Activity-based kinase profiling of approved tyrosine kinase inhibitors

Authors


  • Communicated by: Kozo Kaibuchi

Correspondence: daisuke.kitagawa@carnabio.com

Abstract

The specificities of nine approved tyrosine kinase inhibitors (imatinib, dasatinib, nilotinib, gefitinib, erlotinib, lapatinib, sorafenib, sunitinib, and pazopanib) were determined by activity-based kinase profiling using a large panel of human recombinant active kinases. This panel consisted of 79 tyrosine kinases, 199 serine/threonine kinases, three lipid kinases, and 29 disease-relevant mutant kinases. Many potential targets of each inhibitor were identified by kinase profiling at the Km for ATP. In addition, profiling at a physiological ATP concentration (1 mm) was carried out, and the IC50 values of the inhibitors against each kinase were compared with the estimated plasma-free concentration (calculated from published pharmacokinetic parameters of plasma Ctrough and Cmax values). This analysis revealed that the approved kinase inhibitors were well optimized for their target kinases. This profiling also implicates activity at particular off-target kinases in drug side effects. Thus, large-scale kinase profiling at both Km and physiological ATP concentrations could be useful in characterizing the targets and off-targets of kinase inhibitors.

Introduction

Protein kinases play a crucial role in a wide range of cellular processes (Cohen 2001). Aberrant activation of kinase signaling pathways is involved in many types of disease, such as cancer, inflammation, and neurological disorders (Cohen 2002). Protein kinases, therefore, have become one of the most intensively investigated target classes for therapeutic intervention. So far, more than ten classes of small-molecular-weight protein kinase inhibitors have been approved for cancer treatment, and over 100 kinase inhibitors are currently in clinical development (Zsila et al. 2009). In most cases, these inhibitors compete with ATP for the ATP-binding site on target kinases. Because of the structural conservation of the ATP-binding pocket between diverse kinases, these compounds can also have unintended inhibitory actions at nontarget kinases. To understand the efficacies and side effects of the kinase inhibitors, it is crucial to characterize the selectivity of these inhibitors against kinases (Accili et al. 1996). In this study, we investigated the selectivity of nine approved tyrosine kinase inhibitors (imatinib, dasatinib, nilotinib, gefitinib, erlotinib, lapatinib, sorafenib, sunitinib, and pazopanib) alongside the pan-kinase inhibitor, staurosporine.

Imatinib was approved as the first inhibitor of breakpoint cluster region (BCR)—v-abl Abelson murine leukemia viral oncogene homolog (ABL) tyrosine kinase fusion protein (BCR-ABL) and used for the treatment of several malignancies, such as Philadelphia chromosome-positive chronic myelogenous leukemia (CML) (Cohen et al. 2002), acute lymphoblastic leukemia (ALL) (Piccaluga et al. 2007), and KIT-positive gastrointestinal stromal tumor (GIST) (Dagher et al. 2002). Dasatinib (Khorashad et al. 2008) and nilotinib (Weisberg et al. 2006) are second-generation BCR-ABL tyrosine kinase inhibitors and inhibit all mutant BCR-ABL forms, except for T315I, a gatekeeper mutant that is resistant to imatinib at physiologically relevant concentrations. Gefitinib (Cohen et al. 2003) and erlotinib (Minna & Dowell 2005) are used for treatment of epidermal growth factor receptor (EGFR)-positive nonsmall-cell lung cancer (NSCLC). Lapatinib (Ryan et al. 2008) is a dual inhibitor of EGFR and human epidermal growth factor receptor (HER)2 (also known as ErbB2) and used for treatment of patients with advanced breast cancer or metastatic breast cancer in which HER2 is over-expressed. For treatment of patients with advanced renal cell carcinoma (RCC), sorafenib (Wilhelm et al. 2006), sunitinib (Goodman et al. 2007; Rock et al. 2007), and pazopanib (Bukowski et al. 2010) have been approved. Their target kinases are vascular endothelial growth factor receptor (VEGFR) and other angiogenic growth factor receptors (Motzer & Bukowski 2006).

In 2002, 518 putative protein kinase genes were identified in the human genome, of which approximately 50 were predicted to be catalytically inactive (Manning et al. 2002). We have undertaken gene cloning and protein expression of over 90% of the putative active kinase genes. Consequently, 340 kinases were produced as active human recombinant kinases, which were used to establish 310 activity-based kinase assays (consisting 278 protein kinases, three lipid kinases, and 29 disease-relevant mutant kinases). The activities of these kinases were measured by nonradioactive methods including ELISA, IMAPTM, and off-chip mobility shift assay (MSA). These assay systems were used for kinase profiling with two distinct concentrations of ATP, these being the Km concentration for each individual kinase and a fixed physiological concentration (1 mm).

Results

Kinase profiling at the Km concentration of ATP

To investigate the inhibition profiles of kinase inhibitors, we established a panel of activity-based kinase assays for 310 human kinases (consisting of 278 protein kinases, three lipid kinases, and 29 disease-relevant mutant kinases (Figure 1)). Using this panel, nine of the currently approved kinase inhibitors were screened at a single concentration of 1 μm to determine the kinase targets of these compounds (Supplementary Table S1). The rate of inhibition was calculated for each inhibitor-kinase pair as described in Experimental procedures. For those inhibitor–kinase pairs in which the inhibition rate exceeded 40%, IC50 values were determined using a 10-point half-log dilution series of the inhibitor at the Km concentration of ATP. In parallel, IC50 values of a well-known pan-kinase inhibitor, staurosporine, were determined for each of the 310 kinases.

Figure 1.

Kinase-profiling panel. The phylogenetic tree of the human kinome is shown, colored according to whether they were assayed by MSA (blue), IMAPTM (green), or ELISA (red). All atypical kinases (EEF2K, PDHK2, PDHK4) and lipid kinases (PIK3CA/PIK3R1, SPHK1, SPHK2) were assayed by MSA but are not included in the phylogenetic tree. All mutant kinases were assayed using the same platform as that used for their wild type kinase.

To examine the reproducibility of the IC50 determination, we compared the IC50 values of staurosporine against each kinase determined twice independently. As shown in Figure 2, IC50 values from two independent assays were highly reproducible, regardless of the magnitude of the values.

Figure 2.

Reproducibility of the IC50 determination. The IC50 values of staurosporine from two independent experiments are shown on a scatter plot. Each dot represents a kinase. Both axes are shown on a logarithmic scale.

The IC50 values determined for the nine approved inhibitors are shown as a heat-map (Figure 3, Supplementary Table S2). Consequently, the target kinases (excluding mutant kinases) of an ABL inhibitor, imatinib, were (in rank order of IC50 values): Platelet-derived growth factor receptor (PDGFR)α < PDGFRβ < Discoidin domain receptor (DDR)2 < DDR1 < KIT < lymphocyte-specific protein tyrosine kinase (LCK) < lck/yes-related novel PTK (LYN) B < LYN A < ABL. The IC50 values of imatinib against these nine kinases were all within a hundredfold of that against the top target, PDGFRα (2.5 nm). On this criterion, dasatinib and nilotinib had 38 and 22 target kinases, respectively. The target kinases of dasatinib and nilotinib include the nine targets of imatinib, indicating that, among this group of related ABL inhibitors, imatinib should be the most selective. The target kinases of EGFR inhibitors were (in rank order of IC50 values): EGFR < HER4 < DDR1 for gefitinib; EGFR < HER4 < lymphocyte-oriented kinase (LOK) for erlotinib; EGFR < HER2 < HER4 for lapatinib, demonstrating that these three EGFR inhibitors are highly selective for EGFR family tyrosine kinases. The target kinases of the VEGFR inhibitor, sorafenib, were (in rank order of IC50 values): PDGFRα < DDR2 < rearranged during transfection (RET) < homeodomain-interacting protein kinase (HIPK)4 <  fms-like tyrosine kinase (FLT)4 < FLT1 < kinase insert domain receptor (KDR) < PDGFRβ < RAF1 < FLT3. The IC50 values of sorafenib at these 10 kinases were no more than a hundredfold of that against the top target, PDGFRα. The VEGFRs investigated in this study are FLT1 (also known as VEGFR1), KDR (also known as VEGFR2), and FLT4 (also known as VEGFR3). As expected, most of the target kinases of sorafenib were receptor tyrosine kinases for the angiogenic VEGFs and PDGFs. In contrast, sunitinib and pazopanib each had 30 kinase targets, which included the receptor tyrosine kinases for angiogenic factors, suggesting that sorafenib might be the most selective of these VEGFR inhibitors.

Figure 3.

Heat map chart of the activities of nine approved kinase inhibitors. Kinases for which the inhibitors have IC50 values lower than 1 μm were clustered based on the inhibition profile. Inhibition is color-graded, where the depth of blue coloring is proportional to the potency of its inhibitory activity (i.e. deep blue denotes low IC50 values). Noncolored cells indicate inhibition of less than 50% inhibition at 1 μm compound.

Imatinib, the first-generation ABL inhibitor used for the treatment of GIST and newly diagnosed CML, was followed by the second-generation ABL inhibitors, dasatinib and nilotinib, which are used in cases where patients experience resistance or intolerance to imatinib. Resistance to kinase inhibitors is often conferred by mutations in the target kinases. The selectivity of these nine kinase inhibitors was investigated against a number of mutant kinases (Table 1). In many cases, mutants exhibiting resistance to a certain inhibitor were still sensitive to other inhibitors. For example, although KIT T670I and D816V mutants were insensitive to imatinib, both mutants were comparatively sensitive to sunitinib, and KIT D816V mutant was potently inhibited by dasatinib (Table 1). Two exceptions were the gatekeeper mutants, ABL T315I and EGFR T790M. These mutants were highly resistant to all of the kinase inhibitors examined in this study (Table 1).

Table 1. Comparison of IC50 values of kinase inhibitors against ABL, KIT, EGFR, and PDGFRα and their disease-relevant mutant kinases
ABLABL inhibitors
ImatinibDasatinibNilotinib
Wild type1900.2719
E255K16000.63140
T315I> 10,00011009200
KITKIT inhibitors
ImatinibDasatinibSunitinib
Wild type1373.322
V560G6.90.366.0
V654A3500174.5
T670I> 10,000> 10,00049
D816V81002.437
EGFREGFR inhibitors
GefitinibErlotinibLapatinib
Wild type0.510.754.9
Del 746–7500.400.4737
Del 746–750/T790M320370> 10,000
T790M3401900780
T790M/L858R5105608500
L858R0.670.7710
L861Q0.731.22.5
PDGFRαMulti-TK inhibitors
SorafenibSunitinibPazopanib
  1. The IC50 (nm) values of each kinase inhibitor were determined by activity-based kinase assays at the Km concentration of ATP, as described in Experimental procedures.

Wild type1.01320
V561D5.61421
T674I10019170

Kinase profiling at a physiological ATP concentration

In the kinase assay, IC50 values of ATP-competitive compounds are affected by the concentration of ATP. Intracellular ATP levels are reported to be in the millimolar range (Imamura et al. 2009), so we carried out kinase profiling of compounds at a physiologically relevant ATP concentration (1 mm). The pIC50 values for the nine inhibitors against the fifty most strongly inhibited kinases were arranged and illustrated in Figure 4. The blue area shows the Cmax and Ctrough of the free plasma concentration of the parent compound or main metabolite in patients with cancer after dosing, as described in published literature (Table 2). Imatinib inhibited three kinases: PDGFRα (and its mutant V561D), DDR1, and KIT (and its mutant V560G) at concentrations lower than 290 nm (the maximum plasma-free concentration estimated from the Cmax) (Table 2), but did not inhibit ABL, its main target, at these concentrations, despite being shown to inhibit ABL in patients with cancer. This low potency of imatinib in the 1 mm ATP profiling is explained by its relatively low affinity for the active form of ABL; imatinib is known to preferentially bind and inhibit an inactive form of ABL (Schindler et al. 2000). Similar observations were made for nilotinib, a derivative of imatinib. Nilotinib inhibited tyrosine kinases in the DDR, PDGFR, and ephrin (EPH) families at concentrations below 87 nm, the estimated maximum plasma-free concentration (Table 2). Dasatinib inhibited multiple tyrosine kinases in the sarcoma kinase (SRC), ABL, DDR, tyrosine kinase expressed in hepatocellular carcinoma (TEC) and EPH families, and the salt-inducible kinase (SIK), a serine/threonine kinase, at concentrations below 4.1 nm. In contrast to imatinib and nilotinib, dasatinib is known to bind with approximately equal affinity to both the active and inactive forms of ABL and other kinases.

Figure 4.

Top 50 kinases inhibited by nine approved kinase inhibitors in the activity-based kinase assay at an ATP concentration of 1 mm. The pIC50 values calculated from the IC50 values of the compounds were arranged in order. The two horizontal lines denote the Cmax and Ctrough of the free plasma concentration of the parent compound or main metabolite in patients with cancer after dosing as described in Table 2. The magenta bar in the graph of lapatinib shows the pIC50 value estimated after a 30 min pre-incubation of inhibitor and kinase.

Table 2. Summary of Cmax and Ctrough plasma concentration and protein-binding values for the approved kinase inhibitors
InhibitorsRegimenCmax (μg/mL)Ctrough (μg/mL)PB (%)Estimated plasma-free concentration (nm)aReferences
  1. QD, quaque die/as prescribed; BID, bis in die/twice daily; PB, protein-binding rate.

  2. a

    The putative plasma-free concentration of each tyrosine kinase inhibitor, calculated from the Cmax and Ctrough plasma concentrations and the protein-binding (PB) rate.

  3. b

    Sunitinib+SU12662 from Study A6181033, FDA Approval review (http://www.accessdata.fda.gov/scripts/cder/drugsatfda/).

Imatinib400 mg QD1.60.9891–96180–290Widmer et al. (2008)
Dasatinib70 mg BID0.050≈0.010960.8–4.1Demetri et al. (2009)
Nilotinib400 mg BID2.31.09838–87Kantarjian et al. (2006)
Gefitinib250 mg QD0.350.199043–78Herbst et al. (2002)
Erlotinib150 mg QD2.51.595190–320Tan et al. (2004)
Lapatinib1,200 mg QD1.40.4996.9–24Burris et al. (2009)
Sorafenib400 mg BID10.0≈6.099130–220Awada et al. (2005)
Sunitinibb50 mg QD0.0360.025953.1–4.5 
Pazopanib800 mg QD452499.955–100Hurwitz et al. (2009)

Gefitinib and erlotinib were shown to specifically inhibit EGFR and its mutants, with the exception that erlotinib also inhibited the serine/threonine kinase, LOK. Lapatinib did not inhibit any kinase at a concentration below 24 nm. As lapatinib is known to bind slowly to EGFR and HER2, we repeated the assay with a 30 min pre-incubation of the kinases with the compound before initiation of the kinase reaction (Figure 4, red column). This pre-incubation potentiated the inhibitory activity of lapatinib against EGFR, the EGFR mutants L861Q and L858R, and HER2, but did not change its effect on EGFR del746–750 (Supplementary Table S3). Thus, lapatinib proved to be a specific inhibitor of EGFR and HER2. The other EGFR inhibitors, gefitinib and erlotinib, did not show any enhancement of potency against these kinases by pre-incubation.

Sorafenib and pazopanib inhibited KIT and/or a KIT mutant at concentrations lower than 220 and 100 nm, the maximum plasma-free concentration estimated from the Cmax, respectively. In addition, sorafenib inhibited FLT1, KDR, and FLT4, which are known to be key factors for tumor angiogenesis. Sunitinib inhibited FLT3 and KIT at the lower concentration than the maximum plasma-free concentration estimated from the Cmax. Sunitinib has been reported to be metabolized to N-desethyl-sunitinib (SU12662) in the liver. Thus, the in vivo kinase inhibition profile may be influenced by the properties of this metabolite (Houk et al. 2009). Nevertheless, the potent inhibition of KIT by sunitinib and its metabolites may help to rationalize its application in patients with imatinib-resistant GIST.

Discussion

It is well known that most kinase inhibitors inhibit multiple kinases simultaneously (Davies et al. 2000; Bain et al. 2003, 2007). Therefore, it is important to know the selectivity of compounds against a large panel of kinases. In this study, we carried out activity-based profiling of nine approved kinase inhibitors and a pan-kinase inhibitor, staurosporine, against 310 kinases to examine the specificity of these compounds. Although some novel interactions were revealed, the results generally reproduced those of previous studies using competition binding, thermal shift, and chemical proteomics assays (Fabian et al. 2005; Bantscheff et al. 2007; Fedorov et al. 2007; Karaman et al. 2008). For example, the top 10 targets for imatinib, defined by competition-binding assay, are reported to be ABL, Abelson-related gene (ARG), B-cell lymphocyte kinase (BLK), DDR1, DDR2, colony-stimulating factor 1 receptor (CSF1R, also known as FMS), KIT, LCK, PDGFRα, and PDGFRβ (Karaman et al. 2008), which is remarkably similar to our ranking of ABL, ARG, DDR1, DDR2, KIT, LCK, LYN A, LYN B, PDGFRα, and PDGFRβ.

In physiological conditions, most kinases exist in equilibrium between at least two different conformations: active and inactive. The cellular potency generally correlates well with data from biochemical assays, in some cases predicting the kinase to be in an active form, whereas in other cases predicting it to be in an inactive form (Knight & Shokat 2005). Although imatinib has been demonstrated to show preferential inhibition against the inactive forms of ABL and other kinases (Schindler et al. 2000; Walter et al. 2007), the enzymes targeted by imatinib in this study using active kinases were generally identical to previous studies that also included inactive kinases. It thus seems that the kinase selectivity of compounds in a cellular context can be adequately predicted using an active kinase panel, although profiling using inactive kinases would provide additional complementary information and a better resolution picture of the activity profiles of these compounds.

This large-scale activity-based kinase profiling highlighted putative novel targets for clinically relevant tyrosine kinase inhibitors. Nilotinib is known to inhibit p38 mitogen-activated protein kinase (MAPK) (Manley et al. 2010) but our finding that it also inhibits the upstream kinase, mitogen-activated protein kinase (MAP2K)3, was novel. Combined inhibition of multiple kinases within the same pathway could have a synergistic effect on that pathway and on the inhibition of tumor progression. Except for lapatinib, all of the inhibitors studied were relatively potent against DDR1, a receptor tyrosine kinase involved in cancer-cell invasion. The inhibition of DDR1 kinase could be a key element in determining the anti-cancer efficacy of these drugs. Erlotinib has been demonstrated to bind LOK in a thermal stability shift assay (Fedorov et al. 2007) and, in this study, showed potent inhibition against LOK activity. Dasatinib has been demonstrated to inhibit SIK and Qin-induced kinase (QIK) (Ahmed et al. 2010), which had been identified as putative targets for dasatinib by cell-based quantitative chemical proteomics (Bantscheff et al. 2007). Consequently, we confirmed in this study that dasatinib was a relatively potent inhibitor of SIK and QIK.

For many of these target candidates, little is known about whether they remain as targets at physiological concentrations of ATP. Each kinase protein has an intrinsic Km value for ATP, and the apparent Ki value is usually determined using this concentration during kinase profiling. Although the average Km value for ATP at 281 wild type kinases was 58 μm, intracellular ATP levels are estimated to be between 1 and 10 mm. It is also reported that ATP levels in the cytoplasm and nucleus are higher than in mitochondria (Imamura et al. 2009). The IC50 values of inhibitors against each kinase are likely to be affected by the apparent Km value of ATP in the in vitro assay. The use of subphysiological ATP concentrations during profiling is likely to give the impression that the inhibitors are more potent than they would be at physiological ATP concentrations. Thus, it appears necessary to estimate kinase inhibition at a physiological ATP level during profiling to assess properly the physiological relevance of each compound, hence our use of 1 mm ATP in our profiling of these tyrosine kinase inhibitors. In Table 3, the comparison of IC50 values between at the ATP concentration of Km and 1 mm concerning kinases strongly inhibited by some of nine kinase inhibitors is shown. It is to be anticipated that kinases with high Km values such as DDR1 and KIT showed small differences between two IC50 values, whereas kinases with low Km values such as ABL and EGFR indicated large differences between these IC50 values.

Table 3. Summary of IC50 values at the ATP concentration of Km and 1 mm for 20 kinases
KinaseATP conc. (μm)ImatinibDasatinibNilotinibGefitinibErlotinibLapatinibSorafenibSunitinibPazopanib
K m 1 mm K m 1 mm K m 1 mm K m 1 mm K m 1 mm K m 1 mm K m 1 mm K m 1 mm K m 1 mm
  1. The IC50 (nm) values of each kinase inhibitor against 20 kinases at the ATP concentration of Km and 1 mm are summarized. The actual ATP concentrations in the Km assay are indicated in the second column. n.d. means that IC50 value was greater than 10 μm, and blank cells indicate that the inhibition rate at the inhibitor concentration of 1 μm was below 40% at the ATP concentration of Km.

ABL2519021000.272.3184702600n.d.5407100  990n.d.10015007107500
DDR11001001301621544437130833504400n.d.1601603215065140
DDR250794400.90.842.34.8570440011008700  10336366047200
EGFR58700n.d.262100  0.51220.75404.9160    7700n.d.
FLT1150        4401100  278222591332
FLT310085023004400n.d.56015007302400200780  712500.512.0100450
FLT475        2901800  26130126916100
HER210      31001900  9.8200      
HER425720n.d.3.972470n.d.7.61503875024550  3100n.d.  
HIPK451100n.d.    320n.d.    19510160n.d.  
KDR75        3101700  2915020987.151
KIT4001401703.33.7120170      18024022263.86.4
LCK1016015000.170.48578603906000780n.d.  9004900252103102900
LOK100  22004100  43081065110  110017002203404101500
LYN A1019016000.281.367170035036004806300  5102800333101201200
LYN B2518013000.281.978190041037006605500  6502900444301201300
PDGFRα252.5292.6192.43560075007507200  1.0181311020250
PDGFRβ25689301.1136015006709500890n.d.  344005.79442670
RAF10.52100n.d.7704300770n.d.      51300  130n.d.
RET10  3506400  770n.d.4409600  143201.739461200

As shown in Figure 4, profiling at 1 mm ATP can reveal which kinases are susceptible to inhibition by tyrosine kinase inhibitors in physiological conditions. The ABL inhibitors imatinib, dasatinib, and nilotinib have been demonstrated to inhibit BCR-ABL activity and the phosphorylation of its substrate, Crk-like protein (CRKL), one of the most highly phosphorylated proteins found in patients with CML (White et al. 2007). In this study, although dasatinib was shown to inhibit ABL activity in the range of clinically estimated plasma-free concentrations, imatinib and nilotinib were not. Some kinase inhibitors are known to be sensitive to the phosphorylation state of the kinase, and therefore, these compounds can have variable potency depending on the state of the target protein. Imatinib has been reported to inhibit preferentially the unphosphorylated (inactive) form of ABL (Schindler et al. 2000), as does nilotinib (Weisberg et al. 2006; Manley et al. 2010), and this preferential inhibition of the inactive form of ABL may explain their low potency in this activity-based assay. Generally, imatinib and nilotinib are classified as inhibitors of inactive kinases (type II inhibitors), in contrast to sunitinib, which is an inhibitor of active kinases (type I inhibitor). Sunitinib, however, has demonstrated to have high affinity to a members of the PDGFR family kinases and bind preferentially to the inactive forms of KIT (Quintas-Cardama & Cortes 2008; Gajiwala et al. 2009) and FMS (Kitagawa et al. 2012), indicating that the type I (ATP-competitive) kinase inhibitors binds preferentially inactive form of some kinases. A concurrent profiling assay using inactive kinases may be advisable to generate comprehensive data regarding compound affinity for targets.

To explore the function of genes of interest, a wide variety of knockout mice has been generated. As shown in the references for Table S2, knockout mice have also been created for approximately 70% of the kinases investigated in this study. A considerable number of mice in which a specific kinase is knocked out exhibit impaired cardiac function. Studies of functional knockout mice show that KIT is required for cardiomyocyte terminal differentiation (Li et al. 2008), VEGFR and PDGFR signaling are required for maintaining cardiac homeostasis during a pressure load or ischemia (Thirunavukkarasu et al. 2008; Chintalgattu et al. 2010), AMP-activated protein kinase (AMPK) exerts a cardioprotective effect against pressure-overload-induced ventricular hypertrophy and dysfunction (Zhang et al. 2008), and skeletal muscle myosin light chain kinase (skMLCK) is necessary for myosin regulatory light chain phosphorylation and cardiac performance (Ding et al. 2010). Therefore, inhibitors targeting these kinases potentially result in cardiotoxicity, as is the case with imatinib, dasatinib, sunitinib, and sorafenib, but not gefitinib, erlotinib, and lapatinib (Hasinoff 2010; Force & Kolaja 2011). These cardiotoxic properties will typically be due to kinase inhibition, although nonkinase targets may also be relevant. The approved EGFR family kinase inhibitors are highly specific to their target kinases and have no effects on these cardiac enzymes. The VEGFR inhibitors, sorafenib, sunitinib, and pazopanib target tyrosine kinases relevant to VEGFR and other angiogenic growth factor receptors, blockade of which can lead to hypertension. Treatment with an anti-VEGF monoclonal antibody, bevacizumab, causes hypertension in a considerable number of patients. Sunitinib has been reported to be clearly associated with clinical cardiotoxicity (Chu et al. 2007). In this kinase profiling, sunitinib showed inhibitory effects against kinases related to cardiac function (e.g. AMPKs and skMLCK) in addition to its target VEGFRs (FLT1, KDR, FLT4), PDGFRs and KIT (Figure 2). It also inhibited tyrosine activation motif (TAM) receptor tyrosine kinases, TYRO3, AXL, and MER that are receptors of GAS6 and Protein S (Hurtado et al. 2011), although this observation does not decisively implicate these as ‘off-targets’ of sunitinib. Sorafenib is also known to induce cardiac toxicity but at a lower incidence than sunitinib. In this study, sorafenib-inhibited VEGFRs and PDGFRs with almost the same magnitude as sunitinib, but crucially showed no inhibition of AMPKs. The difference in the severity of cardiac dysfunction evoked by the two drugs may therefore be due to their divergent ability to inhibit AMPKs.

In summary, activity-based kinase profiling using 281 distinct kinases at Km and physiological ATP concentrations revealed that nine approved kinase inhibitors are well optimized for their target kinases. These profiles are likely to benefit patients with cancer by improving the specificity of their therapies. Such kinome-wide profiling could provide useful information on kinase targets in the process of drug discovery. Owing to their relatively high sensitivity, profiling at the Km concentration of ATP should be useful in revealing unknown targets of an inhibitor. Subsequent profiling at a physiological ATP concentration could be used to investigate its likely efficacy in vivo and its potential adverse side effects. Large-scale activity-based profiling should play an important role in the discovery of novel kinase inhibitors.

Experimental procedures

Compounds

Imatinib mesylate (imatinib) was extracted from capsules, and gefitinib, erlotinib hydrochloride (erlotinib) and lapatinib ditosylate (lapatinib) were extracted from their tablet preparations. Dasatinib (Lombardo et al. 2004), sorafenib tosylate (sorafenib) (Bankston et al. 2002), and sunitinib malate (sunitinib) (Sun et al. 2003) were synthesized at Carna Biosciences, Inc. Nilotinib was purchased from Axon Medchem BV (Groningen, The Netherlands). Pazopanib was purchased from LC Laboratories (Woburn, MA).

Protein expression and purification

The kinase-coding region was amplified by PCR from a human tissue cDNA mixture. For expression in Escherichia coli; the genes were subcloned into the pGEX-6P1 vector (GE Healthcare) that contains N-terminal GST tag. This allowed them to be expressed in Escherichia coli strain BL21 or DH5α or JM109 by IPTG (0.1 mm) induction for 20 h at 25 °C. The cells were lysed by sonication in lysis buffer A (50 mm Tris-HCl, pH 7.5, 150 mm NaCl, 5 mm DTT, 1 mm phenylmethansulfonylfluoride (PMSF), 2 μg/mL leupeptin, 2 μg/mL aprotinin, 1 mm NaF, 100 μm sodium orthovanadate, 1 mm cantharidine, 0.5 mm EDTA, 0.5 mm EGTA, and 1 mg/mL lysozyme) before adding 1% Triton X-100. The supernatant of the expressed protein was purified by standard glutathione affinity chromatography. For expression in insect cells, the genes were subcloned into a pFastBAC vector (Invitrogen) that contains N-terminal GST or His(×6) tag. The recombinant bacmid DNA was prepared according to the manufacturer's instructions for the Bac-to-Bac baculovirus expression system (Invitrogen) and transfected into Spodoptera frugiperda 9 (Sf9) insect cells to amplify the recombinant baculovirus. Sf21 cells in Grace's insect media supplemented with 10% fetal calf serum were infected with the recombinant baculovirus at a multiplicity of infection of 0.1–3 per cell and cultured for 48–72 h at 27 °C. The cells were then lysed on ice in lysis buffer A for GST-tag proteins or lysis buffer B (50 mm Tris-HCl, pH 7.5, 300 mm NaCl, 10 mm imidazole, 1% Nonidet P-40, 5 mm DTT, 1 mm PMSF, 2 μg/mL leupeptin, and 2 μg/mL aprotinin) for His(×6)-tagged proteins. The supernatant of the expressed protein was purified by standard one-step affinity chromatography using glutathione SepharoseTM or Ni-NTA. All purification processes thereafter were carried out at 4 °C.

Confirmation of protein identity

To identify the proteins, they were applied to SDS-PAGE and then stained with Coomassie Brilliant Blue. The band on the gel corresponding to the target protein was digested with trypsin and analyzed with Matrix Assisted Laser Desorption Ionization-Time of Flight-Mass Spectrometry (MALDI-TOF-MS) (AutoFlex III, Bruker). The data were analyzed using MASCOT Peptide Mass Fingerprinting as described previously (Kinoshita et al. 2008).

Kinase assays and IC50 determination

The assay protocols of each kinase are published on the Carna Biosciences website (http://www.carnabio.com/english/index.html). Kinase activities were measured by a nonradioisotopic method, such as ELISA, IMAP or MSA (Kinoshita et al. 2008). For each kinase, the apparent Km for ATP was determined by a substrate (ATP)-velocity plot and the Michaelis–Menten equation. Km bin was expediently used for the kinase assay as the Km value for each kinase. Compounds were dissolved in DMSO and diluted in a half-log scale for use in IC50 determination. For each compound, the DMSO solution was diluted in assay buffer to yield a final concentration of 1% DMSO. The kinase assays were carried out using ATP either at its Km or at 1 mm (the physiological concentration of ATP in cells has been demonstrated to be in the millimolar range (Imamura et al. 2009)). The inhibition of kinase activity by each compound was calculated as follows: inhibition (%) = [1-(A-B)/(C-B)] × 100 where A is the response with compound, B is the background response with no kinase, and C is the response with vehicle (1% DMSO). The IC50 value of each compound was calculated by interpolation on a log-concentration-response curve fitted with a four-parameter logistic equation. The pIC50 values were given as −log10(IC50) values.

Acknowledgements

The authors thank Dr. Yusuke Kawase for a critical reading of the manuscript and helpful discussion; Tetsuya Myojin, Yui Iwamae, Hiromi Iwamori and Tomoe Bunno for expert technical assistance.

Ancillary