Proteomic Complex Detection Using Sedimentation - American

teomic Complex Detection using Sedimentation) for pro- filing the sedimentation of a large number of proteins through a rate zonal centrifugation grad...
0 downloads 0 Views 198KB Size
Anal. Chem. 2007, 79, 2078-2083

Proteomic Complex Detection Using Sedimentation Nicholas T. Hartman, Francesca Sicilia,† Kathryn S. Lilley, and Paul Dupree*

Department of Biochemistry, University of Cambridge, Building O, Downing Site, Cambridge CB2 1QW, United Kingdom

Protein-protein interactions are important in many cellular processes, but there are still relatively few methods to screen for novel protein complexes. Here we present a quantitative proteomics technique called ProCoDeS (Proteomic Complex Detection using Sedimentation) for profiling the sedimentation of a large number of proteins through a rate zonal centrifugation gradient. Proteins in a putative complex can be identified since they sediment faster than predicted from their monomer molecular weight. Using solubilized mitochondrial membrane proteins from Arabidopsis thaliana, the relative protein abundance in fractions of a rate zonal gradient was measured with the isotopic labeling reagent ICAT and electrospray mass spectrometry. Subunits of the same protein complex had very similar gradient distribution profiles, demonstrating the reproducibility of the quantitation method. The approximate size of the unknown complex can be inferred from its sedimentation rate relative to known protein complexes. ProCoDeS will be of use in screening extracts of tissues, cells, or organelle fractions to identify specific proteins in stable complexes that can be characterized by subsequent targeted techniques such as affinity tagging. Protein-protein interactions play a fundamental role in almost every physiological process in the cell. With the identification of many proteins in a sample now routine using mass spectrometry (MS) technologies, there is an increasing need for the screening and analysis of protein complexes to understand the function and role of specific proteins in the cell. A wide range of established techniques exist for examining protein complexes, but they are often aimed at characterizing the interactions of a specific test protein. The simplest and most widely used methods involve antibody-based “pull-downs”. Antibodies specific to either the protein of interest or an attached epitope tag are used to purify specific proteins from a heterogeneous sample. Any proteins that exist in a complex with the one being purified will also be extracted from the sample and can be identified using MS techniques. More recently, tags such as the tandem affinity purification (TAP) tag have been used to increase the purity of the isolated complexes, and quantitative MS techniques, such as using isotopic labeling with isotope coded * Corresponding author. E-mail: [email protected]. Tel: +44 1223 333340. Fax: +44 1223 333345. † Current address: Dipartimento di Biologia Vegetale, Universita ` di Roma La Sapienza, 00185 Rome, Italy.

2078 Analytical Chemistry, Vol. 79, No. 5, March 1, 2007

affinity tags (ICAT), have been applied to differentiate between real and background copurified proteins.1,2 Further techniques have been developed to report the presence of an interaction between two proteins in living cells. The yeast two-hybrid system and the split-ubiquitin system allow proteins of interest that interact to give a visible or selectable signal.3,4 Fluorescence resonance energy transfer detects protein-protein interactions by imaging the energy transfer from a donor fluorophore on one protein to an acceptor fluorophore on another.5 Technologies such as TAP tagging have been utilized in a highthroughput manner in simple organisms, such as yeast, to yield large data sets on protein-protein interactions;6 however, such an approach suffers from several drawbacks. The fusion proteins required for most of these analysis methods can potentially interfere with the normal protein interactions, subcellular localization, or functions in the cell. Moreover, genetic constructs must be created for each protein, and in higher organisms, it may not be easy to express the tagged proteins at appropriate endogenous levels or in the correct cell type. In such cases, it would be useful to have evidence for a stable protein complex prior to these laborious individual protein-targeted approaches of affinity isolation. Blue native polyacrylamide gel electrophoresis (BN-PAGE) is a method that separates protein complexes, including membranebound complexes, coated with Coomassie G250 dye and in the absence of ionic detergents.7 BN-PAGE gels are also commonly arranged in a two-dimensional format where complexes are separated by native molecular weight and then individual subunits of those complexes are separated by denatured molecular weight. The resulting protein spots can be identified by mass spectrometric methods, thus making BN-PAGE an excellent way to study the composition of high-abundance protein complexes where stoichiometric protein spots can be analyzed. However, it can be misleading to assign a protein to a specific band by identification of proteins from single gel slices, because proteins often smear across large regions of the gel during BN-PAGE. Consequently, (1) Rigaut, G.; Shevchenko, A.; Rutz, B.; Wilm, M.; Mann, M.; Seraphin, B. Nat. Biotechnol. 1999, 17, 1030-1032. (2) Ranish, J. A.; Yi, E. C.; Leslie, D. M.; Purvine, S. O.; Goodlett, D. R.; Eng. J.; Aebersold, R. Nat. Genet. 2003, 33, 349-355. (3) Fields, S.; Song, O. Nature 1989, 340, 245-246. (4) Johnsson, N.; Varshavsky, A. Proc. Natl. Acad. Sci. U.S.A. 1994, 91, 1034010344. (5) Fairclough, R. H.; Cantor, C. R. Methods Enzymol. 1978, 48, 347-379. (6) Gavin, A. C.; et al. Nature 2002, 415, 141-147. (7) Schagger, H.; Cramer, W. A.; von Jagow, G. Anal. Biochem. 1994, 217, 220-230. 10.1021/ac061959t CCC: $37.00

© 2007 American Chemical Society Published on Web 01/12/2007

detection by sensitive MS methods does not provide sufficient evidence to show that the protein is most abundant at that portion of the gel and, therefore, migrating in a higher molecular weight complex. Here we present Proteomic Complex Detection using Sedimentation (ProCoDeS)sa technique for the high-throughput identification of soluble and membrane proteins in stable complexes. The relative size of the protein complexes can be estimated from their sedimentation in a rate zonal gradient (RZG).8 During RZG, a sample is layered onto a viscous buffered solution, often with an increasing density gradient. During centrifugation, the different species in the sample sediment into the gradient at varying rates. Molecular weight generally has the largest influence in determining the rate of sedimentation, although other factors such as molecular shape also play a role. The rate or resolution of the separation can be modified by changing the concentration, composition, and temperature of the sedimentation gradient or the speed or duration of centrifugation. During RZG centrifugation, monomer proteins and small complexes move more slowly into the gradient than larger protein complexes. The distribution of a protein of interest is normally visualized by western blot analysis of RZG fractions, but specific antibodies or tagged proteins are required for each protein. In this proof-of-principle experiment, ICAT technology was used for relative quantitation of proteins in RZG fractions.9 Thus, we were able to identify and profile sedimentation of over 100 proteins in parallel, allowing 30-40 proteins in stable complexes to be identified in each experiment. EXPERIMENTAL SECTION Apparatus. Centrifugation was carried out in either an Optima TLX benchtop ultracentrifuge or an Optima L-100 XP ultracentrifuge (Beckman Coulter, Fullerton, CA). All centrifuge rotors and tubes were also from Beckman Coulter. Fractions from the organelle and rate zonal gradients were collected using an Auto Densi-Flow fractionator (Labconco, Kansas City, MO). LC-MS/ MS peptide quantitation and sequencing was conducted on 6.4 µL of each sample using an Ultimate Nano-LC with a Famos autosampler and PepMapC18 (15 cm × 75 µL i.d.) analytical column (Dionex, Sunnyvale, CA) into a Q-STAR XL quadrupole time-of-flight mass spectrometer using the NanoSpray electrospray ionization source (Applied Biosystems, Foster City, CA). The mass spectrometer was operated in information-dependent acquisition mode with the two most abundant doubly or triply charged ions in each survey scan singled out for MS/MS analysis. Each of the five ICAT samples from a gradient was analyzed three times. Reagents. All stock solutions were prepared using deionized water, and HPLC solvents were prepared using G Chromasolv HPLC water (Sigma-Aldrich, St. Louis, MO). The callus homogenization buffer consisted of 250 mM sucrose, 10 mM HEPES pH 7.4, 1 mM EDTA, and 1 mM dithiothreitol and was adjusted to the appropriate concentration of iodixanol for the isopycnic centrifugation using a working solution made from a 60% stock (8) Dudkina, N. V.; Eubel, H.; Keegstra, W.; Egbert, J. B.; Braun, H. Proc. Natl. Acad. Sci. U.S.A. 2005, 102, 3225-3229. (9) Gygi, S. P.; Rist, B.; Gerber, S. A.; Turecek, F.; Gelb, M. H.; Aebersold, R. Nat. Biotechnol. 1999, 17, 994-997. (10) Prime, T. A.; Sherrier, D. J.; Mahon, P.; Packman, L. C.; Dupree, P. Electrophoresis 2000, 21, 3488-3499.

iodixanol solution (Sigma-Aldrich) in homogenization buffer. The gradient buffer for rate zonal centrifugation consisted of 100 mM Tris-Cl, 1 mM EDTA, and 1.25% β-dodecylmaltoside (β-DDM) at pH 7.5. Preparation of Mitochondria-Enriched Arabidopsis Membranes. Callus tissue cultures from Arabidopsis thaliana Col-0 were prepared and maintained as previously described.10 Cells were harvested 4 days after the culture medium was refreshed, and all steps were performed at 4 °C or on ice. Approximately 100 g of callus was used to prepare the membranes. The callus was homogenized in an equal volume of homogenization buffer, the homogenate spun 2 × 5 min at 2200g to remove cell wall components and nuclei, and the supernatant centrifuged in a SW28 rotor at 100000g onto a 6-mL 18% iodixanol (CAS 92339-11-2) cushion for 2 h at 4 °C. Concentrated membranes were collected from the top of the cushion, adjusted to 16% iodixanol, and spun for 3 h at 4 °C in a VTi65.1 rotor with OptiSeal 11.2-mL (16 × 67 mm) tubes at 350000g to separate organelles in a self-forming iodixanol density gradient.11-13 Finally, 0.5-mL fractions were harvested from the top of the gradient with fractions 15-20, enriched in mitochondrial proteins, used in this experiment. Solubilization and Rate Zonal Separation of Protein Complexes. Fractions were diluted in an equal volume of the gradient buffer, without β-DDM, and spun 15 min at 50000g at 4 °C in a TLA 100.3 rotor with 1.5-mL polyallomer tubes (Beckman-Coulter). The pellet was washed with the same buffer and spun again. The resulting membrane pellet was resuspended on ice in 650 µL of the gradient buffer with β-DDM to solubilize the membrane protein complexes. β-DDM was chosen because mitochondrial complexes are effectively solubilized and stable using this detergent.7 This suspension was then spun for 15 min at 50000g at 4 °C in the TLA 100.3 rotor and the supernatant, consisting of solubilized membrane proteins, was retained. Sucrose gradients (5-25%) were prepared in Ultra-Clear centrifuge tubes (13 × 51 mm) by layering 750 µL each of 5, 8, 11, 14, 17, and 25% sucrose in gradient buffer. Gradient layers were allowed to diffuse for 2 h at 37 °C and then prechilled to 4 °C prior to use. The 300 µL of solubilized membrane proteins solution was layered onto the sucrose gradient and spun for 6.25 h in the first experiment and 7.50 h in the second experiment using a SW50.1 rotor at 243000g at 4 °C. Finally, fractions of 300 µL were collected from the top of the gradient. Protein Concentration and ICAT Labeling. ICAT tags consist of a cysteine-reactive iodoacetamide group and an acidcleavable biotin group joined by a mass-variable linker region that exists in either a “light” or “heavy” (13C9) form. Two protein samples are labeled with the light and heavy reagents, respectively, the two samples combined and digested with trypsin, and the resulting peptides purified from the enzyme on an ion exchange column. The ICAT-labeled peptides, containing cysteine, are then isolated via purification on an avidin column. Finally, before MS analysis, the biotin group is cleaved off the tagged (11) Graham, J.; Ford, T.; Rickwood, D. Anal. Biochem. 1994, 220, 367-373. (12) Dunkley, T. P.; Watson, R.; Griffin, J. L.; Dupree, P.; Lilley, K. S. Mol. Cell. Proteomics 2004, 11, 1128-1134. (13) Dunkley, T. P.; Hester, S.; Shadforth, I. P.; Runions, J.; Weimar, T.; Hanton, S. L.; Griffin, J. L.; Bessant, C.; Brandizzi, F.; Hawes, C.; Watson, R. B.; Dupree, P.; Lilley, K. S. Proc. Natl. Acad. Sci. U.S.A. 2006, 103, 65186523.

Analytical Chemistry, Vol. 79, No. 5, March 1, 2007

2079

Figure 1. (A) General experimental setup for separating protein monomers and complexes on an RZG depicted along with the gradient fractionation strategy. The pooled fractions serve as an internal standard to allow for accurate ICAT quantitation over multiple mass spectrometry analyses. (B) The distribution of several protein standards is shown on a silver-stained SDS-PAGE gel. The “64-kDa” protein is bovine serum albumin (a monomer), the “140-kDa” protein is a subunit of the lactate dehydrogenase complex, and the “232-kDa” protein is a subunit of the catalase complex. These proteins were separated in their native state on a RZG, fractions collected from the top of the tube, and proteins denatured and finally run on an SDS-PAGE gel to analyze the distribution of each protein. (C) An outline of the quantitative proteomics strategy shows that each fraction to be analyzed is labeled with the ICAT heavy reagent and compared against the internal standard pool labeled with the ICAT light reagent. (D) An example of the ICAT ratios for a peptide in three analyzed fractions. The peaks have been scaled such that the ICAT light peak, the internal standard, is the same height in each fraction. This allows a direct comparison of the ICAT heavy peaks to visualize the distribution of the parent protein across the fractions.

proteins since its additional fragmentation would interfere with the MS-MS peptide sequencing. In each ICAT labeling pair, one-fifth of the total pool sample was labeled with the light reagent and mixed with one of the individual fractions labeled with the heavy reagent. The Cleavable ICAT Reagent from Applied Biosystems was used according to the manufacturer’s directions with the following exceptions. The prescribed labeling buffer was substituted with a modified labeling buffer consisting of 8 M urea, 1 M Tris pH 8.5, 0.1% SDS, and 2% Triton X-100 for the improved solubilization of membrane proteins, the TCEP reduction was performed for 1 h at room temperature, and the trypsin digestion (Promega, Madison, WI) was performed after dilution at least 10-fold with 25 mM Tris pH 8.5 to a volume of 2 mL to prevent urea and SDS in the modified labeling buffer inhibiting the trypsin. Since the ICAT reagent was used for quantitation, only cysteine-containing peptides exist in the final sample, thus greatly reducing the number of theoretically possible peptides per protein. The final labeled samples were resuspended in 20 µL of 0.1% formic acid for LC-MS/MS analysis. For the ProCoDeS analyses, 15 µL of each RZG fraction was set aside for silver-stained SDS-PAGE monitoring of protein distribution (Figure S-1 Supporting Information), while the re2080

Analytical Chemistry, Vol. 79, No. 5, March 1, 2007

maining amount was acetone precipitated. Based on the resulting protein banding patterns seen in the SDS-PAGE, fractions 5-14 were selected for both experiments. Samples were solubilized in 200 µL of the above modified labeling buffer and an internal standard “pool” consisting of the even (first experiment) or odd (second experiment) fractions was combined to serve as an internal standard in all the ICAT experiments. Processing of Protein Identification and Quantitation Data. Proteins were identified using Mascot v. 2.0.01 (Matrix Science, London, UK) searches against the MIPS Arabidopsis protein database (ftp://ftpmips/gsf.de/cress/arabiprot/ arabi_all_proteins_v090704.gz). The peptide cutoff score for proteins identified with two or more unique peptides was 27 for the first experiment and 22 for the second experiment and was determined by finding the score that produced a false positive rate of 100%) miscalculation of the ratio by the software. If any individual ratios were removed, the remaining ratios were reaveraged, a new standard error was calculated, and a notation was made in the data set indicating that such action was taken. In the case where a calculated ratio is based solely on a single measurement, the spectrum was manually verified or removed. RESULTS AND DISCUSSION Experimental Design To Measure the Sedimentation of Proteins. A schematic diagram of the ProCoDeS experimental strategy is shown in Figure 1. Proteins are sedimented by ultracentrifugation to separate larger structures such as protein complexes from the smaller monomers, which remain near the top of the gradient (Figure 1A,B). Relative protein quantities in the RZG fractions can be measured using ICAT labeling and LCMS/MS to determine the distribution of each polypeptide in the gradient. In preliminary experiments, we compared the RZG fractions in a pairwise manner (e.g., fraction 5 vs 7, 6 vs 8), but found that it was difficult to reconstruct a distribution profile for the protein in the gradient. Comparison of fractions against an internal pooled standard sample allows an easier and a more accurate reconstruction of polypeptide distributions in the gradient (Figure 1C,D). A pooled sample of alternate fractions ensures that the internal standard for each protein is roughly equal to the average concentration of that protein across the gradient. Therefore, the resulting ICAT measurements can include ratios both greater and less than one for each protein, thus maximizing the data collected within the accurate range for quantitation. Figure 1D shows a schematic example of ICAT ratios for an individual peptide across several fractions. Protein Identification and Quantification. An experiment was carried out to separate protein complexes from detergentsolubilized Arabidopsis mitochondrial membranes. Five fractions from a 6.25-h RZG separation were analyzed by labeling each with the cleavable ICAT heavy reagent and comparing to a pool of fractions labeled with the ICAT light reagent. Analysis by LCMS/MS identified a total of 240 unique proteins within the defined confidence limits, including 150 proteins with 2 or more unique peptides. Of the 240 identified proteins, 184 contained at least one measured ICAT ratio and 126 proteins contained enough quantitative data, 2 or more consistent ratios, to visualize their distribution in the gradient. A total of 551 ICAT ratios were measured across the 5 fractions with a mean relative standard deviation of each averaged ICAT ratio of 12.7%. Of the 126 profiled proteins, 33 migrated as larger complexes peaking in abundance in the measured gradient fractions and 93 migrated more slowly and thus are likely either monomers or in smaller complexes (Table S-1 Supporting Information). In addition to these profiles, 58 proteins were quantified with at least one ICAT ratio but with insufficient quantitative data to assign a

Table 1. Proteins Identified in Stable Complexesa accession number At4g05020 At5g08670 At2g33040 At1g51650 At5g13450 orfB atp1 orf25 At3g02090 At5g11770 At3g55410 At3g48680 At4g02580 At5g08530 At5g52840 At5g37510 nad7 At5g65750 At5g47890 At3g42050 At1g49140 At3g06310 At5g18800 At4g11150 At2g02050 At3g18410 At4g34030 At2g33210 At3g23990 At3g13930 At3g13860 At2g30970 At2g28000

protein description

notional peak

putative alternative (type II) NAD(P)H dehydrogenase NDB2 ATP synthase β subunit ATP synthase γ subunit ATP synthase  subunit ATP synthase δ subunit orfB-ATP synthase subunit ATP synthase R subunit orf25-ATP synthase subunit mitochondrial processing peptidase β subunit-related complex I subunit 2-oxoglutarate dehydrogenase, E1 subunit, putative complex I subunit complex I subunit complex I subunit complex I subunit complex I subunit complex I subunit 2-oxoglutarate dehydrogenase, E1 subunit, putative complex I subunit vacuolar ATP synthase subunit H complex I subunit (heavy) complex I subunit (heavy) complex I subunit (heavy) vacuolar ATP synthase subunit E complex I subunit (heavy) complex I subunit (heavy) methylcrotonyl-CoA carboxylase β chain, mitochondrial chaperonin (CPN60 homologue, mitochondrial) chaperonin (CPN60, mitochondrial) pyruvate dehydrogenase E2 subunit chaperonin (CPN60 homologue, mitochondrial) aspartate aminotransferase, mitochondrial (AAT1/Asp1) chaperonin (CPN60) plastid RuBisCO subunit

7.87 9.44 9.50 9.53 9.61 9.72 9.75 9.76 10.03 10.08 10.44 10.45 10.50 10.54 10.57 10.63 10.69 10.69 10.78 11.47 11.59 11.75 11.78 11.81 11.87 11.94 11.97 12.56 12.61 12.64 nd nd nd

a The 33 proteins observed by the 6.25-h RZG experiment to exist in stable protein complexes are listed along with their notional peaking points in the gradient, listed as a fraction number. The notional peaking point was calculated by fitting a Gaussian curve to three data points consisting of the ICAT ratio from the most abundant and two adjacent fractions. nd, not determined.

specific profile. Most of these exhibited ICAT ratios of >1.50 in the first measured fraction and no or little data for the remaining fractions. This suggests that the protein was most abundant in the first fraction and, thus, probably either a monomer or in a smaller complex. Another 56 proteins were identified with the Mascot search but did not contain any measured quantitative data. The experiment was repeated with a centrifugation time of 7.5 h to sediment proteins further down the centrifuge tube and bring smaller complexes into the quantitation window. The sedimentation was again measured by the ICAT quantitation profiles, with 167 proteins identified within the confidence limits. There were 380 ICAT ratios measured across the 5 fractions with the averaged ratios having a mean relative standard deviation of 15.3%. Of the 89 profiled proteins containing sufficient ICAT ratio data to visualize their sedimentation in the gradient, 38 migrated as protein complexes including 17 proteins that migrated to points beyond the analyzed fractions. The remaining 51 proteins migrated Analytical Chemistry, Vol. 79, No. 5, March 1, 2007

2081

Figure 2. (A) Gaussian fitted ICAT ratios for fractions 8, 10, and 12 plotted for the FoF1 ATP synthase subunits R, β, γ, δ, and  in addition to the two proteins of interest, orfB and orf25. (B) Gaussian fitted ICAT ratios for 13 proteins from complex I are plotted with labels for the 5 proteins that were observed to exist in the distinctly higher molecular weight version of the complex.

more slowly and thus are likely either monomers or in small complexes (Table S-2 Supporting Information). An additional 11 proteins contained at least one ICAT ratio but could not be assigned to the above groups, and another 67 proteins were identified with the Mascot search but did not contain any measured quantitative data. The results of the two experiments were similar in the quality of data and number of proteins identified. Of the 33 proteins found in complexes in the first experiment, 31 proteins were reidentified in the second. Of these, 21 had sufficient data for profiling and all were found as protein complexes, thus demonstrating the reproducibility and robustness of the technique. Varying the RZG conditions allowed separation of complexes of different sizes. Some of the proteins (e.g., At1g15120, ubiquinol cytochrome c reductase subunit) that did not sediment into the quantified fractions during the 6.25-h RZG experiment were profiled as small complexes by sedimentation with the 7.5-h RZG experiment. Similarly, many of the proteins that migrated toward the bottom of the gradient in the 6.25-h RZG experiment, migrated off the bottom of the gradient in the 7.25-h experiment. Unless otherwise stated, the remaining data and examples presented in this paper are from the 6.25-h RZG experiment. Modeling of the Sedimentation Data. As protein complexes sediment into the gradient, the initial band of material evolves into a Gaussian as a result of diffusion in the gradient media.14 We fitted a Gaussian distribution to the quantitative data for each protein observed to sediment as a complex, using GraphSight (version 2.0.1, Cradle Fields) to find its notional approximate peaking point in the gradient (Table 1). We investigated whether all the subunits of known complexes follow the same sedimentation profile, as expected for stable protein complexes. The fitted gradient distributions of subunits from the F0F1 ATP synthase complex (native mass ∼580 kDa) and also from NADH dehydrogenase (complex I, native mass ∼1000 kDa) are shown in Figure 2.15 We determined the mean and standard deviation of the notional peak in the gradient of the detected subunits in each complex. The notional peaking fraction of the five ATP synthase (14) Schuck, P. Biophys. J. 2000, 78, 1606-1619. (15) Ja¨nsch, L.; Kruft, V.; Schmitz, U. K.; Braun, H. Plant J. 1996, 9, 357-368.

2082 Analytical Chemistry, Vol. 79, No. 5, March 1, 2007

subunits R, β, γ, δ, and  was 9.57 ( 0.12. Complex I subunits sedimented in two distinct groups, suggesting the presence of two subcomplexes (see below). For the proteins in the main cluster (At5g52840, At5g47890, At5g37510, At5g08530, At5g11770, At4g02580, At3g48680 and nad7), the notional peaking fraction was 10.53 ( 0.20. The low standard deviation indicates that multiple proteins in a single complex have a similar distribution when measured by ProCoDeS and suggests that complexes differing in sedimentation by much less than a single fraction could be distinguished with this technique. Complexes in the Plant Mitochondrial Preparation. The majority of proteins identified in complexes by ProCoDeS were from the mitochondrial ATP synthase or complex I. However, further proteins known to be in complexes were identified, including the endomembrane v-ATPase subunits (previously determined complex mass ∼700 kDa)16 and the mitochondrial CPN60 chaperonin (14 subunit complex, native mass ∼800 kDa).11,17 A second mitochondrial CPN60 homologue sedimented at a similar position. A third mitochondrial CPN60 homologue and the plastidic CPN60 were also detected in complexes (Table 1). The methylcrotonyl-CoA carboxylase subunit B sedimented as a large complex, consistent with previous native gel and gel filtration data that it is in a complex of 900 kDa.18 Subunits of the large multisubunit oxo-acid dehydrogenases were also detected in complexes. Proteins that to our knowledge have not previously been shown in complexes were also identified, including putative alternative (type II) NAD(P)H dehydrogenase subunits. This demonstrates that ProCoDeS can provide data on the presence of novel protein complexes. Two proteins encoded in the mitochondrial genome, orf25 and orfB, were found to have sedimentation profiles with very similar notional peaking points (9.72 and 9.76, respectively). Two groups have recently suggested orf25 and orfB are probable subunits of (16) Kluge, C.; Seidel, T.; Bolte, S.; Sharma, S. S.; Hanitzsch, M.; SatiatJeunemaitre, B.; Roβ, J.; Sauer, M.; Golldack, D.; Dietz, K. BMC Cell Biol. 2004, 5, 29. (17) Hutchinson, E. G.; Tichelaar, W.; Hofhaus, G.; Weiss, H.; Leonard, K. R. EMBO J. 1989, 8, 1485-1490. (18) McKean, A. L.; Ke, J.; Song, J.; Che, P.; Achenbach, S.; Nikolau, B. N.; Wurtele, E. S. J. Biol. Chem. 2000, 275, 5582-5590.

ATP synthase in plants on the basis of BN-PAGE analysis.19,20 Their data showed these two proteins to be in the F0 portion of the complex with orf25 orthologous to the b subunit and orfB orthologous to the ATP8 subunit of the yeast and mammalian complexes. Their notional peaking points in the gradient were within two standard deviations from the five well-characterized ATP synthase subunits, thus providing further evidence that these novel proteins are indeed part of the ATP synthase complex. We found that the sedimentation profile of complex I protein subunits fell into two distinctly different distribution profiles. Several of the subunits had a sedimentation profile consistent with being in a slightly larger complex (Figure 2). A recent study by Heazlewood et al. demonstrated using BN-PAGE that complex I in Arabidopsis exists in two arrangements.21 A slightly larger form of the complex contained several additional small subunits. We observed five complex I subunits that exhibited a heavier sedimentation profile than the others (At1g49140, At2g02050, At3g06310, At3g18410, At5g18800). Of these, two (At1g49140, At2g02050) were found by the earlier study to exist in the heavier complex. The remaining three were not detected by the earlier study and are likely additional novel subunits of this newly observed complex. Of the seven previously described subunits that we observed to have a lighter sedimentation profile, six were found to be in the lighter complex by the previous study, with the remaining protein not detected, thus further supporting our conclusion. Since ProCoDeS detects the average distribution across the gradient, the seven proteins detected with the lighter sedimentation profile would be displaying an average distribution between the light and heavy versions of complex I. This example demonstrates the ability of the ProCoDeS technique to resolve relatively subtle differences in the subunit composition between slightly different subforms of large protein complexes. CONCLUSIONS We have presented a novel application of quantitative proteomics for the analysis of protein distribution across RZGs. ProCoDeS is able to profile the sedimentation of a large number of proteins in parallel. The gradients used in this experiment resolved complexes in the range of ∼400-1200 kDa to focus on specific mitochondrial membrane complexes; however, RZGs can be easily adjusted for sample-specific size range and resolution requirements. We were able to identify 33 proteins in stable complexes in Arabidopsis cells in a single experiment. Using ProCoDeS, we were able to corroborate recently published results concerning the subunits of Arabidopsis mitochondrial protein complexes, and we generated additional complementary data. (19) Sabar, M.; Gagliardi, D.; Balk, J.; Leaver, C. J. EMBO Rep. 2003, 4, 381386. (20) Heazlewood, J. L.; Whelan, J.; Milar, A. H. FEBS Lett. 2003, 540, 201205. (21) Heazlewood, J. L.; Howell, K. A.; Millar, A. H. Biochim. Biophys. Acta 2003, 1604, 159-169. (22) Balbo, A.; Minor, K. H.; Velikovsky, C. A.; Mariuzza, R. A.; Peterson, C. B.; Schuck, P. Proc. Natl. Acad. Sci. U.S.A. 2005, 102, 81-86. (23) Ross, L. P.; Huang, Y. N.; Marchese, J. N.; Williamson, B.; Parker, K.; Hattan, S.; Khainovski, N.; Pillai, S.; Subhakar, D.; Ddaniels, S.; Purkayastha, S.; Juhasz, P.; Martin, S.; Bartlet-Jones, M.; He, F.; Jacobson, A.; Pappin, D. J. Mol. Cell. Proteomics 2004, 3, 1154-1169. (24) Old, W. M.; Meyer-Arendt, K.; Aveline-Wolf, L.; Pierce, K. G.; Mendoza, A.; Sevinsky, J. R.; Resing, K. A.; Ahn, N. G. Mol. Cell. Proteomics 2005, 4, 1487-1502.

ProCoDeS could be adapted to a range of MS-based protein relative quantification techniques. Previous studies have used optical signals to monitor the sedimentation of up to three different protein species in a centrifugation gradient.22 However, when using quantitative proteomics technologies, the number of proteins that can be identified and profiled is only limited by the analytical capabilities of the MS-based sequencing and quantitation. Here, with quantitation using the cysteine-labeling ICAT reagent, relatively few peptides were studied for each protein. If a similar experiment were performed using iTRAQ technology, it would be expected that many more peptides per protein could be studied, increasing the confidence and coverage of the technique.23 The samples could also be analyzed using label-free approaches to quantitative proteomics.24 ProCoDeS is particularly well suited for the screening of a crude cellular extract or enriched subcellular material to identify novel proteins that exist in complexes. Since it studies native complexes, it avoids the generation and expression of fusion proteins, which require considerable experimental effort and may also lead to artifactual results. It could also be used to test conditions, such as choice of detergents, for stable protein complex solubilization. Proteins could then be selected for more conventional targeted analyses such as epitope or TAP tagging or immunoprecipitation, using the same sample extraction conditions. Additionally, changes in the organization of protein complexes in response to stress or mutations could be studied. ACKNOWLEDGMENT N.T.H. designed and performed the experiments and wrote the manuscript, F.S. assisted with the experiments, K.S.L. designed the experiments and directed the mass spectrometry, and P.D. designed the research, wrote the manuscript, and directed the research. We thank Zhinong Zhang for the preparation of the callus cultures and Svenja Hester for assistance with the mass spectrometry analyses. Research funding was provided by the BBSRC, and N.T.H. was funded through a Marshall Scholarship and a NSF Graduate Research Fellowship. SUPPORTING INFORMATION AVAILABLE Figure 1: SDS-PAGE analysis of RZG fractions. Aliquots of the fractions used for ICAT labeling were separated on an SDS-PAGE gel and silver stained to visualize the protein distribution. Table 1: Results for protein identification and quantitation from the 6.25-h RZG experiment. For the quantitation data, boldface Roman numbers indicate ratios where the constituent spectra were manually verified and boldface italic numbers indicate ratios with at least one poorly resolved spectrum removed and the ratio recalculated. Percent error is expressed as relative standard deviation. Gray shading indicates no data. Red shading indicates relative standard deviation of the measurements over 20%. ND, not determined. Table 2: Results for protein identification and quantitation from the 7.50 h RZG experiment. Formatting of this table is the same as Table 1. This material is available free of charge via the Internet at http://pubs.acs.org.

Received for review December 13, 2006.

October

18,

2006.

Accepted

AC061959T Analytical Chemistry, Vol. 79, No. 5, March 1, 2007

2083