Site-directed Mutagenesis of Key Residues Unveiled a Novel Allosteric Site on Human Adenosine Kinase for Pyrrolobenzoxa(thia)zepinone Non-Nucleoside Inhibitors

Article abstract of DOI:10.1111/cbdd.12630

Most nucleoside kinases, besides the catalytic domain, feature an allosteric domain which modulates their activity. Generally, non-substrate analogs, interacting with allosteric sites, represent a major opportunity for developing more selective and safer therapeutics. We recently developed a series of non-nucleoside non-competitive inhibitors of human adenosine kinase (hAK), based on a pyrrolobenzoxa(thia)zepinone scaffold. Based on computational analysis, we hypothesized the existence of a novel allosteric site on hAK, topographically distinct from the catalytic site. In this study, we have adopted a multidisciplinary approach including molecular modeling, biochemical studies, and site-directed mutagenesis to validate our hypothesis. Based on a three-dimensional model of interaction between hAK and our molecules, we designed, cloned, and expressed specific, single and double point mutants of hAK (Q74A, Q78A, H107A, K341A, F338A, and Q74A-F338A). Kinetic characterization of recombinant enzymes indicated that these mutations did not affect enzyme functioning; conversely, mutated enzymes are endowed of reduced susceptibility to our non-nucleoside inhibitors, while maintaining comparable affinity for nucleoside inhibitors to the wild-type enzyme. This study represents the first characterization and validation of a novel allosteric site in hAK and may pave the way to the development of novel selective and potent non-nucleoside inhibitors of hAK endowed with therapeutic potential.

Full text of DOI:10.1111/cbdd.12630

Chem Biol Drug Des 2016; 87: 112–120
Research Article
Site-directed Mutagenesis of Key Residues Unveiled a
Novel Allosteric Site on Human Adenosine Kinase for
Pyrrolobenzoxa(thia)zepinone Non-Nucleoside
Inhibitors
Lida Savi1, Margherita Brindisi2,3, Gloria
to the development of novel selective and potent
non-nucleoside inhibitors of hAK endowed with thera-
peutic potential.
Alfano2,3, Stefania Butini2,3,*, Valeria La Pietra3,4
,
Ettore Novellino3,4, Luciana Marinelli3,4,*, Andrea
Lossani1, Federico Focher1,*, Caterina Cavella2,3
Giuseppe Campiani2,3 and Sandra Gemma2,3
,
Key words: adenosine kinase, Ala-scan mutagenesis,
allosteric binding site, allosteric inhibitors, drug discovery,
non-nucleoside human AK inhibitors, small molecule
1Istituto di Genetica Molecolare, CNR, Via Abbiategrasso
207, 27100 Pavia, Italy
Abbreviations: 5-ITU, 5-iodo-7-b-D-ribofuranosyl-7H-pyrrolo
[2,3-d]pyrimidin-4-amine; Ado, adenosine; CNS, central
nervous system; hAK, human AK; Hcy, homocysteine; IPTG,
isopropylthio-b-D-galactoside; KHK, ketohexokinase; NNhA-
KIs, non-nucleoside hAK inhibitors; PNS, peripheral nervous
system; RK, ribokinase; SAH, S-adenosylhomocysteine; wt,
wild type.
2Dipartimento di Biotecnologie, Chimica e Farmacia,
University of Siena, via Aldo Moro 2, 53100 Siena, Italy
3European Research Centre for Drug Discovery and
Development (NatSynDrugs), University of Siena, via Aldo
Moro 2, 53100 Siena, Italy
4
ꢀ
Dipartimento di Farmacia, Universita di Napoli Federico II,
via D. Montesano 49, 80131 Napoli, Italy
*Corresponding authors: Stefania Butini, butini3@unisi.it;
Luciana Marinelli, luciana.marinelli@unina.it; Federico
Focher, focher@igm.cnr.it
Received 24 April 2015, revised 24 July 2015 and accepted
for publication 27 July 2015
L.S. and M.B. equally contributed to this work.
Adenosine kinase (AK; EC.2.7.1.20) plays an important
role in regulating the intracellular as well as extracellular
concentrations of adenosine (Ado) (1,2), a nucleoside
which exerts a broad range of physiological responses via
interaction with Ado receptors. Due to its diverse physio-
logical actions, abnormal Ado concentrations are involved
in several diseases, especially in those affecting central
and peripheral nervous systems (CNS, PNS), where the
fine-tuning of Ado concentration appears to hold a particu-
lar physiological relevance (3). Human AK (hAK) activity
was also found to play a role in the regulation of cardiac
hypertrophy (4) and diabetes (5). In view of the enormous
pharmacological potential of Ado, there has been much
interest in studying role and modulation of AK over the
years (6). Recently, it has been demonstrated that hAK
activity could also play an important role in the epigenetic
control of nucleic acids (7,8). Cytidine methylation is
strategic for gene expression and regulation and requires
the donation of a methyl group from S-adenosylmethion-
ine, a process that is catalyzed by DNA methyltrans-
ferases. The resulting product, S-adenosylhomocysteine
(SAH), is then converted into Ado and homocysteine (Hcy)
by SAH hydrolase (7). Critically, the equilibrium constant
of the SAH hydrolase enzyme lies in the direction of SAH
formation (9); therefore, the reaction will only proceed
when Ado and Hcy are constantly removed (9,10). In the
Most nucleoside kinases, besides the catalytic
domain, feature an allosteric domain which modulates
their activity. Generally, non-substrate analogs, inter-
acting with allosteric sites, represent a major opportu-
nity for developing more selective and safer
therapeutics. We recently developed a series of non-
nucleoside non-competitive inhibitors of human ade-
nosine kinase (hAK), based on a pyrrolobenzoxa(thia)
zepinone scaffold. Based on computational analysis,
we hypothesized the existence of a novel allosteric
site on hAK, topographically distinct from the catalytic
site. In this study, we have adopted a multidisciplinary
approach including molecular modeling, biochemical
studies, and site-directed mutagenesis to validate our
hypothesis. Based on a three-dimensional model of
interaction between hAK and our molecules, we
designed, cloned, and expressed specific, single and
double point mutants of hAK (Q74A, Q78A, H107A,
K341A, F338A, and Q74A-F338A). Kinetic characteriza-
tion of recombinant enzymes indicated that these
mutations did not affect enzyme functioning; con-
versely, mutated enzymes are endowed of reduced
susceptibility to our non-nucleoside inhibitors, while
maintaining comparable affinity for nucleoside
inhibitors to the wild-type enzyme. This study
represents the first characterization and validation of
a novel allosteric site in hAK and may pave the way
112
ª 2015 John Wiley & Sons A/S. doi: 10.1111/cbdd.12630

Discovery of a Novel Allosteric Site on hAK
adult brain, removal of Ado occurs largely via the astro-
cytic hAK (2,3). If this hAK-mediated metabolic clearance
of Ado is impaired, SAH levels rise (10) and inhibit methyl-
transferases by product inhibition (11). Thus, intracellular
concentration of Ado could be crucial to control DNA
methylation and gene expression (7). Accordingly, it has
been observed that Ado induces hypomethylation of DNA
via biochemical interference with the transmethylation
pathway (7). The Boison’s group effectively treated epilep-
togenesis in multiple seizure models by delivering Ado into
the brain. They observed that global DNA methylation was
significantly reduced (by 51%) in the hippocampus of trea-
ted versus control animals. It was proved in rats that tem-
poral lobe epilepsy correlates with an increase in
hippocampal DNA methylation, DNA methyltransferase
activity, disruption of Ado homeostasis, and spontaneous
recurrent seizures. Thus, pathological changes in DNA
methylation patterns may trigger epilepsy and reversal of
these epigenetic changes with Ado therapy may halt
disease progression. Moreover, modulation of the
‘adenosinergic system’ may be relevant in therapy of other
CNS disorders such as schizophrenia, Alzheimer’s dis-
ease, Huntington’s disease, amyotrophic lateral sclerosis,
and multiple sclerosis (6). However, the systemic adminis-
tration of Ado (unless directly injected in brain) has proved
inefficient because of its short half-life in physiological flu-
ids, poor bioavailability, rapid uptake by erythrocytes, lack
of specificity, and long-term cytotoxic effects (3). At the
same time, Ado receptor agonists may lead to unfavorable
side-effect caused by the activation of Ado receptors in
non-target tissues. Accordingly, the inhibition of Ado-me-
tabolizing enzymes, such as hAK, represents a valuable
alternative to elevate the Ado levels in specific brain areas.
bitors (NNhAKIs) able to inhibit the enzyme activity in the
micromolar range (16). These compounds are character-
ized by a unique molecular scaffold, and interestingly they
showed a different mechanism of action compared to
known NNhAKIs. The kinetic behavior of our pyrroloben-
zoxa(thia)zepines suggested the existence of an allosteric
site, topographically distinct from the alkynylpyrimidine
binding site. Based on these data, a computational analy-
sis, using hAK in different conformations (open and
closed), was performed for identifying the NNhAKIs’ allos-
teric binding site. This original analysis revealed the pres-
ence of a shallow pocket located between the helices a3,
a4, and the carboxylic tail which is upon the Ado and ATP
binding sites (16). Intriguingly, this pocket was clearly
detectable only in the closed conformation of hAK, while it
was only partially evident in the open conformation due to
the shift of the b4-sheet (16).
Starting from the hypothesized binding site and our ligand
binding mode (Figure 1), we report herein an alanine scan-
ning mutagenesis and a biochemical study to support our
original hypothesis and to confirm the presence of an
allosteric binding site in hAK for NNhAKIs. This study deals
with the cloning, expression, and biochemical characteri-
zation of several single and double hAK mutants, the
stereoselective synthesis of the most potent NNhAKIs, the
binding experiments against wild-type (wt) and all mutants.
This approach led to the validation of a novel allosteric
binding site on hAK, the exploitation of which may lead to
new avenues for the discovery of selective and safe thera-
peutics.
hAK is an evolutionarily ancient and highly conserved
enzyme, which is directly related to bacterial ribokinases
and fructokinases (12,13). hAK catalyzes the phosphoryla-
tion of Ado to AMP, using ATP as a phosphate donor. The
hAK structure in complex with two Ado molecules has
been solved by Mathews and colleagues (14). The human
enzyme works as a monomer (13) and is composed by
twelve a-helices and fourteen b-sheets and is featured by
(i) a catalytic Ado binding site, (ii) a regulatory Ado binding
site, (iii) an ATP binding site, and (iv) three Mg2+ binding
sites, one of which essential for the catalytic reaction (15).
A
Due to their relevant role in cell metabolism, Ado and ATP
are molecules which interact with several enzymes. Thus,
synthetic inhibitors that interact with the active site of
enzymes metabolizing Ado or ATP, although potent, are
generally poorly selective and therefore toxic. Their low
specificity could be expected even if Ado- or ATP-metabo-
lizing enzymes do not share amino acid sequences homol-
ogy in their catalytic pockets; structural homology could
be enough.
B
Figure 1: Binding pose of NSD438 in the allosteric site of hAK
showed as cornflower cartoon (panel A). Interacting residues are
shown as sticks and colored by atom type, while the H-bond is
represented by a black dashed line (panel B). Residues are
labeled according to PDB numbering.
Aware of this issue, we recently developed a novel class
of non-toxic non-nucleoside and non-competitive hAK inhi-
Chem Biol Drug Des 2016; 87: 112–120
113

Savi et al.
digestion product was analyzed by agarose gel elec-
trophoresis. The full-length plasmid was quantified using
DNA Molecular Marker III as reference (0.21–21.2 kbp;
Roche Applied Science). Then, 1 ng of mutated DNA was
use to transform E. coli (DH5a strain) TSS chemically
competent cells (18). All mutants were sequenced by
Bmr Genomics (Padova, Italy) to confirm the presence of
the desired mutation and the absence of undesired
mutations.
Methods and Materials
Chemistry
Starting materials and solvents were purchased from com-
mercial suppliers and used without further purification.
Reaction progress was monitored by TLC using silica gel
60 F254 (0.040–0.063 lm) with detection by UV. Silica gel
60 (0.040–0.063 lm) was used for column chromatogra-
phy. Yields refer to purified materials and are not opti-
1
mized. H NMR and 13C NMR spectra were recorded on a
Varian 300 MHz or a Bruker 400 MHz spectrometer using
the residual signal of the deuterated solvent as internal
standard. Splitting patterns are described as singlet (s),
doublet (d), triplet (t), quartet (q), and broad (br); chemical
shifts (d) are given in ppm and coupling constants (J) in
hertz (Hz). ESI-MS spectra were performed by an Agilent
1100 Series LC/MSD spectrometer. Optical rotation values
were measured at room temperature using a Perkin–Elmer
model 343 polarimeter operating at 589 nm, correspond-
ing to the sodium D line or a Jasco P-2000 polarimeter
operating at 436 nm. All moisture-sensitive reactions were
performed under argon atmosphere using oven-dried
glassware and anhydrous solvents. Elemental analyses
were performed in a Perkin–Elmer 240C elemental ana-
lyzer. For (R)- and (S)-NSD438, they were in full agreement
with the values found for racemic NSD438 (16) and the
results were within ꢀ 0.4% of the theoretical values. The
synthesis of (R)- and (S)-NSD438 is reported in Scheme 1.
Compounds 14–19 were synthesized as previously
reported by us (16). Synthetic procedures and characteri-
zation data for the new compounds are given in the Sup-
porting Information (Appendix S1).
Expression and purification of recombinant wt-
hAK and hAK mutants
Wild-type and mutant proteins were purified from E. coli
crude extract in a single chromatographic step using Ni-
NTA Superflow column (Qiagen). Briefly, a fresh overnight
saturated culture of bacteria was diluted 1:100 in 1 L of
2 9 YT medium containing ampicillin (75 lg/mL) and incu-
bated at 37 °C with shaking. At OD600, IPTG was added
to a final concentration of 0.4 mM, and culture was incu-
bated for 2 h at 25 °C.
The bacterial cell pellet was resuspended in four volumes of
lysis buffer (50 mM sodium phosphate, pH 8.0, 300 mM
NaCl, 10 mM imidazole). Lysozyme 1 mg/mL and PMSF
1 mM were also added, and the mixture was incubated on
ice for 30 min. Cells were sonicated (6 9 10 seconds, 85
dCy) and homogenized. Then, the lysate was centrifuged at
20 000 g for 30 min at 4 °C. The supernatant was loaded
into Ni-NTA Superflow column (1 mL) previously equili-
brated in lysis buffer. The column was then washed with
10 mL of lysis buffer, followed by 10 mL of wash buffer con-
taining 50 mM sodium phosphate, pH 8.0, 300 mM NaCl,
20 mM imidazole. The protein was then eluted by a gradient
from 20 to 250 mM imidazole in 50 mM sodium phosphate
buffer, pH 8.0, and 300 mM NaCl. The hAK-rich fractions
were pooled and dialyzed against 50 mM Tris–HCl pH 7.5
containing 1 mM 1,4-dithiothreitol (DTT) and 20% glycerol
and then frozen in liquid nitrogen until used.
Biology
For biological tests analytical grade reagents were exclu-
sively used. Bacterial media components were from Difco.
Ni-NTA Superflow resin was from Qiagen. Restriction and
modification enzymes were from Promega, Sigma, or
Roche Molecular Biochemicals. Vectors were from Invitro-
gen. Isopropylthio-b-D-galactoside (IPTG) was from
Sigma. [2,8-3H]-Ado, 29.1 Ci/mmol, was from Hartmann
Analytic.
Biochemical characterization of recombinant wt-
hAK and hAK mutants
Wild-type and mutated hAK were assayed with a radio-
chemical method which measures the formation of [3H]-
AMP from [3H]-Ado. The enzyme was incubated at 37 °C
in 15 lL of a mixture containing 64 mM Tris–HCl (pH 7.5),
40 mM KCl, 0.1 mM MgCl2, 0.1 mM ATP, and 1 lM
[2,8-3H]-Ado (29.1 Ci/mmol, 750 cpm/pmol; Hartmann
Analytic). The reaction was terminated by spotting 10 lL
of the incubated mixture on DEAE 96-square glass fiber fil-
ter (DEAE Filtermat; Wallac). The filter was rinsed three
times in 1 mM ammonium formate, pH 3.6, to remove
unconverted nucleoside, and finally in ethanol. The filter
was dried and melt on scintillator sheets (MeltiLex A; Per-
kin Elmer), and labeled-AMP trapped on filter was quanti-
fied by Microbeta Trilux (Perkin Elmer) luminometer,
according to the manufacturer’s protocol.
Site-directed mutagenesis of recombinant hAK
Human AK mutants were obtained through PCR mutagene-
sis (17) using the wt-hAK gene (1070 bp), inserted into
pTrcHisA vector (16), and the primers reported in Table S1.
For single-site and double-site mutations, the PCR (50 lL)
contained 50 ng of DNA template (pTrcHisA-wt-hAK),
0.1 lM primer pair, 250 lM dNTPs, and 3U of PfuUltra
High-fidelity DNA polymerase (Agilent Technologies©). The
PCR cycles were initiated at 95 °C for 5 min, followed by 12
amplifications (95 °C for 1 min, 62 °C for 1 min, and 68 °C
for 12 min). The last step was performed at 68 °C for 1 h.
To eliminate parental wt DNA, the PCR products were trea-
ted with 10 U of DpnI at 37 °C for 3 h and then 10 lL of
114
Chem Biol Drug Des 2016; 87: 112–120

Discovery of a Novel Allosteric Site on hAK
Scheme 1: Stereoselective synthesis of compounds (S)- and (R)-NSD438. Reagents and conditions: (a) AlCl3, ClCOCOOEt, 1,2-DCE, from 0
to 30 °C, 3 h; (b) (1S,2R,5S)-(+)-menthol, p-TsOH, dry toluene, 120 °C, 12 h; (c) (1S,2R,5S)-(+)-menthol, Ti[OCH(CH3)2]4, dry toluene, 100 °C,
12 h; (d) EtMgBr, ZnCl2, 5 h, from 0 °C (2 h) to ꢁ78 °C (3 h), then to rt (1 h); (e) KOH, MeOH/H2O 1:1, reflux, 24 h; (f) MeI, K2CO3, dry DMF,
rt, 5 h; (g) 1-(2-fluoro-5-nitrophenyl)-1H-pyrrole, NaH, dry THF, rt, 12 h; (h) SnCl2, EtOH, 80 °C, 5 h; (i) NaNO2, NaHSO3, AcOH, EtOH, rt, 3 h;
(j) 15% NaOH, EtOH/THF, rt, 2 h; (k) PCl5, dry CH2Cl2, 40 °C, 12 h; (l) NBS, AIBN, CCl4, 80 °C, 7 h; (m) NaH, thiophenol, dry THF, rt, 10 h.
Inhibition assay of recombinant wt-hAK and hAK
mutants
positions. The binding pocket was identified by placing a
ꢀ
15 A cube centered on the F338 mass center. Molecular
In inhibition studies, increasing concentrations of each com-
pound were tested in duplicate in two independent assays,
as described above. Stock solutions of compound, in 100%
DMSO, were diluted in water to a final DMSO concentration
of 1%. Controls without inhibitor were also incubated in the
presence of 1% DMSO. Reference inhibitor used 5-iodo-7-
docking calculations were performed with the aid of GLIDE
5.5 in standard precision (SP) mode.
Multiple sequence alignment
Multiple sequence alignment of AK with related proteins,
ribokinase (RK) and ketohexokinase (KHK), was performed
using CLUSTALW2 (http://www.ebi.ac.uk/Tools/msa/clustal-
w2). Although we were interested only on human proteins,
also sequences from rat and mouse were included to
achieve a more reliable alignment.
b-D-ribofuranosyl-7H-pyrrolo[2,3-d]pyrimidin-4-amine
(5-
ITU) purchased from Tocris Bioscience (19).
Molecular modeling
Molecular docking
The docking was performed with the aid of GLIDE software
where the ‘write per-residue interaction score’ feature is
implementeda. The NSD438 inhibitor structures were first
generated through the Dundee PRODRG2 server. Then,
geometry optimized ligands were prepared using LIG-PREP
2.3 as implemented in MAESTRO. The 1BX4 closed confor-
mation of the hAK protein was prepared through the Pro-
tein Preparation Wizard of the graphical user interface
Maestro and the OPLS-2001 force field. Water molecules
were removed. Hydrogen atoms were added, and mini-
mization was performed until the rmsd of all heavy atoms
Results and Discussion
hAK mutants and selection of NNhAKIs
Our early computational analysis of hAK allowed us to
raise the hypothesis on the existence of an unprecedent-
edly described allosteric site on the surface of the closed
conformation of hAK (16).
The computational evidence of an allosteric mode of
action was further strengthened by the fact that the inhibi-
tory potency of NSD438 was unaffected by increasing
concentrations of the AK substrates Ado or ATP, indicat-
ꢀ
was within 0.3 A of the crystallographically determined
Chem Biol Drug Des 2016; 87: 112–120
115

Savi et al.
ing a non-competitive mode of inhibition (16). In the newly
identified pocket, our lead NSD438 was found, thorough
docking studies, to lie on a3 helix sandwiched between
helix a4 and the long carboxylic tail. Analysis of the key
interactions between NSD438 and hAK highlighted F338
as key residue in the interaction with the ligand, as an H-
bond and a p–p interaction was found with NSD438 (16)
(Figure 1).
the structures of the selected compounds and their IC50
values against wt-hAK.
Stereoselective synthesis of compounds (S)- and
(R)-NSD438
Although the racemic mixture rac-NSD438 proved to be
separable in the single enantiomers by HPLC analytical
methodologies, the poor solubility of this compound in typ-
ical mobile phases severely limited the scalability of this
method. Moreover, the oily nature of NSD438 hampered
the possibility to establish the absolute configuration by
X-ray techniques. These issues prompted us to develop a
straightforward stereoselective strategy to generate the
single enantiomers. We already settled and proposed a
suitable chemical strategy (20) which we herein exploit for
the preparation of both NSD438 enantiomers. The chemi-
cal route employed for the stereoselective synthesis of
compounds (R)- and (S)-NSD438 is reported in Scheme 1.
Starting from toluene (1), a standard Friedel–Crafts acyla-
tion led to a-ketoacid 2 and its corresponding ethyl ester
3. These latter were both employed for the preparation of
the (+)-menthyl ester derivative 4. Indeed, acid 2 was con-
densed with (+)-menthol in the presence of p-toluenesul-
fonic acid (21), while ester 3 underwent a modification of
the titanium (IV) alkoxyde-catalyzed transesterification
developed by Seebach (22). Accordingly, treatment of 3
with (+)-menthol in the presence of a catalytic amount of
titanium isopropoxide (20 mol %) in toluene provided the
corresponding (+)-menthyl ester 4 in high yield. Reaction
of 4 with ethyl magnesium bromide in the presence of 2
eq of zinc chloride afforded chiral a-hydroxy ester 5 in
good yield and with a 92:8 diastereomeric ratio (dr) by
NMR determination (the two diastereoisomers being easily
separated by flash chromatography). The ensuing conver-
sion of menthyl ester to its corresponding methyl derivative
was performed in order to overcome subsequent issues
with the cleavage of menthyl ester on more hindered sub-
strates which required harsh conditions and provided
extremely poor yields. Accordingly, saponification was per-
formed providing the corresponding acid 6 in turn con-
verted into the methyl ester 7 by a mild procedure
employing methyl iodide in the presence of potassium car-
bonate (23). SNAr performed on derivative 7 on the acti-
vated aryl fluoride 4-nitro-2-pyrrolylfluorobenzene provided
the aryl-alkyl ether 8 in good yield. The subsequent
reduction performed with tin(II) chloride in refluxing etha-
nol led to aniline derivative 9. The next one-pot reaction
On the carboxylic tail, two residues (F338 and K341) were
found in the proximity of the NSD438 terminal phenyl ring,
and hydrophobic interactions with K341 carbonyl chain
were conceivable. In light of these observations, we
developed F338A and K341A mutants and compared the
affinity of a subset of our NNhAKIs analogs with the affinity
toward the wt enzyme. However, besides the peculiar
interactions that each enzyme residue could establish or
not with NSD438, to localize the site of binding for our
NNhAKIs, we decide to select, for our mutagenesis
studies, some residues that border and define the hypo-
thetical binding site. We decided to focus on the
phenylmercaptomethyl subpocket, as in NSD438, the
phenylmercaptomethyl was predicted to possess a larger
surface of interaction with respect to the pyrrolobenzox-
azepinone scaffold that was found perpendicularly oriented
with respect to a3 helix in our previous computational
studies. On the basis of the above, Q74, Q78, H107 were
selected for Ala scan. A double mutant Q74A-F338A was
also considered.
Thus, specific primers were designed (Table S1) to create
single and double amino acid substitutions of hAK using
site-directed PCR mutagenesis. Each mutant, expressed
in E. coli (DH5a), purified as His6-tag recombinant protein,
once biochemically characterized in vitro, would have
given data allowing us to validate the hypothesized inhibi-
tor binding site.
For exploring the effect of single and double mutations on
hAK on the inhibition potency of our NNhAKIs, besides
our hit compound rac-NSD438, we selected a small set of
analogs characterized by inhibition potencies ranging from
1.0 to 12.5 lM on the wt enzyme. All the compounds were
used as racemic mixtures as computational studies
already predicted no stereoselectivity of interaction for
NSD438 (16). To confirm this hypothesis, we developed a
stereoselective approach for (S)-NSD438 and (R)-NSD438
synthesis. The outcome on hAK wild-type inhibition by
compounds (R)- and (S)-NSD438 was in total agreement
with the computational predictions, and no difference was
observed, in this study, in the inhibition potency of both
enantiomers with respect to the racemic mixture of
NSD438. In the selection process of inhibitors to be
engaged in this biochemical study, structural diversity in
terms of terminal aryl moiety sizes, electronic properties,
and presence of H-bonding donor/acceptor groups
(Table 1) were taken into account. Table 1 summarizes
was
a diazotization–dediazotization and provided the
deaminated product 10 in moderate yield. The reaction
encompassed the use of sodium nitrite in the presence
of acetic acid as a mild diazotization procedure and
sodium bisulfite as the reducing agent (24). Finally, start-
ing from compound 10, the synthetic route for the
achievement of the pyrrolobenzoxazepine scaffold fully
traced out the racemic synthesis already reported, with
alkaline hydrolysis of the methyl ester and followed by an
116
Chem Biol Drug Des 2016; 87: 112–120

Discovery of a Novel Allosteric Site on hAK
Table 1: IC50 (lM) values on recombinant hAK mutants. Each
IC50 value is the mean of at least two determinations. SEM is in
the range between 15 and 25%
partially overlapping primers, complementary to a portion
of the gene of interest, both of them bearing the desired
mutation. Each primer, approximately 40 bases long,
presents a superimposed region of about 20 bp, corre-
sponding to the 50 end of each primer, containing the
specific mutation. Five pairs of primers were synthesized
to replace, with alanine, the following amino acids, puta-
tively in the inhibitor binding site: Q74, Q78, H107,
K341, and F338. By removing or creating unique restric-
tion sites, without modifying the amino acid sequence of
hAK, we were able to distinguish mutant from the wt
plasmid, used as template in the PCR. The PCR prod-
ucts were then treated with DpnI (GmA/TC), to fully
digest the parental methylated DNA (25 restriction sites)
from newly synthesized unmethylated mutant DNA.
Then, 1 ng of plasmid DNA of each mutant was used
to transform E. coli (DH5a) competent cells. Expression
and purification of recombinant hAKs were carried out
as described in Methods and Materials. The enzymes
were purified from E. coli crude extract in a single chro-
matographic step using a Ni-NTA Superflow column (Qi-
agen). To characterize the hAK mutants, the enzyme
was incubated in the presence of different concentra-
tions of labeled substrate ([2,8-3H]-Ado) using ATP as
phosphate donor at 0.1 mM. Table 2 summarizes the
kinetic parameters of recombinant hAK mutants com-
pared with wt-hAK (16). The Km values demonstrate that
the single-site and the double-site mutations of the
amino acid residues present in the novel hypothesized
allosteric site of hAK change the enzyme affinity for the
substrate, suggesting a functional relationship between
the putative allosteric site and the catalytic site of the
enzyme. In particular, Kcat and Kcat/Km, which measure
the enzyme turnover and the efficiency of catalysis,
respectively, indicate that the introduced point mutations
somehow affect the behavior of hAK. For instance, Q74A,
Q78A, and F338A mutants have a reduced turnover
(0.05, 0.05, and 0.08, respectively, compared to 0.5 of
wt-hAK) and a 10-fold lower efficiency of phosphorylation
(0.10, 0.13, and 0.10, respectively, compared to 1.25 of
wt-hAK). As expected, the double mutant Q74A/F338A-
hAK, although possessing a Km slightly lower than that of
Human AK
Q74A-
Compounds
X
Y
S
wt Q74A
1.2 6.4
Q78A H107A K341A F338A F338A
NSD438
O
2.0
2.6
3.5
11.8
>100
14
15
16
O
O
O
S
S
S
3.5 17.4
1.0 3.6
2.0 10.2
5.9
0.8
1.1
5.1
3.0
6.1
23.2
3.7
70.2
13.3
30.6
>100
>100
>100
4.0
17
18
O
O
S
2.0 5.0
0.6
2.3
3.1
3.3
4.5
25.4
60.5
>100
>100
O
1.2 20.3
19.8
19
S
S
2.0 15.4
0.4
3.4
5.2
28.5
>100
intramolecular Friedel–Crafts reaction leading to the
desired (S)-NSD438 (25).
wt-hAK, shows both a lower Kcat (0.02) and a lower Kcat
/
Km (0.05), indicating a slow and inefficient phosphorylating
activity. On the other hand, K341A-hAK displays a higher
affinity for the substrate Ado (Km: 0.17 lM) compared to
wt-hAK, but a lower turnover (Kcat: 0.22); these values
lead to a Kcat/Km ratio similar to that reported for the wt-
hAK. Finally, mutant H107A-hAK possesses a lower affin-
ity for the substrate (Km: 1.66 lM) and a Kcat similar to
the wt-hAK, which results in a fourfold lower efficiency of
phosphorylation.
The chemical route for the stereoselective preparation of
cyclic ketone (R)-12 was already described by us (20).
Starting from this optically active intermediate, a radicalic
bromination and alkylation led to desired compound (R)-
NSD438.
Cloning, expression, and biochemical
characterization of recombinant wt-hAK and hAK
mutants
Each hAK mutant was produced by site-directed mutage-
nesis based on PCR, following the method described in
Methods and Materials. This approach, which allowed us
to directly amplify the recombinant plasmid, utilizes two
It must be underlined that, from a structural point of view,
although the binding clefts of Ado and NSD438 are two
distinct sites, they share a portion of the a3 helix and of
the enzyme carboxylic tail. In fact, the central part of the
a3 helix, from which Q74 protrudes, forms the NSD438
Chem Biol Drug Des 2016; 87: 112–120
117

Savi et al.
Table 2: Kinetic parameter of recombinant wt-hAK and mutants.
Km value is the mean of at least two determinations. Kcat was cal-
culated using 46 kDa as the molecular mass of the enzyme
action of the inhibitors, having a Hill’s coefficient >1 (data
not shown), and (ii) the expected reduced inhibitory activity
of the compounds against hAK mutants (see Figure 2 for
NSD438). The outcome on hAK wild-type inhibition by
compounds (R)- and (S)-NSD438 was in total agreement
with the computational predictions, and no difference was
observed in the inhibition potency of both enantiomers
with respect to the racemic mixture (no stereoselective
mode of interaction).
Vmax
(pmol secondsꢁ1 lgꢁ1
Kcat
(secondsꢁ1
Kcat/Km
(secondsꢁ1/lM)
hAK
Km (lM)
)
)
wt
0.40 ꢀ 0.10
0.48 ꢀ 0.20
0.39 ꢀ 0.07
1.66 ꢀ 0.41
0.17 ꢀ 0.02
0.83 ꢀ 0.24
0.36 ꢀ 0.05
10.7 ꢀ 1.00
1.05 ꢀ 0.13
1.00 ꢀ 0.08
7.74 ꢀ 0.08
4.86 ꢀ 0.02
1.85 ꢀ 0.12
0.36 ꢀ 0.02
0.50
0.05
0.05
0.35
0.22
0.08
0.02
1.25
0.10
0.13
0.27
1.29
0.10
0.05
Q74
Q78
H107
K341
F338
Q74-F338
The previously proposed binding mode of NNhAKIs high-
lighted F338 as a key residue in the interaction with our inhi-
bitors, as both an H-bond and a p–p interaction were found
with NSD438. Accordingly, among all the mutations, the
F338A substitution resulted as the more effective, being
responsible of a substantial reduction in the inhibition
potency of the compounds. Point mutations Q78A and
H107A, both lining the phenylmercaptomethyl subpocket of
the allosteric site, did not significantly affect the inhibition
potency of the compounds if compared to the effect F338A.
The other two explored mutations K341A and Q74A signifi-
cantly influenced the binding of NSD438 analogs being the
Q74A mutant slightly more effective in reducing binding.
Among the tested compounds, the analog 14 was the most
sensible to the mutation of both K341A and Q74A. From
docking calculation, it resulted that the imidazole ring occu-
pies the same position of the mercaptophenyl group (Fig-
ure 1) and sits in proximity of both K341 and Q74.
Differently from NSD438 analogs 15–19, the imidazole ring
of 14 would be able to accept an H-bond from K341 or Q74
side chain, thus justifying lower inhibition potency for the
K341A and Q74A enzyme mutants.
cleft pavement, while its end forms the Ado binding site
border. Along the same lines, F338 is part of the flexible
carboxylic tail, which shapes NSD438 binding site and
represents one edge of the Ado binding site. Interestingly,
in the close conformation of the enzyme (PDB code:
1BX4), the flexible carboxylic tail seems to be held in
place by a tight interaction (distance between the rings
ꢀ
centroids = 4.5 A) between F338 and W75 (this latter sit-
ting next to Q74 on the a3 helix). Thus, we can envisage
that F338A mutation, disrupting this p–p interaction, would
allow a major flexibility of the carboxylic tail, not beneficial
for the AK substrate affinity. Differently, the double mutant
F338A-Q74A somehow regains substrate affinity showing
that the two amino acids co-operate in a structural
change advantageous for Ado affinity. Noteworthy,
together with N68, Q67, and W75, the Q74 residue
seems to be important for a3 helix positioning as its polar
side chain is involved in multiple H-bonds with backbone
CO and NH of P80, A84, and K82. The single mutation of
Q74 does not seem to tangibly perturb the enzyme struc-
ture and conformation; however, the simultaneous muta-
tion of Q74 and F338, which was also important in
stabilizing the a3 helix position (see above), may instead
allow a small shifting of the a3 helix, fruitfully affecting Ado
binding. However, further studies could clarify this issue.
To better explore the influence of the introduced mutations
on the inhibitor–enzyme interaction, we cloned and
expressed a double mutant combining the single muta-
tions which mostly affected the inhibitory activity of the
compounds. Accordingly, double alanine mutant in the key
residues F338 and Q74 was generated to evaluate possi-
ble additive effects in the reduction of the binding of our
inhibitors. As expected from the results obtained with the
single mutants, Q74A-F338A-hAK mutant showed a total
loss of affinity for all the tested compounds (Table 1).
Inhibition assays on recombinant wt-hAK and hAK
mutants and rationalization of inhibitors IC50 on
enzyme mutants
The selected compounds, bearing differently functionalized
arylmercaptomethyl moieties (Table 1), were tested on sin-
gle-mutant enzymes (Q74A, Q78A, H107A, K341A,
F338A) and on the double mutant (Q74A-F338A) (see
Methods and Materials) to validate the proposed hypothe-
sis of a novel allosteric binding site. The inhibition effect on
hAKs activity was examined over a range of inhibitors’
concentrations with both substrates Ado and ATP at 1 lM
and 0.1 mM, respectively (Table 1). The inhibition curve for
all compounds follows a sigmoid shape rather than a right
Michealis Menten rectangular hyperbola. The data were
fitted to the Hill equation (r = n[A]n/(Kd+[A]n)) to determine
whether the inhibition exhibited cooperative characteristics.
In general, our results confirm (i) a cooperative mode of
To further investigate whether the mutations introduced in
hAK were only able to modulate the interaction of our
inhibitors with their putative allosteric pocket, we evalu-
ated the hAK inhibitory activity of 5-iodotubercidine
(5-ITU), a known potent competitive inhibitor which binds
the catalytic site (IC50 value in nM range) (19), against the
most efficient mutant developed Q74A-F338A-hAK
in vitro. Interestingly, like in wt-hAK, the double Q74A-
F338A-hAK mutant was inhibited by 5-ITU in the
nanomolar range (Figure 3), thus providing further evi-
dence of the presence in hAK of an allosteric pocket
independent from the catalytic/regulatory site. We
definitely proved the discovery of an allosteric binding site
in hAK for non-nucleoside inhibitors, which substantially
118
Chem Biol Drug Des 2016; 87: 112–120

Discovery of a Novel Allosteric Site on hAK
Figure 3: In vitro effect of increasing concentrations of 5-ITU on
Q74AF338A-hAK activity (IC50, 25 nM). Each point is the average
of three determinations.
secondary structure, thus making impossible to unambigu-
ously establish a clear amino acid correspondence (see
Sequence Alignment in Figure S1). The same holds true
when a comparison with human ketohexokinase (PDB
code: 2HLZ) is performed (Figure S2B). In summary, in
hAK, the novel allosteric site, herein identified, is definitely
shaped by amino acids which are peculiar of this enzyme.
To the best of our knowledge, no other ligands of RK fam-
ily are known to bind this site.
Conclusion
The computational data herein discussed coupled with
Ala-scan mutagenesis studies contributed to confirm the
existence of a putative allosteric site on hAK, which is
exclusively bound by our recently developed a pyrroloben-
zoxa(thia)zepinone NNhAKIs. The discovery of an extra
allosteric site into the human enzyme may be exploited for
the discovery of more potent and selective NNhAKIs
endowed with therapeutic potential.
Figure 2: Effect of NSD438 on wt-hAK and hAK mutants (each
point is the average of two determinations in two independent
experiments). To standardize the enzyme velocities in parallel
experiments, we plotted Vmax/V against NSD438 concentrations.
When in the assay NSD438 concentration is 0 lM, being
V = Vmax
,
the y-axis value is 1. At NSD438 increasing
concentrations, being V < Vmax, the y-axis value increases up to
∞, when V = 0. For the discussion about the nonlinear shape of
the curve, see the text and (Ref. 16).
Acknowledgments
This work was partially supported by the Consiglio Nazio-
ꢁ
nale delle Ricerche/Ministro dell’Istruzione, dell’Universita e
diverges from the same site in other enzymes of the
same family.
della Ricerca ‘Progetto d’interesse invecchiamento’, and
the FIRB Grant RBAU01LSR4_001 (to F.F.). L.S. was sup-
ported by a doctoral fellowship provided by Regione Lom-
bardia, Project ID 13810040 ‘Plant Cell’.
Indeed, as stated in the introduction, AK belongs to the
ribokinase family (13) of enzymes which phosphorylate the
hydroxymethyl groups of a variety of sugars. Although the
overall sequence identity between the members of the RK
family is <30% (see Sequence Alignment in Figure S1),
their tridimensional structures are indeed similar (13).
Particularly, AK and ribokinase (RK) are two important
ribokinase family of enzymes that share unique secondary
and tertiary structural elements (13). A three-dimensional
superposition of hAK (PDB code: 1BX4) and hRK (PDB
code: 2FV7) demonstrates that the allosteric site identified
by us highly diverges between the two proteins due to
multiple mutations especially in the a3 helix and in the C-
terminal tail. As for example, residues such as Q67, K71,
Q74, W75 within the a3 helix of hAK are replaced in RK
by A56, V60, A63, R64, respectively (see Figure S2A).
Notably, the carboxylic tail of hAK (PDB code: 1BX4) heav-
ily differs from that of hRK (PDB code: 2FV7), even in its
Conflict of Interest
Authors declare no conflict of interest.
References
1. Pignataro G., Simon R.P., Boison D. (2007) Trans-
genic overexpression of adenosine kinase aggravates
cell death in ischemia. J Cereb Blood Flow Metab;
27:1–5.
2. Pignataro G., Maysami S., Studer F.E., Wilz A., Simon
R.P., Boison D. (2008) Downregulation of hippocampal
adenosine kinase after focal ischemia as potential
endogenous neuroprotective mechanism.
Blood Flow Metab;28:17–23.
J
Cereb
Chem Biol Drug Des 2016; 87: 112–120
119

Savi et al.
3. Boison D. (2013) Adenosine kinase: exploitation for
therapeutic gain. Pharmacol Rev;65:906–943.
4. Fassett J.T., Hu X., Xu X., Lu Z., Zhang P., Chen Y.,
Bache R.J. (2011) Adenosine kinase regulation of car-
diomyocyte hypertrophy. Am J Physiol Heart Circ Phys-
iol;300:H1722–H1732.
5. Annes J.P., Ryu J.H., Lam K., Carolan P.J., Utz K.,
Hollister-Lock J., Arvanites A.C., Rubin L.L., Weir G.,
Melton D.A. (2012) Adenosine kinase inhibition selec-
tively promotes rodent and porcine islet beta-cell
replication. Proc Natl Acad Sci USA;109:3915–3920.
6. Park J., Gupta R.S. (2013) Adenosine metabolism,
adenosine kinase, and evolution. In: Masino S., Boison
D., editors. Adenosine: A Key Link between Metabo-
lism and Brain Activity. New York: Springer Science +
Business Media; p. 23–54.
17. Liu H.T., Naismith J.H. (2008) An efficient one-step site-di-
rected deletion, insertion, single and multiple-site plasmid
mutagenesis protocol. BMC Biotechnol;8:91–100.
18. Chung C.T., Niemela S.L., Miller R.H. (1989) One-step
preparation of competent Escherichia-coli – transfor-
mation and storage of bacterial-cells in the same solu-
tion. Proc Natl Acad Sci USA;86:2172–2175.
19. Ugarkar B.G., DaRe J.M., Kopcho J.J., Browne C.E.
III, Schanzer J.M., Wiesner J.B., Erion M.D. (2000)
Adenosine kinase inhibitors. 1. Synthesis, enzyme inhi-
bition, and antiseizure activity of 5-iodotubercidin ana-
logues. J Med Chem;43:2883–2893.
20. Brindisi M., Gemma S., Alfano G., Kshirsagar G., Novel-
lino E., Campiani G., Butini S. (2013) A stereoselective
route to 6-substituted pyrrolo-1,5-benzoxazepinones and
their analogues. Tetrahedron Lett;54:5387–5390.
7. Williams-Karnesky R.L., Sandau U.S., Lusardi T.A.,
Lytle N.K., Farrell J.M., Pritchard E.M., Kaplan D.L.,
Boison D. (2013) Epigenetic changes induced by ade-
nosine augmentation therapy prevent epileptogenesis.
J Clin Invest;123:3552–3563.
21. Basavaiah D., Krishna P.R. (1995) New cyclohexyl-
based chiral auxiliaries – enantioselective synthesis
of alpha-hydroxy acids. Tetrahedron;51:12169–12178.
22. Xiang J.M., Li B.L. (2010) The stereoselective synthesis
of 2-Aryl-2-hydroxybutanoic acid via menthyl chiral
auxiliaries. Helv Chim Acta;93:2015–2022.
8. Jia G., Fu Y., He C. (2013) Reversible RNA adenosine
methylation
Genet;29:108–115.
in
biological
regulation.
Trends
23. Ji F., Wu W., Dai X., Mori N., Wu J., Buchwald P.,
Bodor N. (2005) Synthesis and pharmacological effects
of new, N-substituted soft anticholinergics based on
glycopyrrolate. J Pharm Pharmacol;57:1427–1435.
24. Geoffroy O.J., Morinelli T.A., Meier G.P. (2001)
Chemoselective one-pot reductive deamination of aryl
amines. Tetrahedron Lett;42:5367–5369.
25. Fattorusso C., Gemma S., Butini S., Huleatt P., Cata-
lanotti B., Persico M., De Angelis M. et al. (2005)
Specific targeting highly conserved residues in the
HIV-1 reverse transcriptase primer grip region. Design,
synthesis, and biological evaluation of novel, potent,
and broad spectrum NNRTIs with antiviral activity. J
Med Chem;48:7153–7165.
9. Kredich N.M., Martin D.V. Jr (1977) Role of S-adeno-
sylhomocysteine in adenosine-mediated toxicity in cul-
tured mouse T lymphoma cells. Cell;12:931–938.
10. Boison D., Scheurer L., Zumsteg V., Rulicke T., Lityn-
ski P., Fowler B., Brandner S., Mohler H. (2002)
Neonatal hepatic steatosis by disruption of the adeno-
sine kinase gene. Proc Natl Acad Sci USA;99:6985–
6990.
11. James S.J., Melnyk S., Pogribna M., Pogribny I.P.,
Caudill M.A. (2002) Elevation in S-adenosylhomocys-
teine and DNA hypomethylation: potential epigenetic
mechanism for homocysteine-related pathology.
J
Nutr;132:2361S–2366S.
12. Spychala J., Datta N.S., Takabayashi K., Datta M.,
Fox I.H., Gribbin T., Mitchell B.S. (1996) Cloning of
human adenosine kinase cDNA: sequence similarity to
microbial ribokinases and fructokinases. Proc Natl
Acad Sci USA;93:1232–1237.
13. Park J., Gupta R.S. (2008) Adenosine kinase and
ribokinase–the RK family of proteins. Cell Mol Life
Sci;65:2875–2896.
Note
aGlide (2009). New York, S., LLC.
Supporting Information
Additional Supporting Information may be found in the
online version of this article:
14. Mathews I.I., Erion M.D., Ealick S.E. (1998) Structure
of human adenosine kinase at 1.5 A resolution. Bio-
chemistry;37:15607–15620.
Figure S1. Multiple sequence alignment of AK, Ribokinase
RK and Ketohexokinase (KHK) from humans, mouse and
rat.
15. Parducci R.E., Cabrera R., Baez M., Guixe V. (2006)
Evidence for a catalytic Mg2+ ion and effect of phos-
phate on the activity of Escherichia coli phosphofruc-
tokinase-2: regulatory properties of a ribokinase family
member. Biochemistry;45:9291–9299.
Figure S2. Superposition of the crystal structure of hAK
with that of hRK and of hKHK.
16. Butini S., Gemma S., Brindisi M., Borrelli G., Lossani
A., Ponte A.M., Torti A. et al. (2011) Non-nucleoside
inhibitors of human adenosine kinase: synthesis,
molecular modeling, and biological studies. J Med
Chem;54:1401–1420.
Appendix S1. Experimental details and characterization
for the synthesized compounds.
Table S1. Primer used for site-directed mutagenesis.
Chem Biol Drug Des 2016; 87: 112–120
120