Combinatorial discovery of fluorescent pharmacophores by multicomponent reactions in droplet arrays

Article abstract of DOI:10.1021/ja204016e

Fluorescence imaging in clinical diagnostics and biomedical research relies to a great extent on the use of small organic fluorescent probes. Because of the difficulty of combining fluorescent and molecular-recognition properties, the development of such

Full text of DOI:10.1021/ja204016e

COMMUNICATION
pubs.acs.org/JACS
Combinatorial Discovery of Fluorescent Pharmacophores
by Multicomponent Reactions in Droplet Arrays
Olga N. Burchak,† Laurent Mugherli,† Mariano Ostuni,‡ Jean Jacques Lacapꢀere,‡ and Maxim Y. Balakirev*,†
†Biomics Laboratory, Institut de Recherches en Technologies et Sciences pour le Vivant, CEA, 17 rue des Martyrs, 38054 Grenoble,
France
‡Unitꢁe INSERM U773, Centre de Recherche Biomꢁedicale Bichat Beaujon CRB3, Universitꢁe Paris 7 Denis Diderot, BP 416, F-75018
Paris, France
S Supporting Information
b
neglected or considered an undesirable “false positive” in drug
screening.1,3 Such molecules could serve as starting points for
ABSTRACT: Fluorescence imaging in clinical diagnostics
and biomedical research relies to a great extent on the use of
small organic fluorescent probes. Because of the difficulty of
combining fluorescent and molecular-recognition proper-
ties, the development of such probes has been severely
restricted to a number of well-known fluorescent scaffolds.
Here we demonstrate that autofluorescing druglike mol-
ecules are a valuable source of bioimaging probes. Combi-
natorial synthesis and screening of chemical libraries in
droplet microarrays allowed the identification of new types
of fluorophores. Their concise and clean assembly by a
multicomponent reaction presents a unique potential for
the one-step synthesis of thousands of structurally diverse
fluorescent molecules. Because they are based upon a
druglike scaffold, these fluorophores retain their molecular
recognition potential and can be used to design specific
imaging probes.
combinatorial searches for new fluorescent scaffolds.
Imidazo[1,2-a]pyridines are naturally occurring synthetic
compounds that are widely used in medicinal chemistry. In our
screens for inhibitors of the NS3/4A serine protease of the
hepatitis C virus,4 some of the imidazo[1,2-a]pyridines auto-
fluoresced in UV and interfered with the enzymatic assay.
Because imidazo[1,2-a]pyridine-related heterocycles can be pro-
duced by the three-component Ugi reaction (3-CR) of aromatic
amidines with aldehydes and isocyanides,5 we screened a panel of
3-CRs to determine whether some would yield fluorophores with
fluorescence in the visible region. One advantage of MCRs is the
relative ease of their parallelization, which allows an array of
reactions to be performed in a microplate or on a planar solid
support.6 We set up 3-CRs in 100 nL droplets aligned on a glass
slide with a surface-tension microarray.4 Each microarray had
800 hydrophilic spots patterned on a hydrophobic surface to
maintain the positions of the droplets.4 In parallel, we found that
volatile isocyanides could be added to the arrayed aminopyridineꢀ
aldehyde mixtures via adsorption from the gas phase, which
avoided the need to dispense these malodorous, potentially toxic
compounds. Five 3-CRs were examined to validate synthesis on
the patterned support [Figure S1 in the Supporting Information
(SI)]. Each reaction was conducted on an individual slide with
100 nL droplets of 1:1 dimethyl sulfoxide (DMSO)/water
containing an aminopyridineꢀaldehyde mixture (each 2 mM)
and scandium triflate as the Lewis acid catalyst (5% mol). The array
was placed in a hermetic chamber saturated with isocyanide vapor;
afterward, the droplets were combined and analyzed by high-
performance liquid chromatography (HPLC) and electrospray
ionization mass spectrometry (ESI-MS). At 20 h, the yield of
aminoimidazo[1,2-a]pyridines was generally sufficient (>10%) to
detect the formation of fluorescent compounds (Figure S1).
To discover new fluorophores, eight heterocyclic amidines
(A1ꢀA8), 40 aldehydes (B1ꢀB40), and five isocyanides
(C1ꢀC5) were mixed in arrayed droplets to provide 1600
unique combinations (Figure 1a and Figure S2a). Each glass
slide contained 8 ꢁ 40 = 320 aminopyridineꢀaldehyde mixtures
spotted in duplicate. The slides were then exposed to an
isocyanide, and the fluorescence was analyzed with a microarray
scanner at four different filter settings (Figure 1b and Figure S2b).
luorescent labeling of biologically active molecules is widely
F
employed to study their functions within cells and whole
organisms.1 In drug discovery and chemical genetics, fluorescent
tagging of a pharmacophore allows its intracellular visualization
and identification of its potential targets. Even though fluorescent
labels are relatively small, their incorporation can change the
properties of the pharmacophore, thereby limiting their applica-
tion. The ideal pharmacophore should display its own specific
fluorescent signature to allow “label-free” visualization. Such a
marriage of molecular-recognition and fluorescent properties
within one molecule would create a valuable source of imaging
probes for biomarker discovery. One way to achieve this goal is to
explore the chemical space around the known fluorophore cores
to search for fluorescent druglike molecules with specific recog-
nition properties.2 This approach is illustrated in the diversity-
oriented fluorescence libraries developed by Chang and co-
workers,2 which have been successfully used to create fluorescent
probes for various biological targets. Although combinatorial
synthesis based on a known fluorophore core avoids the synthetic
challenge of discovering fluorophores de novo, it also limits the
structural diversity of the chemical library to specific fluorescent
scaffolds that do not necessarily have druglike properties. An-
other approach is to transform the druglike scaffolds into
fluorescent molecules. In fact, many natural and synthetic
pharmacophores exhibit intrinsic fluorescence that is often
Received: May 2, 2011
Published: June 06, 2011
r
2011 American Chemical Society
10058
dx.doi.org/10.1021/ja204016e J. Am. Chem. Soc. 2011, 133, 10058–10061
|

Journal of the American Chemical Society
COMMUNICATION
Figure 1. Combinatorial screening of fluorophores with a surface-tension microarray. (a) Synthesis of the chemical library by the three-component Ugi
reaction (3-CR). (b) Representative fluorescent scanner images of the 3-CR screen showing the reactions of A1ꢀA4, the four positional isomers of
carboxymethyl-2-aminopyridine, with the 40 aldehydes B1ꢀB40 and cyclohexyl isocyanide (C1). Microarray fluorescence (right) with and (left)
without exposure to isocyanide vapor is shown. Each picture is a merge of four false-color images taken at different wavelength settings (lower panel).
The repeated four-color spot motif is a set of control fluorophores: fluorescein, Cy3, TxR, and Cy5. (c) Cumulative microarray 3-CR profiling results for
eight amidines Ax, 40 aldehydes By, and five isocyanides Cz. Each point corresponds to the formation of a fluorescent compound in the reaction Ax þ By þ
Cz recorded at a given wavelength (coded by the corresponding colors).
The fluorescence of each droplet was compared with that of a
control microarray incubated without isocyanide. We observed
that exposure of the slides to isocyanide resulted in the
appearance of a distinctive pattern of fluorescent spots
(Figure 1b and Figure S2b). The microarray thus functions as
a kind of “chemical nose” that reports the presence of the
volatile isocyanide by its fluorescence signal. Analysis of the
microarray data revealed the crucial role of the amidine
structure in the appearance of fluorescence. The four amidines
A2, A4, A6, and A8 gave rise to fluorescent signals, with A2 and
A8 being the most efficient (Figure 1c and Figure S2b).
Aminopyridine A2 produced mainly yellow and green spots,
while aminopyrazine A8 yielded predominantly green spots.
Both A4 and A6 produced fewer and weaker fluorescent signals
with apparently red-shifted emissions. With the same amino-
pyridine, the color of the 3-CR changed significantly depending
on the aldehyde structure, while the isocyanide component did
not have a major effect on the fluorescence pattern. Other
amidines did not produce visible fluorophores except when
reacted with the inherently fluorogenic aldehydes B10, B14, and
B31 (Figure 1c).
The amidines A2, A4, A6, and A8 identified by microarray
screening were then used in scaled-up 3-CRs to synthesize a
small library of ∼60 compounds. We named these compounds
“Flugis”, which stands for Fluorescent Ugi products. In addition,
amidines A8ꢀA12 and isocyanides C6ꢀC8 were introduced to
investigate more extensively the structureꢀfluorescence rela-
tionship of Flugis (Figure 2). We found that all four fluorogenic
amidines exhibited an intrinsic fluorescence in the UVA region,
albeit with different efficiencies (A2, A4, A6, and A8 in Figure 2).
For amidines A2 and A8, the formation of the Flugi framework
was accompanied by a large (up to ∼150 nm) bathochromic shift
in the fluorescence of the compound (Figure 2 and Figure S3a).
The amino group at position 3 was required for the observed red
shift because the corresponding unsubstituted compounds had
fluorescence similar to the initial amidines (compare compounds
A2, Flugi_0, and Flugi_1 in the table in Figure 2; for spectra, see
Figure S3a). The bathochromic effect was even more pro-
nounced in the case of Flugis derived from amidines A4 and
A6 (Flugi_3 and Flugi_4), resulting in exceptionally large
fluorescence Stokes shifts (Figure 2). For Flugi_3, the Stokes
shift reached 272 nm (12 220 cmꢀ1), which is one of the largest
10059
dx.doi.org/10.1021/ja204016e |J. Am. Chem. Soc. 2011, 133, 10058–10061

Journal of the American Chemical Society
COMMUNICATION
the Flugi fluorescence depends on internal charge transfer
(ICT)7 from the 3-amino group to the heterocyclic framework.
Indeed, blocking the electron-donating ability of the nitrogen by
acylation or replacement of the amino group with a nitroso
residue abolished the fluorescence (not shown), while introduc-
tion of an electron-withdrawing p-fluorophenyl residue induced a
hypsochromic shift of ∼30 nm (Figure 2, Flugi_27 vs Flugi_31).
The presence of an ICT state is also supported by the almost
structureless emission spectra and the solvatochromism of the
Flugi fluorescence (Figure S5). The spectral and chemical
properties of Flugis could be tuned further by varying the
aldehyde-derived substituent at the 2-position. The fluorescence
color changed from blue to red according to the increasing
potential for electron donation and the degree of conjugation of
this residue (Figure 2, Flugi_10ꢀ29). The Flugis derived from
aminopyrazine A8 showed a similar trend, with the fluorescence
shifted toward the blue region (Figure 2, Flugi_11, _23, _28, and
_30). All of the fluorophores exhibited good quantum yields and
large Stokes shifts (up to 158 nm). Moreover, methylation of the
N1 nitrogen red-shifted the Flugi fluorescence by an additional
∼30 nm and further increased the Stokes shift to ∼169 nm
(Figure 2, Flugi_21 vs Flugi_22). Finally, the isocyanide-derived
substituent of the 3-amino group also had some effect on the
Flugi fluorescence (Figure 2, Flugi_31ꢀ37). These data show
that Flugis represent a new type of fluorophore with a wide color
coverage and, as a result of the 3-CR, excellent potential for
combinatorial diversification. Fluorescence sensitivity toward
substituents and the polarity of the environment make Flugis a
promising scaffold for the synthesis of fluorescent probes,
sensors, and enzyme substrates.
To demonstrate the potential of Flugis as imaging probes, we
aimed to develop a fluorescent ligand for one of the recognized
drug targets of imidazo[1,2-a]pyridines. Zolpidem and its analogues
in the imidazo[1,2-a]pyridine family are well-known anxiolitic
and sedative medications (Figure S6). These drugs interact with
GABAA benzodiazepine receptors as well as with the peripheral
benzodiazepine receptor, now known as the translocator protein
(TSPO).8 The structural determinants for this interaction have
been extensively studied.9,10 In addition to an aromatic residue at
the 2-position of the heterocycle, the presence of a carboxamide
group at the 3-position seems to be required for biological
activity. In alpidem and zolpidem drugs, this group is separated
from the imidazo[1,2-a]pyridine heterocycle by one methylene
residue. Ligands with a two-atom spacer between the heterocycle
and the carboxamide residue can still bind to benzodiazepine
receptors, albeit with lower affinity.10 We therefore synthesized a
panel of extended zolpidem analogues by 3-CR with tert-butyl
isocyanoacetate (Figure 3a). To determine whether these com-
pounds could be used as imaging probes for the mitochondrial
benzodiazepine receptor TSPO, we studied their localization in
PC-3 prostate carcinoma cells. These cells show a high level of
TSPO expression that is thought to be a hallmark of prostate
cancer.11 The Flugis were incubated with live cells growing in
glass bottom chambers, and their subcellular distributions were
determined with an inverted fluorescence microscope using
fluorescein filters set at the 60ꢁ oil immersion objective. In
parallel, mitochondria were visualized with MitoTracker dye. All
of the Flugis stained PC3 cells, but the localization of Flugi_42
most closely matched the mitochondrial distribution pattern
(Figure 3b and Figure S7). The Flugi_43 lacking the disubsti-
tuted carboxamide group did not stain mitochondria, demon-
strating a preferential localization to the cytoplasm and the
Figure 2. Fluorescence properties of Flugis. The table at the top shows
absorption and emission maxima of selected fluorophores. The panel at
the bottom shows the fluorescence of representative Flugis in DMSO
solution as visualized with a 365 nm UV lamp. The concentrations of the
fluorophores were 1 μM, except for those of A4B25C1 (Flugi_3) and
A6B25C1 (Flugi_4), which were 100 μM. Notes: [a]Absorption and
fluorescence emission spectra were recorded in DMSO. [b]Quantum
yields were measured using fluorescein (ϕf = 0.92) and 2-aminopyridine
(ϕf = 0.60) as external standards. [c]A2B25ꢀH was synthesized by the
condensation of 2-aminopyridine (A2) with 4-methoxyphenacyl bro-
mide. [d]A2B25ꢀNH2 was synthesized by Pd-catalyzed hydrogenolysis
þ
ꢀ
of A2B25C2. [e]A2B2C1Me I was synthesized by alkylation of A2B2C1
in neat methyl iodide. (For experimental procedures regarding notes [c],
[d], and [e], see the SI).
values reported for a small organic fluorophore. However, these
compounds were significantly less bright than the Flugis derived
from A2 and A8 and less useful as imaging probes (Figure 2 and
Figure S3). The Flugis derived from amidines A2 and A8 had
extinction coefficients (ε) in the range (1.1ꢀ2.7) ꢁ 104
Mꢀ1 cmꢀ1 and quantum yields from 0.2 to 0.9, resulting in
fluorophore brightnesses B of ∼104 Mꢀ1 cmꢀ1, which is
comparable to that of many commercial fluorescent dyes.1
Consequently, we focused our analysis on the Flugis derived
from aminopyridine A2, which exhibited fluorescence in the
visible spectrum with a high quantum yield.
Replacing the carboxymethyl residue at the 7-position with
other substituents influenced both the absorption spectra and the
fluorescence emission spectra of the Flugis (Figure 2 and Figure
S4). We observed an almost linear correlation between the
spectral maxima and the Hammett substituent constant (σp);
both the absorption maximum and fluorescence emission max-
imum shifted to longer wavelengths as the electron-withdrawing
ability of the substituent increased (Figure S4). It is plausible that
10060
dx.doi.org/10.1021/ja204016e |J. Am. Chem. Soc. 2011, 133, 10058–10061

Journal of the American Chemical Society
COMMUNICATION
Flugi_42 is a ligand of TSPO (Figure 3b). To characterize this
interaction further, the affinities of Flugi_38ꢀ43 for TSPO were
determined by competition experiments against [3H]PK11195
performed on rat stomach mitochondrial membranes (Figure S8).
Inthese tests, two Flugis had measurable micromolar affinities, and
the potency of Flugi_42 was comparable to that of zolpidem.
In summary, we have demonstrated that autofluorescing drug-
like molecules can be a valuable source of bioimaging probes.
Combinatorial screening in droplet arrays allowed the de novo
discovery of Flugi dyes. We demonstrated fluorescent ligands for
benzodiazepine receptors, but other applications of Flugis may
be envisaged, such as the development of fluorescent mimics for
tryptophan and nucleotides as well as for related drugs.
’ ASSOCIATED CONTENT
S
Supporting Information. Figures S1ꢀS8, experimental
b
procedures, and compound characterization. This material is
available free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION
Corresponding Author
maxim.balakirev@cea.fr
’ REFERENCES
(1) Lavis, L. D.; Raines, R. T. ACS Chem. Biol. 2008, 3, 142.
(2) (a) Rosania, G. R.; Lee, J. W.; Ding, L.; Yoon, H. S.; Chang, Y. T.
J. Am. Chem. Soc. 2003, 125, 1130. For a review, see: (b) Vendrell, M.;
Lee, J. S.; Chang, Y. T. Curr. Opin. Chem. Biol. 2010, 14, 383.
(3) Simeonov, A.; Jadhav, A.; Thomas, C. J.; Wang, Y.; Huang, R.;
Southall, N. T.; Shinn, P.; Smith, J.; Austin, C. P.; Auld, D. S.; Inglese, J.
J. Med. Chem. 2008, 51, 2363.
(4) Mugherli, L.; Burchak, O. N.; Balakireva, L. A.; Thomas, A.;
Chatelain, F.; Balakirev, M. Y. Angew. Chem., Int. Ed. 2009, 48, 7639.
(5) (a) Groebke, K.; Weber, L.; Mehlin, F. Synlett 1998, 661.
(b) Blackburn, C.; Guan, B.; Fleming, P.; Shiosaki, K.; Tsai, S. Tetrahedron
Lett. 1998, 39, 3635. (c) Bienayme, H.; Bouzid, K. Angew. Chem., Int. Ed.
1998, 37, 2234. (d) D€omling, A. Chem. Rev. 2006, 106, 17. (e) Hulme, C.;
Lee, Y.-S. Mol. Diversity 2008, 12, 1.
(6) (a) Hulme, C.; Bienayme, H.; Nixey, T.; Chenera, B.; Jones, W.;
Tempest, P.; Smith, A. Methods Enzymol. 2003, 369, 469. (b) Lin, Q.;
Blackwell, H. E. Chem. Commun. 2006, 2884.
Figure 3. Synthesis and characterization of fluorescent analogues of
zolpidem. (a) Two-step synthesis of zolpidem analogues from AxByC10
Flugis. Spectral properties of the synthesized fluorophores are shown in
the table. (For notes [a], [b], and [d], see the Figure 2 caption).
(b) Fluorescent staining of live PC-3 cells with Flugis. The first column
shows colocalization of Flugi_42 (visualized in the FITC channel) and
mitochondria (labeled with MitoTracker Deep Red, Cy5 channel). The
colocalization scatter plot with a corresponding Pearson’s coefficient of
0.912 is shown at the bottom. The second column shows the different
patterns of cell staining observed with Flugi_43. The third column
shows how Flugi_42 fluorescent staining was affected by cell preincuba-
tion with 25 μM PK11195. Scale bar: 20 μm. For a zoomed view, see
Figure S7.
(7) (a) de Silva, A. P.; Gunaratne, H. Q. N.; Gunnlaugsson, T.;
Huxley, A. J. M.; McCoy, C. P.; Rademacher, J. T.; Rice, T. E. Chem. Rev.
1997, 97, 1515. (b) Lakowicz, J. R. Principles of Fluorescence Spectroscopy,
2nd ed.; Kluwer Academic/Plenum Publishers: New York, 1999.
(8) Papadopoulos, V.; Baraldi, M.; Guilarte, T. R.; Knudsen, T. B.;
Lacapꢀere, J. J.; Lindemann, P.; Norenberg, M. D.; Nutt, D.; Weizman, A.;
Zhang, M. R.; Gavish, M. Trends Pharmacol. Sci. 2006, 27, 402.
(9) Clayton, T.; Chen, J. L.; Ernst, M.; Richter, L.; Cromer, B. A.;
Morton, C. J.; Ng, H.; Kaczorowski, C. C.; Helmstetter, F. J.; Furtmuller,
R.; Ecker, G.; Parker, M. W.; Sieghart, W.; Cook, J. M. Curr. Med. Chem.
2007, 14, 2755.
(10) (a) Kozikowski, A. P.; Ma, D.; Brewer, J.; Sun, S.; Costa, E.;
Romeo, E.; Guidotti, A. J. Med. Chem. 1993, 36, 2908. (b) Trapani, G.;
Franco, M.; Ricciardi, L.; Latrofa, A.; Genchi, G.; Sanna, E.; Tuveri, F.;
Cagetti, E.; Biggio, G.; Liso, G. J. Med. Chem. 1997, 40, 3109.
(11) (a) Pretner, E.; Amri, H.; Li, W.; Brown, R.; Lin, C. S.;
Makariou, E.; Defeudis, F. V.; Drieu, K.; Papadopoulos, V. Anticancer
Res. 2006, 26, 9. (b) Fafalios, A.; Akhavan, A.; Parwani, A. V.; Bies, R. R.;
McHugh, K. J.; Pflug, B. R. Clin. Cancer Res. 2009, 15, 6177.
plasma membrane. To determine whether mitochondrial stain-
ing by Flugi_42 was due to its interaction with the benzodiaze-
pine receptor, cells were preincubated with PK11195, a highly
specific TSPO ligand. At 25 μM, this compound induced a
significant displacement of fluorescent staining, suggesting that
10061
dx.doi.org/10.1021/ja204016e |J. Am. Chem. Soc. 2011, 133, 10058–10061