Structure-Based Design of Targeted Covalent Inhibitors

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 15

Published on 05 April 2018. Downloaded by Shanghai Institute of Materia Medica Chinese Academy of Science on 11/15/2021 1:34:22 AM.

Chem Soc Rev


View Article Online
TUTORIAL REVIEW View Journal | View Issue

Structure-based design of targeted covalent


inhibitors†
Cite this: Chem. Soc. Rev., 2018,
47, 3816
Richard Lonsdale and Richard A. Ward*

Covalent inhibition is a rapidly growing discipline within drug discovery. Many historical covalent
inhibitors were discovered by serendipity, with such a mechanism of action often regarded as undesirable
due to potential toxicity issues. Recent progress has seen a major shift in this outlook, as covalent
inhibition shows promise for targets where previous efforts to identify non-covalent small molecule
inhibitors have failed. Targeted covalent inhibitors (TCIs) can offer drug discovery scientists the ability to
increase the potency and/or selectivity of small molecule inhibitors, by attachment of reactive functional
Received 7th February 2018 groups designed to covalently bind to specific sites in a target. In this tutorial review we introduce the
DOI: 10.1039/c7cs00220c broader concept of covalent inhibition, discuss the potential benefits and challenges of such an approach,
and provide an overview of the current status of the field. We also describe some strategies and
rsc.li/chem-soc-rev computational tools to enable successful covalent drug discovery.

Key learning points


(1) Targeted covalent inhibition (TCI) offers potential gains in potency, selectivity and duration of action.
(2) Covalent inhibition can be achieved by targeting both catalytic and non-catalytic residues in a binding pocket. Traditionally these have been cysteine and
serine residues, but a broader range of residues is now being considered.
(3) TCI can be achieved using several strategies, including attachment of a reactive moiety (‘warhead’) to a reversible lead compound, and growing out from
covalent fragment hits.
(4) Molecular modelling methods, such as covalent docking, molecular dynamics and free energy perturbation simulations, can be used to guide covalent drug
discovery.

1. Introduction catalytic machinery of a target enzyme, or a non-catalytic side


chain located close to the binding pocket. Since covalent bonds
The traditional strategy for structure-based drug design often are significantly stronger than the non-covalent interactions
focuses on the development of small molecule inhibitors described above, the development of covalent inhibitors offers
(or activators) that bind to a target reversibly and non-covalently.1 the potential for inhibitors with increased potency over non-
Improvements in selectivity and potency are typically achieved by covalent analogues. These covalent bonds are often (but not
optimising shape complementarity and non-covalent interactions always) irreversible, and therefore covalent inhibitors offer
(e.g. hydrogen bonds, salt bridges, van der Waals interactions) longer duration of action than reversible inhibitors, since the
between inhibitor and a target binding site. Such reversible target will remain inhibited until the protein is degraded and
inhibitors are in equilibrium between their (target) bound and then regenerated.
unbound forms. Traditionally, covalent inhibitors have been viewed by
Covalent inhibitors contain specific functional groups, the pharmaceutical industry as a liability, rather than an
designed to react with a corresponding site in the target, typically opportunity, due to potential off-target effects and toxicity
an amino acid side chain. The side chain may be part of the issues.2 Early covalent inhibitors, typically originating from
natural products, were discovered by serendipity and only
Chemistry, Oncology, IMED Biotech Unit, AstraZeneca, Cambridge, CB4 0WG, UK. later confirmed to be covalent. Two well-known and
E-mail: richard.a.ward@astrazeneca.com
widely-used covalent inhibitors are aspirin (1) and penicillin
† Electronic supplementary information (ESI) available: Fig. S1–S7 – two-
dimensional structural representations of covalent attachment between com-
G (2) (Fig. 1). 1 is an irreversible inhibitor of cyclooxygenase,
pounds 5, 10, 11, 12, 21, 22 and 23 and their target residues. See DOI: 10.1039/ and operates via acetylation of an active site serine residue. 2
c7cs00220c covalently binds to the active site serine of DD-transpeptidase,

3816 | Chem. Soc. Rev., 2018, 47, 3816--3830 This journal is © The Royal Society of Chemistry 2018
View Article Online

Tutorial Review Chem Soc Rev


Published on 05 April 2018. Downloaded by Shanghai Institute of Materia Medica Chinese Academy of Science on 11/15/2021 1:34:22 AM.

Fig. 1 Aspirin (1) and penicillin G (2), early covalent drugs discovered from
natural products.

an enzyme important to the formation of peptidoglycan walls


Fig. 2 Clopidogrel (3) and omeprazole (4) activation to form covalent
in bacteria.
inhibitors 3a and 4a, respectively.
Some disclosed drugs become covalent inhibitors after
bioactivation, and are inactive in their administered form.
Examples of such compounds include clopidogrel (3) and often referred to as the ‘warhead’, and a cysteine residue
omeprazole (4) (Fig. 2). 3 is used to prevent heart attack located in the binding pocket. For example, ibrutinib (5) is a
and stroke, and is activated in the liver by cytochrome P450 potent and selective inhibitor of Bruton’s tyrosine kinase (BTK),
enzymes, forming 3a. 4 is a proton-pump inhibitor and under an oncology target associated with mantle cell lymphoma.
the acidic conditions of the stomach reacts to form the active The reactive functional group, highlighted in Fig. 3 is an
sulfenamide, 4a. acrylamide, which is a Michael acceptor and forms an irreversible
Off-target effects associated with covalent inhibitors are covalent bond with Cys481 when bound in the ATP binding
typically due to covalent binding to other proteins, resulting pocket of BTK (Fig. 4 and Fig. S1, ESI†). Acalabrutib (6) is
in cell damage or immunological response. Toxicity effects can reportedly a more selective BTK inhibitor, which covalently
arise from reaction of covalent compounds with drug meta- binds to the same cysteine residue but with a methylpro-
bolizing enzymes, resulting in highly reactive intermediates. pargylamide warhead.4 Afatinib (7) and osimertinib (8) are
Fortunately, the traditional viewpoint towards covalent inhibition inhibitors designed to covalently bind to Cys797 in the epidermal
has begun to shift in recent years,2 with a significant number of growth factor receptor (EGFR), and are described in more detail
purposely designed covalent inhibitors entering clinical studies. in Section 4.1.
This has been largely achieved using functional groups with In this tutorial review, we first outline the benefits and
relatively low reactivity, combined with a highly selective reversible challenges associated with developing covalent inhibitors. We
binding motif. Such inhibitors are commonly referred to as then introduce the two main types of covalent inhibitors: catalytic
targeted covalent inhibitors (TCIs), and some key examples are and non-catalytic residue targeting inhibitors. For each class of
shown in Fig. 3. These inhibitors are US FDA-approved oncology inhibitor we provide some examples of compounds that have
drugs that bind to the ATP binding pocket in protein kinases.3 The proceeded to clinical studies. Lastly, we outline some covalent
covalent bond is formed between the reactive functional group, drug discovery strategies and provide an overview of how

Richard Lonsdale received his Dr Richard A. Ward, PhD is a


MChem from the University of Computational Medicinal Chemist
Oxford and his PhD from the working within the Oncology IMED
University of Bristol. He sub- Biotech Unit at AstraZeneca,
sequently completed postdoctoral Cambridge, UK. He has a parti-
studies at the University of Bristol cular interest in structure-based
and Max-Planck-Institut für drug design and the targeting of
Kohlenforschung, with a focus enzymes by covalent mechanisms.
on applying molecular dynamics Richard co-proposed a drug
and quantum mechanics/molecular discovery project with a colleague
mechanics methods to enzyme which resulted in the discovery of
catalysis. He is currently a the EGFR T790M-mutant selective
Richard Lonsdale postdoctoral fellow in the IMED Richard A. Ward inhibitor osimertinib (Tagrissot),
Biotech Unit at AstraZeneca, of which he is named as co-
within the Oncology therapeutic area. His research focusses on inventor. As part of this team he was recently awarded the
understanding and developing targeted covalent inhibitors using Malcolm Campbell Memorial Prize in 2017 by the RSC and is
computational chemistry methods. joint winner of the 8th Capps Green Zomaya Award in 2018 from
the Biological and Medical Chemistry sector of the RSC.

This journal is © The Royal Society of Chemistry 2018 Chem. Soc. Rev., 2018, 47, 3816--3830 | 3817
View Article Online

Chem Soc Rev Tutorial Review


Published on 05 April 2018. Downloaded by Shanghai Institute of Materia Medica Chinese Academy of Science on 11/15/2021 1:34:22 AM.

Fig. 5 Potential risks and benefits associated with covalent inhibitors.

benefits and challenges, hence fall in the overlapping region of


the diagram in Fig. 5, for the reasons discussed below.
One of the key potential benefits of a TCI is the opportunity
to increase the target potency of a compound, since the
compound will typically bind more strongly to the target than
a non-covalent analogue. This can enable access to levels of
potency beyond the range available to non-covalent inhibitors.
Fig. 3 Examples of US FDA-approved targeted covalent inhibitors (TCIs):
‘Intractable’ targets are those which are considered to be
ibutinib (5) and acalabrutinib (6) (BTK); afatinib (7) and osimertinib (8)
(EGFR). The covalent warheads are highlighted in red boxes. impossible to inhibit via traditional means. Such examples
include enzymes with shallow, or highly charged pockets such
as phosphatases, or binding sites which can potentially disrupt
protein–protein interactions. Covalent inhibitors may have
potential applications in this area, since they are able to bind
strongly to binding sites with relatively few reversible binding
interaction possibilities. Additionally, targets with high affinity
for their natural substrate(s) or cofactor, which may be present
at high concentrations in a cellular environment, are sometimes
considered ‘intractable’, since competition between inhibitor
and substrate is too high to be achieved at an acceptable dosing
level. Covalent inhibition can be of benefit here since irrever-
sible binding negates the need to reversibly compete with the
substrate.
Designing inhibitor selectivity can be challenging when
optimising inhibitors to bind to a single target, especially when
that target is a member of a family of targets with similar
Fig. 4 Covalent binding of ibrutinib (5) to BTK (PDB code: 5P9J). binding pockets. For example, protein kinases are common
oncology targets, of which there are over 500 in the human
genome. Since all use the same cofactor (ATP), the pocket is
structure-based computational chemistry methods can benefit
structurally similar across the kinome, therefore selectively
the development of TCIs.
targeting a single kinase can be challenging. Designing compounds
to covalently bind to poorly conserved residues (such as cysteines)
offers potential gains in selectivity for a target kinase over kinases
2. Benefits and challenges of covalent not containing a cysteine in that position.
inhibitors Covalent inhibitors may offer prolonged duration of action,
particularly for target proteins with slow turnover. This is often
When considering covalent inhibition as an approach there can desirable in drug discovery, but not in all cases. The potential to
be a fine balance between the potential risks and benefits, as inhibit a target for an extended duration, i.e. until the protein is
reviewed by Barf and Kaptein5 and summarized in Fig. 5. Two resynthesized could result in convenient dosing schedules and
of the factors considered here have the potential to be both low dosing required for high exposure. However, if the protein

3818 | Chem. Soc. Rev., 2018, 47, 3816--3830 This journal is © The Royal Society of Chemistry 2018
View Article Online

Tutorial Review Chem Soc Rev

in question is functionally important, and has a slow rate of Covalent compounds tend to undergo extrahepatic clearance
Published on 05 April 2018. Downloaded by Shanghai Institute of Materia Medica Chinese Academy of Science on 11/15/2021 1:34:22 AM.

resynthesis, safety concerns may arise. Hence this is considered more readily than non-covalent compounds, which can result
as both a potential benefit and risk. Likewise, the possibility in lower bioavailability.6 For example, electrophilic covalent
of complete target inactivation can be a potential benefit, but warheads are susceptible to reaction with glutathione, a scavenger
only if that is a desired outcome of the compound profile. for reactive electrophiles that is present in high concentration in
Conversely, if the target protein has a rapid resynthesis rate, cells. Warhead reactivity is therefore an important consideration,
covalent inhibition may result in no significant benefit. representing a balance between maximising target engagement,
One of the major risks associated with covalent inhibitors is whilst minimising extrahepatic clearance.
the possibility of non-specific binding.2 If the covalent warhead
is reactive enough to bind to residues in other proteins,
unpredictable toxicity events can potentially occur. Further- 3. Covalent modification of catalytic
more, covalent warheads are commonly susceptible to meta- residues
bolism, resulting in reactive intermediates that can cause
damage to proteins and DNA. To overcome this challenge, Many disease targets in drug discovery are enzymes. Enzyme
covalent warheads with low intrinsic reactivity are preferred. targets offer the potential for a covalent inhibitor to directly
A further challenge commonly presented by covalent interfere with the reaction mechanism. Covalent inhibitors are
inhibitors is that measuring potency is more complex than often designed to mimic the natural substrate, but have been
for corresponding reversible inhibitors. In the case of reversible modified in such a way that the reaction cannot proceed to
inhibitors, potencies are typically reported in terms of Ki or IC50 completion, and the inhibitor remains covalently bound to the
values. Ki is the inhibitory binding constant and IC50 is enzyme, the latter permanently rendered unable to perform its
the equilibrium concentration of inhibitor at which 50% of the function. A detailed understanding of the enzymatic mechanism
target process has been inhibited. IC50 values for irreversible with the natural substrate is necessary in order to successfully
inhibitors are time-dependent, because no such equilibrium develop such inhibitors. Two example targets that have been
exists; the inhibitor remains in the target binding site until the the subject of covalent inhibition strategies targeting catalytic
protein is degraded, therefore inhibition of the target protein residues are provided here.
will continue to increase until all of the inhibitor is engaged in
binding, or all of the target has been inhibited. Hence, it is 3.1 DPP-IV
challenging to compare the IC50 values of irreversible covalent Proline-specific dipeptidyl aminopeptidase IV (DPP-IV) is a
inhibitors due to the time-dependent nature of the inhibition. serine protease responsible for the degradation of peptide
Therefore it is more useful to consider the potency of covalent hormones, contributing to the regulation of blood glucose
inhibitors in terms of the parameters Ki and kinact (Fig. 6). Here, levels. Inhibition of DPP-IV has been identified as a method
Ki corresponds to the reversible binding of the ligand in the for the treatment of type 2 diabetes.7 The enzyme selectively
binding pocket of the protein and kinact is the kinetic rate cleaves the peptide bond close to the N-terminus where a
constant for the formation of the covalent bond between protein proline (or alanine) is located at the penultimate position, as
and inhibitor. Measuring Ki and kinact is usually more challen- shown in Fig. 7. The binding site of DPP-IV consists of two
ging and labour intensive than measuring IC50 values. pockets (S1 and S2, Fig. 8). The hydrophobic S1 pocket, is highly
The relationship between pharmacokinetics (PK) and specific for proline, and also contains the catalytic triad com-
pharmacodynamics (PD) for irreversible covalent inhibitors prising of Ser630, Asp708 and His740. The S2 pocket can bind
is often more complex to that of reversible inhibitors.5 The any amino acid and contains the Glu205 and Glu206 pair, which
duration of action of a reversible inhibitor will depend on bind to the terminal amino group of the peptide substrate.
how long it remains bound to the target (i.e. residence time) Both covalent and non-covalent inhibitors have been developed
and how quickly it is cleared from the body (i.e. PK). For for DPP-IV. Non-covalent DPP-IV inhibitors, such as sitagliptin (9)
an irreversible inhibitor, the compound will remain bound (Fig. 9), exclusively form non-covalent interactions with the bind-
to the target beyond the point where all of the free drug ing pocket of the enzyme. 9 was the first DPP-IV inhibitor to obtain
has been excreted, and as mentioned above, the PK/PD rela- FDA approval in 2006. Saxagliptin (10) (Fig. 9), is a covalent DPP-IV
tionship is also therefore dependent on the rate of target inhibitor and contains a core resembling proline, but in addition
regeneration. forms a reversible covalent interaction with the enzyme.8 When 10
binds to DPP-IV, Ser630 reacts with the nitrile group, resulting
in the formation of an O–C bond between the enzyme and

Fig. 6 Cartoon showing the binding of a covalent inhibitor to a target


binding pocket. Fig. 7 Mechanism of peptidase activity of DPP-IV.

This journal is © The Royal Society of Chemistry 2018 Chem. Soc. Rev., 2018, 47, 3816--3830 | 3819
View Article Online

Chem Soc Rev Tutorial Review


Published on 05 April 2018. Downloaded by Shanghai Institute of Materia Medica Chinese Academy of Science on 11/15/2021 1:34:22 AM.

Fig. 8 X-ray structure of DPP-IV in complex with the tert-butyl-Gly-L-


Pro-L-Ile tripeptide (shown in green, PDB code 2AJB). The S1 and S2
binding pockets are highlighted in red and blue, respectively. Proline is
bound in the S1 pocket and Ser630 has formed a covalent bond to the
peptide. Fig. 10 Binding poses of saxagliptin (10) (green) and 9 (orange) in DPP-IV
(PDB codes: 1X70 and 3BJM, respectively). 10 forms a covalent bond to
Ser630.

Fig. 11 E-64 (11), a potent irreversible cysteine protease inhibitor, often


used as a probe compound.

Fig. 9 Inhibitors of DPP-IV: sitagliptin (9) (reversible non-covalent) and


saxagliptin (10) (reversible covalent).
undergoes deprotonation by histidine to form a thiolate, which
can perform nucleophilic attack at the carbonyl group of the
peptide bond, forming a tetrahedral intermediate. This inter-
inhibitor (Fig. S2, ESI†). The negative charge formed on the
mediate can then undergo hydrolysis, resulting in amine and
nitrile nitrogen atom is stabilised by Tyr547, which also stabilises
carboxylic acid products. E-64 (11) (Fig. 11) is a natural product
the oxyanion hole formed during the native enzymatic reaction.
found to covalently inhibit cysteine proteases, by undergoing
The differences in binding of the covalent and non-covalent
nucleophilic attack by the cysteine, but forming a complex that
inhibitors are highlighted in the overlays of the co-crystal struc-
is unable to undergo the subsequent hydrolysis step (Fig. S3, ESI†).
tures of 9 and 10 in DPP-IV (Fig. 10). Typically, non-covalent
11 is a non-selective inhibitor but can be used as a template
inhibitors are required to extend further into the S2 pocket in
molecule for development of inhibitors that are selective for
order to increase target potency. The reversible covalent bond
specific cysteine proteases. One of the first X-ray crystal struc-
formed in the covalent inhibitor:enzyme complex leads to slow
tures of cathepsin K to be solved (Fig. 12) reveals the covalent
dissociation of the drug from the binding pocket, resulting in
complex formed between 11 and the active site cysteine.10 The
continuing inhibition well beyond the point at which the free
epoxide of 11 undergoes nucleophilic ring opening by the depro-
drug has been cleared. For this reason, covalent DPP-IV inhibitors
tonated Cys25 sulfur atom.
often have longer duration of action than non-covalent examples.
Leading on from such observations more targeted TCI’s
were identified such as odanacatib (12) (Fig. 13) which is a
3.2 Cathepsin K selective and potent reversible covalent inhibitor of Cathepsin
Cysteine proteases are involved in many biological processes K.11 It contains a nitrile warhead which reacts with Cys25 in
and are common disease targets. Cathepsin K is a cysteine a similar manner to that observed between 10 and Ser630 in
protease involved in bone remodelling and resorption. Inhibi- DPP-IV (Fig. S4, ESI†).
tion of cathepsin K has been identified as a possible therapeutic
treatment for osteoporosis.9 Cathepsin K was initially considered
to be intractable, as it has a small and solvent exposed binding 4. Covalent modification of
pocket, which is difficult to target with a reversible non-covalent non-catalytic residues
inhibitor.
The active site of cysteine proteases contains a cysteine Designing covalent inhibitors that exploit the catalytic machinery
and histidine located in close proximity. The cysteine readily of enzymes can be problematic in terms of achieving selectivity,

3820 | Chem. Soc. Rev., 2018, 47, 3816--3830 This journal is © The Royal Society of Chemistry 2018
View Article Online

Tutorial Review Chem Soc Rev

RAS can exist in active (GTP-bound) and inactive (GDP-bound)


Published on 05 April 2018. Downloaded by Shanghai Institute of Materia Medica Chinese Academy of Science on 11/15/2021 1:34:22 AM.

forms. Inhibition of KRAS is a challenging problem, and has


been broadly considered intractable to small molecule inhibi-
tion, due to the incredibly high GTP/GDP affinity of the binding
pocket, high cellular concentration of GTP and GDP, and lack of
other small molecule binding sites. The G12C mutation is
located close to what was subsequently identified as an allosteric
pocket. The flexibility of the switch II region has been shown to
be higher in the GDP-bound state for KRAS-G12C, therefore
it is believed that targeting this allosteric pocket may serve to
stabilise the GDP-bound (inactive) state of RAS. Whilst it is
difficult to achieve strong reversible binding to this pocket,
covalent attachment to the cysteine has been proposed as a
strategy to increase potency. Covalent inhibitors targeting this
cysteine have been discovered that can selectively target the G12C
mutant. An example covalent KRAS G12C inhibitor is ARS-853
Fig. 12 First X-ray crystal structure (PDB: 1ATK) of Cathepsin K reveals
binding of covalent inhibitor E-64 (11) to the active site cysteine (Cys25). (13), displayed in Fig. 14.13 13 contains an acrylamide covalent
warhead, where the nitrogen is part of an azetidine ring.
Acrylamides act as Michael acceptors, where the cysteine per-
forms a nucleophilic addition to the terminal CQC carbon
atom. The binding mode of 13 in KRAS G12C is depicted in
Fig. 15.
EGFR. The epidermal growth factor receptor (EGFR) is a
transmembrane tyrosine kinase involved in activating cell pro-
liferation pathways. Mutations that maintain EGFR in the
active state are a known driver of certain cancers, in particular,
non-small cell lung cancer (NSCLC).14 Inhibition of the tyrosine
Fig. 13 Odanacatib (12), a reversible covalent inhibitor of Cathepsin K. kinase activity of EGFR is a validated treatment for NSCLC that

particularly for enzymes that are part of large families of related


enzymes, e.g. proteases or kinases. Furthermore, many binding
pockets are not directly involved in enzymatic reactions, yet
covalent binding may be desirable. By targeting non-catalytic
residues that are poorly conserved across enzyme classes, success-
ful efforts have been made towards the development of covalent
inhibitors with high selectivity and potency. Some examples are
described below.
Fig. 14 ARS-853 (13) – acrylamide containing inhibitor of KRAS G12C.

4.1 Cysteine
By far the most effort has been directed towards targeting non-
catalytic cysteine residues, for example in cysteine proteases
and protein kinases, where selectivity for a particular member
of a protein family can be challenging. The appeal of cysteine is
due in part to its relatively low abundance in proteins, and its
high nucleophilicity. The thiolate form of cysteine has been
shown to be capable of forming covalent bonds with covalent
warheads spanning a wide range of reactivity, with Michael
acceptors being key examples.12 For the reasons described
above, warheads with low reactivity are generally preferred.
KRAS. KRAS is a GTPase involved in the early stage of many
signal transduction pathways, and has been linked to the
development of many types of cancer. The G12C mutant of
KRAS is associated with colorectal and lung cancers.13 Prevent- Fig. 15 X-ray structure revealing the binding mode of ARS-853 (13)
ing the interaction between KRAS and its downstream signalling (shown in green) in KRAS G12C (PDB: 5F2E). The inhibitor is covalently
partners has been suggested as a possible targeting mechanism. bound to Cys12.

This journal is © The Royal Society of Chemistry 2018 Chem. Soc. Rev., 2018, 47, 3816--3830 | 3821
View Article Online

Chem Soc Rev Tutorial Review


Published on 05 April 2018. Downloaded by Shanghai Institute of Materia Medica Chinese Academy of Science on 11/15/2021 1:34:22 AM.

Fig. 16 First generation EGFR inhibitors gefitinib (14) and erlotinib (15).

is due to EGFR activating mutations. The first generation of


small molecule inhibitors for EGFR were reversible, binding to
the ATP binding site in a conventional manner (e.g. gefitinib
(14) and erlotinib (15), Fig. 16). However, a resistance mechanism
to these inhibitors often occurs via the T790M gatekeeper
mutation, which increases the affinity of EGFR towards ATP.15
An irreversible covalent inhibitor serves to overcome ATP com- Fig. 19 Modeled binding mode of osimertinib (8) in T790M mutant EGFR.
Reprinted with permission from ref. 18. Copyrightr 2014, American
petition, since it cannot be displaced by ATP once bound. The first
Association for Cancer Research.
covalent inhibitor of EGFR and erbB2 was 20 -thioadenosine (16)
(Fig. 17), and forms a disulphide bond with Cys797,16 which is
present on the edge of the ATP-binding site in a relatively small
WT EGFR. It was therefore hypothesized that such side-effects
group of kinases (see Section 5.2).5 Whilst this inhibitor demon-
might be overcome with a T790M mutant selective covalent
strated the possibility of covalently binding to Cys797, disulphide
inhibitor of EGFR. An example of such an inhibitor is osimertinib
bonds are not irreversible. Another early covalent EGFR covalent
(8) (Fig. 3).18 The modelled binding mode of 8 in the T790M
inhibitor, PD-168393 (17), irreversibly binds to Cys797 via an
mutant is shown in Fig. 19.
acrylamide covalent warhead.17 However, the intrinsic chemical
Whilst covalent targeting of C797 in EGFR has been a
reactivity of the acrylamide warhead in 17 was deemed too reactive,
significant advancement in treating NSCLC, this approach is
with liability towards off-target toxicity.
also potentially vulnerable to resistance mechanisms. One
The next generation of EGFR inhibitors were subsequently
such resistance mechanism is the C797S mutation of EGFR.19
designed around these observations to increase potency by
Michael acceptor warheads, such as acrylamides, do not react
covalently binding to C797 (e.g. 7 and neratinib (18) (Fig. 18).
with serine very readily under physiological conditions, and
Side effects (diarrhoea and rash) are common with such
therefore covalent inhibition does not occur with the C797S
inhibitors, due to inhibition of wild-type (WT) EGFR. Indeed,
mutant. The ability to covalently target residues other than
the T790M mutant displays less affinity towards 7 and 18 than
cysteine may be a useful strategy to overcome such resistance
mechanisms.

4.2 Lysine
Since cysteine residues are not always available in a protein
binding site, attempts have also been made to target lysine
residues.20 Lysine is more abundant than cysteine and is a
harder nucleophile, therefore targeting lysine offers alternative
routes to covalent inhibition. The typical pKa of a lysine residue
is 10.4, and therefore most lysine residues will be protonated
Fig. 17 Early covalent inhibitors of EGFR, 2 0 -thioadenosine (16) and PD- under physiological conditions. Surface lysine residues are
168393 (17). more likely to be in the protonated form, but buried lysines
are typically less solvent accessible and may exhibit perturbations
in pKa favouring the neutral form. Only the neutral form is
nucleophilic, and therefore selectivity can potentially be achieved
despite the high abundance of lysine. However, locating lysine
residues with their pKas sufficiently lowered to exist in the neutral
form within a binding site is challenging. One example of such an
approach is through a surface lysine in the myeloid leukemia cell
differentiation protein (Mcl-1) with a reversible covalent aryl
boronic acid carbonyl warhead (19) (Fig. 20).21
Model reactions using N-a-acetyl-lysine and glutathione have
suggested that lysine is less reactive than cysteine towards soft
Fig. 18 Neratinib (18). electrophiles such as acrylamides, but more reactive towards

3822 | Chem. Soc. Rev., 2018, 47, 3816--3830 This journal is © The Royal Society of Chemistry 2018
View Article Online

Tutorial Review Chem Soc Rev


Published on 05 April 2018. Downloaded by Shanghai Institute of Materia Medica Chinese Academy of Science on 11/15/2021 1:34:22 AM.

Fig. 20 Mcl-1 inhibitor 19, containing a reversible covalent aryl boronic


acid carbonyl warhead.

Fig. 22 Covalent complex formed between 5’fluorosulfonylbenzoyladenosine


vinyl sulfones.22 A covalent inhibitor containing a vinyl sulfone (20) and FGFR1 (PDB: 5O49).
warhead has been developed for CDK2, which forms a covalent
bond to a non-catalytic lysine.23
Sulfonyl fluorides have also been identified as possible
warheads for targeting lysine.20 50 -fluorosulfonylbenzoyl adenosine
(20) is a well-established covalent chemical probe for ATP-binding
proteins (Fig. 21). Mukherjee et al. recently published structure
reactivity relationships concerning sulfonyl fluorides towards
cysteine, tyrosine, lysine and serine.24 As part of this work, a
crystal structure was obtained of 20 covalently bound to the
fibroblast growth factor receptor (FGFR). A covalent attachment
is formed to Lys514, as shown in Fig. 22 and Fig. S5 (ESI†).
Dalton et al. recently reported a selective covalent irreversible
inhibitor of PI3Kd (21, Fig. 23).25 The compound forms a Fig. 23 Covalent inhibitor (21) developed to target Lys779 of PI3Kd.
covalent interaction with the conserved Lys779 (Fig. S6, ESI†) The reactive warhead is an activated ester group.
and contains an activated ester as the reactive warhead. Selec-
tivity for the PI3Kd kinase was achieved, despite the conserved
lysine, due to optimization of reversible binding interactions. be incorporating the described lysine-targeting warheads into
The authors started from a reversible template containing a drug-like molecules with desirable properties. Furthermore,
sulphonamide functional group that formed a reversible inter- the majority of the studies on TCIs for lysine have been
action with Lys779, which was then exchanged for the covalent conducted in vitro. Challenges may arise when transferring into
warhead. in vivo systems. However, if these obstacles can be overcome,
Lysine targeting is a less mature area, and at present appears lysine-targeting may have a promising future in covalent drug
more challenging than targeting cysteines. One challenge will development.

4.3 Other residues


There is increasing interest to look at a broader range of non-
catalytic residues26 e.g. serine, threonine, tyrosine, methionine,27
glutamate and aspartate. For example, an additional oncogenic
KRAS mutation to G12C described in 4.1 is G12D. Shokat et al.
functionalized an optimized ligand for the switch II pocket of
KRAS with a set of electrophiles previously reported to react
with carboxylates, in an attempt to covalently target the G12D
mutant.28 However, low reactivity was observed towards G12D.
Martin-Gago et al. have designed a covalent inhibitor (22)
incorporating Woodward’s reagent K that was demonstrated
to bind to a glutamate residue in the lipoprotein chaperone
Fig. 21 5 0 -Fluorosulfonylbenzoyl adenosine (20), a covalent probe mole- PDE6d (Fig. 24).29 The structure of the proposed covalent
cule that covalently binds to lysine in many protein kinases. attachment is shown in Fig. S7 (ESI†).

This journal is © The Royal Society of Chemistry 2018 Chem. Soc. Rev., 2018, 47, 3816--3830 | 3823
View Article Online

Chem Soc Rev Tutorial Review

then be initiated on these along with wider literature reviews. In


Published on 05 April 2018. Downloaded by Shanghai Institute of Materia Medica Chinese Academy of Science on 11/15/2021 1:34:22 AM.

some cases the identification of a potent and selective covalent


inhibitor may facilitate such validation experiments. Cravatt et al.
used reactive fragments to identify proteins across the human
proteome that could potentially be covalently modified at
cysteine.30 Bourne et al. combined a function-site interaction
fingerprint method with density functional theory (DFT) calcula-
tions to determine the potential of cysteines across the human
kinome to form a covalent interaction with an inhibitor.31 Zhang
et al. used a computational algorithm to predict covalent binding
sites along with hydrogen bond frequency analysis and a
distance-based covalent fragment probing method for detecting
covalent binding sites.32 A manually-curated comprehensive
database has been developed for proteins containing a targetable
cysteine, along with known covalent inhibitors.33

5.2 Selecting the right residue


There may be cases where multiple residues are available in
a target for covalent modification, particularly in the case of
non-catalytic residues. In other cases, the residue to target may
Fig. 24 (a) Compound 22 based on Woodward’s reagent K developed as be more clear. An in-depth analysis of the possible residues and
a covalent inhibitor of PDE6d. (b) Crystal structure reveals covalent bond to their environment may help to increase the rate of success, or
a glutamate residue (PDB: 5NAL). to provide early indications of any likely problems.
The location of the residue is clearly an important con-
sideration; it should be close to, or part of, a pocket to which
5. Strategies for structure-based
an inhibitor can bind. Computational algorithms are available
development of TCIs that can indicate likely binding pockets in protein structures.34
These typically identify a set of potential pockets, along with
There are particular considerations that are often unique to the
their relative ranking and the residues surrounding them.
development of covalent inhibitors. Some of these are discussed
These can be filtered towards those containing the residue of
below. Here we assume that the project in question is either
interest, e.g. cysteine. In the case of protein kinases, the ATP
structure-enabled, or where a compound binding mode can be
binding pocket is typically targeted, and therefore a new kinase
modelled with a high degree of confidence.
project would be assessed for covalent targetability based on
the location of cysteines in proximity of the ATP pocket. Gray
5.1 Target selection et al.3 (and subsequently modified by Barf and Kaptein5)
The decision to opt for a covalent inhibitor can be made at any provided a useful overview of the positioning of cysteines
stage of an early drug discovery project, however, it is often around binding pockets of kinases (Fig. 25). For example,
decided at the target selection phase. Traditionally, covalent C797 in EGFR is located at the ‘3F’ position, along with 9 other
inhibition projects have been biased towards oncology targets, protein kinases. One might expect that achieving selectivity
as a strategy to increase potency and selectivity and to avoid over another kinase with a cysteine at the same position may be
resistance mechanisms. However, as knowledge and confidence challenging. Despite this, there are known covalent inhibitors
grows in this area, more widespread use appears likely. displaying high selectivity against a single member of their
Preliminary biological data is required in order to assist the group. For example, selective covalent inhibitors for EGFR have
decision. The protein half-life (regeneration–balance) may be been successfully developed, as well as for other targets in
considered to inform whether a covalent option may be beneficial. group 3F (e.g. BTK, see Fig. 3). Knowledge of targets with
The desired PD and PK profile of the final candidate can suggest similar binding pockets and cysteine locations is a valuable
whether this is feasible. resource for designing selective compounds, as overlaying
The strength of disease linkage to a target should ideally be structures can often highlight where modifications can be
established before any drug discovery work is initiated. How- made to gain selectivity.
ever, an alternative strategy is to prospectively look for novel Covalently targeted residues may require activation before a
targets that could be targeted covalently to identify a subset for reaction can take place. Indeed, this is often preferable as it
target validation studies. Recent applications have been reduces the likelihood of off-target reactivity. For example,
described where this has been done both experimentally and cysteine acts as a nucleophile when in the deprotonated
computationally. Panels of proteins can be screened against (thiolate) state. The typical pKa of a cysteine thiol side chain
reactive covalent fragments (e.g. iodoacetamide) to identify possible is 8.5, suggesting an equilibrium between the neutral and
targets for a covalent approach. Target validation studies could deprotonated forms under standard conditions (pH 7.4), with

3824 | Chem. Soc. Rev., 2018, 47, 3816--3830 This journal is © The Royal Society of Chemistry 2018
View Article Online

Tutorial Review Chem Soc Rev


Published on 05 April 2018. Downloaded by Shanghai Institute of Materia Medica Chinese Academy of Science on 11/15/2021 1:34:22 AM.

Fig. 25 Schematic representation of the terminology used for the location of cysteine residues surrounding the ATP binding pocket across the human
kinome. Based on crystal structure of interleukin-2 tyrosine kinase domain with staurosporine bound (PDB 1SM2). Reprinted with permission from ref. 5.
Copyright 2012 American Chemical Society.

a preference for the former protonation state. However, whether the cysteine is capable of covalent modification.
cysteines located in more polar environments, where nearby As described in Section 6.4, molecular dynamics simulations
positive charge can stabilize the thiolate state of the side chain can be useful in such cases. Conformational sampling can
can display much lower pKa values. For example, the active site indicate the most favourable side chain conformations and their
cysteine in cysteine proteases (which forms an ion pair with relative occupations.
histidine) falls in the region of 2.5–3.5.35 Whilst the thiolate
form is expected to be accessible to a cysteine with a pKa of 5.3 Finding and developing a covalent lead
8.5, a cysteine with a lower pKa value would spend significantly The most commonly used strategies for developing a covalent
more time in the deprotonated form. It is therefore assumed inhibitor are summarized in Fig. 26. The moment at which the
that targeting cysteines with lower pKas will facilitate covalent covalent warhead is added varies from early on (i.e. present in
binding. Experimental determination of cysteine pKa is non- the initial hit), to right at the end of compound optimisation.
trivial, and has been achieved for a relatively small number of The decision of which strategy (or strategies) to select will
proteins. There are a number of computational approaches depend on the amount of prior knowledge and chemical equity
for calculating the pKa of titratable amino acid residues available when embarking on a new project.
(e.g. empirical, Poisson–Boltzmann, thermodynamic integra- Phenotypic or ‘black-box’ screening of covalent inhibitors is
tion, constant-pH molecular dynamics), some of these were the method by which many historic covalent inhibitors were
compared in a study by Rowley et al.36 on protein systems where discovered. In such an approach, compounds are screened
experimental cysteine pKa measurements have been made. against a suitable cellular assay to identify those that cause a
Whilst computational approaches generally do not quantita- desirable change in phenotype. The biological target of such
tively predict pKa values of cysteine, they may be used for molecules may be established after the compounds have been
comparing relative values both between targets and between discovered, although this is not always possible or required.37
cysteines of the same target. The benefits of this method are that it is (a) a direct way to
Assuming that a cysteine is close to the binding pocket and identify an efficacious compound and (b) minimal information
appears to be in a sufficiently polar environment to enable of the target is required. The main drawback of this method is
deprotonation, the side-chain orientation should ideally be that there are relatively few potentially covalent compounds
considered. As it is commonly accepted that reversible binding available in screening collections, as historically more attention
occurs before the covalent bond formation, the target cysteine has been directed towards non-covalent drug development.
side chain should be oriented towards the binding pocket, in Hence, the amount of diversity is typically limited within covalent
order for a covalent bond to form to the inhibitor. X-ray libraries. Additionally, it may be difficult to achieve selectivity for a
structures can be valuable for indicating the average orienta- target without target or structural information available.
tion of amino acid side chains, but do not reveal dynamic Covalent fragment screening may be a useful approach
motion, which is important when considering proteins. Also for where there are no known reversible inhibitors of a target. A
poorly resolved regions, such as those located on flexible loops, selection of low molecular weight covalent compounds is typically
crystal structures may not provide a definitive answer as to screened against a target, with covalent binding confirmed by

This journal is © The Royal Society of Chemistry 2018 Chem. Soc. Rev., 2018, 47, 3816--3830 | 3825
View Article Online

Chem Soc Rev Tutorial Review


Published on 05 April 2018. Downloaded by Shanghai Institute of Materia Medica Chinese Academy of Science on 11/15/2021 1:34:22 AM.

Fig. 26 Covalent inhibitor design strategies.

X-ray crystallography and mass spectrometry and digestion


studies to confirm the residue targeted. Once a hit covalent
fragment has been found, and the binding mode is confirmed,
medicinal chemistry optimisation is deployed to optimize Fig. 27 Selection of covalent warheads that have been used to target
reversible binding and selectivity, in addition to modulating cysteine.
the physical properties of the compound. The approach requires
a library of covalent fragments, which ideally would cover a
range of reactivity. However, as such fragments are relatively An alternative strategy is to identify a reversible lead and to
simple, the affinity of these fragments to a target is commonly use this as a scaffold for designing a covalent inhibitor. Some
driven by reactivity rather than reversible binding. Hence, the commonly used covalent warheads that react with cysteine, and
initial hits may have relatively high reactivity, and it may be can be added at a late stage in synthesis, are shown in Fig. 27.
necessary to reduce the reactivity of the covalent warhead in A key benefit to this approach is that the initial reversible
order to reduce the risk of off-target toxicity. Covalent fragments binding provides a starting point for increasing affinity and
are also likely to bind at more than one residue in the target of selectivity. Reversible non-covalent compound libraries are
interest, potentially in regions that will not result in inhibition. larger and more accessible than covalent libraries, therefore
Reversible covalent fragments have also been used to target non- the chance of locating a reversible hit is higher. The approach
catalytic cysteine residues in protein kinases.38 also enables the simultaneous optimization of the reversible
Another covalent screening method involves high-throughput and covalent interactions. This strategy, however, can only be
screening of a focussed subset of known covalent inhibitors. successful if a suitable reversible lead can be found from
This approach is similar to the fragment-based approach except testing a screening library. Furthermore, structural information
that the compounds screened in this case tend to be larger and is key to this approach in order to know where on the reversible
more drug-like. They may also contain warheads that cover a lead to attach the covalent warhead. Additional linker regions
broader range of reactivity, or those that form (reversible) dis- can be added, if required, in order to bring the warhead and
ulphide linkages to the target. This approach was particularly target residue in closer proximity and/or tune reactivity.
successful in terms of the identification of covalent inhibitors for An additional covalent design strategy may be to add the
RAS G12C via a library of disulphides followed by conversion to covalent warhead to a reversible molecule that has been fully
an irreversible warhead.39 Once a number of hits have been optimized for the target of interest. This is a potentially rapid
identified, selectivity and potency can be optimized by structure- way of identifying a covalent inhibitor, but the inhibitor properties
based drug design. This can also be a valuable approach for may require re-optimization after the addition of the warhead.
identifying potential tool compounds, allowing the biological This approach is also fully dependent on structural information of
impact of inhibiting the target to be explored. As with the how the compound binds to the target.
covalent fragment approach, a benefit of this approach is that With all of these strategies outlined, covalent engagement
it can be used even if a reversible inhibitor of the target is not between the target and the inhibitor should ideally be confirmed.
known. Potential issues that may arise from this approach This is often evaluated by assessment of time dependency of the
include the necessity to have a sufficiently large and diverse IC50. If no time-dependence is observed, then it can be assumed
enough library, and the availability of suitable assays and protein that significant (irreversible) covalent inhibition has not taken
constructs to confirm that covalent binding has taken place. place within the time scale of the experiment. For residues that
All of the approaches described thus far involve screening are of low reactivity, e.g. cysteines that are not easily deprotonated,
compounds that already have a covalent warhead attached. covalent bond formation may be slow. Other approaches

3826 | Chem. Soc. Rev., 2018, 47, 3816--3830 This journal is © The Royal Society of Chemistry 2018
View Article Online

Tutorial Review Chem Soc Rev

commonly used include mass spectrometry studies, testing molecular dynamics (MD) and Monte Carlo (MC), allow con-
Published on 05 April 2018. Downloaded by Shanghai Institute of Materia Medica Chinese Academy of Science on 11/15/2021 1:34:22 AM.

compounds against cysteine to serine mutants and X-ray formational sampling to be performed on X-ray structures, or
crystallography. docking poses. MD simulations allow the time dependent
motion between different conformational minima to be
explored. A major drawback to performing MD simulations
6. Computer-aided covalent drug using an MM treatment of the atoms, is that such methods are
design not capable of modelling chemical reactivity. Hence the cova-
lent bond formation event cannot be modelled with such an
Molecular modelling is used extensively in drug development, approach.
and can play a part in prioritising compounds for synthesis, MD simulations can be used within a covalent drug design
post-rationalising experimental observations, and providing project to sample the side chain conformations of a binding
atomic level explanations that can be used to guide further site reactive residue, in order to evaluate its accessibility to a
design efforts. Many of the modelling methods used in bound ligand. MD can also be performed on covalent docking
drug design were developed for application to non-covalent poses, to observe whether non-covalent binding interactions
inhibitors, and have been extensively reviewed elsewhere.40 are maintained once the covalent attachment has formed. For
Although most can also be applied to covalent drug design,41 example, an important part of the reversible recognition of
special considerations, or modified software packages, are protein kinase inhibitors arises from hydrogen bonding inter-
sometimes required, in order to account for the covalent bond actions with the kinase hinge region.
formed between ligand and substrate. Here we briefly intro- Free energy perturbation (FEP) methods have become a
duce some molecular modelling techniques that are useful for useful tool in drug design for calculating relative binding
covalent drug design. affinities of novel compounds to protein binding pockets,
allowing potency improvements to be predicted.43 Relative
6.1 Covalent docking
binding free energies of covalent inhibitors have also been
Docking is a valuable tool in structure-based drug design. predicted using FEP.44 At present, covalent FEP depends on
However, conventional docking approaches have limited use MM force fields, and therefore can only predict changes in
for covalent inhibitors, due to the inability to consider the affinity for the reversible binding component of the inhibitor.
covalent bond formation. Steric repulsion between the covalent Hence, the effects of substituents on warhead reactivity
warhead and the reactive residue may prevent relevant docking cannot be modelled. However, FEP is potentially useful for
poses from being identified. Fortunately, several docking optimizing non-covalent recognition during lead optimiza-
programs now have functionality for covalent docking. Provid- tion, particularly for regions distal to the warhead. This can
ing the suitable functional group has been coded into the be especially valuable as the synthesis of covalent inhibitors
docking program, docking poses are located in which the can be challenging.
covalent bond between receptor and ligand has been formed.
A strain penalty can be included to account for deviations from 6.3 Warhead reactivity
the optimal geometry around the covalent bond. Whilst this Often the reactivity of a covalent warhead will not be in an
approach can be used to assess whether a covalent inhibitor can appropriate reactivity range. When this is the case, it may be
be accommodated in a binding pocket, as well as for scoring ligands possible to tune the reactivity by making structural changes
based on potential non-covalent ligand-receptor interactions, it around the warhead, or modify/replace it entirely. For
cannot currently consider the reactivity of the covalent warhead. example, in the case of cysteine-targeting warheads, making
This needs to be considered separately, as described below. the warhead more electrophilic will be expected to increase
When designing covalent libraries for screening campaigns, the reactivity. This can be achieved by the addition of
covalent docking can be used to filter out compounds that are electron-withdrawing substituents on the inhibitor at a posi-
unlikely to fit in the binding pocket of interest. London tion close to the warhead. A commonly used experimental
et al. have used a covalent docking protocol to screen assay for estimating the reactivity of cysteine-targeting
large virtual libraries of electrophilic fragments to identify warheads is measurement of the half-life towards adduct
reversible covalent fragments that target non-conserved formation with glutathione (GSH). However, such GSH assays
cysteine residues in several protein kinases.42 The authors are often relatively low-throughput and require compounds to
reported sub-micromolar active compounds, indicating the be synthesized. Several groups have demonstrated that GSH
potential use of such an approach for the rapid discovery of half-life can be predicted using computational methods.45,46
covalent probe molecules. These methods include Hammett parameters, quantum
mechanical (QM) calculations and pKa predictions. Such
6.2 Molecular mechanics (MM) based simulations methods can be used to rank compounds based on predicted
Whilst X-ray crystal structures are invaluable for structure- reactivity and help to prioritise ‘virtual’ compounds for synth-
based drug design, they may be misleading in terms of esis that fall in the desired range. Understanding substituent
conformational preferences of amino acid side chains. Mole- effects on warhead reactivity therefore enables knowledge-
cular mechanics (MM) based simulation methods, such as driven design of covalent inhibitors.

This journal is © The Royal Society of Chemistry 2018 Chem. Soc. Rev., 2018, 47, 3816--3830 | 3827
View Article Online

Chem Soc Rev Tutorial Review

In addition to reactivity with GSH, alternative model 7. Conclusions


Published on 05 April 2018. Downloaded by Shanghai Institute of Materia Medica Chinese Academy of Science on 11/15/2021 1:34:22 AM.

chemical systems have been used to measure covalent warhead


reactivity, including pyridinium and sulfonium ylides,47 and Covalent drug design is a growing and promising area. The
N-a-acetyl-L-lysine.22 number and diversity of covalent inhibitors, and knowledge of
In addition to substituent effects, the effects of warhead how to discover them, is constantly increasing. There is further
exchange can be investigated computationally. Matched mole- scope for extensive development in this area, particularly
cular pairs analysis (MMPA) revealed that structure activity around targeting non-catalytic residues other than cysteine.
relationships are transferrable between the warheads shown There are numerous strategies for covalent drug design
in Fig. 27.46 This enabled the calculation of a warhead correc- projects, the choice will be influenced by the amount of prior
tion factor that can be used to estimate the change in GSH information and/or screening library resources. Computational
half-life when switching from e.g. an acrylamide to a vinyl chemistry techniques can have a crucial impact on covalent
sulphonamide warhead. However, this may not necessarily inhibitor design and will continue to be an important aspect to
result in a more potent compound, as the geometry of the new the development of this exciting area.
warhead may be less favourable with respect to the reactive
residue. Conflicts of interest
Albeck et al. have described the Enzyme Mechanism
Based Method (EMBM) for rational design of reversible There are no conflicts to declare.
covalent inhibitors based on their predicted reactivity with
the target.48 They have used the EMBM method to calculate Acknowledgements
descriptors which are able to discriminate between reactivity of
covalent compounds towards serine and cysteine hydrolases. The authors thank Drs M. R. V. Finlay, N. P. Grimster and
M. J. Packer for useful discussion. This work was funded by the
6.4 Mechanistic modelling AstraZeneca Postdoctoral Programme.
Molecular mechanics (MM)-based methods are useful for
modelling non-covalent interactions in large systems, such as Notes and references
protein–ligand complexes. However, as discussed in 6.2, they
are unable to model reactivity effects involved in covalent complex 1 R. S. Bohacek, C. McMartin and W. C. Guida, Med. Res. Rev.,
formation. QM methods are required in order to accurately model 1996, 16, 3–50.
chemical reactions, but these methods are typically limited to 2 J. Singh, R. C. Petter, T. A. Baillie and A. Whitty, Nat. Rev.
small model systems. A well-established method to model Drug Discovery, 2011, 10, 307–317.
chemical reactions in large systems is to use the hybrid quantum 3 J. Zhang, P. L. Yang and N. S. Gray, Nat. Rev. Cancer, 2009, 9,
mechanics/molecular mechanics approach (QM/MM).49 Here, a 28–39.
small region of interest (i.e. the inhibitor and target residue) is 4 T. Barf, T. Covey, R. Izumi, B. van de Kar, M. Gulrajani,
treated with a QM method, allowing the chemical reaction to be B. van Lith, M. van Hoek, E. de Zwart, D. Mittag, D. Demont,
modelled. The rest of the protein and solvent are treated with a S. Verkaik, F. Krantz, P. G. Pearson, R. Ulrich and
MM representation. The advantage of the QM/MM approach A. Kaptein, J. Pharmacol. Exp. Ther., 2017, 363, 240–252.
over typical small model QM methods is that the effect of the 5 T. Barf and A. Kaptein, J. Med. Chem., 2012, 55,
surrounding protein and solvent is included, both in terms of 6243–6262.
their effects on geometry, but also on the local electrostatic 6 Y. Shibata and M. Chiba, Drug Metab. Dispos., 2015, 43,
environment. 375–384.
QM/MM methods can be used to explore mechanistic 7 D. J. Drucker and M. A. Nauck, Lancet, 2006, 368,
details of enzymatic reactions, which can be useful in design- 1696–1705.
ing covalent inhibitors. Detailed structural and electrostatic 8 D. J. Augeri, J. A. Robl, D. A. Betebenner, D. R. Magnin,
knowledge of the transition state for the reaction with A. Khanna, J. G. Robertson, A. Wang, L. M. Simpkins,
the natural substrate can be used to identify important P. Taunk, Q. Huang, S. P. Han, B. Abboa-Offei, M. Cap,
binding interactions. For example, Ma et al. applied QM/MM L. Xin, L. Tao, E. Tozzo, G. E. Welzel, D. M. Egan,
simulations to study the hydrolytic mechanism of a peptide J. Marcinkeviciene, S. Y. Chang, S. A. Biller, M. S. Kirby,
substrate catalysed by human Cathepsin K.50 They observed R. A. Parker and L. G. Hamann, J. Med. Chem., 2005, 48,
the initial nucleophilic attack by Cys25 on the carbonyl 5025–5037.
carbon of the substrate to be the rate limiting chemical step, 9 D. Bromme and F. Lecaille, Expert Opin. Invest. Drugs, 2009,
with a barrier 10 kcal mol1 lower than the corresponding 18, 585–600.
reference reaction in solution. Furthermore, they calculated 10 M. E. McGrath, J. L. Klaus, M. G. Barnes and D. Brömme,
the individual contributions of the residues surrounding the Nat. Struct. Biol., 1997, 4, 105–109.
active site to the transition state stabilisation energy, thus 11 J. Y. Gauthier, N. Chauret, W. Cromlish, S. Desmarais,
revealing the key residues with which to form interactions L. T. Duong, J. P. Falgueyret, D. B. Kimmel, S. Lamontagne,
with an inhibitor. S. Leger, T. LeRiche, C. S. Li, F. Masse, D. J. McKay, D. A.

3828 | Chem. Soc. Rev., 2018, 47, 3816--3830 This journal is © The Royal Society of Chemistry 2018
View Article Online

Tutorial Review Chem Soc Rev

Nicoll-Griffith, R. M. Oballa, J. T. Palmer, M. D. Percival, 27 S. Lin, X. Yang, S. Jia, A. M. Weeks, M. Hornsby, P. S. Lee,
Published on 05 April 2018. Downloaded by Shanghai Institute of Materia Medica Chinese Academy of Science on 11/15/2021 1:34:22 AM.

D. Riendeau, J. Robichaud, G. A. Rodan, S. B. Rodan, C. Seto, R. V. Nichiporuk, A. T. Iavarone, J. A. Wells, F. D. Toste and
M. Therien, V. L. Truong, M. C. Venuti, G. Wesolowski, C. J. Chang, Science, 2017, 355, 597–602.
R. N. Young, R. Zamboni and W. C. Black, Bioorg. Med. 28 L. M. McGregor, M. L. Jenkins, C. Kerwin, J. E. Burke and
Chem. Lett., 2008, 18, 923–928. K. M. Shokat, Biochemistry, 2017, 56, 3178–3183.
12 P. A. Jackson, J. C. Widen, D. A. Harki and K. M. Brummond, 29 P. Martin-Gago, E. K. Fansa, M. Winzker, S. Murarka, P. Janning,
J. Med. Chem., 2017, 60, 839–885. C. Schultz-Fademrecht, M. Baumann, A. Wittinghofer and
13 J. M. Ostrem and K. M. Shokat, Nat. Rev. Drug Discovery, H. Waldmann, Cell Chem. Biol., 2017, 24(589-597), e585.
2016, 15, 771–785. 30 K. M. Backus, B. E. Correia, K. M. Lum, S. Forli,
14 S. V. Sharma, D. W. Bell, J. Settleman and D. A. Haber, B. D. Horning, G. E. Gonzalez-Paez, S. Chatterjee,
Nat. Rev. Cancer, 2007, 7, 169–181. B. R. Lanning, J. R. Teijaro, A. J. Olson, D. W. Wolan and
15 C. H. Yun, K. E. Mengwasser, A. V. Toms, M. S. Woo, B. F. Cravatt, Nature, 2016, 534, 570–574.
H. Greulich, K. K. Wong, M. Meyerson and M. J. Eck, 31 Z. Zhao, Q. Liu, S. Bliven, L. Xie and P. E. Bourne, J. Med.
Proc. Natl. Acad. Sci. U. S. A., 2008, 105, 2070–2075. Chem., 2017, 60, 2879–2889.
16 J. Singh, E. M. Dobrusin, D. W. Fry, T. Haske, A. Whitty and 32 Y. Zhang, D. Zhang, H. Tian, Y. Jiao, Z. Shi, T. Ran, H. Liu,
D. J. McNamara, J. Med. Chem., 1997, 40, 1130–1135. S. Lu, A. Xu, X. Qiao, J. Pan, L. Yin, W. Zhou, T. Lu and
17 D. W. Fry, A. J. Bridges, W. A. Denny, A. Doherty, K. D. Greis, Y. Chen, Mol. Pharmaceutics, 2016, 13, 3106–3118.
J. L. Hicks, K. E. Hook, P. R. Keller, W. R. Leopold, J. A. Loo, 33 S. Wu, H. Luo Howard, H. Wang, W. Zhao, Q. Hu and Y. Yang,
D. J. McNamara, J. M. Nelson, V. Sherwood, J. B. Smaill, Biochem. Biophys. Res. Commun., 2016, 478, 1268–1273.
S. Trumpp-Kallmeyer and E. M. Dobrusin, Proc. Natl. Acad. 34 Z. Guo, B. Li, L. T. Cheng, S. Zhou, J. A. McCammon and
Sci. U. S. A., 1998, 95, 12022–12027. J. Che, J. Chem. Theory Comput., 2015, 11, 753–765.
18 D. A. Cross, S. E. Ashton, S. Ghiorghiu, C. Eberlein, C. A. 35 S. Pinitglang, A. B. Watts, M. Patel, J. D. Reid, M. A. Noble,
Nebhan, P. J. Spitzler, J. P. Orme, M. R. Finlay, R. A. Ward, S. Gul, A. Bokth, A. Naeem, H. Patel, E. W. Thomas, S. K.
M. J. Mellor, G. Hughes, A. Rahi, V. N. Jacobs, M. Red Sreedharan, C. Verma and K. Brocklehurst, Biochemistry,
Brewer, E. Ichihara, J. Sun, H. Jin, P. Ballard, K. Al-Kadhimi, 1997, 36, 9968–9982.
R. Rowlinson, T. Klinowska, G. H. Richmond, M. Cantarini, 36 E. Awoonor-Williams and C. N. Rowley, J. Chem. Theory
D. W. Kim, M. R. Ranson and W. Pao, Cancer Discovery, Comput., 2016, 12, 4662–4673.
2014, 4, 1046–1061. 37 D. C. Swinney and J. Anthony, Nat. Rev. Drug Discovery, 2011,
19 K. S. Thress, C. P. Paweletz, E. Felip, B. C. Cho, 10, 507–519.
D. Stetson, B. Dougherty, Z. Lai, A. Markovets, 38 R. M. Miller, V. O. Paavilainen, S. Krishnan, I. M. Serafimova
A. Vivancos, Y. Kuang, D. Ercan, S. E. Matthews, and J. Taunton, J. Am. Chem. Soc., 2013, 135, 5298–5301.
M. Cantarini, J. C. Barrett, P. A. Janne and G. R. Oxnard, 39 J. M. Ostrem, U. Peters, M. L. Sos, J. A. Wells and
Nat. Med., 2015, 21, 560–562. K. M. Shokat, Nature, 2013, 503, 548–551.
20 J. Pettinger, K. Jones and M. D. Cheeseman, Angew. Chem., 40 G. Sliwoski, S. Kothiwale, J. Meiler and E. W. Lowe, Jr.,
Int. Ed., 2017, 56, 15200–15209. Pharmacol. Rev., 2014, 66, 334–395.
21 G. Akcay, M. A. Belmonte, B. Aquila, C. Chuaqui, A. W. Hird, 41 E. Awoonor-Williams, A. G. Walsh and C. N. Rowley,
M. L. Lamb, P. B. Rawlins, N. Su, S. Tentarelli, N. P. Biochim. Biophys. Acta, 2017, 1865, 1664–1675.
Grimster and Q. Su, Nat. Chem. Biol., 2016, 12, 931–936. 42 N. London, R. M. Miller, S. Krishnan, K. Uchida, J. J. Irwin,
22 U. P. Dahal, A. M. Gilbert, R. S. Obach, M. E. Flanagan, O. Eidam, L. Gibold, P. Cimermancic, R. Bonnet, B. K. Shoichet
J. M. Chen, C. Garcia-Irizarry, J. T. Starr, B. Schuff, and J. Taunton, Nat. Chem. Biol., 2014, 10, 1066–1072.
D. P. Uccello and J. A. Young, MedChemComm, 2016, 7, 43 L. Wang, Y. Wu, Y. Deng, B. Kim, L. Pierce, G. Krilov,
864–872. D. Lupyan, S. Robinson, M. K. Dahlgren, J. Greenwood,
23 E. Anscombe, E. Meschini, R. Mora-Vidal, M. P. Martin, D. L. Romero, C. Masse, J. L. Knight, T. Steinbrecher,
D. Staunton, M. Geitmann, U. H. Danielson, W. A. Stanley, T. Beuming, W. Damm, E. Harder, W. Sherman,
L. Z. Wang, T. Reuillon, B. T. Golding, C. Cano, D. R. Newell, M. Brewer, R. Wester, M. Murcko, L. Frye, R. Farid, T. Lin,
M. E. Noble, S. R. Wedge, J. A. Endicott and R. J. Griffin, D. L. Mobley, W. L. Jorgensen, B. J. Berne, R. A. Friesner and
Chem. Biol., 2015, 22, 1159–1164. R. Abel, J. Am. Chem. Soc., 2015, 137, 2695–2703.
24 H. Mukherjee, J. Debreczeni, J. Breed, S. Tentarelli, 44 B. Kuhn, M. Tichy, L. Wang, S. Robinson, R. E. Martin,
B. Aquila, J. E. Dowling, A. Whitty and N. P. Grimster, Org. A. Kuglstatter, J. Benz, M. Giroud, T. Schirmeister,
Biomol. Chem., 2017, 15, 9685–9695. R. Abel, F. Diederich and J. Hert, J. Med. Chem., 2017, 60,
25 S. E. Dalton, L. Dittus, D. A. Thomas, M. A. Convery, 2485–2497.
J. Nunes, J. T. Bush, J. P. Evans, T. Werner, M. Bantscheff, 45 M. E. Flanagan, J. A. Abramite, D. P. Anderson, A. Aulabaugh,
J. A. Murphy and S. Campos, J. Am. Chem. Soc., 2018, 140(3), U. P. Dahal, A. M. Gilbert, C. Li, J. Montgomery,
932–939. S. R. Oppenheimer, T. Ryder, B. P. Schuff, D. P. Uccello,
26 D. A. Shannon and E. Weerapana, Curr. Opin. Chem. Biol., G. S. Walker, Y. Wu, M. F. Brown, J. M. Chen, M. M. Hayward,
2015, 24, 18–26. M. C. Noe, R. S. Obach, L. Philippe, V. Shanmugasundaram,

This journal is © The Royal Society of Chemistry 2018 Chem. Soc. Rev., 2018, 47, 3816--3830 | 3829
View Article Online

Chem Soc Rev Tutorial Review

M. J. Shapiro, J. Starr, J. Stroh and Y. Che, J. Med. Chem., 48 M. Shokhen, M. Hirsch, N. Khazanov, R. Ozeri, N. Perlman,
Published on 05 April 2018. Downloaded by Shanghai Institute of Materia Medica Chinese Academy of Science on 11/15/2021 1:34:22 AM.

2014, 57, 10072–10079. T. Traube, S. Vijayakumar and A. Albeck, Isr. J. Chem., 2014,
46 R. Lonsdale, J. Burgess, N. Colclough, N. L. Davies, 54, 1137–1151.
E. M. Lenz, A. L. Orton and R. A. Ward, J. Chem. Inf. Model., 49 M. W. van der Kamp and A. J. Mulholland, Biochemistry,
2017, 57, 3124–3137. 2013, 52, 2708–2728.
47 D. S. Allgauer, H. Jangra, H. Asahara, Z. Li, Q. Chen, H. Zipse, 50 S. Ma, L. S. Devi-Kesavan and J. Gao, J. Am. Chem. Soc., 2007,
A. R. Ofial and H. Mayr, J. Am. Chem. Soc., 2017, 139, 13318–13329. 129, 13633–13645.

3830 | Chem. Soc. Rev., 2018, 47, 3816--3830 This journal is © The Royal Society of Chemistry 2018

You might also like