Deconstruction of a nutlin: Dissecting the binding determinants of a potent protein-protein interaction inhibitor

Article abstract of DOI:10.1021/ml400062c

Protein-protein interaction (PPI) systems represent a rich potential source of targets for drug discovery, but historically have proven to be difficult, particularly in the lead identification stage. Application of the fragment-based approach may help tow

Full text of DOI:10.1021/ml400062c

Letter
pubs.acs.org/acsmedchemlett
Deconstruction of a Nutlin: Dissecting the Binding Determinants of a
Potent Protein−Protein Interaction Inhibitor
David C. Fry,*,† Charles Wartchow,† Bradford Graves,† Cheryl Janson,† Christine Lukacs,†
Ursula Kammlott,† Charles Belunis,† Stefan Palme,‡ Christian Klein,§ and Binh Vu†
†Roche Research Center, 340 Kingsland Street, Nutley, New Jersey 07110, United States
‡Roche Diagnostics GmbH, Nonnenwald 2, Penzberg 82377, Germany
§Roche Glycart AG, Wagistrasse 18, Schlieren CH-8952, Switzerland
S
* Supporting Information
ABSTRACT: Protein−protein interaction (PPI) systems represent a
rich potential source of targets for drug discovery, but historically have
proven to be difficult, particularly in the lead identification stage.
Application of the fragment-based approach may help toward success
with this target class. To provide an example toward understanding the
potential issues associated with such an application, we have
deconstructed one of the best established protein−protein inhibitors,
the Nutlin series that inhibits the interaction between MDM2 and p53, into fragments, and surveyed the resulting binding
properties using heteronuclear single quantum coherence nuclear magnetic resonance (HSQC NMR), surface plasmon
resonance (SPR), and X-ray crystallography. We report the relative contributions toward binding affinity for each of the key
substituents of the Nutlin molecule and show that this series could hypothetically have been discovered via a fragment approach.
We find that the smallest fragment of Nutlin that retains binding accesses two subpockets of MDM2 and has a molecular weight
at the high end of the range that normally defines fragments.
KEYWORDS: Nutlin, protein−protein interaction inhibitor, p53, MDM2, binding affinity
fragment library will emerge as future drug discovery projects
nhibiting protein−protein interactions (PPI) with small
Imolecules is a difficult objective but could potentially lead to on this target class are pursued and reported upon. In the
a wide variety of novel and important therapeutics.1,2 There are
meantime, a complementary way of adding to our knowledge
base is to perform retrospective analyses of successful programs.
That is, to deconstruct known protein−protein inhibitors into
successively smaller fragments and survey their potency and
binding locations, and then compare these attributes to those of
the parent compounds.
several possible pathways toward the discovery of such
molecules, including high-throughput screening of large
compound libraries to obtain initial leads. Traditionally, these
libraries have consisted of compounds in the molecular weight
range 200−500 Da. More recently, a strategy employing
libraries comprised only of small compounds, the fragment-
based approach,3,4 has been gaining popularity. Practitioners
have settled on a similar set of characteristics for the fragments
comprising their libraries, with a molecular weight range
established at 100−300 Da. However, protein−protein
interaction systems represent a unique class of drug target,
and it has already been shown that successful inhibitors of
protein−protein interactions tend to have certain properties
that distinguish them from drugs that act against more
conventional target classes. For example, they are larger and
more three-dimensional.5,6 Therefore, it is an open and vital
question whether fragments meant to serve as potential leads
for protein−protein interaction targets should also have
properties distinct from those of conventional fragments.
For selected PPI targets, the results of fragment screens have
been reported,7 and hits have been described, but no overriding
analysis has appeared comparing the properties of these PPI
fragment hits to fragment hits from non-PPI systems. An
answer to the question of what constitutes an optimal PPI
This strategy has already been applied.8−10 At Abbott, a very
potent inhibitor of the Bcl-2 protein family was developed,
designated ABT-737, and it ultimately entered the clinic as a
potential cancer therapeutic. As commonly found for protein−
protein inhibitors, its molecular weight, 813 Da, was
substantially higher than what is commonly expected for a
drug. In a retrospective study, compounds comprising portions
of ABT-737 were obtained and were checked for activity, and
the smallest piece that still exhibited binding was identified.8
The molecular weight of this smallest active fragment was 293
Da.
Interestingly, a plot of binding affinity vs molecular weight
for this series of fragments produced a linear slope, and this
relationship was confirmed in studies with additional targets.
Therefore, one can use these data to predict the kinds of
Received: February 13, 2013
Accepted: May 24, 2013
© XXXX American Chemical Society
A
dx.doi.org/10.1021/ml400062c | ACS Med. Chem. Lett. XXXX, XXX, XXX−XXX

ACS Medicinal Chemistry Letters
Letter
fragments that should be screened to find a good lead for a
protein−protein interaction target. It was found that the
affinities of the smallest active fragments were all in the range of
50−300 μM. If it is assumed that an acceptably potent drug
candidate (1−10 nM) for a protein−protein target will have a
typically high molecular weight (700−800 Da), then the
fragment lead will need to have a molecular weight of about
300 Da, which is at the upper limit of the size range typical of
fragment libraries.
In a related study, the Krimm group at the University of
Lyon performed deconstruction analyses of ABT-737 and a
variety of other published Bcl-2 family inhibitor scaffolds.9
While the previous Abbott study considered only scaffolds that
were eventual successes, that is, were optimized into true drug
candidates with desirable potency and PK properties, the
Krimm study did not apply this restriction. In studies using
ligand- and protein-based nuclear magnetic resonance (NMR)
methods, binding was observed for several fragments derived
from nine parent scaffolds, and some of these fragments
possessed low molecular weights, in the range 120−230 Da.
A recent report10 described fragmentation of a small
molecule inhibitor of the interaction between the von
Hippel−Lindau protein (pVHL) and the alpha subunit of
hypoxia-inducible factor 1 (HIF-1α), where the parent
compound had a molecular weight of 410 Da. The study
found that the smallest fragment that exhibited detectable
binding by ligand-based NMR and thermal shift methods had a
molecular weight of 262 Da. Further, binding could only be
detected for fragments capable of occupying two adjacent
subpockets at the interface. This collection of deconstruction
studies of protein−protein systems is quite limited, so it is hard
to draw practical generic conclusions, and consequently,
guidelines for PPI fragment screening are currently undefined.
On one hand, it appears that small (<200 Da) fragments can be
found that bind to PPI targets. However, how promising are
these fragment hits, that is, are they capable of evolving into
bona fide drug candidates?
In order to contribute an example toward the understanding
of how fragment size and structure relate to ultimate success in
a drug discovery program, we have performed a deconstruction
study of the Nutlins. The Nutlins constitute a distinct class of
protein−protein inhibitor with a unique chemotype, and they
have achieved high potency and successfully entered clinical
trials.11 These compounds bind to the protein MDM2 and
block its interaction with the multifunctional transcription
factor p53. This enhances the overall level of p53 activity and
thereby prevents cancer cells from evading apoptosis. In the
study we report here, we have systematically deconstructed
RG7112, the first member of the Nutlin family to enter clinical
trials,12 into successively smaller fragments, and investigated the
ability of these fragments to bind to MDM2 using surface
plasmon resonance (SPR) (Figure 1), NMR (Figure 2), and X-
ray crystallography (Figure 3). This investigation represents a
valuable additional case toward answering the question: for a
fragment library targeting protein−protein interactions, what
key properties are shared by successful fragment leads?
Figure 1. SPR sensograms (frame 1 on left) and plot of response vs
concentration (frame 2 on right) for the binding of 1 to MDM2.
1
Figure 2. Representative 2D H−15N-HSQC NMR spectra, showing
MDM2 in the empty state (black) and following the binding of 10
(red).
MDM2 has been reported.14 It shows that, upon binding, the
peptide adopts an alpha helical conformation and achieves
affinity by inserting three hydrophobic side-chains from one
side of this helix, Phe19, Trp23, and Leu26, into three
corresponding subpockets of MDM2 (Figure 3A). An X-ray
structure of 1 bound to MDM2 has been solved12 and
establishes that 1, like all of the Nutlins, utilizes its tripod-like
shape to efficiently insert substituents into these three
subpockets. In the case of 1, the Phe, Trp, and Leu side-chains
of p53 are effectively mimicked by, respectively, an ethoxy
group and two chlorophenyl groups (Figure 3A).
Nutlins also feature an appendage, in this case a piperazine
derivative, extending from the N1 atom of the imidazoline core,
which has been shown to exert a major influence on activity.17
This appendage typically does not make significant contact with
the surface of MDM2, but rather projects outwardly into
solvent. It is not obvious exactly how this moiety contributes to
binding affinity, but its role may be to sterically direct a
substituent into the Phe subpocket and/or to provide a cap that
shields the hydrophobic interactions at the binding interface
from solvent.
We have performed a systematic deconstruction of 1,
obtaining compounds that represent successively smaller
fragments of the parent (their structures are shown in Table
1). Because of practical limitations of chemical synthesis, not
every conceivable fragment was achievable. However, this is a
minor compromise, and the set of compounds provides an
adequate coverage of the fragmentation pathways possible for
1.
The parent compound chosen for this study, RG7112 (1), is
one of the most potent Nutlins developed to date, and its
structure is shown in Table 1. Nutlins exert their inhibitory
activity by binding to MDM2 and directly competing with the
binding of the p53 protein. The binding of p53 to MDM2 can
be fully replicated by a peptide fragment of p53 composed of
residues 15−29.13 The X-ray structure of this peptide bound to
B
dx.doi.org/10.1021/ml400062c | ACS Med. Chem. Lett. XXXX, XXX, XXX−XXX

ACS Medicinal Chemistry Letters
Letter
Figure 3. X-ray crystal structures of MDM2 complexed with the parent Nutlin (1) (A) and various fragments: 9 (B); 10 (C); and 5 (D). In panel A,
there is also a superimposed complexed p53 peptide14 depicted as a green ribbon with selected side chains shown as sticks.
For the set of fragments, SPR and NMR were applied
independently to examine binding to MDM2. SPR was used to
assess binding and to determine Kd values (Table 1) and off-
rates. A representative SPR sensogram, showing binding of the
parent compound to MDM2, is shown in Figure 1. In the SPR
studies, the binding profiles for the noninteracting compounds
were in general very distinct from the compounds that showed
responses with MDM2, and nonspecific binding of the
fragments to MDM2 was not observed. Two-dimensional
1H−15N-HSQC NMR was also used to assess and verify
binding (Table 1). This form of protein-observe NMR is well
established as one of the gold standard methods for sensitively
detecting binding and for determining whether that binding is
authentic, that is, does not involve unfolding, aggregation, or
other undesirable effects on the protein. The HSQC method
does not suffer from the higher false positive rates seen with
SPR, ligand-observe NMR (in cases where verification by
competition is not possible), and other biophysical techniques
used to monitor binding in PPI systems. Furthermore, the
pattern and magnitude of chemical shift changes observed in
the HSQC experiment allows an experienced practitioner to
judge between a binding mode based on significant insertion
into the protein versus a more superficial type of binding
involving only surface interactions.18 This latter form of
binding, while resulting in classification as a hit in various
methods, is unlikely to be capable of advancement toward
improved potency. In addition, the HSQC NMR experiment
provides a low-resolution footprint that indicates where on the
target protein the ligand is interacting. This is important for
determining if a fragment is binding in the same region it
occupies when it is part of the intact parent compound. An
exemplary superposition of two HSQC spectra, showing a
spectrum of empty MDM2 overlaid with a spectrum acquired
following the addition of one of the fragments, is pictured in
Figure 2.
We also attempted X-ray crystallography for the fragments
that were found to bind. While an NMR footprint is adequate
for distinguishing whether a compound is located in the
primary binding cleft or elsewhere, it is hard to determine the
exact position of a compound and its orientation using chemical
shift changes alone. X-ray crystallography can provide a precise
picture of where and how the compound is bound to the
C
dx.doi.org/10.1021/ml400062c | ACS Med. Chem. Lett. XXXX, XXX, XXX−XXX

ACS Medicinal Chemistry Letters
Letter
Table 1. Structures, Molecular Weights, and MDM2 Binding Activity for RG7112 and Its Fragments
compd
1
2
3
4
5
6
7
8
9
10
11
MW
727.78
yes
70.09
no
112.56
no
246.35
no
305.2
yes
236.31
no
356.89
yes
478.65
no
495.48
yes
565.56
yes
554.75
yes
NMR binding?
SPR binding? (Kd, μM)
ligand efficiency
0.22
0.18
no
no
no
26
no
no
no
20
14
1000
0.10
0.31
0.19
0.18
protein. We were able to obtain structures for several of the
fragments in complex with MDM2 via cocrystallization (Figure
3).
The Nutlin molecule can be conceptualized as an imidazoline
core with four R groups, as depicted on the far right in Figure 4.
Toward the other extreme, fragments that retained three of
the four R groups were all found to bind (compounds 9, 10,
and 11). NMR indicated that these compounds were all
binding in the active site cleft of MDM2. However, affinities
varied among these fragments. Compound 10 comprising the
Trp and Leu mimics coupled with the hydrophilic cap was the
most potent, exhibiting a Kd of 14 μM. Similarly, the other
fragment containing the Trp and Leu mimics, in this case
coupled with the Phe mimic (9), was also relatively potent,
exhibiting a Kd of 20 μM. The fragment containing the Phe
mimic and the cap (11), with the Trp mimic added, in this case
lacking the para-chloro substituent, was a much weaker binder,
with a Kd of 1 mM. Nevertheless, this set of compounds
indicates that a Nutlin fragment containing three R groups, in
any combination, is capable of binding to MDM2. X-ray
structures of compounds 9 and 10 bound to MDM2 (Figure
3B,C) show that these fragments attain a position as expected
based on the Nutlin binding paradigm, exhibiting the same
orientation and utilizing the same subpocket-filling strategy as
the parent (Figure 3A).
The binding of fragments that retained only two of the four
R groups was more varied. A fragment consisting of the
combination of the Phe mimic and the cap (8) did not bind
significantly to MDM2. However, a fragment combining the
Trp mimic and the Leu mimic (5) was able to bind. NMR
indicated that the binding of this fragment was authentic and
that it was located in the active site cleft. The affinity of this
compound, which was found to be 26 μM, was not far from
that exhibited by fragments possessing three R groups. X-ray
crystallography verified that 5 maintains the binding strategy of
the parent, as it inserts the expected substituents into the Trp
and Leu subpockets (Figure 3D). A fragment (7) was produced
that combined one chlorophenyl substituent with the Phe
mimic; because of rapid interconversion, the chlorophenyl
would be capable of occupying either the Trp or Leu
subpocket. This compound produced a chemical shift
Figure 4. Hypothetical pathways by which a Nutlin could have been
developed from a fragment hit, in accordance with the experimental
data from the present study.
There is a substituent (designated RTrp, RLeu, and RPhe) to
occupy each of the three key subpockets of MDM2 (as shown
schematically in Figure 1) and one substituent (RCap) that
projects toward solvent and ostensibly caps the binding site.
Our results show that when the parent Nutlin (1) was broken
down to the level of individual R groups (compounds 3 and 4)
or to the isolated imidazoline core (compound 2) none of these
exhibited significant binding to MDM2, within the concen-
tration limits of the SPR and NMR experiments (specific values
are found in the Methods section of the Supporting
Information). It appears that these fragments are too small to
form sufficient interactions with the protein to sum up to a
measurable binding affinity.
D
dx.doi.org/10.1021/ml400062c | ACS Med. Chem. Lett. XXXX, XXX, XXX−XXX

ACS Medicinal Chemistry Letters
Letter
perturbation in NMR indicative of binding to MDM2, with its
location in the active site cleft. However, the binding was
apparently too weak to be confirmed by SPR performed at the
testable limit of 1 mM. Correspondingly, attempts to obtain
cocrystals of this compound with MDM2 were unsuccessful.
Overall, this set of compounds established that Nutlin
fragments containing two of the four parent R groups were
capable of binding to MDM2. The particular combination of R
groups was found to have a dramatic influence on activity,
causing affinities to range from an undetectable level to a Kd of
26 μM.
hypothetical pathway would start with detection of binding of
compound 5 by NMR, SPR, or some other method, and
verification as a true binder by HSQC NMR. Subsequent
chemical elaboration could have led to one of the versions
containing three R groups, all of which were found to be
capable of binding. The addition of the fourth R group would
have resulted in a complete and highly potent Nutlin. The
optimization process could have been guided throughout by X-
ray structures, as the initial hit and all the key derivatives along
the pathway were found to be competent for cocrystallization
with MDM2.
In conclusion, our study supports the notion that protein−
protein interaction systems should be highly amenable to a
fragment-based lead discovery approach, although these
systems will likely require some specialized choice of library
composition.
The ligand efficiency (LE)15 value of 0.31 found for fragment
5 is within the range recommended for an acceptable fragment
hit.16 However, expansion to the larger fragments 9, 10, and 11
resulted in a drop to LE values of 0.10−0.19, and although a
commonly accepted goal during drug optimization is to
maintain LE values near 0.3, these values are comparable to
those of the parent Nutlin, which despite its low efficiency
possessed all of the properties needed for entry into clinical
trials.
The conformation of the MDM2 protein was basically the
same among all of the complexed structures, with the exception
of the area near the Phe subpocket. As can be seen in Figure 3,
when the bound ligand has an appendage filling the Phe
subpocket, Tyr63 of MDM2 is flipped away, and Met58 is
oriented inward (Figure 3A,B); while, when the ligand is
lacking a Phe mimic, Tyr63 flips inward and Met58 is pushed
away (Figure 3C,D).
ASSOCIATED CONTENT
* Supporting Information
■
S
Experimental methods for chemical synthesis, SPR, NMR, and
X-ray crystallography. This material is available free of charge
via the Internet at http://pubs.acs.org.
AUTHOR INFORMATION
Corresponding Author
*(D.C.F.) Tel: 973-235-3709. Fax: 973-235-8897. E-mail: fry.
davidc@gmail.com.
■
Notes
Having identified 5 as the smallest Nutlin fragment capable
of binding to MDM2, we investigated whether a trimmed-down
version of 5 would still be able to bind. A derivative of 5 was
prepared, which lacked the para-chloro substituents (com-
pound 6). It was found to be incapable of binding to MDM2.
This underscores the high importance of the para-chloro
substituents in the dissection of binding determinants for
RG7112.
The authors declare the following competing financial
interest(s): The authors are all current or former employees
of Hoffmann-La Roche, Inc.
ACKNOWLEDGMENTS
■
We gratefully acknowledge the efforts of Santina Russo and
Joachim Diez of Expose GmbH in collecting the diffraction data
at the SLS.
The molecular weight of the smallest fragment of Nutlin that
retains binding competency was established at 305 Da. This
value is at the high end of the molecular weight range that
normally defines fragments. This finding compares closely with
that obtained for ABT-737, another small molecule inhibitor of
a protein−protein interaction that has achieved clinical entry.
Further, the finding that the smallest Nutlin fragment
competent to bind occupies two adjacent subpockets on
MDM2 is consistent with the earlier deconstruction study of
the pVHL inhibitor,10 where only fragments capable of
accessing two subpockets exhibited detectable binding. While
this pair represents a small sample size, there may be a trend
emerging with respect to the composition of fragment
screening libraries aimed against protein−protein interaction
targets, namely, it may be advantageous to skew them toward
higher molecular weights. As fragment screening is performed
increasingly against protein−protein systems, it will be
educational to collectively analyze the results to see if this
trend is subtantantiated. Since the purpose of employing small
compounds is to more efficiently sample chemical space, an
increase in size is undesirable as it will offset this advantage by
requiring many more compounds to achieve suitable coverage.
Nevertheless, this may turn out to be a necessary trade-off.
In retrospect, it appears that the Nutlin series of MDM2
inhibitors could have been discovered via a fragment-based
approach (Figure 4), although the library would require
compounds with molecular weights over 300 Da. One
REFERENCES
■
(1) Wells, J. A.; McClendon, C. L. Reaching for high-hanging fruit in
drug discovery at protein−protein interfaces. Nature 2007, 450, 1001−
1009.
(2) Fry, D. C. Targeting Protein−Protein Interactions for Drug
Discovery. In Protein−Protein Interactions: Methods and Applications,
2nd ed.; Fu, H., Meyerkord, C., Eds.; Springer: New York, 2013; in
press.
(3) Murray, C. W.; Rees, D. C. The rise of fragment-based drug
discovery. Nat. Chem. 2009, 1, 187−192.
(4) Scott, D. E.; Coyne, A. G.; Hudson, S. A.; Abell, C. Fragment-
based approaches in drug discovery and chemical biology. Biochemistry
2012, 51, 4990−5003.
(5) Sperandio, O.; Reynes, C. H.; Camproux, A. C.; Villoutreix, B. O.
Rationalizing the chemical space of protein−protein interaction
inhibitors. Drug Discovery Today 2010, 15, 220−229.
(6) Fry, D. C.; So, S.-S. Modulators of Protein−Protein Interactions:
The Importance of Three-Dimensionality. In Protein−Protein Inter-
actions in Drug Discovery; Doemling, A., Ed.; Wiley-VCH: Weinhelm,
Germany, 2012; pp 55−62.
(7) Bower, J. F.; Pannifer, A. Using fragment-based technologies to
target protein−protein interactions. Curr. Pharm. Des. 2012, 18, 4685−
4696.
(8) Hajduk, P. J. Fragment-based drug design: how big is too big? J.
Med. Chem. 2006, 49, 6972−6976.
(9) Barelier, S.; Pons, J.; Marcillat, O.; Lancelin, J.-M.; Krimm, I.
Fragment-based deconstruction of Bcl-XL inhibitors. J. Med. Chem.
2010, 53, 2577−2588.
E
dx.doi.org/10.1021/ml400062c | ACS Med. Chem. Lett. XXXX, XXX, XXX−XXX

ACS Medicinal Chemistry Letters
Letter
(10) Van Molle, I.; Thomann, A.; Buckley, D. L.; So, E. C.; Lang, S.;
Crews, C. M. Dissecting fragment-based lead discovery at the von
Hippel−Lindau Protein:Hypoxia Inducible Factor 1α protein−protein
interface. Chem. Biol. 2012, 19, 1300−1312.
(11) Vassilev, L. T.; Vu, B. T.; Graves, B.; Carvajal, D.; Podlaski, F.;
Filipovic, Z.; Kong, N.; Kammlott, U.; Lukacs, C.; Klein, C.; Fotouhi,
N.; Liu, E. In vivo activation of the p53 pathway by small-molecule
antagonists of MDM2. Science 2004, 303, 844−848.
(12) Tovar, C.; Graves, B.; Packman, K.; Higgins, B.; Xia, M.;
Tardell, C.; Garrido, R.; Lee, E.; Kolinsky, K.; To, K.-H.; Linn, M.;
Podlaski, F.; Wovkulich, P.; Vu, B.; Vassilev, L. MDM2 small molecule
agonist RG7112 activates p53 signaling and regresses human tumors in
preclinical cancer models. Cancer Res. 2013, 73, 2587−2597.
(13) Picksley, S. M.; Vojtesek, B.; Sparks, A.; Lane, D. P.
Immunochemical analysis of the interaction of p53 with MDM2:
fine mapping of the MDM2 binding site on p53 using synthetic
peptides. Oncogene 1994, 9, 2523−2529.
(14) Kussie, P. H.; Gorina, S.; Marechal, V.; Elenbaas, B.; Moreau, J.;
Levine, A. J.; Pavletich, N. P. Structure of the MDM2 oncoprotein
bound to the p53 tumor suppressor transactivation domain. Science
1996, 274, 948−953.
(15) Hopkins, A. L.; Groom, C. R.; Alex, A. Ligand efficiency: a
useful metric for lead selection. Drug Discovery Today 2004, 9, 430−
431.
(16) Schultes, S.; de Graaf, C.; Haaksma, E. E. J.; de Esch, I. J. P.;
Leurs, R.; Kramer, O. Ligand efficiency as a guide in fragment hit
selection and optimization. Drug Discovery Today Technol. 2010, 7,
e157−e162.
(17) Vu, B.; Wovkulich, P.; Pizzolato, G.; Lovey, A.; Ding, Q.; Jiang,
N.; Liu, J.-J.; Zhao, C.; Glenn, K.; Wen, Y.; Tovar, C.; Packman, K.;
Vassilev, L.; Graves, B. Discovery of RG7112: a small-molecule
MDM2 inhibitor in clinical development. ACS Med. Chem. Lett. 2013,
4, 466−469.
(18) Williamson, M. P. Using chemical shift perturbation to
characterise ligand binding. Prog. Nucl. Magn. Reson. 2013, 73, 1−16.
F
dx.doi.org/10.1021/ml400062c | ACS Med. Chem. Lett. XXXX, XXX, XXX−XXX