Mechanistic Enzymology in Drug Discovery: a Fresh Perspective

Home , EZH2, Hit to lead, Mixed inhibition, Mode of action, Product (chemistry), Uncompetitive inhibitor

REVIEWS

Mechanistic enzymology in drug discovery: a fresh perspective

Geoffrey A. Holdgate1, Thomas D. Meek2 and Rachel L. Grimley1 Abstract | Given the therapeutic and commercial success of small-molecule enzyme inhibitors, as exemplified by kinase inhibitors in oncology, a major focus of current drug-discovery and development efforts is on enzyme targets. Understanding the course of an enzyme-catalysed reaction can help to conceptualize different types of inhibitor and to inform the design of screens to identify desired mechanisms. Exploiting this information allows the thorough evaluation of diverse compounds, providing the knowledge required to efficiently optimize leads towards differentiated candidate drugs. This review highlights the rationale for conducting high-quality mechanistic enzymology studies and considers the added value in combining such studies with orthogonal biophysical methods.

Mechanism Enzyme inhibitors and inactivators comprise roughly a useful compendium of enzymes that are the targets of 1,2 A process by which a reaction half of all marketed drugs and have transformed marketed drugs and makes the case that unlike drug takes place, fully determined human medicine. For example, angiotensin-converting targets of other protein classes, knowledge of the chem- when all the intermediates, enzyme (ACE) inhibitors3, including captopril4,5 and ical mechanism of an individual enzyme, including the complexes and conformational lysinopril6, which emerged in the late 1970s, now con- structural characterization of its transition state14–17, states of an enzyme are characterized and the rate stitute perhaps the most important class of drugs for the exploitation of nucleophilic, active-site residues to 18,19 constants associated with the the treatment of hypertension. ACE inhibitors arose form covalent adducts with electrophilic inactivators conversion between them are from rational design based on the mechanism of action or the chemical conflation of two enzyme substrates quantified. The term is often (MoA) of this metalloproteinase. In addition, in the into a bi‑substrate analogue20,21, affords the opportu- used in inhibition studies to 7 distinguish between different 1980s, statins, exemplified by atorvastatin and lovasta- nity to design de novo enzyme inhibitors based on the 8 modes of inhibition. tin , emerged as natural product-derived inhibitors of chemical species that emerge during catalysis. Drugs 3‑hydroxy‑3‑glutaryl CoA reductase and are now the have therefore been designed that target a number of most commonly used drugs to treat hypercholesterol different enzyme forms via covalent and noncovalent aemia9. Beginning in the 1990s, saquinivir10 and indina- inhibition (FIG. 1). vir11 were identified as rationally designed inhibitors of Understanding the role of enzymes in disease states the aspartic protease of HIV‑1 (REF. 12). HIV‑1 protease and the implementation of strategies to modulate their inhibitors have greatly contributed to the conversion of activities for therapeutic benefit remains a key focus for HIV/AIDS from a ‘death sentence’ to a manageable dis- drug discovery. However, while enzymology is a power ease. Furthermore, in the past 20 years, ATP-competitive ful and established discipline, it often fails to receive inhibitors of protein kinases have provided an arsenal of recognition in providing scientific insight into problems anticancer and anti-inflammatory drugs. experienced along the route from target identification 1Discovery Sciences, IMED Unsurprisingly, enzymes now typically comprise one- to proof of concept in the clinic. This is particularly true Biotech Unit, AstraZeneca, third or more of the discrete drug targets found within at the onset of a drug discovery campaign that targets Building 310, Cambridge Science Park, Milton Road, the portfolios of large pharmaceutical companies. This an enzyme. It is uncommon to have a detailed under- Cambridge, CB4 0WG, UK. is true even for those pharmaceutical companies that standing of the kinetic mechanism (the order of substrate 2Department of Biochemistry have increased their reliance on biopharmaceutical addition and product release) and the chemical mecha & Biophysics, Texas A&M therapies. Because biopharmaceutical agents are gener- nism of catalysis of an enzyme target before the initia- University, College Station, ally confined in their utility to extracellular drug targets, tion of ‘hit’ discovery. Ironically, at this early stage of a Texas 77843, USA. intracellular enzymes still may prove to be a growth area programme, an X‑ray structure of the enzyme target may Correspondence to G.A.H. geoff.holdgate@astrazeneca. for drug discovery. be extant even in the absence of an understanding of its com By their very nature, enzymes are dynamic, with a sin- detailed mechanism of catalysis. As a result, the oppor- gle enzyme representing a number of different targets as tunity for the use of rational design to afford enzyme doi:10.1038/nrd.2017.219 Published online 1 Dec 2017; a consequence of binding substrates, intermediates and inhibitors or inactivators based on the enzyme mecha- corrected online 15 Dec 2017 products during the catalytic cycle. Robertson13 provided nism may have been eclipsed by a medicinal chemistry

REVIEWS

Catalysis

Aspirin: COX1 or COX2 S + E E·S E·X‡ E·P E + P β-Lactams (amoxicillin): transpeptidases Bortezomib: proteasome peptidase Fluorouracil (FdUMP): dTMP synthase Selegiline: mono-amine oxidase E–I E·I E·S·I E·I‡ E·P·I Triclosan: FabI Clavulonate: β-lactamase Allopurinol: xanthine oxidase Osimertinib: EGFR tyrosine kinase Immucillin-H: E·I E·S–I purine nucleoside phosphorylase

Methotrexate: dihydrofolate reductase Avodart: steroid 5α-reductase Allopurinol: OMP decarboxylase Mycophenolate mofetil: IMP dehydrogenase Mupirocin: Ile-tRNA synthetase Lexivir: HIV protease

Figure 1 | The kinetics of drug–target interactions. A single enzyme (E) may represent a number of different targets Nature Reviews | Drug Discovery that may be the focus of chemical intervention to modulate activity. By their very nature, enzymes are dynamic and bind substrates (S), intermediates (X) and products (P) during the catalytic cycle. Enzyme inhibitors (I) may bind to these different species either covalently or non-covalently. It is possible to modulate the abundance of these different enzyme species by changing assay conditions to obtain a balance between enzyme forms (to maximize the chances of finding any type of inhibitory mechanism) or to increase the relative concentration of some forms (in order to focus on particular mechanisms). Here, we show a range of different drug molecules eliciting their effects by binding to different enzyme forms via non-covalent (E·I) or covalent (E–I) binding. ‡ represents the enzyme’s transition state. An early understanding of the enzyme mechanism is advantageous for the identification and subsequent characterization of hits and leads. COX, cyclooxygenase; dTMP, deoxythymidine monophosphate; EGFR, epidermal growth factor receptor; FabI, enoyl-acyl carrier protein reductase; IMP, inosine‑5ʹ-monophosphate; OMP, orotidine 5ʹ-phosphate.

effort that focuses on a limited number of chemotypes potency, selectivity versus homologous targets and their that in no way resemble the substrates of the enzyme developability properties, and MoA data are delivered target. Additionally, following the knowledge gleaned for all lead and drug candidates. This practice provides from more than 20 years of ‘random’ high-throughput a full ‘data dossier’ for the advancement of lead com- screening (HTS) applied to hundreds of enzyme targets, pounds to clinical candidates with a well-characterized pharmaceutical companies are increasingly capable of MoA, cellular efficacy and the appropriate safety and predicting which classes of enzyme targets in their active pharmacokinetic properties. This is particularly true portfolios will provide durable lead compounds from when one considers the difficulty of registering a drug HTS campaigns and which are likely to fail to do so. that has no known functional target23. Additionally, it For these and other reasons, mechanistic studies of allows the identification and prioritization of not just enzyme targets are now enjoying a renaissance, with a single mechanism but potentially a range of diverse many pharmaceutical companies increasing their focus mechanisms, thoroughly characterized using increas- on using detailed, more physiologically relevant enzyme ingly complex assays (isolated protein through to tis- kinetic studies during the lead identification and optimi- sues), to produce a number of differentiated options to zation phases as another measure to address the increas- achieve the desired biological effect in vivo. ingly high cost and low success rates of drug discovery Of course, enzymology alone cannot deliver all of within the industry. There is renewed recognition that the required data or information needed in early drug (mechanistic) enzymology is essential for establishing discovery, and its combination with biophysical meth- robust, reliable and appropriate assays for screening ods and protein structural analysis remains essential for campaigns to identify hit compounds that are also adapt- providing an integrated pharmacological view of the able to comprehensively characterize inhibitors more thermodynamics and kinetics of enzyme catalysis and fully than traditional half-maximal inhibitory concentration inhibition24. Furthermore, integrating the understand-

(IC50) values alone. While useful, the information inher- ing derived from detailed mechanistic characterization

ent in IC50 values does not indicate a change in inhibi- in isolated enzyme assays coupled with cell-based assays tory mechanism (for example, a shift from competitive and, ultimately, with metabolic pathways and systems to mixed-type inhibition) in a series of lead compounds knowledge will continue to identify enzymology as one and moreover, the data are poorly correlative with their of the most important quantitative disciplines within pharmacodynamic action, as they give no measure of drug discovery. Half-maximal inhibitory the period of time the compound ‘resides’ on its target22. This article will discuss the importance of conduct- concentration The purpose of these MoA studies is to character- ing high-quality mechanistic enzymology studies, con- (IC ).The inhibitor 50 ize the interaction of a compound with its target and sider how detailed and early knowledge of an enzyme concentration that gives a 50% decrease in rate under the to understand how natural ligands at physiological mechanism may add value throughout a drug-discovery specific assay conditions concentrations will modulate this activity. Ideally, data programme and assess the value of combining such employed. from a series of hit compounds are annotated in terms of knowledge with orthogonal biophysical methods.

REVIEWS

Pre-screening considerations tyrosine protein kinase ITK/TSK (ITK) as a model sys- What is the appropriate biochemical nature of an tem to understand the regulation of TEC family kinases, enzyme target and its substrate? When a decision is identified a 17‑residue linker sequence between the SH2 made to pursue a discrete enzyme as a drug target, the and kinase domains as a critical component that main- optimal biochemical nature of that enzyme and its sub- tains ITK kinase activity, along with several specific resi strates becomes the subject of debate. Does one insist dues that mediate interactions between this regulatory on a purely cellular environment for the enzyme and region and the catalytic domain of ITK28. In this system, its substrates to maintain biological integrity or is one the use of the isolated kinase domain would be poorly able to proceed with a purified recombinant enzyme and representative of the kinetic characteristics of the enzyme. substrates that may not resemble its native substrates? Furthermore, demonstrating also the impact of protein

As one example, the development of HIV‑1 protease tags, the kcat value for the hexahistidine-tagged cytoplas- inhibitors as drugs was achieved with a recombinant mic domain of hepatocyte growth factor receptor (HGFR; decrement of the protease and with using non-protein, also known as c-MET) has been reported to be 0.0095 s−1 (REF. 29) oligopeptide substrates resembling cognate cleavage , whereas the kcat value for the untagged catalytic sites in retroviral polyproteins15. However, as elaborated domain has been reported to be 1 s−1 (unphosphorylated below, protein kinases provide interesting examples in c-MET) and 31 s−1 (autophosphorylated c-MET)30. which a recombinant ‘biochemical’ enzyme species must With respect to small-molecule inhibitors, U0126 has more closely conform to its native ‘biological’ entity25. been reported to inhibit the truncated, constitutively active Classical drug-discovery campaigns generally include recombinant kinase domain of dual specificity mitogen- a combination of biochemical, biophysical and cell biol- activated protein kinase (MAPK) kinase 1 (MKK1) with

ogy assays and methodologies to identify and character- an IC50 of ~70 nM, whereas it inhibits the full-length ize modulators of key targets and pathways of interest. endogenously activated wild-type enzyme isolated from (REF. 31) In order to derive the most relevant insights from the cells with an IC50 of ~500 nM . In an even more various approaches with the goal of translating in vitro striking example, the addition of a 19-amino acid juxta data to a clinical disease setting, a number of important membrane domain to the catalytic domain of vascular considerations are required. endothelial growth factor receptor (VEGFR) increases the affinity of a small-molecule inhibitor from 1,100 pM to Physiological environment 28 pM (REF. 32). In another example, the catalytic domain A comprehensive understanding of the molecular envi- of the human β‑amyloid precursor protein cleaving ronment of the target of interest is crucial to enable the enzyme (BACE1) was reported to essentially function design of more physiologically relevant assays. Recent independently of the rest of the protein and hence can be times have seen the pharmaceutical industry moving away used as a surrogate for the full-length protein33. from the use of isolated, usually recombinant enzyme Detailed characterization and comparison of the domains, model or generic substrates and returning mechanism and key kinetic parameters between the dif- towards a focus on systems likely to be of higher rele- ferent enzyme forms are required in order to be confident vance and translatability. This has resulted in a greater of the relevance of the isolated domain with regard to the number of assays employing full-length enzymes, com- pathophysiological state of the enzyme. Domain archi- plex partners and native substrates, requiring a more tecture, post-translational modification status, activation detailed analysis of the underlying enzymology. state and artificial modifications such as affinity tags can It is important to consider that in the complex all affect the kinetic integrity of the enzyme. physiological environment, full-length, multi-domain enzymes comprise catalytic domains that can be regu- Binding partners lated by non-catalytic domains in a manner that controls Complex binding partners can also exert significant activity, substrate specificity and potentially inhibi- influence on the catalytic activity of the associated tor structure–activity relationships (SARs). The AKT enzyme, a classic example being the cyclin partners of serine/threonine protein kinases (AKT1–3) provide an cyclin-dependent kinase (CDK) enzymes, where activa- example. A research strategy that utilized a recombi- tion of the cyclin-dependent kinases is a two-step process nant fragment encompassing only the catalytic domain comprising cyclin binding, followed by phosphorylation of this membrane-associated protein kinase, rather at a conserved threonine residue within the kinase acti- than the native, full-length protein, would have over- vation loop34. Such modifications have the potential to looked the allosteric site on the amino-terminal pleck- also affect inhibitor binding. For example, targeting pro- strin homology domain that ultimately provided an tein for XKLP2 (TPX2) is a physiological binding part- exploitable site for inhibitor discovery26. While one phar- ner and positive regulator of Aurora A kinase (AurA); maceutical company could have decided that recombi- while the presence of TPX2 has no effect on the reaction nant expression of the kinase domain of the AKT kinases mechanism and enzyme turnover, it does affect the bind- would prove more expeditious and more amenable to ing affinity of the kinase’s substrates and also the SAR structural characterization, Merck chose the path of of inhibitors binding to a hydrophobic pocket adjacent producing the ‘native’ kinase (including the pleckstrin to the ATP binding site35 (TABLE 1). Profiling in the pres- homology domain) and thereby discovered a new site for ence of the native binding partner can clearly enhance enzyme inhibition, defining a new chemical equity26,27 the probability of identifying molecules that are more (TABLE 1). In another example, a kinetic analysis using likely to have an impact in physiological environments.

REVIEWS

Table 1 | Examples of small-molecule modulation observed under differential profiling conditions, with specific reference to kinases Scenario Example Key observations Small-molecule modulation Refs

Effect of profiling AKT +/− PH domain Discovery of allosteric AKT–I-1: Ki ≈ 3 μM (AKT + PH), >250 μM (AKT – PH) 26,27 full-length versus inhibitors that modulate truncated enzyme only ‘native’ kinase (AKT + PH domain) N N HN

HN O

Effect of profiling Aurora A kinase Rapid-equilibrium, Differential inhibition of AurA versus AurA + TPX2 (TPX2 35 against enzyme +/− +/− TPX2 (‘native’ random-order mechanism: binding decreases the size and accessibility of the differential binding binding partner) TPX2 binding has limited hydrophobic pocket adjacent to the ATP site to inhibitors)

partners effect on kcat; however, AurA + TPX2 has enhanced affinity (~14‑fold decrease A: GW801372X: Ki ≈ 220 nM (AurA), 220 nM (Aur – TPX2) in Kd, ~3‑fold decrease in Km) for ATP + phosphoacceptor N H substrates N N O

N O N N

B: VX‑680: Ki ≈ 1.4 nM (AurA), 5.9 nM (Aur + TPX2)

O N NH N

HN N S

HN N

C: GSK623906A: Ki ≈ 85 nM (AurA), 1,500 nM (Aur + TPX2)

O F S N

N NH

O O

REVIEWS

Table 1 (cont.) | Examples of small-molecule modulation observed under differential profiling conditions, with specific reference to kinases Scenario Example Key observations Small-molecule modulation Refs

Effects of profiling P38 kinase (ATF2 ATP-competitive allosteric CMPD1: Ki,app ≈ 330 nM (MK2a), >20 μM (ATF2) 46,47 against enzyme +/− versus MK2a) inhibition (BIRB‑796) differential substrates prevents phosphorylation OH of all substrates versus non-competitive, MK2a HN substrate-selective inhibition (CMPD1) O

BIRB 796: Ki,app ≈ 0.1 nM (all substrates including MK2a and ATF2)

O O N N N N H H

Characterizing ERK1 and ERK2 (0‑P Two inhibitors of ERK1 and 2 Vertex‑11e: Ki,app ≈ 0.34 nM (0‑P), 2.5 nM (2‑P); residence 83 slow-binding inhibitors and 2‑P forms) (Vertex‑11e and SCH772984) time ≈ 4.8 h and inhibition of with the greatest reported F different modified cellular efficacy demonstrate forms (phosphorylation slow binding and slow states) of the enzyme dissociation from ERK2. Slow dissociation is an important Cl factor that may predict higher N NH drug potency due to greater cellular retention N

NH O

abs NH

OH Cl

SCH772984: Ki,app ≈ 0.12 nM (0‑P), 0.12 nM (2‑P); residence time ≈ 0.9 h

H N N N O HN N abs N N O N N

REVIEWS

−1 −1 −1 Characterizing EGFR (wild-type and Understanding the kinact and Neratinib: Ki = 0.14 nM; kinact = 1.1 ms ; kinact/Ki = 7 μM s 97 irreversible covalent mutant) Ki values is crucial to aid in inhibitors the dissection of molecular determinants underlying overall inhibitor potency O N

HN N HN Cl O

O N N

Effect of promiscuous PKC and MKK1 Many kinase inhibitors Often these compounds are nonspecific, 71 compounds have been shown to inhibit aggregate-forming small molecules inhibiting a unrelated enzymes (for range of enzymes in a time-dependent manner example, β‑lactamase, that is sensitive to the concentration of the target chymotrypsin and malate enzyme as well as the presence of detergent.

dehydrogenase) with IC50 values similar to those of the original targets Rottlerin: IC50 ≈ 3 μM (PKCδ), 1 μM (chymotrypsin, MDH)

O O

HO OH HO O

OH OH

K-252c: IC50 ≈ 2.5 μM (PKC), 9 μM (chymotrypsin, MDH)

NH H N

O N H

UO126: IC50 ≈ 13 μM (MKK1), 100 μM (chymotrypsin, MDH)

H 2N

NH2 H2N S NH2

S N

AKT, RAC-alpha serine/threonine protein kinase; ATF2, cyclic AMP-dependent transcription factor ATF2; EGFR, epidermal growth factor receptor;

ERK, extracellular-signal-regulated kinase; IC50, half-maximal inhibitory concentration; kcat, rate constant for catalysis; Kd, dissociation constant; Ki, inhibition constant; Ki,app, apparent inhibition constant, dependent on the inhibition mechanism; kinact, rate constant for inactivation; Km, Michaelis constant; MDH, malate dehydrogenase; MKK1, dual specificity mitogen-activated protein kinase kinase 1; PH, pleckstrin homology; PKC, protein kinase C; TPX2, targeting protein for XKLP2.

REVIEWS

More than a single enzyme subunit may come into play. to be inhibited in a cellular context. For example, in For example, the histone-lysine N-methyltransferase the case of p38, differential substrate selectivity is able EZH2 can fully catalyse the methylation of histone H3 to define the particular pathways affected by small lysine 27 only in the presence of its associated partners molecules, offering exquisite control and, potentially, comprising its full pentameric Polycomb repressive fewer unwanted side effects46,47 (TABLE 1). complex 236. Hence, an understanding of the physio For a number of years in the epigenetic field, the logically relevant protein complex is key to ensuring that importance of using native nucleosome substrates in the correct form of the enzyme is represented. Indeed, preference to short, histone N‑terminal tail peptides there may be specific complexes associated with health- as surrogates has been realized in order to identify relevant and disease-relevant states of the enzyme, the small-molecule inhibitors that translate through to knowledge of which can help to drive selectivity. A study the cellular context34 and target the desired activ- at Agios Pharmaceuticals used metabolomics to discover ity of the enzyme. In the case of the histone-lysine that 2‑hydroxyglutarate, an agent that causes malignant N-methyltransferase NSD2, for example, the specific- gliomas, results from the neomorphic activity of human ity of lysine methylation is directed by the choice of a isocitrate dehydrogenase (IDH1), which, owing to a nucleosome substrate versus a histone octamer sub- single point mutation (R132H), catalyses its formation strate in the presence of short single-stranded DNA or from α-ketoglutarate37. Immunoprecipitation from double-stranded DNA48. In order to elucidate the true disease-relevant cells coupled with mass spectrometry structure of the transition states of the NSD2‑catalysed analysis can help to identify complex partners and their methylation of Lys36 of histone H3, methylated, full- associated stoichiometry38. length nucleosomes were required as substrates to provide fidelity of the biological target in a panel of kinetic Enzyme substrates isotope effect studies49. This reiterates an important The biochemical nature, or identity, of enzyme sub- observation that while many enzymes can be observed strates is another important consideration. There to catalyse low-level activity using surrogate sub- are numerous examples where binding interactions strates, the physiological relevance of this needs to be between macromolecular substrates and enzymes verified50. These examples illustrate the importance of occur at sites distal to the active site and indeed can understanding the most physiologically relevant form contribute to the overall binding energy for the for- of the enzyme to study and, indeed, target. Moreover, mation of the initial enzyme–substrate (ES) encounter if it is not always feasible to perform high-throughput complex39. Substitution of the native substrate with, for hit-finding campaigns with full-length protein and example, in the case of a protease or kinase, a shorter native substrates, it is crucial to explore the level of peptide could affect both the kinetic mechanism of compromise imbued by the use of simpler systems the enzyme and the ability to identify inhibitors that and define a strategy to relate any activities back to the might prevent formation of the ES complex by binding cellular context. to distal sites, particularly if specific conformational A clear advantage of having detailed knowledge changes accompany binding of the native substrate. For of the molecular environment of the enzyme in the example, specific inhibitor pockets are revealed upon disease-relevant setting is the ability to prioritize spe- substrate binding in bacterial Glu-tRNAGln amido cific mechanisms of interest to target. For example, it transferases40. Among protein kinases are numer- may be preferable to identify inhibitors that are not ous examples in which the kinetic mechanism and competitive with the native substrate where the cellular small-molecule affinity are directly influenced by the concentration of the latter is high; an emerging focus in choice of substrate. When MAPK p38α catalyses phos- the field of kinase drug discovery is on the identification phoryl transfer to the cAMP-dependent transcription of inhibitors that are either non-competitive inhibitors or factor ATF2, the data support a rapid-equilibrium, uncompetitive inhibitors with respect to ATP, in order to random-order mechanism of substrate addition41 (or circumvent the challenge of high cellular concentrations ordered with ATF2 protein binding first42). However, of ATP51. While rare, enzyme inhibitors that operate by in the presence of a short peptide as a phospho uncompetitive inhibition are ideal for disrupting enzyme acceptor, the reaction proceeds through an ordered targets within metabolic pathways because the resulting Random-order mechanism mechanism with MgATP binding first43. A similar sce- accumulation of the substrate cannot thwart inhibition A reaction mechanism in which 52 either of the two substrates nario is recognized for 3-phophoinositide-dependent as it is not competitive . may bind to the enzyme first to protein kinase 1 (PDK1), in which phosphorylation An example of an uncompetitive inhibitor is the nat- form a binary complex, of an extended peptide substrate containing a single ural product mycophenolic acid, which targets inosine followed by the other to form distal recognition element reacts through a rapid- 5ʹ‑monophosphate dehydrogenase (IMPDH). IMPDH a ternary complex. equilibrium, random-order mechanism44. However, catalyses hydride transfer from C2 of IMP to the co‑ + Non-competitive inhibitors phosphorylation of the native downstream substrate substrate NAD , and XMP is produced by addition of Inhibitors that bind with equal S6K1 protein kinase occurs through a steady-state water to IMP. The rate-limiting step in the mechanism affinity both before and after ordered mechanism in which binding of S6K1 pre- of mammalian IMPDH is the hydrolysis of a covalent the varied substrate. cludes association of MgATP45. Detailed understand- thioimidate complex formed between the nascent XMP Uncompetitive inhibitors ing of the catalytic mechanism and full consideration product and an active-site cysteine. Mycophenolate Inhibitors that bind only after of the various enzyme–substrate interactions can binds to this IMPDH–IMP covalent complex to exploit the substrate. afford control over the specific downstream pathways this slow step in catalysis53.

REVIEWS

Enzyme kinetics the majority of the enzyme species sampled in the assay Is an understanding of the rate-limiting step of a exists in the ES form) or to target the rate-limiting step of target enzyme before the launch of a drug discov- catalysis. Assays can be balanced to ensure that a holistic ery programme essential? The rates of the majority breadth of modes of inhibition is captured (for exam- of enzyme-catalysed reactions are limited by either ple, by ensuring that the population of free enzyme E is chemical or product release steps. In the latter case, equivalent to the population of the ES complex under an enzyme–product complex will be the predominant conditions in which the concentration of substrate 59 enzyme form under steady-state conditions, especially equates to the Michaelis constant (Km)) . They can also at high concentrations of substrate. One could envision discriminate against unwanted mechanisms (for exam- a compound screening strategy wherein an enzyme– ple, bias away from competitive inhibition by ensuring an product complex or a mimic thereof is enriched in the excess of substrate is present in the reaction). This abil- assay conditions. In this way, HTS would survey an ity to control the transitory species of enzymes can exist enzyme species with a partially blocked active site, such only as a consequence of a detailed understanding of that a ‘hit’ compound, upon binding to a discrete subsite, the kinetic mechanism in relation to the physiological would inhibit the enzyme by preventing desorption of substrate conditions. This is particularly important for the product ligand. Such an inhibitor would generally multisubstrate enzyme reactions, where a comprehen- effect uncompetitive inhibition of the target enzyme. sive dissection of the mechanism and understanding Importantly, screening an enzyme–product complex of the relative proportions of enzyme species existing at samples a different inhibitor space than the free enzyme, any point during the reaction require determination of and the formation of an inhibited ternary complex may the kinetic mechanism, order of substrate addition and provide higher selectivity than simply binding to the free product release and apparent substrate cooperativity; enzyme. A single plate in such an HTS campaign could this is achieved through a combination of product and be screened twice, once with no added product analogue, dead-end inhibition studies, kinetic isotope effects and once with such an analogue. The first plate would and, where relevant, identification of covalent reaction reveal competitive or mixed-type inhibitors, while the intermediates. A detailed description is beyond the scope latter would preferentially uncover uncompetitive inhib- of this review; however, interested readers are directed to itors. For example, the rate-limiting step of the bacterial Segal60 for further description and to Schneck et al.61 for a enoyl reductases encoded by the FABI and INHA genes comprehensive analysis of cathepsin C using such studies. is the release of the product NAD+ (REF. 54). If one were A sophisticated example showing detailed analysis of four to conduct HTS in the presence and absence of added enzymes in the shikimate pathway (Aro B, D, E and K) NAD+ or its analogue, such as acetyl-pyridine adenine of Streptococcus pneumoniae illustrates how such stud- dinucleotide, hit compounds that bind more tightly to ies enable balanced assay conditions to be generated62. the enzyme–nicotinamide complex rather than the free Similarly, elegant kinetic studies of focal adhesion kinase enzyme might provide a more attractive series of hit 1 (FAK1)63 demonstrated that drug discovery efforts for compounds to lead optimization. this or kinetically similar protein kinases could focus on Consideration of the rates of binding and desorp- developing inhibitors that trap the enzyme in the ‘locked-

tion, kon and koff, can also be key in designing inhibitors down’ ATP-bound or ADP-bound closed form. Clearly, that exhibit kinetic selectivity in which the drug–target an understanding of the molecular enzymology of our complex has a longer half-life than off-target–drug drug discovery targets enables fine tuning of the assays in complexes. Slow-binding inhibition kinetics are a feature an intelligent and sophisticated manner to enable optimi- of many marketed drugs (as summarized in REF. 55), zation of mechanistically novel chemical equity that will and indeed, the concept of residence time has been an translate to be efficacious against physiologically relevant important consideration in the translation of such kinet- enzyme species. ics into cellular and in vivo contexts22,56. A prospective

mechanistic pharmacodynamic model to predict dose– Post-screening: understanding IC50 response curves for inhibitors of the LpxC enzyme from Following the identification of hits, their inhibitory Pseudomonas aeruginosa in an animal model of infection, effect must then be quantified, so that chemistry design encompassing drug–target kinetic parameters, including efforts may be applied to those compounds most likely the on and off rates for the formation and breakdown to have the best chance of progression. Historically, this 57 Slow-binding inhibition of the drug–target complex, has recently been described . has been achieved by measuring the concentration at An inhibition that occurs Irreversible covalent inhibition is also gaining attention, which the compound elicits 50% of its maximal inhib- slowly on the timescale of the where tuning the balance between the initial molecular itory effect (IC50 ). To obtain this value, concentration– assay as the enzyme–inhibitor recognition event and the chemical reactivity can lead response analyses are undertaken, whereby the degree complex concentration increases to its steady-state to the desired target selectivity without compromising of enzymatic inhibition is monitored at increasing con- 19,58 level. safety and concern for off-target toxicity . centrations of the inhibitory test compound. Although these types of experiments are fairly straightforward to Competitive inhibition Screening assays perform, it should be remembered that the IC50 itself A type of inhibition where the Screening assays can be built in a number of ways. may not be a true reflection of inhibitor affinity, as the inhibitor binds only before the varied substrate (see also As mentioned above, they can be devised to identify measured value may be significantly affected by experi- non-competitive and inhibitors with a specific mechanism (for example, mental conditions, particularly the concentration of the uncompetitive inhibition). the substrate-uncompetitive mechanism by ensuring substrate to which it is competitive.

REVIEWS

As well as potential differences associated with buffer There are several key steps that can easily be employed components, pH and ionic strength effects, binding often to help to identify PAINS, including assessing the involves protein flexibility64 and changes in conformation steepness of the slope of the inhibition curve, investi- during induced fit, where the binding site adapts to the gating the time dependency and enzyme concentration-

shape of the ligand processes. The value of the IC50 may dependency of the observed inhibition, monitoring the be modified by changes in substrate identity and concen- effect of increasing detergent concentration, measuring tration as well as enzyme concentration. Additionally, in the potential for inhibition in other assays with unrel

traditional analyses of IC50 values, the enzyme, substrates ated enzymes, characterizing the reactivity with reduc- and inhibitor are co‑mingled for a common period of ing agents and monitoring the effects on protein stability exposure. Any effects of time-dependent inhibition are and unfolding. undetectable in this assay format. An enzyme target Often PAINS give characteristic responses in these inside a cell will experience frequent exposure to its sub- types of experiments, such as steep slopes, as occurs by strates but, owing to pharmacokinetics, only temporary rottlerin inhibition of β‑lactamase76, and time-dependent exposure to an added inhibitor. Under these conditions, and/or enzyme-dependent inhibition. Several promis-

the IC50 values of the primary enzyme target and its cuous kinase inhibitors were demonstrated to have off-target homologues in separate assays receive similar time-dependent inhibitory effects on a number of unre- temporal exposure to a panel of inhibitors, regardless of lated enzymes77. Often, the inhibition may be reduced by whether these targets are all colocalized in a germane cell. adding detergent, and sometimes, these compounds will A more useful measure of enzyme inhibition of hit or lead be active at the same affinity range as unrelated enzymes. compounds may involve pre-incubation of the enzyme The effect on protein stability can be investigated using a and inhibitor, followed by its dilution, before assessment modification to Selwyn’s test78 or by employing biophysical of residual activity by added substrates. Such an approach methods. would unveil tight-binding or time-dependent inhibitors Of course, it is essential to position these assays soon and possibly demonstrate that while time-dependent after initial hits have been identified, so that compounds inhibition is operative for a primary target, this does potentially displaying these unwanted mechanisms can not occur with secondary targets. The effects of some be assessed, annotated and prioritized accordingly.

commonly observed situations in determining IC50 are It should be noted that this does not mean that com- shown in FIG. 2. pounds possessing mechanisms such as those listed above are automatically ruled out from further opti- Mechanism of action mization. Sometimes the promiscuous activity occurs Spurious mechanisms. There has been a growing appre- only at concentrations much greater than those that are ciation of the requirement to identify and remove com- required for true, specific, binding-based inhibition, pounds showing activity in primary screens via undesired and in this case, compounds can move forward in the mechanisms from primary screening hit lists65. Perhaps drug-discovery process. The important factor is that an unsurprisingly, there are many specific mechanisms understanding of the potential liabilities is generated by which so-called nonspecific inhibition, sometimes early, so that an informed decision-making process can known as promiscuous inhibition, can occur. These take place. include common technology hitters50, impurities (such as heavy metals)66, precipitators67, aggregators68, redox Slow-binding and tight-binding enzyme inhibition. active compounds69, intrinsically reactive compounds70 Enzyme inhibitors can act either reversibly or irre- and compounds that bind preferentially to unfolded pro- versibly, and the basic principles are well-described teins. Promiscuous compounds effective against protein in a number of classical enzymology textbooks79. The kinase C (PKC) and MKK1 by virtue of an aggregation time-dependent properties of enzyme inhibitors consti- affect, which effectively reduces the concentration of tute a continuum, wherein some compounds that cova- available active enzyme, have been identified71 (TABLE 1). lently inactivate an enzyme actually demonstrate more Collectively, these compounds have been termed reversibility than a noncovalent inhibitor that achieves a pan-assay interference compounds (PAINS)72 and their remarkably tight binary complex. These modes of inhi-

presence may result in misleading IC50 values. This bition show very different behaviour both in vitro and is because the inhibitory activity is not related to the in vivo80. However, reversibility is not an absolute phe- stoichiometric inhibition of the protein but results from nomenon, since inhibition considered to be irreversible other effects that may lead to inactivation, denaturation may actually be reversible over a period of time that is or removal of active protein from its usual catalytic or much greater than the usual assay time. This leads to an binding function73,74. Several computational approaches75 operational definition that classifies inhibition as irre- have been designed to identify such compounds, and versible if the loss of enzyme activity caused by an inhib- Reversible inhibition suggestions to remove them from screening sets have itor is not restorable over the timescale of the enzyme An inhibition that can be been made. However, the activity of such compounds is activity assay. reversed on the timescale of often dependent on both protein and assay composition, For reversible inhibition, full inhibition is usually the assay by competition or and so these approaches may not be accurate or required. obtained rapidly because there is no chemical reaction dilution. Reversible inhibitors are not precluded from This means that enzyme kinetics-based methods applied involved, only a simple noncovalent interaction. It forming covalent bonds with for a particular protein under relevant assay conditions should be noted, however, that irreversible inhibitors the enzyme. are the preferred route for annotating such compounds. are not required to form covalent bonds and that there

REVIEWS

a b 12 12

10 10

8 8

50 6 50 6 IC IC 4 4

2 2

0 0 10 × Km Km Km/10 Ki 10 × Km Km Km/10 Ki [S] [S]

[S] Km IC50 = Kiʹ = Ki 1 + IC50 = Kiʹ = Ki 1 + Km [S] c d 1.2 8

1.0 7 6 0.8 5 50 50 0.6 IC 4 IC 3 0.4 2 0.2 1 0 0 10 × Km Km Km/10 Ki 10 × Ki Ki Ki/10 Ki [S] [S]

[E]t [S] [E]t IC50 = Kiʹ = Ki IC50 = Kiʹ + = Ki 1 + + 2 Km 2 e f 18 2.5 16 14 2 12

50 1.5

50 10 IC

IC 8 1 6 4 0.5 2 0 0 t = 1200 s t = 600 s t = 300 s t = 5 s K 10 × Km Km Km/10 Ki i [S] [S]

– k t [E]t [S] [E]t [S] 2 – 2e ηIC50 inact IC50 = Kiʹ + = Ki 1 + + IC50(t) = Ki 1 + –1 k t 2 Km 2 Km ηIC50 inact where

IC50(t) ηIC50 = [S] Ki 1 + + IC50(t) Km

Figure 2 | Translating IC50 into reality. Graphical representation of the effects of changes in assay conditions (substrate concentration [S], enzyme concentration [E] or time t) on the magnitude of the measuredNature half-maximal Reviews inhibitory | Drug Discovery

concentration (IC50) for selected types of inhibition. Panels a, b and c show the expected response for standard competitive, uncompetitive and non-competitive inhibition. Panels d and e display the consequences of tight-binding

inhibition for the competitive model. Panel f shows the effect of time on the apparent IC50 for an irreversible inhibitor. In all

instances, the inhibitor constant Ki = 1 nM. The effect of changing substrate concentration is shown for concentrations at

the Michaelis constant (Km) and 10‑fold above and below Km. Where enzyme concentration is varied, it is shown at Ki and

10‑fold above and below Ki. For the time-dependent effects, time is varied between 5 and 1,200 s.

REVIEWS

may be covalent bond formation between a reversible also be seen that if [E] is very low, possibly because the

inhibitor and the enzyme. The distinction between majority of [E]t is actually in the form ES due to the pres- reversible and irreversible inhibitors is made only ence of a very highly competing substrate, slow binding on kinetic grounds. Compounds that fall in the con may result. tinuum of reversible and irreversible inhibition are known Slow-binding inhibition can also arise from the as slow-binding inhibitors57, and slow-binding inhibi- rapid formation of an initial collision complex (E·I), tion, which usually accompanies tight-binding inhibition, which subsequently undergoes slow isomerization to a is a widely occurring phenomenon during drug discov- higher affinity complex (E·I*). A scheme for this type of ery13. This nomenclature may, however, be misleading, slow-binding inhibition, often termed mechanism B, is as many of these compounds associate rapidly with the shown in FIG. 3 (mechanism B). enzyme and then slowly isomerize into a higher affinity It has been suggested that the majority of slow-binding form — for example, this is the mechanism responsi- and slow, tight-binding inhibitions occur according to ble for the isoniazid–NAD adduct inhibition of INHA81 mechanism B85. It is envisaged that inhibitors initially and is also observed for some metal binding inhibitors combine with enzymes at their active sites and subse- of HIV‑1 integrase82. Dissociation of the inhibitor is quently induce conformational changes that cause an also slow. Hence, for a slow-binding inhibitor, on rates increase in affinity and the formation of a more stable EI can be fast or slow, but off rates are always slow. Slow complex, from which inhibitor is slowly released.

binding is operationally defined as an increase in inhi- Hence, for inhibition to conform to mechanism B, Ki*

bition during the time of the assay. Irreversible inhibi- must be lower than Ki, and therefore, k6 must be lower

tion can be considered a special type of slow binding than k5. In addition, the values of k5 and k6 must be of a in which recovery of the enzyme activity is too slow magnitude that allows observation of the attainment of to detect. equilibrium between E·I and E·I*. By following the progress curves for inhibition, in Slow-binding inhibition. Because the degree of inhibi- reactions starting with enzyme, it is possible to dis- tion for a slow-binding inhibitor increases with time, the tinguish between these two different mechanisms of

IC50 also changes with time, and so an IC50 value obtained slow-binding inhibition using nonlinear regression

at a single time point may not be a reliable measure of analysis aided by secondary plots of kobs versus [I], which inhibitory potency. are linear for the one-step mechanism and hyperbolic for It is crucial to recognize the importance of analysing the two-step mechanism, and to estimate the individual time courses for inhibitors suspected to be slow binding rate constants. or irreversible, rather than using single time points for Another useful approach is to carry out jump- dose–response analysis, as failure to do this could result dilution experiments, where the pre-equilibrated E·I in the misannotation of potency and misleading infor- complex is diluted rapidly into the assay and recovery of mation as to the MoA. Instead, it is necessary to charac- enzyme activity is monitored as the inhibitor dissociates terize the rate constants of the steps involved in forming from the complex86. These experiments are aided by the the enzyme–inhibitor (EI) complexes. An example of use of high substrate concentrations, which may serve such characterization is that of extracellular signal- to compete with the inhibitor for rebinding to the free regulated kinase 1 (ERK1; also known as MAPK3) enzyme and to increase the detectable signal. and ERK2, in which both slow association and dissociation of Vertex‑11e and SCH772984 were observed, Irreversible inhibition −1 −1 An inhibition that cannot be with apparent kon values of 0.21 and 2.8 μM s , a Mechanism A b Mechanism B reversed on the timescale of respectively, and dissociation rate constants of 0.21 and k7 k7 the assay. Truly irreversible 1.1 h−1, respectively83 (TABLE 1). E·S Products E·S Products k [S] k [S] inhibitors never dissociate There are two important, distinct mechanisms that 1 1 from the enzyme and are k k account for the majority of slow-binding interactions84. 2 2 characterized by an E E The first occurs because, according to the law of mass inactivation rate constant, k3I k3[I] not a dissociation constant. action, reactions at low concentrations will proceed k k4 k4 5 Irreversible inhibitors do not more slowly. When an inhibitor has a low inhibition E·I E·I E·I* necessarily have to be k constant (K ) and the inhibitor concentration [I] is var- 6 covalently bound. i Slow ied in the region of Ki, the values of the pseudo-first- Tight-binding inhibition order rate constant for the formation of the non-covalent Figure 3 | Mechanisms of slow-binding inhibition. A type of inhibition occurring Two potential mechanisms have been proposed to enzyme–inhibitor (E·I) complex (k3 × [I]t) and the dis- Nature Reviews | Drug Discovery under conditions when the describe the observation of slow-binding enzyme sociation rate constant (k4) would be low. Slow-binding concentration of inhibitor inhibition may also occur because of a low value of k , inhibition. These are shown in parts a and b. Part a required to cause inhibition 3 illustrates the single-step slow-binding mechanism, FIG. 3 is similar to the enzyme and such a mechanism is represented in (mecha- in which the formation of the non-covalent enzyme– concentration, leading to nism A). These low values of association and dissocia- inhibitor (E·I) complex is slow, and part b illustrates the depletion of the free inhibitor tion would lead to slow binding even though k3 may be two-step slow-binding mechanism, in which E I is formed concentration. This results in · of the order expected for a diffusion controlled reaction rapidly but is then slowly converted to the higher affinity breakdown of the usual 8 −1 −1 assumptions leading to simple (>10 M s ). It is the product of the true second-order E·I* complex. E·S, non-covalent enzyme–substrate inhibition kinetics and requires rate constant (k3), [I]t and the enzyme concentration [E] complex; [I], inhibitor concentration; k1–7, rate constant a more complex rate equation. that determines the rate of E·I complex formation. It can of the specific step; [S], substrate concentration.

REVIEWS

Tight-binding inhibition. The drug-discovery process, In practice, it may mean that biophysical and/or kinetic- by its very nature, often results in the identification of based methods need to be used to quantify the dissocia-

high-affinity molecules. Owing to this high affinity, tion constant (Kd), and efforts need to be made to relate

these compounds are often difficult to characterize. In the constant to Ki. order to rank compounds correctly, it is important to

realize that true Ki values must be compared and that Different enzyme forms

IC50 values do not give an accurate representation of Enzyme kinetic assays must, by definition, sample a affinity under certain conditions, for example, situations number of different enzyme species during the catalytic in which ligand depletion occurs87. This situation is often cycle, as substrates are bound, intermediates are formed termed ‘tight binding’, although it really relates to a situ- and products are released during the enzyme-catalysed ation where the active enzyme concentration is approx- reaction, and inhibitors may potentially access a num- imately equal to, or even higher than, the apparent ber of these forms91 (FIG. 1). Additionally, other enzyme inhibition constant. This means that tight binding could forms, such as different post-translational modifications, occur even for weaker compounds at elevated enzyme redox states or partial degradation products may also be concentrations. This phenomenon has been known for present even if the experimenter is unaware of their pres-

many years through the relationship among IC50, Kiʹ (the ence (although efforts should be made to identify such 88 apparent Ki) and [E]t described by Goldstein in 1944 additional enzyme forms). These potential additional (equation 1): forms may have varying catalytic efficiencies and sensi- tivities to the test compound and thus may contribute to

IC50 = Kiʹ + [E]t/2 (1) the measured IC50 in undefined ways. It is therefore pru- dent to undertake rigorous quality control assessments

where Kiʹ = Ki/(1 + [S]/Km). It is clear from equation 1 as detailed in order to understand the contribution of Nature Reviews | Drug Discovery that the measured IC50 is limited to half of the func- different enzyme forms to the total enzyme concentra-

tional enzyme concentration as the Kiʹ value becomes tion and the total catalytic rate, so that the magnitude of

very small and, conversely, that the IC50 approaches the this issue can be estimated. It is also important to repeat

apparent Ki value only when the functional enzyme these assessments on new batches of protein to ensure

concentration is small relative to the apparent Ki value. that the degree of any changes in the amount of modified

The consequence for drug discovery is profound, IC50 enzyme forms is known. cannot be used to rank compounds when the affinity approaches the enzyme concentration. When this situa- Irreversible covalent compounds tion occurs, additional approaches must be taken to fully Irreversible, covalent compounds19,92 are, by definition, understand the true affinity. unable to be ranked in terms of effectiveness using an

Although there are data-fitting approaches that equilibrium dissociation constant, Kd, or Ki. This means

do not assume that the free concentration of inhibi- that IC50 values are also not useful in ranking these types tor equals the total concentration, such as the quad- of compounds. This should be obvious, as any irrevers- ratic equation described by Morrison89 in equation 2, ible inhibitor added at half the enzyme concentration even these methods become unreliable as the affinity will, given long enough for all of the added inhibitor to increases beyond around 10‑fold below the functional covalently modify the enzyme, lead to a 50% decrease in enzyme concentration. the control rate. This is the extreme case in equation 1,

where Kiʹ = 0 and so IC50 = [E]t/2. However, it is still possi- 2 –([I]t–[E]t + Kiʹ) + ([I]t–[E]t+Kiʹ) –4Kiʹ[E]t ble to carry out concentration responses with irreversible vi = v0 (2) 2[E]t compounds under fixed conditions of substrate concentration and time, although this is not recommended to where vi is the inhibited rate,Nature v0 is Reviews the uninhibited | Drug Discovery rate and assess irreversible inhibitor effectiveness. Instead, this [I]t is the total inhibitor concentration. When [I]t = IC50, should be assessed using the kinetic parameter kinact/Ki,

vi = 0.5v0, and equation 2 can be rearranged to provide which is ascertained from time-course experiments equation 1. with varying inhibitor concentrations93, where the rate

Experimental approaches to decrease the functional constant, kobs, represents the conversion from the initial

enzyme concentration or raise the apparent Ki by the complex to the irreversible covalent complex. Here, kinact

use of increased concentrations of competing substrate is the rate constant for inactivation, while the Ki term may have limited utility, as assays are often initially describes the concentration of inhibitor required for configured with low concentrations of enzyme to avoid one-half of the maximum rate of covalent bond forma-

this issue; therefore, further decreases can cause issues tion and should not be confused with Ki describing the

with the biosignal sensitivity. Shifting Kiʹ is possible dissociation of the E·I complex, which is not affected

only for compounds that have a competitive compo- by covalent bond formation. For a 2-step reaction, kinact nent to binding, and often limited shifts may be pos- represents the maximum observed rate constant for sible due to limited solubility restricting the achievable inactivation at saturating [I].

concentration:Km ratio. Experimental conditions for Attempts have also been made to derive these param- 94 accurate determination of Ki values for tight-binding eters from time-dependent IC50 values . The advantage

enzyme inhibitors, following an in silico study of exper- of using a kinetic approach to measure kinact/Ki is two- imental error and assay design, has been described90. fold: it uses valid parameters to describe the inhibition,

REVIEWS

and it allows the effectiveness of the irreversible inhibi- assay system102 and to investigate instances of cooper- tion to be dissected into initial molecular recognition ative binding, which may also require more complex and chemical reaction steps under the conditions of equations when the slope does not equal 1 (REF. 103).

mechanism B (Scheme 2), from which kinact/Ki and kinact can be ascertained. The ability to characterize irrevers- Data analysis ible inhibitors in this way is essential to modulate the Concomitant with the issues described above, mislead-

balance between intrinsic compound reactivity, which ing IC50 values may also arise via improper data analy- may lead to a lack of selectivity, and initial noncovalent sis104. Although many software packages are available interactions, which help improve recognition of the tar- and are used routinely to apply nonlinear regression get protein95,96. TABLE 1 also illustrates the dissection of analysis to calculate parameter values, it is often dif- the irreversible inhibition of epidermal growth factor ficult to know whether an appropriate concentra- receptor (EGFR) tyrosine kinase97. Thus, having these tion–response equation has been fitted to the data. key parameters available for lead series during a lead For example, in many companies, four‑parameter optimization campaign allows the medicinal chemist to logistic equations or variations thereof are automat- tailor the requirements of initial affinity and reactivity in ically used to analyse HTS concentration–response order to mitigate concerns over reactive warheads and data, even if the system is adequately described by a potential off-target toxicity. simpler equation or should be described by an alternative, more complex equation. Of course, the equa- Mechanism of inhibition tion used should ideally directly relate to the physical Test compounds may have a number of different mecha- processes occurring in the assay to allow meaningful nisms with respect to the effects of substrates or, indeed, estimates of relevant parameters, such as inhibition other ligands on the degree of inhibition98. Competitive of a specific enzyme species. Additionally, the exper- inhibition is common, as inhibitors are often either imenter should ideally assess and record whether the designed to mimic the substrates of an enzyme reac- equation provides an adequate quality of fit, gener- tion or to bind at the substrate-binding site, frequently ates precise and accurate parameter values and high- resulting in the observation of competitive kinetics. lights any redundant parameters105. Typically, and Other mechanisms, however, are also possible, includ- certainly historically, this is not possible during the ing non-competitive kinetics, where a test compound concentration–response analysis of several thousand may bind both before and after substrate binding (and compounds, and so the reality is that many company

in the case in which the affinities for the free enzyme databases may contain large numbers of IC50 values and ES complex are different, the inhibition is classified that do not correspond to the true potency of the as mixed inhibition). As mentioned above, uncompetitive test compound. inhibition is rarer and occurs when the inhibitor binds only after the substrate has bound to the enzyme, such Combining enzymology and biophysics as the case in which mycophenolic acid binds to an ino- As we have seen above, mechanistic enzymology is sine 5ʹ‑monophosphate dehydrogenase that is covalently a powerful approach, which when applied carefully, complexed to its nucleotide substrate52,99,100. can allow for extensive characterization of protein– Once inhibitors have been identified during primary modulator activity. screening, it is important to understand how the con- However, combining or supplementing enzyme

centration of substrates affects the measured IC50 values, kinetic studies with biophysical methods can add value as this allows the mechanism of inhibition to be deter- at many stages in the drug discovery process (TABLE 2). mined and provides a basis for extrapolating the activity Since the mid‑1990s, biophysical methods have been seen in isolated protein assays to effects observed in cells, increasingly used in this way, and there are now many where the concentration of substrates may be different to methods106 that can be routinely applied in a num- the biochemical enzyme assay. It also allows for effective ber of settings to enhance the understanding of the assessment of selectivity, as the magnitude and direction molecular MoA of test compounds. of inhibitor selectivity can vary according to the mech- This combination extends through initial reagent anism of inhibition and the substrate concentration at characterization (including protein and tool compound which selectivity is assessed. These observations are well evaluations), assay development and choice of HTS described in a number of classical enzymology texts101. conditions to more detailed MoA studies during lead Initial experiments to determine the mechanism are optimization (TABLE 2). often focused on two substrate concentrations, above Combining these highly informative studies early

and below the Km of the varied substrate. While this may in each of these phases has the advantage of increased give an indication of the mechanism of inhibition, it is impact beyond that delivered by each alone and cru- recommended that full matrix-based experiments cov- cially allows for cost savings to be realized due to ering a range of substrate and inhibitor concentrations unsuitable reagents, incorrect assumptions regarding are undertaken, rather than this sparse data approach, so tool molecules and inappropriate assay configura- that reliable K values can be estimated. tion to detect desired MoAs. A combination of bio Mixed inhibition i A type of inhibition in which Of course, it is also important to use a relevant mod- physical methods to interrogate binding kinetics and the inhibitor binds both before ified form of the Cheng–Prusoff equation to ensure that protein integrity, coupled with detailed dissection and after the varied substrate. the estimated parameter values reflect the nature of the of the enzyme catalytic mechanism and effects of

REVIEWS

Table 2 | Examples of the value of combining enzyme kinetic studies with biophysical methods at various stages of the drug-discovery process Drug-discovery Goal of enzyme Parameters Additional information Examples Refs stage assay measured from biophysics

Assay Understand the Apparent and true Identification of kon, Assessment of recombinant protein 125,126 development target biochemistry Km and kcat values koff and Kd values for quality using thermal unfolding methods; and influence screen for substrates and ligands (tool compounds, identification of a ligand and subsequent design by evaluation order of substrate substrates and generation of fluorescent probes for assay of kinetic mechanism addition and cofactors) for different development targeting HI‑0033 product release isolated enzyme forms; measurement of binding affinity to inactive enzyme forms

Primary screening Identify compounds IC50 for inhibitory Orthogonal hit evaluation Several examples from Abbott on the 127–130 with the desired effects to identify true binding use of NMR after HTS; NMR-based biochemical compounds, primary counterscreen used to characterize mechanism of action screening of weak xanthine oxidase inhibitors; overview compounds (fragments) of biophysical methods for fragment screening

Hit to lead Initial Change in IC50 Effect of different ligands Overview of biophysical methods applied 131–133 characterization with different on measured Kd, binding in the hit‑to‑lead phase; overview of and prioritization of conditions thermodynamics and combinations of enzyme assays with different chemical (including [S] and kinetics biophysics to select the most promising series time) hits and leads; use of surface plasmon resonance to improve lead optimization Mechanistic Detailed Mechanism of As above, orthogonal Mechanism of inhibition, structural 119,

characterization understanding of inhibition, Ki, assessment of parameters studies and binding kinetic 134,135 during lead the kinetics and rate constants describing the same characterization of (R)-PFI‑2 inhibition optimization thermodynamics for reversible physical process and of SETD7; biochemical, biophysical and

of binding and inhibition and kinact potentially including cellular studies on inhibition of IDH1; mechanism of for irreversible low-resolution or characterization of allosteric inhibition of inhibition inhibition high-resolution binding spleen tyrosine kinase site information

HTS, high-throughput screening; IC50, half-maximal inhibitory concentration; IDH1, isocitrate dehydrogenase 1; kcat, catalytic rate constant for the conversion of substrate to product; Kd, dissociation constant; Ki, inhibition constant; kinact, rate constant for inactivation; Km, Michaelis constant; koff, rate constant for desorption; kon, rate constant for binding; [S], substrate concentration; SETD7, SET domain-containing protein 7.

intervention by small-molecule modulators, provides a gel filtration, as well as more biophysical approaches comprehensive level of detail at the molecular level. This such as dynamic light scattering and analytical ultra- interplay is impactful at all stages of target (+/− ligand) centrifugation107. Usually, the criteria to confirm protein characterization. purity is the observation of a single band or peak, indi- cating that the protein is homogenous, monodisperse Combining catalytic and binding assays and in a single, defined oligomeric state. Coupling this Reagent evaluation. Before target proteins are used information to enzyme kinetic data such as active site for assays, there are several important quality control titration is critical, as the observed catalytic activity may measures that should be conducted. The identity, purity, arise from a small proportion of very active enzyme or concentration, functionality and stability of the protein a larger proportion of less active protein, and being able must be confirmed. Biophysical methods may be used to to distinguish between these differences is required to impact all of these areas, but consideration alongside the ensure that the protein is fully competent to bind ligands. enzyme kinetic data provides crucial context-dependent Concomitant to the measurements of catalytic rate, information. it is also important to measure the concentration of the Amino-acid analysis (AAA), liquid chromatography enzyme, so that the activity can be carefully quantified coupled to mass spectrometry (LC–MS) and peptide and related to previous batches or similar enzymes. mapping to understand post-translational modifications Typically, ultraviolet–visible spectrophotometry and/or (PTMs) can be used to ensure that the correct sequence is Lowry, Bradford or bicinchoninic acid (BCA) assays108 present, the expected mass is observed and that any rele- are used to estimate protein concentrations, although vant PTMs are characterized. Of course, coupling this to one of the best, if somewhat laborious, methods is the expected catalytic activity, in terms of converting the quantitative AAA109. particular substrate(s) to the corresponding product(s), Although measuring total protein concentration is confirms that the protein that was expressed and purified important, it must be understood that not all of the pro- is the correct one and is suitable for further study. tein in the sample may be functional, in terms of binding Purity is often a key consideration for biophysics to ligands or carrying out catalysis. Particularly useful in (and structural) work, but is often, quite wrongly, under- this respect may be biophysical binding methods, such appreciated in enzyme kinetics studies. Methods to as isothermal titration calorimetry and surface plas- assess purity include SDS or native PAGE and analytical mon resonance (SPR)110. These methods allow reference

REVIEWS

ligands to be characterized in terms of affinity and other Hit evaluation. The purpose of the hit-evaluation pro- parameters, for example, enthalpy of binding, as well as cess is to identify those hits that elicit their activity via in certain circumstances, the stoichiometry of the bind- a valid mechanism and to remove those compounds for ing interaction. Assuming that the concentration of the which their primary screening activity is undesirable. ligand is accurately known, this can allow an estimation Combining an enzyme assay-based primary screen with of the binding-competent component of the total protein a biophysical-based follow-up assay has some advantages. concentration, which can be compared with functional For example, substrate turnover may be monitored using measurements in catalytic assays. Comparison of param- mass spectrometry or NMR-based methods to mitigate

eters including Kd, ΔH, stoichiometry, Km, kcat, kcat/Km and artefacts encountered within optical-based biochemical

Ki allows the evaluation of the functional protein in differ- assays. Another advantage of biophysical methods is ent preparations or across different proteins that may be that they allow the direct verification of binding to the important to assay during a drug discovery programme. target protein, allowing problems such as binding to Other components in addition to the protein reagent additional components of the enzyme assay to be ruled can be evaluated in both biochemical and biophysical out. For example, compounds may bind to other pro- approaches before assay development. It is important to tein components, such as capturing antibodies, coupling understand whether tool compounds highlighted in the enzymes, artificial protein substrates or even fluorescent literature behave as expected before using them as stand- peptides116,117. The use of biophysical techniques to probe ard reference compounds in hit-finding campaigns. This direct target engagement under a range of assay condi- is a crucial early activity, especially given that it is often tions can help avoid wasting time following spurious difficult to reproduce published data111–113, and one that mechanisms. often takes relatively little effort but potentially results Biophysical methods also can, at this early stage, pro- in high gain, reducing time spent on spurious tools or vide additional information on the mode of action of the unwanted mechanisms in later phases. initial hits. For example, information on the binding specificity (to related or relevant off-target proteins) as well Assay development. HTS assays are usually developed to as the binding stoichiometry can be acquired relatively search for chemical equity having the desired inhibitory straightforwardly, and additional quantitative data on effect via an appropriate mechanism. Combining bio- binding kinetics and thermodynamics are generated to physical measurements with enzyme kinetic data allows afford an understanding of potential binding mechanisms. balanced or discriminatory assays to be developed, whereby a range of different mechanisms or a particular Orthogonal approaches to lead generation individual mechanism can be identified. This requires a The goals of lead generation are to choose a suitable start- thorough understanding of the mechanism of catalysis, ing point for the chemistry, identify tools that will allow which defines the order of addition of substrates and the investigation of the biological hypothesis and use these release of products. Biophysical methods allow the char- tools to uncover issues with the identified chemical series. acterization of binding affinities of substrates and prod- Subsequently, efforts are made to resolve those problems ucts to a range of different enzyme forms (usually as long in order to develop one or two chemical series that will as a catalytically competent species is not formed), which progress further. Thus, an understanding of SARs and the can often facilitate more detailed multivariate enzyme potential for further optimization of potency and selec- kinetic studies. This information can, for instance, be tivity that will drive compound design are key desirables used to adjust assay concentrations of substrates to bias during lead generation. Mechanistic enzymology has towards a particular type of inhibition mechanism (for had a role in delivering this information for many years,

example, high concentrations of substrate bias away and measurements of IC50 are still prominent in this pro-

from weak substrate-competitive inhibitors). cess today. However, reliance on IC50 is problematic, as Biophysical methods, such as SPR or NMR, are also the values may change with the assay conditions118, and employed at AstraZeneca as an orthogonal approach understanding the impact of changes in assay condi- to evaluate hits found during pre-screening of a small tions on the measured binding affinity of the test com- number of compounds designed, as much as possible, pound allows for better predictions of cellular potency to represent the diversity of the full screening deck114. and in vivo effects. This becomes increasingly important The aim of this pre-screening activity is to give an initial when novel targets, novel mechanisms or new binding assessment of the likely hit rate, assess factors includ- sites are the focus of the drug discovery programme, and ing false positive and negative rates, identify potential understanding the cause of the differences between the

technology interference, evaluate reproducibility and Kd or Ki values and IC50 is perhaps more important than gauge the likelihood of other assay artefacts, such as plate for well-characterized or established targets. pattern effects. An example of this is in the characterization of com- The number of hits expected in this process is usually pounds binding to allosteric sites119. A combination of low enough for biophysical methods to evaluate them biophysical methods and enzyme kinetic and structural all for binding to the target protein. The focus of this studies allowed characterization of inhibitors of spleen activity is to inform on the possible reasons for generat- tyrosine kinase (SYK), which were hypothesized to ing false positives, leading to improvements in the assay function by reinforcing the natural regulatory contacts methodology or the screening cascade and sometimes between the SH2 and kinase domains. Understanding the discontinuation of planned screens115. this novel mode of inhibition provides a potentially new

REVIEWS

opportunity for the design of compounds with improved It is not surprising that modern drug discovery has selectivity profiles. In this study, following purity and also moved forward substantially over the same time concentration checks as described above, fluorescence scale123, as the incisive hypotheses, concepts and metho binding studies were combined with analytical ultra dical rigour of the enzymologist have been applied to centrifugation (AUC) and small-angle X‑ray scattering, scientific problems faced within drug discovery. As as well as detailed enzyme kinetic experiments in which with many time-honoured disciplines, sometimes the the mechanism of inhibition was assessed for compounds value of such stringent studies can be overlooked as inhibiting SYK activation by immunoreceptor tyrosine- the sequential processes and timelines that industrial based activation motif (ITAM) binding, LynB phos drug discovery often demands begin to dominate. Of phorylation or autophosphorylation. The compound X1 course, the development of new medicines is a complex was shown to inhibit SYK activation by all three activa- process, and application of the principles of detailed tion mechanisms: autophosphorylation, LynB phospho- mechanistic enzymology will not solve all of the issues, rylation and ITAM binding. The compound showed but increased focus on scientific rigour at the early stages non-competitive inhibition against ATP and peptide will provide a solid foundation for increased success. substrates, thus preventing autophosphorylation by trap- The concept of focusing on science was exempli- ping SYK in its inactive basal state. The observation of fied recently when AstraZeneca published a framework competitive inhibition with respect to LynB activation founded on project quality and depth of understanding as was postulated to be due to a direct overlap of binding a key driver of success124. This ‘5 Rs’ framework (the right sites on SYK, although a more complicated kinetic model target, the right tissue, the right safety, the right patient and could not be ruled out. Another study on the regulation the right commercial potential) can be modified to encom- of tyrosine kinase activity identified an allosteric net- pass the influence of mechanistic enzymology in the early work of dynamically coupled residues in proto-oncogene stages of drug discovery: right target + right reagents + right tyrosine protein kinase SRC that connect regulatory sites assay + right mechanism = right compound. to the ATP and protein substrate-binding sites120. It was A detailed understanding of target biochemistry the combination of data for substrate analogues and before hit identification allows the selection of more product binding at different peptide concentrations from physiologically relevant reagents and assay formats and fluorescence anisotropy and isothermal titration calorim- thus supports the identification of desired mechanisms etry (ITC) experiments, along with the observation of by appropriate screening (allowing a choice among negative cooperativity for peptide binding in the pres- balanced, focused and discriminatory assays). It also ence of ATP from enzyme kinetics studies, coupled with underpins conclusions derived from detailed follow-up molecular dynamics simulations based on the kinase characterization once suitable compounds have been domain from an X‑ray structure of the SRC system in identified or designed. Additionally, it enhances the an active conformation, which allowed this signal-relay quality of systems biology modelling, enabling more mechanism to be characterized. detailed pathway analysis and quantitative interrogation of required compound characteristics. Conclusion Detailed understanding of the molecular interactions Molecular enzymology has come a long way since its foun- of test compounds enables the selection of truly differenti- dations over a century ago121,122. Access to purified target ated series. Early prioritization of these studies thus facili- enzymes in large quantities, utilizing recombinant expres- tates the identification of unwanted mechanisms, allowing sion technology, has permitted detailed characterization of specific deprioritization of compounds that would other mechanistic details in the absence and presence of inhib- wise delay development due to issues such as lack of true itors. The combination of these mechanistic studies with target engagement or action via mechanisms that are the increasing access to high-resolution individual protein not suitable for progression. This mechanism-based, and complex structures, along with biophysical methods informed decision-making process ultimately leads to that can help probe particular enzyme forms, has had a better-quality compounds with a greater chance of clinical broad impact on the understanding of disease processes success, enhances the quality of regulatory submissions and their modulation. and produces higher-quality publications.

1. Imming, P., Sinning, C. & Meyer, A. Drugs, their 6. Patchett, A. A. et al. A new class of angiotensin- 11. Vacca, J. et al. L-735,524: an orally bioavailable targets and the nature and number of drug targets. converting enzyme inhibitors. Nature 288, 280–283 human immunodeficiency virus type 1 protease Nat. Rev. Drug Discov. 5, 821–834 (2006). (1980). inhibitor. Proc. Natl Acad. Sci. USA 91, 4096–4100 2. Hopkins, A. L. & Groom, C. R. The druggable genome. 7. Roth, B. D. The discovery and development of (1994). Nat. Rev. Drug Discov. 1, 727–730 (2002). atorvastatin, a potent novel hypolipidemic agent. 12. Meek, T. D. Inhibitors of HIV‑1 protease. J. Enzyme 3. Skeggs, L. T., Kahn, J. R. & Shumway, N. P. The Progress Med. Chem. 40, 1–22 (2002). Inhib. 6, 65–98 (1992). preparation and function of the hypertensin-converting 8. Alberts, A. et al. Mevinolin: a highly potent competitive 13. Robertson, J. G. Mechanistic basis of enzyme- enzyme. J. Exp. Med. 103, 295–299 (1956). inhibitor of hydroxymethylglutaryl-coenzyme A targeted drugs. Biochemistry 44, 5561–5571 4. Ondetti, M. A., Rubin, B. & Cushman, D. W. Design of reductase and a cholesterol-lowering agent. Proc. Natl (2005). specific inhibitors of angiotensin-converting enzyme: Acad. Sci. USA 77, 3957–3961 (1980). 14. Patel, M. P. et al. Kinetic and chemical mechanisms new class of orally active antihypertensive agents. 9. Nakamura, C. E. & Abeles, R. H. Mode of interaction of the fabG-encoded Streptococcus pneumoniae Science 196, 441–444 (1977). of β-hydroxy-β-methylglutaryl coenzyme A reductase β‑ketoacyl-ACP reductase. Biochemistry 44, 5. Cushman, D. W., Cheung, H., Sabo, E. & Ondetti, M. with strong binding inhibitors: compactin and related 16753–16765 (2005). Design of potent competitive inhibitors of compounds. Biochemistry 24, 1364–1376 (1985). 15. Dryer, G. et al. Hydroxyethylene isostere inhibitors of angiotensin-converting enzyme. Carboxyalkanoyl and 10. Roberts, N. A. et al. Rational design of peptide-based HIV‑1 protease structure-activity analysis using mercapto-alkanoyl amino acids. Biochemistry 16, HIV proteinase inhibitors. Science 248, 358–362 enzyme kinetics, X‑ray crystallography, and infected 5484–5491 (1977). (1990). T‑cell assays. Biochemistry 31, 6646–6659 (1992).

REVIEWS

16. Schramm, V. L. Transition states, analogues, and drug 40. Harpel, M. R. et al. Mutagenesis and mechanism- 60. Segal, I. H. Enzyme Kinetics: Behavior and Analysis of development. ACS Chem. Biol. 8, 71–81 (2013). based inhibition of Streptococcus pyogenes Rapid Equilibrium and Steady-State Enzyme Systems 17. Schramm, V. L. Transition states and transition state Glu-tRNAGln amidotransferase implicate a serine- (Wiley, 1975). analogue interactions with enzymes. Accounts Chem. based glutaminase site. Biochemistry 41, This book is an essential text for any academic or Res. 48, 1032–1039 (2015). 6398–6407 (2002). industrial enzymologist, as it provides a 18. Pham, T. V. et al. Mechanism-based inactivator of 41. Szafranska, A. E. & Dalby, K. N. Kinetic mechanism for comprehensive introduction to enzyme kinetics. isocitrate lyases 1 and 2 from Mycobacterium p38 MAP kinase α. A partial rapid-equilibrium 61. Schneck, J. L. et al. Chemical mechanism of a cysteine tuberculosis. Proc. Natl Acad. Sci. USA 114, random-order ternary-complex mechanism for the protease, cathepsin C, as revealed by integration of 7617–7622 (2017). phosphorylation of a protein substrate. FEBS J. 272, both steady-state and pre-steady-state solvent kinetic 19. De Cesco, S., Kurian, J., Dufresne, C., Mittermaier, A. 4631–4645 (2005). isotope effects. Biochemistry 47, 8697–8710 (2008). & Moitessier, N. Covalent inhibitors design and 42. LoGrasso, P. V. et al. Kinetic mechanism for p38 MAP 62. Noble, M. et al. The kinetic model of the shikimate discovery. Eur. J. Med. Chem. 138, 96–114 (2017). kinase. Biochemistry 36, 10422–10427 (1997). pathway as a tool to optimize enzyme assays for high- 20. Radzicka, A. & Wolfenden, R. Transition state and 43. Chen, G., Porter, M. D., Bristol, J. R., Fitzgibbon, M. J. throughput screening. Biotechnol. Bioeng. 95, multisubstrate analog inhibitors. Methods Enzymol. & Pazhanisamy, S. Kinetic mechanism of the p38‑α 560–573 (2006). 249, 284–312 (1995). MAP kinase: phosphoryl transfer to synthetic 63. Schneck, J. L. et al. Kinetic mechanism and rate- 21. Byers, L. D. & Wolfenden, R. Binding of the by‑product peptides. Biochemistry 39, 2079–2087 (2000). limiting steps of focal adhesion kinase‑1. Biochemistry analog benzylsuccinic acid by carboxypeptidase A. 44. Gao, X. & Harris, T. K. Steady-state kinetic 49, 7151–7163 (2010). Biochemistry 12, 2070–2078 (1973). mechanism of PDK1. J. Biol. Chem. 281, 64. Teague, S. J. Implications of protein flexibility for drug 22. Copeland, R. A., Pompliano, D. L. & Meek, T. D. 21670–21681 (2006). discovery. Nat. Rev. Drug Discov. 2, 527–541 (2003). Drug-target residence time and its implications for lead 45. Keshwani, M. M., Gao, X. & Harris, T. K. Mechanism of 65. Thorne, N., Auld, D. S. & Inglese, J. Apparent activity optimization. Nat. Rev. Drug Discov. 5, 730–739 PDK1‑catalyzed Thr‑229 phosphorylation of the S6K1 in high-throughput screening: origins of compound- (2006). protein kinase. J. Biol. Chem. 284, 22611–22624 dependent assay interference. Curr. Opin. Chem. Biol. This seminal paper introduces an alternative (2009). 14, 315–324 (2010). approach to drug optimization, focusing on drug– 46. Davidson, W. et al. Discovery and characterization of a 66. Hermann, J. C. et al. Metal impurities cause false target residence time rather than the optimization substrate selective p38α inhibitor. Biochemistry 43, positives in high-throughput screening campaigns. of thermodynamic affinity alone. 11658–11671 (2004). ACS Med. Chem. Lett. 4, 197–200 (2013). 23. Overington, J. P., Al‑Lazikani, B. & Hopkins, A. L. How 47. Pargellis, C. et al. Inhibition of p38 MAP kinase by 67. Di, L. & Kerns, E. H. Biological assay challenges from many drug targets are there? Nat. Rev. Drug Discov. utilizing a novel allosteric binding site. Nat. Struct. compound solubility: strategies for bioassay 5, 993–996 (2006). Mol. Biol. 9, 268 (2002). optimization. Drug Discov. Today 11, 446–451 24. Smith, G. F. 1‑Medicinal chemistry by the numbers: 48. Li, Y. et al. The target of the NSD family of histone (2006). the physicochemistry, thermodynamics and kinetics of lysine methyltransferases depends on the nature of 68. McGovern, S. L., Helfand, B. T., Feng, B. & modern drug design. Progress Med. Chem. 48, 1–29 the substrate. J. Biol. Chem. 284, 34283–34295 Shoichet, B. K. A specific mechanism of nonspecific (2009). (2009). inhibition. J. Med. Chem. 46, 4265–4272 (2003). 25. Sachsenmaier, C. & Schachtele, C. Integrated technology 49. Poulin, M. B. et al. Transition state for the This paper shows that aggregates formed by platform protein kinases for drug development in NSD2‑catalyzed methylation of histone H3 lysine 36. promiscuous compounds can reversibly sequester oncology. Biotechniques 33, S101–S106 (2002). Proc. Natl Acad. Sci. 113, 1197–1201 (2016). enzymes, resulting in apparent inhibition. It 26. Lindsley, C. W. et al. Allosteric Akt (PKB) inhibitors: 50. Bar-Even, A. et al. The moderately efficient enzyme: presents post-screening approaches for detecting discovery and SAR of isozyme selective inhibitors. evolutionary and physicochemical trends shaping and avoiding such compounds. Bioorg. Med. Chem. Lett. 15, 761–764 (2005). enzyme parameters. Biochemistry 50, 4402–4410 69. Johnston, P. A. Redox cycling compounds generate

27. Barnett, S. F. et al. Identification and characterization (2011). H2O2 in HTS buffers containing strong reducing of pleckstrin-homology-domain-dependent and 51. Garuti, L., Roberti, M. & Bottegoni, G. Non-ATP reagents — real hits or promiscuous artifacts? isoenzyme-specific Akt inhibitors. Biochem. J. 385, competitive protein kinase inhibitors. Curr. Med. Curr. Opin. Chem. Biol. 15, 174–182 (2011). 399–408 (2005). Chem. 17, 2804–2821 (2010). 70. Dahlin, J. L., Baell, J. & Walters, M. A. in Assay 28. Joseph, R. E., Min, L. & Andreotti, A. H. The linker 52. Cornish-Bowden, A. Why is uncompetitive inhibition so Guidance Manual (eds Sittampalam, G. S. et al.) (2004). between SH2 and kinase domains positively regulates rare?: A possible explanation, with implications for 71. McGovern, S. L. & Shoichet, B. K. Kinase inhibitors: catalysis of the Tec family kinases. Biochemistry 46, the design of drugs and pesticides. FEBS Lett. 203, not just for kinases anymore. J. Med. Chem. 46, 5455–5462 (2007). 3–6 (1986). 1478–1483 (2003). 29. Hays, J. L. & Watowich, S. J. Oligomerization-induced This paper provides insight into why uncompetitive 72. Baell, J. B. & Holloway, G. A. New substructure filters modulation of TPR-MET tyrosine kinase activity. inhibition is rarely encountered and suggests that for removal of pan assay interference compounds J. Biol. Chem. 278, 27456–27463 (2003). it is a useful approach, as these inhibitors may (PAINS) from screening libraries and for their exclusion 30. Timofeevski, S. L. et al. Enzymatic characterization of have greater pharmacological effects. in bioassays. J. Med. Chem. 53, 2719–2740 (2010). c‑Met receptor tyrosine kinase oncogenic mutants and 53. Hedstrom, L. IMP dehydrogenase: structure, 73. Baell, J. & Walters, M. A. Chemical con artists foil kinetic studies with aminopyridine and triazolopyrazine mechanism, and inhibition. Chem. Rev. 109, drug discovery. Nature 513, 481 (2014). inhibitors. Biochemistry 48, 5339–5349 (2009). 2903–2928 (2009). 74. Nissink, J. W. M. & Blackburn, S. Quantification of 31. Favata, M. F. et al. Identification of a novel inhibitor of 54. Ward, W. H. et al. Kinetic and structural frequent-hitter behavior based on historical high- mitogen-activated protein kinase kinase. J. Biol. characteristics of the inhibition of enoyl (acyl carrier throughput screening data. Future Med. Chem. 6, Chem. 273, 18623–18632 (1998). protein) reductase by triclosan. Biochemistry 38, 1113–1126 (2014). 32. Solowiej, J. et al. Characterizing the effects of the 12514–12525 (1999). 75. Saubern, S., Guha, R. & Baell, J. B. KNIME workflow juxtamembrane domain on vascular endothelial 55. Swinney, D. C. Biochemical mechanisms of drug to assess PAINS filters in SMARTS format. growth factor receptor‑2 enzymatic activity, action: what does it take for success? Nat. Rev. Drug Comparison of RDKit and Indigo cheminformatics autophosphorylation, and inhibition by axitinib. Discov. 3, 801–808 (2004). libraries. Mol. Informat. 30, 847–850 (2011). Biochemistry 48, 7019–7031 (2009). 56. Dahl, G. & Akerud, T. Pharmacokinetics and the drug- 76. Shoichet, B. K. Interpreting steep dose-response 33. Kopcho, L. M. et al. Comparative studies of active target residence time concept. Drug Discov. Today 18, curves in early inhibitor discovery. J. Med. Chem. 49, site-ligand interactions among various recombinant 697–707 (2013). 7274–7277 (2006). constructs of human β-amyloid precursor protein This paper demonstrates that the extension of 77. Shoichet, B. K. Screening in a spirit haunted world. cleaving enzyme. Arch. Biochem. Biophys. 410, binding due to a long drug–target residence time Drug Discov. Today 11, 607–615 (2006). 307–316 (2003). occurs only when the binding dissociation rate is 78. Selwyn, M. J. A simple test for inactivation of an 34. Stevenson, L. M., Deal, M. S., Hagopian, J. C. & slower than the PK elimination rate. Exemplary enzyme during assay. Biochim. Biophys. Acta 105, Lew, J. Activation mechanism of CDK2: role of cyclin data for many drugs and/or drug candidates 193–195 (1965). binding versus phosphorylation. Biochemistry 41, indicate that the opposite scenario is commonly 79. Copeland, R. A. in Enzymes: A Practical Introduction 8528–8534 (2002). observed. to Structure, Mechanism, and Data Analysis 2nd edn 35. Anderson, K. et al. Binding of TPX2 to Aurora A alters 57. Walkup, G. K. et al. Translating slow-binding inhibition (ed. Copeland, R. A.) 266–304 (Wiley, 2002). substrate and inhibitor interactions. Biochemistry 46, kinetics into cellular and in vivo effects. Nat. Chem. 80. Singh, J., Petter, R. C., Baillie, T. A. & Whitty, A. 10287–10295 (2007). Biol. 11, 416–423 (2015). The resurgence of covalent drugs. Nat. Rev. Drug 36. Kuzmichev, A., Nishioka, K., Erdjument-Bromage, H., This paper describes a mechanistic Discov. 10, 307–317 (2011). Tempst, P. & Reinberg, D. Histone methyltransferase pharmacodynamic model that includes drug–target 81. Rawat, R., Whitty, A. & Tonge, P. J. The isoniazid-NAD activity associated with a human multiprotein complex kinetic parameters. It has been applied to predict adduct is a slow, tight-binding inhibitor of InhA, the containing the Enhancer of Zeste protein. Genes Dev. dose–response curves for inhibitors in an animal Mycobacterium tuberculosis enoyl reductase: adduct 16, 2893–2905 (2002). model of infection. affinity and drug resistance. Proc. Natl Acad. Sci. USA 37. Dang, L. et al. Cancer-associated IDH1 mutations 58. Johnson, D. S., Weerapana, E. & Cravatt, B. F. 100, 13881–13886 (2003). produce 2‑hydroxyglutarate. Nature 462, 739–744 Strategies for discovering and derisking covalent, 82. Garvey, E. P. et al. Potent inhibitors of HIV‑1 integrase (2009). irreversible enzyme inhibitors. Future Med. Chem. 2, display a two-step, slow-binding inhibition mechanism 38. Smits, A. H., Jansen, P. W., Poser, I., Hyman, A. A. & 949–964 (2010). which is absent in a drug-resistant T66I/M154I Vermeulen, M. Stoichiometry of chromatin-associated 59. Copeland, R. A. Mechanistic considerations in high- mutant. Biochemistry 48, 1644–1653 (2009). protein complexes revealed by label-free quantitative throughput screening. Anal. Biochem. 320, 1–12 83. Rudolph, J., Xiao, Y., Pardi, A. & Ahn, N. G. Slow mass spectrometry-based proteomics. Nucleic Acids (2003). inhibition and conformation selective properties of Res. 41, e28 (2013). This paper describes the steps required to design extracellular signal-regulated kinase 1 and 2 39. Krishnaswamy, S. & Betz, A. Exosites determine the best hit-finding assay possible, taking into inhibitors. Biochemistry 54, 22–31 (2014). macromolecular substrate recognition by account the requirements for a successful 84. Sculley, M. J., Morrison, J. F. & Cleland, W. W. prothrombinase. Biochemistry 36, 12080–12086 screening campaign and also post-screening lead Slow-binding inhibition: the general case. Biochim. (1997). evaluation. Biophys. Acta 1298, 78–86 (1996).

REVIEWS

85. Morrison, J. F. & Walsh, C. T. The behavior and 101. Copeland, R. A. Evaluation of Enzyme Inhibitors in 120. Foda, Z. H., Shan, Y., Kim, E. T., Shaw, D. E. & significance of slow-binding enzyme inhibitors. Drug Discovery: A Guide for Medicinal Chemists and Seeliger, M. A. A dynamically coupled allosteric Adv. Enzymol. Relat. Areas Mol. Biol. 61, 201–301 Pharmacologists (John Wiley & Sons, 2013). network underlies binding cooperativity in Src kinase. (1988). 102. Munson, P. J. & Rodbard, D. An exact correction to Nat. Commun. 6, 5939 (2015). This article describes the detection and analysis of the “Cheng-Prusoff” correction. J. Recept. Res. 8, 121. Brown, A. J. XXXVI.—Enzyme action. J. Chem. Soc., slow-binding inhibition. 533–546 (1988). Trans. 81, 373–388 (1902). 86. Copeland, R. A., Basavapathruni, A., Moyer, M. & This paper describes the effect of the 122. Michaelis, L. & Menten, M. L. Die kinetik der Scott, M. P. Impact of enzyme concentration and concentration of labelled drug and the invertinwirkung. Biochem. Z. 49, 333–369 (1913). residence time on apparent activity recovery in jump concentration of the binding site on the measured 123. Ehrlich, P. Address in pathology on chemiotherapy:

dilution analysis. Anal. Biochem. 416, 206–210 ED50 and hence Ki values, as well as provides an delivered before the Seventeenth International (2011). explicit correction to overcome these issues. Congress of Medicine. Br. Med. J. 2, 353–359 87. Copeland, R. A., Lombardo, D., Giannaras, J. & 103. Cheng, H. C. The influence of cooperativity on the (1913).

Decicco, C. P. Estimating KI values for tight binding determination of dissociation constants: examination 124. Cook, D. et al. Lessons learned from the fate of inhibitors from dose-response plots. Bioorg. Med. of the Cheng-Prusoff equation, the Scatchard analysis, AstraZeneca’s drug pipeline: a five-dimensional Chem. Lett. 5, 1947–1952 (1995). the Schild analysis and related power equations. framework. Nat. Rev. Drug Discov. 13, 419–431 88. Goldstein, A. The mechanism of enzyme-inhibitor- Pharmacol. Res. 50, 21–40 (2004). (2014). substrate reactions: illustrated by the cholinesterase- 104. Johnson, K. A. A century of enzyme kinetic analysis, 125. Mezzasalma, T. M. et al. Enhancing recombinant physostigmine-acetylcholine system. J. Gen. Physiol. 1913 to 2013. FEBS Lett. 587, 2753–2766 protein quality and yield by protein stability 27, 529–580 (1944). (2013). profiling. J. Biomol. Screen 12, 418–428 (2007). 89. Morrison, J. F. Kinetics of the reversible inhibition of 105. Cornish-Bowden, A. Analysis and interpretation of 126. Hajduk, P. J. et al. A strategy for high-throughput enzyme-catalysed reactions by tight-binding inhibitors. enzyme kinetic data. Persp. Sci. 1, 121–125 assay development using leads derived from nuclear Biochim. Biophys. Acta 185, 269–286 (1969). (2014). magnetic resonance-based screening. J. Biomol. This paper highlights the challenges associated 106. Renaud, J. P. et al. Biophysics in drug discovery: Screen 7, 429–432 (2002). with the determination of inhibition constants for impact, challenges and opportunities. Nat. Rev. Drug 127. Hajduk, P. J. & Burns, D. J. Integration of NMR and tight-binding inhibitors and presents the theory Discov. 15, 679–698 (2016). high-throughput screening. Comb. Chem. High behind the analysis. 107. Holdgate, G. et al. in Protein-Ligand Interactions: Throughput Screen 5, 613–621 (2002).

90. Murphy, D. J. Determination of accurate KI values for Methods and Applications (eds Williams, M. & 128. Genick, C. C. et al. Applications of biophysics in high- tight-binding enzyme inhibitors: an in silico study of Daviter, T.) 327–355 (Springer, 2013). throughput screening hit validation. J. Biomol. Screen experimental error and assay design. Anal. Biochem. 108. Olson, B. J. & Markwell, J. UNIT 3.4: Assays for 19, 707–714 (2014). 327, 61–67 (2004). determination of protein concentration. Curr. Protoc. 129. Evenas, J. et al. HTS followed by NMR based 91. Copeland, R. A., Harpel, M. R. & Tummino, P. J. Protein Sci. http://dx.doi.org/10.1002/0471140864. counterscreening. Discovery and optimization of Targeting enzyme inhibitors in drug discovery. Expert ps0304s48 (2007). pyrimidones as reversible and competitive inhibitors Opin. Ther. Targets 11, 967–978 (2007). 109. Dunn, B. UNIT 3.2: Quantitative amino acid analysis. of xanthine oxidase. Bioorg. Med. Chem. Lett. 24, 92. Bauer, R. A. Covalent inhibitors in drug discovery: from Curr. Protoc. Protein Sci. http://dx.doi. 1315–1321 (2014). accidental discoveries to avoided liabilities and designed org/10.1002/0471140864.ps0302s00 (2001). 130. Ciulli, A. Biophysical screening for the discovery of therapies. Drug Discov. Today 20, 1061–1073 (2015). 110. Holdgate, G. A., Anderson, M., Edfeldt, F. & small-molecule ligands. Methods Mol. Biol. 1008, 93. Strelow, J. M. A. Perspective on the kinetics of Geschwindner, S. Affinity-based, biophysical methods 357–388 (2013). covalent and irreversible inhibition. SLAS Discov. 22, to detect and analyze ligand binding to recombinant 131. Geschwindner, S. in Lead Generation: Methods 3–20 (2017). proteins: matching high information content with high and Strategies Ch. 14 (ed. Holenz, J.) (Wiley, This paper provides a detailed explanation of the throughput. J. Struct. Biol. 172, 142–157 (2010). 2016). characterization of irreversible inhibition. 111. Prinz, F., Schlange, T. & Asadullah, K. Believe it or not: 132. Bergsdorf, C. & Ottl, J. Affinity-based screening 94. Krippendorff, B. F., Neuhaus, R., Lienau, P., Reichel, A. how much can we rely on published data on potential techniques: their impact and benefit to increase the & Huisinga, W. Mechanism-based inhibition: deriving drug targets? Nat. Rev. Drug Discov. 10, 712 (2011). number of high quality leads. Expert Opin. Drug

KI and kinact directly from time-dependent IC50 values. 112. Frye, S. V. et al. Tackling reproducibility in academic Discov. 5, 1095–1107 (2010). J. Biomol. Screen 14, 913–923 (2009). preclinical drug discovery. Nat. Rev. Drug Discov. 14, 133. Huber, W. A new strategy for improved secondary 95. Jöst, C., Nitsche, C., Scholz, T., Roux, L. & Klein, C. D. 733–734 (2015). screening and lead optimization using high- Promiscuity and selectivity in covalent enzyme 113. Mullane, K. & Williams, M. Unknown unknowns in resolution SPR characterization of compound-target inhibition: a systematic study of electrophilic biomedical research: does an inability to deal with interactions. J. Mol. Recognit. 18, 273–281 fragments. J. Med. Chem. 57, 7590–7599 (2014). ambiguity contribute to issues of irreproducibility? (2005). 96. Lanning, B. R. et al. A road map to evaluate the Biochem. Pharmacol. 97, 133–136 (2015). 134. Barsyte-Lovejoy, D. et al. (R)-PFI‑2 is a potent and proteome-wide selectivity of covalent kinase inhibitors. 114. Edfeldt, F. N., Folmer, R. H. & Breeze, A. L. Fragment selective inhibitor of SETD7 methyltransferase Nat. Chem. Biol. 10, 760–767 (2014). screening to predict druggability (ligandability) and activity in cells. Proc. Natl Acad. Sci. USA 111, 97. Schwartz, P. A. et al. Covalent EGFR inhibitor analysis lead discovery success. Drug Discov. Today 16, 12853–12858 (2014). reveals importance of reversible interactions to 284–287 (2011). 135. Davis, M. I. et al. Biochemical, cellular, and biophysical potency and mechanisms of drug resistance. Proc. 115. Folmer, R. H. Integrating biophysics with HTS-driven characterization of a potent inhibitor of mutant Natl Acad. Sci. USA 111, 173–178 (2014). drug discovery projects. Drug Discov. Today 21, isocitrate dehydrogenase IDH1. J. Biol. Chem. 289, 98. Ring, B., Wrighton, S. A. & Mohutsky, M. in Enzyme 491–498 (2016). 13717–13725 (2014). Kinetics in Drug Metabolism: Fundamentals and 116. Kaeberlein, M. et al. Substrate-specific activation Applications (eds Nagar, S., Argikar, U. & Tweedie, D.) of sirtuins by resveratrol. J. Biol. Chem. 280, Acknowledgements 37–56 (Springer, 2014). 17038–17045 (2005). The authors would like to thank S. Thrall for critical input 99. Fleming, M. A. et al. Inhibition of IMPDH by 117. Borra, M. T., Smith, B. C. & Denu, J. M. Mechanism of into FIG.1. mycophenolic acid: dissection of forward and reverse human SIRT1 activation by resveratrol. J. Biol. Chem. pathways using capillary electrophoresis. Biochemistry 280, 17187–17195 (2005). Competing interests statement 35, 6990–6997 (1996). 118. Kalliokoski, T., Kramer, C., Vulpetti, A. & Gedeck, P. The authors declare competing interests: see Web version

100. Hedstrom, L. & Wang, C. C. Mycophenolic acid and Comparability of mixed IC50 data – a statistical for details. thiazole adenine dinucleotide inhibition of analysis. PloS ONE 8, e61007 (2013). Tritrichomonas foetus inosine 5ʹ‑monophosphate 119. Hall, J. et al. Biophysical and mechanistic insights into Publisher’s note dehydrogenase: implications on enzyme mechanism. novel allosteric inhibitor of spleen tyrosine kinase. Springer Nature remains neutral with regard to jurisdictional Biochemistry 29, 849–854 (1990). J. Biol. Chem. 287, 7717–7727 (2012). claims in published maps and institutional affiliations.