Two-dimensional Blue Native/SDS Gel Electrophoresis of Multi-Protein Complexes from Whole Cellular Lysates

Identification and characterization of multi-protein complexes is an important step toward an integrative view of protein-protein interaction networks that determine protein function and cell behavior. The limiting factor for identifying protein complexes is the method for their separation. Blue native PAGE (BN-PAGE) permits a high-resolution separation of multi-protein complexes under native conditions. To date, BN-PAGE has only been applicable to purified material. Here, we show that dialysis permits the analysis of multi-protein complexes of whole cellular lysates by BN-PAGE. We visualized different multi-protein complexes by immunoblotting including forms of the eukaryotic proteasome. Complex dynamics after γ interferon stimulation of cells was studied, and an antibody shift assay was used to detect protein-protein interactions in BN-PAGE. Furthermore, we identified defined protein complexes of various proteins including the tumor suppressor p53 and c-Myc. Finally, we identified multi-protein complexes via mass spectrometry, showing that the method has a wide potential for functional proteomics.

In the post-genomic era, there is an increasing need to develop methods that allow the analysis of multi-protein complexes (MPCs). 1 This would provide an integrative view of the protein-protein interaction networks that define protein function and cell behavior. Methods to analyze the interactions of proteins with each other and to study MPCs in terms of their composition, dynamics, post-translational modifications, size, and abundance of their different subunits are highly desirable. So far, protein interactions in yeast have been determined at large scale using either a two-step affinity purification (1), immunoprecipitations (2) of each individual protein, or alternatively using comprehensive two-hybrid screens (3,4). These approaches allow the detection of individual protein-protein interactions (interactomics), and combination of these datasets have led to the development of a network of protein-protein interactions (5,6). The drawback is that these approaches do not study protein complexes directly (complexomics). The properties of MPCs such as their relative abundance and exact composition are in many cases still unknown.
Identification and analysis of MPCs requires their separation under native conditions. Native isoelectric focusing is one way to separate protein complexes based on their isoelectric point. The method is limited by the fact that many proteins tend to be insoluble close to their isoelectric point, hydrophobic proteins tend to precipitate at the basic pole, and the matrixes that are currently available (acrylamide or agarose) have a discrete pore size that limits the size of the complexes that can enter the gel (Ͻ1 MDa). Blue native (BN)-PAGE is a charge shift method, in which the electrophoretic mobility of a MPC is determined by the negative charge of the bound Coomassie dye and the size and shape of the complex (7)(8)(9). Coomassie does not act as a detergent and preserves the structure of MPCs. Importantly, the resolution of BN-PAGE is much higher than that of other methods such as gel filtration or sucrose-gradient ultracentrifugation (7,9). BN-PAGE was developed for the separation of membrane proteins from membrane or organelle fractions and so far has been limited to purified protein preparations. For unknown reasons, it has not been possible to separate whole-cell lysates (WCL).
Here, we show that dialysis of a WCL allows analysis by BN-PAGE, indicating that a small substance in the lysate prevents separation by BN-PAGE. We describe the combination of BN-PAGE and SDS-PAGE in a two-dimensional (2D) approach to produce a protein complex "footprint" of mammalian cells under defined conditions and the identification of MPCs using 2D BN/SDS-PAGE in combination with protein sequencing by mass spectrometry (MS).
Next, 125 l of the WCL was placed on a Microcon® YM50 or YM10 centrifugal filter column (Amicon Bioseparations; Millipore, Bedford, MA) and 375 l of BN buffer (500 mM ⑀-amino caproic acid, 20 mM Bistris, pH 7.0, 2 mM EDTA, 12 mM NaCl, and 10% glycerol) with the corresponding detergent and 1 mM phenylmethylsulfonyl fluoride. The sample was centrifuged 45 min at 15,000 ϫ g at 4°C. The flow-through was removed, and 450 l of cold BN buffer were added to the column. The procedure was repeated four times. After the final spin, the column was inverted in a microcentrifuge tube and centrifuged 1 min at 350 ϫ g to collect the sample. Different dilutions of the sample were made in BN buffer including the corresponding detergent and loaded on the BN gel.
BN-PAGE-BN gels were prepared as described (7,9,10). First, 5.5-14% or 5.5-17% gradient gels cast on either the Bio-Rad protean II minigel system or protean II xi gel system (Bio-Rad, Hercules, CA) were used. Gels were overlaid with 1ϫ gel buffer and stored at 4°C until further use. Then 40 g of ferritin (440 and 880 kDa; Sigma, St. Louis, MO) was used as a marker. No sample buffer (7,9) was added to the samples before loading. Gels were run overnight at 4°C. The entire gel run was performed with the blue cathode buffer (50 mM Tricine, 15 mM Bistris, pH 7.0) and the anode buffer (50 mM Bistris, pH 7.0), i.e. the cathode buffer was not exchanged with a colorless cathode buffer (7,9) during the run.
For the antibody-based gel shift assay, 30 l of the dialyzed WCL was mixed with 2 g of the anti-Mcp21 antibody (see below) and incubated for 30 min on ice before loading of the mixture on the first-dimension gel.
2D SDS-PAGE and Immuoblotting-For further separation in a second-dimension SDS-PAGE, the lanes from first-dimension BN-PAGE were cut out and, if necessary, frozen at Ϫ20°C. Gel strips were equilibrated for 30 min in 2ϫ SDS Laemmli loading buffer with or without reducing agents (␤-mercapthoethanol) and placed into a second-dimension SDS-PAGE of the same thickness and overlaid with 1ϫ SDS Laemmli loading buffer. The second-dimension run and immunoblotting were performed according to standard protocols.
Identification of Proteins by MS-Bands were excised from silverstained gels (14) and dehydrated twice in CH 3 CN. For reduction and alkylation, 10 mM dithiothreitol in 100 mM NH 4 HCO 3 was added and the sample incubated at 56°C for 45 min. The solution was removed and 55 mM iodoacetamide in NH 4 HCO 3 was added. Following incubation for 30 min in the dark, the sample was washed once with 100 mM NH 4 HCO 3 and once with CH 3 CN. After 15-min incubation, the gel fragments were dried and reconstituted with 10 ng/l modified trypsin (Promega, Madison, WI) in 50 mM NH 4 HCO 3 . Samples were incubated on ice for 40 min and overlaid with 50 mM NH 4 HCO 3 for overnight incubation at 37°C. After centrifugation, the supernatant was extracted once with milli-Q H 2 O and three times with 5% HCOOH in 50% CH 3 CN. All four extractions were pooled, the solution vacuum-dried, and the peptides reconstituted in 0.005% heptafluorobutyric acid, 0.4% acetic acid in H 2 O. Analysis of peptides by microelectrospray liquid chromatography (LC) tandem MS (LC-MS/MS) was performed essentially as described (15). Microelectrospray columns were packed with 200 Å, 5-m C18 beads (Michrom BioResources Inc., Auburn, CA). The flow through the column was set to a flow rate of 250 nl/min. The mobile phase used for a linear gradient elution of 15-35% B in 45 min followed by 35-70% B in 15 min consisted of (A) 0.4% acetic acid, 0.005% heptafluorobutyric acid, and 5% acetonitrile and (B) 0.4% acetic acid and 0.005% heptafluorobutyric acid in acetonitrile. Tandem mass spectra were recorded on a LCQ ion trap mass spectrometer (Thermoquest Corp., San Jose, CA) equipped with an in-house microelectrospray ionization source. Needle voltage was set at 1.6 kV. Ion signals above a predetermined threshold automatically triggered the instrument to switch from MS to MS/MS mode for generating collision-induced dissociation spectra (data-dependent MS/MS). The collision-induced dissociation spectra were searched against the IPI human protein sequence database using the computer algorithm Sequest (16). Statistical analysis of the data was performed using a combination of different software (Pep-tideProphet TM , ProteinProphet TM , and Interact) (17)(18)(19). Additional information on these software tools can be found on the Proteomics pages at www.systemsbiology.org.

2D BN/SDS-PAGE of WCL-So far, separation of WCL by
BN-PAGE has not been successful. We reasoned that it might be the presence of a small, unidentified component in the lysate that prevents separation of the sample by BN-PAGE. Our approach was to dialyze the WCL against BN buffer and to resolve it by BN-PAGE followed by a second-dimension SDS-PAGE (2D BN/SDS-PAGE, Fig. 1A). For the dialysis, spin columns with a molecular mass cut-off of 50 and 10 kDa as well as dialysis bags with a cut-off of 30 and 3.5 kDa were used. After dialysis of the WCL, we were able to effectively separate MPCs via BN-PAGE (see below). This demonstrates that an unknown cytosolic compound smaller than 3.5 kDa interferes with BN-PAGE and, when removed, BN-PAGE yields high-resolution separation of whole cellular MPCs.
In 2D BN/SDS-PAGE, monomeric proteins will migrate within the hyperbolic diagonal, whereas protein spots below the diagonal indicate their MPC nature (Fig. 1B). Proteins that are components of the same MPC will be found in one vertical line in the second dimension. Several spots of the same protein in a horizontal line are indicative of the protein being present in several distinct MPCs.

Analysis of the Proteasome Complexes and their Dynamics by 2D BN/SDS-PAGE-We dialyzed WCLs of HEK293 cells and used immunoblotting on a 2D BN/SDS-PAGE membrane
to analyze a well-characterized MPC, the eukaryotic proteasome (Fig. 2). Probing with antibodies against the ␣ subunit Mcp21 or the ␤2 subunit of the 20S proteasome (20), the PA28␣ subunit of the PA28 regulator (21), or the S4 ATPase subunit of the 19S cap (13) revealed specific protein complexes ( Fig. 2A): the 26S proteasome (20S plus 19S cap), the 20S proteasome together with PA28, the 20S proteasome alone, and the PA28 regulator alone (Fig. 2A). By treating WCL with SDS and boiling it before loading the BN-PAGE dimen- sion, the complexes are disrupted and their components migrate as monomers within the diagonal (Fig. 2B). Thus, the spots seen in Fig. 2A correspond to genuine MPCs.
We visualized MPC dynamics by stimulating HeLa cells with ␥ interferon (IFN-␥) for 3 or 5 days to induce the synthesis of subunits of the immunoproteasome such as LMP2 (22). As expected, LMP2 was only found in stimulated cells (Fig. 2C). The mature LMP2 protein was detected exclusively within the 20S core proteasome and the 20S proteasome together with PA28, whereas the LMP2 precursor was found either in the 13S pre-proteasome complex (23) or as free LMP2 precursor within the running front of BN-PAGE (Fig. 2C). Interestingly, the LMP2 precursor monomer was present only after 3 days of IFN-␥ induction, but not after 5 days. Mature LMP2 was only found assembled within the 26S proteasome after 5 days (Fig. 2C).
Detection of Protein-Protein Interactions by Antibody Shift in BN-PAGE-The position of two protein spots in one vertical line in a 2D BN/SDS gel (membrane) could be indicative of their presence in the same MPC. However, each protein could also be part of separate complexes that migrate at the same position in BN-PAGE. The unequivocal demonstration that several proteins are part of the same MPC is possible through an antibody-based gel shift assay (24). Incubation of WCL with an anti-Mcp21 antibody before BN-PAGE resulted in a shift of the Mcp21-containing MPCs toward higher molecular mass in the first dimension (i.e. to the left in the second dimension) (Fig. 2D). The shift of the ␤2 subunit and PA28 indicated that they interacted with Mcp21. The shift is complex specific, because PA28 alone and the free PA28␣ subunit did not shift.
Identification of Novel MPC by Immunoblotting-To test the versatility of our technique and to identify so far unknown protein complexes, we probed immunoblots for different intracellular proteins such as transport proteins, transcription factors, and chaperones (Fig. 3). Different detergents were used (0.1% Triton X-100, 1% saponin, or 0.5% Brij 96), and the same complexes were found, although the ratio of the complexes sometimes was different depending on the detergent used (data not shown). Importin-␣ is an adapter molecule that mediates the nuclear import of proteins carrying a classical nuclear localization signal (25) and thus is present in multiple MPCs (Fig. 3). Interestingly, importin-␤, which interacts with importin-␣/cargo complexes and transports them through the nuclear pore complex, does not show the same pattern as importin-␣, but rather forms two discrete complexes. Further analysis of components of the nuclear import machinery should yield interesting information on the composition of complexes involved in nuclear transport. The tumor suppressor p53 showed three complexes in our system (Fig. 3). The 200-kDa complex probably corresponds to the p53 tetramer (26), and the 800-kDa complex may correspond to the cytoplasmic p53-Parc complex (27). The identity of the third, very large complex (Ͼ5 MDa) is currently unknown. We do not know whether this is a defined complex or whether large p53 aggregates form during the gel run. We observed three complexes for the proto-oncogene c-Myc (Fig.  3). Astonishingly, a large part of the c-Myc protein is present in the form of a discrete complex of at least 1 MDa. The small complex at the diagonal could correspond to c-Myc monomer and/or c-Myc-Max heterodimer. Because many interactions of Myc with other proteins have been described (28), the approach presented here should be informative to determine the predominant interaction partners of Myc within the cell. Transcription factor 2b and heat shock protein 40 (Hsp40) were found to be monomers or present in very small complexes that migrated within the diagonal (Fig. 3). The chaperone Hsp60 is known to form heptameric ring-like structures of 420 kDa (29) that were also detected by our novel method (Fig. 3). Hsp70 and Hsc70 bind to various substrates as well as to cofactors. They do not form identical complexes (Fig. 3), indicating that they might bind to different substrates/cofactors. The heavy chain binding protein (BiP) is a chaperone of the endoplasmic reticulum lumen complexed to distinct substrates (30). Indeed, most, if not all, BiP was found in different MPCs of similar sizes. The BAP membrane proteins form large MPCs in the mitochondria (BAP32/37) (31) and the endoplasmic reticulum (BAP29/31) (32), respectively. The presence of these complexes was confirmed by our approach (Fig. 3). highly desirable to use WCL BN/SDS-PAGE to identify and characterize MPCs using a systematic proteomics approach. Silver staining of a 2D BN/SDS gel showed that that quite a number of spots migrate below the diagonal (white dotted lines), indicating their MPC nature (Fig. 4A). The MPCs disappeared when the WCL was treated with SDS and boiled before loading on the BN gel (Fig. 4B).
Immunoblotting of a 2D WCL BN/SDS-PAGE membrane had revealed the exact positions on which the different subunits of the proteasome migrate ( Fig. 2A). To identify the proteasome from a silver-stained 2D WCL BD/SDS-PAGE gel, we isolated the spots corresponding to the positions identified by immunoblotting (Fig. 4A, spots A and B). The proteins present in these spots were sequenced by LC-MS/MS (15). Spot A corresponded to the proteasome ␣7 subunit, and spot B contained both the proteasome ␤4 and the ␣2 subunit.
In order to facilitate sequencing, we loaded more material onto 2D WCL BN/SDS-PAGE than previously (Fig. 4A). Protein spots were picked from this new gel, labeled A-J in Fig.  4A. The identities of the proteins contained in the gel spots were determined by LC-MS/MS (Table I). Two spots in the same horizontal line (spots C and D) were identified as Hsp60. These spots correspond to the ones detected by immunoblotting (Fig. 3). The monomeric protein present in spot E (Fig.  4A) was identified as protein disulfide isomerase. Spot F contained the ovarian cancer-related tumor marker CA125 and a hypothetical protein corresponding to a seven transmembrane receptor. We identified four proteins in spot G, the most prominent being Hsp90␣, which is predominantly known to be present as a homodimer (33), thus migrating below the diagonal. In addition, spot G contained several proteins that, to date, were only known from large-scale sequencing projects (34). One of these proteins is a novel member of the G protein-coupled seven transmembrane helix receptor superfamily. Interestingly, some members of this family have been shown to form constitutive dimers/oligomers, whereas other members are present as monomers (35). Our approach shows that the new member sequenced here (Q8NH02, spot G) is most likely a constitutive dimer/oligomer, whereas the other novel member (XP_292247, spot F) was present as a monomer. The human period circadian protein 3 was also identified in spot G. It belongs to the basic helix-loop-helix family of transcription factors, containing two PAS (Per, Amt, Sim) dimerization domains, which might be the reason that we detected it as a dimer/oligomer (Fig. 4A). Spot H contained the subunit TCP-1 of the cytosolic chaperonin TriC-CCT MPC (36). This MPC contains two rings, each consisting of eight similar subunits of ϳ60 kDa each. The TriC-CCT complex identified here migrated with an apparent molecular mass of ϳ1 MDa (Fig. 4A), matching the predicted mass of the complex (960 kDa). Spots I and J were identified as the B chain of lactate dehydrogenase. This enzyme is a tetramer consisting of A and B chains (37), both of which have ϳ37 kDa. Thus, the two spots below spots I were most likely the A chain. The calculated molecular mass of the tetramer (150 kDa) corresponds to the largest of the complexes identified, suggesting that this MPC indeed is the tetrameric form of the complex. Interestingly, we detected a smaller complex that contained both subunits (probably a trimer) and a complex that only comprised the B chain (spot J, probably a homodimer). DISCUSSION We describe a new procedure that allows the identification and analysis of MPCs under native conditions from WCLs. The ability to analyze MPCs from WCLs considerably increases the usefulness of BN-PAGE as an analysis tool for functional proteomics. Complexes of known or presumed interaction partners can be studied by immunoblotting. Endogenous MPC composition and dynamics between different cell types, e.g. stimulated and nonstimulated, normal and tumor, wild-type and knockout, can be easily analyzed. BN/ SDS-PAGE combined with protein sequencing by MS allows the identification of new MPCs and the assignment of proteins to a functional context ( Fig. 4 and Table I). When compared with other methods for the identification and analysis of MPCs, such as those depending on affinity purification (1) or immunoprecipitation (2), the approach presented here has several advantages. First, it does not rely on the addition of a tag or on the binding of an antibody, which could potentially influence protein-protein interactions and thus MPC composition and recovery. Second, by WCL BN/SDS-PAGE, multiple MPCs can be analyzed on a single immunoblot, whereas purification methods necessitate a separate purification reaction for each given target protein. Third, WCL BN/SDS-PAGE can provide rapid information on the number, size, composition, and relative abundance of complexes a given protein forms. By coshifting proteins using an antibody against a defined target protein, we were able to demonstrate that proteins that migrate in one vertical line are part of the same MPC (Fig.   2D). One caveat of this approach is that the epitope recognized by the antibody could be located in the protein-protein interaction interface and therefore the complex could be destroyed or no shift will be detected. Thus, we would recommend the use of a panel of antibodies in order to validate the results.
Like other gel-based methods, our approach is limited by the detection levels of protein staining and LC-MS/MS and the availability of suitable antibodies to the proteins of interest. This approach could be combined with very sensitive protein detection methods as e.g. the BIG method (38). Sequencing MPCs directly from gel slices of the first dimension without staining could improve sensitivity. One of the goals of the Human Proteome Organization is to generate antibodies to all human proteins (39). This will greatly expand the use of 2D BN/SDS-PAGE. The novel method presented here in combination with LC-MS/MS is suitable for high-throughput screens (especially if the second dimension is omitted) to identify the MPCs of distinct cell populations and thus help us understand the function of protein complexes within living cells.