Druggable protein–protein interactions – from hot spotsto hot segmentsNir London1, Barak Raveh2 and Ora Schueler-Furman3
Available online at www.sciencedirect.com
ScienceDirect
Protein–Protein Interactions (PPIs) mediate numerous
biological functions. As such, the inhibition of specific PPIs has
tremendous therapeutic value. The notion that these
interactions are ‘undruggable’ has petered out with the
emergence of more and more successful examples of PPI
inhibitors, expanding considerably the scope of potential drug
targets. The accumulated data on successes in the inhibition of
PPIs allow us to analyze the features that are required for such
inhibition. Whereas it has been suggested and shown that
targeting hot spots at PPI interfaces is a good strategy to
achieve inhibition, in this review we focus on the notion that the
most amenable interactions for inhibition are those that are
mediated by a ‘hot segment’, a continuous epitope that
contributes the majority of the binding energy. This criterion is
both useful in guiding future target selection efforts, and in
suggesting immediate inhibitory candidates – the dominant
peptidic segment that mediates the targeted interaction.
Addresses1 Department of Pharmaceutical Chemistry, University of California, San
Francisco, San Francisco, CA 94158, USA2 Department of Bioengineering and Therapeutic Sciences, University of
California, San Francisco, San Francisco, CA 94158, USA3 Department of Microbiology and Molecular Genetics, Institute for
Medical Research Israel-Canada, Hadassah Medical School, The
Hebrew University, POB 12272, Jerusalem 91120, Israel
Corresponding author: Schueler-Furman, Ora ([email protected])
Current Opinion in Chemical Biology 2013, 17:952–959
This review comes from a themed issue Synthetic biomolecules
Edited by Shang-Cheng Hung and Derek N Woolfson
For a complete overview see the Issue and the Editorial
Available online 31st October 2013
1367-5931/$ – see front matter, # 2013 Elsevier Ltd. All rights
reserved.
http://dx.doi.org/10.1016/j.cbpa.2013.10.011
IntroductionAccurate communication between different proteins is
critical to the proper functioning of any living cell. The
interaction between protein partners not only establishes
macromolecular machineries such as ribosomes, poly-
merases and proteasomes, but also mediates transient
signaling pathways, and timely regulatory processes.
Protein–Protein Interactions (PPIs) can be mediated by
the classical interaction between two globular protein
domains. Alternatively, as much as 40% of the PPIs are
estimated to involve peptide–domain interactions,
through the binding of a short peptidic linear binding
Current Opinion in Chemical Biology 2013, 17:952–959
motif to a globular protein domain [1], or even by the co-
folding of these linear motifs within two unstructured
regions [2].
For some time, protein-protein interfaces were considered
relatively flat and featureless (inspired among others by the
seminal analysis of protein–protein interfaces by Thornton
in 1996 [3]), and therefore hard to ‘drug’ using small
molecules that prefer well-defined binding pockets [4].
However, from early on it was also shown that even
seemingly featureless protein–protein interfaces contain
‘hot spot’ residues [5–7]. These often involve large amino
acids such as Tyrosine, Arginine and Tryptophan that bind
in small pockets across the interface and contribute the
major part of the binding interaction energy. Additional
studies that characterized protein–protein interfaces
revealed a higher-level organization of these interfaces
[8]. The prevalence of hot spot residues in both
domain–domain and peptide–domain interfaces [9�] made
it theoretically feasible to disrupt protein–protein (and
peptide-protein) interactions with small molecules that
target their binding sites. In recent years, the number of
successful attempts at inhibiting PPIs with small mol-
ecules has increased considerably (and also been exten-
sively reviewed by e.g. [10–13,14��] and many others). In
this review, we survey distinctive features of protein–protein interfaces that are amenable for inhibition, and
posit that these druggable PPIs are often dominated by ‘hot
segments’, an extension of the concept of hot spots to a
continuous binding epitope that dominates the interaction
[15��]. The prevalence of hot segments in druggable PPIs
highlights peptides and their derivatives as a suitable class
of PPI inhibitors. We discuss various techniques for further
optimization that can propel these starting points into
potential drugs.
Characteristics of successfully inhibited PPIsFollowing the accumulation of examples of small mol-
ecules that inhibit PPIs, databases such as TIMBAL [16],
2P2I [17] and iPPI-DB [18] have been dedicated to the
collection of small molecule PPI inhibitors and relevant
structural data on these PPIs. The 2P2I database
organizes inhibited PPIs in two classes: peptide–domain
interactions and domain–domain interactions [19�] (see
Figure 1 and Table 1). A structural comparison of these
complexes to other heterodimeric complexes suggested
that the druggable 2P2I interfaces involve a much smaller
buried surface area, are more hydrophobic, and do not
show major conformational changes upon binding. Inter-
estingly, we previously found that these features, in
www.sciencedirect.com
Hot spot segments define druggable protein interactions London, Raveh and Schueler-Furman 953
Figure 1
Protein-Protein Interactions
Domain-Domain Domain-Peptide
Estimated15-40%
Estimated>50%
No DominantHot Segment
DominantHot Segment
Amenable to inhibitionby:
&Small
molecules
Peptides
Current Opinion in Chemical Biology
Protein–protein interactions that are peptide-mediated or contain an
interface hot-segment are suitable targets for inhibition. A substantial
fraction of interactions (estimated up to 40% [1]) are mediated by a linear
peptide stretch (domain–peptide interactions). Among the remaining
domain–domain interactions, about half of the protein–protein
interactions with solved structures contain one contiguous interface hot
segment that contributes the dominant part of binding energy (as
estimated using the PeptiDerive protocol [15��], see text).
particular the latter, are also characteristic of peptide–protein interactions [9�]. Another structural analysis of
inhibited PPIs divided the interactions into four classes,
based on whether they are narrow/wide and tight/loose
[14��]. Among these, the narrow (surface area <2500 A2)
and tight (Kd < 200 nM) PPIs are more amenable to
inhibition. In turn, the inhibition of narrow and loose
interactions often involves conformational changes and
allosteric effects. Since these are more difficult to model
and design, they are usually identified in a posterioristructural analyses of inhibitors detected by high through-
put screening (HTS) approaches [14��].
A large share of PPIs are dominated by a hotsegmentAs evident from the large proportion of peptide-
mediated interactions that are targeted by small mol-
ecule inhibitors (see Table 1), this prevalent form of
interaction seems particularly suitable for inhibition.
Peptide-mediated interactions are obviously dominated
by a continuous binding epitope (namely, the peptide).
More surprisingly however, this holds also for domain–domain interactions (e.g. interactions between two glob-
ular protein partners): these too are often dominated by
one hot segment.
www.sciencedirect.com
We have previously developed a protocol that, based on a
solved complex structure, identifies the contiguous pep-
tide epitope within a protein that contributes most to
binding (Rosetta PeptiDerive [15��]). In a systematic
analysis of standard benchmarks of domain–domain inter-
actions using PeptiDerive, we estimated that more than
50% of those interactions are mediated by one hot seg-
ment that contributes the majority of the binding inter-
action energy [15��] (Figure 1). Application of the same
methodology to the entire Protein Data Bank (PDB; [20])
has led Teyra et al. [21�] to estimate that almost 700
protein families are dominated by a hot segment at the
interface of the protein–protein interaction. In a similar
study that was restricted to a-helix mediated interactions,
Jochim and Arora [22�] highlighted a set of �400 com-
plexes that contain a dominant a-helix at the interface.
Hot segments are good predictors of PPIdruggabilityTo examine the relation between druggability and the
existence of hot segments, we apply here the aforemen-
tioned PeptiDerive protocol to the druggable protein–protein complexes enlisted in the 2P2I database. The
results suggest that indeed, one contiguous peptide dom-
inates the interaction in practically all of these reported
cases (Table 1). Trivially, for all 7 peptide-protein inter-
actions, one short dominant epitope contributes 64–99%
of the estimated total interaction energy. Less trivially so,
however, PPIs of the second domain–domain class are
also dominated by a single short continuous epitope that
contributes 43–71% of the interaction energy. Remark-
ably, in all cases, we found that this epitope binds to the
same site that is targeted by a known small molecule
inhibitor (Figure 2). We conclude that it is the presence of
a dominant hot segment at a protein–protein interface
that often renders this PPI druggable.
Additional computational approaches for theprediction of PPI druggabilityAccumulating knowledge about druggable PPIs has
spurred the development of computational approaches
to predict the ‘druggability’ of a query PPI. The most
straightforward approach is to detect hot spot residues,
either by experiment [5] or by computation [23]. This
usually requires a solved structure of the PPI, or a very
accurate model. Such models may be built using protein
docking protocols of PPIs [13] or peptide-docking proto-
cols [24].
The structure of a protein complex is not always available,
but luckily it is also not always necessary in order to
determine the druggability of an interaction. Using
‘pocket biased’ conformational sampling of unbound
proteins on a subset of 2P2I, Johnson and Karanicolas
find that druggable protein interaction sites display low-
energy pocket-containing conformations that are not
found elsewhere on the protein surface [25��].
Current Opinion in Chemical Biology 2013, 17:952–959
954 Synthetic biomolecules
Table 1
Hot segments in the 2P2I database of druggable interactions. The interactions are divided into domain-peptide and domain–domain
interaction, and sorted according to % of binding energy contributed by the hot segment
Class Partner 1 Partner 2 Hot segmenta Energy (REU)b % of total
binding energy
PDB
Domain–peptide
Menin MLL 4–13 �28.7 99 4gq6
Xdm2 p53 17–26 �16.6 95 1ycq
ZipA FtsZ 3–12 �18.8 95 1f47
Mdm4 p53 19–28 �17.4 91 3dab
Mdm2 p53 19–28 �17.5 86 1ycr
Bcl-xL Bak 573–582 �14.6 79 1bxl
Bcl-2 Bax 62–71 �22.0 64 2xa0
Domain–domain
IN LEDGF 360–369 �15.4 71 2b4j
XIAP Smac 1–10 �15.8 57 1g73
TNFR1Ac TNFb 44–53 �8.0 54 1tnr
IL2 IL2Ra 34–43 �13.9 50 1z92
TNFa (B)d TNFa (A) 91–100 �8.6 48 1tnf
HPV E2 HPV E1 452–461 �9.5 45 1tue
XIAP-Bir3 Casp9 316–325 �16.7 45 1nw9
TNFa (A)d TNFa (B) 113–122 �7.8 43 1tnf
a A continuous stretch of 10 amino acids derived from Partner 2 that contributes a dominant part of the estimated binding energy. Partner 1 is the
receptor against which a small molecule was designed. For cases where Partner 1 contains a corresponding continuous stretch that contributes even
more to binding affinity, the PDB code is underlined. In these cases a small molecule (or peptide) might be designed to bind to Partner 2.b As estimated by the PeptiDerive protocol (see [15��]; REU: Rosetta Energy Units).c We should note that in this case, the small molecule inhibitor discovered against the interaction is a covalent modifier, and while it does show
overlap with the predicted dominant peptide, the reversible binding of the molecule is quite weak (40–100 mM).d TNFa forms a symmetric homotrimer; due to small structural differences peptides derived from different monomers have slightly different predicted
binding energies. In this case the most dominant peptide derived from chain A does not overlap with the PPI inhibitor, while the dominant peptide
derived from chain B does.
Computational solvent mapping of apo structures of 15
PPI target proteins [26��] showed that the druggable sites
combine subsites with properties such as concavity and a
mixture of hydrophobicity and polarity that confer a
tendency to bind organic compounds, unique from other
surface regions.
Residues that participate in protein binding in one
protein and in small-molecule (or peptide) binding in
its homologue could be expected to be part of drug-
gable sites. Two studies analyzed such ‘bi-functional’
[27] or ‘multibinding site’ residues [28]. Surprisingly,
while the former found these residues to be less con-
served, the latter found them to be more conserved than
other (mono-functional) interface residues. Nonethe-
less, both approaches seem to indicate that these sites
are prevalent (more than 8000 proteins were annotated
with ‘bi-functional’ residues in the human proteome
[29]).
Finally, a database of interface residue clusters that
provide Small-Molecule Inhibitor Starting Points
(SMISPs) – not necessarily continuous – was used in
an orthogonal machine learning based approach [30] to
train a classifier to identify druggable sites on protein
structures. Application of this approach also estimated
that 48% of the non-redundant PPIs in the PDB would be
amenable for small-molecule inhibition.
Current Opinion in Chemical Biology 2013, 17:952–959
Peptide based inhibitors of PPIsTraditionally, the pharmaceutical industry has focused on
the discovery of small molecules to inhibit PPIs. How-
ever, chemical analyses of successful small molecule PPI
inhibitors suggest that these do not comply with classic
drug-like properties: they are larger, more hydrophobic,
mediate less hydrogen bonds, and have lower ligand
efficiency values [16,31��,32]. Whereas small molecules
seem more appropriate for targeting isolated hot spot
binding sites, peptides may very well be the natural
choice for competing for binding with hot segments.
Peptides may also be readily subjected to large-scale
proteomic methods that are not appropriate for small
molecules, and thanks to their size and increased infor-
mation content, they may be expected to bind with higher
specificity.
Discovery and design approaches
Perhaps the most naıve, yet effective method to discover
peptidic PPI inhibitors is their derivation from the inter-
face of their native PPIs. The use of peptides derived
from proteins to modulate interactions has been exten-
sively used to, for example, investigate G protein sig-
naling [12��], apoptosis [33], cancer related targets such as
survivin [34], angiogenesis [35], host–pathogen inter-
actions [36–38] and countless other systems. It is worth
noting that the discovery of these derived peptides has
mostly been based on the primary sequences of the
www.sciencedirect.com
Hot spot segments define druggable protein interactions London, Raveh and Schueler-Furman 955
Figure 2
(a) (b) (c)
(d) (e) (f)
Current Opinion in Chemical Biology
Small molecule PPI inhibitors overlap in space with a continuous dominant peptide at the interaction interface. Examples of dominant peptides
detected at the interfaces of protein complex structures are compared to the structures of bound small molecule PPI inhibitors. The protein target for
the inhibitor (Partner 1 in Table 1) is given in surface representation. The interacting protein (Partner 2) is depicted as cartoon. The dominant peptide
detected at the interface by PeptiDerive [15��] is colored red. The small molecule PPI inhibitor is colored purple. The PDB IDs for the structures of the
PPIs are reported in Table 1. Here we include the PDB IDs of the small molecule complexes in parentheses. (A) The complex between Bcl-xL and a
peptide from Bak (1ysi) represents the class of peptide-protein interactions in which trivially the dominant peptide that is derived by the crystal
structure corresponds to the inhibitor binding site. (B–F) Non-obvious multiple epitope interfaces: here too the dominant peptide overlaps with the
small molecule inhibitor: (B) XIAP-Bir3 interaction with Casp9 (1tfq); (C) Human papilloma virus (HPV) proteins E2–E1 interaction (1r6n); (D) LEDGF-
Integrase interaction (3lpt); (E) IL2/IL2Ra interaction (1m48), and (F) TNFa trimer disrupter (2az5).
targets and only rarely on structural models (e.g. by using
peptide arrays to test all overlapping peptides in a protein
sequence for binding [39], or by identification of com-
mon/unique peptide sequences among a set of proteins to
inhibit a general/specific interaction [40,41]).
Display methods are a good source for the discovery of
usually completely novel binding sequences. Phage dis-
play, in particular, has been the source of numerous
peptide PPI inhibitors [42]. Lastly, computational design
of peptide binders is increasingly showing promise as a
source of new inhibitory peptide sequences [43�,44,45,
46�,47,48].
Stabilization and optimization
Short and isolated peptides may be conformationally
unstable, susceptible to proteolysis, and deficient in their
bioavailability [49]. Consequently, a swath of stabiliz-
ation, optimization and mimicry strategies has been
developed for rigidifying and improving the bioactivity
of peptide molecules.
Secondary structure stabilization
a-Helices play a major role in PPIs. An analysis of PPIs in
the PDB found that 62% of protein complexes feature an
a-helix at their interface [22�]. 20% of dominant peptides
at PPI interfaces also adopt a helical conformation [15��].For these reasons, stabilization of a-helices holds great
www.sciencedirect.com
promise for peptide based PPI inhibition. Two of the
main approaches for helix stabilization are helix ‘staples’
(Figure 3A; recently reviewed by Verdine and Hilinski
[50]), and Hydrogen Bond Surrogate (HBS) helices
(Figure 3B; comprehensively reviewed in Patgiri et al.[51]). b-Hairpins are also common interface motifs
(Figure 3C), and as such, b-hairpin mimetics can be
useful for inhibition of PPIs [52,53].
For coiled epitopes that do not adopt a well-defined
secondary structure, peptide cyclization methods can be
used to rigidify the peptide conformation and protect it
against proteolysis (Figures 3D and E). Cyclization can
be mediated by head-to-tail covalent linkage of the
peptide backbone or terminal residues [54,55], disulfide
bridging by pairs of cysteine residues, or other side-chain
tethering techniques (e.g. [56]). Like linear peptides,
libraries of cyclic peptides can be screened for activity
using standard proteomic screens such as reverse yeast
two-hybrid and phage display assays (e.g. [55,57–59]),
and combined with, for example, antibody competition
assays [59,60].
Lastly, peptide based molecules such as peptoids [61,62]
(Figure 3F) and foldamers [63] represent a more chem-
istry-centered direction. These take advantage of pepti-
dic properties such as specificity and compatibility with
protein binding, while at the same time they address
Current Opinion in Chemical Biology 2013, 17:952–959
956 Synthetic biomolecules
Figure 3
(a)
(d) (e) (f)
(b) (c)
Current Opinion in Chemical Biology
Peptide stabilization approaches. Examples of various peptide stabilization technologies. For clarity, only the peptide backbones and non-peptidic
moieties are displayed. PDB IDs are given in parentheses. (A) An i � i+4 hydrocarbon stapled helix designed to bind Estrogen Receptor b (2yjd) [72].
(B) Hydrogen Bond Surrogate (HBS) seeding a 10 amino-acid helix [73]. (C) A b-hairpin peptide solved in complex with CXCR4 (3odu) [74]. (D) A
backbone cyclic, disulfide constrained, protease inhibitor peptide (1sfi) [75] (E) A bi-cyclic human urokinase-type plasminogen activator, discovered by
screening a combinatorial library (3qn7) [76]. (F) A five-mer peptoid [77]. Note that the non-canonical amino-acid side-chains, (R)-N-1-cyclohexylethyl,
are connected to the backbone nitrogen rather than the Ca atom.
typical peptide-related challenges and improve, for
example, bio-stability and permeability.
Summary and conclusionsTargeting PPIs with small molecules has come a long
way. Computational discovery methods [64] together
with major advances in experimental techniques, such
as fragment based drug design [65], allow efficient dis-
covery of hits against this challenging class of targets, and
have led to an ever increasing number of PPI inhibitors
that are making their way to the clinic [66]. These
successes, in combination with the accumulation of struc-
tural data about these interactions, make it possible to
elucidate basic features of protein interactions that can be
used to guide and improve target selection and inhibitor
design.
In this review, we have focused on hot segments –continuous peptide stretches that dominate the PPI –as starting points for PPI inhibition. These hot segments
seem to be a general feature of druggable interfaces
(Table 1), as supported by our observation that a range
of small-molecule PPI inhibitors, discovered by a variety
of methods, overlap with these suggested hot segments
in the binding site (Figure 2). That said, we found that
such a predominant peptide is present in just over half of
the analyzed complex structures [15��] (Figure 1), but
Current Opinion in Chemical Biology 2013, 17:952–959
missing from the rest. This remaining portion of pre-
dominantly non-continuous interfaces might prove more
challenging for inhibition. The same may hold for ‘tight
& wide’ interactions (according to the classification of
Smith and Gestwicki [14��]). Targeting these more
challenging interfaces might require development and
improvement of alternative approaches, for example the
combination of several non-continuous epitopes (e.g.
[67]), or approaches not directly targeting the PPI inter-
face, such as allosteric modulators [68] or peptides that
shift the oligomeric state of the target [69,70]. Another
challenge is encountered by competitive peptide based
inhibition in the context of signal transduction, where
the peptide that competes with the original partner
might as well bind and induce the same functional
response as the full protein (e.g. [12��]).
The use of peptides to modulate PPIs goes way back in
time. In nature, living organisms rely on naturally occur-
ring peptides to inhibit or compete with and regulate
interactions. As an example, the Extracellular Death
Factor (EDF) peptide prevents the binding of the MazE
antitoxin to the toxin MazF, thereby activating this
enzyme and leading to cell death. Our modeling studies
indicate that this is achieved by EDF binding to the same
site as the toxin [71]. Similarly, BH3 domain peptides
[33] in anti-apoptotic proteins compete with internal
www.sciencedirect.com
Hot spot segments define druggable protein interactions London, Raveh and Schueler-Furman 957
interactions of a helix with the same binding groove in
apoptotic proteins such as Bcl-xL, thereby preventing
apoptosis [33]. Finally, peptide motifs in regulatory
regions of a range of kinases mediate interactions both
between and within proteins, by for example occluding
the active site, and serving as a scaffold to bind target
proteins [40,41].
Successful inhibition of protein–protein interactions not
only holds great promise, but also great challenges. PPIs
that are dominated by a hot segment might well represent
the lowest and ripest of these high-hanging fruits, and
peptide-based inhibitors might be one ladder on the way
to pick them.
AcknowledgementsThis work was funded by the Israel Science Foundation, founded by theIsrael Academy of Science and Humanities (grant number 319/11) and theUSA-Israel Binational Science Foundation (grant number 2009418) toOSF. NL was supported by an EMBO long-term fellowship (ALTF 1121-2011).
References and recommended readingPapers of particular interest, published within the period of review,have been highlighted as:
� of special interest�� of outstanding interest
1. Petsalaki E, Russell RB: Peptide-mediated interactions inbiological systems: new discoveries and applications. CurrOpin Biotechnol 2008, 19:344-350.
2. Tompa P, Fuxreiter M, Oldfield CJ, Simon I, Dunker AK,Uversky VN: Close encounters of the third kind: disordereddomains and the interactions of proteins. Bioessays 2009,31:328-335.
3. Jones S, Thornton JM: Principles of protein–proteininteractions. Proc Natl Acad Sci U S A 1996, 93:13-20.
4. Hopkins AL, Groom CR: The druggable genome. Nat Rev DrugDiscov 2002, 1:727-730.
5. Clackson T, Wells JA: A hot spot of binding energy in ahormone–receptor interface. Science 1995, 267:383-386.
6. Bogan AA, Thorn KS: Anatomy of hot spots in proteininterfaces. J Mol Biol 1998, 280:1-9.
7. DeLano WL: Unraveling hot spots in binding interfaces:progress and challenges. Curr Opin Struct Biol 2002, 12:14-20.
8. Reichmann D, Rahat O, Cohen M, Neuvirth H, Schreiber G: Themolecular architecture of protein–protein binding sites. CurrOpin Struct Biol 2007, 17:67-76.
9.�
London N, Movshovitz-Attias D, Schueler-Furman O: Thestructural basis of peptide-protein binding strategies.Structure 2010, 18:188-199.
This study compared structural features of protein–protein interactionswith peptide-mediated interactions. The main relevant finding to thisreview is that hotspots are prevalent also in peptide–protein interactionsand can be used as starting point for inhibitor design.
10. Fletcher S, Hamilton AD: Protein surface recognition andproteomimetics: mimics of protein surface structure andfunction. Curr Opin Chem Biol 2005, 9:632-638.
11. Arkin MR, Whitty A: The road less traveled: modulating signaltransduction enzymes by inhibiting their protein–proteininteractions. Curr Opin Chem Biol 2009, 13:284-290.
12.��
Shpakov AO: Signal protein-derived peptides as functionalprobes and regulators of intracellular signaling. J Amino Acids2011:656051.
www.sciencedirect.com
Very detailed and comprehensive review on the use of peptides tounderstand biological interactions using intervening peptides and smallmolecules derivatives.
13. Bienstock RJ: Computational drug design targeting protein–protein interactions. Curr Pharm Des 2012, 18:1240-1254.
14.��
Smith MC, Gestwicki JE: Features of protein–proteininteractions that translate into potent inhibitors: topology,surface area and affinity. Expert Rev Mol Med 2012, 14:e16.
Thorough and insightful analysis of known small molecule PPI inhibitorsand the features of the inhibited interfaces.
15.��
London N, Raveh B, Movshovitz-Attias D, Schueler-Furman O:Can self-inhibitory peptides be derived from the interfaces ofglobular protein–protein interactions? Proteins 2010,78:3140-3149.
This study describes the development and application of PeptiDerive:Starting from a solved protein complex structure, this protocol identifies‘‘hot segments’’, that is, the continuous peptide stretch within a protein thatcontributes most to binding (computed, estimated DDG). Analysis of thesepeptides indicates that for over 50% of the protein interactions in 2 bench-marks of protein complex structures, one such peptide contributes most ofthe binding energy. See also references [21�,22�] that have applied this or asimilar protocol to a larger set of protein complex structures. Current large-scale analyses of the structural and functional features of these peptides areunderway, and a dedicated Webserver is currently being set up.
16. Higueruelo AP, Schreyer A, Bickerton GR, Pitt WR, Groom CR,Blundell TL: Atomic interactions and profile of small moleculesdisrupting protein–protein interfaces: the TIMBAL database.Chem Biol Drug Des 2009, 74:457-467.
17. Basse MJ, Betzi S, Bourgeas R, Bouzidi S, Chetrit B, Hamon V,Morelli X, Roche P: 2P2Idb: a structural database dedicated toorthosteric modulation of protein–protein interactions. NucleicAcids Res 2013, 41:D824-D827.
18. Labbe CM, Laconde G, Kuenemann MA, Villoutreix BO,Sperandio O: iPPI-DB: a manually curated and interactivedatabase of small non-peptide inhibitors of protein–proteininteractions. Drug Discov Today 2013.
19.�
Bourgeas R, Basse MJ, Morelli X, Roche P: Atomic analysis ofprotein–protein interfaces with known inhibitors: the 2P2Idatabase. PLoS ONE 2010, 5:e9598.
Databases of structures of inhibited interactions provides a good startingpoint to identify critical features for successful inhibition (see references[16–18]). This study analyses one of these databases: 2P2I. In particular, itdivides the set into peptide-mediated and protein domain–domain inter-actions. See also reference [31��].
20. Berman HM, Westbrook J, Feng Z, Gilliland G, Bhat TN, Weissig H,Shindyalov IN, Bourne PE: The Protein Data Bank. Nucleic AcidsRes 2000, 28:235-242.
21.�
Teyra J, Sidhu SS, Kim PM: Elucidation of the bindingpreferences of peptide recognition modules: SH3 and PDZdomains. FEBS Lett 2012, 586:2631-2637.
Large scale analysis of hot segments. See reference [15��].
22.�
Jochim AL, Arora PS: Systematic analysis of helical proteininterfaces reveals targets for synthetic inhibitors. ACS ChemBiol 2010, 5:919-923.
Large scale analysis of hot segments. See reference [15��].
23. Potapov V, Cohen M, Schreiber G: Assessing computationalmethods for predicting protein stability upon mutation: goodon average but not in the details. Protein Eng Des Sel 2009,22:553-560.
24. London N, Raveh B, Schueler-Furman O: Peptide docking andstructure-based characterization of peptide binding: fromknowledge to know-how. Curr Opin Struct Biol 2013 http://dx.doi.org/10.1016/j.sbi.2013.07.006.
25.��
Johnson DK, Karanicolas J: Druggable protein interaction sitesare more predisposed to surface pocket formation than therest of the protein surface. PLoS Comput Biol 2013, 9:e1002951.
References [25��,26��] describe approaches to identify druggable sites onthe free protein structures. This reference presents a very nice computa-tional approach that is based on the reduced stability and increasedstructural dynamics of these regions. This dynamic fluctuation allows toidentify the open binding site conformation. See reference [26��] foradditional approach.
Current Opinion in Chemical Biology 2013, 17:952–959
958 Synthetic biomolecules
26.��
Kozakov D, Hall DR, Chuang GY, Cencic R, Brenke R, Grove LE,Beglov D, Pelletier J, Whitty A, Vajda S: Structural conservationof druggable hot spots in protein–protein interfaces. Proc NatlAcad Sci U S A 2011, 108:13528-13533.
References [25��,26��] describe approaches to identify druggable sites onthe free protein structures. This reference presents another complemen-tary approach based on computational solvent mapping: druggable sitesare identified where solvent molecules cluster. As reference [25��], thisstudy is also focussed on application to the free, apo structure.
27. Davis FP, Sali A: The overlap of small molecule and proteinbinding sites within families of protein structures. PLoSComput Biol 2010, 6:e1000668.
28. Thangudu RR, Bryant SH, Panchenko AR, Madej T: Modulatingprotein–protein interactions with small molecules: theimportance of binding hotspots. J Mol Biol 2012, 415:443-453.
29. Davis FP: Proteome-wide prediction of overlapping smallmolecule and protein binding sites using structure. Mol Biosyst2011, 7:545-557.
30. Koes DR, Camacho CJ: Small-molecule inhibitor starting pointslearned from protein–protein interaction inhibitor structure.Bioinformatics 2012, 28:784-791.
31.��
Morelli X, Bourgeas R, Roche P: Chemical and structurallessons from recent successes in protein–protein interactioninhibition (2P2I). Curr Opin Chem Biol 2011, 15:475-481.
Comprehensive review on the inhibition of PPIs. As in reference [19�], theinteractions are classified according to their size and affinity (small/largeand tight/loose), and successful examples are described along failuresthat highlight the current challenges in the field.
32. Wells JA, McClendon CL: Reaching for high-hanging fruit indrug discovery at protein–protein interfaces. Nature 2007,450:1001-1009.
33. Letai A, Bassik MC, Walensky LD, Sorcinelli MD, Weiler S,Korsmeyer SJ: Distinct BH3 domains either sensitize oractivate mitochondrial apoptosis, serving as prototype cancertherapeutics. Cancer Cell 2002, 2:183-192.
34. Plescia J, Salz W, Xia F, Pennati M, Zaffaroni N, Daidone MG,Meli M, Dohi T, Fortugno P, Nefedova Y et al.: Rational design ofshepherdin, a novel anticancer agent. Cancer Cell 2005, 7:457-468.
35. Rosca EV, Koskimaki JE, Rivera CG, Pandey NB, Tamiz AP,Popel AS: Anti-angiogenic peptides for cancer therapeutics.Curr Pharm Biotechnol 2011, 12:1101-1116.
36. Galdiero S, Falanga A, Tarallo R, Russo L, Galdiero E, Cantisani M,Morelli G, Galdiero M: Peptide inhibitors against herpes simplexvirus infections. J Pept Sci 2013, 19:148-158.
37. Maes M, Loyter A, Friedler A: Peptides that inhibit HIV-1integrase by blocking its protein–protein interactions. FEBS J2012, 279:2795-2809.
38. Castel G, Chteoui M, Heyd B, Tordo N: Phage display ofcombinatorial peptide libraries: application to antiviralresearch. Molecules 2011, 16:3499-3518.
39. Katz C, Levy-Beladev L, Rotem-Bamberger S, Rito T, Rudiger SG,Friedler A: Studying protein–protein interactions using peptidearrays. Chem Soc Rev 2011, 40:2131-2145.
40. Mochly-Rosen D, Khaner H, Lopez J, Smith BL: Intracellularreceptors for activated protein kinase C. Identificationof a binding site for the enzyme. J Biol Chem 1991,266:14866-14868.
41. Mochly-Rosen D, Das K, Grimes KV: Protein kinase C, an elusivetherapeutic target? Nat Rev Drug Discov 2012, 11:937-957.
42. Sidhu SS, Fairbrother WJ, Deshayes K: Exploring protein–protein interactions with phage display. Chembiochem 2003,4:14-25.
43.�
Sammond DW, Bosch DE, Butterfoss GL, Purbeck C, Machius M,Siderovski DP, Kuhlman B: Computational design of thesequence and structure of a protein-binding peptide. J AmChem Soc 2011, 133:4190-4192.
One of a few examples of the successful design of a peptide binders tothe target.
Current Opinion in Chemical Biology 2013, 17:952–959
44. Lee J, Sammond DW, Fiorini Z, Saludes JP, Resch MG, Hao B,Wang W, Yin H, Liu X: Computationally designed peptideinhibitors of the ubiquitin E3 Ligase SCF(Fbx4). Chembiochem2013.
45. Rubinstein M, Niv MY: Peptidic modulators of protein–proteininteractions: progress and challenges in computationaldesign. Biopolymers 2009, 91:505-513.
46.�
Vanhee P, van der Sloot AM, Verschueren E, Serrano L,Rousseau F, Schymkowitz J: Computational design of peptideligands. Trends Biotechnol 2011, 29:231-239.
This is a comprehensive review on the computational design of peptideligands, their derivatives, and experimental high throughput techniques.
47. Chandra D, Morrison CJ, Woo J, Cramer S, Karande P: Design ofpeptide affinity ligands for S-protein: a comparison ofcombinatorial and de novo design strategies. Mol Divers 2013,17:357-369.
48. Donsky E, Wolfson HJ: PepCrawler: a fast RRT-based algorithmfor high-resolution refinement and binding affinity estimationof peptide inhibitors. Bioinformatics 2011, 27:2836-2842.
49. Adessi C, Soto C: Converting a peptide into a drug: strategiesto improve stability and bioavailability. Curr Med Chem 2002,9:963-978.
50. Verdine GL, Hilinski GJ: Stapled peptides for intracellular drugtargets. Methods Enzymol 2012, 503:3-33.
51. Patgiri A, Jochim AL, Arora PS: A hydrogen bond surrogateapproach for stabilization of short peptide sequences inalpha-helical conformation. Acc Chem Res 2008, 41:1289-1300.
52. Obrecht D, Chevalier E, Moehle K, Robinson JA: b-Hairpinprotein epitope mimetic technology in drug discovery. DrugDisc Today: Technol 2012, 9:e63-e69.
53. Robinson JA: Beta-hairpin peptidomimetics: design,structures and biological activities. Acc Chem Res 2008,41:1278-1288.
54. Conlan BF, Gillon AD, Craik DJ, Anderson MA: Circular proteinsand mechanisms of cyclization. Biopolymers 2010, 94:573-583.
55. Tavassoli A, Benkovic SJ: Split-intein mediated circular ligationused in the synthesis of cyclic peptide libraries in E. coli. NatProtoc 2007, 2:1126-1133.
56. Angelini A, Morales-Sanfrutos J, Diderich P, Chen S,Heinis C: Bicyclization and tethering to albuminyields long-acting peptide antagonists. J Med Chem 2012,55:10187-10197.
57. Horswill AR, Benkovic SJ: Identifying small-moleculemodulators of protein–protein interactions. Curr Protoc ProteinSci 2006. Chapter 19:Unit 19 15.
58. Joo SH, Xiao Q, Ling Y, Gopishetty B, Pei D: High-throughputsequence determination of cyclic peptide library members bypartial Edman degradation/mass spectrometry. J Am ChemSoc 2006, 128:13000-13009.
59. Hall PR, Hjelle B, Njus H, Ye C, Bondu-Hawkins V, Brown DC,Kilpatrick KA, Larson RS: Phage display selection of cyclicpeptides that inhibit Andes virus infection. J Virol 2009,83:8965-8969.
60. Ramanujam P, Tan WS, Nathan S, Yusoff K: Novel peptides thatinhibit the propagation of Newcastle disease virus. Arch Virol2002, 147:981-993.
61. Dohm MT, Kapoor R, Barron AE: Peptoids: bio-inspiredpolymers as potential pharmaceuticals. Curr Pharm Des 2011,17:2732-2747.
62. Butterfoss GL, Yoo B, Jaworski JN, Chorny I, Dill KA,Zuckermann RN, Bonneau R, Kirshenbaum K, Voelz VA: De novostructure prediction and experimental characterization offolded peptoid oligomers. Proc Natl Acad Sci U S A 2012,109:14320-14325.
63. Goodman CM, Choi S, Shandler S, DeGrado WF: Foldamers asversatile frameworks for the design and evolution of function.Nat Chem Biol 2007, 3:252-262.
www.sciencedirect.com
Hot spot segments define druggable protein interactions London, Raveh and Schueler-Furman 959
64. Kruger DM, Jessen G, Gohlke H: How good are state-of-the-artdocking tools in predicting ligand binding modes in protein–protein interfaces? J Chem Inf Model 2012, 52:2807-2811.
65. Valkov E, Sharpe T, Marsh M, Greive S, Hyvonen M: Targetingprotein–protein interactions and fragment-based drugdiscovery. Top Curr Chem 2012, 317:145-179.
66. Mullard A: Protein–protein interaction inhibitors get into thegroove. Nat Rev Drug Discov 2012, 11:173-175.
67. Eichler J: Peptides as protein binding site mimetics. Curr OpinChem Biol 2008, 12:707-713.
68. Nussinov R, Tsai CJ: Allostery in disease and in drug discovery.Cell 2013, 153:293-305.
69. Hayouka Z, Rosenbluh J, Levin A, Loya S, Lebendiker M,Veprintsev D, Kotler M, Hizi A, Loyter A, Friedler A: Inhibiting HIV-1 integrase by shifting its oligomerization equilibrium. ProcNatl Acad Sci U S A 2007, 104:8316-8321.
70. Gabizon R, Brandt T, Sukenik S, Lahav N, Lebendiker M,Shalev DE, Veprintsev D, Friedler A: Specific recognition of p53tetramers by peptides derived from p53 interacting proteins.PLoS ONE 2012, 7:e38060.
71. Belitsky M, Avshalom H, Erental A, Yelin I, Kumar S, London N,Sperber M, Schueler-Furman O, Engelberg-Kulka H: TheEscherichia coli extracellular death factor EDF induces theendoribonucleolytic activities of the toxins MazF and ChpBK.Mol Cell 2011, 41:625-635.
www.sciencedirect.com
72. Phillips C, Roberts LR, Schade M, Bazin R, Bent A, Davies NL,Moore R, Pannifer AD, Pickford AR, Prior SH et al.: Design andstructure of stapled peptides binding to estrogen receptors. JAm Chem Soc 2011, 133:9696-9699.
73. Liu J, Wang D, Zheng Q, Lu M, Arora PS: Atomicstructure of a short alpha-helix stabilized by a mainchain hydrogen-bond surrogate. J Am Chem Soc 2008,130:4334-4337.
74. Wu B, Chien EY, Mol CD, Fenalti G, Liu W, Katritch V,Abagyan R, Brooun A, Wells P, Bi FC et al.: Structuresof the CXCR4 chemokine GPCR with small-moleculeand cyclic peptide antagonists. Science 2010, 330:1066-1071.
75. Luckett S, Garcia RS, Barker JJ, Konarev AV, Shewry PR,Clarke AR, Brady RL: High-resolution structure of a potent,cyclic proteinase inhibitor from sunflower seeds. J Mol Biol1999, 290:525-533.
76. Angelini A, Cendron L, Chen S, Touati J, Winter G, Zanotti G,Heinis C: Bicyclic peptide inhibitor reveals large contactinterface with a protease target. ACS Chem Biol 2012, 7:817-821.
77. Wu CW, Kirshenbaum K, Sanborn TJ, Patch JA, Huang K,Dill KA, Zuckermann RN, Barron AE: Structural andspectroscopic studies of peptoid oligomers withalpha-chiral aliphatic side chains. J Am Chem Soc 2003,125:13525-13530.
Current Opinion in Chemical Biology 2013, 17:952–959