Optimized Chemical Proteomics Assay for Kinase Inhibitor Profiling

Feb 8, 2015 - ... Paul Otto , Mei Cong , Carrow I. Wells , Benedict-Tilman Berger , Thomas Hanke , Carina Glas , Ke Ding , David H. Drewry , Kilian V...
0 downloads 5 Views 6MB Size
Article pubs.acs.org/jpr

Optimized Chemical Proteomics Assay for Kinase Inhibitor Profiling Guillaume Médard,*,† Fiona Pachl,† Benjamin Ruprecht,† Susan Klaeger,†,‡ Stephanie Heinzlmeir,†,‡ Dominic Helm,† Huichao Qiao,† Xin Ku,† Mathias Wilhelm,† Thomas Kuehne,† Zhixiang Wu,† Antje Dittmann,§ Carsten Hopf,§ Karl Kramer,† and Bernhard Kuster*,†,‡,∥ †

Chair of Proteomics and Bioanalytics, Technische Universität München, Freising, Germany German Cancer Consortium (DKTK) and German Cancer Research Center (DKFZ), Heidelberg, Germany § Cellzome GmbH, Heidelberg, Germany ∥ Center for Integrated Protein Science (CIPSM), Munich, Germany ‡

S Supporting Information *

ABSTRACT: Solid supported probes have proven to be an efficient tool for chemical proteomics. The kinobeads technology features kinase inhibitors covalently attached to Sepharose for affinity enrichment of kinomes from cell or tissue lysates. This technology, combined with quantitative mass spectrometry, is of particular interest for the profiling of kinase inhibitors. It often leads to the identification of new targets for medicinal chemistry campaigns where it allows a two-in-one binding and selectivity assay. The assay can also uncover resistance mechanisms and molecular sources of toxicity. Here we report on the optimization of the kinobead assay resulting in the combination of five chemical probes and four cell lines to cover half the human kinome in a single assay (∼260 kinases). We show the utility and large-scale applicability of the new version of kinobeads by reprofiling the small molecule kinase inhibitors Alvocidib, Crizotinib, Dasatinib, Fasudil, Hydroxyfasudil, Nilotinib, Ibrutinib, Imatinib, and Sunitinib. KEYWORDS: kinobeads, chemical proteomics, drug profiling, affinity matrix, kinase inhibitors



INTRODUCTION

preclinical development as well as approval by the health authorities would be complicated. The classical approach for selectivity profiling uses kinase panels relying on recombinant proteins.15 These assays are valuable and valued tools,16 but they do not capture the molecular complexity of the target proteins in cells. The chemical proteomics concept intends to circumvent the choice of kinases to include in a panel and the limitations of the recombinant proteins assays by covering the native kinome in, ideally unbiased, pull-down experiments.17−19 Examples for such ideas are drug-centric profiling,20−22 the KiNativ technology and other covalent acyl nucleotide probes,23,24 the kinobeads technology,25,26 or other multiplexed kinase inhibitors beads.27,28 Today, these experiments are typically analyzed using quantitative mass spectrometry as an assay readout.29 One attraction of chemical proteomic approaches for kinase inhibitor profiling is that it promotes serendipitous findings as less studied or unexpected kinases are more likely to be identified in an unbiased way. On the basis of the kinobeads technology, we have engaged in a public−private partnership with the goal to unlock the untargeted kinome by profiling

The discovery of new drugs or new targets for a given drug not only requires the definition of their mode of action but also the understanding of their potential off-target effects. In targetbased small molecules drug discovery, protein kinases are among the most frequently targeted proteins.1 Nearly half of all kinases can be mapped to disease loci,2 and one-quarter may be involved in oncogenesis.3 The highly conserved ATP pocket which the inhibitors usually address renders the development of selective drugs challenging.4−6 This has provoked the rise of multitarget kinase inhibitors in cancer therapy.7 Although these molecules may be regarded to be chemists’ failure to develop selective molecules, they have nonetheless proven successful in therapy notably owing to their synergistic effect of blocking several pathways concomitantly. However, only partial knowledge has been gained regarding the activity profiles of many of these drugs, and new targets of even well studied molecules are regularly discovered. Such findings can be a first step toward drug repurposing8,9 or aid in finding molecular mechanisms for drug toxicity. As we are witnessing the revival of phenotypic screening strategies in drug discovery,10−12 the necessity for robust drug profiling becomes even more striking. Without them, no target deconvolution13,14 would be possible, and the © 2015 American Chemical Society

Received: December 5, 2014 Published: February 8, 2015 1574

DOI: 10.1021/pr5012608 J. Proteome Res. 2015, 14, 1574−1586

Article

Journal of Proteome Research

Figure 1. Kinobeads competitive pull-down assay. (A) Kinobeads competitive assay workflow. The native kinase repertoire obtained from the lysis of different cell lines is incubated with the drug to be profiled in increasing concentrations. The affinity matrix captures proteins whose binding pocket is not occupied by the drug. Label-free LC−MS/MS readout provides dose-dependent binding inhibition curves. (B) Structures of the probes constituting the three generations of kinobeads; the wavy lines cut the covalent bond with the Sepharose beads. The frames delimit the probes of the three sets of affinity matrices: in light blue the seven probes of KBα, in purple the eight probes of KBβ and in darker blue the five probes of KBγ.

more than a thousand advanced molecules generated by the pharmaceutical industry. One ambition of this project is to identify chemical probes for poorly studied kinases and potential starting points for medicinal chemistry efforts.30 An ideal kinase profiling assay would allow for the unbiased measurement of drug−protein interactions in vivo. As this is not yet experimentally tractable, cancer cell lines are the method of choice today. An in cellulo assay would have the advantage of, e.g., accounting for the endogenous abundance levels of the kinases thus avoiding the detection of irrelevant off-targets. However, analysis of more than one cell line will generally be necessary in order to determine the full target spectrum of a molecule within the protein family. Similarly, the chemical affinity probe should ideally address all kinases at once. However, no genuine such pan-kinase probe has been identified to date.28,31−34 Instead, mixtures of broad kinase inhibitors are used in order to cover a substantial part of the kinome. The initial implementation of the kinobead workflow25 used an affinity resin containing seven immobilized chemical probes, one cell line or tissue as the source for proteins and an iTRAQ labeling strategy for quantification. We have reconsidered several features of this workflow (Figure 1A) in order to improve overall performance. In this study, we report on our efforts to maximize the number of kinases that can be profiled using the kinobead approach. We detail how kinome coverage can be optimized by combining complementary chemical probes and cellular lysates and illustrate the utility of the

optimized regime by the selectivity profiling of nine clinical kinase inhibitors.



RESULTS

Single Cell Line Kinomes

We first analyzed the recently published kinome profiles of the NCI60 cancer cell line panel35 for which the originally reported kinobeads25 were used (referred to as KBα, seven immobilized probes) (Figure 1B) in terms of kinase protein expression diversity. To do so, we developed a heuristic search algorithm based java application coined kinaseblender.36 It uses a beam search with a beam width defined as the number of sources (here cell lines) multiplied by the logarithmic growing factor ln(e+(number of sources)). By inputing a table of all the kinases experimentally enriched in each single cell line, we could obtain a rank ordered list of in silico cell line combinations with their respective number of identified kinases. According to the analysis, the seven most complementary cell lines would theoretically cover more than 80% (175 kinases) of all kinases identified in the NCI60 panel: COLO205 (colon), OVCAR8 (ovary), U251 (CNS), SKOV3 (ovary), MALME3M (skin), K562 (blood), and SR (blood) (Supplemental File 2, Supporting Information). Perhaps not surprisingly, these cell lines cover a wide range of human biologies. We next extended the analysis to additional cell lines not included in the NCI60 panel. This identified A431 (vulva), HCC827 (lung), and 1575

DOI: 10.1021/pr5012608 J. Proteome Res. 2015, 14, 1574−1586

Article

Journal of Proteome Research

decreased when more than four cell lines were combined due to the dilution of low abundance or low affinity kinases in the combined lysates: too many high abundance and potent binders are present in the lysate leading to the loss of kinases either because of insufficient abundance and/or affinity for the matrix. For instance, only nine and five kinases fewer than predicted were identified in the pull-downs of mix 2 (COLO205, SKNBE2, K562, and OVCAR8) and mix 3 (COLO205, SKNBE2, K562, and SKOV3) respectively, whereas in the case of mix 6 (nine cell lines) 27 kinases were lost with 151 kinases identified out of the predicted 178 (Supplemental File 3). When considering the number of identified kinases, mix 2 (158) and mix 3 (162) outperformed all the others (150 to 155 kinases). But the margin was not sufficient to use these numbers as the only selection criterion, especially since the amount of recovered protein is not encompassed in this comparison. For this purpose, the number of unique spectra was used as a semiquantitative measure of abundance (Figure 2B). Overall, the rationale behind the creation of a cell line mix was supported by the experimental results. Indeed, all combinations of cell lines proved superior to any cell line alone. Not only were the numbers of identified kinases greater for the mixtures than for the single cell lines, but the quantities of the captured kinases were also higher, a pledge of robustness for protein quantification. This was particularly obvious for kinases that are of low or moderate abundance in single cell lines. For these, the dilution effect caused by mixing several lysates was compensated for by a higher abundance of the respective kinase in another cell line. This is illustrated, e.g., by the comparison of the blood cell line K562 and the ovarian cell line SKOV3 where many low abundance kinases in one of the two cell lines are highly represented in the respective other ones (e.g., BTK, PRKCB, CDK5, PTK2B, EPHA2, PRKG1, EPHB2, CAMK2D, RIPK2, EPHB4, EGFR, and CAMK2G) (Supplemental File 3). At some point, however, at constant protein amount input, the dilution effect becomes dominant (e.g., illustrated by mix 6), and the highly abundant binders prevent low abundant kinases to be captured (indicated by the higher number of dark blue rectangles in the heatmap). Lysate mixtures composed of four cell lines exhibited the best overall performance. The final selection criterion was the practical ease with which cell lines can be cultivated and processed to enable chemical proteomics profiling of a large number of drugs. This parameter led to the exclusion of SKOV3 cells (included in mixtures 3 and 4) as they exhibited slow proliferation and formation of a thick extracellular matrix, resulting in poor protein recovery. Taken together, the above line of experiments and considerations led to the selection of COLO205, SKNBE2, K562, and OVCAR8 (mix 2) as the source of protein for the subsequent evaluation of chemical probes.

SKNBE2 (brain) as additional candidates in terms of kinase repertoire and diversity of anatomical origin. As the authenticity of U251 cells has been questioned,37 this particular line was removed from further considerations. The kinomes of these nine cell lines were then determined by pull-down experiments using the second generation of kinobeads developed in our laboratory (referred to as KBβ containing the seven probes of KBα and cpd15, a probe covering the AGC branch) (Figure 1B).33 Finally, kinaseblender analysis of these experiments (Supplemental File 2) determined the best cumulative combination of all of these cell lines (Figure 2A). While

Figure 2. Single cell lines and cell mixture pull-downs. (A) Stacked bar plot representing the in silico prediction of the number of kinases of the best cell mixes for any number of cell lines, based on experimental single cell line pull-downs. Each cell line is represented by a color, whereas the beige portion represents the kinases captured in at least two different cell lines of the considered combination of cells. (B) Heatmap (based on the number of unique spectra as a semiquantitative measure of abundance) of the experimentally captured kinases in the nine considered cell lines and combinations thereof. Mix 1 is a combined lysate of COLO205, SKNBE2, and SKOV3; Mix 2 of COLO205, SKNBE2, OVCAR8, and K562; Mix 3 of COLO205, SKNBE2, SKOV3, and K562; Mix 4 of COLO205, SKNBE2, SKOV3, and OVCAR8; Mix 5 of COLO205, SKNBE2, SKOV3, K562, and OVCAR8; Mix 6 of the nine considered cell lines (white: not identified, then a gradient from light blue to dark blue for 1 to more unique spectra per protein).

OVCAR8 cells expressed the highest diversity of kinases (124 out of a total of 178 kinases identified in all 9 cell lines), the highest in silico kinome coverage for a combination of, e.g., three cell lines is predicted with the cell lines COLO205, SKNBE2, and SKOV3 (163 identifiable kinases).

Kinome Purification Using Single Probe Matrices

Akin to the above optimization of the combination of cell lines, broad kinome coverage also depends on the spectrum of kinases the immobilized chemical probes are able to capture. We therefore subjected the seven affinity probes of KBα (compounds 1−7), compound 15 (part of KBβ) as well as 4 newly synthesized additional probes (compounds 18, F2, 19 and 20b) (structures in Supplemental Figure 1, Supporting Information) to single compound pull-downs, in order to identify the most potent, yet unselective and complementary probes (Supplemental File 3). Compound 18 and F2 are linkable analogues of Nintedanib and PD173074 respectively

Combined Cell Line Kinase Repertoires

The above in silico combination of cell lines suggests that kinase coverage would saturate at around four cell lines. To investigate the experimental reality of these in silico lysate poolings, pulldowns using KBβ on mixtures of three, four, five, and all nine cell line lysates were performed and compared to the most diverse single cell line experiments (Figure 2B). As expected, the predicted maximal number of kinases could not be matched in the experimental setting. In fact, the number of kinases 1576

DOI: 10.1021/pr5012608 J. Proteome Res. 2015, 14, 1574−1586

Article

Journal of Proteome Research and were developed in our laboratory34,38 because they address the growth factor receptor branch (notably, VEGFR, FGFR, and PDGFR) of the human kinome. Compounds 1932 and 20b (VI16743)31 were identified from the literature as very unselective kinase binders suitable for chemoproteomics and were thus also included in the evaluation. The number of kinases identified by the respective probes ranges from 28 for compound F2 to 226 for compound 19 (Figure 3A), the latter

(Supplemental File 3). This prompted us to include this probe in the evaluation of mixtures of probes (see below), in order not to lose this crucial kinase by probe dilution. Kinome Purification Using Mixed Probe Matrices

Having characterized the individual probes, we evaluated eight probe combinations in comparison to KBβ (Supplemental File 3). All these affinity matrices were based on the three most complementary probes (19, 15, and 1), which, in combination as the first matrix, were predicted to enrich 315 kinases (Figure 3B). One or two more good probes among probes 5, 7, 18, and 20b were added to this core matrix to generate seven other matrices. The quantity of functionalized beads was kept identical for all the pull downs, i.e., the amount of each single probe beads decreased with higher number of probes. As visualized by heatmaps (Figure 3C,D), all tested affinity matrices were better than the previous generation of kinobeads: they could enrich more kinases, and they could enrich them in higher amounts. The presence of compound 5 in the matrices granted the enrichment of EGFR without provoking any obvious dilution effect. We therefore chose to keep the 3 most complementary probes (19, 15, and 1) supplemented by probe 5 for the new generation of kinobeads. This four probe matrix was however inferior to five probe matrices in terms of abundance of the captured kinases (Figure 3C). Compounds 7, 18, or 20b performed similarly well as the fifth probe (Figure 3D and Supplemental File 3). Because probe 7 is commercially available, we elected this probe to be part of the KBγ version of kinobeads ultimately consisting of the five probes 19, 15, 1, 7, and 5 (Figure 1B).

Figure 3. Single probes and probe combinations pull downs. (A) Table showing the number of kinases captured by each probe when used in a pull-down experiment of a cell mix composed of OVCAR8, SKNBE2, COLO205, and K562. (B) Stacked bar plot displaying the prediction of the best probe combinations based on the result of single probe pull downs of a cell mix composed of OVCAR8, SKNBE2, COLO205, and K562. This ranking does not take into account intensities. The saturation is predicted to occur after four to five probes. Two probes (Cpd3 and Cpd5) do not appear in the graphic, since the kinases they are able to enrich are all also enriched by other probes. (C and D) Heatmaps (based on the number of unique spectra) of the experimentally captured kinases with affinity matrices consisting of various combinations of probes (10 mg total protein; 100 μL beads in mobicols) (C) in the COLO205, SKNBE2, K562, and OVCAR8 cell mix (D) in COLO205, SKNBE2, K562, and MV4-11 cell mix (white: not identified, then a gradient from light blue to dark blue for 1 to more unique spectra per protein).

Optimized Kinase Repertoire for Kinobeads KBγ

With KBγ in hand, we proceeded to another round of optimization of the cell lysate mixture. Additional single cell line pull-downs were performed with the four selected cell lines (see above) and using a shortlist of further seven cell lines (JURKAT, RAMOS, HELA-S3, HELA013, HEK293, HEK293T/13, and MV4-11) identified as promising candidates among the more than 100 cell lines studied by kinobeads in our laboratory thus far. MV4-11, HEK293, and HELA-S3 cells proved the best among the newly considered cell lines in terms of kinase diversity and orthogonality of repertoires (Supplemental File 3). We again considered practicality and wished to favor cell lines that could be grown in a high density fermentation device in order to be able to access large amounts of cell lysates. K562, COLO205, and MV4-11 qualified in this regard but not SKNBE2 cells due to their semiadherent nature. HEK293 and OVCAR8 failed altogether to grow in our fermenter even when using microcarriers. Out of these three cell lines, the repertoire constituted by K562 and MV4-11 was the widest for two cell lines according to a kinaseblender analysis. COLO205 did not appear as orthogonal to this repertoire as the other cell lines of our subselection. We therefore went on to evaluate six different cell lysate mixtures (Supplemental File 3), all comprising MV4-11 and K562. Three mixtures composed of only three cell lines were prepared in order to evaluate if the higher efficiency of KBγ over KBα or KBβ would translate into a quicker saturation. The other three mixtures were composed of four cell lines (all three fermentable lines, K562, COLO205, MV4-11, and HEK293 or HELA-S3 or SKNBE2). Comparison of the pull-downs results (Figure 4A) confirmed that mixing four cell lines was better than three, and the data showed that SKNBE2 was the best choice as a fourth cell line. COLO205 lysate appeared to be responsible for the

outperforming KBβ, which exhibited affinity for 213 kinases in this experimental setting. Intuitively, it is expected that a multiprobe matrix would provide better capturing capacities by concatenation of the kinomes enriched by the probes composing it. Using kinaseblender, we could hence determine the best combination of probes as defined by their capture potential, i.e., the widest concatenated kinome for a chosen number of probes. For 12 probes, it amounts to 354 kinases. This number would ideally be reached with 10 of the 12 evaluated probes rendering 2 probes superfluous (Figure 3B). The rapid saturation of the number of additionally captured kinases with increasing number of probes is obvious in this graphical representation. Experimentally, this effect would likely be more pronounced due to the accumulation of a core kinome caused by the binding of kinases with strong affinity for several probes. Probes 3 and 5 were identified as superfluous as they offered no benefit in terms of kinome coverage. However, a more refined analysis of the pull-downs showed that only compound 5 had a good affinity for EGFR, whereas other probes could enrich this protein but not in large amounts 1577

DOI: 10.1021/pr5012608 J. Proteome Res. 2015, 14, 1574−1586

Article

Journal of Proteome Research

Figure 4. Optimized kinase repertoire and affinity matrix. (A) The heatmap (based on the number of unique spectra) shows the captured kinases by KBγ in six different cell mixtures (white: not identified, then a gradient from light blue to dark blue for 1 to more unique spectra of the protein). In green, the kinases whose amount is largely provided by COLO205, in violet by SKNBE2 and in orange by HEK293 or HELAS3. (B) Heatmap of the identified kinase-domain containing proteins from the optimized kinase repertoire. The two replicates for KBα, KBβ, and KBγ show the efficiency of the three different affinity matrices, with the number of unique spectra as semiquantitative measure of abundance (white: not identified, then a gradient from light blue to dark blue for 1 to more unique spectra per protein).

Figure 5. Comparison of Cpd19 and CTx-0294885. (A and B) Structures and Venn diagrams of the kinases captured in two biochemical pull down replicates with single probe affinity matrices using the COLO205, SKNBE2, K562, and MV4-11 cell mix. The structural differences between the two probes are colored. (C) Heatmap showing the kinase enrichment efficiency of Cpd19 and CTx-0294885 as well as KBγ with the number of unique spectra as semiquantitative measure of abundance (white: not identified, then a gradient from light blue to dark blue for 1 to more unique spectra of the protein). Each affinity matrix evaluation was performed in biochemical duplicate (R1 and R2). (D) Venn diagrams showing the overlap between the kinases captured by Cpd19 and CTx-0294885 (left: overlap between the kinases captured in both replicates for each probe; right: overlap between the sum of the two replicates for both probes).

main contribution of EphA1, EphA4, FRK, and LCK to the repertoire, whereas SKNBE2 cells contributed with EphA5, PDGFRB, DDR2, RSK4, and ATM (Figure 4A). The addition of HEK293 or HELA-S3 lysates contributed high amounts of PRKAA2, BRK, and PKG1. However, the total number of kinases in the respective mixtures did not add substantially to overall kinome coverage. The dendogram shown in Supplemental Figure 2 visualizes the coverage of the human kinome with this combination of probes and cell lines. It is apparent that the kinome is covered over all branches of the phylogenetic tree. As a result, we chose a 1:1:1:1 mixture of lysates from MV4-11 (fermenter or roller bottles), COLO205 (fermenter), K562 (fermenter or roller bottles), and SKNBE2 (dishes) cells to support our ongoing profiling efforts of hundreds of kinase inhibitors.

AKT2, CSNK1A1, CSNK1D, EGFR, LATS1, and MELK, all kinases that would also be pulled down in a satisfactory manner by KBγ. TTK, STK38 and, to a lesser extent, PNK3, CHK2, PRKCQ, and MAST3 were captured by CTx-0294885 but not Cpd19 or KBγ. Conversely, the protein kinases CDK6, CDK16, CDK17, MAP4K2, ERN1, ERN2, MAP3K6, mTOR, MST4, TGFBR2, STK11, TAOK2, CAMK4, and, to a lesser extent, ICK, MAP2K3, MAP4K5, MAP3K5, MAP2K6, and BMPR2 were found to be good targets of Cpd19 but not CTx-0294885 (Supplemental Figure 3, Supplemental File 3). Taken together, CTx-0294485 is an interesting chemical probe but overall inferior to probe 19 and KBγ, so there was no compelling reason to add this compound to our set of kinobead probes. We could therefore focus on further improving the workflow by optimizing the MS readout.

Validation of KBγ

To confirm that the optimization of kinobeads was successful, we compared the merits of the three sets of probes KBα, KBβ, and KBγ directly by performing duplicate pull-down experiments from the same lysates, and back to back mass spectrometry measurements. This experiment clearly showed the superiority of the newly developed affinity matrix over the older generations of kinobeads with a greater number and intensities of captured kinases (Figure 4B; Supplemental File 3). As we were reaching this conclusion, a new probe, CTx0294885, was published, which is chemically almost identical to probe 19 (structures in Figure 5A,B). It was described to capture more than half of the expressed kinome in MDA-MB231 and, therefore, could be a beneficial addition to the affinity probe repertoire.28 Hence we synthesized CTx-0294885 and compared its kinase binding potential to probe 19 and KBγ (Supplemental File 3). The kinase profiles of CTx-0294885 and probe 19 were comparable with differences in detail (Figure 5C,D). Notably, CTx-0294885 was able to enrich AKT1,

Optimization of the MS Readout

As mass spectrometry measurement is the bottleneck of the workflow, changes in the measurement conditions were investigated to allow for larger throughput. All the kinobeads and cell mix developments were implemented with a 215 min acidic gradient from 7% to 35% acetonitrile in water (0.1% formic acid). Following our work showing the advantage of DMSO in LC−MS/MS gradients,39 we were very pleased to be able to reduce the LC−MS gradient time from 215 to 90 min on a Thermo Orbitrap Elite using 5% DMSO as an additive to the gradients (4−32% acetonitrile in water with 0.1% formic acid and 5% DMSO) with no deleterious effect on the identification of kinases (Supplemental Figure 4). The carryover between successive measurements was found to be negligible: we found the ratio of the sum of kinase intensities identified in a blank to the one of the preceding pull-down 1578

DOI: 10.1021/pr5012608 J. Proteome Res. 2015, 14, 1574−1586

Article

Journal of Proteome Research

Figure 6. Targets of nine drugs. Protein kinase targets of the nine drugs profiled with the kinobeads competitive assay represented on the kinome tree (illustration reproduced courtesy of Cell Signaling Technology, Inc (www.cellsignal.com)). The size of the dots reflects the affinity of Alvocidib (blue), Crizotinib (dark blue), Dasatinib (yellow), Fasudil (red), Hydroxyfasudil (mauve), Nilotinib (orange), Ibrutinib (dark red), Imatinib (turkese), and Sunitinib (light green) for the kinases. Only drug−protein affinities below 1 μM are shown.

by an intensity-based label-free approach using MaxQuant,40

measurement to be only 0.2%. The dose−response samples (Figure 1A) could therefore be measured without in-between blanks, allowing for a precious gain of time. The mass spectrometer was run in data-dependent acquisition mode with an inclusion list containing the three most intense peptides for each kinase, with a 20 min window centered on the observed retention time. The implementation and regular update of this list contribute to the better identification of kinases. Quantification and data normalization were achieved

which is an easy, cheap yet reliable quantification technique with no limit in the multiplexing capacities.29 Thanks to the implementation of maximal peptide ratio extraction and delayed normalization algorithms, the MaxLFQ program included in MaxQuant provides accurate XIC-based quantification.41 1579

DOI: 10.1021/pr5012608 J. Proteome Res. 2015, 14, 1574−1586

Article

Journal of Proteome Research

Figure 7. Heatmap and dose−response curves. (A) Heatmap showing the protein kinase targets of Alvocidib, Crizotinib, Dasatinib, Fasudil, Hydroxyfasudil, Nilotinib, Ibrutinib, Imatinib, and Sunitinib found in the kinobeads competition assay. Colored rectangles (the darker the blue, the stronger the binding) are only shown for kinases with Kd values (or IC50 for Ibrutinib) of less than 1 μM. (B and C) Nine dose kinobeads competition assay binding inhibition curves of native PKN1, PRKX, ROCK1, and ROCK2 fitted with the 4 parameters log−logistic model, and 11 dose activity assay inhibition curves of recombinant PKN1, PRKX, ROCK1, and ROCK2 fitted with the 3 parameters log−logistic model for Fasudil (B) and Hydroxyfasudil (C). The error bars for the IC50 represent a ± standard deviation.

Affinity Pull-Downs Can Be Parallelized in 96-Well Plates

which were manually compiled from 20 different sources4,6,16,25,43−58 (Supplemental Tables 1−9). We believe that this drug target compilation is of further value as an extensive target catalogue of these nine kinase inhibitors. Figure 6 shows the individual protein kinase targets (Kd cutoff of 1 μM) for all nine inhibitors plotted on a phylogenetic kinase tree. From this illustration it appears for instance obvious that Dasatinib is quite unselective but still preferentially targets tyrosine kinases, whereas Sunitinib targets can be found across many families. To further compare the targets of the nine drugs, a heatmap was generated showing the various protein kinase targets of the molecules with a Kd cutoff of 1 μM (Figure 7A). This simple representation gives a clear idea of selectivity target overlaps. As this study was carried out using the same assay procedure and lysate, we believe that this data set enables accurate assessment of the relative target space. We will not describe in details the profiles of the drugs, as this is beyond the purpose of this assay development article, but we wish to illustrate hereafter the potential of our assay for drug profiling by the example of Fasudil and Hydroxyfasudil. Further considerations about Alvocidib as a disruptor of glycogen metabolism and about CDK complexes as evidenced by Alvocidib dose−response can be found in the Supporting Information.

To enhance the potential drug profiling throughput, we reevaluated our initial procedure25 to make it amenable to a 96well plate assay: the amount of beads was reduced to 35 μL of settled beads per well of a 96-well plate instead of 100 μL when using mobicol columns, while the amount of proteins was reduced from 10 mg (for the development of the cell mix and probe mix) to 1, 2, 3, 4, and 5 mg. The number of kinases identified with 5 mg of protein (258 kinases) was larger than with lower amounts of incubated proteins and compared well with the 10 mg proteins/100 μL beads mobicol procedure (261 kinases) (Supplemental Figure 5). Moreover, it has to be noted that the scale-up of KBγ does not present any significant difficulty since less than 1 mg of each compound is necessary for 150 pull-downs in these conditions (35 μL of settled kinobeads, i.e., 7 μL of each single probe beads loaded at less than 1 mg/mL beads). Profiling of Nine Kinase Inhibitors

To prove the large scale applicability and performance of our optimized 96-well plate assay, we profiled the nine approved or very advanced kinase inhibitors Alvocidib, Crizotinib, Dasatinib, Fasudil, Hydroxyfasudil, Nilotinib, Ibrutinib, Imatinib, and Sunitinib. The competitive assays with doses of 0 nM, 3 nM, 10 nM, 30 nM, 100 nM, 300 nM, 1 μM, 3 μM, and 30 μM allowed us to establish the profiles of the inhibitors (Supplemental File 4). IC50 curves were generated using an in-house R script for all quantifiable proteins and manually checked for quality. In order to account for the depletion in kinases and allow a conversion from IC50 to Kd for the interaction of each kinase with the drug, a correction factor26,42 was applied. This depletion factor which is only valid for reversible (not for Ibrutinib) ATP competitive inhibition was derived from a pull down of pull down experiment. The obtained kinobead profiles were then compared against known targets of the nine drugs,

PRKX Is the Main Target of Fasudil, but It Is Not Inhibited by Hydroxyfasudil

The main target of Fasudil was found to be PRKX with a Kd of 201 nM (uncorrected IC50 of 330 nM), whereas it showed a Kd of 632 and 597 nM (uncorrected IC50 of 1.1 μM and 1.0 μM respectively) for ROCK1 and 2. The drug additionally displayed a Kd of 774 nM for PKN1 (uncorrected IC50 of 1.0 μM). These four targets were found to be the only proteins bound with a Kd inferior to 2 μM (Figure 7B; Supplemental Tables 4 and 5). This activity on PRKX had already been observed, but no IC50 or Kd was available since all reported experiments were performed with a single or two high doses 1580

DOI: 10.1021/pr5012608 J. Proteome Res. 2015, 14, 1574−1586

Article

Journal of Proteome Research

many other potential areas of application also be of great interest for laboratories who want to extract proteotypic peptides for MRM kinome studies.

(Supplemental Table 4). Indeed, the kinase activity was reduced by 77% at 10 μM in a split-luciferase screen focusing on 27 kinases of the AGC group, with no comparison possible with the activity on ROCK.49 This inhibition was also evidenced in the comprehensive assay of recombinant kinase catalytic activity conducted with the radiometric HotSpot assay (23% residual activity at 500 nM) with lesser potency for ROCK2 (38% residual activity), PKN1 (41%) and a very low one for ROCK1 (60% remaining activity)47 and in an Ambit KinomeScan binding assay (6.9% remaining binding compared to control).58 This finding was confirmed more recently in two recombinant protein activity assays at 1 and 10 μM placing PRKX (respectively 21% and 2% remaining activity) as the main target at the same level as ROCK2 (25% and 4%).16 We decided to also profile Hydroxyfasudil, which is a known active metabolite of Fasudil.59 The in vivo hydroxylation occurs quickly: it has been reported that Fasudil is cleared from the blood with a half-life of 15 min after intravenous injection in humans, while Hydroxyfasudil remained present as long as 8 h after injection.60 The effect of the drug might then only relate to the activity of the metabolite which is eliminated more slowly than the parent molecule. It is therefore interesting to observe that, for Hydroxyfasudil, ROCK1 binding (Kd) is improved (268 nM), ROCK2 activity maintained (641 nM) but PRKX binding is largely reduced (1.85 μM) (Figure 7C). We corroborated these findings by activity assays (Supplemental File 5) whose results follow the exact same trend: the activity for PRKX decreased from 84 nM for Fasudil to 1.7 μM for Hydroxyfasudil, while the activity for ROCK1, ROCK2, and PKN1 changed from 379, 250, and 394 nM to respectively 171 nM, 231 nM, and 1.26 μM (Figure 7B,C; Supplemental Tables 4 and 5).



Potential of the Assay

We hope that the tedious work of optimization reported here will provide a stepstone for others to start their chemical proteomics efforts. With this 96-well plate assay, the pull downs of eight drug dose responses can now easily be done in one plate with less than 4 h spent for biochemical sample handling. The mass spectrometry readout for one drug profile (nine doses and one pulldown of pulldown) covering half the native human kinome amounts to only 1 day. With nine concentrations, the dose−response curves are robust enough to calculate IC50’s and deduce Kd’s. This offers a distinct advantage over the very often performed single dose profiling assays. With this enhanced practicality and robustness, the potential of the chemical proteomic approach can be fully exploited. Here, we have illustrated its possibilities by profiling nine already well-known and studied drugs. In the frame of a medicinal chemistry program for newly developed molecules, the assay has the potential to act as a primary and selectivity assay with full-length native proteins. New insights can be gained using this technology with even thoroughly studied small molecules, such as Fasudil, which is marketed and considered as a selective ROCK inhibitor. We were indeed first surprised to observe that the main target of the drug is actually PRKX, while the main targets of its metabolite, Hydroxyfasudil, are indeed ROCK1 and ROCK2. After confirming these results by an activity assay, we can deduce the following: whereas the core of Hydroxyfasudil as compared to Fasudil can rotate inside the ATP pocket for the hydroxy moiety to build another binding with the hinge of the ROCKs,54 this seems impossible in the PRKX pocket without a loss of affinity. Depending on the kinetics of action of Fasudil, the activity for PRKX could be of no relevance, but it could also, as an off-target, help explain some effects of the drug. Indeed, this kinase has been shown to be involved in some resistance mechanism61 and to regulate the three main processes of angiogenesis in endothelial cells: proliferation, migration, and vascular-like structure formation.62 When studying the effect of Fasudil as a ROCK inhibitor in the context of tumor progression,63,64 this extra player might have to be also considered.

DISCUSSION

Useful Probe and Cell Kinome Resource

In the design of an improved chemical proteomics based assay suitable for selectivity screening of kinase inhibitors, cell lysate composition as well as affinity probes needed to be improved by iteration. A better cell mix helped define a better probe mix which in turn allowed choosing better cell lines. A discussion on the criteria for selecting probe-candidates and cellcandidates for routine screenings of inhibitors can be found in the Supplemental Results and Discussion. Overall, we found Cpd19 to be the best potent unselective kinase inhibitor of our selection of 12 compounds, and it even compared favorably to the very similar recently disclosed CTx-0294885. Together with Cpd15, Cpd1, Cpd7, and Cpd5, it constitutes what we believe to be the best kinase affinity matrix to date. We must stress out that the five probes were selected to cover the most diverse repertoire of kinases based on the results of single probe affinity pull downs. Our constant monitoring and development of new probes over the last 7 years have led us to believe that the probes evaluated in this study are the most promising affinity probes reported. Hence, the database of single probe affinity pull downs constitutes a very useful resource for rationally tailoring other affinity matrices following the same progression but with another outcome than increased kinome coverage in mind. The database of cell line kinomes generated during the study (freely available at https://www.proteomicsdb.org/) will be of great help to generate other cell line combinations that would be suited for studies of drug interactions with a particular subkinome. With all the raw files available, this work will among



CONCLUSIONS For the design of drugs and tool compounds as well as for target deconvolution, chemical biologists and medicinal chemists need practical and unbiased selectivity assays. As kinase inhibitors constitute a large part of the pharmacopeia (28 approved drugs, 250 in phase I−III), particularly for the treatment of cancers, the promiscuity of these molecules constitutes a challenge faced by the academic groups and the pharmaceutical industry. Avoiding or understanding off-target activity necessitate kinome-wide selectivity assays. Panel screening with recombinant proteins are traditionally used for this purpose but cannot encompass the complexity of the native kinome. Chemical proteomics can overcome this limitation. The described affinity matrix containing five unselective kinase inhibitors allows for the efficient capture of native kinases distributed across the whole kinome tree. An optimized combination of four cell lines displays a large repertoire of kinases, which can be pulled down by this new generation of kinobeads. Associated with label-free quantitative nanoLC− 1581

DOI: 10.1021/pr5012608 J. Proteome Res. 2015, 14, 1574−1586

Article

Journal of Proteome Research

triethylamine (15 μL/mL beads) in DMSO (1 vol of DMSO for 1 vol of beads). Aminoethanol (50 μL/mL beads) was then added, and the mixture was further shaken for an extra 20 h at r.t. in the dark. The beads were then washed with DMSO (10 mL/mL beads) and ethanol (3 × 10 mL/mL beads) and stored in ethanol (1 mL/mL beads) at 4 °C. For the carboxylic acid linkable inhibitors, NHS-Sepharose beads were washed with DMSO and reacted with a 4:1 mixture of aminoethanol (9.7 μL/mL beads) and ethylenediamine (2.7 μL/mL beads) for 20 h on an end-over-end shaker at r.t. in the dark in the presence of triethylamine (15 μL/mL beads) in DMSO (1 vol of DMSO for 1 vol of beads). The beads were then washed with DMSO (3 × 10 mL/mL beads) and DMF (15 mL/mL beads) and reacted with the compounds (1−4 μmol/mL beads), Hünig’s base (3.5 μL/mL beads), triethylamine (20 μL/mL beads), and PyBrOP (4.7 mg/mL beads) for 20 h at r.t. in the dark. After being washed with DMSO (3 × 10 mL/mL beads), the beads were reacted with NHS acetate (10 μmol/mL beads) in the presence of triethylamine (20 μL/mL beads) in DMSO (1 vol DMSO for 1 vol beads) overnight at r.t.. The beads were then washed with DMSO (10 mL/mL beads) and isopropanol (3 × 10 mL/mL beads) and stored in isopropanol (1 mL/mL beads) at 4 °C. Aliquots of the supernatants before and after coupling were controlled by LC−MS to conclude of the completion of the reactions. The affinity matrices were prepared by mixing the functionalized beads in equal volume amounts, i.e., for KBγ, a 1:1:1:1:1 mixture of the probes 1, 5, 7, 15, and 19 (all beads loaded with 2 μmol of probe/mL of beads, but probe 1 beads loaded at 1 μmol/mL beads).

MS/MS, the method can be used to perform unbiased drug profilings: the affinities of one small molecule for more than 250 human native kinases can be determined in one chemical proteomics experiment, as can binding to other proteins and information about complexes be obtained. The finding that PRKX is the main target of Fasudil but not of its metabolite Hydroxyfasudil showcases the potential of the assay. Moreover, the systematic study of the enrichment capabilities of 13 probes provides a useful resource to tailor other affinity matrices for targeted studies of human subkinomes.



EXPERIMENTAL PROCEDURES

Cell Culture and Lysis

K562 and MV4-11 cells were grown in roller culture in RPMI1640 medium supplemented with 10% FBS and antibiotic antimycotic solution (Sigma A5955). Cells were harvested upon density (approximately 5e6), centrifuged, and washed with cold PBS. SKNBE2 cells were grown in stationary culture (15 cm dishes) in DMEM/HAMS medium supplemented with 10% FBS and antibiotic antimycotic solution (Sigma A5955). Cells were harvested upon confluence by mechanical detachment followed by centrifugation and washing with cold PBS. COLO205 cells were produced by fed-batch fermentation in 1.8 L scale in a Braun Biostat B2 (37 °C, pH = 7.1 CO2 regulated, 30 rpm, O2 maintained at starting value) in RPMI1640 medium supplemented with 10% FBS and antibiotic antimycotic solution (Sigma A5955). Cells were harvested upon density (approximately 5e6), centrifuged, and washed with cold PBS. The cells were lysed in lysis buffer containing 50 mM Tris/ HCl, pH 7.5, 5% glycerol, 1.5 mM MgCl2, 150 mM NaCl, 1 mM Na3VO4, 0.8% NP40, 0.375 mM NaF, 1 mM DTT including protease inhibitors (SigmaFast protease inhibitor tablet S8820) and phosphatase inhibitors (Phosphatase Inhibitor Cocktail 3, Sigma-Aldrich, Munich, Germany). The lysate was ultracentrifuged for 1 h at 4 °C and 145000g.

96-Well Plate Kinobead Competition Assay

The cell lysates were diluted with equal volumes of 1× compound pull down (CP) buffer (50 mM Tris/HCl, pH 7.5, 5% glycerol, 1.5 mM MgCl2, 150 mM NaCl, 20 mM NaF, 1 mM sodium ortho-vanadate, 1 mM DTT), protease inhibitor (SigmaFast protease inhibitor tablet S8820) and phosphatase inhibitors (Phosphatase Inhibitor Cocktail 3, Sigma-Aldrich, Munich, Germany). If required, lysates were further diluted to a final protein concentration of 5 mg/mL using 1× CP buffer supplemented with 0.4% NP-40 (CP-0.4). The cell mixes were prepared as 1:1:...:1 mixtures regarding the total amount of proteins as determined by Bradford assay. For selectivity profiling experiments in 96-well plates, the diluted cell mix lysates (5 mg of total proteins/well) were incubated for 45 min at 4 °C in a end-over-end shaker with 0 nM (DMSO control), 3 nM, 10 nM, 30 nM, 100 nM, 300 nM, 1 μM, 3 μM, or 30 μM of the kinase inhibitor dissolved in DMSO. KBγ (35 μL settled beads resuspended with 50% glycerol, washed with water, CP buffer and equilibrated with CP-0.4 buffer) were incubated with the lysates at 4 °C for 30 min. The DMSO control lysate was recovered and incubated similarly with KBγ as a pull down of pull down experiment in order to calculate the depletion factor. The beads were then washed (3 mL of CP-0.4 buffer followed by 2 mL of CP-0.2) and the bound proteins subsequently eluted by incubation for 30 min at 50 °C with 60 μL of 2× NuPAGE LDS sample buffer (Invitrogen, Darmstadt, Germany) containing 50 mM DTT and centrifugation. The reduced eluates (30 μL) were alkylated with chloroacetamide (3 μL, 550 mM), and the proteins were desalted and concentrated by a short electrophoresis (about 0.5

Affinity Matrices Preparation

Compounds 1 (linkable PD173955), 2 (CZC8004), 3 (linkable Sunitinib), 4 (Staurosporine), 5 (linkable Vandetanib), 6 (bisindolylmaleimide III), and 7 (Purvalanol B) were commercially sourced. Cpd15 (AKT probe),33 F2 (linkable PD173074),38 and Cpd18 (linkable Nintedanib)34 were synthesized as reported previously. Cpd19 and CTx-0294885 were synthesized in two steps: (1) regioselective displacement of the 4-Cl of 2,4,5-trichloropyrimidime (1 equiv) with the relevant aniline (2-amino-Nmethylbenzenesulfonamide obtained in two steps from 2nitrobenzenesulfonyl chloride or commercially available 2amino-N-methylbenzamide) (1 equiv) in DMF in the presence of cesium carbonate (2 equiv) at r.t.; (2) camphorsulfonic acid (1 equiv) catalyzed 2-Cl displacement with the second aniline (4-[(N-Boc)aminoethyl]aniline or 1-(4-aminophenyl)piperazine) (1 equiv) in dry isopropanol at 80 °C. Cpd20b (VI16743) was generously provided by Dr. Josef Wissing in Helmholtz Centre for Infection Research, Braunschweig, Germany. For the amino linkable inhibitors, NHS-Sepharose beads (Amersham Biosciences) were washed with DMSO and reacted with the compounds (1−4 μmol/mL beads) for 20 h on an end-over-end shaker at r.t. in the dark in the presence of 1582

DOI: 10.1021/pr5012608 J. Proteome Res. 2015, 14, 1574−1586

Article

Journal of Proteome Research

the beads (f) remains constant, then the amount of kinase captured in the first pull down is f T, and it is f(T − f T) in the pull down of pull down, where T is the total amount of the kinase in the lysate and hence (T − f T) the amount of free kinase in the lysate during the first pull down. Then

cm) on a 4−12% NuPAGE gel (Invitrogen). In-gel trypsin digestion was performed according to standard procedures. LC−MS/MS Analysis

The optimized conditions for the liquid chromatography tandem mass spectrometry measurements feature an Eksigent nanoLC-Ultra 1D+ (Eksigent, Dublin, CA) coupled to an Orbitrap Elite instrument (Thermo Scientific, Bremen, Germany). The peptides were delivered to a trap column (100 μm × 2 cm, packed in-house with Reprosil-Pur C18-AQ 5 μm resin, Dr. Maisch, Ammerbuch, Germany) at a flow rate of 5 μL/min in 100% solvent A (0.1% formic acid, FA, in HPLC grade water). After 10 min of loading and washing, peptides were transferred to an analytical column (75 μm × 40 cm, packed in-house with Reprosil-Gold C18, 3 μm resin, Dr. Maisch, Ammerbuch, Germany) and separated at a flow rate of 300 nL/min using a 90 min gradient ranging from 4% to 32% solvent C in B (solvent B: 0.1% FA and 5% DMSO in HPLC grade water, solvent C: 0.1% FA and 5% DMSO in acetonitrile). The eluent was sprayed via stainless steel emitters (Thermo) at a spray voltage of 2.2 kV and a heated capillary temperature of 275 °C. The Orbitrap Elite instrument was operated in data-dependent mode, automatically switching between MS and MS2. Full scan MS spectra (m/z 360−1300) were acquired in the Orbitrap at 30 000 (m/z 400) resolution using an automatic gain control (AGC) target value of 1e6 charges. Tandem mass spectra of up to 15 precursors were generated in the multipole collision cell by using higher energy collisional dissociation (HCD) (AGC target value 2e4, normalized collision energy of 30%) and analyzed in the Orbitrap at a resolution of 15 000. Precursor ion isolation width was set to 2.0 Th, the maximum injection time for MS/MS was 100 ms, and dynamic exclusion was set to 20 s.

r = f (T − fT )/(fT ) = 1 − f

and f=1−r

Using the definition of KdKB, [KB] = KdKB[KB· kinase]/[kinase] = KdKB·fT /(T − fT ) = KdKB(1 − r )/r

with [KB] the “concentration” of free kinobeads, [kinase] the concentration of the free kinase, and KdKB the dissociation constant of the kinase with the kinobeads. For a competitive binding assay with a constant initial “concentration” of kinobeads, the Cheng-Prusoff equation can be used. The dissociation constant for the inhibitor to the kinase is therefore K i = IC50/(1 + [KB]/KdKB) = IC50 /(1 + KdKB(1 − r )/(r·KdKB)) = r ·IC50



ASSOCIATED CONTENT

S Supporting Information *

Peptide and Protein Identification and Quantification

Supplemental File 1 comprises six supplemental figures (structures of the 12 probes; kinome tree representation showing the coverage of KBγ; comparison of the differentially expressed kinases enriched by immobilized CTx-0294885, Cpd19, or KBγ; Venn diagram showing the overlap of the kinases identified using two different gradients for the MS readout of a pulldown experiment; Venn diagrams showing the overlaps between the kinases enriched with KBγ in different conditions; Alvocidib dose response curves for selected complexes), supplemental results and discussion (choice of the probes; choice of the cell lines; the assay informs about complexes: the example of CDK complexes as evidenced by Alvocidib dose−response; a more global view of the effect of drugs: the example of Alvocidib as a disruptor of glycogen metabolism) and nine supplemental tables (Alvocidib, Crizotinib, Dasatinib, Fasudil, Hydroxyfasudil, Nilotinib, Ibrutinib, Imatinib, and Sunitinib profiles compared to literature data). Supplemental File 2 shows the output of the kinaseblender analysis of the NCI60 panel. Supplemental File 3 comprises all the scaffold outputs for the single cell line and single probe pulldowns, as well as for the combinations of cells and probes. Supplemental File 4 is the assembly of the MaxQuant search Protein Groups output for each drug profiling. Supplemental File 5 shows the results of the activity assays for Hydroxyfasudil and Fasudil. This material is available free of charge via the Internet at http://pubs.acs.org. Raw MS data, Scaffold files, and pdf files containing dose−response curves are to be found in ProteomicsDB67 at https://www. proteomicsdb.org/ identifier PRDB004111.

For the cell lines and probes selection process, peak lists were extracted from MS data files using Mascot Distiller v2.2.1 (Matrix Science, U.K.) and subsequently searched against the Human IPI database version v3.68 using carbamidomethyl cysteine as a fixed modification, and N-terminal protein acetylation and methionine oxidation as variable modifications. Trypsin was specified as the proteolytic enzyme, and up to two missed cleavages were allowed. The mass tolerance of the precursor ion was set to 10 ppm and for fragmentations to 0.05 Da. The Dat files were further processed with the Percolator algorithm (version 1.14)65 and then loaded into Scaffold (version 3.6.3) using the mascot threshold to adjust the FDR to less than 1% for proteins and peptides. For drug profiling, MaxQuant software (version 1.4.0.5) and Uniprot database (v22.07.143) were used for intensity based label free quantification. For identification 0.01 peptide and protein FDRs were used. Feature matching between raw files was enabled, using a match time window of 1 min. Averaged LFQ intensity values were used to calculate protein ratios with the DMSO sample as reference. MQ data were filtered for reverse identifications (false positives) and contaminants. IC50 curves were drawn using an in-house software tool adapting an open source statistical software tool.66 Depletion Factor

The depletion factor (r) for a kinase is defined as the ratio of the amount of a kinase captured in the pull down of pull down to the amount of the kinase captured in the pull down experiment. Assuming that the fraction of a kinase captured by 1583

DOI: 10.1021/pr5012608 J. Proteome Res. 2015, 14, 1574−1586

Article

Journal of Proteome Research



(12) Moffat, J. G.; Rudolph, J.; Bailey, D. Phenotypic screening in cancer drug discovery - past, present and future. Nature Rev. Drug Discovery 2014, 13 (8), 588−602. (13) Terstappen, G.; Schlüpen, C.; Raggiaschi, R.; Gaviraghi, G. Target deconvolution strategies in drug discovery. Nature Rev. Drug Discovery 2007, 6 (11), 891−903. (14) Ziegler, S.; Pries, V.; Hedberg, C.; Waldmann, H. Target Identification for Small Bioactive Molecules: Finding the Needle in the Haystack. Angew. Chem., Int. Ed. Engl. 2013, 52, 2744−2792. (15) Drewry, D. H.; Bamborough, P.; Schneider, K.; Smith, G. K. The Kinome and its Impact on Medicinal Chemistry. Kinase Drug Discovery 2012, 1−29. (16) Gao, Y.; Davies, S.; Augustin, M.; Woodward, A.; Patel, U.; Kovelman, R.; Harvey, K. A broad activity screen in support of a chemogenomic map for kinase signalling research and drug discovery. Biochem. J. 2013, 451 (2), 313−328. (17) Schirle, M.; Bantscheff, M.; Kuster, B. Mass spectrometry-based proteomics in preclinical drug discovery. Chemistry & biology 2012, 19 (1), 72−84. (18) Rix, U.; Superti-Furga, G. Target profiling of small molecules by chemical proteomics. Nat. Chem. Biol. 2009, 5 (9), 616−624. (19) Xiao, Y.; Wang, Y., Global discovery of protein kinases and other nucleotide-binding proteins by mass spectrometry. Mass Spectrom Rev., DOI: 10.1002/mas.21447. (20) Rix, U.; Hantschel, O.; Dürnberger, G.; Remsing Rix, L.; Planyavsky, M.; Fernbach, N.; Kaupe, I.; Bennett, K.; Valent, P.; Colinge, J.; Köcher, T.; Superti-Furga, G. Chemical proteomic profiles of the BCR-ABL inhibitors imatinib, nilotinib, and dasatinib reveal novel kinase and nonkinase targets. Blood 2007, 110 (12), 4055−4063. (21) Rix, U.; Remsing Rix, L.; Terker, A.; Fernbach, N.; Hantschel, O.; Planyavsky, M.; Breitwieser, F.; Herrmann, H.; Colinge, J.; Bennett, K.; Augustin, M.; Till, J.; Heinrich, M.; Valent, P.; SupertiFurga, G. A comprehensive target selectivity survey of the BCR-ABL kinase inhibitor INNO-406 by kinase profiling and chemical proteomics in chronic myeloid leukemia cells. Leukemia 2010, 24 (1), 44−50. (22) Ong, S. E.; Schenone, M.; Margolin, A. A.; Li, X.; Do, K.; Doud, M. K.; Mani, D. R.; Kuai, L.; Wang, X.; Wood, J. L.; Tolliday, N. J.; Koehler, A. N.; Marcaurelle, L. A.; Golub, T. R.; Gould, R. J.; Schreiber, S. L.; Carr, S. A. Identifying the proteins to which smallmolecule probes and drugs bind in cells. Proc. Natl. Acad. Sci. U. S. A. 2009, 106 (12), 4617−22. (23) Patricelli, M.; Nomanbhoy, T.; Wu, J.; Brown, H.; Zhou, D.; Zhang, J.; Jagannathan, S.; Aban, A.; Okerberg, E.; Herring, C.; Nordin, B.; Weissig, H.; Yang, Q.; Lee, J.-D.; Gray, N.; Kozarich, J. In situ kinase profiling reveals functionally relevant properties of native kinases. Chem. Biol. 2011, 18 (6), 699−710. (24) Xiao, Y.; Guo, L.; Wang, Y. A Targeted Quantitative Proteomics Strategy for Global Kinome Profiling of Cancer Cells and Tissues. Mol. Cell Proteomics 2014, 13 (4), 1065−75. (25) Bantscheff, M.; Eberhard, D.; Abraham, Y.; Bastuck, S.; Boesche, M.; Hobson, S.; Mathieson, T.; Perrin, J.; Raida, M.; Rau, C.; Reader, V. r.; Sweetman, G.; Bauer, A.; Bouwmeester, T.; Hopf, C.; Kruse, U.; Neubauer, G.; Ramsden, N.; Rick, J.; Kuster, B.; Drewes, G. Quantitative chemical proteomics reveals mechanisms of action of clinical ABL kinase inhibitors. Nature Biotechnol. 2007, 25 (9), 1035− 1044. (26) Lemeer, S.; Zörgiebel, C.; Ruprecht, B.; Kohl, K.; Kuster, B. Comparing Immobilized Kinase Inhibitors and Covalent ATP Probes for Proteomic Profiling of Kinase Expression and Drug Selectivity. J. Proteome Res. 2013, 12 (4), 1723−1731. (27) Duncan, J.; Whittle, M.; Nakamura, K.; Abell, A.; Midland, A.; Zawistowski, J.; Johnson, N.; Granger, D.; Jordan, N.; Darr, D.; Usary, J.; Kuan, P.-F.; Smalley, D.; Major, B.; He, X.; Hoadley, K.; Zhou, B.; Sharpless, N.; Perou, C.; Kim, W.; Gomez, S.; Chen, X.; Jin, J.; Frye, S.; Earp, H.; Graves, L.; Johnson, G. Dynamic reprogramming of the kinome in response to targeted MEK inhibition in triple-negative breast cancer. Cell 2012, 149 (2), 307−321.

AUTHOR INFORMATION

Corresponding Authors

*(G.M.) E-mail: [email protected]. *(B.K.) E-mail: [email protected]. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS We would like to express our gratitude to Michaela KroetzFahning, Andrea Hubauer, and Andreas Klaus for their excellent technical assistance. We also thank Prof. Thomas Hofmann of the Chair of Food Chemistry and Molecular Sensory Science, Technische Universität München for performing NMR experiments.



REFERENCES

(1) Cohen, P. Protein kinases–the major drug targets of the twentyfirst century? Nature Rev. Drug Discovery 2002, 1 (4), 309−315. (2) Manning, G.; Whyte, D.; Martinez, R.; Hunter, T.; Sudarsanam, S. The protein kinase complement of the human genome. Science (New York, N.Y.) 2002, 298 (5600), 1912−1934. (3) Greenman, C.; Stephens, P.; Smith, R.; Dalgliesh, G.; Hunter, C.; Bignell, G.; Davies, H.; Teague, J.; Butler, A.; Stevens, C.; Edkins, S.; O’Meara, S.; Vastrik, I.; Schmidt, E.; Avis, T.; Barthorpe, S.; Bhamra, G.; Buck, G.; Choudhury, B.; Clements, J.; Cole, J.; Dicks, E.; Forbes, S.; Gray, K.; Halliday, K.; Harrison, R.; Hills, K.; Hinton, J.; Jenkinson, A.; Jones, D.; Menzies, A.; Mironenko, T.; Perry, J.; Raine, K.; Richardson, D.; Shepherd, R.; Small, A.; Tofts, C.; Varian, J.; Webb, T.; West, S.; Widaa, S.; Yates, A.; Cahill, D.; Louis, D.; Goldstraw, P.; Nicholson, A.; Brasseur, F.; Looijenga, L.; Weber, B.; Chiew, Y.-E.; DeFazio, A.; Greaves, M.; Green, A.; Campbell, P.; Birney, E.; Easton, D.; Chenevix-Trench, G.; Tan, M.-H.; Khoo, S.; Teh, B.; Yuen, S.; Leung, S.; Wooster, R.; Futreal, P.; Stratton, M. Patterns of somatic mutation in human cancer genomes. Nature 2007, 446 (7132), 153− 158. (4) Karaman, M.; Herrgard, S.; Treiber, D.; Gallant, P.; Atteridge, C.; Campbell, B.; Chan, K.; Ciceri, P.; Davis, M.; Edeen, P.; Faraoni, R.; Floyd, M.; Hunt, J.; Lockhart, D.; Milanov, Z.; Morrison, M.; Pallares, G.; Patel, H.; Pritchard, S.; Wodicka, L.; Zarrinkar, P. A quantitative analysis of kinase inhibitor selectivity. Nat. Biotechnol. 2008, 26 (1), 127−132. (5) Huang, D.; Zhou, T.; Lafleur, K.; Nevado, C.; Caflisch, A. Kinase selectivity potential for inhibitors targeting the ATP binding site: a network analysis. Bioinformatics (Oxford, England) 2009, 26 (2), 198− 204. (6) Davis, M.; Hunt, J.; Herrgard, S.; Ciceri, P.; Wodicka, L.; Pallares, G.; Hocker, M.; Treiber, D.; Zarrinkar, P. Comprehensive analysis of kinase inhibitor selectivity. Nature Biotechnol. 2011, 29 (11), 1046− 1051. (7) Morphy, R. Selectively nonselective kinase inhibition: striking the right balance. J. Med. Chem. 2010, 53 (4), 1413−1437. (8) Ashburn, T.; Thor, K. Drug repositioning: identifying and developing new uses for existing drugs. Nature Rev. Drug Discovery 2004, 3 (8), 673−683. (9) Dakshanamurthy, S.; Issa, N.; Assefnia, S.; Seshasayee, A.; Peters, O.; Madhavan, S.; Uren, A.; Brown, M.; Byers, S. Predicting new indications for approved drugs using a proteochemometric method. J. Med. Chem. 2012, 55 (15), 6832−6848. (10) Lee, J.; Uhlik, M.; Moxham, C.; Tomandl, D.; Sall, D. Modern phenotypic drug discovery is a viable, neoclassic pharma strategy. J. Med. Chem. 2012, 55 (10), 4527−4538. (11) Zheng, W.; Thorne, N.; McKew, J. Phenotypic screens as a renewed approach for drug discovery. Drug Discovery Today 2013, 18 (21−22), 1067−73. 1584

DOI: 10.1021/pr5012608 J. Proteome Res. 2015, 14, 1574−1586

Article

Journal of Proteome Research

graphic analysis, and biological activities. J. Med. Chem. 2002, 45 (18), 3905−27. (45) Lu, H.; Chang, D. J.; Baratte, B.; Meijer, L.; Schulze-Gahmen, U. Crystal structure of a human cyclin-dependent kinase 6 complex with a flavonol inhibitor, fisetin. J. Med. Chem. 2005, 48 (3), 737−43. (46) Huber, K. V.; Salah, E.; Radic, B.; Gridling, M.; Elkins, J. M.; Stukalov, A.; Jemth, A. S.; Gokturk, C.; Sanjiv, K.; Stromberg, K.; Pham, T.; Berglund, U. W.; Colinge, J.; Bennett, K. L.; Loizou, J. I.; Helleday, T.; Knapp, S.; Superti-Furga, G. Stereospecific targeting of MTH1 by (S)-crizotinib as an anticancer strategy. Nature 2014, 508 (7495), 222−7. (47) Anastassiadis, T.; Deacon, S.; Devarajan, K.; Ma, H.; Peterson, J. Comprehensive assay of kinase catalytic activity reveals features of kinase inhibitor selectivity. Nature Biotechnol. 2011, 29 (11), 1039− 1045. (48) Kitagawa, D.; Yokota, K.; Gouda, M.; Narumi, Y.; Ohmoto, H.; Nishiwaki, E.; Akita, K.; Kirii, Y. Activity-based kinase profiling of approved tyrosine kinase inhibitors. Genes Cells 2013, 18 (2), 110−22. (49) Jester, B.; Gaj, A.; Shomin, C.; Cox, K.; Ghosh, I. Testing the promiscuity of commercial kinase inhibitors against the AGC kinase group using a split-luciferase screen. J. Med. Chem. 2012, 55 (4), 1526−1537. (50) Morwick, T.; Buttner, F. H.; Cywin, C. L.; Dahmann, G.; Hickey, E.; Jakes, S.; Kaplita, P.; Kashem, M. A.; Kerr, S.; Kugler, S.; Mao, W.; Marshall, D.; Paw, Z.; Shih, C. K.; Wu, F.; Young, E. Hit to lead account of the discovery of bisbenzamide and related ureidobenzamide inhibitors of Rho kinase. J. Med. Chem. 2010, 53 (2), 759−77. (51) Wu, F.; Buttner, F. H.; Chen, R.; Hickey, E.; Jakes, S.; Kaplita, P.; Kashem, M. A.; Kerr, S.; Kugler, S.; Paw, Z.; Prokopowicz, A.; Shih, C. K.; Snow, R.; Young, E.; Cywin, C. L. Substituted 2H-isoquinolin-1one as potent Rho-Kinase inhibitors. Part 1: Hit-to-lead account. Bioorg. Med. Chem. Lett. 2010, 20 (11), 3235−9. (52) Uehata, M.; Ishizaki, T.; Satoh, H.; Ono, T.; Kawahara, T.; Morishita, T.; Tamakawa, H.; Yamagami, K.; Inui, J.; Maekawa, M.; Narumiya, S. Calcium sensitization of smooth muscle mediated by a Rho-associated protein kinase in hypertension. Nature 1997, 389 (6654), 990−4. (53) Rikitake, Y.; Kim, H. H.; Huang, Z.; Seto, M.; Yano, K.; Asano, T.; Moskowitz, M. A.; Liao, J. K. Inhibition of Rho kinase (ROCK) leads to increased cerebral blood flow and stroke protection. Stroke 2005, 36 (10), 2251−7. (54) Jacobs, M.; Hayakawa, K.; Swenson, L.; Bellon, S.; Fleming, M.; Taslimi, P.; Doran, J. The structure of dimeric ROCK I reveals the mechanism for ligand selectivity. J. Biol. Chem. 2006, 281 (1), 260−8. (55) Honigberg, L. A.; Smith, A. M.; Sirisawad, M.; Verner, E.; Loury, D.; Chang, B.; Li, S.; Pan, Z.; Thamm, D. H.; Miller, R. A.; Buggy, J. J. The Bruton tyrosine kinase inhibitor PCI-32765 blocks B-cell activation and is efficacious in models of autoimmune disease and B-cell malignancy. Proc. Natl. Acad. Sci. U. S. A. 2010, 107 (29), 13075−80. (56) Mendel, D. B.; Laird, A. D.; Xin, X.; Louie, S. G.; Christensen, J. G.; Li, G.; Schreck, R. E.; Abrams, T. J.; Ngai, T. J.; Lee, L. B.; Murray, L. J.; Carver, J.; Chan, E.; Moss, K. G.; Haznedar, J. O.; Sukbuntherng, J.; Blake, R. A.; Sun, L.; Tang, C.; Miller, T.; Shirazian, S.; McMahon, G.; Cherrington, J. M. In vivo antitumor activity of SU11248, a novel tyrosine kinase inhibitor targeting vascular endothelial growth factor and platelet-derived growth factor receptors: determination of a pharmacokinetic/pharmacodynamic relationship. Clin. Cancer Res. 2003, 9 (1), 327−37. (57) Cui, J. J.; Tran-Dubé, M.; Shen, H.; Nambu, M.; Kung, P. P.; Pairish, M.; Jia, L.; Meng, J.; Funk, L.; Botrous, I.; McTigue, M.; Grodsky, N.; Ryan, K.; Padrique, E.; Alton, G.; Timofeevski, S.; Yamazaki, S.; Li, Q.; Zou, H.; Christensen, J.; Mroczkowski, B.; Bender, S.; Kania, R. S.; Edwards, M. P. Structure based drug design of crizotinib (PF-02341066), a potent and selective dual inhibitor of mesenchymal-epithelial transition factor (c-MET) kinase and anaplastic lymphoma kinase (ALK). J. Med. Chem. 2011, 54 (18), 6342− 63.

(28) Zhang, L.; Holmes, I.; Hochgräfe, F.; Walker, S.; Ali, N.; Humphrey, E.; Wu, J.; de Silva, M.; Kersten, W.; Connor, T.; Falk, H.; Allan, L.; Street, I.; Bentley, J.; Pilling, P.; Monahan, B.; Peat, T.; Daly, R. Characterization of the novel broad-spectrum kinase inhibitor CTx0294885 as an affinity reagent for mass spectrometry-based kinome profiling. J. Proteome Res. 2013, 12 (7), 3104−3116. (29) Bantscheff, M.; Lemeer, S.; Savitski, M.; Kuster, B. Quantitative mass spectrometry in proteomics: critical review update from 2007 to the present. Anal. Bioanal. Chem. 2012, 404 (4), 939−965. (30) Knapp, S.; Arruda, P.; Blagg, J.; Burley, S.; Drewry, D.; Edwards, A.; Fabbro, D.; Gillespie, P.; Gray, N.; Kuster, B.; Lackey, K.; Mazzafera, P.; Tomkinson, N.; Willson, T.; Workman, P.; Zuercher, W. A public-private partnership to unlock the untargeted kinome. Nat. Chem. Biol. 2013, 9 (1), 3−6. (31) Oppermann, F. S.; Gnad, F.; Olsen, J. V.; Hornberger, R.; Greff, Z.; Keri, G.; Mann, M.; Daub, H. Large-scale proteomics analysis of the human kinome. Mol. Cell Proteomics 2009, 8 (7), 1751−64. (32) Schirle, M.; Petrella, E.; Brittain, S.; Schwalb, D.; Harrington, E.; Cornella-Taracido, I.; Tallarico, J. Kinase inhibitor profiling using chemoproteomics. Methods Mol. Biol. (Clifton, N.J.) 2012, 795, 161− 177. (33) Pachl, F.; Plattner, P.; Ruprecht, B.; Médard, G.; Sewald, N.; Kuster, B. Characterization of a Chemical Affinity Probe Targeting Akt Kinases. J. Proteome Res. 2013, 12, 3792−3800. (34) Ku, X.; Heinzlmeir, S.; Helm, D.; Médard, G.; Kuster, B. New Affinity Probe Targeting VEGF Receptors for Kinase Inhibitor Selectivity Profiling by Chemical Proteomics. J. Proteome Res. 2014, 13 (5), 2445−2452. (35) Gholami, A.; Hahne, H.; Wu, Z.; Auer, F.; Meng, C.; Wilhelm, M.; Kuster, B. Global Proteome Analysis of the NCI-60 Cell Line Panel. Cell Rep. 2013, 4 (3), 609−620. (36) https://github.com/ThomasKuehne/kinaseblender. (37) Ishii, N.; Maier, D.; Merlo, A.; Tada, M.; Sawamura, Y.; Diserens, A. C.; Van Meir, E. G. Frequent co-alterations of TP53, p16/ CDKN2A, p14ARF, PTEN tumor suppressor genes in human glioma cell lines. Brain Pathol. 1999, 9 (3), 469−79. (38) Ku, X.; Heinzlmeir, S.; Liu, X.; Médard, G.; Kuster, B. A new chemical probe for quantitative proteomic profiling of fibroblast growth factor receptor and its inhibitors. J. Proteomics 2014, 96 (16), 44−55. (39) Hahne, H.; Pachl, F.; Ruprecht, B.; Maier, S.; Klaeger, S.; Helm, D.; Médard, G.; Wilm, M.; Lemeer, S.; Kuster, B. DMSO enhances electrospray response, boosting sensitivity of proteomic experiments. Nat. Methods 2013, 10, 989−991. (40) Cox, J.; Mann, M. MaxQuant enables high peptide identification rates, individualized p.p.b.-range mass accuracies and proteome-wide protein quantification. Nat. Biotechnol. 2008, 26 (12), 1367−72. (41) Cox, J.; Hein, M. Y.; Luber, C. A.; Paron, I.; Nagaraj, N.; Mann, M. Accurate proteome-wide label-free quantification by delayed normalization and maximal peptide ratio extraction, termed MaxLFQ. Mol. Cell Proteomics 2014, 13 (9), 2513−26. (42) Sharma, K.; Weber, C.; Bairlein, M.; Greff, Z.; Kéri, G.; Cox, J.; Olsen, J.; Daub, H. Proteomics strategy for quantitative protein interaction profiling in cell extracts. Nat. Methods 2009, 6 (10), 741− 744. (43) Fabian, M.; Biggs, W.; Treiber, D.; Atteridge, C.; Azimioara, M.; Benedetti, M.; Carter, T.; Ciceri, P.; Edeen, P.; Floyd, M.; Ford, J.; Galvin, M.; Gerlach, J.; Grotzfeld, R.; Herrgard, S.; Insko, D.; Insko, M.; Lai, A.; Lélias, J.-M.; Mehta, S.; Milanov, Z.; Velasco, A.; Wodicka, L.; Patel, H.; Zarrinkar, P.; Lockhart, D. A small molecule-kinase interaction map for clinical kinase inhibitors. Nature Biotechnol. 2005, 23 (3), 329−336. (44) Kim, K. S.; Kimball, S. D.; Misra, R. N.; Rawlins, D. B.; Hunt, J. T.; Xiao, H. Y.; Lu, S.; Qian, L.; Han, W. C.; Shan, W.; Mitt, T.; Cai, Z. W.; Poss, M. A.; Zhu, H.; Sack, J. S.; Tokarski, J. S.; Chang, C. Y.; Pavletich, N.; Kamath, A.; Humphreys, W. G.; Marathe, P.; Bursuker, I.; Kellar, K. A.; Roongta, U.; Batorsky, R.; Mulheron, J. G.; Bol, D.; Fairchild, C. R.; Lee, F. Y.; Webster, K. R. Discovery of aminothiazole inhibitors of cyclin-dependent kinase 2: synthesis, X-ray crystallo1585

DOI: 10.1021/pr5012608 J. Proteome Res. 2015, 14, 1574−1586

Article

Journal of Proteome Research (58) Wen, Q.; Goldenson, B.; Silver, S. J.; Schenone, M.; Dancik, V.; Huang, Z.; Wang, L. Z.; Lewis, T. A.; An, W. F.; Li, X.; Bray, M. A.; Thiollier, C.; Diebold, L.; Gilles, L.; Vokes, M. S.; Moore, C. B.; BlissMoreau, M.; Verplank, L.; Tolliday, N. J.; Mishra, R.; Vemula, S.; Shi, J.; Wei, L.; Kapur, R.; Lopez, C. K.; Gerby, B.; Ballerini, P.; Pflumio, F.; Gilliland, D. G.; Goldberg, L.; Birger, Y.; Izraeli, S.; Gamis, A. S.; Smith, F. O.; Woods, W. G.; Taub, J.; Scherer, C. A.; Bradner, J. E.; Goh, B. C.; Mercher, T.; Carpenter, A. E.; Gould, R. J.; Clemons, P. A.; Carr, S. A.; Root, D. E.; Schreiber, S. L.; Stern, A. M.; Crispino, J. D. Identification of regulators of polyploidization presents therapeutic targets for treatment of AMKL. Cell 2012, 150 (3), 575−89. (59) Satoh, S.; Utsunomiya, T.; Tsurui, K.; Kobayashi, T.; Ikegaki, I.; Sasaki, Y.; Asano, T. Pharmacological profile of hydroxy fasudil as a selective rho kinase inhibitor on ischemic brain damage. Life Sci. 2001, 69 (12), 1441−53. (60) Shibuya, M.; Suzuki, Y.; Sugita, K.; Saito, I.; Sasaki, T.; Takakura, K.; Nagata, I.; Kikuchi, H.; Takemae, T.; Hidaka, H.; et al. Effect of AT877 on cerebral vasospasm after aneurysmal subarachnoid hemorrhage. Results of a prospective placebo-controlled double-blind trial. J. Neurosurg. 1992, 76 (4), 571−7. (61) Bender, C.; Ullrich, A. PRKX, TTBK2 and RSK4 expression causes Sunitinib resistance in kidney carcinoma- and melanoma-cell lines. Int. J. Cancer 2012, 131 (2), E45−E55. (62) Li, X.; Iomini, C.; Hyink, D.; Wilson, P. D. PRKX critically regulates endothelial cell proliferation, migration, and vascular-like structure formation. Dev. Biol. 2011, 356 (2), 475−85. (63) Ying, H.; Biroc, S. L.; Li, W. W.; Alicke, B.; Xuan, J. A.; Pagila, R.; Ohashi, Y.; Okada, T.; Kamata, Y.; Dinter, H. The Rho kinase inhibitor fasudil inhibits tumor progression in human and rat tumor models. Mol. Cancer Ther 2006, 5 (9), 2158−64. (64) Deng, L.; Li, G.; Li, R.; Liu, Q.; He, Q.; Zhang, J. Rho-kinase inhibitor, fasudil, suppresses glioblastoma cell line progression in vitro and in vivo. Cancer Biol. Ther 2010, 9 (11), 875−84. (65) Brosch, M.; Yu, L.; Hubbard, T.; Choudhary, J. Accurate and sensitive peptide identification with Mascot Percolator. J. Proteome Res. 2009, 8 (6), 3176−81. (66) Ritz, C. Toward a unified approach to dose-response modeling in ecotoxicology. Environ. Toxicol. Chem. 2010, 29 (1), 220−9. (67) Wilhelm, M.; Schlegl, J.; Hahne, H.; Moghaddas Gholami, A.; Lieberenz, M.; Savitski, M. M.; Ziegler, E.; Butzmann, L.; Gessulat, S.; Marx, H.; Mathieson, T.; Lemeer, S.; Schnatbaum, K.; Reimer, U.; Wenschuh, H.; Mollenhauer, M.; Slotta-Huspenina, J.; Boese, J. H.; Bantscheff, M.; Gerstmair, A.; Faerber, F.; Kuster, B. Massspectrometry-based draft of the human proteome. Nature 2014, 509 (7502), 582−7.

1586

DOI: 10.1021/pr5012608 J. Proteome Res. 2015, 14, 1574−1586