Quantitative Proteomics Analysis of Cell Cycle-regulated Golgi Disassembly and Reassembly*

During mitosis, the stacked structure of the Golgi undergoes a continuous fragmentation process. The generated mitotic fragments are evenly distributed into the daughter cells and reassembled into new Golgi stacks. This disassembly and reassembly process is critical for Golgi biogenesis during cell division, but the underlying molecular mechanism is poorly understood. In this study, we have recapitulated this process using an in vitro assay and analyzed the proteins associated with interphase and mitotic Golgi membranes using a proteomic approach. Incubation of purified rat liver Golgi membranes with mitotic HeLa cell cytosol led to fragmentation of the membranes; subsequent treatment of these membranes with interphase cytosol allowed the reassembly of the Golgi fragments into new Golgi stacks. These membranes were then used for quantitative proteomics analyses by combining the isobaric tags for relative and absolute quantification approach with OFFGEL isoelectric focusing separation and liquid chromatography-matrix assisted laser desorption ionization-tandem mass spectrometry. In three independent experiments, a total of 1,193 Golgi-associated proteins were identified and quantified. These included broad functional categories, such as Golgi structural proteins, Golgi resident enzymes, SNAREs, Rab GTPases, cargo, and cytoskeletal proteins. More importantly, the combination of the quantitative approach with Western blotting allowed us to unveil 84 proteins with significant changes in abundance under the mitotic condition compared with the interphase condition. Among these proteins, several COPI coatomer subunits (α, β, γ, and δ) are of particular interest. Altogether, this systematic quantitative proteomic study revealed candidate proteins of the molecular machinery that control the Golgi disassembly and reassembly processes in the cell cycle.

The Golgi complex is the central organelle in the secretory pathway, essential for post-translational modifications, sorting, and trafficking of newly synthesized secretory and membrane proteins and lipids in all eukaryotic cells. The unique feature of the Golgi in almost all eukaryotic cells is the densely packed stacks of the flattened cisternal membranes. Processing enzymes in the Golgi complex, including those involved in modifying bound oligosaccharides, are arranged across the stack in the cis to trans order in which they function. In animal cells, stacks are often interconnected to form a ribbon-like structure that is localized adjacent to the nucleus. Despite its complicated morphology and function, the Golgi apparatus is dynamic, capable of rapid disassembly and reassembly during mitosis or upon drug treatment. At the onset of mitosis, the characteristic stacked organization of the Golgi apparatus undergoes extensive fragmentation. The mitotic Golgi fragments generated by this disassembly process are subsequently evenly distributed to the daughter cells, where they are reassembled into new Golgi stacks during cytokinesis (1,2).
It is believed that the cell cycle-regulated Golgi disassembly and reassembly processes involve interactions between cytosolic and membrane proteins, and many studies have been performed in efforts to identify these proteins using biochemical and cell biology approaches. However, the underlying mechanism that controls the Golgi disassembly and reassembly processes is still far from clear. For example, it is generally accepted that the microtubule cytoskeleton plays a role in organizing the Golgi structure (3). The Golgi interacts with the microtubules to maintain its perinuclear localization within the cytoplasm in interphase cells (4). During mitosis, rearrangement of the cytoskeleton facilitates the dispersal of the Golgi ribbon (1,2). However, the fragmentation of the Golgi during mitosis is far more extensive than cytoskeletal rearrangement. In addition, phosphorylation of Golgi structural proteins as well as vesicle formation is involved in mitotic Golgi fragmentation (5). Recently, several labs have attempted to comprehensively identify Golgi membrane proteins using organellar proteomics, a fundamental and fast expanding research technology in proteomics and cell biology that combines biochemical fractionation and comprehensive protein identification (6 -11). However, quantitative studies documenting the comprehensive protein changes in the Golgi membrane during the cell cycle have not been reported thus far. Analysis of the dynamics of proteins in the Golgi membrane during disassembly has the advantage of identifying not only those proteins that may be involved in this process but also peripheral membrane proteins that exhibit changes, which are more likely to be specific regulators for the morphological change rather than contaminating membrane components.
Biochemical reconstitution experiments have provided powerful tools with which we can dissect biological processes. One widely used method to study the Golgi disassembly and reassembly processes during the cell cycle involves purified Golgi membranes to which mitotic or interphase cytosol is added (12)(13)(14). After incubation, the membranes are separated from the cytosol by centrifugation through a sucrose cushion and then processed for biochemical and morphological analyses. This approach has contributed to the discovery and examination of many of the currently identified proteins that mediate Golgi membrane tethering (15,16), fusion (13,(17)(18)(19)(20), and cisternal stacking (21)(22)(23). When combined with modern quantitative proteomics approaches, this in vitro reconstitution system is expected to provide a powerful tool to further dissect the molecular mechanism that regulates Golgi membrane dynamics during the cell cycle. In this study, we applied isobaric tags for relative and absolute quantification (iTRAQ) 3 and LC-MALDI-MS/MS analysis to quantify protein changes on Golgi membrane stacks during the cell cycle using membranes provided by the in vitro reconstitution system. iTRAQ is a recently developed chemical labeling reagent (24) that has quickly gained popularity in proteomics within the past few years (25,26). It has the advantage of simultaneous identification and quantitative comparison of several (up to eight) protein samples in the same experiment. This mass spectrometry-based technology allows us to quantitatively analyze proteins that are associated with the Golgi when the membranes are intact during interphase and when they are fragmented during mitosis. The ability to multiplex samples minimizes experimental variations and allows control samples to be run at the same time, resulting in the ability to detect relatively small changes in protein levels. This unbiased strategy also helps us to understand the global changes in the protein composition of the Golgi when its morphology is changed during cell division. This study revealed candidate proteins involved in the regulation of Golgi morphological changes during the cell cycle.
Cell-free Golgi Disassembly and Reassembly Assay-Golgi membranes were purified from rat liver as described previously (28); interphase (IC) and mitotic (MC) cytosols were prepared from HeLa S3 cells as described in the supplemental "Experimental Procedures" (14). Cytosols were changed into related buffers using Bio-Spin 6 columns and cleared by centrifugation before they were used to treat Golgi membranes. The Golgi disassembly assay was performed as described previously (5). In brief, purified Golgi membranes (200 g) were mixed with 10 mg of mitotic cytosol (the MC:Golgi ratio was based on the quantitation results of these two components in rat liver cells), 1 mM GTP and an ATP-regenerating system (10 mM creatine phosphate, 0.1 mM ATP, 20 g/ml creatine kinase, and 20 g/ml cytochalasin B) in MEB buffer (50 mM Tris-HCl, pH 7.4, 0.2 M sucrose, 50 mM KCl, 20 mM ␤-glycerophosphate, 15 mM EGTA, 10 mM MgCl 2 , 2 mM ATP, 1 mM GTP, 1 mM glutathione, and protease inhibitors), with a final volume of 1 ml. After incubation for 60 min at 37°C, mitotic Golgi fragments were isolated, and soluble proteins were removed by centrifugation (55,000 rpm for 30 min in a TLA55 rotor) through a 0.4 M sucrose cushion in KHM buffer (20 mM Hepes-KOH, pH 7.0, 0.2 M sucrose, 60 mM KCl, 5 mM Mg(OAc) 2 , 2 mM ATP, 1 mM GTP, 1 mM glutathione, and protease inhibitors) onto a 6 l 2 M sucrose cushion. The membranes were resuspended in KHM buffer, and aliquots were succeeded either to fixation and processing for electron microscopy (EM) (5,20,23) or to reassembly reactions described below or kept frozen at Ϫ80°C until further use for the proteomics analysis. Incubation of Golgi membrane with interphase cytosol in KHM buffer was used as a control. For Golgi reassembly, 200 g of mitotic Golgi fragments were resuspended in 4 mg of IC in KHM buffer in the presence of an ATP regeneration system in a final volume of 300 l and incubated at 37°C for 60 min. Membranes were pelleted by centrifugation as described above, processed for EM, or kept at Ϫ80°C for the proteomic analysis or Western blotting. All results were confirmed by three or more independent experiments.
For EM analysis, the percentage of membranes in cisternae or in vesicles was determined by the intersection method (12,23). Cisternae were defined as long membrane profiles with a length greater than four times their width, the later being not Ͼ60 nm. Normal cisterna ranges were 20 -30 nm in width and longer than 200 nm. Stacks were defined as two or more cisternae that were separated by no more than 15 nm and overlapped in parallel by Ͼ50% of their length.
In-solution Digestion and iTRAQ TM Labeling-For quantitative proteomics analysis, membranes were collected under four experimental conditions: 1) incubation with buffer alone; 2) with interphase cytosol alone; 3) with mitotic cytosol alone; and 4) sequential incubation with mitotic and interphase cytosols. After incubation, Golgi membrane was separated from cytosol by centrifugation through a sucrose cushion, and the Golgi membrane pellets were stored at Ϫ80°C until use.
For in-solution digestion, Golgi membrane pellets were solubilized on ice in 50 l buffer containing 0.5 M triethyl ammonium bicarbonate, 8 M urea and 0.4% SDS. Protein concentrations were determined with the Bio-Rad Bradford assay kit. Golgi membrane proteins (30 g) from each of the previously mentioned four groups were reduced with 5 mM TCEP (tris(2carboxyethyl)phosphine) for 1 h at 37°C. Cysteines were then blocked with 10 mM methyl methanethiosulfonate for 20 min at room temperature. The protein solution was diluted four times with 0.5 M triethyl ammonium bicarbonate containing 3 g trypsin (1:10 w/w) and incubated at 37°C overnight. To quantitatively compare Golgi membrane treated with mitotic cytosol to those treated with interphase cytosol, multiplexed isobaric tags (iTRAQ TM reagents) were used to label tryptic peptides from different conditions: 115 for interphase cytosol treatment, 116 for mitotic cytosol treatment, and 117 for sequential mitotic and interphase cytosol treatments. In some experiments, a 114 tag was also used to label Golgi membranes treated with buffer alone. The labeling procedure was performed according to the protocol provided by the manufacturer. The reaction was stopped by diluting the mixture with 10 volumes of H 2 O.
OFFGEL Isoelectric Focusing and LC-MALDI-MS/MS-The combined peptide mixture was first fractionated based on isoelectric points (pIs) on a 3100 OFFGEL Fractionator (Agilent Technologies) using an immobilized pH gradient strip pH 3-10 (GE Healthcare) with a 12-well manifold according to the manufacturer's instructions. The iTRAQ-labeled peptides were diluted in the peptide-focusing buffer to a final volume of 1.8 ml, 150 l of which was loaded into each well. The samples were focused with a maximum current of 50 A until 50 kilovolt hours were achieved. Peptide fractions were harvested and desalted using Pierce C-18 spin columns, and the resulting elute was dried using a vacuum centrifuge and resuspended in 40 l 0.1% trifluoroacetic acid. The resuspended peptide sample from each OFFGEL fraction was concentrated on a reversedphase cartridge (Zorbax C-18; 5 mm ϫ 0.3 mm inner diameter; 5 m particles; Agilent) and then separated by a reversed-phase column (Zorbax 300 SB C-18 column, 75 m ϫ 150 mm, 3.5 m particles) on an Agilent 1100 high performance liquid chromatography system with the following binary gradient (flow rate: 300 nl/min): 0 min, 6.5% B; 9 min, 6.5% B; 12 min, 15% B; 92 min, 45% B; 97 min, 60% B; 102 min, 100% B; 104 min, 100% B; 105 min, 6.5% B; 115 min, 6.5% B. Solvent A is 0.1% trifluoroacetic acid, and solvent B is 90% acetonitrile and 0.1% trifluoroacetic acid. The column effluent was mixed with MALDI matrix (2 mg/ml ␣-cyano-4-hydroxycinnamic acid) through a 25-nl mixing tee and spotted on 1,536 well OptiTOF TM MALDI target plates, which were later analyzed by tandem mass spectrometry. The MALDI matrix was delivered to the mixing tee by external infusion pump (Harvard Apparatus, Holliston, MA) at 800 nl/min.
The MS and MS/MS spectra were acquired on an Applied Biosystems 4800 Proteomics Analyzer (tandem time-of-flight) (Applied Biosystems/MDX Sciex, Foster City, CA) in positive ion reflection mode with a 200 Hz Nd:YAG laser operating at 355 nm. The accelerating voltage was 20 kV with a 400-ns delay. For MS/MS spectra, the collision energy was 2 keV, and the collision gas was air. Each MALDI plate was calibrated on nine calibration wells using standards from Applied Biosystems with a 20 ppm mass accuracy in the MS mode. Both MS and MS/MS data were then acquired in the sample wells using the instrument default calibration. Typical MS spectra were obtained with the minimum possible laser energy to maintain the best resolution. Single-stage MS spectra for the entire samples were first collected, and in each sample well MS/MS spectra were acquired from the 12 most intense peaks above the signal to noise ratio threshold of 30.
Data Base Search and Bioinformatics Analysis-Data base searching was performed using ABI's ProteinPilot TM (version 3.0) using the Paragon algorithm (29). This software interacts directly with the Oracle database, in which the mass spectrometer stores its data and submits monoisotopic peak lists in batch to a local instance of the search engine for protein identities. No additional peak list filtering was specified. Peak lists were generated by the mass spectrometer during data acquisition based on a specified signal to noise threshold (30 in this case). ProteinPilot TM also performed protein grouping to remove redundant hits and comparative quantifications using iTRAQ ratios. In an effort to comprehensively identify both rat proteins (originated from the Golgi membranes) and human cytosolic proteins (recruited from HeLa cell cytosol), a combined human and rat data base was created by manually concatenating International Protein Index rat (version 3.49, with 40,131 of Rattus norvegicus proteins) and International Protein Index human databases (version 3.53, with 73,748 of Homo sapiens proteins). All data sets were then searched against this data base in Pro-teinPilot. All reported proteins were identified with 90% or greater confidence as determined by ProteinPilot TM unused scores (Ն1.0) with the corresponding false positive discovery rate below 1%. The Paragon TM algorithm in ProteinPilot software was used as the default search program with iTRAQ-labeled peptide as the sample type, trypsin as the digestion agent, methyl methanethiosulfonate for cysteine modification, and 4800 TOF/TOF as the instrument. The protein and peptide summary results obtained from ProteinPilot version 3.0 software were exported to Microsoft Excel. The iTRAQ peak area data were normalized for loading error by biased corrections calculated using ProteinPilot TM . The default ion intensity threshold in ProteinPilot TM for calculating peptide ratios was 40 counts.
Hierarchical clustering was performed based on Pearson correlation using the Cluster 3.0 program, and clusters were viewed with the JAVA TREEVIEW program. The theoretical pIs of the peptides in each OFFGEL fraction were calculated using the "Compute pI/Mw" tool accessible on the ExPASy Web site. The identified peptides with a percentage confidence Ͼ90% were included in the peptide list for calculation.

RESULTS
Applying iTRAQ-based Quantitative Proteomics to Cell-free Golgi Disassembly and Reassembly Assay-In this study, large amounts of stacked Golgi membranes were isolated from rat liver as described previously (28). These membranes were routinely purified 80 -100-fold (97.4 Ϯ 4.2 from 10 preparations) over the homogenate, examined by measuring ␤-l,4-galactosyltransferase activity. When analyzed by EM, the membranes were exhibited as stacked structures (Fig. 1A), and Ͼ60% of the cisternal membrane profiles remained in stacks, with the rest in vesicles and tubular structures (Fig. 1, A and E). As demonstrated by EM, preparations of Golgi stacks were contaminated with a minimal amount of single membrane sheets, presumably plasma membrane or endoplasmic reticulum; no mitochondria or nuclei contamination has been observed in the Golgi preparations. To mimic the interphase and mitotic conditions in the Golgi membranes, we used an established cell-free system (5) to mimic the Golgi disassembly and reassembly process during the cell cycle. In this in vitro assay, the isolated Golgi membranes were treated with interphase or mitotic HeLa cell cytosol. Interphase cytosol was prepared from large scale cultures of HeLa S3 cells cultured in suspension. Mitotic cytosol was prepared from spinner cells that had been synchronized by nocodazole treatment for 20 -24 h, which arrested 96 -98% of the cells in prometaphase (for procedure see supplemental "Experimental Procedures").
When Golgi stacks were incubated with mitotic cytosol for 60 min at 37°C in the presence of ATP and an ATP regenerating system, the morphology of the membranes underwent dramatic changes. They were transformed into vesicles and short tubules (Fig. 1, C and E), similar to that reported for mitotic Golgi profiles in HeLa cells (30). Incubation with buffer alone, or with interphase cytosol (Fig. 1B), did not result in dramatic morphological change, as most of the membranes remained in cisternae (Fig. 1E) and most of which were found in stacks (Fig.  1F). To reconstitute the reassembly of fragmented Golgi membrane, mitotic Golgi fragments (Fig. 1C) were isolated by centrifugation through a sucrose cushion and reincubated with interphase cytosol in the presence of ATP, GTP, and an ATPregenerating system. Golgi stacks reassembled after a 60-min incubation at 37°C (see Fig. 1D); most of the membranes were fused to form new cisternae Fig. 1E), which subsequently generated stacked structures (Fig. 1F). This system allowed us to prepare large amounts of Golgi membranes of interphase and mitotic conditions with high purity. Because treatment of rat Golgi membranes with human cell cytosol faithfully mimics the morphological changes of the Golgi in the cell cycle, this experiment showed that the underlying mechanism in rat and human is conserved. More importantly, it also suggested that in this system, rat and human proteins are interchangeable. Thus, by quantifying the amount of rat (from the Golgi membranes) and human (from HeLa cell cytosol) proteins associated with the Golgi membranes from different reactions, we could identify candidate proteins involved in the morphological change of the Golgi. In addition, the reversibility of the disassembly and reassembly process also allowed us to apply proteomic techniques to identify candidate proteins whose association with the Golgi membranes is regulated during the cell cycle.
Taking advantage of this established in vitro reconstitution approach, we analyzed the global changes of proteins that were associated with interphase or mitotic Golgi membranes in order to delineate the underlying mechanism for the dramatic morphological changes of the Golgi structure during the cell cycle. For this purpose, equal amounts of isolated Golgi membranes were treated separately in four different conditions as shown in Fig. 1, and reisolated by centrifugation through a sucrose cushion to remove the cytosolic proteins. The membranes were then solubilized and digested with trypsin, and the resulting tryptic peptides were labeled with iTRAQ reagents (as described in the "Experimental Procedures"), followed by OFFGEL isoelectric focusing and LC-MALDI-MS/MS analysis. This allowed us to comprehensively quantify the protein changes in the Golgi membranes under different conditions. In this study, a total of three independent experiments were performed, and the flow chart of these experiments is illustrated in Fig. 2. In these experiments, Golgi membranes treated with buffer alone were also labeled with the iTRAQ reagent 114 as negative controls. However, since the membranes in this reaction were not treated with cytosol, there were few cytosolic proteins associated with the membranes. On the contrast, a number of cytosolic proteins ubiquitously associated with Golgi membranes when treated with either interphase or mitotic cytosol. For this reason, Golgi membranes treated with interphase cytosol were used as controls in all experiments to better uncover mitosis specific Golgi protein changes.
To quantitatively measure the amount of protein in the membranes, considerations were taken into the experimental design to reduce variations. First, we conducted duplicate biological experiments that separated the peptide mixture in the first dimension into eight fractions based on their pIs using an OFFGEL isoelectric focusing system. This allowed us to assess the reproducibility of the experimental results. Next, to confirm the results from the duplicate experiments and also extend our findings, we conducted a third experiment identical to the duplicates, except that there were 12 fractions of peptides collected from the first dimensional OFFGEL separation. Using the third experiment as an example, we analyzed the characteristics of peptide separation by OFFGEL isoelectric focusing. It was found that the numbers of total and unique peptides identified from each OFFGEL fraction were similar except in the middle fractions, 6 and 7 (Fig. 3A). To evaluate the efficient separation of peptides by their pIs, the peptides identified in each OFFGEL fraction were first filtered by their confidence scores to exclude any hit below 90% confidence, after which the theoretical pIs were calculated using an on-line software tool, Compute pI/MW. As a representative, the results from the third experiment are shown in Fig. 3B. As expected, the calculated pIs increased in a nearly linear fashion from the acidic end   MARCH 5, 2010 • VOLUME 285 • NUMBER 10 JOURNAL OF BIOLOGICAL CHEMISTRY 7201 (fraction 1) to the basic end (fraction 12) with the middle four fractions (5 to 8) at a slower slope.

Comparative Analysis of Golgi Disassembly and Reassembly
Proteomic Survey of an Enriched Stacked Golgi Membrane Fraction-When the peptides were analyzed by LC-MS/MS, three data sets from three independent biological experiments were collected. Over 14,000 MS/MS spectra were acquired and searched in a combined International Protein Index rat and International Protein Index human database using the ProteinPilot TM software. Protein identifications required an unused score of 1.0 or greater corresponding to a 90% or more confidence and Ͻ1% false discovery rate. 60 -70% of these spectra were identified with Ͼ90% confidence. The numbers of proteins identified in each data set and common between different data sets are reported in supplemental Fig. S1. The protein grouping function in ProteinPilot TM and the combined data base not only allowed us to comprehensively identify both rat and human proteins but also to distinguish between them when there are unique peptide sequences detected for rat or human. If the identified peptides for a given gene name matched sequences identical between the corresponding human and rat protein, this protein was then assigned as a "common" protein (supplemental Table 1). Using these criteria, a total of 1,193 unique proteins were identified in three data sets (supplemental Table  2), 632 were assigned as rat proteins, 334 were assigned as human proteins, and 227 were assigned as common proteins. These proteins were then categorized based on their annotated subcellular localizations in the data bases and literature (supplemental Table 2) and an overview is shown in Fig. 4A. Half (50%, 606 proteins) of these proteins are known or candidate bona fide Golgi (30%) or cargo proteins (18%) or proteins in transit to other subcellular organelles such as the lysosome (2%). The second largest category includes cytosolic proteins (29%), followed by cytoskeletal proteins (8%). Because isolated Golgi membrane were incubated with excess cytosolic proteins as experimentally designed, it was expected that a significant number of cytosolic proteins would be present in the Golgi sample mainly through protein interactions with Golgi resident proteins as well as certain degree of nonspecific adsorption. Ubiquitin-conjugating proteins (Fig. 4B, 52 proteins, 4% of total) have been related to protein sorting and trafficking (31)(32)(33) and postmitotic Golgi membrane reassembly (19,34). These cytosolic proteins may be involved in regulation of Golgi membrane dynamics during the cell cycle. The rest of the proteins are those with unknown identity or with uncharacterized subcellular localizations (3%) or those likely to be contaminants from the ER (2%), mitochondria (1%), and nucleus (7%).
Quantitative Proteomics Analysis of Protein Changes in Mitotic Golgi Disassembly and Reassembly-In all three data sets, the iTRAQ ratios of identified proteins were normalized using the automatic bias correction function in the ProteinPilot TM software to compensate for the minor differences during sample mixing. When necessary, the data between the different cytosol reactions were further adjusted according to the mean ratios of a group of Golgi enzymes that are expected to remain unchanged during the "Experimental Procedures." The overall quality of the quantification and normalization is evaluated with the histogram of iTRAQ ratio distribution from identified proteins above the confidence threshold, 90%. Fig. 5 shows the histograms of the protein ratios for MC versus IC (116:115) (Fig.  5A) and MC 3 IC versus IC treatment (117:115) (Fig. 5B). These histograms indicate normal distributions centered ϳ1.0 (mean Ϯ S.D. are 0.97 Ϯ 0.33 and 0.99 Ϯ 0.37, respectively). The protein ratio histograms exhibit quite broad distributions (0.2-2.2 or more, which are relatively large deviations) indicating the presence of differentially regulated proteins. Supplemental Table 3 shows protein changes from 84 unique proteins detected in three experiments that showed significant changes (MC versus IC) with a p value Յ0.05 for statistical significance and a threshold of 20% (mean ratios Ͼ1.20 or Ͻ0.80) for biological significance. This cut- off value is supported by our experience in a large number of projects where changes with 20% threshold and a p value Յ0.05 have been validated by Western blots (37).
We next applied the hierarchical clustering method, a commonly used method for microarray analysis, to the proteomics results of the differentially regulated proteins (supplemental Table 3). The iTRAQ ratio of each protein was used to perform the cluster analysis using the Pearson correlation, and the results were visualized using Java TreeView (Fig. 6A). It became obvious that different clusters of proteins are present. Some proteins were up-regulated with mitotic cytosol treatment (Fig.  6A, bright yellow squares in the middle lane), whereas other proteins were down-regulated with mitotic cytosol treatment, and the changes were then reversed under Golgi reassembly conditions (Fig. 6A, dark squares in the middle lanes and lighter squares in the third lanes). As an indication of efficient clustering, the different subunits of the same protein complexes (e.g. chaperonin-containing T-complex subunits, Ccts and COPI coatomers, Cops) were clustered together.
Closer analysis of selected clusters ( Fig. 6B; Table 1) revealed proteins clustered together in several major patterns. One group of proteins including actin (Actg1), Rab11a, and subunits of the Tcp1 (chaperonin-containing T-complex and Ccts) showed a significant increase in association with Golgi upon MC treatment and then returned to the control level in the cases of actin and Rab11a or below the control level in the cases of T-complex subunits (Fig. 6B) after subsequent IC treatment. In contrast, a cluster containing COPI coatomer subunits (Cops), the heat shock protein Hsc70 (Hspa8), and tubulin sub-units (Tubs) showed a dramatic decrease upon mitotic cytosol treatment then returned back to control level (Fig. 6B). It has been shown that COPI vesicle formation is involved in mitotic Golgi fragmentation (5,20); whether other proteins mentioned above are part of this machinery requires further investigation. Several golgins (e.g. Golgas and Golphs) and GRASP (Gorasp) proteins such as GRASP65 have been shown to be involved in Golgi structure formation, and phosphorylation of these proteins is required for Golgi membrane unstacking during mitosis (1,23). However, their total abundance did not appear to have any dramatic change (Table 2). In addition, the Golgi structural proteins, enzymes, and SNARE proteins (e.g. Golgi SNAP receptor complex member 1 (Gosr1) and syntaxin 5a (Stx5a)), as expected, maintained the same level during the experimental procedures ( Fig. 6D; Table 2).
To validate the iTRAQ-based results, we conducted Western blot analysis on a large number of proteins, including many in the major clusters shown above (Fig. 7). We first compared the protein levels in the two cytosols. Most proteins, including ARF and coat proteins, Rabs and tethering complexes, some kinases, heat shock proteins, actin, and tubulin, have a comparable amount in both interphase and mitotic cytosols (Fig. 7A). These results indicate that the change of Golgi-associated proteins was not due to the availabilities of the proteins in interphase or mitotic cytosol. We then compared the proteins associated with either interphase or mitotic Golgi membranes (Fig. 7B). As expected, Golgi resident proteins such as TGN38 and enzymes such as ␣-mannosidase II had no change across all conditions. Consistent with our iTRAQ results, the total amount of proteins such as GM130, golgin 84, and GRASP65 did not change, although GRASP55 (indicated by the band shift) and GRASP65 (recognized by the phospho-specific antibody, pGRASP65) were highly phosphorylated (Fig. 7B). Previous studies demonstrated that these proteins are modified during mitosis, but they do not detach from the membranes, nor are they degraded during mitosis (22,23). SNARE proteins and cargos (e.g. rat serum albumin) were among the unchanged proteins as indicated by the iTRAQ results (Fig. 6D). The unique patterns of changes of COPI coatomer subunits were confirmed in the Western blot results (Fig. 7B) with antibodies against five of all six subunits identified by iTRAQ. COPI vesicle budding has been shown to be part of the process to sequester the Golgi membranes into vesicles during mitosis (5). The reduction of coat proteins on the membranes is likely due to uncoating after the generation of the vesicles. The different patterns of the heat shock proteins and the cytoskeleton proteins actin and tubulin were also confirmed (Fig. 7B). Altogether, the comprehensive Western analysis results are consistent with the iTRAQ-based quantitative proteomics results and provided strong support to our findings in this study.

DISCUSSION
Organellar proteomics combines subcellular fractionation with mass spectrometry and provides a powerful approach to identifying the protein compositions of each subcellular compartment. Because of the central role of the Golgi apparatus in cell biology and physiology, it has been a very active research subject of modern organellar proteomics. However, due to some limitations in gel-based protein separation techniques and relatively lower sensitivity of mass spectrometers in the past, early studies of Golgi proteomes reported a relatively small number of identified proteins. In an early twodimensional gel study (10), 73 proteins were identified from isolated rat liver Golgi stacks. In another study, the protein compositions of Golgi fractions isolated from basal and maximal functional states of the mammary glands were compared using two-dimensional gel electrophoresis, and 45 proteins were identified (8). A subsequent mass spectrometry analysis using one-dimensional SDS-PAGE led to the identification of 81 proteins from a purified hepatic Golgi apparatus (6). Recent progress in proteomic technology has enabledmorecomprehensivehighthroughput profiling strategies of enriched Golgi fractions, resulting in hundreds of protein identifications. In a shotgun proteomics study, 421 proteins were characterized in Golgi fractions isolated from rat liver (7). In a more recent study using the redundant peptide-counting approach, a quantitative map of the rough ER, smooth ER, and Golgi apparatus isolated from rat liver homogenates was conducted (38). In this study, 193 proteins unique to Golgi were reported together, with another 405 proteins found in both the ER and the Golgi. For the 193 proteins found only in Golgi/COPI vesicles, terminal sugar transferases, CASP, golgin, SNAREs, and a subset of Rabs and COGs (the conserved oligomeric Golgi complex proteins) were prominent (38).
The overall goal of the current study was to identify proteins associated with Golgi membranes in different phases of the cell cycle and therefore to unveil the molecular mechanisms underlying the morphological change of Golgi membranes. For this purpose, proteins in isolated Golgi membranes treated with mitotic cytosol were comprehensively compared with those treated with interphase cytosol using the iTRAQ-labeling technique combined with LC-MALDI-MS/ MS analysis. This allowed us to identify and quantitatively analyze a total of 1,193 of Golgi membrane and associated proteins. The identified known and potential Golgi intrinsic and associated proteins include the following major functional categories: 42 Golgi structural proteins, such as GRASP55, GRASP65, and GM130, that are involved in Golgi structure formation; 50 Golgi resident enzymes involved in various steps of glycoprotein modifications such as mannosidases I and II; seven SNARE proteins for membrane fusion; 25 Rab GTPases for vesicle tethering and 36 COPI and COPII vesicle coat protein subunits. More importantly, the combination of the quantitative approach with Western blotting allowed us to unveil 84 proteins with significant changes in abundance under mitotic condition compared with interphase condition (Figs. 6 and 7 and supplemental Table 3). The specificities of these protein changes to Golgi membrane disassembly process in the cell cycle were further sup-FIGURE 6. Hierarchical clustering of Golgi membrane and associated proteins that showed significant changes upon treatment with mitotic cytosol. Hierarchical clustering using Pearson correlation was performed on the proteins in supplemental Table 3 to generate the heat map. As an overview, part of the clustering result is shown in A. Two representative clusters showed unique patterns were selected and the averaged fold changes (MC and MC 3 IC versus IC) from three separate experiments were plotted for each protein (B and C). In contrast to the proteins showed significant changes, many Golgi structural proteins showed little change. Several such proteins were selected as representatives for a very distinct cluster, and their average ratios under the three indicated conditions were also plotted (D).
ported by the following evidence: 1) many of these protein changes were detected in more than one experiment; 2) multiple or all subunits of several protein complexes were found changed in the same direction and to a similar degree; and 3) the protein changes under mitotic condition were reversed upon a subsequent interphase cytosol treatment, which is consistent with the reversible morphological changes under these conditions. Among these proteins, several COPI coatomer subunits (e.g. ␣, ␤, ␥, and ␦) are of particular interest. COPI vesicle formation in interphase cells is believed to mediate intra-Golgi transport and retrograde transport for the Golgi to the endoplasmic reticulum-Golgi intermediate compartment (39). In interphase cells, COPI coat proteins are enriched on Golgi membranes. COPI vesicle formation is also thought to be part of the mitotic fragmentation process to sequester the Golgi membranes into vesicles (5). As the COPI coat is not a stable structure, it falls off the vesicles for preparation of subsequent membrane fusion. As shown in Fig. 7B, the membrane-associated COPI coat proteins were reduced after The iTRAQ were used to label tryptic peptides: 115 for interphase cytosol treatment, 116 for mitotic cytosol treatment, and 117 for sequential mitotic and interphase cytosol treatments. Three independent biological experiments were performed and indicated in the "Exp." column. In ProteinPilot software, the true ratio for the protein is expected to be found between mean ϫ EF (error factor) and mean Ϭ EF with 95% confidence. The unused ProtScore is a measurement of all the peptide evidence for a protein that is not better explained by a higher ranking protein. ProtScore ϭ Ϫlog (1 Ϫ % confidence); therefore, ProtScore 1.0 corresponds to 90% confidence, and 2.0 corresponds to 99% confidence. Based on the protein grouping results in ProteinPilot, the number of peptides unique to rat or human protein sequences or common between rat and human protein sequences is listed for each protein in the "Unique peptides " and "Common peptides" columns.  were only peptides unique to rat sequences detected, all of these proteins are assigned to rat origin. Therefore, only the numbers of peptides unique to rat protein sequences are listed in this column. b Uso1 showed decrease in both experiments. This was also confirmed by Western blotting (Fig. 7B). mitotic cytosol treatment and returned to control level after subsequent interphase cytosol treatment.
The signaling protein Rab11a is a small GTPase associated with the trans Golgi network (TGN) and endosomes and is thought to be involved in endosome to TGN recycling (40 -42). Recent studies showed that Rab11 is required for the formation of the cleavage furrow, or contractile ring, for separation of the two daughter cells in cytokinesis (43)(44)(45)(46). Although it has been thought that the membrane is likely from the endosomes, as Rab11 is located on both endosome and TGN membranes, it is unclear whether the Golgi can also provide the required membranes for cytokinesis. Our results showed that the association of Rab11a is increased during mitosis (Fig. 6B), demonstrating that Golgi membrane could function as the donor membrane for the formation of the cleavage furrow.
Another interesting group of proteins are the cytoskeletal proteins. For example, the amount of membrane-associated actin ␥ 1 (Actg1) was dramatically increased upon mitotic cytosol treatment compared with control interphase cytosol treatment, and this amount was reduced when the mitotic Golgi fragments were further treated with the interphase cytosol to allow Golgi assembly to occur (Figs. 6B and 7B). In contrast, the microtubule components, such as tubulin isoforms ␣ 1a (Tuba1a), 1b (Tuba1b), 3a (Tuba3a) and ␤5 (Tubb5), were decreased when the Golgi membranes were fragmented upon mitotic cytosol treatment and were restored when further treated with interphase cytosol (Figs. 6C and 7B). The formation and positioning of the Golgi ribbon-like structure is thought to be dependent on microtubule organization and the centrosome (47)(48)(49)(50). The centrosome nucleates polarized microtubule assembly and creates a radial array of microtubules with minus ends embedded in the centrosome and plus ends extended in all directions toward the cell periphery. Golgi membrane-associated minus enddirected microtubule motors, mainly dynein, move the Golgi stacks inward and cause concentration of the Golgi stacks around the centrosome. ER-derived vesicles track along microtubules to allow collection of the cargo proteins in the Golgi apparatus positioned at the cell center and then move back toward the cell periphery along microtubules to various destinations including the plasma membrane. Thus, the ribbon formation and localization of the Golgi may facilitate protein transport in mammalian cells. Our results demonstrated that the interaction between the Golgi membranes and the microtubule cytoskeleton is reduced during mitosis, suggesting that dissociation of Golgi membranes from the microtubule cytoskeleton may be part of the mechanism underlying Golgi fragmentation at the onset of mitosis. On the other hand, the role for actin microfilaments in Golgi organization is so far unclear, although it has been indicated that actin plays a role in ER to Golgi transport (51). Our results suggest that the actin cytoskeleton is selectively associated with mitotic Golgi vesicles instead of intact Golgi cisternal membranes, a feature important for vesicle transport. These possibilities need further investigation. Taken together, this quantitative proteomics study has provided a near complete accounting of proteins in the Golgi apparatus of rat liver under both interphase and mitotic conditions. The data provide an important foundation for the resolution of long standing problems regarding the secretory pathway and biogenesis of this membrane organelle. The Golgi enzymes and SNARE proteins were not changed, whereas some of the Golgi structural proteins (e.g. GRASP65, indicated by the phospho-specific antibody, pGRASP65; and GRASP55, indicated by the band shift) were phosphorylated. The membrane association of COP proteins and some other cytosolic proteins changed after subsequent treatments. RSA, rat serum albumin.