Rose Bengal analogs and vesicular glutamate transporters (VGLUTs)

Article abstract of DOI:10.1016/j.bmc.2010.06.069

Vesicular glutamate transporters (VGLUTs) allow the loading of presynaptic glutamate vesicles and thus play a critical role in glutamatergic synaptic transmission. Rose Bengal (RB) is the most potent known VGLUT inhibitor (K _i 25 nM); therefore we designed, synthesized and tested in brain preparations, a series of analogs based on this scaffold. We showed that among the two tautomers of RB, the carboxylic and not the lactonic form is active against VGLUTs and generated a pharmacophore model to determine the minimal structure requirements. We also tested RB specificity in other neurotransmitter uptake systems. RB proved to potently inhibit VMAT (K_i 64 nM) but weakly VACHT (K_i >9.7 μM) and may be a useful tool in glutamate/acetylcholine co-transmission studies.

Full text of DOI:10.1016/j.bmc.2010.06.069

Bioorganic & Medicinal Chemistry 18 (2010) 6922–6933
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry
journal homepage: www.elsevier.com/locate/bmc
Rose Bengal analogs and vesicular glutamate transporters (VGLUTs)
Nicolas Pietrancosta a, Albane Kessler a, , Franck-Cyril Favre-Besse a, Nicolas Triballeau a,à
,
Thomas Quentin b,c,d, Bruno Giros b,c,d,e, Salah El Mestikawy b,c,d,e, Francine C. Acher a,
*
a Laboratoire de Chimie et Biochimie Pharmacologiques et Toxicologiques, UMR8601 CNRS, Université Paris Descartes, 45 rue des Saints-Pères, 75270 Paris 06, France
b Centre National de la Recherche Scientifique (CNRS) UMR 7224, Université Paris 6, Pathophysiology of Central Nervous System Disorders, 9 quai Saint Bernard, 75005 Paris, France
c Institut National de la Santé et de la Recherche Médicale (INSERM), U952, Université Paris 6, Pathophysiology of Central Nervous System Disorders, 9 quai Saint Bernard,
75005 Paris, France
d Université Pierre et Marie Curie, UPMC, Université Paris 6, Pathophysiology of Central Nervous System Disorders, 9 quai Saint Bernard, 75005 Paris, France
e Douglas Hospital Research Center, Department of Psychiatry, McGill University, 6875 Boulevard Lasalle Verdun, QC, Canada
a r t i c l e i n f o
a b s t r a c t
Article history:
Vesicular glutamate transporters (VGLUTs) allow the loading of presynaptic glutamate vesicles and thus
play a critical role in glutamatergic synaptic transmission. Rose Bengal (RB) is the most potent known
VGLUT inhibitor (Ki 25 nM); therefore we designed, synthesized and tested in brain preparations, a series
of analogs based on this scaffold. We showed that among the two tautomers of RB, the carboxylic and not
the lactonic form is active against VGLUTs and generated a pharmacophore model to determine the min-
imal structure requirements. We also tested RB specificity in other neurotransmitter uptake systems. RB
Received 29 January 2010
Revised 21 June 2010
Accepted 21 June 2010
Available online 25 June 2010
Keywords:
VGLUT
VMAT
VACHT
proved to potently inhibit VMAT (Ki 64 nM) but weakly VACHT (Ki >9.7 lM) and may be a useful tool in
glutamate/acetylcholine co-transmission studies.
Ó 2010 Elsevier Ltd. All rights reserved.
Fluorescein
Glutamate
Non-competitive inhibitors
Tautomers
Active binding form
Pharmacophore model
1. Introduction
brane.1,2 Once in the cytoplasma, neurotransmitters accumulates
into synaptic vesicles for additional cycles of release.3–5 The com-
bined actions of the plasma membrane and vesicular transporters
maintain the proper concentrations of extracellular neurotransmit-
ter before, during and after neurotransmission.6,7 Glutamate, the
major excitatory neurotransmitter in the mammalian central ner-
vous system (CNS), is involved in most, if not all, physiological
functions as well as neuropathologies of the CNS.8–17
The neurotransmitter cycle is an essential step in synaptic
transmission. Neurotransmitters released from synaptic vesicles
by exocytosis bind to various receptors to produce their biological
response. Extraneuronal neurotransmitters are then removed from
the synaptic cleft by uptake sites located in the plasma mem-
In this study, we focus our attention on one step of the glutamate
cycle: the vesicular accumulation uploading by vesicular glutamate
transporters (VGLUTs). Three different isoforms of VGLUTs have
been identified (VGLUT1–3) sharing a high degree of sequence
homology (more than 75%) but with almost complementary distri-
butions of VGLUT1 and VGLUT2, and a more restricted expression
of VGLUT3 in a small set of neurons releasing other neurotransmit-
ters.1,18–21 Recently, a relevant correlation between Alzheimer asso-
ciated dementia and VGLUT1 disappearance has been described.22
This study contributes to investigations on VGLUT markers and/or
modulators as potential early diagnostic tools in AD.23,24
Abbreviations: ACPD, trans-(1S,3R)-1-amino-1,3-cyclopentanedicarboxylic acid;
AD, Alzheimer disease; ADP, adenosine-diphosphate; ATP, adenosine-triphosphate;
CNS, central nervous system; Glu, glutamate; DBU, 1,8-diazabicyclo[5.4.0]undec-7-
ene; DMF, dimethylformamide; DMSO, dimethylsulfoxide; DCM, dichloromethane;
ESI, electrospray ionization; HPLC, high performance liquid chromatography; 5-HT,
serotonine; IC50, concentration for 50% of inhibition; NMR, nuclear magnetic
resonance; PLC, preparative layer chromatography; RB, Rose Bengal; RBL, Rose
Bengal lactone; SAR, structure–activity relashionship; VACHT, vesicular acetylcho-
line transporter; VGLUT, vesicular glutamate transporter; VMAT, vesicular mono-
amine transporter.
* Corresponding author. Tel.: +33 (0)1 42 86 33 21; fax: +33 (0)1 42 86 83 87.
E-mail address: francine.acher@parisdescartes.fr (F.C. Acher).
Present address: GSK, Severo Ochoa 2, 28760 Tres Cantos, Madrid, Spain.
Present address: Galápagos SASU, 102 Avenue Gaston Roussel, 93230 Romain-
à
Despite their key role in excitatory transmission, very few
chemical compounds efficiently target this family of glutamate
ville, France.
0968-0896/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmc.2010.06.069

N. Pietrancosta et al. / Bioorg. Med. Chem. 18 (2010) 6922–6933
6923
at 180 °C before refluxing in methanol. The expected product is ob-
tained in a good crude yield after extraction and engaged directly
in the next step without further purifications. The yields for these
condensations varied from 21% to 59%.
The mechanism of the condensation step involves a multiple
and successive Friedel and Craft acylation (Scheme SI-1). This
sequential mechanism allows the synthesis of asymmetrical43
and symmetrical compounds such as 6 and 7 in which the first at-
tack is performed by the 1,3-dihydroxynaphthalene followed by
resorcinol attack for 6 (Fig. 2).
Cl
O
Cl
Cl
Cl
OH NH2
N
CO2H
I
N
I
HO3S
SO3H
HO
O
I
2
I
Trypan Blue
IC50= 50 nM
Rose Bengal
IC50= 19 nM
HO
N
NH2
CO2H
HO2C
Additionally, this mechanism (Scheme SI-1) determines the posi-
tions of the R1–4 substituents in compounds 1 and subsequently in 2
in Scheme 1. Starting with mono-substituted phthalic anhydrides, a
single or a mixture of two regioisomers could be formed with sub-
stituents at R1,4 (1h, 1j) or R2,3 (1i, 1k, 1l) positions. The first nucleo-
philic attack of resorcinol on the phthalate derivative occurred at the
most electrophilic carbonyl (Scheme SI-1), but in the case of com-
pounds 1h, 1i, 1k, 1l the two phthalic carbonyls appeared to display
an equivalent electrophilicity and a mixture of two isomers for each
compound was obtained40 (Scheme SI-2A). In order to confirm this
hypothesis, we performed the 13C NMR spectra of starting materials
(i.e., phthalate derivatives) assuming that the chemical shifts of the
carbonyl carbons could be correlated to their reactivity against
nucleophilic attack. We observed that when the difference of car-
bonyl 13C chemical shift of starting phthalate anhydride was less
than 2 ppm, two regioisomers were obtained (i.e., 1h, 1i, 1k, 1l)
but in the case of compound 1j only one isomer was identified by
1H NMR (Table SI-2). This is probably due to the delocalization of
oxygen electrons of the hydroxyl group through the beta carbonyl
resulting in its deactivation (Scheme SI-1B). The different isomers
were not separated and each mixture was evaluated for its uptake
inhibitory activity. As none of the tested compounds appeared to
have significant activity (see below), the separation was definitively
not carried out although some recrystallization protocols have been
described in which protection deprotection steps were re-
quired.40,44–46
HO2C
CO2H
(1S,3R)-ACPD
IC50 = 230 µM
Quinolein dicarboxylic acid
IC50= 40 µM
Figure 1. Different families of known VGLUTs ligands.
transporters.1 The need of pharmacological tools is now crucial to
understand better the functional implication of VGLUT1–3 in nor-
mal and pathological conditions and to use them as biomarkers of
human ‘synaptopathologies’ for diagnosis or therapeutic purposes.
Presently, three types of VGLUT inhibitors have been identified1
(Fig. 1): dyes5,25–33 with Ki in the 20 nM to 10
l
M range (e.g., Rose
Bengal and Trypan Blue), substituted quinolines29,34,35 with Ki in
the 40–300 M range and glutamate analogs with IC50 >230
l
lM
(e.g., (1S,3R)-ACPD). Dyes display higher affinities, probably be-
cause of their hydrophobic aromatic part, but a lower selectivity
than quinolines.2 Among them, Rose Bengal (RB) is the only non-
competitive inhibitor and it exhibits the highest affinity of all (Ki
19 nM);36 The third type includes the glutamate-like inhibitors
known to interact with glutamate receptors but these compounds
show very poor activity on vesicular glutamate uptake.1 We may
explain these results by the low affinity of VGLUTs against gluta-
mate, its physiological substrate (Km(Glu) >1 mM).
The original aim of this study was the design of new VGLUT
binding compounds with improved affinity and selectivity. Since
RB was the most potent known inhibitor, we chose to synthesize
various derivatives 1–8 and to evaluate them as glutamate uptake
inhibitors in a 96-well plate assay. Their activities were merged
with those of commercially available dyes that had been previ-
ously tested (BU1 to BU15 in Table 1)33 to establish a structure–
activity relationship and to generate a pharmacophore model. This
model allowed to define the chemical and spatial features of RB
that are required for potent vesicular glutamate transport inhibi-
tion. To assess its selectivity, we also examined the efficacy of RB
(2b) on the vesicular monoamine transporter (VMAT) and vesicular
acetylcholine transporter (VACHT). Since RB is a non-competitive
inhibitors, we discuss its possible binding to other proteins in-
volved in glutamate transport.
The last step of the synthesis consists in the insertion of iodine
in the xanthene moiety. Among the different classical methodolo-
gies, we have selected one of the smoothest (I2/HIO3 in EtOH).41
This method leads to the desired compounds 2 without detection
of any partial iodination (Scheme 1). Moreover, this strategy is
suitable for the final purification processes. Indeed, one of the ma-
jor difficulties of this synthesis consists in the purification and the
characterization of the compounds that revealed not to be trivial:
these molecules are highly aromatic and planar, they are subjected
to
p-stacking which also explains their low solubility in common
organic solvents.47 Introduction of iodine or bromine atoms con-
fers to them a high degree of anisotropy. In most cases, the remain-
ing protons of the molecules do not relax properly and the 1H NMR
spectra were very sluggish with broad signals. Similarly, 13C NMR
spectra could not be recorded. Variation of temperature could
not improve the quality of the spectra, neither did dilution of the
samples. Moreover, only one 1H NMR peak is expected with the
more soluble lactonic tautomer.48–50 Compounds were thus char-
acterized by mass spectrometry and their purity assessed by HPLC.
A HPLC method was set up using a C18 column and a methanol/
water gradient as eluent (see below). Mass spectroscopy (negative
electrospray ESIꢀ) of the crude precipitated compound showed in
most cases a single peak corresponding to [MꢀH]ꢀ. In few cases,
we have observed the corresponding positive peak in ESI+ mode.
A tautomeric equilibrium between a carboxylic and lactonic
form of fluorescein derivatives is known to take place (Scheme
2). Christophersen et al. have demonstrated the structure of red
and orange fluorescein, that is, tautomeric forms, by a combination
of solution and solid state 13C NMR methods.51 Importantly this
equilibrium is pH dependent. Fluorescein derivatives exist in four
2. Results and discussion
2.1. Chemistry
All synthesis of fluorescein analogs that were described follow
the same strategy37–39 that we also adopted. Synthesis of com-
pounds 2 was performed in two steps: coupling of phthalic anhy-
dride with resorcinol to afford fluorescein derivatives40 1 and
subsequent introduction of iodine onto the xanthene moiety
(Scheme 1).41 For the poly-condensation, two methods have been
described in the literature with Brønsted or Lewis acids. After com-
paring these two methods (methane sulfonic acid42 or ZnCl243), we
selected the zinc chloride catalysis protocols due to higher yields
and easier workup. In this protocol, the two reagents are crushed
together, pooled with zinc chloride and the neat mixture heated

6924
N. Pietrancosta et al. / Bioorg. Med. Chem. 18 (2010) 6922–6933
Table 1
Activities of newly synthesized (1–8) compounds and of those previously tested and numbered BU1–1533
R3
R2
R1
O
R4
R7
CO2H
O
I
I
I
I
R6
R6
O
HO
O
HO
O
OH
O
HO
I
I
I
I
R5
R5
BU4
BU13
1-3,6,7,8 and BU1-BU3, BU5-BU12, BU14, BU15
Entry
R1
R2
R3
R4
R5
R6
R7
IC50 nM (n)
2b
2c
2a
2d
BU1
BU6
BU2
BU9
BU14
2f
Cl
Cl
H
Cl
H
H
Cl
H
H
Cl
Cl
H
I
I
I
I
I
Br
I
Br
I
I
I
Br
I
I
Br
I
I
Br
CH3
I
I
I
I
I
Br
I
I
I
I
I
I
I
I
I
I
I
I
Br
I
Br
I
I
I
Br
I
I
NO2
I
I
Br
I
I
I
I
H
I
NO2
I
I
I
I
I
CO2H
CO2H
CO2H
CO2H
CO2Me
CO2H
CO2Bn
CO2Et
H
25 3.2 (9)
27 4.1 (3)
36 6.5 (6)
41 8.4 (3)
44
51
65
65
68
74 11.7 (3)
102 5.4 (3)
117
123 17.4 (4)
192 5 (4)
238 9 (2)
288 71 (3)
H
Cl
Cl
Cl
Cl
H
Cl
Br
F
Cl
Cl
Cl
Cl
H
Cl
Br
F
H
Cl
Cl
Cl
H
Cl
Br
F
Cl
Cl
Cl
H
Cl
Br
F
CO2H
CO2H
CO2Me
CO2H
CO2Et
CO2Et
CO2H
CO2H
CO2H
CO2H
CO2H
CO2H
CO2H
CO2H
CO2H
CO2H
CO2H
CO2H
CO2H
CO2H
CO2H
CO2H
CO2H
CO2H
SO3H
CO2H
CO2H
CO2H
CO2H
2e
BU10
H
H
H
H
4a
Cl
Cl
H
H
H
H
H
H
H
H
H
—
H
H
H
H
NO2
H
H
Cl
Cl
H
CO2H
H
H
H
NCS–
CH3
H
H
—
H
H
tBu
H
H
H
H
Cl
Cl
H
H
CO2H
H
H
H
H
CH3
H
—
H
H
H
tBu
H
H
H
Cl
Cl
H
Cl
Cl
H
H
H
H
H
H
H
H
H
—
H
H
H
H
H
NO2
OH
Cl
Cl
H
H
H
Cl
Cl
3
8
2ib
BU7
2g
406
531 175 (3)
689
BU3
2kb
987 137 (3)
BU5
BU4
BU8c
BU13c
2lb
1060
2200
2445
3850
5925 3685 (2)
2hb
6690 2048 (2)
2j
I
I
H
H
Cl
H
7800 648 (2)
8824 1119 (2)
15,000
>32,000
>64,000
Inactive (3)
Inactive (3)
Inactive (3)
5c,d
BU11
BU15
BU12
1a
Cl
Cl
H
H
H
Cl
Cl
H
H
H
H
OH
H
H
H
H
H
H
Cl
Cl
6
7
Cl
Cl
Cl
Cl
Mono-naphthyl
Di-naphthyl
Data from this study and literature may be compared on the same scale since Erythrosin B (2a) and RB (2b) display similar activities in both studies.
a
RB monoether.
Mixture of the two diastereoisomers.
Lactone form.
b
c
d
RB diether. Each concentration has been measured in triplicate and data are the mean SEM of n separate experiments.
protonated states (neutral form, monoanion A or B, dianion,
Scheme 3) because of two ionizable groups. In the fluorescein 1
series, substituents present in the xanthene moiety (R5 and R6,
Scheme 3) highly influence the pKa of the phenolic oxygen, while
when in the phenyl ring (R1–R4, Scheme 3) they only modulate
the pKa of the carboxylic acid. Several authors have shown that
the lactone form is only found with the neutral form and not with
mono- or dianion forms.39,48,52 Consequently at physiological pH
most tested derivatives are only found in the carboxylic tautomeric
form because of deprotonation of their carboxylic and phenolic
groups. However some exceptions were described. In fact, fluores-
cein analogs bearing mesomer acceptor substituents (e.g., NO2) in
ortho positions to the phenol groups (R5 and R6 in Scheme 3, BU8
in Table 1) as well as phenolphthalein derivatives (BU13 in Table
1) are only found as lactones at neutral pH.53–56
Thus a key point was to identify which form between carboxylic
acid and lactone of all tested analogs, occurs in biological conditions
(pH 7.2). Indeed the two tautomers exhibit different 3D conforma-
tions as well as diverse ability to contract some protein interactions
and thus reveal different pharmacological effects. As RB and RBL
(Rose Bengal lactone) are commercially available, and their spectral
properties described, UV absorption of these two samples in a pH 7.2
phosphate buffer were measured. The two spectra are totally identi-
cal and the same as described for RB open form (different from RBL),
showing that in these conditions, the RBL opens up (Scheme 3). In or-
der to further characterize which forms of compounds 2 are present
insolution, they wereanalyzedby HPLCona C18columnelutedwith
a methanol/water gradient. Open compounds showed retention
timesaround15 minwhereascyclizedformswerelessretainedwith
a retention time ranging from 5 to 8 min. RB and RBL dissolved in a

N. Pietrancosta et al. / Bioorg. Med. Chem. 18 (2010) 6922–6933
6925
R3
R3
R2
R1
R2
R1
R4
CO2H
R4
R5
R1
O
R2
R3
HO
OH
Method C
method A or B
CO2H
O
+
I
I
R4
O
HO
O
HO
O
O
O
R5
R5
R5
R5
2(a-l)
1(a-l)
Cl
Cl
Cl
Cl
Cl
Cl
Cl
Cl
Cl
Cl
Cl
O
O
method D
method E
CO2H
O
Cl
O
I
I
I
I
I
I
R8
O
O
O
HO
O
O
HO
O
O
I
I
I
I
I
I
3
2b
4 R8 = H
5 R8 = CH3
O
O
1) KOH
2) method D
O
O
O2N
HO
NO2
OH
O2N
HO
NO2
O
O
O
Br
Br
Br
Br
BU8
8
Scheme 1. Formation of fluorescein derivatives by condensation of phthalic anhydride and resorcinol and subsequent iodination. 1: R5 = H, 2: R5 = I, a: R1 = R2 = R3 = R4 = H; b:
R1 = R2 = R3 = R4 = Cl; c: R1 = R4 = Cl, R2 = R3 = H; d: R1 = R4 = H, R2 = R3 = Cl; e: R1 = R2 = R3 = R4 = F; f: R1 = R2 = R3 = R4 = Br; h: R1 = NO2, R2 = R3 = R4 = H and R1 = R2 = R3 = H,
R4 = NO2; i: R1 = R3 = R4 = H, R2 = CO2H and R1 = R2 = R4 = H, R3 = CO2H; j: R1 = R2 = R3 = H, R4 = OH; k: R1 = R3 = R4 = H, R2 = CH3 and R1 = R2 = R4 = H, R3 = CH3; l: R1 = R3 = R4 = H,
R2 = t-Bu and R1 = R2 = R4 = H, R3 = t-Bu; g: R1 = R2 = R3 = R4 = H R5 = CH3. Reagents and conditions: (a) method A: CH3SO3H, 80 °C; method B: ZnCl2, 180 °C, neat; method C:
HIO3/I2/EtOH; method D: Et-Br, DBU, THF, 65 °C; method E: CH3I/DMF, rt.
phenolderivativeasMitsunobureactionormethyliodidealkylation.
Compound 4 was obtained by direct alkylation of RB lactone by
methyl iodide in THF. To block Rose Bengal in its cyclized form, we
performed the O-methylation of both phenolic oxygens with 2 equiv
of methyl iodide affording diether 5. Unlike 3 and 5, 4 may still un-
dergo the tautomeric equilibrium (Scheme 3) which is shifted to
the carboxylic form at neutral pH.57,58
Cl
Cl
Cl
Cl
Cl
Cl
Cl
Cl
CO2H
CO2H
HO
O
6
O
HO
O
O
7
Figure 2. Structure of compounds 6 and 7.
2.2. VGLUTs Pharmacology
In order to rationalize the synthesis of the new compounds, we
thought to use the structure–activity study (SAR) performed by Bole
and Ueda (entries BU1 to BU15 in Table 1).33 As mentioned previ-
ously, all Rose Bengal derivatives may exist in either carboxylic or
lactonic forms (Scheme 2). A large number of physicochemical stud-
ies (pH- and solvent-dependent) on fluorescein derivatives, focused
on the equilibrium between the different tautomers, have been pub-
lished.52,59–61 Yet, Bole and Ueda33 left out this point and considered
that carboxylic acid and lactone of RB analogs remained as two dif-
ferent chemical entities in aqueous solution. Consequently, we
thought that the RB SAR study of VGLUT inhibition needed to be
revisited. The new and former pharmacological data were analyzed
and allowed to determine the structural requirements for VGLUT
inhibition by RB analogs. The RB analogs were first tested with the
commonly used [3H]-Glu vesicular uptake assay in the presence
(or absence) of inhibitors at various concentrations, as previously
described.62 This type of assay was technically incompatible with
structure–activity studies because of the limited number of com-
pounds that could be tested simultaneously. Hence, a 96-well plate
O
CO2H
O
HO
O
HO
O
OH
O
carboxylic form
lactonic form
Scheme 2. Tautomeric equilibrium of fluorescein.
pH 7.2 phosphate buffer eluted both as RB at 15.80 min (Fig. SI-1).
Carboxylic form of fluorescein and iodinated analogs in basic condi-
tions was also confirmed by HPLC. Finally to identify which of the
tautomeric forms binds to the transporters VGLUTs, we synthesized
various analogs locked in the carboxylic or lactonic form (3–5, 8
scheme 1). The esterified compounds 3 and 8 were obtained by reac-
tion of ethyl bromide with RB in the presence of DBU in THF. Differ-
ent methodologies have been described for the etherification of

6926
N. Pietrancosta et al. / Bioorg. Med. Chem. 18 (2010) 6922–6933
R3
R3
R3
R2
R1
R2
R1
R2
R1
R4
R4
R4
-
-
CO2H
CO2
CO2
R6
R6
R6
R6
R6
-O
R6
pK1
pK2
HO
O
HO
O
O
O
O
O
R5
R5
R5
R5
R5
R5
neutral form
monoanion A
dianion
or
R3
R3
R4
O
R2
R1
R4
CO2H
R6
R2
R1
O
R6
R6
OH
R6
-O
HO
O
O
O
R5
R5
R5
R5
lactone
monoanion B
Scheme 3. Protonation state of fluorescein derivative tautomers at various pH. Monoanion A or B are present according to R5 and R6 substituents that influence the pKa of the
phenol. For R1–R6 see Scheme 1 and Chart SI.
assay was set up and validated with the following controls. Activity
measures in this up-scaled assay and classical assay proved to be
comparable as demonstrated with RB (2b) and Erythrosin B (2a)
(Table 1). IC50 values of 25 nM and 36 nM for 2b and 2a, respectively,
were obtained with the 96-well plate assay that compare to 22 nM
and 35 nM obtained by Bole and Ueda.33 Moreover, the 96-well plate
assay increased the robustness of data and allowed higher precision
in the IC50 determination. It also highlighted the solubility troubles
encountered with RB derivatives even at low micromolar concentra-
tions while this problem was not easily detected in the traditional
assay (reduced filter diameter in the 96-well plate assay). As men-
tioned previously, the Rose Bengal analogs are sparingly soluble in
various organic solvents such as dimethylsulfoxide (DMSO) tradi-
tionally used for compound solubilization. In order to be sure about
thetested concentration, our analogswere dissolved in thetransport
biological buffer. The compounds were then filtered and the concen-
tration confirmed by HPLC analysis as described by Bole and Ueda.33
The activities of all new compounds are reported in Table 1 together
with literature data of compounds BU1–BU15.33 These two sets of
IC50’s can be directly compared since two compounds 2b (RB) and
2a (Erythrosin B) which were used to calibrate the assay displayed
similar values in both studies (see above). Altogether the improve-
ment of the uptake assay resulted in rapid and reliable evaluation
of our new compounds.
potent as RB ester 3 and one log more potent than BU8. We can
thus conclude that the bioactive form of Rose Bengal is the carbox-
ylic form which is indeed the predominant form in aqueous solu-
tion buffered at pH 7.2 for biological assays. The loss of planar
structure and loss of conjugation of the xanthyl moiety with C90
sp3 hybridized may be the cause of the reduced activity of RB lac-
tones and BU13. Such an observation would be in line with a flat
hydrophobic binding site.
As previously reported,33 positions 20, 40, 50, 70 (R5 and R6
Table 1) of the fluorescein xanthyl part require iodine or bromine
substitutions. Comparison of IC50’s of 2b and BU6 with BU11
revealed a decrease of over two orders of magnitude when I or Br
were replaced by H. This critical role is confirmed by the reduced
effect of 1a and BU5 compared to 2a. Any other substitutions with
smaller or larger groups (Fig. SI-2) decreased the inhibitory potency:
compound 2g in which two iodines were replaced by a methyl group
was 20 times less active than the tetraiodo analog 2a (Erythrosin B)
and the bulky compounds 6 and 7 were totally inactive at a 2.5 lM
concentration.
Regarding the phenyl part of fluorescein analogs, it may not be
replaced by a linear chain since BU4 is poorly active but the pres-
ence of four chlorine atoms at positions 4–7 is not mandatory since
2a (4xH), 2b (4xCl) and 2c (2xCl) are equipotent and 2f (4xBr) and
2e (4xF) only slightly weaker. Because of this relative lenience, we
designed a series of 2a analogs varying the phenyl substituents.
However while halogens were well accommodated decreasing
from chlorine (2b and 2c) to bromine (2f) and fluorine (2e) and
even hydrogen (2a), other substitutions were deleterious. Intro-
ducing a polar substitution at position 4, 7 (R1, R4) such as NO2
in 2h, OH in 2j or at position 5, 6 (R2, R3) such as CO2H in 2i re-
sulted in a drop of activity. Hydrophobic groups at position 4, 7
(R1, R4) induced an even more dramatic loss with a methyl group
in 2k and a tert-butyl in 2l. The carboxylic group at position 3 of
the phenyl ring (R7 in Table 1) is unnecessary for binding but useful
for aqueous solubility. Actually this group may be esterified to
methyl, ethyl or benzyl ester or even removed (BU14) without ma-
jor loss of potency. This observation suggests the design of a new
ester library that could be obtained by easy esterification of 2b
with diverse alcohols. Previous studies have taken advantage of
that carboxylic function in 2b to link fluorescent RB to various ac-
tive principles (e.g., peptides, alkyl chains).63,64 Such a substitution
2.3. Structure–activity relationship
VGLUT inhibitors listed in Table 1 (structures displayed in Chart
SI) were ranked in order of potency. Among the 17 new compounds
none was more efficient than RB (2b). The monoether derivative 4
that may still undergo the tautomeric equilibrium and is mostly
found as carboxylic acid at neutral pH, was used as a probe of phe-
nolic methylation when its activity was compared to that of 5. On
the other hand, the role of the RB carboxylic group was defined by
comparison of the inhibitory effect of descarboxy RB (BU14) with
that of RB (2b). While 3, 4, 8, BU1, BU2, BU9, BU10 and BU14 still
reasonably inhibited glutamate transport, analogs 5 and BU8
required at least one or two log higher concentration. Although
not as hydrophobic, the R6 nitro group of BU8 is neutral, electron
attracting and sterically similar to the iodo substituent of 2a and
2g (Fig. SI-2). Moreover, ester 8 of opened lactone BU8, was as

N. Pietrancosta et al. / Bioorg. Med. Chem. 18 (2010) 6922–6933
6927
may thus enable to reach a specific pocket that would increase the
specificity of the inhibitor.
A subtractive phase then discards the solutions that can be mapped
by a majority of inactive compounds. Finally, an optimization
phase relaxes the remaining models via a simulated annealing
algorithm. During this last phase, features are to some extent mod-
ified (their type, position, and, optionally, their weights and toler-
ances) in order to find a fit that better correlates with the measured
activities. In the present case the generated models are combina-
tions of four types of features: hydrophobic, ring aromatic, H-bond
acceptor and electron-withdrawing group exhibiting either fixed
or variable weights and tolerances. The retained models are those
that exhibit a limited computational cost to describe the SAR, but
they should also predict the activities within the correct log range
(i.e., root mean square ‘RMS’ below 1 log of activity) and show
acceptable discriminatory power between most active and least
active compounds (area under the ROC curve >0.8). The best theo-
retical model (‘fixed cost’) and the worst (‘null cost’) serve as refer-
ences to determine the quality of the generated models (Table 2).
Here, the use of variable weights allowed to slightly improve
the RMS of the best model from 1.02 to 0.98 log of activity while
retaining the same features and tri-dimensional dispositions as of
the best model obtained with fixed weights (data not shown). This
means that the additional complexity (i.e., cost) introduced while
allowing weight variation is mostly due to the use of variable
weights. It was therefore decided to retain this model because
weights introduce an extra facet in the SAR analysis providing in-
sight into the contribution of each feature to the activity. This mod-
el (Fig. 3) consists of four features: two ring aromatic A1 and A2
(one for each side ring of the xanthene system with weights of
2.9 and 2.25) and two hydrophobic features H1 at R5 position of
the xanthene moiety (weight: 2.25) and H2 located on the orthog-
onal phenyl substituent (weight 1.58). The presence of the two ring
aromatic features A1 and A2 together with their total weight of
5.15 underscores the importance of the planarity of the tricyclic
xanthene system. Indeed open structures such as BU13 exhibit a
strong angle between its lateral rings. Other structures like 7 adopt
a twisted helicene-like conformation thus limiting the fit of such
compounds and explaining their inactivity on VGLUTs.
Altogether halogen substitutions at the xanthyl and phenyl ring
of RB (2b) are the most critical. A specific interaction between hal-
ogens and oxygen atoms from proteins and in particular from
backbone carbonyls has been recently reported.65 This halogen
bonding was demonstrated to stabilize several protein kinase–
inhibitor complexes. Such interactions may also be involved in
RB–VGLUT complex and would explain the unique properties of
halogen substituents in RB. Alternatively, the halogen atoms may
be binding in a hydrophobic cavity that would be best fitted with
iodine.
2.4. VGLUT pharmacophore model generation
Overall the SAR data allowed us to generate a pharmacophore
model of RB VGLUT inhibitors. The modeling part of this study
was performed with version 2.5 of the Discovery Studio suite avail-
able from Accelrys Inc. (San Diego CA, USA).
Each molecule was drawn in its predominant tautomer and its
major protonated state at the pH used for the biological tests (pH
7.2), as determined above. To ensure proper conformational cover-
age, each isomer of compounds available in mixtures (2h, 2i, 2k, 2l)
was drawn separately yielding a total of 37 molecules. Conforma-
tional expansion was performed with Catalyst-CatConf66,67 using
the BEST mode and an energy window of 10 kcal/mol above the
global minimum. It is worth to note that the opened structures
of fluorescein derivatives can be drawn with two major tautomeric
forms with one phenolic and one quinonic form on each side of the
central pyran ring of the xanthene system yielding a totally sym-
metric structure. Despite this symmetry DS Pharmacophore (i.e.,
the embedded Catalyst engine) is unable to perceive the two side
rings as equivalent when feature mapping is performed. In order
to address this point, the default Ring_Aromatic feature was edited
to allow equivalent perception of the phenolic and quinonic parts
of the molecular graph. Secondly, the default HB_Acceptor feature
was modified to forbid the mapping of some weak H-bond accep-
tors present in the dataset (ether oxides, divalent oxygen atoms of
esters, nitro groups and phenolic oxygens hindered by dual substi-
tution in ortho and ortho0). Finally, a novel feature was designed to
account for the presence of electron-withdrawing groups at certain
positions of many compounds of the dataset and which could play
a significant role in explaining the SAR. This new feature matches
with nitro, cyano, trifluoromethyl, fluoryl, chloryl, ketonic and
acidic moieties.
Pharmacophore model perception was performed using Cata-
lyst-HypoGen executable from the ‘3D QSAR Pharmacophore Per-
ception’ protocol available from Discovery Studio 2.5 which
invokes Catalyst-HypoGen via a PipelinePilot server.68
The training set consists in a selection of 18 compounds that
regularly span the activity range: 2c, 2a, BU1, BU6, 2e, BU10, 4,
BU7, 2g, BU5, BU4, BU13, 2j, BU11, BU15, BU12, 1a and 7. Their
respective activities were kept as measured and their uncertainties
set to 3. HypoGen was then left free to select the most relevant fea-
tures among HYDROPHOBIC, Ring_Aromatic, HB_Acceptor and
electon-withdrawing group with a minimum interfeature distance
set at 2 Å. Several runs were performed with weights and/or toler-
ances kept fixed or authorized to vary.
The second most important feature in terms of weight is H1. If
one considers the planar compounds which can fit the two ring
aromatics, H1 correctly discriminates between the most active
compounds exhibiting iodine, bromine or methyl at the R5 position
from the compounds bearing a hydrogen or a hydrophilic substitu-
ent at this very position: roughly, compounds that correctly map
A1 and A2 but not H1 have an IC50 above 2.5 lM. H2 finally ex-
plains the limited activity of compounds such as BU4 which does
not have the orthogonal phenyl moiety or lactonic compounds
such as BU13 which are endowed with such a group but cannot
project it to the correct location due to the tetrahedric geometry
of the central carbon atom. Its lowered weight however does not
penalize enough some compounds such as 2i and 2j which, despite
hydrophilic substituents in R4, are still overestimated by the mod-
el. This is a known limitation of pharmacophore perception with
Catalyst-HypoGen which focuses only on the presence (or absence)
of features contributing to the activity but does not account for the
presence of features penalizing the activity (here hydrophilic sub-
stituents at the R4 position).
It is worth to note that neither HB_Acceptor nor electron-with-
drawing group features were retained in this model. In particular,
Catalyst-HypoGen could have used such features to depict the SAR
of the various orthogonal phenyl substituents. A comparison be-
tween 2b (RB), BU1, BU2 and BU14 which all have IC50 of the same
order of magnitude and only differ by their various HB_Acceptor
moiety allows to understand the irrelevance of this feature. It is
quite remarkable in fact that there is such little activity differences
between a strong negatively charged function (2b, RB), a neutral
weaker acceptor (BU1), a bulky substituent (BU2) or a simple
The next step consists in the model validation. Indeed, Catalyst-
HypoGen is a pharmacophore model perception method that
determines minimum sets of features capable to explain the differ-
ences in activities among a collection of compounds spanning typ-
ically 3 to 4 logs of activities.68 Its heuristic is divided in three
phases. First of all, a constructive phase builds the pharmacophore
space from the two most potent compounds keeping the configu-
rations that matches with the rest of the most active molecules.

6928
N. Pietrancosta et al. / Bioorg. Med. Chem. 18 (2010) 6922–6933
Table 2
Results obtained with the best models (model with lowest cost of 10 retained
hypotheses after each run) obtained during the different runs performed with fixed
and/or variable weights and tolerances
Parameters/model
Cost
RMS
ROC (AUC)
Fixed Best Null Fixed Best Null Fixed Best Null
Default (fixes weights 75
and tolerances)
85
1.02
0.94
Variable tolerances
Variable weights
Var. tolerances
and weights
88
87
99 119
96
0
1.02 2.56
0.98
0.99
1
0.95 0.5
0.94
0.93
100 109
Figure 4. Inhibition of vesicular neurotransmitter uptake: glutamate, 5-HT and
acetylcholine by Rose Bengal. These experiments were performed with crude
synaptic vesicles from rat brains. Assays could not be run at RB (2b) concentrations
above 10ꢀ6 M because of solubility limitation.
VMAT inhibition, RB may be a useful tool to study the mechanism
differences between VMAT and VACHT.72
Finally, our observations led us to question about the binding
site of RB and its analogs. While Ueda has partially addressed that
issue and eliminated Na+/K+ channel as well as V-ATPase involve-
ment,36 no unambiguous answer has been provided. Indeed, the
neurotransmitter uptake was performed in an entire synaptic ves-
icle. As Rose Bengal (2b) has been described as a non-competitive
inhibitor, we could reasonably envisage the possibility for Rose
Bengal (2b) to indirectly block glutamate uptake by interfering
with another protein located upstream of VGLUT in the uptake
pathway. For that purpose we have analyzed more carefully re-
ported biological activities of fluorescein derivatives. A brief over-
view revealed that Erythrosin B (2a) was known to inhibit several
neurotransmitter uptake58,73 but also choline acetyltransferase by
competitive binding at the adenosine binding site.74 Interestingly,
several publications have reported on the competitive75–79 and
non-competitive kinase inhibition80 and non-specific enzymatic
inhibition81 of RB (2b) but in all these studies, RB acted at higher
concentrations than at VGLUTs (micromolar concentrations com-
pared to 25 nM). Since RB is a non-competitive inhibitor of VGLUTs
with respect to glutamate and exhibits almost no inhibition of ATP
hydrolysis in the uptake assay,36 it is likely to bind to a specific
allosteric site of the transporter. Such ligands are considered to
be advantageous as they usually afford higher selectivity than
those binding at endogenous substrate (e.g., glutamate or ATP) site.
In addition, it was also observed that kinase inhibitors may simul-
taneously bind at several sites that may be of high and low affin-
ity.76,82 This situation may also occur presently. Furthermore,
Schoichet and collaborators suggested that this non-specific effect
could be due to the formation of aggregates that are well known
with this kind of dyes.81 We suggest that RB could act as well as
an ATP- or ADP/AMP- competitive binding site inhibitor (but not
at the V-ATPase binding site) as it has been observed for some
other fluorescein analogs.75–79
Figure 3. Selected VGLUTs pharmacophore model with Rose Bengal displayed
(atom colors: carbon gray, oxygen red, chlorine green, iodine purple). Specific
features are represented as colored spheres (dark blue: A1 and A2 ring aromatic,
cyan: H1 and H2 hydrophobic features).
hydrogen (BU14). To explain these observations, we propose that
the carboxyl moiety is pointing toward the solvent. Although we
can not comment the mapping of 2h, 2i, 2k and 2l because the
activity of each isomer could not be resolved from the obtained
mixtures, the remaining test set compounds (RB, 2d, 2f, 4, BU2,
BU3, BU8, BU9, BU14, 6, 8) were all predicted in their actual log
of activity at the exception of 5 (overestimated). Clearly the lac-
tonic form of compounds such as BU8, BU13 and 5 seem detrimen-
tal to the activity but, as stated before, unfavorable features are not
properly considered during HypoGen runs.
2.5. VMAT2 and VACHT Pharmacology
In parallel to the structure/activity studies, and in order to eval-
uate more carefully the potentiality of this scaffold as VGLUT
markers/modulators, we decided to examine the specificity of Rose
Bengal (2b) with regards to other families of vesicular neurotrans-
mitter transporters.69 We assessed the efficacy of RB (2b) to inhibit
the vesicular H+-dependent accumulation of glutamate, mono-
amine (5-HT)69,70 and acetylcholine mediated by the respective
transporters (VGLUTs, VMAT2 and VACHT) (Fig. 4).
As previously reported, Rose Bengal inhibited glutamate vesicu-
lar uptake (IC50 25 1.1 nM, with a Hill number of 1.06 0.08,
Fig. 4). However, unexpectedly Rose Bengal was also able to inhibit
H+-dependent [3H]-5-HT accumulation with an IC50 64 1 nM and
Hill number of 1.7 0.12. On the other hand, the efficiency of RB
(2b) in acetylcholine uptake experiments was markedly low with
Indeed, numerous studies have been performed to elucidate the
structure–activity requirement for efficient ATP competitive inhi-
bition,83 and have shown, among other characteristics, the involve-
ment of a flat hydrophobic feature into the ATP binding pocket.
Such a feature is actually found in the fluorescein scaffold and in
no significant effect between 0.01 and
1 lM (extrapolated
IC50 ꢁ 9.7 2.3
l
M). Such results are surprising since VMAT and
VACHT belong to the same SLC18 family and VGLUTs to a related
but distinct SLC17 family.71 Yet, taking advantage of its preferred

N. Pietrancosta et al. / Bioorg. Med. Chem. 18 (2010) 6922–6933
6929
our pharmacophore model.84–87 However pharmacophore models
of kinase inhibitors are also characterized by at least one hydrogen
bond donor or acceptor not present in our model, therefore sup-
porting the hypothesis of an allosteric binding site. We could also
envisage a possible interaction of RB with other proteins that take
part in the transport of neurotransmitters and are present in the
system such as chloride, potassium channels or kinase such as
pyruvate kinase88 or phosphodiesterase.89 Altogether our data
are consistent with a similar high nM affinity binding site of RB
at VGLUTs and VMAT2 that is distinct from the weaker affinity
and non-specific microM sites that have been described for numer-
ous proteins. With VACHT, RB would only bind to such a weak
affinity site.
electrospray ionization (ionization tension 4.5 kV, injection temper-
ature 240 °C). HPLC analyses were carried out on a Gilson analytical
instrument with a 321 pump, CC 125/4.0 Nucleosil 100-5 C18 col-
umn (Macherey-Nagel, 125 mm ꢂ 4 mm), temperature was con-
trolled with an Igloo-CIL Peltier effect thermostat, eluted peaks
were detected at 210 nm and 277 nm by a UV–vis 156 detector
and retention times are reported in minutes. Products were eluted
witha water/methanolmobile phase at 40 °C, a 0.5 mL/min flow rate
and detected by UV. The following gradient conditions were used
with solvent A (H2O) and solvent B (methanol): 80% A and 20% of B
for 3 min, linear increase from 20% to 80% B between 3 and 15 min,
20% A–80% B from 15 min to 25 min. Purity of the tested compounds
was established by analytical HPLC and HPLC–MS and was at least
95%.
3. Conclusions
4.2. General methods for the formation of fluorescein
derivatives
Although the RB analog synthesis may appear to be relatively
straightforward in view of the few synthetic steps required, their
poor solubility and characterization difficulty encountered during
the preparation, made it not so trivial. The structure–activity anal-
ysis revealed that the chemical optimization of RB (2b) may not be
an easy task since halogen substitution remains optimal and intro-
ducing any other groups on the xanthyl or phenyl moiety de-
creased the inhibitory potency. Nevertheless, position 3 of the
phenyl ring appeared to be less stringent indicating that the SAR
could be extended with a new series of RB esters. These observa-
tions were confirmed by the modeling study identifying only
hydrophobic features as seems to be required for a good VGLUT
inhibition. Our pharmacological studies on VGLUTs, VMAT2 and
VACHT, revealed the lack of selectivity of RB (2b) between VGLUTs
and VMAT2. Yet, our results point out the presence of a high affin-
ity binding site on these two transporters that would deserve fu-
ture investigations. This site may be an allosteric VGLUT site or
situated on another protein required in the vesicular uptake sys-
tem. Altogether, these results confirm the interest of Rose Bengal
and analogs both as potent inhibitors of vesicular glutamate up-
take and as potential useful tools for neurotransmitter uptake
studies. In particular, if we consider investigation of mechanisms
in which both VGLUTs (or VMAT2) and VACHT are involved to-
gether.72 Moreover, we also open the way for selectivity optimiza-
tion and provide some insights to elucidate the inhibition binding
mode of this scaffold.
4.2.1. Method A
Phthalic anhydride or tetrachlorophthalic anhydride (1 equiv)
was added to a solution of resorcinol (2 equiv) in methane sulfonic
acid (1 M). The resulting mixture was heated under nitrogen at
80–85 °C for 36–48 h. The cooled mixture was poured into six vol-
umes of ice water followed by filtration. The filtrate containing
fluorescein or derivatives was dried in vacuo to constant weight.
4.2.2. Method B
Resorcinol (2 equiv) was ground together with phthalic anhy-
dride (1 equiv) and ZnCl2 (0.3 equiv) with a pestle and mortar,
and the resulting mixture was heated in a 190 °C oil bath for
30 min. Methanol (about 5 mL/mmol) was added before the med-
ium was cooled down to rt, and the suspension was sonicated for
maximum solubilization. Water was then added (about 15 mL/
mmol). When a precipitate formed (depending on MeOH and water
volumes added, and nature of aromatic substituents), it contained
the fluorescein derivative and was filtered. Otherwise the mixture
was extracted with DCM. The combined organic layers were
washed with brine, dried over MgSO4, and evaporated to afford a
yellow to dark orange-brown solid residue. In both methods fluo-
rescein derivatives were obtained in their carboxylic form.
For clarity, atom numbering of lactonic forms has been used.
5
5
4
6
4
6
7
same numbering
O3a
8' O
3a
HO2C
7
4. Experimental section
4.1. General
7a
9'
7a
9'
1'
4'
1'
4'
2'
3'
2'
3'
OH
7'
7'
HO 6'
O
OH
HO 6'
O
5'
5'
All chemicals and solvents were purchased from commercial
suppliers (Acros, Aldrich) and used as received. 1H (250.13 or
500.16 MHz) and 13C (62.9 or 125.78 MHz) NMR spectra were re-
corded on an ARX 250 or an Avance II 500 Bruker spectrometers.
Chemical shifts (d, ppm) are given with reference to residual 1H or
13C of deuterated solvents (CDCl3 7.24, 77.00; CD3OD 3.30, 49.0;
D2O 4.80). Product visualization was achieved by thin-layer chroma-
tography and was performed on Merck Silica Gel 60 plates with fluo-
rescent indicator. The plates were visualized with UV light (254 nm)
or with a 3.5% solution of phosphomolybdic acid in ethanol or with
2% (w/v) ninhydrin in ethanol. Column chromatography was per-
formed with Biotage FLASH40i chromatography module (prepacked
cartridge system). Elemental analyses were performed at the ICSN
CNRS, Gif-sur-Yvette, France. Melting points were determined in
capillary tubes on a Büchi apparatus and are uncorrected. IR spectra
of samples were obtained neat with a FT-IR spectrophotometer.
Mass spectra (MS) were recorded with a LCQ-advantage (Thermo-
Finnigan) mass spectrometer with positive (ESI+) or negative (ESIꢀ)
4.2.2.1. 7a-(60-Hydroxy-30-oxo-3H-xanthen-90-yl)benzoic acid
(1a). Method B: 78% yield, yellow solid, mp 320 °C. 1H NMR
(DMSO-d6) d: 10.11 (s, 2H, OH), 8.02 (d, J = 7.3 Hz, 1H, H4), 7.83–
7.70 (m, 2H, H40, H50, H6), 7.27 (d, J = 7.3 Hz, 1H, H7), 6.69 (br s,
2H, H10, H80), 6.56 (m, 4H, H40, H50, H20, H70). 1H NMR (MeOD-d4)
d: 8.20 (d, J = 7.3 Hz, 1H, H4), 7.98–7.85 (m, 2H, H5, H6), 7.40 (d,
J = 7.3 Hz, 1H, H7), 6.87 (d, J = 2.3 Hz, 2H, H10, H80), 6.82 (d,
J = 8.7 Hz, 2H, H20, H70), 6.73 (dd, 2H, H40, H50). 13C (MeOD-d4) d:
171.5 (Cq), 161.2 (Cq), 161.1 (Cq), 154.0 (Cq), 136.4, 131.0, 130.1,
128.2 (Cq), 125.7, 125.3, 113.4, 111.3 (Cq), 103.5. IR 1597 cmꢀ1. MS
(ESI): m/z 331.1 [MꢀH]ꢀ.
4.2.2.2. 4,5,6,7-Tetrachloro-7a-(60-hydroxy-30-oxo-3H-xanthen-
90-yl)benzoic acid (1b). Method A: 76% yield, brown solid, mp
>360 °C. 1H NMR (DMSO-d6) d: 10.22 (s, 2H, OH), 6.94 (d,
J = 8.5 Hz, 2H, H10, H80), 6.73 (d, J = 1.9 Hz, 2H, H40, H50), 6.57 (dd,

6930
N. Pietrancosta et al. / Bioorg. Med. Chem. 18 (2010) 6922–6933
2H, H20, H70). 13C NMR (DMSO-d6) d: 163.5, 159.9, 152.0, 148.3,
138.8, 134.9, 134.7, 130.3, 128.9, 127.2, 124.1, 112.6, 106.5,
102.3, 81.7. MS (ESI): m/z 425.3 [M–CO2H]ꢀ.
der. 1H NMR (DMSO-d6) d: 10.12 (s, 2H, OH), 7.94–7.60 (m, 3H, H7,
H5, H4), 6.69 (br s, 2H, H10, H80), 6.60–6.52 (m, 4H, H40, H50, H20,
H70), 1.40, 1.33, 1.25 (3s, 9H, tBu). MS (ESI): m/z 387.1 [MꢀH]ꢀ;
ESI+: 389.3 [M+H]+.
4.2.2.3. 4,7-Dichloro-7a-(60-hydroxy-30-oxo-3H-xanthen-90-yl)ben-
zoic acid (1c). Method B: 67% yield, yellow solid, mp >360 °C. 1H NMR
(DMSO-d6) d: 10.60 (s, 2H, OH), 7.82 (br s, 1H, H5 or H6), 7.73 (br s, 1H,
H5 or H6), 6.79 (d, J = 1.9 Hz, 2H, H10, H80), 6.69 (br s, 2H, H40, H50), 6.58
(m, 2H, H20, H70). MS (ESI): m/z 401.3 [M+H]+; MS (ESI): m/z 398.9
[MꢀH]ꢀ.
4.2.2.13. 4,5,6,7-Tetrachloro-7a-(90-hydroxy-50-oxo-5H-benzo[a]
xanthen-120-yl)benzoic acid (6). Method B: 27% yield, brown
powder. 1H NMR (DMSO-d6) d: 11.79 (s, 2H, OH), 8.01–7.52 (m,
4H, H90, H100, H110, H120), 6.75 (br s, 1H, H10), 6.60–6.38 (m, 3H,
H20, H40, H50). HPLC tR 15.37 min .MS (ESI): m/z 518.8 [MꢀH]ꢀ.
4.2.2.4. 5,6-Dichloro-7a-(60-hydroxy-30-oxo-3H-xanthen-90-yl)ben-
zoic acid (1d). Method B: 89% yield, yellow solid, mp >360 °C. 1HNMR
(DMSO-d6) d: 10.18 (s, 2H, OH), 8.29 (s, 1H, H4 or H7), 7.76 (s, 1H, H4 or
H7), 6.74–6.69 (m, 4H, H40, H50, H10, H80), 6.58 (dd, J = 8.5 Hz, J = 2.3 Hz,
2H, H20, H70). MS (ESI): m/z 401.3 [M+H]+.
4.2.2.14. 4,5,6,7-Tetrachloro-7a-(90-hydroxy-50-oxo-5H-dibenzo
[a,j]xanthen-140-yl)benzoic acid (7). Method B: 63% yield,
brown powder. 1H NMR (DMSO-d6) d: 12.07 (s, 2H, OH),
8.01–7.32 (m, 8H, H90, H100, H110, H120, H130, H140, H150, H160),
6.65–6.38 (m, 2H, H40, H50). HPLC tR 15.28 min MS (ESI): m/z
569.2 [MꢀH]ꢀ.
4.2.2.5. 4,5,6,7-Tetrafluoro-7a-(60-hydroxy-30-oxo-3H-xanthen-90-
yl)benzoic acid (1e). Method B: 69% yield, brown solid. 1H NMR
(DMSO-d6) d: 10.28 (s, 2H, OH), 7.03 (d, J = 8.1 Hz, 2H, H10, H80), 6.73
(br s, 2H, H40, H50), 6.63 (d, J = 8.1 Hz, 2H, H20, H70). MS (ESI): m/z
405.3 [M+H]+.
4.3. General method for the iodination of fluorescein
derivatives
4.3.1. Method C
4.2.2.6. 4,5,6,7-Tetrabromo-7a-(60-hydroxy-30-oxo-3H-xanthen-
90-yl)benzoic acid (1f). Method B: 60% yield, brown solid. 1H
NMR (DMSO-d6) d: 10.25 (s, 2H, OH), 6.87 (d, J = 8.7 Hz, 2H, H10,
H80), 6.67 (br s, 2H, H40, H50), 6.56 (d, J = 8.7 Hz, 2H, H20, H70). MS
(ESI): m/z 646.8 [MꢀH]ꢀ.
Iodic acid (2 equiv) was dissolved in a minimum of water and
added dropwise to a solution of compound 1 (1 equiv) and iodine
(2.5 equiv) in absolute ethanol. This mixture was stirred over 2 h
during which the dark brown solution slowly turned red orange.
The mixture was then heated to 60 °C for an hour. After cooling,
the mixture was filtered and the red solid washed with water
and ethanol.
4.2.2.7. 7a-(60-Hydroxy-40,50-dimethyl-30-oxo-3H-xanthen-90-yl)ben-
zoic acid (1g). Method B: 76% yield, orange powder. 1H NMR (DMSO-d6)
d: 9.94 (s, 2H, OH), 7.99 (d, J = 7.3 Hz, 1H, H4), 7.83–7.69 (m, 2H, H5, H6),
7.28(d, J = 7.3 Hz, 1H, H1), 6.62 (d, 2H, J = 8.9 Hz,H10,H80), 6.38 (d, 2H, H20,
H70), 2.30 (s, 6H, CH3). MS (ESI): m/z 359.2 [MꢀH]ꢀ.
4.3.1.1. 7a-(60-Hydroxy-20,40,50,70-tetraiodo-30-oxo-3H-xanthen-
90-yl)benzoic acid (2a). MethodC: 59% yield, red powder, mp 290–
292 °C. 1H NMR (DMSO-d6) d: 10.24 (br s, 2H, OH), 8.18 (d, J = 7.3 Hz,
1H, H4), 7.86–7.78(m, 2H, H5, H6),7.46(d, J = 7.3 Hz, 1H, H7), 7.27(s,
1H, H10 or H80), 7.02 (s, 1H, H10 or H80). HPLC tR 13.35 min. MS (ESI):
m/z 834.7 [MꢀH]ꢀ; MS (ESI): m/z 836.8 [MꢀH]+.
4.2.2.8. 7a-(60-Hydroxy-30-oxo-3H-xanthen-90-yl)-7-nitrobenzoic
acid and2-(60-hydroxy-30-oxo-3H-xanthen-90-yl)-4-nitrobenzoic
acid (1h). Method B: 63% yield, yellow powder. 1H NMR (MeOD-
d4) d: 8.08 (d, J = 7.8 Hz, 1H, H4), 7.93 (t, J = 7.8 Hz, 1H, H5), 7.50 (d,
J = 7.8 Hz, 1H, H6), 6.70 (d, J = 2.3 Hz, 2H, H10, H80), 6.38 (d,
J = 8.4 Hz, 2H, H40, H50), 6.58 (dd, 2H, H20, H70). MS (ESI): m/z
376.2 [MꢀH]ꢀ.
4.3.1.2. 4,5,6,7-Tetrachloro-7a-(60-hydroxy-20,40,50,70-tetraiodo-
30-oxo-3H-xanthen-90-yl)benzoic acid (2b). Method C: 43%
yield, red powder, mp 297–300 °C. 1H NMR (MeOD-d4) d: 7.59 (s,
2H, H10, H80). HPLC tR 15.80 min. MS (ESI): m/z 972.4 [MꢀH]ꢀ;
MS (ESI): m/z 974.6 [MꢀH]+.
4.2.2.9.7a-(60-Hydroxy-30-oxo-3H-xanthen-90-yl)terephthalicacid
2-(60-hydroxy-30-oxo-3H-xanthen-90-yl)isophthalic acid
4.3.1.3. 4,7-Dichloro-7a-(60-hydroxy-20,40,50,70-tetraiodo-30-oxo-
3H-xanthen-90-yl)benzoic acid (2c). Method C: 49% yield, red
powder, mp >320 °C. 1H NMR (DMSO-d6) d: 10.25 (s, 2H, OH),
7.86 (br s, 2H), 7.43 (br s, 2H). HPLC tR 13.97 min. MS (ESI): m/z
902.5 [MꢀH]ꢀ; MS (ESI): m/z 904.7 [MꢀH]+.
and
(1i). Method B: 51% yield, orange powder, mixture of two isomers,
mp >360 °C. 1H NMR (DMSO-d6) d: 13.55 (br s, 2H, COOH), 10.27 (s,
4H, OH), 8.41–7.39 (m, 6H, H4, H5, H6, H7), 6.75 (br s, 4H, H10, H80),
6.64–6.55 (m, 8H, H40, H50, H20, H70). MS (ESI): m/z 375.1 [MꢀH]ꢀ.
4.2.2.10. 4-Hydroxy-7a-(60-hydroxy-30-oxo-3H-xanthen-90-yl)ben-
zoic acid (1j). Method B: 39% yield, dark yellow powder. 1H NMR
(DMSO-d6) d: 10.93 (s, 1H, OH), 10.05 (s, 2H, OH), 7.53 (t, J = 7.8 Hz,
1H, H6), 7.00 (d, J = 7.8 Hz, 1H, H7), 6.65–6.52 (m, 7H, H5, H40, H50,
H20, H70, H10, H80). MS (ESI): m/z 347.0 [MꢀH]ꢀ; ESI+: 349.2 [M+H]+.
4.3.1.4. 5,6-Dichloro-7a-(60-hydroxy-20,40,50,70-tetraiodo-30-oxo-
3H-xanthen-90-yl)benzoic acid (2d). Method C: 48% yield, red
powder, mp >320 °C. 1H NMR (CDCl3) d: 8.16 (s, 1H, H4), 7.25 (s,
2H, H10, H80), 7.14 (s, 1H, H7). HPLC tR 14.80 min. MS (ESI): m/z
902.6 [MꢀH]ꢀ; MS (ESI): m/z 904.7 [MꢀH]+.
4.2.2.11.
7a-(60-Hydroxy-30-oxo-3H-xanthen-90-yl)-5-methyl-
4.3.1.5. 4,5,6,7-Tetrafluoro-7a-(60-hydroxy-20,40,50,70-tetraiodo-
30-oxo-3H-xanthen-90-yl)benzoic acid (2e). Method C: 56% yield,
red powder, mp 294 °C. 1H NMR (DMSO-d6) d: 10.33 (s, 2H, OH),
7.66 (s, 2H, H10, H80). HPLC tR 12.07 min. MS (ESI): m/z 906.7
[MꢀH]ꢀ; ESI+: 908.6 [MꢀH]+.
benzoic acid and 2-(60-hydroxy-30-oxo-3H-xanthen-90-yl)-6-
methylbenzoic acid (1k). Method B: 87% yield, orange-red pow-
der. 1H NMR (DMSO-d6) d: 10.13 (s, 2H, OH), 7.87 (d, 1H, H7 or H4),
7.73–7.53 (m, 1H, H5), 7.14 (d, 1H, H7 or H4), 6.74 (br s, 2H, H10,
H80), 6.67–6.62 (m, 4H, H40, H50, H20, H70), 2.42 (s, 3H, CH3). MS
(ESI): m/z 345.1 [MꢀH]ꢀ; ESI+: 347.3 [M+H]+.
4.3.1.6. 4,5,6,7-Tetrabromo-7a-(60-hydroxy-20,40,50,70-tetraiodo-
30-oxo-3H-xanthen-90-yl)benzoic acid (2f). Method C: 37% yield,
red powder, mp 284–286 °C. 1H NMR (DMSO-d6) d: 10.22 (s, 2H,
OH), 7.51 (s, 2H, H10, H80). HPLC tR 14.53 min. MS (ESI): m/z
1150.2 [MꢀH]ꢀ.
4.2.2.12.
5-tert-Butyl-7a-(60-hydroxy-30-oxo-3H-xanthen-90-yl)
benzoic acid and 6-tert-butyl-2-(60-hydroxy-30-oxo-3H-xanthen-
90-yl)benzoic acid (1l). Method B: quantitative yield, yellow pow-

N. Pietrancosta et al. / Bioorg. Med. Chem. 18 (2010) 6922–6933
6931
4.3.1.7.
7a-(60-Hydroxy-20,70-diiodo-40,50-dimethyl-30-oxo-3H-
carbonate (2.2 equiv) at 0 °C. Methyl iodide (2.2 equiv) was then
added dropwise, the mixture was warmed to room temperature
and stirred for 17 h. Residual methyl iodide was quenched by a
mixture of ice/water and extracted with AcOEt/H2O. The organic
layer was dried over MgSO4, filtered and evaporated to give a violet
solid (31.3 mg, 24% yield). HPLC tR 10.35 min. MS (ESI): m/z 986.5
[MꢀH]ꢀ.
xanthen-90-yl)benzoic acid (2g). Method C: 37% yield, red pow-
der, decomp., mp 190 °C. HPLC tR 17.87 min. MS (ESI): m/z 610.7
[MꢀH]ꢀ.
4.3.1.8. 7a-(60-Hydroxy-20,40,50,70-tetraiodo-30-oxo-3H-xanthen-
90-yl)-4-nitrobenzoic acid and 7a-(60-hydroxy-20,40,50,70-tetraiod-
o-30-oxo-3H-xanthen-90-yl)-7-nitrobenzoic acid (2h). Method C:
45% yield, red powder, mp >320 °C. HPLC tR 11.82 min. MS (ESI):
m/z 879.7 [MꢀH]ꢀ; MS (ESI): m/z 881.7 [MꢀH]+.
4.3.3.2. 4,5,6,7-Tetrachloro-20,40,50,70-tetraiodo-30,60-dimethoxy-
3H-spiro[isobenzofuran-1,90-xanthen]-3-one (5). Compound 4
(0.13 mmol, 1 equiv) was suspended in 5 mL DMF with potassium
carbonate (2.2 equiv) at 0 °C. Methyl iodide (2.2 equiv) was then
added dropwise; the mixture was warmed to room temperature
and stirred for 17 h. K2CO3 (5 equiv) was then added and the mix-
ture was stirred for two more days. Residual methyl iodide was
quenched by a mixture of ice/water and extracted with AcOEt/
H2O. The organic layer was dried and evaporated to afford a violet
solid (13.5 mg, 42.6% yield). HPLC tR 7.36 min. MS (ESI): m/z 1000.6
[M+H]+.
4.3.1.9. 7a-(60-Hydroxy-20,40,50,70-tetraiodo-30-oxo-3H-xanthen-
90-yl)terephthalic acid and 7a-(60-hydroxy-20,40,50,70-tetraiodo-
30-oxo-3H-xanthen-90-yl)isophthalic acid (2i). Method C: 56%
yield, red powder, mp >320 °C. HPLC tR 12.61 min. MS (ESI): m/z
878.7 [MꢀH]ꢀ; MS (ESI): m/z 880.7 [MꢀH]+.
4.3.1.10. 4-Hydroxy-7a-(60-hydroxy-20,40,50,70-tetraiodo-30-oxo-
3H-xanthen-90-yl)benzoic acid (2j). Method C: 21% yield, red
powder, mp >320 °C. HPLC tR 12.42 min. MS (ESI): m/z 850.6
[MꢀH]ꢀ.
4.4. Subcellular fractionation and amino acid uptake assay
4.3.1.11. 7a-(60-Hydroxy-20,40,50,70-tetraiodo-30-oxo-3H-xanthen-
90-yl)-5-methylbenzoic acid and 7a-(60-hydroxy-20,40,50,70-tetraiod-
o-30-oxo-3H-xanthen-90-yl)-6-methylbenzoic acid (2k). Method
C: 43% yield, red powder, mp >320 °C. HPLC tR 13.05 min. MS
(ESI): m/z 848.7 [MꢀH]ꢀ; MS (ESI): m/z 850.8 [MꢀH]+.
Rats or mice were killed by decapitation. Brains were removed
and the striatum, cerebral cortex, hippocampus and brainstem were
dissected on ice and resuspended in ice-cold sucrose (0.32 M).
Striatal synaptic vesicles were purified as described.90 Synaptic ves-
icle pellets from 30 rat brains or 61 mouse brains were resuspended
in ice-cold 0.32 M sucrose, 4 mM KCl, 4 mM MgSO4 and 10 mM
HEPES–KOH, pH 7.4 (uptake buffer).
4.3.1.12. 5-tert-Butyl-7a-(60-hydroxy-20,40,50,70-tetraiodo-30-oxo-
3H-xanthen-90-yl)benzoic acid and 6-tert-butyl-7a-(60-hydroxy-
20,40,50,70-tetraiodo-30-oxo-3H-xanthen-90-yl)benzoic acid (2l).
Method C: 39% yield, red powder, mp >320 °C. HPLC tR
14.62 min. MS (ESI): m/z 890.7 [MꢀH]ꢀ; MS (ESI): m/z 892.8
[MꢀH]+.
4.5. Vesicular neurotransmitter transport assay
The Vesicular Glutamate uptake was assayed by a modification of
the method of Naito and Ueda25 as described by Kish and Ueda.91 Ali-
quots (10
10 min at 30 °C in a solution containing 20 mM HEPES–KOH, pH
7.4, 0.25 mM sucrose, 4 mM MgSO4, 4 mM KCl, 2 mM -aspartic acid,
50
M Glu, 2 mM ATP, and 2 mCi (40 mM) [3H]Glu (Amersham) in a
final volume of 0.1 mL. Test compounds were dissolved in DMSO and
added in 0.6 L aliquots to give final concentrations between
0.14 nM and 25 M. Synaptic vesicles were preincubated with test
compounds for 30 min at 0 °C. Uptake was initiated by addition of
a mixture A of 2 mM ATP and 50 M Glu (final concentration) con-
taining 1
Ci of [3H]Glu. After 10 min at 30 °C, 2.5 mL of ice-cold
0.15 m KCl were added and the mixture immediately filtered
through Whatman GF/C glass–fiber filter paper (25 mm) using a
manifold under vacuum. The filters were washed five times with
2.5 mL of 0.15 M KCl and placed in scintillation vials with scintilla-
tion cocktail (Cytoscint, ICN). The vials were shaken overnight and
radioactivity determined in a Beckman LS 6500 scintillation spectro-
photometer. Values obtained in the absence of ATP were subtracted
from those in the presence of ATP to calculate ATP dependent uptake
activity. The amount of radioactive Glu trapped on GF/C glass–fiber
filter paper was indistinguishable from that trapped on the Millipore
HAWP filter (data not shown). Since the rate of filtration with GF/C
glass-fiber filter paper is greater, this type of filter was used in all
experiments described here.
lg) of crude rat synaptic vesicles were incubated for
4.3.2. Method D. Esterification
4.3.2.1. Ethyl 4,5,6,7-tetrachloro-7a-(60-hydroxy-20,40,50,70-tetrai-
odo-30-oxo-3H-xanthen-90-yl)benzoate (3). Compound 2b (0.06
mmol, 1 equiv) was suspended in 10 mL THF. Ethylene bromide
L
l
solution (1 equiv in 250
DBU solution (1 equiv in 50
l
L of THF) was then added as well as a
L). The solution was heated at 65 °C
l
l
l
overnight. The mixture was extracted with DCM/H2O, the organic
layer was dried and purified by extraction with aq. HCl/DCM. A
red powder was obtained in a low yield (15.6 mg, 26.5%). HPLC tR
18.12 min. MS (ESI): m/z 1000.6 [MꢀH]ꢀ.
l
l
4.3.2.2. Ethyl 2-(20,70-dibromo-60-hydroxy-40-50-dinitro-30-oxo-
3H-xanthen-90-yl)benzoate (8). Eosin B (BU8, 580 mg, 1 mmol)
and potassium hydroxide (200 mg, 3.5 mmol) were dissolved in
distilled THF (10 mL). The reaction mixture was stirred under ar-
gon for 30 min. Then ethyl bromide (272.4 mg, 2.5 mmol) and
DBU (372 lL, 2.5 mmol) were added and the reaction was heated
under argon at 70 °C for 72 h. After cooling down to room temper-
ature, the reaction mixture was extracted with ethyl acetate
(2 ꢂ 100 mL). The organic layer was washed with two portions of
brine and dried over anhydrous MgSO4. The solvent was evapo-
rated at reduced pressure. The red crude product was purified by
PLC (Silica Gel 60 F254, 2 mm). Elution with DCM/MeOH (90/10)
afforded 8 as a red-violet powder (0.284 g, yield 46.7%). HPLC tR
3.27 min. MS (ESI): m/z 606.9 [MꢀH]ꢀ. Mp >320 °C.
The glutamate uptake was also performed on 96-well plate
(Millipore MultiScreenHTS-HA Plate) using MultiScreen-HTS Vac-
uum Manifold as filtration support, all the previous concentrations
were conserved and we processed as follow: first, a pre-incubation
4.3.3. Method E. Etherification
of synaptic vesicles with/without inhibitors in a total of 50
buffer was performed; next, uptake was initiated by addition of
solution A containing ATP and1
Ci of [3H]Glu. The uptake was
stopped by addition of 0.15 M of ice-cold KCl (each well was
lL of
4.3.3.1. 4,5,6,7-Tetrachloro-7a-(20,40,50,70-tetraiodo-60-methoxy-
30-oxo-3H-xanthen-90-yl)benzoic acid (4). Rose Bengal lactone
(0.13 mmol, 1 equiv) was suspended in 5 mL DMF with potassium
l

6932
N. Pietrancosta et al. / Bioorg. Med. Chem. 18 (2010) 6922–6933
7. Burger, P. M.; Mehl, E.; Cameron, P. L.; Maycox, P. R.; Baumert, M.; Lottspeich,
F.; Decamilli, P.; Jahn, R. Neuron 1989, 3, 715.
8. Meldrum, B. Clin. Sci. 1985, 68, 113.
washed five times). The same controls were used for the evaluation
of inhibitory efficiency.
Other neurotransmitter uptake was performed with similar
protocols. Briefly, Rat synaptic vesicles are incubated with appro-
priated neurotransmitter, buffer and inhibitor. The uptake is then
stopped by addition of cold potassium chloride.62 The residual
radioactivity is measured and the inhibition is calculated from 0%
given by uptake without inhibitor and 100% given by the radioac-
tivity measure in absence of ATP.
9. Choi, D. W. Neuron 1988, 1, 623.
10. Javitt, D. C.; Zukin, S. R. Am. J. Psychiatry 1991, 148, 1301.
11. Zorumski, C. F.; Olney, J. W. Pharmacol. Ther. 1993, 59, 145.
12. Lipton, S. A.; Rosenberg, P. A. N. Eng. J. Med. 1994, 330, 613.
13. Bradford, H. F. Prog. Neurobiol. 1995, 47, 477.
14. Olney, J. W.; Farber, N. B. Arch. Gen. Psychiat. 1995, 52, 998.
15. Coyle, J. T. Harvard Rev. Psychiatry 1996, 3, 241.
16. Chapman, A. G. Prog. Brain Res. 1998, 116, 371.
17. Lepicard, E.; Poirel, O.; Kashani, A.; Amilhon, B.; Hirsch, E.; Epelbaum, J.; David,
J. P.; Delacourte, J.; Giros, B.; Betancur, C.; El Mestikawy, S. Eur.
Neuropsychopharmacol. 2007, 17, S534.
18. Bellocchio, E. E.; Reimer, R. J.; Fremeau, R. T.; Edwards, R. H. Science 2000, 289,
957.
4.6. VGLUT pharmacophore model generation
Molecule sketching, conformation generation, feature edition
and pharmacophore model generation were performed using Dis-
covery Studio version 2.5 (Accelrys Inc., San Diego CA) which
launched the suitable Catalyst standalone executables via a Pipe-
linePilot server.
19. Bai, L. Q.; Xu, H.; Collins, J. F.; Ghishan, F. K. J. Biol. Chem. 2001, 276, 36764.
20. Herzog, E.; Bellenchi, G. C.; Gras, C.; Bernard, V.; Ravassard, P.; Bedet, C.;
Gasnier, B.; Giros, B.; El Mestikawy, S. J. Neurosci. 2001, 21, 22.
21. Gras, C.; Herzog, E.; Bellenchi, G. C.; Bernard, V.; Ravassard, P.; Pohl, M.;
Gasnier, B.; Giros, B.; El Mestikawy, S. J. Neurosci. 2002, 22, 5442.
22. Kashani, A.; Lepicard, E.; Poirel, O.; Videau, C.; David, J. P.; Fallet-Bianco, C.;
Simon, A.; Delacourte, A.; Giros, B.; Epelbaum, J.; Betancur, C.; El Mestikawy, S.
Neurobiol. Aging 2008, 29, 1619.
4.6.1. Statistical significance
23. Takamori, S. Neurosci. Res. 2006, 55, 343.
24. Francis, P. T. Int. J. Geriatr. Psychiatry 2003, 18, S15.
25. Naito, S.; Ueda, T. J. Neurochem. 1985, 44, 99.
26. Winter, H. C.; Ueda, T. Neurochem. Res. 1993, 18, 79.
27. Carlson, M. D.; Kish, P. E.; Ueda, T. J. Neurochem. 1989, 53, 1889.
28. Moriyama, Y.; Amakatsu, K.; Yamada, H.; Park, M. Y.; Futai, M. Arch. Biochem.
Biophys. 1991, 290, 233.
To evaluate if the SAR content of the training set was indeed re-
flected by the selected model, a Fischer test was performed by
shuffling the measured activity values of the training set 49 times
(Catalyst-CatScramble) and running Catalyst-Hypogen with these
artificial datasets. Since none of the models generated with the
49 artificial datasets could perform better than the original set, it
was concluded that the significance of the model obtained with
our training set was at least of 98%.
29. Carrigan, C. N.; Esslinger, C. S.; Bartlett, R. D.; Bridges, R. J.; Thompson, C. M.
Bioorg. Med. Chem. Lett. 1999, 9, 2607.
30. Roseth, S.; Fykse, E. M.; Fonnum, F. J. Neurochem. 1995, 65, 96.
31. Roseth, S.; Fykse, E. M.; Fonnum, F. Biochem. Pharmacol. 1998, 56, 1243.
32. Moriyama, Y.; Yamamoto, A. J. Biol. Chem. 1995, 270, 22314.
33. Bole, D. G.; Ueda, T. Neurochem. Res. 2005, 30, 363.
34. Bartlett, R. D.; Esslinger, C. S.; Thompson, C. M.; Bridges, R. J.
Neuropharmacology 1998, 37, 839.
35. Carrigan, C. N.; Bartlett, R. D.; Esslinger, C. S.; Cybulski, K. A.;
Tongcharoensirikul, P.; Bridges, R. J.; Thompson, C. M. J. Med. Chem. 2002, 45,
2260.
36. Ogita, K.; Hirata, K.; Bole, D. G.; Yoshida, S.; Tamura, Y.; Leckenby, A. M.; Ueda,
T. J. Neurochem. 2001, 77, 34.
37. von Bayer, A. Chem. Ber. 1871, 5, 255.
38. Burdette, S.; Walkup, G.; Spingler, B.; Tsien, R.; Lippard, S. J. Am. Chem. Soc.
2001, 123, 7831.
39. Sun, W. C.; Gee, K. R.; Klaubert, D. H.; Haugland, R. P. J. Org. Chem. 1997, 62,
6469.
4.6.2. ROC curve analysis
A ROC curve analysis92 was performed in order to evaluate the
ability of the model to correctly discriminate the most active mol-
ecules (below 900nM) from the least active ones (above 900nM).
The selected model exhibited areas under the curve (AUCs) of
0.94 and 0.86 when the training set and the whole dataset were,
respectively, considered. Last a Fisher test was performed on the
overall dataset to assess the likelihood that the ROC performance
of the retained model was due to shear chance. Hence, by shuffling
99 times the measured activities within the whole dataset one can
calculate 99 ROC AUCs. Since none of the determined AUC was
higher than 0.86, the Fischer statistical test allowed to assert that
the significance of the model is at least 99%.
40. Ueno, Y.; Jiao, G. S.; Burgess, K. Synthesis (Stuttgart) 2004, 15, 2591.
41. Schmidt, H. W. Synthesis (Stuttgart) 1985, 8, 778.
42. Hirano, T.; Kikuchi, K.; Urano, Y.; Nagano, T. J. Am. Chem. Soc. 2002, 124, 6555.
43. Lukhtanov, E. A.; Vorobiev, A. V. J. Org. Chem. 2008, 73, 2424.
44. Rossi, F. M.; Kao, J. P. Y. Bioconjugate Chem 1997, 8, 495.
45. Woodroofe, C. C.; Lim, M. H.; Bu, W. M.; Lippard, S. J. Tetrahedron 2005, 61,
3097.
46. Lyttle, M. H.; Carter, T. G.; Cook, R. M. Org. Process Res. Dev. 2001, 5, 45.
47. Coan, K. E. D.; Shoichet, B. K. J. Am. Chem. Soc. 2008, 130, 9606.
48. Klonis, N.; Sawyer, W. H. Photochem. Photobiol. 2000, 72, 179.
49. Alvarez-Pez, J. M.; Ballesteros, L.; Talavera, E.; Yguerabide, J. J. Phys. Chem. A
2001, 105, 6320.
50. Hwang, S. G.; Jang, Y. H.; Chung, D. S. Chem. Lett. 2001, 11, 1182.
51. Anthoni, U.; Christophersen, C.; Nielsen, P. H.; Puschl, A.; Schaumburg, K.
Struct. Chem. 1995, 6, 161.
52. Lavis, L. D.; Rutkoski, T. J.; Raines, R. T. Anal. Chem. 2007, 79, 6775.
53. Samoilov, D. V.; Mchedlov-Petrosyan, N. O.; Martynova, V. P.; El’tsov, A. V. Russ.
J. Gen. Chem. 2000, 8, 1259.
Acknowledgments
This work was supported by a grant from the Agence Nationale
pour la Recherche (Grant ANR-06-NEURO-048-02). The authors
thank Drs. M. Hamon, L. Lanfumey and R. Mongeau from U677
for their help and availability, Drs. R. Azerad and M. Maurs
(UMR8601)for helpful advice in chromatographyand Drs. B. Gasnier
and I. McCort for critical reading of the manuscript.
54. Mchedlov-Petrossyan, N. O.; Vodolazkaya, N. A.; Surov, Y. N.; Samoylov, D. V.
Spectrochim. Acta A 2005, 61, 2747.
55. Brenzovich, W. E.; Odell, M. D.; Soares, J. P.; Abelt, C. J. Dyes Pigments 2003, 59,
251.
Supplementary data
56. Tamura, Z.; Terada, R.; Ohno, K.; Maeda, M. Anal. Sci. 1999, 15, 339.
57. Xu, D.; Vanloon, A.; Linden, S. M.; Neckers, D. C. J. Photochem. 1987, 38, 357.
58. Amatguerri, F.; Lopezgonzalez, M. M. C.; Martinezutrilla, R.; Sastre, R. Dyes
Pigments 1990, 12, 249.
Supplementary data associated with this article can be found, in
the online version, at doi:10.1016/j.bmc.2010.06.069.
59. Klonis, N.; Clayton, A. H. A.; Voss, E. W.; Sawyer, W. H. Photochem. Photobiol.
1998, 67, 500.
References and notes
60. Mchedlov-Petrosyan, N. O.; Rubtsov, M. I.; Lukatskaya, L. L. Russ. J. Gen. Chem.
2000, 8, 1177.
61. Mchedlov-Petrosyan, N. O.; Vodolazkaya, N. A.; Martynova, V. P.; Samoilov, D.
V.; El’tsov, A. V. Russ. J. Gen. Chem. 2002, 5, 785.
62. Gras, C.; Amilhon, B.; Lepicard, E. M.; Poirel, O.; Vinatier, J.; Herbin, M.; Dumas,
S.; Tzavara, E. T.; Wade, M. R.; Nomikos, G. G.; Hanoun, N.; Saurini, F.; Kemel,
M. L.; Gasnier, B.; Giros, B.; El Mestikawy, S. Nat. Neurosci. 2008, 11, 292.
63. Carreon, J. R.; Roberts, M. A.; Wittenhagen, L. M.; Kelley, S. O. Org. Lett. 2005, 7,
99.
1. Thompson, C. M.; Davis, E.; Carrigan, C. N.; Cox, H. D.; Bridges, R. J.; Gerdes, J. M.
Curr. Med. Chem. 2005, 12, 2041.
2. Shigeri, Y.; Seal, R. P.; Shimamoto, K. Brain Res. Rev. 2004, 45, 250.
3. Nicholls, D. G. J. Neurochem. 1989, 52, 331.
4. Maycox, P. R.; Deckwerth, T.; Jahn, R. EMBO J. 1990, 9, 1465.
5. Ozkan, E. D.; Ueda, T. Jpn. J. Pharmacol. 1998, 77, 1.
6. Stormmathisen, J.; Leknes, A. K.; Bore, A. T.; Vaaland, J. L.; Edminson, P.; Haug, F.
M. S.; Ottersen, O. P. Nature 1983, 301, 517.

N. Pietrancosta et al. / Bioorg. Med. Chem. 18 (2010) 6922–6933
6933
64. Sugita, N.; Kawabata, K.; Sasaki, K.; Sakata, I.; Umemura, S. Bioconjugate Chem.
2007, 18, 866.
78. Neslund, G. G.; Miara, J. E.; Kang, J. J.; Dahms, A. S. Curr. Top Cell. Regul. 1984, 24,
447.
65. Lu, Y. X.; Shi, T.; Wang, Y.; Yang, H. Y.; Yan, X. H.; Luo, X. M.; Jiang, H. L.; Zhu, W.
L. J. Med. Chem. 2009, 52, 2854.
79. Ogan, J. T.; Reifenberger, M. S.; Milanick, M. A.; Gatto, C. Blood Cell Mol. Dis.
2007, 38, 229.
66. Smellie, A.; Kahn, S. D.; Teig, S. L. J. Chem. Inf. Comput. Sci 1995, 35, 285.
67. Smellie, A.; Teig, S. L.; Towbin, P. J. Comput. Chem. 1995, 16, 171.
68. Li, H.; Sutter, J.; Hoffmann, R., HypoGen: an automated system for generating
3D predictive pharmacophore models. In Pharmacophore Perception,
Development, and Use in Drug Design, Guner, O. F. Ed. International University
Line: La Jolla, CA: 2000; Vol. 2, pp 171.
80. Hopkins, S. C.; Vale, R. D.; Kuntz, I. D. Biochemistry-Us 2000, 39, 2805.
81. McGovern, S. L.; Caselli, E.; Grigorieff, N.; Shoichet, B. K. J. Med. Chem. 2002, 45,
1712.
82. Bachmann, A.; Russ, U.; Waldegger, S.; Quast, U. Br. J. Pharmacol. 2000, 131, 433.
83. McGregor, M. J. J. Chem. Inf. Model. 2007, 47, 2374.
84. Knight, Z. A.; Shokat, K. M. Chem. Biol. 2005, 12, 621.
85. Xie, Q. Q.; Xie, H. Z.; Ren, J. X.; Li, L. L.; Yang, S. Y. J. Mol. Graphics Modell. 2009,
27, 751.
69. Chaudhry, F. A.; Edwards, R. H.; Fonnum, F. Annu. Rev. Pharmacol. Toxicol. 2008,
48, 277.
70. Amilhon, B.; El Mestikawy, S. M S-Med. Sci. 2008, 24, 1009.
71. Fredriksson, R.; Nordstrom, K. J.; Stephansson, O.; Hagglund, M. G.; Schioth, H.
B. FEBS Lett. 2008, 582, 3811.
86. Baldwin, I.; Bamborough, P.; Haslam, C. G.; Hunjan, S. S.; Longstaff, T.; Mooney,
C. J.; Patel, S.; Quinn, J.; Somers, D. O. Bioorg. Med. Chem. Lett. 2008, 18, 5285.
87. Noble, M. E. M.; Endicott, J. A.; Johnson, L. N. Science 2004, 303, 1800.
88. Ishida, A.; Noda, Y.; Ueda, T. Neurochem. Res. 2009, 34, 807.
89. Reyes-Irisarri, E.; Perez-Torres, S.; Mengod, G. Neuroscience 2005, 132, 1173.
90. Huttner, W. B.; Schiebler, W.; Greengard, P.; De Camilli, P. J. Cell Biol. 1983, 96,
1374.
72. Parsons, S. M. FASEB J. 2000, 14, 2423.
73. Logan, W. J.; Swanson, J. M. Science 1979, 206, 363.
74. Mautner, H. G.; Merrill, R. E.; Currier, S. F.; Harvey, G. J. Med. Chem. 1981, 24,
1534.
75. Deweille, J. R.; Muller, M.; Lazdunski, M. J. Biol. Chem. 1992, 267, 4557.
76. Cai, Z. W.; Sheppard, D. N. J. Biol. Chem. 2002, 277, 19546.
77. Karlish, S. J. D. J. Bioenerg. Biomembr. 1980, 12, 111.
91. Kish, P. E.; Ueda, T. Method Enzymol. 1989, 174, 9.
92. Triballeau, N.; Acher, F.; Brabet, I.; Pin, J. P.; Bertrand, H. O. J. Med. Chem. 2005,
48, 2534.