HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text (PDF) Supplemental Data All Versions of this Article: M600106-MCP200v1 5/9/1543    most recent Submit a response Alert me when this article is cited Alert me when eLetters are posted Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Glossary Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Bisle, B. Articles by Oesterhelt, D. Search for Related Content PubMed PubMed Citation Articles by Bisle, B. Articles by Oesterhelt, D. Social Bookmarking                      What's this?

Molecular & Cellular Proteomics 5:1543-1558, 2006.
© 2006 by The American Society for Biochemistry and Molecular Biology, Inc.

---

Research

Quantitative Profiling of the Membrane Proteome in a Halophilic Archaeon^*,S

Birgit Bisle^,, Alexander Schmidt^,¶, Burghardt Scheibe^||, Christian Klein, Andreas Tebbe, Joseph Kellermann^¶, Frank Siedler, Friedhelm Pfeiffer, Friedrich Lottspeich^¶ and Dieter Oesterhelt^,**

From the Department of Membrane Biochemistry and ^¶ Protein Analysis Group, Max Planck Institute of Biochemistry, 82152 Martinsried, Germany and ^|| GE Healthcare Life Sciences, Protein Sciences, Oskar-Schlemmer-Strausse 11, 80807 München, Germany

	ABSTRACT

TOP ABSTRACT EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

 
We present a large scale quantitation study of the membrane proteome from Halobacterium salinarum. To overcome problems generally encountered with membrane proteins, we established a membrane preparation protocol that allows the application of most proteomic techniques originally developed for soluble proteins. Proteins were quantified using two complementary approaches. For gel-based quantitation, DIGE labeling was combined with two-dimensional gel electrophoresis on an improved 16-benzyldimethyl-n-hexadecylammonium chloride/SDS system. MS-based quantitation was carried out by combining gel-free separation with the recently developed isotope-coded protein labeling technique. Good correlations between these two independent quantitation strategies were obtained. From computational analysis we conclude that labeling of free amino groups by isotope-coded protein labeling (Lys and free N termini) is better suited for membrane proteins than Cys-based labeling strategies but that quantitation of integral membrane proteins remains cumbersome compared with soluble proteins. Nevertheless we could quantify 155 membrane proteins; 101 of these had transmembrane domains. We compared two growth states that strongly affect the energy supply of the cells: aerobic versus anaerobic/phototrophic conditions. The photosynthetic protein bacteriorhodopsin is the most highly regulated protein. As expected, several other membrane proteins involved in aerobic or anaerobic energy metabolism were found to be regulated, but in total, however, the number of regulated proteins is rather small.

The definition of a theoretical membrane proteome is commonly performed by bioinformatic algorithms to predict TMDs using the program TMHMM (5) or lipid anchors using the Prosite motif PS00013 (www.expasy.org/prosite) as adapted to halophilic proteins (6). Membrane-associated proteins as e.g. membrane complex subunits can only be determined by gene annotation (HaloLex database, www.halolex.mpg.de). As many protein functions are yet unknown this hampers the discrimination of specific and unspecific membrane association.

A large number of cytosolic proteomes have been characterized with the high resolving techniques of 2D PAGE MALDI-TOF, whereas membrane proteomes are still a minority in such studies particularly with respect to transmembrane proteins (3, 4, 7, 8). For identification of integral membrane proteins shotgun proteomic approaches were established as a powerful tool by combining tandem mass spectrometry with various protein or peptide separation procedures such as 1D PAGE (3, 9, 10) or multidimensional chromatography (11–14).

Extension of proteomic techniques toward protein quantitation allows detailed information about the dynamics of proteomes to be obtained. Strategies widely used in "quantitative" proteomics are based on the comparison of signal intensities of proteins or peptides from different cellular states either by mass spectrometric measurements or by comparison of stain intensities for matching spots on 2D gels. The latter gel-based method became highly sophisticated with the introduction of the DIGE procedure (15, 16), which allows differential labeling of two cellular states with matched CyDye^TM DIGE minimal fluorophores (GE Healthcare) and the subsequent joint separation on a single gel. The high dynamic range of fluorophores guarantees the quantitative evaluation of relative protein abundance by imaging techniques (17). Although most studies refer to cytosolic proteins separated by 2D gel electrophoresis (for a review, see Ref. 18), recent experiments expand the technology to membrane proteins in combination with alternative gel-based methods such as Blue Native PAGE (19).

A variety of quantitation techniques based on mass spectrometry have been developed as an alternative to the DIGE technology. Many of these methods use differential protein labeling with stable isotopes resulting in pairs of mass peaks with a characteristic mass difference. Comparing the peak area of corresponding peaks allows the calculation of the relative protein abundance in the two samples (for reviews, see Refs. 20 and 21).

The isotopic label that conveys the quantitative information can be introduced into the proteins metabolically during cell culture with labeled amino acids (SILAC (22)) or after protein isolation by chemically tagging certain amino acid residues as exemplified by the ICAT approach (labeling cysteine), the first technique developed in this field (23). In addition labeling techniques have been developed that require independent digestions of the samples to be compared. These are peptide-labeling strategies such as iTRAQ (24) or N-terminal nicotinylation (25) as well as enzyme-facilitated isotope labeling (26).

The recently developed isotope coded protein labeling (ICPL) technology (27) introduces a label to free amino groups of proteins (labeling lysine and the N terminus). Compared with ICAT, this increases the number of peptides contributing to protein quantitation. As protein samples from organelles, cells, or tissues can be easily labeled and basically all downstream protein or peptide separation techniques are applicable after labeling, this is a promising general quantitation strategy. Although quantitative proteomic approaches are widely used for the investigation of cytosolic proteins or whole cell lysates, only a few studies so far have focused on membrane proteins (28–35).

The halophilic archaeon Halobacterium salinarum is a specialist in adaptation to extreme environments: optimal growth is observed in saturated brines. To understand the adaptation of this organism to these extreme conditions the protein inventory of the cell was investigated in intensive proteomic studies resulting in the identification of 68.3% (3, 36)² of the predicted open reading frames. The organism can grow under aerobic as well as under anaerobic conditions. In the latter case, energy production is possible via the use of alternative terminal electron acceptors like dimethyl sulfoxide (37, 38), via fermentation of arginine to ornithine (39, 40), or photosynthetically via the retinal protein bacteriorhodopsin (41). As these processes are related to the membrane, H. salinarum is a good model to study quantitative membrane proteomics.

In this report we describe a strategy for membrane protein quantitation by two different techniques, DIGE-16-BAC/SDS-PAGE and more extensively with the ICPL approach based on a sample preparation that allows the most versatile use of proteomic methods for membrane proteins. We chose aerobic and anaerobic/phototrophic growth as maximally different bioenergetic culture conditions of H. salinarum to follow changes in the membrane proteome.

	EXPERIMENTAL PROCEDURES

TOP ABSTRACT EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

 
Bacterial Strain and Growth Conditions—
H. salinarum (strain R1, DSM 671) was grown in complete medium (42) as described previously (41). Briefly for preparation of a starter culture, Halobacterium was grown aerobically in the dark at 37 °C in 1 liter of complete medium to the stationary phase. For protein preparation, Halobacterium was grown through three successive transfers. For the first two transfers, 35 ml of fresh medium were inoculated with 1 ml of the previous culture; for the third transfer 1 liter of medium in a 2-liter flask was inoculated with 35 ml of the previous culture. The cells were grown to late log phase (30–40 Klett units) either aerobically in the dark or phototrophically with light as energy source. For the latter, flasks were closed after inoculation so that residual oxygen was consumed, and growth continued under anaerobic conditions. For a time course experiment, H. salinarum was grown aerobically in the dark to the late logarithmic phase and subjected to phototrophic growth for time periods of 3, 10, or 24 h, respectively.

Membrane Protein Preparation—
Two liters of cell culture were centrifuged for 50 min at 4000 x g, and cells were resuspended in 40 ml of basal salt (BS; 4.3 M NaCl, 81 mM MgSO₄, 27 mM KCl) before cell rupture by sonication (3 x 1 min on ice, 50% duty cycle, Branson sonicator). The vesicles were centrifuged (30 krpm, 1 h, 4 °C, 65,000 x g), and the pellet was resuspended in 2 ml of BS layered over a linear sucrose density gradient (10–60% sucrose in BS (w/w)) and centrifuged for 14 h (25 krpm, 4 °C, 80,000 x g). The colored vesicle band was collected, and sucrose was removed by dilution with 1 M NaCl and pelleting of vesicles by centrifugation. This step was repeated with 500 mM NaCl, and the final pellet was resuspended in 2 ml of H₂O. The membranes were delipidated with chloroform/methanol as described previously (43), and precipitated proteins were subsequently lyophilized.

DIGE Labeling and 16-BAC/SDS-PAGE—
Delipidated and lyophilized membrane proteins isolated from cells grown under aerobic and phototrophic conditions were separately resuspended in 6 M urea, 10 mM Tris, pH 8.8, to a concentration of 20 mg/ml. Then 5 µl (100 µg) of each sample were labeled with 200 pmol of Cy^TM3 CyDye DIGE minimal dye or Cy5 CyDye DIGE minimal dye (GE Healthcare). DIGE fluors were added to the protein solution and incubated after vortexing for 30 min on ice in the dark. To stop the reaction 1 µl of 10 mM lysine was added, and after vortexing the samples were left for another 10 min on ice in the dark. Both Cy3- and Cy5-labeled proteins were combined, and from each sample (aerobic and phototrophic) 300 µg (15 µl) of unlabeled proteins were added to allow subsequent protein identification. After 1:1 dilution with a modified sample buffer (55 mM DTT, 1% (w/v) 16-BAC, 1% (v/v) glycerol, 0.05% Pyronin Y) proteins originating from both conditions were co-separated on a single gel. The first dimension was 12% 16-BAC-PAGE as described previously (44, 45). The corresponding lane was excised from the gel without staining and equilibrated for 15 min in reducing solution (65 mM DTT, 2% SDS, 80 mM Tris, pH 8.8) followed by incubation in alkylation solution (260 mM iodoacetamide, 2% SDS, 80 mM Tris, pH 8.8) for another 15 min. Electrophoresis in the second dimension was carried out by 10% SDS-PAGE.

Image Acquisition and Data Analysis—
DIGE gels were scanned within the gel cassettes using Typhoon^TM imager series (GE Healthcare) to prevent any changes in the dimensions or gel cracking during large format gel handling. Cy3 was excited by green (532 nm) and Cy5 was excited by red (633 nm) laser using appropriate emission filters (Cy3, 580 band pass 30; Cy5, 670 band pass 30) to minimize cross-talk. DIGE gel analysis was performed with DeCyder^TM software (GE Healthcare) utilizing a proprietary co-detection algorithm that permits automatic spot detection, background subtraction, normalization, and relative quantitation of proteins from different images.

MALDI-TOF PMF Analysis—
The 16-BAC/SDS gel was stained with silver, and all visible spots were picked using a robot, digested with trypsin, and analyzed with MALDI-TOF PMF, and protein identifications were considered as reliable as described previously (36).

ICPL and Digestion—
Isotopic labeling of proteins was performed as described previously (27). Briefly equal amounts of each sample were separately dissolved in 6 M guanidine HCl, 0.1 M HEPES, pH 8.5, to obtain a total protein concentration of 5 mg/ml (100 µg/20 µl). The disulfide bonds were reduced (tris-2-carboxyethyl phosphine) and alkylated (iodoacetamide) as described previously (27). The protein samples obtained from the aerobically grown cells were labeled with 6-[¹²C]nicotinoyl-N-hydroxysuccinimide (light label). All mixtures from phototrophic growth states were labeled with 6-[¹³C]nicotinoyl-N-hydroxysuccinimide (heavy label). After destroying any excess of reagents with 4.5 µM hydroxylamine, the two samples were combined, diluted with 25 mM Tris, pH 8.5, to a final guanidine HCl concentration of 0.5 M and digested overnight at 37 °C with trypsin (substrate-to-enzyme ratio = 50:1). The resulting peptide mixture was acidified with 10 µl of 1% TFA, the volume was reduced by evaporation to 30 µl, and the sample was stored at –80 °C until further use. For enzymatic cleavage using endoproteinase Glu-C, the protein mixture was diluted (1:5) with 25 mM Tris, pH 7.8, to a final guanidine HCl concentration of 1 M and digested overnight at 25 °C (substrate-to-enzyme ratio = 30:1).

Protein samples to be digested with both enzymes were first cleaved with trypsin overnight as described above. The volume of the peptide solution was reduced by evaporation to 1 M guanidine HCl thus inactivating trypsin. Finally the sample was pH adjusted to pH 7.8 with 1% TFA, and enzymatic cleavage with endoproteinase Glu-C was performed as described above. An aliquot of 30 µg of digested proteins was applied for each LC-MS/MS experiment.

For the time series experiment membrane proteins isolated from the aerobic culture were labeled with light ICPL tags, and membrane proteins isolated from cultures phototrophically grown for 3, 10, or 24 h were each labeled with heavy ICPL tags. Three preparations were made in which the light isotopically labeled sample (aerobic) was mixed individually with each of the phototrophic samples carrying the heavy labels. Protein mixtures were digested with trypsin as well as with trypsin + Glu-C as described above, and 30 µg of digested proteins were each applied for LC-MS/MS measurement.

HPLC and MALDI Plate Spotting—
All peptide separations were performed utilizing a capillary liquid chromatography system (Ultimate, LC Packings) containing a reversed-phase column (LC Packings PepMap reversed-phase C₁₈ column, 75-µm inner diameter, 15 cm) coupled directly on line with a MALDI target spotter (Probot, LC Packings). A sample volume of 50 µl was injected, and the peptides were trapped on a short reversed-phase column (300-µm inner diameter, 5 mm). The mobile phase consisted of 0.05% TFA (A) and 0.04% TFA in 80% (v/v) ACN (B). For the separation of the peptides, a 65-min linear gradient from 10 to 45% B at a flow rate of 200 nl/min was used followed by a 20-min wash step of the column with 100% B. The analytical column was directly connected to a MicroTee (Upchurch, Oak Harbor, WA) where the eluent was mixed with MALDI matrix solution (5 mg/ml -cyano-4-hydroxycinnamic acid, 50% (v/v) ACN, 5 mM ammonium dihydrogen phosphate, 0.1% (v/v) TFA) at a flow rate of 1.3 µl/min and deposited onto a blank MALDI plate. The LC eluent was automatically spotted in 10-s fractions over a time period of 66.66 min resulting in 400 spots per MALDI target plate. The sample spots were allowed to dry at room temperature.

MALDI-TOF/TOF Analysis—
Mass spectrometric analysis was performed on a 4700 Proteomics Analyzer from Applied Biosystems (Framingham, MA) equipped with an neodymium-yttrium aluminium garnet (Nd-YAG) laser that produces pulsed power at 355 nm at pulse rates of 200 Hz. MS spectra were acquired by accumulation of 2500 laser shots using a positive reflector mode with a deflection cutoff range of m/z 700. Subsequently the MS spectra were automatically analyzed by the Peakpicker software (a generous gift from Applied Biosystems) to detect and quantify isotopic peptide pairs and generate a list of precursor ions for MS/MS analysis. A total of 1500 laser shots were carried out for each high energy MALDI-TOF/TOF CID spectrum utilizing collision energy of 1 keV and nitrogen as collision gas. For MS/MS analysis of bacteriorhodopsin (BR), the sample as digested for MALDI-TOF PMF was manually spotted, and spectra were acquired accordingly.

Data Analysis—
All MS/MS spectra obtained were searched against the Halobacterium protein sequence database, which was exported from the HaloLex database (www.halolex.mpg.de) (36) using an in-house version of Mascot (46) in combination with the GPS-Explorer^TM 2.0 software (Applied Biosystems). For the database search, carbamidomethylation was set as a required cysteine modification, whereas oxidation of methionine was considered as a variable modification. Further potential modifications include [¹²C]- and [¹³C]nicotinoylation of lysine and the protein N terminus. Because all lysines are blocked by the ICPL tag, trypsin only cleaves after arginine residues, and consequently this was mimicked by selecting the enzyme Arg-C for the Mascot search. MS/MS spectra with a Mascot score above the confidence threshold score using a confidence level of 98% in case of singlets and 90% in case of pairs were considered to be correct calls (Supplemental Table II). To further estimate the rate of false positive peptide identifications, the MS/MS data set of experiment "trypsin 1" was searched against a database containing all protein sequences in reverse order (Supplemental Table III). All internal peptides generated by the reversed database have the same mass as their equivalent in the standard database, whereas only the mass of the N- and C-terminal peptides differ. Subsequent data validation was carried out using the Spotfire® DecisionSite^TM Browser 8.0 (Spotfire AB Europe, Göteborg, Sweden) where all incorrectly identified modified peptides can be easily detected and erased. In detail the number of ICPL tags per identified peptide must match the number of lysines in the sequence and an eventual N terminus of the protein. Secondly the mass of the precursor peptide is increased by the number of tags, and the mass difference between the corresponding pairs must have the expected value. To minimize potential errors in ratio determination, e.g. due to overlapping peaks or incorrect peak assignments, the automatically acquired quantitative data were further manually verified (Data Explorer, Applied Biosystems) in the following cases: (i) only one labeled peptide was available for protein quantitation or (ii) the calculated peptide ratios showed a standard deviation of more than 20% regarding other quantified peptides of the same protein.

The ratio for each pair was calculated by the program Peakpicker. Ratios for each protein were determined comprising all quantified peptides of one protein (Supplemental Table I, column 8), and the standard derivation in percent was calculated by the program Spotfire on the basis of the raw data (Supplemental Table I, column 10). The median of the complete set of quantified peptides was calculated and used for a computational normalization of the original ratios (Supplemental Table I, column 11). Finally regulation factors were computed for each protein such that the same extent of positive or negative regulation results in an identical absolute value of the regulation factor (Supplemental Table I, column 12). Initially the ratio was computed by dividing the phototrophic over the aerobic value. For proteins that were more abundant under phototrophic conditions, this directly results in a regulation factor between 1 and infinity. For proteins that were less abundant under phototrophic conditions (more abundant under aerobic conditions) this results in ratios between 1 and 0. To provide symmetric regulation factors these ratios were inversed and multiplied by –1. This scale excludes any values between 1 and –1.

	RESULTS

TOP ABSTRACT EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

 
Sample Preparation—
The halophilic archaeon H. salinarum was grown in complex medium under aerobic and phototrophic growth conditions. Culture conditions were identical except that the aerobic sample was grown in open flasks in the dark, whereas phototrophic samples were grown in flasks that were illuminated and closed so that cells consumed residual oxygen by the end of the lag phase, which usually was about 30 h (data not shown), and growth continued under anaerobic conditions. After reaching the late log phase cells were harvested, and membranes were prepared for quantitative proteomic analysis (Fig. 1A).

View larger version (35K): [in this window] [in a new window]   FIG. 1. Flow chart of the membrane protein preparation and quantitation strategy. A, membrane preparation. Membranes of cells cultured differently were isolated by sonication, membranes were delipidated, and precipitated proteins were lyophilized. B, quantitation by DIGE after 16-BAC/SDS separation. Samples were labeled with Cy3/Cy5. The same weighed amounts of proteins from both conditions were combined, separated by 16-BAC-PAGE, and quantified with the DeCyder software. After silver staining, spots were picked, and proteins were identified by MALDI-TOF PMF. C, quantitation by ICPL using LC-MS/MS. Proteins were labeled with 6-[12C]nicotinoyl-N-hydroxysuccinimide or 6-[13C]nicotinoyl-N-hydroxysuccinimide, samples were combined after labeling and digested, and peptides were separated by HPLC. Fractions were directly spotted on a MALDI target, analyzed by MALDI-TOF/TOF, and quantified by the Peakpicker software. RP, reversed-phase.

16-BAC/SDS-PAGE and DIGE—
In a first approach we attempted quantitation by a two-dimensional gel-based strategy that gives an immediate overview of regulatory phenomena. By using 16-BAC/SDS, it is possible to avoid IEF, which results in irreversible precipitation of integral membrane proteins (3). Changes in the membrane proteome were followed by comparing spot intensities on two-dimensional 16-BAC/SDS gels. An optimized protocol for these gels allows a high resolving separation of membrane proteins. In a first experiment we analyzed different states (aerobic and anaerobic/phototrophic) on individual gels with silver staining (not shown), but hardly any differences could be detected. This approach has several disadvantages (difficult spot correlation, low dynamic range of silver stain, and insufficient reproducibility of overall stain intensity). Quantitation by the DIGE system was applied to overcome these problems. Lyophilized membrane protein samples were subjected to labeling with CyDye DIGE fluorophores, combined, and co-separated by two-dimensional 16-BAC/SDS-PAGE (Fig. 1B). The results obtained by the DIGE experiment were acquired by differential in-gel analysis without exploiting the internal standard.

Fig. 2 shows the separated mixture of Cy3- (aerobic sample) and Cy5 (phototrophic sample)-labeled membrane proteins. The different colors indicate quantitative differences under the two growth conditions: red represents up-regulation in the phototrophic sample, green represents up-regulation in the aerobic sample, and yellow indicates no regulation. For quantitation fluorescence intensities were analyzed by the program DeCyder 2D (GE Healthcare), after subsequent silver staining spots were picked, and proteins were identified by MALDI-TOF PMF. With this method all gel spots can be quantified even if subsequent protein identification is not successful. The list of identified and quantified proteins is presented in Table I (quantitative data for gel spots without identification are not shown).

View larger version (76K): [in this window] [in a new window]   FIG. 2. 2D 16-BAC/SDS separation of membrane proteins labeled by DIGE. Membrane proteins of aerobically cultured cells were labeled with Cy3 (green), and membrane proteins of phototrophically cultured cells were labeled with Cy5 (red). Samples were combined, separated by 16-BAC-PAGE, and quantified by the DeCyder software. After subsequent silver staining, in-gel tryptic digestion and protein identification was followed by MALDI-TOF PMF. a–c, three spots of BR.

The three BR spots showed regulation factors of 2.3 (a), 3.5 (b), and 4.4 (c), respectively. In the ICPL experiments described below BR was quantified by a single peptide resulting in a regulation factor of 3.1. The correlation between DIGE and ICPL quantitation is considered to be good when taking into account that the gel-free method cannot distinguish between different protein isoforms and that the regulation factor of 2.3 for the most intense spot (a) was obtained. A good correlation was also found for the regulated protein OE4311F and for the large number of unregulated proteins (Table I). In our subsequent work, we complemented the DIGE/CyDye-based quantitative analysis of the membrane proteome by the ICPL approach, which provides enhanced protein identification properties.

Quantitation of Membrane Proteins by ICPL—
The ICPL technology (27) introduces a label (6-nicotinoyl-N-hydroxysuccinimide) to all free amino groups of intact proteins, explicitly to lysines and free N termini. This allows mass spectrometry-based protein quantitation by comparison of the peak areas of MS mass peak pairs in combination with peptide identification by MS/MS. As illustrated in Fig. 1C, membrane proteins from aerobically grown cells (light label with ¹²C) and phototrophically grown cells (heavy label with ¹³C) labeled using the ICPL method were combined and digested. Peptides separated by HPLC on a reversed-phase C₁₈ column were then analyzed by MS. Results were evaluated by the program Peakpicker, which selects potential labeled mass peak pairs by their characteristic mass difference between the ¹²C and ¹³C ICPL tags. Such pairs can be used for both identification and quantitation. Peakpicker communicates with the instrument to preferentially select the more intense peak from a mass peak pair for subsequent MS/MS analysis. Residual MS/MS capacity is used for analysis of singlets. Singlets may either originate from unlabeled peptides or from a labeled peptide for which the partner peak escaped peak detection.

Using this approach, certain difficulties specially attributed to integral membrane proteins are encountered. The size of tryptic peptides tends to increase with the number of TMDs (3). This is aggravated by the fact that the ICPL reagent modifies lysine residues, and thus tryptic cleavage occurs only after arginine, resulting in even longer peptides that are not detectable by MALDI-TOF/TOF as they exceed the scanned mass range of 800–4000 Da. As an example, BR does not yield a single peptide after tryptic cleavage having both required properties for quantitation, a detectable mass and an ICPL tag at lysine. To increase the probability of integral membrane protein quantitation, different digestion protocols were applied. The samples were digested by either trypsin or Glu-C (cutting C-terminally of glutamic acid) or by a combination of both. With the latter combination, BR could be quantified by a 1125-Da peptide representing a part of a TMD with a GRAVY index of 0.77. The weak signal intensity of the MS peak (Fig. 3) derived from a highly abundant protein illustrates the difficulties of proteomic analysis for TMD-containing peptides.

View larger version (17K): [in this window] [in a new window]   FIG. 3. Quantitation of the seven-TMD protein BR (OE3106F). Membrane proteins were isolated, labeled with ICPL tags as described under "Experimental Procedures," and digested with a combination of Glu-C and trypsin. Peptides were separated by C18 reversed-phase LC and spotted on a MALDI target, pairs were selected, and peptides were identified by MS/MS. The expanded view shows one labeled peptide pair representing a part of a TMD of BR. The ratio of the peak areas derived from the 13C heavy (phototrophic sample) and 12C light (aerobic sample) labeled peptide in the MS spectra yields the relative quantitation of BR.

ICPL Is Well Suited for Quantitative Proteomics on Membrane Proteins—
In a previous study (3) we showed that the number of tryptic peptides identifiable by MALDI-TOF is much lower for integral membrane proteins than for cytosolic proteins. To estimate the applicability of several quantitation strategies, we computed the number of quantifiable peptides for Cys (ICAT and HysTag) or Lys (ICPL) labeling. Such peptides must fulfill two criteria: (i) the mass must be within the scanned range of 800–4000 Da, and (ii) the labeled amino acid must be present in the peptide sequence. As peptides containing predicted transmembrane regions are severely underrepresented in the set of identified peptides from integral membrane proteins (unpublished data),³ we also computed data for peptides derived exclusively from loop regions. The average peptide-to-protein ratio is shown for several proteases and combinations thereof (Fig. 4A). For all digestion protocols, the theoretical number of quantifiable peptides per protein is much lower for integral membrane proteins than for soluble proteins (2–3-fold reduction) and even further reduced when restricting the analysis to loop peptides (3–4-fold reduction).

View larger version (29K): [in this window] [in a new window]   FIG. 4. Statistic analysis of quantifiable peptides per protein. A, the average number of peptides per protein containing either Cys or Lys in the peptide sequence with a mass of 800–4000 Da was determined for various digestion scenarios. The figure distinguishes between non-TMD and TMD proteins, and for the latter data were computed after restriction to loop peptides. The letters B and C indicate for which of the data details are specified in the corresponding figure section. B and C, the fraction of proteins having a given number of quantifiable peptides was computed. The bar to the left shows the fraction of proteins that cannot be quantified at all (no quantifiable peptide) (bottom, dark diagonal striped bars) and the fraction of proteins that can be quantified (one or more quantifiable peptides) (top, light diagonal striped bars). The plot to the right shows the fraction of proteins having a given number of quantifiable peptides, omitting proteins lacking a quantifiable peptide. In B, Lys and Cys labeling are compared for tryptic digestion of proteins labeled using the ICPL method (mimicked by Arg-C) as indicated by letter B in A. In C, details are specified for non-TMD and TMD proteins as well as loop peptides for Lys labeling of the same digestion scenario as indicated by letter C in A. Raw data for all digestion protocols are given in Supplemental Table VI.

Based on the calculations for this proteome, the ICPL approach resulted in more peptides that contribute to the quantitation of the membrane proteins than an approach tagging at Cys like ICAT (23) or HysTag (29). But even ICPL is far less efficient than for cytosolic proteins.

Assessing the Reliability of Protein Identification—
It is important to minimize false positive identifications by MS/MS while simultaneously trying to avoid false negatives. The data set of experiment trypsin 1 was analyzed to select the confidence level that meets these goals (Table II). The data were searched against two databases: one contained the standard protein sequences (Supplemental Table II, experiment trypsin 1), and the other was a "reversed database" in which all protein sequences are written from C to N terminus (Supplemental Table III) (48, 49). In such a reversed database, general characteristics like amino acid composition and peptide mass distribution are nearly unchanged, whereas the sequence-derived fragmentation patterns that form the basis for identification are maximally different. The false positive rate was computed under the assumption that all identifications in the reversed database are false positives and that all identifications in the standard database are correct calls. The number of identified peptides depends on the setting of the confidence level, which is the percentage of hits considered by Mascot to be correct calls. For unlabeled peptides, the number of false positives by our evaluation closely correlates with the applied Mascot confidence level (Table II, column 5).

Large Scale Quantitation of Membrane Proteins—
Table III summarizes the data of four experiments while further details are given in Supplemental Table II. After the automated recording, selection of pairs, and identification of unique proteins (Table III, columns 3, 4, and 6) the data were inspected manually. After removing data that do not fulfill the criteria specified under "Experimental Procedures," the list of final candidate peptides and proteins was obtained (Table III, columns 5 and 7). Up to 20% of the quantified peptides were considered to be false positives in the individual experiments, reducing the number of quantified proteins by 12%.

In Fig. 5 identified (A) and quantified (B) proteins from the four experiments are grouped into transmembrane proteins, lipid-anchored membrane proteins, components of membrane protein complexes, and other proteins. Of the 206 identified proteins, 175 (85%) could be quantified by ICPL on average by four peptides per protein. A total of 163 membrane proteome constituents were identified, and quantitative data for 136 (83%) of these could be obtained. Similar identification and quantitation ratios for TMD and non-TMD proteins show that this technique is applicable to membrane protein quantitation despite the plethora of difficulties encountered for this class of proteins. In addition 13 of 39 quantified non-TMD proteins are yet functionally uncharacterized and could be either cytoplasmic contaminants or true membrane-associated proteins.

View larger version (30K): [in this window] [in a new window]   FIG. 5. Categories of 206 identified (A) and 175 quantified (B) proteins by the ICPL approach. Shown are TMD prediction by TMHMM (5), lipid anchor prediction by the identification of the lipobox motif (6), and membrane protein complex subunits according to protein annotation (www.halolex.mpg.de).

The statistical variance of the regulation factor was computed by the program Spotfire. We attempted to estimate (i) the influence of the number of quantitated peptides per protein and (ii) whether data from repetition of LC-MS/MS measurements differ in variance from all experiments. In the following, the standard deviation (S.D.) is given as a percentage of the regulation factor and ranges from 6.9 to 10.4% (Fig. 6). The variability was slightly lower when only two rather than three or more peptides could be quantified (7.9 versus 10.4% for all four experiments). Data from all experiments are slightly more variable than those for repetition of LC-MS/MS measurements (10.4 versus 6.9% for proteins identified with at least three peptides). Nevertheless the reproducibility of the quantitation was sufficiently high to allow the combination of the data from all four experiments. With the S.D. below or only slightly above 10%, a value of 30% is 3 S.D. above mean, and thus we use 30% increase or decrease in average peak area (i.e. regulation factors above +1.3 or below –1.3) to indicate regulation. An assignment based on data exceeding 3 S.D. should be correct with 99%.

View larger version (27K): [in this window] [in a new window]   FIG. 6. Variance of ICPL quantitation. The standard deviation, given in percentage of the regulation factor, was calculated for proteins quantified by at least three peptides in four experiments (black bars) and the LC-MS/MS repetition (white bars). In addition proteins quantified by two peptides are shown (diagonal striped bars). The inset shows the average standard deviation for each group.

View larger version (28K): [in this window] [in a new window]   FIG. 7. Regulated proteins sorted by function class. Proteins up-regulated in aerobic (aerob) or anaerobic (anaerob)/phototrophic (phototroph) growth were grouped into proteins involved in aerobic or anaerobic bioenergetics, transport, translation, and various other functions as well as unknown functions according to Supplemental Table I.

Other proteins involved in the transcriptional regulation and biosynthesis of BR like the transcription factor homolog and the bacterio-opsin activator were not quantified in this study as they are not membrane-associated. However, these proteins as well as the phytoene dehydrogenase, a key enzyme of retinal biosynthesis, were found to be up-regulated under phototrophic growth conditions when analyzing the cytosolic proteome⁴ (50).

In addition to BR, two protein complexes involved in anaerobic bioenergetics were found to be more abundant under anaerobic/phototrophic growth conditions. One complex is the dimethyl sulfoxide reductase (37) for which we quantified the A and B subunits. The other complex is the Halobacterium homolog of the Escherichia coli anaerobic dehydrogenase for which we quantified the A and C subunits. In contrast, under aerobic growth conditions, three subunits of two distinct cytochrome c-type terminal oxidases were found to be more abundant. Several subunits from respiratory chain complexes I, II, and III as well as of the ATP synthase could be quantified but were found not to be regulated.

A predominant group of regulated proteins, having a reduced level under anaerobic/phototrophic conditions, are involved in transport processes. A total of 10 subunits from nine distinct ABC-type transport systems were found to be regulated. A conspicuous group that shows a significantly increased level under anaerobic/phototrophic growth conditions are ribosomal proteins. This may, however, indicate more extensive membrane association of the ribosomes under conditions of enhanced membrane protein biosynthesis rather than quantitation of the absolute amount of ribosomal proteins. Although frequently considered to be mere contaminations of membrane samples, ribosomes and other components of the translation machinery like the 7 S RNA of the signal recognition particle have been described to be pulled down with membranes especially under conditions in which membrane protein biosynthesis is extensive (51).

Finally a number of "conserved hypothetical proteins" were found in different amounts under the two growth conditions. Up to now, the function of these proteins is enigmatic, and it is not yet evident in which of the many cellular processes being affected by the shift from aerobic to anaerobic/phototrophic growth they might be involved.

To analyze the regulation of BR biosynthesis more closely, we performed a time series experiment using cells that were grown aerobically in the dark. One of the cultures was stopped (0-h sample), the other cultures were closed and grown for an additional 3, 10, or 24 h under phototrophic growth conditions. Membranes were isolated, labeled with the ¹²C light (0 h) or ¹³C heavy (3, 10, and 24 h) ICPL label, and digested either with trypsin or with trypsin + Glu-C. Relative abundance of proteins was determined by the comparison of each phototrophically grown membrane protein sample with the aerobic membrane protein sample (Supplemental Table V).

Switching a high density cell culture from aerobic to anaerobic/phototrophic growth by this procedure is physiologically not equivalent to growth of a freshly inoculated culture under aerobic versus anaerobic/phototrophic growth conditions. This is indicated by the fact that cell density remained constant for 24 h. Under these conditions, the amount of BR, which was again found to be the most intensively regulated protein, linearly increased up to 5-fold within 24 h (Fig. 8). Only one subunit of one of the cytochrome c-type terminal oxidases could be quantified and was found not to be affected. In addition, several ribosomal proteins were found to linearly decrease over 24 h.

View larger version (26K): [in this window] [in a new window]   FIG. 8. Time course experiment monitoring changes in the relative abundance of BR, COX 1B, and four ribosomal proteins (Rsp7, Rsp3aR, Rpl15R, and Rsp5). Four individual cultures were grown aerobically in the dark up to the late logarithmic phase. Growth was arrested in one of the cultures used for the preparation of membrane proteins that serves as a reference. The rest of the cultures were subjected to phototrophic growth for time periods of 3, 10, and 24 h, respectively. Membrane proteins isolated from the aerobic culture were labeled with light ICPL tags, and those from phototrophic cultures were labeled with heavy ICPL tags. Equal amounts of isotopically labeled proteins were then mixed and digested, and the relative abundance of proteins was determined by LC-MS/MS.

	DISCUSSION

TOP ABSTRACT EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

A large set of quantitative data could be collected by applying the gel-free and mass-based ICPL technique (27). ICPL introduces a label to free primary amines, thus tagging Lys and the free N terminus of a protein. The labeling of Lys rather than Cys as in other approaches (ICAT and HysTag) proved to be advantageous for the study of membrane proteins (Fig. 4). This increased the number of quantifiable peptides per protein by a factor of about 3 especially when concentrating on integral membrane proteins where peptides from loop regions are preferentially detected. Even with Lys labeling, the number of quantifiably peptides was low for integral membrane proteins as compared with soluble proteins. Thus, quantitations by only one peptide, a situation resembling "one-hit wonders," cannot be avoided, and several integral membrane proteins did not even have a single quantifiable peptide.

The SILAC approach can basically operate with any amino acid being used as label (22), and thus this technique was successfully applied to membrane proteins (30). By choosing an amino acid of appropriate frequency one can overcome the limited set of quantifiable peptides for membrane proteins, but the technique has several severe drawbacks. For multicellular organisms, it is restricted to cell cultures. For prokaryotes, synthetic media are required, and we experienced very poor growth of Halobacterium in synthetic media under phototrophic conditions. Thus, although statistically favorable, the SILAC approach was not applied in this study.

Quantitation strategies such as iTRAQ (24), N-terminal nicotinylation (25), or enzyme-mediated heavy atom incorporation (26) allow labeling of basically every peptide. These techniques require an additional processing step that must be executed before samples can be mixed, namely independent proteolytic cleavage. Thus, reproducibility of proteolytic digestion is a fundamental prerequisite for these approaches, and for integral membrane proteins this is known to be a difficult task. In comparison, the isotopic tagging is performed directly after membrane preparation (for which we simplified the protocol) and thus at the earliest possible stage to enhance reproducibility and accuracy. Isotope labeling can be achieved by using ²H or ¹³C. One disadvantage of using ²H is the different retention times of heavy and light labeled peptides during reversed-phase LC separation (52), resulting in a rather complex data analysis for accurate ratio calculation. In contrast, using ¹³C labeling results in co-elution of modified peptides, which greatly facilitates data interpretation and accurate protein quantitation.

One approach to overcome the problem of having a low number of quantifiable peptides with the ICPL approach is to repeat the analysis and to apply different digestion protocols. The latter does not increase the quantifiable peptides per protein in the individual experiment but increases the number of quantified peptides when data are pooled. Additionally application of several digestion protocols extends the range of quantifiable proteins as some are accessible by only one digestion procedure, e.g. the combination of trypsin and Glu-C for BR (Fig. 3).

The complexity of the sample may be a limiting factor in protein quantitation. In contrast to the ICAT approach, where tagged peptides are enriched, the ICPL technique does not allow such enrichment. We found that the separation capacity is not exceeded despite using one-dimensional HPLC as the only separation dimension as an additional separation step on the protein level by 1D PAGE did not lead to an increase of quantitative data (not shown). We attribute this to the limited complexity of the peptides derived from the membrane proteome of prokaryotes. In contrast, the complexity of cytosolic preparations exceeds the separation capacity, and thus prefractionation on the protein level by 1D PAGE is advantageous.⁴ Tryptic digestion theoretically generates 4176 peptides in the scanned mass range from 582 proteins with predicted TMD but 32,180 peptides from 2247 proteins without TMD. Thus the complexity of the integral membrane proteome is reduced by 3.9-fold at the protein and by 7.7-fold at the peptide level, although complexity is somewhat higher due to the inclusion of peripheral membrane proteins in the membrane proteome.

BR is frequently considered as the archetypical membrane protein that has been used as reference protein for many technical developments including membrane proteomic techniques (53–57). Accordingly high sequence coverage has been reported for this protein. It should be noted, however, that these studies deal with protein identification rather than quantitation and also that highly pure samples (purple membrane with 99% protein purity) have been used. These results are difficult to transfer to a genome-wide quantitation strategy, which is the major goal of our analysis. In conclusion, the gel-free analysis remains the method of choice for quantitative membrane proteomics.

In addition to the mass spectrometry-based quantitation we used a gel-based strategy and combined the DIGE minimal labeling (18) of membrane proteins with a subsequent separation on 16-BAC/SDS two-dimensional gel (44, 45). This approach allows the quantitation on the protein level and is well suited to analyze the dynamics of the proteome on a global scale as it is independent of mass spectrometric identification of individual gel spots, which may be a cumbersome task for integral membrane proteins. The cells responded to the switch from aerobic to anaerobic/phototrophic growth with minor changes, which is in concordance with the data of the ICPL approach. Only a few slightly regulated proteins were identified; among them, as expected, was BR, which is the basis for archaeal photosynthesis. With gel-based quantitation one can distinguish between isoforms due to differential processing and posttranslational modification when these isoforms migrate differently in a two-dimensional gel system. In contrast, these proteins are often quantified by the same peptides upon mass-based quantitation. In the gel-based DIGE approach BR was measured in three different spots, whereas it was quantified by one peptide in the ICPL approach. It is known from the literature that the BR precursor is processed in two steps (58). Although not analyzed in detail, one can assume that the spot with the higher apparent mass represents the BR precursor, which is expected to show a higher regulation factor than the mature protein, which is known to be extremely stable. Beyond the sole information regarding protein regulation the DIGE data are also suitable as reference for the mass-based quantitation by the ICPL approach. The results of the quantitative information for the two different technologies correlate very well, therefore confirming each other reciprocally.

The application of two completely different strategies is important to validate quantitation data resulting in a "double standard in proteomics" (59). Recently applied to cytosolic proteomes (59–61), this is especially important for membrane proteins because they have an increased tendency to escape mass spectrometry-based quantitation (62).

In total, 175 proteins could be quantified in this analysis of which 155 are membrane proteins, and 101 of these are integral membrane proteins with predicted TMDs. Thus, the relative abundance of a large subset of the H. salinarum membrane proteome for the two growth conditions, aerobic and anaerobic/phototrophic, has been determined. Although these two states represent a dramatic shift in energy supply, quantitative differences in the membrane proteome affected rather few proteins, and those were only slightly regulated. Quantitative differences were below our threshold for 75% of the identified proteins.

As expected BR, the light-driven H⁺ pump responsible for halobacterial photosynthesis (41), was preferentially found under phototrophic conditions, a phenomenon that is well documented in the literature (50, 63). In our study, there was 3 times more BR under phototrophic than under aerobic conditions. In a time series experiment, the relative abundance of BR reached 5-fold concentration within 24 h with a linear increase during this time period. This ratio may seem small as BR can become the most prominent membrane protein of the cell, but that refers to the overproducing strain S9, whereas the wild-type strain R1 was used in this study. Other studies provide highly variable regulatory factors for bacteriorhodopsin, depending on the aerobic growth conditions with which the phototrophic condition is compared (64).

Under aerobic conditions, increased amounts of subunits from two cytochrome c-type terminal oxidases (which oxidize halocyanin and correspond to complex IV of the respiratory chain) were found, but no differences for the other complexes of the respiratory chain could be obtained. In the time series experiment, only one subunit from one of the cytochrome c-type terminal oxidases could be quantified and was found not to decrease within 24 h. This can be explained if the decrease in cytochrome c-type terminal oxidase, as found under steady-state conditions, results from "dilution out" by cell division rather than from a specific elimination of the protein.

Under anaerobic/phototrophic conditions, we found increased amounts for two subunits of the DMSO reductase, which functions as a terminal oxidase under anaerobic conditions and was described to function with DMSO as well as trimethylamine-N-oxide (37, 38). As neither DMSO nor trimethylamine-N-oxide was added to the medium, the regulation is most likely due to lack of oxygen rather than substrate induction. Another protein complex that shows increased amounts under anaerobic/phototrophic conditions is the H. salinarum homolog of anaerobic glycerol-3-phosphate dehydrogenase, which has been characterized in E. coli (65). Although the sequences of the three subunits from this complex do not provide evidence for membrane anchoring, subunit B is quite hydrophobic. In addition subunits of this complex have been preferentially identified in membrane samples (3), and it is known that the E. coli complex is loosely associated with the membrane (66, 67). Both of the regulated complexes involved in anaerobic energy metabolism interact with the lipid-soluble quinone pool of the respiratory chain. This may indicate that the respiratory chain is used under both aerobic and anaerobic conditions with exchange of the terminal oxidase and thus may explain why only the terminal oxidases are found to be regulated. Thus, with regulation of a relatively small protein subset, the physiological status of the cell can be drastically changed.

Similarly Halobacterium is able to switch to photosynthesis with regulation of only a few proteins. This contrasts with the events in chlorophyll-dependent photosynthetic bacteria. Here the chromatophoric membranes harboring reaction centers and light-harvesting complexes are absent under aerobic conditions and only built in light under anaerobic conditions. This process involves the synthesis of dozens of protein components and lipid constituents of membranes (for a review, see Ref. 68). Such an all-or-none response in biosynthesis of the photosynthetic apparatus is essential due to the formation of highly toxic superoxide radicals upon the combined action of light and oxygen on chlorophyll systems. In the retinal-based photosynthesis bacteriorhodopsin acts as a "mechanically" light-driven proton pump involving only a thermoreversible cis-trans isomerization without any electron transfer reactions (69, 70). Therefore, it can be active in the presence of oxygen without the danger of toxic side reactions.

We found several regulated proteins that are not directly linked to photosynthesis or aerobic and anaerobic energy metabolism. Reduced amounts were obtained for nine different ABC transporters under anaerobic/phototrophic conditions. This may be either due to limitations in membrane space related to formation of the BR-containing purple membrane or may be a specific down-regulation of the affected transporters. Under anaerobic conditions, the cell uses BR-driven photosynthesis as an efficient energy source. It was found that under these conditions respiration as well as anaerobic arginine fermentation is inhibited (71, 72), whereas a light-dependent fixation of CO₂ could be measured (73). Accordingly there may be a reduced requirement for metabolic substrates and correspondingly for transport processes.

Quantitative differences were found for several ribosomal proteins, which often are considered to be mere contaminants of membrane preparations. However, it was found that up to 50% of the ribosomes in Archaea are attached to the membranes by a specific interaction, and the fraction of attached ribosomes varies with the growth phase (74).

Cells grown continuously under anaerobic/phototrophic conditions showed an increased amount for some of the ribosomal proteins (although others remained unaffected). This may be explained by an increased membrane attachment of the ribosomes as a consequence of increased synthesis of the membrane protein BR rather than by an increased concentration of the ribosomal proteins themselves. It has been reported that the amount of ribosomes isolated with the membrane increases when BR synthesis is induced (51).

A different situation was found in the time course experiment where the amount of ribosomal proteins was found to linearly decrease over 24 h. Measurements of cell density showed that growth suddenly stopped when the high density cell culture was switched from aerobic to anaerobic/phototrophic growth conditions. Such a sudden stop will strongly effect protein biosynthesis and may result in degradation or reduced membrane attachment of ribosomes.

More than 75% of the quantified proteins did not show differences in their expression level. This is consistent with other quantitative proteomic studies on H. salinarum⁴ where only small protein subsets are found to be regulated. Upon a systematic proteomic approach by 2D gel electrophoresis we found that about 80% of all genes are expressed at a level to reach visibility on a 2D gel under just a single growth condition (36). Our interpretation is that halobacterial cells in their natural habitat are exposed to drastic changes of nutrition, oxygen, and light in relatively short periods of time. As a survival strategy it seems to be better to have the full set of proteins readily available than to be forced to synthesize these under shortage of energy or starving conditions. Such a strategy, which is not energy-efficient, may be well suited for organisms that inhabit ecological niches with reduced selective pressure and that are adapted to extreme environments. In times of high energy a full set of proteins is made to survive in times of low energy supply during which biosynthesis becomes difficult or impossible.

	ACKNOWLEDGMENTS

 
We are grateful to Monica Zobawa for excellent technical assistance.

	FOOTNOTES

 
Received, March 28, 2006, and in revised form, June 26, 2006.

Published, MCP Papers in Press, June 27, 2006, DOI 10.1074/mcp.M600106-MCP200

¹ The abbreviations used are: TMD, transmembrane domain; 16-BAC, 16-benzyldimethyl-n-hexadecylammonium chloride; 1D, one-dimensional; 2D, two-dimensional; BS, basal salt; ICPL, isotope-coded protein labeling; PMF, peptide mass fingerprinting; SILAC, stable isotope labeling with amino acids in cell culture; BR, bacteriorhodopsin; ABC, ATP-binding cassette.

² F. Siedler, unpublished data.

³ B. Bisle, C. Klein, F. Siedler, F. Pfeiffer, and D. Oesterhelt, unpublished data.

⁴ A. Tebbe, A. Schmidt, K. Konstantinidis, M. Falb, B. Bisle, C. Klein, J. Kellermann, F. Siedler, F. Pfeiffer, F. Lottspeich, and D. Oesterhelt, manuscript in preparation.

^* The work was supported in part by Bundesministerium fuer Bildung und Forschung Grants 031U108C/031U208C and 031U101C. 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.

^S The on-line version of this article (available at http://www.mcponline.org) contains supplemental material.

Both authors contributed equally to this work.

^** To whom correspondence should be addressed: Dept. of Membrane, Biochemistry, Max Planck Institute of Biochemistry, Am Klopferspitz 18, D-82152 Martinsried, Germany. Tel.: 49-89-8578-2386; Fax: 49-89-8578-3557; E-mail: oesterhe{at}biochem.mpg.de

	REFERENCES

TOP ABSTRACT EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

 

1. Wallin, E., and Von Heijne, G. (1998) Genome-wide analysis of integral membrane proteins from eubacterial, archaean, and eukaryotic organisms. Protein Sci. 7, 1029 –1038[Abstract]

2. Stevens, T. J., and Arkin I. T. (2000) Do more complex organisms have a greater proportion of membrane proteins in their genomes? Proteins Struct. Funct. Genet. 39, 417 –420[CrossRef][Medline]

3. Klein, C., Garcia-Rizo, C., Bisle, B., Scheffer, B., Zischka, H., Pfeiffer, F., Siedler, F., and Oesterhelt, D. (2005) The membrane proteome of Halobacterium salinarum. Proteomics 5, 180 –197[CrossRef][Medline]

4. Santoni, V., Molloy, M., and Rabilloud, T. (2000) Membrane proteins and p