 mg, 1.22 mmol) and ethyl bromoacetate (0.42 mL, 3.66 mmol)
in DMF (10 mL), and the mixture was stirred at 758C for 2 h. The
reaction mixture was allowed to cool to RT and then partitioned
between EtOAc and water. The organic layer was washed with
water and then brine, separated, dried over MgSO4, filtered, and
concentrated in vacuo. The residue was purified by column chro-
matography on silica gel (EtOAc/petroleum ether, 0–30%) to afford
semi-pure product, which was further purified by preparative HPLC
Keywords: allosterism · drug discovery · enzyme regulation ·
fragment screening · structural biology
1
to afford 7 as a colourless solid (23 mg, 6.5%): H NMR (400 MHz,
[D6]DMSO): d=7.80 (dt, 1H), 7.65 (t, 1H), 7.55 (t, 1H), 7.35 (dt, 1H),
6.60 (br s, 2H), 4.90 (s, 2H), 4.20 (q, 2H), 1.23 ppm (t, 3H); MS (ESI):
m/z=292 [M+H]+.
[1] J. R. Casey, Biochem. Cell Biol. 2006, 84, 930–939.
[2] M. S. Wutke, J. Buck, L. R. Levin, JOP 2001, 2, 154–158.
[3] J. Buck, M. L. Sinclair, L. Schapal, M. J. Cann, L. R. Levin, Proc. Natl. Acad.
Sci. USA 1999, 96, 79–84.
[4] Y. Chen, M. J. Cann, T. N. Litvin, V. Iourgenko, M. L. Sinclair, L. R. Levin, J.
Buck, Science 2000, 289, 625–628.
[5] M. Kamenetsky, S. Middelhaufe, E. M. Bank, L. R. Levin, J. Buck, C. Steeg-
born, J. Mol. Biol. 2006, 362, 623–639.
[6] G. S. McKnight, Curr. Opin. Cell Biol. 1991, 3, 213–217.
[7] P. Sassone-Corsi, Cold Spring Harbor Perspect. Biol. 2012, 4, a011148.
[8] T. N. Litvin, M. Kamenetsky, A. Zarifyan, J. Buck, L. R. Levin, J. Biol. Chem.
2003, 278, 15922–15926.
[9] R. A. Bundey, P. A. Insel, Sci. STKE 2004, pe19.
[10] M. Tresguerres, L. R. Levin, J. Buck, Kidney Int. 2011, 79, 1277–1288.
[11] B. S. Jaiswal, M. Conti, J. Biol. Chem. 2001, 276, 31698–31708.
[12] J. Roelofs, P. J. M. Van Haastert, Mol. Biol. Evol. 2002, 19, 2239–2246.
[13] B. S. Jaiswal, M. Conti, Proc. Natl. Acad. Sci. USA 2003, 100, 10676–
10681.
[14] E. J. Neer, J. Biol. Chem. 1978, 253, 5808–5812.
[15] J. O. Gordeladze, V. Hansson, Mol. Cell. Endocrinol. 1981, 23, 125–136.
[16] G. Esposito, B. S. Jaiswal, F. Xie, M. A. Krajnc-Franken, T. J. Robben, A. M.
Strik, C. Kuil, R. L. Philipsen, M. van Duin, M. Conti, J. A. Gossen, Proc.
Natl. Acad. Sci. USA 2004, 101, 2993–2998.
[17] C. Steegborn, T. N. Litvin, L. R. Levin, J. Buck, Nat. Struct. Mol. Biol. 2005,
12, 32–37.
[18] S.-Z. Yan, Z.-H. Huang, R. S. Shaw, W.-J. Tang, J. Biol. Chem. 1997, 272,
12342–12349.
[19] J. J. G. Tesmer, R. K. Sunahara, R. A. Johnson, G. Gosselin, A. G. Gilman,
S. R. Sprang, Science 1999, 285, 756–760.
[20] M. J. Cann, A. Hammer, J. Zhou, T. Kanacher, J. Biol. Chem. 2003, 278,
35033–35038.
Scheme 4. Reagents and conditions: a) NaH, DMF, 758C, o/n.
(4-Aminofurazan-3-yl)-[3-(1H-benzoimidazol-2-ylmethoxy)phe-
nyl]methanone (8): NaH (60 wt% in mineral oil, 29 mg, 0.73 mmol)
was added to a mixture of (4-aminofurazan-3-yl)-(3-hydroxyphe-
nyl)methanone [prepared as outlined above for the synthesis of 7]
(75 mg, 0.37 mmol) and 2-chloromethyl-1H-benzimidazole (122 mg,
0.73 mmol) in DMF (10 mL), and the mixture was stirred at 758C
overnight. The reaction mixture was allowed to cool to RT and was
then partitioned between EtOAc and water. The organic layer was
washed twice with water, separated, dried over anhydrous MgSO4,
filtered and concentrated in vacuo. The residue was purified by
preparative HPLC to afford 8 as a colourless solid (5 mg, 4.1%):
1H NMR (400 MHz, [D6]DMSO): d=12.70 (br s, 1H), 7.80 (m, 2H),
7.63 (dd, 1H), 7.58 (t, 1H), 7.50 (td, 2H), 7.20 (m, 2H), 6.60 (br s,
2H), 5.43 ppm (br s, 2H); MS (ESI): m/z=336 [M+H]+.
[21] M. Kobayashi, J. Buck, L. R. Levin, Dev. Genes Evol. 2004, 214, 503–509.
[22] J. J. G. Tesmer, R. K. Sunahara, A. G. Gilman, S. R. Sprang, Science 1997,
278, 1907–1916.
[23] G. Zhang, Y. Liu, A. E. Ruoho, J. H. Hurley, Nature 1997, 386, 247–253.
[24] W. Jahnke, J. M. Rondeau, S. Cotesta, A. Marzinzik, X. Pellꢃ, M. Geiser, A.
Strauss, M. Gçtte, F. Bitsch, R. Hemmig, C. Henry, S. Lehmann, J. F. Glick-
man, T. P. Roddy, S. J. Stout, J. R. Green, Nat. Chem. Biol. 2010, 6, 660–
666.
Author Contributions
S.M.S-B. designed the protein constructs, and expressed, purified,
characterized and crystallized the protein. C. W. M. analysed the
ꢀ 2014 The Authors. Published by Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 2014, 9, 823 – 832 831

CHEMMEDCHEM
FULL PAPERS
www.chemmedchem.org
[25] D. I. Rhodes, T. S. Peat, N. Vandegraaff, D. Jeevarajah, G. Le, E. D. Jones,
J. A. Smith, J. A. Coates, L. J. Winfield, N. Thienthong, J. Newman, D.
Lucent, J. H. Ryan, G. P. Savage, C. L. Francis, J. J. Deadman, Antiviral
Chem. Chemother. 2011, 21, 155–168.
[26] I. Krimm, J.-M. Lancelin, J.-P. Praly, J. Med. Chem. 2012, 55, 1287–1295.
[27] S. M. Saalau-Bethell, A. J. Woodhead, G. Chessari, M. G. Carr, J. Coyle, B.
Graham, S. D. Hiscock, C. W. Murray, P. Pathuri, S. J. Rich, C. J. Richard-
son, P. A. Williams, H. Jhoti, Nat. Chem. Biol. 2012, 8, 920–925.
[30] Collaborative Computational Project, Number 4, Acta Crystallogr D Biol
Crystallogr. 1994, 50, 760–763.
[31] P. R. Evans, Proceedings of the CCP4 Study Weekend, 1997.
[32] C. W. Dessauer, A. G. Gilman, J. Biol. Chem. 1997, 272, 27787–27795.
[33] A. Sali, T. L. Blundell, J. Mol. Biol. 1990, 212, 403–428.
[34] M. J. Sutcliffe, I. Haneef, D. Carney, T. L. Blundell, Protein Eng. 1987, 1,
377–384.
[28] C. W. Murray, D. C. Rees, Nat. Chem. Biol. 2009, 1, 187–192.
[29] A. G. W. Leslie, Joint CCP4+ESF-EAMCB Newsletter on Protein Crystal-
lography 1992, 26; http://www.ccp4.ac.uk/newsletters/No26.pdf.
Received: November 25, 2013
Published online on February 24, 2014
ꢀ 2014 The Authors. Published by Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemMedChem 2014, 9, 823 – 832 832