Iterative in situ click chemistry assembles a branched capture agent and allosteric inhibitor for akt1

Article abstract of DOI:10.1021/ja2064389

We describe the use of iterative in situ click chemistry to design an Akt-specific branched peptide triligand that is a drop-in replacement for monoclonal antibodies in multiple biochemical assays. Each peptide module in the branched structure makes unique contributions to affinity and/or specificity resulting in a 200 nM affinity ligand that efficiently immunoprecipitates Akt from cancer cell lysates and labels Akt in fixed cells. Our use of a small molecule to preinhibit Akt prior to screening resulted in low micromolar inhibitory potency and an allosteric mode of inhibition, which is evidenced through a series of competitive enzyme kinetic assays. To demonstrate the efficiency and selectivity of the protein-templated in situ click reaction, we developed a novel QPCR-based methodology that enabled a quantitative assessment of its yield. These results point to the potential for iterative in situ click chemistry to generate potent, synthetically accessible antibody replacements with novel inhibitory properties.

Full text of DOI:10.1021/ja2064389

ARTICLE
pubs.acs.org/JACS
Iterative in Situ Click Chemistry Assembles a Branched Capture Agent
and Allosteric Inhibitor for Akt1
Steven W. Millward,† Ryan K. Henning,† Gabriel A. Kwong,‡ Suresh Pitram,§ Heather D. Agnew,§
Kaycie M. Deyle,† Arundhati Nag,† Jason Hein,^,|| Su Seong Lee,# Jaehong Lim,# Jessica A. Pfeilsticker,†
K. Barry Sharpless,^ and James R. Heath*,†
†Nanosystems Biology Cancer Center, Division of Chemistry and Chemical Engineering, MC-127-72, California Institute of
Technology, Pasadena, California 91125, United States
‡Division of Health Sciences and Technology, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge,
Massachusetts 02139, United States
§Integrated Diagnostics, 6162 Bristol Parkway, Culver City, California 90230, United States
^Department of Chemistry and The Skaggs Institute for Chemical Biology, The Scripps Research Institute, 10550 North Torrey Pines
Road, La Jolla, California 92037, United States
#Institute of Bioengineering and Nanotechnology, 31 Biopolis Way, The Nanos, Singapore 138669
S Supporting Information
b
ABSTRACT: We describe the use of iterative in situ click chemistry to design an
Akt-specific branched peptide triligand that is a drop-in replacement for
monoclonal antibodies in multiple biochemical assays. Each peptide module
in the branched structure makes unique contributions to affinity and/or
specificity resulting in a 200 nM affinity ligand that efficiently immunoprecipi-
tates Akt from cancer cell lysates and labels Akt in fixed cells. Our use of a small
molecule to preinhibit Akt prior to screening resulted in low micromolar
inhibitory potency and an allosteric mode of inhibition, which is evidenced
through a series of competitive enzyme kinetic assays. To demonstrate the
efficiency and selectivity of the protein-templated in situ click reaction, we
developed a novel QPCR-based methodology that enabled a quantitative assessment of its yield. These results point to the potential
for iterative in situ click chemistry to generate potent, synthetically accessible antibody replacements with novel inhibitory
properties.
’ INTRODUCTION
genetic analyses of the antigen recognition surface have shown
that the majority of the molecular diversity of the variable loops is
contained in a single highly variable loop (CDR-H3).8 In hu-
mans, this loop ranges in size from 1 to 35 residues (15 on
average),9 can adopt a wide range of structural conformations,10
and is responsible for most of the interactions with the antigen
molecule. The other five loops are significantly less diverse and
adopt only a handful of conformations. This suggests that a
carefully selected “anchor” peptide can dominate the mode and
strength of a multiligand interaction. It also suggests that other
peptide components, while providing only modest contributions
to the total interaction energy, can supply important scaffolding
features and specificity elements. Analogously, the iterative
in situ click chemistry approach utilized here begins with an
initial short-chain anchor peptide and then proceeds by adding
additional, covalently coupled peptide branches via a process that
is promoted by the target protein. The specificity and inhibitory
A critical bottleneck in the transition from a potential cancer
biomarker to a clinical diagnostic tool is the availability of high-
affinity, high-selectivity molecular entities to recognize and capture
biomarkers from complex biological mixtures. Almost all current
platforms employ antibodies as capture agents despite their high
cost, poor stability,1,2 and the batch-to-batch variability that often
characterizes biologicals. Many antibodies are poorly characterized,
and reports have called into question the high specificity that is a
perceived hallmark of antibodies.3,4 Developing protein capture
agent approaches, such as phage display peptides5 or nucleic acid
aptamers, can potentially represent a powerful alternative to anti-
bodies for certain diagnostic arrays.6,7 However, the challenge of find-
ing a general and robust approach to produce protein capture agents
that match the performance of monoclonal antibodies remains
daunting. An ideal antibody replacement would be synthetically
facile, be stable to a range of thermal and chemical conditions, and
display high affinity and specificity for the target of interest.
A high quality monoclonal antibody possesses a low-nanomo-
lar affinity and high target specificity. Interestingly, structural and
Received: July 11, 2011
Published: September 30, 2011
r
2011 American Chemical Society
18280
dx.doi.org/10.1021/ja2064389 J. Am. Chem. Soc. 2011, 133, 18280–18288
|

Journal of the American Chemical Society
ARTICLE
potency of the original anchor are augmented in the final
multiligand by the peripheral peptide branches.
such a product screen can be utilized to increase the affinity and/
or selectivity of the final multiligand capture agent. Finally, we
report on the affinity, selectivity, and inhibitory characteristics of
the triligand capture agent/inhibitor, and we demonstrate its use
in cell-based biological assays.
We recently reported on the technique of iterative in situ click
chemistry for the production of a triligand capture agent against
the model protein carbonic anhydrase II (CAII).11 The approach
built upon the method of in situ click chemistry,12 which is a
technique in which a small molecule enzymatic inhibitor is
separated into two moieties, each of which is then expanded
into a small library, one containing acetylene functionalities and
the other containing azide groups. The enzyme itself then
assembles the ‘best fit’ inhibitor from these library components
by selectively promoting the 1,3-dipolar cycloaddition between
the acetylene and azide groups to form a triazole linkage (the
‘click’ reaction). The enzyme promotes the click reaction only
between those library components that bind to the protein in just
the right orientation. The resultant inhibitor can exhibit far
superior affinity characteristics relative to the initial inhibitor
that formed the basis of the two libraries.13,14 Iterative in situ click
chemistry extends this concept to enable the discovery of multi-
ligand protein capture agents, and it has several advantages. First,
structural information about the protein target is replaced by the
ability to sample a very large chemical space to identify the ligand
components of the capture agent. For example, an initial ligand
binder (an anchor ligand) may be identified by screening the
protein against a large (>106 element) one-beadꢀone-com-
pound (OBOC)15 peptide library, where the peptides them-
selves may be composed of natural, non-natural, and/or artificial
amino acids. That anchor ligand is then utilized in an in situ click
screen, again using a large OBOC library, to identify a biligand
binder. A second advantage is that the process can be repeated, so
that the biligand is used as an anchor to identify a triligand and so
forth. The final capture agent can then be scaled up using
relatively simple and largely automated chemistries and deriva-
tized with biochemical handles, such as biotin, as an intrinsic
part of its structure. The approach permits the exploration of
branched, cyclic, and linear capture agent architectures. While
many strategies for protein-directed multiligand assembly have
been described,16,17 most require detailed structural information
on the target to guide the screening strategy, and most (such as
the original in situ click approach) are optimized for low-diversity
small molecule libraries.
Here we extend the use of iterative in situ click chemistry to
synthesize a high-specificity branched, triligand capture agent/
inhibitor for the Akt kinase. Akt is a critical molecular router that
mediates signal transduction from the plasma membrane
(cytokine receptors, GPCRs) to downstream effector molecules
that control cell growth, apoptosis, and translation.18 Akt over-
expression and/or hyperactivation has been observed in numer-
ous cancer types.19 Such ubiquitous and aberrant Akt activity has
made Akt a target for cancer diagnostics and therapeutics.20 For
developing a capture agent/inhibitor against Akt, we developed
two novel screening strategies. First, we report on the use of a
preinhibited form of the kinase as a screening target, which provides
an approach toward developing an allosteric site inhibitor.
Second, we take advantage of the fact that an in situ click screen,
in which an anchor ligand and protein target are screened against
a large OBOC library, will selectively generate triazole-linked
products on the hit beads. We characterize the efficiency of this
process through the use of a novel QPCR assay to quantify the
on-bead product. We also expand this concept in the form of
‘product screens’, in which the presence of on-bead clicked
product is taken to be the signature of a hit bead. We show that
’ EXPERIMENTAL SECTION
There are a number of affinity, selectivity, biochemical, and biological
assays described in this work. Full experimental procedures, synthetic
protocols, and additional experimental data are discussed in the Supporting
Information.
Three different types of screens were employed for this work. The first
screen, called a target screen, is analogous to most screening strategies
described in the literature. Hit beads are identified by selecting those
beads onto which the target protein has bound. The second type of
screen, called an inhibited target screen, is based on selecting those beads
on to which to which the target protein has bound, but the screen itself is
carried out in the presence of a small molecule inhibitor. The third type
of screen is the product screen; hit beads are identified by the presence of
on-bead triazole-linked product.
Screening: General. All screens utilized OBOC peptide libraries
that were synthesized on TentaGel (Rapp Polymere) using standard
Fmoc SPPS protocols. Blocking conditions were carried out in Akt
Blocking Buffer: 25 mM Tris-Cl (pH = 7.5), 150 mM NaCl, 10 mM
MgCl2, 0.1% (v/v) β-mercaptoethanol, 0.1% (v/v) Tween-20, and
1 mg/mL BSA. The results of each screen are presented in the
Supporting Information.
Screening: Anchor Ligand. An initial anchor ligand against Akt
was identified through the use of an inhibited target screen. The OBOC
library was synthesized manually and was of the form H2N-AzX-
XXXXX-GYM-TG where TG = TentaGel resin, X = one of eighteen
natural L-amino acids (Cys, Met), and AzX = one of three azido amino
acids Az2, Az4, and Az8 (Supporting Information, Figure S2). The
syntheses of Az2 is described in the Supporting Information; the
synthesis of Az4 and Az8 were carried out as previously described.11
Inhibited Target Screen: The initial screen was carried out in the
presence of 21 nM Akt1-S473E-T308P and 500 μM of the inhibitor
Ac7 under blocking conditions for 75 min at room temperature. Binding
was detected by probing with a monoclonal antibody specific for
phosphorylated T308, followed by anti-mouse secondary antibodies
conjugated with alkaline phosphatase. Purple “hit” beads (defined by
color change in the presence of 5-bromo-4-chloro-3-indolyl phosphate/
nitro blue tetrazolium (BCIP/NBT) substrate) were washed in water,
stripped with 7.5 M Guad-Cl (pH = 2), and sequenced by Edman
degradation (Supporting Information, Figure S3). The initial hit
sequences defined a focused library which was subjected to an inhibited
target screen as described above (Supporting Information, Figures S4
and S5). Candidate sequences were scaled up and tested for their ability
to inhibit Akt1-S473E-T308P activity. The most potent sequence was
resynthesized as a C-terminal biotinylated peptide and used as the
anchor in the biligand screen (Scheme 1).
Screening: Biligand Branch. The biligand branch was identified
through a two-step screening process of a target screen to identify
potential hits, followed by a product screen of those hit beads to identify
€
true hits. Target Screen: A naive library of the form H2N-Pra-XXXXX-
GM-TG where TG = TentaGel resin, X = one of eighteen natural
L-amino acids (lacking Cys and Met), and Pra = L-propargylglycine was
synthesized by standard split-mix protocols on a Titan 357 automated
peptide synthesizer (AAPPTEC). The library was incubated with 90 μM
biotinylated anchor peptide and either Akt1-S473E (9 nM) or Akt1-
S473E-T308P (37 nM) under blocking conditions for 90 min at room
temperature. Binding was detected by antibodies against the phosphory-
lated or nonphosphorylated form of Akt1. After color development,
18281
dx.doi.org/10.1021/ja2064389 |J. Am. Chem. Soc. 2011, 133, 18280–18288

Journal of the American Chemical Society
ARTICLE
Scheme 1. Selection of Anchor Peptidea
diethyldithiocarbamate (trihydrate), 29 mM DIEA, in DMF), three times in
NMP, three times in water, and once in Akt Blocking Buffer. The beads
were then stripped in 7.5 M Guad-HCl (pH = 2), washed in dH2O, and
blocked in QPCR Blocking Buffer for 2 h (0.3% (w/v) BSA, 0.1% (v/v)
Tween-20, 150 μg/mL sheared salmon sperm DNA (Ambion) in
phosphate-buffered saline). The beads were then probed with 0.5 μg/
mL streptavidinꢀoligonucleotide (SA-oligo) for 1 h at room tempera-
ture. The beads were washed five times in QPCR Blocking Buffer and
three times in PBS. Three sets of five beads for each reaction condition
were placed in PCR tubes and subjected to QPCR on an Applied
Biosystems 7300 instrument. Additional experimental details can be
found in the Supporting Information.
a The kinase domain of Akt1 (grey) is preincubated with an ATP-
competitive small molecule inhibitor, Ac7 (1). The inhibited kinase is
then screened with a comprehensive solid-phase pentamer library on
TentaGel (yellow circle) with an N-terminal azido-amino acid. The
structure of the resulting biotinylated anchor peptide is also shown.
’ RESULTS AND DISCUSSION
The Akt1 kinase domain used here contained a S473E
mutation which mimics phosphorylation at the critical S473
residue.21 Expression of the kinase in insect cells (Hi5) produced
∼20 mg/L of partially active enzyme (Akt1-S473E) as well as a
small amount of fully active kinase phosphorylated at Thr308
(Akt1-S473E-T308P). Both were readily separated by anion-
exchange chromatography, were enzymatically active, and were
used as screening targets.
purple hit beads were stripped, decolorized in DMF, and reprobed with
the primary and secondary antibodies in the absence of target protein.
Beads that remained clear were washed and stripped prior to the product
screen. Product Screen: The beads were reblocked and incubated with
streptavidin-AlexaFluor647 (Invitrogen) at a concentration of 0.4 μg/
mL for 30 min at room temperature. The beads were washed exhaus-
tively with Akt Blocking Buffer and imaged on an Axon Genepix 4400A
scanner (MDS). The beads displaying saturated fluorescence signal in
the product screen were sequenced by Edman degradation (Supporting
Information, Figure S8).
In Situ Click Screens. The sequential in situ click chemistry
strategy begins with the identification of an anchor ligand.
Although the anchor ligand can have low affinity for the target,
the nature of how it binds to the target is likely the factor that
most influences the rest of the multiligand development process.
The dominant hot spot on Akt is the catalytic site which contains
the ATP and peptide substrate binding sites. These sites and the
catalytic residues provide the standard epitope for most kinase
inhibitors. We hypothesized that blocking the ATP-binding site
with a small molecule inhibitor (Scheme 1) would remove it as a
thermodynamic sink for library binding. A second possibility,
that was not observed, was that the kinase would direct the
conjugation of the acetylene-bearing small molecule and the
anchor peptide to form a bivalent inhibitor similar to those
previously described.22,23 The absence of the in situ click reaction
between these components may be attributable to suboptimal
proximity and/or spatial arrangement on the protein surface.
Our initial screen against Akt1-S473E-T308P generated a set
of hits which informed the design of a focused library. Upon
screening, this library yielded a set of highly similar sequences
(Supporting Information, Figure S5). Interestingly, one of these
peptides (Az8-VFYRLGY-CONH2) showed almost 95% inhibi-
tion of the Akt1 kinase in the absence and presence of the
conjugated small molecule inhibitor (Supporting Information,
Figure S6). This peptide showed little resemblance to any known
Akt1 peptide substrate24 (e.g., RPRAATF) and was not phos-
phorylated by Akt in vitro (Supporting Information, Figure S7).
The Ac7-independent inhibitory potency of the peptide sug-
gested a novel mechanism for binding and inhibition which led us
to employ it as an anchor for biligand development.
Screening: Triligand Branch. The triligand branch was identified
€
solely through a product screen. The initial naive library was the same as
€
in the initial anchor screen. The naive library (100 mg) was precleared
against alkaline phosphatase-conjugated streptavidin (SA-AP) and de-
veloped with BCIP/NBT, and the purple beads were removed from the
pool. The remaining library was stripped overnight, decolorized with
NMP, and blocked. Akt biligand was synthesized with a C-terminal
biotin and a 5-hexynoic acid cap and used as the anchor compound for
the tertiary ligand screen (5HA-Biligand-Bio, Supporting Information,
Figure S9). Product Screen: The 5HA-Biligand-Bio was incubated in the
presence of the library at a concentration of 30 μM along with Akt-
S473E at a concentration of 110 nM for 90 min at room temperature.
The beads were washed as before, probed with SA-AP, and developed in
BCIP/NBT, and the purple beads were sequenced by Edman degrada-
tion (Supporting Information, Figure S10). A focused library (30 mg)
based on the first round hit sequences was precleared against SA-AP and
subjected to a second round of product screening. The beads were
developed in the presence of BCIP/NBT, and the purple beads were
sequenced by Edman degradation (Supporting Information, Figure
S11).
QPCR-Based Assay for Quantitating in Situ Click Effi-
ciency. The secondary peptide (H2N-Pra-FWFLRG) and the anchor
peptide (H2N-Az8-VFYRLGY) were synthesized on TentaGel. A 0.45
mg amount (∼1000 beads) of each was combined with Akt-S473E
(22 μM) and the corresponding biotinylated peptide (200 μM), the
biotinylated peptide alone (200 μM), or DMSO. The Cu+-catalyzed
click reaction contained 0.45 mg of immobilized peptide, the biotiny-
lated peptide (200 μM), CuI (9 mM), L-ascorbic acid (30 mM), and
tris[(1-benzyl-1H-1,2,3-triazol-4-yl)methyl]amine (TBTA, 4 mM) in a
final volume of 50 μL of 4:1 NMP:dH2O. For immobilized secondary
peptide, the corresponding soluble biotinylated peptide was Ac-Az8-
VFYRLGY-biotin. For immobilized anchor peptide, the corresponding
soluble biotinylated peptide was Ac-Pra-FWFLRG-biotin.
OBOC libraries have been previously employed to isolate
protein kinase substrate sequences25 as well as low affinity
inhibitors for nonprotein kinases.26 However, we are unaware
of previous attempts to select an inhibitory peptide for a protein
kinase in the presence of an ATP-competitive small molecule.
Inhibition of kinases by ligands that bind outside the sub-
strate sites is a well-known phenomenon and so the approach
reported here may have some generality for the develop-
ment of inhibitors.27 Given the structural conservation of the
After incubating the in situ click reactions at 25 ꢀC for 18 h with
strong agitation, the beads were removed and washed exhaustively
in Akt Blocking Buffer. The Cu+ reactions were washed three times
with NMP, ten times in copper chelation solution (22 mM sodium
18282
dx.doi.org/10.1021/ja2064389 |J. Am. Chem. Soc. 2011, 133, 18280–18288

Journal of the American Chemical Society
ARTICLE
Scheme 2. Strategy for Biligand Screensa
positively charged amino acids at positions 2 and 4, negatively
charged amino acids at position 3, and hydrophobic amino acids
at position 5 (Supporting Information, Figure S11). From this
pool, we selected the tertiary peptide Ac-Az8-RHERI-CONH2
and conjugated it to the biligand to form a biotinylated branched
triligand (Figure 3A).
Characterizing the On-Bead in Situ Click Reaction. The
target screen is an important and unique aspect of the meth-
odologies reported here. Thus, we sought to determine the
efficiency and selectivity of the in situ click reaction between
the on-bead secondary peptide and the soluble anchor peptide in
the presence and absence of Akt1-S473E. Previous work in our
laboratory11 suggested that the in situ reaction is low-yielding
relative to the Cu(I)-catalyzed process. Thus, we developed an
analytical assay based on immuno-PCR31 as a means of amplify-
ing and detecting the in situ product. This method takes
advantage of the exquisite sensitivity and large dynamic range
of quantitative PCR (QPCR). We carried out variations of the
in situ click reaction between the biotinylated anchor peptide and
resin-bound secondary peptides. We then used the biotin label to
attach a streptavidinꢀoligonucleotide (SA-oligo) onto only
those beads that contained biligand product. In order to account
for variable sizes, five beads were individually selected and added
to each QPCR reaction, the threshold cycle (Ct) was determined
for each reaction condition, and a standard curve was used to
calculate the amount of biligand present on the bead for each
reaction condition (Figure 1). The click reaction between the
two peptides was approximately 10-fold more efficient in the
presence of Akt1 than in its absence, confirming the requirement
for the target protein to catalyze the click reaction. When the
anchor peptide was immobilized and the biotinylated secondary
peptide was in solution, the efficiency of the in situ click process
was reduced by a factor of 4 (but still above background level),
suggesting that the in situ click reaction observed in the screen is
dependent on the identities of the anchor and secondary peptides
and the manner in which they are displayed (on-bead or in
solution). The copper catalyzed click reaction did not display any
orientation dependence, providing further evidence that the click
reaction observed in the screening process is highly target-
dependent.
a A comprehensive pentapeptide library is synthesized on TentaGel
(yellow circle) and appended with propargylglycine, an acetylene-
containing amino acid. This library is incubated with the Akt1 kinase
domain (grey) and biotinylated anchor peptide (black). The solid-phase
library is probed with an anti-Akt antibody followed by a secondary
antibody conjugated to alkaline phosphatase (purple). Hit beads are
reprobed with the antibodies alone to eliminate antibody binders. The
hit beads are washed, stripped, and reprobed with AlexaFluor647-
labeled streptavidin (red) and imaged for fluorescence. The most highly
fluorescent beads are sequenced to obtain the biligand candidates. The
target screens resulted in hit frequencies between 0.001% and 0.01% of
€
the beads in the naive library while the product screen validated between
23% and 37% of these beads for sequencing.
ATP-binding pocket across kinases, and hence the often-
observed poor selectivity of inhibitors targeted to such sites,28
an approach toward developing off-site inhibitors may have
unique advantages.
While the anchor peptide displayed excellent selectivity for
Akt1-S473E, its affinity (>25 μM) was unsuitable for high-
sensitivity capture of Akt. To increase the affinity, a biligand
branch was selected through a sequence of target and product
screens (Scheme 2). For these screens, the anchor peptide was
modified with a C-terminal biotin for use in the product screen.
We hypothesized that the combination of an initial target screen,
followed by a product screen of those initial hit beads, would
identify only the library members that bound the target and
participated in a successful click event. The target screens
resulted in hit frequencies between 0.001% and 0.01% of the
€
beads in the naive library. The product screen with fluorescent
streptavidin validated between 23% and 37% of these beads for
sequencing.
The biligand branch was selected in the context of an 8-carbon
linker between the two peptide ligands. We assayed for the
selectivity of the in situ click screen by measuring the inhibition
characteristics of biligands prepared with 1, 4, or 8 carbon atom
linkers, but otherwise identical peptide components (Figure 2A).
We found that the selected biligand with the longest linker
(n = 8) possessed a significantly higher inhibitory potency
(Figure 2B) relative to n = 4 and n = 1 linkers. This again
suggests that the secondary peptide was selected in the context of
spatial constraints that define the proteinꢀpeptide interactions.
We note that only multiligands linked by 1ꢀ4 triazoles (anti)
were synthesized. The in situ reaction can result in formation of
both the 1ꢀ4 (anti) and 1ꢀ5 (syn) triazole. It is possible that the
1ꢀ5 triazole (syn) may be the favored linkage and may confer
additional affinity and/or selectivity to the multiligand. However,
on the basis of the fact that our linker between the separate
peptide ligands is a flexible eight-carbon bridge, we would not
expect a significant difference between the two triazole isomers.
Taken together, these experiments suggest that the “hit”
sequences that emerged from our biligand screens were deter-
mined by (1) the target protein, (2) the soluble anchor peptide,
and (3) the relative spacing between the azide and acetylene
The sequences obtained in the secondary ligand screen
showed a preference for aromatic amino acids in the first three
random positions (Supporting Information, Figure S8). We
selected three peptide sequences and synthesized the corre-
sponding biligands with the 1,4-triazole using the Cu(I)-cata-
lyzed azideꢀalkyne cycloaddition.29,30 Two of the three showed
increased binding to Akt1 in immunoprecipitation experiments,
and the most promising candidate (Ac-Pra-FWFLRG-CONH2,
Pra = propargylglycine) was scaled up for additional character-
ization and for the development of the triligand.
For triligand development, the selected biligand was appended
with 5-hexynoic acid at the N-terminus and used as the anchor
ligand (Supporting Information, Figure S9). We omitted the
target screen in these experiments, electing to employ only the
€
product screen for triazole formation. Results from the naive
library revealed weak consensus for the tertiary peptide, although
Az8 was the preferred amino acid at position 1, and positions 2, 3,
and 4 showed a propensity for positively charged amino acids
(Supporting Information, Figure S10). We designed a focused
library based on the amino acid frequencies observed in the initial
screen and identified a tertiary ligand consensus sequence with
18283
dx.doi.org/10.1021/ja2064389 |J. Am. Chem. Soc. 2011, 133, 18280–18288

Journal of the American Chemical Society
ARTICLE
Figure 2. Effect of biligand linker length on Akt inhibition. (A)
Biligands were synthesized with one-, four-, and eight-carbon linkers
between the anchor peptide and triazole. (B) Each variant was titrated
against activated Akt1-S473E-T308P, the activity was measured by
immunoblotting and quantitated by densitometry. The n = 8 linker
which was present in the screen clearly yields the biligand with the
highest inhibitory potency.
Figure 1. Determination of on-bead in situ click reaction efficiency. (A)
The secondary ligand (red letters) is synthesized on TentaGel (yellow)
with an N-terminal propargylglycine and the soluble anchor peptide is
appended with a C-terminal biotin (black) and an N-terminal azido-
amino acid (Az8). These are incubated together under conditions
described below. After the reaction is completed, the beads are washed,
stripped, and probed with a streptavidinꢀoligonucleotide template (SA-
Oligo) (red) to detect the formation of the triazole. The beads were then
subjected to on-bead QPCR. (B) Results of QPCR. The reaction
conditions were (1) Akt1 + biotinylated anchor peptide + bead, (2)
biotinylated anchor peptide + bead, (3) bead, and (4) CuI/ascorbic
acid + biotinylated anchor peptide + bead. The red bars represent the
reaction configuration described in panel a; the black bars represent
reactions where the anchor peptide is synthesized on bead and the
biotinylated secondary peptide is in solution (inverted configuration).
The error bars represent standard error. In this experiment, the
efficiency of the on-bead in situ click reaction is approximately 10-fold
higher in the presence of the Akt1 target than in its absence. For
comparison, the efficiency of the Cu(I)-catalyzed click reaction is
approximately 3 orders of magnitude higher than the protein-templated
reaction. (C) Standard curve of SA-oligo template: Each point is the
mean Ct of duplicate experiments.
The use of long, conformationally restricted linkers in the anchor
and library peptides, or perhaps the use of N-methylated amino
acids34 in the OBOC libraries, may reduce this effect and capture
more of the individual binding energies in the biligand product.
Affinity and Specificity Characterization of Multiligand
PCC Agents. We tested the affinity of the anchor, biligand, and
triligand by ELISA with immobilized Akt-S473E (Figure 3B). As
a detection agent, the biligand showed >100-fold improvement
in its affinity for Akt relative to the anchor peptide while the
triligand showed only a modest affinity gain (2ꢀ3 fold). When
the triligand is employed as an immobilized capture agent,
analysis by surface plasmon resonance (SPR) confirms mid- to
low-nanomolar affinity for Akt1-S473E (KD = 200 nM,
Figure 3C). We next sought to address the specificity of the
triligand for a panel of His-tagged protein kinases (Figure 3D).
For these assays, the multiligands were used as immobilized
capture agents for Akt1-S473E as well as a set of commercially
available active, His-tagged protein kinase domains from the
AGC family (Akt1, PDK1, and p70s6 kinase), the STE family
(MEK1), and the GMGC family (GSK3β). The relative affinity
of each kinase was determined by probing with a horseradish
peroxidase-conjugated anti-His6 antibody and normalizing the
response to Akt1-S473E. The anchor was very specific for the
Akt1 protein, with only modest binding to GSK3β. The sig-
nificantly higher affinity biligand showed reduced selectivity, with
significant cross-reactivity to GSK3β. For the triligand, however,
binding to GSK3β was significantly reduced, bringing it close to
the level observed for the anchor peptide. Additionally, the off-
target binding to MEK1 is completely eliminated at the triligand
functionalities in each component. Because the signal-to-noise
ratio of the in situ screen (true hit/false positive) depends on the
relative rates of the target-mediated and background reactions,
QPCR-guided optimization of the screening conditions may
significantly increase the robustness of the sequential in situ click
approach. In contrast to previous work with small molecule
biligands,12,32,33 we expect the high conformational entropy of
the anchor and secondary peptides to result in a smaller popula-
tion of ternary complexes with the correct conformational and
geometrical orientation required by the cycloaddition reaction.
18284
dx.doi.org/10.1021/ja2064389 |J. Am. Chem. Soc. 2011, 133, 18280–18288

Journal of the American Chemical Society
ARTICLE
Figure 3. Characterization of Akt triligand. (A) Structure of the biotinylated Akt triligand. (B) Relative affinity of Akt triligand and its components. Akt-
S473E was immobilized on Ni-NTA plates and incubated with varying concentrations of biotinylated peptide. All values were normalized to the binding
observed at saturation. (C) Absolute affinity of the triligand by surface plasmon resonance. For this experiment, the biotinylated triligand was
immobilized on a streptavidin-derivatized Biacore chip and probed with Akt-S473E. Fits for the sensograms are shown as solid lines, and the KD was
found to be 200 nM by determination of the kinetic parameters. (D) Specificity of anchor, biligand, and triligand. Biotinylated ligand was immobilized on
streptavidin plates and probed with an array of His-tagged kinases. Values represent the mean A450 obtained from three experiments after normalization
to Akt1-S473E binding. The error bars show the standard error. Note that although the affinity of the triligand is only marginally improved over the
biligand, the selectivity for Akt1 is clearly enhanced.
stage. To quantify the changes in selectivity for Akt1 and GSK3β,
we carried out a similar ELISA-based binding analysis using
a range of kinase concentrations (Supporting Information,
Figure S18). As we translate from biligand to triligand, the affinity
for Akt1 increases modestly while the affinity for GSK3β is
decreased by approximately 2-fold. This demonstrates that Akt1
specificity, which is lost at the biligand stage, is partially regained
at the triligand stage, mainly through destabilizing interactions
with GSK3β.
The biligand exhibited a significant affinity enhancement for
Akt1, but also a reduced selectivity when assayed against off-
target kinases. The triligand, by contrast, exhibited only a modest
affinity improvement but a significant selectivity enhancement
relative to the biligand. These observations suggest that large
increases in affinity may come at the cost of reduced specificity,
but the increased specificity need not be accompanied by increased
affinity. This is consistent with previous studies on antibodyꢀsmall
molecule interactions,35 DNAꢀprotein interactions,36 and pro-
teinꢀprotein interactions.37ꢀ39 Indeed, a recent study demon-
strated an inverse correlation between affinity and selectivity for
small molecule protein kinase inhibitors.40 For the case of the
capture agents reported here, the tertiary peptide contributes
destabilizing interactions to GSK3β while making only marginal
stabilizing interactions to Akt1. This suggests that the in situ click
product screen approach described here has the potential to
increase the specificity of the multiligand in a manner reminiscent
of negative design.41 This makes intuitive sense. Unlike most
screening methodologies, the product screening strategy employed
here does not define hits as those that bind the protein most
strongly, but instead it defines hits as those that are compatible with
the protein-catalyzed process.
Triligand Inhibition of Akt1. We carried out two standard
enzyme kinetic assays to determine the mode of Akt1 inhibition
by the triligand.42 For such assays, the kinase activity is measured
under varying substrate and triligand concentrations (Figure 4).
The resulting data can be interpreted so that the nature of
the competition between the triligand and the substrate for the
relevant kinase binding site is determined. For example, if
the triligand and ATP competed for the same binding pocket,
the maximum velocity (Vmax) of the kinase would be unchanged
while the Michaelis constant (KM) would increase. The plots can
also be used as a means of determining the inhibition constant
(Ki) of the triligand. The results (Figure 4) show reduction in the
enzyme Vmax for both substrates, as a function of increasing
triligand concentration. This indicates that the triligand does not
directly compete with either substrate for binding to their
respective binding sites. When ATP is the varied substrate,
triligand addition appears to have no effect on KM, consistent
18285
dx.doi.org/10.1021/ja2064389 |J. Am. Chem. Soc. 2011, 133, 18280–18288

Journal of the American Chemical Society
ARTICLE
with noncompetitive inhibition. When the peptide concentration
is varied, KM decreases in a manner consistent with uncompe-
titive inhibition. However, when inhibition constants are deter-
mined from the Dixon plot (varied [ATP]) and the
CornishꢀBowden plot (varied [peptide]), the similarity in
values (4 μM vs 2.6 μM, respectively, Table 1) suggests that
triligand inhibition occurs in a noncompetitive fashion. Regard-
less of the exact mechanism, these experiments rule out compe-
titive inhibition with respect to the ATP and peptide substrates.
The competitive enzyme kinetics studies demonstrate that the
triligand binds to a site distinct from the active (ATP or peptide-
binding) site of the kinase and so achieves its inhibitory
characteristics via an allosteric mechanism. This supports our
hypothesis that blocking the ATP-binding site with Ac7 enabled
the library to sample alternate binding sites with unexpected
functional properties. While the precise mechanism for inhibi-
tion is unknown, competition ELISA experiments suggest that
the multiligand binding site may partially overlap with that of an
anti-Akt1 antibody directed toward the C-terminus (Supporting
Information, Figure S17). The C-terminus is effectively on the
‘backside’ of the kinase domain, opposite the active site.
Biological Applications of the Branched Triligand. The
ultimate test of a protein capture agent is whether or not it can be
used to detect its cognate protein from a biologically relevant
sample. We tested whether the multiligand capture agents could
recognize full-length Akt from cancer cell lines. Previous studies
have shown that Akt2 is overexpressed in the OVCAR3 ovarian
cancer cell line, and so we utilized this cell line as an experimental
platform for immunoprecipitation (IP) and immunocytochem-
ical experiments (ICC).43 For the IP experiments, we immobi-
lized the anchor, biligand, and triligand on streptavidin-agarose
and panned each resin with OVCAR3 cell lysates obtained from
cells stimulated with a combination of epidermal growth factor
(EGF) and insulin or from untreated cells (Figure 5A). Probing
the elutions with pan-Akt antibody, which detects all three
isoforms of Akt, confirms the increased affinity of the biligand
relative to the anchor peptide in lysates from both induced and
noninduced cells. The triligand shows somewhat increased immuno-
precipitation of Akt relative to the biligand in induced cell lysates
but not in the noninduced control cell lysate. This effect is also
observed when the immunoprecipitations are probed with an
antibody specific for Akt phosphorylated Ser473 (Ser474 in
Akt2). Analysis of the total immunoprecipitated protein by
SDS-PAGE showed low nonselective binding for all the ligands
(Supporting Information, Figure S15). A protein at approxi-
mately 50 kDa is immunoprecipitated by the triligand and, to a
lesser extent, by the anchor and biligand. On the basis of our
characterization of triligand specificity, this likely corresponds to
the GSK3 kinase which has a predicted molecular weight of
47 kDa. We are currently investigating the basis for off-target
GSK3 binding which is particularly surprising given the low
sequence homology between Akt1 and GSK3.
These experiments confirm the increase in capture efficiency
as the combined affinity/selectivity metrics of the capture agent
are increased by translating from anchor to biligand to triligand,
particularly in cells stimulated through the PI3K pathway by
insulin and EGF. We compared the performance of our capture
agent against an antibody that was similarly raised against the
kinase domain of Akt1 (Figure 5A, lane 5) for an ‘apples to
apples’ comparison of immunoprecipitation efficiency. Interest-
ingly, the commercial anti-Akt1 antibody shows almost no
immunoprecipitation of Akt. The poor performance of the
antibody may be due to nonoptimal presentation on streptavi-
dinꢀagarose, or some other experimental artifact, but that poor
performance was observed across many assays and across multi-
ple batches of antibody. Even for an equivalently performing
antibody, the ability to use large amounts of triligand at relatively
Figure 4. Inhibition of Akt1 activity. (A) Velocity vs [peptide sub-
strate]with varying concentration of inhibitory triligand. (B) Velocity vs
[ATP] with varying concentration of inhibitory triligand. In both experi-
ments, the Vmax of Akt1 is decreasing as the [triligand] increases, evidence
that the triligand is not competing with either substrate. The figures to the
right of each graph illustrate the conclusion that the triligand was not
competitive with respect to either the peptide or ATP binding site.
Table 1. Kinetic Parameters from Competition Experimentsa
V vs [peptide], variable [triligand]
V vs [ATP], variable [triligand]
K
M (μM) (this work)
9 ( 0.7 (peptide)
3.5 (peptide)
45764 ( 1486
(2.6 ( 0.7)
213 ( 25 (ATP)
155 (ATP)
KM (μM) (literature)
Vmax (pmol/min/mg)
Ki(Ki0) (μM)
98996 ( 8242
4 ( 1
inhibition mode
uncompetitive
noncompetitive
a KM and Vmax values for Akt1-S473E-T308P were obtained by nonlinear regression from the results of the inhibition experiments described above
(GraphPad). The literature values for the Michaelis constants for ATP and substrate peptide were taken from Klein et. al.21 The triligand Ki
(noncompetitive inhibition) was obtained from the negative X-intercept of the Dixon plot of 1/v vs [triligand] with varying [ATP]. The Ki is the mean
value of the observed X-intercepts, and the error range is the standard deviation. The triligand Ki0 (uncompetitive inhibition) was obtained from the
intercept of the CornishꢀBowden plot of [peptide]/v vs [triligand] with varying [peptide]. The Ki0 is the mean value of the intercepts, and the error
range is the standard deviation. These plots as well as additional kinetic analyses are presented in Supporting Information Figures S12 and S13.
18286
dx.doi.org/10.1021/ja2064389 |J. Am. Chem. Soc. 2011, 133, 18280–18288

Journal of the American Chemical Society
ARTICLE
iterative on-bead in situ click chemistry. The triligand can
function as both a capture and detection agent in standard ELISA
and biological assays. The allosteric inhibitory properties of the
triligand highlight an additional therapeutic potential of these
novel capture agents that we are currently investigating.
’ ASSOCIATED CONTENT
S
Supporting Information. Full citations for references 6
b
and 28, synthetic protocols, compound characterization, experi-
mental protocols, and additional data. This material is available
free of charge via the Internet at http://pubs.acs.org.
’ AUTHOR INFORMATION
Corresponding Author
heath@caltech.edu
Present Addresses
School of Natural Sciences, University of California, Merced,
5200 North Lake Rd., Merced, California, 95343, United States.
’ ACKNOWLEDGMENT
This work was supported by the National Cancer Institute
grant No. 5U54CA119347 (J.R.H., P.I.), the Institute for Collabora-
tive Biotechnologies (contract no. W911NF-09-D-0001 from the
U.S. Army Research Office), The Institute of Bioengineering and
Nanotechnology (Biomedical Research Council, Agency for
Science, Technology and Research, Singapore), and the Grand
Duchy of Luxembourg via a subcontract from the Institute for
Systems Biology. S.W.M. acknowledges support from an NRSF
postdoctoral fellowship 1F32CA136150-01. We thank Prof. Carl
Parker for the generous use of his equipment and expertise. The
Akt1-S473E plasmid was a gift from Dr. Shoshana Klein (The
Hebrew University of Jerusalem). Expression of the Akt1-S473E
was carried out by the Protein Expression Center at Caltech.
Figure 5. Immunoprecipitation and immunocytochemical experiments
with OVCAR3 cells. (A) Biotinylated ligands were immobilized on
streptavidinꢀagarose and incubated with lysates from OVCAR3 cell
lines treated with EGF and insulin (induced) or untreated (control).
After 18 h at 4 ꢀC, resins were washed exhaustively, eluted with SDS-
PAGE sample buffer, and analyzed by Western blotting. (1) Blank resin,
(2) anchor, (3) biligand, (4) triligand, (5) [5G3] mAb, (L) lysate. (B)
Immunofluorescent images of Akt in fixed OVCAR3 cells stained with
either a fluorescein-conjugated anti-AKT antibody (Ab) or a fluorescein-
conjugated triligand (TL). Each imaging agent distinguishes cytoplasmic
or membrane-bound AKT in unstimulated or EGF+insulin- treated
cells, respectively. Contrast is enhanced equally in all images, but the
fluorescence images of the triligand stained cells were achieved at a
3-fold lower excitation laser intensity relative to the (much dimmer)
antibody-stained cells. Scale bar = 100 μm.
’ REFERENCES
(1) Borrebaeck, C. A. K. Immunol. Today 2000, 21, 379.
(2) Kodadek, T.; Reddy, M. M.; Olivos, H. J.; Bachhawat-Sikder, K.;
Alluri, P. G. Acc. Chem. Res. 2004, 37, 711.
(3) Kijanka, G.; IpCho, S.; Baars, S.; Chen, H.; Hadley, K.; Beveridge,
A.; Gould, E.; Murphy, D. J. Immunol. Methods 2009, 340, 132.
(4) Michaud, G. A.; Salcius, M.; Zhou, F.; Bangham, R.; Bonin, J.;
Guo, H.; Snyder, M.; Predki, P. F.; Schweitzer, B. I. Nat. Biotechnol.
2003, 21, 1509.
(5) Smith, G. P.; Petrenko, V. A. Chem. Rev. 1997, 97, 391.
(6) Gold, L.; et al. PLoS One 2010, 5, e15004.
(7) Ostroff, R. M.; Bigbee, W. L.; Franklin, W.; Gold, L.; Mehan, M.;
Miller, Y. E.; Pass, H. I.; Rom, W. N.; Siegfried, J. M.; Stewart, A.; Walker,
J. J.; Weissfeld, J. L.; Williams, S.; Zichi, D.; Brody, E. N. PLoS One 2010,
5, e15003.
low cost relative to a monoclonal antibody confers a significant
advantage.
To further explore the interaction between the multiligand
capture agent and full-length Akt, we synthesized a fluorescein-
labeled triligand (Supporting Information, Figure S16) and em-
ployed it as an ICC imaging agent infixed, permeabilized OVCAR3
cells. This construct was benchmarked against a fluorescein-labeled
monoclonal antibody specific for Akt. For this assay, Akt is expected
to locate to the cell membrane following stimulation of receptor
tyrosine kinases by EGF and insulin. Images of unstimulated and
stimulated cells clearly reveal this effect (Figure 5B). Both com-
pounds show localization of Akt at the plasma membrane after cell
stimulation, presumably mediated through binding of the pleck-
strin homology domain (PHD) to phosphatidylinositol (3,4,5)-
trisphosphate at the inner face of the membrane.
(8) Xu, J. L.; Davis, M. M. Immunity 2000, 13, 37.
(9) Zemlin, M.; Klinger, M.; Link, J.; Zemlin, C.; Bauer, K.; Engler,
J. A.; Schroeder, H. W., Jr.; Kirkham, P. M. J. Mol. Biol. 2003, 334, 733.
(10) Chothia, C.; Lesk, A. M.; Tramontano, A.; Levitt, M.; Smith-
Gill, S. J.; Air, G.; Sheriff, S.; Padlan, E. A.; Davies, D.; Tulip, W. R.; et al.
Nature 1989, 342, 877.
(11) Agnew, H. D.; Rohde, R. D.; Millward, S. W.; Nag, A.; Yeo,
W. S.; Hein, J. E.; Pitram, S. M.; Tariq, A. A.; Burns, V. M.; Krom, R. J.;
Fokin, V. V.; Sharpless, K. B.; Heath, J. R. Angew. Chem., Int. Ed. 2009,
48, 4944.
’ CONCLUSIONS
In this work we have demonstrated the design of a peptide-
based drop-in replacement for Akt monoclonal antibodies using
18287
dx.doi.org/10.1021/ja2064389 |J. Am. Chem. Soc. 2011, 133, 18280–18288

Journal of the American Chemical Society
ARTICLE
(12) Manetsch, R.; Krasinski, A.; Radic, Z.; Raushel, J.; Taylor, P.;
Sharpless, K. B.; Kolb, H. C. J. Am. Chem. Soc. 2004, 126, 12809.
(13) Jencks, W. P. Proc. Natl. Acad. Sci. U.S.A. 1981, 78, 4046.
(14) Murray, C. W.; Verdonk, M. L. J. Comput.-Aided Mol. Des. 2002,
16, 741.
(15) Lam, K. S.; Salmon, S. E.; Hersh, E. M.; Hruby, V. J.; Kazmierski,
W. M.; Knapp, R. J. Nature 1991, 354, 82.
(16) Erlanson, D. A.; Braisted, A. C.; Raphael, D. R.; Randal, M.;
Stroud, R. M.; Gordon, E. M.; Wells, J. A. Proc. Natl. Acad. Sci. U.S.A.
2000, 97, 9367.
(17) Shuker, S. B.; Hajduk, P. J.; Meadows, R. P.; Fesik, S. W. Science
1996, 274, 1531.
(18) Vivanco, I.; Sawyers, C. L. Nat. Rev. Cancer 2002, 2, 489.
(19) Altomare, D. A.; Testa, J. R. Oncogene 2005, 24, 7455.
(20) Garcia-Echeverria, C.; Sellers, W. R. Oncogene 2008, 27, 5511.
(21) Klein, S.; Geiger, T.; Linchevski, I.; Lebendiker, M.; Itkin, A.;
Assayag, K.; Levitzki, A. Protein Expr. Purif. 2005, 41, 162.
(22) Meyer, S. C.; Shomin, C. D.; Gaj, T.; Ghosh, I. J. Am. Chem. Soc.
2007, 129, 13812.
(23) Shomin, C. D.; Meyer, S. C.; Ghosh, I. Bioorg. Med. Chem. 2009,
17, 6196.
(24) Obata, T.; Yaffe, M. B.; Leparc, G. G.; Piro, E. T.; Maegawa, H.;
Kashiwagi, A.; Kikkawa, R.; Cantley, L. C. J. Biol. Chem. 2000,
275, 36108.
(25) Souroujon, M. C.; Mochly-Rosen, D. Nat. Biotechnol. 1998,
16, 919.
(26) Samson, I.; Kerremans, L.; Rozenski, J.; Samyn, B.; Van Beeumen,
J.; Van Aerschot, A.; Herdewijn, P. Bioorg. Med. Chem. 1995, 3, 257.
(27) Kirkland, L. O.; McInnes, C. Biochem. Pharmacol. 2009,
77, 1561.
(28) Karaman, M. W.; et al. Nat. Biotechnol. 2008, 26, 127.
(29) Rostovtsev, V. V.; Green, L. G.; Fokin, V. V.; Sharpless, K. B.
Angew. Chem., Int. Ed. 2002, 41, 2596.
(30) Tornøe, C. W.; Christensen, C.; Meldal, M. J. Org. Chem. 2002,
67, 3057.
(31) Niemeyer, C. M.; Adler, M.; Wacker, R. Trends Biotechnol.
2005, 23, 208.
(32) Mocharla, V. P.; Colasson, B.; Lee, L. V.; Roper, S.; Sharpless,
K. B.; Wong, C. H.; Kolb, H. C. Angew. Chem., Int. Ed. 2004, 44, 116.
(33) Whiting, M.; Muldoon, J.; Lin, Y. C.; Silverman, S. M.; Lindstrom,
W.; Olson, A. J.; Kolb, H. C.; Finn, M. G.; Sharpless, K. B.; Elder, J. H.;
Fokin, V. V. Angew. Chem., Int. Ed. 2006, 45, 1435.
(34) Chatterjee, J.; Gilon, C.; Hoffman, A.; Kessler, H. Acc. Chem.
Res. 2008, 41, 1331.
(35) Schildbach, J. F.; Panka, D. J.; Parks, D. R.; Jager, G. C.;
Novotny, J.; Herzenberg, L. A.; Mudgett-Hunter, M.; Bruccoleri,
R. E.; Haber, E.; Margolies, M. N. J. Biol. Chem. 1991, 266, 4640.
(36) Jen-Jacobson, L. Biopolymers 1997, 44, 153.
(37) Nguyen, J. T.; Porter, M.; Amoui, M.; Miller, W. T.; Zuckermann,
R. N.; Lim, W. A. Chem. Biol. 2000, 7, 463.
(38) Nguyen, J. T.; Turck, C. W.; Cohen, F. E.; Zuckermann, R. N.;
Lim, W. A. Science 1998, 282, 2088.
(39) Zarrinpar, A.; Park, S. H.; Lim, W. A. Nature 2003, 426, 676.
(40) Posy, S. L.; Hermsmeier, M. A.; Vaccaro, W.; Ott, K.-H.;
Todderud, G.; Lippy, J. S.; Trainor, G. L.; Loughney, D. A.; Johnson,
S. R. J. Med. Chem. 2011, 54, 54.
(41) Mason, J. M.; Muller, K. M.; Arndt, K. M. Biochemistry 2007,
46, 4804.
(42) Segel, I. H. Enzyme kinetics: behavior and analysis of rapid
equilibrium and steady state enzyme systems; Wiley: New York, 1975.
(43) Yang, L.; Dan, H. C.; Sun, M.; Liu, Q.; Sun, X. M.; Feldman,
R. I.; Hamilton, A. D.; Polokoff, M.; Nicosia, S. V.; Herlyn, M.; Sebti,
S. M.; Cheng, J. Q. Cancer Res. 2004, 64, 4394.
18288
dx.doi.org/10.1021/ja2064389 |J. Am. Chem. Soc. 2011, 133, 18280–18288