E. coli strains (Table 1) were grown at 37°C in TY medium containing 1% Bacto trypton, 0.5% Bacto yeast extract, 0.5% NaCl and 3 mM NaOH. Expression of the constructs was induced by adding up to 1 mM IPTG or 0.02% L-arabinose, depending on the plasmid vector. Antibiotics were ampicillin (100 µg/ml) or Chloramphenicol (25 µg/ml). LMC500 (MC4100 lysA) was made chemically competent using the calcium chloride method. MC1061 and its derivative MC1061 ΔOmpA were transformed using electroporation.

All constructs (Table 1) were cloned into a pTrc99A vector (Pharmacia Biotech, USA), a pBR322 derivative plasmid, of which the trc promoter was modified with a down mutation to reduce expression levels [16]. All DNA manipulation, analysis and bacterial transformations were performed according to standard protocols (Sambrook et al., 1989). All PCR fragments were sequenced, either at Baseclear (Leiden) or at the AMC DNA sequencing facility (Amsterdam Medical Centre). Primers were ordered from MWG or Biolegio, and Advantage DNA polymerase (Clontech) or pfuTurbo DNA polymerase (Stratagene) was used for the PCR reactions. The cloning steps performed to obtain the plasmids are described in the Materials and Methods S1.

Fresh overnight cultures grown at 37°C were diluted 1000x into 50–100 ml fresh TY medium and cultured at 37°C. Growth was monitored by measurement of the optical density at 600 nm with a spectrophotometer (Perkin-Ellmers). IPTG was added at around an OD600 of 0.1, and when the cells reached an OD600 of 1.0, they were transferred to a 50 ml Falcon tube and put on ice. The cells were then collected by centrifugation for 15 min at 4000 rpm in a tabletop centrifuge at 4°C (Eppendorf). The supernatant was carefully removed, and the cells were resuspended in ice-cold sonication buffer (10 mM Tris-HCl buffer, pH 7.9, supplemented with 1 mM EDTA and 1 tablet of Roche Protease Inhibitor Cocktail), at a concentration corresponding to an OD600 of 250. This cell suspension was transferred to a 2 ml Eppendorf tube, and sonicated on ice with a tip sonicator (Branson) in 4–5 10-second bursts with 10 second cooling in between each burst. Debris and intact cells were pelleted in a 4°C cooled centrifuge at 2700 x g for 2 min. The supernatant was transferred to a 1.5 ml Eppendorf tube and frozen at −20°C as total cell lysate.

After thawing, the cell lysate was diluted to 4 ml (corresponding to an OD600 of 12.5), and 100 µl of this was saved as “total cell lysate”. The samples were pelleted at 45000 rpm (corresponding to 200.000 x g) for 45 min in an ultracentrifuge (Beckman-Coulter). After centrifugation, 500 µl was saved as “supernatant”. The membrane pellet was resuspended in 100 µl sonication buffer and frozen at −20°C.

For SDS-PAGE, samples were mixed with sample buffer (end concentration: 62.5 mM Tris pH 6.8, 2% SDS, 10% glycerol, 2% 2-mercaptoethanol) and either heated to 99°C for 5 min or heated to 50°C for 15 min and electrophoresed on 15% polyacrylamide slabs. Anti-FLAG and anti-myc monoclonal antibodies used for the immunoblots were obtained from Sigma and Roche, respectively. The polyclonal anti-OmpA antibody was a kind gift from A. Driessen (University of Groningen, Netherlands). The bands were detected using the ECL+ chemiluminescence kit (Amersham) and scanning with a STORM 860 fluorescence imager. Densitometry was performed using ImageJ (http://rsb.info.nih.gov/ij/). The mean pixel value of a rectangular region was calculated close but outside a band of interest to calculate the mean background pixel value. The same selection rectangle was positioned to include the band of interest, and again a mean pixel value is calculated. Subtraction then gives a band intensity value. All band comparisons were performed using the same selection rectangle.

Cells were fixed in 2.8% formaldehyde (FA) and 0.04% glutaraldehyde (GA) in growth medium for 15 min at room temperature, then washed and resuspended in PBS (140 mM NaCl, 27 mM KCl, 10 mM Na₂HPO₄·2H₂O, 2 mM KH₂PO₄ pH 7.2). Cell concentration was adjusted to an OD600 of 0.6 and samples were incubated in 75 µl PBS containing 30 mg/ml BSA to block non-specific sites on the cell surface for 30 min at 37°C. Then antibodies were added, either anti-FLAG (M2, Sigma) or anti-myc (9E10, Roche) at an end concentration of 20 µg/ml, and samples were incubated at 37°C for 30 min. The cells were washed 3 times with 2 volumes of PBS containing 30 mg/ml BSA, and then incubated in 1 volume with Donkey-anti-Mouse-Cy3 conjugate (Jackson ImmunoResearch) at 10 µg/ml end concentration for 30 min at 37°C, washed 3 times with 2 volumes PBS and imaged.

Cells were put on ice, and an amount of cells equivalent to 1 ml OD600 of 0.3 (around 2·10⁸ cells) was taken for labeling. Cells were collected in all cases by centrifugation at 20.000 x g for 5 min at 4°C. The pellet was resuspended in 75 µl PBS at room temperature (RT) with 0.1% BSA. The cells are left at RT for 10 min to block aspecific sites on the cell surface. Then either biotinylated anti-FLAG (Sigma) was added (50 µg/ml) (FLAG constructs), or streptavidin-Alexa 488 (Molecular Probes) was added directly (40 µg/ml) (SA-1 constructs). Cells were incubated at RT for 30 min. The cells were spun down and washed twice with 0.5 ml PBS, and resuspended in 150 µl PBS. For the cells labeled with biotinylated FLAG, streptavidin-Alexa 546 (Molecular Probes) was added (5 µg/ml), and samples were incubated for 30 min at RT. Then, PBS (0.85 ml) was added and the cells were pelleted. After a second wash with 0.5 ml PBS, the cells were fixed in 1 ml PBS with 2.8% formaldehyde and 0.042% glutaraldehyde, washed in 1 volume of PBS and resuspended in 0.1 volume PBS. The cells were either imaged directly or stored at 4°C over night before imaging.

Cells were immobilized on 1% agarose in water slabs-coated object glasses as described by [17] and photographed with a CoolSnap fx (Photometrics) CCD camera mounted on an Olympus BX-60 fluorescence microscope through a UPLANFl 100x/1.3 oil objective (Japan). Images were taken using the public domain program Object-Image2.19 by Norbert Vischer (University of Amsterdam, http://simon.bio.uva.nl/object-image.html), which is based on NIH Image by Wayne Rasband. In all experiments the cells were first photographed in the phase contrast mode. Then a fluorescence image was taken using either a green excitation/red emission (U-MNG, ex. 530–550 nm), or a blue excitation/green emission filter cube (U-MNB or EGFP, ex. 470–490 nm).

A topology model of the transmembrane domain of OmpA is shown in Figure 1. For locations in loop 2 and loop 4 (after Y63, G70 and I153 respectively), it has been shown that small (up to 21 residue) peptides can be inserted without any reduction in protein levels [3], [18] and membrane incorporation (after G70 and I153, [3]). For loop 2, reported inserted peptides are listed in Table 2. Initially, we used only the OmpA TM domain, but later also the periplasmic domain was added. 3xFLAG and 2xmyc peptides were chosen as epitope tags (Table 2). We will refer to them as FLAG and myc from now on. High-affinity monoclonal antibodies are commercially available for these epitopes. SWISS-Model [19] was used to predict OmpA folding after peptide insertion. First, a continuous model was generated of the first 176 residues, based on the published crystal structures of OmpA-171 (Figure S1A). Then, models were generated of loop insertions after different residues in the protein, and the resulting (static) loop conformations were examined for their propensity to extend away from the membrane normal axis. Models were generated of insertions in loop 2, 3 and 4. Finally, loop 3 was chosen, since the computer-generated model of an inserted FLAG peptide after N109 predicted the largest distance away from the surface (Figure S1C).

To be able to compare the performance of our constructs with results reported in literature, we also constructed loop 2 insertions with the FLAG and myc epitopes. For loop 2, insertions after G65, G70, Y72 and Q75 were modeled. In the end G70 was chosen, because (a) as mentioned, it was shown that at this location, a 21-residue peptide could be inserted without any negative effects on membrane insertion [3], and (b) modeling with SWISS-Model of these four locations predicted that at G70, the loop would be extended away from the surface more than at the other three positions in loop 2 (Figure S1B).

During the course of this work, a loop 1 peptide insertion (SA-1 tag) in full-length OmpA was described in the literature [5]. The position of this insertion is also indicated in Figure 1 (after N26). As mentioned in the introduction, the SA-1 peptide tag (listed in Table 2) binds streptavidin directly with high affinity. Since this peptide was neutrally charged and could be conveniently labeled with fluorescent streptavidin, we decided to compare this loop insertion variant with our constructs. Therefore, it was cloned into our expression vector, both as a truncated OmpA consisting only of the TM domain and as a full-length protein, and assayed in the same way as the FLAG/myc insertions.

The constructs were tested for expression in LMC500 (MC4100 lysA), a well characterized laboratory strain [20], [21], MC1061 ΔOmpA [5], an OmpA knockout strain, and in its parental strain MC1061 [22]. To test to what extent the proteins could be expressed without affecting the growth rate, the growth was followed by measuring the optical density before and after addition of the inducer IPTG. Growth curves from a typical experiment are shown in Figure 2. In this experiment, cells carrying an empty vector were compared with cells expressing an OmpA-177 protein with a FLAG insertion either in loop 2 or loop 3. Without IPTG induction, all growth curves appear identical. Approximately 15 minutes after addition of IPTG however, the OmpA-177 protein with a FLAG insertion in loop 2 shows a lag phase where growth stops for approximately 30 min, after which growth continues normally. In contrast, for the growth curve of the OmpA-177 variant containing a FLAG insertion in loop 3, no lag phase is observed after induction with IPTG. Instead, the growth curve is similar to the growth curve of uninduced cells.

The induction experiment was performed for all the OmpA TM domain variants and the results are listed in Table 3. It can be concluded that in strain LMC500, induction of the loop 2 (FLAG/myc), and to a lesser extent, the loop 1 (SA-1), but not the loop 3 (FLAG/myc) insertions caused a lag phase of 30–60 min after which growth continued normally. Surprisingly, this lag phase was absent both in MC1061 and in MC1061ΔOmpA. Induction with 0.3 mM IPTG of the loop 2 (FLAG), or loop 1 (SA-1) in these strains had no effect on the growth rate at all. Because in LMC500 and MC1061, similar protein levels were detected on immunoblots, it appears that strain MC1061 and its derivative MC1061ΔOmpA are more able to cope with the perturbation caused by the induction (see also discussion).

Samples, harvested at the end of the induction experiment described above were analyzed by SDS-PAGE and immunoblotting. For the FLAG and myc constructs, a blot is shown in Figure S2. Based on the immunoblots it was clear that all four variants (either FLAG or myc in Loop 2 or 3) were expressed, no degradation bands were observed, and all ran at a similar height that was retarded with respect to their calculated molecular weight (approx. 3–4 kDa). This was also observed for full-length constructs that carried a FLAG insertion (see below). The reduced mobility on gel was attributed to the high amount of negative charge in the FLAG and myc peptides. Unexpectedly, unprocessed (precursor) proOmpA-177 was also detected in uninduced samples as well as in induced samples expressing either FLAG or myc in loop 3 (see also Figure S3). The difference in protein levels between induced and uninduced samples was a factor 2–5 fold (all protein levels were determined from immunoblot by densitometry using ImageJ, data not shown). Without induction, for FLAG as well as myc, about 25% of the loop 3 constructs detected were not processed.

Induced OmpA-177 without loop insertion was directly visible on a Coomassie Brilliant Blue (Coomassie) stained SDS-PAGE (see Figure 3) and could be detected on immunoblot using a polyclonal antibody against OmpA, but the band was very weak compared to endogenous (full-length) OmpA (data not shown). From this, we conclude that this polyclonal antibody primarily recognizes the periplasmic domain of OmpA. Also, an attempt was made to detect the streptavidin binding peptide SA-1 construct on blot using a Streptavidin-HRP conjugate, but only the endogenous cytoplasmic biotinylated biotin carrier protein BCCP was detected (data not shown). Apparently, streptavidin does not bind sufficiently strong to the denatured conformation of SA-1.

To be able to directly compare the amounts of protein in the membrane, membrane fractions were isolated of cells expressing the various OmpA TM domain variants (Figure 3). Induction of both OmpA-177 and OmpA-177 SA-1 constructs resulted in strong bands at roughly their expected height (calculated MW of OmpA-177 is 19.3 kDa and of OmpA-177 SA-1 is 21.1 kDa). Surprisingly, no bands could be identified that corresponded to the FLAG or myc constructs.

To determine how much of the expressed construct is present in the envelope fractions put on gel, we compared soluble (S) to membrane (M) fractions. Although not visible on a Coomassie stained gel, immunoblots of the fractions showed that the FLAG and myc constructs were present, and fractionated predominantly to the membrane fraction (Figure S3). We conclude that these constructs are present in the cell at greatly reduced amounts (less than 10% as judged from the Coomassie stained gel) compared to wild type OmpA-177 or the OmpA-177-SA1. This could either be due to a reduced synthesis rate (not expected for FLAG and myc tags), due an increased degradation rate (see also below) or both.

To our knowledge, all OmpA loop insertions reported in the literature have been made in full-length OmpA. To establish whether the periplasmic domain might play a role in preventing a reduction in protein levels after FLAG or myc insertion in the transmembrane domain, full-length constructs were made for the FLAG epitope in loop 2 and 3, and the SA-1 insertion in loop 1. These constructs were transformed to LMC500 and MC1061 ΔOmpA and were subjected to the same induction experiment as for the OmpA TM domain variants. Without induction, all full-length constructs grew normally in both genetic backgrounds. After induction, the growth rate remained unaffected for all full-length OmpA loop insertion constructs in the MC1061 ΔOmpA background whereas expression in LMC500 resulted in growth curves similar to those obtained for the OmpA-177 constructs (only tested for the loop 3 FLAG insertion).

The full-length constructs were detected both on Coomassie stained gel (Figure 4A) as well as on immunoblot using a polyclonal antibody that recognizes the periplasmic domain of OmpA (Figure 4B, Figure S4). As with the FLAG OmpA-177 proteins, full-length OmpA FLAG constructs had a higher apparent molecular weight than calculated. Comparing soluble to membrane fractions using immunoblots yielded similar results as for the OmpA-177 variants, with the majority of each construct fractionating to the membrane fraction, except loop 3, which was divided over the soluble and membrane fraction (Figure S4).

Again, the FLAG constructs appear greatly reduced compared to both the OmpA without insertion and the OmpA with an SA-1 insertion, all expressed from plasmid. The intensities of the anti-OmpA bands from Figure S4 were quantified using ImageJ and plotted as bar graphs in Figure 5. Together with Figure 4A, it can be concluded that full-length OmpA constructs with FLAG insertions either in loop 2 or loop 3, are present at approximately 5–10% compared to full-length OmpA without insertion or the full-length OmpA carrying a SA-1 insertion, respectively.

Apart from intact construct, the immunoblots probed with anti-OmpA antibody revealed a strong band around 17 kDa (the expected size of the periplasmic domain) that fractionates to the soluble fraction (Figure 4B). This band was absent from samples containing induced full-length OmpA without insertion and weakly detected in induced full-length OmpA with the SA-1 insertion (data not shown). From this we conclude that the degradation band is specific and that it is related to the reduced protein levels of the FLAG constructs. In addition, the FLAG constructs were detected using anti-FLAG antibody (Figure S4 and not shown). The 17 kDa degradation band did not react with anti-FLAG (the smallest degradation band detected with anti-FLAG ran around 27 kDa). Apparently, the periplasmic domain of OmpA was cleaved from the TM domain and the latter was to a large extent degraded.

To test whether the 17 kDa degradation band was present in periplasmic aggregates, we fractionated the cells into soluble periplasmic contents and insoluble pellet by osmotic shock (See Figure S6 and Materials and Methods S1). We find that the degradation band was not aggregated but present in the soluble periplasmic fraction, consistent with the idea that the FLAG/myc OmpA variants are to a large extent degraded by periplasmic proteases.

Taken together, it can be concluded that both the full-length and the OmpA-177 constructs with a FLAG insertion are present in greatly reduced amounts, compared to without insertion or with a SA-1 insertion. Therefore, addition of the periplasmic domain does not improve protein levels of the FLAG insertion. Furthermore, the reduction in protein levels is, at least partly, caused by degradation, as observed on immunoblots.

To study whether the OmpA variants, present in the membrane fraction, have obtained their native form, we examined their heat-modifiability. The native form of the OmpA TM domain is a compact β-barrel that has a particularly tight fold with a half life of 30 min when heated to 72°C in 2% SDS [23]. When heated in SDS at lower temperatures (e.g. 50°C) the β-barrel becomes soluble without unfolding and migrates faster through the gel (30 kDa) relative to its denatured form, which runs at the expected molecular weight (35 kDa). This effect is called heat modifiability [14] and it is a general property of β-barrels. Various techniques have been used to confirm that the 30 kDa form corresponds to the native fold of OmpA (for references consult [24]. It is generally assumed that in vivo, the native form of OmpA is generated only after proper insertion into the outer membrane (Ried, 1994).

Cell membranes containing IPTG-induced OmpA-177 or OmpA-177 SA-1 proteins, expressed in LMC500 or MC1061 ΔOmpA were either heated in sample buffer for 5 min at 99°C or for 15 min at 50°C, then applied on gel, separated by SDS-PAGE and stained with Coomassie (Figure 6). Note that both OmpC and OmpF do not become soluble at 50°C and are therefore only visible in samples heated at 99°C [10]. As expected, the endogenous OmpA of LMC500 is fully heat-modifiable. Furthermore, the OmpA-177 protein, but not the OmpA-177 containing the SA-1 insertion, shows the aberrant heat-modifiability already observed in the literature for OmpA-171 [10], where the folded protein migrates slower than the unfolded protein (see also discussion). Finally, the OmpA-177 SA-1 protein is also fully heat-modifiable. We conclude that the majority, if not all, of both constructs have reached their native form.

For the FLAG and myc constructs, because of their low levels in the cell, immunoblotting was used to visualize their heat-modifiability (Figure 7). The OmpA TM domain constructs with FLAG or myc are all predominantly heat-modifiable (Figure 7A, indicated “mature”). As expected, the loop 3 precursor bands (indicated “unprocessed”), present in the membrane fraction, are not at all heat-modifiable. Also a full-length OmpA construct carrying a loop 2 FLAG insertion was found to be fully heat-modifiable (Figure 7B). We conclude that all FLAG and myc constructs are at least partially heat-modifiable, and thus can be properly incorporated in the outer membrane.

To determine the accessibility on the cell surface of the inserted antigenic peptides, both fixed and living cells were labeled. Cells carrying the FLAG and myc in either loop 2 or 3 of OmpA-177 were induced with 0.3 mM IPTG, fixed and stained with monoclonal anti-FLAG or monoclonal anti-myc. As a negative control, the primary antibody was left out, and no fluorescence was observed for any of these samples. All four constructs were detected on the bacterial cell surface (Figure S5). Loop 2 insertions show more staining at the poles, whereas loop 3 insertions are more homogeneous.

Living cells were labeled using a biotinylated variant of the anti-FLAG antibody. Labeling of living cells was performed because it was found that fluorescent streptavidin (needed for the SA-1 peptide) penetrated fixed cells to show a nucleoid-like staining (data not shown). The cells could be fixed after labeling to preserve the staining.

Results of labeled, uninduced cells carrying either OmpA-177 FLAG (loop 3) or the full-length OmpA-FLAG (loop 3) showed that both have comparable levels of antibody-accessible FLAG epitope on their surface (Figure 8A, B). This provides further evidence that adding the periplasmic domain does not result in increased stability of the protein. The limited increase in fluorescence after induction correlates with a modest increase of intact protein, and a larger increase of the 17 kDa degradation band as detected on immunoblot (data not shown).

Finally, living LMC500 cells expressing OmpA-177-SA1 were directly labeled with fluorescent streptavidin, fixed and imaged (Figure 8C). To be able to compare with and without induction directly, a short exposure time (150 ms) was chosen. Without induction staining was homogeneous along the perimeter of the cells. After induction (and after recovery from the lag phase that occurs in LMC500, see earlier), fluorescence increased markedly, and during labeling, strong streptavidin-mediated cross-linking between cells occurred, leading to clumps. Overall, we conclude that all the constructs are detected on the cell surface, but for FLAG, and likely also myc, the increase after induction is reduced due to degradation, whereas for SA-1, both efficient labeling and a strong effect of induction on the surface display is observed.

Most bacterial display studies focus on either identifying novel scaffolds or novel applications. In this work, a thorough analysis was performed of constructs that contain similar-sized peptides inserted into the OmpA display scaffold. The effect of their expression on E. coli growth, as well as their relative amount, sub-cellular localization, folding state and surface accessibility has been determined. In doing so, we believe that several useful observations have been made that are relevant for the fields of surface display, (outer) membrane protein over-expression and purification as well as for applications that make use of bacteria as antigen-displaying whole-cell adsorbents.

First, we conclude that OmpA displays not all small peptides equally efficiently, since the highly charged and hydrophilic 3xFLAG and 2xmyc epitopes reduce protein levels 10-fold due to degradation by periplasmic proteases. Second, the growth curves obtained after inducing high-level expression of the OmpA epitope variants in two standard laboratory strains MC4100 (LysA) and MC1061 suggest that strain MC1061 is a strain of choice when expressing OMPs in large amounts in E. coli. Finally, our observation that the heat modifiability of the OmpA TM domain can be changed from ‘aberrant’ to normal by insertion of small epitopes predicts that some β-barrels might not exhibit significant heat-modifiability at all.