Approved and Experimental Small‐Molecule Oncology Kinase Inhibitor Drugs: A Mid‐2016 Overview

Kinase inhibitor research is a comparatively recent branch of medicinal chemistry and pharmacology and the first small‐molecule kinase inhibitor, imatinib, was approved for clinical use only 15 years ago. Since then, 33 more kinase inhibitor drugs have received regulatory approval for the treatment of a variety of cancers and the volume of reports on the discovery and development of kinase inhibitors has increased to an extent where it is now difficult—even for those working in the field—easily to keep an overview of the compounds that are being developed, as currently there are 231 such compounds, targeting 38 different protein and lipid kinases (not counting isoforms), in clinical use or under clinical investigation. The purpose of this review is thus to provide an overview of the biomedical rationales for the kinases being targeted on the one hand, and the design principles, as well as chemical, pharmacological, pharmaceutical, and toxicological kinase inhibitor properties, on the other hand. Two issues that are especially important in kinase inhibitor research, target selectivity and drug resistance, as well as the underlying structural concepts, are discussed in general terms and in the context of relevant kinases and their inhibitors.


INTRODUCTION
Kinases are now one of the most important drug development target classes in several therapeutic indications, but until very recently all kinase inhibitors that were approved for clinical use were anticancer drugs, with the exception of fasudil, a (diazepanesulfonyl)isoquinoline ROCK inhibitor (for protein kinase nomenclature refer to Table S1). 1 Fasudil has been approved for some years in Japan for the treatment of certain cardiovascular indications 2 but even this agent-and ROCK inhibitors in general-may in fact also have potential uses in oncology. 3 However, the situation changed with the U.S. Food and Drug Administration (FDA) approval in 2012 of the pyrrolopyrimidine JAK inhibitor tofacitinib for the treatment of rheumatoid arthritis. 4 Again JAK inhibitors are also relevant in cancer, as we shall see later. At the time it was expected that the first kinase inhibitor approval outside oncology would act as a watershed and that pharmaceutical companies would accelerate kinase inhibitor development for a range of different therapeutic indications. In fact, kinase inhibitors are undergoing clinical development in inflammatory, cardiovascular, neurodegenerative, liver, and metabolic diseases 5 but since the approval of tofacitinib there has been only one additional approval outside oncology: nintedanib, a multitargeted kinase inhibitor (MTKI), in idiopathic pulmonary fibrosis (IPF). 6 Interestingly, this compound was also approved in the European Union for lung cancer patients with advanced adenocarcinoma after first-line chemotherapy. Here, we shall confine our discussion to kinase inhibitors in oncology, not only approved agents (Fig. 1), but all compounds currently undergoing clinical development.

r FISCHER
According to an analysis of new anticancer therapeutics that entered the clinic during the period 1990-2006, kinase inhibition was the most prevalent mode of action of experimental agents, whereas previously interference with DNA replication was the most common mechanism. 7 A different analysis of new clinical oncology agents during a similar period (1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007) noted that while clinical development attrition was high (82%) overall, kinase inhibitors performed significantly better (53% attrition). 8 Only a few years ago it was easy for those working in the field to keep track of the kinase inhibitor oncology drugs under development but the sheer number of agents that have recently entered clinical studies now renders this difficult. The main purpose of the present review is therefore to provide an up-to-date summary of some of approved and experimental (as verified in relevant publicly accessible literature, corporate disclosures, and https://clinicaltrials.gov/) kinase inhibitor cancer drugs. Salient points regarding kinase inhibitor targets, selectivity, medicinal chemistry, pharmacology, and toxicology will be discussed. Biological agents, including antibodies and antisense oligonucleotides, are not included but the focus is on small-molecule agents.

SELECTIVITY
Following the success with the first approved small-molecule kinase inhibitor anticancer drug, imatinib ( Fig. 1), 9 especially in the treatment of chronic myelogenous leukemia (CML), highly specific targeted kinase inhibitors rapidly became a desirable goal in oncology drug discovery. However, demonstration of clear-cut clinical efficacy has been much more elusive with a range of other experimental kinase inhibitor drugs, and this situation has led to the on-going dispute about selective versus so-called multitargeted agents.
The question is whether "magic bullets" are preferable to "magic shotguns" 10 and arises not so much because one might be superior to the other, but because upon closer examination even the most carefully designed small-molecule kinase inhibitors often inadvertently turn out to be multitargeted. In the early days of kinase inhibitor development, it was not possible to profile the kinase selectivity of candidate molecules extensively, and unexpected kinase targets were frequently discovered for drugs late in the development process. This situation has changed with the advent of kinase panel 11 and chemical proteomics 12 technologies that permit kinome-wide analysis of candidate molecules.
Nevertheless, multitargeted agents are frequently more efficacious than selective ones, 13 although this may be due to selective drugs lacking efficacy because they block targets that are inappropriate per se or inappropriate for a specific cancer. Furthermore, network models suggest that partial inhibition of a small number of targets can be more efficient than the complete inhibition of a single target. 14 Certainly, the task of devising a kinase inhibitor that exclusively inhibits a range of predetermined kinases-which may not necessarily be structurally relatedwould appear to be formidable. Although strategies for the rational design of multitargeted agents by multiple structure-based design and pharmacophore combination 15,16 have been described, these have not yet been implemented widely for the design of kinase inhibitors with predetermined multiple selectivity.

A. Active versus Inactive Kinases
Because the vast majority of small-molecule kinase inhibitors target the ubiquitous ATPbinding site, target selectivity represents the biggest challenge for kinase inhibitor drug design and development. Protein and lipid kinases catalyze the transfer of the γ -phosphoryl group of ATP to their specific macromolecular substrates. This requires an architectural arrangement of the kinase to allow simultaneous binding of both substrates and positioning of the catalytic residues for productive phosphoryl-transfer. For protein kinases, this canonical arrangement can be illustrated using PKA as an example (Fig. 2). ATP is bound in a cleft between the N-and C-lobes of the kinase. These lobes are connected by the hinge region, which contains the so-called gatekeeper residue. The adenine portion of ATP forms H-bonds with the mainchain peptide bonds of the two hinge residues distal to the gatekeeper. The activation (A) loop adopts a conformation that supports binding of the macromolecular substrate and in which the side chain of the Asp residue of the conserved DFG motif is positioned toward the ATP γphosphate (DFG in). Furthermore, the active (i.e., catalytically competent) kinase architecture is characterized by a near-linear arrangement of the regulatory (R) spine, which is composed of five residues (R0-R4) that interlock through H-bonding (R0-R1), CH-π (R1-R2-R3), and hydrophobic (R3-R4) interactions. Distortion of the linear spine, which contains residues from both the N-(R3 and R4) and C-lobes (R0, R1, and R2), is characteristic of inactive kinase architectures and can result from tilting of the kinase lobes about the hinge, by conformational preferences of the A-loop as a result of its phosphorylation status, or movement of the αC-helix or the catalytic (C) loop. It is sometimes stated that type-II inhibitors are not ATP competitive and that for this reason the drop-off in cellular potency from biochemical potency is frequently smaller than is the case for type-I inhibitors. 21 This is not entirely correct, since inactive forms of kinases can still bind ATP, although the difference in K M (ATP) between inactive and activated (phosphorylated) kinase forms can be large. 19 The ratio of biochemical and cellular potency of a kinase inhibitor will therefore depend on whether it competes with ATP for the active or an inactive state of the target kinase, and in the latter case on the K M (ATP) of the inactive state. Regardless of inhibition mode, the relationship between biochemical and cellular potency is of course also affected by factors such as solubility, protein binding, and membrane permeability.

D. Selectivity Based on Receptor Occupancy
For kinase inhibitor drugs, we are interested not so much in selectivity based on differential kinase affinity under equilibrium conditions (SI 1 ), but in selectivity under physiological conditions in vivo. Here, we need to take into account that the magnitude of physiological responses is proportional to the amount of inhibitor bound by a particular kinase over time. To assess true selectivity of kinase inhibitors it is therefore more useful to look at fractional kinase (receptor) r FISCHER occupancy as a function of time (fo t ). Based on this we can formulate a second SI ( Fig. 3A; For a type-I kinase inhibitor that reversibly binds a kinase by mechanism A, fo t depends on ligand concentration and K d as shown in Figure 3A. To illustrate the relationship between SI 1 and SI 2 , let us assume that we have kinase inhibitors that all bind a particular on-target kinase with a K d value of 10 nM but that bind an undesirable (leading to toxicity) off-target kinase with K d values of 50 nM (SI 1 = 5), 250 nM (SI 1 = 25), 1 μM (SI 1 = 100), and 10 μM (SI 1 = 1000), respectively. If we assume that these hypothetical kinase inhibitors have identical pharmacokinetic profiles and are dosed repeatedly so as to achieve continuous exposure, measured as the concentration of free drug fraction at steady state (C ssf ), then we can substitute C ssf for [L] t in the expression for fo t and evaluate SI 2 as a function of C ssf . This analysis (Fig. 3B) shows that at C ssf = K d (on-target) = 10 nM we have approximately 50% on-target receptor occupancy. At that C ssf value, we have SI 2 values of 3, 13, 50, and 500 for the inhibitors with SI 1 values of 5, 25, 100, and 1000, respectively. Depending on how physiological responses are coupled to on-target kinase occupancy, we may require much higher fo to achieve efficacy. If 90% occupancy is required (fo on-target = 0.9, C ssf = 90 nM), then the SI 2 values for our inhibitors are 1.4 (SI 1 = 2.5), 3.4 (SI 1 = 25), 11 (SI 1 = 50), and 100 (SI 1 = 1000), respectively. Depending on how off-target kinase occupancy is coupled to toxicity, the drop-off from SI 1 to SI 2 may be smaller or larger but the relationship between selectivity in terms of binding versus occupancy as a function of C ssf clearly shows that an inhibitor with an SI 1 value of < 100 is unlikely to provide a useful margin between on-and off-target activities in vivo.

E. Selectivity Based on Residence Time
High-affinity kinase inhibitors, especially those of type II, frequently exhibit slow binding kinetics. This is due to induced-fit binding, where the kinase undergoes a time-dependent change from a conformation R to a conformation R*, where the latter has higher affinity for the ligand L than the former ( Fig. 3A; mechanism B). Here, dissociation (off-rate) and residence time are not defined by a single rate constant as in mechanism A, but by a composite of rate constants, where k 4 (and sometimes k 2 ) are limiting for dissociation. For inhibitors that bind kinases in inactive-like conformations, mechanism C is sometimes invoked, especially when these inhibitors selectively bind the kinase in its A-loop nonphosphorylated state. There are currently insufficient kinetic data to distinguish clearly between inhibitors that act by mechanisms B and C, however. 22 In any case, the situation often arises where an inhibitor exhibits kinase residence time that is significant in comparison to its in vivo elimination half-life. Under these conditions, receptor occupancy can be maintained at levels leading to pharmacological responses even after a drug has been largely eliminated from systemic circulation. Under these circumstances, we need to take the dynamics of binding into account when assessing selectivity. To illustrate this point, let us assume that we have a kinase inhibitor that interacts with three target kinases and that the rate constants k 1 (1 × 10 8 M −1 s 1 , i.e., tenfold below the diffusional limit), k 2 (1 s −1 ), and k 3 (6 × 10 −3 s −1 ) are identical in each case, and that only k 4 varies: 8.5 × 10 −3 s −1 (target 1), 2.8 × 10 −4 s −1 (target 2), and 1.26 × 10 −5 s −1 (target 3), that is, typical values for type-I, type-II, and covalent reversible type-VI inhibitors. 17,23 With these values we have residence times τ of 2 min (target 1), 1 h (target 2), and 22 h (target 3). If we further assume that the inhibitor was dosed orally once, reaching a maximum plasma concentration (C max ) of 10 μM and being eliminated with a half-life of 3 h, then we can calculate free plasma levels, as well as fo t with respect to the three targets, as a function of time as shown (Fig. 3C). This analysis reveals that here selectivity is not driven by apparent affinity under equilibrium, but by residence time. Thus, >50% receptor occupancy is maintained for up to 12 h (target 1), 24 h (target 2), and 36 h (target 3). For targets 2 and 3, significant receptor occupancy persists at a time when plasma levels would be undetectable using conventional bioanalysis methods (24 h, C f = 4.2 × 10 −10 M). Because the inhibitor dissociates from target 3 much more slowly than for targets 1 and 2, it can be expected to be functionally selective for target kinase 3. Again the exact situation will depend on the magnitude and duration of receptor occupancy required to elicit physiological responses from the three targets, but it is clear that residence time is important in terms of selectivity and should be considered in the drug discovery process. 24

TARGET VALIDITY
It is important to remember that the pioneer kinase inhibitor drug imatinib itself is not monoselective for BCR-ABL, the oncogenic kinase uniquely present in CML, but also inhibits a range of other kinases with high potency. Clearly, the therapeutic effectiveness of imatinib in CML is not principally a result of drug selectivity, but the vital dependence of CML cells on BCR-ABL kinase activity. So for the narrow versus broad drug selectivity discussion to make any sense, we need to consider target validity.
If pharmacological modulation of a specific cellular component is sufficient to eradicate a diagnostically distinct cancer type, then surely an agent highly selective for that target should be preferable to a more promiscuous one on basic pharmacological grounds. If, on the other hand, the viability of a particular cancer type only appears susceptible to interference with two or more aberrant components, then multitargeted agents would seem to be called for. Because we want to devise new cancer therapies that are inherently less toxic than conventional chemotherapies, the important thing is specificity for cancerous cells rather than inherent target selectivity.
If we accept the premise that at least for some cancer indications modulation of multiple targets is necessary to achieve the desired therapeutic effect, then the question that poses itself is whether it would be better to employ an appropriate cocktail of highly selective drugs or a single multitargeted drug. 25,26 Considering the complexity of human cancers, one would expect that the availability of a range of highly targeted agents would allow a more effective and modular approach to personalized therapy than multitargeted agents. However, combination therapies are prone to dosing errors and drug-drug interactions. 27 Furthermore, highly targeted drugs by definition will not find broad chemotherapeutic application but will only work in the (perhaps small) subset of cancers that possess the specific targeted molecular lesion, whereas multitargeted drugs may find wider therapeutic use, that is, they may be preferable commercial propositions for pharmaceutical companies.
As it turns out, the main clinical indication for imatinib, CML, is unusual since most cancers, especially nonhematological tumors, including, for example, common epithelial neoplasms of the lung, breast, colon, and prostate, display a much more complex transformed genotype than CML and may lack the critical vulnerability centered on a single cellular component characteristic of early-stage CML. In fact, mouse models suggest that the BCR-ABL translocation is the only genetic abnormality required to cause CML. 28 It thus follows that imatinib probably represents an exception rather than a new paradigm in the area of kinase inhibitors. 29 However, target validation work in several areas suggests that there are probably many other cancers that may also possess therapeutic Achilles heels. Typically this situation arises as a result of oncogene addiction, that is, the apparently paradoxical dependence of a tumor cell for sustained proliferation and survival on a single oncogenic pathway or protein, despite the presence of numerous genetic alterations. 30 A number of such oncogene addictions have been identified and frequently the relevant oncogene products are kinases. 31,32 Such kinases are attractive as cancer drug targets, and several are now being targeted with experimental drugs, for example, BRAF in melanoma, HER2 in breast cancer, HER1 in nonsmall cell lung cancer (NSCLC), MET in gastric cancer, and ARK in colon cancer. However, unlike in the case of CML, these addictions are not usually present in all cancers of a particular histological type, but the potentially responsive cancers need to be identified by examination of appropriate biomarkers.

DRUG RESISTANCE
Although not themselves mutagenic like many conventional chemotherapy agents, kinase inhibitor drugs, just like most anticancer drugs, are nevertheless prone to both intrinsic and acquired resistance. The manifestation of drug resistance in at least a proportion of cases involves direct mutation of the target kinase in such a way that function is preserved, while inhibition by the drug is circumvented. Although it remains unclear how these mutations are caused, it is likely that genetic selection for insensitive target kinase mutants under the influence of the kinase inhibitor drug plays a role. This phenomenon was first observed with imatinib and is in part responsible for the current emphasis on MTKIs, which may not be vulnerable to the same extent to this selection pressure. 33 In the case of the HER1 inhibitors gefitinib and erlotinib, which are used for the treatment of NSCLC with certain somatic mutations in the HER1 Tyr kinase or with amplified wild-type HER1, it has also been observed that all patients eventually develop resistance, either through a secondary HER mutation T790M in about half of all cases, by MET oncogene amplification, or via other mechanisms. 34 The HER1 T790M mutation concerns the gatekeeper residue (Thr in most Tyr kinases), which is not required for the binding of the natural cofactor ATP, but which is almost invariably involved in the recognition of ATP-antagonist kinase inhibitors. Resistance to HER2 inhibitors in breast cancer is also frequent and also develops through multiple mechanisms but unlike for HER1 inhibitors, for example, lapatinib resistance due to the HER2 gatekeeper T798I mutation appears to be less prevalent. 35 Resistance mutations corresponding to those observed with HER Tyr kinase inhibitors are known to be important in the case of many other kinases and their inhibitors, as we shall see later.
It has long been known that cancer cells can develop drug resistance much more rapidly than would be predicted from the rates of conventional mutation, by differential utilization of the genome, while normal cells in the same organism remain sensitive. There is now much evidence that not only point mutations, but also deletions, translocations, amplifications, epigenetic changes, and even altered microRNA levels, can generate cancer cell progeny tolerant to individual or multiple drugs. 36 It is also now believed that clonal karyotype alterations are generated auto-catalytically by cancer-specific aneuploidy and that these specific chromosomal alterations cause drug resistance by changing the stoichiometry and integrity of multigenic transcriptomes. 37 Even in the case of imatinib over half of all resistance cases are associated with clonal karyotype alterations rather than kinase mutations. 38

TOXICOLOGY
For the discussion of kinase inhibitor toxicology we need to return to the question of selectivity. Provided a hypothetical mono-selective kinase inhibitor anticancer drug does not give rise to toxic metabolites, one would expect such a compound to possess a clean toxicology profile. Obviously this will only be the case, however, if on-target effects in tissues other than the tumor can be tolerated. Again we can use the pioneer kinase inhibitor drug imatinib to illustrate this point. Although one of the targets of imatinib, BCR-ABL, is uniquely present in transformed granulocytes as a result of an oncogenic chromosomal translocation, the normal nonfused ABL kinase, which is widely expressed in the body, is of course also inhibited by imatinib. This kinase turns out to be especially important for cardiomyocyte function and survival, and ABL inhibition has been identified as the root cause of the cardiotoxicity of imatinib. 39 Similarly, hypo-phosphatemia is probably another on-target toxicity of imatinib, resulting from inhibition of PDGFR kinase activity in osteoclasts. 40 Various other toxicities are target related, such as the skin toxicity of HER inhibitors. 41 Dermatological toxicities are also a feature of the MTKI sorafenib, currently approved for the treatment of metastatic clear-cell renal cell carcinoma (RCC) and hepatocellular carcinoma, but it is not understood if this is target related. 42 Animal studies show that chronic pharmacologic inhibition of HER leads to cardiac dysfunction. 43 Cardiac toxicity-drugs causing heart muscle valve damage or potentially fatal arrhythmias-has been implicated in 28% of drug withdrawals in the United States over the last 30 years. That cardiac-and skin-related toxicities have been observed for many Tyr kinase inhibitors both in oncology and other indications suggests the possibility of a class effect. 44 Another on-target toxicity is hyperglycemia, which has often been reported as a side effect of agents targeting IGF-IR (where IR is insulin receptor), PI3K, AKT, and mTOR. All these kinases are important components of the insulin signaling pathway and their inhibition leads to altered hepatic glycogen metabolism and blockage of peripheral glucose uptake. It was shown recently, however, that drug-induced hyperglycemia might be attenuated by decreasing liver glycogen storage using fasting, combined with a low-carbohydrate diet. 45

APPROVED AND EXPERIMENTAL KINASE INHIBITOR DRUGS
An overview of approved and experimental oncology kinase inhibitor drugs is given in Figure 4 and individual agents are discussed below by the kinases they act on. Details for compounds are summarized in Tables S2-29.

A. BCR-ABL Kinase
The BCR-ABL oncogene is generated in CML by the so-called Philadelphia chromosome that results from a reciprocal translocation that juxtaposes the Abelson murine leukemia viral oncogene homolog 1 (ABL) gene with the breakpoint cluster region (BCR) gene. 46 Since its approval, the first ABL kinase inhibitor imatinib has been used successfully in the control of progression from CML chronic phase to blast crisis. Although medial survival in CML was 3-6 years prior to the clinical use of imatinib, a 10-year survival rate of 68% has been reported for patients with CML who receive imatinib after failure of interferon. 47 A second generation of ABL inhibitors has now been approved: dasatinib and nilotinib, predominantly because many patients with CML undergo clinical relapse due to resistance to imatinib while these new agents are effective against most imatinib-resistant BCR-ABL mutants (Fig. 5). 48 Neither imatinib nor the second-generation ABL inhibitors are selective, but also inhibit a range of other kinases, especially PDGFR and SCFR. In fact, some of these agents have been used in the treatment of dermatofibroma sarcoma protuberans and gastrointestinal stromal tumors (GIST), in which disorders PDGFR and SCFR, respectively, are implicated. 49 Bosutinib is a third-generation ABL inhibitor. Like the second-generation inhibitors, it is also active against the majority of imatinib-resistant BCR-ABL mutants, but it is a much more selective inhibitor, with only SRC kinases as significant additional targets. 50 The other approved third-generation inhibitor, ponatinib, on the other hand, is an MTKI, but it is the only approved ABL inhibitor that has activity against the imatinib-resistant T315I gatekeeper r FISCHER BCR-ABL mutant kinase (Fig. 5). 51 Another BCR-ABL inhibitor, radotinib (structurally closely related to imatinib), was approved in South Korea in 2012.
Although there are currently six approved ABL inhibitors used successfully for the treatment of CML, there are at least another two ABL inhibitors at various stages of clinical development, including flumatinib (another compound closely related to imatinib) and ABL 001, an allosteric (type IV) inhibitor that targets the myristoyl-binding site in the SH1 catalytic kinase domain (Table S2).

B. SRC Kinases
A number of ABL kinase inhibitors approved in CML, such as dasatinib, nilotinib, bosutinib, and ponatinib, also inhibit the closely related SRC kinases. SRC family kinases are cytosolic The ABL kinase (green) binding mode of the type-II inhibitor imatinib (gray) shows an H-bond interaction (dashed line) with the gatekeeper residue T315 (yellow; secondary mutation sites Y253 and E255 in the Gly-rich loop are indicated in purple). 222 (C) The binding of ATP does not involve the T351 residue, hence the catalytic competence of the T315I mutant. 223 (D) Secondgeneration drugs such as the type-I compound dasatinib are not effective against T315I mutant kinase, as they also make polar interactions with T315. 224 (E) The recently approved ponatinib, which also inhibits the T315I mutant, evades the altered gatekeeper (Ile) through the placement of an ethynyl linkage between its front-and back-pocket-binding portions. 225 Constructed from PDB entries 1IEP, 2SRC, 2GQG, and 3IK3. The substrate peptide SRC kinase inhibitor KX 01 was shown by saturation transfer difference NMR spectroscopy, using an ATP-site-binding paramagnetic spinlabeled pyrazolopyrimidine probe containing the 2,2,6,6-tetramethyl-piperidine-1-oxyl (TEMPO) radical, to bind to SRC kinases in an ATP noncompetitive manner. 226 (B) An X-ray crystal structure (PDB entry 3KXZ) of this probe (green) bound to the SRC-family kinase LCK (surface with electrostatic potential coloring) is shown. The extended substrate-binding cavity, where KX-01 (cyan, modeled (Schrödinger Glide, https://www.schrodinger.com/Glide) binding pose) is likely to bind, is evident proximal to the ATP-binding site.
Tyr kinases that have a multitude of functions in cellular signal transduction, whose activated forms are aberrantly expressed in many solid tumors, especially metastatic cancers. 52 This has led to clinical evaluation of such dual ABL and SRC agents, including saracatinib and AZD 0424 (closely related compounds that share the (6-chloro-2H- [1,3]dioxolo[4,5-b]pyridin-7-yl)aminoquinazolin-7-yl substructure), in advanced solid tumors such as NSCLC, as well as in lymphomas. 53 Saracatinib is also currently being evaluated in Alzheimer's disease as an inhibitor of the SRC family kinase FYN, which is implicated in the pathogenesis of this disease. 54 The most SRC-selective experimental kinase inhibitor drug (Table S3) appears to be a compound known as KX 01, which is a type-III kinase inhibitor, that is, a peptidomimetic of SRC kinase substrate peptides (Fig. 6).
To date, therapeutic success in clinical trials of SRC kinase inhibitors has been limited and it has been postulated that a paradigm shift is required in clinical trial design for such agents in terms of patient selection, biomarker development, and rational design of combinatorial regimens with SRC inhibitors. 55 However, recent insights suggest that highly selective SRC kinase inhibitors may possess significant advantages over dual ABL/SRC inhibitors, since tumor suppressor signaling through ABL appears to occur at least in some cancers. 56 Inhibitors that are highly selective for SRC over ABL have recently been reported, and the clinical development of such compounds may overcome this issue, as well as the cardiotoxicity associated with ABL inhibition. 57

C. Epidermal Growth Factor Receptor Kinases
The EGFR Tyr kinase family consists of four transmembrane receptors, HER1-4, and signaling through this receptor family is important for many cell functions and for cell survival. At least HER1-3 are involved in tumorigenesis, and hetero-dimerization of these receptors plays an important role in their function. 58 Inhibitors of the EGFR Tyr kinase family have been studied mostly in breast cancers and NSCLCs driven by HER2 and HER1, respectively. As a consequence, activating mutations in, for example, EGFR (such as exon 19 in-frame deletions and L858R point mutations) serve not only as the rationale for the use of EGFR Tyr kinase inhibitors, but also as biomarkers for patient selection. 59 Of the approved EGFR inhibitors, both gefitinib and erlotinib (Fig. 7B) are HER1 specific and are used in NSCLCs, many of which overexpress HER1. 60 The trend in new EGFR Tyr kinase inhibitors is now toward pan-HER-targeted compounds, including the approved agent lapatinib (Fig. 7D), which inhibits both HER1 and HER2 and is used for the treatment of advanced HER2-positive breast cancer. 61 Icotinib, which is approved in China, is also an inhibitor of both HER1 and HER2. Although HER2 lacks an extracellular ligand-binding domain, it is nevertheless involved in hetero-dimerization with other ligand-bound EGFR partners. Because EGFR signaling that affects cell proliferation and survival depends on both HER1 and HER2, inhibitors that block the catalytic activity of both have been sought, in the hope of enhanced clinical efficacy. However, clinical studies of pan-HER inhibitors have been difficult due to a narrow therapeutic margin with such agents, with dose-limiting on-target pan-HER toxicities such as diarrhea and skin rash. 62 Although the HER2-specific antibody drug trastuzumab has been used successfully in HER2-overexpressing breast cancer for some time, 63 it is perhaps surprising that more progress has not been achieved to date with HER2-selective small-molecule kinase inhibitors. Highly HER2-selective inhibitors are in fact known (Fig. 7C), 64 but none seems to be under clinical evaluation at present.
Both acquired and innate drug resistance appear to play an important role in limiting the response rates to EGFR Tyr kinase inhibitor drugs. The comparatively unique presence of an unpaired Cys residue in the ATP-binding pocket of the EGFR Tyr kinases has allowed the development of irreversible ATP-site inhibitors. 65 In preclinical studies, it has been shown that cancers that have become resistant to gefitinib and erlotinib through the secondary EGFR T790M mutation (in ca. 60% of cases) remain sensitive to second-generation type-VI inhibitors. 66 This, and the observation that such inhibitors may be less likely to induce drugresistant kinase mutants, might be due to the fact that covalent inhibitors can still block a mutant kinase that has reduced affinity for noncovalent inhibitors because with covalent binders inhibition is rate-rather than affinity-driven. 67 Two such agents that can overcome T790M mutation, afatinib (Fig. 7F) and osimertinib (Fig. 7G), were recently approved in metastatic NSCLC. Afatinib is a pan-HER inhibitor that blocks the catalytic activity of HER1, HER2, and HER4, as well as phosphorylation of the EGFR-trans-signaling partner HER3, 68 which is itself kinase-inactive. 69 Osimertinib, on the other hand, is selective for T790M-mutant HER1, as well as L858R and exon-19 deletion mutant forms of HER1. Because osimertinib spares wild-type HER1, it may have reduced toxicity compared to afatinib and other EGFR inhibitors used in NSCLC. 70,71 Apart from the six small-molecule EGFR Tyr kinase inhibitors already approved, no fewer than four type-VI inhibitors are currently in late-stage clinical development, including dacomatinib and neratinib (pan-HER and T790M HER1), as well as ASP 8273 and rociletinib (T790M HER1-selective). At least another 16 agents, with varying selectivity profiles, are undergoing clinical development (Table S4). (Table S5) The colony-stimulating factor-1 (CSF1)/platelet-derived growth factor (PDGF) Tyr kinase subfamily contains a number of closely related kinases that are relevant to cancer, including CSF1R, VEGFR/FLT, PDGFR, and SCFR (c-Kit). Vascular endothelial growth factors (VEGFs) act on three main receptor Tyr kinase receptors, that is, VEGFR1, VEGFR2, and VEGFR3, whereas PDGFs act on the PDGFRα and PDGFRβ receptors, which contain similar Tyr kinase domains. Both VEGFRs and PDGFRs activate cellular proliferation and survival  227 The phenylacetylene group of erlotinib interacts with the side chain of M766 (magenta stick model) at the base of the αC-helix, as well as the gatekeeper residue T790 (green). (C) Similarly, the HER2-selective compound SYR 127063 binds to HER2 in a way that involves an active-like conformation of that protein. Here, the ligand trifluoromethylphenyl and chloro groups interact with M766 (EGFR numbering) and T790, respectively. 228 It is evident that in active HER2 the inward position of the αC-helix is less pronounced than in active HER1. In active HER2, the cavity proximal to the gatekeeper is thus larger than in active HER1 and can accommodate the larger bis(aryl)ether group of SYR 127063, whereas the phenylacetylene group of erlotinib would be unable to make contacts with the αC-helix in HER2. Conversely the bis(aryl)ether group of SYR 127063 is too large to fit the gatekeeper-αC-helix cavity of active EGFR. (D) Dual HER1 and HER2 inhibitors such as lapatinib are generally observed to bind preferentially to inactive forms of the protein (HER1 shown), in which a short α-helix in the A-loop supports an outward position of the αC-helix. 229 (E) TAK 285 is the only EGFR inhibitor for which crystal complexes with both HER1 (ligand and colored protein elements in light coloring) and HER2 (dark coloring) have been solved. 228,230 Although similar in structure to SYR 127063, TAK 285 inhibits HER1 and HER2 with similar potency. It can be seen that it binds to inactive forms of both kinases, which are highly similar. (F) Second-generation type-VI inhibitors are active against T790M-mutant kinase forms and, for example, afatinib binds to wild-type (ligand and colored protein elements in light coloring) and the T790M form of HER1 (dark coloring) in a very similar manner, with slight repositioning of the substituted aniline to optimize interactions with T790 or T790M. 68 (G) Osimertinib, on the other hand, is a type-VI inhibitor that selectively binds to the T790M-mutant EGFR kinase. It was observed that the main determinant for this selectivity in the design of osimertinib was the indole N-methyl group, which can be observed to interact closely with the thiomethyl group of M790, an interaction that is not possible with T790 (modeled pose from docking using the Schrödinger Glide program with the covalent docking protocol, using the 3IKA HER1 structure as the receptor). 231 Constructed from PDB entries 1M17, 3PP0, 1XKK, 3POZ, 3RCD, 4G5J, and 4G5P.

D. CSF1/PDGF Receptor Subfamily Kinases
signaling pathways, such as the RAF-MAP-MEK-ERK and the PI3K-AKT pathways. These receptors are implicated in tumor angiogenesis, metastasis, and progression. 72,73 All currently approved (and many of the late-stage development) compounds that inhibit CSF1/PDGF Tyr kinases are MTKIs and contain some of the original kinase inhibitor core templates, 74 for example, sunitinib and nintedanib are 3-methylidene-2-oxo-1H-indoles, vandetanib is a 4-(anilino)quinazoline, and pazopanib is a bis(arylamino)pyrimidine. These agents are variously used in the treatment of advanced RCC (sunitinib, axitinib, pazopanib), GIST (sunitinib), advanced or metastatic pancreatic neuroendocrine tumors (sunitinib), advanced soft tissue sarcoma (pazopanib), late-stage medullary thyroid cancer (MTC; vandetanib), and progressive, radio-iodine-refractory differentiated thyroid cancer (lenvatinib). Nintedanib is another approved MTKI with activity against CSF1/PDGF kinases. It received its first global approval not in oncology but for the treatment of IPF. The activity in this indication may be linked to the ability of nintedanib, among other kinases, potently to inhibit FGFR Tyr kinases, as aberrant FGFRs appear to contribute to the pathogenesis of IPF (see below). 75 Nintedanib is also now used for the treatment of advanced, metastatic NSCLC in combination with docetaxel. Numerous additional VEGFR inhibitors with varying selectivity profiles are now being developed. For example, fruquitinib and cediranib are selective pan-VEGFR/PDGFR inhibitors.
Although there is abundant evidence that angiogenesis inhibitors can suppress tumor development in preclinical models, recent findings show that pharmacological as well as genetic VEGFR suppression can in fact alter tumor development by increasing invasion and metastasis. 76,77 In the clinic, the response to VEGFR inhibitors is often limited and followed by disease progression that is at least in part due to tumor escape as a result of VEGFR inhibition. For these reasons, research and clinical trialing of drug combination strategies that may prevent such escape mechanisms are on-going. 78 Closely related to the VEGFR Tyr kinases is FLT3, which is commonly overexpressed in most B-lineage leukemias (acute lymphocytic leukemia, ALL; acute myelogenous leukemia, AML), in subsets of T-cell ALL, and CML in blast crisis. Furthermore, FLT3 activating mutations are present in about a third of AML patients. 79 A number of multitargeted CSF1/PDGF TK inhibitors with activity against FLT3 are being developed, for example, midostaurin, AKN 028, and PLX 7486. Quizartinib, a benzo[d]imidazo[2,1-b]thiazol-2-yl)phenyl)urea compound, is an example of a late-stage selective type-II 80 FLT3 inhibitor being trialed in the treatment of ALL. 81 Unusually for a type-I TK inhibitor, crenolanib has high selectivity for FLT3, SCFR, and PDGFR kinases, including resistance-conferring kinase domain mutant forms. 82 This finding supports the emerging notion that type-II inhibitors may not necessarily be intrinsically more selective than type-I inhibitors, as has been thought in the past. 83 Among the CSF1/PDGF Tyr kinases is SCFR. Oncogenic signaling from mutant KIT is especially common in GIST 84 and currently the standard therapy for unresectable or metastatic GISTs is first-line imatinib, second-line sunitinib, and third-line regorafenib. 85 All three of these agents are MTKIs that were originally developed for alternative indications, but more selective SCFR inhibitors are also now being developed, for example, crenolanib and PLX 9486. The D816V KIT activating mutation occurs in >90% of patients with systemic mastocytosis and whereas imatinib is not efficacious in this indication, dasatinib has been shown to be active. 86 CSF1R is a cell-surface receptor for both macrophage CSF1 and interleukin-34. It plays an important role as a regulator of tissue and tumor-associated macrophages. 87 Because CSF1R activation promotes tumor progression through suppression of the antitumor immune response, as well as promotion of angiogenesis and metastasis, CSF1R inhibitors may be particularly useful in many cancer indications. In fact, many MTKIs from the CSF1/PDGF class also inhibit CSF1R and this activity may contribute to their efficacy. However, selective CSF1R r FISCHER inhibitors are also now under development, for example, pexidartinib and ARRY 382 (undisclosed structure).

E. Fibroblast Growth Factor Receptor Kinases
The four isoforms (FGFR1-4) of this receptor Tyr kinase family are structurally and functionally related to the HER family and bind diverse FGF ligands. FGFR signaling appears to be implicated in the pathogenesis of several cancers and FGFR aberrations are frequent (7.1% of cancers), especially in urothelial, breast, endometrial, and lung cancers. Apart from finding potential therapeutic application against tumors with FGFR genetic alterations, FGFR inhibitors may also be useful to overcome resistance against HER inhibitors, as alternative signaling through the FGFR pathway apparently represents an important escape mechanism when HER signaling is suppressed. Conversely, FGFR kinase inhibitors may be able to potentiate HER inhibition. 88 Many of the approved MTKIs in the CSF1/PDGF class also inhibit FGFR kinases but more selective FGFR kinase inhibitors are also now under development. AZD 4547, infigratinib, and erdafitinib ( Fig. 8A) are potent and selective pan-FGFR inhibitors currently undergoing phase-II trials in a range of cancers. Another seven FGFR kinase inhibitors (structures not disclosed) are currently in phase-I trials (Table S6). At least two of these are irreversible inhibitors (PRN 1371; TAS 120, refer Fig. 8E), whereas another is an isoform-selective FGFR4 inhibitor (FGF 401).
One of the mechanisms of acquired resistance to FGFR inhibition is gatekeeper mutation. Thus, V561M mutation in FGFR1 confers significant resistance to, for example, lucitanib ( Fig. 8C) but not dovitinib (Fig. 8D). Interestingly, AZD 4547 retains affinity for the V561M mutant FGFR1 by adopting an alternative binding mode in which the flexible dimethoxyphenethyl group-which is in contact with the wild-type V561 gatekeeper-is able to bend away from the larger M561 gatekeeper in the mutant kinase. 89

F. Hepatocyte Growth Factor Receptor (HGFR or MET) Kinase
The MET Tyr kinase is a cell-surface receptor for hepatocyte growth factor, whose downstream signaling pathways are important for invasive growth in terms of motility, proliferation, and protection from apoptosis. These processes are important in embryonic development and tissue morphogenesis but often become reactivated in cancer, especially in tumor metastasis. Constitutive MET activation is a feature of many cancers, 32 including NSCL carcinomas with acquired resistance to EGFR inhibitors. 90 MET amplification can be a primary or secondary transformation event and, for example, in some patients with colorectal carcinomas, MET amplification has been observed not in primary tumors, but in liver metastases. 91 The first MET Tyr kinase inhibitor to be approved for the treatment of patients with progressive metastatic MTC is cabozantinib, an MTKI with MET and VEGFR2 as its main targets. 92 A range of functionally different (Fig. 9) selective and MTKI MET inhibitors are currently under clinical evaluation in a number of cancers, including NSCLC resistant to EGFR inhibitors. 93 The most advanced of these is tivantinib, a highly selective MET inhibitor. Secondgeneration inhibitors that are active against mutant forms of MET, such as tepotinib, are also being developed (Table S7). (Table S8) ALK is a receptor Tyr kinase in the IR family and ALK rearrangement is second only to HER mutation as an oncogene addiction in NSCLC. Modification of ALK is present in only a small proportion of NSCLC patients but specific targeting of this population for treatment with ALK inhibitors is particularly effective. 94 The first ALK inhibitor approval in NSCLC was crizotinib, an MTKI originally discovered and developed as a MET inhibitor and now prescribed successfully as first-or second-line therapy in ALK-positive NSCLC. ALK rearrangements also occur in >50% of anaplastic large cell lymphomas (ALCLs) and crizotinib has been shown to have clinical activity in advanced chemoresistant ALCL. 95 r FISCHER  234 is shown in cyan, the αC-helix, which contains key catalytic residues, is shown in magenta, and the A-loop, which is important for substrate recognition, in yellow, with the Phe residue of the DFG motif shown as a stick model. An example of an inhibitor that is selective for A-loopphosphorylated active form of MET is the derivative of MK 2461. It is a type-I inhibitor and the shape of the ATP-binding site in its complexed structure (C) is similar to that of the apo-form, except that the G-loop folds over the ligand. 234 The late-stage development MET inhibitor tivantinib is an unusual type-II inhibitor, which does not bind deep into the cavity formed by the DFG-out conformation, but it induces a major shape change in the kinase, involving both the A-loop and the αC-helix (D). 235 Like tivantinib, merestinib is selective for unphosphorylated inactive MET; it is a conventional type-II inhibitor (E). 236 Constructed from PDB entries 3Q6U, 3Q6W, 3RHK, and 4EEV.

G. Anaplastic Lymphoma Kinase
The development of acquired resistance is the main reason why patients with ALK-positive NSCLC eventually relapse on crizotinib. Secondary resistance mutations in the ATP-binding site have been identified 96 and one of the main drivers for the development of second-generation ALK inhibitors has been to overcome crizotinib resistance. 97 Ceritinib was approved later and is a more potent ALK inhibitor that overcomes the most common secondary mutation to crizotinib (gatekeeper L1196M), as does the recently approved selective ALK inhibitor alectinib.
These second-generation ALK inhibitors also show activity against brain metastases, which is important, as almost half of ALK-positive NSCLC patients on crizotinib exhibit such metastases, even when systemic disease is under control. 98 Three more third-generation ALK inhibitors are currently undergoing clinical evaluation: brigatinib, belizatinib, and lorlatinib. At least the latter compound was designed specifically to have high CNS penetrance (Fig. 10). 99

H. IGF1R and IR
The insulin-like growth factor 1 (IGF1) signaling pathway regulates many aspects of the proliferation, differentiation, and growth of normal cells. However, its receptor, IGF1R, as well as its ligands and adaptor proteins are highly expressed in many malignancies. Furthermore, the IGF1 pathway is also implicated in carcinogenesis and cancer risk. IGF1R activates the RAS-RAF-MAPK and PI3K-AKT-mTOR pathways, both of which are associated with cancer cell proliferation and survival. For these reasons, IGF1R inhibitors have been developed, especially in NSCLC. 100 The first generation of IGF1R inhibitors that have been tested in clinical trials are humanized monoclonal antibodies that block ligand binding to IGF1R. Although early results were promising, later studies using combinations with chemotherapy or EGFR inhibitors in NSCLC have been disappointing. 101,102 Currently, the only clinical small-molecule IGF1R inhibitor targeted at the Tyr kinase activity of this receptor is the phenylquinoline compound linsitinib, a dual IGF1R and IR inhibitor (Table S29). (Table S9) TRK was originally identified as an oncogene product that arises as a result of mutations and rearrangements of the NTRK genes and has been detected in various tumors, especially in the colon, lung, and thyroid gland. 103 The three isoforms of wild-type TRK are expressed predominantly in neurons, however, where their activation by neurotrophins plays important roles in cell development, maintenance, and function. It is now clear that TRKs also regulate processes in nonneuronal cells. Furthermore, activation of cell survival pathways through TRKs contributes to tumorigenesis and resistance to cytotoxic chemotherapy not only in neuroblastoma, but also in many other cancers. 104 The TRKs belong to the same IR Tyr kinase subfamily as ALK and ROS, and entrectinib, a pan-TRK inhibitor with additional potent ALK and ROS inhibitory potency, is currently undergoing phase-II trials in colorectal carcinoma and NSCLC. 105 Another agent at a similar stage of development is LOXO 101, a selective pan-TRK inhibitor. Entrectinib and LOXO 101 are structurally related, the former contains an indazole, the latter a pyrazolo[1,5-a]pyrimidine core, both compounds containing a meta-difluorophenyl substituent.

I. Tropomyosin Receptor Kinase
AXL belongs to the TYRO3, AXL, and MERTK (TAM) family of receptor Tyr kinases and although it does not appear to be a strong oncogenic driver, it is overexpressed in many tumors and overexpression correlates with poor prognosis. Aberrant TAM signaling in cancer suggests that AXL inhibitors may confer antitumor immunity and tumor cell survival, as well as enhance chemosensitivity and suppress metastases. 106 It is somewhat difficult to decide which development compounds should be designated as AXL inhibitors, since several MTKIs, including approved agents such as cabozantinib, bosutinib, crizotinib, and sunitinib, are also potent AXL inhibitors. 107 One of the most potent AXL inhibitors (sub-nM IC 50 ) is the pyrazinecarboxamide gilteritinib, which also potently inhibits FLT3 and ALK. 108 The first selective (at least in terms of cell-based kinase inhibition activity) AXL inhibitor to enter the clinic is BGB 324, a substituted triazole-3,5-diamine that contains two instances of the unusual tetrahydrobenzannulene substructure (Table S10).
REarranged during Transfection (RET) is a receptor Tyr kinase whose activity supports cell survival and proliferation, as well as cell migration, differentiation, and chemotaxis. Activating mutations in RET (C634W and M918T) and a number of different gene translocations leading to oncogenic fusions involving RET occur in various cancers, particularly in MTC and NSCLC. 109 RET is closely related to other receptor Tyr kinases, especially VEGFR, and many MTKIs possess RET-inhibitory activity (refer Tables S5, S7, S9). As we have seen, vandetanib, lenvatinib, and cabozantinib are approved for use in MTC and the activity of these agents in thyroid cancers-as well as in some lung cancers-is probably due to a large extent because they inhibit RET. [110][111][112][113] However, the efficacy of MTKIs, especially those that also block VEGFR activity, is limited by toxicities 114 and selective RET inhibitors are highly desirable for the treatment of cancers that are driven by RET activation. Although research aimed at the discovery of such agents is underway, 115,116 there do not appear to be any RET-specific kinase inhibitors under clinical evaluation at present.

J. Transforming Growth Factor-β Receptor
A range of cellular responses, including proliferation, differentiation, motility, and survival, are elicited upon activation by the TGF-β cytokines of the three TGFR isoforms through receptor hetero-dimerization. However, in normal cells TGF-β actually functions as a tumor suppressor but TGF-β signaling is high-jacked by cells during tumor development and is particularly important during the epithelial-mesenchymal transition. 117 TGFRs contain Ser/Thr kinase domains and signal transduction occurs through TGFR heterodimer-mediated phosphorylation of cellular proteins such as SMADs, which activate the TGF-β signaling pathway. 118 Drugs, including antisense oligonucleotides, antibodies, and small molecules targeting this pathway, have been under development in both neoplastic and nonneoplastic indications, especially in fibrosis. 119 Two TGFR1 kinase inhibitors are currently being developed in oncology, galunisertib and TEW 7197, both of which contain the 2-methyl-6-(4H,5H,6H-pyrrolo[1,2b]pyrazol-2-yl)pyridine substructure (Table S11).

K. Janus Kinases
The JAK family includes the three cytoplasmic Tyr kinases JAK1, JAK2, and JAK3, as well as the nonreceptor Tyr kinase TYK2. These kinases transmit signaling from a number of cytokines and growth factors important in hematopoiesis and cellular immune functions. Although JAK3, which is critical to signal transduction via common γ -chains of interleukin receptors, is pursued primarily as a drug target in inflammation, especially JAK2 is also an oncology drug target. The interest in JAK2 inhibitors stems from the discovery that many myeloproliferative neoplasms display an activating mutation (V617F) in the JAK2 gene. 120 Ruxolitinib, a JAK1/2-selective compound, 121 is the first JAK inhibitor in oncology, approved for the treatment of myelofibrosis. A recent survey of the clinical use of this agent concluded that ruxolitinib, currently the only approved therapy for myelofibrosis, is associated with a survival benefit and has changed the treatment landscape for this disease. 122 Nevertheless, clinical responses have been attenuated and the original hope of curative effects with JAK2 inhibition similar to those seen with imatinib in CML has not been fulfilled. It has recently been suggested that this may be due to an intrinsic resistance to JAK2 inhibition in myelofibrosis. 123 A number of other (structurally diverse) compounds (Table S12), including pacritinib (JAK2/FLT3-selective), momelotinib (JAK1-3-selective), are also currently undergoing clinical development in myeloproliferative neoplasms. 124 It is not as yet clear how these agents will be differentiated from ruxolitinib and if they will have to be compared to ruxolitinib in randomized studies for the purposes of approval. 122

L. Bruton Tyrosine Kinase
BTK is a nonreceptor kinase in the B-cell antigen receptor (BCR) signaling cascade that mediates B-cell activation, proliferation, migration, and survival. In B-cell non-Hodgkin's lymphomas (NHL), constitutive BCR activation through BTK supports cell proliferation and survival, rendering BTK a relevant pharmacological target in NHL. This is especially attractive in terms of specificity, since the phenotype of loss-of-function BTK, as occurs in the human disease X-linked agammaglobulinemia, is restricted to B cells. 125 The first potent and selective type-VI inhibitors of BTK were reported in 2007 and it took only 6 years before the first inhibitor of this type, ibrutinib, obtained initial global regulatory approval in the treatment of mantle cell lymphoma. The first BTK inhibitors of any kind were actually reported already in 1999: terreic acid (a quinone epoxide) 126 and leflunomide metabolite analogues (2-cyano-3-hydroxybut-2-enamides). 127 Since these compounds possess highly electrophilic functions, it is likely that they are also type-VI inhibitors that react covalently with the thiol of C481, an unpaired Cys residue in the ATP-binding site of BTK (corresponding to C773 in HER kinases), although this was not apparently realized at the time of their discovery. Interestingly, terreic acid was originally studied in the context of mast cell activation and it was realized that BTK inhibitors may find therapeutic application in immunology and inflammation, whereas leflunomide metabolite analogues were reported as antileukemic drug leads. Ibrutinib, too, was discovered in a rheumatoid arthritis drug discovery program, 128 and was later repositioned in B-cell malignancies. 129 Apart from ibrutinib (a pyrazolo [3,4-d]pyrimidine) there are currently five additional BTK inhibitors under clinical development in cancer (Table S13). Of these, the chemical structures of two are in the public domain and both are structurally closely related to ibutinib. Acalabrutinib is an imidazo[1,5-a]pyrazine derivative with a reactive but-2-enoyl function, whereas GS 4059 is a purin-8-one derivative with-like ibrutinib-an acryloyl electrophile.

M. Focal Adhesion Kinase
FAK1 is a cytoplasmic nonreceptor Tyr kinase that transmits cell motility, survival, and proliferation signals from integrin-mediated cell extracellular matrix attachment and growth factor receptors. Because metastatic propensity and poor clinical outcomes of many cancers are correlated with enhanced FAK1 expression and activity, FAK1 is a promising target for cancer therapy. 130 A number of FAK1 inhibitors have been under clinical investigation in the past, 131 but currently only three remain active (Table S14). GSK 2256098 and VS 4718 are selective FAK1 inhibitors, whereas defactinib is a dual FAK1 and FAK2 inhibitor. Such dual selectivity is believed to be potentially beneficial due to functional overlap and redundancy of these two closely related kinases. 132 All three compounds are structurally related ATP-competitive (presumably type-I) inhibitors, but allosteric FAK1 inhibitors have also been reported, 133 although these have not yet reached clinical evaluation.

N. Spleen Tyrosine Kinase
SYK is a cytoplasmic nonreceptor Tyr kinase expressed predominantly in hematopoietic cells. Phosphorylation of cytoplasmic immunoreceptor tyrosine-based activation motifs leads to recruitment of SYK, which propagates signaling by activation of downstream pathways, including PI3K, MAPK, and BTK. 134 Because SYK mediates cellular responses to antigens and antigen-immunoglobulin complexes, it is especially relevant as a drug target to suppress the consequences of acute and chronic inflammation. 134 However, SYK, together with BTK, is also involved in dysregulation of BCR signaling in hematological malignancies such as diffuse large B-cell lymphoma (DLBCL), follicular lymphoma, and chronic lymphocytic leukemia (CLL). 135 A number of SYK inhibitors have been evaluated clinically, mainly for the treatment of immune disorders, but were found to exhibit dose-limiting toxicities attributed predominantly to lack of sufficient selectivity for SYK. 136 The first of these was fostamatinib (a methylene phosphate oral prodrug of a poorly soluble dianilinopyrimidine compound), which failed in phase-III trials in rheumatoid arthritis and has also been trialed in B-and T-cell lymphoma, but apparently has not progressed. 137 Second-generation SYK inhibitors with improved kinome selectivity profiles have now been developed and entered the clinic (Table S15). The most advanced of these is entospletinib, a highly SYK-selective oral disubstituted imidazo[1,2a]pyrazine compound, currently in phase-II studies in both inflammation and cancer (hematological malignancies). Cerdulatinib is a dual JAK and SYK inhibitor (Table S12) that has activity in autoimmunity and B-cell cancers like other SYK inhibitors. 138 (Table S16) The PI3K pathway regulates protein synthesis, cell growth, and cell proliferation in response to nutrient availability and mitogenic growth factors. Activation of this pathway has long been associated with malignant transformation and antiapoptotic signaling. 139 Amplification or mutation of many receptor Tyr kinases, mutation of downstream signaling components, including PI3K itself, loss or mutation of the PTEN tumor suppressor protein, as well as mutation and amplification of AKT, can lead to prosurvival signaling through the PI3K-AKT-mTOR-eIF4E pathway in cancer cells. At least four kinases in this pathway, PI3K itself (i.e., the class-I PI3K isoforms α-δ), AKT, its activating kinase PDK1, and mTOR, are being actively pursued as oncology drug targets.

O. PI3K Pathway Kinases
The first-generation mTOR inhibitors are derived from the macrolide rapamycin (sirolimus), a natural product from a fungus indigenous to the Easter Island (Rapa Nui). Studies based on the observation that rapamycin inhibited the proliferation of T cells led to the discovery of the mTOR (mechanistic or mammalian target of rapamycin) genes and the development of sirolimus as a targeted drug. Although the anticancer activities of rapamycin had been known for a long time, sirolimus was first developed as an immunosuppressive agent. However, a rapalogue (rapamycin analogue), temsirolimus, was approved for the use in advanced RCC. Objective responses of kidney cancer patients to rapalogues had been observed already in phase-I studies 140 and the demonstration that loss of the von Hippel-Lindau tumor suppressor gene (VHL) sensitizes kidney cancer cells to mTOR inhibitors subsequently enabled biomarker-driven development of temsirolimus in RCC. 141 Another rapalogue, everolimus, was approved later for the same indication.
mTOR participates in at least two functionally distinct multiprotein complexes termed mTORC1/RAPTOR and mTORC2/RICTOR. Rapalogues, which block mTOR activity indirectly in complex with the immunophilin receptor known as FK506-binding protein-12 (FKBP12), specifically inhibit mTORC1. 142 It was reported recently that the rapalogue everolimus can also act as a MET inhibitor (refer section on HGFR kinase), since METlike mTOR-also requires FKBP12 binding for full activity. 143 Because of the presence of mTORC1-dependent negative feedback loops via the mTOR phosphorylation substrate ribosomal protein S6 kinase β-1 leading to upregulation of PI3K and RAS activities, inhibition of mTORC1 can lead to activation of AKT-and MAPK-dependent survival pathways. 144 For this reason, selective ATP-antagonist mTOR inhibitors, which block both mTOR complexes, were developed and are also now entering clinical trials. Clinical compounds belonging to this group include vistusertib, sapanisertib, and CC 223.
Unsurprisingly considering the similarity of mTOR and PI3K, the majority of ATPantagonist mTOR inhibitors turned out to inhibit PI3K as well, or they were designed to possess dual specificity. Again several compounds of this type are being developed, for example, the dual pan-class-I PI3K and mTOR kinase inhibitors voxtalisib, LY 3023414, PQR 309, gedatolisib, and VS 5584.
Inhibitors that do not modulate mTOR at all but inhibit various combinations of class-I PI3K isoforms are also now being pursued. This appears to be a valid approach as all four functionally nonredundant isoforms can generate the signaling molecule phosphatidylinositol-3,4,5-trisphosphate (PIP 3 ), which recruits AKT to the cell membrane, and because all are oncogenic in model systems. 145 Furthermore, all four isoforms are frequently genetically altered in a variety of cancers. 146 At present, it remains unclear what PI3K/mTOR inhibition selectivity (Fig. 11) will be therapeutically optimal. It is likely that the preferred class-I PI3K target will depend on the cellular context. 147 The main driver for seeking isoform-selective PI3K inhibitors is probably because broad selectivity may limit the therapeutic margin of an mTOR/PI3K inhibitor, due to the important role of the PI3K pathways in normal cells. Currently, there are several pan-class-I PI3K inhibitors undergoing clinical trials, including buparlisip, copanlisip, pictilisib, ZSTK 474, and SF 1126, the latter being a vascularly targeted prodrug.
That PI3Kδ is only expressed in hematopoietic cells provides the therapeutic rationale for the use of the PI3Kδ-selective compounds in hematological cancers 148 and this concept has now led to the first approval of a PI3Kδ-selective inhibitor, idelalisib as monotherapy for follicular and small lymphocytic lymphoma. Another compound, duvelisib, with a similar selectivity profile (tenfold to 40-fold selectivity over PI3Kγ and higher over other PI3Ks) is in late-stage development. Other clinical compounds with more or less-pronounced PI3Kδ selectivity include AMG 319, HMPL 689, INCB 050465, and CDZ 173. A recent study showed that PI3Kδ (and PI3Kγ ) is important for T-cell-mediated immune tolerance to cancer and PI3Kδ inhibitors may therefore find therapeutic applications not just in blood cancers but perhaps much more widely. 149 Apart from PI3Kδ-selective inhibitors, a growing number of experimental PI3K inhibitor drugs with a range of PI3K isoform selectivity profiles are being investigated. The interest in PI3Kα-selective agents stems predominantly from the observation that the gene encoding this isoform is mutated particularly frequently in certain cancers (e.g., HER2-and KRAS-driven tumors). 150 The most advanced PI3Kα-selective compounds are alpelisib and talelisib; others include MLN 1117 and AZD 8835. Because PI3Kα specifically regulates glucose hemostasis, r FISCHER The PI3Kδ-selective idelalisib 238 (green) binds to PI3Kδ through H-bonding interactions (broken lines) with the backbone amide bonds of the hinge region (cyan). Its selectivity derives predominantly from induced-fit binding of the quinazolinone substructure between the side chains of M752 and W760 (spheres). (C) An apo-structure of PI3Kδ 239 shows that the residues that differ between the four PI3K isoforms (magenta) are quite remote from the ligand in the PI3Kδ-idelalisib complex (and other ligands in other complexes) and that the slot between the M752 and W760 residues is closed. Although the residues corresponding to M752 and W760 are conserved in all four PI3K isoforms, induced-fit binding as observed with idalalisib appears to be a selectivity feature of PI3Kδ-selective compounds. 239 (D) The PI3Kα-selective alpelisib 240 interacts closely (broken lines) with the side chain of Q859, one of the few residues around the ATP-binding pocket that are not conserved across the PI3Ks. (E) The pan-PI3K-selective buparlisip 241 binds to PI3Kγ in a similar mode as alpelisib binds to PI3Kδ but any selectivity-imparting features are absent. (F) The mTOR-selective compound torin 2 242 binds to mTOR without any of the usual hinge-region H-bonds. Its selectivity is probably due in large part to interaction of the naphthyridinone substructure with the indole of W2239 in the hinge, which is not conserved in PI3Ks. 243

(G)
The dual mTOR-and pan-PI3K inhibitor PI 103 244 interacts with mTOR through polar interactions involving the morpoline O and the phenol OH groups. 243 Similar interactions are also possible with any of the PI3K isoforms. Furthermore, PI 103 adopts an almost completely planar conformation, which can be accommodated in the mTOR and the PI3K ATP-binding sites, regardless of the width of the binding clefts. Constructed from PDB entries 4XE0, 2WXR, 4JPS, 3SD5, 4JSX, and 4JT6. this must be carefully managed with PI3Kα-selective inhibitors in the clinical setting. 151 Dependence on PI3Kβ has been observed in certain PTEN-deficient tumors and the first PI3Kβselective agent, GSK 2636771 is now being trialed in this setting. Other PI3Kβ-selective clinical compounds include AZD 8186 and KA 2237.
Because of the importance of the PI3K pathway to cancer cell survival, it was originally thought that PI3K inhibitors might have anticancer efficacy in monotherapy, but with the clinical experience amassed so far with over 20 different PI3K/mTOR inhibitors, it is increasingly becoming clear that this is unlikely to be the case. Like most kinase inhibitors, PI3K inhibitors will probably work best in combination with other anticancer drugs. 152 AKT (also known as PKB) is a Ser/Thr kinase that acts downstream of PI3K and is activated by recruitment to the cell membrane by interaction with PIP 3 through its pleckstrin homology (PH) domain. Once activated, AKT mediates downstream responses by phosphorylation of multiple proteins. The lipid phosphatase PTEN, which is frequently inactivated in human tumors, negatively regulates AKT by dephosphorylation of PIP 3 , whereas AKT is amplified or overexpressed in many tumors. 153 Until recently, the clinically most advanced AKT inhibitor was the alkylphospholipid perifosine. It is not a direct kinase inhibitor but interferes with membrane localization of AKT. 154 Phase-II clinical trials of perifosine as a single agent in several tumor types failed to show objective responses. 153 Although promising activity in patients with RCC, who have failed prior Tyr kinase inhibitor therapy, was reported, 155 perifosine does not appear to be under development any longer. Several type-I pan-AKT inhibitors, including AZD 5363, ipatasertib (Fig. 12A, C), LY 2780301, and afuresertib, are now in phase-II clinical studies and several more compounds are in early stage trials (Table S17), including the allosteric (type-III) AKT inhibitor MK 2206 (Fig. 12A). This compound binds to an AKT site formed only in the presence of the PH domain and binding of the inhibitor promotes the formation of an inactive kinase conformation. 156 MK 2206 is similar to compound VIII of ref. 157, whose experimental binding mode is known (Fig. 12B, C). ARQ 092 and BAY 1125976 are also allosteric AKT inhibitors. The structure of the latter compound has not yet been disclosed but the compound series that encompasses ARQ 092 is structurally related to compound VIII and MK 2206. 158 PDK1 plays an important role in the PI3K-AKT pathway. It activates AKT, as well as a number of other related AGC kinases, thus controlling several signaling pathways important in proliferation, apoptosis, and angiogenesis. Like PI3K, PDK1 is upregulated in more than half of all tumors due to constitutive growth factor receptor activation and PTEN mutations. As expected, transgenic mice expressing low levels of PDK1 are protected from tumorigenesis that normally results from loss of PTEN. The search for potent and selective PDK1 inhibitors has been pursued for some time but as with many other kinases, achieving selectivity in inhibitors has proven challenging. 159 Nevertheless, the first PDK1 inhibitor, AR 12, recently entered clinical trials (Table S29). This compound was designed from celecoxib, a cyclooxygenase-2 inhibitor, following the assignation of tumor-preventive properties of celecoxib to PDK inhibition. However, the activity and selectivity with which AR 12 inhibits PDK1 is actually uncertain. 159

P. Protein Kinase C
PKC overexpression has been linked to several types of cancer, with the PKCβ isoform believed to be involved in VEGF-induced tumor development and angiogenesis and in the apoptosisregulating PI3K-AKT pathway. The macrocyclic bisindolylmaleimide enzastaurin selectively inhibits PKCβ at low concentrations, and also inhibits other PKC isozymes at higher concentrations that are reached or surpassed in clinical trials. 160 Antitumor activity of enzastaurin has been primarily attributed to the inhibition of AKT and its downstream targets, but its precise mechanism of action is not well understood. 161 Both PKC and AKT are activated by PDK, and several PKC isozymes can also directly phosphorylate AKT at S473, which is essential for AKT activity. It was recently reported that a phase-III clinical trial of enzastaurin in monotherapy in DLBCL failed to show a statistically significant increase compared to placebo in disease-free survival in patients at high risk of relapse following rituximab-based chemotherapy, 162 and apparently development of this compound has been discontinued.
Enzastaurin is not the first staurosporin derivative to undergo clinical development, for example, UCN-01 (7-hydroxystaurosporin) and midostaurin, which also inhibit PKC but have been developed as MTKIs (main targets are CHK1 and PDK1 for UCN-01, and FLT3 and other CSF-1/PDGF receptor subfamily kinases for midostaurin). 163,164 Other staurosporine derivatives are being developed in nononcology indications. 165 The only PKC inhibitor that remains under clinical study appears to be LXS 196, an oral compound whose structure or properties have not yet been disclosed (Table S29).

Q. Choline Kinase-α
CK is a lipid kinase that phosphorylates free choline to give phosphocholine, enabling the formation of phosphatidylcholine, the main phospholipid in mammalian cell membranes, through the Kennedy pathway. Many tumors display altered lipid metabolism during development and as a consequence of chemotherapy CK is upregulated in many tumors and is associated with various malignant phenotypes. 166 Furthermore, CK was recently reported as an androgen receptor chaperone and potentially valuable drug target in prostate cancer. 167,168 TCD 717 is the first CK inhibitor to have recently entered clinical trials in solid tumors (Table S29). It inhibits CK by virtue of competing with the substrate choline; its structure has not been disclosed but is related to hemicholinium-3 ((2S,2 S)-2,2 -biphenyl-4,4 -diylbis(2-hydroxy-4,4dimethylmorpholin-4-ium), a choline re-uptake inhibitor. 169 Casein kinase II is a constitutively active hetero-tetrameric Ser/Thr kinase that sustains many cellular signaling pathways. Although CK II function is not apparently altered in cancer cells by mutation, it is nevertheless overexpressed in many different cancers. 170 Tumor cells often exhibit CK II-driven nononcogene addiction 171 and it has been shown that ablation of CK II through antisense oligonucleotides has profound antiproliferative and proapoptotic effects in a rodent xenograft model. 172 A number of small-molecule CK II inhibitors with anticancer activity have been reported and the first such compound, silmitasertib, an CK II-selective ATP-competitive benzo[c] [2,6]naphthyridine-8-carboxylic acid compound, has recently entered clinical evaluation (Table S29). 173 MELK is an AMPK family kinase with several cellular functions related to survival and proliferation and it is overexpressed in many tumors. These functions appear to be especially important in cancer stem cells, that is, undifferentiated cancer cells. The first clinical MELK inhibitor is OTS 167, a disubstituted (naphthyridin-3-yl)ethanone compound, whose binding mode was reported recently. 174 Another clinical compound, the (indol-3-ylidene)methyl)pyrrole amcasertib, also targets (undisclosed) cancer stem cell pathway kinases (Table S29).

R. PIM
The proviral insertion in murine (PIM) lymphoma family of Ser/Thr kinases (PIM1-3) is expressed predominantly in hematopoietic cells and signals downstream from ABL, JAK2, and FLT3. PIM kinases contribute to the regulation of the cell cycle, apoptosis, and the proliferation and migration of cells. All three isoforms have oncogenic potential and are aberrantly expressed in various tumors. 175 Although mouse knock-out of any of the three PIM kinases results in mild phenotypes, deficient PIM1 signaling is associated with cardiac function and PIM triple knock-out mice develop heart failure by 6 months. These observations may be relevant to the withdrawal of the first-in-class clinical PIM inhibitor SGI 1776 due to cardiotoxicity. 176 Despite this observation, it would appear that both PIM inhibitors currently undergoing clinical development, that is, INCB 053914 and PIM 447 (Table S18) are pan-selective PIM inhibitors. PIM kinases are somewhat unusual insofar as they recognize ATP, as well as most inhibitors, in a manner that does not involve the typical H-boding interactions between the kinase hinge region and the ligand, due to the presence of two Pro residues in the PIM hinge, which imparts an atypical conformation to this region. 177 This unusual recognition mode may explain the high kinome-wide selectivity of compounds such as PIM 447.

S. Mitogen-Activated Protein Kinase Pathway Kinases
The RAS-RAF-MEK-ERK MAPK pathway is one of the main signaling pathways that cancer cells use for the purposes of proliferative and survival advantages. Activation of the MAPK pathway, through activating mutations in either NRAS or BRAF, is particularly common in melanoma. 178 Below we shall discuss clinical RAF (Table S19) and MEK inhibitors (Table S20), which have recently been joined by the first ERK inhibitors (Table S21).
Currently, there are two RAF kinase inhibitors approved for the treatment of unresectable or metastatic melanoma with BRAF V600E mutation, vemurafenib and dabrafenib. Both compounds display selectivity for V600E BRAF over other kinases, and to a lesser extent over wild-type RAF kinase isoforms. 179,180 With these compounds, inhibition of the MAPK pathway is specific to tumor cells with mutant BRAF and this is likely due to their selectivity. In the majority of melanoma patients treated with vemurafenib or dabrafenib significant tumor regression and progression-free survival have been observed, a result that compares favorably with previously available therapies for metastatic melanoma. 178 Neither agent is indicated for the treatment of patients with wild-type BRAF melanoma due to the potential risk of tumor promotion, and both were approved concurrently with assays for the detection of BRAF V600E mutations. This contraindication is due to paradoxical activation of the MAPK pathway by RAF inhibitors in cells with wild-type BRAF. MAPK pathway activation occurs as a result of RAS-dependent MEK phosphorylation by homo-and hetero-dimerized RAF isoforms. In mutant BRAF melanoma cells, in which MEK activation is driven by BRAF, RAF inhibition is effective because of increased affinity of BRAF V600E for ATP and lack of CRAF activation. Paradoxical MAPK pathway activation in cells with normal BRAF, where RAS signaling occurs predominantly through CRAF, is thought to arise due to binding of type-I RAF inhibitors to wild-type RAF isoforms, which leads to their loss of auto-inhibition in a manner that is independent of kinase inhibition, enhanced RAF dimerization, membrane localization, and eventually RAS-dependent MEK activation. 181,182 The same phenomenon, that is, paradoxical MAPK pathway activation in healthy keratinocytes, is probably responsible for the significant cutaneous toxicities observed with RAF inhibitors. 183 Despite the approval of vemurafenib and dabrafenib, a whole range of additional RAF kinase inhibitors are currently being developed in advanced melanoma and other solid tumors (Table S19). 184 The first RAF inhibitor studied in melanoma, however, was sorafenib, a type-II MTKI with pan-RAF activity. This compound, which also has activity against several CSF1/PDGF receptor subfamily kinases (Table S5), was being developed-and is now approved in-advanced RCC at the time when the BRAF V600E mutation, which is present in over half of all malignant melanomas, was first discovered. 185 Despite several clinical trials of sorafenib in melanoma, it was not possible to show significant patient benefit, however. Presumably, this was because at the maximum tolerated dose, defined by toxicities emanating from inhibition of kinases other than RAF, there was insufficient MAPK pathway inhibition. 178 The first MEK inhibitor (Table S20) approval was trametinib in the treatment of patients with unresectable or metastatic melanoma with BRAF V600E or BRAF V600K mutation as detected by an FDA-approved test. This offers another treatment modality in this disease, although it is not as yet clear what the best treatment options with RAF and MEK inhibitors are. 186 Nevertheless, acquired resistance to RAF inhibitors occurs frequently and through multiple genetic mechanisms, which suggests that simultaneous inhibition of MEK and mutant RAF kinases may be a better strategy in terms of overall efficacy, prevention of MEK-driven acquired resistance, and modulation of toxicity emanating from paradoxical MAPK-pathway activation with RAF-inhibitor monotherapy. 187 On this basis, cobimetinib was recently approved for the treatment of mutant BRAF melanoma in combination with vemurafenib.
Although treatment of advanced melanoma with both RAF and MEK inhibitors has been observed to be effective initially, resistance to both inhibitor types eventually occurs. 189 The main mechanisms of acquired resistance appear to be reactivation of the MAPK pathway in a number of ways, as well as compensatory signaling through the parallel PI3K pathway. 178 The latter observation has led to the initiation of clinical trials in melanoma and other solid tumors, where RAF/MEK and PI3K inhibitors are used in combination. 190 Direct targeting of ERK1/2 in order to block signaling through the RAS-RAF-MEK-ERK pathway has lagged behind the upstream inhibition strategies, but may be of benefit in several ways. 191 Thus, it has been demonstrated that dual pharmacological inhibition of MEK and ERK can be synergistic by inhibiting the emergence of resistance and overcoming acquired resistance to MEK inhibitors. 192 There are currently three ERK1/2 inhibitors under clinical evaluation (Table S21), the most advanced being ulixertinib. Both ulixertinib and GDC 0994 contain the N-((1S)-2-hydroxy-1-phenylethyl)formamide substructure, whereas the chemical structure of CC 90003 has not been disclosed.

T. p38 MAPK
Several different MAPK signaling pathways are used by cells to respond to extracellular stimuli. Apart from the RAS-RAF-MEK-ERK we have just discussed, another important MAPK Figure 13. Structural basis of MEK inhibition. MEK1 and MEK2 are comparatively unique insofar as these kinases contain an allosteric-binding site adjacent to, but not overlapping, with the ATP-binding pocket. This exosite is evident in ATP/Mg 2+ -bound structures (A), as well as in complexes with type-I inhibitors (e.g., K252a, B). 245 A number of ATP noncompetitive MEK inhibitors have been discovered that occupy this exosite, thus distorting the catalytic site and inactivating the enzyme (e.g., rafametinib (gray), superposed with ATP (cyan), C). 246 The 2-fluoro-4-iodo-aniline system of rafametinib has high shape complementarity with the lipophilic back-pocket of the exosite (D) and is in fact present in most known type-III MEK inhibitors, including seven of eight current clinical MEK inhibitors (E). Protein structures are shown as green surfaces or meshes; ATP and related nucleosides are shown as cyan sticks (Mg 2+ as green spheres), allosteric inhibitors as gray sticks, and K252a as yellow sticks. Constructed from PDB entries 1S9J, 2P55, 3DV3, 3DY7, 3E8N, 3EQB, 3EQC, 3EQD, 3EQF, 3EQG, 3EQH, 3EQI, 3MBL, 3ORN, 3OS3, 3PP1, 3SLS, 3V01, 3V04, 4AN2, 4AN3, 4AN9, 4ANB, and 1S9I. pathway is the p38 MAPK pathway, which mediates stress responses and responses to cytokines, chemokines, hormones, and growth factors. The four p38 Ser/Thr kinase isoforms p38α (MAPK14), p38β (MAPK11), p38δ (MAPK13), and p38γ (MAPK12) can be activated in a number of ways involving both posttranslational modification and protein interactions, thus integrating many different signals. In turn, the p38 kinases have numerous downstream targets and can activate different transcriptional programs in a spatially and temporally controlled manner. 193 p38 MAPK is pursued mainly as a drug target in chronic inflammatory disorders, especially p38α, which is overactivated in inflamed tissues. Many different p38 MAPK inhibitors are being developed in such disorders, including rheumatoid arthritis, atherosclerosis, and asthma. 194,195 However, p38 MAPKs also play important roles in the responses of cancer cells to oncogenic stress, radiation, and chemotherapy and altered expression of p38 MAPKs is frequently observed in tumors, where they promote tumorigenesis. 196 Currently, two structurally related p38 MAPK inhibitors are under clinical investigation in oncology indications: ralimetinib and pexmetinib (Table S22).

U. MNK
The Ser/Thr kinases MNK1 and MNK2 are activated through the mitogen-and stress-activated MAP kinase pathways and they phosphorylate the eukaryotic translation initiation factor 4E (eIF4E). Genetic studies indicate that whereas MNK activity is required for eIF4E-mediated oncogenic cell transformation, such activity is dispensable for normal development. 197 Until recently, very few MNK inhibitors were known but selective type-I/II and type-VI MNK inhibitors have now been reported. 198,199 The first MNK inhibitor to reach phase-I clinical trial is BAY 1143269 (Table S29; undisclosed structure). (Table S23) The CDKs comprise a large group of Ser/Thr kinases involved predominantly in the control of the cell cycle and the regulation of transcription. Recruitment into and progression through the cell cycle are tightly regulated and the interphase CDKs control the first gap phase (G 1 ; CDK4, CDK6, CDK1, CDK2) by activating E2F-responsive genes required for entry to the DNA synthesis phase (S-phase), as well as the second gap phase (G 2 ; CDK1, CDK2) by activating the FoxM1 gene program required for execution of mitosis. Similarly, the different phases of RNA polymerase II (RNAPII)-based transcription are regulated by CDKs. Here, sequential phosphorylation of the C-terminal domain of RNAPII occurs: by CDK7 and CDK8 in transcription initiation, and by CDK9 in RNA elongation. Additionally, CDK activity is required (at least CDK11) for RNA processing (splicing). 200 However, our understanding of CDK biology is still incomplete due to extensive functional overlap and redundancy between the many CDK-cyclin pairs, not only in the cell cycle and RNAPII transcription, but also between the two.

V. Cyclin-Dependent Kinases
Because cancer is characterized by uncontrolled cell division, interference with CDKs as key regulators of the cell cycle appears attractive as a therapeutic strategy. However, the complexity of CDK-cyclin biology has confounded interpretable genetic target validation of CDKs. As a result uncertainty exists about which CDK function(s) should be blocked to achieve effective and specific anticancer activity in a therapeutic setting, and this has long beset the development of CDK inhibitor drugs. A large number of CDK inhibitors had been trialed clinically in the past-mostly with disappointing results. 201 Nevertheless, the first CDK inhibitor, palbociclib, has now been granted FDA approval for use in combination with letrozole (a nonsteroidal aromatase inhibitor) in ER-positive, HER2negative metastatic breast cancer. Palbociclib is a highly CDK4/6-selective agent and preclinical studies with this and several other CDK4/6-selective compounds show anticancer activity in a range of models, with this activity being contingent on the presence of functional retinoblastoma protein (pRb; the CDK4/6 phosphorylation substrate), and leading to cytostatic G 1 arrest. Rapidly reversible neutropenia is the main dose-limiting toxicity of palbociclib, as one would expect on mechanistic grounds. 201 Other CDK inhibitors that remain under clinical development include two more CDK4/6selective agents: ribociclib, abemaciclib, as well as number of pan-CDK-selective compounds: dinaciclib, AT 7519, milciclib, roniciclib, and CYC 065. Additionally, MTKIs that may be classed as CDK inhibitors are TG 02 and RGB 286638.
The kinase inhibitor that probably has a longer clinical development history than any other is alvocidib (flavopiridol). This compound has been difficult to develop due to target promiscuity and limitations in formulation and disposition, but marked efficacy was observed some time ago in CLL, and this activity was ascribed to potent CDK9 inhibition by alvocidib. CLL-and many other cancers-are known selectively to depend on efficient RNA synthesis in order to support antiapoptotic signaling, which in turn requires CDK9 activity. 202 Alvocidib has now re-entered clinical development, based on promising activity in AML, also likely due to potent CDK9 inhibition. The first selective CDK9 inhibitor, BAY 1143572, has also now entered clinical trials. This compound contains the unusual sulfoximine solubilizing substructure. 203

W. Aurora Kinases
ARKs are a group of mitotic Ser/Thr kinases and in humans there are three homologues. ARK-A regulates mitotic entry, spindle assembly, alignment of metaphase chromosomes, and completion of cytokinesis. ARK-B and ARK-C, on the other hand, are chromosomal passenger proteins predominantly involved in chromosomal bi-orientation. ARKs play important roles in the maintenance of genetic stability and their aberrant expression leads to genomic instability or aneuploidy. Because ARKs are overexpressed in many cancers, thus generating aggressive tumors, they are regarded by many as attractive targets for cancer therapy. 204 There is an on-going debate in the field about what ARK selectivity is optimal in terms of efficacy and on-target toxicity. Currently, six ARK inhibitors (Table S24)  Although some therapeutic effects have been seen with most experimental ARK inhibitor drugs, it would appear that all of them suffer the same limitations as classical antimitotic agents, that is, cytostatic or cytotoxic effects against proliferating cells only, without good discrimination between transformed and normal cells, which manifest clinically as neutropenia and other hematological toxicities. Interestingly, several ARK inhibitors have been found to be effective against imatinib-and second-generation Tyr kinase inhibitor-resistant forms of the BCR-ABL kinase, especially the T315I mutant form. Some ARK inhibitors are now being tested clinically in patients with imatinib-resistant CML. 205

X. Polo-Like Kinases
PLK1 is one of four related Ser/Thr kinases associated with cell proliferation, especially in the mitotic phase of the cell cycle. Most cancers exhibit higher PLK1 expression than corresponding normal tissues, and PLK1 inhibition results in G 2 -M phase arrest, activation of the mitotic checkpoint, spindle dysfunction, and apoptosis in cancer cells. To date, several PLK inhibitors have entered clinical evaluation, so far with disappointing results in terms of efficacy and with neutropenia as the most prominent toxicity. 206 The most advanced experimental PLK inhibitor drug is volasertib, a potent and selective PLK1 inhibitor with pronounced antitumor activity in preclinical models, including taxane-resistant colorectal cancer. One of the possible reasons for the limited anticancer activity observed with this agent in humans may be its limited bioavailability in tumors compared to normal tissues. 207 Using a genome-wide synthetic lethality screen, it was shown recently that KRAS-driven tumors are selectively sensitive to PLK1 inhibition. 208 This finding might be used profitably to guide patient selection in efficacy trials of experimental PLK1 inhibitor drugs. PLK4, which is involved in centriole duplication, has not as yet received significant attention as an oncology target, but a selective PLK4 inhibitor, CFI 400945, has recently entered clinical evaluation (Table S25).

Y. DNA Damage Response Pathway Kinases
The PI3K-related kinase (PIKK) family includes ATM, ATR, DNA-PK, hSMG1, mTOR (discussed above under PI3K pathway inhibitors), and TRRAP, all large proteins that contain a C-terminal kinase domain closely related to the lipid kinase PI3K, but which function as Ser/Thr kinases, whereas CHK1 and CHK2 are Ser/Thr kinases in the CAMK family (NIM1 subfamily). Several PIKKs, including ATM, ATR, and DNA-PK, as well as the CHK kinases, are implicated in DNA replication and damage checkpoint pathways.
The DNA damage response helps cells to maintain genomic stability and the two main signaling routes are the ATM-CHK2 pathway, which controls DNA double-strand break repair, and the ATR-CHK1 pathway, which responds primarily to single-stranded DNA. DNA-PK, on the other hand, plays a role in DNA repair by nonhomologous end joining. ATM and ATR act on their CHK substrates, which in turn promote cell cycle arrest to allow time for repair, as well as on numerous other components involved in DNA repair, apoptosis, and the cell cycle. Since efficient DNA repair allows cancer cells to tolerate not only oncogenic replication stress, but also exogenous genotoxic stress from radio-and chemotherapy, the DNA response pathway may represent a promising target for cancer therapy. At least ATR and CHK1 inhibitors are believed to be potentially useful for monotherapy of certain cancers, for example, those with p53 and other DNA damage response deficiencies, whereas ATM, ATR, CHKs, and DNA-PK inhibitors are all potentially valuable as radio-and chemo-sensitizing agents. 209 Although a number of CHK inhibitors had been clinically evaluated in the past, 210 currently only the CHK1 inhibitors prexasertib and GDC 0575 remain under active development (Table S26). The first selective ATR inhibitors to enter the clinic are VX 970, VX 803, and AZD 6738 (Table S27). Additionally, the ATM inhibitor AZD 0156 (Table S29) and the DNA-PK inhibitors VX 984 and M 3814 (Table S28) are in early clinical trials. Finally, the first-in-class WEE1 inhibitor AZD 1775 (Table S29) should be mentioned here, as WEE1 is another kinase associated with cell cycle checkpoint control following DNA damage.

Z. CDC7
CDC7 is a Ser/Thr kinase that is involved in regulation of the S-phase and mitosis during the cell cycle. CDC7 inhibition has been reported to block DNA replication without activation of the S-phase checkpoint and directly to induce cellular apoptosis in transformed cells but not in normal cells. 211 A number of CDC7 inhibitors have been reported and several have undergone early clinical evaluation. 212 However, at present the only CDC7 inhibitor in the clinic is TAK 931 (Table S29; undisclosed structure).

MEDICINAL CHEMISTRY
Of the 231 (at the time of writing) clinical kinase inhibitor anticancer drugs, chemical structures for 172 are in the public domain. These molecules display structural and physicochemical properties as summarized and compared to those of all currently approved small-molecule drugs in Table I.
In terms of physicochemical properties, it can be seen that on average oncology kinase inhibitors are significantly larger (by almost a third; M r ) and more lipophilic (by around twothird; logP, logD) than small-molecule drugs in general, and they have much lower (predicted) intrinsic water solubility (by almost half; logS). Increased lipophilicity and lower solubility stem predominantly from relatively abundant aliphatic and aromatic systems (nAr) present in oncology kinase inhibitors. Furthermore, kinase inhibitors are comparatively more flexible (nRot) and contain more polar functionalities for H-bonding interactions (TPSA, HBA, HBD) than approved drugs in general. Structurally, kinase inhibitors on average contain more cyclic systems (nRing, nAr) but fewer chiral centers (nChir) than other drugs, presumably because kinase inhibitors, unlike many other drugs, are less frequently derived from natural products.
Correlation of logP and logS (Fig. 14) shows that most of the kinase inhibitor drugs are expected to be highly permeable but poorly soluble, some of them with exceptionally low (predicted) intrinsic aqueous solubility. With very few exceptions kinase inhibitor drugs are being developed as immediate release products for administration by the oral route and while the physicochemical properties in Table I are consistent with oral bioavailability, 213,214 average M r and logS suggest an average thermodynamic aqueous solubility of about 2 μg/mL (75 μg/mL for all approved drugs), which is only expected to be consistent with very low clinical doses, and provided permeability is high. 215 For many oncology kinase inhibitors absorption is likely to be variable and in vivo drug dissolution rate-limiting. It can therefore be expected that many kinase inhibitor drugs will be challenging to develop from a formulation viewpoint. As can be seen from Figure 15, both the maximum and average structural dissimilarity among oncology kinase inhibitors is significantly lower than that of all approved drugs, indicating that many kinase inhibitors, which generally target the conserved ATP-binding site, are pharmacophorically related. As expected, dissimilarity of compounds that are targeted to subgroups or individual kinases is diminished compared to that of all experimental kinase inhibitors. The groups of kinase inhibitors with the highest and lowest average structural dissimilarity are the PI3K and CSF-1/PDGF inhibitors on the one hand, and the BCR-ABL inhibitors on the other.
Another way of looking at structural similarity of oncology kinase inhibitor drugs is to assess the frequency at which common substructures occur in these compounds (Fig. 16). This shows that by far the most common substructure is the formanilide group (31 instances). This is probably not a distinguishing structural feature of kinase inhibitors, as this group occurs frequently in many different drug classes. The next most frequent substructure, however, is the quinazoline group (14 instances), a commonly used kinase inhibitor core, which occurs in several approved drugs and many experimental agents, especially in EGFR inhibitors. Other r FISCHER Figure 14. Predicted biopharmaceutical properties of kinase inhibitors compared to all approved drugs. According to the Biopharmaceutics Classification System (BCS), 247 oral drugs belong to class I if they are highly permeable and highly soluble, class II if highly permeable but of low solubility, class III if poorly permeable but highly soluble, and class IV if both poorly permeable and soluble. For the purposes of bioequivalence and other preclinical and clinical drug development studies, class boundaries are based on solubility of the highest dose strength and permeability as assessed by pharmacokinetic comparison of oral and intravenous dosing. Here, an attempt is made to predict such classifications based on calculated solubility (logS derived from intrinsic aqueous molar solubility) and permeability (logP; predicted octanol-water partition coefficient). LogS and logP values of -4.52 (30 μM) and 1.35, respectively, were used to set the class boundaries. 248 It can be seen that using this approximation it is predicted that the majority of kinase inhibitors fall into class II, some into class I, a few into class III, and very few into class IV.
Although kinase inhibitor lead compounds were frequently discovered using highthroughput screening in the past, a much more common current strategy is structure-based redesign of known kinase inhibitor templates, as is evident from the preceding similarity and property comparisons of clinical kinase inhibitors. More recently, fragment-based drug discovery has started to play an increasingly important role in the design of kinase inhibitors and the first approved kinase inhibitor arrived at by this strategy is vemurafenib. 216 A more recent example is AZD 5363 (Fig. 17).
Using a virtual screen of fragment compounds (M r < 250) against a model of AKT based on a crystallographic complex of an active form of that protein, followed by crystallography  Predicted logP values were calculated using Instant JChem (www.chemaxon.com). Ligand efficiency (LE) was calculated from pIC 50 values and lipophilic ligand efficiency (LLE) values were calculated from pIC 50 and predicted logP. 249 Binding mode illustrations were constructed from PDB entries 2JDO, 2UVX, 2UVY, 2VNW, 2 × 39, and 4GV1. Ligands are shown as cyan stick models, H-bonds as broken yellow lines, and ligand H-bonded water molecules as red spheres. Hinge-region residues involved in ligand H-bonding, charged residues in the ribose-binding region involved in polar ligand interactions, as well as a mobile Phe residue in the G-loop are shown as line models. of putative binding fragments with a PKA-AKT chimera construct, 7-azaindole was identified as an AKT hinge-binding fragment (Fig. 17A). 217 The binding mode of the isoquinoline-5sulfonamide AKT inhibitor CCT 077373 discovered earlier 218 suggested that determinants for AKT inhibition activity other than hinge binding were a basic amine group interacting with acidic groups in the ribose-binding regions, as well as a lipophilic group (chlorophenyl) interacting with the G-loop (Fig. 17F). For synthetic ease, the azaindole was replaced with purine, which was arylated at the 6-position to afford benzylamine intermediate 1 as one of the most active and ligand-efficient derivatives (Fig. 17B). Replacement of the phenyl group in that intermediate with piperidine gave rise to compounds with improved polar interactions with the acidic residues in the ribose-binding region and intermediate 2 (Fig. 17C) was identified as a compound with improved activity and excellent lipophilic ligand efficiency. 219 However, this compound was too polar and lacked cellular activity. Polarity was modulated by replacement of the azaindole core with pyrrolo[2,3-d]pyrimidine, repositioning of the primary amine group, and extension with a 4-chlorophenyl group to give rise to intermediate 3 (Fig. 17D), a highly potent AKT inhibitor with well-balanced physicochemical properties. 219,220 . Intermediate 3 was subsequently further elaborated in terms of kinase selectivity, hERG activity, and in vivo disposition properties to afford AZD 5363 (Fig. 17E). 221

CONCLUSIONS
Kinase inhibitor discovery is a comparatively recent medicinal chemistry endeavor and has progressed almost exponentially over the last ß25 years. Although early kinase inhibitor approvals were for MTKIs, the preponderance of more selective agents has increased gradually and now predominates. The advent of kinome-wide kinase assays and associated molecular pharmacology technologies, and the ever-expanding knowledge of how kinase inhibitors engage their molecular targets at the atomic level, now permit the rational design and validation of highly selective agents. This, together with modern clinical investigation strategies incorporating patient diagnosis, prognosis, and stratification based on genetic and mechanistic biomarkers, will continue to expand the oncologist's arsenal of highly specific kinase inhibitor drugs and will hopefully permit truly effective, modular, personalized treatments of cancers with less reliance on toxic chemotherapy. Because in general highly target-selective agents are also highly specific in vivo, it is hoped that progress can be made to optimize the therapeutic margin (therapeutic vs. toxic effects) of kinase inhibitor drugs, not only in oncology, but also in other therapeutic indications with unmet medical need. Based on current information, it is likely that this may not be achievable in many cases with inhibitors that target the catalytic activity of kinases, except in those cases where a particular cancer is addicted completely to the activity of a given kinase. In principle, pharmacological modulation of kinase activities is also possible by allosteric means, as we have seen. This is especially attractive if the possibility of such allostery arises in a manner that is specific for a particular context in terms of kinase environment and location, for example, through specific disease-associated interactions of a kinases, but such approaches remain challenging. Maximization of kinase inhibitor efficacy by these or other means is highly important in order to avoid emergence of drug resistance, which is a feature of many current kinase inhibitors, especially those with high target selectivity. As we have seen, the first kinase inhibitors for inflammatory disease have now been approved and several kinase inhibitors are under clinical investigation in indications other than cancer. A major challenge will be to apply what we have learned from oncology kinase inhibitors to other major disorders that currently lack effective therapies, such as neurodegeneration and (drug-resistant) microbial and viral infections.

SUPPORTING INFORMATION
Additional Supporting Information may be found in the online version of this article at the publisher's web site: Table S1. Protein kinase nomenclature. Table S2. BCR-ABL inhibitors. Table S3. SRC kinase inhibitors. Table S4. Epidermal growth factor receptor tyrosine kinase inhibitors. Table S5. Inhibitors of tyrosine kinases in the CSF-1/PDGF receptor subfamily. Table S6. Fibroblast growth factor receptor kinase inhibitors. Table S7. Hepatocyte growth factor receptor kinase (MET) inhibitors.