Searching for Biomarkers of Aurora-A Kinase Activity - ACS Publications

Luisa Rusconi, and Barbara Valsasina. Nerviano Medical Sciences, Dept of Biology, 20014 Nerviano (MI), Italy. Received January 31, 2005. Aurora-A, -B,...
0 downloads 0 Views 375KB Size
Download PDF
Searching for Biomarkers of Aurora-A Kinase Activity: Identification of in Vitro Substrates through a Modified KESTREL Approach Sonia Troiani,*,† Mauro Uggeri,† Ju1 rgen Moll, Antonella Isacchi, Henryk M. Kalisz, Luisa Rusconi, and Barbara Valsasina Nerviano Medical Sciences, Dept of Biology, 20014 Nerviano (MI), Italy Received January 31, 2005

Aurora-A, -B, and -C are members of a small family of mitotic serine/threonine kinases that regulate centrosome maturation, chromosome segregation, and cytokinesis. They are often overexpressed in different human tumor types and have been identified as attractive targets for anticancer drug development. As specific inhibitors of the Aurora kinases are entering phase I clinical trials, there is a high need for appropriate Aurora-A biomarkers to follow mechanism of action or response. To identify novel Aurora-A substrates potentially useful as specific biomarkers we applied several modifications to the original KESTREL (Kinase Substrate Tracking and Elucidation) method in conjunction with gel electrophoresis and MALDI-MS and LC-MS/MS. The major modifications to the method included the introduction of a heating step to inactivate endogenous kinases after cell lysis and the execution of the in vitro kinase reaction in the presence of 5 mM Mg2+ and at high (1 mM) ATP concentration. Total and fractionated extracts from nocodazole-treated HeLa cells were used as a source of Aurora-A substrates. Using this approach, we were able to detect a number of Aurora-A specific phospholabeled signals and to identify vimentin as a putative Aurora-A substrate. Vimentin was then confirmed as an in vitro substrate of Aurora-A by the phosphorylation of the recombinant protein followed by MS and antibody detection. Keywords: aurora • substrate • phosphorylation • KESTREL • vimentin

Introduction Aurora proteins form a family of mitotic serine/threonine kinases comprising three members in mammals, namely Aurora-A, -B, and -C. Their activity is cell-cycle regulated, with highest levels of expression and activity being detected during the G2/M phase. These proteins play important roles in the orchestration of the mitotic process, operating in the regulation of centrosome function, bipolar spindle assembly and chromosome segregation.1-4 Moreover, Aurora kinases are found to be overexpressed in diverse human cancers and in some instances their elevated expression is correlated with chromosomal instability and is a prognostic factor.5 Of the three kinases, only Aurora-A has been shown to contribute to oncogenic transformation, probably due to its role in regulating the centrosome cycle and its essential functions in the mitotic spindle.6,7 These findings suggest that Aurora kinases represent attractive targets for anticancer drug development and indeed small molecule inhibitors of Aurora kinases are currently entering phase I clinical trials.8 The emergence of novel targeted therapies for cancer treatment creates a need for specific biomarkers that not only contribute to the exhaustive understanding of the mechanism * To whom correspondence should be addressed. Sonia Troiani, Nerviano Medical Sciences, Viale Pasteur 10, 20014 Nerviano (MI), Italy. Tel: +390331-581292. Fax: +39-0331-581267. E-mail: [email protected]. † These two authors contributed equally to the work.

1296

Journal of Proteome Research 2005, 4, 1296-1303

Published on Web 05/27/2005

of action of the compound of interest, but can be used also to evaluate the clinical response after drug treatment. A number of substrates and interacting proteins were identified for Aurora-A and -B, including a kinesin-like motor protein,9 histone H310 and kinetochore proteins.11 Phosphorylation of histone H3 at Ser 10 is at present the best characterized candidate biomarker for proof of MOA of Aurora-B inhibition. However, histone H3 Ser 10 is a direct substrate for Aurora-B, while it represents a rather indirect mitotic marker to follow the activity of Aurora-A inhibitors. Hence, there is a need to discover novel MOA-related biomarkers that could represent a clear readout of the activity of Aurora-A following drug treatment.12 Various methodologies have been developed to date with the aim of identifying kinase substrates. Some examples include the screen of peptide libraries for phosphorylation consensus identification,13,14 the use of antibodies reactive against phosphorylated motifs to identify the phosphorylated substrates of the kinase of interest15 or the modification of the ATP pocket of kinases in order to use ATP analogues that can only be exploited by the mutated kinase of interest, allowing the identification of specific substrates.16,17 Another systematic approach that can predict or prioritize potential targets in silico was developed by Tien et al. and was leading to the identification and validation of TACC3, survivin, Hec1 and hsNuf2 as potential substrates of Aurora kinases.18 However, these meth10.1021/pr050018e CCC: $30.25

 2005 American Chemical Society

Biomarkers of Aurora-A Kinase Activity

ods require a detailed characterization of the enzyme of interest, possibly including 3D structure, consensus sequence for phosphorylation or precise localization and cell-cycle stage of activation, which is not always available. A simpler method for the identification of kinase substrates in crude cell extracts was developed by Knebel et al.19 The method, termed KESTREL (Kinase Substrate Tracking and Elucidation), involves the incubation of a total cell extract with an active recombinant kinase in the presence of [γ-32P]ATP, usually in the nanomolar concentration range, and using Mn2+, that is typically a cofactor of just a small number of protein kinases, primarily tyrosine kinases, to reduce the background of phosphorylation due to endogenous kinase activity. This method, however, can only be applied to kinases endowed with a high specific activity, thus yielding a signal well above the background, and to kinases capable of functioning in the presence of Mn2+ as cofactor. An alternative way to reduce the high background phosphorylation involves sample heat treatment prior to the kinase reaction. This approach was used to identify DNA Topoisomerase II as a substrate of Aurora-B by using a heat treated chromosomal scaffold preparation.20 In the present study, we have developed a modified KESTREL approach in which we introduced a heating step in order to inactivate endogenous protein kinases present in cell extract, thus reducing the assay background without the need to change the cofactor. This modified KESTREL method was applied to the recombinant Aurora-A kinase, which is characterized by a relatively low specific activity and a high sensitivity to a Mg/ Mn shift, both features being hardly compatible with the original KESTREL conditions. Here, we report our tailored KESTREL approach and the results obtained using both total and fractionated cell extracts as substrate source. In these experiments, a number of putative Aurora-A substrates were detected and among them Bip protein and vimentin were identified.

Experimental Section Cell Culture. Human HeLa cells were cultured in Dulbecco’s modified Eagle’s medium (Gibco Invitrogen Corporation, Carlsbad, CA) supplemented with 10% FBS (Gibco Invitrogen Corporation). For G2/M synchronization, cells were treated with 400 ng/mL nocodazole for 16-18 h. Protein Extraction. For total protein extract preparation 4 × 107 nocodazole-treated cells were recovered by flask shaking. Proteins were extracted with RIPA buffer (20 mM Tris-HCl pH 7.4, 150 mM NaCl, 1 mM EDTA, 1% NP40, 0.1% SDS, 0.25% Na deoxycholate, protease inhibitors). On the other hand, the fractionated extract was prepared starting from 8 × 107 nocodazole-treated cells that were collected by shaking. The cell pellet was washed once with PBS then protein extract was prepared according to the hexylene glycol-based procedure previously described21 followed by sucrose gradient fractionation. SDS-PAGE and Western Blotting. Proteins were separated by SDS-PAGE following standard protocol and detected with Colloidal Coomassie Blue (Pierce, Rockford IL). Blotting onto nitrocellulose was performed in Tris-Gly buffer containing 10% Me-OH at 90 V, 300 mA for 90 min. Antibodies used: anti-PSer and anti-PThr (Cell Signaling, Beverly, MA), anti-IAK1, anti-MCM2, and anti-vimentin (BD Biosciences, San Jose, CA), anti-PSer 72 vimentin (Santa Cruz Biotechnology, Santa Cruz, CA). Working solution dilution: Anti-PSer/PThr 1:1000 in 5% BSA, TTBS (0.1% Tween 20 in

research articles TBS), anti-IAK1, anti-MCM2, anti-vimentin and anti-PSer 72 vimentin 1:1000 in TTBS containing 5% milk. Antibody incubation was performed overnight at 4 °C (antiPSer/PThr) or 2 h at room temperature (anti-IAK1, anti-MCM2, anti-vimentin and anti-PSer 72 vimentin). Secondary antimouse and anti-rabbit (Bio-Rad Laboratories, Hercules, CA) antibodies were used at 1:4000 in TTBS with 5% BSA (or milk) and for 1 h at room temperature. The Amersham ECL system was used for chemiluminescence signal detection. Kinase Assay. Endogenous phosphatase activation was carried out by incubation at 30 °C for 20 min. Heat treatment consisted of 10 min incubation at 65 °C followed by at least 5 min incubation on ice. After heat treatment a 1:100 dilution of phosphatase inhibitor cocktail (Sigma-Aldrich, St. Louis, MO) was added. Recombinant Aurora-A was expressed in insect cells and purified to >80% homogeneity as judged by SDS-PAGE analysis and Coomassie staining. Kinase assay reaction was performed in Mg2+ buffer (50 mM HEPES pH 7.9, 5 mM MgCl2, 10 mM DTT, 3 µM NaVO3) or Mn2+ buffer (50 mM HEPES pH 7.9, 5 mM MnCl2, 10 mM DTT, 3 µM NaVO3) at 37 °C for 30 min using 0.05 µM [γ-32P]ATP (1 µCi) or a mixture of [γ-32P]ATP and 1 mM ATP (1 µCi total activity) and 200 ng of recombinant Aurora-A. The reaction was stopped by addition of Sample Buffer (25 mM Tris-HCl pH 6.8, 4% SDS, 20% glycerol, 10% β-mercaptoethanol, 0.002% bromophenol blue) and incubation at 100 °C for 5 min. Two-Dimensional Gel Electrophoresis. Proteins were precipitated using 2D Clean Up Kit (Amersham Biosciences Corp., San Francisco, CA). The protein pellet was resuspended in Rehydration Buffer (9M Urea, 3% Chaps, 40 mM DTT, 1% IPG buffer). Proteins were separated by isoelectric focusing using the IPGphor apparatus (Amersham Biosciences Corp.). In-gel rehydration was done on 7 cm IPG strips pH 4-7 or IPG strips pH 3-10 NL (Amersham Biosciences Corp.). Isoelectric focusing was performed for a total of 23 000 V/hours. For the second dimension the IPG strips were equilibrated for 10 min in 6 M urea, 30% glycerol, 1% SDS, 2% DTT and 10 min in 6 M urea, 30% glycerol, 1% SDS, 3% iodoacetamide. SDS-PAGE was performed following standard protocols. Radioactive signals were detected by Storm 860 PhosphoImager (Amersham Biosciences Corp.). Gel images were analyzed using the Z3 software (Compugen, San Jose, CA). In-Gel Tryptic Digestion. Protein digestion was performed with trypsin by using the Digest Pro system (Intavis, Koeln, Germany) following the standard protocol. The elution mixture was then dried down in a speed vacuum and redissolved in 50% acetonitrile/0.1% TFA for MALDI-MS analysis or in deionized water for LC-MS/MS analysis. In Vitro Vimentin Phosphorylation. To identify vimentin phosphorylated residues by mass spectrometry, 10 µg of recombinant vimentin head domain (GST-1-84, Santa Cruz Biotechnology) were phosphorylated by 1.5 µg of recombinant Aurora-A. The kinase reaction was performed in Mg2+ buffer for 2 h at 37 °C. The protein mixture was loaded on a 10% polyacrylamide gel and Colloidal Coomassie Blue staining was used for protein visualization. The band at 35 kDa corresponding to vimentin head domain was excised from gel and trypsin digested by using the Digest Pro system. For autoradiography, 1 µg of vimentin head domain was phosphorylated in the presence of 200 ng of Aurora A recombinant kinase and 0.05 µM of [γ-32P]ATP. The sample was then run on SDS-PAGE and radioactive bands were visualized with Journal of Proteome Research • Vol. 4, No. 4, 2005 1297

research articles Storm 860 PhosphoImager. The same protocol, with the exception that cold instead of [γ-32P]ATP was used, was employed for Western blot with anti-PSer 72 vimentin antibody. MALDI-MS. Samples for matrix-assisted laser desorption ionization-time-of-flight (MALDI-TOF) analysis were prepared by spotting 0.5 µL of peptide mixture with 0.5 µL of R-cyano4-hydroxycinnamic acid (10 mg/mL in 50:50 acetonitrile:water containing 0.1% TFA) and analyzed on a Voyager DE-PRO (Applied Biosystems, Warrington, UK). All spectra were collected in reflector mode using four peptides of known mass as external calibration standards. Protein identification was done with Mascot (Matrix Science Ltd, London, UK) software. The following parameters were used for MASCOT searching: NCBInr database for Homo sapiens sequences, up to two tryptic missed cleavages allowed, fixed cysteine carbamidomethyl modification, methionine oxidation as variable modification, 100 ppm tolerance for monoisotopic peptide species. Probability-based MASCOT scores were used to evaluate protein identifications. HPLC-ESI-MS/MS. Tandem mass spectrometry was performed on a hybrid quadrupole-time-of-flight instrument (QToF2, Micromass, Manchester, UK) equipped with a Z-spray source and calibrated by injecting a solution of Glu-Fibrinopeptide (Sigma-Aldrich) (0.5 pmol/µL) at 0.5 µL/min, applying 3.2 kV to the spraying capillary, and a collision energy of 29 V. For liquid chromatography coupled to electrospray tandem mass spectrometry, low flowrate reverse phase HPLC was used on-line to the Q-ToF2. The mobile phase gradient was mixed and delivered by the Ultimate HPLC pump (LC Packings, Amsterdam, NL) equipped with a nano-flow splitter that was delivering in column a flow rate of about 0.2 µL/min from a primary flow rate of 200 µL/min. Mobile phase A composition was 95:5 water:acetonitrile (v/v), 0.05% formic acid and mobile phase B composition was 5:95 water:acetonitrile (v/v), 0.06% formic acid. 5 µL of sample were injected onto a C18 precolumn (PepMap 0.3 × 5 mm, LC Packings) and was preconcentrated and desalted with mobile phase A at a flow rate of 20 µL/min for 5 min. The separation of the peptides was performed on a 75 µm I. D. × 150 mm C18 PepMap column (LC Packings) with a gradient from 5 to 60% mobile phase B in 58 min followed by a washing step at 60% B in 15 min at a flow rate of 0.2 µL/ min. The UV detection was registered at λ1 ) 214 nm and λ2 ) 280 and the MS acquisition was performed in survey scan mode. The HPLC outlet was coupled to the MS nanospray capillary at which 3.2 kV were applied. The MS acquisition was performed in survey scan mode where the Q-Tof selected doubly, triply, quadruply charged ions above an intensity threshold of 10 counts/second (CPS) for MS/MS analysis from each survey scan (auto function switching). The instrument performed the MS/MS analysis under user-defined parameters of collision energy. For the subsequent protein identification, the raw data files were converted into .pkl files that were used in Mascot searches of the NCBInr database for mammalian sequences allowing the recognition of doubly and triply charged precursor ions with up to two tryptic missed cleavages, fixed carbamidomethyl modification, variable Ser/Thr phosphorylation, ( 0.5 Da peptide tolerance and ( 0.3 Da MS/MS tolerance for monoisotopic species. Protein identifications were confirmed by manual MS/MS data analysis for at least one peptide per protein.

Results and Discussion Aurora-B kinase activity can be clearly followed in vitro and in vivo by monitoring the phosphorylation status of histone 1298

Journal of Proteome Research • Vol. 4, No. 4, 2005

Troiani et al.

H3 on Ser 10. A decrease in the phosphorylation of this marker correlates with inhibition of the kinase activity in cell lines and animal models. It can be easily detected by immunofluorescence and immunohistochemical studies, showing good premises as a possible tool in clinical trials on Aurora kinase inhibitors8 (Moll et al., unpublished results). Conversely, a marker with similar favorable properties and degree of validation is not presently available for Aurora-A. In fact, while Aurora-A is able to phosphorylate histone H3 Ser 10 as efficiently as Aurora-B in vitro, no modulation of this phosphorylation has been observed when Aurora-A was depleted by RNAi.22 In addition, immunofluorescence experiments at different cell cycle stages clearly indicate that the two proteins do not co-localize, thereby making their direct interaction improbable. A number of Aurora-A substrates have been described so far. They include activators such as Tpx2, which was shown to positively regulate Aurora-A activity by protecting it from dephosphorylation by phosphatase PP1 at the critical Thr 288 residue.23,24 Additional known substrates include phosphatases, such as PP1 and Cdc25B,25 the kinesin-like motor protein Eg59 and Rho-associated protein kinase ROCK whose suppression was recently shown to bypass the cell cycle arrest induced by Aurora-A depletion,26 the mammalian homologue of Drosophila transforming acidic coiled-coil (D-TACC) protein TACC3,18 and other cancer related proteins, such as BRCA127 and p53.28 However, application of these substrates as biomarkers for Aurora-A kinase activity in clinical settings is hampered by the fact that the majority of these proteins are present at very low concentrations and/or are cell-cycle regulated. In addition, more promising candidates, such as the Rho-associated protein kinase, have not been unambiguously validated as tools for the detection of Aurora-A pharmacological inhibition. Therefore, we decided to identify conditions suitable for the identification of new Aurora-A substrates by a modified Kestrel approach. Cell Extract Preparation. In principle, as Aurora-A associates with centrosomes and the microtubule region proximal to them, a centrosome extract may be considered the most suitable material in the search for physiological Aurora-A substrates. However, low levels of Aurora-A have also been associated with the spindle midzone and midbody late in mitosis.1 Moreover, in resting cells, Aurora-A is located in centrosomes and it continuously exchanges with a cytoplasmic pool. By the way, other proteins usually located in different sub-cellular compartments, such as HSP90, have also been demonstrated to shuttle between the cytoplasm and the centrosome,29 so that localization may be considered a dynamic process rather than a static definition. Furthermore, centrosomes are very difficult organelles to isolate as demonstrated by Andersen et al.,30 who had to apply a sophisticated mass spectrometry approach termed Protein Correlation Profiling in order to discriminate true centrosomal components from contaminants in classical “centrosome preparations”. Therefore, restricting our investigation to a centrosome preparation, rather than being an advantage, could have instead impoverished the analysis. Consequently, we concentrated our efforts on fractionated and total extracts from nocodazole-blocked Hela cells in parallel. As Aurora-A expression and activity peaks during the G2/M phase of the cell cycle,1 cells were blocked in pro-metaphase by treatment with nocodazole, to provide a starting material enriched in active Aurora-A and thus presumably also in its endogenous substrates. In fact, Aurora-A protein levels were

Biomarkers of Aurora-A Kinase Activity

research articles

Figure 1. (A) Aurora-A enrichment in nocodazole-treated cells. 30 µg of total cell extract from asynchronous or nocodazole-treated cells were analyzed by SDS-PAGE and Western blotting against Aurora-A kinase, MCM2 protein western blot was used as loading control. (B) 2D gel electrophoresis analysis of fractionated extract (left) and total HeLa extract (right), 200 µg of total proteins were loaded and detected by Colloidal coomassie staining. Circles are used to highlight in total extract several cytosolic proteins, which appear to be highly depleted in the fractionated one, and conversely in the fractionated extract a highly enriched nuclear protein. (C) Pie charts indicated distribution of most abundant proteins identified by MALDI-MS in the two different preparations. 18 proteins were identified in the scaffold preparation and 27 proteins in the total extract. Protein classification was based on subcellular location reported in the Swiss-Prot database annotations and other public domains.

dramatically increased in HeLa cells after nocodazole block with respect to asynchronous cells, as clearly demonstrated by immunoblotting of the cell lysate with an Aurora-A specific antibody (Figure 1A). Nocodazole-treated HeLa cells were lysed in hexylene glycol buffer and the pellet, containing chromosome-bound proteins together with mitochondrial proteins and cytoskeleton components, was fractionated on a sucrose gradient31 resulting in the elimination of membrane debris and most cytosolic proteins. The fractionated and total extracts were analyzed by 2D electrophoresis using colloidal Coomassie staining (Figure 1B), and the most abundant protein spots were excised from both gels. Tryptic digestion and analysis by mass spectrometry of the excised protein spots demonstrated a significant reduction in the content of cytosolic proteins, such as enolase, aldolase and glutathione S-transferase, and a notable enrichment of nuclear proteins, including nucleophosmin, in the fractionated extracts (Figure 1B). Moreover, comparative categorization of the proteins identified by mass spectrometry in the two different extracts according to their typical sub-cellular localization allowed us to verify a visible change in the distribution of the most abundant proteins, with a major depletion of cytoplasmic proteins accompanied by a significant increase in the fraction of nuclear components in the fractionated extract (Figure 1C). KESTREL Approach. For the in vitro detection of putative substrates of Aurora-A in extracts from nocodazole-blocked HeLa cells, we applied the KESTREL approach.19 The KESTREL method was originally developed with highly active recombinant protein kinases, using very short incubation times for the kinase reaction and very low concentrations of [γ-32P]ATP in order to select substrates with a high affinity for the kinase of interest. A further distinguishing feature was the replacement of Mg2+ with Mn2+ as cofactor, to reduce the high phosphorylation background due to endogenous kinase activity in the

extract. Additional interesting features, such as depletion of endogenous ATP and activation of endogenous phosphatases, were also applied to enhance the availability of potential substrates to phosphorylation by the recombinant kinase. Although this method was successfully applied for the identification of MKK, SAPK, and PKB family substrates,19,32,33 the assay conditions are limiting for kinases having medium-tolow specific activity, such as Aurora-A. Consequently, the KESTREL method had to be partially revised in order to make the procedure suitable for this kinase. Since the fractionation procedure is quite labor intensive, we decided to use total cell extract to optimize the conditions for in vitro kinase assay. As a first step, in accordance with the typical KESTREL protocol, the nocodazole-blocked HeLa total cell extracts were desalted by Sephadex G25 chromatography to deplete endogenous ATP and other low molecular weight cofactors. The relevance of endogenous phosphatase activation by mild heating was then evaluated. The cell extract was incubated at 30 °C for 20 min, and the overall phosphorylation level was assessed by immunoblotting with an anti-PSer/PThr antibody. As demonstrated in Figure 2, no significant reduction in the overall phosphorylation level was observed in the treated extract in comparison to the untreated sample. One possible explanation is that the protein extraction was performed in the absence of phosphatase inhibitors, which may have enabled the endogenous phosphatases to work during the extraction process, making the subsequent thermal activation step irrelevant. Furthermore, we evaluated the background phosphorylation in the presence of low (0.05 µM) vs high (1 mM) ATP concentrations, using only radioactive ATP in the first case while combining hot and cold ATP for the latter. Generally speaking, ATP concentration in the reaction environment is highly critical for any kinase, so that using a concentration of ATP well below its Km will drastically reduce the velocity of the Journal of Proteome Research • Vol. 4, No. 4, 2005 1299

research articles

Figure 2. SDS-PAGE and western blotting using PSer and PThr antibodies. 30 µg of total cell extract from nocodazole-treated HeLa cells were analyzed before and after incubation at 30 °C for 20 min to activate endogenous phosphatases. The treatment does not change the level of protein phosphorylation. Panel A western blotting anti-PSer/PThr, panel B anti MCM2 (loading control).

Figure 3. Effects of different reaction conditions on background levels. Autoradiography of 30 µg of total HeLa cell extracts after incubation with [γ-32P]ATP. Different concentrations of ATP and Mg2+ or Mn2+ cofactors were evaluated. Low ATP: 0.05 µM [γ-32P]ATP, high ATP: 0.05 µM [γ-32P]ATP + 1 mM ATP; Mn2+ reduces background levels as reported by Knebel et al. while ATP concentration does not impact on background.

substrate phosphorylation reaction. Although this is not a key issue when dealing with highly active kinases, it becomes crucial when working with kinases of low or even medium activity, to the point that kinase activity may be impaired. The Km of Aurora-A for ATP is reported to be in the range of 30-50 µM34 meaning that the classical KESTREL conditions are performed well below the optimal conditions for Aurora-A enzyme activity. Working at higher ATP concentrations may enhance the stoichiometry of substrate phosphorylation, simplifying subsequent identification of the phosphorylated residues by mass spectrometry. As shown in Figure 3, increasing the ATP concentration in the kinase reaction did not affect the background signal. Another critical feature of the KESTREL approach, namely the reduction of background phosphorylation due to endogenous kinases through the use of Mn2+ as cofactor, was evaluated by assessing the level of phosphorylation in the total 1300

Journal of Proteome Research • Vol. 4, No. 4, 2005

Troiani et al.

Figure 4. Aurora-A kinase reaction: Autoradiography of 30 µg of total cell extract phosphorylated with/without 200 ng (3 mU) of Aurora-A recombinant kinase and in the presence of Mg2+ or Mn2+ as counterions and 0.05 µM [γ-32P]ATP. No difference was observed in the presence or absence of Aurora-A either using Mg2+ or Mn2+.

Figure 5. (A) Effect of heat treatment on background of phosphorylation. Autoradiography of 30 µg of total cell extract phosphorylated in the presence/absence of Aurora-A recombinant kinase after extract incubation at 65 °C. After heat inactivation background due to endogenous kinases is completely removed both in the presence of Mg2+ and Mn2+ as cofactor (lanes 1, 2). Aurora-A activity appears to be remarkably higher in the presence of Mg2+ cofactor.

extract from nocodazole-blocked HeLa cells in the presence of either Mg2+ or Mn2+. In our case, the use of Mn2+ instead of Mg2+ as cofactor reduced the background phosphorylation levels, in agreement with previous observations.19 The extract was then submitted to phosphorylation with recombinant Aurora-A using the standard KESTREL conditions. However, no signal from Aurora-A induced phosphorylation could be detected above the high background level, irrespective of using higher ATP concentrations or using different cofactors (Figure 4). To solve this issue, we introduced an additional incubation step, where the extract was heated at 65 °C for 15 min prior to the phosphorylation reaction. When the heated extract was incubated with ATP in absence of the recombinant enzyme no background signal was detectable, irrespective of whether Mg2+ or Mn2+ was employed as cofactor (Figure 5, lanes 1-2), demonstrating that heating the extract at 65 °C for 15 min completely inactivated the endogenous kinases. When the phosphorylation reaction with Aurora-A was carried out on the

Biomarkers of Aurora-A Kinase Activity

extract after heat treatment, subsequent autoradiography revealed a clear series of bands specifically phosphorylated by Aurora-A (Figure 5, lanes 3-4). These bands were seen in the presence of 5 mM Mg2+ as well as 5 mM Mn2+, although the intensity of the phosphorylation signals was significantly higher in the presence of Mg2+. Introducing the additional step of extract heating was clearly effective, allowing detection of Aurora-A substrate bands following the in vitro kinase reaction. We repeated the experiment more than three times with distinct cell extracts confirming background elimination. However, a caveat of using such a treatment is that it may generate artifacts, possibly through heat-induced protein aggregation or precipitation, leading to a loss of potential substrates. Also, disruption of protein complexes and partial unfolding of isolated proteins, making sites available for phosphorylation that would not be accessible in a more physiological environment, could lead to false positive bands. Nevertheless, it has to be stressed that no clear discrimination of Aurora-A-dependent phosphorylation signals from background due to endogenous kinase activity could be achieved without the heat pretreatment step. Consequently, we decided to apply the KESTREL approach to Aurora-A using the most favorable conditions, namely, preheating of the extract at 65 °C for 15 min, followed by the in vitro kinase assay in the presence of 5 mM Mg2+ as cofactor and 1 mM total (cold plus hot) ATP. Substrate Identification. Although 1D gel electrophoresis provides a rapid assessment of the extent of in vitro phosphorylation by the kinase, unambiguous identification of the authentic in vitro substrates resolved by 1D gel electrophoresis is complicated by the comigration of several proteins under a seemingly single band. Consequently, as 2D gel electrophoresis is endowed with a much higher resolving power thus enabling a more reliable assignment of the true in vitro substrates, we selected this approach in order to simplify protein identification. For this experiment, we decided to analyze total and fractionated extracts in parallel. Nocodazole-blocked HeLa cell extracts were phosphorylated by Aurora-A in parallel in the presence of [γ-32P] and cold ATP under the optimized conditions, and subsequently resolved by 2D gel electrophoresis. The autoradiography images obtained from the radiolabeled samples were superimposed onto the Coomassie stained preparative gels derived from the nonlabeled experiment. Protein spots corresponding to the radioactive signals were excised from the preparative gel, digested with trypsin and identified by MALDIMS and LC-MS/MS. Figure 6 shows the 2D-gels images resulting from two experiments performed on the fractionated and total extracts. Among the phosphorylated spots detected in 2D gels of the fractionated and total extracts, BiP protein and vimentin were unambiguously identified by MALDI-MS and LC-MS/MS. Sequence coverage by MALDI-MS was very high and more than 10 different peptides were sequenced for each protein by LC-MS/MS to confirm the MALDI identification (Table 1). In addition, LC-MS/MS analysis identified one single protein per spot ruling out the possibility that the assignment of the actual substrate was hindered by comigrating higher abundance proteins. No signal corresponding to phosphorylated peptides was revealed during mass spectrometry analysis. The absence of phosphopeptide signals could be explained by the fact that phosphorylated peptides have a much lower ionization response factor than their unphosphorylated counterparts as well

research articles

Figure 6. 2D gel analysis of extracts phosphorylated in vitro with Aurora-A. (A) Fractionated extract 200 µg. (B) Total extract 200 µg. The radioactive signal obtained after kinase assay (left) is compared to the Coomassie staining of a preparative gel obtained under the same experimental conditions but without radioactive ATP (right). Protein spots comigrating with the radioactive signals were excised from gel and identified by MALDI-MS and LC-MS/MS analysis. Highlighted spots correspond to vimentin and GRP78 (Bip).

as by the possible low stoichiometry of phosphorylation. However, several residues in both proteins (Ser 4, Ser 441, Thr 525 in BiP protein; and Ser 39, Ser 66, Ser 72, Ser 226, Ser 412 in vimentin) reside within sequences that are compliant with the yeast Aurora kinase-Ipl1 phosphorylation consensus [R,K][X]1-2-[S,T]-[I,L,V].11 A similar consensus has been recently identified for the human Aurora-A kinase by using alanine scanning mutagenesis on a peptide from RalA, a downstream effector of Ras recognized as a putative substrate of the kinase.35 Interestingly, both BiP and vimentin have been previously described as putative Aurora-B substrates. BiP was actually identified through a proteomic screen exploiting a heat-treated chromosomal scaffold extract as a source of potential Aurora-B substrates, which were submitted to in vitro phosphorylation by the kinase.20 In the same study, DNA topoisomerase II was identified as an Aurora-B substrate and subsequently confirmed to be a substrate for Aurora-B in cells, thus validating the overall approach. On the other hand vimentin, which was first demonstrated to be an in vitro substrate for Aurora-B, was subsequently shown to be phosphorylated in vivo on Ser 72 in an Aurora-B dependent manner.36 Indeed, since Aurora-A and -B share the same phosphorylation consensus sequence, the identification of shared substrates was expected. Specific validation experiments involving, for example, the investigation of possible colocalization of kinase and substrate during different cell cycle phases, need to be carried out in order to discriminate and fully understand the relevance of the two Aurora isoforms with respect to the substrates identified so far. Besides these known substrates, in our in vitro screen we have identified several novel substrates. The validation of the most promising candidates as well as a careful investigation of their suitability as Aurora-A biomarkers is ongoing. Validation of Vimentin as Aurora-A in Vitro Substrate. A first step in the validation of the obtained hits should be the Journal of Proteome Research • Vol. 4, No. 4, 2005 1301

research articles

Troiani et al.

Table 1. Protein Identification by MALDI-MS and LC-MS/MS of Two In-Gel Digested Phosphorylated Spots Detected in 2D Gels of the Fractionated and Total HeLa Cell Extracts protein name/ accession no.

a

LC-MS/MS pI/MW

GRP78 human/ P11021

5.01/70kDa

Vimentin human/ P08670

5.06/53kDa

MALDI-MS

ID peptides

42%a

25%a

I60TPSYVAFTPEGER73

162b 21 ppmc

816b

80%a 161b 18 ppmc

32%a 442b

N81QLTSNPENTVFDAK95 N81QLTSNPENTVFDAKR96 T123KPYIQVDIGGGQTK13 V164THAVVTVPAYFNDAQR180 I197INEPTAAAIAYGLDKR213 I306EIESFYEGEDFSETLTR323 S447QIFSTASDNQPTVTIK463 D474NHLLGTFDLTGIPPAPR491 K619KELEEIVQPIISK632 E621LEEIVQPIISK632 L633YGSAGPPPTGEEDTAEKDEL653 S52LYASSPGGVYATR65 L80LQDSVDFSLADAINTEFK98 L80LQDSVDFSLADAINTEFKNTR101 E198EAENTLQSFRQDVDNASLAR218 N284LQEAEEWYK293 F296ADLSEAANRNNDALR311 Q323VQSLTCEVDALKGTNESLER343 E383YQDLLNVK391 L404LEGEESRISLPLPNFSSLNLR425 I412SLPLPNFSSLNLR425 E426TNLDSLPLVDTHSKR441

Sequence coverage. b MASCOT score. c Error distribution (accuracy).

Figure 7. Analysis of vimentin phosphorylation induced by Aurora-A. (A) Top: 1 µg of vimentin head domain was phosphorylated in the presence of 200 ng of Aurora-A recombinant kinase and 0.05 µM [γ-32P]ATP. A band at the expected MW (35 kDa) can be observed by autoradiography after Aurora-A induced phosphorylation. Bottom: Western blot with anti-PSer 72 vimentin antibody on samples processed with the same protocol described in A but replacing [γ-32P] ATP with cold ATP. (B) 30 µg of extract from nocodazole-treated HeLa cells were phosphorylated by Aurora-A with modified Kestrel method (see text). Vimentin was detected by Western blotting using total vimentin (top) antibody and anti-PSer 72 vimentin antibody (bottom). (C) MALDI-TOF-MS spectrum of Coomassie stained recombinant GST vimentin head domain after in vitro phosphorylation with Aurora-A after excision and in-gel digestion of the band. Phosphorylated vimentin peptide (m/z 1050.7) and its unphosphorylated counterpart (m/z 970.7) are highlighted.

identification of phosphorylated residue(s) in the endogenous protein when feasible or in a recombinant form if not. We applied this procedure to vimentin that was found highly phosphorylated in both fractionated and total extracts after phosphorylation with Aurora-A. Vimentin is a 53 kDa protein that constitutes the main part of intermediate filaments. It is organized in three domains: the head (residues 1-94), the rod (95-406) and the tail (407-465). The head domain is phosphorylated by several kinases, including PKC, ROCK, MAPK.37-39 In addition, Ser 72 was demonstrated to be a substrate for PAK40 and Aurora-B. Vimentin was shown to partially co-localize with Aurora-B at the cleavage furrow boundaries.34 Silencing mutation of Ser 71 and of Ser 72 to Ala impairs filament formation 1302

Journal of Proteome Research • Vol. 4, No. 4, 2005

in vitro, suggesting that phosphorylation at these two sites positively regulates vimentin function. Following identification of vimentin as an Aurora-A substrate through our modified KESTREL approach, we tested the ability of Aurora-A to phosphorylate vimentin in vitro. Recombinant vimentin head domain fused with GST was subjected to the kinase reaction with Aurora-A using either hot or cold ATP and substrate phosphorylation was evaluated by autoradiography and mass spectrometry. Autoradiography after incubation with recombinant Aurora-A showed a clear phosphorylated band at the expected MW (Figure 7A). The corresponding band in a Coomassie-stained gel was excised, digested and analyzed by MALDI-MS. A peak at m/z 1050.7 corresponding to vimentin

research articles

Biomarkers of Aurora-A Kinase Activity

phosphopeptide 69-77 was detected by MALDI-MS (Figure 7C). The peptide was then sequenced by LC-MS/MS and Ser 72 was identified as the phosphorylated residue (see Supporting Information). The same preparation was analyzed by immunoblotting against PSer 72 and a specific band at the expected 35 kDa MW was observed only after phosphorylation with recombinant Aurora-A. While in the starting extracts it was not possible to detect the endogenously phosphorylated vimentin, probably due to sensitivity issues or to endogenous phosphatase action during cell extract preparation in a total extract from nocodazole-blocked HeLa cells submitted to in vitro phosphorylation with Aurora-A a positive band at ∼50 kDa was recognized by the anti-PSer 72 vimentin antibody, corresponding to the protein band detected with an anti-vimentin antibody. Phosphopeptide 69-77 could not be identified in the digest of this band, possibly due to the low amount of protein and the presence in the total cell extract of many additional proteins competing for phosphorylation by Aurora-A. This substrate competition may lead to a phosphorylation stoichiometry in endogenous vimentin remarkably lower than in the shorter recombinant fragment, which is assayed as single substrate in the in vitro reaction. Taken together, our data confirmed vimentin as a true Aurora-A in vitro substrate. Further studies are required to clarify the possible physiological relevance of this finding. In conclusion, we have developed an improved KESTREL approach for kinase substrate identification by introducing a step where the material to be submitted to phosphorylation is treated with heat in order to inactivate the endogenous kinases and drastically reduce background phosphorylation. This procedure allowed us to apply KESTREL to the Aurora-A kinase under conditions optimal for the performance of the kinase under study and ultimately to identify several in vitro substrates. Our results indicate that already known substrates of the Aurora kinase family could be unambiguously identified by combining fractionation of the extract from nocodazoleblocked HeLa cells, which yielded a suitable substrate source, with our optimized KESTREL procedure, 2D gel electrophoresis and LC-MS/MS analysis. Abbreviations: ATP, adenosin triphosphate; BIP, immunoglobulin heavy chain binding protein; FBS, fetal bovine serum; GRP78, glucose regulated protein 78 kDa; GST, Glutathione S-Transferase; MCM2, minichromosome maintenance protein 2; MeOH, methanol; MOA, mechanism of action; RIPA, radioimmunoprecipitation assay buffer.

Acknowledgment. We thank Vanessa Battistella for recombinant Aurora-A kinase production, Silvia Messali for HeLa cell cultivation and Paola Storici for helpful discussion. Supporting Information Available: MS/MS spectrum of the m/z 525.842+ ion corresponding to vimentin phosphorylated peptide 69-77. This material is available free of charge at http://pubs.acs.org. References (1) Carmena, M.; Earnshaw, W. C. Nat. Rev. Mol. Cell Biol. 2003, 4, 842-854. (2) Bischoff, J. R.; Plowman, G. D. Trends Cell Biol. 1999, 9, 454459.

(3) Bischoff, J. R.; Anderson, L.; Zhu, Y.; Mossie, K.; Ng, L.; Souza, B.; et al. EMBO J. 1998, 17, 3052-3065. (4) Terada, Y.; Tatsuka, M.; Suzuki, F.; Yasuda , Y.; Fujita, S.; Otsu, M. EMBO J. 1998, 17, 667-676. (5) Meraldi, P.; Honda, R.; Nigg, E. A. Curr. Opin. Genet. Dev. 2004, 14, 29-36. (6) Giet, R.; Prigent, C. J. Cell Sci. 1999, 112, 3591-3601. (7) Warner, S. L.; Bearss, D. J.; Han, H.; Von Hoff, D. D. Mol. Cancer Ther. 2003, 2, 589-595. (8) Harrington, E. A.; Bebbington, D.; Moore, J.; Rasmussen, R. K.; Ajose-Adeogun, A. O.; Nakayama, T.; et al. Nat. Med. 2004, 10, 262-267. (9) Giet, R.; Uzbekov, R.; Cubizolles, F.; Le Guellec, K.; Prigent, C. J. Biol. Chem. 1999, 274, 15005-15013. (10) Pascreau, G.; Arlot-Bonnemains, Y.; Prigent, C. Prog. Cell Cycle Res. 2003, 5, 369-374. (11) Cheeseman, I. M.; Anderson, S.; Jwa, M.; Green, E. M.; Kang, J.; Yates, J. R., 3rd; et al. Cell 2002, 111, 163-172. (12) Keen, N.; Taylor, S. Nat. Rev. Cancer 2004, 4, 927-936. (13) Obata, T.; Yaffe, M. B.; Leparc, G. G.; Piro, E. T.; Maegawa, H.; Kashiwagi, A.; et al. J. Biol. Chem. 2000, 275, 36108-36115. (14) Hutti, J. E.; Jarrell, E. T.; Chang, J. D.; Abbott, D. W.; Storz, P.; Toker, A.; et al. Nat. Methods 2004, 1, 27-29. (15) Zhang, H.; Zha, X.; Tan, Y.; Hornbeck, P. V.; Mastrangelo, A. J.; Alessi, D. R.; et al. J. Biol. Chem. 2002, 277, 39379-39387. (16) Maly, D. J.; Allen, J. A.; Shokat, K. M. J. Am. Chem. Soc. 2004, 126, 9160-9161. (17) Shah, K.; Shokat, K. M. Methods Mol. Biol. 2003, 233, 253-271. (18) Tien, A. C.; Lin, M. H.; Su, L. J.; Hong, Y. R.; Cheng, T. S.; Lee, Y. C.; et al. Mol. Cell. Proteomics 2004, 3, 93-104. (19) Knebel, A.; Morrice, N.; Cohen, P. EMBO J. 2001, 20, 4360-4369. (20) Morrison, C.; Henzing, A. J.; Jensen, O. N.; Osheroff, N.; Dodson, H.; Kandels-Lewis, S. E.; et al. Nucleic Acids Res. 2002, 30, 53185327. (21) Marsden, M. P.; Laemmli, U. K. Cell 1979, 17, 849-858. (22) Hsu, J. Y.; Sun, Z. W.; Li, X.; Reuben M.; Tatchell, K.; Bishop, D. K.; et al. Cell 2000, 102, 279-291. (23) Bayliss, R.; Sardon, T.; Vernos, I.; Conti, E. Mol. Cell. 2003, 12, 851-862. (24) Eyers, P. A.; Maller, J. L. J. Biol. Chem. 2004, 279, 9008-9015. (25) Dutertre, S.; Cazales, M.; Quaranta, M.; Froment, C.; Trabut, V.; Dozier, C.; et al. J. Cell Sci. 2004, 117, 2523-2531. (26) Du, J.; Hannon, G. J. Proc. Natl. Acad. Sci. U.S.A. 2004, 101, 89758980. (27) Ouchi, M.; Fujiuchi, N.; Sasai, K.; Katayama, H.; Minamishima, Y. A.; Ongusaha, P. P.; et al. J. Biol. Chem. 2004, 279, 1964319648. (28) Liu, Q.; Kaneko, S.; Yang, L.; Feldman, R. I.; Nicosia, S. V.; Chen, J.; et al. J. Biol. Chem. 2004, 279, 52175-52182. (29) Lange, B. M.; Bachi, A.; Wilm, M.; Gonzalez, C. EMBO J. 2000, 19, 1252-1262. (30) Andersen, J. S.; Wilkinson, C. J.; Mayor, T.; Mortensen, P.; Nigg, E. A.; Mann, M. Nature 2003, 426, 570-574. (31) Adolphs, K. W.; Cheng, S. M.; Paulson, J. R.; Laemmli, U. K. Proc. Natl. Acad. Sci. U.S.A. 1977, 74, 4937-4941. (32) McNeill, H.; Knebel, A.; Arthur, J. S.; Cuenda, A.; Cohen, P. Biochem. J. 2004, 384, 391-400. (33) Murray, J. T.; Campbell, D. G.; Peggie, M.; Mora, A.; Cohen, P. Biochem. J. 2004, 384, 489-494. (34) Sun, C.; Newbatt, Y.; Douglas, L.; Workman, P.; Aherne, W.; Linardopoulos, S. J. Biomol. Screen. 2004, 9, 391-397. (35) Wu, J. C.; Chen, T. Y.; Yu, C. T.; Tsai, S. J.; Hsu, J. M.; Tang, M. J.; et al. J. Biol. Chem. 2005, in press. (36) Goto, H.; Yasui, Y.; Kawajiri, A.; Nigg, E. A.; Terada, Y.; Tatsuka, M.; et al. J. Biol. Chem. 2003, 278, 8526-8530. (37) Yasui, Y.; Goto, H.; Matsui, S.; Manser, E.; Lim, L.; Nagata, K.-i.; Inagaki, M. Oncogene 2001, 20, 2868-2876. (38) Goto, H.; Kosako, H.; Tanabe, K.; Yanagida, M.; Sakurai, M.; Amano, M.; et al. J. Biol. Chem. 1998, 273, 11728-11736. (39) Cheng, T. J.; Lai, Y. K. J. Cell. Biochem. 1998, 71, 169-181. (40) Goto, H.; Tanabe, K.; Manser, E.; Lim, L.; Yasui, Y.; Inagaki, M. Genes Cells 2002, 7, 91-97.

PR050018E

Journal of Proteome Research • Vol. 4, No. 4, 2005 1303