The substrate-activity-screening methodology applied to receptor tyrosine kinases: A proof-of-concept study

Article abstract of DOI:10.1016/j.ejmech.2012.08.038

Protein kinases are widely recognized as important therapeutic targets due to their involvement in signal transduction pathways. These pathways are tightly controlled and regulated, notably by the ability of kinases to selectively phosphorylate a defined set of substrates. A wide variety of disorders can arise as a consequence of abnormal kinase-mediated phosphorylation and numerous kinase inhibitors have earned their place as key components of the modern pharmacopeia. Although "traditional" kinase inhibitors typically act by preventing the interaction between the kinase and ATP, thus stopping substrate phosphorylation, an alternative approach consists in disrupting the protein-protein interaction between the kinase and its downstream partners. In order to facilitate the identification of potential chemical starting points for substrate-site inhibition approaches, we desired to investigate the application of Substrate Activity Screening to kinases. We herein report a proof-of-concept study demonstrating, on a model tyrosine kinase, that the key requirements of this methodology can be met. Namely, using peptides as model substrates, we show that a simple ADP-accumulation assay can be used to monitor substrate efficiency and that efficiency can be optimized in a modular manner. More importantly, we demonstrate that structure-efficiency relationships translate into structure-activity relationships upon conversion of the substrates into inhibitors.

Full text of DOI:10.1016/j.ejmech.2012.08.038

European Journal of Medicinal Chemistry 57 (2012) 1e9
Contents lists available at SciVerse ScienceDirect
European Journal of Medicinal Chemistry
journal homepage: http://www.elsevier.com/locate/ejmech
Original article
The Substrate-Activity-Screening methodology applied to receptor tyrosine
kinases: A proof-of-concept study
,
a
a
a
Julien Chapelat a, Frédéric Berst a , Andreas L. Marzinzik , Henrik Moebitz , Peter Drueckes ,
*
Joerg Trappe a, Doriano Fabbro a, Dieter Seebach b
a Novartis Institutes for BioMedical Research, Novartis Campus, CH-4002 Basel, Switzerland
b Eidgenössische Technische Hochschule, CH-8092 Zurich, Switzerland
a r t i c l e i n f o
a b s t r a c t
Article history:
Protein kinases are widely recognized as important therapeutic targets due to their involvement in signal
transduction pathways. These pathways are tightly controlled and regulated, notably by the ability of
kinases to selectively phosphorylate a defined set of substrates. A wide variety of disorders can arise as
a consequence of abnormal kinase-mediated phosphorylation and numerous kinase inhibitors have
earned their place as key components of the modern pharmacopeia. Although “traditional” kinase
inhibitors typically act by preventing the interaction between the kinase and ATP, thus stopping substrate
phosphorylation, an alternative approach consists in disrupting the proteineprotein interaction between
the kinase and its downstream partners. In order to facilitate the identification of potential chemical
starting points for substrate-site inhibition approaches, we desired to investigate the application of
Substrate Activity Screening to kinases. We herein report a proof-of-concept study demonstrating, on
a model tyrosine kinase, that the key requirements of this methodology can be met. Namely, using
peptides as model substrates, we show that a simple ADP-accumulation assay can be used to monitor
substrate efficiency and that efficiency can be optimized in a modular manner. More importantly, we
demonstrate that structureeefficiency relationships translate into structureeactivity relationships upon
conversion of the substrates into inhibitors.
Received 2 May 2012
Received in revised form
22 July 2012
Accepted 26 August 2012
Available online 4 September 2012
Keywords:
Kinase inhibitor
Substrate-Activity-Screening
Proteineprotein interactions
MichaeliseMenten kinetics
Insulin-like Growth Factor 1 Receptor
Ó 2012 Elsevier Masson SAS. All rights reserved.
1. Introduction and design
between the kinase and its peptidic substrates represents another
attractive paradigm to modulate kinase function and several
The role of protein kinases in signal transduction pathways has
been extensively studied over the past years, revealing their ability
to switch various cellular processes on or off. As a consequence of
abnormal kinase-mediated phosphorylation, many diseases arise
such as cancer, diabetes, rheumatoid arthritis and hypertension [1].
This diversity of therapeutic indications has elevated kinases to the
rank of high priority targets for the pharmaceutical industry.
Although most of the efforts in the development of inhibitors
concern ATP mimetic compounds [2], many disadvantages have
been identified such as kinome selectivity, competing against high
concentrations of ATP in cells and an increasingly congested
intellectual property landscape. As a consequence, alternative
approaches for kinase inhibition have been envisaged such as the
elaboration of potent and selective non-ATP site-directed ligands
able to alter kinase function [3]. Disrupting the interactions
substrate-site ligands have been reported over the last decade [4].
However, a significant hurdle hampering the identification of
kinase-substrate proteineprotein interaction inhibitors often lies in
characterizing the binding site and ruling out other allosteric
actions on either the kinase or its substrate [5].
In 2007, Ellman et al. reported a novel approach for the identi-
fication of substrate-targeted phosphatase inhibitors
e the
Substrate Activity Screening (SAS) methodology e that relied on
the identification and optimization of PtpB substrates that were
subsequently turned into inhibitors through replacement of the
reactive phosphate by an inert isostere (Fig. 1) [6]. We envisaged
that the application of this strategy to kinases would have signifi-
cant advantages over the common inhibitor screening methodol-
ogies. Indeed, substrate-site inhibitor starting points are generally
weak, which imposes technical limitations on their detection
(displacement of a weak reference substrate required for detection
and lack of dynamic range). In addition, inhibitor screening typi-
cally provides no information on mode-of-action and additional
studies must be carried out to rule out inadvertent competition
* Corresponding author. Tel.: þ41 (0)61 324 6011; fax: þ41 (0)61 324 8001.
E-mail address: frederic.berst@novartis.com (F. Berst).
0223-5234/$ e see front matter Ó 2012 Elsevier Masson SAS. All rights reserved.
http://dx.doi.org/10.1016/j.ejmech.2012.08.038

2
J. Chapelat et al. / European Journal of Medicinal Chemistry 57 (2012) 1e9
Fig. 2. General strategy for the identification of substrate-targeted tyrosine kinase
inhibitors.
Fig. 1. Identification of PtpB substrate-targeted inhibitor by the SAS methodology.
with ATP, interaction with the competition substrate or an allosteric
mechanism. Conversely, detection of substrate phosphorylation is
a direct, mechanistic readout ensuring that the recognition event
happens in the substrate site. Furthermore, assays can be designed
that amplify the signal resulting from even minute amounts of
phosphorylation, thus facilitating the establishment of structuree
activity relationships (SAR).
As part of a larger program aiming at identifying non-ATP site-
directed inhibitors of tyrosine kinases, we envisaged to adapt the
SAS methodology to this class of receptors and obtain a proof of
concept. The methodology can be summarized in three steps
(Fig. 2): (Step 1) a diverse, modular initial library of phenols is
synthesized by decoration of suitable low molecular weight scaf-
folds. The library is then screened for phosphorylation by the
tyrosine kinase of interest, (Step 2) phenolic hits are optimized in
a multidimensional fashion to increase their substrate efficiency
(the phenol is retained as the assay readout depends on its pres-
ence) and (Step 3) finally, the phenolic moiety is removed and the
substrate is thereby turned into an inhibitor.
Fig. 3. Tetrapeptide 1 identified from IGF-1R substrate profiling [7].
the discovery phenol library can easily be added to routine
screening campaigns. Poor dissociation of the phosphorylated
product from the kinase (product inhibition) was perceived as
another potential source of false negatives, albeit mitigated by the
fact that phosphorylated substrates usually rapidly dissociate from
the protein, thus ensuring that the binding site is free for the next
substrate molecule.
In order to obtain a proof of concept, it must be shown that: (a)
an assay with sufficient dynamic range can be identified to detect
minute amounts of phosphorylation, (b) substrate phosphorylation
can be optimized in a modular fashion, (c) a correlation can be
established, under certain conditions, between the phosphoryla-
tion efficiency of the substrate (KM, substrate) and its ability to bind to
the kinase (KD, substrate) and (d) the binding SAR observed for the
substrate can be transferred to the corresponding inhibitors (KD,
inhibitor). It is important to note that this methodology has the
ambition to provide a suitable starting point for further medicinal
chemistry optimization and does not claim to be a comprehensive
screening method. Indeed, the scaffold decoration approach
inherently biases chemical matter. Moreover, poor substrates but
otherwise suitable ligands, in other words competent inhibitors,
would not be identified by the methodology. We did not foresee
this to be an issue, since kinase inhibition assays are typically part
of the usual arsenal of a kinase medicinal chemistry program and
2. Results and discussion
To set the stage for our study, we required a well-characterized
kinase-substrate pair as well as an enzymatic assay with a suitable
dynamic range. We therefore decided to make use of previous work
carried out in our laboratory, which led to the identification of an
efficient tetrapeptide substrate 1 of Insulin-like Growth Factor 1
Receptor (IGF-1R) (Fig. 3) [7]. This allowed us to circumvent the
costly synthesis of
a diverse, modular low-molecular-weight
phenol library to provide the initial substrate hit from “step 1”.
Moreover, we had already shown that the kinetic enzymatic assay
(ADP QuestÔ [8]) used to profile the substrate specificity of IGF-1R
enabled the determination of KM constants down to 2.5 mM while
still satisfying the MichaeliseMenten assumptions [9], thus making
it an ideal tool for our purpose.
Module 1
Module 2
Fig. 4. Modified optimization strategy: using tetramer 1 as initial hit.

J. Chapelat et al. / European Journal of Medicinal Chemistry 57 (2012) 1e9
3
Scheme 1. Solid phase synthesis of the M2-library of substrates. a) DMA:Piperidine (4:1), 3 ꢀ 30 min then Fmoc-Trp(Boc)-OH, HCTU, DIEA, NMP, 2 ꢀ 90 min; b) DMA:Piperidine (4:1),
3 ꢀ 30 min then Fmoc-Glu(tBu)-OH, HCTU, DIEA, NMP, 90 min; c) DMA:Piperidine (4:1), 3 ꢀ 30 min then Fmoc-Phe-OH, HCTU, DIEA, NMP, 90 min; d) DMA:Piperidine (4:1), 3 ꢀ 30 min
then Fmoc-Tyr(tBu)-OH, HCTU, DIEA, NMP, 90 min; e) DMA:Piperidine (4:1), 3 ꢀ 30 min then Module 2-COOH, HCTU, DIEA, NMP, 90 min; f) TFA:TIS:H2O (95:2.5:2.5), 3 ꢀ 30 min.
According to our proposal, the tetrapeptide 1 was formally
decomposed into a phenol-containing scaffold (tyrosine) flanked
by two diversity elements: module 1 (M1: FEW-NHMe) and
module 2 (M2: acetyl, Fig. 4). We set out to demonstrate that M1
and M2 could be optimized independently and that the combina-
tion of optimal moieties would lead to the most efficient substrate.
The next logical step to obtain final proof-of-concept would be to
confirm the transfer of substrate SAR to inhibitor SAR upon
replacement of the phenol moiety.
We therefore synthesized a first library incorporating 125
structurally diverse carboxylic acid modules 2 by Fmoc/tBu-Boc
solid-phase peptide synthesis on a Prelude peptide synthesizer
(Protein Technologies Inc.). The indole-functionalized acid-labile
solid support 2 ensured that each substrate candidate would be
Fig. 5. Primary screen of module 2 building blocks. Substrates have been grouped in bins according to their V0(10
normalized with M2 ¼ acetyl (x ¼ 1).
m
M)

4
J. Chapelat et al. / European Journal of Medicinal Chemistry 57 (2012) 1e9
Table 1
Module 2 optimization: MichaeliseMenten constants for a selection of substrates.
Entry Module 2
Substrate KM
(
m
M)a VMax
(nM sꢁ1
No phosphorylationb
V0 10
Bin
m
M
5
4
3
a
a
)
(nM sꢁ1
)
1
2
3
4
5
H-
Ac-
4
1
3
5
6
232 ꢂ 29 4.4 ꢂ 0.9
16.0 ꢂ 4.2 13.7 ꢂ 0.3
114 ꢂ 15 27.6 ꢂ 5.7
0.17 ꢂ 0.07
6.2 ꢂ 0.1
2.5 ꢂ 0.1
1
4
3
2
Ac-Asp-
Ac-Glu-
COOHe(CH2)2e
r² 0.8277
144 ꢂ 19 23.3 ꢂ 1.5 1.45 ꢂ 0.4
6
7
> 1000c
n.d.c
0.4 ꢂ 0.1
1.1 ꢂ 0.1
1
7
8
8
9
204 ꢂ 27 21.7 ꢂ 2.9
2
6
8
9
10
log (1/V0(10 µM)) (s.M-1)
8.6 ꢂ 1.2
12.8 ꢂ 2.5 10.9 ꢂ 0.1
Fig. 6. Correlation between log (1/V0(10 M)) and pKM values for a selection of 50
m
substrates. Dot colour correlates with the bins in Fig. 5. (For interpretation of the
references to colour in this figure legend, the reader is referred to the web version of
this article.)
9
10
11
12
230 ꢂ 19 24.3 ꢂ 3.3 0.84 ꢂ 0.04
2
4
4
and largely derived from purely hydrophobic or positively-charged
moieties. A small portion of the library exhibited initial phos-
phorylation velocities comparable to that of the optimized
substrate Ac-DYFEW-NHMe 3 (Table 1, entry 3) with five candidates
being phosphorylated at a faster initial rate.
10
11
15.1 ꢂ 0.2 11.1 ꢂ 2.2
28.7 ꢂ 6.9 16.9 ꢂ 1.1
5.4 ꢂ 0.1
4.6 ꢂ 0.1
In order to verify our assumption that a correlation could be
found between initial velocities at 10 mM and KM, the Michaelise
Menten constant was recorded for a diverse selection of 50
substrates randomly selected from each bin. A good correlation
(r2 ¼ 0.82, Fig. 6) was obtained between pKM and log (1/V0(10 M)),
m
12
13
36.9 ꢂ 0.9 13.3 ꢂ 1.1
4.3 ꢂ 0.1
4
thus confirming that the initial rate of phosphorylation was an
adequate, easily accessible surrogate readout for substrate
efficiency.
a
The most efficient substrate in the library made use of (S,S)-
cyclohexyl-1,2-dicarboxylic acid as module 2 (9, Table 1, entry 8),
affording a 16-fold improvement of substrate 1 (Table 1, entry 2)
and a 2-fold improvement of positive-control substrate 3 (Table 1,
entry 3) [10]. Interestingly, rigidification of the vector delivering
a carboxylate anion to the protein formally allowed for the removal
of a hydrogen bond donoreacceptor moiety in the form of the N-
acetyl group of 3 (Fig. 7). It is worth noting that the enantiomeric
module 2 derived from (R,R)-cyclohexyl-1,2-dicarboxylic acid led to
a very poor substrate of IGF-1R (8, Table 1, entry 7). A similar
situation was observed in the cyclopentyl series (9 and 10, Table 1,
entries 9 and 10), providing us with first indications of a responsive
SAR for substrates not entirely composed of proteinogenic amino
acids.
Having observed dynamic effects on substrate efficiency upon
variation of module 2, we turned our attention to alterations of
module 1. For this exercise, we decided to keep module 2 constant
as N-acetyl aspartic acid. We anticipated that tryptophan variations
were most likely to give rise to responsive SAR in a proof of concept
study, thus making this residue an ideal point of diversity. However,
the nature of the solid-phase peptide synthesis used for the elab-
oration of the substrates imposed that the tryptophan replacement
be introduced as the first chemical step on the resin. We envisaged
that a fragment coupling strategy would greatly reduce the
numbers of individual manipulations necessary for the elaboration
of each substrate. It is however well known that fragment coupling
Kinetics were recorded as 3 independent experiments using the ADP QuestÔ
assay (See Experimental section).
b
No phosphorylation was detected with substrate concentration up to 1 mM.
An accurate value could not be determined due to the limited solubility of the
c
substrate above 1 mM.
released from the resin as the N-methyl carboxamide (Scheme 1).
In addition, Ac-Asp-OH was added to the diversity set to serve as an
internal control for the methodology since we had previously
identified the pentapeptide Ac-DYFEW-NHMe 3 (Scheme 1, Module
2 ¼ Ac-Asp-) as a very efficient IGF-1R substrate [7].
Each library member was then subjected to phosphorylation by
IGF-1R in our optimized ADP QuestÔ assay. Although the deter-
mination of KM values is the method of choice, it would be
impractical for larger libraries. As we accepted that our readout
sacrifice those substrate candidates for which the phosphorylation
reaction was sluggish (either through poor recognition, slow
phosphate transfer or slow product dissociation), we chose to
classify the library members according to initial phosphorylation
velocity at a concentration of 10 mM. We indeed surmised that good
substrate recognition would be a prerequisite for fast kinase-
catalysed reactions and that this plate-based readout would allow
rapid identification of the best candidates.
Gratifyingly, a broad range of substrate behaviour was detected
by the assay (Fig. 5). As expected, a majority of modules 2 led to
relatively incompetent substrates falling within the “acetyl” bin

J. Chapelat et al. / European Journal of Medicinal Chemistry 57 (2012) 1e9
5
Fig. 7. Structure of the M2-optimized substrate 9.
can lead to racemisation of the fragment’s carboxy terminus upon
activation. Replacement of the glutamic acid residue in module 1 by
glycine would enable a fragment coupling strategy without any risk
of racemisation. We therefore synthesized the substrate candidate
Ac-DYFGW-NHMe 15 as a new reference substrate (M2: Ac-D; M1:
FGW-NHMe).
little value to our effort. Thus, a collection of 35 substrates was
synthesized according to Scheme 2. The fragment 14 common to all
substrate candidates was obtained on large scale by Fmoc/tBu-Boc
solid-phase peptide synthesis on (2-chloro)-trityl chloride poly-
styrene resin. A set of 35 N-Fmoc-protected amino acid tryptophan
replacements were independently loaded onto resin 2. Following
N-Fmoc deprotection, the 35 resins were derivatized with fragment
14. Subsequent deprotection and cleavage from the support affor-
ded the substrate array.
Satisfactorily, the KM value of 15 was determined to be 70 mM,
a value well situated in the dynamic range of the ADP QuestÔ assay
and offering a suitable window for the determination of tryptophan
replacement SAR. We decided to limit ourselves to hydrophobic
variations, assuming from previous work [7] and known structural
information [11] that ablation of the key Van-der-Waals interac-
tions would largely lead to inactive substrates, which would add
Phosphorylation rates at 10 mM substrate concentration were
then recorded [12]. Perhaps not surprisingly, it appeared difficult to
improve upon nature’s choice of a hydrophobic moiety using
essentially the same delivery vector in the form of an amino acid
Scheme 2. Strategy for the synthesis of the M1 library. a) Fmoc-Gly-OH, DIEA, DCM, 2 ꢀ 90 min; b) DMA:Piperidine (4:1), 3 ꢀ 30 min then Fmoc-Phe-OH, HCTU, DIEA, NMP, 90 min;
c) DMA:Piperidine (4:1), 3 ꢀ 30 min then Fmoc-Tyr(tBu)-OH, HCTU, DIEA, NMP, 90 min; d) DMA:Piperidine (4:1), 3 ꢀ 30 min then Fmoc-Asp(tBu)-OH, HCTU, DIEA, NMP, 90 min; e)
DMA:Piperidine (4:1), 3 ꢀ 30 min then DMA:Ac2O:pyridine (8:1:1), 90 min; f) DCM:HFIP (4:1), 2 ꢀ 60 min; g) DMA:Piperidine (4:1), 3 ꢀ 30 min then Fmoc-Xaa-OH, HCTU, DIEA,
NMP, 2 ꢀ 90 min; h) DMA:Piperidine (4:1), 3 ꢀ 30 min then 14, HCTU, DIEA, NMP, 2 ꢀ 90 min; i) TFA:TIS:H2O (95:2.5:2.5), 3 ꢀ 30 min.

6
J. Chapelat et al. / European Journal of Medicinal Chemistry 57 (2012) 1e9
Table 2
Module 1 optimization: Replacement of the tryptophan residue on Ac-DYFG-Xaa-
NHMe substrates.
Entry W-replacement
Substrate
KM
(
m
M)a
VMax
V0 10
mM
a
a
(nM sꢁ1
)
(nM sꢁ1
1.7 ꢂ 0.1
5.2 ꢂ 1.2 0.09 ꢂ 0.01
)
1
2
-Trp-
-D-Trp-
15
16
70.3 ꢂ 1.1
907 ꢂ 14
8.9 ꢂ 0.8
3
17
18
19.3 ꢂ 3.7
6.9 ꢂ 0.7
2.1 ꢂ 0.1
4
5
473 ꢂ 21
12.7 ꢂ 2.1 0.24 ꢂ 0.05
Efficiency
19
20
106 ꢂ 9
4.7 ꢂ 1.3 0.55 ꢂ 0.02
6
156 ꢂ 21
8.2 ꢂ 1.2 0.47 ꢂ 0.05
a
Kinetics were recorded as 3 independent experiments using the ADP QuestÔ
Fig. 9. Selection of the nine combination substrates 22(R31; R23).
assay (See Experimental section).
side chain. Nevertheless, a handful of candidates actually under-
went phosphorylation faster than the tryptophan-containing
parent substrate. Among those, the substrate obtained with 6-
chlortryptophan underwent phosphorylation at the fastest rate.
Again, we determined the MichaeliseMenten constants of a repre-
sentative set to check the V0eKM correlation (Table 2) and confirm
that 6-Cl-tryptophan substrate 17 was indeed the most efficient
one (Table 2, entry 3).
Finally, glutamic acid was re-introduced in place of glycine and,
as expected, an improvement of the KM value was observed across
all peptides, with the 6-chlorotryptophan derivative 21 being again
the most efficient substrate (Fig. 8) [13].
Fig. 8. Structure of the W-optimized substrates 17 and 21.

J. Chapelat et al. / European Journal of Medicinal Chemistry 57 (2012) 1e9
7
Table 3
a consequence, the combination of the best modules led to the best
substrate 22(R31; R23) in terms of V0 and KM. Due to a limitation
inherent to the assay format [9], an accurate value for the KM of
22(R31; R32) could unfortunately not be precisely determined.
Nevertheless, the stage was set for applying the final step of our
methodology: the conversion of substrates into inhibitors and the
confirmation that substrate SAR translates into inhibitor SAR.
Two tyrosine phosphorylation transition state mimetics have
been described in the literature: the hydroxymethyltyrosine
phosphate 23 of Hangauer et al. [14] and the meta-iodo-tyrosine 24
used by Lam et al. [15] to convert c-Src kinase substrates into
inhibitors (Fig. 10). We decided to use the hydroxymethyltyrosine
phosphate derivative 23 to eliminate the risk that residual phos-
phorylation of meta-iodo-tyrosine complicate assay readouts. A
protected version of 23 was synthesized according to published
procedures [16], starting from Cbz-Tyr(Tf)-Bzl 26. The final
compound 25 was obtained as a mixture of two diastereoisomers.
The nine substrates 22(R1i ; Rj ) chosen for the SAR additivity
study were converted to the corr2esponding nine inhibitors (Fig. 11,
MichaeliseMenten constants of the M1-M2 combination substrates 22(R31; R23).
Entry
Substrate
KM
(
m
M)a
VMax
V0 10
mM
a
a
(nM sꢁ1
)
(nM sꢁ1
)
1
2
3
4
5
6
7
8
9
22(R31; R32)
22(R31; R22)
22(R31; R12)
22(R21; R32)
22(R21; R22)
22(R21; R12)
22(R11; R32)
22(R11; R22)
22(R11; R12)
<5
m
Mb
11.1 ꢂ 0.4
10.9 ꢂ 0.5
5.2 ꢂ 0.9
9.9 ꢂ 1.0
8.7 ꢂ 1.8
n.dc
7.5 ꢂ 0.1
3.3 ꢂ 0.1
1.0 ꢂ 0.1
4.0 ꢂ 0.1
0.26 ꢂ 0.04
n.dc
23.5 ꢂ 1.0
35.3 ꢂ 4.0
15.4 ꢂ 3.9
278 ꢂ 72
>1 mMc
188 ꢂ 18
9.1 ꢂ 1.9
0.51 ꢂ 0.01
>1 mMc
n.dc
n.dc
>1 mMc
n.dc
n.dc
a
Kinetics were recorded as 3 independent experiments using the ADP QuestÔ
assay (See Experimental section).
b
The lower limit of the assay was reached.
An accurate value could not be determined due to the limited solubility of the
c
substrate above 1 mM.
At this stage of the study, we had demonstrated that the rapid
initial velocity readout in our simple ADP accumulation assay
system was able to provide information on the structureeefficiency
relationship around variations of two independent diversity
elements. We decided to validate the modular substrate optimisa-
tion approach by verifying the additivity of SAR across modules. To
this end we combined three modules 1 with three modules 2,
making sure to choose moieties having a marked influence on
substrate efficiency. Nine combination substrates were synthesized
on indole-functionalized resin 2 (Fig. 9). The substrates were
submitted to kinase-mediated phosphorylation, and initial veloci-
27(Ri1; Rj )) by substitution of tyrosine with mimetic 23. The initial
2
peptide hit 3 was also converted to the corresponding inhibitor 28
(R1 ¼ W, R2 ¼ Ac-D) and served as benchmark. The ability of each
inhibitor to impair IGF-1R-mediated phosphorylation of a common
generic substrate was then assessed in a standard ATP depletion
assay (Table 4) [17]. To our delight, the efficiency trends observed
for the substrates translated into comparable inhibitory trends with
the most efficient substrate 22(R31; R23) giving rise to the best
inhibitor 27(R31; R32). This finding was the definite proof that the SAS
methodology can be applied to tyrosine kinases and has value as
a modular methodology for the generation of substrate-site kinase
inhibitors.
ties at 10
mM as well as MichaeliseMenten parameters were
determined (Table 3). We were satisfied to see that the synergy
between the two modules was exclusively additive and that, as
Fig. 10. Tyrosine mimetics described in the literature and the building block 25 used in our synthetic sequence.
Fig. 11. Inhibitors synthesized for the validation of the SAS methodology.

8
J. Chapelat et al. / European Journal of Medicinal Chemistry 57 (2012) 1e9
Table 4
submitted to two consecutive 30 min treatments with a solution of
the first Fmoc-protected amino acid (3 equiv., 0.3 M), HCTU (3
equiv., 0.3 M) and DIEA (4.5 equiv. 0.9 M) in NMP. The resin was
washed twice with DMA and capped with a solution of DMA:A-
c2O:pyridine (8:1:1) for 30 min. The resin was finally washed three
times with DMA. This protocol was repeated for the loading of each
amino acid or M2eCOOH. In the case of an amino acid as N-
terminus, the resin was acetylated by removing the Fmoc group
with 20% piperidine in DMA (three times 15 min) followed by
treatment with a solution of DMA:Ac2O:pyridine (8:1:1) for 30 min.
The compound was then released by two consecutive treatment of
the beads with a solution of TFA:TIS:H2O (95:2.5:2.5) for 2 h. The
cleavage solutions were then combined and the solvents were
removed under reduced pressure. The crude product was then
purified by preparative reversed-phase HPLC (Waters, SunfireÔ C18
IC50 values of the M1eM2 combination inhibitors 27(Ri1; Rj ).
2
Entry
Inhibitor
IC50 (m
M)a
1
2
3
4
5
6
7
8
9
27(R31; R32)
27(R31; R22)
27(R31; R12)
27(R21; R32)
27(R21; R22)
27(R21; R12)
27(R11; R32)
27(R11; R22)
27(R11; R12)
28
2.5 ꢂ 0.8
21.6 ꢂ 6.6
94 ꢂ 29
14.0 ꢂ 2.5
40.5 ꢂ 4.6
489 ꢂ 29
22.8 ꢂ 5.7
51.3 ꢂ 8.1
78.2 ꢂ 6.5
15.6 ꢂ 3.0
10
a
IC50’s were recorded as 3 independent experiments using KinaseGloÔ assay
(See Experimental section).
OBDÔ 5
m
M, 30 ꢀ 100 mm) using gradients of water (þ0.1% TFA)
3. Conclusion
and acetonitrile at a flow rate of 30 mL/min and using a DAD UV
detector. Product purity was assessed by UPLC on Waters Acquity
analytical instruments. The structure of the pure peptidic
substrates and inhibitors were confirmed by ESI-MS and by HR-MS
spectroscopy. Conditions, instrument specifications and results are
listed in the Supplementary material.
To summarize, we have shown that a substrate of a receptor
tyrosine kinase can be optimized in a modular, combinatorial
manner. Notably, we have shown that two independent points of
diversity can be easily varied to increase substrate efficiency and
that the combination of favourable elements led to increased turn-
over in an additive manner. Finally, we have shown that structuree
efficiency relationships of the substrates translated well into
structureeactivity relationships of the inhibitors upon replacement
of the phosphorylation warhead by an inactive mimetic. Having
successfully shown, with model peptidogenic compounds, that this
methodology has potential in the field of kinase inhibition, we now
turn our attention to the generation of large, modular libraries of
phenol-containing small-molecules. We are anticipating that these
libraries will provide initial, drug-like compounds to feed the
optimization methodology. A second research arm in our labora-
tories aims at identifying new tyrosine mimetics compatible with
cell wall permeation, and results will be disclosed in due course.
Compound stock solutions were then prepared using
DMSO:H2O (75:25).
4.1.3. Synthesis of fully protected fragment 14
(2-Chloro)-trityl chloride resin (15.2 g,1 mmol/g) was swelled in
DCM and then treated with Fmoc-Gly-OH (9.02 g, 30.3 mmol, 2
equiv.) and DIEA (10.6 mL, 60.7 mmol, 4 equiv.) in DCM (150 mL) for
4 h at 25 ꢃC. The resin was washed twice with DCM:MeOH:DIEA
(17:2:1), twice with DCM and twice with DMA. The next amino
acids (Fmoc-Phe-OH, Fmoc-Tyr(tBu)-OH and Fmoc-Asp(tBu)-OH)
were loaded using classical SPPS procedures: the Fmoc protecting
group was removed using 20% piperidine in DMA (150 mL, three
times 15 min) and the resin was washed twice with DMA. The resin
was then shaken with the Fmoc-protected amino acid (45.6 mmol,
3 equiv.), HCTU (14.6 g, 45.6 mmol, 3 equiv.) and DIEA (11.9 mL,
68.4 mmol, 4.5 equiv.) in NMP (150 mL) for 15 h at rt. The beads
were washed twice with DMA and treated with a capping solution
made of DMA:Ac2O:pyridine (8:1:1, 150 mL) for 30 min. The resin
was finally washed thrice with DMA. After the last amino acid
coupling, the Fmoc group was removed using 20% piperidine in
DMA (150 mL, three times 15 min) and the resin was washed twice
with DMA. The resin was then shaken with DMA:Ac2O:pyridine
(8:1:1, 150 mL) for 30 min and washed thrice with DMA. The pro-
tected fragment was cleaved off the resin by treatment with
DCM:hexafluoroisopropanol (4:1, 150 mL, two times 60 min) and
the solvents were removed under vacuum. The crude product was
purified by reverse phase chromatography on a Büchi Purification
System using a gradient of acetonitrile (35e100% in 20 min) in
water (þ0.1% TFA) at a flow of 50 mL/min to afford 14 (6.36 g) as
a white powder after removal of solvents and lyophilization.
4. Experimental section
4.1. Chemistry
4.1.1. General remarks
Reagents were used as received and chemicals of the quality
purum, purum p. a. or >98% were used without further purification.
All solvents were at least reagent grade. The {3-[(methyl-Fmoc-
amino)-methyl]-indol-1-yl}-acetyl AM resin 2 was purchased from
EMD Chemicals Inc. Evaporations were carried out under reduced
pressure. MS (ESI) were obtained from a Waters SQ Detector and
the accurate mass measurements (HR-MS) were performed by
using electrospray ionization in positive ion modus after separation
by liquid chromatography. The elemental composition was derived
from the averaged mass spectra acquired at the high resolution of
about 30,000 on an LTQ Orbitrap XL mass spectrometer (Thermo
Scientific). The high mass accuracy below 1 ppm was obtained by
using a lock mass. NMR spectra’s of compounds 14 and 27(R31; R23)
were recorded on a Bruker spectrometer pulsed at 600 MHz at
298 K in DMSO-d6.
Yield: 64%; 1H NMR (600 MHz, DMSO-d6)
d ppm 13.2e11.9 (brs,
1H, COOH), 8.16 (t, J ¼ 5.67 Hz, 1H), 8.04 (dd, J ¼ 11.2, 8.6 Hz, 2H),
7.67 (d, J ¼ 8.05 Hz, 1H), 7.24e7.21 (m, 4H), 7.18 (m, 1H), 7.03 (d,
J ¼ 8.42 Hz, 2H), 6.80 (d, J ¼ 8.42 Hz, 2H), 4.57e4.50 (m, 2H), 4.43e
4.39 (m,1H), 3.78 (d, J ¼ 5.85 Hz, 2H), 3.05 (dd, J ¼ 13.9, 4.76 Hz,1H),
2.89 (dd, J ¼ 13.9, 4.39 Hz, 1H), 2.83 (dd, J ¼ 13.9, 9.15 Hz, 1H), 2.70
(dd, J ¼ 13.9, 9.15 Hz, 1H), 2.53 (dd, J ¼ 10.3, 5.49 Hz, 1H), 2.30 (dd,
J ¼ 15.7, 8.78 Hz, 1H), 1.79 (s, 3H), 1.36 (s, 9H), 1.26 (s, 9H); UPLC
4.1.2. General procedure for the SPPS on the indole-functionalized
resin 2
The peptides and semi-peptidic compounds were synthesized
on a Prelude peptide synthesizer (Protein Technologies Inc.)
following standard SPPS procedures. The indole-functionalized
solid support 2 was swelled in DMA and the Fmoc group
removed using three consecutive treatments with 20% piperidine
in DMA for 15 min. The resin was washed twice with DMA and then
(Waters Acquity, UPLC HSS T3 1.8
1.2 mL/min, 2
NH4OAc) in acetonitrile (þ0.05% HCOOH) in 9.4 min, DAD TIC 210
e 350 nm): Rt ¼ 3.75 min, >95% DAD TIC; MS (ESI) 655.4 for
[M þ H]þ (calcd 655.3 for C34H46N4O9).
m
m, 2.1 ꢀ 50 mm, flow rate:
m
L injection, 2e98% water (þ0.05% HCOOH þ 0.05%

J. Chapelat et al. / European Journal of Medicinal Chemistry 57 (2012) 1e9
9
4.1.4. Characterization of fully optimized inhibitor 27(R31; R23)
1H NMR (600 MHz, DMSO-d6)
ppm 10.98 (brs, 1H), 8.15 (dd,
1 m
M f.c. and incubated at 25 ꢃC for 25 min. KinaseGloÔ reagent
d
was added and the reaction was incubated for 30 min at 25 ꢃC.
Relative luminescence was recorded on a Victor 2Ô plate reader
(PerkineElmer). IC50 values were directly obtained by using the
GraphPad PrismÒ software. n ¼ 3.
J ¼ 12.08, 7.68 Hz, 1H), 7.93 (d, J ¼ 6.59 Hz, 1H), 7.85 (d, J ¼ 7.68 Hz,
1H), 7.83e7.78 (m, 2H), 7.56 (d, J ¼ 8.78 Hz, 1H), 7.36 (s, 1H), 7.31e
7.23 (m, 6H), 7.19 (t, J ¼ 7.00, 6.95 Hz, 1H), 7.17 (brs, 1H), 7.09 (d,
J ¼ 7.68 Hz, 2H), 6.99 (d, J ¼ 8.42 Hz, 1H), 4.65 (d, J ¼ 13.54 Hz, 1H),
4.51e4.46 (m,1H), 4.43 (q, J ¼ 6.95 Hz,1H), 4.39e4.33 (m,1H), 4.30e
4.24 (m, 1H), 3.13e3.01 (m, 2H), 2.99e2.88 (m, 3H), 2.74e2.65 (m,
1H), 2.54 (d, J ¼ 4.02 Hz, 3H), 2.48e2.42 (m, 1H), 2.42e2.36 (m, 1H),
2.22 (t, J ¼ 7.87 Hz, 2H), 2.01e1.95 (m, 1H), 1.84e1.85 (m, 1H), 1.82e
1.73 (m,1H),1.71e1.55 (m, 3H),1.29e1.10 (m, 3H),1.09e0.98 (m,1H);
Acknowledgements
We thank Novartis Pharma AG and the Program Office for
funding this research, Dr. Patrick Chêne, Dr. Philipp Grosche, Ste-
phanie Pickett, Lionel Muller, Nicole Martin and Daniela Manfrina
for their collaboration and contribution to this project. We are also
grateful to Dr. Christian Guénat, Dr. Ingo Muckenschnabel and Régis
Denay for analytical support.
UPLC (Waters Acquity, UPLC HSS T3 1.8
1.2 mL/min, 2
m
m, 2.1 ꢀ 50 mm, flow rate:
m
L injection, 2e98% water (þ0.05% HCOOH þ 0.05%
NH4OAc) in acetonitrile (þ0.05% HCOOH) in 9.4 min, DAD TIC 210
e 350 nm): Rt ¼ 2.97 min, >95% DAD TIC; MS (ESI) 939.2 for
[M
þ
H]þ (calcd 939.3 for C44H52ClN6O13P); HR-MS (ESI)
Appendix A. Supplementary data
939.30873 for [M þ H]þ (calcd 939.30917 for C44H52ClN6O13P).
Supplementary data related to this article can be found at http://
dx.doi.org/10.1016/j.ejmech.2012.08.038.
4.2. Enzymatic assays
4.2.1. Expression and purification of IGF-1R tyrosine kinase
References
The cloning, expression and purification of both GST-His6-tag-
ged kinases was performed as reported previously [18]: Briefly, the
cDNA encoding for the entire cytoplasmic domain of human IGF-1R
(aa 960e1367) was cloned into the pFastBacGST2-PreScission
vector, and a His6-tag was introduced at the C-terminus of the
coding sequences. Proteins were expressed in Spodoptera frugiperda
Sf9 cells. For the production of unphosphorylated protein, vector
encoding for the phosphatase YopH was co-expressed with the IGF-
1R vector. The soluble protein obtained from cell lysates was
purified by two-step affinity chromatography purification (Gluta-
thione sepharose 4B and His-Trap HP columns). Protein concen-
tration was determined by the Bradford method, and the purity of
the protein was determined by SDS-PAGE analysis and HPLC.
[1] P. Cohen, Nat. Rev. Drug Disc 1 (2002) 309e315.
[2] (a) S.P. Davies, H. Reddy, M. Caivano, P. Cohen, Biochem. J. 351 (2000) 95e105;
(b) J. Bain, H. McLauchlan, M. Elliott, P. Cohen, Biochem. J. 371 (2003) 199e
204.
[3] For a review on allosteric inhibitors see R.M. . Eglen, T. Reisine, Expert Opin.
Drug Discov. 5 (2010) 277e290.
[4] (a) S.M. Sebti, Q. Cheng, A.D. Hamilton, K. Kayser-Bricker US2010/0009397 A1
(2010). (b) Litman, et al., Biochemistry 46 (2007) 4716e4724;
(c) Steiner, et al., Eur. J. Pharm. 562 (2007) 1e11.
[5] G. Ye, R. Tiwari, K. Parang, Curr. Opin. Investig. Drugs 9 (2008) 605e613.
[6] M.B. Soellner, K.A. Rawls, C. Grundner, T. Alber, J.A. Ellman, J. Am. Chem. Soc.
129 (2007) 9613e9615.
[7] J. Chapelat, F. Berst, A.L. Marzinzik, H. Moebitz, P. Drueckes, J. Trappe,
D. Seebach, Bioorg. Med. Chem. Lett. 21 (2011) 7030e7033.
[8] N.W. Charter, L. Kauffman, R. Singh, R.M. Eglen, J. Biomol. Screen. 11 (2006)
390e399.
[9] The assay system is based on the detection of ADP formed during the course
of the phosphorylation reaction and the lowest ADP concentration required
for detection is 500 nM. This implied that the minimal substrate concentration
still enabling accurate measurements while maintaining low substrate
4.2.2. Assessment of MichaeliseMenten kinetics with ADP QuestÔ
assay
IGF-1R kinase was pre-activated as follows: 15 mM HEPES (pH
7.4), 20 mM NaCl, 1 mM EGTA, 0.02% Tween 20, 10 mM MgCl2,
conversion (<20%) was situated around 2.5
[10] The KM value measured for substrate 9 approached the technical limit of
2.5 M imposed by the ADP-QuestÔ assay (see refs. [8,9]), which would hinder
mM.
m
0.1 mg/mL BSA, 1850 ng/mL GST-His6-IGF-1R and 1.5 mM ATP were
any further optimization of module 2. Although we did not want to increase the
efficiency at this point in time in our peptide-based proof-of-concept study, we
recognized that this limitation might prove problematic when the technology
is transferred to small molecule chemical space. We therefore decided to
artificially raise the KM of the substrate by alteration of module 1. This was
expected to serve the dual purpose of increasing the dynamic range of the
assay, should this be necessary, and of ascertaining that modification of both
modules lead to additive SAR. We had previously determined that the tryp-
tophan residue in substrate 3 was key for recognition by IGF-1R. We therefore
chose ten substrates from the 125-membered library covering a wide range of
KM values, synthesized the tryptophan-to-phenylalanine mutants and deter-
mined their MichaeliseMenten constants. The values are available in the
Supplementary material. As anticipated, given the relatively linear conforma-
tion required for kinase-mediated peptide phosphorylation, we observed that
the KM values were uniformly shifted towards less efficient substrates upon
substitution of the tryptophan residue by phenylalanine.
incubated for 60 min at 25 ꢃC, then aliquoted and frozen at e 80 ꢃC.
Kinase reactions were carried out in 384-well plates: 15 mM HEPES
(pH 7.4), 20 mM NaCl, 1 mM EGTA, 0.02% Tween 20, 10 mM MgCl2,
0.1 mg/mL BSA, 0.4 ng/
semi-peptidic substrate at concentration ranging from 1 to 1000
and ADP QuestÔ detection reagents were mixed together. Reactions
were then initiated by the addition of ATP at 60 M f.c. and relative
mL pre-activated GST-His6-IGF-1R, peptidic or
mM
m
fluorescence was recorded every 2 min for 60 min on a Victor 2Ô
plate reader (PerkineElmer). MichaeliseMenten constants were
directly obtained by using the GraphPad PrismÒ software. n ¼ 3.
For assessment of initial phosphorylation rates at 10
mM
substrate concentration V0(10 M), the same protocol was used with
m
[11] The importance of the tryptophan residue and its interaction with the kinase
was based on IGF-1R crystal structure PDB ID 1k3a.
peptidic or semi-peptidic substrate at 10 mM f.c.
[12] For details see the Supplementary material.
4.2.3. Assessment of inhibitor IC50’s with KinaseGloÔ assay
IGF-1R kinase was pre-activated as follows: 15 mM HEPES (pH
7.4), 20 mM NaCl, 1 mM EGTA, 0.02% Tween 20, 10 mM MnCl2,
[13] In order to verify the additivity of the SAR observed with glycine, a set of 6
Trp-modified substrates was synthesized with glutamic acid and the
MichaeliseMenten kinetics were measured and compared with the corre-
sponding glycine substrates. The results are available in the Supplementary
material.
[14] M.H. Kim, J.H. Lai, D.G. Hangauer, Int. J. Pept. Protein Res. 44 (1994) 457e465.
[15] J. Alfaro-Lopez, W. Yuan, B.C. Phan, J. Kamath, Q. Lou, K.S. Lam, V.J. Hruby,
J. Med. Chem. 41 (1998) 2252e2260.
[16] Z.-J. Yao, Y. Gao, T.R. Burke Jr., Tetrahedron: Asymmetry 10 (1999) 3727e3734
(The detailed synthetic sequence is described in the Supplementary material).
[17] The KinaseGloÔ assay from Promega Corp. was used to determine the IC50’s
of IGF-1R inhibitors. For detailed procedure the Supplementary material.
[18] D. Erdmann, C. Zimmermann, P. Fontana, J.-C. Hau, A. De Pover, P. Chène,
J. Biomol. Technol. 21 (2010) 9e17.
0.1 mg/mL BSA, 277 ng/mL GST-His6-IGF-1R and 50 mM ATP were
incubated for 30 min at 25 ꢃC, then aliquoted and frozen at ꢁ80 ꢃC.
Kinase reactions were carried out in 384-well plates: 15 mM HEPES
(pH 7.4), 20 mM NaCl, 1 mM EGTA, 0.02% Tween 20, 10 mM MnCl2,
0.1 mg/mL BSA, 0.5 ng/
mL pre-activated GST-His6-IGF-1R, 5 mM Ac-
EAEDEPEGDYFEWLE-NHMe substrate [7] and inhibitor 27(Ri1; Rj )
2
at concentrations ranging from 250
mM to 114 nM were mixed
together. Reactions were then initiated by the addition of ATP at