Frontal affinity chromatography-mass spectrometry useful for characterization of new ligands for GPR17 receptor

Article abstract of DOI:10.1021/jm901691y

The application of frontal affinity chromatography-mass spectrometry (FAC-MS), along with molecular modeling studies, to the screening of potential drug candidates toward the recently deorphanized G-protein-coupled receptor (GPCR) GPR17 is shown. GPR17 is dually activated by uracil nucleotides and cysteinyl-leukotrienes, and is expressed in organs typically undergoing ischemic damage (i.e., braheart and kidney), thus representing a new pharmacological target for acute and chronic neurodegeneration. GPR17 was entrapped on an immobilized artificial membrane (IAM), and this stationary phase was used to screen a library of nucleotide derivatives by FAC-MS to select high affinity ligands. The chromatographic results have been validated with a reference functional assay ([³⁵S]GTPγS binding assay). The receptor nucleotide-binding site was studied by setting up a column where a mutated GPR17 receptor (Arg255Ile) has been immobilized. The chromatographic behavior of the tested nucleotide derivatives together with in silico studies have been used to gain insights into the structure requirement of GPR17 ligands.

Full text of DOI:10.1021/jm901691y

J. Med. Chem. 2010, 53, 3489–3501 3489
DOI: 10.1021/jm901691y
Frontal Affinity Chromatography-Mass Spectrometry Useful for Characterization of New Ligands for
GPR17 Receptor
Enrica Calleri,†,¥ Stefania Ceruti,‡,¥ Gloria Cristalli,§,¥ Claudia Martini, ,¥ Caterina Temporini,† Chiara Parravicini,‡
Rosaria Volpini,§ Simona Daniele, Gabriele Caccialanza,† Davide Lecca,‡ Catia Lambertucci,§ Maria Letizia Trincavelli,
Gabriella Marucci,§ Irving W. Wainer,^ Graziella Ranghino,ꢀ Piercarlo Fantucci,ꢀ,# Maria P. Abbracchio,‡,¥ and
Gabriella Massolini*,†,¥
†Department of Pharmaceutical Chemistry, University of Pavia, Via Taramelli 12, 27100 Pavia, Italy, ‡Laboratory of Cellular and Molecular
Pharmacology of Purinergic Transmission, Department of Pharmacological Sciences, Universitaꢀ degli Studi di Milano, Via Balzaretti 9,
20133 Milan, Italy, §Department of Chemical Sciences, University of Camerino, Via S. Agostino 1, 62032 Camerino, Italy, Department of
Psychiatry, Neurobiology, Pharmacology, and Biotechnology, University of Pisa, Via Bonanno 6, 56126 Pisa, Italy, ^Gerontology Research
Center, National Institute on Aging, National Institutes of Health, Baltimore, Maryland 21224, ꢀDelos S.r.l., Via Lurani 12, 20091 Bresso (MI),
Italy, and #Department of Biotechnology and Biosciences, Universitaꢀ di Milano-Bicocca, Piazza della Scienza 2, 20126 Milan, Italy.
¥ These authors equally contributed to this paper.
Received November 16, 2009
The application of frontal affinity chromatography-mass spectrometry (FAC-MS), along with
molecular modeling studies, to the screening of potential drug candidates toward the recently
deorphanized G-protein-coupled receptor (GPCR) GPR17 is shown. GPR17 is dually activated by
uracil nucleotides and cysteinyl-leukotrienes, and is expressed in organs typically undergoing ischemic
damage (i.e., brain, heart and kidney), thus representing a new pharmacological target for acute and
chronic neurodegeneration. GPR17 was entrapped on an immobilized artificial membrane (IAM), and
this stationary phase was used to screen a library of nucleotide derivatives by FAC-MS to select high
affinity ligands. The chromatographic results have been validated with a reference functional assay
([35S]GTPγS binding assay). The receptor nucleotide-binding site was studied by setting up a column
where a mutated GPR17 receptor (Arg255Ile) has been immobilized. The chromatographic behavior of
the tested nucleotide derivatives together with in silico studies have been used to gain insights into the
structure requirement of GPR17 ligands.
Introduction
be done to expand and validate the application of this
methodology to the screening of combinatorial libraries to-
ward important biological targets such as G-protein-coupled
receptors (GPCRs).
The combination of frontal affinity chromatography (FACa)
and mass spectrometry (MS) for binding studies is a rather
new analytical concept that has been first described in 1998 by
Schreimer and Hindsgaul.1-3 On one side, FAC allows us to
study molecular interactions in a simple fashion, is amenable
to automation, and provides precise and accurate binding
data. On the other hand, MS allows the identification of
compounds solely based on their mass-to-charge ratio (m/z)
and is a label-free detection system. The principles of FAC-
MS have been well described.4-8 Nonetheless, the use of
FAC-MS is not yet so widespread and more work should
GPCRs are important appealing pharmaceutical targets,
but useful potent and selective ligands do not exist for the
majority of these receptors and several of them are still
“orphan”. In addition, from a drug discovery and design
prospective, structural information on GPCRs can represent
a fundamental tool for structural-based drug design. How-
ever, the X-ray structure has been determined for only four
GPCRs (i.e., the visual receptor rhodopsin, β1 and β2 adreno-
receptors, and very recently the A2A adenosine receptor).9,10
The feasibility of the FAC approach has already been
demonstrated for some GPCRs upon immobilization on
artificial membranes (IAM);11,12 however, the application of
these particular stationary phases to screening studies has not
been yet reported. In this respect, we have recently described
the development of a chromatographic stationary phase by
immobilizing on IAM particles membrane fragments from a
cell line expressing the recently deorphanized G-protein-
coupled receptor GPR17.13
*To whom correspondence should be addressed. Phone: þ39 0382
987383. Fax: þ39 0382 422975. E-mail: g.massolini@unipv.it.
a Abbreviations: 1P, monophosphate; 2P, diphosphate; 3P, triphos-
phate; 3D, three-dimensional; ATP, adenosine triphosphate; CDI, N,N0-
carbonyldiimidazole; COM, center of mass; CysLT, cysteinyl-leukotriene;
DAD, diode array detector; DEAE, diethylaminoethyl; DMF, N,N-
dimethylformamide; ESI, electrospray ionization; FAC, frontal affinity
chromatography; GPCR, G-protein-coupled receptor; GTPγS, guanosine
50-O-[γ-thio]triphosphate); HEPES, 4-(2-hydroxyethyl)piperazine-1-etha-
nesulfonic sodium salt; IAM, immobilization on artificial membranes; LB,
Luria-Bertani; LTQ, linear trap quadrupole; MS, mass spectrometry;
PCR, polymerase chain reactions; PMSF, phenylmethylsulfonyl fluoride;
SEM, standard error of the mean; SMD, steered molecular dynamics
simulations; TPCK, tosylamido-2-phenylethyl chloromethyl ketone; UDP,
uridine diphosphate; WT, wild type.
This is an interesting “dual “ receptor responding to two
distinct families of inflammatory mediators, namely, uracil
nucleotides and cysteinyl-leukotrienes (cysLTs), which are
massively releasedin the brainupon different typesofinjury.14
r
2010 American Chemical Society
Published on Web 04/15/2010
pubs.acs.org/jmc

3490 Journal of Medicinal Chemistry, 2010, Vol. 53, No. 9
Calleri et al.
We have recently reported that GPR17 is pathologically
activated during acute CNS injury, thus contributing to early
necrotic death inside the lesion,14 but also participates in the
subsequent remodeling and repair of the lesioned area occur-
ringinthe days and weeksafter injury.15 Thus, GPR17 ligands
could represent a new class of neuroprotective and neuro-
reparativeagents in several types of humanneurodegenerative
diseases, such as stroke, trauma, and multiple sclerosis. Since,
as mentioned above, GPR17 dually responds to two unrelated
families of endogenous ligands and may represent just the first
example of a series of “dual” GPRCs,16 we hypothesize that
the identification of selective GPR17 ligands could lead to the
development of novel dualistic pharmacological agents with
previously unexplored therapeutic potential.
In this context, elucidation of the structure of GPR17 and
of ligand binding mechanisms is a necessary steps to obtain
selectiveand potent drugsforthis newpotentialtarget. Onthis
basis, a 3D molecular model of GPR17 embedded in a sol-
vated phospholipid bilayer and refined by molecular dyna-
mics simulations has been created.17
receptor.27 Hence, compounds consisting of 18 members, ran-
ging from 479 to 675 in molecular weight, have been evaluated.
The synthesis of the known tested compounds was carried
out following already described procedures, which are listed
in Table 1. New nucleotides 7, 8, and 14 were prepared by
phosphorylation of the corresponding nucleosides 22,28 23,29
and 2630 (Scheme 1). Nucleosides were first phosphorylated
to the monophosphates, and from these the corresponding
di- and triphosphates were prepared.
Monophosphates 24, 25, and 27 were synthesized by
reacting the corresponding nucleosides 22, 23, and 26 with
POCl3 in trimethylphosphate at room temperature, follow-
ing the Yoshikawa method (Scheme 1).31 Synthesis of triphos-
phates 7 and 8 and diphosphate 14 were carried out by a
modification of the Hoard-Ott method (Scheme 1).32 To
this purpose, the tributylammonium salts of monopho-
sphates were activated as phosphoroimidazolates by reacting
with N,N0-carbonyldiimidazole (CDI). These unstable inter-
mediates were coupled with bis(tri-n-butylammonium) pyro-
phosphate at room temperature to give the corresponding
nucleoside triphosphates 7 and 8 (Scheme 1). The nucleoside
diphosphate 14 was prepared by reacting the corresponding
phosphorimidazolate with tri-n-butylammonium phosphate
at room temperature (Scheme 1).
Molecular dynamics simulations suggest that the GPR17
nucleotide binding pocket is similar to that described for the
other P2Y receptors, although only one of the three basic
residues that have been typically involved in ligand recogni-
tion is conserved (Arg255).
The phosphorylating reagents were freshly prepared start-
ing from the inorganic salts, which were first converted to
inorganic acid with Dowex Hþ form resin and then into the
tributylammonium salts by adding tri-n-butylamine. The
removal of water under vacuum, followed by coevaporation
with dry ethanol, furnished the dry organic salts, which were
suspended in dry DMF after coevaporation of the residue
ethanol with DMF.
In this work we report, for the first time, the application of
FAC-MS to the screening of a small library of compounds
toward GPR17 immobilized on a chromatographic support.
This approach allowed the rapid and accurate ranking of the
selected compounds. The chromatographic data have been
validated by a reference functional assay ([35S]GTPγS binding
assay). In addition, the simplicity of the method has allowed
us to explore the binding mode of the nucleotide site of the
receptor by extending this approach to a column where an
artificial GPR17 receptor bearing a mutation in a key amino
acid for ligand recognition has been immobilized. Compar-
ison of the different chromatographic behavior of the nucleo-
tide derivatives in the library together with in silico studies has
been used to gain insight into the structure requirement of
GPR17 ligands. We expect that this integrated approach can
contribute to GPCR-based drug discovery process.
All the new nucleotide analogues were obtained as ammo-
nium salts after ion exchange chromatography on a Sepha-
-
dex diethylaminoethyl (DEAE) column (HCO3 form),
eluted with a linear gradient of ammonium bicarbonate,
which gives a gradual increase of ionic strength.
Synthesized nucleotides were characterized with 1H NMR
and the phosphoric groups identified through 31P NMR
analysis in D2O. In fact mono-, di-, and triphosphates show
characteristic 31P NMR spectra. Monophosphate nucleo-
tides show a characteristic peak at about 0 ppm; diphos-
phates show characteristic doublet at -9 to -10 ppm, and
triphosphates show a series of peaks at -6, -11, and -22,
corresponding to the γ, R, and δ phosphorus atoms, respec-
tively. The phosphorus chemical shift is highly dependent on
the pH of the solution and the counterion of the nucleotide.
FAC-MS Screening of the Chemical Libraries. To explore
the ability of FAC-MS to rank mixtures containing multiple
ligands, a first column (column GPR17-IAM-I) was pre-
pared by immobilizing crude membranes of 1321N1 cells
expressing wild type GPR17, following the acquisitions from
previous studies on the optimization of the GPCR stationary
phase where the nonspecific interactions were minimized by
analyzing the retention of the three reference compounds on
a blank column only containing IAM and an additional
column containing membranes from cells transfected with an
empty vector.13,33 Because our previous studies demon-
strated that nonspecific interactions contribute to the total
retention time but do not interfere with the specific displace-
ment study, we have reasoned that it was no longer necessary
to run a blank column in parallel for every single ranking
experiment. The preparation of the receptor stationary phase
is described in the Experimental Section.
Results and Discussion
Chemical Library Design and Synthesis. The compound
library for GPR17 was selected taking into account the chemi-
cal structures of three known receptor ligands with different
potencies: the antagonists 1 (cangrelor; IC50=0.7 nM) and 2
(MRS 2179; IC50=508 nM) and the agonist 3 (UDP; EC50=
1.14 μM) (Table 1).14 Starting from the cangrelor structure
and with the aim of investigating the structural requirements
for ligand recognition at the nucleotide receptor active site, a
series of ATP derivatives 4-7, substituted at the 2- and/or
N6-position, were selected. Furthermore, in order to evaluate
the importance of the 3-nitrogen purine ring and the grade of
nucleotide phosphorylation, the 2,N6-disubstituted ATP
3-deaza analogue 8 and some diphosphate nucleotides 9-11
were evaluated. Furthermore, compound 12, a bisphosphate
derivative analogue to MRS 2179, and the two 5-substituted
UDP analogues 13 and 14 were selected. Moreover, some
phosphorylated acyclonucleotides 15-21, bearing an adenine
base and a four-carbon alkyl spacer in the 9-position, which
can mimic the ribose moiety of nucleotide ligands, were added
to the selection to obtain additional information about the
structural requirements of the sugar portion that binds the

Article
Journal of Medicinal Chemistry, 2010, Vol. 53, No. 9 3491
Table 1. Chemical Structures of the Compounds Selected for FAC-MS Studies
The selected 18 compounds were split into two series:
analogues of nucleotides are included in library A, while
acyclic nucleotides are part of library B (see Tables 1 and 2).
Solutions containing the ligands belonging to the two libraries

3492 Journal of Medicinal Chemistry, 2010, Vol. 53, No. 9
Calleri et al.
Scheme 1a
a Reaction conditions: (a) (i) POCl3, (CH3O)3PO; (ii) NH4HCO3; b) (i) nBu3N, DMF; (ii) CDI, DMF; (iii) CH3OH, DMF; (iv) (nBu3NH)2(P2O7H2),
DMF; (v) NH4HCO3; (c) (i) n-Bu3N, DMF; (ii) CDI, DMF; (iii) CH3OH, DMF; (iv) (n-Bu3NH)(HPO4), DMF; (v) NH4HCO3.
and the three already known GPR17 ligands (UDP, MRS
2179, and cangrelor) all at 1 μM were prepared in mobile
phase buffer and continuously infused through the column.
The effluent was analyzed in an ESI mass spectrometer to
detect each components. The mass analysis of the library was
performed in negative mode to ensure efficient ionization
and detection of the compounds. Detailed chromatographic
experimental conditions are reported in the Experimental
Section. The potential ligands were retained in the column
based on their affinity, and the total run time was 55 and
20 min for libraries A and B, respectively. Because the mole-
cular weight for each compound is known, the individual
breakthrough front of each compound was easily identified
as shown in Figure 1 reporting the extracted breakthrough
curves for each analyte of library A and library B. The
breakthrough times reported in Table 2 are the results of a
single determination carried out on the freshly prepared
stationary phase. To compare the chromatographic results
obtained with the screening of the two libraries, the percen-
tage of breakthrough time with respect to cangrelor was
calculated for all the analytes (Table 2).
As expected, the elution order of the three reference
compounds in the two experiments was consistent with our
previous results13 and in agreement with their potency at this
receptor (cangrelor>MRS 2179>UDP). Moreover, a very
good correspondence between the relative breakthrough
% time of the less potent reference compounds (MRS 2179
and UDP) with respect to cangrelor was obtained in the two
experiments (relative % frontal time of 29 and 13 in the first
experiment and of 28 and 12 in the second experiment for
MRS 2179 and UDP, respectively).
Table 2. Frontal Affinity Chromatography-Mass Spectrometry (FAC-
MS) Results for Library A and Library B Screened with Immobilized
GPR17a
c
compd
t b(min)
t/tC ꢀ 100
Library A
3.20
4.06
UDP
13
13.03
16.54
19.47
19.67
25.99
28.92
30.67
36.09
37.56
45.05
49.90
68.43
100.00
120.37
9
14
4.78
4.83
12
MRS 2179
6.38
7.10
5
11
7.53
8.86
7
9.22
8
11.06
12.25
16.80
24.55
29.55
6
10
cangrelor
4
Library B
1.50
UDP
15
12.30
12.38
13.11
19.84
23.20
26.56
27.87
29.34
75.57
100.00
1.51
1.60
2.42
2.83
3.24
3.40
3.58
9.22
12.20
18
19
16
17
MRS 2179
20
21
cangrelor
a The breakthrough times are from a single raking experiment carried
out on the freshly prepared column. Column: GPR17-IAM-I (20.72
milion cells). Breakthrough times of the three reference compounds
(UDP, MRS 2179, and cangrelor) are marked in bold. See text for
detailed chromatographic conditions. b t is the breakthrough time of the
ligand with the immobilized GPR17. c tC represents the breakthrough
time of the indicator cangrelor.
In general, the first set of compounds (library A) was
significantly retained by the column, while the second set
(library B) weakly interacted with GPR17-column, with the
exception of the acyclonucleotide triphosphate 21, which
was eluted immediately before cangrelor with a relative %
frontal time of 75, which indeed represents one of the highest
values observed.
Moreover, in both series, the binding order of the phos-
phorylated compounds was consistent with the degree of
phosphorylation; hence, in library A the binding order was
3P>2P (compare 4 and 7 with 10 and 9, respectively) and in
library B the order was 3P > 2P > 1P (Table 2).
The large breakthrough volumes for compounds 4, 10,
and 21 indicate that they can be identified as high binding
candidates. In particular, for 4, which was the strongest
binding ligand, a subnanomolar Kd value can be hypothe-
sized from the chromatographic data, as it elutes after
cangrelor. On the other hand, compounds 5-8 and 11 seem
to be medium-high binding ligands, since they were eluted

Article
Journal of Medicinal Chemistry, 2010, Vol. 53, No. 9 3493
Figure 1. Frontal affinity chromatography-mass spectrometry of library A (a) and library B (b). Shown are extracted breakthrough curves
for each analyte of library A and library B. Mixtures of ligands in the presence of the three reference compounds each at 1 μM were infused
through the GPR17-IAM-I column using the mass spectrometer in negative mode.
with breakthrough times between those of cangrelor and
MRS 2179.
Compound 12 elutes with a breakthrough time close that
of MRS 2179. This finding is in agreement with the structure

3494 Journal of Medicinal Chemistry, 2010, Vol. 53, No. 9
Calleri et al.
of the two compounds that are very similar bisphosphate
analogues. The same consideration can be drawn for the
5-substituted UDP derivatives 13 and 14, which elute close to
UDP. These compounds can thus be classified as weak
binders toward GPR17, like UDP itself. Interestingly, can-
grelor and MRS 2179 are antagonists at GPR17 while UDP
acts as an agonist;14 thus, as already demonstrated for many
other receptors, also for GPR17, antagonists display higher
binding affinity with respect to agonists.
Table 3. Comparison of the Breakthrough Times of Derivatives in
Library A Obtained with Two Different GPR17-IAM Columnsa
GPR17-IAM-IIb
t (min)
t/tC ꢀ 100
2.69 25.38
GPR17-IAM-Ic
library A compd
t (min)
t/tC ꢀ 100
UDP
13
3.20
4.06
13.03
16.54
19.47
19.67
25.99
28.92
30.67
36.09
37.56
45.05
49.90
68.43
100.00
120.37
2.79
3.03
2.97
3.55
3.62
3.69
4.26
4.50
4.77
5.67
6.45
10.60
13.07
26.32
28.58
28.02
33.49
34.15
34.81
40.19
42.45
45.00
53.49
60.85
100.00
123.30
9
14
4.78
4.83
12
MRS 2179
6.38
7.10
On the basis of these results, library A containing the most
promising ligands was selected for further experiments.
The reproducibility of the method was assessed by pre-
paring and testing a second GPR17 column (GPR17-IAM-
II) starting from a new batch of transfected cells. Table 3
reports the data in comparison with the data previously
obtained on the GPR17-IAM-I column. An overall decrease
of the breakthrough times for all the compounds analyzed
with GPR17-IAM-II column was evident, but remarkably,
the relative ranking of the known and potential ligands was
highly consistent for both columns, with the only exception
of 9 and 14, for which an apparent reversal of the break-
through time order was observed. However, it must be
emphasized that for these two analytes, the elution time
difference (approximately 0.05 min) is too small to draw
such a conclusion, and a more plausible explanation is that
the two analytes coelute. The different run times between the
two GPR17 columns can be ascribed to different numbers of
the expressed receptors in the two different cell preparations,
as the transfection efficiency ranged from 30% to 50% (data
not shown). Globally, these results confirm that FAC-MS is
a powerful tool that can be used as a qualitative and
reproducible method for ligand discovery.
Structure-Affinity Relationships (SARs). The obtained
screening results have been used to draw preliminary binding
SAR speculations at the hit discovery stage that would reveal
ways of designing potent and selective GPR17 ligands. The
data from FAC-MS experiments suggest that lipophilic
substituents on the C-2 and on the N6 amino group of purine
ring can lead to a tight interaction with the hydrophobic
residues in the binding pocket. In fact, in both series listed in
library A and library B, compounds that strongly bind to the
receptor bear lipophilic groups in the 2-position like the
phenylalkynyl chain or halogens (see 4, 10, and 21) and/or an
alkyl/cycloalkyl substituent on the N6 amino group (see 6, 7,
and 8). The replacement of the phenyl group with a n-butyl
moiety brings about a reduction of binding affinity to
GPR17 (compare 10 and 11). On the other hand, it could
be that the presence of a nitrogen in 3-position is not essen-
tial, since its isosteric replacement with a CH does not signi-
ficantly change the breakthrough time (compare 7 and 8).
It is clear that in compounds with the same substituents on
the purine moiety, the rank order of potency is 3P>2P>1P
(i.e., 17>16>15 and 20>19>18). On the other hand, the
presence of an intact ribose seems to be crucial; in fact,
compounds of library B show, in general, lower retention
time compared with nucleotides analogues. It is worthwhile
to note that acyclonucleotide triphosphate 21, the only
compound of library B binding quite strongly to the column,
bears in the 2-position an iodine atom with high lipophilic
feature. Accordingly, it is clear that the nature and the
position of substituents on the purine ring modulate the
receptor binding affinity, the presence of rigid lipophilic
groups in 2-position being the most favorable for the inter-
action of these compounds with the receptor.
5
11
7.53
8.86
7
9.22
8
11.06
12.25
16.80
24.55
29.55
6
10
cangrelor
4
a The breakthrough times are from a single ranking experiment car-
ried out on the freshly prepared column. b Column: GPR17-IAM-II
(19.5 milion cells). c Column: GPR17-IAM-I (20.72 milion cells).
Functional Assay. The strength of a FAC-MS method is
the ability to rank multiple ligands simultaneously, offering
the possibility to discriminate between low, medium, and
high affinity candidates; however, it identifies compounds
simply on the basis of their binding affinities, irrespective of
their ability to activate the target.
In order to verify the reliability of the obtained results and
to evaluate the functional activity of the studied compounds
on GPR17, some derivatives were tested in a well established
GPCR assay, the [35S]GTPγS binding, based on the ability of
agonists to increase the binding of radioactive GTP to the
activated GPCR. Antagonist activity has been assessed by
the ability to counteract the increase of [35S]GTPγS binding
induced by the agonist UDP-glucose.14,15
To validate the FAC-MS data, the higher binding ligand
4, some of the presumed medium-high binding ligands 6-8,
the bisphosphate 12, and the UDP derivative 13 were selec-
ted for the functional study. The GPR17 antagonists can-
grelor, MRS 2179, and the agonist UDP were also assayed in
parallel as reference compounds.
The [35S]GTPγS binding results are reported in Table 4.
The obtained potency values of the tested ligands versus
wild-type GPR17 closely reflect the elution order of the ana-
lytes in the GPR17 wild-type column. A strong correlation
was found by comparing these results, expressed as half-
maximal response concentrations (EC50) or half-maximal
inhibition (IC50) values for agonists and antagonists, respec-
tively, with the data obtained with the GPR17 column by
FAC-MS, thus confirming the validity of our approach in
the ranking of new potential ligands for GPR17.
The ligands 13, 6, 8, 7, and 4 were able to increase
[35S]GTPγS binding, with potency values in the micromolar
and subnanomolar range, showing an agonist profile (Figure 2a).
In particular, the ATP analogue 4, bearing a lipophilic steric
hindered substituent in the 2-position, was a very potent
agonist of GPR17 with an EC50 of 36 pM, in agreement with
the fact that this is the compound with the longer break-
through time from the GPR17-column. Also, a lipophilic
substituent in the N6-position of ATP favored, although
to a less extent, the interaction with the GPR17 receptor.
In fact, comparison of potency of 6 (EC50=1.4 nM), bearing
a N6-cyclopentyl group, and 7 (EC50 = 11 nM), bearing a
smaller methyl group, demonstrated clearly the contribution

Article
Journal of Medicinal Chemistry, 2010, Vol. 53, No. 9 3495
Table 4. FAC-MS Data in Comparison with [35S]GTPγS Binding Results Obtained on GPR17-IAM-II and Mutated GPR17-IAMa
GPR17-IAM-II
EC50
mutated-GPR17-IAM
EC50
compd
UDP
13
t b (min)
IC50c (nM)
t b (min)
IC50 c (nM)
2.69
2.79
3.55
3.62
3.69
4.5
1.14 ( 0.2 μM
945 ( 48 nM
2.15
2.51
3.0 ( 0.3 μM
1.87 ( 0.20 μM
12
MRS 2179
582 ( 57
508 ( 29
112 ( 7
3.47
227 ( 22
5
7
11 ( 1 nM
8
4.77
5.67
10.6
13.07
1.7 ( 0.1 nM
1.4 ( 0.1 nM
6
cangrelor
0.7 ( 0.02
13.16
16.07
0.15 ( 0.01
4
36 ( 3 pM
25.4 ( 1.0 pM
a Columns: GPR17-IAM-II (19.5 million cells), mutated GPR17-IAM (19.5 million cells) b t is the breakthrough time of the ligand with the
immobilized GPR17. c Versus UDP-glucose 10 μM.
Figure 2. [35S]GTPγS experiments: (a) dose-response curves of some compounds of library A in 1321N1 cells expressing wild-type GPR17;
(b) antagonistic effect of 12 on UDP-glucose stimulation of wild-type GPR17 in the [35S]GTPγS binding, where each point is the mean ( SEM
of three independent experiments run in triplicate; (c) dose-response curves of UDP (3), 4, and 13 in 1321N1 cells expressing mutated GPR17;
(d) antagonistic effect of cangrelor (1) and MRS 2179 (2) on UDP-glucose stimulation of mutated GPR17 in the [35S]GTPγS binding, where
each point is the mean ( SEM of three independent experiments run in triplicate.
of the more lipophilic substituent. With regard to the purine
ring, we can confirm that the presence of a nitrogen in the
3-position is not essential (see above), since its isosteric
replacement by a CH induced even a 7-fold increase of
activity as shown by comparing 7 (EC50 = 11 nM) and 8
(EC50 = 1.7 nM).
Compound 12 did not induce any significant increase in
[35S]GTPγS binding; conversely, it was able to counteract
[35S]GTPγS binding stimulation induced by the purinergic
agonist UDP-glucose (used at 10 μM), with an antagonist
profile (Figure 2b) and an affinity constant (IC50) in the
nanomolar range, comparable to that reported for its analo-
gue derivative MRS 2179.14
As expected, the UDP derivative 13 showed an activity com-
parable to that of UDP itself (EC50 = 0.945 versus 1.140 μM,
respectively).
Comparison between Wild Type and Mutated GPR17.
Binding of extracellular nucleotides to P2Y receptors is
critically dependent on the basic arginine residue belonging
to the conserved motif H-X-X-R/K in transmembrane do-
main (TM) 6 of P2Y and CysLT1 and CysLT2 receptors.34
Our previous computational studies suggest that this also
holds true for Arg255 of GPR17.17 Here, to assess the actual
role of this residue in receptor binding, starting from the
previously published model of GPR17, we mutated this basic
amino acid to isoleucine and generated an in silico mutant
GPR17 receptor (R255I). Using steered molecular dynamics
simulations (SMD), we then simulated a forced unbinding of
the endogenous ligand UDP from both the wild type (WT)
and the R255I receptor models of GPR17.
Figure 3 shows a global view of the SMD trajectories
where the “extraction” of UDP from the WT and R255I
receptor models is compared. We found that the energy
required to unbind UDP from its pocket was higher for the
WT than for the mutated R255I receptor and the exit of the
ligand from its intracellular cavities occurred earlier in the
R255I model with respect to the WT receptor. The different
kinetics of UDP unbinding from the WT and the R255I

3496 Journal of Medicinal Chemistry, 2010, Vol. 53, No. 9
Calleri et al.
Figure 3. Forced unbinding profile of UDP. To elucidate the effect of a single point mutation of the arginine residue (Arg255) of GPR17,
potentially involved in nucleotide binding, simulations of the forced unbinding of UDP from the wild type (WT) and the mutant (R255I)
receptor models are compared. Panel a shows the pulling energy developed to unbind UDP from WT (in magenta) and from R255I (in black).
The detected significant difference in energy peak intensities suggests that the mutation actually affects the binding of UDP. The distances
between the couples of atoms involved in the formation of hydrogen/electrostatic bonds occurring between the ligand and either WT or
R255I receptors are reported in panels b and c. In the case of WT, for almost the whole duration of the simulation, residue Arg255 holds
the ligand near the pocket, thus allowing UDP to maintain interactions with other residues, among which are Tyr185, Tyr251, and Tyr262
(b). The same interactions are loosen up at the early stages of simulation in R255I receptor (c). In panel d, some representative frames of the
unbinding pathway of UDP (red sticks) from GPR17 (blue cartoon) are reported.
Table 5. FAC-MS Results for Library a Screened on GPR17-IAM and
mutated receptors are evident in our MPEG video where
visual representations of the two different SMD trajectories
for the WT and the mutated receptors (in purple and red,
respectively) are reported as a function of time. The video is
available in Supporting Information.
Mutated GPR17-IAM Columnsa
mutated
GPR17-IAM
(tm/tw ꢀ 100) -
compd
UDP
13
GPR17-IAM-II
100 b (%)
2.69
2.79
2.97
3.03
3.55
3.62
3.69
4.26
4.50
4.77
5.67
6.45
10.60
13.07
2.15
2.51
2.90
2.96
3.54
3.47
4.08
4.68
5.27
5.79
7.08
8.42
13.16
16.07
-20.1
-10.0
-2.4
-2.3
-0.3
-4.1
To experimentally validate the above in silico simulations
and to challenge the reliability of the FAC-MS analysis, we
generated and expressed the mutated receptor in 1321N1
cells and prepared a column from these cells (mutated
GPR17-IAM), in parallel with a column loaded with mem-
branes from cells expressing the WT receptor (GPR17-
IAM), containing the same number of cells (19.5 millions)/
column. An equimolar solution containing all the analytes of
library A (1 μM each) and the three reference compounds
(1 μM each) was then profiled by FAC-MS on both columns.
As far as the comparison of the elution order on the two
columns is concerned (Table 5), the same trend was observed
with the exception of the structurally similar compounds 12
and MRS 2179, whose breakthrough times were reversed in
mutated GPR17 column. Again, considering that the break-
through curves of the two analytes are very close, this appa-
rent change of elution order needs confirmation by using a
column with a higher binding site density. Table 5 also
reports the % variation of the breakthrough time determined
on mutated GPR17-IAM with respect to the GPR17-IAM-II
column. From these data, it is clear that the mutation is not
silent. The ligands can be divided into two groups depending
14
9
12
MRS 2179
5
11
þ10.6
þ10.1
þ17.1
þ21.4
þ24.9
þ30.5
þ24.1
þ23.0
7
8
6
10
cangrelor
4
a The breakthrough times are the result of a single ranking experiment
carried out on the freshly prepared column. Columns: GPR17-IAM-II
and mutated GPR17-IAM (19.5 million cells). b The breakthrough time
variation % was calculated on mutated GPR17-IAM (tm) with respect to
GPR17-IAM-II column (tw).
on the sign of the breakthrough time variation. It is specu-
lated that analytes with a positive variation most likely inter-
act with the receptor binding pocket through hydrophobic
interactions, while for compounds with a negative variation
electrostatic interactions are predominant. A variation % of

Article
Journal of Medicinal Chemistry, 2010, Vol. 53, No. 9 3497
less than (10% has been considered as nonsignificant, since
this change is linked to the experimental variability.
Di- and triphosphate derivatives show an opposite beha-
vior. The diphosphate compounds UDP and 13 show a lower
affinity (negative variation %) on the mutated GPR17-
column: for UDP this is in line with the predicted “in silico”
modeling results (Supporting Information video). Conver-
sely, triphosphate derivatives (5-8) show an increased affi-
nity (positive variation %). The exception is given by 10 and
11, diphosphate derivatives structurally characterized by the
presence of a lipophilic chain on the adenine moiety.
We also performed in silico docking studies using a virtual
library of the synthetic nucleotide derivative ligands tested in
vitro to allow a theoretical interpretation of results and to get
more information regarding the molecular interactions of
the most interesting ligands with GPR17. Two of the most
representative ligand candidates (MRS 2179 and 4, a bisphos-
phate and a triphosphate compound, respectively) were chosen,
and their docking poses are shown in Figure 4. As evident
from these docking results, while Arg255 is involved in GPR17
binding to MRS 2179, another residue (Arg87) seems to be
primarily involved in binding of the receptor to the phosphate
chain of 4(see arrows in Figure 4). These different docking poses
could account for the different effect of the Arg255 mutation on
the behavior of bisphosphate and triphosphate ligands on the
GPR17 column. Specifically, it can be hypothesized that,
different from bisphosphate derivatives like MRS 2179, for
triphosphate derivatives such as 4, mutation of Arg255 is not
enough to reduce the compound affinity for the column, as the
longer phosphate chain can interact with other arginine residues
(e.g., Arg87) through electrostatic interactions.
Figure 4. Binding pose hypothesis of nucleotide derivatives on
GPR17. The picture shows the two different possible binding
modalities hypothesized for bisphosphate and triphosphate com-
pounds on GPR17. Panels a and b show a macroscopic view of a
typical diphosphate compound (MRS 2179, panel a, spheres repre-
sentation) and of a triphosphate one (4, panel b, spheres repre-
sentation) docked to the receptor model (cartoon representation).
Panels c and d show the molecular surface of the pocket identified
for MRS 2179 and 4, respectively. Two possible target arginine
residues of ligand phosphate moieties are highlighted (yellow, stick
representation): Arg255 for MRS 2179 phosphates and Arg87 for 4
phosphates. Regions with hydrophobic, hydrophilic, and mild polar
features are reported in green, purple, and gray, respectively.
Phosphates of MRS 2179 and 4 involved in arginine binding are
reported in magenta and highlighted with arrows. It is evident that
different arginine residues (Arg255 and Arg87) participate in ligand
binding in the cases of MRS 2179 and 4, respectively (see text for
more detail).
Moreover, it could be hypothesized that triphosphate 4,
bearing a lipophilic phenylethynyl chain, can be better accom-
modated into the more lipophilic pocket of mutated GPR17,
thus leading to an increased breakthrough time. This effect is
less evident for diphosphate derivatives where the loss of
electrostatic interaction, due to the mutation, is preponder-
ant with respect to the increased lipophilicity.
mutated receptor (Table 5). Among all the tested com-
pounds, despite a significant decrease in the IC50 value to-
ward the mutated receptor, no significant shift in the reten-
tion time was observed for MRS 2179 (Table 5).
Functional Assay on Mutated Receptor. To further confirm
the data reported above, some analytes from library A (i.e., 4
and 13, demonstrating the highest and the lowest affinity
toward wild-type GPR17, respectively) together with the
three standard compounds (UDP, MRS 2179, and cangrelor)
were also tested in parallel on wild-type and mutated GPR17
in the [35S]GTPγS functional binding assay. All the tested
compounds maintained the same pharmacological profile
(agonist/antagonist) at both the wild-type (Figure 2a and
Figure 2b) and the mutated GPR17 receptor (Figure 2c and
Figure 2d).
No significant changes in receptor affinity (i.e., EC50) were
demonstrated for all tested agonists, but a trend toward a
reduced affinity was observed for UDP and compound 13,
whereas a trend toward an increased affinity was evident for
compound 4 (Table 5). In contrast, the antagonists cangrelor
and MRS 2179 showed a significant increase in affinity
toward the mutated vs the wild type receptor (P < 0.05; see
Table 4).
Conclusions
In this work, we have developed a new strategy based on the
FAC-MS method coupled to in silico docking studies for
deeper structural information on the GPR17 receptor and for
the screening of new and potent agonist and antagonist
ligands acting on this recently deorphanized GPCR that plays
a crucial role in the development of ischemic brain damage
and in the later remodeling/repair processes.
The use of FAC-MS as a reliable method for screening
compound libraries toward important pharmaceutical targets
such as GPCRs is still a challenge; however, in this study, we
have demonstrated the applicability of this technique to the
screening of drug candidates toward GPR17.
Unlike other affinity-based screening strategies, which are
not specific enough for chemical identification and/or suffer
from low throughput, FAC-MS methodology is highly
specific and can be used for the discovery of classes of small
molecule ligands (agonists and antagonists) in a single fast
screening campaign. The strength of the FAC-MS appro-
ach has been validated using a reference functional assay
([35S]GTPγS assay), and the chromatographic data allowed
the drawing of preliminary structure-activity relationships.
However, the design and synthesis of selective GPR17 agonist/
antagonist ligands would greatly benefit from the knowledge
Data from [35S]GTPγS functional experiments paralleled
those obtained in affinity chromatography studies. In fact, a
decrease in the breakthrough time was observed for those
compounds showing a trend toward a reduced affinity for the
mutated receptor (e.g., UDP and 13; Table 5), whereas
a higher breakthrough time was observed for cangrelor
and compound 4, showing an increased affinity toward the

3498 Journal of Medicinal Chemistry, 2010, Vol. 53, No. 9
Calleri et al.
of receptor 3D structure and from the definition of its ligand
binding mode. In the case of GPCRs, modeling crucially
depends on biological validation that is, in turn, typically
supported by site-directed mutagenesis of the binding site.
Moreover, a further refinement of the 3D model also came
from in silico mutagenesis of the predicted key amino acid
residues in or near the target binding site.
To the best of our knowledge, this is the first time that
FAC-MS in combination with molecular modeling simula-
tions is applied to the study of membrane proteins. The
integration between analytical, biochemical/pharmacologi-
cal, and computational assays will drive a progressive reduc-
tion of the number of potential candidate ligands by selecting
only the most promising lead compounds. Moreover, the data
obtained will address the design and the synthesis of new
refined compounds acting as new potent and selective ligands.
27, in turn dissolved in 1 mL of dry DMF, an amount of 36 μL of
tri-n-butylamine (28 mg, 0.15 mmol) was added. The solution
was stirred for 20 min at room temperature and then evaporated
to dryness under anhydrous conditions. After resuspension in
1.4 mL of dry DMF, N,N0-carbonyldiimidazole (122 mg, 0.75
mmol) was added and the mixture was stirred for 3 h at room
temperature. Methanol (49 μL, 1.2 mmol) was added and the
mixture stirred for 30 min at room temperature. Then a total of
6 mL (3 mmol) of a 0.5 M solution of bis(tri-n-butylammonium)
pyrophosphate or tri-n-butylammonium phosphate in DMF
was added (for the synthesis of the triphosphate or the dipho-
sphate derivatives, respectively). The mixture was stirred for
14 h at room temperature. The solvent was removed in vacuo.
The mixture, dissolved in H2O, was purified by means of ion-
exchange chromatography.
2-Chloro-N6-methyladenosine Triphosphate (7). Reaction of
24 furnished 7 as white solid; yield 37%. 1H NMR (D2O) δ 2.88
(s, 3H, CH3), 4.08 (m, 2H, CH2-50), 4.21 (m, 1H, H-40), 4.40 (m,
1H, H-30), 4.58 (m, 1H, H-20), 5.85 (d, 1H, J=6.6 Hz, H-10), 8.25
(s, 1H, H-8). 31P NMR (D2O) δ -21.50 (m), -9.23 (m). Anal.
Calcd for C11H29ClN9O13P3: C, 21.18; H, 4.69; N, 20.21. Found:
Experimental Section
Synthesis. 1H NMR spectra were obtained with Varian
Mercury 400 MHz spectrometer; δ is in ppm, J in Hz. All ex-
changeable protons were confirmed by addition of D2O. 31P
NMR spectra were recorded at room temperature using a
Varian Mercury 400 MHz spectrometer. Elemental analyses
were determined on a Fisons model EA 1108 analyzer and are
within (0.4% of theoretical values. Purity of the compounds
was g95% according to elemental analysis data. TLCs were
carried out on precoated TLC plates with silica gel 60 F-254
(Merck). For ion exchange chromatography, Sephadex DEAE
A-25 (Fluka) was used. Mass analyses were performed on a
quadrupole ESI mass apparatus (HP 1100 MSD).
C, 21.03; H, 4.46; N, 20.46. ESI-MS m/z: 276.5 [M - 2H]2-
554.0 [M -H]-.
,
2-Chloro-3-deaza-N6-methyladenosine Triphosphate (8). Re-
1
action of 25 furnished 8 as white solid; yield 69%. H NMR
(D2O) δ 2.86 (s, 3H, CH3), 4.09 (m, 2H, CH2-50), 4.22 (m, 1H, H-
40), 4.41 (m, 1H, H-30), 4.51 (m, 1H, H-20), 5.77 (d, 1H, J =6.6
Hz, H-10), 6.87 (s, 1H, H-3), 8.21 (s, 1H, H-8). 31P NMR (D2O) δ
-21.58 (t), -10.34 (d), -7.47 (d). Anal. Calcd for C12H30-
ClN8O13P3: C, 23.14; H, 4.86; N, 17.99. Found: C, 22.82; H,
4.51; N, 18.36. ESI-MS m/z: 276.0 [M-2H]-2, 552.8 [M-H]-.
5-Hexynyluridine Diphosphate (14). Reaction of 27 furnished
14 as white solid; 60% yield. 1H NMR (D2O) δ 0.74 (t, 3H, J=
7.3 Hz, CH3), 1.26 (m, 2H, CH2CH3), 1.39 (m, 2H,
CH2CH2CH3), 2.25 (t, 2H, J = 6.9 Hz, CH2CtC), 4.08 (m,
3H, H-40 and CH2-50), 4.22 (m, 2H, H-20 and H-30), 5.77 (d, 1H,
J=4.5 Hz, H-10), 7.85 (s, 1H, H-6). 31P NMR (D2O) δ -10.49
(d), -8.21 (d). Anal. Calcd for C15H31N5O12P2: C, 33.65; H,
5.84; N, 13.08. Found: C, 33.37; H, 5.65; N, 13.43. ESI-MS m/z
483.0 [M -H]-.
General Procedure for the Synthesis of the Nucleoside 50-Mono-
phosphates 24, 25, and 27. To 0.56 mmol of the nucleosides 22,28
23,29 and 26,30 in turn dissolved in 3.0 mL of trimethylphos-
phate, an amount of 4 equiv of POCl3 (209 μL, 2.24 mmol) was
added. The solution was stirred at room temperature for 3 h.
H2O (3 mL) was added, and the solution was neutralized by
adding triethylamine dropwise. The mixture was purified by ion
-
exchange chromatography on Sephadex DEAE A-25 (HCO3
Preparation of Bis(tri-n-buthylammonium) Pyrophosphate.
Sodium pyrophosphate decahydrate (3.34 g, 7.5 mmol) was
dissolved in 75 mL of water and put under stirring. Dowex
50x8, 20-50 mesh, Hþ form (21 g), was added, and the suspen-
sion was stirred 20 min. A solution of ethanol (30 mL) and tri-n-
buthylamine (3.57 mL, 15 mmol) was put in an ice bath, and the
pyrophosphate solution was filtered, leaving the filtrate to go
directly into the solution. The resin on the filter was washed with
water until the filtrate showed a pH of 7. The solvent was
removed under vacuum (temperature below 35 ꢀC) and coeva-
porated three times with ethanol and three times with dry
DMF. The residue was put in 15 mL of DMF in order to
obtain a bis(tri-n-buthylammonium) pyrophosphate solution,
0.5 M in DMF. The solution was stored over molecular sieves
at 4 ꢀC.
Preparation of Tri-n-buthylammonium Monophosphate. So-
dium phosphate monobasic monohydrate (1.38 g, 10 mmol) was
dissolved in 70 mL of water and put under stirring. Dowex 50x8,
20-50 mesh, Hþ form (21 g), was added, and the suspension was
stirred 20 min. A solution of ethanol (50 mL) and tri-n-buthyl-
amine (2.40 mL, 10 mmol) was put in an ice bath and the
orthophosphate solution was filtered, leaving the filtrate to go
directly into the solution. The resin on the filter was washed with
water until the filtrate showed a pH of 7. The solution was stirred
for 30 min in an ice bath. The solvent was removed under
vacuum (temperature below 35 ꢀC) and coevaporated three
times with ethanol and three times with dry DMF. The residue
was put in 20 mL of dry DMF in order to obtain a tri-n-
buthylammonium monophosphate solution, 0.5 M in DMF.
The solution was stored over molecular sieves at 4 ꢀC.
form) equilibrated with H2O and eluted with a linear gradient of
H2O/NH4HCO3 0.5 M, after washing the resin with H2O.
2-Chloro-N6-methyladenosine Monophosphate (24). Reaction
of 22 furnished 24 as white solid; yield 68%. 1H NMR (D2O) δ
2.77 (s, 3H, CH3), 3.89 (m, 2H, CH2-50), 4.16 (m, 1H, H-40), 4.28
(d, 1H, J=4.4 Hz, H-30), 4.50 (d, 1H, J=5.2 Hz, H-20), 5.76 (d,
1H, J=5.6 Hz, H-10), 8.16 (s, 1H, H-8). 31P NMR (D2O) δ 2.59
(s). Anal. Calcd for C11H18ClN6O7P: C, 32.01; H, 4.40; N, 20.36.
Found: C, 31.79; H, 4.37; N, 20.65. ESI-MS m/z: 394.0 [M -
H]-, 789.0 [2M -H]-.
2-Chloro-3-deaza-N6-methyladenosine Monophosphate (25).
Reaction of 23 furnished 25 as white solid; yield 88%. 1H
NMR (D2O) δ 2.85 (s, 3H, CH3), 3.84 (m, 2H, CH2-50), 4.19
(m, 1H, H-40), 4.25 (m, 1H, H-30), 4.53 (m, 1H, H-20), 5.76 (d,
1H, J = 6.4 Hz, H-10), 6.82 (s, 1H, H-3), 8.29 (s, 1H, H-8). 31P
NMR (D2O) δ 0.87 (s). Anal. Calcd for C12H19ClN5O7P: C,
35.01; H, 4.65; N, 17.01. Found: C, 34.86; H, 4.43; N, 17.39. ESI-
MS m/z: 393.0 [M -H]-, 786.9 [2M -H]-.
5-Hexynyluridine Monophosphate (27). Reaction of 26 furn-
ished 27 as white solid; 45% yield. 1H NMR (D2O) δ 0.75 (t, 3H,
J = 7.2 Hz, CH3), 1.27 (m, 2H, CH2CH3), 1.43 (m, 2H,
CH2CH2CH2), 2.27 (t, 2H, J = 6.9 Hz, CH2CꢁC), 3.94 (m,
2H, CH2-50), 4.15 (m, 1H, H-40), 4.20 (m, 2H, H-20 e H-30), 5.70
(d, 1H, J= 5.1 Hz, H-10), 7.86 (s, 1H, H-6). 31P NMR (D2O) δ
1.75 (s). Anal. Calcd for C15H24N3O9P: C, 42.76; H, 5.74; N,
9.97. Found: C, 42.62; H, 5.65; N, 10.21. ESI-MS m/z 201.1 [M-
2H]2-, 403.0 [M- H]-, 807.0 [2M- H]-.
General Procedure for the Synthesis of the Nucleoside 50-
Triphosphates (7 and 8) and Nucleoside 50-Diphosphates (14).
To 0.15 mmol of the nucleoside 50-monophosphates 24, 25, and

Article
Chromatographic Procedures for the Monitoring and the Puri-
fication of the Phosphorylation Reactions. The reactions were
monitored by TLC, using precoated TLC plates with silica gel
60 F-254 (Merck) and isopropanol-H2O-NH4OH (30%)
(5.5:1:3.5) as mobile phase. The nucleotides were purified by
means of ion-exchange chromatography on a Sephadex DEAE
A-25 (Fluka) column (HCO3- form) equilibrated with H2O and
eluted with a linear gradient of H2O/0.5 M NH4HCO3.
Preparation of the DEAE-Sephadex Resin. Swelling of the
Resin. An amount of 10 g of the resin was put into a beaker with
400-500 mL of deionized water (R>10 MΩ) for 24 h at 4 ꢀC.
After that the resin was swelled with 500 mL of NaHCO3, 1 M.
The resin was kept for 3 days at 4 ꢀC, changing the supernatant
buffer solution every day.
Preparation of Chromatographic Columns. Cell Culture and
Transfection. Human astrocytoma cells (1321N1) were cultured
as described previously.13 Cells were seeded on 100 mm Petri
dishes (106 cells/dish) and transfected with pcDNA3.1 vector
containing the construct encoding for either the human wild
type or the R255I mutated GPR17 receptor (pcDNA3.1-
wtGPR17h or pcDNA3.1-muGPR17h) with FuGENE 6
(Roche Diagnostics, Monza, Italy), according to the manufac-
turer’s instructions. Control cultures received in parallel the
corresponding pcDNA3.1 empty vector. In selected experi-
ments, transfection efficiency was checked under a fluorescence
microscope equipped with a fluorescein filter after cotransfec-
tion of cultures with pcDNA3.1-GPR17h and a plasmid encod-
ing for the enhanced green fluorescent protein (eGFP).
Site Directed Mutagenesis of hGPR17. The previously con-
structed pcDNA3.1-hGPR17 plasmid14 was mutated using
the QuikChange II site-directed mutagenesis kit (Stratagene,
La Jolla, CA). The arginine at position 255, in the DNA binding
domain of GPR17, was mutated to a isoleucine, by amplifica-
tion of a fragment from cloned pcDNA3.1-hGPR17 plasmid
using polymerase chain reactions (PCR) primers (50-gccctac-
cacgtcaacatctccgtctacgtgctg-30 and 50-cagcacgtagacggagatgtt-
gacgtggtagggc-30) according to the kit manufacture’s direc-
tions. The mutated plasmids were then chemically transformed
into the E. coli R-DH5 strain. The cells were grown overnight in
5 mL of Luria-Bertani (LB) medium containing ampicillin,
50 μg/mL. The overnight culture was used to prepare a miniprep
by the method described in the Qiagen manual. The presence of
the correct mutation and the absence of PCR-derived altera-
tions to the coding sequence were confirmed by complete
sequencing of the receptor construct.
Membrane Preparation. For membrane preparation, the
method developed for P2Y1 receptors was followed.33 Two full
days (48 h) after transfection, 2 ꢀ 107 1321N1 cells transiently
transfected with either pcDNA3.1-wtGPR17h or pcDNA3.1-
muGPR17h were homogenized for 3 ꢀ 30 s at a setting of 11 on a
model PT-2100 homogenizer (Kinematica, Luzern, Switzerland)
in 10 mL of Hepes buffer (20 mM, pH 8.0) containing 500 mM
NaCl, 5 mM 2-mercaptoethanol, 100 μM benzamidine, 10 μg/
mL aprotinin, 10 μg/mL leupeptin, 50 μg/mL tosylamido-
2-phenylethyl chloromethyl ketone (TPCK), 100 μg/mL phenyl-
methylsulfonyl fluoride (PMSF), and 100 μM adenosine triphos-
phate (ATP). The homogenate was centrifuged at 700g for
5 min, the pellet was discarded, and the supernatant was centri-
fuged at 100000g for 30 min at 4 ꢀC. The resulting pellet was
suspended in 10 mL of Hepes buffer (20 mM, pH 8.0) containing
2% (w/v) n-octyl-β-D-glucopyranoside, 500 mM NaCl, 5 mM
2-mercaptoethanol, 100 μM benzamidine, 10 μg/mL aprotinin,
10 μg/mL leupeptin, 50 μg/mL TPCK, 100 μg/mL PMSF, 100 μM
ATP, and 10% glycerol. The resulting mixture was rotated at
150 rpm using an Orbit shaker for 18 h at 4 ꢀC and was centri-
fuged at 100000g for 25 min. The supernatant was collected and
mixed with 70 mg of IAM particles and gently rotated at room
temperature for 1 h at 150 rpm. The suspended particles were
then dialyzed for 2 days against Hepes buffer (20 mM, pH 8.0)
containing 500 mM NaCl, 1 mM ethylenediaminetetraacetic
Journal of Medicinal Chemistry, 2010, Vol. 53, No. 9 3499
acid (EDTA), and 0.01% sodium azide, changing the buffer
every 24 h. The suspension was then centrifuged at 700 rpm for
3 min (4 ꢀC), and the supernatant was discarded. The pellet was
washed with 10 mL of chromatographic running buffer (5 mM
ammonium acetate, pH 7.4) and the supernatant removed by
centrifugation at 700 rpm for 3 min (4 ꢀC).
The pellet was resuspended in 1 mL of chromatographic buf-
fer and packed into an Omnifit glass column to yield a 3 mm ꢀ
6.6 mm i.d. chromatographic bed.
FAC-MS Screening. The immobilization of wild-type and
mutated GPR17 to IAM particles was carried out as previously
described.13,33
The frontal affinity chromatography experiments were car-
ried out on a chromatographic system consisting of a thermo-
stated column oven Surveyor autosampler controlled at 20 ꢀC
(Thermo Finnigan, San Jose, CA), a quaternary gradient
Surveyor MS pump (Thermo Finnigan, San Jose, CA) equipped
with a diode array detector (DAD), and a LTQ linear ion trap
mass spectrometer (MS) with electrospray ionization (ESI) ion
source controlled by Xcalibur software 1.4 (Thermo Finnigan,
San Jose, CA).
The mobile phase was composed of 90% ammonium acetate
(5 mM, pH 7.4) containing 10% methanol, and the experiments
were carried out at a flow rate of 500 μL/min at ambient
temperature. A postcolumn mobile-phase splitting valve was
also inserted so that only 200 μL/min were diverted to MS. Mass
spectra were generated in negative ion mode under constant
instrumental conditions: source voltage 4.0 kV, capillary vol-
tage 46 V, sheath gas flow 38 (arbitrary units), auxiliary gas flow
8 (arbitrary units), sweep gas flow 1 (arbitrary units), capillary
temperature 270 ꢀC, tube lens voltage -85.06 V.
For the ranking experiment, MS conditions were optimized
so that all the compounds were detected simultaneously as
negative [M -H]- ions.
Stock solutions of each analyte were prepared in water (the
concentrations were from 0.2 to 0.6 mM depending on the
available amount of each compound). The stock solutions were
stored at -20 ꢀC before the preparation of the screening
solution. Each screening solution was prepared to contain the
library members each at 1 μM in 5 mM ammonium acetate,
pH 7.4, containing 10% methanol. The mixture also contained
cangrelor, MRS 2179, and UDP (1 μM) as affinity reference
compounds.
Each extracted ion breakthrough curve was analyzed with a
polynomial equation of degree 3 (y=ax3 þ bx2 þ cx þ d) to fit
the chromatographic data, and the inflection point, correspond-
ing to the breakthrough time, was determined by the second
derivative.
GPR17 Functional Assay. In order to investigate the pharma-
cological profile and the potency of the selected compounds on
purinergic GPR17 binding site, GTPγS binding assay has been
performed on 1321N1 cells transiently transfected with either
the wild type or the mutated human protein (see above).
The pharmacological profile of the new ligands (agonists or
antagonists) toward the GPR17 receptor was evaluated by
assessing the effect of different ligand concentrations to mod-
ulate GPR17 receptor-G protein coupling. In particular, the
effect of the compounds alone (acting as agonist) and in the
presence of the purinergic agonist UDP-glucose (behaving as an
antagonist) in modulating GTPγS binding was evaluated.
In parallel, in order to investigate if compound-mediated
effects were really ascribed to the interaction with GPR17
binding site, GTPγS binding was also performed in 1321N1
cells transfected with the empty vector (data not shown).
Because GPCR stimulation by agonists results in increased
binding of GTP to G-proteins (which can, in turn, be quantified
by measuring [35S]GTPγS binding to purified membranes),35
activation of the hGPR17 in transiently transfected 1321N1 cells
was determined by testing the increase of [35S]GTPγS binding by
exogenously added compounds.35-37 1321N1 cells (control and

3500 Journal of Medicinal Chemistry, 2010, Vol. 53, No. 9
Calleri et al.
transfected cells) were homogenized in 5 mM Tris-HCl, 2 mM
EDTA, pH 7.4, and centrifuged at 48000g for 15 min at 4 ꢀC.
The resulting pellets were washed in 50 mM Tris-HCl, 10 mM
MgCl2, pH 7.4, and stored at 80 ꢀC until use.
and ligands) components of the system were separately coupled
to a temperature bath at 310 K, with a coupling constant τt of
0.1 ps. The pressure coupling was set as independent in the x, y,
and z directions (semiisotropic coupling), with a constant
pressure of 1 bar and a coupling constant τp of 1 ps. A 2 fs time
step was used for the integration of the equations of motions,
and all bond distances involving hydrogen atoms were con-
strained using LINCS. Configurations were saved for every 1 ps
for analysis.
Optimal [35S]GTPγS binding conditions were determined in
preliminary experiments.14 Briefly, aliquots of cell membranes
(20 μg) were incubated with GDP (3 μM), [35S]GTPγS (0.15 nM,
1,250 Ci/mmol, Perkin-Elmer) and different compound concen-
trations (10 nM to 10 μM). After incubation at room tempera-
ture in a shaking water bath for 30 min, cells were harvested by
rapid filtration and assayed for 35S radioactivity. Nonspecific
binding of [35S]GTPγS was measured with 10 μM GTPγS.
Statistical Analysis. For the analysis and graphic presentation
of [35S]GTPγS binding data, a nonlinear multipurpose curve
fitting computer program Graph-Pad Prism (GraphPad) was
used. All data are presented as the mean ( SEM of three
different experiments. Data were tested for statistical signifi-
cance with the paired Student’s t test or by analysis of variance
(one-way ANOVA), as appropriate. When significant differ-
ences were observed, the Newman-Keuls multiple comparison
test (one-way ANOVA) was done. A value of P < 0.05 was
considered significant.
Acknowledgment. Cangrelor was a kind gift from The
Medicine Company (Parsippany, NJ). This work was sup-
ported by a grant from the Italian Ministry for University and
Research (Project FIRB RBN503YA3L).
Supporting Information Available: MPEG video showing the
different kinetics of UDP unbinding from the WT and the R255I
mutated receptors, where a visual representation of the two
different SMD trajectories for the WT and the mutated recep-
tors (in purple and red, respectively) are reported as a function of
time. This material is available free of charge via the Internet at
http://pubs.acs.org.
Molecular Modeling. Steered Molecular Dynamics Simula-
tion. A previously published rhodopsin-based homology model
of the human GPR17 receptor bound to its endogenous ligand
UDP and refined by means of molecular dynamics (MD)
simulations17 was used as a starting point for docking and
SMD studies (see also the Computational Details section). In
parallel, the Arg255 residue of the same receptor model was
mutated to Ile in order to investigate its role in ligand recogni-
tion, according to in vitro studies. The effect of this mutation on
ligand binding was then studied by simulating a forced unbind-
ing of the docked UDP from both the WT and the R255I
receptor models using SMD simulations. The ligand unbinding
was induced by applying an external force to the center of mass
(COM) of the ligand that simulates a retracting cantilever
directed along a selected pathway. Because of the action-reac-
tion principle, the spring acts as a sensor of all interactions
occurring between the protein and its ligand. The elastic force is
proportional to the spring elongation relative to its equilibrium
References
(1) Schriemer, D. C.; Bundle, D. R; Li, L.; Hindsgaul, O. Micro-scale
frontal affinity chromatography with mass spectrometric detec-
tion: a new method for the screening of compound libraries. Angew.
Chem. 1998, 110, 3625–3628; Angew. Chem., Int. Ed. 1998, 37,
3383-3387.
(2) Ng, E.; Schriemer, D. Emerging challenges in ligand discovery: new
opportunities for chromatographic assay. Expert Rev. Proteomics
2005, 2, 891–900.
(3) Calleri, E.; Temporini, C.; Caccialanza, G.; Massolini, G. Target-
based drug discovery: the emerging success of frontal affinity
chromatography coupled to mass spectrometry. ChemMedChem
2009, 4, 905–916.
(4) Slon-Usakiewicz, J. J.; Dai, J. R.; Ng, W.; Foster, J. E.; Deretey, E.;
Toledo-Sherman, L.; Redden, P. R.; Pasternak, A.; Reid, N.
Kinase screening. Applications of frontal affinity chromatography
coupled to mass spectrometry in drug discovery. Anal. Chem. 2005,
77, 1268–1274.
(5) Ng, E. S.; Chan, N. W.; Lewis, D. F.; Hindsgaul, O.; Schriemer,
D. C. Frontal affinity chromatography-mass spectrometry. Nat.
Protoc. 2007, 2, 1907–1917.
(6) Slon-Usakiewicz, J. J.; Ng, W.; Dai, J.-R.; Pasternak, A.; Redden,
P. R. Frontal affinity chromatography with MS detection
(FAC-MS) in drug discovery. Drug Discovery Today 2005, 10,
409–416.
(7) Toledo-Sherman, L.; Deretey, E.; Slon-Usakiewicz, J. J.; Ng, W.;
Dai, J. R.; Foster, J. E.; Redden, P. R.; Uger, M. D.; Liao, L. C.;
Pasternak, A.; Reid, N. L. Frontal affinity chromatography with
MS detection of EphB2 tyrosine kinase receptor. 2. Identification
of small-molecule inhibitors via coupling with virtual screening.
J. Med. Chem. 2005, 48, 3221–3230.
(8) Ng, W.; Dai, J. R.; Slon-Usakiewicz, J. J.; Redden, P. R.; Pasternak,
A.; Reid, N. Automated multiple ligand screening by frontal
affinity chromatography-mass spectromatry (FAC-MS). J. Bio-
mol. Screening 2007, 12, 167–174.
(9) Jaakola, V. P.; Griffith, M. T.; Hanson, M. A.; Cherezov, V.;
Chien, E. Y.; Lane, J. R.; IJzerman, A. P.; Stevens, R. C. The 2.6
angstrom crystal structure of a human A2A adenosine receptor
bound to an antagonist. Science 2008, 322, 1211–1217.
(10) Warne, T.; Serrano-Vega, M. J.; Baker, J. G.; Moukhametzianov,
R.; Edwards, P. C.; Henderson, R.; Leslie, A. G.; Tate, C. G.;
Schertler, G. F. Structure of a beta1-adrenergic G-protein-coupled
receptor. Nature 2008, 454, 486–491.
(11) Pidgeon, C.; Marcus, C.; Alvarez, F. In Applications of Enzymes
Biotechnology; Baldwin, T. O., Kelly, J. W., Eds.; Plenum Press: New
York, 1992; pp 201-237.
(12) Moaddel, R.; Wainer, I. W. Development of immobilized mem-
brane-based affinity columns for use in the online characterization
of membrane bound proteins and for targeted affinity isolations.
Anal. Chim. Acta 2006, 564, 97–105.
position, and it is given by the expression F =k(v t-x ), where v t
B
B
B
B
-x is the displacement of the restrained atom with respect to its
B
original position, t is time elapsed from the beginning of the
simulation; v = 0.004 nm/ps corresponds to the velocity of the
retracting cantilever, and k = 2000 kJ/(mol nm2) is its force
B
constant. The combination for v and k for our simulations was
B
chosen among different values tested in a previous series of
simulations (v = 0.001-0.01 nm/ps; k = 2000-5000 kJ/(mol
B
nm2)) so that the complete unbinding of the ligands occurs
within 1 ns for each SMD simulation. For the interpretation of
the results the registered pulling force was converted into energy
values by applying the following expression: E= F (v t-x ).
B B
B
Docking studies of some representative ligands of the avail-
able chemical library of synthetic nucleotide derivatives were
also performed on both the WT and the R255I receptor models
in order to provide a theoretical interpretation of the results
obtained by the in vitro screening. These docking studies were
performed using the semiflexible docking module provided by
DELOS package (http://www.delos-bio.it).
Computational Details. All the simulations were run on a
Linux cluster Blade with Xeon processors. All the simulations
and concerning analysis were carried out using the Gromacs 3.3
package,17 using the Gromacs force field, modified by all the
parameters necessary for the description of each component
(protein, lipids, solvent, and ligands) and their reciprocal inter-
actions, based on manufacturer’s instructions. Periodic bound-
ary conditions were applied in all three x, y, and z dimensions.
The isothermal isobaric NPT ensemble (constant number of par-
ticles, pressure, and temperature) was applied. Solvent (water
molecules and chlorine ions) and nonsolvent (lipids, protein,
(13) Temporini, C.; Ceruti, S.; Calleri, E.; Ferrario, S.; Moaddel, R.;
Abbracchio, M. P.; Massolini, G. Development of an immobilized
GPR17 receptor stationary phase for binding determination using

Article
frontal affinity chromatography coupled to mass spectrometry.
Journal of Medicinal Chemistry, 2010, Vol. 53, No. 9 3501
relationships of bisphosphate nucleotide derivatives as P2Y1 recep-
tor antagonists and partial agonists. J. Med. Chem. 1999, 42, 1625–
1638.
Anal. Biochem. 2009, 384, 123–129.
(14) Ciana, P.; Fumagalli, M; Trincavelli, M. L.; Verderio, C.; Rosa, P.;
Lecca, D.; Ferrario, S.; Parravicini, C.; Capra, V.; Gelosa, P.;
Guerrini, U.; Belcredito, S.; Cimino, M.; Sironi, L.; Tremoli, E.;
Rovati, G. E.; Martini, C.; Abbracchio, M. P. The orphan receptor
GPR17 identified as a new dual uracil nucleotides/cysteinyl-leuko-
trienes receptor. EMBO J. 2006, 25, 4615–4627.
(15) Lecca, D.; Trincavelli, M. L.; Gelosa, P.; Sironi, L.; Ciana, P.;
Fumagalli, M.; Villa, G.; Verderio, C.; Grumelli, C.; Guerrini, U.;
Tremoli, E.; Rosa, P.; Cuboni, S.; Martini, C.; Buffo, A.; Cimino,
M.; Abbracchio, M. P. The recently identified P2Y-like receptor
GPR17 is a sensor of brain damage and a new target for brain
repair. PLoS One 2008, 3, e3579.
(16) Mellor, E. A.; Maekawa, A.; Austen, K. F.; Boyce, J. A. Cysteinyl
leukotriene receptor 1 is also a pyrimidinergic receptor and is
expressed by human mastcells. Proc. Natl. Acad. Sci. U.S.A.
2001, 98, 7964–7969.
(17) Parravicini, C.; Ranghino, G.; Abbracchio, M. P.; Fantucci, P.
GPR17: molecular modeling and dynamics studies of the 3-D
structure and purinergic ligand binding features in comparison
with P2Y receptors. BMC Bioinf. 2008, 9, 263.
(25) Michelson, A. M.; Dondon, J.; Grunberg-Manago, M. Action of
polynucleotide phosphorylase on 5-halogenouridine 50-pyrophos-
phates. Biochim. Biophys. Acta 1962, 55, 529–540.
ꢁ
(26) Besada, P.; Shin, D. H.; Costanzi, S.; Ko, H.; Mathe, C.; Gagner-
on, J.; Gosselin, G.; Maddileti, S.; Harden, T. K.; Jacobson, K. A.
Structure-activity relationships of uridine 50-diphosphate analo-
gues at the human P2Y6 receptor. J. Med. Chem. 2006, 49, 5532–
5543.
(27) Volpini, R.; Mishra, R. C.; Kachare, D. D.; Dal Ben, D.; Lambertucci,
C.; Antonini, I.; Vittori, S.; Marucci, G.; Sokolova, E.; Nistri, A.;
Cristalli, G. Adenine based acyclic-nucleotides as novel P2X3
receptor ligands. J. Med. Chem. 2009, 52, 4596–4603.
(28) Venkata Palle, E. E.; Wu, Y.; Dewan Zeng, T. M.; Zablocki, J.
2-Pyrazolyl-N6-Substituted adenosine derivatives as high affinity
and selective adenosine A3 receptor agonists. J. Med. Chem. 2004,
47, 4766–4773.
(29) May, J. A., Jr.; Townsend, L. B. General synthesis of 4-substituted
1-(β-D-ribofuranosyl)imidazo[4,5-c]pyridines. J. Chem. Soc., Per-
kin Trans. 1 1975, 2, 125–129.
(18) Cristalli, G.; Podda, G. M.; Costanzi, S.; Lambertucci, C.; Lecchi,
A.; Vittori, S.; Volpini, R.; Zighetti, M. L.; Cattaneo, M. Effects
of 50-phosphate derivatives of 2-hexynyl adenosine and 2-phenyl-
ethynyl adenosine on responses of human platelets mediated by
P2Y receptors. J. Med. Chem. 2005, 48, 2763–2766.
(19) Fischer, B.; Boyer, J. L.; Hoyle, C. H.; Ziganshin, A. U.; Brizzolara,
A. L.; Knight, G. E.; Zimmet, J.; Burnstock, G.; Harden, T. K.;
Jacobson, K. A. Identification of potent, selective P2Y-purinocep-
tor agonists: structure-activity relationships for 2-thioether deri-
vatives of adenosine 50-triphosphate. J. Med. Chem. 1993, 36,
3937–3946.
(20) Burnstock, G.; Fischer, B.; Hoyle, C. H. V.; Maillard, M.; Ziganshin,
A. U.; Brizzolara, A. L.; von Isakovics, A.; Boyer, J. L.; Harden,
K.; Jacobson, K. A. Structure activity relationships for derivatives
of adenosine-50-triphosphate as agonists at P2 purinoceptors:
heterogeneity within P2X and P2Y subtypes. Drug Dev. Res.
1994, 31, 206–219.
(21) Tarun, M. K.; Timothy, J. M. Allele-specific activators and
inhibitors for kinesin. Proc. Natl. Acad. Sci. U.S.A. 1999, 96,
9106–9111.
(22) Gough, G.; Maguire, M. H.; Penglis, F. Analogues of adenosine
50-diphosphate-new platelet aggregators. Influence of purine ring
and phosphate chain substitutions on the platelet-aggregating potency
of adenosine 50-diphosphate. Mol. Pharmacol. 1972, 8, 170–177.
(23) Camaioni, E.; Boyer, J. L.; Mohanram, A.; Harden, T. K.;
Jacobson, K. A. Deoxyadenosine bisphosphate derivatives as
potent antagonists at P2Y1 receptors. J. Med. Chem. 1998, 41,
183–190.
(30) Ghilagaber, S.; Hunter, W. N.; Marquez, R. Efficient coupling of
low boiling point alkynes and 5-iodo-nucleosides. Tetrahedron
Lett. 2006, 48, 483–486.
(31) Yoshikawa, M.; Kato, T.; Takenishi, T. A novel method for
phosphorylation of nucleosides to 50-nucleotidase. Tetrahedron
Lett. 1967, 50, 5065–5068.
(32) Hoard, D. E.; Ott, D. G. J. Conversion of mono- and oligodeoxy-
ribonucleotides to 50-triphosphates. J. Am. Chem. Soc. 1965, 87,
1785–1788.
(33) Moaddel, R.; Calleri, E.; Massolini, G.; Frazier, C. R.; Wainer,
I. W. The synthesis and initial characterization of an immobilized
purinergic receptor (P2Y1) liquid chromatography stationary
phase for online screening. Anal. Biochem. 2007, 364, 216–218.
(34) Abbracchio, M. P.; Burnstock, G.; Boeynaems, J. M.; Barnard,
E. A.; Boyer, J. L.; Kennedy, C.; Knight, G. E.; Fumagalli, M.;
Gachet, C.; Jacobson, K. A.; Weisman, G. A. International Union
of Pharmacology LVIII: Update on the P2Y G protein-coupled
nucleotide receptors: from molecular mechanisms and pathophy-
siology to therapy. Pharmacol Rev. 2006, 58, 281–341.
(35) Marteau, F.; Le Poul, E.; Communi, D.; Labouret, C.; Savi, P.;
Boeynaems, J. M.; Gonzalez, N. S. Pharmacological characteriza-
tion of the human P2Y13 receptor. Mol. Pharmacol. 2003, 64 (1),
104–112.
ꢁ
(36) Kotani, M.; Mollereau, C.; Detheux, M.; Le Poul, E.; Brezillon, S.;
Vakili, J.; Mazarguil, H.; Vassart, G.; Zajac, J. M.; Parmentier, M.
Functional characterization of a human receptor for neuropeptide
FF and related peptides. Br. J. Pharmacol. 2001, 133, 138–144.
(37) Fumagalli, M.; Trincavelli, L.; Lecca, D.; Martini, C.; Ciana, P.;
Abbracchio, M. P. Cloning, pharmacological characterisation and
distribution of the rat G-protein-coupled P2Y(13) receptor. Bio-
chem. Pharmacol. 2004, 68 (1), 113–124.
(24) Nandanan, E.; Camaioni, E.; Jang, S. Y.; Kim, Y. C.; Cristalli, G.;
Herdewijn, P.; Secrist, J. A., 3rd; Tiwari, K. N.; Mohanram, A.;
Harden, T. K.; Boyer, J. L.; Jacobson, K. A. Structure-activity