in vivo profiling endogenous interactions with knock-out in mammalian cells

In Vivo Profiling Endogenous Interactions withKnock-Out in Mammalian Cells

Ling Xie,† Linhong Jing,‡ Yanbao Yu,†,§ Kazuhiro Nakamura,| Carol E. Parker,‡ Gary L. Johnson,|

and Xian Chen*,†,‡

Department of Biochemistry & Biophysics and Department of Pharmacology, School of Medicine,University of North Carolina, 120 Mason Farm Road, Genetic Medicine, Ste 3010, Campus Box No. 7260,Chapel Hill, North Carolina 27599-7260, UNC-Duke Proteomics Center, Chapel Hill, North Carolina 27599-7260, andInstitutes for Biomedical Sciences, Fudan University, Shanghai, China 200032

To precisely identify and screen target-specific protein-protein interactions at the endogenous level, here weintroduce a novel quantitative proteomic method we havetermed in vivo Profiling Endogenous Interactions withKnock-out (iPEIK). In our design, mouse embryonicfibroblasts (MEFs) derived from target gene knockout(KO) mice can be stable isotope-tagged and serve as atarget-free background to “light-up” the target protein-specific protein complex formed in the correspondingwild-type (WT) cells. In mass spectrometric analysis ofthe pairs of non-labeled versus heavy isotope-labeledpeptide signals derived from WT versus KO cells, respec-tively, we then quantitatively measured the abundancedifferences of the proteins in the complex immunopre-cipitated (IP) from the target-expressing WT versus target-absent KO cells, respectively. Those proteins detectedwith little or no presence in the cells of KO origin weredetermined as target-specific interacting partners. Fur-ther, dynamic interactors could be identified throughdifferent IP mixing schemes. Using iPEIK we identifiedmultiple interacting partners both previously known andunknown to be associated with mitogen-activated proteinkinase kinase kinase 2 (MEKK2). Because of the avail-ability of a large library of knockout mice models withvarious target proteins of biological interests our methodis generally applicable to screen any endogenous target-specific PPIs of physiological relevance.

In mamalian cells the low-abundance, transient, and dynamicnature of endogenous protein-protein interactions (PPIs) demandfor highly accurate and sensitive methods to distinguish targetprotein-specific interactions. Immunoprecipitation (IP) is a com-monly used method to pull down a target protein and itsinteracting partners. However, the lack of a threshold to distin-guish target-specific binding partners in an IP product with non-specific contaminants often results in a high degree of false-positive identifications, and this situation can be pronounced with

antibodies of low specificity. Tandem affinity purification (TAP)was then developed to isolate protein complexes in high purityprior to mass spectrometric (MS) identification.1,2 In general,multiple steps of epitope affinity-based purifications are requiredto reduce the contaminating proteins bound non-specifically to abait protein. As a tradeoff, the repetitive affinity washings oftenlead to the loss of weak or transient interactions of biologicalrelevance. Also, unlike in yeast cells in which it can be engineeredto have only the exogenously tagged target expressing, inmammalian cells the untagged endogenous target protein alwaysexpress which represents the background affecting the specificityof distinguishing target-specific interactors and the yield of thecomplex pulled-down through the tagged target protein.3

Because of still limited sensitivity and accuracy of currentlyavailable abundance-based mass spectrometric (MS) methods inanalyzing low-abundance proteins, most of the PPIs were identifiedin the protein complexes with a target protein expressing at non-physiologically relevant levels when the identification of endog-enous PPIs is highly challenging.4-6 As an effective way toimprove signal specificity, stable isotope labeling (SIL) can assistMS for large-scale protein quantification.7-10 Since 1999, we havebeen developing a quantitative proteomic strategy of amino acid-coded mass tagging (AACT) 8,11 or SILAC as named by anothergroup12 which has provided a high throughput quantitative

* To whom correspondence should be addressed. E-mail: xian_chen@med.unc.edu. Fax: (919) 966-2852.

† Department of Biochemistry & Biophysics, University of North Carolina.‡ UNC-Duke Proteomics Center.§ Fudan University.| Department of Pharmacology, School of Medicine, University of North

Carolina.

(1) Zhou, Q.; Lieberman, P. M.; Boyer, T. G.; Berk, A. J. Genes Dev. 1992, 6,1964–1974.

(2) Zhou, Q.; Lieberman, P. M.; Boyer, T. G.; Berk, A. J. Genes Dev. 1992, 6,1964–1974.

(3) Rigaut, G.; Shevchenko, A.; Rutz, B.; Wilm, M.; Mann, M.; Seraphin, B.Nat. Biotechnol. 1999, 17, 1030–1032.

(4) Domon, B.; Aebersold, R. Science 2006, 312, 212–217.(5) Domon, B.; Alving, K.; He, T.; Ryan, T. E.; Patterson, S. D. Curr. Opin.

Mol. Ther. 2002, 4, 577–586.(6) Aebersold, R.; Mann, M. Nature 2003, 422, 198–207.(7) Chen, X. In Methods in Molecular Biology/Methods in Molecular Medicine;

Human Press, 2005.(8) Chen, X.; Smith, L. M.; Bradbury, E. M. Anal. Chem. 2000, 72, 1134–

1143.(9) Oda, Y.; Huang, K.; Cross, F. R.; Cowburn, D.; Chait, B. T. Proc. Natl. Acad.

Sci. U.S.A. 1999, 96, 6591–6596.(10) Goodlett, D. R.; Keller, A.; Watts, J. D.; Newitt, R.; Yi, E. C.; Purvine, S.;

Eng, J. K.; von Haller, P.; Aebersold, R.; Kolker, E. Rapid Commun. MassSpectrom. 2001, 15, 1214–1221.

(11) Zhu, H.; Pan, S.; Gu, S.; Bradbury, E. M.; Chen, X. Rapid Commun. MassSpectrom. 2002, 16, 2115–2123.

(12) Ong, S. E.; Blagoev, B.; Kratchmarova, I.; Kristensen, D. B.; Steen, H.;Pandey, A.; Mann, M. Mol. Cell. Proteomics 2002, 1, 376–386.

Anal. Chem. 2009, 81, 1411–1417

10.1021/ac802161d CCC: $40.75 2009 American Chemical Society 1411Analytical Chemistry, Vol. 81, No. 4, February 15, 2009Published on Web 01/21/2009

solution8,13-20 for any comparative analysis of global changes indisease-related protein expression,13,18,21 post-translational modi-fications,14 and lately PPIs.15 Particularly, to increase the sensitivityand accuracy in distinguishing specific PPI in IP pull-downcomplexes, we have previously introduced a “dual-tagging”(epitope and AACT isotope tags) quantitative proteomic method22

that integrates the capabilities of natural complex formation,epitope affinity isolation, and “in-spectra” quantitative markers todistinguish systematically those interacting proteins from a largenon-specific binding background.22,23 However, in addition to thetedious procedures involved in establishing the cell lines thatstably express the bait/target protein close to endogenous levels,the use of exogenous tagging strategy for PPI screening alwayscauses concerns about the epitope-tagged protein that could bedifferent from its endogenous version both at expression leveland the controlling elements at transcription step. Recently amethod of quantitative IP combined with knockdown (QUICK)was introduced,24 which uses RNA intereference (RNAi) to knock-down target gene for the background control and SILAC asquantitative markers to measure abundance changes of target-specific interactors in the complexes immunoprecipitated from awild-type cell line. Inevitably, the sensitivity and accuracy of thisquantitative method in judging target-specific interactors versusnon-specific background primarily rely on the efficiency of RNAi-based bait knock-down as the leftover bait protein in the controlline may increase the background for quantitative analysis.Furthermore, protein expression is not always proportional to theamount of RNA, especially for those proteins with long turnoverrate. Moreover, when either epitope tagging or QUICK approachto pull down bait-specific immunoprecipitates in mammalian cellsare used, the interference and competition for binding fromuntagged/remaining endogenous counterpart could be a majorconcern to be addressed. To precisely characterize PPIs occurringin real time in the cell lines derived directly from tissue cells oreven later in primary cells, here we report a novel quantitativeproteomic method of in vivo Profiling Endogenous Interactionswith Knockout (iPEIK). In mass spectrometric analysis, theimmunoprecipitates originated from bait knockout mouse cellsserve as a “clean” background to distinguish the specific bait-

interacting proteins immunoprecipated from those of wild-typemice.

Mitogen-activated protein kinases (MAPKs) are ubiquitouslyexpressed and regulate a wide variety of functions in virtually all celltypes (see review, ref 25). Dysregulated MAPK activity is associatedwith a variety of pathological states, including those arising frominflammation, such as arthritis and inflammatory bowel disease,26,27

as well as the syndromes that include the uncontrolled cellularproliferation and tissue remodeling characteristic of cancer.28 How-ever, little is known about the molecular mechanisms underlying theMAPK-related disease pathogenesis. In this regard, the identificationof key components and their interactions in MAPK pathways isessential to understand novel functions of this superkinase family. Itis known that MAPKs are the terminal kinase in a three kinasephosphor-relay module, in which MAPKs are phosphorylated andactivated by mitogen-activated protein kinase kinase MKKs, whichthemselves are phosphorylated and activated by mitogen-activatedkinase kinase kinases (MKKKs) (see review, ref 29). MEKK2 is oneof more than 20 MKKKs known so far. MEKK2 is involved in bothextracellular-related kinase 5 (ERK5) and c-Jun N-terminal kinaseJNK signal pathways as its binding to different partners such asMEK5 and MKK7 can coordinately control ERK5 and c-Jun N-terminal kinase activation.30 Phosphorylated MEKK2 is regulated bySmurf1, a HECT domain ubiquitin ligase, which controls osteoblastactivity and bone homeostasis by targeting MEKK2 for degradation.31

As the first step to systematically reveal MEKK2-specific functionalnetwork, as well as to evaluate the effectiveness of our design, wechose to perform an unbiased screening on identifying the interactingpartners of MEKK2 at the endogenous level. As MEKK2 knockoutmice is vital and fertile, suggesting that it is developmentallydispensable,32 the mouse MEKK2-/- and wild-type MEF cell lineswhich are derived from the embryo of MEKK2 knockout and wildtype mice, respectively, 32 are used in our study.

EXPERIMENTAL METHODSCell Culture and Reagents. Wild-type and MEKK2-/- mouse

embryonic fibroblasts (MEFs) were isolated as describedpreviously and grown in Dulbecco’s Modification of Eagle’sMedium with 10% fetal bovine serum, 100 U/mL penicillin, and100 µg/mL streptomycin at 37 °C with 5% CO2.32 For a SILAC/AACT experiment, MEKK2-/- MEF cells were grown in theregular leucine-depleted DMEM later supplemented with 110mg/L of D3-L-Leucine (Cambridge Isotope Laboratories) whilewild-type MEF cells were cultured in the regular medium. Amouse monoclonal antibody (mAb) for MEKK2 was generatedagainst recombinant MEKK2.

Immunoprecipitation and Immunoblotting Analysis. TheMEF cells were washed with ice-cold phosphate-buffered saline

(13) Zhu, H.; Pan, S.; Gu, S.; Bradbury, E. M.; Chen, X. Rapid Commun. MassSpectrom. 2002, 16, 2115–2123.

(14) Zhu, H.; Hunter, T. C.; Pan, S.; Yau, P. M.; Bradbury, E. M.; Chen, X. Anal.Chem. 2002, 74, 1687–1694.

(15) Gu, S.; Pan, S.; Bradbury, E. M.; Chen, X. Anal. Chem. 2002, 74, 5774–5785.

(16) Chen, X. Methods in Molecular Biology/Methods in Molecular Medicine BookSeries; Humana Press: Totowa, NJ, 2004.

(17) Gu, S.; Pan, S.; Bradbury, E. M.; Chen, X. J. Am. Soc. Mass Spectrom. 2003,14, 1–7.

(18) Pan, S.; Gu, S.; Bradbury, E. M.; Chen, X. Anal. Chem. 2003, 75, 1316–1324.

(19) Hunter, T. C.; Yang, L.; Zhu, H.; Majidi, V.; Bradbury, E. M.; Chen, X. Anal.Chem. 2001, 73, 4891–4902.

(20) Harris, M. N.; Ozpolat, B.; Abdi, F.; Gu, S.; Legler, A.; Mawuenyega, K. G.;Tirado-Gomez, M.; Lopez-Berestein, G.; Chen, X. Blood 2004, 104, 1314–1323.

(21) Gu, S.; Chen, J.; Dobos, K. M.; Bradbury, E. M.; Belisle, J. T.; Chen, X.Mol. Cell. Proteomics 2003, 2, 1284–1296.

(22) Wang, T.; Gu, S.; Ronni, T.; Du, Y. C.; Chen, X. J. Proteome Res. 2005, 4,941–949.

(23) Du, Y. C.; Gu, S.; Zhou, J.; Wang, T.; Cai, H.; Macinnes, M. A.; Bradbury,E. M.; Chen, X. Mol. Cell. Proteomics 2006, 5, 1033–1044.

(24) Selbach, M.; Mann, M. Nat. Methods 2006, 3, 981–983.

(25) Pearson, G.; Robinson, F.; Beers Gibson, T.; Xu, B. E.; Karandikar, M.;Berman, K.; Cobb, M. H. Endocr. Rev. 2001, 22, 153–183.

(26) Johnson, G. L.; Lapadat, R. Science 2002, 298, 1911–1912.(27) Hollenbach, E.; Neumann, M.; Vieth, M.; Roessner, A.; Malfertheiner, P.;

Naumann, M. FASEB J. 2004, 18, 1550–1552.(28) Gollob, J. A.; Wilhelm, S.; Carter, C.; Kelley, S. L. Semin. Oncol. 2006, 33,

392–406.(29) Cuevas, B. D.; Abell, A. N.; Johnson, G. L. Oncogene 2007, 26, 3159–3171.(30) Nakamura, K.; Johnson, G. L. Mol. Cell. Biol. 2007, 27, 4566–4577.(31) Yamashita, M.; Ying, S. X.; Zhang, G. M.; Li, C.; Cheng, S. Y.; Deng, C. X.;

Zhang, Y. E. Cell 2005, 121, 101–113.(32) Garrington, T. P.; Ishizuka, T.; Papst, P. J.; Chayama, K.; Webb, S.; Yujiri,

T.; Sun, W.; Sather, S.; Russell, D. M.; Gibson, S. B.; Keller, G.; Gelfand,E. W.; Johnson, G. L. EMBO J. 2000, 19, 5387–5395.

1412 Analytical Chemistry, Vol. 81, No. 4, February 15, 2009

and lysed with solubilizing buffer (1% NP-40, 10 mM Tris [pH7.5], 150 mM NaCl, 0.4 mM EDTA, 2 mM Na3VO4, 1× phospho-tase inhibitor cocktail (Pierce), 1× protease inhibitor cocktail(sigma-aldrich)) on ice. The cell lysate was centrifuged at100,000g for 60 min, and supernatants were retained for furtherprocessing.

For IP, the monoclonal anti-MEKK2 was first coupled toprotein G agarose (Genscript) with dimethyl pimelimidate. For apost-mixing IP experiment, 20 µg of conjugated antibody wasincubated with the lysate from approximately 1-2 × 108

MEKK2-/- and wild-type MEF cells, separately, for 6 h at 4°C. The immunoprecipitates were collected by centrifugationfollowed by washing with solubilization buffer three times andPBS twice. The immunoprecipitated complex was eluted with0.1 M Glycine (pH2.5). The protein concentration was deter-mined by BCA assay. The eluted proteins from WT and KOwere mixed by 1:1 and precipitated with 1/100 (v/v) of 2% Nadeoxycholate and 1/10 of 100% Trichloroacetic acid overnightat 4 °C. The precipitates were redissolved in 1× SDS gel samplebuffer at 70 °C and separated on NuPAGE 4-12% SDS gel(Invitrogen). The gel was stained by G-250 and contiguouslycut for in-gel digestion. For a premixing IP experiment, theWT and KO cell lysates were mixed by the same protein massfollowed by immunopreciptation as described above.

For immunobloting, 5 mg of cell lysate was incubated with 1µg of anti-MEKK2 tethered on protein G agarose. The immunecomplexes were washed, eluted, separated by SDS-PAGE, andtransferred to a PVDF membrane. After blocking, membraneswere blotted with selected antibodies and visualized using the ECLplus detection system (GE).

In-Gel Digestion and Reverse-Phase NanoLC-MS/MSAnalysis. The gel bands were loaded on ProGest autodigester(Genomic Solutions) for tryptic digestion. The extracted peptidesolution was dried with speed-vac. The peptide pellets wereredissolved in 0.5% acidic acid, desalted with reversed phaseC18-packed tips, and eluted with 80% acetonitrile/0.5% aceticacid. The organic solvent was removed by speed-vac. Thevolume of digested peptides was brought up in 0.1% formic acid.Each sample was separated with online Eksigent nanoLCsystem and analyzed by a LTQ-Orbitrap hybrid mass spectrom-eter (Thermo Electron, San Jose, CA), which was equipped witha nano electrospray source (New Objective, Inc., Woburn, MA).The peptides were loaded on IntegraFrit Sample trap (Pro-teoPep II C18, 300 Å, 5 µm, 75 µm × 25 mm, New Objective,Inc., Woburn, MA) by using mobile phase of 50% acetonitrilein 0.1% formic acid. The retained peptides were washedisocratically with the same buffer to remove any excessreagents. The cleaned peptides were resolved on a PicoFritAnalytical Columns (ProteoPep II C18, 300 Å, 5 µm, 50 µm ×100 mm, tip ID ) 10 µm, New Objective, Inc., Woburn, MA)or Dionex 75 um × 150 mm column with a multistep gradientof solvent 2A (water premixed with 0.1% formic acid) andsolvent 2B (acetonitrile premixed with 0.1% formic acid).

Data Analysis. The LC-MS-MS/MS raw data was convertedto DTA files using ThermoElectron Bioworks 3.3.1 and correlatedto theoretical fragmentation patterns of tryptic peptide sequencesfrom the Fasta databases using SEQUEST. Search parametersincluded (1) variable modifications allowing mass increase of 80

Da for possible phophorylation and 3 Da for 3-deuterium labeledLeucine; (2) restricted to trypsin generated peptides, allowing fortwo missed cleavages; (3) The criteria for peptide were based ontop hit(s) with individual cross correlation exceeding a thresholddependent on the precursor charge state. The proteins matchedwith at least two peptides and probability smaller than 0.001 (P <0.001) were considered as positive identification. PepQuan wasused in SILAC protein quantification.

RESULTS AND DISCUSSIONGeneral iPEIK Experimental Design. As illustrated in

Figure 1, KO cells are cultured in a “heavy” medium supplementedwith particular type of heavy amino acids, for example, a selectedtype of either deuterium, or/ 13C, or/ 15N-enriched amino acidswhile WT cells are grown in the “light/regular” medium. EachWT or KO cell lysate is incubated with monoclonal anti-MEKK2covalently coupled to the protein G/A-agarose or magneticbeads for IP. The heavy and light immunoprecipitates are thencombined at a 1:1 ratio of total protein mass in a “post-IPmixing” scheme (Figure 1 right). Also to further distinguishthose specific interactors with high on- and off-rate in their bindingto the target protein, we also perform a single IP experiment onthe mixed WT and KO cell lysates in a “pre-IP mixing” scheme(Figure 1 left). The proteins in each set of immunoprecipitatesare separated on 1D-SDS gels, digested, and the eluted peptidesfrom each gel slice are subjected to LC-MS/MS analysis. In MSspectra, the peptide signals from target/bait protein should showonly the light isotope peak. Its specific binding proteins shouldpresent high light-to-heavy isotope ratio (L/H) while non-specificcontaminants show equal intensity for both light and heavy isotopesignals.

Moreover, if some interactors are detectable in both post-IPand pre-IP mixing scheme, the dynamic interacting componentscan display different L/H ratios for their peptide signals, depend-ing on the approach taken through either post-IP or pre-IP mixingscheme. Because of possible back-exchanges between light andheavy version of a protein during the IP incubation step, we expectto observe a relatively lower L/H for a target-specific interactorin the pre-IP mixing than that in the post-IP mixing.

Profiling the Immunoprecipitated Complex Associatedwith MEKK2. As shown in Figure 2, in both pre-IP and post-IPpair mixing runs, the bait MEKK2 was identified with no signalat the m/z corresponding to its heavy leucine-containing peptides,indicating the absence of the bait protein from MEKK2-/- cells(Figure 2a). Multiple peptides of MEKK2 representing ap-proximately 40% of its full-length sequence coverage were identi-fied with similar infinite L/H ratio. Using either post-IP or pre-IPmixing approach, we identified 96 and 130 proteins respectivelywith at least two peptides sequenced for each protein identificationat high confidence (p < 0.001) (see Supporting Information, Tables1 and 2). Among them, 86 and 114 proteins had multiple leucine-containing peptides for quantification of L/H ratios, respectively.The peptide signals from most of those proteins such as Actinshowing their L/H ratio around 1 (Figure 2b) were consideredas non-specific contaminants because they were randomly andequally associated with the antibody beads regardless of thepresence or absence of MEKK2 in the cells.

In reference to the measured L/H ratio of certain proteinspreviously known to interact with MEKK2, we set the threshold for

1413Analytical Chemistry, Vol. 81, No. 4, February 15, 2009

determining new MEKK2-interacting proteins identified by MS asfollows: an L/H ratio over 2 or a 100% increase in abundance bycomparing the peptides extracted from wild-type versus KO cellsthrough post-IP mixing scheme, while following pre-IP mixing

scheme, the L/H ratio was set at over 1.3 because of possible back-exchange effect. The proteins with relatively high L/H ratios werelisted in Table 1. The distribution of L/H ratios versus the numberof the identified proteins was given in Figure 2e.

Figure 1. Schematic diagram of iPEIK design for screening endogenous PPIs. Both MEF cell lines with either the target gene knockout or thewild-type with endogenous gene retained can be generated from particular tissue or organ of the corresponding knockout and wild-type mice,respectively. When a selected MEF cell line is cultured in the media containing a particular type of stable isotope-enriched or heavy aminoacids, AACT can be metabolically incorporated into the cellular proteins in a residue-specific way. Depending upon the purpose of learning thenature of the interactions, either post-IP mixing or pre-IP mixing scheme can be separately used or both used for comparative analysis. The“bait” protein (indicated by blue oval), its direct binding partners (triangle/stable, moon/dynamic) and indirect interacting protein (rectangular) arepulled down along with some non-specific contaminants (diamond). Color: blue coded proteins are from wild-type cells in light media, red onesfrom knockout cells in heavy media.

1414 Analytical Chemistry, Vol. 81, No. 4, February 15, 2009

Figure 2. Identification of endogenous interacting partners in the immnuoprecipitates. (a) Representative MS spectra for MEKK2 peptidesshowed that MEKK2 is only present in light form from wild-type cells and there is no signal originated from knockout cells. (b) The majorityof protein components, such as Actin, in the complex were identified with no abundance change between their light and heavy forms; (c,d) MEK5 and 14-3-3 are known interacting proteins with MEKK2. (e) The open circles are the L/H ratios of proteins identified followingpost-IP mixing scheme, and the filled circles correspond to those following pre-IP mixing scheme. The details of all identified proteins arelisted in Supporting Information, Tables 1 and 2. (f) Some proteins, such as Heat Shock protein 90 R, are identified with significantlyhigher ratio in post-IP scheme than that in pre-IP scheme. This may indicate their dynamic or weak interactions with the bait protein.

1415Analytical Chemistry, Vol. 81, No. 4, February 15, 2009

Detection of Known MEKK2-Interacting Proteins in theImmunoprecipitates Derived from WT Cells in Contrast tothe KO Background. By using iPEIK approach a few proteinspreviously known for their involvements in the MAPK signalingpathway were identified as the MEKK2 interactors, which includeMEK5, heat shock protein 90R and 14-3-3 proteins (Table 1).Those proteins have been previously found to bind MEKK2 byeither two-hybrid screening assay, or tandem affinity purification,or in vitro pull-down assay with over-expressed MEKK2 protein.33-35

In the post-IP mixing complex, the L/H ratios of 3.98 and 4.55were found for Hsp90aa1 and14-3-3 zeta/delta, respectively,suggesting their specific interactions with MEKK2 (Figure 2d and2f left). Interestingly, MEK5 was identified in the pre-IP mixingwith L/H at 1.6 after averaging all populations of light and thedeuterium-containing MEK5 peptide signals possibly offset duringthe µLC elution. The heavy population of the MEK5 peptides wasprobably due to fast exchange between “light” MEK5 bound toMEKK2 in wild-type cells and the free “heavy” unbound MEK5expressed in KO cells when the heavy and light cell lysates weremixed. Indeed, the high on-rate and off-rate of in vitro bindingbetween MEK5 and MEKK2 measured by Biacore indicated thesehighly dynamic interactions between both proteins (ka ) 5.8 ×105 M-1 s-1, kd ) 0.055 s-1, unpublished data). Anotherpossibility is that free MEKK2 in WT cell lysate interacted with

free MEK5 in the MEKK2-/- cell lysate when they are mixedfor IP. Similarly, the L/H ratio of HSP90aa1 was found closeto 1 in the pre-IP mixing complex in comparison with that of4.55 in the post-IP mixing one (Figure 2f). This may also betrue for other proteins identified in the immunoprecipitatesfollowing pre-IP mixing scheme, such as TRAF7 and isoform1aof C-jun-amino-terminal kinase-interacting protein 3 (Mapk8ip3)(Supporting Information, Table 2). As shown in Figure 2e, moreproteins with high L/H ratios were identified in the post-IPscheme than those using the pre-IP mixing approach. Apparently,pre-IP mixing simplifies the experimental procedure and decreasesthe abundance difference for non-specific binding proteins fromtwo cell populations, but it could be problematic for those proteinsbound to the enzymes through weak and/or dynamic interactions.If an interactor is detected in both approaches, the comparativeanalysis of its ratios could suggest its binding strength to thetarget/bait protein.

Newly Identified Components in the MEKK2-InteractingComplex. In addition to these known interactors, the proteinsidentified with L/H ratios larger than 2 were considered aspossible novel MEKK2-interacting partners. These identificationsinclude Zipper interacting protein kinase (ZIPK, also known asdeath-associated protein kinase 3 [DAPK3]), Ro 52 protein(Trim21), serum deprivation response protein (SDPR), and soforth. (Figure 3 and Table 1). Similar to the bait MEKK2, ZIPKwas identified with an infinite L/H ratio or no signal detected fromits heavy version immunoprecipitated from the KO cell line(Figure 3a). Ro52 was also identified with L/H at 2.49 (Figure3b). Because of its L/H ratio at close to the threshold observedin MS analysis, Ro52 was chosen for further validation by using

(33) Bouwmeester, T.; Bauch, A.; Ruffner, H.; Angrand, P. O.; Bergamini, G.;Croughton, K.; Cruciat, C.; Eberhard, D.; Gagneur, J.; Ghidelli, S.; Hopf,C.; Huhse, B.; Mangano, R.; Michon, A. M.; Schirle, M.; Schlegl, J.; Schwab,M.; Stein, M. A.; Bauer, A.; Casari, G.; Drewes, G.; Gavin, A. C.; Jackson,D. B.; Joberty, G.; Neubauer, G.; Rick, J.; Kuster, B.; Superti-Furga, G. Nat.Cell Biol. 2004, 6, 97–105.

(34) Fanger, G. R.; Widmann, C.; Porter, A. C.; Sather, S.; Johnson, G. L.;Vaillancourt, R. R. J. Biol. Chem. 1998, 273, 3476–3483.

(35) Nakamura, K.; Johnson, G. L. J. Biol. Chem. 2003, 278, 36989–36992.

Table 1. Summary of Proteins Identified in the MEKK2 Complex

no. of peptides

accession no. protein total quantification L/H ratioPost-IP Mixing Scheme

IPI00117088.2 Mitogen-activated protein kinase kinase kinase 2 14 10 ∞IPI00117846.1 Death-associated protein kinase 3 2 2 ∞IPI00330804.3 Heat shock protein HSP 90-alpha 2 2 3.96IPI00129517.1 Isoform Mitochondrial of Peroxiredoxin-5, mitochondrial precursor 3 1 3.80IPI00114407.2 THO complex 4 2 1 3.56IPI00131871.1 COP9 signalosome complex subunit 4 2 1 3.55IPI00331738.4 52 kDa Ro protein 5 5 2.49IPI00135660.3 Serum deprivation-response protein 3 2 2.08

pSDPR 1 1 24.4

IPI00116498.1 14-3-3 protein zeta/delta 4 3 4.55IPI00227392.4 14-3-3 protein eta 2 2 3.46IPI00230682.6 14-3-3 protein beta/alpha 3 2 2.15IPI00230707.5 14-3-3 protein gamma 3 3 2.04IPI00118384.1 14-3-3 protein epsilon 4 3 1.85

IPI00116281.2 T-complex protein 1 subunit zeta 2 1 3.16IPI00118678.1 T-complex protein 1 subunit alpha A 2 1 2.70IPI00320217.8 T-complex protein1 subunit beta 3 2 2.25IPI00116277.2 T-complex protein 1 sununit delta 4 2 2.25IPI00469268.4 T-complex protein 1 subunit theta 5 2 2.08

Pre-IP Mixing SchemeIPI00117088 Map3k2 Mitogen-activated protein kinase kinase kinase 2 17 14 ∞IPI00121471 Serpinb6a 2 2 1.64IPI00126447 Map2k5 Isoform 1 of Dual specificity mitogen-activated protein kina 3 1 1.60IPI00177038 Actr2 Actin-related protein 2 2 2 1.60IPI00330497 Kank2 Isoform 1 of KN motif and ankyrin repeat domain-containing 3 2 1.51IPI00230435 Lmna Isoform C2 of Lamin-A/C 2 1 1.50IPI00400300 Lmna Isoform C of Lamin-A/C 3 2 1.37IPI00119478 Tmod3 Tropomodulin-3 3 3 1.30

1416 Analytical Chemistry, Vol. 81, No. 4, February 15, 2009

co-IP and Western blotting. As shown in Figure 4 (left panel),Ro52, as well as MEK5, was detected only in the immunoprecipi-tates from wild-type MEF along with MEKK2. Meanwhile, thecellular expression of both MEK5 and Ro52 were found un-changed comparing their abundances available in WT versus KOcells (Figure 4 right panel). This observation explains why theresidual population of either MEK5 or Ro52 was observed at theirheavy isotope signals, that is, because of their expressions alsoin KO cells, small populations of these proteins may come downduring the IP experiments performed in target-absence KO cells.Meanwhile MS is more sensitive to detect these populations.

Interestingly among those newly identified interactors, bothnon- and phospho-peptides belonging to SDPR were identified.The L/H ratio of the SDPR nonphosphopeptides was at 2 whencomparing its population in MEKK2 versus MEKK2-/- immu-

noprecipitates (Figure 3c left). Meanwhile a significantly high L/Hat approximately 24 was observed for the SDPR phosphopeptide,suggesting that the phosphorylated SDRP is predominantlyinvolved in the MEKK2 complex (Figure 3c right). Further, thevalidation and characterization of other possible MEKK2-specificcomponents in the identified MEKK2 complex are carried out byusing a variety of biological assays, and the corresponding resultswill be reported else.

In conclusion, the success of our method relies on the taggingwith stable isotopes at the complexes immunoprecipitated fromthe cells derived from wild-type or target protein-depleted knock-out mice. In MS analysis, the isotopic signals observed for thebait protein derived from knockout mouse cells are serving as a“clean” background from which to distinguish the specific bait-interacting proteins immunoprecipitated from those of wild-typeorigin. To decode biologically relevant protein interaction networksin vivo, iPEIK is generally applicable to those biological systemswith available knockout mice models.

ACKNOWLEDGMENTThis work was supported by U.S. NIH 1R01AI064806-01A2,

and U.S. Department of Energy, the Office of Science (BER),Grant DE-FG02-07ER64422.

SUPPORTING INFORMATION AVAILABLESummary of proteins in MEKK2 complex with pre- and post-

IP mixing scheme. This material is available free of charge viathe Internet at http://pubs.acs.org.

Received for review October 12, 2008. Accepted January6, 2009.

AC802161D

Figure 3. Newly identified MEKK2-interacting partners. Possible novel MEKK2-interacting partners are identified with L/H ratio higher than 2,such as Zipper interacting protein kinase (Zipk) (a), Ro52 protein (Trim21) (b), serum deprivation response protein (SDPR) and its phosphorylatedform (pSDPR) (c, left and right) (see Supporting Information).

Figure 4. Validation of selected components in MEKK2 immuno-precipitate using immunoblotting. Left panel: The immunoprecipitatesobtained from MEKK2-/- or wild type MEF cells by using anti-MEKK2 antibody were immunoblotted with the corresponding antibodyof each selected protein. Right panel: whole cell lysate (WCL) wereimmunoblotted with the corresponding antibody of each selectedprotein.

1417Analytical Chemistry, Vol. 81, No. 4, February 15, 2009