Development and initial application of a hybridization-independent, DNA-encoded reaction discovery system compatible with organic solvents

Article abstract of DOI:10.1021/ja074155j

We have developed and applied an approach to reaction discovery that takes advantage of DNA encoding, DNA-programmed assembly of substrate pairs, in vitro selection, and PCR amplification, yet does not require reaction conditions that support DNA hybridiz

Full text of DOI:10.1021/ja074155j

Published on Web 11/10/2007
Development and Initial Application of a
Hybridization-Independent, DNA-Encoded Reaction Discovery
System Compatible with Organic Solvents
Mary M. Rozenman, Matthew W. Kanan, and David R. Liu*
Contribution from the Howard Hughes Medical Institute and the Department of Chemistry and
Chemical Biology, HarVard UniVersity, Cambridge, Massachusetts 02138
Received June 7, 2007; E-mail: drliu@fas.harvard.edu
Abstract: We have developed and applied an approach to reaction discovery that takes advantage of
DNA encoding, DNA-programmed assembly of substrate pairs, in vitro selection, and PCR amplification,
yet does not require reaction conditions that support DNA hybridization. This system allows the simultaneous
evaluation of >200 potential bond-forming combinations of substrates in a single experiment and can be
applied in a range of solvent and temperature conditions. In an initial application, we applied this system
to explore Au(III)-mediated chemistry and uncovered a simple, mild method for the selective Markovnikov-
type hydroarylation of vinyl arenes and trisubstituted olefins with indoles.
Introduction
We recently implemented a selection-based approach to the
discovery of bond-forming reactions.6 Reaction discovery is a
central endeavor in chemistry because it provides new tools for
chemical synthesis, facilitates the discovery of functional
synthetic molecules, and can reveal new principles of reactivity
when coupled with mechanistic investigation. Most methods for
reaction discovery search for conditions that enable a specific
desired product structure to be formed from potential precursors.
Our approach is complementary because it does not focus on
one particular product, but instead simultaneously evaluates bond
formation between any two members of a large collection of
substrates under one of many different reaction conditions.
Because this approach uses a selection for bond formation that
is independent of substrate or product structure, it does not rely
on specific reactivity predictions and enables a broad search
for reactivity among a range of substrates.7
New functional molecules emerge in nature through iterated
cycles of translation, selection, amplification, and diversification
of genetic material. In the laboratory, researchers can carry out
the same evolutionary process in a directed fashion to access
molecules possessing desired properties. Evolution-based ap-
proaches have primarily been applied to the discovery of
biomolecules with a broad range of function.1,2 More recently,
the techniques of molecular evolution have been applied
to problems in the chemical sciences, including the synthesis
and discovery of functional small molecules and the discovery
of new chemical reactions.3,4 Such approaches can be parti-
cularly powerful because they are compatible with a selection,
a process that simultaneously evaluates all members of an
arbitrarily large population of molecules and separates functional
molecules from inactive variants. Selections can be more
efficient than conventional screens in which molecules or
reactions are individually evaluated in a low- or high-throughput
manner because selections allow for en masse evaluation with-
out requiring spatial separation of candidate molecules. Selec-
tions have proven especially effective when the molecules
under selection are associated with nucleic acids that encode
each molecule’s identity4 because nucleic acids can be readily
amplified and decoded. As a result, selections carried out
on nucleic acids or nucleic acid-small molecule conjugates
require only minute quantities of material (typically, sub-
nanomolar) and can be iterated to multiply their net effective-
ness.5
Our first-generation reaction discovery system (Figure 1a)6
used DNA hybridization to organize many potential bond-
forming substrate combinations into discrete pairs. Following
the exposure of DNA-duplex localized reactants to a given set
of reaction conditions, DNA-templated bond formation between
substrates in each reactive pair induced the transfer of a biotin
group to the DNA strand encoding those two substrates. In vitro
selection using immobilized streptavidin, PCR amplification,
and DNA microarray analysis subsequently revealed the identi-
ties of reactive substrate pairs. This system uncovered a mild
and efficient Pd(II)-mediated coupling reaction between alkyna-
mides and alkenes to generate trans-R,â-unsaturated ketones.6,8
(6) Kanan, M. W.; Rozenman, M. M.; Sakurai, K.; Snyder, T. M.; Liu, D. R.
Nature 2004, 431, 545-9.
(1) Lin, H.; Cornish, V. W. Angew. Chem., Int. Ed. 2002, 41, 4402-25.
(2) Yuan, L.; Kurek, I.; English, J.; Keenan, R. Microbiol. Mol. Biol. ReV.
2005, 69, 373-92.
(7) Porco and coworkers have recently reported promising results from a LC//
MS-based reaction discovery system that screens the ability of one substrate
to undergo a bond-forming reaction with one of many different potential
partners. See: Beeler, A. B.; Su, S.; Singleton, C. A.; Porco, J. A., Jr. J.
Am. Chem. Soc. 2007, 129, 1413-9.
(3) Gartner, Z. J. Pure Appl. Chem. 2006, 78, 1-14.
(4) Rozenman, M. M.; McNaughton, B. R.; Liu, D. R. Curr. Opin. Chem.
Biol. 2007, 11, 259-68.
(5) Doyon, J. B.; Snyder, T. M.; Liu, D. R. J. Am. Chem. Soc. 2003, 125,
12372-3.
(8) Momiyama, N.; Kanan, M. W.; Liu, D. R. J. Am. Chem. Soc. 2007, 129,
2230-1.
9
10.1021/ja074155j CCC: $37.00 © 2007 American Chemical Society
J. AM. CHEM. SOC. 2007, 129, 14933-14938
14933

A R T I C L E S
Rozenman et al.
Figure 1. DNA hybridization-dependent (a) and DNA hybridization-independent (b) systems for reaction discovery. The modular assembly strategy for the
hybridization-independent system is shown in (c).
Past applications of evolutionary principles to problems in
the chemical sciences have largely been limited to contexts that
mimic the biological milieu. For example, our first-generation
reaction discovery system was limited to aqueous, high salt,
low-temperature conditions that facilitate DNA hybridization.
Here, we describe a hybridization-independent reaction discov-
ery system (Figure 1b) that offers the advantages of a selection-
based approach but allows for discovery in organic solvents, at
high temperatures, and in the presence of additives that may
preclude DNA base pairing. In addition, we report the use of
this second-generation system to discover a Au(III)- or acid-
mediated alkene hydroarylation reaction.
the completed pool of n × m substrate pairs, each linked to a
single strand of DNA that uniquely identifies its two attached
substrates.
Selection Design. Selection for bond formation in the
hybridization-independent reaction discovery system (Figure 2)
uses concepts developed by the molecular evolution com-
munity.9 A single solution containing ∼1 pmol total of the
completed substrate pool is incubated under a chosen set of
reaction conditions. After cleavage of the disulfide bond, only
those oligonucleotides encoding a productive bond-forming
substrate combination retain a covalently attached biotin group
(Figure 2). DNA strands encoding bond-forming combinations
are selected using streptavidin-linked magnetic beads. PCR
amplification and DNA microarray analysis reveals the identity
of substrate pairs that undergo bond formation under the chosen
reaction conditions as green microarray spots (Figure 2).6
A distinguishing feature of the current and former DNA-
encoded reaction discovery systems is the ability to detect bond-
forming reactions between substrates that would preferentially
homocouple under the reaction conditions. When the substrate
pool is exposed to reaction conditions at extremely low
concentrations (∼nanomolar), random intermolecular reactions
including substrate homocoupling do not take place at an
appreciable rate. In contrast, DNA-encoded substrate pairs
experience a much higher effective molarity (∼millimolar),
enabling them to react and survive the selection for bond
formation.
For our initial construction of this system, we chose 14 pool
A and 14 pool B substrates (Figure 3) to represent simple,
readily accessible organic functional groups. We assembled the
system starting with 4 nmol of A-linked oligonucleotides (Figure
1c, left), resulting in >300 pmol of the 196 heterocoupling
species and 28 homocoupling species that comprise the fully
assembled pool of 224 substrate combinations. This quantity
of material is sufficient to assess the reactivity of these substrate
combinations under several hundred different reaction condi-
tions, representing the evaluation of >50,000 potential reactions.
Results and Discussion
Preparation of DNA-Encoded Substrate Pools. To remove
the requirement for DNA hybridization, we replaced the
complementary strands in the first-generation system with a
single strand attached to both substrates in a manner that enables
selection for bond formation (Figure 1b). These single strands
that each contain a substrate pair are assembled modularly by
enzyme-catalyzed primer extension and ligation reactions that
minimize the number of required starting components.
Briefly, small-molecule substrates are chosen for each of two
substrate pools. Each of n substrates from pool A is linked
through an internal adenine to an oligonucleotide that contains
both a coding region for that substrate as well as a constant
region (Figure 1c). The combined set of n pool A DNA-
substrate conjugates is hybridized with one pool B primer, which
contains a region complementary to the pool A constant region,
as well as a unique coding sequence for one pool B substrate.
Enzyme-catalyzed primer extension and ligation reactions result
in a set of n finished substrates in which one pool B substrate
is covalently linked through a biotinylated disulfide linker to
an oligonucleotide bearing one of the n pool A substrates (Figure
1c and Supporting Information Figures S4-S6). These primer
extension and ligation steps are repeated for each of m different
pool B substrates. The resulting samples are combined to provide
(9) Wilson, D. S.; Szostak, J. W. Annu. ReV. Biochem. 1999, 68, 611-47.
9
14934 J. AM. CHEM. SOC. VOL. 129, NO. 48, 2007

Development of a DNA-Encoded Reaction Discovery System
A R T I C L E S
Figure 2. A one-pot selection and analysis method for the detection of bond-forming reactions between DNA-linked small-molecule substrates.
Selection for Bond Formation in Organic Solvents and
at High Temperatures. To validate the ability of the system
to detect bond-forming events under conditions that do not
support DNA hybridization, we performed several selections
for known reactions in a variety of organic solvents and at a
range of temperatures. We exposed 1.5 pmol of substrate pool
to 1 mM Cu(I) for 10 min at 25 °C in acetonitrile, DMF,
methanol, and dioxane; in all cases, the final solvent composition
was 90% organic solvent and 10% H2O. In all four cases,
selection for bond formation, PCR, and microarray analysis
resulted exclusively in a strong signal from the substrate
combination corresponding to the Cu(I)-catalyzed cycloaddi-
tion10 between an azide (A7) and alkyne (B5) (Figure 3 and
Supporting Information Figure S8). To test for the ability of
the system to detect a less efficient reaction, we exposed the
substrate pool to NaBH3CN in 9:1 acetonitrile:water. Microarray
analysis reflected the expected reductive amination reaction
between the aryl aldehyde (A2) and allyl amine (B14) (Figure
3). Finally, we exposed the system to 1 mM Na2PdCl4 in
aqueous solvent for 20 min at 25, 50, 65, or 95 °C. Microarray
analysis reported expected Pd-mediated reactions at all four
temperatures, including those that do not support DNA hybrid-
ization (Supporting Information Figure S9). Taken together,
these findings validate the hybridization-independent reaction
discovery system as a means of revealing bond-forming
reactivity in organic solvents and across a range of temperatures.
We then applied this system to explore the reactivity of Au-
(III) salts. Recent reports suggest that Au(III) can activate a
broad range of organic functional groups.11,12 We exposed 1
pmol of the substrate pool to 10 mM AuCl3 in 9:1 acetonitrile:
water. After 2 h at 25 °C, selection, PCR amplification, and
DNA microarray analysis were carried out as described above.
The strongest green spots on the microarray suggested bond
formation between the indole (B9) and substrates including
styrene (A14) and a terminal alkyne (A13) (Figure 3).
The B9+A13 indole-alkyne coupling signal is consistent
with previous reports.13-15 To the best of our knowledge,
however, a gold-mediated coupling between indoles and styrenes
has not been previously reported, although gold has been
observed to mediate the addition of activated methylenes to
styrene16 and the intramolecular hydroarylation of allenes with
indoles.17
Characterization of the Indole-Styrene Coupling Reac-
tion. The putative indole-styrene reaction was characterized
both in a DNA-linked format and in a conventional flask-based
format. To mimic the reaction discovery environment, we
exposed 100 pmol of DNA-linked indole to 50 mM N-propyl-
4-vinylbenzamide and 10 mM AuCl3 in 9:1 acetonitrile:water.
After 2 h at 25 °C, the DNA-linked material was treated with
S1 nuclease to cleave the DNA strand into mononucleotide
adducts (Supporting Information). High-resolution mass spec-
trometry of the resulting material was consistent with the
formation of a redox-neutral coupling product with a molecular
weight corresponding to the sum of the indole and styrene
substrates (Supporting Information Figure S10). These findings
validated the B9+A14 selection result and indicated that all
atoms present in the starting material were present in the reaction
product.
We subsequently optimized reaction conditions using non-
DNA-linked substrates. Although product formation was inef-
ficient in MeCN, slow addition of styrene to N-phenylsulfonyl-
(Bs)-protected indole (1) in the presence of stoichiometric AuCl3
in methylene chloride resulted in 82% isolated yield of the
Markovnikov-type hydroarylation product (2) without the exclu-
sion of air or moisture from the system (Table 1, entries 1-3).
High yields of product were also observed with the use of 10
mol % Au(III) in the presence of 30 mol % AgOTf15 (Table 1,
entries 4 and 5). While changing the gold counterion did not
affect product formation (Table 1, entry 5), the silver counterion
proved to be an important determinant of reaction efficiency
(10) Rostovtsev, V. V.; Green, L. G.; Fokin, V. V.; Sharpless, K. B. Angew.
Chem., Int. Ed. 2002, 41, 2596-9.
(11) Gorin, D. J.; Toste, F. D. Nature 2007, 446, 395-403.
(12) Hashmi, A. S.; Hutchings, G. J. Angew. Chem., Int. Ed. 2006, 45, 7896-
936.
(13) Ferrer, C.; Echavarren, A. M. Angew. Chem., Int. Ed. 2006, 45, 1105-9.
(14) Li, Z.; Shi, Z.; He, C. J. Organomet. Chem. 2005, 690, 5049-5054.
(15) Sommer, K.; Reetz, M. T. Eur. J. Org. Chem. 2003, 3485-3496.
(16) Yao, X.; Li, C. J. J. Am. Chem. Soc. 2004, 126, 6884-5.
(17) Liu, C.; Widenhoefer, R. A. Org. Lett. 2007, 9, 1935-8.
9
J. AM. CHEM. SOC. VOL. 129, NO. 48, 2007 14935

A R T I C L E S
Rozenman et al.
Figure 3. Substrates used in this work and microarray analysis of selection experiments carried out after exposure to reactions conditions listed under each
array. Green spots suggest bond formation between corresponding substrates.
(Table 1, entries 4-7). On the basis of this observation18 and
recent literature reporting parallels in reactivity between metal
triflates and triflic acid,19,20 we investigated triflic acid as a
potential catalyst of this olefin hydroarylation reaction. We
found that 5 mol % triflic acid generated product in 91% yield
(Table 1, entry 9). Silver triflate alone (Table 1, entry 8) did
not mediate the reaction. Exposure of the reaction substrates to
5 mol % HCl or to 1 equiv of HCl (Table 1, entries 10 and 11)
also failed to generate product, indicating that the reaction can
be mediated either by triflic acid or by AuCl3.
The prospect of accessing aryl-functionalized indole scaffolds
under very mild conditions prompted further exploration of the
substrate scope of the triflic acid-catalyzed hydroarylation
reaction (Table 2). In all cases, reactions were run open to the
air with no precautions taken to exclude moisture. Moderate to
Table 1. Optimization of Indole-Styene Couplinga
entry
additive
solvent
% yield
1
2
3
1 equiv of AuCl3
1 equiv fo AuCl3
1 equiv of AuCl3
CH3CN
1,2-C2H4Cl2
CH2Cl2
<1
74
82
93
90
52
0
0
91
0
4
5
6
10 mol % AuCl3/30 mol % AgOTf
10 mol % AuBr3/30 mol % AgOTf
10 mol % AuCl3/30 mol % AgBF4
CH2Cl2
CH2Cl2
CH2Cl2
7
8
9
10
11
10 mol % AuCl3/30 mol % Ag(O2CCF3) CH2Cl2
1 equiv of AgOTf
5 mol % TfOH
5 mol % HCl
CH2Cl2
CH2Cl2
CH2Cl2
CH2Cl2
1 equiv of HCl
0
(18) Anderson, L. L.; Arnold, J.; Bergman, R. G. J. Am. Chem. Soc. 2005, 127,
14542-3.
a Reactions were carried out with 1 equiv of styrene added slowly to a
solution of 1 and the specified additive. No care was taken to exclude air
or moisture from the system. The yields shown are isolated yields of pure
product.
(19) Li, Z.; Zhang, J.; Brouwer, C.; Yang, C. G.; Reich, N. W.; He, C. Org.
Lett. 2006, 8, 4175-8.
(20) Rosenfeld, D. C.; Shekhar, S.; Takemiya, A.; Utsunomiya, M.; Hartwig, J.
F. Org. Lett. 2006, 8, 4179-82.
9
14936 J. AM. CHEM. SOC. VOL. 129, NO. 48, 2007

Development of a DNA-Encoded Reaction Discovery System
A R T I C L E S
Table 2. Acid-Catalyzed Indole Hydroarylationsa
olefins with indole nucleophiles are also of significant interest,
they have remained scarce. Widenhoefer and co-workers have
reported a Pt-catalyzed intramolecular hydroarylation of unac-
tivated alkenes.26 While this chemistry was recently extended
to include the intermolecular hydroarylation of styrenes and
simple R-olefins with 1,2-disubstituted indoles,27 it requires high
temperatures and exhibits e2:1 selectivity for the Markovnikov
product when styryl olefins are used. Friedel-Crafts-type
hydroarylations of styrenes with anisole, thiophene, and xylene
nucleophiles under Lewis acid catalysis have also been recently
reported,28-30 but require the use of superstoichiometric arene
(g4 equiv) and have not been extended to indole nucleo-
philes.
The Au or acid-mediated regioselective hydroarylation of
styrenes with Bs-protected indole reported here therefore
provides an efficient route to indole-containing diaryl scaffolds
not readily accessed by other hydroarylation reactions.31 In
addition, the product scaffolds accessed by these reactions are
of particular interest because they are featured prominently in
a wide array of broadly applied pharmacophores including
peroxisome proliferator-activated receptor gamma (PPARγ)
agonists, endothelin (ETA) receptor agonists, and monoamine
reuptake inhibitors.
Conclusion
The application of selection-based principles to address
chemical problems is a relatively new but promising area.4
Here, we have implemented an approach to reaction discovery
that takes advantage of DNA encoding, DNA-programmed
assembly of substrate pairs, in vitro selection, and PCR
amplification, yet does not require reaction conditions that
support DNA hybridization. This hybridization-independent
reaction discovery system allows the simultaneous evaluation
of >200 potential bond-forming combinations of substrates in
a single experiment and has led to the discovery of a simple,
mild method for the selective Markovnikov-type hydroarylation
of olefins with indoles. Although the discovered transformation
is consistent with previous notions of indole and styrene
reactivity, its selectivity, efficiency, and mildness make it a
potentially useful addition to known methods for accessing
diaryl scaffolds. The scope and mechanism of this chemistry
are under further investigation in our laboratory. Furthermore,
we are in the process of applying the hybridization-independent
reaction discovery system to a broad-scale exploration of
transition metal-mediated and organocatalyst-mediated reactivity
of simple organic functional groups in aqueous and organic
solvent.
a Reactions were carried out with 1 equiv of 4 added slowly to a solution
of 3 in the presence of 5 mol % TfOH. No care was taken to exclude air
or moisture from the system. The yields shown are isolated yields of pure
product.
good yields of Markovnikov-type hydroarylation products were
obtained with substituted styrene variants as well as with
2-methyl-2-pentene (Table 2, entries 2-4). However, mono-
and disubstituted olefins such as 1-pentene, trans-2-heptene, or
cyclohexene are not productive substrates for this transformation
(Table 2, entry 5 and Supporting Information).
The reactions described above represent a significant addition
to Friedel-Crafts-type indole hydroarylation chemistry.21 Re-
searchers have previously developed organocatalytic, Brønsted-
acid-mediated, and Lewis-acid-mediated hydroarylation reac-
tions that couple indoles to electron-deficient olefins such as
nitroalkenes22 and R,â-unsaturated carbonyls,21,23 or to electron-
rich olefins such as enamines.24,25 Although methods for the
hydroarylation of styrenes and unactivated alkyl-substituted
(26) Liu, C.; Han, X.; Wang, X.; Widenhoefer, R. A. J. Am. Chem. Soc. 2004,
126, 3700-1.
(27) Zhang, Z.; Wang, X.; Widenhoefer, R. A. Chem. Commun. 2006, 3717-
9.
(28) Kischel, J.; Jovel, I.; Mertins, K.; Zapf, A.; Beller, M. Org. Lett. 2006, 8,
19-22.
(21) Bandini, M.; Melloni, A.; Tommasi, S.; Umani-Ronchi, A. Synlett 2005,
1199-1222.
(29) Rueping, M.; Nachtsheim, B. J.; Scheidt, T. Org. Lett. 2006, 8, 3717-9.
(30) Sun, H.-B.; Li, B.; Hua, R.; Yin, Y. Eur. J. Org. Chem. 2006, 4231-
4236.
(22) Herrera, R. P.; Sgarzani, V.; Bernardi, L.; Ricci, A. Angew. Chem., Int.
Ed. 2005, 44, 6576-9.
(23) Evans, D. A.; Fandrick, K. R.; Song, H. J.; Scheidt, K. A.; Xu, R. J. Am.
Chem. Soc. 2007, 129, 10029-41.
(24) Terada, M.; Sorimachi, K. J. Am. Chem. Soc. 2007, 129, 292-3.
(25) Wang, Y. Q.; Song, J.; Hong, R.; Li, H.; Deng, L. J. Am. Chem. Soc. 2006,
128, 8156-7.
(31) Kobayashi and coworkers have reported a Bronsted acid-catalyzed dehy-
drative coupling of l-methylindole with activated benzyl alcohols that
provides access to diaryl scaffolds similar to the diaryl hydroarylation
products reported here. See: Shirakawa, S.; Kobayashi, S. Org. Lett. 2007,
9, 311-4.
9
J. AM. CHEM. SOC. VOL. 129, NO. 48, 2007 14937

A R T I C L E S
Rozenman et al.
Acknowledgment. We thank Abigail Doyle for helpful dis-
cussions. This work was supported by NIH grant RO1GM065865
and the Howard Hughes Medical Institute. M.M.R. is an NSF
Graduate Research Fellow.
Supporting Information Available: Experimental details and
supporting data. This material is available free of charge via
the Internet at http://pubs.acs.org.
JA074155J
9
14938 J. AM. CHEM. SOC. VOL. 129, NO. 48, 2007