Two-dimensional Blue Native/SDS Gel Electrophoresis of Multiprotein Complexes from Helicobacter pylori*

The study of protein interactions constitutes an important domain to understand the physiology and pathogenesis of microorganisms. The two-dimensional blue native/SDS-PAGE was initially reported to analyze membrane protein complexes. In this study, both cytoplasmic and membrane complexes of a bacterium, the strain J99 of the gastric pathogen Helicobacter pylori, were analyzed by this method. It was possible to identify 34 different proteins grouped in 13 multiprotein complexes, 11 from the cytoplasm and two from the membrane, either previously reported partially or totally in the literature. Besides complexes involved in H. pylori physiology, this method allowed the description of interactions involving known pathogenic factors such as (i) urease with the heat shock protein GroEL or with the putative ketol-acid reductoisomerase IlvC and (ii) the cag pathogenicity island CagA protein with the DNA gyrase GyrA as well as insight on the partners of TsaA, a peroxide reductase/stress-dependent molecular chaperone. The two-dimensional blue native/SDS-PAGE combined with mass spectrometry is a potential tool to study the differences in complexes isolated in various situations and also to study the interactions between bacterial and eucaryotic cell proteins.

Helicobacter pylori is a spiral, microaerophilic, Gram-negative bacterium that colonizes the gastric epithelium (1) in 40 -60% of the world's population. Approximately 10 -20% of these infected individuals suffer from diseases, such as peptic ulcer disease, or from conditions such as chronic atrophic gastritis, which can then evolve toward gastric cancer (2)(3)(4).
Identification and characterization of multiprotein complexes are important steps in obtaining an integrative view of protein-protein interaction networks that determine protein function and cell behavior. The availability of complete DNA sequences of two H. pylori strains (5,6) has led to the devel-opment of reliable proteome-wide approaches for a better understanding of the virulence mechanisms of the bacterium. One of the strategies of functional proteomics, a method used to identify gene function at the protein level, is the comprehensive analysis of protein-protein interactions related to the functional linkage among proteins and the analysis of functional cellular machinery to better understand the basis of the organism's functions.
Recent progress in high throughput technologies has allowed the characterization of protein-protein interactions more directly than ever before using procedures such as the two-hybrid assay (7), co-purification (8,9), or co-immunoprecipitation (10). The workhorse of experimental proteomics has been the two-hybrid screening method, although it has been criticized for its limited accuracy of results and its laborintensive nature (11,12). Indeed, it is currently the most reliable technique for large scale characterization of protein interactions in complete genomes (13). Protein chips may eventually provide large scale simultaneous protein-protein interaction data (14,15), but technical problems (denaturation and substrate biocompatibility) must be overcome to scale up for high throughput analysis. To date, these technologies have generated large interaction networks for bacteria (16), yeast (8,9), fruit flies (17), and nematode worms (18).
Other approaches will undoubtedly become prominent as proteomics technology continues to evolve. The limiting factor for identifying protein complexes is the separation method that must be performed under native conditions to prevent protein dissociation. Because of these limits, the two-dimensional blue native (2D BN) 1 /SDS-PAGE method was applied to study the H. pylori reference strain J99 complexome. Protein identification was performed by using LC-MS/MS. This highly resolvent separation method was initially described for the separation under native conditions of the membrane protein complexes of mitochondria (19). Later on, numerous studies focused on the protein complexes of the respiratory chain (20 -27). More recently, because this method is reproducible, it was successfully used to study the detection of protein complex deficiencies of mitochondria (28 -32). In the same manner, this method was applied to the protein complexes of the chloroplast membranes of cyanobacteria (33) and protein complexes of the mitochondrial and chloroplast membranes of plants (34 -41). Moreover a similar method using agarose instead of acrylamide was developed to study protein complexes of approximate molecular mass greater than 1200 kDa (42). The anionic dye used in this method (Coomassie Brilliant Blue G-250) binds to the surface of all proteins, particularly on aromatic residues and on arginines. This binding of a large number of negatively charged dye molecules to proteins facilitates the multiprotein complex migration in a first dimension native electrophoresis (BN-PAGE), and the tendency for protein aggregation is thus reduced considerably. Multiprotein complexes are separated according to their size and shape. Each multiprotein complex is denatured in a second dimension electrophoresis (SDS-PAGE), and the protein alignment allows the identification of interactive proteins.
Recent modifications have made it possible to apply this method to whole protein complexes (43)(44)(45). Indeed this method was recently used to study the Escherichia coli cell envelope complexome (44), human embryonic kidney 293 cells (43), and human platelets (45). A dialysis must be performed on cytoplasmic extracts to eliminate salt and small molecules (43).
This method was applied to analyze both H. pylori cytoplasmic and membrane complexes. Purification steps such as liquid IEF or chromatography fractionation and enrichment were used to improve the multiprotein complex separation from the cytoplasm.

EXPERIMENTAL PROCEDURES
Bacterial Growth Conditions-H. pylori strain J99 (ATCC 700824) (6) was used for the experiments. H. pylori cells were cultured for 48 h on Wilkins Chalgren agar (Oxoid Ltd., Hampshire, UK) plates supplemented with 10% human blood and the following antibiotics: 10 mg/ml vancomycin (Lilly France S.A., Fergesheim, France), 2 mg/ml cefsulodin (Takeda France S.A., Puteaux, France), 5 g/ml Fungizone (Bristol-Myers Squibb Co.), and 5 mg/ml trimethoprim (GlaxoSmith-Kline). A bacterial suspension was grown in brucella broth (BD Biosciences) supplemented with the same antibiotics cited above and with 10% fetal calf serum (Eurobio, Les Ulis, France). The plates were incubated at 37°C under microaerobic conditions (5% O 2 , 10% CO 2 , 85% N 2 ), and the broth was incubated in a 1-liter loosely capped container, with a 150 rpm agitation, in a microaerobic atmosphere at 37°C for less than 72 h. Bacteria harvested from two or three agar plates were suspended in 250 ml of brucella broth. The resulting liquid culture showed an approximate bacterial growth of 0.1% (w/v). For the development of the 2D BN/SDS-PAGE applied to H. pylori cytoplasmic extract and to obtain the final results for both the cytoplasmic and the membrane extract, a total of 10 g of H. pylori strain J99 was frozen.
Bacterial Lysate, Cytoplasmic, and Membrane Preparations-All of the bacteria and sample manipulations were performed at 4°C. Bacteria were harvested from culture by a centrifugation at 2,500 ϫ g for 10 min and washed twice in PBS buffer. Bacteria were suspended (v/v) in native extraction buffer A (750 mM 6-amino-n-caproic acid, 50 mM Tris/HCl, pH 7.0, at 4°C) supplemented with a 1 mM final concentration of PMSF and passed three times through a One Shot disruptor at 2 kilobars. The lysate was centrifuged at 6,000 ϫ g for 20 min, the supernatant was kept, a 0.2 mg/ml final concentration of DNase I was added, and digestion was carried out for 1 h at 25°C. Then the supernatant was centrifuged at 100,000 ϫ g for 30 min at 4°C; the resulting supernatant and pellet contained the cytoplasmic protein and the membrane protein, respectively. For the cytoplasmic sample preparation, the supernatant was filtered with a Miracloth membrane (Calbiochem). and this sample was named fraction I. The concentration of 125-150 mg/ml protein was obtained, and the sample appeared limpid.
For the membrane sample preparation, a bacterial lysate was centrifuged at 6,000 ϫ g for 20 min, and the pellet was resuspended in buffer A and passed three times through a One Shot disruptor at 2 kilobars. The resulting lysate was centrifuged at 100,000 ϫ g for 30 min, and the pellet was resuspended in buffer A and centrifuged once more. The washing step was performed three times. The protein extraction was then carried out by resuspending the membrane in 1-2 ml of buffer A supplemented with 2% ␤-D-dodecyl-n-maltoside detergent (Sigma). This sample was then centrifuged at 100,000 ϫ g for 30 min, and the membrane multiprotein complexes contained in the supernatant were separated by 2D BN/SDS-PAGE. Concerning the membrane extract, the detergent used interfered with the Bradford protein dosage. For this reason and to obtain sufficiently resolvent gels, a preliminary experiment was carried out with various sample dilutions directly loaded onto the first dimension of the 2D BN/SDS-PAGE gel to determine the quantity of material needed for the electrophoresis.
Purification Steps on the Cytoplasmic Sample-All of the steps were carried out at 4°C. Purification steps using physical or chemical properties of the multiprotein complexes constituted an important aspect to confirm protein-protein interactions. Therefore, liquid IEF or exclusion filtration methods were used as purification steps before applying the 2D BN/SDS-PAGE.
Liquid IEF purification was used to separate the multiprotein complexes according to their pI in a pH range from 3.5 to 10. An aliquot of fraction I (containing ϳ60 mg of protein) was analyzed in a Rotofor system (Rotofor Prep IEF Cell, Bio-Rad). The protein mixture was prepared according to the manufacturer's recommendations before filling the Rotofor chamber. The IEF method produced many protein precipitates in the most abundant protein fractions with a pI of ϳ5-6. Very low protein concentrations were found in the basic fractions although a concentration step with the Centricon kit (Amicon, Inc., Beverly, MA) was used.
After the IEF of the protein sample, the resulting fractions were desalted in buffer A using a 5-ml HiTrap TM desalting column (Amersham Biosciences). Multiprotein complexes were recovered in 300-l fractions. Indeed for H. pylori cytoplasmic extracts, a preliminary dialysis is necessary to obtain highly resolvent gels. Here, dialysis was replaced by a desalting step, which allows the elimination of small molecules and salts, as was described for the purification of the human embryonic kidney cell line HEK293 (43).
Gel filtration purification was also tested. An aliquot of fraction I (300 l containing ϳ60 mg of protein) was loaded on a Superdex TM 200 column (Amersham Biosciences). Buffer A was run at a flow rate of 0.3 ml/min using the FPLC Akta (Amersham Biosciences). Multiprotein complexes were recovered in 250-l fractions. The cytoplasmic sample was separated into five peaks of major interest: 1000, 580, 340, 100, and 93 kDa. This method allowed the adaptability of the 2D BN/SDS-PAGE acrylamide gradient according to the mass of interest of the complexes. The Centricon kit (Amicon, Inc.) with a cutoff of 50 kDa was also used to concentrate the purified samples when the sample concentration was lower than 50 mg/ml.
First Dimension: BN-PAGE-Sample preparation and BN-PAGE were carried out as described by Schagger and von Jagow (19) with the following minor modifications. The gel dimension was 22 cm ϫ 16.5 cm ϫ 1 mm. Separating gels with linear 4 -9% acrylamide gradient gels were used. Anode and cathode buffers contained 50 mM Tris, 75 mM glycine, and only the cathode buffer was supplemented with 0.002% Serva blue G (Serva, Heidelberg, Germany). Before loading the sample, 1 l of sample buffer (500 mM 6-amino-n-caproic acid, 5% Serva blue G) was added. The gel was run overnight at 4°C at 1 watt. Thyroglobulin (669 kDa) and bovine serum albumin (66 kDa) were used for each BN-PAGE analysis as molecular weight size standards (Sigma).
Different acrylamide gradients were tested for the BN-PAGE to improve the multiprotein complex separation. We found that the best multiprotein complex separation was obtained with a linear 4 -9% acrylamide gradient. A certain balance needs to be found to optimize both focalization and separation of complexes with a mass greater than 60 kDa. The 4 -9% acrylamide gradient was performed three times on the same sample and two more times on a sample originating from another extraction; this experiment showed that the migration distance of the multiprotein complexes was the same (data not shown). This is proof that the multiprotein complexes of H. pylori maintain their global conformation during the electrophoresis. Therefore, a molecular mass could be attributed to the cytoplasmic complexes based on the two different marker proteins mentioned above.
Second Dimension: SDS-PAGE-Individual lanes from BN-PAGE were equilibrated for 5 min in an equilibrating buffer containing 1% SDS (w/v), 0.125 mM Tris/HCl, pH 6.8, and then dipped into equilibrating buffer supplemented with 100 mM dithiothreitol (Acros Organics, Morris Plains, NJ) for 15 min. Individual lanes were subsequently soaked in equilibrating buffer supplemented with 55 mM iodoacetamide (Sigma) for 15 min. An ultimate washing step lasting 5 min was performed in the equilibrating buffer without supplement. Individual lanes were placed on a glass plate at the usual position for stacking gels. After positioning the spacers and covering with the second glass plate, the gel was brought into a vertical position. Then the 10% separating gel mixture was poured. After polymerization the stacking gel mixture was poured.
Gel Staining-Silver staining was performed using a silver staining kit (Sigma) according to the manufacturer's instructions.
In-gel Protein Digestion-Silver-stained proteins separated by SDS-PAGE were excised and destained using the PROTSIL2 silver staining kit (Sigma) according to the manufacturer's instructions. Spots were subsequently washed in distilled water/methanol/acetic acid (47.5:47.5:5) until complete destaining. The solvent mixture was removed and replaced by acetonitrile. After shrinking of the gel pieces, acetonitrile was removed, and gel pieces were dried in a vacuum centrifuge. Gel pieces were rehydrated in 10 ng/l trypsin (Sigma) and 50 mM bicarbonate and incubated overnight at 37°C. The supernatant was removed and stored at Ϫ20°C, and the gel pieces were incubated for 15 min in 50 mM bicarbonate at room temperature under rotary shaking. This second supernatant was pooled with the previous one, and a distilled water/acetonitrile/acetic acid (47.5:47.5:5) solution was added to the gel pieces for 15 min. This step was repeated again twice. Supernatants were pooled and concentrated in a vacuum centrifuge to a final volume of 30 l. Digests were finally acidified by the addition of 1.8 l of acetic acid and stored at Ϫ20°C.
On-line Capillary HPLC Nanospray Ion Trap MS/MS Analysis-Peptide mixtures were analyzed by on-line capillary HPLC (LC Pack-ings, Amsterdam, The Netherlands) coupled to a nanospray LCQ ion trap mass spectrometer (ThermoFinnigan, San Jose, CA). Peptides were separated on a 75-m-inner diameter ϫ 15-cm C 18 PepMap TM column (LC Packings). The flow rate was set at 200 nl/min. Peptides were eluted using a 5-50% linear gradient of solvent B for 30 min (solvent A was 0.1% formic acid in 5% acetonitrile, and solvent B was 0.1% formic acid in 80% acetonitrile). The mass spectrometer was operated in positive ion mode at a 2-kV needle voltage and a 38-V capillary voltage. Data acquisition was performed in a data-dependent mode consisting of, alternatively in a single run, a full scan MS over the range m/z 300 -2000 and a full scan MS/MS in an exclusion dynamic mode. MS/MS data were acquired using a 3 m/z unit ion isolation window, a 35% relative collision energy, and a 5-min dynamic exclusion duration.
Data Analysis-Data were analyzed by SEQUEST (ThermoFinnigan, Torrance, CA) against a subset of the National Center for Biotechnology Information (NCBI) database consisting of H. pylori strain J99 and 26695 protein sequences. Carbamidomethylation of cysteines (ϩ57) and oxidation of methionines (ϩ16) were considered as differential modifications. Only peptides with an Xcorr values greater than 1.5 (single charge), 2 (double charge), and 2.5 (triple charge) were retained. In all cases, ⌬C n values have to be greater than 0.1.

RESULTS
Global Presentation of the Results-Multiprotein complexes from the cytoplasm were named "C," and those from the membrane were named "M." Putative multihomooligomeric protein complexes were named "H." Certain identifications (noted S1-S5) could not be attributed to complexes; they are not presented in Table I but are noted in the legends of the figures only. Proteins that are components of the same multiprotein complex co-migrate in the first dimension and are found aligned with a similar shape in the second dimension. This is the basic condition for the identification of a multiprotein complex. As an example, several spots were found to be perfectly aligned (Figs. 1A and 2), and the five proteins with a similar shape (spots 21-25) were attributed to the multiprotein complex named C8. However, the elongated form of these spots in Fig. 1A has a more rounded form in Fig. 2. In a further example, spots 4 -6 ( Fig. 3) fulfilled these criteria, and the three corresponding proteins were considered to be subunits of a multiprotein complex named C2. Aligned spots but with different shapes were considered to belong to different multiprotein complexes. Obvious examples are indicated by arrows (spots A1-A6) in Figs. 2 and 3. The spots A1 and A2 indicated by arrows (Fig. 2) are perfectly aligned; however, spot A1 has a rounded form, whereas spot A2 has a more elongated shape. For this reason, they were attributed to different complexes. Other examples of aligned spots with different shapes are indicated by arrows: 1) spots A3 and A4 in Fig. 2 and 2) spots A5 and A6 in Fig. 3. One can also notice that only spot A6 was identified in the crude extract (Fig. 1A). All these spots (spots A1-A6) were considered to belong to distinct complexes.
The crude cytoplasmic and membrane protein complexes separated by 2D BN/SDS-PAGE are shown in Fig. 1, A and B, respectively. To increase the number of multiprotein com-

Description of the protein complexes identified in H. pylori reference strain J99 using 2D BN/SDS-PAGE
Multiprotein complexes named "C" and "M" correspond to complexes isolated from the cytoplasm and from the membrane compartments, respectively. Sample preparation before applying the 2D BN/SDS-PAGE is indicated for each protein complex. Putative multihomooligomeric protein complexes were named "H." The multiprotein complexes were all localized on 2D BN/SDS-PAGE gels performed in this study and are represented in Figs Protein complex identified when a liquid isoelectrofocalization purification (Rotofor system) was performed before applying 2D BN/SDS-PAGE (Fig. 2).
d Protein complex identified when a gel filtration purification (Superdex 200 column) was performed before applying 2D BN/SDS-PAGE (Fig. 3).
e Protein complex identified when no purification step was performed before applying 2D BN/SDS-PAGE (Fig. 1).
plexes observed, non-denaturing purification steps were performed on the cytoplasmic extract before applying the 2D BN/SDS-PAGE separation, e.g. liquid IEF (Fig. 2) or exclusion filtration (Fig. 3). All of the multiprotein complexes identified in this study are described in Table I. Attempts to identify proteins were made on a large number of spots, but because some LC-MS/MS identifications failed, certain complexes have not been reported in this study. Different types of complexes were observed. Multiheterooligomeric complexes fulfilling the previously defined criteria were clearly identified, and in certain cases, these complexes were found several times. Complex C1 (spots 1-3) is an example; it is found in Figs. 2 and 3. The aligned spots are elongated in Fig. 3 but have a more marked curve in Fig. 2. This complex is no doubt present in Fig. 1A, but only spots 1 and 2 can be seen. Because spots 1 and 2 are more intensely colored than spot 3 (in Figs. 2 and 3) and the spots 1 and 2 are very light in Fig. 1, a plausible explanation is that the latter could not be visualized in Fig. 1. One can also notice that the shape of spot 3 is more difficult to interpret because the gel migration is diffuse in this molecular mass range. Interestingly the genes jhp0631, jhp0632, and jhp0633 that encode these three proteins are located on the same operon (6). Complex C2 (spots 4 -6) is clearly visualized in Fig. 3 but to a lesser   FIG. 1. Analysis of the crude cytoplasmic and membrane samples of H. pylori reference strain J99. The first dimension gel (BN-PAGE) was performed with an acrylamide gradient of 4 -9% for the crude cytoplasmic sample (A) and 6 -13% for the crude membrane sample (B). C and M represent complexes isolated from the cytoplasm and from the membrane compartments, respectively. H represents putative multihomooligomeric complexes. * represents spots where different proteins were identified (see Table I). A second migration of the C3 and M1 complexes is shown in Boxes 1 and 3, respectively. The migration of the C5 complex identified in Fig. 3 was also performed using the crude cytoplasmic extract, but the quantity of protein loaded on the gel was increased (Box 2). A second migration of spots S3 and S4 is shown in Box 4. Arrow A6 refers to Fig. 3. Two proteins were identified in spot S1: AlpB (or HopB; two peptides, coverage ϭ 2%) and AlpA (or HopC; three peptides, coverage ϭ 2%). Spot S2 corresponds to TsaA (five peptides, coverage ϭ 31%). Spot S3 contains FrpB3 (22 peptides, coverage ϭ 30%), CagA (four peptides, coverage ϭ 2%), and ClpB (four peptides, coverage ϭ 2%), and spot S4 contains BabB (seven peptides, coverage ϭ 13%), HydB (or HyaB; 11 peptides, coverage ϭ 18%), and HopM (or OMP5, six peptides, coverage ϭ 7%). Spots D1-D14 correspond to monomers of AtpA, AtpD, GroEL, Tig, RecA, AddB, Tfs, TufA, JHP0971, JHP1356, Adk, PorC, CagL, and JHP1494, respectively. extent than in Fig. 1A because spot 4 is not perfectly aligned with spots 5 and 6 in this figure. Indeed a mixture of proteins containing DnaN (23 peptides, coverage ϭ 83%) and PorA (six peptides, coverage ϭ 18%) was identified in spot 4 in the crude extract. In fact, spot 4 probably corresponds to two different spots with the spot corresponding to DnaK masking FIG. 2. Protein complexes identified from H. pylori reference strain J99 when the cytoplasmic sample was purified using the liquid isoelectrofocalization method (Rotofor system) before applying 2D BN/SDS-PAGE. Analysis of the sample fraction containing protein complexes with a global pI of 5 is shown. The first dimension gel (BN-PAGE) was performed with an acrylamide gradient of 4 -9%. H represents putative multihomooligomeric complexes. * represents spots where different proteins were identified (see Table I  the less intense spot corresponding to PorA. The complex C3 represented by spots 7-10 was identified several times in the crude extract (Fig. 1, A and Box 1). However, after the exclusion filtration purification step, not all of the spots comprising the complex were found. Only the most intense spots, i.e. spots 9 and 10, were found at a molecular mass of 500 kDa (Fig. 3). They both correspond to alkyl hydroperoxide reductase TsaA (Table I). After the liquid IEF purification step, only three intense spots were found at a molecular mass of 500 kDa, and they were also all identified as TsaA (Fig. 2, spots S5-S7). These results are in agreement with the fact that TsaA has been described as an abundant protein in H. pylori (46). The results also strongly suggest that spots 7 and 8 in complex C3, which were already very light in Fig. 1A, were probably not detected after the different purification steps due to their weak intensity or due to a loss of these two subunits during the purification process. In addition, after the liquid IEF purification step, all of the visible spots on the Fig. 2 gel were analyzed by LC-MS/MS, and the proteins corresponding to spots 7 and 8 were not found in the 75 analyzed spots (data not shown).
Among the multiheterooligomeric complexes, certain ones were not identified until after the purification step. The exclusion filtration purification of the samples enabled the identification of complexes C4, C5, and C6 (Fig. 3). The analysis of the sample fraction containing protein complexes eluted in the range of 580 kDa from the exclusion filtration allowed the identification of complex C4, comprised of two subunits. Indeed spots 11 and 12 are aligned and have the same elongated form. Similarly spots 13 and 14 with their lengthened form belong to complex C5 and were also found in the crude cytoplasmic extract when the first dimension gel was performed with a preparation containing more proteins (Fig. 1, A, H, and Box 2). Furthermore this complex was also found in other 2D BN/SDS-PAGE gels (data not shown). Lastly, complex C6 contains three subunits represented by the similar and aligned spots 15-17 (Fig. 3).
Problems were encountered for certain identifications (Figs. 1A and 2). For example, spots 18 -20 from complex C7 were found to be aligned with a similar shape as were spots 21-25 from complex C8. Spots 18 -20 (UreB, GroEL, and UreA) from complex C7 were also found in complexes C8 and C9, but the quantities of the corresponding proteins appeared lower in complexes C7 and C8 compared with complex C9, although this amount was estimated on a silver-stained gel considered as not adequate to provide a quantitative measure. This may suggest that the C7-"specific" subunits are present in C8 and C9 but are near or below the detection limit of the silver stain. However, these three complexes migrate at different molecular mass, suggesting different states of oligomerization. Consequently three different complexes were attributed. In addition, spots 26 -32 that were found to be aligned and to all have the same shape in Fig. 1A were also identified in Fig. 2 but with the additional spots 33-35. These spots also seem to be present after purification by exclusion filtration (Fig. 3); unfor-tunately their identification after this purification step was not sufficient to attribute spots 26 -32 and 33-35 either to a single complex or to two different complexes. A hydroxyapatite affinity chromatography purification (HA Ultrogel column) was also performed, but the complex corresponding to spots 26 -32 did not show an affinity to calcium and was directly eluted as demonstrated by the 2D BN/SDS-PAGE analysis of this fraction (data not shown). In fact, the analyses by 2D BN/SDS-PAGE and LC-MS/MS of the unretained fraction and the fractions eluted with the 1-300 mM linear gradient of sodium phosphate buffer did not allow the confirmation of the presence of all of these spots in only one complex or in two different complexes because spots 26 -32 were never identified in these different fractions (data not shown). The absence of spots 33-35 on the gel in Fig. 1A can be explained in two ways: 1) either the proteins present in spots 26 -32 and 33-35 are all part of the same complex and spots 33-35 that were less intense were not identified in the crude extract or 2) spots 26 -32 belong to a different complex than that of spots 33-35 but with the same apparent pI (which would explain why they were not found in the crude extract except after the liquid IEF purification). As a conservative measure to avoid describing a false complex, these spots were attributed to two different complexes, i.e. complexes C9 and C10 containing spots 26 -32 and 33-35, respectively.
Another problem relates to the identification of a mixture of proteins in certain spots, and as a result, the same shape criterion was more difficult to apply. Different complexes can indeed co-migrate in the first dimension, and therefore the spots are aligned on the second dimension. Accordingly complex C11 is comprised of spots 36 and 37 with a protein mixture in spot 37 (Fig. 1A). As a result, it is not possible to evaluate whether the protein present in spot 36 interacts with only one or with two proteins present in spot 37. To the contrary, because the proteins present in spot 37 are the subunits A and B of succinyl-CoA transferase (ScoA and ScoB) of H. pylori and are known to interact together (16,47), this complex was considered to contain JHP0739, with the two partners ScoA and ScoB (Table I).
Furthermore spots H1-H5 (Figs. 1A and 2) that were very intense were identified several times. The molecular mass observed during the migration in the first dimension of the complexes corresponding to these spots did not correspond to the molecular mass of the proteins identified in the second dimension. No other similarly formed spot was found in alignment with spots H 1 -H 5 either before or after different purifications. Several hypotheses could explain these incoherent migrations in the first and second dimensions: these spots could correspond to multihomooligomeric complexes, or they could also belong to multiheterooligomeric complexes that contain weakly expressed proteins that are not visible. Because these spots were identified several times after different purifications, it can be assumed that they are multihomooligomeric complexes. Furthermore these five complexes named H1-H5 all have already been described as multihomooligomers in other organisms (Table I), and most of the proteins involved in H1-H4 complexes, e.g. TsaA, Pfr, SodB, and AroQ, have already been reported in multihomooligomeric protein complexes in H. pylori. For each of these five putative multihomooligomeric protein complexes named H1-H5, the putative number of subunits was estimated to be 35, 31, 12, 14, and 14, respectively. Concerning the crude membrane sample (Fig. 1B), the complex M1 was easily identified with the aligned spots 38 -41 having a slightly elongated and oval form. The spots 42 and 43 were aligned; however, it was difficult to compare their form in the second dimension clearly because the molecular mass corresponding to spot 43 is low, and the gel migration is diffuse in this molecular mass range. The co-migration in the first dimension seems to be correct because these two spots were observed several times (data not shown). Moreover, these proteins (spots 42 and 43) have been identified previously as the subunits A and B of the fumarate reductase (FrdA and FrdB) of H. pylori (48). This complex was named M2. We had difficulties identifying another complex. Indeed spots S3 and S4 were aligned and presented a similar, very elongated shape, and they were identified several times in the crude extract (Fig. 1, B and Box 4). However, the molecular mass observed during the first migration (ϳ250 kDa) was lower than the molecular mass of all the proteins identified in the second dimension because protein mixtures were observed in these two spots: spot S3 contained the predicted iron-regulated outer membrane protein FrpB_3 (or FrpB3, JHP1405; 97 kDa), the cag pathogenicity island protein A CagA (or Cag26, JHP0495; 129 kDa), and the predicted ATPdependent protease binding subunit/heat shock protein ClpB (JHP0249; 96 kDa); spot S4 contained the predicted adhesin BabB (or OMP19, JHP1164; 76 kDa), the large subunit of the quinone-reactive Ni/Fe-hydrogenase HydB (or HyaB, JHP0575; 64 kDa), and the predicted outer membrane protein HopM (or OMP5, JHP0212; 75 kDa). The difference between the molecular mass of the complex observed in the first dimension or deduced from the migration of the proteins in the second dimension is too important for these proteins to belong to the same complex. These incoherent migrations in the first and second dimensions could be explained by the presence of at least two complexes that co-migrate in the first dimension and whose subunits have the same mass. Given these results, no complex including these proteins, considered to be important in the virulence of H. pylori, can be described yet, and complementary experiments must be carried out. This example clearly reflects the various difficulties of interpretation of the results obtained by the 2D BN/SDS-PAGE technique as well as the frequent difficulty in assigning certain proteins to a complex when a mixture of proteins is present.
A total of 13 multiprotein complexes containing 34 different proteins (46 in total) were identified from H. pylori strain J99. Among these complexes, 11 were obtained from the cyto-plasm, and two were from the membrane. Seven multiprotein complexes identified in this study were described previously (Table I) either totally or partially, confirming the interest of this method to study the H. pylori complexome. As an example, several interactions between the different proteins from the cytoplasmic complex C9 (Figs. 1A and 2) were partially described in the literature (Table II). These interactions included the interaction between the two structural subunits, UreA and UreB, of the well known urease enzyme (49). Moreover, previous studies described the urease enzyme interacting with GroEL (Hsp60), a chaperone and heat shock protein (50). On the other hand, some protein interactions described in the complex C9 and reported by Hybrigenics (Rain et al. (16)) involved an intermediary protein (Table II) not identified in the current study, e.g. the interaction between GroEL and RpoB included the predicted outer membrane protein JHP0600. This cytoplasmic complex C9 that includes nine different proteins contained the greatest number of interacting proteins observed in this study. In addition to heterooligomeric complexes, five putative multihomooligomeric protein complexes were identified from the cytoplasm (H1-H5, Table II). Concerning the membrane, very few complexes are reported in this study due to a poor resolution and/or visualization of the spots observed on the numerous gels performed, although the 2D BN/SDS-PAGE was initially reported for the separation of membrane protein complexes (51). This suggests that membrane purifications should also improve the gel quality.
Different oligomerization states were sometimes observed in certain complexes. For example, TsaA was found in different oligomerization states and in both cytoplasmic and membrane samples: C3 and spots H1 and S2 (Figs. 1, A and B; 2; and 3 and Table I). These results are in agreement with those Solid and dotted brackets represent direct protein-protein interactions and interactions including an intermediary protein (indicated in italics), respectively. Uvra corresponds to HP0705 and JHP0644 in H. pylori reference strains 26695 and J99, respectively. *, Dunn et al. (49), †, Evans et al. (50); §, Hybrigenics (Rain et al. (16)); ¶, the intermediary protein reported in this interaction is specific to H. pylori reference strain 26695, and no homolog exists in H. pylori reference strain J99.
of Backert et al. (46) who also identified TsaA, both in structure-bound and in soluble fractions of H. pylori, in eight oligomerization states in their 2D gel electrophoresis and mass spectrometry study. In addition, different migrations of TsaA were observed during the separation in the second dimension (Fig. 2, spots S5-S7; and Fig. 3, spots 9 and 10) suggesting that multiple isoforms of TsaA exist. Their occurrence can be explained by probable post-translational modifications modifying the physicochemical criteria (pI, molecular mass, and binding affinity) of the protein, as reported previously for numerous proteins (52,53), thus altering the migration in the second dimension. In fact, H. pylori proteins are subject to a high degree of post-translational modification, and TsaA is among several proteins present in multiple isoforms in the bacterium (54). TsaA, originally identified as a 26-kDa antigen, belongs to a family of antioxidants called AhpC/TSA (alkyl hydroperoxide reductase/thiol-specific antioxidant) involved in the defense against oxidative stress and survival of the bacterium in the host. More recently, TsaA was shown to switch from a peroxide reductase to a stress-dependent molecular chaperone function (55). H. pylori TsaA was initially reported to induce an immune response in H. pylori-infected patients with gastric adenocarcinoma (56) and also an immune response in H. pylori-infected Mongolian gerbils that is associated with the emergence of gastric diseases such as chronic active gastritis, gastric ulcer, and gastric cancer (57). TsaA belongs to the five immunodominant H. pylori proteins (58). Pentameric arrangements of homodimers of TsaA were described in H. pylori cytoplasm (59,60). In our study, TsaA was found in homooligomeric form but also in a complex including proteins JHP1030 and PepQ. The presence of JHP1030 with TsaA is not really surprising because JHP1030 is a predicted zinc-dependent mannitol dehydrogenase that possesses an oxidoreductase activity (inferred from electronic annotation). On the other hand, the presence of PepQ in this complex is more difficult to interpret because it is a predicted proline peptidase that has high homology with prolidases and X-prolyl dipeptidyl aminopeptidases and can be involved either in protein maturation or in nitrogen metabolism. Other experiments should make it possible to determine the exact role of these two partners of TsaA.
Complexes Involved in H. pylori Virulence-Before considering virulence factors stricto sensu, complexes containing proteins involved in the energy metabolism (electron transport pathway, tricarboxylic acid cycle, and gluconeogenesis), which is a prerequisite for virulence, will be briefly presented. Three pyruvate ferredoxin oxidoreductases, PorA, PorB, and PorC (ex-PorG), involved in electron transport were found together in complex C2 (Table I and Figs. 1A and 3). The heterotetrameric POR complex is comprised of the subunits PorA, PorB, PorD, and PorC (61) and was reported previously to be essential because inactivation of porB appeared to be lethal for H. pylori (62). PorD, the fourth subunit of the POR complex, was not found in this study probably because its molecular mass (14.9 kDa) is near the border limit of the gel migration. POR has been implicated in the bioreduction of nitroimidazole drugs, particularly metronidazole used for H. pylori eradication (61,63). To understand the mechanisms of metronidazole activation and resistance, which are currently poorly characterized, a more detailed knowledge of the electron transport pathways of the enzymes involved would be helpful.
Fumarate reductase catalyzes the reduction of fumarate to succinate in the Krebs cycle and appears to play an important role in the energy metabolism of H. pylori. The two catalytic subunits of H. pylori fumarate reductase, FrdA and FrdB (48), were found in the membrane complex M2 (Fig. 1B and Table  I). Fumarate reductase is not necessary for H. pylori survival in vitro but is essential for H. pylori colonization of the mouse stomach (64). It was shown to be strongly immunogenic in sera from H. pylori-positive patients, suggesting that this protein involved in H. pylori energy metabolism could also be used as an anti-H. pylori vaccine candidate (65).
Examples of multiprotein complexes including known H. pylori virulence factors are described below.
The Urease Enzyme-H. pylori produces large amounts of urease that catalyzes the hydrolysis of urea to yield ammonia and carbon dioxide, neutralizing hydrogen ions before they can lower the intracellular pH, thus enabling the bacterium to survive in gastric acid and to colonize the gastric mucosa (66). This urease activity is thus required for colonization in vivo (67). The native urease complex was found in both the membrane and the cytoplasm of the bacteria. Different oligomerization forms of the urease were identified in the cytoplasm (Table I). Four urease complexes named C7, C8, C9, and M1 with a global pI of 5 and a molecular mass range of 900 -1300 kDa were identified with crude and purified extracts (Figs. 1, A and B, and 2 and Table I). In fact, UreA and UreB are known to be present in multiple isoforms in H. pylori (54) and are among the predominant H. pylori proteins identified during cell infection (68). H. pylori is known to produce a 550-kDa heterohexameric enzyme composed of three ␣ (UreA) and three ␤ subunits (UreB). Four heterohexamers (2200 kDa) form a 16-nm spherical complex also produced by the bacteria (69). This kind of oligomerization may be adapted for acid resistance. The separation resolution of the 2D BN/SDS-PAGE did not allow the visualization of large complexes with a molecular mass greater than 1500 kDa. All of the urease complexes identified in this study included the UreA/UreB/ GroEL core. These three proteins were also reported to be present in structure-bound and in soluble fractions of H. pylori (46) and were strongly reactive to specific antibodies (58), indicating that this core is highly immunodominant in H. pylori. Under all pH conditions, the most abundant proteins observed were the urease structural subunit UreB and the chaperonin GroEL (70). A close interaction between GroEL, the co-chaperone protein GroES, and the urease enzyme had been suspected (50,71), but until now no interaction between GroEL and the urease has been confirmed. The C8, C9, and M1 complexes included additional proteins besides the core UreA/UreB/GroEL. The C8 complex included the IlvC protein, a putative ketol-acid reductoisomerase involved in isoleucinevaline biosynthesis and the thiol peroxidase Tpx involved in detoxification; the M1 complex included the FrpB_3 protein, a putative iron-regulated outer membrane protein; and the C9 complex included six other proteins (Tables I and II).
cag Pathogenicity Island (PAI) Proteins-CagA is the marker and one of the effectors of the cagPAI, translocated in gastric epithelial cells and conferring increased virulence to H. pylori strains (72). CagA was identified both in the cytoplasm and in the membrane: complex C5 and spot S3, respectively (Figs. 1,  A and B, and 3 and Table I). This localization of CagA in the membrane was reported previously (46) and is probably due to its presence in the type IV secretion system in the H. pylori cell wall (73,74). The multiprotein complex C5 including CagA and the DNA gyrase subunit A (GyrA) was found in the cytoplasmic sample with a molecular mass of 475 kDa (Figs. 1A and 3) suggesting a probable oligomerization of CagA and/or GyrA in this complex. This CagA-GyrA complex was also observed on other 2D BN/SDS-PAGE (data not shown). Cultures of CagA-negative H. pylori strains show a slower growth than the CagA-positive strains (72,75). This phenomenon could be explained by the CagA/GyrA interaction described in this study because this interaction may have a favorable effect on the normal DNA replication process, leading to a better development of the bacterial cell. However, Hybrigenics (Rain et al. (16)) described an interaction between GyrA and the cagPAI protein I (CagI), including a small intermediary cysteine-rich protein B (HcpB), which is a weak ␤-lactamase.

DISCUSSION
The classical methods widely used for large scale analysis of protein interactions are the yeast two-hybrid and the tandem affinity protein tag methods with the main limitation of expressing chimerical proteins. With the yeast two-hybrid method, mostly binary interactions are identified. Furthermore the proteins of interest are often expressed in a heterologous system, the yeast, and the interactions are observed in a particular compartment, the nucleus. Moreover the reliability of high throughput yeast two-hybrid assays is ϳ50% (76). In the tandem affinity protein tag method, the protein complexes are analyzed in vivo, but the organism being studied needs to be competent for exogenous DNA. Another disadvantage of this method is that multihomooligomeric complexes cannot be analyzed.
The 2D BN/SDS-PAGE used for the study of the H. pylori complexome also has some limitations. First, it can be difficult to easily assign each spot to independent protein complexes when they are located on the same vertical line in the second dimension; complexes C8 and C9 are very representative of this problem as well as spots containing a mixture of proteins such as spots S3 and S4 for which no complex could be attributed. For this reason, it is sometimes necessary not only to analyze the crude extract but also to carry out various methods of purifications based on different physicochemical criteria (pI-, molecular mass-, or affinity-based purification). This makes it possible to validate the identified complex and to identify complexes sometimes with a better resolution and/or visualization. However, subunits of unstable complexes can be lost during these various purification steps, therefore leading to the description of incomplete complexes. Second, proteins that are weakly expressed in certain complexes may not be visible, again leading to the description of incomplete complexes or to false multihomooligomeric complexes. In addition, mixtures of proteins can sometimes be identified in the same spots, making it difficult to determine whether all of the identified proteins belong to the same complex. This was a frequently found case for the membrane samples analyzed in this study where the outer membrane proteins often had similar molecular mass and pI. For this reason, several complexes were not reported in this study. Third, the extraction is carried out under non-denaturing conditions, and therefore the complexes requiring denaturing conditions for extraction are not easily found. Examples include 1) H. pylori vacuolating cytotoxin VacA, which is quickly exported over the membrane (77, 78) but was not detected with this method, and 2) flagellar sheaths, anchored in the membrane and requiring denaturing extraction conditions (79), which were not detected either. However, the use of non-denaturing conditions is, in fact, the major advantage of this technique because it provides the native form of protein complexes and thus allows an analysis of the physiologic state of the organism. The method was highly reproducible, and even if it is not possible to determine the totality of the H. pylori complexes, this technique is of great interest to compare the complexes of a bacterium in two different physiological states. Moreover immunoblot applied to the 2D BN/SDS-PAGE would lead to the identification of specific protein complexes, particularly those including strain-specific proteins and virulence factors. Therefore, the comparison of the currently available results obtained with all of the different methods used to analyze protein complexes will allow an accurate identification of the H. pylori protein-protein interaction network. Acknowledgment-We are grateful to Corinne Asencio for technical support in bacterial cell culture.
* This work was supported in part by the Conseil Ré gional d'Aquitaine and the Association pour la Recherche sur le Cancer. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
¶ Recipient of a predoctoral fellowship from the Fondation pour la Recherche Mé dicale.