(2009) 23, 477–485 & 2009 Macmillan Publishers Limited All rights reserved 0887-6924/09 $32.00 www.nature.com/leu ORIGINAL ARTICLE

Global target profile of the inhibitor bosutinib in primary chronic myeloid leukemia cells

LL Remsing Rix1, U Rix1, J Colinge1, O Hantschel1, KL Bennett1, T Stranzl1,AMu¨ller1, C Baumgartner2, P Valent2, M Augustin3, JH Till3 and G Superti-Furga1

1Center for Molecular Medicine of the Austrian Academy of Sciences (CeMM), Vienna, Austria; 2Division of Hematology and Hemostaseology, Department of Internal Medicine I, Medical University of Vienna, Vienna, Austria and 3Millipore UK Ltd, Gemini Crescent, Dundee Technology Park, Dundee, UK

The detailed molecular mechanism of action of second- AMN107) are two such inhibitors that have been approved for generation BCR–ABL inhibitors, including the treatment of patients who are resistant or intolerant to perturbed targets and pathways, should contribute to rationa- 5,6 lized therapy in chronic myeloid leukemia (CML) or in other . An up-and-coming next-generation BCR–ABL affected diseases. Here, we characterized the target profile of inhibitor is bosutinib (SKI-606), a dual Sarcoma/Abelson (SRC/ the dual SRC/ABL inhibitor bosutinib employing a two-tiered ABL) kinase inhibitor that has recently been introduced in approach using chemical proteomics to identify natural binders clinical trials against CML (phase III) and breast (phase in whole cell lysates of primary CML and K562 cells in parallel to II).7–9 Bosutinib has been shown to target several imatinib in vitro kinase assays against a large recombinant kinase panel. resistance-causing BCR–ABL mutants, except the T315I mutant, The combined strategy resulted in a global survey of bosutinib targets comprised of over 45 novel tyrosine and serine/ and is unique in that it does not inhibit either the mast/stem cell threonine . We have found clear differences in the (KIT) or platelet-derived (PDGFR) receptors, 7,8 target patterns of bosutinib in primary CML cells versus the possibly explaining its higher therapeutic index. Furthermore, K562 cell line. A comparison of bosutinib with across bosutinib results in a delayed compensatory upregulation in the whole kinase panel revealed overlapping, but distinct, -activated kinase (MAPK) activity as compared inhibition profiles. Common among those were the SRC, with imatinib, suggesting further differences in their target ABL and TEC family kinases. Bosutinib did not inhibit KIT 10 or platelet-derived , but prominently profile. targeted the -linked STE20 kinases. Although in vivo Recent reports have demonstrated the success of large-scale bosutinib is inactive against ABL T315I, we found this clinically profiling of kinase inhibitors to determine their full spectrum of important mutant to be enzymatically inhibited in the mid- target in, for example, cell lines representing CML nanomolar range. Finally, bosutinib is the first kinase inhibitor (K562) and (Hs578T).11–15 In our laboratory, we shown to target CAMK2G, recently implicated in myeloid emphasize identifying drug targets in their clinically relevant leukemia cell proliferation. Leukemia (2009) 23, 477–485; doi:10.1038/leu.2008.334; patient cell populations. Thus, we previously used a chemical published online 27 November 2008 proteomic approach in which we immobilized coupleable Keywords: chronic myeloid leukemia; bosutinib; dasatinib; kinase analogs of imatinib, dasatinib and and exposed them profiling; chemical proteomics to total cell lysates not only of the BCR–ABL-positive K562 cell line, but also to extracts of primary CML patient cells, thereby identifying and subsequently validating both novel kinase (BTK, DDR1) and nonkinase (NQO2) targets.14,16 Here, we used a combined chemical proteomic and in vitro profiling method to evaluate (i) the natural target profile of Introduction bosutinib in primary CML and K562 cells and (ii) the activity of bosutinib versus dasatinib against over 270 recombinant Chronic myeloid leukemia (CML) is a three-phase hemato- kinases, thereby maximizing the cross-section of the kinome poietic disease originating in the stem cell compartment. under investigation. This combined approach revealed that A reciprocal translocation between 9 and 22 bosutinib, similar to dasatinib, is also a broad specificity kinase results in the expression of the BCR–ABL fusion oncoprotein, a inhibitor, targeting not only SRC and ABL kinases, but also TEC deregulated tyrosine kinase that activates several proliferative 1 family kinases, including BTK, which is known to have host and antiapoptotic signaling pathways. The BCR–ABL inhibitor immunomodulatory activity.17 In contrast to dasatinib, bosutinib imatinib (Gleevec, STI-571) entered the market in 2001 as the 2 does not inhibit KIT or PDGFR, but rather displays a striking first targeted therapeutic against CML; however, it soon affinity toward the STE family of kinases, in particular the STE20 became clear that both primary and secondary, or acquired, subfamily. CAMK2G, a Ca2 þ /calmodulin-dependent protein resistance toward imatinib can occur and thus necessitated the 3,4 kinase recently shown to be intricately involved in regulating development of second-generation BCR–ABL inhibitors. signaling networks controlling the proliferation of myeloid Dasatinib (Sprycel, BMS-354825) and nilotinib (Tasigna, leukemia cells,18 was also identified as a novel kinase target inhibited by bosutinib, but not dasatinib. Furthermore, using Correspondence: Dr G Superti-Furga, Center for Molecular Medicine chemical proteomics, clear differences were observed between of the Austrian Academy of Sciences (CeMM), Lazarettgasse 19, the bosutinib target profiles within primary CML cells and the Vienna 1090, Austria. E-mail: [email protected] K562 cell line. In addition to the correlation found between Received 25 August 2008; revised 26 September 2008; accepted 14 inhibitor affinity purification and inhibition of kinase activity, October 2008; published online 27 November 2008 this study exemplifies the complementary use of such kinase Bosutinib target profile in CML LL Remsing Rix et al 478 panels together with chemical proteomics for (near) kinome- were obtained from two CML patients with accelerated-phase wide identification of possible targets that could be responsible disease upon clinical admission. Upon taking the blood for clinically observed (side) effects or represent secondary samples, these patients had previously only received hydro- therapeutic points of intervention. xyurea, but no kinase inhibitor therapy (additional patient information, for example, blood makeup, is provided in Supplementary Table S1). In addition to the patient samples, Materials and methods K562 cells were used to allow for comparisons between the bosutinib profile obtained here and the profiles of bosutinib and Biological material other BCR–ABL inhibitors generated in this cell line.11,14 The list K562 cell pellets were generated by Cilbiotech (Mons, Belgium). of identified proteins (Supplementary Tables S2–S4) obtained Unfractionated peripheral blood leukocytes (buffy coat) were from cell lysates of the two CML patient samples and K562 obtained from patients in accelerated-phase CML. Peripheral following incubation with the bosutinib affinity matrix was blood mononuclear cells were obtained using Ficoll. Studies filtered against a ‘core proteome’ data set (comprised of the most were approved by the institutional review board (Medical prevalent proteins in the total cell lysate of K562, for example, University of Vienna). Written informed consent in accordance DNAPK, GCN1L1, HSPs, , tubulin) and a list of ‘frequent with the Declaration of Helsinki was obtained before blood hitters’ (that is, nonspecific proteins, such as mTOR or COG6, donation. identified in purifications using seven unrelated drugs).14 Table 1 lists the protein kinases present in this filtered list (see also Supplementary Table S5). Compounds and immobilization In parallel to the chemical proteomic investigation, bosutinib Bosutinib was a gift from Oridis Biomed (Graz, Austria). and c-bosutinib were assayed at 1 and 10 mM against 272 c-Bosutinib was synthesized by Vichem Chemie Ltd (Budapest, recombinant kinases, including several clinically relevant kinase Hungary) and acetylated or immobilized on NHS-activated mutants (Supplementary Table S6). This analysis revealed a Sepharose 4 Fast Flow (GE Healthcare Bio-Sciences AB, strong correlation (Po2.2E-16 for 1 mM and Po2E-16 for 10 mM) Uppsala, Sweden) through its amino-functionality as described 14 between inhibition by bosutinib and c-bosutinib further validat- previously. Dasatinib was synthesized by WuXi PharmaTech ing the use of c-bosutinib as a probe for detecting bosutinib- (Shanghai, China). binding proteins in physiological samples (Figure 1b). A correlation was also found between the two profiling Kinase inhibition analysis techniques in that the stronger a kinase was purified using the chemical proteomic approach (that is, higher sequence coverage), Bosutinib, c-bosutinib and its N-acetyl derivative were the stronger it was found to be inhibited (that is, lower IC )inthe assayed in vitro for inhibition of recombinant full-length 50 kinase panel (Figure 1c). Indeed, over half of the overlapping c-ABL (Millipore, Dundee, UK). of 5FAM- kinases were completely inhibited (less than 5% remaining KKGEAIYAAPFA-NH2 was monitored using the IMAP kinase activity) by 1 mM bosutinib. The three outliers, TAOK3, AMPK assay and a SpectraMax M5 plate reader (Molecular Devices, and MEK1, were still affected in a concentration-dependent Eugene, OR, USA). In addition, kinase assays were conducted manner, with less than 50% activity remaining at 10 mM.IC using Millipore’s KinaseProfiler according to the protocols 50 values were also obtained for several kinases listed in Table 1. detailed at http://www.millipore.com/drugdiscovery/dd3/ In addition to the significant link between kinase affinity KinaseProfiler (see also the protocol manual supplied in purification and kinase activity inhibition, the complementary Supplementary Figure S1), where adenosine triphosphate value of each technique was revealed, as several cell-type concentration for each specific kinase assay was set within 15 mM specific targets were only identified by affinity chromatography of the apparent Km for adenosine triphosphate where determined. (see Table 1), while many additional targets of these drugs, probably not expressed in the cells used here, but important in Affinity purification terms of creating a comprehensive target profile, were revealed Affinity chromatography was performed as described previously exclusively in the large-scale kinase inhibition assay (Figure 1d). except that, in addition to our normal inhibitor mix, Roche’s Another benefit of this two-tiered approach is that the weakness Complete Protease Inhibitor Cocktail tablet (1 tablet/50 ml lysis of one technique is alleviated by the strength of the other. On buffer) was included during experiments with the patients’ the one hand, we have used a chemical proteomic method cells.14,19 representing a post-genomic adaptation of classical drug affinity purification using high-end mass spectrometry and bioinfor- matics. The major advantage of this method when applied to Sample preparation, mass spectrometry and mass disease-relevant cells is that a drug is exposed in an unbiased spectrometry data processing manner to the entire human proteome in a ‘natural’ setting. See Supplementary Materials and methods. In contrast to most large-scale activity assays where recombi- nant proteins are used, this approach identifies drug-binding proteins in their natural state and environment, that is, level of Results and discussion abundance, post-translational modifications (for example, phos- phorylation), splice variants and presence of natural binding Chemical proteomics and kinase inhibition data sets partners. Such important considerations can all have a crucial display good correlation and complementarity impact on drug–protein interactions. However, the approach On the basis of the bosutinib binding mode to c-ABL and c-SRC, is only semiquantitative, as it does not directly allow for where the 4-methylpiperazine moiety is oriented toward the discrimination between affinity and abundance of the target, solvent region,14,20 the methyl group on the piperazine ring of thus highlighting the synergistic potential for integration with a bosutinib was replaced with a butylamine to form coupleable quantitative orthogonal approach as successfully shown by the (c)-bosutinib (Figure 1a). Peripheral blood mononuclear cells recent kinobeads/iTRAQ approach, which measures protein

Leukemia Bosutinib target profile in CML LL Remsing Rix et al 479

Figure 1 Good correlation and complementarity observed between chemical proteomics and kinase inhibition data sets. (a) Chemical structures of bosutinib, its coupleable analog c-bosutinib and the acetylated derivative thereof (N-Ac-c-bosutinib), which is indicative of the immobilized compound. All three compounds displayed equal potency in a c-ABL kinase assay (data not shown). (b) c-Bosutinib analog is a valid probe for detecting bosutinib-binding proteins as shown by the strong correlation between c-bosutinib and the parent drug across the entire kinase panel at 1 and 10 mM.(c) There is a good correlation between the two different profiling techniques (less than 50% remaining activity at 1 mM (Po8.1E-9) for kinases seen in both assays). Differences are mainly attributable to the fact that chemical proteomics data imply, for example, also protein expression levels and post-translational modifications. The curve is drawn to suggest the logarithmic dependence of range of IC50 with respect to the proteomics observations (sequence coverage) (curves of the form y ¼ aeÀbx , a, b40). Scatter plots were similar for all cell samples. Therefore, all proteomic observations were included in a single plot and each kinase is represented by several data points along its IC50 depending on the number of cell samples in which it was observed. (d) The two techniques strongly complement each other as several targets are only identified using chemical proteomics (CP, black), whereas others are observed only in the kinase assay panel (KA, white). Gray circles represent kinases seen in both analyses. The Human Kinome Map is adapted with permission from Technology (www.cellsignal.com).

binding competition.11 In contrast, we here apply the above- kinome expressed in the disease-related cells. Thus, quantitative mentioned large-scale assays using recombinant kinases, which data are also provided for ‘off-target’ kinases, which could be are advantageous in that (i) the inhibitory potential of the drug is responsible for clinically observed side effects in other tissues or determined quantitatively and directly on an enzymatic level cell types or represent secondary therapeutic targets (Table 2 and (ii) the panel of kinases is most likely different than the and Supplementary Table S7).

Leukemia Bosutinib target profile in CML LL Remsing Rix et al 480 Table 1 Average sequence coverage and IC50 values for kinase targets of bosutinib

IPI-ID name Patient 1 Patient 2 K562 IC50 (nM)

IPI00294842 MAP4K5 (KHS) F 1.5% 10.5% 0.3* IPI00013981 YES1 17.0% 15.0% 23.0% 0.4 IPI00216969 ABL1 4.0% 5.0% 22.0% 0.5 IPI00472302 BCR FF14.0% F IPI00830056 BCR–ABL1 (fusion region) FF13.0% F IPI00329488 ABL2 (ARG) (isoforms 1B & 4) 1.0% 2.5% 19.8% 0.5 IPI00328867 SRC 42.0% 24.0% 27.0% 1.0* IPI00016871 FGR 40.0% 36.5% F 1.1 IPI00555672 4.5% 4.0% (3.0%) 1.3 IPI00219012 FYN 15.0% 11.5% (11.0%) 1.8 IPI00000885 FRK (PTK5) 1.5% FF 2.2 IPI00029132 BTK 34.5% 44.0% 51.5% 2.5* IPI00552750 TNK2 (ACK1) 13.0% 14.0% F 2.7 IPI00029769 HCK 42.5% 39.5% (5.0%) 3.2 IPI00335281 STK24 (MST3) (12.5%) (13.0%) 3.0% 3.9 IPI00186826 EPHB4 F 4.5% 12.5% 5.5* IPI00020899 BMX 4.0% 6.0% F 7.9 IPI00298625 LYN (A and B) 56.3% 50.5% 41.5% 8.0* IPI00149094 MAP4K2 (GCK) FF3.0% 9.9* IPI00166680 MINK1 (3.0%) (2.5%) F 22 IPI00783394 STK10 (LOK) 8.5% 10.0% F 52 IPI00218278 SYK 4.0% 8.0% 3.0% 52* IPI00013212 CSK 52.0% 54.0% 44.5% 63 IPI00170675 CAMK1D 13.5% 11.0% 4.0% 92 IPI00216435 PTK2B (FAK2, PYK2) 5.5% 8.0% 3.5% 134 IPI00216378 CAMK2G 10.0% 10.0% 18.0% 184 IPI00011488 STK4 (MST1) 13.5% 7.0% F 191 IPI00000878 TEC FF30.5% 282 IPI00293613 TBK1 1.0% 5.5% 7.5% 471 IPI00029263 FER 9.0% 3.0% 8.0% 490 IPI00413961 PTK2 (FAK) FF5.5% 538 IPI00410485 TAOK3 (JIK) 3.0% 3.5% F 69% inhib. IPI00410287 PRKAA1 (AMPK) FF12.5% 60% inhib. IPI00219604 MAP2K1 (MEK1) 23.5% 20.0% 14.5% 56% inhib. IPI00844141 MST4 (MASK) 35.0% 36.5% 15.5% NA IPI00003783 MAP2K2 (MEK2) 24.5% 27.0% 15.5% NA IPI00479760 AAK1 20.5% 16.0% 7.0% NA IPI00513803 MAP3K2 17.5% 18.5% 7.0% NA IPI00017801 MAP3K3 16.0% 17.0% 9.0% NA IPI00029643 ZAK (Isoforms 1 & 2) 13.8% 11.8% 14.3% NA IPI00291457 STK35 (CLIK1) 13.5% F 14.5% NA IPI00002857 MAPK14 (p38a) 12.5% 13.0% F NA IPI00337426 BMP2 K (BIKE) 10.5% 7.0% 12.0% NA IPI00013219 ILK 9.0% FF NA IPI00012093 STK25 (YSK1) (9.0%) (10.0%) 3.5% NA IPI00218905 TNIK 5.5% (2.5%) F NA IPI00022827 SLK 5.0% 5.0% 3.0% NA IPI00006752 MAP4K4 (HGK, NIK) 5.0% 3.5% F NA IPI00787127 MAP3K1 1.0% 1.5% 3.0% NA IPI00298949 GAK F 7.0% 18.5% NA IPI00784270 MAP4K1 (HPK1) F 4.0% F NA IPI00384765 PKMYT1 F 2.5% 7.5% NA IPI00158248 MAP2K5 FF14.0% NA IPI00386260 MAP3K4 FF3.5% NA Abbreviation: inhib., inhibition. IPI-ID indicates IPI protein database entries to which identified peptides were assigned. Unless otherwise indicated, kinases present in this list have been specifically identified in duplicate in at least one of the three cell samples. The average sequence coverage (SC) is based on the unique peptides and takes into account the size of the protein. If a protein was identified in only one of the two repeats per cell type, then the SC is given in italics. Where no peptide is found to specifically confirm that a protein is present, the SC is given in parenthesis (%). Proteins are sorted first by IC50, where available, and then by average SC. The BCR–ABL is split into its components BCR and ABL1 plus a short peptide sequence 11 characteristic of the fusion region of BCR–ABL. IC50s marked with an asterisk were obtained from Bantscheff et al. % inhibition is at 10 mM. The top five hits for each cell type are highlighted in bold.

Bosutinib potently inhibits not only the SRC and ABL being most prominent (shown in bold in Table 1). The IC50 for family kinases, but also CAMK2G and the TEC and inhibition of all SFKs was less than 10 nM (see also Bantscheff STE20 kinases et al.11). The negative regulator of the SRC kinases, CSK, was As expected, almost all SRC family kinases (SFKs) were also one of the top hits across all cell types (IC50 ¼ 63 nM). ARG identified with LYN (up to 58% sequence coverage), FGR and peptides and peptides mapping to both the BCR and ABL HCK, the three principle SFKs in myeloerythroid progenitor cells portions of BCR–ABL, resulting in a sequence coverage of 36%,

Leukemia Bosutinib target profile in CML LL Remsing Rix et al 481 Table 2 IC50 (nM) values or activity (%) remaining at 10 mM for ties. However, the STE20 kinase most prominently identified, selected kinases MST4 (sequence coverage of greater than 35% in the CML patient samples), has been indicated not only in proapoptotic Kinase Bosutinib Dasatinib signaling, as are the related MST1 and MST3 proteins, but also in upregulating proliferative and transformation signals.25,26 More- ABL(T315I) 26 1648 over, it is unknown if the MSTs are active in these cells or if their AXL 174 86% BLK (mouse) 7.3 8 proapoptotic activity is overruled by BCR–ABL signaling. Also c-KIT 6313 93 part of the MAPK signaling pathway, but associated with c-KIT(D816 V) 2772 132 antiapoptotic signaling, the TRKA and TRKB kinases were E c-KIT(D816 H) 32 2.6 inhibited by bosutinib (IC50 25 nM) as well. These kinases are c-KIT(V560G) 181 1.5 not only involved in neuronal development, but have also been c-KIT(V654A) 132 4.2 shown to be either tumor supportive or suppressive, depending c-RAF 78% 164 27 EGFR 53 322 on the tumor type. At this point, it is unclear what impact EGFR(T790 M) 491 41% nanomolar inhibition of kinases involved in MAPK signaling by EPHB2 8.5 4.4 bosutinib will have on leukemic or healthy cells in CML FES 358 83% therapy; however, this possibly complementary pro- (TRKs) and IRAK4 574 84% anti- (MSTs) apoptotic effect of bosutinib may be one aspect LIMK1 84% 114 underlying the positive therapeutic index so far exhibited by PDGFRa 86% 63 RIPK2 28% 109 bosutinib in clinical trials. TIE2 1191 48% As in previous studies, we also identified several TRKA 22 35% nonkinase proteins, such as GRB2, STS-1, SHC, NCK2 and TRKB 27 52% SHIP2, which are known to interact with BCR–ABL and other TXK 40 o0.3 tyrosine kinases and therefore are most likely to be copurified 14 Abbreviations: EGFR, receptor; PDGFRa, with these. We furthermore detected some nonprotein platelet-derived growth factor receptor-a. kinases, such as pyruvate kinase and hexokinase-3, which upon Differentially inhibited kinase targets of bosutinib and dasatinib not further investigation turned out to be nonspecific binders, as identified in CML cells, which might have implications for side effects their enzymatic activities were not inhibited by bosutinib (data and/or therapy of unrelated diseases. Kinases are of human origin unless otherwise stated. not shown). Finally and most notably, we have identified CAMK2G, a Ca2 þ /calmodulin-dependent recently shown to were identified from K562 total cell lysates, thus serving as an be involved in controlling myeloid leukemia cell proliferation, 18 important internal positive control. Despite high protease as a novel target inhibited by bosutinib (IC50 ¼ 184 nM). activity particularly affecting BCR–ABL in the primary CML Hence, besides BCR–ABL and the SFKs, inhibition of CAMK2G patient cell lysates,19 we were still able to identify ABL and ARG is another important route through which bosutinib could exert peptides in both patient samples. As with dasatinib,13,14,16 the its beneficial effect in CML. Furthermore, in contrast to other TEC family kinases are also prominently targeted by bosutinib, CAMK inhibitors, which interfere with Ca2 þ /calmodulin com- with three of five members identified. In particular, BTK plex formation, bosutinib represents, to our knowledge, the first (IC50 ¼ 2.5 nM) was a major binding protein in all cell types. compound directly targeting the adenosine triphosphate binding Thus, the use of bosutinib could result in effects on immune site of CAMK2G. This is an important aspect, as activated function. Other tyrosine kinases potently inhibited by bosutinib CAMK2G is relatively independent of levels of the Ca2 þ / include the tumor suppressor EPHB4, FER, PTK2 (FAK), PYK2, calmodulin complex.18 SYK and TNK2, a nonreceptor protein tyrosine kinase whose overexpression in many human is often associated with poor prognosis and knockdown of which leads to increased Chemical proteomic approach discerns differences apoptosis in transformed cells.21–23 In addition, several serine/ between primary patient and K562 cells threonine kinases, for example, the NAK family members AAK1, In this study, we have observed significant differences between BIKE and GAK, were identified. the bosutinib protein target profiles in the K562 cell line versus Interestingly, the MAPK super family was found to be a major the two primary patient cells, with the latter showing a much target group of bosutinib. Although MEK1 (and presumably also closer relationship to each other than to K562 (Figure 2). K562, MEK2) is not significantly inhibited, the fact that bosutinib hits a an established model cell line for CML most closely resembling combination of many other kinases in this pathway (for blast crisis, is known to have alterations rendering it different example, MST1 and GCK) could explain why, compared with from primary cell samples.28,29 Thus, several relevant targets imatinib, it exhibits a delayed increase in MAPK signaling in would have been overlooked if only the K562 cell line was CML progenitor cells.10,24 Most remarkably, as many as 20 studied. For instance, numerous proteins were only identified in members of the STE group of serine/threonine kinases were patient cells including 7 of the 13 STE20s, p38a, TNK2, the SFKs found to interact with the bosutinib matrix, including 13 STE20 FGR and HCK and the TEC family kinase BMX. On the other kinases. In this regard, five of the six STE20 kinases that were hand, several kinases were purified on the bosutinib matrix only purified with the bosutinib matrix and included in the kinase from K562 cells, namely PRKAA1 (AMPK), MAP2K5, MAP3K4, assay have IC50 values below 200 nM. Whereas the PAK TEC and FAK. This disparity may highlight differences in protein subfamily of STE20 kinases does not seem to be sensitive expression, mutational status or differential secondary modifica- toward bosutinib inhibition (IC50 values greater than 1 mM for tions affecting bosutinib binding. It is also possible that the those tested), the GCK subfamily was significantly targeted. This proteins identified only in K562 cells do not actually play a role subgroup of kinases is involved in the proapoptotic c-Jun in the early stages of CML. Nonetheless, such proteins may still N-terminal kinase and p38 MAPK signaling pathways, suggest- be of importance during blast crisis and/or present interesting ing that bosutinib may actually also have antiapoptotic proper- targets for other disease manifestations.

Leukemia Bosutinib target profile in CML LL Remsing Rix et al 482

Figure 2 Bosutinib kinase target signature in CML patient and K562 cells as determined by chemical proteomics. (a) Emphasizing the differences between CML patient cells (red) and K562 cells (blue), the kinases targeted by bosutinib are displayed on the kinome tree. Purple circles represent kinases targeted in both. The Human Kinome Map is adapted with permission from Cell Signaling Technology (www.cellsignal.com). (b) Heat map representing bosutinib kinase signature in K562 and CML patient cells as determined by chemical proteomics. K562 shows major differences to both CML patient cell samples as illustrated by the long dendrogram length. Peptide count increases from white (no peptides) to red. Heat map was obtained by using the square root of the average peptide count of each sample type and plotted with R (http://www.r-project.org) by using the Heatmap_2 package. Dendrograms are computed by using the Euclidean distance and averages. Square root was applied to scale the counts, and avoid dendrograms dominated by a few proteins only.

Biological pathway analysis of kinases inhibited by highest hit pathway with a score of 12.4 is not involved in bosutinib emphasizes impact on MAPK signaling hematopoiesis, but rather in the regulation of the endocrine To identify biological processes in the hematopoietic system that system. This possible influence on the gonadotropin-releasing are particularly affected by bosutinib, we submitted all kinases hormone signaling pathway may hint at reproductive side effects identified through chemical proteomics to KEGG pathway in patients taking bosutinib. analysis and scored the resulting pathways on the basis of the cumulative kinase IC50s, that is, the lower the IC50 and therefore the higher the score, the stronger the pathway is considered Bosutinib protein target profile differs significantly from affected by bosutinib. Using the above-mentioned correlation that of dasatinib between sequence coverage and kinase inhibition (Figure 1c), To allow for direct comparison between the global kinase the IC50 values were estimated for those kinases not included in profiles of bosutinib and dasatinib, both dual SRC/ABL inhibitors the kinase assay panel. As expected, and the used against CML, dasatinib was also assayed against the kinase immune system are highly represented among the top 10 KEGG panel at 1 and 10 mM, and IC50 values were determined for pathways identified (Table 3). Accordingly, MAPK signaling selected kinases for both drugs (Tables 1 and 2, Supplementary prominently heads the list with a score of 15.6 followed by Fc Tables S6 and S7). As with bosutinib, the IC50 for inhibition of all 11 epsilon RI signaling and natural killer cell-mediated cytotoxi- SFKs for dasatinib is less than 10 nM (see also Bantscheff et al. ). city. T- and B-cell receptor signaling are also among the top The TEC family kinases TEC, BTK and BMX are potently affected immune system pathways. Interestingly, the second inhibited by both drugs (although slightly better by dasatinib).

Leukemia Bosutinib target profile in CML LL Remsing Rix et al 483

Figure 3 Targeting the kinome: bosutinib versus dasatinib. Emphasizing the differences found between bosutinib (red) and dasatinib (blue) as illustrated on the kinome tree. Purple circles represent kinases targeted by both drugs. Data points represent a compilation of targets obtained via chemical proteomics in K562 and CML patient cells, kinases inhibited at least 50% at 1 mM in the kinase panel (including AMPK, TAOK3 and MAP2K1, which were not sufficiently inhibited enzymatically, but still identified by chemical proteomics) and targets published by Bantscheff et al.11 The Human Kinome Map is adapted with permission from Cell Signaling Technology (www.cellsignal.com).

In agreement with the profiles obtained through chemical inhibited by both drugs at 1 mM, the ABL T315I mutant was, proteomics for dasatinib14 and bosutinib, the kinase inhibition under applied assay conditions, inhibited by bosutinib almost 70 profile also illustrates that there are distinct differences in the times more potently than by dasatinib (IC50 of 26 versus kinases targeted by these two broad-spectrum SCR/ABL inhibi- 1648 nM, respectively), which curiously does not translate to tors (Figure 3). This distinction seems to emphasize drug cellular activity, as previous reports describe no effect of selectivity on the basis of the gatekeeper residue. For instance, bosutinib on Ba/F3 cells, animal models or patients with the 8,30 whereas ABL (IC50o0.5 nM) and most of the assessed ABL T315I . The epidermal growth factor receptor mutants (H396P, M351T, Q252H and Y253F) were completely gatekeeper mutant T790M was also more strongly inhibited by

Leukemia Bosutinib target profile in CML LL Remsing Rix et al 484 Table 3 Pathway analysis of kinases targeted by bosutinib

KEGG pathway Kinases in pathway Score KEGG ontology

MAPK signaling MAP2K1, MAP2K2, MAP2K5, MAP3K1, MAP3K2, MAP3K3, 15.6 Signal transduction MAP3K4, MAP4K1, MAP4K2, MAP4K4, MAPK14, STK4, TAOK3, ZAK Fc epsilon RI signaling BTK, FYN, LYN, MAP2K1, MAP2K2, MAPK14, SYK 9.6 Immune system Natural killer cell-mediated cytotoxicity FYN, LCK, MAP2K1, MAP2K2, PTK2B, SYK 8.6 Immune system FYN, ILK, MAP2K1, PTK2, SRC 7.0 Cell communication T-cell receptor signaling FYN, LCK, TEC 6.2 Immune system B-cell receptor signaling BTK, LYN, SYK 6.0 Immune system ABL1, PKMYT1 4.4 Cell growth & death CML ABL1, MAP2K1, MAP2K2 4.2 Cancer VEGF signaling MAP2K1, MAP2K2, MAPK14, PTK2, SRC 4.1 Signal transduction Calcium signaling CAMK2G, PTK2B 1.6 Signal transduction GnRH signaling CAMK2G, MAP2K1, MAP2K2, MAP3K1, MAP3K2, 12.4 Endocrine system MAP3K3, MAP3K4, MAPK14, PTK2B, SRC Abbreviations: CML, chronic myeloid leukemia; GnRH, gonadotropin-releasing hormone; MAPK, mitogen-activated protein kinase; VEGF, vascular endothelial growth factor. Kinases identified through chemical proteomics were submitted to KEGG Pathway analysis. KEGG pathway is a collection of pathway maps representing known molecular interaction and reaction networks (http://www.genome.ad.jp/kegg/pathway.html).P The recognized pathways were then weighted according to the kinase IC50s, where known, using the equation ScoreðPÞ¼À logðIC50ðiÞ=max IC50Þ for all kinases (i) we have i2P identified in a pathway P. Unknown IC50s were estimated to be the median IC50 of kinases with ±5% sequence coverage of the kinase of interest on the basis of the correlation seen between kinase purification through chemical proteomics (sequence coverage) and inhibition in kinase assays (IC50) (Figure 1c). Listed are those pathways involved in biological processes in the hematopoietic system. Although nonhematopoietic, GnRH signaling was the second top hit pathway identified and may hint at reproductive side effects in patients taking bosutinib.

bosutinib (IC50 ¼ 491 nM) than by dasatinib (41% activity Conclusion remaining at 10 mM). Thus, bosutinib may represent an interest- ing structural scaffold from which new inhibitors that are better Bosutinib (SKI-606), originally designed as an SRC inhibitor, is able to target these clinically important mutants could be also a potent ABL inhibitor and is at present in clinical trials for developed. Epidermal growth factor receptor was also inhibited treatment of CML patients who are resistant or intolerant to 7,35 to a higher degree by bosutinib (IC50 ¼ 53 nM) than dasatinib imatinib therapy. In this study, we have shown the benefits of (322 nM), which could have important implications for the use of applying a two-tiered approach combining chemical proteomics bosutinib in additional (novel) indications, such as treatment of and large-scale kinase inhibition analysis to establish a global colorectal, pancreatic or nonsmall cell lung cancer.31,32 Further protein target profile of bosutinib in a disease-relevant setting complementing the profiles obtained by chemical proteomics, using cells obtained from patients with CML and the K562 cell dasatinib was far less active than bosutinib against the STE20s, line. Thus, we identified almost 20 new bosutinib kinase targets which predominantly carry a methionine gatekeeper residue, in CML patient cells. Overall, almost 50 novel targets were exhibiting in most cases an IC50 of greater than 10 mM for identified using the combined techniques. inhibition. As expected, neither KIT nor PDGFR are targeted by Our approach illustrates the significant difference in target bosutinib; however, the FES tyrosine kinase, which is important profiles from the experimental cell line K562 and the samples for signaling by KIT D816V,33 was inhibited, suggesting that obtained from human patients, thus signifying the importance of despite resistance of this mutant to bosutinib, there may still be a generating such profiles in disease-specific primary cell popula- clinical effect on D816V-positive malignancies. Finally, the tions. Furthermore, we have demonstrated a clear divergence of Ca2 þ /calmodulin-dependent protein kinases CAMK1D and targets for bosutinib and dasatinib, including identifying several CAMK2G, the neurotrophin-binding receptor tyrosine kinases novel bosutinib targets, such as CAMK2G and the STE20 TRKA/B and AXL, a recently reported target of bosutinib found kinases. This aids in providing a molecular rationale for to be involved in breast cancer cell motility and invasivity,15 discriminating between these two rather broad specificity SRC/ were selectively inhibited by bosutinib, whereas RIPK2 was ABL tyrosine kinase inhibitors in terms of the diverse clinical inhibited only by dasatinib. observations regarding both their efficacy and related side The different target profiles of bosutinib and dasatinib effects. It also helps to provide a basis for patient-specific use of presented here may help in understanding some of the clinically such kinase inhibitors as single agents or in combination therapy observed side effects of these drugs. In particular, dasatinib against CML. inhibits both KIT and PDGFR, which are believed to play a role in the occurrence of serious pleural effusions frequently seen during dasatinib treatment.34 Bosutinib, on the other hand, Acknowledgements neither inhibits these kinases nor causes this particular side effect. However, inhibition of these two kinases alone cannot This work was supported by the Leukemia and Society account for this effect, as they are also inhibited with similar (Grant number 5081-05), the Austrian federal ministry for science potency by imatinib and nilotinib, which generally do not cause and research bmwf under the GEN-AU program (GZ200.142/I-VI/ pleural effusions.7 Therefore, a more complex explanation such I/2006 and GZ200.145/I-VI/I/2006), the Austrian Science Fund as a combined inhibition of KIT or PDGFR and the TEC kinase (FWF; P18737-B11), the Austrian National Bank (O¨ NB) and the BTK may be responsible. It is also possible that the receptor Austrian Academy of Sciences (O¨ AW). We thank Melanie tyrosine kinase DDR1 plays a role in this regard.14 Planyavsky for preparation of the SDS-PAGE gels, Gerhard

Leukemia Bosutinib target profile in CML LL Remsing Rix et al 485 Du¨rnberger for assistance in bioinformatics analyses and the 18 Si J, Collins SJ. Activated Ca2+/calmodulin-dependent protein Superti-Furga group for helpful discussions. kinase IIgamma is a critical regulator of myeloid leukemia cell proliferation. Cancer Res 2008; 68: 3733–3742. 19 Maxwell SA, Kurzrock R, Parsons SJ, Talpaz M, Gallick GE, Kloetzer WS et al. Analysis of P210bcr- tyrosine protein kinase References activity in various subtypes of Philadelphia -positive cells from chronic myelogenous leukemia patients. Cancer Res 1 Jabbour E, Cortes JE, Giles FJ, O’Brien S, Kantarjian HM. Current 1987; 47: 1731–1739. and emerging treatment options in chronic myeloid leukemia. 20 Thaimattam R, Daga PR, Banerjee R, Iqbal J. 3D-QSAR studies on Cancer 2007; 109: 2171–2181. c-Src kinase inhibitors and docking analyses of a potent dual 2 Druker BJ. Imatinib as a paradigm of targeted therapies. Adv kinase inhibitor of c-Src and c-Abl kinases. Bioorg Med Chem Cancer Res 2004; 91: 1–30. 2005; 13: 4704–4712. 3 Druker BJ. Circumventing resistance to kinase-inhibitor therapy. 21 van der Horst EH, Degenhardt YY, Strelow A, Slavin A, Chinn L, N Engl J Med 2006; 354: 2594–2596. Orf J et al. Metastatic properties and genomic amplification of the 4 Shah NP, Sawyers CL. Mechanisms of resistance to STI571 in tyrosine kinase gene ACK1. Proc Natl Acad Sci USA 2005; 102: -associated . 2003; 15901–15906. 22: 7389–7395. 22 Mahajan NP, Liu Y, Majumder S, Warren MR, Parker CE, Mohler JL 5 Kantarjian H, Giles F, Wunderle L, Bhalla K, O’Brien S, Wassmann et al. Activated Cdc42-associated kinase Ack1 promotes prostate B et al. Nilotinib in imatinib-resistant CML and Philadelphia cancer progression via tyrosine phosphoryla- chromosome-positive ALL. N Engl J Med 2006; 354: 2542–2551. tion. Proc Natl Acad Sci USA 2007; 104: 8438–8443. 6 Lombardo LJ, Lee FY, Chen P, Norris D, Barrish JC, Behnia K et al. 23 Howlin J, Rosenkvist J, Andersson T. TNK2 preserves epidermal Discovery of N-(2-chloro-6-methyl- phenyl)-2-(6-(4-(2-hydro- growth factor receptor expression on the cell surface and enhances xyethyl)- piperazin-1-yl)-2-methylpyrimidin-4- ylamino)thiazole- migration and invasion of human breast cancer cells. Breast 5-carboxamide (BMS-354825), a dual Src/Abl kinase inhibitor Cancer Res 2008; 10: R36. with potent antitumor activity in preclinical assays. J Med Chem 24 Chu S, Holtz M, Gupta M, Bhatia R. BCR/ABL kinase inhibition 2004; 47: 6658–6661. by imatinib mesylate enhances MAP kinase activity in 7 Quintas-Cardama A, Kantarjian H, Cortes J. Flying under the radar: chronic myelogenous leukemia CD34+ cells. Blood 2004; 103: the new wave of BCR-ABL inhibitors. Nat Rev Drug Discov 2007; 3167–3174. 6: 834–848. 25 Ling P, Lu TJ, Yuan CJ, Lai MD. Biosignaling of mammalian Ste20- 8 Puttini M, Coluccia AM, Boschelli F, Cleris L, Marchesi E, Donella- related kinases. Cell Signal 2008; 20: 1237–1247. Deana A et al. In vitro and in vivo activity of SKI-606, a novel 26 Pombo CM, Force T, Kyriakis J, Nogueira E, Fidalgo M, Zalvide J. Src-Abl inhibitor, against imatinib-resistant Bcr-Abl+ neoplastic The GCK II and III subfamilies of the STE20 group kinases. Front cells. Cancer Res 2006; 66: 11314–11322. Biosci 2007; 12: 850–859. 9 Vultur A, Buettner R, Kowolik C, Liang W, Smith D, Boschelli F 27 Kruttgen A, Schneider I, Weis J. The dark side of the NGF family: et al. SKI-606 (bosutinib), a novel Src kinase inhibitor, suppresses neurotrophins in neoplasias. Brain Pathol 2006; 16: 304–310. migration and invasion of human breast cancer cells. Mol Cancer 28 Gribble SM, Roberts I, Grace C, Andrews KM, Green AR, Nacheva Ther 2008; 7: 1185–1194. EP. Cytogenetics of the chronic myeloid leukemia-derived cell line 10 Konig H, Holyoake TL, Bhatia R. Effective and selective inhibition K562: karyotype clarification by multicolor fluorescence in situ of chronic myeloid leukemia primitive hematopoietic progenitors hybridization, comparative genomic hybridization, and - by the dual Src/Abl kinase inhibitor SKI-606. Blood 2008; 111: specific fluorescence in situ hybridization. Cancer Genet Cyto- 2329–2338. genet 2000; 118: 1–8. 11 Bantscheff M, Eberhard D, Abraham Y, Bastuck S, Boesche M, 29 Guo JQ, Lin H, Kantarjian H, Talpaz M, Champlin R, Andreeff M Hobson S et al. Quantitative chemical proteomics reveals et al. Comparison of competitive-nested PCR and real-time PCR in mechanisms of action of clinical ABL kinase inhibitors. Nat detecting BCR-ABL fusion transcripts in chronic myeloid leukemia Biotechnol 2007; 25: 1035–1044. patients. Leukemia 2002; 16: 2447–2453. 12 Fabian MA, Biggs III WH, Treiber DK, Atteridge CE, Azimioara 30 Gambacorti-Passerini C, Kantarjian H, Bruemmendorf T, Martinelli MD, Benedetti MG et al. A small molecule-kinase interaction map G, Baccarani M, Fischer T et al. Bosutinib (SKI-606) demonstrates for clinical kinase inhibitors. Nat Biotechnol 2005; 23: 329–336. clinical activity and is well tolerated among patients with AP and 13 Hantschel O, Rix U, Superti-Furga G. Target spectrum of the BCR- BP CML and Ph+ ALL. ASH Ann Meeting Abstr 2007; 473. ABL inhibitors imatinib, nilotinib and dasatinib. Leuk Lymphoma 31 Engelman JA, Janne PA. Mechanisms of acquired resistance to 2008; 49: 615–619. epidermal growth factor inhibitors in non- 14 Rix U, Hantschel O, Durnberger G, Remsing Rix LL, Planyavsky small cell lung cancer. Clin Cancer Res 2008; 14: 2895–2899. M, Fernbach NV et al. Chemical proteomic profiles of the BCR– 32 Ciardiello F, Tortora G. EGFR antagonists in cancer treatment. ABL inhibitors imatinib, nilotinib, and dasatinib reveal novel N Engl J Med 2008; 358: 1160–1174. kinase and non-kinase targets. Blood 2007; 110: 4055–4063. 33 Voisset E, Lopez S, Dubreuil P, De Sepulveda P. The tyrosine 15 Zhang YX, Knyazev PG, Cheburkin YV, Sharma K, Knyazev YP, kinase FES is an essential effector of KITD816V proliferation signal. Orfi L et al. AXL is a potential target for therapeutic intervention in Blood 2007; 110: 2593–2599. breast cancer progression. Cancer Res 2008; 68: 1905–1915. 34 Talpaz M, Shah NP, Kantarjian H, Donato N, Nicoll J, Paquette R 16 Hantschel O, Rix U, Schmidt U, Burckstummer T, Kneidinger M, et al. Dasatinib in imatinib-resistant Philadelphia chromosome- Schutze G et al. The Btk tyrosine kinase is a major target of the Bcr- positive leukemias. N Engl J Med 2006; 354: 2531–2541. Abl inhibitor dasatinib. Proc Natl Acad Sci USA 2007; 104: 35 Golas JM, Arndt K, Etienne C, Lucas J, Nardin D, Gibbons J et al. 13283–13288. SKI-606, a 4-anilino-3-quinolinecarbonitrile dual inhibitor 17 Lindvall JM, Blomberg KE, Valiaho J, Vargas L, Heinonen JE, of Src and Abl kinases, is a potent antiproliferative agent against Berglof A et al. Bruton’s tyrosine kinase: cell biology, sequence chronic myelogenous leukemia cells in culture and causes conservation, mutation spectrum, siRNA modifications, and regression of K562 xenografts in nude mice. Cancer Res 2003; expression profiling. Immunol Rev 2005; 203: 200–215. 63: 375–381.

Supplementary Information accompanies the paper on the Leukemia website (http://www.nature.com/leu)

Leukemia