M. P. Molloy, Department of Biological Chemistry, University of Michigan Medical School, Ann Arbor, MI, 48109-0606, USA. Fax: + 734 6470951, E-mail: email@example.com
Outer membrane proteins (OMPs) of Gram-negative bacteria are key molecules that interface the cell with the environment. Traditional biochemical and genetic approaches have yielded a wealth of knowledge relating to the function of OMPs. Nonetheless, with the completion of the Escherichia coli genome sequencing project there is the opportunity to further expand our understanding of the organization, expression and function of the OMPs in this Gram-negative bacterium. In this report we describe a proteomic approach which provides a platform for parallel analysis of OMPs. We propose a rapid method for isolation of bacterial OMPs using carbonate incubation, purification and protein array by two-dimensional electrophoresis, followed by protein identification using mass spectrometry. Applying this method to examine E. coli K-12 cells grown in minimal media we identified 21 out of 26 (80%) of the predicted integral OMPs that are annotated in SWISS-PROT release 37 and predicted to separate within the range of pH 4–7 and molecular mass 10–80 kDa. Five outer membrane lipoproteins were also identified and only minor contamination by nonmembrane proteins was observed. Importantly, this research readily demonstrates that integral OMPs, commonly missing from 2D gel maps, are amenable to separation by two-dimensional electrophoresis. Two of the identified OMPs (YbiL, YeaF) were previously known only from their ORFs, and their identification confirms the cognate genes are transcribed and translated. Furthermore, we show that like the E. coli iron receptors FhuE and FhuA, the expression of YbiL is markedly increased by iron limitation, suggesting a putative role for this protein in iron transport. In an additional demonstration we show the value of parallel protein analysis to document changes in E. coli OMP expression as influenced by culture temperature.
Escherichia coli has been adopted as a model organism for structural and functional studies aimed at understanding the biophysical and biochemical organization of proteins in Gram-negative cell walls. The cell walls of these bacteria comprise of three morphologically defined layers . The innermost layer is the cytoplasmic membrane that regulates passage of metabolites into and out of the cytoplasm. External to this membrane, a thin peptidoglycan layer encases the cytoplasmic membrane providing mechanical rigidity. The most exterior layer, the outer membrane is an additional barrier consisting of lipids, polysaccharides and proteins that easily distinguishes Gram-negative from Gram-positive bacteria that lack this additional membrane.
Studies of the outer membrane has revealed the presence of both integral and lipid-linked membrane proteins. In an early account using two-dimensional electrophoresis with carrier ampholyte generated pH gradients, Ames and Nikaido  provided an indication of the complexity of the E. coli cell envelope, reporting 150 proteins from this preparation. Some bacterial membrane proteins are expressed in high copy number (e.g. OmpA is present at 105 copies per cell ), which have aided their detection and biochemical characterization. These include the porins (e.g. OmpC, OmpF), a family of integral outer membrane proteins (OMPs) that form a hydrophilic channel, allowing nonspecific diffusion of small molecules across the outer membrane barrier (reviewed in [4–6]).Proteins responsible for transport of specific nutrients have also been identified. These include channel-forming transmembrane proteins for the passage of phosphate (PhoE), maltose and maltodextrins (LamB), and nucleosides (Tsx). Other transmembrane proteins have been implicated as high affinity receptors for the transport of ferric iron (FepA, FhuA), vitamin B12 (BtuB), and fatty acids (FadL). Furthermore, some of these OMPs are target receptors for bacteriophages and colicins. Finally, due to their location as interfaces between the cell and the environment, OMPs are important candidate antigens for developing strategies aimed at providing protection against bacterial pathogens [7–10].
The elucidation in 1997 of the nucleotide sequence encoding the E. coli K-12 genome is an invaluable resource for microbiologists . One of the opportunities this achievement presents is the chance to fully define the protein composition of a bacterial outer membrane. As of July 1999, all but approximately 10% of the open-reading frames (ORFs) had been annotated in the SWISS-PROT protein database (A. Bairoch, personal communication). At the present time, examination of the predicted ORFs incorporated into SWISS-PROT Release 37, indicates that E. coli K-12 has the potential to express 62 OMPs (discounting OMPs derived from plasmids). Proteins are annotated as OMPs based upon literature reports, computer prediction programs or by sequence similarity with known related OMPs (A. Bairoch, personal communication). When the known lipoproteins, which are attached by a N-acyl diglyceride linkage are omitted, 39 integral or transmembrane proteins are predicted in the E. coli outer membrane. However, for many of these ORFs it is not known whether they are transcriptionally active, with the potential gene-products known only as ‘hypothetical’ proteins. Furthermore, it is of interest to microbiologists to understand which gene-products are expressed under various states of growth and environmental stresses. One approach to examine these issues is to isolate and identify the entire complement of membrane proteins in parallel using a combination of 2D electrophoresis and mass spectrometry.
Until recently such an approach has been impractical due to technical difficulties in both extracting and solubilizing membrane proteins. This issue was highlighted by investigators recording proteins identified on 2D gels, and comparing these to proteins predicted by genome sequencing . Interestingly, for each of the three model organisms examined (E. coli, Bacillus subtilis, Saccharomyces cerevisiae) the majority of identified proteins displayed hydrophilic hydropathies, while hydrophobic proteins were vastly under-represented on these same gels. In addition, examination of current E. coli 2D gel reference maps (e.g. SWISS-2DPAGE) indicates the absence of many predicted OMPs. The confounding element in this case is that unlike the hydrophobic protein cluster missing from 2D gels , E. coli OMPs possess hydrophilic hydropathy values comparable to many cytoplasmic proteins that are easily separable by 2D electrophoresis. This suggests that extraction and solubilization of OMPs prior to 2D electrophoresis is an important area that requires exploration.
Solving this aspect of the puzzle is important, as excellent progress in downstream steps has been achieved. These include improved solubilization of some intractable proteins, combined with reproducible, high loading 2D gels [13,14] and improved protein microanalysis techniques such as mass spectrometry [15,16], intertwined with powerful database identification packages. This science has been formalized as ‘proteomics’. Put simply, proteomics is a formalized approach for obtaining a rapid ‘snap-shot’ of the protein complement of a tissue, cell or cell component [17,18]. Such an approach is powerful in that it allows a parallel assessment of temporal protein fluxes. This is an important concept in view of the dynamic nature of protein expression. Undoubtedly, changes in protein expression are essential in any study aimed at investigating cellular networks.
In this report we describe a proteomic approach to study the protein composition of the E. coli outer membrane. We propose a simple method for the rapid isolation, separation by 2D electrophoresis and identification of predominantly integral OMPs. We compare predicted and experimentally identified E. coli OMPs and evaluate the potential of a protein-based approach for large-scale analyses of membrane proteins. In addition we present two cases demonstrating the value of a proteomic approach for generating insight into protein function. In the first instance we propose a role for a previously hypothetical protein (YbiL) as an iron-binding protein related to FoxA, while in the second experiment we identify two temperature inducible OMPs. Finally, the significance of a proteomic approach for the study of bacterial OMPs is discussed.
Materials and methods
E. coli cell culture
Unless stated otherwise, E. coli K-12, strain W3110 were grown aerobically at 37 °C with shaking (200 r.p_m. in 2 L baffled flasks; Nalgene, NY, USA) from a 1 : 100 dilution of overnight culture to late exponential phase at a D420 of 1.5.
Culture medium was buffered with 50 mm Mes, 20 mm KOH adjusted to pH 6.6 with NaOH. The buffer was supplemented with: 6 g·L−1 glycerol, 1.2 g·L−1 NH4Cl, 2 mm K2HPO4, 0.5 mm CaCl·2H2O, 0.5 mm MgCl·7H2O, 18 µm FeNa EDTA.H2O and trace elements as described . Immediately before use the medium was supplemented with 5 µg·mL−1 thiamine.
For the iron limitation experiment the concentration of FeNa EDTA.H2O in the culture medium was reduced 100-fold to 0.18 µm. All other conditions were unchanged.
Cells were collected by centrifugation at 2500 g for 8 min. The cells were resuspended in 50 mm Tris/HCl, pH 7.3 for washing, then pelleted by centrifugation at 2500 g for 8 min. The equivalent of 400 mL of culture was resuspended in 5 mL of 50 mm Tris/HCl, pH 7.3 and stored at −80 °C until required.
P. aeruginosa cell culture
P. aeruginosa strain PAO1 was purchased from The University of New South Wales Culture Collection, Sydney, Australia. An overnight culture was diluted 1 : 100 and grown in M9 minimal media using 2% (w/v) glucose as a carbon source at 37 °C with shaking (200 r.p_m.) to D420 of 1.5. Cells were collected by centrifugation (6000 g for 30 min at 4 °C) and washed twice with NaCl/Pi. The equivalent of 400 mL of culture was resuspended in 5 mL NaCl/Pi and stored frozen at −80 °C until required.
Protein quantitation was determined by Fmoc amino acid analysis as previously described .
Isolation of cell membranes by carbonate extraction
This method was modified from Fujiki et al. . Twenty milligrams of cellular protein (as determined by amino acid analysis) was diluted to 6 mL with 50 mm Tris/HCl, pH 7.3 containing 0.7 mg of DNase I (Sigma). The cells were ruptured in an Aminco French press with two presses at 9.65 × 107 Pa and the unbroken cells removed by centrifugation at 2500 g for 10 min. The supernatant was diluted with ice cold 0.1 m sodium carbonate (pH 11) to a final volume of 60 mL and stirred slowly on ice for 1 h. The carbonate treated membranes were collected by ultracentrifugation in a Beckman 55.2 Ti rotor at an average of 115 000 g for 1 h at 4 °C. The supernatant was discarded and the membrane pellet resuspended and washed in 2 mL of 50 mm Tris/HCl, pH 7.3. The pellet was collected by centrifugation at an average of 115 000 g for 20 min (Beckman 55.2 Ti rotor) and solubilized for 2D electrophoresis with 1.5 mL of IEF solution [7 m urea, 2 m thiourea, 1% (w/v) ASB14 , 40 mm Tris, 2 mm tributyl phosphine (TBP) and 0.5% (v/v) Biolytes 3–10 (Bio-Rad, CA, USA)].
An 18-cm pH 4–7 immobilized pH gradient (IPG) (Amersham-Pharmacia) was rehydrated overnight with 430 µL of IEF solution that contained 500 µg of the solubilized membrane proteins. Isoelectric focusing was conducted for 30 000 V·h at a maximum of 5000 V using the Multiphor II system (Amersham-Pharmacia, Sweden). For the second dimension, the IPG was equilibrated for 20 min by rocking in a solution of 0.15 m bisTris/0.1 m HCl, 6 m urea, 2% (w/v) SDS, 20% (v/v) glycerol, 5 mm TBP, 2.5% (w/v) acrylamide. The IPG was then embedded onto a 10% SDS/PAGE gel with 1% (w/v) agarose, 0.4% (w/v) SDS, 0.15 m bisTris/0.1 m HCl. Gel buffer consisted of 0.179 m Tris/0.1 m HCl, cathode buffer contained 0.2 m taurine/25 mm Tris, 0.1% (w/v) SDS and the anode buffer was 0.384 m glycine/50 mm Tris, 0.1% (w/v) SDS . Gels were run at 5 °C and 25 V for 2 h then 85 V overnight until complete. The gels were stained for 24 h with colloidal Coomassie G-250 (Merck, Germany) as previously detailed  and were imaged using a Molecular Dynamics laser Personal Densitometer SI.
Peptide mass fingerprinting (PMF)
Protein spots were excised from the gel using the ARRM 214 spot excision robot (ARRM, Adelaide, Australia) and delivered into a polypropylene 96 well microtitre plate (Nunc). Tryptic digestion and MALDI-TOF MS was performed as previously described . The spectra was calibrated using the porcine trypsin auto-digestion peptide at 2211.09 Da [M + H]+. Peptide masses were searched against the SWISS-PROT Release 37 protein database using the PeptIdent program (http://expasy.proteome.org.au/tools/peptident.html) or NCBInr database using the MS-Fit program (http://prospector.ucsf.edu) using a mass cut-off of 100 p.p_m.
Predicted E. coli OMPs
SWISS-PROT Release 37 was interrogated to retrieve all E. coli K-12 proteins annotated as OMPs (including hypothetical proteins). Proteins derived from plasmids were not considered. Unless annotated otherwise, it was assumed each protein would be expressed under the culture conditions described. Table 1 lists the 58 potential OMPs and is subdivided into 4 sections: predicted integral membrane proteins with pI 4–7 (37 proteins), predicted integral membrane proteins pI > 7 (2 proteins), predicted outer membrane lipoproteins pI 4–7 (10 proteins), predicted outer membrane lipoproteins pI > 7 (9 proteins). No E. coli OMPs are predicted with a pI less than 4.
Table 1. OMP list. Sixty-two E. coli proteins are annotated as outer membrane proteins in SWISS-PROT release 37. 58 of these could be expressed under the described culture conditions. The list was subdivided into 4 sections: integral OMPs pI 4–7, integral OMPs pI > 7, outer membrane lipoproteins pI 4–7, outer membrane lipoproteins pI > 7. Hypothetical proteins known only from an open reading frame are shown in bold. Proteins in italic were identified following carbonate treatment, separation by 2D electrophoresis and PMF. GRAVY values indicate relative hydrophobicity. Increasing GRAVY indicates increasing hydrophobicity. In total 78% of all predicted integral OMPs (excluding hypotheticals) were identified.
Integral OMPs including hypothetical proteins
Bacteriophage N4 adsorption
Hypothetical TonB dependent receptor
Colicin I receptor
General secretion pathway protein D
Vitamin B12 receptor
Antigen 43 Fluffing protein beta chain
Antigen 43 Fluffing protein alpha chain
Long-chain fatty acid transport
General secretion pathway
Integral outer membrane proteins pI > 7
Outer membrane lipoproteins
Putative Polysaccharide export protein
Hypothetical polysaccharide export
Outer membrane lipoproteins pI > 7
Murein transglycosylase D
Murein transglycosylase C
Murein transglycosylase A
Murein transglycosylase B
Outer membrane lipoproteins pI > 7 (cont.)
Major outer membrane lipoprotein
SWISS-PROT Release 37 contains a number of additional predicted E. coli lipoproteins that were not included in Table 1 due to incomplete annotation describing their subcellular location. Four integral proteins predicted by the DNA sequence were not included in Table 1 as these were not induced under the culture conditions. These four proteins were HtrE, a porin-like protein induced by heat-shock ; PhoE, a porin induced by phosphate limitation ; OmpG, a porin expressed after a rare chromosomal deletion (termed cog-192) of 13.1 kb immediately upstream of OmpG ; FecA, induced by citrate .
Of the 37 integral membrane proteins with pI 4–7, none were predicted to be less than 10 kDa, while 26 were predicted to lie within the typical SDS/PAGE separation window of 10–80 kDa. Four of these proteins were hypothetical translations predicted from ORFs. Above 80 kDa, 8 out of 11 proteins are known only as hypothetical proteins.
Separation and identification of OMPs
Of the 20 mg of cellular protein that was used for the carbonate extraction, 1.2 mg (± 0.1 mg, n = 4) protein was recovered in the insoluble pellet after carbonate extraction. This represents 6% of the cellular protein. Figure 1 shows a micropreparatively loaded (500 µg) 2D gel of carbonate insoluble E. coli membrane proteins. Approximately 100 protein spots were observed, including various intensely stained spots such as OmpA, known to be present at 105 copies per cell . Many of the OMPs resolved into at least two-charged isoforms, although in some cases only a single charged species was observed (e.g. OmpX), while for others (e.g. OmpA), multiple charged isoforms exist. Presently, we have been unable to conclusively establish the nature of the charge variance, although we have observed some deamidated peptides that may contribute to protein heterogeneity. Furthermore, OmpX that was observed only as a single spot does not contain a typical Asn-Gly deamidation site.
Overall, good spot resolution was obtained with only minimal streaking, indicating high protein solubility attributed to the use of the surfactant ASB14 . We have also tested the separation using an IEF solution that contained a commercially available sulfobetaine surfactant (SB 3–10) that had shown good results with some E. coli OMPs . However, whilst we recovered the dominant OMPs (including OmpA, OmpX, OmpF, OmpC), the solubilization was less effective using SB 3–10, with the overall amount of protein lower and many of the less abundant OMPs missing (data not shown).
All Coomassie blue stained protein spots were excised from the gel (Fig. 1), digested in situ with trypsin and subjected to PMF for identification (e.g. Figure 2). The identified proteins are labelled on Fig. 1 and listed in Table 2. In most cases, mass spectra showed two or more peptides that were matched to a SWISS-PROT E. coli database homologue, thus establishing protein identity. The very high mass accuracy obtained (commonly less than 50 p.p_m.), plus the high number of peptides matched provides strong confidence in our assignments. In some cases protein identification was compromised by unreliable peptide spectra (either few peptides or poor spectra). Most often this was associated with weakly stained Coomassie blue spots, reflecting the decline in analytical sensitivity with decreased protein analyte. Interestingly, on a few occasions a heavily stained spot would yield poor spectra, or in approximately 5% of analyses, good peptide spectra did not result in a database match of two or more peptides to the same protein.
Table 2. E. coli proteins identified fromFig. 1following PMF. With the exception of flagellin, all highly abundant proteins were identified as integral OMPs. Five outer membrane lipoproteins were also identified. Ten other proteins not described as membrane proteins were identified and may represent proteins that associate with membranes. The number in parentheses indicates the number of additional peptides matched with a mass < 1000 Da or > 2200 Da.
In total, 40 unique proteins were identified. 75% of these proteins are annotated in SWISS-PROT as either membrane proteins or membrane associated. Of the 26 integral OMPs predicted to lie within the separation window of pH 4–7 and Mr 10–80 kDa, we identified 18 of these on the single gel (Fig. 1). This constitutes 69% of the integral OMPs predicted in this window. We also confirmed the synthesis of two of the four hypothetical predicted proteins in this separation window (YbiL, YeaF).
As indicated in Table 2, a range of functionally diverse integral OMPs including porins, specific transport proteins, enzymes and two previously hypothetical proteins were identified. In addition, 5 outer membrane associated lipoproteins were identified, as well as three proteins (AcrA, AtpB, NuoC) associated with the cytoplasmic membrane. With the exception of flagellin, the remaining nonmembrane proteins were detected in far lower amounts than the integral OMPs. Nonetheless, it may be important to consider that some of these proteins of lower abundance may have a functional association with the membrane, and should not be viewed as poor separation, as no abundant cytoplasmic marker proteins (e.g. DnaK, LivJ, EF-Tu) were detected.
Induced expression of iron receptors
To account for the missing integral OMPs we proposed two possibilities; some of the proteins may not be amenable to the isolation and separation procedure, or alternatively some proteins may not be expressed in sufficient amounts to enable their detection and identification. Three of the proteins that were expected, although not observed in Fig. 1 were annotated in SWISS-PROT as iron receptors. Therefore we designed experiments to induce expression of these proteins by reducing the concentration of iron in the culture media. Figure 3 displays a portion of the 2D gel of the carbonate insoluble E. coli OMPs from bacteria grown under conditions of iron limitation. Three new, intensely stained protein spots were detected, and analysed by PMF. We identified these three OMP as iron receptors (CirA, FhuE, FepA), each not identified in Fig. 1. In this case it is clear that the absence of the iron receptors from the initial gel (Fig. 1) is due to low expression levels and not due to technical difficulties in isolation and separation of these proteins.
Apart from the known iron receptors, little effect was seen to influence the expression of the remaining OMPs upon iron restriction. However, the expression of the previously hypothetical protein YbiL was an exception, as it was greatly stimulated by growth under iron restriction, implicating YbiL as a putative iron receptor. Furthermore, a BLAST 2 search indicates that YbiL displays closest homology with iron receptors from numerous species and a molecular mass consistent with other E. coli iron receptors (≈ 80 kDa) (Table 3). In summarizing the findings using the two culture conditions, we identified 80% of the predicted integral OMPs annotated in SWISS-PROT Release 37, predicted to separate in the window of pH 4–7 and 10–80 kDa.
Table 3. Top 10 SWISS-PROT matches to E. coli YbiL using Blast 2 sequence matching. The molecular mass of YbiL (78.3 kDa) is consistent with other E. coli iron receptors.
Swiss-Prot accession No.
Ferric-Pseudobactin M114 Receptor
Ferrioxamine B Receptor
Ferric-Pseudobactin 358 Receptor
Outer-Membrane Receptor For Fe(III)-Coprogen, Fe(III)-Ferrioxamine B, Fe(III)-Rhodotrulic Acid
Iron(III) Dicitrate Transport Protein
Isolation and separation of OMPs from P. aeruginosa
In order to examine the validity of this method to isolate OMPs from other Gram-negative bacteria we tested the procedure with P. aeruginosa. The proteome map of OMPs isolated by carbonate treatment is shown in Fig. 4. We chose five on the most intense Coomassie blue stained protein spots and analysed them by PMF. The identities of these proteins, as well as the closest matching E. coli protein are listed in Table 4. All five of these proteins are described as integral OMPs, providing compelling evidence for the reproducible action of this isolation procedure when applied to P. aeruginosa, and offers potential for similar results in related bacteria. Currently, a more thorough investigation of the proteins on this gel is indicating very high coverage of OMPs (A. Nouwens, M. Molloy, S. Cordwell, M. Larsen, M. Willcox, M. Gillings & B. Walsh, unpublished results).
Table 4. Identity of five P. aeruginosa proteins labelled inFig. 4following PMF. A FastA search was conducted to determine the closest matching E. coli protein. The number in parentheses indicates the number of additional peptides matched with a mass < 1000 Da or > 2200 Da.
Number of tryptic peptides predicted between 1000 and 2200 Da considering no missed cleavages.
Outer membrane Porin F
Temperature regulated OMP expression
It is well documented that expression of the E. coli porins OmpC and OmpF are regulated by temperature . However, in regards to expression of most other OMPs their temperature-dependent status is unclear. To investigate this issue we used 2D electrophoresis to observe the effect of growth temperature upon OMP expression. Cells were cultured at either 21 °C or 37 °C and their OMPs isolated as described above. A portion of the 2D gels displaying the most striking changes in protein expression is shown in Fig. 5. A highlight of this experiment is the absence of three proteins when cells are cultured at 21 °C compared to 37 °C growth. These proteins were identified by PMF as the cell surface protease OmpT and the E. coli specific, dual-subunit OMP, Antigen 43 (α and β subunits). Increases or decreases in expression of some minor proteins were also observed, although moreover the protein pattern between these samples was highly reproducible. Clearly, this approach is an ideal method to identify candidate molecules whose expression alters with environmental changes. This is an important first step that facilitates more direct experimentation to establish the roles of these proteins.
We have developed a simple and rapid method for the isolation and separation of OMPs from E. coli. The most abundant proteins isolated with this technique were identified as transmembrane cell surface proteins, many previously missing from E. coli 2D gel reference maps. Significantly, the isolation and separation procedure appeared equally applicable when trialed with the Gram-negative microbe P. aeruginosa. Furthermore, this protocol is amenable to micropreparative separations. This is important in the context of high-throughput analysis, where each protein spot can then be subjected to a microanalytical identification/characterization technique, removing any need to pool identical protein spots from multiple gels. We view this study as an important first step to encourage proteomic-based investigations of bacterial membranes. As we have demonstrated such an approach can provide profound results (i.e. identification of a new iron receptor) even with widely studied model species such as E. coli.
One of the key procedures for the isolation of integral OMPs is incubating the cells in an alkaline solution of sodium carbonate. This approach avoids the need to isolate membranes by traditional methods such as density centrifugation, which is often time-consuming. In the initial report describing carbonate treatment, Fujiki et al.  reported that this procedure converted closed microsomal vesicles into open sheets, concomitantly stripping loosely attached peripheral proteins. In contrast to eukaryotic glycosylphosphatidylinositol linked proteins that are removed with carbonate treatment, a number of E. coli N-acyl diglyceride attached lipoproteins were recovered with the integral OMPs, albeit in much lower abundance than the integral OMPs (Table 2). It is anticipated that both the OM and CM are insoluble with this treatment. Nonetheless, our 2D gel indicates that OMPs are preferentially solubilized by the IEF solution, with proteins of the CM (predicted with considerably higher hydrophobic characteristics) most likely remaining insoluble. This is despite using a solubilizing solution amenable to 2D electrophoresis that we consider contains some of the most powerful reagents, outside of SDS. SDS/PAGE of the insoluble pellet showed the presence of some additional bands not detected in a sample that provided Fig. 1 (data not shown).
We recovered 21 out of 37 of all the predicted E. coli integral OMPs between pH 4–7 and independent of molecular mass. This takes into account 10 hypothetical proteins that were not observed. If one considers that these ORFs are not transcriptionally active under the conditions described or are not authethic transcripts, then we identified 78% of the integral OMPs with known transcriptionally activity. This data supports the view that during exponential growth in a defined minimal media, E. coli K-12 transcribes and translates greater than three-quarters of its genes encoding authentic integral OMPs. In addition, it is important for protein-based analyses of membrane proteins to note this level of protein coverage. We view this result as encouraging, and anticipate that with correct experimental design the large-scale analysis offered by proteomics will support investigation of complex biological networks.
At this point it is prudent to offer explanations for the missing 22% of predicted proteins. Large molecular mass proteins (> 80 kDa) are commonly missing from micropreparative 2D maps. The three large predicted proteins not detected here (NfrA, 108 kDa; FimD, 92 kDa; SfmD, 92 kDa) may have been lost in the separation procedure, either at the point of IPG reswelling or during IEF by precipitation at the isoelectric point. Of the remaining three unidentified proteins less than 80 kDa, OmpN is known as a quiescent porin, whose expression is diminished by the presence of the classical porins, OmpC and OmpF . It is likely the levels of expression of OmpN are too low for detection in the W3110 strain used in this experiment. Finally, GspD and HofQ, both annotated as general secretion pathway proteins are known only from their DNA sequence. These proteins may not be expressed in minimal media where secretion may be low, or alternatively are only lowly expressed thereby escaping detection. Of the hypothetical proteins, clearly further work is required to understand their absence. Monitoring of mRNA expression may confirm if these proteins are transcriptionally active, and therefore could be expected on 2D gels. We were surprised that we did not identify YiaT (pI 4.48, 25 kDa) as it is expected in a region of the gel where its presence should be obvious. This may indicate YiaT is not expressed, or at best lowly expressed in W3110 under the growth conditions described here.
In an interesting aspect of this study we monitored OMP expression under conditions of iron limitation. This led to the observation of increased expression of a protein (YbiL) with no previous known function. This response was in accord with expression increases from other known outer membrane iron binding proteins and implicates YbiL as an iron receptor. This approach of monitoring comparative expression levels has been valuable, enabling us to postulate a role for a previously hypothetical protein. Importantly, this opinion is supported by amino acid sequence similarity with other proteins of known function (Table 4). This provides an example of how proteomics can be used to enhance and strengthen information generated by genomics. In a second experiment we were able to readily document changes in protein composition of the outer membrane as cells were cultured at different temperatures. This approach identified an OMP (Ag43) that had previously shown phase-variable expression , although it had not been reported that Ag43 is repressed at low growth temperatures. This provides one example of how such an approach is a useful first step to identify interesting proteins prior to launching more direct functional studies. The protease OmpT was also missing from cells cultured at 21 °C. In this case, previous reports have shown OmpT to display temperature dependant expression [33,34]. Thus, our approach has demonstrated both discovery and confirmation of a previous finding.
The 2D gel (Fig. 1) exhibits a unique library of OMPs. We are of the opinion that there is great potential for uses of such a library in biochemistry and biotechnology. For example, the 2D gel pattern represents a bacterial ‘finger-print’ that may be useful in discerning bacterial species. Additionally, these proteins could be tested for their ability to interact with various ligands. To extrapolate from the current report, it should be possible to create a similar library of OMPs from many other Gram-negative bacteria. Even for species with undefined genomes, the near exclusive isolation of integral OMPs using this method means an abundant protein spot on a gel is highly likely to be an integral OMP and is worth the effort to identify and characterize it, albeit using techniques that are slower than PMF (i.e. Edman sequencing) or more complex (i.e. MS/MS). Continuing with this argument, for the immunologist there is enormous potential to use this approach to ‘fast-track’ identification of likely candidate antigens that may elicit immune responses in invaded host organisms. Integral and lipid-linked OMPs may be key molecules in designing strategies to guard against bacterial infection.
This research has been facilitated by access to the Australian Proteome Analysis Facility established under the Australian Government's Major National Research Facilities Program. M. P. M. is the recipient of an Australian Proteome Industry Research and Development postgraduate scholarship. We acknowledge the expertise of Michel Satre for assistance during the initial phase of this work. We thank Femia Hopwood, Peter Hains and Angela Connolly for their contribution to this study.